Kuo
Laplace Transform Table
Laplace Transfonn F(s)
1
Time Functionftt)
Unit~impulse
Unit~step
function Bet)
1 8
1
function us(t)
Unit...ramp function t
s2
n! sn+l
1
~
('(n
= positive integer)
e C1t
te at
8+a
1
(8+ (1.)2
n!
(5 + at+ 1
1
t" e at (n
= positive integer)
eP1)(a 1= IJ)
(s + a)(s + 13) s
{JOI
_l_(e at 
(8 + a}(s +,8)
1
_l_(f3ept  aeat) (a =F fJ)
fJa
s(s + a)
1 8(8+ a)2
1
.!.(1  e al ) a
012
2(1  e ctt  ateat)
s2(8+ a)
,
~(at  1 + e a1 ) 2
a
0i2 t 
1
82 (S
+ (1)2
s
1[ a+(t+(; e at] 2 2)
(1  at)e at
(s + a)2
Wn
2 8
+ w;
s + ti)~
sin wnt
COSl.rJ12 t
s2
laplace Transform Table (cant.)
Laplace Transform F(s)
s(s2 +llJ~)
Time Functionflt)
I  cos(J)nt
fJi n
w~(s +a)
s2
Wn
J
0:1
ot2
+ w~ sinewill + 9)
+w;
where 8 = tan 1 (oon/a)
2
Wn
Wn (s + a}(s2 +~) 2 wn
a
+WIl
2e
1 + J a2+w~
J
. ( sm Waf

8)
where
Wn
0= tan (wn/a)
n
s2 +2twns+~
2
~
etwn/sinw ~ I
(~<
1)
s(s2 + 2~wns + w~)
w"
1
v'I="?
wit
2
I
et;wn/
Sirl(Wn~ t + e)
({< 1)
whereO
= cosl~
s2
+ 2~wns + w;,
sw;
~
et;wnt
sin(Wn v'I="? t  f))
({< 1)
whereB = coSl~
w;(s +a)
s2
ci
WII
?a{w + {Ji (  ~ _ ,) net;wj,tsin 0011
VI  t;2 t + 0)
(~< 1)
+ 2~wns+~
wheref)
= tanI W II v'I="?
aCwn
J
w; s2(s2 + 2?;'w'Js +~)
t+
W"
2~
Wn~
el;wlftsin(Wn~ t+O)
(t < 1)
where() = cos 1 (2t 2  1)
9TH
EDITION
Automatic Control Systems
FARID GOLNARAGHI
Simon Fraser University
BENJAMIN C. KUO
University of Illinois at UrbanaChampaign
@
WILEY JOHN WILEY & SONS. INC.
VI' & Executive Puhli'her D<m Fordey Associate Publisher /)(/IIleI Sayr. Senior I'rodllctiou Editor Nicole Repasky Marketing Manage r Christopher R,,,,I Senior DeSigne r Kevin MlII7,hy Production Manilgel11ent Se rvic'Cs Elm Street Puhlishing Servlce~ Edilorial Assistant Carolyn IVeisman Media Editor Lmmm Saplra Cover Pholo Science Source/Plmto Researchers n,is book was set in Times 1I0lllan by Thomson Digital and printed and bound hy Que becurl Versailles. noe cover Wa.' printed by Qucbe(.'orNer.;ililies. nils book is printed
011
acid free paper.€)
<D 2010. 2003. 2000. 1991. 19S7. 19S2. 1975. looi. 1962 John Wiley & SoIL'. Inc. All rights No part of this publication 111:'Y be reproduced. stored in u rctrit. ';11 ~'Sl cm or InmsmiUt.."(1 in any ... ronn or hy an)' means. electronic. IIK.oclmnica'. photocopying. recording, 5Canning or otherwise. cx(.'ept us pcnnllt(."(1 under Se<.llon.s 107 or 108 of the 19;6 Unitl.>d Slutf..>$ Copyright Act. wUhout either the prior written [M!nnission of the Publisher, or authoriz.ulion through payment of tile appn'l'riutc pcrcopy fee to thc Copyright Clearall(''C Center, IIIC., 222 Hoscwood Drive, Dan\'e~ . MA 01923. \\'ebsite w\\",'.(.·opyrigllt.cum. RC<luests to the Publisher for permission shoukl be addres~'tl to the l'ennissions Department. Jolm 'Viley &Sons. Inc .. III RiveI' Streel. Hoboken. NJ 07030..5774. (201 )748.0011 . (201)748liOO8. websile
Copyright
rt. SCr\'cd. "
r""
www.wilt:y.(.''Om/golpermissions.
To on:le r books or (or cu~t{)T11Cr service. pk"35C call JSOOCA U . WILEY (2255945).
MATI...AH'll) ami SlflllIllI,k1 all: lro(lernnrk, (JITI"~ i\ltllllWooo, Inc. ami are u:sell (('itll l>enllIuiOfI , TIle t(' MalhlVvri..., (Ioes 110' u:nmmt II,e accumClj of,IIe text or enrcisl!'.S In "lis book, TIlls book's u<w or tlbeus, n" of d AlA Tf.AB'1 .i('.ft1t·(/t'C vr rclll(etl procillcts tltHJN rlvt cons,itufe endonJC"rr.II' or '~/Hm$m"S/dl) by 11,e Ma'''Wmis 1 ' of (I parllcflllI r lJI!tillgogil'lli alJproach tJr IHlrllCllllIr use of 'he MATI.A8 J( ,I4iftu:ort!. f
ISBN I:! 9i8047Q.O.\8\lf;.2
Printed in the Unitc:d Stutt.s of Amerk.., 10 9 8 7 6 5 4 3 2 1
To my wife, Mitra, and to Sophia and Carmen, the joys of my life.
M. Farid Golnaraghi
Preface (Readme)
This is the ninth edition of the text but the first with Farid Golnaraghi as the lead author. For this edition, we increased the number of examples~ added MATLAB'R""'l toolboxes, and enhanced the MATLAB GUI software, ACSYS. We added more computer~aided tools for students and teachers. The prepublication manuscript was reviewed by many professors, and most of the relevant suggestions have been adopted. In this edition, Chapters I through 4 are organized to contain all background material t while Chapters 5 through 10 contain material directly related to the subject of control. In this edition, the following materials have been moved into appendices on this book's Web site at www.wiley.comlcollege/golnaraghi. Appendix A; Elementary Matrix Theory and Algebra Appendix B: Difference Equations Appendix C: Laplace Transfonn Table Appendix D: zTransfonn Table Appendix E: Properties and Construction of the Root Loci Appendix F: General Nyquist Criterion Appendix G: ACSYS 2008: Description of the Software Appendix H: DiscreteData Control Systems In addition, the Web site contains the MATLAB files for ACSYS, which are software tools for solving controlsystem problems. and PowerPoint files for the illustrations in the text. The following paragraphs are aimed at three groups: professors who have adopted the book or who we hope will select it as their text; practicing engineers looking for answers to solve their daytoday design problems; and. finally, students who are going to live with the book because it has been assigned for the control..systems course they are taking.
To the Professor: The material assembled in this book is an outgrowth of seniorlevel
controlsystem courses taught by the authors at their universities throughOllt their teaching careers. The first eight editions have been adopted by hundreds of universities in the United States and around the world and have been translated into at least six languages. Practically all the design topics presented in the eighth edition have been retained. This text contains not only conventional MATLAB toolboxes, where students can learn MATLAB and utilize their programming skills, but also a graphical MATLABbased software, ACSYS. The ACSYS software added to this edition is very different from the software accompanying any other control book. Here, through extensive use of MATLAB aUI programming. we have created software that is easy to llse. As a result, students will need to focus only on learning control problems. not programming! We also have added two new applications, SIMLab and Virtual Lab, through which students work on realistic problems and conduct speed and position control labs in a software environment. In SIMLab, students have access to the system parameters and can alter them (as in any simulation). In Virtual Lab, we have introduced a blackbox approach in which the students
I
MATLAB It is a registered trademark of The MathWorks. Inc.
iv
Preface. v
have no access to the plant parameters and have to use some sort of system identification technique to find them. Through Virtual Lab we have essentially provided students with a realistic online lab with all the problems they would encounter in a real speed or positioncontrol labfor example, amplifier saturation, noise. and nonlinearity. We welcome your ideas for the future editions of this book. Finally, a sample sectionbysection for a onesemester course is given in the Instructor's Manual, which is available from the publisher to qualified instructors. The Manual also contains detailed solutions to aU the problems in the book.
To Practicing Engineers: This book was written with the readers in mind and is very suitable for selfstudy. Our objective was to treat subjects clearly and thoroughly. The book does not use the theoremproofQ.E.D. style and is without heavy mathematics. The authors have consulted extensively for wide sectors of the industry for many years and have participated in solving numerous controlsystems problems, from aerospace systems to industrial controls, automotive controls, and control of computer peripherals. Although it is difficult to adopt all the details and realism of practical problems in a textbook at this level, some examples and problems reflect simplified versions of reallife systems. To Students: You have had it now that you have signed up for this course and your
professor has assigned this book! You had no say about the choice, though you can form and express your opinion on the book after reading it. Worse yet, one of the reasons that your professor made the selection is because he or she intends to make you work hard. But please don't misunderstand us: what we really mean is that. though this is an easy book to study (in our opinion), it is a nononsense book. It doesn't have cartoons or nicelooking photographs to amuse you. From here on, it is all business and hard work. You should have had the prerequisites on subjects found in a typical linearsystems course, such as how to solve linear ordinary differential equations, Laplace transform and applications, and timeresponse and frequencydomain analysis of linear systems. In this book you will not find too much new mathematics to which you have not been exposed before. What is interesting and challenging is that you are going to learn how to apply some of the mathematics that you have acquired during the past two or three years of study in college. In case you need to review some of the mathematical foundations, you can find them in the appendices on this book's Web site. The Web site also contairts lots of other goodies, including the ACSYS software, which is GUI software that uses MATLABbased programs for solving linear control systems problems. You will also find the Simulink(R·2based SIMLab and Virtual Lab, which will help you to gain understanding of realworld control systems. This book has numerous illustrative examples. Some of these are deliberately simple for the purpose of illustrating new ideas and subject matter. Some examples are more elaborate, in order to bring the practical world closer to you. Furthennore. the objective of this book is to present a complex subject in a clear and thorough way. One of the important learning strategies for you as a student is not to rely strictly on the textbook assigned. When studying a certain subject, go to the library and check out a few similar texts to see how other authors treat the same subject. You may gain new perspectives on the subject and discover that one author may treat the material with more care and thoroughness than the others. Do not be distracted by writtendown coverage with oversimplified examples. The minute you step into the real world, you will face the design of control systems with nonlinearities and/or timevarying elements as well as orders that can boggle your mind. It
2 Simulink'):I
is a registeted trademark of The MathWorks. Inc.
vi '. Preface
discouraging and firstorder exist may be discouraging to tell you now that strictly linear and firstorder systems do not exist in world. in the real world. Some Some advanced engineering students in college do not believe that the material they of learn in the classroom is ever going to be applied directly in industry. Some of our students come back from field and interview trips totally surprised to find that the material they learned in courses on control systems is actually being used in industry today. They are surprised to find that this book is also a popular reference for practicing engineers. engineers. Unfortunately, these factfinding, eyeopening, and selfmotivating trips usually occur near selfmotivating of often motivated. the end of their college days, which is often too late for students to get motivated. There are many learning aids available to you: the MATLAB .. based ACSYS software MATLABbased software of will assist you in solving all kinds of controlsystems problems. The SIMLab and Virtual of Lab software can be used for simulation of virtual experimental systems. These are all found on the Web site. In addition, the Review Questions and Summaries at the end of each useful site, chapter should be useful to you. Also on the Web site~ you will find the errata and other supplemental material. enjoy major We hope that you will enjoy this book. It will represent another major textbook acquisition (investment) in your college career. Our advice to you is not to sell it back to the bookstore at the end of the semester. If you do so but find out later in your professional If professional career that you need to refer to a control systems book, you will have to buy it again at a refer higher price.
Special Acknowledgments: The authors wish to thank the reviewers for their invaluable comments and suggestions. The prepublication reviews have had a great impact on the prepublication revision project. Dr. Earl Foster, Dr. Vahe Caliskan, The authors thank Simon Fraser students and research associates Michael Ages, Fraser Jamalian, Jennifer Johannes Minor, Linda^Franak, Arash Jamalian7 Jennifer Leone, Neda Parnian, Sean Franak MacPherson, Amin Kamalzadeh, and Nathan (Wuyang) Zheng for their help. Farid Golnaraghi Professor Benjamin of Golnnraghi also wishes to thank Professor Benjamin Kuo for sharing the pleasure of for writing this wonderful book, and for his teachings, patience, and support throughout this wonderful teachings. experience. experience.
7
^
M. F. Golnaraghi, Golnaraghi, Vancouver, British Columbia, Vancouver. British ColUJnbia. Canada Canada B. C. Ku(), Kuo, CiulInpaign. Illinois, U.S.A. Champaign, Illinois. U.S.A.
2009 2009
Contents
Preface iv
• CHAPTER 1
228
Introduction 1
11 Introduction 111 Basic Components of a Control System 2 1~12 Examples of ControlSystem Applications 2 OpenLoop Control Systems 113 (Nonfeedhack Systems) 5 114 ClosedLoop Control Systems (Feedback Control Systems) 7 \Vhat Is Feedback. and What Are Its Effects? 8 121 Effect of Feedback 011 Overall Gain 8 122 Effect of Feedback on Stability 9 123 Effect of Feedback on External Disturbance or Noise 10 Types of Feedback Control Systems 11 131 linear versus Nonlinear Control Systems 11 132 TimeInvmiant versus TimeVarying Systems 12 Summary 14
1~2
24
13
26
14
~
CHAPTER 2
Mathematical Foundation 16
21 ComplexVariable Concept 16 Complex Numbers 16 211 Complex Variables 18 212 213 Functions of a Complex Variable 214 Analytic Function 2() 215 Singularities and Poles of a Function 20 216 Zeros of a FUIlction 20 217 Polar Representation 22
27
19
22
FrequencyDomain Plots 26 221 ComputerAided Construction of the FrequencyDomain Plots 26 222 Polar Plots 27 223 Bode Plot (Corner Plot or Asymptotic Plot) 32 224 Real Constant K 34 225 Poles and Zeros at the Origin, Uw)±P 34 22~6 Simple Zero, I + jwT 37 227 Simple Pole. 1/(1 + jltJT) .'39
28
29
2JO 211
212 21:3
Quadratic Poles and Zeros 39 Pure Time Delay, ejwTd 42 MagnitudePhase Plot 44 Gain and PhaseCrossover Points 46 Minimum.Phase and Nonminimum2212 Phase Functions 47 Introduction to Differential Equations 49 231 Linear Ordinary Differential Equations 49 Nonlinear Differential Equations 49 2.'32 FirstOrder Differential 233 Ec!uations; State Equations 50 234 Definition of State Variables 50 23.5 The Output Equation .51 Laplace Transfonn 52 241 Definition of the Laplace Transform 52 Inverse Laplace Transformation .54 Important Theorems of the Laplace Transform .54 Inverse Laplace Transform by PartialFraction Expansion 57 251 PartialFraction Expansion ,57 Application of the Lapla(.'e Transform to the Solution of Linear Ordinary Differential Equations 62 FirstOrder Prototype System 63 261 SecondOrder Prototype 262 System 64 Impulse Response and Transfer Functions of Linear Systems 67 271 Impult;e Response 67 272 Transfer Function (Single~Input. SingleOutput Systems) 70 Proper Transfer Functions 71 273 Characteristic Equation 71 274 Transfer Function (Multivariable 275 Systems) 71 Stability of Linear Control Systems 72 BoundedInput. BoundedOutput (BIBO) S tahility~Continuous Data Svstems 73 ReI.ltioJlship between Characteristic Eqllution Roots and Stability 74 ZeroInput and ~ymptotic Stability of ContinuousData Systems 74 Methods of Determining Stability 77 RouthHurwitz Criterion 78 229 2210 2211
vii
viii
Contents
Routh's Tabulation 79 Special Cases when Routh's Tabulation Terminates Prematurely 80 MATLAB Tools and Case Studies 84 2141 DescIiption and Ufie of Transfer Function Tool 84 2142 MATLAB Tools for Stability 85 Summmy 90
415
42
2la2
214
43
21.5
 CHAPTER
3
Block Diagrams and SignalFlow Graphs 104
31
Backlash and Dead Zone (Nonline~lr Characteristics) 164 Introduction to Modeling of Simple Electrical Systems 16.5 421 Modeling of Passive Electrical Elements 165 422 Modeling of Electrical Networks 165 Modeling of Active Electrical Eleinents: Operational Amplifiers 172 431 The Ideal OpAmp 173 432 Sums and Differences 173
4.'33
FirstOrder OpAmp
32
Block Diagrams 104 311 Typical Elements of Block Diagrams in Control Systems 106 312 Relation between Mathemntica.I Equations and Block Diagrams 109 31.'3 Block Diagram Reduction 113 314 Block Diagram of MultiInput SystemsSpecial Case: Systems with a Disturbance 115 315 Block Diagmms and Transfer Functions of Multivariable Systems 117 SignalFlow Graphs (SFGs) 119 :321 Basic Elements of an SFG 119 322 SummalY of the Basic Properties of SFG 120 323 Definitions of SFC Terms 120 324 SFG Algebra 123 32,5 SFC of a Feedback Control
46
4~8
System
326
124
33 34
Relation he tween Block Diagrams and SFGs 124 32~7 Gain Formula for SFG 124 328 Application of the Gain Fonllula between Output Nodes and Noninput Nodes 127 .'329 Application of the Gain Formula to Block Diagrams 128 3210 Simplified Gain Formula 129 MATLAB Tools and Case Studies 129 Summary 133
49
410
411
412 413
Configurations 174 Introduction to Modeling of Thermal Systems 177 441 Elementary Heat Transfer Properties 177 Introduction to Modeling of Fluid Systems 180 451 Elementary Fluid and Gas System Properties 180 Sensors and Encoders in Control Systems 189 461 Potentiometer 189 . 4·6·2 Tachometers 194 463 Incremental Encoder 195 DC Motol's in Conh"ol Systems 198 471 Basic Operational Principles of DC Motors 199 4i·2 Basic Classifications of PM DC Motors 199 473 Mathematical Modeling of PM DC Motors 201 Systems \"ith Tnmsportation Lags (Time Delays) 205 481 Approximation of the Time~DcIay Function hy Rational Functions 206 Linearization of Nonlinear Systems 206 491 Linearization Using Taylor Series: Classical Representation 20i 492 Lineari7.ation Using the State Space Approach 207 Analogies 213 Case Studi('~s 216 MATLAB Tools 222 Summary 22.'3
 CHAPTER 4
Theoretical Foundation and Background Material: Modeling of Dynamic Systems 147
41
Introduction to Modeling of Mechanical Systems 148 411 Translational Motion 148 Rotntlonal Motion 1.57 412 413 Conversion between Tmnslational and Rotational Motions 161 414 Gear Trains 162
> CHAPTER 5
TimeDomain Analysis of Control Systems 253
51
52
53
54
Time Response of ContinuousData Systems: Introduction 253 Typical Test Signals for the Time Response of Control Systems 254 The UnitStep Response and TimeDomain Specifications 256 SteadyState Error 258
Contents .... ix
.541
SteadyState Error of Linenr ContinuousData Control Systems 2.58 SteadyState Error Caused fly Nonlinear System Elements 272 Time Response of a Pr~t()type FirstOrder System 274 Transient Response of a Protol)pe SecondOrder System 275 561 Damping Ratio und Damping Factor 277 562 Natural Undamped Frequency 278 563 Ma.xiU1UIIl Overshoot 280 564 Delay Time amI Rise Time 2R3 565 Settling TiIl1(~ 285 Speed and P()sitio~ Control of a DC Motor 289 571 Speed Response and the Effects of Inductance and DisturbanceOpen Loop Response 289 Speed Control of DC Motors: ClosedLoop Response 291 573 PositioIl Control 292 TimeDomain Analysis of a PositionControl
!D CHAPTER. 6
The Control Lab 337
61 62 Introduction 3.'37 Description of the Virtual Experimental System .'338 621 Motor 339 Position Sensor or Speed Sensor 3,39 622 62~3 Power Amplifier 340 624 Interface 340 Description of SIMLnh and Virtual Lab Software 340 Simulation .:md Virtunl EJo.'Periments 345 641 OpenLoop Speed 345 OpenLoop Sine Input :347 64 ..2 6·4 ....'3 Speed Control 3.50 644 Position Control 352 D(~sign Project lHohotic Arm 354 Design Project 2~QuarterCur Model 3.57 661 Introduction to the QuarterCar
66~2
56
63
6~4
5i
65
66
58
Model 357 C}nseclLoop Acceleration
System 293 • 581 UnitStep Transient Response 294 582 The SteadyState Response 298
583
Time Response to a UnitRamp Input 298 Time Response of a ThirdOrder System 300 Basic Control Systems and Effects of Adding Poles ml(l Zeros to Transfer FUIlctions 304 5Hl Addition of a Pole to the ForwardPath Transfer FUI1C.·tion: UnityFeedback Systems a05 Addition of a Pole to tIw 592 CloseclLoop Trnnsft"r Function :30; 59...'3 Addition of a Zero to tilt' ClosedLoop Transfer Function aw; Addition of a Zero to the ForwardPath Tmnsfer Function: UnityFeedhack Systems 309 Dominant Poles and Zeros of Tnlllsfer Functions 311 SumnmlY Elrect~ of Polps and 5101 Zeros 313 The nelative Dumping Hatio :313 5102 510..3 The Proper \Vay of Negl(~eting the Insignificant Poles \\ith Consideration of the SteadyState Hesponse .313 Basic Control Systems Utilizing Addition of P()l(~s and Zeros 314 MATLAB Tools 3H) SUmmal)' 320
663
67
Control :359 DescriptIon of Quarter Car Modeling Tool 360 664 Passive Suspension 364 665 ClosedLoop Relative Position Con trol 365 666 Closed ..Loop Acceleration Control 366 Summmy 367
,. CHAPTER 7
Root Locus Analysis 372
71 72
7 3
510
or
Introduction 372 Basic Properties of the Root Loci (RL) 373 Properties of the Root Loci 377 731 K = 0 and K ±oo Points 317 732 Numh(~r of Branches OIl the Root Loci .'378 7:33 Svmmet..v of the RL 378 Angles of Asymptotes of the RL: 734 Beha\ior of the RL at lsI = 00 378 Intt~r::;C'ct of the Asymptotes 735 (Centroid) 379 73(j Root Loci OIl the Heal A'<is :380 737 Angles of Departure and Angles of Arriva1 of the RL 3S0 Intersection of the RL \\ith the ImagiIlary Ax.is 380 7:39 Breakaway Points (Snddle Points)
=
512 513
OIl
the RL 380
:382
7:310
TIle Root Sensitivity
x
~
Contents
Design Aspects of the Root Loci 38.1) 741 Effects of Adding Poles and Zeros to G(s) H(s) 385 Root Contours (HC): MultipleParameter Variation 393 MATLAB Tools and Case Studies 400 Summary 400 812 Relative Stability Related to the Slope of the Magnitude CUlve of the Bode Plot 459 8121 Conditionally Stable System 459 Stability Analysis with the MagnitudePhase Plot 462 ConstantM Loci in the MagnitudePhase Plane: The Nichols Chart 463 . Nichols Chart Applied to NonnnityFeedback Systems 469 Sensitivity Studies in the Frequency Domain 470 MATLAB TooL') and Case Studies 472 Summary 472
74
75 76 77
~
813 814 815 816 817
818
CHAPTER
a
FrequencyDomain Analysis 409
81
82
8....'3
84
85
86
8~7
88
89
810
811
Introduction 409 811 Frequency Response of ClosedLoop Systems 410 812 FrequencyDomain Specifications 412 Mn W r• and Bandwidth of the Prototype SecondOrder System 413 821 Resonant Peak and Reson~mt Frequency 413 8M22 Bandwidth 416 Effects of Adding a Zero to the ForwardPath Transfer Function 418 Effects of Adding a Pole to the ForwardPath Transfer Function 424 Nyquist Stability Cliterioll: Fundamentals 426 851 Stability Problem 427 852 Definition of Encircled and Enclosed 428 853 Number of Encirclements and Enclosures 429 854 Principles of the Argument 429 8.55 Nyquist Puth 433 856 Nyquist Criterion and the L(s) Of the G(s)H(s) Plot 434 Nyquist Criterion for Systems with MinimumPhase Transfer Functions 435 86 1 Application of the NyqUist Criterion to MinimumPhase Tranfer Functions That Are Not Strictly Proper 436 Relation between the Root Loti and the Nyquist Plot 437 Illustrative Examples: NyqUist Criterion for MinimumPhase Transfer Functions 440 Effects of Adding Poles and Zeros to L(s) on the Shape of the NyqUist Plot 444 Relative StabiUty: Gain Margin and Phase Margin 449 8101 Gain Margin (GM) 451 8102 Phase Margin (PM) 453 Stability Analysis with the Bode Plot 455 8111 Bode Plots of Systems with Pure Time Delays 458
w
t. CHAPTER 9
Design of Control Systems 487
91 Introduction 487 911 Design Specifications 487 912 Controller ConfiguTIltions 489 913 Fundamental Principles of Design 491 Design with the PD Controllel' 492 921 TimeDomain Interpretation of PD Control 494 922 Frequem.yDoml1in Interpretation of PD Control 496 923 Summary of Effects of PD Control 497 Design with the PI Controller .511 9.'31 TimeDomain Interpretation and Design of PI Control 513 932 FrequencyDomain Interpretation and Design of PI Control 514 Design with the PID Controller 528 Design with PhaseLead Controller 532 951 TimeDomain Interpretation and Design of PhaseLead Control .534 952 FrequencyDomain Interpretation and Design of PhaseLead Control 535 953 Effects of PhaseLead Compensation ,1)54 9.54 Limitations of SingleStage PhaseLead Control 555 955 Multistage PhaseLead Controller ,555 9.56 Sensitivity Considerations 559 Design with PhaseLag Controller 561 961 TimeDomain Interpretation and Design of PhaseLag Control .561 962 Frequenc.yDomain Interpretation and Design of PhaseLag Control 563 963 Effects and Limitations of PhaseLag Control 574 Design \'vith LeadLag Controller 574 PoleZeroCuneellation Design: Notch Filter 576 981 SecondOrder Active Filter .579 982 Freql.lency. . Domain Interpretation and Design .580
92
93
94 95
96
97 98
Contents
9~9
<0lIl
xi
9~10
911
912
913
914 915 916
~
Forward and Feedforward Controllers 588 Design of Robust Control Systems 590 Minor~Loop Feedback Control 601 9~11 ~ 1 RateFeedback or Tachometer~Feedback Control 601 9~112 MinorLoop Feedback Control with Active Filter 603 A Hydraulic Control System 605 9121 Modeling Lineal' Actuator 60.5 9122 Four\Vay ElectroHydraulic Valve 606 g~123 Modeling the Hydraulic System 612 9124 Applications 613 Controller Design 617 9131 P Control 617 913~2 PD Control 621 9133 PI Control fi2.fi 9·134 PID Control 628 MATLAB Tools and Case Studies 631 Plotting Tutorial 647 Summary 649
1081 1082
10~83
1084
Characteristic EqU~ltioIl from a Differential Equation 695 Characteristic Equation from a Transfer Function 696 Characteristic Equation from State Equations 696 Eigenvalues 697
1085
109
Eigenvectors 697
CHAPTER 10
1011
State Variable Analysis 673
1O~ 1 102
103 104
105
106 107 108
Introduction 673 Block Diagrams. Transfer Functions. and State Diagrams 673 1021 Transfer Functions (Multivariable Systems) 673 1022 Block Diagrams and Transfer Functions of M ultivariable Systems 674 State Diagram 676 From Differential Equations to State Diagrams 678 From State Diagrams to Transfer Function 679 1026 From State Diagrams to State and Output Equations 680 VectorMatrix Representation of State Equations 682 StateTransition MatriX 684 1O4~1 Signiflcance of tbe StateTransition Matrix. 685 1042 Properties of the StateTransition Matrix 685 StateTransition Equation 687 1051 StateTransition Equation Determined from the State Diagram 689 Relationship between State Equations and HighOrder Differential Equations 691 Relationship between State Equations and Transfer Fundions 693 Chamcteristic Equations, Eigenvalues, lUld Eigenvectors 695
1012
1013
1016 1017
1018 1019
10~20
1086 Generalized Eigenvectors 698 Similarity Transformation 699 1091 Invariance Propertie..., of the Similarity Transformations 700 1092 Controllability Canonical Form (CCF) 701 1093 ObservabilityCmlonicru Fonn (OCF) 703 1094 Diagonal Canonical Form (DCF) 704 1095 Jordan Canonical Fonn (JCF) 706 Decompositions of TnUlsfer Functions 707 Direct Decomposition 707 10101 lO1O~2 Cascade Decomposition 712 10103 Parallel Decomposition 713 Conhoollability of Control Systems 714 10111 General Concept of Controllability 716 10112 Definition of State COIltrollability i16 10113 Altemute Tests 011 Controllability 717 Ohservability of Linear Systems 719 10121 Definition of ObseIVability 719 10122 Alternate Tests on Obsetvability 720 Relationship among Controllability, . ObservabiliLy, and Transfer Functions 721 Invariant Theorems on Controllability and o bservability 72.'3 Case Study: MagneticBall Suspension System 725 StateFeedback Control 728 PolePlacement Design Through State Feedback 730 State Feedhack with Integral Control 735 MATLAB Tools and Case Studies 741 10191 Description and Use of the State~Space Analysis Tool 741 1019~2 Description and Use of t£"ynl for StateSpace Applications 748 Summary 751
77'3
... INDEX
Appendices can be found on this hook's companion 'Veb site: www.wileyocom/collegelgoinaraghi.
~
APPENDIX A
Elementary Matrix Theory and Algebra Al
AI Elementary Matrix Theory AI A~I1 Definition of a Matrix A2
xii • Contents
A2 Matrix Algebra A~.5 A21 Equality of Matrices A.5 A22 Addition and Subtraction of Matrices A~6 A2~3 Associative Law of Matrix (Addition and Subtraction) A6 A24 Commutative Law of Matrix (Additioll and Subtraction) A6 A2~.5 Matrix Multiplicutiml A6 A2~6 Rules of Matrix Multiplication 1\7 A2.. 7 Multiplication by a Scalar k A8 A28 Inverse of a Matrix (Mahi.x Division) A8 A2~9 Rank of a Matrix A9 ComputerAided Solutions of Matrices A9
> APPENDIX G
ACSYS 2008: Description of the Software 61
Gl G2
Installation of ACSYS Gl Description of the Software GI C21 t(1)ym G2 C2~2 Statetool G3 G·2..3 Controls G3 G·2~4 SIMLab and Virtual Lab Final Comments G4
G4
G·,'3
~
APPENOIX H
DiscreteData Control Systems H'
HI H2
Introduction HI The zTmnsfonn HI Definition of the z.Transform III Relationship between the Laplace Transform and the zTransform H2 H2..3 Some Important Theorems of the zTransform H3 H·2..4 Inverse zTransform H.5 H2..5 Computer Solution of the Partial~ Fraction E:\1>unsion ofY(z)/z B7 H26 ApplicatioIl of the zTransform to the Solution of Lineal' Difference Equations H7 Transfer Functions of DiscreteData Syst(~ms HS H3~1 Tr~msfer Functions of Discrete~Data Systems with Cascade Elements H12 II 3~2 Transfer Function of the ZeroOrderHold H13 H33 Transfer Functions of ClosedLoop DiscreteData Systems H14 State Equations of Linettr DiscreteData Systems H16 H4~1 Discrete State Equations H16 H 4.. 2 Solutions of the Discrete State Equations: Discrete StateTransition E(luations H18 H~43 zTr.lnsform Solution of Discrete State Equations H19 H4..4 Tmnsfer~Functi()n Matrix and the Characteristic Equation H20 H4~5 State Diagrams of Discrete~Data Systems H22 H4~6 State Diabtrams f(Jr SampledData Systems H23 Stability of DiscreteData Systems H·26 H.5 .. 1 BIBO Stability H26 H52 ZeroInput Stahility H26 II·5~3 Stahility Tests of DiscreteDatll Systems H27 TinwDomain Properties of DiscreteData H21
A3
,.. APPENDIX B
Difference Equations 81
B1
H22
Difference Equations
B1
,. APPENDIX C
Laplace Transform Table
~
C1
APPENDIX D
zTransform Table D1
~
APPENDIX E
Properties and Construction of 1he Root Loci El
El E2
E3 E4 E5 E6 E7
E~8
K = 0 and K = ±oo Points El Number of Brunches on the Root Loci E2
Symmetry of the Root Loci E2 Angles of Asymptotes of the Root Loci and Behtlvior of the Root Loci at lsi = 00 E4 Intersect of the Asymptotes (Centroid) E5 Root Loei on the Real Axis E8 Angles of Departure Hnd Angles of Arrival of the Root Loci E9 Intersection of the Root Loci with the Imaginary Axis Ell Breakaway Points Ell E9~ 1 (Saddle Points) on the Root Loci Ell E9..2 The Angle of Arrival and Departure of Root Loci at the Breakaway Point E12 Calculation of K on the Root Loci E16
B3
H4
E9
EwlO
~
APPENDIX F
General Nyquist Criterion F1
FI
F 2 F:3
Formulation of Nyquist Criterion F~l Fl~l System with MinimumPhase Loop Transfer Functions F4 F 1 ~2 Systems vvith Improper Loop Transfer Functions F 4 Illustrative EXfUllplesGenel'al 1\yquist CIiteIion Minimum ruul Nonminimum Transfer Functions F4 Stability Analysis of Multiloop Systems F13
H5
H6
Contents
11:31 Time Response of DiscreteData Control Systems H31 H 62 Mapping iletween .vPlane amI zPlane Trajectories H34 Ho3 Relation between CharacteristicEquatioIl Roots and Transient Hespollse H38 SteadyState Error Analysis of DiscreteData Control Systems H41 Root Loci of DiscreteData Systems H4.5 FrequencyDomain Analysis of DiscreteData Control Systems H49 Systems H61 HI02
~
xiii
Hll
H7 118 H9
1112
1191
HIO
Bode Plot with tlw
1151
H13 H14
wTmnsformation H 50 Design of DiscreteData Control Sy~iems 11101 Introduction II51
Digital lmplt~mentati()n of Analog Contl'OIlers 1152 HI03 Digital Implementation of the PID Controller H54 H1O4 Digital Implementation of Lead and Lug ControlIers H 57 Digital Controllers II58 11111 Physical Realizahility of Digital Conttolle."s 11.58 Design of DiscreteData Control Systems in the Frequency Domain and the zPlane H51 H121 Phase~Lead and PhaseLag Controllers in the tvDomain H61 Design of DiscreteData Control Systems with Deadbeat Response Ii68 PolePlac(~mellt Design with State
Feedback H70
, CHAPTER
"
~..'
Introduction
11 INTRODUCTION
The main objectives of this chapter are:
1.
2.
To define a control system. To explain why control systems are important. To give some examples of controlsystem applications. To explain why feedback is incorporated into most control systems. To introduce types of control systems.
3. To introduce the basic components of a control system .
4. 5.
6.
• Control systems are in abundance in modern
civilization .
One of the most commonly asked questions by a novice on a control system is: What is a control system? To answer the question , we can say that in our daily lives there are numerous "objectives" that need to be accomplished. For instance, in the domestic domain, we need to regulate the temperature and humidity of homes and buildings for comfortable living. For transportation, we need to control the automobile and airplane to go from one point to another accurately and safely. Industrially. manufacturing processes contain numerous objectives for products that will satisfy the precision and costeffectiveness requirements. A human being is capable of performing a wide range of tasks, including decision making. Some of these tasks, such as picking up objects and walking from one point to another, are commonly carried out in a routine fashion. Under certain conditions, some of these tasks are to be performed in the best possible way. For instance, an athlete running a IOOyard dash has the objective of running that distance in the shortest possible time. A marathon runner, on the other hand. not only must run the distance as quickly as possible, but, in doing so, he or she must control the consumption of energy and devise the best strategy for the race. The means of achieving these "objectives" usually involve the use of control systems that implement certain control strategies. In recent years, control systems have assumed an increasingly important role in the development and advancement of modern civilization and technology. Practically every aspect of our daytoday activities is affected by some lype of (;Untrol system. Connor systems are found in abundance in all sectors of industry, such as quality control of manufactured products, automatic assembly lines. machinetool control, space technology and weapon systems, computer control, transportation systems, power systems, robotics, MicroElectroMechanical Systems (MEMS), nanotechnology, and many others. Even the control of inventory and social and economic systems may be approached from the theory of automatic controL
1
Potential applications of control of these systems may benefit the following areas: • Machine tools. or controlled variables. and space industries. 112 Examples of ControlSystem Applications Intelligent Systems Applications of control systems have significantly increased through the development of new materials. In all these areas. 3.g. which provide unique opportunities for highly efficient actuation and sensing. Control in Virtual Prototyping and Hardware in the Loop The concept of virtual prototyping has become a widely used phenomenon in the automotive. • Photolithography. engines~ flight control mechanisms. steering. Objectives of control. In more technical terms~ the objectives can be identified with inputs. 11. Introduction Objectives Results Figure 11 Basic components of a control system. defense. Design tools such as MATLAB and Simulink enable companies to design and test controllers for different components (e. For example. ABS. or actuating signals~ u. onloff shape control of solar reflectors or aerudynamic surfaces. aeronautics. marine. y. the objective of the control system is to control the outputs in some prescribed manner by the inputs through the elements of the control system. Here the physical controller hardware is interfaced with the computer and replaces its mathematical model within the computer! . robotics. biomedical.. Hardware in the loop terminology is a new approach of testing individual components by attaching them to the virtual and controller prototypes. • Process control. 111 Basic Components of a Control System The basic ingredients of a control system can be described by: 1. and specialized devices) within the system and examine the behavior of the control system on the virtual prototype in real time. surgical. and endoscopic. pressure to cut costs has forced manufacturers to design and test an entire system in a computer environment before a physical prototype is made. Stateoftheart actuators and sensors may be implemented in virtually any system. • Biomechanical and biomedical. and the defense and space industries. and the results are also called outputs. In general. aerospace. locomotion. Results or outputs. The basic relationship among these three components is illustrated in Fig. Controlsystem components. Enable the manufacture of smaller microelectronic circuits by controlling vibration in the photolithography circuitprinting process. suspension. • Flexible robotics.2 Chapter 1. including biological propulsion. Improve precision and increase productivity by controlling chatter. material handling. thereby reducing energy lusses and environmental impacts. 2. Artificial muscles. drug delivery systems. This allows the designers to change or adjust controller parameters online before the actual hardware is developed. and other assistive technologies. Enable faster motion with greater accuracy. landing gear.
11 Introduction ~ 3
Smart Transportation Systems
The automobile and its evolution in the last two centuries is arguably the most transformative invention of man. Over years innovations have made cars faster, stronger, and aesthetically appealing. We have grown to desire cars that are . 'intelligent" and provide maximum levels of comfort. safety, and fuel efficiency. Examples of intelligent systems in cars include climate control, cruise control, antilock brake systems (ABSs), active suspensions that reduce vehicle vibration over rough terrain or improve stability. air springs that selflevel the vehicle in highG turns (in addition to providing a better ride). integrated vehicle dynamics that provide yaw control when the vehicle is either over or understeering (by selectively activating the brakes to regain vehicle control), traction control systems to prevent spinning of wheels during acceleration, and active sway bars to provide "controlled" rolling of the vehicle. The following are a few examples.
Drivebywire and Driver Assist Systems The new generations of intelligent vehicles will be able to understand the driving environment, know their whereabouts, monitor their health. understand the road signs, and monitor driver perrormance, even overriding drivers to avoid catastrophic accidents. These tasks require significant overhaul of current designs. Drivebywire technology replaces the traditional mechanical and hydraulic systems with electronics and control systems; using electromechanical actuators and humanmachine interfaces such as pedal and steering feel emulatorsotherwise known as haptic systems. Hence, the traditional componentssuch as the steering column, intermediate shafts, pumps, hoses, fluids. belts, coolers, brake boosters, and master cylindersare eliminated from the vehicle. Haptic interfaces that can offer adequate transparency to the driver while maintaining safety and stability of the system. Removing the bulky mechanical steering wheel column and the rest of the steering system has clear advantages in tenns of mass reduction and safety in modern vehicles, along with improved ergonomics as a result of creating more driver space. Replacing the steering wheel with a lk'lptic device that the driver controls through the sense of touch would be useful in this regard. The haptic device would produce the same sense to the driver as the mechanical steering wheel but with improvements in cost. safety. and fuel consumption as a result of eliminating the bulky mechanical system. Driver assist systems help drivers to avoid or mitigate an accident by sensing the nature and significance of the danger. Depending on the significance and timing of the threat, these onboard safety systems will initially alert the driver as early as possible to an impending danger. Then, they will actively assist or, ultimately, intervene in order to avert the accident or mitigate its consequences. Provisions for automatic over~ride features when the driver loses control due to fatigue or lack of attention will be an important part of the system. In such systems, the socalled advanced vehicle control system monitors the longitudinal and lateral control, and by interacting with a central management unit. it will be ready to take control of the vehicle whenever the need arises. The system can be readily integrated with sensor networks that monitor every aspect of the conditions on the road and are prepared to take appropriate action in a safe manner. Integration and UtiliUltion ofAdvanced Hybrid Powertrains Hybrid technologies offer improved fuel consumption while enhancing driving experience. Utilizing new energy storage and conversion technologies and integrating them with powertrains would be prime objectives of this research activity. Such technologies must be compatible with current platfonns and must enhance, rather than compromise, vehicle function. Sample applications would include developing plugin hybrid technology. which would enhance the vehicle cruising distance based on using battery power alone, and utilizing sustainable
4 .... Chapter 1. Introduction
energy resources, such as solar and wind power, to charge the batteries. The smart plugin vehicle can be a part of an integrated smart home and grid energy system of the future t which would utilize smart energy metering devices for optimal use of grid energy by avoiding peak energy consumption hours.
High Performance Realtime Control, Health Monitoring, and Diagnosis Modern
vehicles utilize an increasing number of sensors, actuators. and networked embedded computers. The need for high performance computing would increase with the introduction of such revolutionary features as drivebywire systems into modern vehicles. The tremendous computational burden of processing sensory data into appropriate control and monitoring signals and diagnostic information creates challenges in the design of embedded computing technology. Towards this end, a related challenge is to incorporate sophisticated computational techniques that control. monitor. and diagnose complex automotive systems while meeting requirements such as low power consumption and cost effectiveness. The following represent more traditional applications of control that have become part of our daily lives. Steering Control of an Automobile As a simple example of the control system, as shown in Fig. 11, consider the steering control of an automobile. The direction of the two front wheels can be regarded as the controlled variable~ or the output, y; the direction of the steering wheel is the actuating signal~ or the input, u. The control system, or process in this case, is composed of the steering mechanism and the dynamics of the entire automobile. However, if the objective is to control the speed of the automobile, then the amount of pressure exerted on the accelerator is the actuating signal. and the vehicle speed is the controlled variable. As a whole, we can regard the simplified automobile control system as one with two inputs (steering and accelerator) and two outputs (heading and speed). In this case. the two controls and two outputs are independent of each other, but there are systems for which the controls are coupled. Systems with more than one input and one output are called multivariable systems.
Idle~Speed
Control of an Automobile As another example of a control system, we consider the idlespeed control of an automobile engine. The objective of such a control system is to maintain the engine idle speed at a relatively low value (for fuel economy) regardless of the applied engine loads (e.g., transmission, power steering, air conditioning). Without the idlespeed control, any sudden engine~load application would cause a drop in engine speed that might cause the engine to stall. Thus the main objectives of the idle~speed control system are (1) to eliminate or minimize the speed droop when engine loading is applied and (2) to maintain the engine idle speed at a desired value. Fig. 12 shows the block diagram of the idlespeed control system from the standpoint of inputssystemoutputs. In this case. the throttle angle a and the load torque TL (due to the application of air conditioning, power steering, transmission. or power brakes. etc.) are the inputs, and the engine speed w is the output. The engine is the controlled process of the system. SunTracking Control of Solar Collectors To achieve the goal of develuping economically feasible nonfossilfuel electrical power, the U .S~ government has sponsored many organizations in research and development of solar power conversion methods. including the solar~cell conversion techniques. In most of
11 Introduction
5
~
,
~
000
,=
•
lDLESPEED CONTROL
\~TOR
Load torque TL
Thro(tle angle a
ENGINE
Engine speed w
Figure 12 Idlespeed control system.
Figure 13 Solar collector field .
these systems, the need for high efficiencies JiCLaleS the use of devices for sun tracking. Fig. 13 shows a solar collector field. Fig. 14 shows a conceptual method of efficient water extraction using solar power. During the hours of daylight, the solar collector would produce electricity to pump water from the underground water table to a reservoir (perhaps on a nearby mountai.n or hill), and in the early morning hours, the water would be released into the irrigation system. One of the most important features of the solar collector is that the collector dish must track the sun accurately. Therefore, the movement of the collector dish must be contro ll ed by sophisticated control systems . The block diagram of Fig. 15 describes the general philosophy of the suntracking system together with some of the most important components. The controller ensures that the tracking collector is pointed toward the sun in the morning and sends a "start track" command. The controller constantly calculates the sun's rate for the two axes (azimuth and elevation) of control during the day. The controller uses the sun rate and sun sensor infOlmation as inputs to generate proper motor commands to s1ew the collector.
113 Openloop Control Systems (Nonfeedback Systems)
• Openloop systems .u·e econo mical but usu ally in accurate.
The idlespeed control system illustrated in Fig. 12, shown previollsly, is rather unsophisticated and is called an openloop control system. It is not difficult to see that the system as shown would not satisfactorily fulfill critical performance requirements. For instance, if the throttle angle ex is set at a certain initial value that corresponds to a certain
6 .. Chapter 1. Introduction
HILL
8
ELECTRICITY
IRRIGATION
r~ ~ MOTOR
WATER TABLE
~SOLAR COLLECTOR
/
Figure 14 Conceptual method of efficient water extraction using solar power.
()i 
.

Sun's Rate
,
Command MotocRate
Trim SUN
SENSOR
e
Position Error
CONTROLLER
Rate
/
MOTOR
DRIVER
+
r
Torque Dislurbance Td
SPEED REDUCER
Figure 1·5 Important components of the suntracking control system.
engine speed, then when a load torque TL is applied, there is no way to prevent a drop in the engine speed. The only way to make the system work is to have a means of adjusting a in response to a change in the load torque in order to maintain (J) at the desired level. The conventional electric washing machine is another example of an openloop control system because, typically, the amount of machine wash time is entirely determined by the judgment and estimation of the human operator. The elements of an openloop control system can usually be divided into two parts: the controller and the controlled process, as shown by the block diagram of Fig. 16. An input signal, or command, t, is applied to the controller, whose output acts as the actuating signal u; the actuating signal then controls the controlled process so that the controlled variable y will perform according to some prescribed standards. In simple cases, the controller can be
Reference ,._ _ _ _, Controlled input I' CONTROLLED variable y   .tCONTROLLER I~ PROCESS
Figure '·6 Elements of an openloop control system.
, , Introduction ~ 7
an amplifier, a mechanical linkage, a filter, or other control elements, depending on the nature of the system. In more sophisticated cases, the controller can be a computer such as a microprocessor. Because of the simplicity and economy of openloop control systems, we find tlus type of system in many noncritical applications.
114 Closedloop Control Systems (Feedback Control Systems)
What is missing in the openloop control system for more accurate and more adaptive control is a link or feedback from the output to the input of the system. To obtain more accurate control, the controlled signal y should be fed back and compared with the reference input, and an actuating signal proportional to the difference of the input and the output must be sent through the system to correct the error. A system with one or more feedback paths such as that just described is called a closedloop system. A closedloop idlespeed control system is shown in Fig. 17. The reference input Wr • Closedloop systems have many advantages over open sets the desired idling speed. The engine speed at idle should agree with the reference value loop systems. w,., and any difference slIch as the load torgue TL is sensed by the speed transducer and the error detector. The controller will operate on the difference and provide a signal to adjust the throttle angle a to correct the enor. Fig. 18 compares the typical performances of openloop and closedloop idlespeed control systems. In Fig. I8(a), the idle speed of the openloop system will drop and settle at a lower value after a load torque is applied. Tn Fig. 18 (b), the idle speed of the closedloop system is shown to recover quickly to the preset value after the application of TL . The objective of the idlespeed control system illustrated, also known as a r egulator system, is to maintain the system output at a prescribed level.
Error de tector
+
CONTROLLER
+
~~TRANSDUCER ~~
SPEED
Figure 17 Block diagram of a closedloop idlespeed control system.
1
Desired idle speed
Application of TL Desired idle s peed Time
OJ,
1
Appl ic'llion of TL Time
w,.
(a)
(b)
Figure 18 (a) Typical response of the openloop idlespeed control system . (b) Typical response of the closedloop idlespeed control system.
8
Chapter 1. Introduction
12 WHAT IS FEEDBACK, AND WHAT ARE ITS EFFECTS?
The motivation for using feedback, as ill ustrated by the examples in Section II. is somewhat overSimplified. In these examples, feedback is used to reduce the error between the reference input and the system output. However, the significance of the effects of feedback in control systems is more complex than is demonstrated by these simple examples. The reduction of system error is merely one of the many important effects that feedback may have upon a system. We show in the following sections that feedback also nas effects on such system performance characteristics as stability, bandwidth, overall gain, impedance, and sensitivity. To understand the effects of feedback on a control system, it is essential to examine this phenomenon in a broad sense. When feedback is deliberately introduced for the purpose of control, its existence is easily identified. However, there are numerous situations where a physical system that we recognize as an inherently nonfeedback system turns out to have feedback when it is observed in a certain manner. In general, we can state that whenever a closed sequence of causeandeffect relationships exists among the variables of a system, feedback is said to exist. This viewpoint will inevitably admit feedback in a large number of systems that ordinarily would be identified as non feedback systems. However, controlsystem theory allows numerous systems, with or without physi cal feedback, to be studied in a systematic way once the existence of feedback in the sense mentioned previously is established. We shall now investigate the effects of feedback on the various aspects of system performance. Without the necessary mathematical fou ndation of linearsystem theory. at this point we can rely only on simple staticsystem notation for our discussion. Let us consi der the simple feedback system configuration shown in Fig. 19, where r is the input signal ; y, the output signal ; e, the error; and b, the feedback signal. The parameters G and H may be considered as constant gains. By simple algebraic manipul ations, it is simple to show that the inputoutput relation of the system is
• Feedback cxists whencver there is a dosed sequence of causcanueffect rel atio nships.
M =~ = _ G_ _
r
1 +GH
(11 )
Using this basic relationship of the feedback system structure, we can uncover some of the significant effects of feedback .
121
Effect of Feedback on Overall Gain
As seen from Eq . (1  1), feedback affects the gain G of a nonfeedback system by a factor of 1 + GH. The system of Fig. 19 is said to have negative feedback. because a minus sign is assigned to the feedback signaL The quantity GH may itself include a minus sign, so the general effect ojjeedback is that it may increase or decrease the gain G. In a practical control system, G and H are functions of frequency, so the magnitude of 1 + GH may be
+
• Feedback may increase the gain of a system in one frequency nUlge hut ucc rease it in another.
~
+
e
~
+
r
0
G
~
y
.ro.
+b '
l
J
H
~
Figure 19 Feedback system.
12 What Is Feedback, and What Are Its Effects?  9
greater than I in one frequency range but less than 1 in another. Therefore,feedback could increase the gain of system in one frequency range but decrease it ill another.
122 Effect of Feedback on Stabilitv
• A system is unstable if its Stability is a notion that describes whether the system will be able to follow the input output is out of control. command, that is, be useful in general. In a nonrigorous manner, a system. is said to be unstahle if its output is out of control. To investigate the effect of feedback on stability, we can again refer to the expression in Eq. (11). If GH = 1, the output of the system is infinite for any finite input, and the system is said to be unstable. Therefore, we may state that feedback can cause a system that is originally stable to become unstable. Certainly, feedback is a doubleedged sword; when it is improperly used, it can be harmful. It should be pointed out, however, that we are only dealing with the static case here, and, in general, GH = 1 is not the only condition for instability. The subject of system stability will be treated formally in Chapters 2, 5, 7, and 8. It can be demonstrated that one of the advantages of incorporating feedback is that it can stabilize an unstable system. Let us assume that the feedback system in Fig. 19 is unstable because GH =  1. If we introduce another feedback loop through a negative feedback gain of F, as shown in Fig. 1 I 0, the inputoutput relation of the overall system is
y r
• Feedback can improve stability or be harmful to stability.
G
1 + GH +GF
(I 2 )
It is apparent that although the properties of G and H are such that the innerloop feedback system is unstable, because GH =  1, the overall system can be stable by properly selecting the outerloop feedback gain F. In practice, GH is a function of frequency, and the stability condition of the closedloop system depends on the magnitude and phase of GH. The bottom line is thatfeedhack can improve s tahilit.v or he harmful to stability if it is not pruper/y applied. Sensitivity considerations often are important in the design of control systems. Becau~e all physical elements have properties that change with environment and age, we cannot always consider the parameters of a control system to be completely stationary over the entire operating life of the system. For instance, the winding resistance of an electric motor changes as the temperature of the motor rises during operation. Control systems with electric components may not operate normally when first turned on because
r
0
e
+
+

+
G
y
~

b
+
= +1
~
1
H
,..
F
,..
Figure 110 Feedback system with two feedback loops.
10
~
Chapter 1. Introduction
of the stillchanging system parameters during warmup. This phenomenon is sometimes called "moming sickness." Most duplicating machines have a warmup period during which time operation is blocked out when first turned on. In general, a good control system should be very insensitive to parameter variations but sensitive to the input commands. We shall investigate what effect feedback has on sensitivity to parameter variation s. RefelTing (0 the system in Fig. 19, we consider G to be a gain parameter that may vary. The sensitivity of the gain of the overall system M to the variation in G is defined as M 8MI M percentage change in M So = oGIG = percentage change in G (13)
• Note: Feedback can increase or decrease the sensitivity of a system.
where oM denotes the incrementa} change in M due to the incremental change in G, or oG. By using Eg. (11), the sensitivity function is written M 8M G 1 So = oG M = 1 + GH (14) This relation shows that if GH is a positive constant, the magnitude of the sensitivity function can be made arbitrarily small by increasing GH, provided that the system remains stable. It is apparent that, in an openloop system, the gain of the system will respond in a onetoone fashion to the variation in G (i.e., S~ = 1). Again, in practice, GH is a function of frequency; the magnitude of 1 + GH may be less than unity over some frequency ranges, so feedback could be harmful to the sensitivity to parameter variations in certain cases. In general, the sensitivity of the system gain of a feedback system to parameter variations depends on where the parameter is located. The reader can derive the sensitivity of the system in Fig. 19 due to the variation of H.
123 Effect of Feedback on External Disturbance or Noise
All physical systems are subject to some types of extraneous signals or noise during operation. Examples of these signals are thermalnoise voltage in electronic circuits and brush or commutator noise in electric motors. External disturbances, such as wind gusts acting on an antenna, are also quite common in control systems. Therefore, control systems should be designed so that they are insensitive to noise and disturbances and sensitive to input commands. The effect of feedback on noise and disturbance depends greatly on where these extraneous signals occur in the system. No general conclusions can be reached, but in many situations, feedback can reduce the effect of noise and disturbance on system peliormance. Let us refer to the system shown in Fig. 1 11 , in which r denotes the command
• Feedback can reduce the effect of noise.
+
II
+
r
0
_J
e 
I
+
e2 
+

G,
+

e,
+
G2
y
~
b
+
H
Figure 111 Feedback system with a noise signal.
13 Types of Feedback Control Systems ~ 11
signal and n is the noise signal. In the absence of feedback, that is, H = O. the output y due to n acting alone is
(15)
With the presence of feedback, the system output due to n acting alone is
y=
• Feedback also can affect bandwidth. impedance, transient responses, and frequency responses.
Gz n I + GIGzH
(16)
Comparing Eq. (16) with Eq. (15) shows that the noise component in the output of Eq. (16) is reduced by the factor 1 + G 1GzH if the latter is greater than unity and the system is kept stable. In Chapter 9, the feedforward and forward controller configurations are used along with feedback to reduce the effects of disturbance and noise inputs. In general. feedback also has effects on such performance characteristics as bandwidth, impedance, transient response, and frequency response. These effects will be explained as we continue.
.. 13 TYPES OF FEEDBACK CONTROL SYSTEMS
Feedback control systems may be classified in a number of ways, depending upon the purpose of the classification. For instance, according to the method of analysis and design~ control systems are classified as linear or nonlinear, and timevarying or timeinvariant According to the types of signal found in the system. reference is often made to continuousdata or discretedata systems, and modulated or unmodulated systems. Control systems are often classified according to the main purpose of the system. For instance, a positioncontrol system and a velocitycontrol system control the output variables just as the names imply. In Chapter 9, the type of control system is defined according to the fonn of the openloop transfer function. In general, there are many other ways of identifying control systems according to some special features of the system. It is important to know some of the more common ways of classifying control systems before embarking on the analysis and design of these systems.
131 Linear versus Nonlinear Control Systems
• Most reallife control systems have nonlinear characteristics to some extent.
This classification is made according to the methods of analysis and design. Strictly speaking, linear systems do not exist in practice, because all physical systems are nonlinear to some extent. Linear feedback control systems are idealized models fabricated by the analyst purely for the simplicity of analysis and design. When the magnitudes of signals in a control system are limited to ranges in which system components exhibit linear characteristics (i.e., the principle of superposition applies), the system is essentially linear. But when the magnitudes of signals are extended beyond the range of the linear operation, depending on the severity of the nonlinearity, the system should no longer be considered linear. For instance, amplifiers used in control systems often exhibit a saturation effect when their input signals become large; the magnetic field of a motor usuany has saturation properties. Other common nonlinear effects found in control systems are the backlash or dead play between coupled gear members. nonlinear spring characteristics, nonlinear friction force or torque between moving members. and so on. Quite often, nonlinear characteristics are intentionally introduced in a control system to improve its performance
12 ... Chapter 1. Introduction
or provide more effective control. For instance, to achieve minimumtime control, an 00off (bangbang or relay) type controller is used in many missile or spacecraft control systems. Typically in these systems, jets are mounted on the sides of the vehicle to provide reaction torque for attitude control. These jets are often controlled in a fullon or fulloff fashion, so a fixed amount of air is applied from a given jet for a certain time period to control the attitude of the space vehicle. • There arc no general For linear systems, a wealth of analytical and graphical techniques is available for methods for solving a wide design and analysis purposes. A majority of the material in this text is devoted to the class of nonlinear systems. analysis and design of linear systems. Nonlinear systems, on the other hand, are usually difficult to treat mathematically, and there are no general methods available for solving a wide class of nonlinear systems. It is practical to first design the controller based on the linearsystem model by neglecting the nonlinearities of the system. The designed controller is then applied to the nonlinear system model for evaluation or redesign by computer simulation. The Virtual Lab introduced in Chapter 6 is mainly used to model the characteristics of practical systems with realistic physical components.
132 TimeInvariant versus TimeVarying Systems
When the parameters of a control system are stationary with respect to time during the operation of the system, the system is called a timeinvariant system. In practice. most physical systems contain elements that drift or vary with time. For example, the winding resistance of an electric motor will vary when the motor is first being excited and its
temperature is rising. Another example of a
time~varying
system is a guidedmissile
control system in which the mass of the missile decreases as the fuel on board is being consumed during flight. Although a timevarying system without nonlinearity is still a linear system, the analysis and design of this class of systems are usually much more complex than that of the linear timeinvariant systems.
Continuous.. Data Control Systems
A continuousdata system is one in which the signals at various parts of the system are all functions of the continuous time variable t. The signals in continuousdata systems may be further classified as ac or dc. Unlike the general definitions of ae and de signals used in electrical engineering, ac and de control systems carry special significance in control systems terminology. When one refers to an ac control system, it usually means that the signals in the system are modulated by some form of modulation scheme. A de control system, on the other hand, simply implies that the signals are unmodulated~ but they are still ac signals according to the conventional definition. The schematic diagram of a closedloop dc control system is shown in Fig. 112. Typical waveforms of the signals in response to a stepfunction input are shown in the figure. Typical components of a de control system are potentiometers, dc amplifiers, dc motors, de tachometers, and so on. Figure 113 shows the schematic diagram of a typical ac control system that performs essentially the same task as the dc system in Fig. 112. In this case, the signalS in the system are modulated; that is~ the information is transmitted by an ac carrier signal. Notice that the output controlled variable still behaves similarly to that of the dc system. In this case, the modulated signals are demodulated by the lowpass characteristics of the ac motor. Ac control systems are used extensively in aircraft and missile control systems in which noise and disturbance often create problems. By using modulated ac control systems with carrier frequcncies of 400 IIz or higher, the system will be less susceptible to lowfrequency noise. Typical components of an ac control system are synchros, ac amplifiers, ac motors, gyroscopes, accelerometers~ and so on.
13 Types of Feedback Control Systems
13
+
e
+
DC
AMPLIAER
+
+
=E
+
0,.
Controlled variable
Reference input
o
Errol' detector
o
Figure 112 Schematic diagJ'am of a typical dc closedloop system.
Synchro transmitter
Synchro control tra nsfonner
AC
AMPLIFIER
AC servomotor
o
Figure 113 Schematic diagram of a lypical ac closedloop control system.
In practice, not al l control systems are strictly of the ac or dc type. A system may incorporate a mixture of ac and dc components, using modulators and demodulators to match the signals at various points in the system.
DiscreteData Control Systems Discretedata control systems differ from the continuousdata systems in that the signals at one or more points of the system are in the form of either a pulse train or a digital code. Usually. discretedata control systems are subdi vided into sampleddata and digital control systems. Sampleddata control sys tem s refer to a more general class of
14 .. Chapter 1. Introduction
DATA HOLD (FILTER)
e*(i)
•
h(r)
CONTROLLED .1'(1) PROCESS rr''+
Figure 114 Block diagram of a sampleddata control system .
Digital coded
input
DIGITALTOANALOG CONVERTER ANALOGTODIGITAL CONVERTER
Attitude of
AIRFRAME
missile
SENSORS
Figure 115 Digital autopilot system for a guided missile.
• Digital control systems an.! lIsually less susceptible to noise.
discretedata systems in which the signals are in the form of pulse data. A digital control system refers to the use of a digital computer or controller in the system so that the signals are digitally coded, such as in binary code. In general, a sampleddata system receives data or information only intermittently at specific instants of time. For example, the error signal in a control system can be supplied only in the form of pulses, in which case the control system receives no information about the error signal during the periods between two consecutive pulses. Strictly, a sampleddata system can also be classified as an ac system, because the signal of the system is pulse modulated. Figure 114 illustrates how a typical sampleddata system operates. A continuousdata input signal r(t) is applied to the system. The error signal e(l) is sampled by a sampling device, the sampler, and the output of the sampler is a sequence of pulses. The sampling rate of the sampler mayor may not be uniform. There are many advantages to incorporating sampling into a control system. One important advantage is that expensive equipment used in the system may be timeshared among several control channels. Another advantage is that pulse data are usually less susceptible to noise. Because digital computers provide many advantages in size and flexibility, computer control has become increasingly popular in recent years. Many airborne systems contain digital controllers that can pack thousands of discrete elements into a space no larger than the size of this book. Figure 115 shows the basic elements of a digital autopilot for guidedmissile control.
14 SUMMARY
In this chapter, we introduced some of the basic concepts of what a control sy~tem is and what it is supposed to accomplish. The basic components of a control system were described. By demonstrating the effects offeedback in a rudimentary way, the question of why most control systems are cJosedloop systems was also clarified. Most important, it was poi nted out that feedback is a doubleedged sword it can benefit as well as hW'm the system to be controlled. This is part of the challenging task of designing a control system, which involves consideration of such performance criteria as stability,
Review Questions
~
15
sensitivity. bandwidth, and accuracy. Finally, various types of control systems were categorized according to the system signals, linearity. and control objectives. Several typical controlsystem examples were given to illustrate the analysis and design of control systems. Most systems
encountered in real life are nonlinear and timevarying to some extent. Tin:: cOIlcentration on the
studie..o;; of linear systems is due primarily to the availability of unified and simpletounderstand analytical methods in the analysis and design of linear systems.
~
REVIEW QUESTIONS
1. List the advantages and disadvantages of an openloop system.
2.
List the advantages and disadvantages of a closedloop system. Give the definitions of ac and de control systems. Give the advantages of a digital control system over a continuousdata control system.
A closedloop control system is usually more accurate than an openlOOp system. (T) (F)
(F)
3. 4.
5.
6.
Feedback is sometimes used to improve the sensitivity of a contr'ol
sy~tem.
(T)
7. If an openloop system is unstable. then applying feedback will always improve (T) its stability.
(F)
8. Feedback can increase the gain of a system in one frequency range but decrease it in another. (T)
9.
10.
(F) (F)
Nonlinear elements are sometimes intentionally introduced to a control system to improve its performance.
(T)
Discretedata conLrol systems are more susceptible to noise due to the nature of their signals. (T)
(F)
Answers to these review questions can be found on this book's companion Web site: www.wiley.comlcollege/golnaraghi.
CHAPTER
2
Mathematical Foundation
The studies of control systems rely to a great extent on applied mathematics. One of the major purposes of controlsystem studies is to develop a set of analytical tools so that the designer can arrive with reasonably predictable and reliable designs without depending solely on the drudgery of experimentation or extensive computer simulation. In this chapter, it is assumed that the reader has some level of familiarity with these concepts through earlier courses. Elementary matrix algebra is covered in Appendix A. Because of space limitations, as well as the fact that most subjects are considered as review material for the reader, the treatment of these mathematical subjects is not exhaustive. The reader who wishes to conduct an indepth study of any of these subjects should refer to books thac are devoted to them. The mai n objectives of this chapter are: 1. 2. 3. 5. 6. 7. 8. To introduce the fundnmentals of complex valiables. To introduce freq uency domain analysis ,md frequency plots. To introduce differential equations and state space systems. To demonstrate the applications of Laplace transforms to solve linear ordinary differential equations. To introduce the concept of transfer functions and how to apply them to the modeling of linear time invariant systems. To discuss stability of linear timeinvariant systems and the RouthHurwitz criterion.
To demonstrate the MATLAB tools using case studies.
4. To introduce the fundamentals of Laplace transforms.
21 COMPLEXVARIABLE CONCEPT
To underswnd complex variables, it is wise to start with the concept of complex numbers and their mathematical properties.
211
Complex Numbers
A complex number is represented in rectangular form as
z = x + jy
(21)
where, j = . =T and (x , y) are real and imaginary coefficients of;: respectively. We can J treat (x , y ) as a point in the Cartesian coordinate frame shown in Fig. 21. A point in a
16
2~1
ComplexwVariable Concept
~
17
Imaginary
splane
z=x+j...,.
R
       F     +  t     + . Real
,
I
\"
I I
R
... ____,' ~
z*=xjy
Figure 21 Complex number = representation· in rectangular and polar forms.
rectangular coordinate frame may also be defined by a vector R and an angle easy to see that
e. It is then
(22)
x = Reose
y
= RsinO
where. R
= magnitude of z
of z and is measured from the x axis. Righthand rule convention: positive phase is in counter clockwise direction. R=JX2+y2
() = phase
Hence,
() = tan 1 ~
Introducing Eq. (22) into Eq. (21), we get
(23)
x
z = Reose + jRsine
Upon comparison of Taylor series of the terms involved. it is easy to confirm
(24)
ejO = cos e+ j sin ()
(25)
Eq. (25) is also known as the Euler formula. As a result, Eq. (21) may also be represented in polar form as
z = Re j9 = R L8
We define the conjugate of the complex number z in Eq. (21) as
(26)
Or, alternatively,
z* =x jy
(27)
(28)
z* = Rcose  jR sine = R e j9
Note:
zz*
= R2 = Xl + y2
(29)
Table 21 shows basic mathematical properties of complex numbers.
18
r",o
Chapter 2. Mathematical Foundation
TABLE 2·1
Addition
Basic Properties of Complex Numbers
ZI { Z2

=x, + jYl = Xz + jY2
Z = (Xl
+ X2) + j(YI + )'2)
Subtraction
Zl = .tl + 1:V1 { Z2 = X2 + JYz z:=: (.tlX2)+j(YIYZ)
 z =R) elf:/(_ R2e jH2
Multiplicatio1l
Zl { Zz

l
Z
= (XIX2 =1
= XI + jYI = X2 + jyz
YIY2)
ZI { Zz
= Rle Ph jiJ2
= Rze
(RIRz)ej(BI+8z)
+ j(XIY2 + X2)'1 )
 z=
Z= (RIRZ)L(81 +(2)
Division
{
Zl
= XI + ~YI
+ JY2
jYl
Z2 = X2
zi = Xl * { z2 = X2 z, z=Z2
. Complex Conjugate
JYz
2= (~~)ej(ihe2)
z =
(~~) I (81 
82)
EXAMPLE 2·1·'
Find/ andj4.
J = VI
o
r7l
Jr . rr = COS "2 +. stn"2 = e ,o:r J ~
i
.3
=
vClJ=ivCT = vCl =  j
,o~
F=e2=e
_Joa
2
/=/)=/=1
L,
EXAMPLE 21"2 Find z" using Eq.
(26),
(210)
212 Complex Variables
A complex variable s has two components: a real component a and an imaginary component w. Graphically, the real component of s is represented by a (J axis in the horizontal direction, and the imaginary component is measured along the vertical jw axis, in the complex splane. Fig. 22 illustrates the complex splalle, in which any arbitrary point s = Sl is defined by the coordinates a = al, and Ct) = WI, or simply Sl = at +jWt·
2~ 1 Complex~Variable
Concept . 19
j{))
~~·a
Figure
Z~2
Complex
s~plane.
213 Functions of a Complex Variable
The function G(s) is said to be a function of the complex. variable s if. for every value of s.
there is one or more corresponding values of G(s). Because s is defined to have real and imaginary parts, the function G(s) is also represented by its real and imaginary parts; that is.
G(s)
= Re[G(s)] + j
lrn[G(s)]
(211)
where Re[G(s)] denotes the real patt of G(s), and Im[G(s)] represents the imaginary part of G(s). The function G(s) is also represented by the complex G(s)plane, with Re[G(s)] as the real axis and Im[G(s)] as the imaginary axis. If for every value of s there is only one corresponding value of G(s) in the G(s)plane. O(s) is said to be a single.. valued function. and the mapping from points in the splane onto points in the G(s)plane is described as single..valued (Fig. 23). If the mapping from the G(s)plane to the splane is also singlevalued. the mapping is called oneto .. one. However~ there are many functions for which the mapping from the function plane to the complexvariable plane is not singlevalued. For instance, given the function
G(s) = s(s
+ I)
1
(212)
jw
j ImG
splane
WI
~ __ _
G(s)pIane
o
... ...
ReG
I
  =':.....
G(sl)
Figure 23 Singlevalued mapping from the splane to the
G(s)~plane.
20 .. Chapter 2. Mathematical Foundation
it is apparent that, for each value of s, there is only one unique corresponding value for G(s). However, the inverse mapping is not true; for instance, the point G(s) = 00 is mapped onto two points, s ;;;;;: and s = 1, in the splane.
°
2~1~4
Analytic Function
A/unction O(s) of the complex variable s is called an analyticfunctionin a region of the splane if the function and all its detivatives exist in the region. For instance, the function
given in Eq. (212) is analytic at every point in the splane except at the points 8 = 0 and = 1. At these two points, the value of the function is infinite. As another example, the function G(s) = 8 + 2 is analytic at every point in the finite splane.
s
215 Singularities and Poles of a Function
The singularities of a function are the points in the splane at which the function or its derivatives do not exist. A pole is the most common type of singularity and plays a very important role in the studies of classical control theory. The definition of a pole can be stated as: If a function G(s) is analytic and single .. valued in the neighborhood ofpoint Ph it is said to have a pole of order r at s = Pi if the limit lim [(s  Pi)rO(S)] has afinite, nonzero value. In other words, the denominator of
G(s) must include the factor ($ s"""" Pi
If r
= 1, the pole at s =
PiY. so when s = Pi. the function becomes infinite. Pi is calIed a simple pole. As an example, the function
G(s) =
IO(s + 2) s(s + 1)(8 + 3)2
(213)
has a pole of order 2 at s = 3 and simple poles at s == 0 and s = 1. It can also be said that the function G(s) is analytic in the splane except at these poles. See Fig. 24 for the graphical representation of the finite poles of the system.
21 ..6 Zeros of a Function
The definition of a zero of a function can be stated as: If the junction G(s) is analytic at S = Zh it is said to have a zero of order r at s = Zi if the limit lim [(s  Zi)rG(S)J has a
finite. nonzero value. Or, simply, O(s) has a zero oj order r at S = Zj if l/G(s) has an rthorder pole at s = Zi. For example, the function in Eq. (213) has a simple zero at s = 2. If the function under consideration is a rational function of s. that is, a quotient of two polynomials of s, the total number of poles equals the total number of zeros, counting the
multipleorder poles and zeros and taking into account the poles and zeros at infinity. The function in Eq. (213) has four finite poles at s = 0,  I,  3, and 3; there is one finite zero at s = 2, but there are three zeros at infinity, because
s 10 Zi
S
lim G(s)
t
=
(X)
S ..... 00
lim 1? = 0 s·
(214)
Therefore, the function has a total of four poles and four zeros in the entire splane, including infinity. See Fig. 24 for the graphical representation of the finite zeros of the system.
.21 ComplexVariable Concept . use "zpk" to create zeropolegain models by the following sequence of MATLAB functions »G = zpk( [2].. Toolbox 2. [0 1 3 3] r 10) Alternatively use: » s »clear all =tfC J S r).. (213).. 21 jro splane 3 2 1 () 1 2 s(s+ 1)(s+3f Figure 24 Graphicru representation of G(s) = 10(..polegain form » pole (Gp) ans = » Gzpk ~ zpk(Gp) Zero/pole/gain: 10 Cs + 2) S(8+3)"2(s+1) o 1 3 3 » zero(Gp) ans = 2 .5+2) in the splane: x poles and 0 zeros. 1 For Eq. 1. »Gp = 10*(8 + 2)/(S*(9 + 1)*(8 + 3)"2) Zero/pole/gain: 10 (s + 2) 8(s+1)(8+3)"2 Transfer function: 10 s + 20 Convert the transfer function to polynomial fonn »Gp = tf(G) Transfer function: 10 s + 20 s"4 + 7 s"3 + 15 s"2 + 9 8 Use "polen and "zero" to obtain the poles and zeros of the transfer function Convert the transfer function Gp to zero .
jco+ 2= \ / 2 2 + co2 «?#' co = R\ sin c\>\ 2 = R\co$<p\ R{ = \ / 2 2 + co2 0 . G W 16 .22 Chapter 2. .splane 5. where co is a constant varying from v zero to infinity. Thus. . = tan JO) (219) (220) (221) (222) (223) — . = 1 +J2 R= \/\+2 2 )=tan' 1  Figure 25 Graphical representation of S\ = 2j + I in the . Mathematical Foundation 217 Polar Representation To find the polar representation of G(s) in Eq.vplane.v + 10.46 0 = tan" 1 .f 1 = \/5 1 2 0. That is G(s) = 1 (215) s = 2j = Rej0 = 2ej% s + 1 .v + 2)(. (218) at s = jco. 25 for a graphical representation of s\ = 2.. \ 1 1 G{2j) 1 = l e =2J(2jTT) 2 v5' (217) 2\/5 See Fig. we look at individual components..v + 8) (218) To evaluate Eq.= 1.v+16 2 16 (.. we look at individual components.2j + 1 = R ej$ R = ^ 2 2 .vplane. V . (212) at s — 2j.11 rad{(216) 26.57 63. EXAMPLE 213 Find the polar representation of G{s) given below for „ = jco.43 c 1 _ ../ + 1 in the .
= rpl .W '. . 8) wJ+~ wJ'" jw+ 2 = R1(jsinl/J1 + cosl/Jd (224) (225) (226) jw + 2 = Rleii/J. . 26 for a graphIcal representation of components Hence. The phase goes from 0° to 180°. ~ (232) Table 22 describes different Rand rp values as (V changes. 23 j(J) splane (d tPl = tan1 ~ 0' 2 Figure 26 Graphical representation of components of l . we can define l/J = tan _IImG(jw) . + 2)( wj + 8) 16 1 jw+2 jw + 8 = )22 +#eNr ====J 82 + (J)2 ejth (229) = As a result.. ReG{jw) = LG(s = JW) .21 ComplexVariable Concept. As shown.rp. G(s = jOJ) becomes: (230) where Similarly. the magnitude decreases as the frequency increases. jtt) + 8 = V8 2 + w2 eflth tP2 = tan 16 1 W/R2 8/R2 (227) (228) 0 == 16ell f (wj See Fig. .
1 I R ¢ 3.24 . .123 0. ( .0016 IO 100 Alternative Approach: If we multiply both numerator and denominator of Eq.58 33. .. Mathematical Foundation TABLE 22 Numerical Values of Sample Magnitude and Phase of the System in Example 213 wrad/s 0. Chapter 2. 2~~ .28 0.ti)  jIOw] = Real + Imaginary (233) = 16V(16 .I.999 0. the frequency response can be determined graphically.jw + 8) 1 2)(' 8):=. (218) by the .03 174.tan 16 R =. I.w2)2+(IOw)2 e (tif + 4)(cv2 + 64) 16 iiP = J(Wl = Re jrjJ _ 1 + 4)(w2 + 64) el'l' '. I \ ¢ cr _ • _ 1 lOlO/R (16 __ (02)/R q). (wJ + wJ + So as you have noticed.jw + 2)( . .888 0. Consider the following second order system: K (234) G(s) = (S+PI)(S+P2) j(Q . f h d . ( . 2~~' Wj+ Wj+8) for a fixed value of w.4)((Q2+ 64) Figure 21 Graphical representation of ( .::======' VUi}2 I.LG(JW.' we get comp Iex conjugate 0 t e enommator. 27 for a graphical representation of.tan Im(G(jw)) Re(G(jw» See Fig. Jw+ Jw+ G( 'w) J = 16(jw+2}(jw+8) (w2 + 22)(w2 + 82) = (w2 +4)1(~ + 64) [(16 .69 130. . JOw/R (16 __ ( 2 )/R where t/> .""~'" . 8) for a fixed value of w.e.
. Z. OOOOi »R = abs(Z) R= 3...21 ComplexVariable Concept . OOOOi »ZC = conj (Z) ZC== 3.2. in radians. 2) Z= 3.. conjugate of the elements of Z X == real (Z) returns the real part of the elements of the complex array Z y= imag CZ) returns the imaginary part of the elements of array Z R == abs (2) returns the complex modulus (magnitude). from the two real inputs Z ZC = conj (Z) = a + hi returns the complex.5880 »ZRT.2 Here are MATLAB commands to treat complex variables: Z == complex (a b) I creates a complex output. OOOOi . R.0000 + 2. *exp(i*theta) converts back to the original complex Z » Z = complexe3. "2 + imag(Z) . *exp(i*theta) ZRT= 3.. for each element of complex array Z The angles lie between the "real axis" in the splane and the magnitude R Z = R. 25 Toolbox 21 .:::.1\2) theta = angleeZ) returns the phase angles.6056 » theta = angle(Z) theta = 0. which is the same as R == sqrt(real(Z). 0000 . 0000 + 2.
if s = jw.26 Chapter 2. 1. O( j€V) is the frequency response function of O(s). where ! jll> = 4>1 . I For the formal definition of a "transfer function. Polar plot. we use MATLAB and the ACSYS software for this purpose. The function G( jw) is generally a complex function of the frequency wand can be written as G(jw) = IG(jw)ILG(jw) (240) where IG( jw) I denotes the magnitude of G( jw).. and ! G( jru) is the phase of G( jw). The frequencydomain analysis of the closedloop system can be conducted from the frequencydomain plots of G(s) with s replaced by jru.I jm + P2. By definition. Bode plot.. + VIm) . A plot of the magnitude in decibels versus w (or IoglOw) in semilog (or rectangular) coordinates Magnitudewphase plot. 3." refer to Section 272.(c/J] + ..k'=""". The following frequencydomain plots of G( jw) versus (J) are often used in the analysis and design of linear control systems in the frequency domain.1 l~_ _ (238) I)s + Pi) i=l The magnitude and phase of G(s) are as follows R = 1GUru) 1 =K Ijw+zI1"'1jw+zm[ 1Jw + pI!'" Jjw + Pnl ¢ = /G(jw) = (VII + .. +c/J. . 1 jm + Pill JW + P21 + 1'1 . In this textbook. A plot of the magnitude (in decibels) versus the phase on rectangular coordinates..) 22 FREQUENCYDOMAIN PLOTS Let G(s) be the forwardpath transfer function I of a linear control system with unity feedback. Mathematical Foundation where (PI) and (pz) are poles of the function G(s)... K.. because €V has a unit of frequency (rad/s): K G(s) = (2~35) (jm + pt)(jm + 112) The magnitude of O( jw) is R = 1G(jw)1 and the phase angle of G( jm) is =... especially if the function is of high order. A plot of the magnitude versus phase in the polar coordinates as w is varied from zero to infinity 2.¢2 11l 2:)s + Zk) O(s) = K . = / G( j(d) = / K For the general case.. (2~36) q. with (vas a variable parameter on the curve 2~2~1 ComputerAided Construction of the Frequency~Domain Plots The data for the pJotting of the fre{)uencydomain plots are usually quite time consuming to generate if the computation is carried out manually.
1 + T. and the phase reaches 900 • This is presented by a vector with an infinitesimally small length directed along the 90°axis in the G( jw)wplane. the exact plot of G( jw) turns out to be a semicircle.f = jw. Thus. For any frequency w = WI. the magnitude of G( jw) decreases~ and the phase becomes more negative. the process can be regarded as the mapping of the positive half of the imaginary axis of the splane onto the G( jw)plane.jclJ:. j~ ~ 27 jImG GUCLI)plane jill) 0 (j ReG Figure 28 Polar plot shown as a mapping of the positive half of the jwaxis in the splane onto the G(jw)plunc. (243). counterclockwise is referred to as positive. the length of the vector in the polar coordinates decreases and the vector rotates in the clockwise (negative) direction.y where T is a positive constant. domain plots so that proper interpretations can be made on these computergenerated plots. By substituting other finite values of w into Eq.22 FrequencyDomain Plots jaJ splane . the magnitude of G( jw) becomes zero. 1 L tanI wT In terms of magnitude and phase. Setting . 29. the analyst and designer should be familiar with the properties of the frequency . 28. . at w = 0.. 222 Polar Plots The polar plot of a function of the complex variable s. and clockwise is negative. consider the function (241) = 1 +JwT 1..c) is zero. In measuring the phase . the magnitude and phase of G( jwd are represented by a vector in the G( jw)plane. G( jw) is represented by a vector of unit length directed in the 0° direction. When w approaches infinity. From an analytical standpoint. we have G( jw) G(s). EXAMPLE 221 To illustrate the construction of the polar plot of a function 1 G(s) = . A simple example of this mapping is shown in Fig. As w increases. G(s). From a mathematical viewpoint. (242) is rewritten as G(jw) = Jl + w 2T2 (243) When (. the magnitude of G(jw) is unity. As lJ) increases. as shown in Fig. and the phase of G(jw) is at 0°. is a plot of the magnitude of G( jw) versus the phase of G( jw) on polar coordinates as w is varied from zero to infinity. Eq.
(1 Figure 210 Polar plots of G( JW) = (1 + jwT2) + jwTl r . 2. ReG .28 ~ Chapter 2. in this case. 210. 1. consider the function G( 'w) = 1 + jwTz J 1 + jwTl where TI and Tz arc positive real constants. Mathematical Foundation jlmG Figure 2 9 Polar plot of G{jw) M = (1+}WTf ~ EXAMPLE 222 As a second illustrative example.plane O~. Eq. depends on the relative magnitudes ofTt and T2• If Tz is greater than Tit the magnitude of G(jw) is always greater than unity as (f) is varied from zero to infinity. and the phase of G( jw) is always positive. The general shape of the polar plot of a function G( jw) can be determined from the following information. The intersections of the polar plot with the real and imaginary axes. (245) that correspond to these two conditions are shown in Fig. If T2 is less than Tl ~ the magnitude of G( jw) is always less than unity.~r~. and the values of w at these intersections jlmG GUco). and the phase is always negative. The behavior of the magnitude and phase of G( jw) at w = 0 and (J) = 00. The polar plots of G( jw) of Eq. (244) is rewritten as (244) G( jw) = The polar plot of G( jw}.
5 o 0....5~~~~~~~~~ 00 to + oc.§ 0.5 1.~_.2 ~ ~ 0.... .22 FrequencyDomain Plots . it is clear that the polar plot reflects only a portion of the Nyquist diagram.. den! = [T1l] . numl = [T2 1J .. Often.3 0..2 0... G2 = tf (num2 . where (i) is varying from Nyquist diagram of G1 and 02 O.l'_ _ _.:.. an exact plot of the frequency response is not essential..4 0..1 ~ 0.. such as the Nyquist stability criterion (see Chapter 8). = hold on.denl). _ __ _ ' 0.. functions: Tl = 10..5 2 Comparing the results in Toolbox 221 and Fig.. num2 = [T1l] .!iii ~ 0 ..I.. (2~44) for two cases is obtained by the following sequence of MATLAB T2 5.4 0. Gl = tf(numl. a rough sketch of the polar plot of the transfer function is adequate for stability analysis in the frequency domain.. titIe ( I Nyquist diagram of Gl and G2 I ) Note: The "nyquist" jime/ion provides a complete polar diagram. 0.3 0.5 "_ _ _ 1 L. 210. den2) . nyquist(Gl)......_ _ .1( 29 Toolbox 221 The Nyquist diagram for Eq.1 ._ _ _ _...._ _ _ . nyquist (G2)... den2 == [T2 1] ..L. In many controlsystem applications.
at the point of intersection. Next. EXAMPLE 223 In frequencydomain analyses of control systems. we are interested in the properties of the G(jw) at infinity. often we have to detenrune the basic properties of a polar plot. we determine the intersections. the magnitude and phac.. Mathematical Foundation }. Similarly. Consider the following transfer function: 10 G(s) = s(s + 1) (246) By substituting s = jw in Eq. Chapter 2.jw)( .. the properties ofthepolarplotofG( jw) atw = Oandw = 00 are ascertained. Therefore. From Eq. meaning that the G( jw) plot intersects only with the real axis of the G( jw)plane at the origin. as shown in Fig...~ ( Jev) . G( jto) is written G' _ lO(jw)(jw+ 1) _ IOw2 .j(Q + 1) (04 + w2 J (04 + (02 (252) = Re[G(jev)] + jlmIG( jw)] When we set Irn[G( jev)] to zero. (246). which corresponds to ev = 0 in this case. the imaginary part of G( jw) is zero. GUID)plane jImG o ReG Figure 211 Polar plot of G(s) = .. orthe polar plot with the two axes of the G( jev }plane. which corresponds to the origin of the G( jw)plane. The only real solution for (J) is also w = 00. (2~52) to Zero. the intersection of G(jev) with the imaginary axis is found by setting Re[G(jev)] of Eq. Based on this information as well as knowledge of the angles of G( j(t» at ltJ ::::: 0 and w = 00. the polar plot of G( jw) is easily sketched without actual plotting. we see that Im[G( jw)J = 00 and Re [G( jw)] = 10 at w = O..~(l~tr . Under certain conditions. if any. oc LGUro) = W+oc lim LIO/Uw)2= 180° Thus. = W > 00 lim IG( j{J»1 lim (J)~ 1~ = 0 lim w. If the polar plot of G( jev) intersects the real axis.~ . (252).>0 lim IG( jw)! = lim 10 = w>Ow 00 0 (247) (248) (249) (250) lim LG(jw) W + 00 = wlimO LIOI j{J) = 90 . 211.30 . we must rationalize G( jw) by multiplying its numerator and denominator by the complex conjugate of its denominator. ImIG( jlV}] =0 (251) To express G( jw) as the sum of its real and imaginary parts.e of G( j{J») at (J) = 0 and w = 00 are computed as follows: w+O ... The conclusion is that the polar plot of G( j(o) does not intersect anyone of the axes at any finite nonzero frequency.jw(jw + 1)( . that is. we get w = 00.
the polar plot of G( jw) for negative values of (i) is the mirror image of that for positive w. 30 jlO(2 . lim IG(jw)\ = lim 00 W' 1~ = 0 (256) To find the intersections of the G( jw) plot on the real and imaginary axes of the G( jcv)plane. From Eq.2~2 Frequency~Domain Plots • 31 . has no physical meaning. This gives the point of intersection on the real axis at = G( ±jVZ) = 5/3 (259) The result. We shall show in the next section that approximate information on the polar plot can always be obtained from the Bode plot. we also see that Re[G( jO)] co and hu[G( jO)] = 00.lL)Z) 9m2 (258) Setting ReIG(jw)] to zero. Setting 1m [G(jw)J to zero. (253). it simply represents a mapping point on the negative jwaxis of the s. it is now possible to make a sketch of the polarplotforthe transfer fUIlctiun in Eq. the last equation is written G(jw) = Re[G(jw)J + jIm[G(jw)] :::. it is somewhat awkward for design purposes. plane. with the mirror placed on the real axis of the G( jw )~plane. (258).. In general.:: 00 (2"54) coo lim /G( jW) = liIll 151jw = 900 coo llJ+ (255) co+::x.:~''=" + (2 . The following calculations are made for the properties of the magnitude and phase of G(jw) at (tJ := 0 and (0 = 00: co+O lim IO( jw}l:::. and G(joo) == 0.((. which can be sketched = = ReG Figure 212 Polar plot of G(s) = 10). Furthermore.. lim co+Ow ~ . because the frequency is negative. 212.jw + 2) 2 . With this information. . Although the method of obtaining the rough sketch of the polar plot of a transfer function as described is quite straightforward. as shown in Fig. we have CJ) ±:v/2 rad/sec. we have w = 00. s(s + l)(s + 2) 10 (253) we want to make a rough sketch of the polar plot of G( jw). the polar plot is basically a tool for analysis. the algebraic manipUlation may again be quite involved. in general. S{Hl)(s+2 . for complicated transfer functions that may have multiple crossings on the real and imaginary axeS of the transfer~function plane.. ) = (jw IO(jw)(jw+ 1)(jw+2) jcu(jw + 1)( jw + 2)( jw)( . which means that the G(jw) plot intersects the imaginary axis only at the origin.. we rationalize G( jcu) to give G .jw + 1)( .wZ)2 (257) After simplification.12) 90)3 + w(2 . w /2 rad/sec. EXAMPLE 224 Given the transfer function O(s) :::. if G(s) is a rational function of s (a quotient of two polynomials of s).
because the poles and zeros of G(s) are easily identified. such as the polar plot and the magnitudeversusphase plot. Tio T2 . sketches of the polar plots can be obtained with the help of the Bode plots. For constructing the Bode plot manually. Consider the function G(s) = ~(s + Zl)(S + Z2)' •• sJ(s + PI)(S + P2)' . G(s) is preferably written in the following form: G(s) = K~ (I + TIs)(l + T2S) . T'l' ~. Tc/.20log lO II + jwTal. for more complicated transfer functions.20 log 10 Ijwl. we have IG(jw)ldB = 20Iog w IG(jw)1 = 2010g IO IKI + 20log lO 11 + jwTti + 20 loglOlI + jmT21 . If the Bode plot is to be constructed with a computer program. which allow the simple sketching ofthe plot without detailed computation. and Wn are real constants.20log lO il + j2t. and the z's and the p's may be real or complex (in conjugate pairs) numbers. unless MATLAB is used.32 Chapter 2. then either form of Eq. then without loss of generality. (s (s + Zm) e. Because the straightline approximation of the Bode plot is relatively easy to construct.. G(s) = K(I $( I + Tas) (I + TIs)(l + T2S) eht + 2~s / w" + s2 / m~) (262) where K.. In simple tenns. 2~2~3 Bode Plot (Corner Plot or Asymptotic Plot) The Bode plot of the function G( jrv) lS composed of two plots. the data necessary for the other frequencydomain plots. can be easily generated from the Bode plot. (261) are of the same fonn. the Ts may be real or complex (in conjugate pairs) numbers. It is assumed that the secondorder polynomial in the denominator has complexconjugate zeros. Because practically all the terms in Eq. (260) is the preferred form for rootlocus construction... Because the magnitude of G( jm) in the Bode plot is expressed in dB. These names stem from the fact that the Bode plot can be constructed by using straightline approximations that are asymptotic to the actual plot. Thus. (260) or Eq. The phase relations are also added and subtracted from each other algebraically. A Bode plot is also known as a corner plot or an asymptotic plot of G( jw). we can use the following transfer function to illustrate the construction of the Bode diagram. (1 + Tm s ) esJ(1 + Tas)(l + TbS) . the Bode plot has the following features: 1. In Chapter 7. and Td is the real time delay. (1 + ~lS) TdS (261) where KI is a real constant. 2. Eq.w  w2 /m~1 (263) . respectively. one with the ampJitude of G(jm) in decibels (dB) versus loglOm or m and the other with the phase of G(jw) in degrees as a function 'of loglOUJ or w. Mathematical Foundation without any calculations. (261) can be used. product and division factors in G( jw) became additions and subtractions. The magnitude of G( jm) in dB is obtained by multiplying the logarithm (base 10) of IG(jw)1 by 20.. The magnitude plot of the Bode plot of G( jw) can be approximated by straightline segments.TdS + fJll) where K and Tel are real constants.
m "'C '' G) 40 '13 =e O'l :E ~ os 60 00 100 0 45 't5 .. G = tf(num.. the function G( jw) may be of higher order than that of Eq.. G=16!Cs"2 + 10*8 + 16) .22 FrequencyDomain Plots ~ 33 The phase of G( jw) is LG(jw) = LK + L(1 . G) (. bode(G). Approach 1 num= [16].. . Bode Diagram 0 20 .180 101 10° 10 1 10 2 10 3 . Compare the results to the values in Table 22. The "bode" function computes the magnitude and phase of the frequency response of linear time invariant models. is obtained by the following sequence of MATLAB functions. (263) and (264) indicate that additional terms in G( jro) would simply produce more similar terms in the magnitude and phase expressions..... However.. so the basic method of construction of the Bode plot would be Toolbox 222 The Bode plot for Example 213. bode(G). den=[11016].. Eqs.t:: a. den) ..L(1 fad + jwTa) (264) + 2tw/wn   WTd In general. The magnitude is plotted in decibels (dB) and the phase in degrees..L(1 + jwTt) + L(1 + jwT2) {J} /(t)~) Ljro . (262) and have many more factored terms.) Oi (I) IIJ!l 90 . Approach 2 s = tf(' s')..135 . using the MATLAB "bode" function.
Chapter 2. G(jw) can contain just five simple types of factors: 1. (267) with respect to 10gtoW. (263) and (264) verify one of the unique characteristics of the Bode plot in that each of the five types of factors listed can be considered as a separate plot. and so on. and r are positive integers Eqs. a unit change in log toW corresponds to a change of ±20 p dB in magnitude. dl d oglOw (±20ploglOW) = ±20p dB/decade (268) These lines pass through the OdB axis at W = 1. Poles or zeros at s = 1fT of order q: (1 + jwT)±q 4. a unit change in log lOW in the rectangular coordinates is equivalent to one decade of variation in w. We have also indicated that. the slopes of the straight lines described by Eq.jwTd • where T(t. . Because KdB = 20 10g1O K = constant (265) and OO LK = { 180 0 K>O K <0 (266) the Bode plot of the real constant K is shown in Fig.34 . (jw)± P The magnitude of (jw)±p in dB is given by (267) for W 2:: O. Pure time delay e. 225 Poles and Zeros at the Origin. The slopes of these lines are detennined by taking the derivative of Eq. Thus. Complex poles and zeros of order r: (1 + j2~w/wn . Mathematical Foundation the same. p. 213 in semilog coordinates. Constant factor: K 2. Thus. depending on whether w or log lOw is used as the abscissa. q. We shall now investigate sketching the Bode plot of different types of factors. that is. (268) are said to be ±20 p dB/decade of frequency. from 1 to 10. The curves can be plotted on semilog graph paper or linear rectangularcoordinate graph paper.w2 /w~)±r 5. that is. Furthermore. the individual plots are then added or subtracted accordingly to yield the total magnitude in dB and the phase plot of G(jw). Poles or zeros at the origin of order p: (jw)±P 3.. in the semilog coordinates. 10 to 100. The last expression for a given p represents a straight line in either semilog or rectangular coordinates. in general.
LK(K>O) 0 S 30 ~ ~ f. The frequencies WI and W2 are separated by one octave if W2/ WI = 2. the number of octaves between W2 and WI is number of octaves = loglO(W2/W}) 1 1 = 0.2. sometimes octaves are used to represent the separation of two frequencies.:l 'l 60 90 120 /LK(K<O) 150 I I 180 0.1 I (jJ (rad/sec) 10 100 Figure 213 Bode plot of constant K.:) ~ 10 0 10 20 30 40 0. the relation between octaves and decades is number of octaves = 1/0.32 decades (271) .301 decades = 3. The number of decades between any two frequencies WI and CO2 is given by (269) Similarly.1 m(radlsec) 10 100 90 60 30 'b:Q ¥.log 10 ogIO 2 .~2 Frequency~Oomain Plots 4 35 40 30 IIIII I I 20 20 loglO K dB (K = to) ~ ~ '§ f.301 (CO2) WI (270) Thus. Instead of decades.
. 60 40 20 ~ ~ ~ Q :::> '§' 0  """'" ~ ..1 OJ (rad/sec) 10 100 Figure 214 Bode plots of ( jll)) P • .\. (267). 0.t....I.1 (t) II "Q(. the magnitude of G( jw) is a straight line with a slope of 20 dB / decade. ~ ~.~I~  ~ 1 20 40 L  IV I . we have ±20pdBjdecade = ±20p x 0. i' 10 I r.Q\ "'" ~~c ""~ ~"'to I II ~ 60 0.~t1[:::: ~~ II""~ ~~ '<l. V 11.) ¢. (271) into Eq. which has a simple pole at s = 0. Mathematical Foundation Substituting Eq.0 V d~~r"(I)) I'r. The phase of (jw)±p is written (273) The magnitude and phase curves of the function (jw)±p are shown in Fig.lI'N 'fr!'b/q.36 ~ Chapter 2.~ L(jOJr i L(jOJ) " ~ 90 0 S '§" :::> ". and it passes through the OdB axis at w = 1 rad/sec..I ~ 0 L( 1/j(J) 90 ~ L(lIjf1)2 180 L(lljf1)3 270 . ~ rl'QJ)< too (rad/sec) 180 .::.e"<1q. . 1I11J J(r>. 2wl4 for several values of P.r/::.301 ~ 6p dB/octave (272) For the function G( s) = 1/s.. C:~~rr ~1.V vI" /' /" ~I'~rt~~d8/decad r.
then Eq. Locate the comer frequency w = 1IT on the frequency axis. The actual values and the straightline approximation of II + jwTldB as functions of wT are tabulated in Table 23. because the asymptotic plot forms the shape of a corner at this frequency. with the two lines intersecting at w = I IT. (277). Because the phase of G( jw) varies from 0° to 90°. a straightline approximation can be made for the phase curve. The frequency given in Eq. The magnitude of G( jw) in dB is (274) JG(jW)ldB = 201og lO IG(jw)1 = 20 Iog lO Jl +w2T2 dB (275) To obtain asymptotic approximations of IG(jw)[dB' we consider both very large and very sman values of w. we can draw a line from 0° at 1 decade below the corner frequency to 90° at 1 decade above the corner frequency. It is useful to remember that the error is 3 dB at the corner frequency. (274) is a smooth curve and deviates only slightly from the straightline approximation. As shown in Fig. Table 23 gives the values of L( 1 + jwT) versus wT. The intersect of these two lines is found by equating Eq. The phase of G{ jw) == 1 + jwT is (279) Similar to the magnitude curve. At very low frequencies. a smooth curve can be sketched simply by locating the 3dB point at the corner frequency and the IdB points at 1 octave above and below the corner frequency. 215. At very high frequencies~ wT» I. which is the OdB axis. (276) represents a straight line with a slope of 20 dB/decade of frequency. the actual magnitude curve is obtained by adding the errors to the asymptotic plot at the strategic frequencies. 3. as shown in Fig. (276) to Eq. 37 226 Simple Zero. 215. 1 + jw T Consider the function G(jw) = 1 + jwT where T is a positive real constant. 2. and it is 1 dB at 1 octave above (tv = 2/T) and 1 octave below (w = 1/2T) the comer frequency. wT« I. The errurbetween the actual magnitude curve and the straightline asymptotes is symmetrical with respect to the comer frequency w = 1IT. the maximum deviation between the straightline approximation and the actual curve is less than 6°.3 dB. Based on these facts. we can approximate 1 + w2 T2 by (j)2T2. which gives w=l/T (278) This frequency is also the intersect of the highfrequency approximate plot and the lowfrequency approximate plot. Usually. the procedure of drawing 11 + jwTl dB is as follows: 1. the error is dropped to approximately 0. (274).nown as the corner frequency of the Bode plot of Eq. Draw the 20dB/decade (or 6dB/octave) line and the horizontal line at 0 dB. (275) is approximated by IG(jw)ldB e:! 201og 10 1 = 0 (276) because w 2T2 is neglected when compared with 1. At 1 decade above and below the corner frequency. If necessary. The actual IG(jw) IdB plot ofEq. (275) becomes (277) Eq.22 Frequency~Domain Plots. Eq. . (278) is also k.
.. .043 2 3 11 + jwTldB 0 0 0 0 0 2. ai' g '§ 0 ::> _..043 1..12 1.00 loglOwT + jwT] 1.~I'\I V ~ ~ ~~I' ~ ...76 1. .. 1' ..1 100 ruT 90 60 ~ i 30 0 ~ ~ ::> 0 'J ~ .50 0.. r.4 45....23 10.. 1' ~ .0 0. .01 0..7 2 1 0. ~ ~/ ~~ ~ :::::"'~ G(s) = 1 + Ts ~ :::::::::: 30 60 ~ ~ .4 ..0 40.000043 0. ..12 0 0..3 0..00 10....41 1..0 2.6 2 3 37.04 11 + jwTldB 0.1 ~" I"=: ruT = (l~Ts)' ~ ~ ~~ 10 100 0. Chapter 2. G(s) Asy~ptotes I III = I + Ts I I .31 2.00043 0.. I' O(s) =_1_" I +Ts 90 I 0.043 40.10 0.7 63. ~ i'..3 7.__.117 0.4 84.0 20..00043 89...01 Figure 215 Bode plots of G(s) = 1 + Ts and 0(8) TABLE 2·3 Values of !(1 +jwt) versus Cd T StraightLine Approximation 11 Error L(l + JwT) wT 0. .043 0..0 (dB) (deg) 0..01 0.... / ..00 1...3 1.00043 2 1 0.::::::... 1 +Ts ' 10 / .3 6.26 1..00 100. 1" 40 0..4 V / " ..._.005 26. ....0 1.. O(s)=_I~ 20  __.65 2..._.38 ..... _ .4 100..0 52.5 5.3 4. Mathematical Foundation 40 20 :9...0 20..
Therefore. The magnitude in dB and phase of the Bode plot of Eq. the error between the straight~line approximation and the actual magnitude curve is . By letting s = jm. is given by the negative of the right side of Eq.3 dB. (275). 228 Quadratic Poles and Zeros Now consider the secondorder transfer function (283) We are interested only in the case when ~ ~ 1.2~2 FrequencyDomain Plots ~ 39 2~27 Simple Pole. For instance. The asymptotic approximations of IG( jw) IdB at low and high frequencies are wT« I IG(jeo)ldB ~ OdB (281) (282) Thus~ the corner frequency of the Bode plot of Eq. (280) are shown in Fig. (283) becomes (284) The magnitude of G( jeo) in dB is At very low frequencies. 1/(1 +if» T) For the function G( jw) = 1 1. (280) is still at UJ :::: I /T~ except that at high frequencies the slope of the straightline approximation is 20dB/decade. At the corner frequency. w/wn « 1. Eq. it is simple to extend all the analysis for the case of the simple zero to the Bode plot of Eq. The phase of G( jw) is 0 degrees at eo = O~ and 90 0 when (U = 00. (285) can be approximated as . Eq. the numbers in 11 + ja>TldB' the straightline approximation of II + jwT\dB' the error (dB)9 and the L(I + jwT) columns should all be negative. and the phase LGe jm) is the negative of the angle in Eq. T +Jm (280) the magnitude. la( jw)! in dB. The data in Table 23 are still useful for the simplepole case if appropriate sign changes are made to the numbers. because otherwise G(s) would have two unequal real poles~ and the Bode plot can be obtained by considering G(s) as the product of two transfer functions with simple poles. (280). (279). 215.
= 40 30 20 /~=?05 0.40 • Chapter 2.5 3 $ '§' :=:> to () I IJrf v 0. 1'"0.2 )2' t = 0.3 t.boo .0 0.. (286) to Eq.. 2M 16 for the corresponding ~.. The actual and the asymptotic curves of IG(jw)ldB are shown in Fig./ .. (287).4 0.: f :. / ~ t' I:'~I']'I f"o..... Mathematical Foundation Thus. 217 for the same set of values of {..:> C. The errors between the two sets of curves are shown in Fig.. yielding the corner frequency at (J) W/l' The actual magnitude curve of G( jlV) in this case may differ strikingly from the asymptotic curve.01 ... (283) becomes » !G(jw)!dB ~ .0 u = mlOJ..3 180 I 0. (283) is a straight line that 1ies on the OdB axis..2010g lO V(ev/lVn )4 = 40 log 10 (ev/w.6 // V ><~ i' I / ~ "'0... The actual curve is obtained by making corrections to the asymptotes by using either the data from the error curves of Fig..i ~ '.() ~ ~~I .707/ '" ~ ~ I l..1 """""0. 1. the magnitude in dB of G( jw) in Eq.. v O. I ~ N .. 217 or the curves in Fig. the lowfrequency asymptote of the magnitude plot ofEq.2 ....01 ~ ~ :::: r.... .1 U 1...) dB (287) This equation represents a straight line with a slope of 40 dB / decade in the Bodeplot coordinates. lV/ lVn 1..4 0."'" "".011 .I 10 100 Figure 216 Bode plot of G(s) = I+2t'( / SWII+·~(tI/' \ ("/ .1 ... ~= . ~ . At very high frequencies. The intersection of the two asymptotes is found by equating Eq.I\ I r\ r'\ I' (\ ~ 0..1 0.0 = OJ/run 10 100 () ___ 45 li= .05 I I 0... 216 for several values of ~. The standard procedure of constructing the secondorder IG( jw) IdB is to first locate the corner frequency Wn and 40dB/decade line to the right of Wfl.:> 10 20 30 40 0..= 0. The rea80n for this is that the amplitude and phase curves of the secondorder G( jw) depend not only on the corner frequency Wn but also on the damping ratio ~~ which does not enter the asymptotic curve.:) K90 '§' " 135 1..
The errors between the actual and the asymptotic curves in Fig. 0. The analysis of the Bode plot of the secondorder transfer function ofEq.. rs + 1) j . function~ is obtained by Approach 1 num= [1]. ~ ~~ " . using the MATLAB "bode" the following sequence of MATLAB functions. For (289) the magnitude and phase curves are obtained by inverting those in Fig. Toolbox 223 The Bode plot for Fig. 217 are also inverted. 216.0 IO 100 Figure 217 Errors in magnitude curves of Bode plots of G(s) =I +~(' "( / s rot. den) . \ +(I (dn) 2' S The phase of G( jw) is given by (288) and is plotted as shuwn in Fig.. s = tf(' s'). G = tf'(num.5 ..01 0..2 0.~t=O. Approach 2 bodeCG)... den = [1 ..11] .l ~O.05 and (J) = 1.. G = 11 (s"2 + .22 FrequencyDomain Plots ~ 41 25 20 15 10 "C ~ ~V' t=O~O: .1 1. bode(G). 217 when { = 0. 216 for various values of {...6 0.3 ~ 1 J '§' Q ~ 5 0 5 ~ :J(?<O.707 10 IS 0. (283) can be applied to the secondorder transfer function with two complex zeros.
jwTd = WTd (290) which decreases linearly as a function of w. Mathematical Foundation Bode Diagram 20 40~~~~~~~~L~~~~~~~~ Or~~~~~~~~~~~~~ 45 i ~ 0. (292) shows that G( jet)) has corner frequencies at w = 2. (261) and set s computer plotting. consider the function G(s) 10(s + 10) . for the transfer function G(jw) = Gl (jw)ejlt)Td (291 ) the magnitude plot IO( jw) IdB is identical to that of IGI (jw) IdB • The phase plot LG( jw) is obtained by subtracting WTd radians from the phase curve of Gl (jw) at various w. e. this step is unnecessary).5w)(l + jO. .2w) Eq. The complete Bode plot of the magnitude and phase of O{ jw) is obtained by adding the component curves together. The pole at s = 0 gives a magnitude curve that is a straight line with a slope of 20dB/decade.s(s + 2)(s + 5) The first step is to express O(s) in the form of Eq. passing through the (J) = 1 radl sec point on the O~dB axis. ) G( JW = (292) = jw (keeping in mind that. for (293) jw(1 10(1 + jO.jCJlTd The magnitude of the pure time delay term is equal to unity for all values of w. as shown in Fig. 61 90 . point by point.135 180~ ____~__~~~~~~__===~~~====~=x~d 10D 101 Frequency (radfsec) 229 Pure Time Delay.'1) ! 0.5..leu) + jO.. 218. 218. we have . The phase of the pure time delay term is Le. and 10 rad/sec.42' Chapter 2. The actual curves can be obtained by a computer program and are shown in Fig. EXAMPLE 225 As an illustrative example on the manual construction of the Bode plot. Thus.
den) .) "'~ '\ ::> 0 §' iSH .:..210 " Phase crossover V 5.... ISO C1 'S C.. 20 Gain crossover / 3.. 218.::: ¥" 20 dB/sec I . 7 1 0) .t. function. 40 dB/sec r..78 rad/sec .. . num::.. using the MATLAB •'bode.". (293). 1 . ~ !. r. G = tfCnum.88 tad/sec I'k~ v~ . . bode(G).rr.. den = [ . """ to (rad/sec) 100 ~ ~~ ~tI'to '"1()(K) 90 120 r.10 5 (JJ 10 100 1000 (rad/sec) Figure 2 18 Bode plot 0 f G () = s N lO(s+ lOJ sts+2)t:s+S)..III 40 dB/sec It .10 "'"~ " 2 5 (J) l.60 dB/sec ~r::: ~~ ~~ S2 40 60 80 100 0. Toolbox 224 The Bode plot for Eq. is obtained by the following sequence of MATLAB functions.  240 270 0. ~ i===:c==.22 FrequencyDomain Plots 40 ~ 43 20 0 r.. 12 f'\ ~~ p. [110]..l I ~ ~ :.. The result is a graph similar to Fig.. "'r.
219.44 ~ Chapter 2. The relationships among these three plots are easily identified by comparing the curves in Figs. However. When constant coefficient K of the transfer function varies. jlmG Figure 219 . in the construction of the plot. One of the most important applications of this type of plot is that. the plot is simply raised or lowered vertically according to the value of K in dB. (292) are shown in Fig. the plot can be superposed on the Nichols chart (see Chapter 8) to give information on the relative stability and frequency response of the system. 220. when G( jw) is the forwardpath transfer function of a unityfeedback control system. ~ EXAMPLE 226 As an illustrative example. Mathematical Foundation 2210 MagnitudePhase Plot The magnitudephase plot of G( jw) is a plot of the magnitude of G( jll)) in dB versus its phase in degrees. respectively. 218. Thus. and 220. it is best to make the magnitudephase plot by computer or transfer the data from the Bode plot. 219 and Fig. The Bode plot of the function is already shown in Fig. the polar plot and the magnitudephase plot of Eq. with ill as a parameter on the curve. the property of adding the curves of the individual component~ of the transfer function in the Bode plot does not carry over to this case. 218.
= 1 rad/sec Phase crossover I~. [0 2 5] ..Gain crossover co= 5.5 180..:::: 3.. 2 .78 rad/sec ~ OJ. lO(s+ 10) F" Igure 220 Magmtu dephase pIot 0 f G( S ) = s(s+2)(s+5) • Toolbox 2 ./ V il 10 0 .2~2 FrequencyDomain Plots ..5 The magnitude and phase plot for Example 226 may be obtained using the MATLAB '"nichols" function.0 157.88 rad/sec 1"I .0 247. .5 225... »G = zpk(r .10) Zero/pole/gain: 10 (8 + 10 s (s+2) (s+5) » nichols (G) See Fig. by the following sequence of MATLAB functions. 220. 20 yo.0 8 * 80 270.0 202. 10J.l:Q / V /' :s Q) ~ 10 J_ ~ a 20 'E eo = 40 50 I I I OJ::::: 10 rad/sec ... 60 70 {J} = 100 rad/sec a 90.0 112...30 OJ =30 rad/sec .. 45 40 30 .5 Phase (deg) .5 135.
7 deg (at 3. • Pbasecrossover point.77 rad/sec) Pm =10. Mathematical Foundation Toolbox 226 The phase and gain margins for Eq. ~ 0 ~ 20 ~40 _60~ __ ~ __ ~~~~~L____~_ _~L~~~L~ ____~~~~~ 90~~~~~~~r.~~~~~~~~~~~~ ! f '@ 135 I 180 2211 Gain. margin(G1). G1 = tf(num.and PhaseCrossover Points Gain. (292) are obtained by the following sequence of MATLAB functions.46 ~ Chapter 2. margin(G1).88 rad/sec) I 40 m 20 :g §.36 dB (at 5. The frequency at the gaincrossover point is called the gaincrossover frequency 6>g. These are defined as follows. The gaincrossover point on the frequencydomain plot of GUm) is the point at which IG( jro) I = 1 or IG( jw) IdB = 0 dB. • Gaincrossover point. den=[17100].and phasecrossover points on the frequencydomain plots are important for analysis and design of control systems. Approach 1 num = [10100] . "Margin" produces a Bode plot and displays the margins on this plot. Approach 2 s = tfC' s'). den) . The phasecrossover point on the frequencydomain plot of G( jw) is the point at which LG( jw} = 1800 • The frequency at the phasecrossover point is called the phasecrossover frequency wpo . Gl=(lO"'s + lOO)/(sA3 + rsA2 + 10"8). Bode Diagram 13m =7.
m)rr/2.90°) over the same frequency range. 218. • Bode plot. The gain~crossover point (or points) is where the G( jw) curve crosses the O"dB axis. (295) is shown in Fig. The phasecrossover point (or points) is where the G( jw) curve crosses the 1800 axis (see Fig. The gaincrossover point (or points) is where IG(jtll) I = 1. when s == jw and as w varies from 00 to 0. The phasecrossover point (or points) is where the phase curve crosses the 1800 axis (see Fig. In other words. Conversely. M Gs _ ( ) . (292) has a net phase change of only 90° (from 180° to . Notice that the nonminimum~ phase function has a net phase shift of 2701) (from 1801) to + 90~) as w varies from 00 to O. it is called a nonminimumphase transfer function. • Magnitudephase plot. as shown in Fig. 220). (2~92). (2 92) is in the righthalf s~plane. if any. LG(jw). 2~21(b). and the polar plot is shown in Fig.22 FrequencyDomain Plots 4 47 The gain and phase crossovers are interpreted with respect to three types of plots: • Polar plot. The phase curve of the Bode plot ofG(jit)) ofEq. . However. IG(jw) I is completely defined. 22 12 MinimumPhase and NonminimumPhase Functions 00 A majority of the process transfer functions encountered in linear control systems do not have poles or zeros in the righthalf splane. • A nonminimumphase transfer function will always have shift as w is varied from 00 to O. 218).10) + 5) (295) The magnitude plOl of the Bode diagram of G( jw) is identical to that of the minimum~phase transfer function in Eq. When a transfer function has either a pole or a zero in the righthalf splane. that is. T (294) the magnitude of G( jw) is the same whether T is positive (nonminimum phase) or negative (minimum phase). The gaincrossover point (or points) is where the magnitude curve IG( jtll) IdB crosses the OdB axis. a more positive phase EXAMPLE 2~27 As an il1ustrative example of the properties of the nonminimumphase transfer fUIlction. consider that the zero of the transfer function of Eq. given a minimumphase function G(s)y knowing its magnitude characteristics IG(jw) I completely defines the phase characteristics. 219). given LG(jw) . • The value of a minimumphase transfer function cannot become zero or infinity at any finite nonzero frequency.s(s + 2)(s 1O(s . 221(a). This class of transfer functions is called the minimumphase transfer function. Nonminimumphase transfer functions do not have the unique magnitudephase relationships. the phase of G( jw} is different for positive and negativ~ T. the total phase variation of G(jm) is (n . Additional properties of the minimum~phase transfer functions are as follows: • For a minimumphase transfer function G(s) with m zeros and n poles. The phasecrossover point (or points) is where LG(jcv) = 1800 (see Fig. For instance~ given the function G(jw)=l 1 JW . excluding the poles at s = D. Minimumphase transfer functions have an important property in that their magnitude and phase characteristics are uniquely related. whereas the minimumphase transfer function of Eq.
• Do not use the Bode plot and the gainphase plot of a nonminimumphase transfer function for stability studies.::. Bode diagrams of nonminimumphase forwardpath transfer functions should not be used for stability analysis of closedloop control systems.1. the polar plot.r. " i' i\ r\.48 ... . . (b) Polar plot. The topic of frequency response bas a special importance in the study of control systems and is ~ revisited later in Chapter 8. • The magnitude of the pure time delay tenn is unity for all w.. Chapter 2. ~ o i' 'l 90 135 ~ 10 (j) ". Mathematical Foundation 90 45 r. . .1  100 1000 (rad/sec) (a) Gplane ReG (b) Figure 2·21 (a) Phase curve of the Bode plot. is more convenient for nonminimumphase systems.. l o '§' 45 :. For stability studies. The same is true for the magnitudephase plot... • The magnitude and phase characteristics of a minimumphase function are uniquely related. 1' 180 0. .. G(s) = S(~~~~~T5)' Care should be taken when using the Bode diagram for the analysis and design of systems with nonminimumphase transfer functions. when used along with the Nyquist criterion discussed in Chapter 8. Here are some important notes: • A Bode plot is also known as a corner plot or an asymptotic plot.
at ~ . For instance. (296) is referred to as a secondorder differential equation.. such as in heattransfer systems.. e(t) is the forcing function. a series electric RLC (resistanceinductancecapacitance) network can be represented by the differential equation: RiCI) diCt) 1 + Ldt + c J i(t)dt = e(t) (296) where R is the resistance. C.dt + aoy(t) = let) dy(t) (297) which is also known as a linear ordinary differential equation if the coefficients ao. A first~order linear ordinary differential equation is therefore in the general form: dt + aoy(t) = [(t) and the secondorder general form of a linear ordinary differential equation is d y(t) dt 2 2 dy(t) (298) +a 1 dy(t) dt +a yet) = J(t) 0 (299) In this text. (296) should be referred to as an integroditlerential equation. Eq. t. the following differential equation that describes the motion of a pendulum of mass m and length /. the applied voltage. partial differential equations are used. Eq. the differential equations encountered are all of the ordinary type... the capacitance. (2100) is nonlinear.. 231 Linear Ordinary Differential Equations In general. the dependent variable or unknown that is to be determined by solving the differential equation. "anI are not functions of y(t). .+ anl d ~. and i(t). and e(t). Eq.1 p + .3. i(t). and we refer to the system as a secondorder system. the independent variable. the current in the network. + a 1 . These equations generally involve derivatives and integrals of the dependent variables with respect to the independent variableusually time. the inductance. is d28(t) . because an integral is involved. and the system is called a nonlinear system. 2. later discussed in this chapter. because we treat only systems that contain lumped parameters. In this case. ml(jj2+ mgsm8(t) = 0 (2100) Because e(t) appears as a sine function. For instance. For systems with distributed parameters.23 Introduction to Differential Equations 41 49 ~ 23 INTRODUCTION TO DIFFERENTIAL EQUATIONS A wide range of systems in engineering are modeled mathematically by differential equations.2 Nonlinear Differential Equations Many physical systems are nonlinear and must be described by nonlinear differential equations. Strictly speaking. L. the differential equation of an nthorder system is written dny(t) dn1y(t) d..
order differential equations are simpler to solve than higherorder ones.. and future conditions of the system. for Eq.t) = X2(t) . if we let Xl (t) = J i(t)dt (2101) and X2(t) = d:q (t) lit = . let us define Xl (t) = y(t) X2(t) = .. (297) to the rest of the terms..a ll 2Xll\ (t ) (l1llXn (t ) dX2(t) dt dX (t) . the set of firstorder differential equations in Eq. X2(t). an nthorder differential equation can be decomposed into n firstorder differential equations.. (296) is decomposed into the following two firstorder differential equations: dx~.. first . .. Mathematical Foundation 233 FirstOrder Differential Equations: State Equations2 In general.. present. In control systems theory..ttl are caned the state variables.50 . it is convenient to define a set of state variables and state equations to model dynamic systems.]I = Il + f (t ) Notice that the last equation is obtained by equating the highestordered derivative term in Eq.X2. (2106) is called the state equations. l(t) (2102) then Eq.. From a mathematical perspective.alx2 (t) . the variables XI (t). and XI. (296). (2104) In a similar manner..= . firstorder differential equations are used in the analytical studies of control systems. 234 Definition of State Variables The state of a system refers to the past. d~~) (2105) cJllly(t) xll(t) = dt n  I Il then the nthorder differential equation is decomposed into equations: dXI (t) firstorder differential dt = X2(t) = X3(t) (2106) oOxl (t) . Because.. As it turns out. . Chapter 2. For the differential equation in Eq. in princi pie. (297). . (2105) are the state variables of the nthorder system :2 Please refe!' to Chapter 10 for more in~depth study of Stl1le Space Systems. .X I (t)  (2103) 1 dt2(t) dt 1 LC X2(t) + eel) L L R . . ~Xll(t) defined in Eq.
. ll b21 al2 a21 a22 a2n (n x n) (2110) ani a n2 bl2 b22 a.. rotor velocity.... present. in an electric motor. and displacement can be measured physically. An output of a system is a variable that can be measured. but a state variable does not always satisfy this requirement. xll(t) u(t) = [=~~~ll u p'(!) aln (2109) The coefficient matrices A and B are defined as: A= [all . Hence. the state variables x I (to). In general. Xl (t).l 235 The Output Equation bn2 b'~'J One should not confuse the state variables with the outputs of a system. but it cannot be measured directly during operation and therefore does not ordinarily qualify as an output variable. the state variables should completely define the future behavior of the system. x(t) = . On the other hand.X2(t) • The state variables of a system are defined as a minimal set of variables.m p B= [b b2p bl 1 (n x p) (211l) b. and future states of the motor. such that knowledge of these variables at any time to and information on the applied input at time to are sufficient to determine the state of the system at any time t > to. the space state form for n state variables is x(t) = Ax(t) + Bu XI(t) X2(t} (2. • Once the inputs of the system for t 2 to and the initial statesjust defined are specified. such state variables as the winding current. '=2 (to). . The state variables must satisfy the following conditions: • At any initial time t = to. Xn (to) define the initial states of the system. magnetic flux can also be regarded as a state variable in an electric motor.2 3 Introduction to Differential Equations M 51 described by Eq. In general. and the n first~order differential equations are the state equations. ..xn(t). . (2..97). because it represents the past. an output variable can be expressed as an algebraic . 107) where x(t) is the state vector having n rows. For instance. 1 (2108) [ and u(t) is the input vector with prows.. and these variables all qualify as output variables. there are some basic rules regarding the definition of a state variable and what constitutes a state equation.
In contrast with the classical method of solving linear differential equations. (2117) is also known as the onesided Laplace transform. Mathematical Foundation combination of the state variables. the Laplace transform method has the following two features: 1. (297). The defining equation in Eq. The Laplace transform converts the differential equation into an algebraic equation in sdomain. The final solution is obtained by taking the inverse Laplace transform. For the system described by Eq. that is~ S = U + jOJ.52 . . f(t)eS1dt or Jo (2116) F(s) = Laplace transform of J(t) = . if y(t) is designated as the output~ then the output equation is simply y(t) = XI (t). which is a complex variable. 24 LAPLACE TRANSFORM The Laplace transform is one of the mathematical tools used to solve linear ordinary differential equations. In general.c[ f(t)] (2117) The variable s is referred to as the Laplace operator.. It is then possible to manipulate the algebraic equation by simple algebraic rules to obtain the solution in the sdomain. 2. [ell C21 Cql Cq2 : Cin ] (2113) [dll d21 D= . 00 (2115) the Laplace transfonn of fit) is defined as F(s) = (ex. This simply means that all information contained .. where (f is the real component and w is the imaginary component. 241 Definition of the Laplace Transform Given the real function fit) that satisfies the condition 10'' ' If(t)eutldt < for some finite. dI2 d22 diP] d2p (2114) d qp dql d q2 We will utilize these concepts in the modeling of various dynamical systems .. y(t) = [YI(t)] Y2(t) : = ex(t) + Do yq(t) Cl2 C22 (2112) c= . as the integration is evaluated from t = 0 to 00.. Chapter 2. The homogeneous equation and the particular integral of the solution of the differential equation are obtained in one operation. real CT..
to t = 00. (2117) is used for the evaluation of the Laplace transform of f(l). response does not precede excitation.or 0+ never needs to be addressed. that is. (2117) is almost never used in problem solving. The Laplace transform of ft. (2119) is valid if which means that the teal part of s. for a physical system when an input is applied at t = 0.st IX = (2118) The Laplace transform of fit) is obtained as F(s) = £[us(t)) = l 'X.sided Laplace transform should be defined from t = 0. the fine point of using 0. such as the one given in Appendix C. since the transform expressions encountered are either given or can be found from the Laplace transform table. = . time reference is often chosen at t = O. Strictly. the response of the system does not start sooner than 1= 0. () u. This assumption does not impose any limitation on the applications of the Laplace transform to linear systems. the one. must be greater than zero. we simply refer to the Laplace transform of the unitstep function as liS. For the subjects treated in this text.24 Laplace Transform ~ 53 inj{t) prior to t = 0 is ignored or considered to be zero. . This limiting process will take care of situations under which the functionj{t) has ajump discontinuity or an impulse at I = O.~(t)estdt s 0 s (2119) Eq. Jo erx1estdt I = e(s+ll'}tl'X. Such a system is also known as being causal or simply pbysicalJy realizable..41 Use the MATLAB symbolic toolbox to find the Laplace transforms.t) is written F(s) = ( ox.yms t »f= t A 4 f= t A 4 » laplace (f) ans = . For simplicity. The following examples illustrate how Eq. »::. In practice. The symbol I = 0.implies the limit of t l> 0 is taken from the left side oft = O. CT. since in the usual timedomain studies. the defining equation of the Laplace transform in Eq. Thus.. EXAMPLE 241 Letflt) be a unitstep function that is defined as /(t) = us(t) = = 0 1 t~O t<O 1 I = __ e. we shall simply use t = 0 or 1= 10("2: 0) as the initial time in all subsequent discussions. s +a () oS +a (2122) Toolbox 2. Furthermore. and rarely do we have to be concerned with the "'1 region in the splane in which the transform integral converges absolutely. EXAMPLE 242 Consider the exponential function t~O where a is a real constant.
For simple functions.s'f . for which no proofs are given here.lim j(t) t t+O = sF(s)  f(O) (2127) In generat for higherorder derivatives of J(t). 2~4~3 Important Theorems of the Laplace Transform The applications ofthe Laplace transfonn in many instances are simplified by utilization of the properties of the transform. Eq.1/(0) .. . . 27rJ c joo l c jOQ (2124) where c is a real constant that is greater than the real parts of all the singularities of F(s). Then (2126) • Theorem 3. Mathematical Foundation 242 Inverse Laplace Transformation Given the Laplace transform F(s). the operation of obtaining j(t) is tenned the inverse Laplace transformation and is denoted by I(t) = Inverse Laplace transfonn of F(s) = £1 [F(s)] (2123) The inverse Laplace transform integral is given as 1 + F(s) est ds jet) =~. (2124) represents a line integral that is to be evaluated in the splane. £[d I(t)] = s"F(s) _ lim [snl f(t) dtn t_ 0 n + 1'2 dl(t) + .22. andj(O) is the limit ofj(t) as t approaches O.}'<s) be the Laplace transform of!t(t) andlz(t).. You may also use the ACSYS "Transfer Function Symbolic" Toot Tfsymt for partialfraction expansion and inverse Laplace transformation. Multiplication by a Constant Let k be a constant and F(s) be the Laplace transfonn of j(t). + d dt n  l dfl  I(t)] 1 (2128) = snF(s) . Differentiation Let F(s) be the Laplace transfonn ofj(t). • Theorem 1.t(ItI)(O) where 1(1)(0) denotes the ithorder derivative of /(t) with respect to t. evaluated at t = O.s.(1)(0) . The Laplace transform of the time derivative of j{t) is £[dfd(t)] = sF(s) . Then £[kf(t)] = kF(s) (2125) • Theorem 2... These properties are presented by the following theorems. the inverse Laplace transform operation can be carried out simply by referring to the Laplace transform table. respectively. Sum and Difference Let F1(s) and F.S4 ~ Chapter 2. the inverse Laplace transform can be carried out by first performing a partialfraction expansion (Section 2~5) on F(s) and then using the Transform Table from Appendix D.. For complex functions. such as the one given in Appendix C and on the inside back cover.
Final. Integration The Laplace transfonn of the first integral off(t) with respect to t is the Laplace transfonn of ft.. that is. The following examples illustrate the care that must be taken in applying the theorem..t. that is. then 10 lim /(t) = lim sF(s) $00 (2132) if the limit exists.T)] = eTSF(s) (2131) where us(t . Shift in Time The Laplace transform of fit) delayed by time T is equal to the Laplace transform f(t) multiplied by eT.T)us(t . (2133).. III Theorem 7.t) divided by s. . [Jrll! Jr 1J1  1 o o . (h 10 f(t)drdtldt2'" dtnl 1=. tI Theorem 6. Initial.3 Consider the function F{s) = s(""""s2+s +2) Because sF(s) is analytic on the imaginary axis and in the righthalf splane. as stated in the theorem. ~ EXAMPLE 24. (2129) For nthorder integration.T) denotes the unitstep function that is shifted in time to the right by T. The finalvalue theorem is not valid if sF{s) contains any pole whose real part is zero or positive. Using Eq. we have tex.. the finalvalue theorem may be applied.Value Theorem If the Laplace transfonn of f(t) is F(s). and if sF(s) is analytic (see Section 214 on the definition of an analytic function) on the imaginary axis and in the right half of the splane. then Ico lim f(t) =$0 sF(s) lim (2133) The finalvalue theorem is very useful for the analysis and design of control systems. which is equivalent to the analytic requirement of sF(s) in the righthalf splane.c[ f(t . 5 lim f(t) = lim sF(s) 50 = sOS2 +s+ 2 = ~2 lim 5 (2135) 4111 .Value Theorem If the Laplace transform of f( t) is F{s). £.24 Laplace Transform ~ 55 III Theorem 4.nF(s) (2130) II Theorem 5. because it gives the final value of a time function by knowing the behavior of its Laplace transform at s = O.
T)J = eT. [dl'f(t)] where j(k) (0) = s" F(s)  s'.e[ e=Faf f(t)] = F(s ± a) (2137) TABLE 24 Theorems of Laplace Translorms Multiplication by a constant Sum and difference Differentiati on £[kf(t)] = kF(s) £. EXAMPLE 244 Consider the function (V F(s) ="2 s~ +w (2136) which is the Laplace transfonn of f(t) = sin (df.2/(0) = dkf~t) dt I 1"..56 '.. . the finalvalue theorem cannot be applied in this case.e[fot It (r)h(t ~ T)dT] =£ Complex convolution [1 1 f2(r) It (t ~ r)dr] = £[Ji (I) * 12(/)] .. although the finalvalue theorem would yield a value of zero as the final value of fit).elfl (t) 12(1)) = Fl (s) * Fz(s) . lim /(t) (+x = s>O sF(s) if sF(s) does not have poles on orto the right of the imaginary axis in lim the s~plane. Theorem 8. In other words.s'1. that is.l = s" ] C[ f{t .(t . Complex Shifting The Laplace transform of fit} multiplied by e T IXI ~ where a is a constant. with s replaced by s ± a.0 Integration £. Because the function sF(s) has two poles on the imaginary axis of the splane. . Chapter 2. is equal to the Laplace transform F(s). Mathematical Foundation .iF(s) tO lim 1(1) = lim sF(s) .. 1t1 o 0 0 F(s) f(t)drdlJdt2·· ·dt.T)u. the result is erroneous. [d~~)] = sF(s) dIll /(0) [..'ix.1 f(O) . Complex shifting C[ e=fCltf f(t)] = F(s ± a) Fl (S)F2(S) = Real convolution . Shift in time Initialvalue theorem Finalvalue theorem [1 '1l1tnJ .
(2138) shows that multiplication of two transformed functions in the complex sdomain is equivalent to the convolution of two corresponding real functions of t in the Idomain. called the complex convolution. in general. .. Table 24 summarizes the theorems of the Laplace transforms represented. Real Convolution (Complex Multiplication) Let Fl (s) and F 2(s) be the Laplace transforms of fl(t) and f2(t). An important fact to remember is that the inverse Laplace transform of the product of two junctions in the Sdol1zuin is not equal to the product of the two corresponding real functions in the ttionlain. the theorem states that multiplication in the real tdomain is equivalent to convolution in the complex sdomain.. then Fl (S)F2(S) == . it can be written as Q(s) G(s) = P(s) where P(s) and Q(s) are polynomials of s.. The methods of partialfraction expansion will now be given for the cases of simple poles.. or real multiplication. Eq..] =£[fo' 12(.. anl are real coefficients...25 Inverse Laplace Transform by PartialFraction Expansion • 57 • Theorem 9. ... 251 PartialFraction Expansion When the Laplace transfonn solution of a differential equation is a rational function in s. 12(t) = 0. Rathert the inverse Laplace transfonn operation involving rational functions can be carried out using a Laplace transform table and partialfraction expansion. multipleorder poles. the evaluation of the inverse Laplace transform does not rely on the use of the inversion integral of Eq..elfl (t) * 12(t)] (2138) =£[fo' M r)!2(t. for t< 0. that is. and complexconjugate poles of G(s). both of which can also be done by computer programs. The polynomial P(s) may be written (2142) where ao.)d... )d. (2139) There is also a dual relation to the real convolution theorem.... and II (t) = 0.. respectively.] where the symbol * denotes convolution in the time domain. Details of the complex convolution formula are not given here. that is. It is assumed that the order of pes) in s is greater than that 0/ Q(s).. Essentially. * • 25 INVERSE LAPLACE TRANSFORM BY PARTIALFRACTION EXPANSION In a majority of the problems in control systems. (2124). (ll •.)11 (t .. (2140) where denotes complex convolution in this case.
s+ 1 8+2 s+3 (2147) The coefficients K_I.. we mUltiply both sides of Eq. 2.(s + l)(s + 2)(s + 3) . . ..58 . (2117) can be written as O(S) = Q(s) P(s) where Sl = Q(s) (s + $1 )(s + S2) ..\'+2 s+3 1 7 6 5(3} + 3 (2 .al coefficients » [r t p.. k] = residue(b.. =f.vI) and let S Sl. 150) Thus.S...)(S3 . (2143) by the factor (s + Si) and then setting s equal to $... » a = [16116] % denominator polynomial coefficients You can calculate the partial fraction expansion as » b = [5 3] % numerator po~ynomi. Eq. (2146) becomes (2151) Toolbox 251 For Example 251.0000 1... To find the coefficient Ksh for instance. (2143) by (s +. . (2143) is written KS/l i. Eq. Mathematical Foundation G(s) Has Simple Poles If all the poles of 0(8) are simple and real.. (s + sn) expansion. = Ksl = [(S + 81) Q(S)] I pes) S=SL = Q( sd (S2 .st) . Eq. +S + Sl S + S2 S + Sn The coefficient Ksi (i = 1.n) is determined by multiplying both sides of Eq.$2 f:. and K3 are determined as follows: K_I = [(8 + I)G(s)] K2 = [(s + 2)0(s)] s=2 I I s=I = (2 _ 1)(3 _ 1) = 1 5(2) + 3 = (1 _ 2)(3 _ 2) = 7 5(1)+3 (2148) (2149) K3 = [(s + 3)G(s)] Is=3 = (1 _ 3)(2 _ 3) = 6 G(s)=+s+1 .0000 . K2.sn. (2146) is a ratio of two polynomials. (SIl ~ sI) (2145) ~ EXAMPLE 251 Consider the function O(s) _ 58+3 _ 5s+3 . Applying the partiaI~fraction O(s) Ksl Ks2 =++ . a) r= 6.. Chapter 2. Eg. Thus.s3 + 6s2 + lls + 6 which is written in the partialfraction expanded form: G() s (2146) =KI +K2 +K3 .0000 7..
The determination of the coefficients that correspond to the multipleorder poles is described as follows. G(s) is written G(s) = Q(s) = P(s) Q(s) r (s + st}(s + 82) . + Ks(nr) sISl s+sz 3+Sllr I +.0000 11.n . convert the partial fraction expansion back to polynomial coefficients. may be evaluated by the method described by Eq..r terms of repeated poles I )0 + . is of mUltiplicity r.r=_" d "] = ds [ (s + Si) 'G(s) I 2 $=.0000 1.. ArI = [(s+ Si)' G(S)ll.0000 Note that the result is normalized for the leading coefficient in the denominator.0000 6. (2145).p.0000 a= 5.25 Inverse Laplace Transform by PartialFraction Expansion ~ S9 p= 3.0000 k= [] Now.Ks(nr).. (S + Snr)(S + Si) (2~152) (i =I 1. n . (2154) (2155) Ar2 = 21 ds2 1 d [( S + Si) .G(s) ] I s=s. then G(s) can be expanded as G(s) = Ksl + Ks2 + . 2.0000 3.r).<. which correspond to simple poles..0000 1. Ks2. Ksl. ....a] =residue(r.. (2156) (2157) ..0000 2.r teons of simple poles + I Al ++ S+Si A2 (s+sil (S+SiY I+. •. A. . or we say that the pole at s = S.0000 6. . »[b..k) b= 0.r) coefficients. G(s) Has MultipleOrder Poles If r of the n poles of G(s) are identical. +:~::: Ar (2153) Then (n ..
Eq.0000 1. [(s + ds 2 1 I )3G (s)] = 5=1 !!...0000 1.0000 .60 ~ Chapter 2.\.'I + 1)3 G (s)] I s=I 1d 1 = 2d 2 [ ~(2)] S + I s=I =1 (2164) The completed partialfraction expansion is I 1 s+ I (s + 1)' G(s) =2s+2(s+2) . EXAMPLE 252 Consider the function G~)= I s(s + 1)3(s + 2) = $5 + 5. 158) is a ratio of two polynomials.1 .3 (2165) Toolbox 252 For Example 2~52. a) % b is the numerator and a is the denominator r= 1. S [(.ents b= 1 » [r. k] = residue (b. p.... Mathematical Foundation . ds s{s + 2) 2 [_1_] I 0 = sI $ S (2163) d At = 21d 2 '.. (2. (2153). » a a= »clear all = (1 5 9 7 21 % coeffi.4 + 9s3 + 7s" + 25' ? (2158) By using the format of Eq...] %polynornial coeffici.t=:o = ~ = K2 = [(s + 2)G(s)] and those of the thirdorder pole are t=2 ~ (2161) (2162) AZ = !!.cients of polynomial s"4 + 5'f:s"g + 9*s"2 + 7*s + 2 9 7 1 5 2 »b = [1.0000 1. G(s) is written G() s =+++?+s s + 2 s + 1 ($ + 1t (s + 1)3 Kn K~2 AI A2 A3 (2159) The coefficients corresponding to the simple poles are Ko = [sG(s)] I.
K_(1+j"w = (s + (J  jw)G(s)1 S=(1 +' jet> (2166) K(1_j"w = (s + + j'w)G(s} I5=(1.25 Inverse Laplace Transform by PartialFraction Expansion p= ~ 61 2.o  (2169) (2170) and (2171 ) . jw (2168) Let uS assume that the value of ~ is less than one.0000 0.0000 5.JW " (J (2167) ~ EXAMPLE 253 Consider the secondorder prototype function G(s) = w'" ') s2 + 2{UJns + Il w. Then.jw The corresponding coefficients of these poles are found by using Eq.0000 a= 0.0000 k= [] » [b.0000 7. a] = residue(r.0000 1.k)% Obtain the polynomial form b= 0. (2144) is valid also for simple complexconjugate poles. they deserve special treatment here.0000 G(s) Has Simple ComplexConjugate Poles The partialfraction expansion of Eq. (2117) contains a pair of complex poles: S = (1 + jw and s = u . G(s) is expanded as follows! G( s ) where = s+a  K_(T+ jw " JW + S + a + Jll) .0000 1. (2145).p . K.0000 2.0000 9.0000 1. Because complexconjugate poles are more difficult to handle and are of special interest in control system studies.0000 1.0000 1. so that the poles of G(s) are complex. Suppose that G(s) of Eq.
The procedure is outlined as follows: I. Manipulate the transformed algebraic equation and solve for the output variable. first.. Chapter 2.'Ii + + jCt) (f Taking the inverse Laplace transform on both sides of the last equation gives (2~17S) Or. we primarily study linear ordinary differential equations with constant coefficients such as the firstorder linear system: dy(t) dt + aoy(t) = J(t) (2177) or the secondorder linear system: .+ al _. the partialfraction expansion. Mathematical Foundation The coefficients in Eq. Perfonn partialfraction expansion to the transfonned algebraic equation.. Transform the differential equation to the sdomain by Laplace transform using the Laplace transform table.. Let us examine two specific cases. 3. (2~168) is 1 ] (2~174) G( ) = 1 _ s 2jw S + (T . (2~169) are determined as K_rr+jw = (s + 0' . ui (2~172) JW K_ajw = (s + C1 + jW)G(s) I = _ s=a.jw)G(s) l s=a+jw = n 2.+ aoy(t) 2 dt dt d2y(t) dv(t) = f(t) (2178) Linear ordinary differential equations can be solved by the Laplace transform method with the aid of the theorems on Laplace transfonn given in Section 24. the mathematical models of most components of control systems are represented by first. 4. [ . and the table of Laplace transforms.jw 2Jw U)~ (2173) The complete paI1ialfraction expansion of Eq.jill w. . 26 APPLICATION OF THE LAPLACE TRANSFORM TO THE SOLUTION OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS As we see later.or secondorder differentia1 equations.62 . 2. Obtain the inverse Laplace transform from the Laplace transfonn table.and secondorder prototype systems.. In this textbook. ).
f(t) If y(O) = Y (0) = = us(t) = { ~: t < 0. t' is known as the time constant of the system. t. which is a measure of how fast the system responds to initial conditions of external excitations.26 Application of the Laplace Transform < 63 261 FirstOrder Prototype System Consider Eq. (2179). (2184). the system response approaches faster to the final value. dy(t) 1 f{/) . (2183) becomes Y(s) = Ko s (2183) Kl/r (2184) +I where.= srY(s) + Y(s) s or (2182) 1 I Y(s) = . As the value of time constant r decreases. (2179). KO = 1 and K_ 1/ t = 1.S is + I Notice that the system has a pole at s = l/r.e~ tiT: (2185) t·x where t is the time for y(t) to reach 63% of its final value of lim )'(/) = I. ~" EXAMPLE 261 Find the solution of the firstorder ditTerential Eq.and L(y(t)) = . y(l) 0. Eq. (2177). we get the time response of Eq. .dt + . 222 for a general value of t. we have (2181) 1 o. SOLUTION For a unit step input /(1) Eq. which may also be represented by the firstorder prototype form: . . 'Cs + Vo(t) = 1 .c (us(t)) = . Using partial fractions. Typical unit~step responses of y(t)are shown in Fig.?: 0 (2180) = T. Applying the inverse Laplace transform to Eq.\' 1 .V(t) + )'(t~ Y(s).63 o Figure 222 Unitstep response of a firstorder RC circuit system. (2179) is written as U.y(t) == '[. (2179) where..
(2183).yf0(O) + 3sY(s) . yet) = 0. for the secondorder prototype of the form: (2186) where ~ is known as the damping ratio. 222. The significance of this representation will become more evident when we study the time response of control systems in Chapter 5. The initial conditions are y(O) = 1 and yll}(O) = t:(v(t)/dtlt=o = 2. is equivalent to: » s = syrn( . (2185).1.sy(O) . Mathematical Foundation Toolbox 261 The inverse Laplace transfonn for Eq.1 is obtained using » clear all. »tau = 0. s' ) . and (Un is the natuml frequency of the system. :» :» syrns s tau. ::» tau = syrne 'tau'). the sym command lets you construct symbolic variables and expressions.01:1. Note.64 Chapter 2.63. 26~2 SecondOrder Prototype System Similarly. we get _S2  s+5 Y(s) = 8(s2 + 38 + 2) = s(s + l)(s + 2) ~s2 .(t) + 3 dt + 2y(t) = dv(t) Sus(t) (2187) where uit) is the unitstep function . The result is Eq.1. we first take the Laplace transform on both sides of Eq.s + 5 . shown in Fig. The prototype forms of differential equations provide a common format of representing various components of a control system. You may wish to confirm that at t = 0.. To solve the differential equation. (2153): s2y(s) .3)'(0) + 2Y(s) == 5/s (2188) Substituting the values of the initial conditions into the last equation and solving for Y(s). ilaplace(1/(tau~"sA2 + s)). » t = 0:0. Time response of Eq. » plot (lexp( t/tau)) . and the command: » syms s tau. for a given value r = 0. (2183) is obtained using the l\1ATLAB Symbolic Toolbox by the following sequence of MATLAB functions. EXAMPLE 262 Consider the differential equation ~ d2}.
Js +~) (2194) where. The result is yet) where = 1..9938) t?= 0 (2197) Eq. the finalvalue theorem of Eq. the last two terms represent the transient or homogeneous solution.25t sin (26. knowing that the poles are at s = 0.w.. (2133) may be applied.51 + 0. we get the complete solution as The first term in Eq. . yet} = 1 . we have first checked and found that the poles of function .:::O (2195) e = cos.5455 and (Un = 31.. where (J = {WII = 17.9938 rad (rrrad) == 56..94 180 D D (2196) Thus.w"t ( ) t. (2193).~Sin wlI~t+e 1.s(s2 + 2t. (2~189) is expanded by partialfraction expansion to give Y(s) = 2s . in order to ensure the validity of the finalvalue theorem.{2 et. (2190).193e. (2194) directly. the Laplace transform method gives the entire solution in one operation.62.1.jw. (2197) can be derived by petforming the partialfraction expansion Qf Eq.5 d~~) + lOOOy(t) = 1000us (t) (2193) The initial values of y(t) and dy(t)/dt are zero. which requires separate steps to give the transient and the steadystate response. (2191) is the steady~state solution or the particular integral. (J + jw. using the secondorder prototype representation. (2194) can be executed in a number of ways.s{s2 + 34.. The inverse Laplace transform of Eq. EXAMPLE 263 Consider the linear differential equation d~~t) + 34.25 (2198) (2199) w == UJIl~ == 26..5s + 1000) . The Laplace transform table in Appendix C provides the timeresponse expression in Eq. we have Y(s) _ 1000 _ w~ .17. Unlike the classical method.1{= 0... and (J . and solving for Y(s). ~ = 0. If only the magnitude of the steadystate solution of yet) is of interest.rY(s) are all in the lefthalf splane.26 Application of the Laplace Transform . or solutions. Thus.. (2l94).s + 1 + 2(8 + 2) 5 5 3 (2190) Taking the inverse Laplace transform of Eq. t~~ Y t r () =s~ S r Y() S =s ~o s2 + 3s + 2 = '2 r _s2  8 +5 5 (2192) where. Taking the Laplace transform on both sides of Eq.5 . 65 Eq.
Kajw (2200) Ko = sY(s) I..cO. s Ko Ka+je. (2194) is written Y() s where =+ s + (T )W + $ + a + JW .1 Sin[wll~t ~ t/J] (2205) Substituting Eq.. we have y(t} = 1 R e~wllt sin [w/I\h  t.9938) t~O (2207) splane 8= cos l~ Figure 223 Root location in the splane.wn t [ej(ltJtr/J) _ ei(ctJtr/J}] + ~ 1 el. (2205) for 1/. (2200) is now written y(t) = 1 + = 1 2J~ .2 Krrjw = ($+0'+ jw)Y(s) I = S=(fjw 2J . (2·203) The angle ifJ is given by and is illustrated in Fig. Mathematical Foundation The partialfraction expansion of Eq..~ ejrjJ 1 t. Chapter 2. . ~.5t+O. 223.t. The inverse Laplace transform of Eq.193e17. (2~204) into Eq. .e1.2 t + cosl~] t ~O (2206) or y(t) = 1 _1.25t sin(26.~=o = 1 Ku+jw = (s (2·201) + 0"  jw)Y(s) I = s=a+jw 2} 'R j4J 1 .66 . 1 .
1 Impulse Response Consider that a linear time~invariant system has the input u(l) and output yet).2 0.. title ( 'Step Response' ) xlabel (I Time (sec' ) ylabel ( . G=1000/(sA2 + 34.15 Time (sec) 0.5 1000] . den = [1 34..5*8+1000) .2 0.S 0. (2194) for a unit"step input may also be obtained using Alternati vel y: nurn = [1000] .den). One way to define the transfer function is to use the impulse response~ which is defined as follows.2~7 Impulse Response and Transfer Functions of linear Systems 67 Toolbox 262 Time response of Eq. Arnpli tude • ) s = t f (' s'). 1.4 0.2 :t::: g is.4~~~~~. III 0. step(G). step (G).8 £ O. step Response 1. a rectangular pulse function u(t) of a very large magnitude l~/2e becomes . As shown in Fig.1 0. 224..3 0. G= tf (num. title (' Step Response f) xlabel ( I Time (sec • ) ylabel (' Amplitude') "step" produces the lime response of II function for a unitstep input.35 27 IMPULSE RESPONSE AND TRANSFER FUNCTIONS OF LINEAR SYSTEMS The classical way of modeling linear timeinvariant systems is to use transfer functions to represent inputoutput relations between variables.05 0.. 27.25 0.
= w"u(t) 2 (2210) O(s) = . with any input. u(t). U(I) 8(t) is also known as unit impulse or Dirac delta function. conditions. ~ EXAMPLE 2~ 7~1 For the second~order prototype system Eq.~2t t 2~53. (2211) is the transfer function of the system in Eq. an impulse function for very small durations as e + O. .r j t+ r i ! Figure 2~24 Graphical representation an impulse function. The important point here is that the response of allY systen'l can be characterized by its impulse response g(t). using Eq. (2~208). We define = G(s) = .e. the impulse response g(t) is (2~21O). 224 is o u(t) t5:TB = { 2e Tf. it is easy to verify that the Laplace transform of ~(t) is unity. Once the impulse response of a linear system is known. shown in Example 2~5~3 as: d2 y(t) dy(t) 2 ~ + 2~ClJI1 cit + wlly(t) Hence. Mathematical Foundation Ii (t) 2e . can be found by using the transfer function. given zero initial g(t) = (Vir etw"t sinwnVl 2:: 0 Jl{2 (2212) . i. . . } ! i ! t. . LIo(t)] = 1 as e ~ O.68 ~ Chapter 2.c(u(t)) = Y(s) F(s) (2209) as the transfer function of the system. For t = 0 in Eq.c(u(t)) = s2 + 2~(Vns + w~ Similar to Example .to t = 00.c(y{t)) . . the output of the system y(t). which is defined as the output when the input is a unitimpulse function o(t).C(y(t)) . The equation representing Fig. w. (2116) and noting the actual limits of the integral are defined from t = 0. (2186). .<t<r+& it (2208) o t~T+e For it = 1.
Toolbox 271 The unit impulse response of Eq. using the convolution properties of Laplace transforms. () = cosI ?:. £[y(t)] = £lus *g(t)] = C t [io usg(t .~Sin {J)1l~t+(} 1 .den). G=1000/(sA2+34.27 Impulse Response and Transfer Functions of Linear Systems For a unitstep input u{t) ~ 69 = u.5*s+1000). den = [1 34 ... (2213). the output y(t) is therefore Jo or etW. Impulse Response 20~~~~~~r~~~' .2 where. [1000] . impulse(G). G = tf (num.?.T)dT = ] s0(8) (2213) From the inverse Laplace transform of Eq.{t).f r llsg(t . s ::: t f (' s'). impulse (G). (2194) may be obtained using Alternatively: num::.T)dr ( ) t~O y(t)=l. 5 1000] .
even computer simulations often start with transfer functions. Let us consider that the inputoutput relation of a linear timeinvariant system is described by the following nth~order differential equation with constant real coefficients: (2217) The coefficients ao.. differential equations of the fonn of Eq..q(t). and impulse response . ant and bo. + hiS + bo U(s) sn + all _1s1. the output response yet) for t ~ to is determined by solving Eq.. rather than with differential equations. although efficient subroutines are available on digital computers for the solution of highorder differential equations. +blS+bo)U(s) (2218) The transfer function between u(t) and yet) is given by G{s) = Y(s) = bmsfll + bm_1S . Once the input u(t) for t ~ to and the initial conditions ofy(!) and the derivatives ofy(t) are specified at the initial time t = to. so it is more convenient to derive the transfer function directly from the differential equation. the inputoutput relation of a linear timeinvariant system with continuousdata input is often described by a differential equation. . .70 Chapter 2. SingleOutput Systems) The transfer/unctioJl of a linear timeinvariant system is defined liS the Laplace transform of the impulse response. The transfer function G(s) is defined as G(s) = £lg(I)] (2215) The transfer function G(s) is related to the Laplace transform of the input and the output through the following relation: Y{s) G(s) = U{s) (2216) with all the initial conditions set to zero. Mathematical Foundation 27. we simply take the Laplace transform on both sides of the equation and assume zero initial conditions.. (2217) are seldom used in their original form for the analysis and design of control systems. (2217). the method of using differential equations exclusively is quite cumbersome. output y(t). Although the transfer function of a linear system is defined in terms of the impulse response.. the basic philosophy of linear control theory is that oj"cieveloping analysis and design tools that lvill avoid the exact solution of the system differential equations.. singleoutput (SISO) system. +als+ao)Y(s) = (blllsm+bm_tSmI + .. The result is (s"+a/l_tsll1 + . It should be pointed out that. and Y(s) and U(s) are the Laplace transforms of y(t) and u(t)t respectively. In classical control theory.. in practice. However.+ . Let G(s) denote the transfer function of a singleinput. + als + ao m 1 (2219) . To obtain the transfer function of the linear system that is represented by Eq. except when computersimulation solutions are desired for final presentation or verification.. witiz all the initial conditions set to zero. from the standpoint of linearsystem analysis and design..2 Transfer Function (SingleInput. al ~ .. hi ~ ..1 + . bm are real constants... Thus. with input u(t).
• All initial conditions of the system are set to zero. time. It is not a function of the real variable. or any other variable that is used as the independent variable. the transfer function between a pair of input and output variables is the ratio of the Laplace transform of the output to the Laplace transform of the input. Alternately.. Becam. + alS + ao == 0 (2·220) Later we shall show that the stability of linear. the transfer function is called proper. In general. the transfer function between the jth input and the ith output is defined as (2221 ) with Rk(S) 0. the transfer function is a function of z when the z~ transfonn is used (refer to Appendix D). 275 Transfer Function (Multivariable Systems) The definition of a transfer function is easily extended to a system with multiple inputs and outputs.p. the ith output transform is written (2222) = . 274 Characteristic Equation The characteristic equation of a linear system is defined as the equation obtained by setting the denominator polynomial of the transfer function to zero. • The transfer function between an input variable and an output variable of a system is defined as the Laplace transform of the impulse response. from Eq. when all other inputs are set to zero. single~output systems is completely governed by the roots of the characteristic equation. k = 1. the total effect on any output due to all the inputs acting simultaneously is obtained by adding up the outputs due to each input acting alone. . If n = m. is said to be strictly proper if the order of the denominator polynomial is greater than that of the numerator polynomial (Le. 273 Proper Transfer Functions The transfer function in Eq. if a linear system has p inputs and q outputs. A system of this type is often referred to as a multivariable system. For discrete~data systems modeled by difference equations. (2~217) may be used to describe the relationship between a pair of input and output variables.. lis defined with only thejth input in effect~ whereas the other inputs are set to zero. sl1 is + an_lsI! 1 + . The transfer function is improper if m > n. singleinput. the characteristic equation of the system described by Eq..e the principle of superposition is valid for linear systems. .2~7 Impulse Response and Transfer Functions of Linear Systems ~ 71 The properties of the transfer function are summarized as foIIows: • The transfer function is defined only for a linear timeinvariant system. Thus. • The transfer function is independent of the input of the system..2. Note that Eq. n > m). In a multi variable system~ a differential equation of the fonn ofEq. It is not defined for nonlinear systems. • The transfer function of a continuousdata system is expressed only as a function of the complex variable s. k¥=j. When all the p inputs are in action...
and GIl (s) G(s) = G2~ (s) [ Glp(S) ] G2J~(S) (2226) Gql (s) Gqp(s) is the q x p transferfunction matrix. An unstable system is generally considered to be useless. in matrixvector form~ yes) = G(s)R(s) where (2223) YleS)] Y2(S) . ~ 28 STABILITY OF LINEAR CONTROL SYSTEMS From the studies of linear differential equations with constant coefficients of SISO systems~ we learned that the homogeneous solution that corresponds to the transient response of the system is governed by the roots ofthe characteristic equation. we can classify stability as absolute stability and relative stability. Basically. it is a yes or no answer. nonlinear. We shall deal only with the stability of linear SISO timeinvariant systems in the following discussions. and timevaryingthe definition of stability can be given in many different forms. RI (s) R2(S) R(s) = (2225) is the p x 1 transformed input vector. timeinvariant. the most important requirement is that the system must be stable. Once the system is found to be stable. yes) = lYq(S) is the q x 1 transformed output vector.. all the initial conditions of the system are zero. In preparation for the definition of stability. . The zerostate response is due to the input only. the design oflinearcontrol systems may be regarded as a problem of arranging the location of the poles and zeros of the system transfer function such that the system will perfonn according to the prescribed specifications. When all types of systems are consideredlinear.. Chapter 2.72 . Among the many forms of performance specifications used in design. and this degree of stability is a measure of relative stability.state response. we define the two following types of responses for linear timeinvariant systems: • Zero. it is of interest to determine how stable it is. For analysis and design purposes. Mathematical Foundation It is convenient to express Eq. Absolute stability refers to whether the system is stable or unstable.
From the principle of superposition. respectively.. output. or Jy(t) I ::.2·9 BDunded~lnput. Bounded~Output (BIBO) StabilityContinuousData Systems ~ 73 • Zeroinput response. (2234) . the total response is written Total response = zerostate response + zeroinput response The definitions just given apply to continuousdata as well as discretedata systems. if its output y(t) is bounded to a bounded input u(t). boundedoutput) stable. Then. all the inputs are zero. The zeroinput response is due to the initial conditions only. for any finite positive Q. and g(/) be the input. With zero initial conditions. The convolution integral relating u(t). BOUNDEDOUTPUT (BI80) STABILITYCONTINUOUSDATA SYSTEMS Let u(t). or simply stable. y(t).r)g(r)dr Jo (2227) Taking the absolute value of both sides of the equation. and the impulse response of a linear timeinvariant system. y(t). and get) is y(t) = (OC. we get (2228) or (2229) If u(t) is bounded~ lu(t)J ~ M where M is a finite positive number. the system iy said to be BIBO (boundedinput. N < 00 where N is a finite positive number. ~ 29 BOUNDEDINPUT. the following condition must hold: (2232) (2233) Or. when a system is subject to both inputs and initial conditions. (2230) (2 .u(t . 231) Thus. if yet) is to be bounded.
as G(s) == .)] == f" g(l)e"dt (2235) Taking the absolute value on both sides of the last equation. Thus. (2234). Then. (2"234) implies that the area under the Ig(r)lversusr"curve must be finite . we shall define zeroinput stability and asymptotic stability and establish their relations with BmO stability. we write the transfer function G(s). (2237) becomes 00 ~ {'X Mlg(t)Jdt = ('X 19(t)ldt ~) ~ (2239) which violates the BIBO stability requirement. A system is said to be unsfable if if is nul BlBO slable. Eq. they must all lie in the le/thalfsplane. sin wot. say. then the output will be of the fonn of t sin oo()t~ which is unbounded. 210 RELATIONSHIP BETWEEN CHARACTERISTIC EQUATION ROOTS AND STABILITY To show the relation between the roots of the characteristic equation and the condition in Eq. for BIBO stability. We shall show that the zeroinput stability also depends on the roots of the characteristic equation. cannot be located in the riglu"ha/fs"piane or on the jooaxis. where a is the real part of s. Zero"inpnt stability refers to the stability condition when the input is zero.74 ~ Chapter 2. y(t) can be expressed as y(t) = Lgdt)y(k)(to) k=O 1l1 (2240) ... = .e[g(. if the input is a sinusoid. (2236) becomes 00 ~ fX Ig(t)llea/Idt Jo (2237) axis~ a If one or more roots of the characteristic equation are in the righthalf splane or on the jw" 2:: 0. Let the input of an nthorder system be zero and the output due to the initial conditions be y(t). and the system is unstable . Mathematical Foundation The condition given in Eq. the roots of the characteristic equation. according to the Laplace transform definition. . when s assumes a value of a pole of = 00. 211 ZERO"INPUT AND ASYMPTOTIC STABILITY OF CONTINUOUSDATA SYSTEMS In this section. G(s) lestl = lentl. or the poles o/G(s). al s = jwo and s jmo. we have (2236) Because G(s). in other wor{l~. then (2"238) Eq. When a syst~1Il has ruuls un lhe jwaxis. and the system is driven only by its initial conditions.
to satisfy the two conditions in Eqs. 11.\:v.jor linear timeinvariallt systerns. + j(f)./or tiny set offinite y<k)(tO). In other words. (·x for all t ~ to (2242) lim ly(t)J = 0 (2243) Because the condition in the last equation requires that the magnitude of y(t) reaches zero as time approaches infinity~ the zeroinput stability is also known at the asymptotic stability. Mathematically. we see that. Then.2 1 • • • .t + i=t m IJmI i:::::O E Liles!t (2246) where KJ and Li are constant coefficients. Because the exponential terms eSil in the last equation control the response y(t) as t ~ 00. and the rest are of multiple order.fJponse y(t). there exists a positive number M~ which depends on y(k)(to). or stable. zeroinput. the condition in Eq. the foregoing definition can be stated: A linear time~i111'aritlllt system is zeroinput stable iJ. Taking the absolute value on both sides of Eq. the system is unslclble. \y(t}\ S M < 00 and 2. yet) will be of the form: y(t) = LKies. (2242) requires that the following condition be true: . and asymptotic stability all have the same requirement that the roots of the characteristic equation must all he located in the lefthalf splll11e. The zeroinput stability is defined as follows: If the ::er()~input re. if a . Thus. the real parts of Sj must be negative. For this reason. (2242) and (2243).~·tem is BIBO stahle. (2240)T we get (2244) Because all the initial conditions are assumed to be finite. subject to the finite initial conditions. From the preceding discussions. it must also he ::eroinput or (lsymptoticctl£v sltlble.1 2:)gk(t)1 < 00 k=O for all t~ 0 (2245) Let the n characteristic equation roots be expressed as Sj = CI. The latter . y{k)(tO)' reaches zero as t approaches infinity. i'= 1. otherwise. BIBO. we shall simply refer to the stability condition of a linear system as stable or unstable...2~ 11 Zero~lnput and Asymptotic Stability of Continuous~Data Systems ~ 75 where dky(t) y (k) ( ) =kto dt I l=tn (2241) and Bk(t) denotes the zeroinput response due to y(k)(to). the roots of the characteristic equation must all be in the lefthalf splane. such that 1. the system is said to be zero~illput stable. if In of the 1l roots are simple.
.) = 0 for any i for simple roots. Chapter 2. n (at least one simple root in the righthalf splane or at least one multipleorder root on the ai Unstable jwaxis) condition refers to the condition that at least one of the characteristic equation roots is not in the lefthalf splane. stable) Unstable due to the pole at s = I 20(s . . it is in complexconjugate pairs. . if the system were designed to be an oscillator. For practical reasons. EXAMPLE 2111 The following closedloop transfer functions and their associated stability conditions are given...76 . 2~ . . (All the roots are in the lefthalf splane. and no ai > 0 For i = I. . no multipleorder roots on the jwaxis. 2.(s+ l)(s + 2)(s + 3) M(s) _ 20(s + 1) . Similarly..(s . An exception to this is if the system wel'e intended to be an integrator (or. Mathematical Foundation TABLE 25 Stability Conditions of Linear ContinuousData TimeInvariant SISO Systems Stability Condition Asymptotically stable or simply stable Marginally stable or marginally unstable Root Values (1i < 0 for all i. 2. If any of the roots is complex. simply.. The following example illustrates the stability conditions of systems with reference to the poles of the system transfer functions that are also the roots of the characteristic equation.n. Let the characteristic equation roots or eigenvalues of A of a linear continuousdata timeinvariant SISO system be Si = Gj + jw.1) + 4) Marginally stable or marginally unstable due to s = ±j2 Unstable due to the multipleorder pole at s Stable if the pole at s M(s) = M(s) _ to (82 + 4)2(s + to) 10 = ±j2 . in the case of control systems. i = 1. i = 1.. I n (at least one simple root.sA + 30s3 + s2 + lOs = 0 is placed intentionally . note exceptions) OJ> 0 for any i.. and the system would be regarded as stable... then the system would have 1'00t(s) at s = 0 and would be considered stable. .. The possible stability conditions of the system are summarized in Table 25 with respect to the roots of the characteristic equation.. we often refer to the situation in which the characteristic equation has simple roots on the jwaxis and none in the righthalf plane as marginally stable or marginally unstable.1)(82 + 2s + 2) M(s) _ . or ai = 0 for allY multipleorder root..(s + 2)(s2 BIBD or asymptotically stable (or. Because the roots of the characteristic equation are the same as the eigenvalues of A of the state equations. M(s) _ 20 . i = 1\ 2. n.. the stability condition places the same restrictions on the eigenvalues. the characteristic equation would have simple roots on the jwaxis.. . a velocity control system). and n roots in the righthalf splane.
Nyquist criterion. the roots of the characteristic equation can be found using MATLAB as demonstrated in various MATLAB Toolbox windows discussed earlier in this chapter. This diagram is a plot of the magnitude of the loop transfer function G(jw)H(jw) in dB and the phase of G(jw)H(jw) in degrees. there is no need to compute the complete system response to determine stability. These programs are discussed in detail in Appendix G. The regions of stability and instability in the sp]ane are illustrated in Fig. This topic is discussed in detail in Chapter 8. 225. RouthHurwitz criterion. This criterion is an algebraic method that provides information on the absolute stability of a linear timeinvariant system that has a characteristic equation with constant coefficients. For all practical purposes. When the system parameters are all known. The methods outlined in the following list are well known for determining the stability of linear continuousdata systems without involving root solving. and the concepts of loop transfer function and closeloop systems are discussed in Chapter 3. which is discussed at the end of this chapter. This criterion is a semigraphical method that gives information on the difference between the number of poles and zeros of the closedloop transfer function that are in the righthalf splane by observing the behavior of the Nyquist plot of the loop transfer function. 2. all versus frequency W· The concepts of loop transfer function and closedloop systems are jm splane Stable region Uns table re gion o Stable re gion Unstable regiull Figure 225 Stable and unstable regions in the .vplane.212 Methods of Determining Stability 77 212 METHODS OF DETERMINING STABILITY The discussions in the preceding secti ons lead to the conclusion that the stability of linear timeinvarian t SISO sys lems can be determlned by checking on the location of the roots of the characteristic equation of the system. The number of roots that lie on the jwaxis and in the righthalf splane is also indicmed. 1. . so a RouthHurwitz stability routine has also been developed for this textbook (tfrouth). The Transfer Function Symbolic Tool (tfsym) developed for this chapter may also be utilized to find the transfer function poles and zeros. For design purposes. The criterion tests whether any of the roots of the characteristic equation lie in the righthalf splane. there will be unknown or variable parameters imbedded in the characteristic equation. See the end of this chapter for some examples. Bode diagram. 3.
To ensure the last equation does not have roots with positive real parts. The two necessary conditions for Eq. ~ 213 ROUTHHURWITZ CRITERION The RouthHurwitz criterion represents a method of determining the location of zeros of a polynomial with constant real coefficients with respect to the left half and right half of the . .. (2247) as follows: which relate the coefficients of a. . for it is quite possible that an equation with all its coefficients nonzero and of the same sign still may not have all the roots in the left half of the splane. depending on the particular situation. Details of the RouthHurwitz stability criterion are presented in the following section. (2247) to have no roots in the righthalf splane can easily be checked by inspection of the equation. these conditions are not sufficient.~ products of the roots taken three at a tIme ~ au (2250) aO .~ .""' all roots ~ all (2248) an .= (1)n prodllct~ of all the roots all (2251) Thus.= .jinding computer programs can solve for the zeros ofa po~vnomial with ease. Because root. . algebra~ These conditions are based on the laws of Eq. Consider that the characteristic equation of a linear time~variant SISO system is of the fonn (2247) where all the coefficients are real. All the coefficients of the equation have the same sign. 2.1 . However. None of the coefficients vanish. it is necessary (but not sUfficient) that the following conditions hold: 1. most of the analysis and design techniques on control systems represent alternate methods of solving the same problem.vplane. as will be evident throughout the text. without actually solving for the zeros.78 ~ Chapter 2.2 = " pro dncts 0 f the roots taken two at a tIme an (2249) ~~ . all these ratios must be positive and nonzero unless at least one of the roots has a positive real part. Thus. This topic is discussed in detail in Chapter 8. Mathematical Foundation discussed in Chapter 3..= . The designer simply has to choose the best analytical tool. The stability of the closedloop system can be determined by observing the behavior of these plots. the value of the RouthHurwitz criterion is at best limited to equations with at least one unknown parameter.
illustrated here for a sixthorder equation: a6s6 + ass5 + .. coefficients. Once the Routh's tabulation has been completed. .Cao =F E Fao .a6 x 0 as A =aO =0 =0 0 0 0 Aa~i .. However.. and the second row consists of the second. The column of s's on the left side is used for identification purposes.1 Routh~Hurwitz Criterion <1 79 Routh's Tabulation The Hurwitz criterion gives the necessary and sufficient condition for all roots of Eq. The fonowing conclusions are made: The roots of the equation are all in the left half of the splane if all the elements of the first column of the Routh's tabulation are of the same sign. or those in the righthalf splane. The number of changes of signs in the elements of the first column equals the number of roots with positive real parts. the last step in the application of the criterion is to investigate the signs of the coefficients in the first column of the tabulation.aSaO =B =D =ao astlo .Ex 0 =ao F 0 0 0 0 0 0 This array is called the Routh's tabulation or Routh's array. the evaluation of the n Hurwitz detenninants is tedious to carry out. fifth. third... + ellS + ao = a4 a3 0 Q2 at (2252) s6 s5 . and the last row of the Routh's tabulation should always be the ..coefficients. .\"4 a6 a5 aSa4 a6 a 3 ao 0 as s3 =A aSa2 .a5B A =c =E x 0 .a6 a l as All! . which contains information on the roots of the equation. as shown in the following tabulation: The next step is to form the following array of numbers by the indicated operations.. fourth.p row. . The criterion requires that the equation's n Hurwitz determinants must all be positive.2·13 2~13 . all couIlting from the highestorder term. sixth. (2247) to lie in the left half of the splane. .. (2247) into two rows. The reference column keeps track of the calculations. is to arrange the coefficients of the equation in Eq. The first row consists of the first. But Routh simplified the process by introducing a tabulation method in place of the Hurwitz determinants. now called the RouthHurwitz criterion. The first step in the simplification of the Hurwitz criterion.as x 0 A A s2 s1 BCAD C Cao A x 0 C CxOAxO C sO ED . .
9331i 2132 Special Cases when Routh's Tabulation Terminates Prematurely The equations considered in the two preceding examples are designed so that Routh's tabulation can be carried out without any complications.1 Consider tl1e equation 2f:j + s3 + 3s2 + 5s + 10 = 0 (2253) Because the equation has no missing terms and the coefficients are all of the same sign. 4444i 1. the following difficulties may occur. which prevent Routh's tabulation from completing properly: .755 ± j1.4444i o . (1)(3) . which cause the system to be unstable. the sufficient condition must still be checked.. (2253)..933 and s = 0. 1 The roots of the polynomial in Eq. 9331i 1.(2)(5) 1 = 7 10 0 (7)(5) .13. Routh's tabulation is made as follows: . 7555 .005 ± jO.1. the last two roots are in the righthalf splane.0055 + O. Sign change sl . Toolbox 213.~4 2 3 5 10 0 $'"" Sign change s. However.7555 + 1. it satisfies the nece!.444. Clearly.p Because there are two sign changes in the first column of the tabulation.0055 . the equation has two roots in the light half of the splane.1. »clear all = [2 1 3 5 10] % Define polynomial 2 * 81\4+sI\3+3 * 81\2+5 i( 8+10 2 1 3 5 10 » p p= » roots(p) ans = 0..sary condition for not having roots in the righthalf or on the imaginary axis of the splane. Mathematical Foundation The following examples illustrate the applications of the RouthHurwitz criterion when the tabulation terminates without complications. Solving for the roots of Eq.80 Chapter 2. (2253) are obtained using the following sequence of MATLAB functions.43 0 0 0 0 . we have the four roots at s = . EXAMPLE 2.O.(1)(10) 7 10 = 6. Depending on the coefficients of the equation.
and Routh's tabulation cannot continue. it indicates that one or more of the following conditions may exist: 1. 0 sO 3 0 Because there are two sign changes in the first column of Routh's tabulation. Starting with the s2 row. we need to apply the RouthHurwitz criterion. we also get some of the roots of the original equation. we rep/ace the zero element in the first column by an arbitrary small jJositilte number £. This is illustrated by the following example. In the first case. the equation in Eq. Form the auxiliary equation A (5) tbe row of zeros. Routh's tabulation is carried out as follows: S4 1 2 3 s3 s2 l 2 0 0 3 Because the first element of the . To overcome this difficulty. 3. (2154) has two roots in the righthalf splane. 3 Sign change Sign change .091 ± jO. and then proceed with the tabulation.406 ± j1. The roM:. 2. of the auxiliary equation also satisfy the original equation. the last two roots are clearly in the righthalf splane.902 and s = 0. To remedy the situation. that is. The elements in one row of Routh's tabulation are all zero.v = 1 ± j 1. ~ 3 [. The situation with the entire row of zeros can be remedied by using the auxiliary equation A(s) = O.293. The auxiliary equation is always an even polynomial. but the others are not.2~ 13 RouthHurwitz Criterion • 81 1. It should be noted that the tmethod described may not give correct results if the equation has pure imaginary roots. if a zero appears in the first element of a row. 2. To continue with Routh's tabulation when a row of zero appears. which is formed from the coefficients of tbe row just above the row of zeros in Routh's tabulation. s = 1 ± jl. the results are as follows: S2 1. only even powers of s appear. Solving for the 1'00ts of Eq. (2254). by solving the auxiliary equation. the elements in the next row will all become infinite. we get II = 0. the elements in the Sl row would aU be infinite. when aU the elements in one row of Routh's tabulation are zeros before the tabulation is properly tcnninated."{2 row is zero.\'1 2c. Thus. In the second special case. 2. . this gives dA(s)/ds = . and then proceed \vith Routh'8 tclbuitltion. Take the derivative of the auxiliary equation with respect to s. The equation has pairs of complexconjugate roots forming symmetry about the origin of the splane. for example. ~ EXAMPLE 2132 Consider the characteristic equation of a linear system s4 +$3 +2$2 +2s+3 = 0 (2254) Because all the coefficients are nonzero and of the same sign.3 r. = 0 by using the coefficients from the row just preceding n. we conduct the following steps: 1. we replace the zero in the s2 row with a small positive number E. The equation has at least one pair of real roots with equal magnitude but opposite signs. The first element in anyone row of Routh's tabulation is zero. The equation has one or more pairs of imaginary roots.
Mathematical Foundation 3.3 sO 0 1.5 x 107 K 410. The following example illustrates the realistic value of the RouthHurwitz criterion in a simple design problem... Chapter 2.382 + 1.3 Consider the following equation.1. ~ = = ~ EXAMPLE 2134 Consider that a thirdorder control system has the characteristic equation (2258) The RouthHurwitz criterion is best suited to determine the critical value of Kfor stability. the equation has two roots on the jataxis. which are also two of the roots of Eq. Solving the auxiliary equation in Eq. Replace the row of zeros with the coefficients of dA(S)/d5 = O.13. Continue with Routh's tabulation in the usual manner with the newly formed row of coefficients replacing the row of zeros. 4. that is.3 . the value of K for which at least one root will lie on the jeuaxis and none in the righthalf spl ane. (2256).5 x 107 K . (2258) is made as follows: s3 + 3408.204. Routh's tabulation of Eq. . Because all zeros occurring in a row that corresponds to an odd power of s creates an auxiliary equation that has only even powers of s. the roots of the auxiliary equation may all tie on thejwaxis.\"1 1. These imaginary roots caused the initial Routh's tabulation to have the entire row of zeros in the Sl row. Interpret the change of signs.82 . we form the auxiliary equation using the coefficients of the S2 row: A(s) = 4s2 + 4 The derivative of A(s) with respect to sis =0 (2257) Sl dA(s) ds =88 = 0 from which the coefficients 8 and 0 replace the zeros in the remaining portion of the Routh's tabulation is Sl sO row of the original tabulation. we get the two roots at s j and s j. and the system is marginally stable. the equation in Eq. For design purposes.204. (2257) does not have any root in the righthalf 8plane. if any.. we can use the allzerorow condition to solve for the marginal value of a system parameter for system stability.5 x 107 K 3408.000 3408. Thus. The 8 0 coefficients of dA(s)j ds 4 Because there are no sign changes in the first column of the entire Routh's tabulation. EXAMPLE 2. of the coefficients in the first column of the Routh's tabulation in the usual manner.36 X 107 .. which may be the characteristic equation of a linear control system: 55 + 4S4 + 8s3 + si + 7s + 4 = 0 55 s4 s3 (2255) Routh's tabulation is I 4 8 7 8 4 4 6 6 0 4 s2 sl 0 0 Because a row of zeros appears prematurely. OOOs + 1. (2255). 5.5 X 107 K = 0 1.
57.57 (2261) If we let K = 273. Thus.. we have K < 273.528 Toolbox 213·3 Refer to Section 2142 for the MATLAB symbolic tool to solve this problem.5 As another example of using the RouthHurwitz criterion for simple design problems. we substitute K :::: 273. and from the Sl row. Routh's tabulation of Eq.213 RouthHurwitz Criterion .27 rad/sec . EXAMPLE 2. To find these roots. (2259). all the coefficients in the first column of Routh's tabulation must have the same sign. (2258) will have two roots on the jwaxis.j1097. all the roots ofEq. .1036 X 109 = 0 (2262) which has foots at s = jl097 and s = . consider that the characteristic equation of a closedloop control system is s3 + 3Ks2 + (K + 2)s + 4 = 0 K+2 4 0 (2263) It is desired to find the range of K so that the system is stable.3s2 + 4. and the corresponding value of K at these roots is 273.3 and (2259) (2260) From the inequality of Eq.. which is obtained from Routh's tabulation by using the coefficients of the S2 row. Therefore. if the system is operated with K = 273.. half splane~ and.57.2 Refer to Section 2142 for the MATLAB symbolic tool to solve this problem. 83 Toolbox 213.. the conclition of stability is K> O.57 in the auxiliary equation.36 x 107 . (2260) gives K > O.4 > 0 or (2264) (2265) K < 2.13. the condition of stability is 3K2 + 6K . the condition of K for the system to be stable is O<K<273..57. and the condition in Eq. (2. thus. the zero~input response of the systenl will be an undamped sinusoid with a frequency of 1097.258) must be in the left. the characteristic equation in Eq. A(s) = 3408.4 3K 4 From the i row. Also. .57.5 x 107K > 0 3408.1. For the system to be stable. This leads to the foHowing conditions: 410.528 or K > 0.. (2263) is s' s2 sl sO 3K 3K(K +2) .
Enter the transfer function. containing exponential functions or sinusoidal functions of s. for example. such as the unit circle in the zplane. It should be reiterated that the RouthHurwitz criterion is valid only if the characteristic equation is algebraic with real coefficients. to get the time response. Mathematical Foundation When the conditions of K > 0 and K > 0. To run Programs. 228. As instructed. The stability boundary is thejwaxis of the splane. the RouthHurwitz criterion simply cannot be applied. Another limitation of the RouthHurwitz criterion is that it is valid only for the determination of roots of the characteristic equation with respect to the left half or the ri ght half of the splane. it is apparent that the latter requirement is more stringent. as shown in Fig. .84 ~ Chapter 2. Transter Function and InveJse laplace StateSpace StateSpace with Init. 226. K must satisfy K > 0. go to MAT LAB Command window after clicking each pushbutton. for the closedloop system to be stable.528 (2266) The requirement of K < . ~ 214 MATLAB TOOLS AND CASE STUDIES Description and Use of Transfer Function Tool If you have access to the MATLAB Symbolic Toolbox. The instructions appear in a Help Dialog window. you may use the ACSYS Transfer Function Symbolic Tool by pressing the appropriate button in the ACSYS window or by typing in tfsym in the MATLAB command window. 2141 You must run this program within the MATLAB command window. Thus. or if the equation is not algebraic. The Symbolic Tool window is shown in Fig.2. Figure 227 The Symbolic Help Dialog window. ) Help Dialog '~ You must have access to MAT LAB Symbolic TQolbol<. 227. If anyone of the coefficients is complex. Condo Close and EKit Figure 226 The Transfer Function Symbolic window. which is the stability boundary of discretedata systems (Appendix H). press the "Transfer Function and Inverse Laplace " button to run the program. as shown in Fig.528 are compared.528 is disregarded because K cannot be negative. Click the "Help for 1st Time User" button to see the instructions on how to use the toolbox. The criterion cannot be applied to any othcr stability boundaries in a complex plane.
John Wiley & Sons. for the transfer function Y(s) S(5 2 (2268) + 34.g.(s2 + 2~a>lrs + a>~) 1000 _ (I). (2267) time representation for a unitstep input. as shown in Fig. 85 Transfer Function Symbolic.53/2*i))"t + (1/2 . as we demonstrated in Toolbox 25l for Example 251.6)/«s+ 1)*(5+2)*(s+3» 5s+3 (s+i)(. (2269) using the tfsym tool. So to obtain Eq. 228. the MATLAB code that appears in Toolbox 2131 is the easiest .5s + 1000) . use the following transfer function: 5s+3 Gs _ ( ) .g..13/40*i)* exp« 69/4+ 53/2"'i)*t) 214~2 4 MATLAB Tools for Stability The easiest way to assess stability of known transfer functions is to find the location of the poles.• Use the following inpllt format: (5+2)*(5"3+2*5.:rr.6) (~+3)·(. To find the time representation of Eq.5(S+ l)(s + 2)(s + 3) and repeat the previous steps..2w14 MATLAB Tools and Case Studies . the time representation of this system is obtained as 1+( 1/2 + 13/40"i)* exp« 69/4 . For that purpose..+2)(~+i) Inverse Laplace transform: Gtime::: exp(t)+7*exp(2*t)6*exp (3"'t) Figure 228 The inverse Laplace transform of Eq. e.i:.. (2267) for an impulse input.. ~ EXAMPLE 2~141 Find the inverse Laplace transform of the transfer function G(s) _ 5s + 3 _ 5s + 3 . in the MATLAB command window. the user may combine the input transfer function (e.(s + 1)(5 + 2)(8 + 3) . or by utilizing the tfsym function. 1)/(s*(5"2+2*s+ I)) Enter 0=5*(s+O. (2267) for a different input function such as a step or a sinusoid. Similarly.+i)w(~+3·) G in polynomial form: Transfer function: .83 + 6s2 + lIs + 6 (2267) You can do this either by using the i/aplace command in the MATLAB command window.i.+6 G factored: 5s+3 Zero/pole/gain: 5(5+0.::'3+6. lis for a unitstep input) with the transfer function in the TFtool input window. © Kuo & Golnaraghi 8th Edition.
. The system is therefore unstable because of two positive poles. including the "Transfer Function Symbolicn (tfsym) and the '~TransferFunction Calculator~' (trcal). Press the URouthHurwitz button to form the Routh table and conduct the RouthHurwitz stability test. sA 3+sA2+s+1) or in vectorial (e.. as shown in Fig. the user must refer to the MATLAB command window. we introduce the tfrouth tool. the poles of the system. say k. 3. ~ . [1 1 1 1]) fonns. many of the other tools within ACSYS software may also be used to find the poles of the closedloop system transfer function. ful1owe:d by clicking the: ""RouthHurwitz" button to get the RouthHurwitz matrix. let's use tfrouth for the following polynomial: 2s4 + s3 + 3s2 + 5s + 10 = 0 (2270) In the MATLAB command module.86 ~ Chapter 2. enter it in the box designated as "Enter Symbolic Parameters. EXAMPLE 2144 Revisiting Example 2133.g. Mathematical Foundation way for finding the roots of the characteristic equation polynomiali." respectively. as shown in Fig. and more importantly it may be utilized for controller design applications where it is important to assess the stability of a system for a controller gain.. 2. we expect to see two unstable poles. S. for sA 3 +k1*SA2+k2*s+ 1. EXAMPLE 2N143 Consider Example 2132 for characteristic equation of a linear system: 8 4 + . 230. type in "tfrouth" and enter the characteristic Eq. let us solve some of the earlier examples in this chapter. you need to enter "k1 k2" in the "Enter Symbolic Parameters n box. Press the "RouthHurwitz" button and check the results in the MATLAB command window. In case you wish to assess the stability of the system for a design parameter.. You may also conduct a more thorough stability study of your system using the root locus and phase and gain margin concepts utilizing the "Controller Design Tool.'13 + 2s2 + 2s + 3 = 0 (2271 ) After entering the transfer function characteristic equation using tfrouth and pressing the "RouthHurwitz" button~ we get the results shown in Fig. Type "tfrouth" in the MATLAB command module within the "tfsymbolic" directory. To better illustrate how to use tfrouth. These topics will be thoroughly discussed in Chapter 9. which may be used to find the Routh array. The steps involved in setting up and then solving a given stability problem using tfrouth are as follows.. U .g.. However. because of the final two sign changes. The results match Example 2142. In this section. use tfrouth to study the following characteristic equation: s5 +4s4 +8s':\ +si +/s+4=0 (2272) to get the results shown in Fig. To see the complete Routh table. 1. 4. (2270) in polynomial form. 229.. As a result. .." For example. ~ . Enter the characteristic polynomial in symbolic (e. EXAMPLE 2142 Recall Example 2131.e. followed by entering the polynomial sl\3 + kl *s1\2 + k2*s + 1 in the "Characteristic Equation" box. The Routh's array first column also shows two sign changes to confirm this result. 232. 231.
ki"2]' in the Characteristic Equation box and 'kp ki' in the symbolic variables text box. 11 Close Figure 229 Entering characteristic polynomial for Example 2142 using the tfrouth module.:" : "~i~~~~~r·~~~~.".:'~: . RouthHurwitz Matrix: [2 [ 3 5 10 0 [ I [ [ 7 [ [45/7 [ 10 0 0 0 0 [10 0 There are two sign changes in the first column. 3*kp.Enter Characteristic Equation: 1 Enter the charac. or in symbolic form with complex variable 's'. Enter Symbolic Parameters (For example:) I Characteristi. 3'kp*s + kiA2): enter '[1.:. ex: The following are all equivalent: '(S"2 + 7*s + 12).' ex: For (sA2 of. Enter All symbolic parameters in the box labeled 'Enter Symbolic Parameters .:'~:':~':. Equation c I [':':: .: :~:" ":·': : : ":":. and (8+4)*(8+3)'. OR: Type in ($A2 + 3*kP"s + ki"2) in the Characteristic Equation box and 'kp 10' in the symbolic variables text box . :'.:·"·: ·" : :.: :'~.214 MATLAS Tools and Case Studies 87 . after using the RouthHurwitz test. . [1 712J.teristic equation using a vector of polynomial coefficients. Figure 230 Stability results for Example 2142.
after using the RouthHurwitz test. Auxiliary polynomial is used. 2~33. after using the Routh~Hurwitz test.. . the system is unstable. and 235 for more details.. RouthHurwitz Matrix: [ 2 2 3 3 0 0 [ [ [ I eps I [3 + 2 eps 0 0 0 0 [ eps I. Figure 232 Stability results for Example 2144. Chapter 2. The unstable poles of the system may be obtained directly by obtaining the roots of the auxiliary polynomial: A(s) = 4s2 + 4::::: 0 ~ EXAMPLE 2·145 Considering the characteristic equation of a closedloop control system s3 + 3Ki + (K + 2)5 + 4 = 0 (2274) It is desired to find the range of K so that the system is stable. because of the final zero sign changes. Row of zeros found at rowS. 234. In the end. RouthHurwitz Matrix: [1 [ [4 [ 8 8 6 7 4 0 0 0 0 [6 [ [4 [ 4 0 0 [8 [ [4 There nre two sign changes in the first column. the program has automatically replaced the whole row of zeros in the fifth row with the coefficients of the polynomial formed from the derivative of an auxiliary polynomial fonned from the fourth row. Further. we expect to see no additional unstable poles.~I [ 3 There are two sign changes in the first column. In this case. Figure 231 Stability results for Example 2143. Epsilon is used. As a result. Mathematical Foundation First element of row3 is zero. the user is encouraged to make use of the software to solve examples and problems appearing in this chapter.88 . See Figs.
3'kp. ex: The following are all equivalent: '(sA2 + 7'3 + 12).:. OR: Type in (s"2 + 3'kp's .~. ·: ·.' ex: For (s"2 + 3"kp"s + ki"2): enter '[1.·"' '·~:·: ·:~: : : :· : ' : ·: : ·: ·. .~~:. and (8+4)' (s+3)'. Enter ALL symbolic parameters in the box labeled 'Enter Symbolic Parameters . RouthHurwitz Matrix: [ [ k +2 1 3k I [ 4 2 I I 4 + 3 k + 6 k 1 1 1 1 1 01 1 11f3 [ k [ [ 1 4 0 1 Figure 234 The Routh's :tnay for Example 2.: : : ': ':. [1 712).':...: .:. ki A 21' in the Characteristic Equation box and 'kp ki' in the symbolic variables text box.145..':.214 MATlAB Tools and Case Studies .:]1 Close Figure 233 Entering characteristic polynomial for Example 2145 using the tfrouth module. 89 II Routh Hurwitz .':. Enter Characteristic Equation:  Enter the characteristic equation using a veclor tl1 polynomial coefficients .: : ~~:~~~~. or in symbolic form with complex variable's' .. Enter Symbolic Parameters (For example:) I I Characteristic Equation sA3+3*k*SA2+(k+2)'s+4 r.~r:0.: .":::'::"... ki"2) in the Characteristic Equation box and 'kp ki' in the symbolic variables text box..
90 • Chapter 2. In this chapter. Mathematical Foundation »k=.2000 4. The final solution in the real domain is obtained by taking the inverse transfonn.4. »RH RH= [ I t. the roots of the characteristic equation must all be located in the left half of the splane.0000 2.0000 3. For engineering problems.=l~ » eval(RH) ans= 1.0000 0 There are two sign changes in the first column. ~215 SUMMARY In this chapter. the definitions of BIBO.0000 3. k+2 4 0 0 » eva1(RH) ans= 1. we presented some fundamental mathematics required for the study of linear control systems.0000 4. The solutions are first obtained in the transform domain by using the familiar methods of solving algebraic equations. The value of the RouthHurwitz criterion is diminished if the characteristic equation can be solved using MATLAB.9333 0 4. . This transform method is characterized by first transfonning the realdomain equations into algebraic equations in the transfonn domain. The Laplace transform is used for the solution of linear ordinary differential equations. I 4. zeroinput. Figure 235 The Routh's array for Example 2145. and asymptotic stability of linear timeinvariant continuousdata and discretedata systems are given. The necessary condition for a polynomial F(s) to have no zeros on the jwaxis and in the right half of the splane is that all its coefficients must be of the same sign and none can "al1ish. the transform tables and the partialfraction expansion method are recommended for the inverse transformation.0000 1. It is shown that the condition of these types of stability is related directly to the roots of the characteristic equation. Specifically. » k.0000 0. [ 1I3*(4+3*kI\2+6*k)/k.4000 1. we started with complex numbers and their basic properties leading to frequency domain mathematics and plots.6667 0 4. The necessary and sufficient conditions of F(s) to have zeros only in the left half of the splane are checked with the RouthHurwitz criterion.0000 0 There are no sign changes in the first column. 3*k. For a continuousdata system to be stable.
.Review Questions'" 91 t~ REVIEW QUESTIONS 1. Do you know how to handle the partialfraction expansion of a function whose denominator order is not greater than that of the numerator... tus(t)? 10. F(s). do you have to perform the partialfraction expansion? I F(s) = . 12. uit). What is a causal system'? Give the defining equation of the onesided Laplace trnnsform. Give the definitions of the poles and zeros of a function of the complex variable s.1'3 + 3s2 + Ks + K2 = 0 third~order 16. pes)  10 (82 + 5s + 1) .. 9. What are state equations'? 4. (b) The equation has two roots on the jltJaxis at s == j and . 2. Do you know how to handle the exponential term in performing the partialfraction expansion of F(s) = 10 e2s (s + l)(s + 2) 13. The third root is in the lefthalf splane.(s + l)(s + 2) ~'. Give the defining equation of the inverse Laplace transform. 6.3 (s + 5)' 15. What is the condition under which the theorem is valid? 8. for example. 5. In trying to find the inverse Laplace transform of the following function.3 system are 2 2 4 4 ..j. What are the advantages of the Laplace transform method of solving linear ordinary differential equations over the classical method·? 3.Ii Select the correct answer from the following choices: (a) The equation has one root in the righthalf splane.clfl(t)]!2(t)] in terms of F\(s) and F2(S).e[h(t)] = Fl(S). Give the Laplace transform of Jtt) shifted to the right (deJayed) by I:I in tenns of the Laplace transform of j(t).: 14. then find . If . 11. .e[fl(t)] = F\(s) and ... Can the RouthHurwitz criterion be directly applied to the stability analysis of the following systems'! (a) Continuousdata system with the characteristic equation s4 + 5s3 + 2s2 + 3s + 2e2s = 0 system with the characteristic equation (b) Continuous~data s4  5. The first two rows of Routh's tabulation of a . 7. Give the Laplace transfonn of the unitstep function. Give the expression of the finalvalue theorem of the Laplace transform. What is the Laplace transform of the unitramp function.
• John Wiley & Sons. 2nd Ed•• McGrawHill Book Company. T. Laplace Transforms PartialFraction Expansion 7. NJ. 1964. L. C. 2002. If the numbers in the first column of Routh's tabulation turn out to be all negative. 2nd Ed. (T) (F) 20. A. 2nd ed . When a row of Routh's tabulation contains all zeros before the tabulation ends. Answers to these review questions can be found on this book's companion Web site: www. 1999. . J. 1997. 2nd Ed . "On the Partial Fmction Expansion of a Rational Function with Multiple Poles by Digital Computer. C.92 . NJ. ~ REFERENCES L F. (d) The equation has two roots on the jwaxis at s = 2j and s righthalf splane. "A Partial Fraction Algorithm.• Prentice Hall.• Advanced Engineering MathemtllicJ. (T) (F) 21. Stubberud. Ogata. K. pp. Mar. Dynamic Systems. Palm. Fluid Power with Applications. 4th ed . 1967. this means (T) (F) that the equation has roots on the imaginary axis of the splane. B. 5th Ed. Additional References 9.ent. 1960. H. McGrawHill Book Company. K. "ChE400: Applied Chemical Engineering Calculations. NJ. C. Englewood Cliffs. Pottle. Esfandiari. The roots of the auxiliary equation. Lawrence. The third root is in the lefthalf . 13.• Addison· Wesley. I. Wylie. 15.• Prentice Hall.ohiou. Hildebrand. 1990. 12. ACI6. J_ L. Ill. Circuit Theory. New York.A(s) must also be the roots of the latter. 1997. = 2j. 1971. CadwaIIender. 2nd Ed . J. pp. 1997. Esposito. Shearer. Modern Control Engineering. Vu and R. 8.wiley. (T) 18. Automatic Control. The following characteristic equation of a continuousdata system represents an unstable system because there is a zero coefficient. J. Ana(ysis. The following characteristic equation of a continuousdata system represents an unstable system because it contains a negative coefficient. 2000. Kulakowski. S. R. PrenticeHall. W. A.comlcollege/golnaraghi. V.vnamit· Systems. Prentice Hall. 10. B. R. Feedback and Control Systems. and I. Vol. Distefano. Williams." IEEE Trans. R.vplane. Oct. and J. 2. NJ. Woods and K. NJ. New York. Modeling and Simuiatilm of Dynamic Systems. 3rd Ed . 1997. Analysis and Design of Dynamic Systems. B. Vol.• McGrawHill. CTl1. 14. of Routh's tabulation ofa characteristic equation 19. New York. L. Modern Control Engineering. J. 3.htm. Modeling. tT) (F) (F) = O. Cochin and W. 11.~ Prentice Hall." IEEE Trans. B. 2002. Gardner. IrwinIMcGrawHill. 6.edu/che/che400lHandouts% 20and%2OClass%20notes. NJ. 5. 4th Ed. Chapter 2. 4. Methods of Applied Mathematics. Jr. Prentice Hall. Watkins. 161162. F. und Control ojD. Ogata. 489491. Complex Variables. The third root is in the 17. O... Dynamic Modeling and Control of Engineering Systems. Linear Netu'orks and Systems. Kuo. then the equation for which the tabulation is made has at least one root not in the left half of the splane." hltp:llwww. 1965. Mathematical Foundation (c) The equation has two roots on the jwaxis at s = 2j and s = 2j.
. e2s (d) G(s) = 10s(s + l)(s + 2) Poles and zeros of a function are given. Golnaraghi. Khatwani.00 (b) Simple poles: I. Automatic Control. Automatic Control Systems. K. 5th Ed . 2003.Problems .1. Gantmacher. OC 23. 7th Ed. ~~. Automatic Control. Find the polar plot of the following functions: (a) GUltJ} = (jeu _ 2) 10 G( jll) (b) = 1 +2?:(i :) + (j:) 2 0 < S< 1 G(jw) (e) = 1 +2~(j . B. poles of order 2: 0. 1995. F. 584. April 1981. 20. Mark the finite poles wich x and the finite zeros with 0 in the splane. Chelsea Publishing Company.• John Wiley and Sons. C. zeros: ±j. Matrix Theory.. Find the polar representation of O(s) given in Problem 21 for s = jltJ. 93 16.. ApriII981." IEEE Tram. Vol. Stability 18. Vol. zeros: 1. p. PROBLEMS FOR SECTION 22 24. B. 25. Pillai. NY. 19. (a) G(s) = s2(s + l)(s + 10) lOts +2) (e) G(s) = 10(s + 2) s(s2 + 2s + 2) IOs(s + 1) (b) G(s) = (s + 2)(s2 + 3s + 2) 22. Automatic Control Systems. Use MATLAB to find the poles and zeros of the functions in Problem 2~ 1. Find the poles and zeros of the following functions (including the ones at infinity. where ltJ is a constant varying from zero to infinity. PROBLEMS PROBLEMS FOR SECTION 21 21. 4. "The £ Method of the RouthHurwitz Criterion. K. Prentice Hall. 17.w + 1) . "On RouthHurwitz Criterion. S. 00. AC26. AC26. C. 2. find the function: (a) Simple poles: 0.:) + (j :) 2 1. if any). New York. NJ.> 1 (d) GUll) = jCJJ(J. zeros: 0 (e) Simple poles: 3. 1964." IEEE Trans. Kuo. R.583. II. Kuo and F. Vol. (j~w + 1) jwL (e) GUw) = . 1. poles of order 2: 3.
calculate. 212. Then.4X2(t) .= Ax(t) + Bu(t).. which is . which is £rgt (t)g2(t)) £[gl (t) * 82 (t)1 = Ot (S)G2(S) = Gt (s) * G2{S) 214.03(e jwt = (eiwt _ 1 )(3ejwt + 1)2 + 1)(ejwl + 0.c{ sin2 2t}. 211.X3(t) PROBLEMS FOR SECTION 24 210.T)us(t ~ T) = enO(s}] 213.5) 28. 2~7. Express the following set of firstorder differential equations in the vectormatrix form of dx(t) . Prove the integration theorem 4 in Section 243. 2. . Prove theorems 6 and 7.X3(t) + U2(t) dxl (t) ~ = XI (I) + 2x2(t) + 2Ul (t) X3(t) (b) ~ tU:2(t) = 2xI (t)  + U2(t) dX3(t) ~ = 3Xt (t) . 15. Prove theorem 3 in Section 243. Use MATLAB to find the polar plot of the functions in Problem 25.94 I> Chapter 2...!:!. Use MATLAB to draw the Bode plot of the functions in Problem 27..c{ sin22t}. Prove the shiftintime theorem.5) (a) G()m) = jw( jw + 10) (jw + 50) (b) G(jw) = jw( jw _ ~~~w2 + 10) · (c) G(J(u) = oo2 (joo _ 25002 (jw .c{ cos2 2t} when you know . Draw the Bode plot of the following functions: · 2000( jm + 0." :::. Use MATLAB to obtain.c[g(t .) + (j~) Wtl (J)n 0 ~ ?. Verify your anSwer by calculating L{cos22t} in MATLAB... PROBLEMS FOR SECTION 23 29. Mathematical Foundation 26.100w2 + 100) + 100) 2 G(jw) (d) = I + 2?:(j. dx) (t) dt = XI (1) + 2x2(t) + 3X3(t) + u) (t) (a) ~ = 2\"2(t) dX2(t) dX3(t) ~ ~ XI (t) . Prove the convolution theorem in both time and s domain. I · (e) G( }m) O.3X2(t) .
2P21. Find the Laplace transfonns of the following functions. Figure ZPZO 221.kT) where oCt) = unitimpulse function 2·17.. Find the Laplace transforms of the functions shown in Fig.)~ 1 _~ ~L2J 11 3 L.. Use MATLAB to solve Problem 216. t 7L!J g(') o 1 2 (b) I . Find the Laplace transfOIm of the following function. 218. First.Problems • 95 2·16. 19. 1+1 get) = o 1 o O:St<l l<t<2 2 t~3 2. f(t) ~ o 1 TI2 T  .~(t) (b) get) = (tsin2t + e.2t sin2t lls(t) (d) g(/) = sin2tcos2t us{t) (c) (e) get) = Lek=O ':'Xl 5kT 8(t . Use the theorems on Laplace transfonns. (3) get) = 5te5tu.. 3 4 t Figure 2P18 2.±J II 5~ (a) II r. 2P20. 2P18. and then take the Laplace transform. if applicable. Find the Laplace transfOlm of the periodic function in Fig. Take the Laplace transfonnof g(t) to get the following: g(. Let gT(t) be the description of the function over the basic period and then delay gT(l) appropriately to getg(t). write a complete expression for g(l).t ~ t <3 220. Find the Laplace transform of the function in Fig. .2r )us (t) get) = 2e.
1} +5 d~~) +4 f(t} = e. Use MATLAB to solve the following differential equation: dZy y dt2 225. (a) d21r. 224. Mathematical Foundation f(t) I L2 0 I L ~ t .2tus (t} Assume zero initial conditions.. (A ' zero Imt..T+2y(t) (c) dt dt2 dt = e1u (t) S 2y d 2 (O) = 1 dt dy(O) = 1 y(O) = 0 dt 223. = et . 2P25 for chemical reaction..L2 Fidure 2P21 222. . Reactor 3 Reactor I 10_ _.. Solve the following differential equations by means of the Laplace transfonn.aI cond'HlOns) ssummg A series of a threereactor tank is arranged as shown in Fig.96 ..'2 +2+'.. Use MATLAB to find the Laplace transform of the functions in Problem 222. d 3v(t) d 2y·(t} d"(r} . Chapter 2... Fi gu re 2P25 The state equation for each reactor is defined as follows: .
A3] when Vj and ki represent the volume and the temperature constant of each tank as shown in the following table: Reactor Vi ki 0. perform partialfraction expansion on G(s). First. PROBLEMS FOR SECTION 25 226.4 Use MATLAB to solve the differential equations assuming CAl = CAZ = CA3 = 0 at t = O.k2 VZCA2] R3 : dCA3 1 dt = y[1000CA2 3 lOOOCA3 .(s + 1)2(5 + 3) 100(s+ 2) s + 4)(s + 1) e 10 (e) G(s) = s(s2 (d) G(s) = 2{s + 1) 8(s2 + S + 2) 1 (s + 1)3 (e) G(s) = . use the Laplace transform table.s + 4e.5)(s2 + 5s + 5) ( ) G(s) g = 2 + 2se. 1 (a) G(s) = s(s + 2}(s + 3) (b) O(s) _ . Find the inverse Laplace transforms of the following functions.1I00CA2 .s2 + lis + 6 s4 + 5s3 + 7s2 + 5s + 6 3s 3 (i) G(s) = + lOs 2 + 8$ + 5 2·27.2s :0. then.k3 v:~r. Use MATLAB to find the inverse Laplace transfonns of the functions in Problem 2w26. use the inverse Laplace transform.2 1000 2 1500 100 3 0.Problems 4 97 R2: dt = VZ[1100CAI dCA2 1 . First.2 + 3s + 2 2s + I (h) G(s) = s3 +6. then. 2(s2+s+l) (f) G(s) = s(s + 1. perform partialfraction expansion on G(s). .1 0.
.v(s + 2)(s2 + 2s + 2) 5(s + 2) s2( s+ 1)(s+ 5) (b) G(s) _ (c) G(s) = 5e.Y'(/) +2_'"2+'+2V(I) + 2 __ d" l1 (1) dv(t) l dt dt dt 230.. Find the transfer function Y(s)IR(s) for each of the systems. Use MATLAB to find the partialfraction expansion to the following functions. convert it to the set of firstorder differential equation. y(r)dt = dr(t . Mathematical Foundation 228.2) +2r{t .) (a) .98 .( ) = 5 ( ) dt2 + dt + Y t r t (e) +10+2+)1(t) +2 2 d1 3 d 3 y·(t) d2 y(t) dt dy·(t) dt· 11 & 0  dr(/) y(r)dT=+2r(t) dt (d) 2~+~+5y(t) = r(t) d 2y(t) dv(t) + 2r(t  1) +2r(t) (e) d 2y{ t + 1) dt 2 +4 ') dy( t + 1) dt +5y(t + 1) = I dr( t) +2 J t ':Xl r( r)dr (0 d.5)2 (g) G(s) = ($2 + 4)(s2 + 2s + 2) = 18 + 9s + 15$2 +! + 2 $2(S + 2)(s + 1)4 3 283 + s2 + 8s + 6 (h) G(s) 232. (a) A = (b) A = [0 1 2] [312] I 1 0 2 I 1 B= [0 1] I 0 0 0 ~l ~ ~ B = [1] g dr(t) dt 229.s (d) G(s) = (s + I )(s2 + s + 1) 100(s2 + s + 3} (') ss"+ 5s+_3) 1 S(52 (e) G(s) = (f) G(s) ~ + 1)(s + 0.2. The following differential equations represent linear timeinvariant systems. Given the state equation orthe system. (Assume zero initial conditions.2) dt Use MATLAB to find Y(s)IR(s) for the differential equations in Problem 229.._0 +2+5+6v(l) = 3+r(t) d3v(t) dt 3 d 2y·(t) dt 2 dy(t) dt  (b) d4y(t) dr + lOd2y(t) dy(t) 5 . Use MATLAB to find the inverse Laplace transforms of the functions in Problem 231. 231. lO(s + 1) (a) G(s) = s2(s + 4)(s + 6) (s + 1) .. Chapter 2. where r(l) denotes the input and yet) the output. 3 J ox .
99 PROBLEMS FOR SECTIONS 27 THROUGH 213 233. Using the RouthHurwitz criterion.. Without using the RouthHurwitz criterion. determine the stability of the closedloop system that has the following characteristic equations. For each of the characteristic equations of feedback control systems given.. llse MATLAB to determine the range of K so that the system is asymptotically stable. + 25s3 + 15s2 + 20s + K = 0 (b) s4 + Ks3 + 2s2 + (K + 1)8 + 10:::: 0 (c) s3 + (K + 2)s2 + 2Ks + 10 = 0 (a) s4 . or unstable. the closedloop system transfer function is given.5s + 10 = 0 (e) s6 + 2\. determine if the following systems are asymptotically stable. + 25s 2 + lOs + 450 = 0 (b) s3 + 25s2 + lOs + 50 = 0 (e) s3 + 25r + 250s + 10 = 0 (d) 284 + 10. Detennine the number of roots of each equation that are in the righthalf splane and on the jwaxis.1) . 237. Determine the value of K so that the system is marginally stable and determine the frequency of sustained oscillation. (a) M(8) = s3 IO~s + 2) s. + 10. + lOs + 130 = 0 (b) + 12s3 + s2 + 28 + 10 = 0 (c) 84 + 12s3 + 1082 + 108+ 10 = 0 (d) 84 + 1283 + 82 + lOs + 1 = 0 (e) s6 + 6ss + 12584 + 100s3 + loos2 + 20s + (a) s3 4 8 10 =0 (f) sS + 12534 I 100s3 j 100s2 + 203 + 10 = 0 2~38. 3 + 5.5 + 884 + 15s3 + 20r + 168 + 16 = 0 (f) 84 + 2s3 + IOs2 + 20s + 5 = 0 (g) s8 + 2s7 + 886 + 12s5 + 20s4 + 16s 3 + 1682 = (a) 8 3 0 236.5) 234. Use MATLAB Toolbox 2131 to find the roots of the following characteristic equations of linear continuousdata systems and detennine the stability condition of the systems.1 + 352 + 5s + 5)(s2 + 2) (b) M(s) = (s K (e) M(s) = $3 +58+5 (d) M(s) _ loo(s . Use MATLAB to solve Problem 235. Use the ROOTS command in MATLAB to solve Problem 233.Problems .(5 + 5)(s2 + 2s + 2) 100 (e) M(s) = S · 2 2 S + 10 3 3 8+ M s _ ()  (I) .~ + 3s3 + 50s2 + s + 106 10(s + 12. marginally stable. In each case.582 + 5. 235. if applicable..
Use MATLAB to solve Problem 240. 2 40. A linear timeinvariant system is described by the following state equations. apply the RouthHurwitz criterion to determine the stability of the closedloop system as a function of K. Mathematical Foundation (d) s3 + 20s2 + 5s + 10K = 0 (e) s4 + Ks' + 5s2 + lOs + 10K = 0 (0 s4+l2. dt = dx(t) Ax(t) + Bu(t) . where K = [kl k2k3] and k ..== 10xI (I) + u(t) u(t) = klx} (I) . Given the forwardpath transfer function of unityfeedback control systems.5s.l+jJ. 244. A controlled process is modeled by the following state equations. t The control u(t) is obtained from state feedback such that . Determine the region in the k 1versusk2 parameter plane in which the closedloop system is asymptotically stable.~ K(s + 4)(s + 20) + too)(s + 500) (b) G(s) = K(s ~ lO)(s + 20) s~(s + 2) (c) G(s) (d) G(s) = s(s + 10){s + 20) K = s3 K{s + 1) + 2s2 + 38 + 1 241.100 ~ Chapter 2. Detennine the constraints on the elements of K so that the closedloop system is asymptotically stable.. Determine R the value Qr K that will cause sustained constantamplitude oscillations in the system.= XI (t) dt dt) (t) dx2(t) 2X2(t) d. Indicate the boundary on which the system is marginally stable. Determine the regions in the TversusK parameter plane where the closedloop system is asymptotically stable and where it is unstable. The loop transfer function of a singleloop feedback control system is given as G(s)H(s) = s(s + 2){1 K(s + 5) + Ts) The parameters K and T may be represented in a plane with K as the horizontal axis and T as the vertical axis.. so that u(t} = Kx(t}.v3(. . (a) G(s) = .It = Ax(t) + Bu(t) where dx(t) A= [~ 4 3 ~ ~] o B= [~] 1 The closedloop system is implemented by state feedback. Given the system in Slale equation form. Determine the frequency of oscillation.+5s+K=O 2R39.k2XZ(t) where kl and kz are real constants. k2~ and k3 are real constants. 242. 243.
The block diagram of a motor·control system with tachometer feedback is shown in Fig. Consider the open· Loop system in Fig. Find lhe range of the tachometer constant K . yes) Figure 2p·45b Assuming f(t) = kpe + kd . so lhal the system is asymptotically stable.· dt de (a) Find the open· loop transfer [unction. = [kl k2k) )? F(s) .Problems 101 (a) A = (b) A = [i ~2l B ~ m [i ~l B ~ m 0 3 0 0 2 0 Can the system be stabilized by state feedback U(l) = Kx(t). (b) Find the closed·loop transfer function. 2P46. 2p·45(b) .+z. 2·46.~ yes) figure 2P·45a where  d2 y g 1 [Y dt = z and f(l) = T.6)(s + 10) Figure 2p·46 . 2P·45(a). where K 2·45. t dz d Our goal is to stabilize this system so the closed· L feedback control will be defined as shown in oop the block diagram in Fig. r(t) 100 y(tl + s(s + 5. (e) Find the range of kp and k" in which the system is stable.
electric power must be brought to the payload through cables. The conventional RouthHurwitz criterion gives information only on the location of the zeros of a polynomial F(s) with respect to the left half and right half of the splane. (Use K as the venical and Cl as the horizontal axis. The total force produced by the magnetic actuators is f(t). The attitude of the payload in the y direction is controlled by magnetic actuators located at the base.1 in the splane. Find the region in the KversusCl plane for the system to be asymptotically stable. so that the RouthHurwitz criterion can be applied to determine whether F(s) has zeros to the right of the line s = Cl. Devise a linear transformation s == f( p.) r(t) e(l) K(s+ 2) y(l) + figure 2P'l7 Tl 248.\2 +405+4 = 0 249.102 ~ Chapter 2. Because there are experiments located on the payload. where Cl is a positive real number. Mathematical Foundation 247. The block diagram of a control system is shown in Fig. The linear spring with spring constant Ks is used to model the cable attachment. a). (a) F(s) = s2 + 5s + 3 = 0 (b) s3 (c) (d) + 3052 + 3s + J = 0 F(s) = s3 + 4s2 + 35 + 10 = 0 s3 +4. where p is a complex variable. The controls of the other degrees of motion are independent and are not considered here. Apply the transformation to the following ch. The payload of a spaceshuttlepointing control system is modeled as a pure mass M . The dynamic system model for the I 't) 2 f 't) 2 Figure 2P4!l .~racteristic equations to determine how many roots are to the right of the line s = . 2P47. The payload is suspended by magnetic bearings so that no friction is encountered in the control.
An inventorycontrol system is modeled by the following differential equations: dx 1(t) = X2(t} + u(t) dt dx2(t) = Ku(t) dt where Xl(t) is the level ofinventory. so that /(t) = Kpy(t) .y(t}.Problems <I 103 control of the yaxis motion is shown in Figure 2P49. the rate of sales of product. a real constant. (0 If F(s) is the forwardpath transfer function of unityfeedback control systems. the production rate. Let u(t) = r(t} . (e) If O(s) is the forwardpath transfer function of unityfeedback control systems. The force equation of motion in the ydirection is f(t) = Ksy(t) + Md~~t} where Ks = 0. (b) Find the characteristic equation of the closedloop system. X2(t). . Determine the constraint on K so that the closedloop system is asymptotically stable.4t . and K. find the transfer function of the closedloop system and apply the RouthHurwitz criterion to determine its stability. 252. Use MATLAB to solve Problem 250. Use MATLAB to (a) Generate symbolically the time function of ft. u(t). (b) Generate symbolically G(s) = s(s + 2)(s2 + 2s + 2) (s + I) (e) Find the Laplace transfonn off(t) and name it F(s). (d) Find the inverse Laplace transform of O(s) and name it g(I).KD dy(t} dt (a) Draw a functional block diagram for the system. The magnetic actuators are controlled through state feedback.t) f(t) =5+2e2tsin(2t+~) 4e2tcos(2t~ +3e. Let the output of the system by y( t) = XI (t) and r( t) be the reference set point for the desired inventory level.5 NmJm and M = 500 kg. 250. 251. find the transfer function of the closedloop system and apply the RouthHurwitz criterion to detennine its stability. (e) Find the region in the KDversusKp plane in which the system is asymptotically stable.
=~ Actual Room Temperature Figure 31 A simplified block diagram representation of a heating system. S. To establish a parallel between block diagrams and signalflow graphs. To use Mason's gain formula for finding transfer function of systems. we use the material presented in this chapter and Chapter 2 to fully model and study the performance of various control systems. 2. The process is relatively straightforward. For example. To introduce the signalflow graphs. Or it can be used. The main objectives of this chapter are: 1. also defined as the input. To study block diagrams. ~~ 4. ~ 31 BLOCK DIAGRAMS The block diagram modeling may provide control engineers with a better understanding of the composition and interconnection of the components of a system. We also utilize the block diagram reduction techniques and the Mason's gain formula to find the transfer function of the overall control system. consider a Simplified block diagram representation of the heating system in your lecture room.. shown in Fig. together with transfer functions. where by setting a desired temperature. A simple electronic circuit within the thermostat compares the actual room temperature to the desired room temperature Heat Loss Desired Room Temperature . 104 . 6. their components. I:'=. we discuss graphical techniques for modeling control systems and their underlying mathematics. to describe the causeandeffect relationships throughout the system. one can set off the furnace to provide heat to the room. The actual room temperature is also known as the output and is measured by a sensor within the thermostat.3 Block Diagrams and SignalFlow Graphs In this chapter. 7. 3. To obtain transfer function of systems through block diagram manipulation and reduction. and their underlying mathematics. To introduce state diagrams . To demonstrate the MATLAB tools using case studies.. Later on in Chapters 4 and 5. 31.CHAPTER .
Output speed LOAD (a) v.. an error voltage will be generated. then the amplifier can be regarded as linear. . The block diagram in this case simply shows how the system components are interconnected.{t) 1 B+Js (J) (I) Q(s) (b) Figure 32 (a) Block diagram of a dcmotor control system. and the amplifier may be described by the transfer function (31) where K is a constant. 32 (a). If the mathematical and functional relationships of all the system elements are known .. . 105 Disturbance torque =:"'+~I Input voltaoe AMPLIFIER . speedcontrol system.. (comparator). realistically. has a nonlinear characteristic. which is the slope of the linear region of the amplifier characteristics. The process of sensing the output and comparing it with the input to establish an error signal is known as feedback.. If the room temperature is below the desired temperature. Note that the error voltage here causes the furnace to turn on. or. naturally. Vls) and Va(s) do not exist. would disturb the heating process (disturbance). The nonlinear amplifier gain can only be described in time domain and between the time variables v. 32 (b).. consider the block diagram of Fig. its dynamics can be represented by transfer functions.31 Block Diagrams . which models an openloop. (b) Block diagram with transfer functions and amplifier characteristics. In this figure . the input voltage to the motor is the output of the power amplifier. In general. and the furnace would finally shut off when the error reaches zero.. which.. Alternatively. The error voltage acts as a switch to open the gas valve and turn on the furnace (or the actuator). If the motor is linear. The room temperature is constantly monitored by the output sensor. and no mathematical details are given.{!) and va(t). block diagrams can be used to model linear as well as nonlinear systems. in this case. However. Opening the windows and the door in the classroom would cause heat loss and. ifit is operated in the linearregion of its characteristics. Laplace transform variables do not apply to nonlinear systems. These topics are discussed later in this chapter.. As another example. hence. dcmotor. more appropriately. For example.. we can use Signalflow graphs or state diagrams to provide a graphical representation of a control system. the block diagram can be used as a tool for the analytic or computer solution of the system. if the magnitude of Vi(t) is limited to the linear range of the amplifier. the inputoutput relations of the dcmotor control system may be represented by the block diagram shown in Fig.
32 to Fig. Note that the addition and subtraction operations in Fig. As a rule. 31). Once the block diagram of a system is fu11y constructed. One of the important components of a control system is the sensing and the electronic device that acts as a junction point for signal comparisonsotherwise known as a comparator. Disturbance Figure 33 Block diagram representation of a general control system.Y(s) (33) As mentioned earlier. . In general. one can study individual components or the overall system behavior. these devices possess sensors and perform simple mathematical operations such as addition and subtraction (such as the thermostat in Fig. in Fig.106 .turbance signal • Feedback loops Fig. 35. blocks represent the equations of the system in time domain or the transfer function of the system in the Laplace domain. These equations are normally in the time domain or preferably (because of ease in manipulation) in the Laplace domain. Three examples of comparators are illustrated in Fig. Block Diagrams and SignalFlow Graphs 311 Typical Elements of Block Diagrams in Control Systems We shall now define the blockdiagram elements used frequently in control systems and the related algebra. You may wish to compare Fig. the block diagram implies e(t) = r(t) . 33 to find the control terminology for each system. 33 shows one configuration where these elements are interconnected. each block represents an element in the control system. 31 or Fig. so the input and output variables of these blockdiagram elements can be timedomain variables or Laplacetransform variables.. The common elements in block diagrams of most control systems include: • Comparators • Blocks representing individual component transfer functions. as demonstrated in Fig. 34 (a).y(t) or (32) £(s) = R(s) . including: • Reference sensor (or input sensor) • Output sensor • Actuator • Controller • Plant (the component whose variables are to be controlled) • Input or reference signals • Output signals • Di:. 34 (a) and (b) are linear. and each element can be modeled by one or more equations. Thus. 34. Chapter 3.
) = r(t) + y(r) R(s) + y(l ) yes) E(s) =RCs)  YCs) R(s) + + y(t) yes) E(s) =R(s) + yes) (a) (b) + + e(/) = rl(l) + r2(t)  yet) yes) . U(s) I G(s) X (s) . EXAMPLE 311 Consider the block diagram of two transfer functions G I (s) and G2(s) that are connected in series. (a) Subtraction. (c) Addition and subtraction. the following inputoutput relationship can be written for the system in Fig. u) XC I) Time domain Figure 35 Time and Laplace domain block diagrams. .31 Block Diagrams r(t) ~ 107 e(t.) = r(/) . the transfer function of the block in Fig. actuator.y(t) r(t) ee. (b) Addition. for signals A(s) and Xes).<~)r £(s) = RI(s) + R2(s)  A comparator performs addition and Sll btracti on Y(.) (c) Figure 34 Blockdiagram elements of typical sensing devices of control systems.. Find the transfer function G(s) of the overall system. controller. 35: x (s) = G (s) U (s) (34) If signal Xes) is the output and signal U(s) denotes the input. we have .) ~I g (x.. Hence.domain Laplace In Laplace domain. SOLUTION The system transfer function can be obtained by combining individual block equations. 35 is X(s) G(s) =U(s) (35) Typical block elements that appear in the block diagram representation of most control systems include plant. and sensor . U (t.
L . it is necessary that the controlled variable be fed back and compared with the reference input.. After the comparison. (s) AJ(s) = A ..U(s) A2(S) = A) (s)G. (s) + C2(S» C(s) = Xes) U(s) Hence.. SOLUTION The system transfer function can be obtained by combining individual block equations. .108 .. (s) and G2 (s). Chapter 3. using Eq. diagrams Gl(s) and Gz(s) connected in series. respectively. A J (s) . (s) A (5) "I . 37. I G.. A feedback control system is also known a closedloop system. we combine the equations as follows. for the overall system.. A system may have multiple feedback loops. as shown in Fig. ~ (36) ~ EXAMPLE 312 Consider a more complicated system of two transfer functions CI(S) and Gis) that are connected in parallel. Block Diagrams an d SignalFlow Graphs U (5) . (36). 36 can be represented by the system in Fig. and AAs) and A3(s) are the outputs. which is used to actuate the control system. C(05) = GJ (s) + G2(S) (37) For a system to be classified as a feedback control system. (s) acts as the input... Xes) = A(s)Gz(s) A(s) = U(s)G)(s) Xes) = G.(S)G2 (S) Xes) Xes) = A2(S) + . U(s} xes) C(s) = C. . (S)G2(S)U(S) C(s) = Helice. A. Hence. As a result. Find the transfer function C(s) of the overall system .. (s)G 2(s) Hence. Further. the system in Fig. 38 shows the block diagram of a linear A J (s) G. note that signal U(s) goes through a branch pointP to become Al (s). 35. (s) A2 (s) U (s) p X (~) Al (s) G2 (s) A3 (s) Figure 37 Block diagrams G\(s) and G 2 (s) connected in parallel.' L . Note for the two blocks G._ G2 (8) _ X(s) ' ~ figure 36 Block. an error signal is generated. the actuator is activated in the presence of the error to minimize or eliminate that very error. A necessary component of every feedback control system is an output sensor.43(S) = U (s)(G. which is used to convert the output signal to a quantity that has the same units as the reference input. Fig._ _.
From Fig. (39) into Eq. The following lenninology is defined with reference to the diagram : ret ). (313) If G and H are constants. B(s) = feedback signal u(t). Ye s) = output (controlled variable ) b et ). they are also called gains. U( s) = actuating signal = error signal e(t ). (3. feedback control system w ith a single feedback loop.12) becomes M(s) = Yes) = G(s) R(s) 1 . when H(s) = 1 H (s) = feedback transfer function G(s)H (s ) = L(s) = looptransferfllnc tion G(8) = forw ardpath transfer function M(s) = Y(s)/ R(s) = closedloop transfer fun ction or system transfer function The closedloop transfer function M(s) can be expressed as a function of G(5) and H(s). and if H = 0 . (310) into Eq. If H = 1 in Fig. (37) and then solving for Y(s)IR(s) gives the closedloop transfer function M 5) = Ye s) = G(s) ( R(s) 1 + G(s)H(s) (312) The feedback system in Fig.G(s)H(s ) unity feedback loop.B(s) Substituting Eq .()~ u (t) (314) . and the transfer function Eq. R(s) = reference input(command ) y (t ). 38.10) y es) = G(s)R(s) .G(s )H(s) (3.31 Block Diagrams ye s) G(s) r (t) 109 R(s) U(s) + u(J ) y(e) be e) H{ s) 8(s) Figure 38 Basic block diagram of a feedback control system. we write yes) = G(s)U(s) and (38) B(s ) = H(s)Y( s) The actuating signal is written (39) U(s) = R(s) . When the comparator adds the fe edback. 38.1l) Substituting Eq. (3 8) yields (3 . E(s). it is called positive feedback. 38 is said 10 ha ve a negative feedback loop because the comparator subtracts. the system is said to be open loop. the system is said to have a 312 Relation between Mathematical Equations and Block Diagrams Consider the following secondorder prototype system: i (t) + 2 ~ wni (t ) + w~ x (t) = (.
(315) consists of constant damping ratio S. If you wish. if A(s) is integrated once. If we rearrange Eg./U(s) + Figure 310 Addition of blocks ~. then we can designate internal variables A(s) and Yes). 310.2 ~wlIs./X(s) Figure 3·9 Graphical representation of Eq.. X(s) Figure 3·11 DIock diagram representation of Eq. which has Laplace representation (assuming zero initial conditions x(O = i(O) = 0): ) Xes) i + 2 (Wn Xes) s +w~x (s) = w. Because the signals Xes) in the righthand side of Fig. It is evident that there is no unique way of representing a system model with block diagrams.1" as shown in Fig. leading to the block diagram representation of the system Eg. as long as the s t OJ. Hence. as illustrated in Fig. 312(a) to obtain Fig. and to the graphical represen~atjon of Eq. we get V(s) . The best way to see this is by recalling that is the integration operation in Laplace domain. respectively. s If the system studied here corresponds to the springmassdamper seen in Fig. respectively. and after integrating V(s). 310 ~re the same. 39. 311. (317). (317) in Laplace domain . (3·17) using a comparator. (315) to w. Chapter 3. they can be connected. (315) input U(s). . 312. w . which represent acceleration and velocity of the system. 311 by factoring out the term 1 as in Fig. The signais 2swn sX(s) and w~X(s) may be conceived as the signal X(s) going into blocks with transfer functions 2 w"s and w~. Block Diagrams and SignalFlow Graphs w. U(s) W II . as shown in Fig./U(s) + s2 X(s) OJ. 45 (see Chapter 4).110 . We may use different block diagram forms for different purposes. and the signal Xes) may be obtained by integrating s2X(s) twice or by postmultiplying by . 312(b).U(s) 2swIIX(s)sw~X(s) =X(s)s2 (3·16) it can graphically be shown as in Fig.. you can further dissect the block diagram in Fig. Eq.. we get the Xes) signal. (317).. constant natural frequency and output Xes).
312 and compare that to the transfer function of system in Eq . (315). 314(b) are equivalent systems. (317) in Laplace domain. EXAMPLE 313 Find the transfer function of the system in Fig. as shown in Fig. yes) Figure 313 Block diagram of Eq. This enables us to determine the behavior of velocity signal with input U(s). 312(b) may be moved to the righthand side of the comparator. to obtain the transfer function V(s)jU(s). This is the same as faclOrizalion of w~ as shown: w~ U (s) . (b) Final block diagram representation of Eq. 3l3. .(tJ~ X (s) = w~ (U(s) . 312 to get yes) as the system output. (317). we may yet rearrange Fig.31 Block Diagrams ' X(s) 111 (a) Xes) (b) Figure 312 (a) Factorization of ~ term in the internal feedback loop of Fig. 311. 312(b) and Fig. which results in a simpler block diagram representation of the system shown in Fig. overall transfer function of the system is not altered.X (s)) (317) Fig. Note that Fig. For example. SOLUTIONS The w~ block at the input and feedback signals in Fig. 314(a) shows the factorization operation ofEq . 314 (b). (317) in Laplace domain with Yes) represented as the output.
V(s) U(s) :::: 2 r w W II l + _~_" S I 2 s I 1+ [ +s 21. or V(s) A(s) = I t + 21. (3.and postmultiplication by w2 and respectively. 313. Simplifying the two feedback loops in Fig. (319) is also the tra nsfer function of system Eq."wl/ (318) After pre."wI1 S 2 W" S V(s) S(v.15). which ultimately results in t."w lI ) Eq .. it is easy to identify the internal feedback loop. s X(s) U(s) s (s + 2l. (3. 312(b). (31 7) in Laplace domain.Xes ) (al Xes) (0) Figure 314 (a) Factorization of (L!~ .12). U(s ) = S2 +2I.112 ."wI1 s s+2l. (319).. (b) Alternative block diagram representation of Eg. which in turn can be simplified llsing Eq.15. Xes) Figure 31 5 A block diagram w~ Considering Fig. " EXAMPLE 314 Find the velocity SOLUTIONS we have tralL~ferfunclion using Fig. the block diagram of the system is simplified to what is shown in Fig. 3.. Chapter 3. starting wilh the internal loop first.. Block Diagrams and SignalFlow Graphs U(s) + A(s) ' .."wl/s +w'i (320) . 313 and compare that to the derivative of Eg.."wn ) 1+ w2 (319) s (s + 21.
as shown in Fig.3' Block Diagrams ~ 113 Eg. we have the following relations: yes) = A(S)G2(S) B(s) = A(s)HI (S)G2(S) (322) But G2 (S) =? B(s) = = Y(s) A(s) Y(S)Hl(S) (323) 2. we have the following relations: yes) = A(S)G2(S) + B(s)Hl (s) (324) (a) p A(s)+t f. it is often necessary to move a comparator or a branch point to make the block diagram reduction process simpler.. should also be done such that the output yes) is unaltered. Try to find the A(s)/ U(s) transfer function. 317(a) aod Fig. Obviously you must get: i X(s) / U(s). 316(a) and Fig. The two key operations in this case are: 1. which is nothing but mUltiplying Eg . . 316(a). 316(b). (3) 9) by • an s term.. (320) is the same as the derivative of Eg. as shown in Fig. In Fig. 313 Block Diagram Reduction As you might have noticed from the examples in the previous section... For complicated block diagrams. Moving a comparator. we have the following relations: B(s) yes) = A(S)G2(S) = Y(s)HI (8) (321 ) In Fig. 316(b). Moving a branch point from P to Q. the transfer function of a control system may be obtained by manipulation of its block diagram and by its ultimate reduction into one block. In Fig. 317(b). This operation mllst be done such that the signals yes) and B(s) are unaltered. (319).. 317(a). Y(s) 8(s) (b) yes) B(s) Figure 316 (a) Branch point relocation from point P to (b) point Q..
114 ~ Chapter 3. (d) Eliminating the inner feedback loop. 318(b). After that. we have the following relations: Y 1(s) = HI (s) A(s) + B(s) G2(S) (325) Y{s) = Y. to the left of block G2 . SOLUTION To perform the block diagram reduction. (c) Combining the blocks G" Gz. 318(c).(s)~ Y(s) (b) Y(. In Fig. and G4 as shown in Fig. 317(a). tbe reduction becomes trivial. Block Diagrams and SignalFlow Graphs (a) . to the left of block G2 . and G3 . (b) Moving the branch point at Y.J Figure 317 (a) Comparator relocation from the righthand side of block G2 (s) to (b) the lefthand side of block G2 (s) . (S)G2(S) So HI (. . first by combining the blocks G2 • G3 . and then by eliminating the two + y (a) Figure 318 (a) Original block diagram.I') Y(s) =A(S)G2(S) + B(s) G2(S) G2(S) (326) => Y(s) = A(S)G2(S) + B(s)HI (s) ~ EXAMPLE 315 Find the inputoutput transfer function of the system shown in Fig. one approach is to move the branch point at Y.4. as shown in Fig. 317(b).
the overall response of the system under multiinputs is the summation of the responses due to the individual inputs. one of the inputs. is known as disturbance. In this case.. it is always important to learn the effects of D(s) on the system. A simple block diagram with two inputs is shown in Fig.e. 31) usually adversely affects the perf0n11anCe of the control system by placing a burden on the controller/actuator components. Super Position: For linear systems. 3l8(d) Y(s) E(s) (327) 314 Block Diagram of MultHnput SystemsSpecial Case: Systems with a Disturbance An important case in the study of control systems is when a disturbance signal is present Disturbance (such as heat loss in the example in Fig. while R(s) is the reference input Before designing a proper controller for the system. sy~tem after the reduction in Fig. in this case. As a result. i. (328) . D(s).31 Block Diagrams '4 115 y (b) y (c) y (d ) Figure 318 (Continued) feedbac k loops. We use the method of superposition in modeling a mUltiinput system. the transfer function of the final become. 319.
321): (330) = 0. 319 when R(s) = O. . the block diagram is rearranged Y(s) D(s) D(s) yes) (a) Yes) (b) Figure 321 Block diagram of the system in Fig. Block Diagrams and SignalFlow Graphs D(s) Controller R(s) Plant yes) + Outpm Sensor Figure 319 Block diagram of a system undergoing disturbance R(s) Y(s ) Figure 320 Block magram of the system in Fig. 319 when D(s) = O.116 Chapter 3. the block diagram is simplified (Fig. 320) to give the transfer function Y(s) R(s) When R(s) (329) to give (Fig. When D(s) = 0.
the individual input and output signals are designated. from Eq. y(t) Figure 322 Block diagram representations of a muItivariable system. 322(b). . there will be a higher burden 011 the controller. whereas in the block diagram of Fig. Fig. and. as a result. The transfer function re lationships of the system are expressed in vectormatrix form (see Appendix A): Y(s) = G(s)U(s) U(s) (332) (3 33) (334) = R(s) = B(s) B(s) H(s)Y(s) . MULTrv ARIABLE SYSTEM (a) r(t) __+1. blll= o bIR =O 31 5 Block Diagrams and Transfer Functions of Multivariable Systems Tn this section. (328) to Eq. 322 (a). 323 shows the block diagram of a multivariable feedback con trol system. it adversely affects the performance of the system. In Fig. (332). The negative sign in the numerator of shows that the disturbance signal interferes with the controller signal.1 . Naturally.G2 + 1 + G] G2 H 1 D(s) Observations: ~ I D=O and have the same denominators if the disturbance signal goes to the forward path. to compensate. we ultimately get YlOtal = Y«(S))! R S R(s) D=O + ~«(S))! S D(s) R=O (331 ) GIG2 Y(s) = 1 + G] G2 H ]R(s) .31 Block Diagrams 117 As a result. 322(£1) and (b). Two blockdiagram representations of a multivariable system with p inputs and q outputs are shown in Fig. 322(b) is preferable in practice because of its simplicity. we shall illustrate the block diagram and matrix representations (see Appendix A) of multi variable systems. The case of Fig. MULTI VARIABLE SYSTEM (b) f.. the multiplicity of the inputs and outputs is denoted by vectors .
(3 9). respectively. (310) to Eg . The closedtoop transfer matrix is defined as M (s) Then Eq.~l ~ [S~l u 2 2 ") 1 (341) 05 + 1 where " s+ 2 D.G(s)H(s)Y(s) (335) Solving for yes) from Eq. The closedloop transfer function matrix of the system is give n by Eq .11) into Eq. where yes) is the q x 1 output vector. 323 are G(s) = [+ ) 1 s+ 2 [ H(s) = [~ ~] [ 05+ 1 S+2 2 ~ (339) respec tive ly. Substituting Eq. and is evaluated as follows: 1 + G(s) H(s) = 1 +. = of + 2 s + 3 + ~ = s2 + 5s + 2 s + 1 .10) and then from Eg. (312) gives Y(s) = [I + G(s)R (s) r 1G(s) R(s) (336) provided that 1 + G(s)H (s) is nonsingular. U(s). (3 14) is written = [I + G(s)R (s)r 1 G(s) (337) Y (s) = M (s) R (s) (338) ~ EXAMPLE 31 6 Consider that the forward path transfer function matrix and the feedbac kpath transfer func tion matrix of the system shown in Fig. we get Y (s) = G(s)R(s ) .118 . +1 I 1+  . and B(s) are all p x 1 vectors. R(s). (3. Block Diagrams and SignalFlow Graphs Y (s) B (s) Figure 323 Block diagram of a multi variable feedback control system.\"+2 s s(s+ 1) (342) . Chapter 3. (315). (3.~J 1S _ s+2  I s +"3 1 (340) s+2 The closedloop transfer function matrix is M(S) =[I + G(S)H(SWJG(S)=~[:' : ~ 05+ 1 . and G(s) and R(s) are q x p and p x q transferfunction matrices.
21 Basic Elements of an SFG When constructing an SFG. Besides the differences in the physical appearance of the SFG and the block diagram.. The SFG was introduced by S. The branches have associated branch gains and directions.''+ 32 SIGNALFLOW GRAPHS (SFGs) A signalflow graph (SPO) may be regarded as a simplified version of a block diagram. . . . the construction of the SPG is basically a matter of .f 4 3s + 2 s(s + 1) 1] s (343) .3~2 Signal~Flow Graphs (SFGs) . (331) or Eq. Consider a linear system that is described by a set of N algebraic equations: N Yj = LakjYk k""'l j = I! 2.. When the system is represented by a set of integrodifferential equations. are used to represent variables. whereas the blockdiagram notation is more liberaL An SFG may be defined as a graphical rrteans of portraying the inputoutput relationships among the variables of a set of linear algebraic equations. . (331). In general. given a set of equations such as Eq.. 119 Thus. 2~ . M(s) = 3s2 + + s(s + J)(s + 2) s2 + Ss + 2 [ 2 . Mason [2] for the causeand~effect representation of linear systems that are modeled by algebraic equations. or nodes..yeS + 1) 9.N (347) 3. or N Yj(s) = LG1g(s)}k(s) k""'l j = I.. 1. we must first transform these into Laplacetransform equations and then rearrange the latter in the form of Eq. A signal can transmit through a branch only in the direction of the arrow. according to the causeandeffect equations. junction points. (347). The nodes are connected by line segments called branches. N (344) It should be pointed out that these N equations are written in the form of causeand~effect relations: N jtheffect = L(gain fromkto j) x (kthcause) k=l (345) or simply Output = L(gain) x (input) (346) This is the single most important axiom in forming the set of algebraic equations for SFGs. the signalflow graph is constrained by more rigid mathematical nlles.
SFG applies only to linear systems.. Block Diagrams and Signal~Flow Graphs ill:! o~~~. 4. Notice that the branch directing from node Yt (input) to node Y2 (output) expresses the dependence of Y2 on YI but not the reverse. between the two variables. Nodes are used to represent variables. a signal of strength al2YI is delivered at node yz. Normally. consider the following set of algebraic equations: yz = al2YI Y3 Y4 :=. 3. 2. in Fig. step by step.~~~o )'1 Y2 Figure 3~24 Signal flow graph of Y2:=. consider that a linear system is represented by the simple algebraic equation (348) where Yt is the input~ Y2 is the output. 323 Definitions of SFG Terms In addition to the branches and nodes defined earlier for the SFO... (348) is shown in Fig. the following tenns are useful for the purpose of identification and execution of the SFG algebra.. following a succession of causeandeffect relations through the system. EXAMPLE 321 As an example on the construction of an SFG. a23Yz + a3zY3 + Cl43Y4 = a24Y2 + a34Y3 + a44)'4 Ys = a25YZ + {l45Y4 The SFG for these equations is constructed. IfEq. 324 does not imply this relationship. or transmittance. 324. (349) is valid as a causeandeffect equation. 6.120 ~ Chapter 3. 325. the nodes are arranged from left to right. 5. from the input to the output. Although algebraically Eq. The SPG representation ofEq. (350) 322 Summary of the Basic Properties of SFG The important properties of the SPO that have been covered thus far are summarized as follows. . (348) can be written as 1 YI = Y2 (349) a12 the SFG of Fig. The brunch between the input node and the output node should be interpreted as a unilateral ampli fier with gain a12. allYl. a new SFG should be drawn with Y2 as the input and Yl as the output. The branch directing from node Yk to Yj represents the dependence of Yj upon Yk but not the reverse. following through the causeandeffect relations of each variable in terms of itself and the others. . so when a signal of one unit is applied at the input Y. For instance. A signal Yk traveling along a branch between Yk and Yj is multiplied by the gain of the branch akj' so a signal akjYk is delivered at Yj. 1. Signals travel along branches only in the direction described by the arrows of the branches. and a12 is the gain. The equations for which an SFO is drawn must be algebraic equations in the form of causeand~effect.
The same procedure is applied to Y3. we cannot convert a noninput node into an input node by reversing the branch direction of the procedure described for output nodes. (350). we simply connect a branch with unity gain from the existing node Y2 to a new node also designated as Y2. The equation that portrays the relationship at node Y2 now reads (351) Y2 = Y2 + al2YI + a32Y3 which is different from the original equation given in Fig. this condition is not always readily met by an output node. in the modified SFG of Fig. For instance. However. 324). 324). 326(a) does not have a node that satisfies the condition of an output node. Input Node (Source): An input node is a node that has only outgoing branches (example: node Y1 in Fig. the SFG in Fig. .32 SignalFlow Graphs (SFGs) "'~ 121 o Y2 Y3 (a) Y2 = al2>'1 + a32~'3 Ys YI ~ ~ o h Y5 (b) Y2 = al2VI + a32v3 Y3 == a23)'2 + a43Y4 Yl y. However. 327 would result. the equations Y2 . 326(a}. In general. If we attempt to convert it into an input node by adding an incoming branch with unity gain from another identical node Y2. as shown in Fig. node Y2 of the SPO in Fig. the SFO of Fig. o YI (d) Complete signalflow graph Figure 3·25 Stepbystep construction of the signalHow graph in Eq. Notice that. 326(b). To make Y2 an output node. 326(a) is not an input node. Output Node (Sink): An output node is a node that has only incoming branches: (example: node Y2 in Fig. we can make any noninput node of an SPG an output by the procedure just illustrated. It may be necessary to regard Y2 andlor Y3 as output nodes to find the effects at these nodes due to the input. For instance.= Y2 and Y3 === Y3 are added to the Original equations. 326(b).
Y3 . The reader should try to detennine the two forward paths between YI and Y4. Loop Gain: The loop gain is the path gain of a loop. Therefore. th~re are three forward paths between YI and Y5. There are two forward paths between Yl and Y3: One contains the branches from Yl to Y2 to Y3~ and the other one contains the branches from Yt to Y2 to Y4 (through the branch with gain a24) and then back to Y3 (through the branch with gain a43). in the SFG of Fig.Y4 in Fig. The forward path between YI and Y2 is simply the connecting branch between the two nodes. 328.Y2 in Fig. it may have numerous paths just by traversing the branches a23 and a32 continuously. For example. These are shown in Fig. ForwardPath Gain: The forwardpath gain is the path gain of a forward path. Figure 327 EIToneous way to make node Y2 an input node. there are four loops in the SFG of Fig. The definition of a path is entirely general. the loops)'2 .Y2 . and the rest of the nodes are all possible output nodes. Similarly.122 • Chapter 3. as simple as the SPG of Fig.Y4 .. For example. Block Diagrams and SignalFlow Graphs (a) Orjginal signalflow graph (b) Modified signal~f1ow graph Figure 326 Modification of a signalflow graph so that Y2 and Ya satisfy the condition as output nodes. For example. 32S(d) is a12 a 23 a 34· Loop: A loop is a path that originates and terminates on the same node and along which no other node is encountered more than once. For example.anches traversed in the same direction. Path Gain: The product of the branch gains encountered in traversing a path is called the path gain. Path: A path is any collection of a continuous succession of b. 326(a) is.Y3 . since it does not prevent any node from being traversed more than once. 325(d).Y3 . 325(d) are nontouching loops. . For example. 325(d). 328 is a24 a 43 a 32· Nontouching Loops: Two parts of an SFG are nontouching if they do not share a common node.Y2 and Y4 . the loop gain of the loop Y2 . at an output node and along which no node is traversed more than once. Yl is the input node. the path gain for the path Yl . Forward Path: A forward path is a path that starts at an input node and end.Y4 of the SFG in Fig..
123 Figure 328 Four loops in the signalflow graph of Fig. = al7Yl a16YI (353) Parallel branches in the same direction connecting two nodes can be replaced by a single branch with gain equal to the sum of the gains of the parallel branches. An example of this case is illustrated in Fig. (352) 2. The value of the variable represented by a node is equa1 to the sum of all the signals entering the node. The value of the variable represented by a node is transmitted through all branches leaving the node. the value of Yl is equal to the sum of the signals transmitted through all the incoming branches. For the SFG of Fig. that is.. we can outline the following manipulation rules and algebra: 1. Y3 Ys Ys Figure 329 Node as a summing point and as a transmitting point. 325(d). In the SFG of Fig. we have Y6 = Y7 Ys =.3~2 SignalFlow Graphs {SFGs} . 330.alSYI 3. . 329. 324 SFG Algebra Based on the properties of the SFG. 329.
Block Diagrams and Signal~Flow Graphs ~~~~~~~~~~~~~~. 332. there is a general gain formula available that allows the determination of the inputoutput relations of an SFG by jnspection. 4. .~~~O~~~..~~~O~~~. (312) can be obtained. alZ a23 a34 O~~~. the task of solving for the inputoutput relations by algebraic manipulation could be quite tedious.7 Gain Formula for SFG Given an SFG or block diagram. G(s) I R?s)~~~·~~E~(S~r(S)~~·~~Y?s) H(s) Figure 332 Signalflow graph of the feedback control system shown in Fig.~~~~~~r~~'4~~Yl ".. A series connection of unidirectional branches.. Fortunately. as shown in Fig. Using the SFG algebra already outlined. as shown in Table 31. can be replaced by a single branch with gain equal to the product of the branch gains. 331.124 ~ Chapter 3. 38 is drawn as shown in Fig. u+b+c / 12 Figure 3~30 Signalflow graph with parallel paths replaced by one with a single branch. 32.~~10 Yl Y4 Figure 331 Signalflow graph with cascade unidirectional branches replaced by a single branch. the closedloop transfer function in Eq. 325 SFG of a Feedback Control System The SFG of the singleloop feedback control system in Fig.~~~O YI Yz Y3 Y4 a 12aZ3a34 O~~I. 38. 326 Relation between Block Diagrams and SFGs The relation between block diagrams and SFGs are tabulated for three important cases.
+. the gain between the input node Yin and output node Yout is r3] M .32 SignalFlow Graphs (SFGs) TABLE 3·1 Block diagrams and their SFG equivalent representations 125 BIot:k Diagram Simple TransFer Function U(. for that part of the SFG that is nontouching with the kth forward path.k k= l ~ N fi (354 ) outputnode variable between Yin and YoU! YOI/l YOUI M Mk = gain = gain N = total number of forward paths between Yin and of the kth forward paths between Yill and fi = 1  L Lil +L j Lj2  L Lu + ' . .} Signal Flow Diagram Y(s) _I G (.5 Parallel Feedhack U(.) G (s) O~+~~O Y(8) U(s) = G( ') 7 11 ~ . ) Y( ..(sum of products of gains of all possible combinations of three nontouching loops) +. !!J.) +<' \ : ' .0U(s) ~  +<0 yr. .. ) I/( s) Res) Given an SFG with N forward paths and K loops. k . " ..k = the !!J. or = i. 0 : > .) possible combination of r non !!J.(sum of the gains of all individual loops) + (sum of products of gains of all possible combinations of two nontouching loops) .. ..I')I Y(s) u(s) Y(s ) R(s) Yes) Y (s ) 0(8) R( s} = 1 +G(s}H(s) ril) + b(t) y(t) R(s ) 0  G(. = 1. k (355) Lint = gain product of the mth (m touching loops (1 ~ r ~ K). where Yin = inputnode variable Y alll = = YOllt Yin = ~ MIJ!l. j .
(aI2 0 23 a 34 a 45) 1 . we have (3~61) Y5 )'1 MI~I +M28. Let us first determine the gain between)'1 and)'5 using the gain formula. the closedloop transfer function is written Y(s) Mt8. However. that is. Furthermore. (354). k are the only terms in the formula that could be complicated if the SPG has a large number of loops and non touching loops. 8. The three forward paths between)'1 and Y5 and the forwardpath gains are Ml :::::: a12 a23 a34 a 4S M2 al2 alS M3 = il12 a24 a45 = Forward path: Forward path: Forward path: YI .a44 Substituting these quantities into Eq.Y2 .3a32044 . 3w 25(d).2 +M3 A3 = 8.Y2 . the loop gain is Lll = G{s}H{s) 3. Block Diagrams and Signal~Flow Graphs The gain formula in Eq.126 ~ Chapter 3. 6. There are no non touching loops since there is only one loop. There is only one loop.Y5 Y5 The four loops of the SFG are shown in Fig.a44) + a12 a 24 a 45 + a34a43 + a24a32a43 + a44) + a2. EXAMPLE 322 Consider that the closedwloop transfer function Y(s)JR(s) of the SPG in Fig.1 = 8.Y4 Thus. . Thus. (354). Thus.)'3 and Y4 .Y5 )'1 )'2 . and ~ = 1 .Y4 . the forward path is in touch with the only loop. . Thus.Y3 . There is only one forward path between R(s) and Y(s). (3~54).(a23a32 + (012 a 25)(1 . 8. The following results are obtained by inspection of the SFG: 1. Care must be taken when applying the gain formula to ensure that it is applied between an input node and an output node. the two loops are Y2 . fl and 8. 1 = I.Y2 and Y4 . (354) may seem fonnidable to use at first glance. (312).a34a43 . Eq. 332 is to be determined by use of the gain formula. 328.3 = 1. Two of the loops are not in touch with forward path M 2 • These loops are )'3 .2 = 1 .LlI = 1 + G(s)H(s) (358) Using Eq. the product of the gains of the two nontouching loops is L12 == a23a32a44 (360) All the loops are in touch with forward pathsMI andM3.)'3 )'1 . EXAMPLE 3"23 Consider the SFG shown in Fig. and the forwardpath gain is Ml = G(s) (356) 2.)'4. .)'4  Y4 .) G(s) R(s) = = 1 + G(s)H(s) x (359) which agrees with Eq.a34 a 43 . The loop gains are LJ J = a23 a 32 L21 = a34 a 43 L31 = a24 a 43 a 32 L41 = a44 There is only one pair of nontouching loops.
the latter is not an input. 333. The following inputoutput relations are obtained by use of the gain formula: Y2 1 + 03 H 2 + H4 = Yl a + G3 HZH 4 (366) (367) where A. Signal"Flow Graphs (SFGs) where 0114 127 A = 1 . EXAMPLE 324 Consider the SPG in Fig. . (363).32. where A 1S given in Eq.(aZ3032 + a34a43 + a24032043 + 044) + a23032044 The reader should verify that choosing)'2 as the output.. which represents the dependence of Y7 upon Y2. For example.::::: 1 + G)HI +G3H2 + GIG2G3H3 +H4+0IG3H)H2 + GIHIH4 + G3H2H4 + G1G2G3H3H4 + GIG'jHI H2H4 (368) Y7 Figure 333 SignalHow graph for Example 324. it may be of interest to find the relation Y7/Y2.(Lll +L21 +L31 +L4d +L12 = 1 . it is of interest to find the relation between an outputnode variable and a noninputnode variable. 328 Application of the Gain Formula between Output Nodes and Noninput Nodes It was pointed out earlier that the gain formula can only be applied between a pair of input and output nodes. in the SFG of Figure 333.'. . Often.
G2G3 + 0. . to be able to identify all the loops and nontouching palts clearly. (374) yes) E(s) G[G2G 3 I GIG4 = 1 + 0) G2H.G4 (372) R(s) = A where Similarly. 3. 334(b). (354) can directly be applied to the block diagram to determine the transfer function of the system. the gain formula in Eq. by including an input node. That is Y(s) G.128 ~ Chapter 3. Block Diagrams and SignalwFlow Graphs We can show that. the gain fonnula can still be applied to find the gain between a noninput node and an output node. 01G2G3. it may be helpful if an equivalent SFG is drawn for the block diagram first before applying the gain formula. That is: Forward Path Gains: 1. 2. GtG2H. ~ EXAMPLE 326 To illustrate how an equivalent SPO of a block diagram is constructed and how the gain formula is applied to a block diagram. the gain between Y2 and Y7 is written Y7 y7/Yl GlG2 G3 G4 + G. GtG4 The closedwloop transfer function of the system is obtained by applying Eq. = + G3 H2H4 (371) 3~29 Application of the Gain Formula to Block Diagrams Because of the similarity between the block diagram and the SFG. consider the block diagmm shown in Fig.EM k J1klfromYin tO Y2 d (369) Because d is independent of the inputs and the outputs. Notice that since a node on the SPG is interpreted as the summing point of all incoming signals to the node. where Y2 is not an input. 2. Let Yin be an input andYout be an output node of a SFG. Gl G2G. the negative feedbacks on the block diagram are represented by assigning negative gains to the feedback paths on the SFG. (354) to either the block diagram or the SPO in Fig. in complex systems.Gs(l + G3 H2) Y2 = Y2/Y. First we can identify the forward paths and loops in the system and their corresponding gains. l'!> EXAMPLE 32~5 From the SPO in Fig. + G2G3H2 + G4H2 (3~70). G) G4 Loop Gains: 1. Yout!Y2. The equivalent SPG of the system is shown in Fig. The gain. However. may be written as =y= Y2 YoU! Yout Yin ~ Yin EMkdk Ifrom Yin lOYuul d . G4H2. 4. (375) The last expression is obtained using Eq. 5. G2G3H2. 334.. the last equation is written YOUl n = EMkdklfro~YIP 1 + G3 H2 + H4 L'Mkdk Ifrom Vin lOYoul tOY2 (370) Notice that d does not appear in the last equation. 333.3. 334(a).
as shown next. Y2 R Y3 Y1 . (376) . which correspond to the block diagrams shown in Fig.H2 Y Y I (b) Figure 334 (a) Block diagram of a control system . 335. + (a) G4 G1 G.... . (354) takes a far simpler look.Yin I . Y2 . 3210 Simplified Gain Formula From Example 326.33 MATLAB Tools and Case Studies ~ 129 r~ G4 1 . 1 s+ 1 I G1 \s) = . as shown in the following example .Y4 in Example 323) in the block diagram or SFG of the system. L (376) Redo Examples 32~2 .. (b) Equivalent signalflow graph . however. EXAMPLE 331 Consider the following transfer functions.. 33 MATLAB TOOLS AND CASE STUDIES There is no specific software developed for this chapter. ..r + Use MATLAB \0 H(s) = lO (377) lind the transfer function Y(s)/R(s) for each case ... YOUI Forward Patiz Gains M..Loop Gains through 326 to continn the validity of Eq .. it was felt that students may master this subject without referring to a computer. MATLAB may be used..Y3 . For simple operations.G2(S) = . then Eg.)'2 and)'4 . if there are no nontouching loops and forward paths (e. As a general rule. Although MATLAB Controls Toolbox offers functions for finding the transfer functions from a given block diagram. ..2' G(s) = (J)' 1' s+ s+ s.. .g .... we can see that all/oops andforward paths are touching in this case. The results are as follows ..
()n : s+2 s+2 .130 ~ Chapter 3. Block Diagrams and SignalFlow Graphs R(s) ~~~I~~~I_~ yes) (a) R(s) ~ (b ) R(s) 1 . G1=1/(s+1) Transfer function: s + l s+l »G2=tf( [ 11] .. + yes) Figure 335 Basic block diagrams (el used for Example 331. [1 2]) Transfer functio n : s+l s+2 s+2 » YR=G1' G2 Transfer function: s+l s"2 + 3 s + 2 » YR=G1' G2 » G2= (s+1)/(s+2) Transfer function: s+l Tr ansfer function: s+l » 1 YR_sirnple=minreal (YR) » YR_ sirnple=rninreal (YR) Transfer function: 1 Transfer func1:i . [11]) Transfer function: 1 Y(s) R(s) » » 1 s+ 1 = s2 + 3s + 2 = 1 (s + 2) Approach 1 » s clear all = tf(' s') ... Toolbox 331 Case (a) Use MATLAB to find G) * G2 Approach 2 »clear all »G1=tf( [1] .
. G2) Transfer function: sA2 + 3 s + 3 Use "minreal(YR)" for pole zero cancellation..5) (s+ l)(s+2) Approach 2 Approach 1 » 8 = tf( . if necessary Alternatively use "YR=parallel(GI. »clear all » Gl=tf( [1] I [11]) Transfer funct:i.el(Gl..ero/pole/Gain format: » zpk(YR) Use "zero{YR. 8 ' ) » G=l/(s+l) »clear all .3 3 MATLAB Tools and Case Studies 11 131 M Use "minreal(YR)" for pole zero cancellation.G2r· instead of ~'YR=G 1*02" Case (b) Usc MATLAB to find Gt + 02 Y(s) R(s) 2s + 3 = s2+3s+2 = 2(s + 1.on: 1 Transfer function: 1 8+1 »G2= (s+1)/(s+2) s+l » G2=1:f( [11] .5000 .5000 + O. 8660i » pole(YR) ans= 2 1 Zero/pole/gain: (s"'2+35+3) (8+2) (8+1) ." to obtain transfer fUIlction zeros.G2)" instead of "YR=GI+G2" Use "zpk{YRr to obtain the real z. if necessary Alternatively use .Gl+G2 Transfer function: sA2 + 3 s + 3 ::» YR=G1+G2 Transfer function: sA2 + 3 s + 3 »YR==paralJ. Use "pole{YR)" to obtain transfer function poles: » zero (YR) ans= 1. G2) Transfer function: s"2 + 3 s + 3 » YR=parallel (G1. [1 2]) Transfer function: s+1 s+2 Transfer function: 8+1 8+2 »YR.O. 4YR=series(G 1. a660i 1.:::.
Block Diagrams and Signal·Flow Graphs Toolbox 3..1225i s"2 + s + 10 .122Si 0.132 ~ Chapter 3. s · ) i » G=1/(s*(s+1)) Transfer function: 1 »clear YR » »clear all G=t f C[ 1] I [1 1 0] ) Transfer function: 1 8"2 + s s"2 + s » H=10 » H=10 H= H= 10 10 » YR=G/(1+G*H) Transfer function: 8"2 » YR=G/ (l+G*H) Transfer function: +8 8 s"2 + s s"4 + 2 8"3 s"4 + 2 s 3 + 11 s"2 + 10 s A + 11 8"2 + 10 » YR_simple=minreal CYR) Transfer function: » YR_simple=minreal(YR) Transfer function: 1 8"2 1 + S + 10 sA2 + s + 10 Use "minreal(YR)" for pole zero cancellation.3.5000 .3·2 Case (b) Use MATLAB to find the closedloop feedback function 1 +GGH Case (c) ( = ) Y(s) Rs 1 0 s+s+1 2 Approach 1 Approach 2 » s =t f ( .5000 + 3. if necessary Alternatively use: Use "pole(YR)" to obtain transfer func tion poles: H) w » YR=feedback(G I » pole (YR) ans Transfer function: 1 = 0.
State the form to which the equations must first be conditioned before drawing the SFG. "Feedback TheorySome Properties of Signal Flow Graphs. 8. Mason. 9. IRE. 13. A powerful method of representing the interrelationships between the signals of a linear system is the signal~flow graph.wiley. . T.comlcollege!golnaraghi." Elec. Transfer functions. 70. Multivariable and singlevatiable systems were examined. Define the output node of an SFG. 41. Sept. or SFG. No. and signalflow graphs were defined. Can signalHow graphs (SPGs) be applied to nonlinear systems'? How can SFGs be applied to systems that are described by differential equations? Define the input node of all SFG. D. REVIEW QUESTIONS 1. : . pp. S. lO. Ellg. 4. 11441156. 15. Does the il of an SFG depend on which pair of input and output is selected? List the advantages and utilities of the state diagram. The state variables of a dynamic system are not equul to the number of energystorage elements under what condition? Answers to these review questions can be found on 1his book's companion Web site: www. Can the gain formula be applied between any two nodes of an SFG'? Explain what the nontouching loops of an SFG are. 1953. MATLAB was used to calculate transfer functions of simple block diagram systems. 16. 2.9 pp. what happens to the initial conditions of the system? Define the characteristic equation of a linear system in terms of the transfer function. Define the transfer function of a linear time~invariant system in terms of its impulse response. 6. diagram representation was shown to be a versatile method of portraying linear and nonlinear systems. Vol. J. define the srate variables? 17." Pro£'. block diagrams. how do you write the state equations of the system? 19. When applied properly. Given the state diagram of a linear dynamic system. 7. REFERENCES Block Diagrams and SignalFlow Graphs 1. Graybeal.1951. "Block Diagram Network Transfonnation. Given the state diagram of a linear dynamic system.References 1'f 133 34 SUMMARY This chapter was devoted to the mathematical modeling of physical systems. how do you find the transfer function between a pair of input and output variables? 18. 14. What does the arrow on the branch of an SFG represent'? 11 + Explain how a noninput node of an SFG can be made into on output node. 2. Vol. At the end of the chapter. 3. 985990. Given the state diagram of a linear dynamic system. The block. The transfer function of a linear sysl~m was defined in terms of impulse response as well as differential equations. how do YOll 12. an SFG allows the derivation of the transfer functions between input and output variables of a linear system using the gain formula. When defining the transfer function. What is referred to as a multivariable system? S. A state diagram is an SFG that is applied to dynamic systems that are represented by differential equations..
McGrawHill. 3P3 and find the Y/X. "Feedback Theory—Further Properties of Signal Flow Graphs. Consider the block diagram shown in Fig. McGrawHill. Englewood Cliffs. StateVariable Analysis of Electric Networks 6. Reduce the block diagram shown in Fig. and J. C. Find: XX tg^ \ _ 1 . 44. Mason. B. New York. L. X 1 (s + 2) +Y Y (s+\) Figure 3P2 33. Robert. C. 3P1. PROBLEMS PROBLEMS FOR SECTION 31 31. NJ. M. The feedback transfer function. 1967. J E E> ' s(s + p) Kps 4 1*. 5. July 1956." Proc. 1967. The error transfer function. Signal Flow Graphs and Applications. J. B. 4. P. H2 X G1 H1 H3 Figure 3P3 G2 G3 Y . The closed loop transfer function. No. Kuo. S. Vol. Linear Circuits and Systems. IRE. Boisvert. 3P2 to unity feedback form and find the system characteristic equation. 32. Block Diagrams and SignalFlow Graphs 3. The forward path transfer function. Kuo. 1962. A. Linear Networks and Systems. New York. r Y ^ Figure 3P1 (a) (b) (c) (d) (e) The loop transfer function. Robichaud. 7 pp.134 Chapter 3. Prentice Hall. Reduce the block diagram shown in Fig. 920926.
consisting of the engine speed and turbineinlet temperature. Use MATLAB to solve Problem 35.s). 3P5(a) is controlled by a closedloop system with block diagram shown in Fig. 3P7. .I G(4 COMBUSTION G2 G3 Y 5 TURBINE PROPELLER Figure 3P5(a) RW. Reduce the block diagram shown in Fig. 3P5(b). E(i') G(i) Y(s) n(s) Figure 3P5(b) 36. (b) Find the closedloop transfer function ®0(s)/@r(s). and output vector Y{s). 3P4 to unity feedback form and find the Y/X. The aircraft turboprop engine shown in Fig. The block diagram of the positioncontrol system of an electronic word processor is shown in Fig. which contains the fuel rate and propeller blade angle. 37. The transfer function matrices are given as 2 10 s(s + 2) G{s) = H(*) = 5 1 s s + lFind the closedloop transfer function matrix [I + G ^ j H ^ ) I . (a) Find the loop transfer function ®0(s)/®e(s) (the outer feedback path is open). H3 X G1 H2 H1 Figure 3P4 35.Problems 135 34. The engine is modeled as a multivariable system with input vector E(.
2 + R( s) N(s) + s+2 10 s(s + I) Y(s) O. and H2(S) so that the output yes) is not affected by the disturbance signal N(s) at all.G4(s).136 > Chapter 3.Hl(S). 3PS. The block diagram of a feedback control system is shown in Fig. Block Diagrams and SignalFlow Graphs Current feedback Tachometer feedback Figure 3P7 3·8.G2(S). Find the following transfer functions : Yes) (a)  R(s) I N= O (b) yes) E(s) I N=O (c) Y(S)I N(s) R=O (d) Find the output yes) when R(s) and N(s) are applied simu ltaneously. .Ss Figure 3P8 3·9. 3P9. (a) Apply the SFO gain formula directly to the block diagram to find the transfer function s Y(S)I R(s) N=O Yes) I N(s) R=O Express Yes) in terms of R(s) and N(s) when both inputs are applied simultaneously. The block diagram of a feedback control system is shown in Fig.G3(S). (b) Find the desired relation among the transfer functions G 1(S).
Apply the finalvalue theorem. (a) Find the transfer function H(s) so that the output yes) is not affected by the disturbance torque N(s). (b) With H(s) as determined in part (a). Fig. Find tne transfer function Y(s) / N(s) IR=o.Problems 137 + Yes) N(s) Figure 3P9 310. R(s) = 1/ s"2. Fig. = = N(s) + Yes) + K(s s(s + + 3) l)(s + 2) Figure 3P11 . rtf) lUs(I) . and N(s) O. N(s) R(s) Ye s) + Figure 3P10 311. 3P1 0 shows the block diagram of the antenna control system of the solarcollector field shown in Fig. 15.1 when the input is a unitramp flillction. The signal N(s) denotes the wind gust di sturbance acted on the antenna. Determine the expression of Gd (S) so that the effect of N(s) is entirely eliminated . The feedforward transfer function Gd(s) is used to eliminate the effect of N(s) on the output Yes). 3Pl l shows the block diagram of a dcmotor control system. The signal N(s) denotes the frictional torque at the motor shaft. find the value of K so that the steadystate value of e(t) is equal to 0.
The feedback path is opened in lhis case.138 Chapter 3. . The block diagram of an electric train control i$ shown in Fig. (1)= voltage representing the desired train speed. find the steadystate speed of the train in feet per second when the input is er(t) 313. we apply a step func lion of 1 volt to the input of the controller.10')u s(t) (a) Find the transfer fl.15 V / ft / sec CONTROLLER G c(s) lI(r) Train speed SPEED DETECTOR Kt Figure 3P12 To determine the transfer function of the controller.61  0.0. 316.=O Y2(S) I R1 (S) R~=O Y) (S)I Y2 (S) I R2(S) R. (d) Assuming thatKis set at a value so that the train will not run away (unstable).. =O R2(S) R.5) when a step input of 1 V is applied to the controller. 3P12.3e. 315. =O PROBLEMS FOR SECTION 32 317.5 l )Us (tO.lnction Gc(s ) of the controller. ec(t) = us(t).Y2 t = r2 t + . d 2Y dl ) + 2 dYI (t) +3Y2(t) 2(1) dl = f J (t) + 1'2(1) ([2Y2(t ) dYI (I) () () () drl(t) . Use MATLAB to solve Problem 3.Jt Find the following transfer functions: Yl(S )1 R](s ) R. A linear timeinvariant multivariable system with inputs 1'1 (t) and 1"2( t) and outputs YI (t) and Y2(t) is described by the following set of differential equations. ft/sec vet) = Mass of trai n = 30.12. V = speed of train .6(. = us(t )Y.[i2" + 3 ~+Y l t . Block Diagrams and SignalFlow Graphs 312. (c) Derive the closedloop transfer function V(s) ! Er (s) of the system. (b) Derive the forward path transfer function V(s) ! E(s) of the system. 314. Use MATLAB to solve Problem 314. The output of the controller is measured and described by the following equation: f(t) = 100(1  0. The system and variables are parameter~ e. 3P4. 000 Ib / sec 2 K = amplifier gain = gain of speed indicator = 0.7e. M K. that is.. Find the stateflow diagram for the system shown in Fig. Repeat Problem 312 when the output of the controller is measured and described by the following expression: f (t) = 100(1  0.3e.
5 ~6 ~1] x+ [O~5 1.3X3 +3 (b) al + 3X2 + X3 Xl  = . (a) Draw an equivalent SFG for the system .5 0. (b) Find the following transfer functions by applying the gain formul a of the SFG directly to the block diagram.J 2X2 .!") . O (c) Compare the answers by applying the gain formu la to the equivalent SFG.2 +AIS +Aos 3·20. 0. Y Y4 Y2 Ys s functIons: Y1 YI Y1 Y2 .Problems 139 318. (a) Xl = x2 .5 o Find the statespace representation of a system with the following transfer function: B1s + Bos G(. 3P22 to find the fo llowi ng transfer . The block diagram of a control system is shown in Fig. Draw signalflow graphs for the fo llowing sets of algebraic equations. These equations should first be arranged in the form of causeandeffect relations before SFGs can be drawn. Draw a signalflow diagram for the system with the following statespace representation: x = [ S 0. 3P2 1. Show that there arc many possible SFGs for each set of equations.5 0. Y(S)I R(s) N=() Y(S)I E(5)1 N(s) R=O R(s) N=() E(S)I N(s) R".X:.5 z=[ 3·19.5 0.5 005]U 0. N(s) R(s) Yes) + Figure 3P21 322. = 1 321.\. Apply the gain formul a to the SFGs shown in Fig.5 0.
140 • Chapter 3. Block Diagrams and Signal~Flow Graphs (a) Ys (b) (d) (e) Figure 3P~22 .
3P24 is the equivalent circuit of an electronic circuit. The objective is to find the value of the constant k so that the output voltage R. Find the transfer functions Y7 /Y' and Y2/Y. The voltage source ed(t) represents a disturbance voltage. eo(c) Figure 3P24 .ke \(t) R. Signalflow graphs may be used to solve a variety of electric network problems. Shown in Fig. 3P23. el R2 i2 c3 i.Problems 323. Go 141 Ca) Cb) Figure 3P23 324. R) + R~ R. of the SFGs shown in Fig. + c. + '"'.
Find the gain eo / ed with all other inputs set to zero. For ed not to affect eo. Block Diagrams and SignalFlow Graphs eo(l) is not affected by ed(t). set ea/ed to zero. 325. (a) 1 G ~~~~~~~~3~~~~ H (b) Figure 3P25 326. This involves a combination of node and loop equations. To solve the problem. Show that the two systems shown in Figs 3P25(a) and (b) are equivalent.142 ~ Chapter 3. Show that the two systems shown in Figs.. (a) Figure 3P27(a) . it is best to first write a set of causeandeffect equations for the network. 3P27. Then constmct an SFG using these equations. 0 ~ 3P~26(a) and (b) are not equivalent. Find the following transfer functions for the SFG shown in Fig.I ff2 (a) 0 I ~ 0 Y(l H~ (b) Figure 3P26 327. Yi ~ HI G1 • 1 0 .
. (e) Find the following transfer functions: Y1(s)1 Rl (s) R~=O YI(s)1 Rz(s) R\=O Y2(S) Rl (s) I R2 =O Y2(S) Rz(s) I R\=O (d) Express the transfer functions in matrix [onn. Find the following transfer functions for the SFG shown in Fig. (a)  YI Y8=O (b) (c) (d) Y7! Y7! Y7! Y4 YB YJ=O Y4 Ys=o Y7! YJ=o Figure 3P28 329.Problems . 3P4{a) is shown in Fig. Comment on why the results for parts (e) and (d) are not the same. The signals are defined as RJ (s) = fuel rate R2 (s) = propeller blade angle Y1 (s) = engine speed Yz(s) = turbine inlet temperature (a) Draw an equivalent SFG for the system. The coupling between the signals of the turboprop engine shown in Fig. 3P29. (b) Find the t:l of the system using the SPG gain formula. 3P~28. yes) = G(s)R(s).. 143 (b) Figure 3PZl(b) 328.
( s.0 and ret) = uAt). Ge(s) (b) Let 100 Gp(s) = Ge(s) = (s+ 1)(s+5) Find the output response y(t) when (c) With N(s) .(s + ) s . N(s) + Y(s) Figure 3P30 331.( + I . . Keep in mind that the poles of the closedloop transfer function must all be in the lefthalf splane for the finalvalue theorem to be valid..144 Chapter 3. and Ge(s) and H(s) are the controller transfer functions. Block Diagrams and SignalFlow Graphs Figure 3P29 330. Gp(s) and Ge(s) as given in part (b).(s) . select H(s) among the following choices such that when net) = us(t) and ret) = 0.. (There may be more than one answer.:) s+:2:7) H(s) = lO(s + 1) 5+2 H(s) = sKI (n = positive integer) Select n. the steadystate value of y(t) is equal to zero. (a) Derive the transfer functions Y(s)/R(s)IN=O and Y(s)/N(s)IR=O' Find Y(s)/R(s)IN = 0 when = G..) 10 10 H (s) = . Figure 3P30 shows the block diagram of a control system with conditional feedback. The transfer function G pes) denotes the controlled process. Use MATLAB to solve Problem 330.1 H(s) = .
Problems 332.(t) . N(s) R(s) + + s+ I s+2 LO s(s + 20) yes) Figure 3P37 . (t) . (d) Perfonn a partialfraction ex pansion of yes)! R(s). (b) Find the characteristic equation and its roots. (t) . and r(t) is the input. The differential equation of a linear system is d 3 y'(t) d 2 v(t) d}. Define the state variables from right to left in ascending order. Repeat Problem 334 for the differential equation given in Problem 335.s (t).' 2 + 6 . Use MATLAB to (a) Find the partialfraction expansion of Y(s) I R(s).dt2 + 5 +y (r) = rt ) + d y( dt 4 3 2 336. 3P37.[t = 5x.dr4 ) 4 . The block diagram of a feedback control system is shown in Fig. (t) + 3X2(1) 145 .[t = 2x.3+ 5 . 334. (d) Plot the step response of the system. 333. 337. (b) Find the transfer functions XI (s) / R(s) and Xz(s )/R(s). (0 Find the final value of y(t) by using the fina lvalue theorem. (e) Find the output yet) for t 2: 0 when r(t) = u. (e) Find the transfer function Y(s) ! R(s) . Repeat Problem 333 for the following differential equation: d y(t d y(t) dy(t ) ( . Consider the differential equation given in Problem 333. Use MATLAB to find the roots.+ 10 (t) = r(/) ' Y dxz(t) dt dt dt where y(t) is the output. (a) Write the state equation of the system.dt3 t) + 3 . (e) Verify the final value that you obtained in Problem 333 part (f).5X2(t) + 21'(t) (a) Find the characteristic equation of the system. 3·35. Consider the following differential equations of a system: dx. (e) Find the output yet) for t 2: 0 when r(/ ) = us(t) . (b) Find the Laplace transform of the system.
Block Diagrams and SignalFlow Graphs SignalFlow (a) Derive the following transfer functions: transfer functions: yes) R(s) I N=() Y(s) E(s)1 N(s) R=O R(s) N=0 R=0 R(S) N=O I (5) is for the reduction the noise N(s).v 238/ + 36s3 + 38s2 + 285+16 of PI P2.PZ)""P1' Ps (P. ADDITIONAL PROBLEMS 339. Assuming 4 P. of N(s). (e) (d) Find the steadystate value of e{t) when the input is a unitstep function.\'2 + 62s + 12 342. (b) Use MATLAB to calculate P3 = P 2 . feedback control for transfer 340. Use MATLAB to calculate the polynomial (a) P 6 = (s + 1)(s2 + 2){s + 3)(2. == 2s6 9 5 15s 4 + 25s / + 25s + 148 + PI = 2s* + 9s / + 15s + 2 5 3 + 2 5 /2 + 145+ 6 P 2 = / ' + 8s5 + 2 3 4 + 3 6 / + 3 8 / + 28s + 16 P2 == s6 + 8.::::: G ( 5 G2(S) (c) G5(S) .s2 + ])(s I 2)(s I 4)(s2 jI 29 + 1) P7 .~z + s + 1) P6 = (5 1 ) ( / 2)(. G4(S) yes) (c) Find the characteristic equation and its roots when G4(S) is as determined in part (b). 340. characteristic G&(s) is as determined in part (b). (. (b) The controller with the transfer function G4 (s) is for the reduction of the effect of the noise N(s). (a) Use MATLAB to find roots of Pj and Pz.(s) ).(5) + G2(5) (b) G4(S) = G) . P4 = P 2 + p).(s) = GI (s) * 02(5) (d) G6 (s) = G[(s)*G2(s) . Use MATLAB to solve Problem 337.v) as determined in part (b).9 + 3 ) ( 2 / + 5 + 1) (b) P 7 = (. transfer function Find G4(s) so that the output Y(s) is totally independent of N(s). Use MATLAB to calculate unity feedback closed loop Problemfunction in Problem 341. G4(s) (c) G5(s) = G4(S) G3(S) (h(s) G4(s) 04(S) (d) G(. Set N(s) = 0.v + 12 ^ = %G + 28s5 + 83s4 + 13Ss3 + 126.::::. and P 5 = (PI .146 ~ Chapter 3. P 2 )*Pi. of e(t) N(s) O./ 4 1)(s·1 2)(* I 4 ) ( / 2^ + 341. P 3 = Pz Pi. y(t) 2: G4(8) 338. 4 z + P\. Use G4(. 343.2 + 4 7 5 + 60 s3+12s / +47s+60 b () a2 2(S) ~ 4s6 + 2855 + 8 3 / + 13553 + 12652 + 62.PI. Use MATLAB to calculate (a) G3(S) = GJ (s) + G2(5) G 3 (5)=G. Use MATLAB to perform partialfraction expansion to the following functions: functions: perform partialfraction (a) 01 (s) = (a) Gi(s) = ( + 1) ( / + 2)(s + 4)(5+10) ($5+ 1 )($2 + 2)(5 + 4)(s + 10) 5(5 + 2)(sZ + 2s + 5)(2s2 + s + 4) s(s 2 ) ( / + 25 + 5 ) ( 2 / + 5 + 4) / + 1 2 .::::.C 2 ( 5 ) G4(5) . (e) Find >'(/) for t > 0 when the input is a unitstep function.
It should also be emphasized that the modern control engineer should place special emphasis on the mathematical modeling of systems so that analysis and design problems can be conveniently solved by computers. and computers. pneumatic. 147 . In this chapter. whereas the state equations can be applied to linear as well as nonlinear systems. sensors and actuators. Using the basic modeling principles such as Newton 's second law of motion or Kirchoff' s law. • To introduce modeling thermal and fluid systems. their counterparts for non1inear systems are usually quite complex. • To discuss sensors and actuators. The main objectives of this chapter are: • To introduce modeling of mechanical systems.CHAPTER ·4 Theoretical Foundation and Background Material: Modeling of Dynamic Systems One of the most important tasks in the analysis and design of control systems is mathematical modeling of the systems. A control system may be composed of various components including mechanical. Therefore. otherwise known as dynamic systems. In this textbook. It is not difficult to understand that the analytical and computer simulation of any system is only as good as the model used to describe it. the models of these dynamic systems are represented by differential equations.. and electrical. more importantly. • To discuss analogies. whenever necessary. The transfer function is valid only for linear timeinvariant systems. • To introduce modeling of electrical systems. we review basic properties of these systems. so that the system may be realistically characterized by a linear mathematical model. we consider systems that are modeled by ordinary differential equations . the controlsystems engineer often has the task of determining not only how to accurately describe a system mathematically but. thermal. how to make proper assumptions and approximations. fluid. The two most common methods of modeling linear systems are the transfer function method and the statevariable method. • To discuss linearization of nonlinear systems. Although the analysis and design of linear control systems have been well developed.
The equations of a linear mechanical system are written by first constructing a model of the system containing interconnected linear elements and then by applying Newton's law of motion to the freebody diagram (FBD). In Chapters 5 and 9. :. Definition: Mass is considered a property of an element that stores the kinetic energy of translational motion. the main objectives of the following sections are: • To demonstrate mathematical modeling of control systems and components. in most cases. Newton's law of motion states that the algebraic sum of extemal forces acting on a rigid body in a given direction is equal to the product of the mass of the body and its acceleration in the same direction. rotational. • To provide examples that improve learning.148 ~. as shown in Section 410. This chapter further is intended to be selfsufficient and will not affect the general flow of the text.174 ftlsec 2 in British units. The variables that are used to describe translational motion are acceleration. • To demonstrate how computer solutions are used to obtain the response of these models. or a combination of both. The latter are modeled by partial differential equations. (4~2). Mass is analogous to the inductance of electric networks. then M is given by M=g W (41) where g is the acceleration of free fall of the body due to gravity (g :::: 32. The motion of mechanical elements can be described in various dimensions as translational. it is easier and therefore preferred to approximate them with lumped mass models and ordinary differential equations. For translational motion. Of coursey in reality all systems are continuous. the equation of motion is Eq. The equations governing the motion of mechanical systems are often directly or indirectly formulated from Newton's law of motion. the coverage here is by no means exhaustive. through various examples and case studies. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Furthermore. but. This chapter represents an introduction to the method of modeling. and displacement. and g :::: 9. Eq. The law can be expressed as L external forces=Ma (42) . Chapter 4. (433) is used. Because numerous types of controlsystem components are available. 411 Translational Motion The motion of translation is defined as a motion that takes place along a straight or curved path. velocity.. If W denotes the weight of a body.~ 41 INTRODUCTION TO MODELING OF MECHANICAL SYSTEMS Mechanical systems may be modeled as systems of lumped masses (rigid bodies) or as distributed mass (continuous) systems. the fundamentals discussed here are utilized to model more complex control systems and to establish their behavior. whereas the former are represented by ordinary differential equations.8066 mlsec 2 in SI units). and for rotational motion.
. • Linear spring. or simply stiffness. and their relative velocity among others. so an exact mathematical description of the frictiona1 force is difficult.. and a is the acceleration in the direction considered.4. These are discussed separately in the following paragraphs. static friction. Viscous friction represents a retarding force that is a linear relationship between the applied force and velocity. frictional forces exist. Three different types of friction are commonly used in practical systems: viscous friction. (44) implies that the force acting on the spring is directly proportional to the displacement (deformation) of the spring.. ~y(t) ~ fl. spring may be a model of an actual spring or a compliance of a cable or a belt. 42. In general. The force equation is written as I{t) = Ma(t) = M d2Y (t) = Mdv(t) dt2 dt (43) where a(t) is the acceleration~ v(t) denotes linear velocity. where M denotes the mass.t) Figure 41 Forcemass system. Fig. a spring is considered to be an element that stores potential energy. 41 illustrates the situation where a force is acting on a body with mass M. y(O Figure 42 Forcespring system. such as that shown in Fig. The model representing a linear spring element is shown in Fig. 43. Eq. then Eq. The frictional forces encountered in physical systems are usually of a nonlinear nature. f(t) = Ky(t) (44) where K is the spring constant. The mathematical expression of viscous friction is f(t) = B dy(t) dt (46) K '11O'1+~ j(t) r. • Viscous friction. the pressure between the surfaces. a linea. (44) should be modified to f(t) . and y(t) is the displacement of mass M. For linear translational motion. . Whenever there is motion or tendency of motion between two physical elements. In practice. and Coulomb friction.T = Ky(t) (45) • Friction for translation motion. in addition to the mass. respectively. The character~ istics of the frictional forces between two contacting surfaces often depend on such factors as the composition of the surfaces. 1 Introduction to Modeling of Mechanicar Systems ~ 149 rMtll___. The schematic diagram element for viscous friction is often represented by a dashpot. If the spring is preloaded with a preload tension of T. the following system elements are also involved.
once motion begins. which is used to model friction in highprecision . The static frictional force can be represented by the expression f(t) = ±(Fs)ly=o (47) which is defined as a frictional force that exists only when the body is stationary but has a tendency of moving. we have to use other frictional models to represent the actual phenomenon accurately. Chapter 4. The mathematical relation for the Coulomb friction is given by (48) where Fe is the Coulomb friction coefficient. f f o o (a) (b) (e) Figure 4. Notice that. They are by no means exhaustive or guaranteed to be accurate. 44(a) shows the functional relation between the viscous frictional force and velocity. The sign of the friction depends on the direction of motion or the initial direction of velocity.. • Coulomb friction. Theoretical Foundation and Background Material: Modeling of Dynamic Systems 11 IfIJ B f+ y(t) J(i) It>. One such example is rolling dry friction [3. but the sign of the frictional force changes with the reversal of the direction of velocity.. Coulomb friction is a retarding force that has constant amplitude with respect to the change of velocity.4 Graphical representation of linear and nonlinear frictional forces. the static frictional force vanishes and other frictions take over. • Static friction. (b) Static friction. 44(c). In many unusual situations. (a) Viscous friction. The forcetovelocity relation of static friction is illustrated in Fig. where B is the viscous frictional coefficient.150 . 44(b). (c) Coulomb friction. The functional description of the frictiontovelocity relation is shown in Fig. It should be pointed out that the three types of frictions cited here are merely practical models that have been devised to portray frictional phenomena found in physical systems.4]. Figure 43 Dasbpot for viscous friction. Static friction represents a retarding force that tends to prevent motionfrom beginning. Fig.
4 mm 1 ft == 0.. The linear motion concerned is in the horizontal direction../tt) Ky(t) : B dy(t) M 41 I+~ j(t) M dt (b) (a) Figure 4·5 (a) Massspringfriction system.37 in 1 in. (b) Freebody diagram..B dy(t) _ Ky(t} dt = Md2y(t) 2 dt The last equation may be rearranged by equating the highestorder derivative term to the rest of the terms: d y(t) := dt2 2 _!!. The force equation of the system is I(t) . .06852 slug 1 m = 3.'" y(t) 1 . It turns out that rolling dry friction has nonlinear hysteresis properties that make it impossible for use in linear system modeling. TabJe 41 shows the basic translational mechanical system properties with their corresponding basic SI and other measurement units...5969 oz(force) 1 N == 1 kgm/s2 1 dyn = 1 gcm/s 2 Spring Consttlnt Viscous Friction Coefficient K B N/m Nlmlsec tb/ft Ib/ftlsec ball bearings used in spacecraft systems. ! . 45(b).. ~ EXAMPLE 411 Consider the massspringfriction system shown in Fig.2808 ft = 39.274oz(mass) Distance y meter (m) ft in mlsec m1sec 2 Newton (N) := 0. 151 TABLE 4·1 Basic Translational Mechanical System Properties and Their Units Parameter Symbol Used M SI Units kilogram (kg) Other Units slug ftlsec 2 Conversion Factors 1 kg:= 1000g = 2.3048 m Velocity Acceleration Force v ftJsee a f in/sec ftlsec 2 in/sec2 pound (lb force) dyne 1 N = O.. 45(a). dy(t) _ K y(t) + ~ /(t) M dt M M tt=1 B 1. = 25. yen ..20461b(mass) Mass := 35..22481b(force) = 3.4·1 Introduction to Modeling of Mechanical Systems . The freebody diagram of the system is shown in Fig.." .
:jl2 y(t) represent velocity and acceleration. 46. the transfer function between Y(s) and F(s) is obtained by taking the Laplace transform on both sides of Eq . Chapter 4. so that x= Ax + Bu where (413) x(t) = [X1(t) ] X2 (t) (414) (415) y(t) = and Xl (t ) yet) = x2 (I) u(t) = f(t) M (416) So using Eqs.MX2(s) . Or.MX I(s) Y(s) = XI (5) Y(s) 1 F (s) = Ms2 + Bs + K B K + MF(s) (419) 1 . where n is the number of stMe variables.t 1) block diagram representation. (411) with zero initial conditions: Y(s) J F(s )= Ms2+ Bs + K (4. alternatively.152 .tate vector x(t) having n rows. (417) may also be written ns a set of firstorder ditTerential equations: dX~?) = y(t) X2(t) . which is shown in Fig. where Y(t) dY(t)) = ( dt and y( t ) = (d 2y(t)) . (413) through (416). (4. the transfer function between Y(s) and F(s) is obtained by taking the Laplace transform on both sides of Eq. respectively.Ni X2 (t ) + M K f(l) (418) = X I (t) For zero initial conditions.Jt =  Ni x i (t) . Elj. (410) is rewritten in vectoral form as [Z~ J dX2(t) = (  Ei)[~~] + j~) B I (4.12) The same result is obtained by applying the gain formula to the block diagram.17) The state Eq. Eq. Theoretical Foundation and Background Material: Modeling of Dynamic Systems F(s) K Figure 46 The massspringfriction system of Eq. the former equation may be rewritten into an inputoutput form as + !y(t) + ~y(t) = ~ f (t ) (411) where y(t) is the output and f~) is considered the input. (4.18): SX I(S) = X2 (S) sX2 (s) = .. For zero initial conditions. (410) may also be represented "in the space state form using a .
(420) or by applying the gain formula to the block diagram representation of the system.XI (S) +. G = tf(num .\"2(0).. num= [lJ. xlabel( 'Time (Second)') . (418) has a different Laplace transform representation that may be wrilten as: SXI (s)  XI (0) = X2(S ) B K 1 = . (419). M = I. (41 9).41 Introduction to Modeling of Mechanical Systems 153 Figure 47 Block diagram representation of massspringfriction system of Eq. 48.den).X2 (S) . M=l. which is shown in Fig. = 1. B=l. Figure 48 Block diagram representation of massspringfrictio n system of Eq. y1 = STEP (G. The same result is obtained by applying the gain formula to the block diagram representation of the system in Eg . 47.ylabel(' Step Respon se') ti tle ( 'Response of the system to step input' ) . shown in Fig.1') M M M SX2(S) .1'"(.02:30.. den = [M B K] . p lot(t. the output becomes 1 Ms M Y(s) = Ms2 + Bs + K F(s) + Ms2 + Bs + K XI (0) + Ms2 + Bs + K X2(O) (42 1) Toolbox 411 Time domain step response jo/' Eq. For nonzero initial conditions. (420) with initial conditions XI (0) and . y1) . Eg. (412) is calculated using MATIAB jor K K=l. B = I: t=O:O.X2(0) (420) Y(s) = XI (s ) Upon simplifying Eg . t).
consider the system shown in Fig. The force equations are f{t) = Kfyl (/) .r D..a I! 0) f.Yt(t)J Figure 49 Mechanical system for Example 412. .t) K 1.2 5 10 15 Time (SecDnd) 20 25 30 l..154 ~ Chapter 4.Y2(t)} (422) K[V2(t) . must be assigned to the end points of the spring. Because the spring is defonned when it is subject to a foreeJt!).4 0.....fi.4~~~~~~~ 1.vI(t)] _ B dY2 (t) = M d2Y2 (t) . EXAMPLE 412 As another example of writing the dynamic equations of a mechanical system with translational motion. .. (b) 'll'l' . Theoretical Foundation and Background Material: Modeling of Dynamic Systems Response ofthe system to step input 1. (a) Massspringdamper system. K fit) dl Kb'z{t) . o. two displacements..2 CD ~ en a. Yl and Y2. (b) Free·body diagram.:. 49(b). dt dt 2 These equations are rearranged in inputoutput form as 2 d Y2(t) dt2 +B M dY2(t) dt + K V2(t) = M' K Yl (t) M I BdYz(l) IJII B M M ~~ (a) ~. 49(a). .6 0.. The freebody diagrams of the system are shown in Fig. ..
(427) and applying the gain formula to the block diagram representation of the system. which is shown in Fig. The state equations are therefore written as (427) The same result is obtained after tak. the transfer function of Eq. Note that in Fig. By using the last two equations. For zero initial conditions. 410. x)(t). respectively.41 Introduction to Modeling of Mechanical Systems I K ISS M YI K S . Note that in Fig. and X 2 (s) are Laplace transfonns of . Y2(S) . 411. and '2(t). (b) Block diagram representation.ing the Laplace transform of Eq. 4] 1. F(s). (425) using Eq. . Yl(t). (424): Y2(S) K Yl(S ) = Ms2 +Bs+K (425) The same result is obtained by applying the gain formula to the block diagram representation of the system. (426) is the same as Eg. For state representation.f{t). X)(s).I \' I KIM (a) (b) Figure 410 Massspringfriclion system of Eg. which is shown in Fig. these equations may be rearranged as (426) For zero initial conditions. (425). 410. the state variables are defined as Xl (t) = Y2(t) and X2(1) = dY2(t) ( dt. Y2(t). the transfer function between Yl(S) and Y2 (s) is obtained by taking the Laplace transform on both sides of Eg. (422) was also used. Eg. (422). YI(S). (a) The signalflow graph representation.
156
Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
Figure 411 Block diagram representatioIl of massspringfriction system ofEq . (427).
EXAMPLE 413
Consider the two degrees of freedom (2DOF) springmass system, with two masses nil and 1/12> two springs kl and k2' and two forces il and.h as shown in Fig. 4[2. Find the equations of motion.
SOLUTION To avoid any confusion, we first draw the freebody diagram (FBD) of the system by assuming the masses are displaced in the posi tive direction, so that YI > Y2 > 0 (i.c., springs are both in tension). The FBD of the system is shown in Fig. 41 3. Applying Newton 's second law to the masses M I and M 2 , we have i l (t)  K IY I
+ K2(Y I 
Y2) = Ml.h
h(t)  K2(Y I  Y2) = M2h
Rearranging the equations inco the standard inputoutput form, we have
Mj);l
(428)
+ (Kt + K2)YI
M2Yl  K2Yl
 K2Y2 = i l (t)
+ K2Y2 = J2(t)
(429)
Alternatively, Eq. (429) lUay be represented in the standard secondorder matrix. form, as
[~t ~2][t~ J + [K~i:2 K~2 ][~~J = [ j~]
XI
(430)
In state space form, assuming the following state vector x(t), the inputs UI(t) and U2(t) , and the output yet), we get
x(t) =
[
(t) X2(t) X3 (t) X4(t)
1 [YI 1
=
(t) Y2(1 ) . )11 (t ) ,Ill = It (t), U2 = h(l) . yet) =
Xl (l )
(431 )
.Yz (t)
Figure 412 A 2DOF springmass system.
Figure 413 FBD of the 2DOF springmass system .
4~1
Introduction to Modeling of Mechanical Systems ' 157
Then, using X3
= YI
,
and X.4
= Y2' we get the statewspace representation as
+
IIMl 0
III
[:: [~:1 [~ ~ ~ ~1 1 [~l [~l
~3
.::::
KIJMI
Kt/Ml
0
0.1:3
+
112
(state equation)
0
X4
KdM\
K2/Mt
0
0
X4
11M2
Y = rIO
0
01
[~: 1 o·
+
X4
UI
+ 0 . U2
(output equation)
where the state equation is a set of four firstworder differential equations.
4..12 Rotational Motion
The rotational motion of a body can be defined as motion about a fixed axis. The extension of Newton's law of motion for rotational motion states that the algebraic sum of moments or torque about a fixed axis is equal to the product of the inertia and the angular acceleration about the axis. Or 2:torques
= JOt
(433)
where J denotes the inertia and Ot is the angular acceleration. The other variables generally used to describe the motion of rotation are torque T, angular velocity w, and angular displacement The elements involved with the rotational motion are as follows:
e.
• Inertia. Inertia. J, is consideretl a property of an element that stores the kinetic energy of rotational motion.. The inertia of a given dement depends on the geometric composition about the axis of rotation and its density. For instance, the inertia of a circular disk or shaft, of radius r and mass M. about its geometric axis is given by
(434)
When a torque is applied to a body with inertia J, as shown in Fig. 414, the torque equation is written
T(t) = Ja(t) = Jdw(t) = J d 9(t)
dt
2
dP
(435)
where 9(t) is the angular displacement; w(t), the angular velocity; and a(t), the angular acceleration.
T(J)
~")+~1
0
(1)
FIgure 414 Torqueinertia system.
158 .. Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
III___________T_(l)~
.. K }
O(t)
Figure 415 Torque torsional spring system.
• Torsional spring. As with the linear spring for trans1ational motion. a torsional spring constant K, in torqueperunit angular displacement, can be devised to represent the compJiance of a rod or a shaft when it is subject to an applied torque. Fig. 415 illustrates a simple torquespring system that can be represented by the equation
T(t) = Kf)(t)
(436)
If the torsional spring is preloaded by a preload torque ofTP, Eq. (436) is modified to
T(t)  TP
= KO(t)
(437)
• Friction for rotational motion. The three types of friction described for translational motion can be carried over to the motion of rotation. Therefore, Eqs. (46), (47), and (48) can be replaced, respectively. by their counterparts: • Viscous friction. T(t) • Static friction.
(439)
= B dB(t)
dt
(438)
• Coulomb friction.
dOCt)
T(f) = Fc
Id~{t)1
(440)
Table 42 shows the SI and other measurement units for inertia and the variables in rotational mechanical systems.
,.. EXAMPLE 414 The rotational system shown in Fig. 416(a) consists of a diskmoutlted on a shaft that is fixed at one
end. The moment of inertia of the disk about the axis of rotation is J. The edge of the disk is riding on the surface, and the viscous friction coefficient between the two surfaces is B. The inertia of the shaft is negligible. but the torsional spring constant is K. Assume that a torque is applied to the disk. as shown; then the torque or moment equation about the axis of the shaft is written from the freebody diagram of Fig. 416(b):
d28 (t) Bd8(t) K8() (441) dt + t ,12 + t Notice that this system is analogous to the translational system in Fig. 45. The state equations may be written by defining the state variables as Xl (t) :::= O(t) and X2(t) = dXI (t)ldt. T( ) = J
t
4~ 1
Introduction to Modeling of Mechanical Systems
~\\\\\\\\\\\\\\\\\\\\\\\\\
~
159
K
~~
o
(Po
\ B dO dt
J
(a)
(b)
Figure
4~16
Rotational system for Example
4~1~4.
TABLE 42
Parameter
Basic Rotational Mechanical System Properties and Their Units
Symbol Used
SI
Units kg_m
2
Other Units
Conversion Factors Igcm = 1.417 x 1O5 0zin.sec 2 Ilbftsec2 = 192ozin.sec2 = 32.2Ibft2 1 ozin.sec2 = 386ozin2 1 gcmsec2 = 980 g~cm2 180 1 rad =  = 57.3deg
7C
Inertia
J
slugftZ
Ib~ftsec2
Oztn.sec
•
2
Angular Displacement Angular Velocity
T
0
Radian radian/sec
Radian
radian/sec
21l' lrpm=60 = O.1047rad/sec 1 rpm = 6 deg/ sec
Angular Acceleration Torque
A
T
radian/sec 2 (Nm) dyneem
radian/sec 2 lbft ozin. ftlb/md ft ~lb/radlsec
1 gcm = 0.0139 ozin. Ilbft = 192 ozin. 1 ozin. = 0.OO521Ib~ft
Spring Constant Viscous FrictiolZ Coefficient
Energy
K B Q
Nm/rad
Nm/radlsec J (joules)
Btu
Calorie
IJ = INm IBtu = 1055J 1 cal = 4.1841
. EXAMPLE 415 Fig. 417(a) shows the diagram of a motor coupled to an inertial load through a shaft with a spring
constant K. A non~rigid coupling between two mechanical components in a control system often causes torsional resonances that can be transmitted to all parts of the system. The system variables and parameters are defined as follows:
Tm(t)
= motor torque
Bm
= motor viscousfriction coefficient
K = spring constant of the shaft
Om(t) = motor displacement wm(t) = motor velocity
160 Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
I MOTOR I _ KHJI 14}
J
III'
B
III
8m
(a)
()t. L..._~...J
(b)
Figure ~17 (a) Motorload system. (b) Freebody diagram.
Jm = motor inertia
£h(t) = load displacement wL( t) = load velocity
h. = load inertia The freebody diagrams of the system are shown in Fig. 417(b). The torque equations of the system are
(442)
(443)
In this case, the system contains three energystorage elements in Jm , h, and K. Thus, there should be three state variables. Care should be taken in constructing the state diagram and assigning the state variables so that a minimum number of the latter are incorporated. Eqs. (442) and (443) nre rearranged as
2 8 m d8m (t) K d tn (t) _ 2
e
dt
Jm
dt
lin
Iemt ()
(J ( )}
Lt
1T +Im ( t ) 1m
(444)
d2 edt) K ~ = JL [em(t) 
8d t )J
Xl (t) =
(445)
The state variables in this case are defined as X3(t) = d(jm(t)fdt. The s.tate equations are
8m (t)  t1[.(t), X2(1) = d8L(t)/dt. and
dtdt)
~
= X3(t) X2(t)
dx2(t)
dt
= K Xl (t)
IL
(446)
 ' =
dt3(t) dt
Xl(t)  X3(t) +T,n(t)
Jm
Jm
K
Bm
1
1m
The SPG representation is shown in Fig. 418.
Figure 418 Rotational system of Eq. (446) signalflow graph representation.
41 Introduction to Modeling of Mechanical Systems
161
413 Conversion between Translational and Rotational Motions
In motioncontrol systems, it is often necessary to convert rotational motion into translational motion. For instance, a load may be controlled to move along a straight line through a rotary motorandlead screw assembly, such as that shown in Fig. 419. Fig. 420 shows a similar situation in which a rackandpinion assembly is used as a mechanical linkage. Another familiar system in motion control is the control of a mass through a pulley by a rotary motor, as shown in Fig. 421. The systems shown in Figs. 419,420, and 421 can all be represented by a simple system with an equivalent inertia connected directly to the drive motor. For instance, the mass in Fig. 421 can be regarded as a point mass that moves about the pulley, which has a radius r. By disregarding the inertia of the pulley, the equivalent inertia that the motor sees is (447) If the radius of the pinion in Fig. 420 is r, the equivalent inertia that the motor sees is also given by Eg. (447). Now consider the system of Fig. 419. The lead of the screw, L, is defined as the linear distance that the mass travels per revolution of the screw. In principle, the two systems in Fig. 420 and Fig. 421 are equivalent. In Fig. 420, the distance traveled by the mass per revolution of the pinion is 2][r. By using Eq. (447) as the equivalent inertia for the system of Fig. 419, we have
J = W (~)2 g 2n:
(448)
Motor
Lead screw
Figure 419 Rotarytolinear motion control system (lead screw).
~ x(t)
[W]
•;;~;:; ~iJ{'J
Rack
motor
DriV~.,
T(t)
K
Figure 420 Rotaryto Linear motion control system (rack and pinion).
Figure 421 Rotarytolinear motion control system (belt and pulley).
162 .. Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
MNI
N'~
4~ 14
ILL
Figure 422 Gear train.
Gear Trains
A gear train, lever, or timing belt over a pulley is a mechanical device that transmits energy from one part of the system to another in such a way that force, torque, speed, and displacement may be altered. These devices can also be regarded as matching devices used to attain maximum power transfer. Two gears are shown coupled together in Fig. 4~22. The inertia and friction of the gears are neglected in the ideal case considered. The relationships between the torques Tl and T2 , angular displacement 81 and 02, and the teeth numbers Nt and N2 of the gear train are derived from the following facts: 1. The number of teeth on the surface of the gears is proportional to the radii, J and '2 of the gears; that is,
(449)
2. The distance traveled along the surface of each gear is the same. Thus,
(450)
3. The work done by one gear is equal to that of the other since there are assumed to be no losses. Thus,
(451)
If the angular velocities of the two gears WI and W2 are brought into the picture. Eqs. (449) through (451) lead to
(452)
In practice, gears do have inertia and friction between the coupled gear teeth that often cannot be neglected. An equivalent representation of a gear train with viscous friction, Coulomb friction, and inertia considered as Jumped parameters is shown in Fig. 423. where T denotes the applied torque. T1 and T2 are the transmitted torque, Fcl and Fc2 are the Coulomb friction coefficients. and B 1 and B2 are the viscous friction coefficients. The torque equation for gear 2 is (453)
4~1
introduction to Modeling of Mechanical Systems . 163
[J) BI:I?Fd )
T.81
T2
N2
NJ
J2
Figure 423 Gear train with friction and inertia.
The torque equation on the side of gear 1 is
d 2 fh (t) del (t) T(t) = J1 ~ + Bl ;[t + Fc1
Using Eq. (452), Eq. (453) is converted to
IWII + Tl (t)
WI
(454)
T ( ) = Nl Y, ( ) = (Nl) J d 01 (t) I t N2 2 t N2 2 dt2
2
2
+
(Nl)2 B
N2
2
d(h (t) dt
+ N2
Nt F. W2
c2
\W2\
(455)
Eq. (455) indicates that it is possible to reflect inertia, friction, compliance, torque~ speed, and displacement from one side of a gear train to the other. The following quantities are obtained when reflecting from gear 2 to gear 1:
Inertia:
(Z:Yh
(Z:)
2 R2
Viscousfriction coefficient:
Nt Torque: N2 T2
Angular displacement: Nt
N2
£02
(456)
(h
Angular velocity:
Z:
Coulomb friction torque:
Z: FeZ 1:1
Similarly, gear parameters and variables can be reflected frorn gear 1 to gear 2 by simply interchanging the subscripts in the preceding expressions. If a torsional spring effect is present~ the spring constant is also multiplied by (Nt /N2)2 in reflecting from gear 2 to gear 1. Now substituting Eq. (455) intu ElJ. (454), we get
(457)
where
(458)
164,
Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
(459)
(460)
EXAMPLE
416 Given a load tbat bas inertia of 0.05 ozin.sec 2 and a Coulomb friction torque of 2 ozin ., find the inertia and frictional tOrque reflected through a 1:5 gear train (NIl N2 1/ 5, with N2 on the load
=
side). The reflected inertia on the side of N J is (1/ 5)2 x O.05 = 0 .002 ozin .sec 2 . Thc reflected Coulomb friction is (l / 5) x 2 = 0.4 ozin.
415 Backlash and Dead Zone (Nonlinear Characteristics)
Backlash and dead zone are commonly found in gear trains and similar mechanical linkages where the coupling is not perfect. In a majority of situations, backlash may give rise to undesirable inaccuracy, oscillations, and instability in control systems. In addition, it has a tendency to wear out the mechanical elements. Regardless of the actual mechanical elements, a physical model of backlash or dead zone between an input and an output member is shown in Fig. 424. The model can be used for a rotational system as well as for a translational system. The amount of backlash is bt2 on either side of the reference position. In general, the dynamics of the mechanical linkage with backlash depend on the relative inertiatofriction ratio of the output member. If the inertia of the output member is very small compared with that of tbe input member. the motion is controlled predominantly by friction. This means that the output member will not coast whenever there is no contact between the two members. When the output is driven by the input, the two members will travel togetber until the input member reverses its direction ; then the output member will be at a standstill until the backlash is taken up on the other side, at which time it is assumed that the output member instantaneously takes on the velocity of the input member. The transfer characteristic between the input and output displacements of a system with backlash with negligible output inertia is shown in Fig. 425 .
.1'(1)
f. x(t)
'InpU(~~
f. y( t)
Figure 424 Physical model of backlash between two mechanical elements.
r
_
. t b.d ~ I+
i
Figure 425 Inputoutpllt characteristic of backlash.
Output _
4·2 Introduction to Modeling of Simple Electrical Systems <: 165
R
~
i(t)
(a)
(b)
Figure 4·26 Basic passive electrical elements. (n) A resistor. (b) An inductor. (c) A capacitor.
.. 42 INTRODUCTION TO MODELING OF SIMPLE ELECTRICAL SYSTEMS
First we address modeling of electrical networks with simple passive elements such as resistors, inductors, and capacitors. Later, in the next section, we address operational amplifiers, which are active electrical elements.
42~1
Modeling of Passive Electrical Elements
Cunsider Fig. 426, which shows the basic passive electrical elements~ resistors, inductors, and capacitors.
Resistors: Ohm's law states that the voltage drop. eR(t), across a resistor R is proportional
to the current i(t) going through the resistor. Or
eR(t) = i(t)R
Inductors:
(461 )
The voltage drop. eL(t). across an inductor L is proportional to the time rate of change of current i(t) going through the inductor. Thus,
edt) = L di(t)
dt
(462)
Capacitor: The voltage drop, ec(t), across a capacitor C is proportional to the integral current i(t) going through the capacitor with respect to time. Therefore,
(463)
422 Modeling of Electrical Networks
The classical way of writing equations of electric networks is based on the loop method or the node method, both of which are formulated from the two laws of Kirchhoff, which state:
Current Law or Loop Method:
node is zero.
The algebraic summation of all currents entering a The algebraic sum of all voltage drops around a
Voltage Law or Node Method:
complete closed loop is zero.
'. EXAMPLE 421 Let us consider the RLC network shown in Fig.
4~27. Using the voltage law
e(t) = eR + e[. + e"
166
Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
R
0
VINI.
~
+
L '000'
i(t)
e(t)
C
0
(a)
1 J
',l"
i(O)
"'
I
 ' s
ec(O)
I
e(t)
.r
I
C
e t'
 [
(b)
1
s I
e,.
I
(e)
Figure 427 RLC network. (a) Electrical schematics. (b) Signalflow graph representation . (c) Block diagram represenlation .
where
= Voltage across the resistor R eL = Voltage across the inductor L e(' = Voltage across the capaci tor C
eR
Or
e(t) = +ec(t)
Using current in C:
+ Ri (l) + LTt
diet)
(465)
(466) dl and taking a derivative ofEq. (454) with respect to time, we get the equation of the RLC network as
c[2i(t) diet) i(t) de(t) L 2 + R  +=  dt dt C dE
Cdec(t) = itt)
(467)
42 Introduction to Modeling of Simple Electrical Systems
~
167
A pmctical approach is to assign the current in the inductor L. l(t), and the voltage across the capacitor C, ec(t). as the state variables. The reason for this choice is because the state variables are directly related to the energystorage element of a system. The inductor stores kinetic energy, ann the capacitor stores electric potential energy. By assigning l(t) and ec(t) as state variables, we have a complete description of the past history (via the illjtial states) and the present and future states of the network. The state equations for the network in Fig. 427 are written by first equating the CUITent in C and the voltage across L in terms of the state variables and the applied voltage e(t). In vectormatrix form, the equations of the system are expressed as
(468)
This format is also known as the state form if we set
[XI (I) 1= X2(t)
Or
[e("(t) i(t)
1
(469)
[XI] ['~l] = [0 b1 + [~] e(l)
X2
I L
R 
X2

(470)
L
L
The transfer functions of the system are obtained by applying the gain formula to the SFG or block diagram of the system in Fig. 4~27 when all the initial states are set to zero.
Ec(s) {l/LC)s2 _ 1 £(s) = 1 + (R/L)sI + (1/ LC)s2  1 + Res + LC.v2
/(s) (l/L)sl E(s) :; 1 + (RjL)sl + (ljLC)s2
Cs
1 +RCs+ LCs2
(472)
~
Toolbox 421
Time domain step responses/or Eqs. (471) and (472) are shown using MATLAB for R = 1. L = 1, C = J:
R=l; L=l; C=l;
t=0:O.02:30;
num1 == [1] ;
den1 == [L*C R*C 1] ;
num2 == [C 0] ;
den2.:= [L*C R*C 1] ; G1 :: tf(num1, den1) j G2 = tf(nurn2, den2) ; yl = step (Gl,t); y2 = step (G2,t); plot(t,yl, 'r'); hold on plot Ct , y2, 'g'); xlabel( 'Time') ylabel( 'Gain')
168
Chapter 4. Theoretica l Foundation and Background Material: Modeling of Dynamic Systems
70~rr~~~~
60
50
AD
~
C'
30
20
.~ \ r I
J
I
\
10~
o o
\
'.
\ , J' /
.                    1
5
10
_10~L~~JJJ
15
20
25
30
Time
EXAMPLE 422 As ano ther exampl e of writing the stale equation s of an electric network, consider the network
shown in Fi g. 428(u) . According to the foregoin g discussion, the voltage across the capacitor, ec(L) , and the currents of the inductors, i le t) and i2(t) , are assigned as state variables, as shown in Fig . 428(a) . The state equations of the network are obtained by writing the voltages across the
e(l)
(al
 lie
1.
L I
s
I
J
C
e,.
s I
e,.
I
L:.
e(l)
il
RI /LI
 R/L z
IILI
(h)
Figure 428 Network of Example 422. (a) Electrical schematic . (b) SFG representation.
42 Introduction to Modeling of Simple Electrical Systems .q 169
inductors and the currents in the capacitor in terms of the three state variables. The state equations are
L)
dil (1) . cit = Rill (I) 
e(.(IJ
+ e(/)
(473) (474) (475)
diz (t) . L2 cit ~ R212(/)
+ ec(t)
dec ( I) . .( C ~=ll (t ) 121 )
In vectormatrix form, the state equations are written as
Rl
LI
0
Rz
o
1
C where
I LI 1
(476)
L2
o
[ ~~] = [~~g~]
X3
(477)
edt)
The signalflow diagram of the network, without the initial states, is shown in Fig. 428(b), The transfer functions between II(s) and E(s), Iz(s) and E(s), and Ec(s) and E(s), respectively. are written from the state diagram
II
(s)
2 L2CS
+ R:zCs + 1
8
E(s)
h(s) £(s)
=
1
(478)
=~
~S+R2
(479)
E(.'(s) E(s) = where
6.
(480)
Toolbox 422
Time domain step response for the gain formula explained by step responses of Eqs. 478480 are shown using MATLAB as illustrated below (for RI = J. R2 1, Ll J, L2 J, C = 1):
=
=
=
Rl=l; R2=1; Ll=l; L2=1; C=l; t=O:O.02:30; num1 = [L2*C R2*C 11; num2 = [1]; num3 = [L2 R2] ; den = [Ll*L2*C Rl*L2*C+R2*Ll*C Ll+L2+Rl*R2*C Rl+R2] ; Gl = tf(nurnl, den) ; G2 = tf(num2,den) i G3 = tf(nurn3 , den) ; y1 = step (Gl,t); y2 = step (G2 , t) ; y3 = step (G3 , t) ;
170
Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
plot:(t,yl, 'r') ; hold on plot(t,y2, ' g') ; hold on plot(t,y3, 'b'); xlabel ( 'Time' ) ylabel ( • Gain' )
0.7
0.6
0.5
I
0.4
';0
c:
~
I
0.3
0.2
0.1
D"
0
! I
5
l
10
15 TIme
20
25
30
EXAMPLE
423 Consider tile RC circuit shown in Fig, 429. Find the differential equation of the system. Using the voltage law
(482)
where
eR =
iR
(483)
and the voltage across the capacitor v,,(t) is
ec(t)
But from Fig. 429
=
C I
J'
ldt
(4·84)
eo(t ) = C I
J'
Idt = edt)
(485)
Figure 4·29 Simple electrical RC circuit.
42 Introduction to Modeling of Simple Electrical Systems
171
If we differentiate Eq. (485) with respect to time, we get _ = de,,(t) C dt
(486)
or (487) This implies that Eq . (485) can be written ill an inputoutput form
eill(t)
= Rce,, (t) + eo(t)
(488)
In Laplace domain, we get the system transfer function as Eo(s ) Eill(s) RC s + 1 (489)
where the r = Re is also known as the time constant of the system. The significance of this term is discussed earlier in Chapter 2, and the initial conditions are assumed to be eill(t = 0) =
ev(t
= 0) = o.
EXAMPLE 424
Consider the RC circuit shown in Fig. 430. Find the differential equation of the system.
c
Figure 430
Simple electrical
Re circuit.
As before, we ha ve (490) or
ein(t) =
But vo(t) = iR. So
~
f
idt + iR
(49l)
(492) is the differential equation of the system. To solve Eq . (492), we differentiate once with respect to time:
. eo (t) ei,,(l) = Re
+ eo
. ( t)
(493)
In Laptace domain, we get the system transfer function as Eo(s) where, aga in, r RC s
Eill(s) = Res + 1
(494)
= Re is the time constant of the system .
EXAMPLE 425
Consider the voltage di vider of Fig. 431. Given an input voltage eo (t), find the output voltage e/(1) in the circuit composed of two resistors R] and R2 .
172 • Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems
ry:, .~
R)
Figure 431 A voltage divider.
The currents in the resistors are
(495)
(496)
The node equation at the eJ(t) node is
it  i2
=0
R2
(497)
Substituting Eqs. (495) and (496) into the previous node equation:
eo,t)  e) (t) _ e) (t) = 0
Rl
(498)
Rearrangement of this equation yields the following equation for the voltage divider:
el (t)
= R2 R2 eo(t) R 1+
= RZ R Eo(s) R 1+ 2
(4100)
In Laplace domain. we get EI (s)
The SI and most other measurement units for variables in electrical systems are the same. as shown in Table 43.
TABLE 43
Parameter Charge Current Voltage Energy Power Resistance Capacitance
Basic Electrical System Properties and Their Units
Notation Units coulomb = newtonmeter/volt ampere (A) volt (V) joule = volt x coulomb joule/sec ohm (0) = volt/amp farad (F) coulomb/volt amp sec/volt = second/ohm henry (H) = volt sec/amp = ohm sec
Q
e
H P R
C
=
=
Inductance
L
t> 43 MODELING OF ACTIVE ELECTRICAL ELEMENTS: OPERATIONAL AMPLIFIERS
Operational amplifiers, or simply op~amps, offer a convenient way to build t implement, or realize continuousdata or sdomain transfer functions. In control systems, opMamps are often used to implement the controllers or compensators that evolve from the controlsystem design process, so in this section we illustrate common opamp configurations. An
43 Modeling of Active Electrical Elements: Operational Amplifiers
113
+
Figure 432 Schematic di:lgram of an opamp.
indepth presentation of opamps is beyond the scope of this text. For those interested, many texts are available that are devoted to all aspects of opamp circuit design and applications [8, 9]. Our primary goal here is to show how to implement nrstorder transfer functions with opamps while keeping in mind that higherorder transfer functions are also important. In fact, simple highorder transfer functions can be implemented by connecting firstorder opamp configurations together. Only a representative sample of the multitude of opamp configurations will be di scussed. Some of the practical issues associateu wilh opamps are demonstrated in Chapters 5 and 9.
431 The Ideal OpAmp
When good engineering practice is used, an opamp circuit can be accurately analyzed by considering the opamp to be ideal. The ideal opamp circuit is shown in Fig. 432, and it has the following properties : 1. The voltage between the + and  terminals is zero, that is, e+ = e . This property is cOl11monly called the virtual ground or virtual short. 2. The currents into the + and  input terminals are zero. Thus, the input impedance is infinite.
3. The impedance seen looking into the output terminal is zero. Thus, the output is an ideal voltage source. 4. The inputoutput relationship is eo = A(e+  e ), where the gain A approaches infinity.
The inputoutput relationship for many opamp configurations can be determined by using these principles . An opamp cannot be used as shown in Fig. 432. Rather, linear operation requires the addition of feedback of the output signal to the  input terminal.
432 Sums and Differences
As illustrated in Chapter 3, one of the most fundamental elements in a block diagram or an SPG is the addi.tion or subtraction of signals. When these signals are voltages, opamps provide a simple way to add or subtract signals, as shown in Fig. 433, where all the resistors have (he same value. Using superposition and the ideal properties given in the preceding section, the inputoutput relat10nship in Fig. 433(a) is \10 =  (Vo  Vb ). Thus, the output is the negative Sllm of the input voltages. When a positive sum is desired, the circuit shown in Fig. 433(b) can be used. Here the output is given by eo = ell + eb. Modifying Fig. 433(b) slightly gives the differencing circuit shown in Fig. 433(c), which has an inputoutput relationship of eo = e"  ell.
l + R + (b) R R ell R rY\1Irri + + eo = eb . Using ideal opamp properties.174 Chapter 4. 433 FirstOrder OpAmp Configurations In addition to adding and subtracting signal s.eo R (c) Figure 433 Opamps used to add and subtract signals. 434.'"Nv. Inductors are not commonly used because they tend to be bulkier and more expensive. In the figure . the inputoutput relationship . we will ex plore only those that use the inverting opamp configuration shown in Fig. Z. opamps can be used to implement transfer functions of continuousdata systems. While many alternatives are available. or . Theoretical Foundation and Background Material: Modeling of Dynamic Systems R R + + eo = (e" + eh) (a) R R . (s) and Z2(S) are impedances commonly composed of resistors and capacitors.
L + sC. 1 of  R:d Rl R:!Cz (ConI inued) . it is possible to implement poles and zeros along the negative real axis as well as at the origin in the splane..J\/'y=R 1 /(1 Rz ](.R2C I ) S Y1 = sC I 1~ RI A. i. such a s (4101) where Yl (s) = I ! 2 j (s) and Y2(S ) = 1/ 22(S) are the admittances associated with the circuit impedm1ces.e..(s) . A.g.J\/'y=R 1 Y1 = sC2 1~ 1?2 R I C2 ~ (c) C1 (. i. an integrator A. of the circuit shown in Fig. = CJ Pole at the origin .J\/'yZ2 = R2 Input Element (a) ZI I?I Transfer Function Comments Invclting gain. Using the invetting opamp configuration shown in Fig. The negative gain is usually not an issue because it is simple to add a gain of .1 ) 1 (b) 21 C. transfer function.e. 434 can be written in a number of ways . TABLE 44 Inverting OpAmp Transfer Functions Feedback Element R2 A.J\/'y Z\ =R\ qgY.. Because the invelting opamp configuration has been used. e. (4101) apply conveniently to the different compositions of the circuit impedances. ( . all the transfer functions have negative gains.. a differentiator Z"2 =R 2 (d) A.J\/'y Zero at the origin.1 c with a de gain R 2 :2 s+ ..43 Modeling of Active Electrical Elements: Operational Amplifiers 175 E.{s) + + EII(s) Figure 434 Inverting opamp configuration. if RJ = R')" I.= .R2 " R~ R tC2 Polc at . . The different transfer function forms givel11n Eq. as shown in Table 44.1 to the input and output signal to make the net gain positive. 434 and using resisrors and capacitors as elements to compose 2 \(s) and 0.
435 are Ep(s) R2 (4103) Proportional: £(s) = .e . 435 is just one of many possible implementations of Eq. =R. By superposition.+ KDS S (4~102) where Kp • K D . In Chapters 5 and 9. Input Element (e) RI Comments Pole at the origin and a zero at 1/ R2C2. (f) RI Lfu. It is important to note that Fig. the negative gains of the proportional. Kp KJ + .. a lead or lag 1 Yz = 2 +sCz controller EXAMPLE 431 As an example of opamp realization of transfer functions. this transfer function will be called the PID controUer. the proportional gain can be implemented using line (a).( 8+ C C2 RlCl 1 s+n2 C2 1) Poles at s at s =R C 2 1 2 and a zero =RC I 1 . 435..e. Also. K .e. the integral tenn can be implemented using line (b).Rl Integral: Derivative: The output voltage is (4104) (4105) (4106) Thus.r J YI = JW'vZ2=R 2 Zero at 8 = RICl' i. a PD 1 controller +sC I I (g) R) R2 Lfu. For example. (4102). + ·\C. The design of the controller should be guided by the availability of standard capacitors and resistors.• and KIJ are matched. and the derivative tenn can be implemented using line (c). 433(a). consider the transfer function G(s) ::.176 " Chapter 4. are real constants. the output of G(s) is the sum of the responses due to each term in G(s). the second an integral term. it is common to add components to limit the highfrequency gain of the differentiator and to limit the . and K. it is possible to implement the PID controller with just three opamps. the transfer function of the PID opamp circuit is Eo(s) Rz G(s) ::.i. the design is completed by choosing the values of the resistors and the capacitors of the opamp circuit so that the desired values of K p .• a PI controller JW'vZt=R t ~t= z. and the third a derivative term. (4102) and (4107).. Rl + RiCi S + RdCtl S 1 (4107) By equating Eqs. By making the sum negative. This sum can be implemented by adding an additional input resistance to the circuit shown in Fig.r J YI =++sC) I l .L ~ ~ R2 R. i.z . and derivative term implementations are canceled. since the first term is a proportional gain. £(s) ::.r Lfu. The transfer functions of the components of the op~amp circuit in Fig. integral. Using Table 44... Theoretical Foundation and Background Material: Modeling of Dynamic Systems TABLE 44 (Continued) Feedback Element Transfer Function C. giving the desired result shown in Fig..
Opamps are also used in control systems for NO and O/A converters. One advantage of the implementation shown in Fig.v) ~~ R Figure 435 Implementation of a PID controller. integrator output magnitude. + R R Ejs) E{)(. Heat transfer is related to the heat flow rate q. Because of the complex mathematics associated with these nonlinear systems. which is often referred to as antiwindup protection. the concept of capacitance in a heat transfer problem is related to storage (or discharge) of heat in a body. 441 Elementary Heat Transfer Properties' The two key variables in a thermal process are temperature T and thermal storage or heat stored Q. which has the units of power. . 4·35 is that each of the three constants K p . which has the same units as energy. The capacitance Cis related to the change of the body temperature T with respect to time and the rate of heat flow q: q= ct (4109) IFor more indepth study of this subject. sampling devices. 44 INTRODUCTION TO MODELING OF THERMAL SYSTEMS In this section. and realization of nonlinear elements for system compensation. we introduce thermal and fluid systems.44 Introduction to Modeling of Thermal Systems 177 R. K" and Ko can be adjusted or tuned individually by varying resistor values in its opamp circuits. we only focus on basic and simplified models. That is q= Q (4108) As in the electrical systems. refer to references [17].
or radiation. as shown in Fig. q . Figure 431 Fluidboundary heat convection. k X i i+ ! 7 III Figure 436 Onedirectional heat conduction flow. At the boundary where the fluid and the solid surface meet.T2 is the difference between the temperatures at x = 0 and x = £~ or T/ and T2 • Note in this case. where the thermal capacitance C can be stated as a product of p material density~ c material specific heat. convection. In general this type of heat transfer happens in solid materials due to a temperature difference between two surfaces.178 ~ Chapter 4. So the rate of heat transfer q may be represented in terms of R as q= AT R (4113) Convection: This type of heat transfer occurs between a solid surface and a fluid exposed to it.T (4114) 1j Fluid flow . and volume V: (4110) In a thermal system. and AT = Tl . heat tends to travel from the hot to the cold region. In this case. the heat conduction in other directions is zero. Theoretical Foundation and Background Material: Modeling of Dynamic Systems i q A. In thermal convection. k is the thermal conductivity related to the material used. 436. 437. Also note that 1 1 DIZ=T=kA R (4112) where R is also known as thermal resistance.. Conduction: Thermal conduction describes how an object conducts heat. as shown in Fig. the heat flow is given by q = hAllT = Dot:. . the rate of heat transfer is given by where q is the rate of heat transfer (flow).. The transfer of energy in this case takes place by molecule diffusion and in a direction perpendicular to the object surface. But once the fluid is exposed to the heat. the heat transfer process is by conduction. e. assuming a perfect insulation. there are three different ways that heat is transferred: by conduction. it can be replaced by new fluid. A is the area normal to the direction of heat flow x. Considering onedirectional steady state heat conduction along x.
667 x 10. TABLE 45 Parameter Basic Thermal System Properties and Their Un. Thus q=R AT (4116) Radiation: The rate of heat transfer through radiation between two separate objects is determined by the StephanBoltzmann law.184 J Temperature Energy (Heat Stored) T Q OF (Fahrenheit) Btu calorie Btu/sec QF/(Btu/hr) Heat Flow Rate l] J/sec W Resistance Capacitance R °elW C1KfW J/(kg °C) J/(kg OK) C Btu/OF BturR . Figure 438 A simple heat radiation system with directly opposite ideal radiators. and T/ and T2 are the absolute temperatures of the two bodies.32) x 5/9 PC =oK + 273 IJ = INm 1 Btu = 1055J 1 cal = 4. 438). Note that Eq. A is the area of heat transfer. The tenn hA may be denoted by Do.T f is the difference between the boundary and fluid temperatures. where Do = hA = R 1 (4115) Again. II is the coefficient of convective heat transfer. where q is the rate of heat transfer or heat flow. .ts Symbol Used SI Units °C (Celsius) PK (Kelvin) J (joule) Other Units Conversion Factors °C = (OF . u is the StephanBoltzmann constant and is equal to 5. and IlT = Tb . (4117) applies to directly opposed ideal radiators of equal surface area A that peliectly absorb all the heat without reflection (see Fig. A is the area normal to the heat lluw. (4117) where q is the rate of heat transfer.8 W/m2 °K4. The SI and other measurement units for variables in thermal systems are shown in Table 45. the rate of heat transfer q may be represented in terms of thermal resistance R.44 Introduction to Modeling of Thermal Systems • 179 ~ q..
(4118) and (4119). (4109) and the convective thennal resistance R from Eq. = p = Material density c = Material specific heat k = Material thermal conductivity It = Coefficient of convective heat transfer SOLUTION The rate of heat storage in the solid from Eq.Tf) The energy balance equation for the system dictates qto be the same in Eqs. Find the equations of the heat transfer process for the following: T 1 f. T. + Tf 1 (4120) In Laplace domain. and the initial conditions are assumed to be TI.180 • Chapter 4. (4119). there are five parameters of importancepressure. as shown in Fig. upon introducing lhennal capacitance C from Eq. we derive the equations of fluid and pneumatic systems. we get ReTE. the convection rate of heat transferred from the fluid is q = hA(Tf . Hence. = Solid object temperature. In fluid systems. 2Por a more indepth study of this subject.EXAMPLE 441 A rectangular object is composed of a material that is in contact with fluid on its top side while being perfectly insulated on three other sides. (4113) and substituting the righthand sides ofEq. to be discussed in more detail in Chapter 5.( dJrE) (4118) Also. 439. Figure 439 Heat transfer problem between a fluid and an insulated solid object. Theoretical Foundation and Background Material: Modeling of Dynamic Systems :. assume that the temperature distribution is uniform temperature Length of the object A = Cross sectional area of the object Tt =Top fluid /. . temperature. (4109) is q = peA/. flow mass (and flow rate). 45 INTRODUCTION TO MODELING OF FLUID SYSTEMS 451 Elementary Fluid and Gas System Properties2 In this section. Understanding the models of these systems will later help in appreciating the models of hydraulic and pneumatic actuators. the transfer function of the system is written as 1J(s) RCs+ 1 (4121) where the RC = T is also known as the time constant of the system. (4118) into Eq. The significance of this term is discussed earlier in Chapter 2. density.(t = 0) = 1"'/(t = 0) = O. = Tj(s} T. refer to references [17].
and V is the container volume. and q is the net fluid flow rate (volume Row rate of the ingoing fluid qi minus volume flow rate of the outgoing fluid qo). (4126) CapacitanceIncompressible Fluids: Similar to the electrical capacitance. To understand these concepts better.o of the fluid flow rate q to the rate of pressure P: (4127) Or q = C? (4. can be modeled by passive components including resistance. we must look at the fluid continuity equation or the law of conservation of mass.128) . and flow volume (and volume rate).qo where rn is the net mass flow. M n is the mass of the control volume (or for simplicity " the container" fluid) . Hence. For the conu'ol volume shown in Fig. Note (4125) which is also known as the conservation of volume for the fluid . The fluid capacitance C is the rati. In case of incompressible fluids. The conservation of mass states: .45 Introduction to Modeling of Fluid Systems 181 nl = q". fluid capacitance relates to how energy can be stored in a fluid system.(pV) dt d (4123) dm . = pV dt + Vi> (4124) where In is the net mass flow rate. p is constant. qlll = nz is the mass flow rate. capacitance. the fluid volume remains constant. we have qlll = pq In =/ pqdt (4122) q = qi . = pq Figure 1lIlO Control volume and the net mass flow rate. and inductance. 440 and the net mass flow rate. just like electrical systems.= dt dm d pq = (Mev) dt = . p is fluid density. Incompressible fluid systems. For an incompressible fluid.
For a pneumatic system. is P= A=. and it defines tbe rate of change of gas stored in a control volume. . Capacitance expresses the rate of change of the fluid mass with respect to pressure. it is customary to use the mass flow rate qm as opposed to volume flow rate q in the equations involving pneumatic systems. That is c= dm/ elt = dm dP / dt dP (4131 ) In general . noting that q = V.= phg pV phgA (4129) A Figure 441 Incompressible fluid flow into an opentop cylindrical container. c ==pgh pg V A (4130) For the general case. respectively. e temperatures of the fluid entering and flowing out of the container are almost the same. pressure. In this case.1 ". what is happening in Fig. (4133) a=  P aT 1 (a PIl1 ) P"f . 4·41 . the fluid density p is nonlinear and may depend on temperature and pressure. the law of conservation of volume does not apply because the volume of a gas varies with pressure or other external effects. only Ihe conservation of mass applies. In this case. may be linearized using the first· order Taylor series relating Pili to P and T: Pill = P + ( DPIII) ap (p P. CapacitancePneumatic Systems: As in the previous case.ef) (4132) I ":1. assuming a springloaded piston of area A traveling inside a rigid cylindrical container. as the fluid flows into the control volume.·" .. (4124) may be rewritten as dm V. (T  T. however. This nonlinear dependency.V) C (4136) where it = AX. As a result.r. P= I (q . Pnl(P.. and temperature. as shown in Fig. 440. In most cases of interest. as .182 Chapter 4. accumulators are fluid capacitors. . J where P. capacitance relates to how energy can be stored in the system. 442. known as the equation of state.'e are constant reference values of density.= pq = VjJ ~ q = P = cp ell f3 EXAMPLE 452 Tn practice. the fluid mass will change. P ni' and T. Theoretical Foundation and Background Material: Modeling of Dynamic Systems EXAMPLE 451 The pressure in the tank shown in Fig. respectively. As a result. he Further. (4135) . which may be modeled as a springloaded piston system. if the container of volume Vis a rigid object.T"f (4134) are the bulk modulus and the thermal expansion coefficient. so does the pressure. the pressure rate is shown as . which is filled to height 11. using Eq. Eq.T'4 Pre! ) + (aPII/) aT > .T ). (4133). In this case.
(4137) may be restated as C==V~=V As a result. and Rg is the gas constant. Notice in this case four parameters p. (4140). . Or TI == T2 PIVt =P2 V2 In this case.. which is a general process for all fluids relating the pressure.4·5 Introduction to Modeling of Fluid Systems . Using a polytropic process. shown in Fig. (4140). n = 0 in the capacitance relation Eq. the gas temperatures at any two given instants are the same. V. the perfect gas law states: PV = mRgT (4138) where V is the volume of a gas with absolute pressure P and mass m. the general gas law may also be defined by PI VI P2 V2 r. For a constant volume container. V C= nRgT As a side note. (4139) where n is called the polytropic exponent and can vary from 0 to relation Eq. if in a polytropic process the mass m is constant. For a constant volume or an isovolumetric process. 440 with respect to pressure. and mass. to solve one.~ 183 q Paml Figure 4·42 A springloaded piston system. n = 1 in the capacitance relation Eq. volume. (4140). and given a process from state 1 to state 2. For a perfect gas under normal temperatures and pressures. As a result. T. T is the absolute temperature of the gas. n == ex: in the capacitance relation Eq. the general gas capacitance relation Eq. the relation becomes VI Pt = Tl P2 Tz In this case. and m are mathematically related. the other three must be somehow related.=~ (4142) For a constant temperature or an isothermal process. we have: P V)n P (m = P (1)" = comito 00. (4138) and knowing m = pV. which depends on the type of gas used. the capacitance dp (p. (4131) becomes dm dp C==V(4~137) dP dP where the container volume V is a constant.) Pnp(nI) dP =V~ p (4141) IlP Or~ using Eq. For a constant pressure or an isobaric process. PI =Pz Vl V2 (4143) (4144) TI = T2 = Vz (4145) In this case.
where k is the ratio of specific heats: k = ('" C.154) where q is the volume flow rate. assuming a frictionless pipe wi th a uniform fluid flow moving at the speed v.149) (4.150) So (PJ . i ~ not discl1ssed here. but it can also occur where an external force (e. is the gas specific heat at constant volume..md A is the crosssectional area of the pipe.152) (4151) where /1P = PI . ResistanceIncompressible Fluids: As in the electrical systems.4 (for air). For the system shown in Fig.P q". the relation becomes P J vf = P2 v i (4146) In this case. k = 1.P2) = lJj where L = p€ A is known as the fluid inductance.g. assuming a laminar fl ow. an ex ternal force F is applied.148) q (4. (4147) where cp is the specific heat of the gas at constant pressure and c. The concept of inductance is rarely discussed in the case of compressible flll id~ and ga$e~ and .184 Chapte r 4. 444. therefore. R Figure 444 Flow of an incompressible fluid through a pipe and a tluid resistor R... Depending on the type of flow (i. 443. (4140). From Newton's second law. In thc case shown in Fig. InductanceIncompressible Fluids: Fluid inductance is also referred to as fluid inertance in relation to the inertia of a moving fluid inside a passage (li ne or a pipe).P = Rqm = Rpq (4. Inertance occurs mainly in long lines.153) R=  6.Pl is the pressure drop .'vi'" = pM. Th eoretica l Foundation and Background Material: Modeling of Dynamic Systems . F = A /1P = . the force resisting the fluid passing through a passage li ke a pipe is (4. For a reversible adiabatic or an isentropic process.e. (4. fluid resistors dissipate energy.. caused by a pump) causes a significant change in the flow rate . in order to accelerate the fluid. Table 46 shows R for va rious passage cross sections.P2) Av  (4." b. . laminar or turbulent) the fl uid resistance relationship can be linear or nonl inear and re lates the pressure drop to the mass flow rate ql/l' For a laminar flow. we define b.P But = (Pl V . n = k in the capacitance relation Eq. In pneumatic systems. F q 4""'):_")+ P2 q Figure 443 Fluid forced through a frictionless pipe of length e.
4 5 Introduction to Modeling of Fluid Systems M .Lf (1 Rectangular cross section: Approximation + h/w)2 R = 12. the conservation of volume also applies. most useful. w4 R=wh:~ 8J. ~. d = diameter penmeter R= 32t..Ll Annular cross section ndodr do (1 .g.Lf wits w/Il = small R= 8. (4 153) is rewritten as M where Rr is the turbulent resistance and n is a power varying depending on the boundary usede. General case = length. the pressure drop relation Eq. di Annular cross section~ Approximation R =_12_IL_R 7rdo dr do/di = small When the flow becomes turbulent. water or any incompressible fluid (Le .. Because the fluid density p is a constant. EXAMPLE 453 For the liquidlevel system shown in Fig. SOLUTION The conservation of mass suggests dm dt = ~ = pqi f)qa d(pV) (4156) where pqi and pq() are the mass flow rate in and out of the valve.~ 185 TABLE 46 Equations of Resistance Rfor Laminar Flows PI Fluid resistance 0ql/l. n = 7/4 for a long pipe and.• fluid density p A OneTank LiquidLevel is constant) enters the tank from the top and exits through the valve in the bottom. and output~ h.P2 = Laminar resistance: R Jl: Fluid viscosity w ::::. width. P2 Symbols used Fluid volume Bow rate: q Pressure drop: AP P12 = Pl . Find the system equation for the input. h = height.. 11 = 2 for a flow through an orifice or a valve. qi. I. which suggests the time rate of . The valve resistance is R. 445. The fluid height in the tank is h and is variable. respectively. The volume flow System rate at the valve inlet and the volume flow rate at the valve outlet are qi and qoo respectively.tR Ad~ d" = hydraulic diameter = 7rd4 ~A Circular cross section Square cross section Rectangular cross section R = 128Jl!' R= 32J.~) = inner diameter = outer diameter.l1.
change of the fluid volume inside the tank is equal to the difference of incoming and outgoing flow rates. =A/g is the capacitance and p = R is the resistance. as shown in Fig. which is variable.p Figure 445 A singletank liquidlevel system. it is not difficult to see pqi ..p is the pressure drop across the valve. SOLUTION Using the same approach as in Example 453. CTt= pqj . we get Pl = Palm P2 = Patm + pgh (4162) where PJ is in the pressure at the valve inlet and pz is the outlet pressure.XR~_!(> h A. Find the differential System equations.. After combining Egs. A is the tank crosssectional area. the fluid valve resistance R.P2 RI = pqi  (Patm + pghd . is defined as 8. 446. (4154). and Iz is the fluid inside the tank height and is a variable. Theoretical Foundation and Background Material: Modeling of Dynamic Systems T 1'. (4157) through (4162).186 Chapter 4.P% dp (4160) Using relation Eq." • EXAMPLE 454 Consider a double tank system.pqI = pqi  PI . assuming a laminar flow. we get the system equation: dh pgh (4163) pA dt =pqi R Or RCh+h=qi g . In this case. R (4164) where C t" = Re. (4158). with hJ and h2 representing the two tank A TwoTank LiquidLevel heights and Rl and R2 representing the two valve resistances t respectively.(Palm + pgh2) RI (4165) . system time constant is . As a result..p pqa =R"= PI  R P2 (4161) where tJ. Relating the pressure to fluid height II. and PClJm is the atmospheric pressure.______L. Recall from Eq. (4131) c= Hence elm/elt dP/dt (4158) dm = Cdp dt dt (4159) Or from Eq.
the valve is modeled as an orifice.I + Pi P P=P~ + (4171) q".:.. the equations of system llre + pghJ) RJ (Pm". We use the next example to better illustrate these concepts.P = Pi . as shown in Fig. ljlll Pi X R ({II) ~ p Figure 447 Air flow through a pipe with an orifice..45 Introduction to Modeling of Fluid Systems .. P is outlet pressure. 1n this case. 447.:. .. the mass flow rate is {flit. For a gas following the perfect gas law Eg . + pgh2) (Parm + pgh2) R2 PIt (4166) (4167) (4168) ResistancePneumatic Systems: The resi~tance for pnellmatic systems is a bit more complicated.::. we have b. is the same. 187  T 1 and A1·p h2 T A 2. by the virtue of continuity.:. pql . which is obtained experimentally.P RT = ~(4170) q".P RL = (4169) Qlll where Pi is the inlet pressure. Not considering the theoretical details.+ X ~PV'P Figure 448 Ga~ flow into a rigid container..I X Figure 446 Twotank liquidlevel system. .. i~ related to the outlet pressure P.p R2 RI ql ~ X . Pi = p . R qlll Pi.. for a lami nar flow if b.. In this case. shown in Fig. we get b. q is the volumetric flow rate.P is small... 448. inlet pressure is Pi. That is.. RJ R2 (Palm Thus. EXAMPLE 455 Consider air passing through a valve lind entering a rigid container system. For a turbulent flow. and RL is the equivalent resistance.. it is customary to think of the pressures in both sides of the valve as a constant pressure Ps (or steady state pressure) plus a variation.P2 1'2 . and the pressure inside the container (or the valve outlet pressure) is P.PQ2 = PI . (4138). where R]' is the turbulent resistance. Note that the mass flow rare 1/1/1 on bOtll sides of the valve.1'3 . the flow through a valve or an orifice of crosssectional areaA. p.
v (4186) . (4169) we have (4176) Pi.P Pi . (4176) and (4173). and (4173).. we get (4182) Vdp =_l_dp dt RgT dt But from Eqs. (4186) and (4179) become the same. the law of conservation of mass dictates that in the container dt dt where Pi is fluid density before reaching the valve. (4~169). the transfer function of the system is written as (41 SO} P(s) 1 Pi(S)::::: RCs+ 1 (4181) where the RC = 1: is also known as the time constant of the system.188 '" Chapter 4. we get dp Pi . we have dp V dt Thus. where temperature is constant. and using R = RL for simplicity. we have = = dpV=dm R T dt dt 8 From Eq. (4182). we have dm dt = d(pV) = vdp (4173) dm Recall from Eq.P (4184) V dp Rg T dt =P. and taking a derivative of Eq. Using an isothermal process.RAfter sUbstituting (4185) c= RaT the differential Eqs. 448 with constant volume V. (4183) (4172).P qm= RL (4177) Thus from Eqs.P c=dt R or (4178) Cdp +!!. and the initial conditions are assumed to be p(t 0) p(t = 0) = O. =PiRL Pi . At the inlet (left side of the valve). Theoretical Foundation and Background Material: Modeling of Dynamic Systems For the rigid container in Fig. (4138) with respect to time. (4131) dt = Piq = qm But dm dt Hence = dp/dt dm/dp (4175) dm= Cdp dt dt But from Eq. (4172) and Eq.= Pi dt R R The differential equation can be rearranged as (4179) RCj>+ P= Pi In Laplace domain. The significance of this term is discussed earlier in Chapter 2.
46 Sensors and Encoders in Control Systems . 45 I shows the equivalent circuit representation of a potentiometer. (4 139). 450 shows a linear potentiometer that also contains a builtin operational amplifier. either linear or rotational. with limited or unlimited rotational motion. is proportional to the input displacement. Fig.. which is measured across the variable terminal and ground.184J Volume Flow Rate Mass Flow Rate Resistance (hydraulic) Resistance (pneumatic) Capacitance (hydraulic) Capacitance (pneumatic) Time Constant q m3/sec qm R kg/sec Nsec/m5 lb/sec Ibr sec/in5 Ibr hr/(ft2 Ibm} R C C 't'=RC sec/m 2 rosIN m2 sec rr insllb Using the polytropic process defined in Eq. When a voltage is applied across the fixed terminals of the potentiometer~ the output voltage. it is easy to see Rep +p == Pi M (4187) where v c=nRgT (4188) The S1 and other measurement units for variables in electrical systems are tabulated in Table 47. 189 TABLE 47 Basic Fluid and Pneumatic System Properties and Their Units Parameter Symbol Used SI Units Other Units of (Fahrenheit) OR (Rankin) Btu calorie ft 3/sec in 3/sec Conversion Factors °C = (OP_ 32) x 5/9 °C =0 K+273 11 = INm 1 Btu = 1055J 1 cal Temperature Energy (Heat Stored) T Q °C (Celsius) QK(Kelvin) I (joule) = 4. 461 Potentiometer A potentiometer is an electromechanical transducer that converts mechanical energy into electrical energy. when a voltage is applied across the fixed terminals. the principle of operation and applications of some of the sensors and encoders that are commonly used in control systems are described. For precision control. Because the voltage across the variable terminal and reference is proportional to the shaft displacement of the potentiometer. the conductive plastic potentiometer is preferable~ because it has infinite resolution~ long rotational life. Fig.. The potentiometers are commonly made with wirewound or conductive plastic resistance material. good output smoothness.. In this section. . linear or rotary... . and Fig. the device can be used to indicate the absolute position of a system or the . The input to the device is in the fonn of a mechanical displacement. Rotary potentiometers are available commercially in singlerevolution or multirevolution fonn. either linearly or according to some nonlinear relation.46 SENSORS AND ENCODERS IN CONTROL SYSTEMS Sensors and encoders are important components used to monitor the performance and for feedback in control systems. and low static noise. 449 shows a cutaway view of a rotary potentiometer.
.. Inc .. .). Fig. . Theoretical Foundation and Background Material: Modeling of Dynamic Systems Precious metal coil contact Dual gold plated _ _ __ sl ipring contacts ~ 51 iderblock with dua l guides Resistance element : r .Bronze chromated brass bush ing position guides terminals Figure 449 Tenturn rotary potentiometer In. Then (4189) I'ixed terminals ~ .190 " Chapter 4.Wl< ewound 0< hVbid ca il Gold plated slipring '. 452(a) shows the arrangement when the housing of the potentiometer is fixed at reference.).4 .<o Variable terminal Figure 451 Electric circuit representation of a potentiometer. (courte~y of Helipot Division of Beckman Figure 450 Linear motion potentiometer with bui ltin operational amplifier (courtesy of Waters Manufacturing.truments.. Rotor molded to shaft =':~ Mechan ical stops t . relative position of two mechanical outputs. the outPllt voltage e(l) will be proportional to the shaft position ec(t) in the case of a rotary motion. Inc.. ' .
positioncontrol system. This arrangement allows the comparison of two remotely located shaft positions. The potentiometers are used in the feedback path to compare the actual load position with the desired reference position. (bj Two pmenliometers used to sense the positions of two shafts. is given by Ks = 27rN E V/ rad (4190) where E is the magnitude of the reference voltage applied to the fixed terminals. Typical waveforms of the signal s in the sys tem when the input e. 452. the error signal is amplified by a de amplifier whose output drives the armature of a permanentmagnet de motor. 4S4(b). These definitions are different from those commonly lIsed in e lectrical engineerin g. where dc simply refers to unidirectional signals and ac indicates alternating signals ....  I Figure 4·52 Potentiometer used as a position indicator. Fig. On the other hand. If there is a discrepancy between the load position and the reference input. where K.>+\0 + e(t) (a) 1 . A more flexible arrangement is obtained by using two potentiometers connected in parallel. Note that the electric signals are all unmodula ted.. 452. As shown in Fig. For an Nturn potentiometer. In dcmotor control systems. 454(a) shows the schematic diagram of a typical dcmotor. In controlsystems terminology. the total displacement of the variable arm is 2nN radians. a de signal usually refers to an unmodulated signal.. The output voltage is taken across the variable terminals of the two potentiometers and is given by (4191) Fig. as shown in Fig.(t) is a step function are shown in Fig. 454 (a). 453 illustrates the block diagram representation of the setups in Fig. potentiometers are often used for position feedback. The proportional constant I<C.e(l) (b) .< is the proportional constant. 4S2(b). an ae signal refers to signals that are modulated by a Inodulation process.46 Sensors and Encoders in Control Systems 191 + OJ1) E . an error signal is generated by the potentiometers tbat will drive the motor in such a way that thi s e rror is mi nimized quickly. (a) el(t)~(l )  Ks ± r:(1) (h) Figure 453 Block diagram representation of potentiometer arrangements in Fig. . .
. (b) Typical waveforms of signals in the control system of part (a).. 454(a)... The frequency of this signal is usually much higher than that of the signal that is being transmitted through the system. positioncontrol system with potentiometers as error sensors. The signal v(t) is referred to as the carrier whose frequency is We.(l) EQI~ 0 ~ ·M o :7 I ~ ~ == (b) " Figure 454 (a) A dcmotor. Typical signals of an ac control system are shown in Fig..:= e. or v{t) = E sin wet (4192) . 455(a) illustrates a control system that serves essentially the same purpose as that of the system in Fig. 455(b). ~ o o 'I eel) L _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ __ _ _ _ _ __ _ 1 0 I~ ~ ==<= =====..Chapter 4... Theoretical Foundation and Background Material: Modeling of Dynamic Systems de motor + e DC AMPLIFIER + Reference inpUl (a) Br(t) I 1 . Fig. the vol tage applied to the error detector is sinusoidal. Control systems with ac signals are usually found in aerospace systems that are more susceptible to noise..192 II. except that ac signals prevail. In this case .
46 Sensors and Encoders in Control Systems Twophase ac motor 193 + e AC AMPLIFIER Reference input (a) (Jr(t) II L _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ o vet) e(r) (b) Figure 455 (a) An ac control system with potentiometers as error detectors. Analytically. or (41 94) For the ee(t) shown in Fig. . e(t) becomes a suppressedcarriermodulated signal. The term suppressedcarrier modulation stems from the fact that when a signal ee(t) is modulated by a carrier signal v(t) according to Eg. (b) Typical waveforms of signals in the control system of part (a). the output of the error signal is given by e(/) = K s8e (t)v(t) (4193) where 8e (t) is the difference between the input displacement and the load displacement. This reversal in phase causes the ac motor to reverse in direction according to the desired sense of correction of the error signal 8e (t). 455(b). A reversal in phase of e(t) occurs whenever the signal crosses the zeromagnitude axis.
In control systems. For instance. normally. the accuracy of the tachometer is highly critical. or an ac device to a dc device through a demodulator. so that the displacement of the load will be of the same form as the dc signal before modulation. or the error. The difference between the two signals. It should be pointed out that a control system need not contain all dc or all ac components. . the resultant signal e(t) no longer contains the original carrier frequency e(t) is also a sinusoid given by illustrate this. Fig. that is. This is clearly seen from the waveforms of Fig. In a positioncontrol system. speed control. e(l) no longer contains the carrier frequency We or the signal frequency Ws but has only the two sidebands We + Ws and Wc . velocity feedback is often used to improve the stability or the damping of the closedloop system. is amplified and used to drive the motor so that the velocity will eventually reach the desired value. with the output voltage proportional to the magnitude of the angular velocity of the input shaft. In this type of application. as the accuracy of the speed control depends on it. velocity feedback. DC tachometers are used in control systems in many ways. they can be used as velocity indicators to provide shaftspeed readout. which represents the desired velocity to be achieved. Ws « We.ws . we get (4196) Therefore. Using familiar trigonometric relations and substituting Eqs. 455(b). Theoretical Foundation and Background Material: Modeling of Dynamic Systems (4. When the modulated signaJ is transmitted through the system. 455(a) may be replaced by an ac amplifier that is preceded by a modulator and followed by a demodulator. or stabilization. the output voltage is a dc signal. 456 is a block diagram of a typical velocitycontrol system in which the tachometer output is compared with the reference voltage. (4193). the motor acts as a demodulator.194 Chapter 4. most of the tachometers used are of the dc variety. 462 Tachometers Tachometers are electromechanical devices that convert mechanical energy into electrical energy. the dc amplifier of the system in Fig. Fig. 457 shows the block diagram of such an de motor CONTROLLER POWER AMPLIF[ER T Tachometer Figure 456 Velocitycontra! system with tachometer feedback . It is quite common to couple a dc component to an ac component through a modulator. let us assume that O W(" To (4195) where.192) and (4195) into Eq. The device works essentially as a voltage generator. (4.193).
The third and most traditional use of a dc tachometer is in providing the visual speed readout of a rotating shaft. respectively. and the accuracy of the tachometer is not so critical. application. the incremental encoder's simplicity in construction. In this case. For the most part. The trunsfer function of a tachometer is obtruned by taking the Laplace transform on both sides of Eq. provide a pulse for each increment of resolution but do not make distinctions between the increments. (4197). e(t). the rotor displacement in radians.46 Sensors and Encoders in Control Systems Il_. In the simplest terms. and K the tachometer constant in V/rad/sec. Mathematical Modeling of Tachometers The dynamics of the tachometer can be represented by the equation (4197) where eM) is the output voltage. and versatility have made it by far one of the most popular encoders in control systems. The encoders that output a digital signal are known as absolute encoders. Tachome ters used in this capacity are generally connected directly to a voltmeter calibrated in revolutions per minute (rpm). The result is (4198) where EM) and 0>(s) are the Laplace transforms of e. Tn practice. the choice of which type of encoder to lise depends on economics and control objectives. w(t). the tachometer feedback forms an inner loop to improve the damping characteristics of the system. absolute encoders provide as output a distinct digital code indicative of each particular least significant increment of resolution. The value of K l is usually given as a catalog " parameter in volts per 1000 rpm (Vlkrpm). 463 Incremental Encoder Incremental encoders are frequently found in modem control systems for converting linear or rotary di splacement into digitally coded or pulse signals. on the other hand. ease of application.(t) and B(t). Incremental encoders. However.!M~otor 195 PREAMP POWER AMPLIFIER 1 Tachometer Figure 457 Po "it ioncontrol system with tachometer feedback . the rotor velocity in rad/sec. Incremental encoders are available in rotary . the need for absolute encoders has much to do with the concern for data loss during power failure or the applications involving periods of mechanical motion without the readout under power. low cost.
the two signals are said to be in quadrature. Tn this case. Fig. as shown in Fig. 462 shows the singlechannel output and the H~~~~~~111= Sensor source (lamp. 459 show typical rotary and linear incremental encoders. photodiodc) Figure 4· 60 Typical incremental optomechanics. The mask is used to pass or block a beam of light between the light source and the photosensor located behind the mask. Squarewave signals compatible with digital logic are derived by using a linear amplifier followed by a comparator. and a sensor. a stationary mask.) . 458 and Fig. A dualchannel encoder with two sets of output pulses is necessary for direction sensing and other control functions. 461 (b). . Inc. depending on the resolution required.196 Chapter 4. pulses are produced for both directions of shaft rotation . Rotating disk Statiomu'Y mask photolransist()l'. The disk has alternate opaque and transparent sectors . Fig. For encoders with relatively low resolution. A typical rotary incremental encoder has four basic parts: a light source. 461(a) shows a typical rectangular output waveform of a singlechannel incremental encoder. The signals uniquely define Oto1 and 1to0 logic transitions with respect to the direction of rotation of the encoder disk so that a directionsending logic circuit can be constructed to decode the signals. The waveforms of the sensor outputs are generally triangular or sinusoidal. a multipleslit mask is often used to maximize reception of the shutter light. When the phase of the twooutput pulse train is 90° apart electrically. and linear forms. Any pair of these sectors represents an incremental period.:cll.). LED) (photovoltaic <. Figure 459 Linear incremental encoder (comtesy of DISC Instruments. For fineresolution encoders (up to thousands of increments per evolution). 460. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Figure 458 Rotary incremental c·ncoder (courtesy of DISC Instruments. the mask is not necessary. a rotary disk. Inc . as shown in Fig. Fig.
The following ex.ample illustrates some applications of the incremental encoder in control systems. (b) Typical dualchannel encoder signals in quadrature. Thus. and the encoder has 480 cycles. These zero crossings can be used for position indication. quadrature outputs with sinusoidal waveforms. 463 over one cycle. position control. Let us assume that the encoder shaft is coupled directly to the rotor shaft of a motor that directly drives the printwheel of an electronic typewriter or word processor. Note that the two encoder signals generate 4 zero crossings per cycle. The sinusoidal signals from the incremental encoder can be used for fine position control in feedback control systems. that is.Qua~rature I 0 AJ I I L J""' L I (b) Figure 4~61 (a) 'Jypical rectangular output waveform of a singlechannel encoder device (bidirectional). this correspOl1(l~ to 1920196 = 20 zero crossings per character. . EXAMPLE 46 Consider an incremental encoder that generates two sinusoidal signals in quadrature as the encoder disk rotates. The output signals of the two channels are shown in Fig.4~6 Sensors and Encoders in Control Systems'" 197 CW shaft rotation CCW shaft rotation LJ (a) /+.Period +/ BLLrtU I 90 I+. (a) (b) Figure 462 (a) Typical sinusoidal output waveform of a singlechannel encoder device. (b) Typical dualchannel encoder signals in quadrature (bidirectional). The printwheel has 96 character positions on its periphery. or speed measurements in control systems. For the 96character printwheel. . there are 480 x 4 : : : : 1920 zero crossings per revolution. there are 20 zero crossings between two adjacent characters.
The advancements made in power electronics have made brushless dc motors quite popular in highperformance control systems. with the development of the rareearth magnet. by using the slope of the sine wave at position A. Lowlimeconstant properties have opened new applications for de motors in computer peripheral equipment such as tape drives. 500 I I .198 . it is possible to achieve very high torquetovolume PM dc motors at reasonable cost. say.. DC motors. and word processors. Before permanentmagnet technology was fully developed. rev sec 1920 zero crossings/rev = 31. and their characteristics are quite nonlinear. : ~ 47 DC MOTORS IN CONTROL SYSTEMS Directcurrent (dc) motors are one of the most widely used prime movers in the industry today. as well as in the automation and machinetool industries. which makes the analytical task more difficult. the majority of the small servomotors used for control purposes were ac. One way of measuring the velocity of the printwheel is to count the number of pulses generated by an electronic clock that occur between consecutive zero crossings of the encoder outputs. Today. the advances made in brushandcommutator technology have made these wearable parts practically maintenancefree. the clock generates 500. especially for position control. the control system should null the error quickly. because of their brushes and commutators. that is. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Output channel 1 Output channel 2 Position 111 1 cycle to! Figure 463 One cycle of the output signals of a dualchannel incremental encoder.2Srpm The encoder arrangement described can be used for fine position control of the printwheel. ac motors are more difficult to control. In reality. = 1000 zero crossmgs/sec pu ses zero crossmg _ 1000 zero crossings/sec _ 0 52083 / (4199) . the torqueperunit volume or weight of a de motor with a permanentmagnet (PM) field was far from desirable. Years ago. printers. are more expensive. thus achieving a very high torquetoinertia ratio. If the counter records. Furthermore. then.000 pulses/sec . 463 correspond to a character position on the printwheel (the next character position is 20 zero crossings away). and the point corresponds to a stable equilibrium point. disk drives. Advanced manufacturing techniques have also produced dc motors with ironless rotors that have very low inertia. The coarse position control of the system must first drive the printwheel position to within 1 zero crossing on either side of position A. on the other hand. and variableflux dc motors are suitable only for certain types of control applications. Let the zero crossing A of the waveforms in Fig. 500 clock pulses while the encoder rotates from the zero crossing to the next.000 pulses/sec. the shaft speed is 500. Chapter 4. . Let us assume that a 500kHz clock is used.
the back emf. an important type of dc motors in which the commutation is done electronically is called brushless de. or a rareearth compound. The armature conductors are placed in the rotor slots. ~riented magnetic ~ flux f/J T=Fr I r [ . According to the armature construction. and K m . As shown in Fig.). t/J. Alnico. or ozin. the magnetic field of a dc motor can be produced by field windings or pennanent magnets. In addition to the torque developed by the arrangement shown in Fig. The relationship between the back emf and the shaft velocity is (4201) where eb denotes the back emf (volts) and Wm is the shaft velocity (rad/sec) of the motor. IronCore PM DC Motors The rotor and stator configuration of an ironcore PM dc motor is shown in Fig. The relationship among the developed torque. The pennanentmagnet material can be bari urn ferrite. Due to the popularity of PM dc motors in control system applications. and the conductor is located at a distance r from the center of rotation. the PM de motor can be broken down into three types of armature design: ironcore. 471 Basic Operational Principles of DC Motors The dc motor is basically a torque transducer that converts electric energy into mechanical energy.4 7 DC Motors in Control Systems <: 199 M _\ \ I / F~ Uniform. and movingcoil motors. (4~200) and (4201) fonn the basis of the dcmotor operation. tends to oppose the current flow. This voltage.lbft. Eqs. the magnetic flux (in webers). 464. radially. Conventional dc motors have mechanical brushes and commutators. flux </1. when the conductor moves in the magnetic field. This type of de motor . 464. PM dc motors can be classified according to commutation scheme and armature design. The magnetic flux produced by the magnet passes through a laminated rotor structure that contains slots. which is proportional to the shaft velocity. "Conductor carrying current I Center of rotation Fjgure 464 Torque production in a dc motor. a voltage is generated across its tenninals. and current ill is (4200) where Tm is the motor torque (in Nm. 465. itl • the armature current (in amperes). 472 Basic Classifications of PM DC Motors In general. a currentcarrying conductor is established in a magnetic field with flux cP. However. we shall concentrate on this type of motor. The torque developed on the motor shaft is directly proportional to the field flux and the armature current. a proportional constant. surfacewound.
MovingCoil DC Motors Movingcoil motors are designed to have very low moments of inertia and very low armature inductance. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Conduc[Qrs (bonded to rotating amlature iron) Rotating flux path Figure 465 Crosssection view of a permanentmagnet (PM) ironcore de motor. The armature COI](]u(. One end of the cy jinder forms a hub. 468. as shown in Fig.200 ' Chapter 4. the conductor structure is supported by nonmagnetic materialusually epoxy resins or fiberglass. SurfaceWound DC Motors Fig. Figure 466 Crosssection view of a surfacewound permanentmagnet (PM) de molor. high inductance. This is achieved by placing the armature conductors in the air gap between a stationary flux return path and the PM structure. so this type of motor has lower inductance than that of the ironcore structure. The conductors are laid out in the air gap between the rotor and the PM field. A crosssection view of such a motor is shown in Fig. is characterized by relatively high rotor inertia (since the rotating part consists of the armature windings). which is made of laminated disks fastened to the motor shaft. low cost. In this case. and high reliability. the annature has no "cogging" effect. which is attached to the motor shaft. . 466 shows the rotor construction of a surfacewound PM dc motor. 467. Because no slots are Ilsed on the rotor in this design.to form a hollow cylinder.lurs are bonded to the surface of a cylindrical rotor structure. Because all Conductors (bonded together by nonmagnetic materials) Stationary flux return path Figure 467 Crosssection view of a surfacewound permanentmagnet (PM) de motor.
permanentmagnet (PM). brush less de motors can be Llsed when a low moment of inertia is needed. as shown in Fig. As it will be demonstrated here. Depending on the specific application.47 DC Motors in Control Systems Air gap 201 Magnet pole Amlalure (hollowcup. ironcore de motor.shaped conductor array) Core Flux path Figure 468 Crosssection side view of a movingcoil de motor. such as the spindle drive in highperformance disk drives used in computers. The most common configuration of brushless dc motorsespecially for incrementalmotion applicationsis one in which the rotor consists of magnets and "backiron" SUppOIt and whose commutated windings are located external to the rotating parts. 470 to represent a PM de Stator Pemlanentmagnet rotor Figure 469 Crosssection view of a brushless. . and values of less than 100 JLH are common in this type of motor. 473 Mathematical Modeling of PM DC Motors Dc motors are extensively used in control systems. We use the equivalent circuit diagram in Fig. such as the one shown in Fig. In this section we establish the mathematical model for dc motors. Its lowinertia and lowinductance properties make the movingcoil motor one of the best actuator choices for highperfOlmance control systems. Brushless DC Motors Brushless dc motors differ from the previously mentioned dc motors in that they employ electIical (rather than mechanical) commutation of the armature current. 468. the motor inductance is very low. Because the conductors in the movingcoil armature are not in direct contact with iron. its moment of inertia is very low. the mathematical model of a dc motor is linear. Compared to the conventional dc motors. unnecessary elements have been removed from the annature of the movingcoil motor. 469. it is an insideout configuration.
(4202) is written (4203) where Ki is the torque constant in NmJA . 470. IbftJA. 470 are (4204) (4205) (4206) (4207) where h(t) represents a load frictional torque such as Coulomb friction . . For linear analysis. or ozinJA. Eq.it).ousfriction coefficient With reference to the circuit diagrrun of Fig. motor. (4202) Because rP is constant. and a voltage source eh representing the back emf (electromotive force) in the aJmature when the rotor rotates. the causeandeffect equations for the motor circuit in Fig. The motor variables and parameters are defined as follows: i'a(t) = annaturecunent La = armature inductance Rn = armature resistance ea (t) = applied voltage e!J(t) = back emf TL(t) = load torque Kb = backemf constant cf> = magnetic flux in the air gap Will Tm (t) Bm(t) Ki = = motortorque rotor displacement (t) = rotor aJ1gular velocity = torque conRtant J m = rotor inertia Bm = vise. Thus. Theoretical Foundation and Background Material: Modeling of Dynamic Systems 1. we assume that the torque developed by the motor is proportional to the airgap flux and the armature current. the control of the de motor is applied at the armature terminals in the form of the applied voltage eaCt).1/ + i I/J Magnetic flux Figure 470 Model of a separately excited de motor. The armature is modeled as a circuit with resistance R" connected in series with an inductance L". Starting with the control input voltage e.202 Chapter 4.
471.. wm(t). By direct substitution and eliminating all the nonstate variables from Egs. (4204) through (4207) consider that the applied voltage ea(t) is the cause.I s I I I . The transfer function between the motor displacement and the input voltage is obtained from the state diagram as 01/1{s) Ea(s) K.47 DC Motors in Control Systems 203 S I J. in Eg. finally. (4207).0] 1 em(t) 0 J" T. in this case. La t "'m(t)  .. the state equations of the dcmotor system are written in vectormatrix form: dia(t) dt dwm(t) dt dem(t) dt Ra La K. and 8m(t). The SFG diagram of the system is drawn as shown in Fig. Fig. (4208). . The advantage of using the block diagram is that it gives a clear picture of the transfer function K" Figure 472 Block diagram of a dcmotor system. 1m 0 Kb La 0 Em 1m 1 0 0 [() 1+ [1 ] 7. (4205). Eq. and. iaCt) causes the torque Tm(t).(t) 0 (4208) Notice that .(t) + [. Eg.I s I KilL" Figure 471 Signalflow graph diagram of a dcmotor system with nonzero initial conditions. = La1ms3 + (Ra11/l + BmLa)s2 + (K"Ki + RaBIII)s (4209) where h(!) has been set to zero. L" I Jm =l . hU) is treated as a second input in the state equations. the torgue TmCt) causes the angular velocity wm(t) and displacement ell/(t). (4204) through (4207). using Eg.. (4206) defines the back emf. Egs. (4204) considers that diaCt)/dt is the immediate effect due to ea(t). The state variables of the system can be defined as ia(t). in Eq.. 472 shows a blockdiagram representation of the dcmotor system.
. This is expected because. To show the relationship. we convert Eq. the values of Kb and K. and equating Eq. As seen from Eq. Physically. (4~210)."".204 !~ Chapter 4. Now. Pis where Tm(t) is in ftIb and wm(t) is in radlsec.= ... that is. in SI units. (4 205) and (4 206).. substituting Eqs..746Ki 550 (4216) .. for a given motor their values are closely related. we get M M KbWm(t)Tm(t) Tm (t)aJ m(t) '. we wri te the mechanical power developed in the armature as (4210) The mechanical power is also expressed as where. tlzebackem/efJect is equivalent to an "electricJriction. (4~214) In terms of torque and angular velocity.. in 51 units. (4205) and (4206) in Eq."'''. are identical if Kb is represented in V/rad/sec and K. the backemf constant Kb represents an added term to the resistance Rli and the viscousfriction coefficientBm • Therefore. the SFG diagram of Fig.. (4209)... we get from which we get Kb(V/rad/sec) = Ki(NmlA) Thus. Theoretical Foundation and Background Material: Modeling of Dynamic Systems relation between each block of the system. In the British unit system. if ea<t) is a constant input. (4209). is in Nm1A.. we see that. 471 and the block diagram of Fig. the stability oj the system. (4214) to Eq. Relation between Ki and Kb Although functionally the torque constant Ki and backemf constant Kb are two separate parameters. (4210) into horsepower (hp). 472 show that the motor has a "builtin" feedback loop caused by the back emf. the significance of the transfer function @)m(s)/Ea(s) is that the de motor is essentially all illtegrating device between these two variables. that is." which tends to improve the stability of the motor and in general. Tm(t) is in Nm and wm(t) is in tad/sec. the output motor displacement wiU behave as the output of an integrator. Because an s can be factored out of the denominator of Eq. it will increase linearly with time. Using Eq. (4215). Although a dc motor by itself is basically an openloop system. the back emf represents the feedback of a signal that is proportional to the negative of the speed of the motor.
the monitoring point is located some distance from the mixing point. . especially systems with hydraulic. since it takes time for the computer to execute numerical operations. the measured quantity is (4219) b(t) Metering point v ~ )'(1. pure time delays may be encountered in various types of systems. If the rate of flow of the mixed solution is v inches per second and d is the distance between the mixing and the metering points. Fig.1 (b) ROller~ ~Steel plate Figure 473 Systems w ith transportation J ag. Fig. A time delay therefore exists between the mixing point and the place where the change in concentration is detected.d~.~\'(t) ~14. the time lag is given by Td = . (4217) where K" is in V/rad/sec and Ki is in ft\b/A . To assure that a homogeneous solution is measured. the systems considered all have transfer functions that are quotients of polynomial s.48 Systems with Transportation Lags (Time Delays) 205 Thus. Systems with computer control also have time delays. or mechanical transmissions. 473 illustrates systems in which transportation lags or pure time delays are observed. Kb = 550Ki 746 = 1.356K. the output will not begin to respond to an input until after a given time interval. In these systems.) Solution A + Valve IThicknesslneasuring gauge d ~I (a) Solution B __l~Y~I~)(r~)________________R_o_k_ l 'r_~~o~~~______ ~D~''~ ________~.seconds d v (4218) If it is assumed that the concentration of the mixing point is y(l) and that it is reproduced without change Td seconds later at the mon itoring point. 473(a) outlines an arrangement where two different fluids are to be mixed in appropriate proportions. pneumatic. In practice. 48 SYSTEMS WITH TRANSPORTATION LAGS (TIME DELAYS) Thus far.
. the approximations are not valid when the magnitude of Tcfofj is large. we may find that some devices have moderate nonlinear characteristics. . Theoretical Foundation and Background Material: Modeling of Dynamic Systems The Laplace transfonn of Eq. which is given in the following for a two~tenn approximation: e.. we should realize that most components found in physical systems have nonlinear characteristics. (4221). Chapter 4. In practice. that is. or nonlinear properties that would occur if they were driven into certain operating regions. 49 LINEARIZATION OF NONLINEAR SYSTEMS From the discussions given in the preceding sections on basic system modeling. (4222) or (4223) where only three terms of the series are used. 473(b) illustrates the control of thickness of rolled steel plates. A better approximation is to use the Pade approximation [5. 481 Approximation of the TimeDelay Function by Rational Functions Systems that are described inherently by transcendental transfer functions are more difficult to handle.206 . The transfer function between the thickness at the rollers and the measuring point is again given by Eq. The transfer function between b(t) and yet) is (4221) Fig.TdS by a rational function. (4224) contains a zero in the righthalf splane so that the step response of the approximating system may exhibit a small negative undershoot near t = O. the modeling by linearsystem models may give quite accurate analytical results over a relatively wide range of operating conditions. There are many ways of approximating e.6]. The rootlocus technique (Chapter 7) is also more easily applied only to systems with rational transfer functions.TdS ~ 1 . For these devices. Many analytical tools such as the RouthHurwitz criterion (Chapter 2) are restricted to rational transfer functions. Apparently.TdS/2 1 + Td s/2 (4224) The approximation of the transfer function in Eq. One way is to approximate the exponential function by a Maclaurin series. (4219) is (4220) where yes) is the Laplace transform of yet).
That is (4226) Or f(x(t)) =f(xo(t)) + 3 df(~~(t))(x(t) _ xo(t)) + ~ d f~. ret). the linear model may contain timevarying elements.( ) ( »3 + . and a linearization scheme may be used by replacing !(x(t» with the first two terms in Eq.(x(t) .49 Linearization of Nonlinear Systems ~ 207 However.~(t))(x(t) _ xoCt))2+ 2 (4227) 1 d !Cxo(t»( . fis a function of the state vector and the input vector. A function f(x(t» can therefore be represented in a form Ii f(x(t)) = Lc.xoCt))i . Taylor series may be used to expand a nonlinear function f(x(t)) about a reference or operating value xo(t). the series Eq. d3 . f(x(t)) ~ f(xo(t)) + d f<:(t)) (x(t)  xo{t» (4228) = Co + CIAx 492 Linearization Using the State Space Approach Alternatively.x t . there are numerous physical devices that possess strong nonlinear characteristics. steady state pressure in a fluid system. An operating value could be the equilibrium position in a springmassdamper. and so on. and f[x(t). (4227) converges. when a nonlinear system is linearized at an operating point.Xo (t »n til If 8(X) = x(t) . (4227). a fixed voltage in an electrical system. More importantly. In general. .=1 (4225) where the constant ci represents the derivatives ofj{x(t)) with respect to x(t) and evaluated at the operating point xo(t). That is. an n x 1 function vector. For these devices a linearized model is valid only for limited range of operation and often only at the operating point at which the linearization is carried out.. the p x 1 input vector. let us represent a nonlinear system by the following vectormatrix state equations: dx(t) dt' = f[x(t).91 Linearization Using Taylor Series: Classical Representation In general. 1 d" !(xo(t»)( ( ) I d x t .Xo t 6 t· + n. . y 4. r(t)] (4229) where x(t) represents the n x 1 state vector.xo(t) is small.. r (t)].
.J [). which corresponds to the nominal input ro(l) and some fixed initial states.::::: 1. All the terms of the Taylor series of order higher than the first are discarded.. Il.ro . A linearization process that depends on expanding the nonlinear state equations into a Taylor series about a nominal operating point or trajectory is now described.208 Chapter 4. Let the nominal operating trajectory be denoted by xo(t). it is desirable to perform a linearization whenever the situation justifies it. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Being able to represent a nonlinear andlor timevarying system by state equations is a distinct advantage of the statevariable approach over the transferfunction method.ft(x.ro  rOj ) (4232) where i.r)1 ~ 8.r)1 L. Expanding the nonlinear state equation ofEq. (4237) may be written in vectormatrix form: ax = A*ax+B*ar (4238) . J xJ ~l ~ (rj x(J. .ro} + ~8. Let axi = Xi XOi and (4233) (4234) Then LlXi Since = Xi  XOi Eq. As a simple example. J ~I (Xj xu.XOj ) + ~afi(x. (4229) into a Taylor series about x(t) = xo{t) and neglecting all the higherorder tenns yields . and the linear approximation of the nonlinear state equations at the nominal point results. (4232) is written (4237) Eq. the following nonlinear state equations are given: dxl (t) ~ = XI(/) +x~(t) 1Gt (4230) dt2(t) = XI (1) + r(t) (4231) Because nonlinear systems are usually difficult to analyze and design. 2. Xi(t) = fi (xo. since the latter is strictly defined only for linear timeinvariant systems. .
(b) Freebody diagram of mass m. For the mass 111. respectively. as shown in Fig.v are the external forces applied to mass m." and F'. Fig.t = max LF.e..ill sin B) ay (4243) where 1and Jare the unit vectors along x and y directions.v = a= (4241) (4242) may where F.. ·2 ~ y \ x \ m mg (a) (b) Figure 474 (a) A springsupported pendulum. as shown in . acceleration vector is (i8Sine102 sine)t+ (l8COSBtilsine)J ax = ( lesine . y) representation.. (4 244) w . 1 EXAMPLE 491 Find the equation of motion of a pendulum with a mass m and a massless rod of length I. Using the rectangular coordinate frame (x. Acceleration of mass m is a vector with tangential and centripetal components.. mass and the rod. = ( . SOLUTION Assume the mass is moving in the positive direction as defined by angle e. respectively. The first step is to draw the freebody diagram of the components of the system. i. Note that eis measured from the x axis in the counterclockwise direction. .4~9 Linearization of Nonlinear Systems·. the equations of motion are LF. 474.ee cos 8  if) sine.209 where 8fl aXI B/1 8X2 8fl OXIl A*= 012 aXt al2 8X2 al2 BXn Bjn aXn (4~239) BIn OXl {jln {)X2 8fI 8fl 8fl 8r t B* = ar2 8f2 8rp aI2 orI ai2 Brp 8r2 ajn 8rz (4240) 81n ar) oln Drp The following examples serve to illustrate the linearization procedure just described. As a result. 474(b). and ax and ay are the components of acceleration of mass m in x and y.
(4253) into (4173) with r(t) ::::: 0. SOLUTION The freebody diagram for the moment equation is shown in Fig. we have LFx = Frcose+mg L F'. (4185) by (. using static equilibrium position e. 474. XOI = O. for small motions the linearization of the system implies df) = e ~ sin e. since there is no input (or extemal excitations) in this case. we define Xl = 0 andx2 = iJ as state variables. Eq. After rearranging. Hence. rederive the differential equation using the moment equation. Alternatively in the state space form. and as a result the state space representation of Eq. Notice that the last two equations are linear and are valid only for small signals. we get (4254) (4255) where Ax} (t) and at2(t) denote nominal values of Xt(t) and X2(t). then a = K. =X2 X2 = I smXt g • (4253) Substituting Eq. (4252) becomes Xl . respectively. Chapter 4. 474. the linear representation of the system is . (4241) and (4242) may therefore be rewritten as Frcosf) + nzg = m( f8sin(J  (4248) (4249) Fr sine = m(f9cosO mg sinO Premultiplying Eq.:::. 0 as the operating point. Applying the moment equation about the fixed point O. (4186) by (cos 8) and adding the two. g e+ i(J = O.210 ~. If XOl is chosen to be at the origin of the nonlinearity. (4255) becomes AX2(t) = KAxI (1) (4258) Switching back to classical representation. we get = mEa (4250) (4251) (4252) where Or (sin 2e + cos 2e = 1). (4250) is rewritten a~ mf9+mgsinf) = 0 9 + ~ sine = 0 f.v = Fr sin e eil sinO) it:f sine) (4246) (4247) Eqs. Eq.~ EXAMPLE 492 For the pendulum shown in Fig. In vectorwmatrix form. In brief. (4257) It is of interest to check the significance of the linearization. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Considering the external forces applied to mass. these linearized state equations are written as (4256) where a = E. = constant I. we get e+ KfJ = 0 (4259) ~. (4260) ..sine) and Eq.
Du(t) = Xt (t) Now evaluating the coefficients of Eq. Integrating both sides of Eq. Fig. X02(t)]. (4237).Mg .49 Linearization of Nonlinear Systems· 211 Rearranging the equation in the standard inputoutput differential equation form.~1 (t)] = [0 AX2(1) 0 0 2] [ru:t (t)] + [ ax2(t) Pet) 0 ]IlU(C) I. For small motions..dt2 yet) lf2y(t) (4276) (4277) e(t) = Ri(t) + LTt diet) . the linearized equations are [ a. me2(j + mgt sin 0 = 0 or (4261) (4·262) (j + l sinO =0 which is the same result obtained previously. (4266) and (4~267) are to be linearized is described by XOt (t) = t + 1 X02 ( t) =1 8/2(t)=U(I) (4. which is the solution to the equations with initial conditions x.(O) = X2(0) = 1 and input u(t) O. we get 811(1)=0 ()XI ( ) 1 8fl(t)=_2_ !~ 3 ilx'J. linearization of a nonlinear system often results in a linear timevarying system. 4~ 75 shows the diagram of a magneticbalisuspension system. The differential equations of the system are M . As mentioned earlier./) = where •• 2 0 (4264) (4265) EXAM PLE 493 In Example 491.:q (t) (4267) These equations are to be linearized about the nominal trajectory [XOt (I).t (4275) which is 11 set of linear state equations with timevarying coefficients. we have = . (t) Xi(t) aXI () t 3 () X"02 1 \ (4272) Eq. sine ~ e The linearized differential equation is (4263) () + w.271 ) 8/2(1) . the nominal trajectory about which Eqs.. (4237) gives Atl (t)::= d. (4266) gives Xl (4268) (4269) (4~270) (1) = t + I Thctcfore.\'2(t) = X2(0) = 1 Then Eq. Consider the following nonlinear system: 1 (4266) ·i:dt) = . (4270) and (4271) into Eqs. (4273) and (4274)..t2(t) 2 ~2(t) (4273) (4274) = llo(t)ati (I) + XOt (I)Au(t) By substituting Eqs. the linearized system turns out to be timeinvariant. (4~267) with respect to t.2() ''\:2 t X2(t) = u(t). as in Example 491. The objective of the system is to control the position of the steel ball by adjusting the current in the electromagnet through the input voltage eel}.
'lit M X I (I) dx3(1) ell = .=g . The state dXl (t) dt (4278) (4279) dX2(1) I x~(t) . X2(r) = dy(t)/dt.fOI = constant. (4282) into Eq.Ielt (4283) Thw.. (4276) eel) = Ri(l) d'(l) + L .~x(t) + ~ e(t) L .' L (4280) Let us linearize the system about the equilibrium point YO( I) = .212 Chapter 4. Theoretical Foundation and Background Material: Modeling of Dynamic Systems + R L e T y Electromagnet 1 Mg Steel Ball Figure 475 Magneticballsuspension system.. io (t) = XQ3(t) = J MgxOl (4284) . where e(t) = input voltage !J(t) = ball position '. = X2(t) and X3(t) = i(t).. (428 1) (4282) X02(t) = dXol (l) = dt 2 0 d yo(t) = 0 dt 2 The nominal value of i(t) is determined by substituting Eq.'(1) = wiIllJillg currelll R L winding resistance winding inductance mass of ball gravitational acceleration M g Let us define the state variables as XI(/) equations of the system are = yet ). Then .
(4287) with Eq. ~ dfk(t} = vet) (4288) d y(t) dt2 2 +!!. (4288) by K. The . (4291) clearly shows the analogies among the mechanical and electrical component~. (467) as di{t) L(jf = ec(t} .t).= . (4288) with Eq.Ri(l) and using the current relation Eq. with the coefficient matrices A" and 8* evaluated as 0 X03. to assign v(t). 49(a). in Example 411. (4~41). in terms of displacement y(t). (411) and (441) are analogous to a series RLC electric network shown in Example 421. Yl and Y2. we have Force on mass: M .Bv(t) elv(t) dt + f(t) (4287) Velocity of spring: K dt The final equadon of motion Eq. (4290) and Eq.4. the force acting on the spring. 213 The linearized state equation is expressed in the form ofEq.10 Analogies c. (4·287) by M and multiplying Eq.fdt) . (411).' fi. Hence. the velocity... two displacements. with this analogy. and the viscousfriction coefficient B is analogous to resistance R.2 01 MX~I o 1[ = g '"01 0 (4285) [ o L 0 B' ~ [i] (4286) 410 ANALOGIES Comparing Eqs. it is not difficult to see that the mechanical systems in Eqs. after rewriting Eq. consider the system shown in Fig. As a result. 2 0 0 2X03 A* = M>. (466): = t C dec(t) I+() + eel) (4291) tit the comparison ofEq. mass M and inertia J are analogous to inductance L. (411) may be obtained by dividing both sides ofEq. and (465). r EXAMPLE 4101 It is logical. Because the spring is deformed when it is subject to a force ft.dy(t) + M yet) = K M dt f{t) M (4289) Considering Example 421. the spring constant K is analogous to the inverse of capacitance lie. must be assigned to the end points of the spring. Writing the force on M and the velocity of the spring as functions of the state variables and the input force j(t). (4238). as state variables. andjk(t).i" EXAMPLE 4102 As another example of writing the dynamic equations of a mechanical system with translational motion. since the former is analogous to the current in L and the latter is analogous to the voltage across C.
49(h). The transfer functions of the system are obtained by applying the gain formula to the state diagram. v(I). and thu s there should be two state variables. (4296) and (4297) clearly show that the system is of the second order. The stale variables are defined as XI (I) = )'2 (/) and X2 (t) = dY2 (r)/dt. R = 200 sec / m:?. (4296) and (4297) give the solutions to Y2(t) and dY2(t)/dl directly. freebody diagram of the system is shown in Fig.)'2 (4294) 2 ~ = B dY2(t) K b () M . The situati on is better explained by refening to the analogous electric network of the system shown in Fig. (4292).(t) the spring as the other state variable. (4292). 476. Eqs. the voltage across the capacitance ee(l) in this case is redundant. as shown in F ig. Eqs. the SFG diagram of the system is drawn in Fig. Assuming a la minar flow. which is the same as dY"2(t)lcit. the valve resistance becomes linear. whereas there me two state variables in v(l) andIket). . 477. (4287). (4298) and (4297) can provide only the sol utio ns to the velocity of M. once fit) is specified. T he state equations are written directly from the state diagram: (4296) (4297) 011 As an alternative. since it is equal Lo the applied volLage e(/). TIle displacement YI(l) is then found using Eq. Then Y2(1) is determined by integrating v(t) with respect to I. The force c!]uatio ns are I(t) = K [y .~ M+B . On the other hand.Y2( t )] dY2(/) (4292) (4293) t KlVI () v~ (t )] = . Y2(S) F(s) I s(Ms + B) MS2 (4300) + Bs+K Ks(Ms + B) (4301 ) EXAMPLE 4103 Dry air passes through a valve into a rigid J m 3 container. Theoretical Foundation and Background Material: Modeling ot Dynamic Systems R + e(t ) c + L Figure 476 Electric network analogous to the mechanical system in Fig. 410. We have dt =  dv (t) B M v(t) +M 1 Ik (t) (4298) (4299) il (t ) = f(t ) One may wonder why there is only one state equation in Eq. The pressure at the lefthand side of the valve is Pi.Jt + Ai 'I I (t )] (4295) B y using the last two equations. Although the network has two energystorage elements in Land C. we can assign the velocity vet) of the mass M as one state vari able and the forceii. (t) . and YI(I) is obtained from Eq. at a constant A Pneumatic System temperature T = 25°C( = 298°K). Find the time constant of the system. which is higher than the pressure in th e tank p.214 Chapter 4. The two state equations of Eqs.dr 2 dt d2Y2(t) These equation s are rearranged as YI (r) = Y2 (/) d Y2( t) + k f(l) . 41O(a).
RairT p = . similar analogies may be obtained. R.! Fluid (incompressible) fl.= >F i(t) > v(t) where e = voltage i(t) = current F = force 'v( t) lineal' ve locity R=B F = ]( /V(t)dt = c=J( v(t) = Mechanical (rotation) 1 l~f J Felt = L = JIy! T = Bw(t) e= >T i(t) R=B T = J( 1 J J = > w(t) w(t)dt where e = voltage i(t) = current T = torque w(t) = angular velocity C=~ J( w=Tdt L=. and low pressures.410 Analogies .7. as shown in Table 48.63(298) .63~ sec 2 °K EXAMPLE 4104 For the liquidlevel system shown in Fig. and electrical System systems.35 ftlbr 0.::.L Analogy F = Bv(t) e.4536 kg °K(9/5) 88. RairT (4302) P = RairT (4303) Thus. 215 Figure 477 A pneumatic system with a valve and a sphel'ical rigid tank. Thermal.. isothermal process. fluid. system time constant isr = RC. from Example 455. R is the A OneTank LiquidLevel resistance. Comparing the thermal. and Fluid Systems and Their Electrical Equivalents R. As a result. C = Ajg is the capacitance and p ::. p I pl' = . = 53.88. SOLUTION Assuming air as an ideal gas. the time constant is _ RV _ "t"  (200){1) _ Rtlil'T .P = Rq(t) (laminar How) R depends on flow regime (! i(t) = > LlP = > q(t) q(t) L = cP C depends on flow regime . 445. from reference [1] at the end of this chapter.p RV . TABLE 48 System Mechanical (translation) Mechanical. . the equation o[ the system is p+p= Pi where air at standard pressure and temperature is represented as an ideal gas. = pf (0 ow m a pipe) A where (:~ = voltage i(t) current = (Omtinueli) .. )sec (4304) where.C.3048m 4..4SN kgm/sec lbln OR = Illr Ibm oR ft lbr N 0.3 .5(1 2 0.
sec .00010zin.sec N = ()y/Om = 1/10 Bt:cause the motor shaft is coupled to the load through a gear U'ain with a gear ratio of N. gear backlash.0050zin. and elevators of the aircraft were all linked to the cockpit through mechanical linkages.sec2 Bm = 0. 479 shows the analytical block diagram of the system using the dcmotor model given in Fig. The reader should refer to Chapter 6.0 ozin. Theoretical Foundation and Background Material: Modeling of Dynamic Systems TABLE 48 (Continued) System Roe. The system is simplified to the extent that saturation of the amplifier gain and motor torque.) The objective of the system is to have the output of the system.sec2 0. (When you get into the real world. Due to the requirements of improved response and reliability. Oy = NOm.015 ozin.01 ozin. Fig. Gone are the days when the ailerons. The purpose of the system considered here is to control the positions of the fins of a modem airship. rudder. the surfaces of modern aircraft are controlled by electric actuators with electronic controls. 9y(t).00020zin. = 1m + N 2JL = 0. some of these nonlinear effects should be incorporated into the mathematical model to come up with a better controller design that works in reality.5 VIA K t = 0 V/radlsec Ra = 5. follow the input.0001 Bf = Bm + N 2BL = + 0. Chapter 4. (4305) the total inertia and viscousfriction coefficient seen by the motor are J.a03H Ki = 9. The socalled flybywire control system used in modern aircraft implies that the attitude of aircraft is no longer controlled entirely be mechanical linkages.0636 V/radlsec .216. 478. Fig.005 + 1/100 = 0. and shaft compliances have all been neglected.sec BL = 1.0ozin. where these topics are discussed in more detail.01/100 = O.sec2 JL = 0. 478 illustrates the controlled surfaces and the block diagram of one axis of such a positioncontrol system. The following system parameters are given initially: Gain of encoder Gain of preamplifier Gain of power amplifier 1(0. = 1 V/rad K = adjustable [(1 = IOVN Gain of current feedback Gain of tachometer feedback Armature resistance of motor Armature inductance of motor Torque constant of motor Back~emf constant of motor Inertia of motor rotor Inertia of load Viscousfriction coefficient of motor Viscousfriction coefficient of load Geartrain ratio between motor and load K2 = 0.lA Kb = 0.L Analogy where A = area of cross section l = length p = fluid density Thermal P = pressure q( t) = volume flow rate R=AT q T=bjqdt i(t) = > q(t) where e = voltage i(t) ::::: current e= >T T = temperature q( t) = heat flow 411 CASE STUDIES EXAMPLE 4111 Consider the system in Fig.lm = 0. 472. 9r (t).00 La = a.
In the system described here. + KJ K2ir)s + RaBt + KJ K2B. = .003 . 478.Ol333 sec Br .= 0. respectively.s [La it s2 ~~K~ (4306) + (Ra i . + La B..411 Case Studies Control surface 217 COlllrul surface SelLsor Command + PREAMP POWER AMPLIRER WITH CURRENT FB Position of control surface TACHOMETER Figure 478 Block diagram of an attitudecontrol system of an aircraft. The forwardpath transfer function of the unityfeedback system is written from Fig.3.= 0.(s) 0 p (s) _ .0002 r..0003 sec (4307) 5+5 The mechanical time constant of the motorload system is i. 479 by applying the SFG gain formula: G(. tracking the sun in only one plane is accomplished. A schematic diagram of the system is shown in .\') 0.. we shall model a sunseeker control system whose purpose is to control the attitude of a space vehicle so that it will track the sun with high accuracy. Gear ratio Power amplifier C urrent feedbac k Tachometer feedback Figure 479 Transferfunction block diagram of the system shown in Fig. The electrical time constant of the amplifiermotor system is 0.015 (4308) EXAMPLE 4112 In this case study.\.= 0. 0. + KiKb + KKI K/Ki 1 The system is of the third order. since the highestorder term in G(s) is .
o:(t) = 0. 481. r ~el1l ~"NI. Tachometer 1:. The cells are mounted in slIch a way that when the sensor is pointed at the sun.. Any difference in the shortcircuit current of the two cells is sensed and amplified by the opamp. The principal elements of the error discriminator are two small rectangular silicon photovoltaic cells mounted behind a rectangular sl it in an enclosure. 48l. or ill(t) = h(t) = O. an error signal will be present at the output of the amplifier when the light from the slit is not precisely centered on the cells.. The description of each part of the system is given in the fol1owing sections. makes an angle 8r (t) with respect to the reference axis. Servo amplifier K Figure 480 Schematic diagram of a sunseeker system.. t ~ Vehicle axis I+L~ I ~B'" I R Ide motor + + + e. 480... This error voltage. Fig. a beam of light from the slit overlaps both cells. and Bo(t) denotes the vehicle axis with respect to the reference axis. and ia(t) = ib(t) = 1. o:(t)._ ~ Sun Ray c 1 Solar axis . near zero: (4309) The coordinate system described is illustrated in Fig.~~ . Error Discriminator When the vehicle is aligned perfectly with the sun. The reference axis is taken to be the fixed frame of the dc motor. From the geometry of the sun ray and the photovoltaic cells shown in Fig. when fed to the servoamplifier. The objective of the control system is to maintain the error between e. Because the current of each cell is proportional to the illumination on the cell . The solar axis. Coordinate System The center of the coordinate system is considered to be at the output gear of the system. or the line from the output gear to the sun. Silicon cells are used as current sources and connected in opposite polarity to the input of an opamp.(t) . and all rotations are measured with respect to this axis.218 Chapter 4. w.(t) and B.ill cause the motor to drive the system back into alignment. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Error discriminator t  Output gear l/n ___ __ lL_ __~~~4_I. we have (4310) ob = 2"  W L tan 0: ( t ) (4311 ) .
the sun ray is completely on cell A. The relationship between the output of the opamp and the currents ia(t) and ib(t) is = (4314) ~~~~~~~~~~ a ~. (C . The abscissa is tan is approximated by Q' Q'. we have ia(t) = 1 + W tana(t) 2LI ib(t) = 1 .: 0 for tana(t) ~ (C + W /2)/L. i!J(t) O. For W /2L :::.~~ 219 Vehicle axis Center of output gear Fixed axis of de motor frame Figure 481 Coordinate system of the sunseeker system. tan aCt) has been approximated by aCt) on the abscissa. for a given a(t). Because the current iit) is proportional to oa and ib(t) is proportional to ob. ia(t) = ib(t) ::. where for small angle a(t). the error discriminator may be represented by the nonlinear characteristic of Fig.W tan aCt) 2LI (4312) (4313) forO ~ tana(t) :::. . but it for small values of Q'. 482. Therefore. W /2L. where oa denotes the width of the sun ray that shines on cell A and ob is the same on cell B. For (C .W /2)/L.W /2}L S tana(t) :S (C + W /2)L.!:!:: L 2L Figure 482 Nonlinear characteristic of the error discriminator. and ia(t) = 2/.411 Case Studies Solar axis . tana(t) :::.!!: L 2L ~+. ia(t) decreases linearly from 21 to zero.
. 483. (4318). 484(a) are modeled in a two degree of freedom (2DOF) system. 483. and damping characteristics of the system as illustrated in Fig. Servoamplifier The gain of the servoamplifier is K. Typically. (4317) DC Motor The dc motor has been modeled in Section 46. EXAMPLE 4113 Classically. The equations are (4318) (4319) (4320) (4321) where J and B are the inertia and viscousfriction coefficient seen at the motor shaft. Tbus. A block diagram that characterizes all the functional relations of the system is shown in Fig. as shown in (c) . as shown in (b). Theoretica l Foundation and Background Material: Modeling of Dynam ic Systems Figure 483 Block diagram of the sunseeker system. With reference to Fig. the inertia. the output of the servoamplifier is expressed as Tachometer The output voltage of the tachometer e t is related to the angular velocity of the motor through the tachometer constant K/: (4316) The angular position of the output gear is related to the motor position through the gear ratio lin. it is sufficient for the following analysis to assume a IDOF model. The inductance of the motor is neglected in Eq. the quartercar model is used in the study of vehicle suspension systems and the resulting dynamic response due to various road inputs.220 Chapter 4. Although a 2DOF sy~tem is a more accurate model. stiffness.
Figure 485 Active control of the IOOF model via a dc motor and rack. .. . (4323) yields the inputoutput relationship = s2 + 2~wns + wii .y(t) lO kg 2. Closed. (b) Two degrees of freedom. where HI K C x(t) yet) :(1) Effective 1A car mass Effective stiffness Effective damping Absolute displacement of the mass Absolute displacement of the base Relative displacement (x(t) .1 m m m the equation of motion of the system is defined as follows: nix(t) + c.t(t) + Ax(t) = cy(t) + ky(t) (4322) which can be simplified by substituting the relation z(t) = x(t) .7135 N/m In 0.. (c) One degree of freedom. Open.. 484(c). (a) Quarter car.y(t) and nondimensionalizing the coefficients to the form z(t) + 2~{J)nz(t) + cv~z(t) = y(t) = a{t) (4323) The Laplace transfOIm of Eq.lx. A(s) Z(s) 1 (4324) where the base acceleration A(s) is the Laplace transform of a(t) and is the input.lX t• . and relative displacement z(s) is the output.Loop Position Control Active control of the suspension system is to be achieved using the same dc motor described in Section 47 used in conjunction with a rack as shown in Fig.Loop Base Excitation Given the system illustrated in Fig. k 1x 1)' (c) ..ly (a) (b) Figure 484 Quartercar model realization.411 Case Studies 221 . 485.9135 Nmls.
: K K ~s+ Ra 1 Ra (4329) La ) ( s+ 1 r Ra 2 mrA(s) KIIlKb a ( s ~ La +1 (Js + Bs + K ) + . .J(t). this is analogous to previous input. The block diagram in Fig.ma(t ) j(t) (4326) (4327) = T (t) . The Automatic Control mrA (s) = 7. 486 can thus be compared to Fig.Bs+K) + ~s V. (4327) into Eq. (4326).(1/11(. + B.i)) r Because z = (Jr. 485. Theoretical Foundation and Background Material: Modeling of Dynamic Systems In Fig. this chapter does not contain any software because of its focus on theoretical development.. T(t) is the torque produced by the motor with shaft rotation andr is the radius of the motor drive gear. mX e.222 Chapter 4. = f(/) . where we address more complex controlsystem modeling and analysis. In Chapters 6 and 9. hence. Eq. (4322) is rewritten to include the active component.s R' ) 412 MATLAB TOOLS Apart from the MATLAB toolboxes appearing with the chapter. (mr2 + lm )s2 + (cr2 + BII/ )s + k1. rearrange.Td(S )).(. + ci + kx = cy + ky + f(t ) (4325) where mz + cz + k. this system is rearranged to the following fom1 : Z(s) ~ (L )(Js2 +R. we will introduce the Automatic Control Systems MATLAB and SIMULINK tools..lIl)i = f(t) . (4328) Noting that Z(s) / r = 8(s) . and K = k1 2 . Using the principle of superposition. and take the Laplace transform to get Z(5) = . B = cr2 + Bm .:. the term mIA (s) is interpreted as a disturbance torque.2 [T(s )  r mrA(s)] .output relationships where 8(s) = Geq(T(s) . Thus.(s) Z(s) Figure 486 Block diagram of an armaturecontrolled dc motor. where J = mr2 + Jm. we can substitute Eq . 485 .
three case study examples were presented that refiect mathematical modeling of practical applications. These simulation tools provide the user with virtual experiments and design projects using systems involving dc motors. sensors. However. and transfer functions. Throughout this chapter. only some of the physical devices used in practice are described. In the end. and methods of approximating the transfer functions by rational ones are described. and fluid systems are described using differential equations. 8. such as error detectors. 9. It can be invoked from the MATLAB command line by simply typing Acsys and then by clicking on the appropriate pushbutton. mechanical. Among the three types of friction described.Review Questions . we have identified subjects that may be solved using ACSYS. and dc motors. Give the advantages of dc motors for controlsystems applications. thermal.' and torques TI and T2 . which type is governed by a linear mathematical relation? 2. Given a two~gear system with angular displacement 61 and 92. The operations and mathematical descriptions of some of the commonly used components in control systems. numbers of teeth Nt and N'l. and how does it affect the perlormance of a control system? 10. The same encoder described in Question 4 and an electronic clock with a frequency of 1 MHz are used for speed measurement. which are discussed in detail in Chapter 6. electronic components~ and mechanical components. What is the angular rotation of the encoder shaft in degrees if 16 zero crossings are detected? 5. Once the linearized model is detennined. REVIEW QUESTIONS 1. What are the electrical and mechanical time constants of an electric motor? 11. What are the sources of nonlinearities in a dc motor? What are the effects of inductance and inertia in a dc motor'? What is back emf in a dc motor. How are potentiometers used in control systems? 4. and the coverage is not intended to be exhaustive. A specific MATLAB tool has been developed for most chapters of this textbook. What is the average speed of the encoder shaft in rpm if 500 clock pulses are detected between two consecutive zero crossings of the encoder? 6. The main purpose of this chapter is to illustrate the methods of system modeling. tachometers. Section 4~9 introduced the linearization of nonlinear systems at a nominal operating point. This chapter includes various examples. and how is it related to the constant Kb? . Analogies were used to relate the equations of these systems.~ 223 Systems software (ACSYS) consists of a number of mfiles and aVIs (graphical user interface) for the analysis of simple control engineering transfer functions. the perfonnance of the nonlinear system can be investigated under the smallsignal conditions at the designated operating point. Digital encoders are used in control systems for position and speed detection. Because nonlinear systems cannot be ignored in the real world." The most relevant components of ACSYS to the problems in this chapter are Virtual Lab and SIMLab. due to space limitations and the intended scope of this text. back~emf Under what condition is the torque constant K. write the mathematical relations between these variables and parameters. Systems with pure time delays are modeled. 7. of a dc motor valid. and this book is not devoted to the subject. Consider that an encoder is set up to output 3600 zero crossings per revolution. with a box in the left margin of the text titled "MATLAB TOOL. 3. The basic mathematical relations of electrical. state equations. are presented in this chapter. 413 SUMMARY This chapter is devoted to the mathematical modeling of physical systems.
4P~2.• Prentice Hall. B. 1995.• John Wiley and Sons. TX. L. J. V. Kennedy. 2002. An inertial and frictional load is driven by a dc motor with torque Tm. 7th Ed . McGraw~Hill. E. Modern Control Engineering. 2nd Ed . Ogata. C. 8. Dynamic. (b) Calculate its natural frequency. 6 Figure 4P~1 r. Lawrence. Does the linearization technique described in this chapter always result in a linear timeinvariant system? The answers to these review questions can be found on this book's companion Web site: www. The dynamic equation of the system is dWm(t) Tm(t) =Jm~+Bmwm If the inertia is doubled. J. Holt. Analysis. (b) Calculate its natural frequency. Woods and K. and Control ofDYlIamic Systems. J. the frictional coefficient Bm is doubled? What is the mechanical constant of the system? 13. A.'" PROBLEMS PROBLEMS FOR SECTION 4·1 4·1. John Wiley & Sons. Kuo and F. Cadwallendel·.V) (a) Find the equation of the motion. 5th Ed . Kom. 11. and how is it used in control systems? 14. What is a tachometer. Give the transfer function of a pure time delay rei 15. F.. 2nd Ed. T. S. Huelsman. 1. B. A. Operational Amplifier Circuits. 2003. NJ. Automatic Control SJslems. New York. REFERENCES 1. Aml~vsis and Design oj Dynamic Systems. 1999. 2. PJ. 8th Ed . 3rd Ed. Palm III.224 Chapter 4. Dynamic Modeling and COlltrol o/Engineering Systems. Esfandiari. J. 1997. Consider the massspring system shown in Fig. B.'er with Applications. Kulakowski. V.• Prentice Hall. 4. NY..• Prentice Hall. 1992. and G. S. New York. Kuo. 4P1.comlcollege/golnaraghi. L P. how will it affect the steadystate speed of the motor? How will the steadystate speed be affected if. Esposito. Automatic Control S)·stems. Shearer. and J. Consider the fivespring onemass system shown in Fig. 1988. Wait.entice Hall. Rinehart and Winston. Irwin/McGraw~Hill. 9.. Golnaraghi.vstems. Introduction to Opemtional Amplifier Theory a1lci Applications. 4th Ed . Modeling anti Simulation o/Dynamic Systems. 2nd Ed. 3. 6.. instead. I. Addison~Wesley. (a) Find its single springmass equivalent. W. i. Vu and R. 42. NJ. Modeling. L. K. NJ. Gardner.S. H. Fluid POl'j. 7. Cochin and W. Fort Worth. C. NJ.wiley. L. 1997. 1997. PI·entice Hall. 10. Theoretical Foundation and Background Material: Modeling of Dynamic Systems 12. 2000. . 1997. R. NI.
• _ fil) r 8] fit) B2 ~ m .t ~ (b) (c) Figure 4p·4 . %~ . respectively. Fig. Write the force equations of the linear translalional systems shown in Fig. (c) Calculate its natural frequency. then: (a) Find the equation of the motion. (b) Determine the transfer function of the system. ~ lvl 2 K ~ M] fil) 8J (a) ~ M 8] ~ . 4P4. Figure 4P3 44. If the mass of the wheel and its mass moment of inertia are III and J. (d) Use MATLAB to plot the step response of the system.Problems 225 Figure 4p2 43.:: . 4p3 shows a simple model of a vehicle suspension system hitting a bump.
and. Write the force equations of the linear translational system shown in Fig. and !1 represents the coefficient of rolling friction. 4P6. F represents the force applied by the engine. respectively. (c) Find the transfer function. (b) Define the state variables as follows : (i) X I = Y2 . X2 = = dyJ/dt Y2 . andx3 = dY2 / d1 andx3 (c) Write the state equations. X2 = dy2Idt . (b) Find the state variables and output equations.\:4 (ij ) X I = dY I / dt = Y2. Find the transfer functions Y1(S) / F(s) and Y2 (s)/ F(s) . If the train only travels in one direction: (a) Draw the freebody diagram. Theoretical Foundation and Background Material: Modeling of Dynamic Systems (a) Draw the system block diagrams or SFGs. 4·5. The mass of the engine and the car are M and m. Draw system block diagrams. Write the state equations. The two are held together by a spring with the stiffness coefficient of K. (iii) X I = )' 1. . Set Mf{ = 0 for the transfer functions . x2 = YI . 4P5 . as shown in Fig. Consider a train consisting of an engine and a car.226 Chapter 4. X3 = Y I . Figure 4P6 A controller is appl ied to the train so that it has a smooth statt and stop. B T YI fi t) +Mg (a) T T fi t) + Mg (b) Figure IIP5 4·6. along with a constantspeed ride. (d) Write the statespace equations of the system. Find the transfer functions YI(s) / F(s) and Y2(S)/ F(s) .
B. andJ\:I). the displacement of th~ trailer. find the state equation of the system. If the rods with the length of L are massless and the spring is attached to the rods ~ from the top. III III Figure 4PB 49. The fo llowing parameters and variables are defined: M is the mass of the trailer. the displaceme nt of the towing vehicle. Fig. o Motor Figure 4P9 .. the forc e of the towing vehicle . (b) Write the state equations by defi ning the following state variables: .Problems 227 47. K. 4P7. the spring constant of the hitch.. 4P9 shows an inverted pendulum on a cart.. the viscousdampi ng coefficient of the hitch. A vehicle towing a trailer through a springdamper coupling hitch is shown in Fig. YI (t). B" the viscousfriction coefficient of the trailer. = YI (t) . figure 4P] (a) Write the diffe rential equation of the system. Y2(t). Assume that the displacement angle of the pendulums shown in Fig. 4P8 are small enough that the spring always remains horizontal.1:1 (t) X2(1) = dY2(t)dt..Y2(t) and 48.
Theoretical Foundation and Background Material: Modeling of Dynamic Systems If the mass of the cart is represented by M and the force f is applied to hold the bar at the desired position . the Lever arm goes up and down. 4Pll shows a wellknown " ball and beam" system in control systems. (c) Find the transfer function. Iff is an impulse signal . (b) Determine the dynamic equation of the motion . 410. As the servo gear turns by an angle e. The change in angle causes the ball to roll along the beam. Fig. 4P1O. then (a) Draw the freebody diagram. and then the angle of the beam is changed by ct. plot the impulse response of the system by using MATLAB. A lever arm is attached to the one end of the beam and a servo gear is attached to the other end of the lever arm. A ball is located on a beam to roll along the length of the beam. (d) Write the state space of the system. (b) Determine the dynamic equation of the motion. then (a) Draw the freebody diagram of mass M. A twostage inverted pendulum on a cart is shown in Fig. (d) Write the state space equations of the system. Figure 4P11 .228 Chapter 4. o Motor Figure 4P10 If the mass of the cart is represented by M and the force f is applied to hold the bar at the desired position . (c) Find the transfer function . A controller is desired to manipulate the ball's position. 411.
and an autopilot is designed to control the pitch of the airplane. Write the torque equations of the system. h. Assume rigid shafts (a) Assume that JI> h. NJ. (d). Find the tota l inerlia the motor sees. The motion equations of an aircraft are a set of six nonlinear coupled differential equations. (a) Determine the longi tudinal eq uations of motion of the aircraft. Write the state equations. Under certain assumptions. 4·14.. (d) Find the ~lep response of the system by using MATLAB. and 1. assume that change in pitch angle does nor affect the speed of an aircraft under any circumstance . (c). Weight Figure 4P1Z Consider that the airplane is in steadycruise at constant altitude and velocity. Find the transfer functions 8. To simplify the problem. Fig.13. and (e).Problems Assuming: 229 m = mass of the baH r = radius of the ball d = lever arm offset g = gravitational acceleration L = length of the beam J = ball's moment of inertia p (X = ball position coordinate = beam angle coordinate () = servo gear angle (a) Determine the dynamic equation of the motion. Write the torque equations of the rotational systems shown in Fig. they can be decoupled and linearized into the longitudinal and lateral equations. T/IJ(t) is the applied torque. and h . 4P. 413. Find the transfer function €')(s) / T(s) for the system in (a) . and h are negligible. N" and N4 are the number of gear tee th. Pitch control is a longitudinal problem.12 shows a simple model of airplane during its !light. N 2. (b) Repeat paJ1 (a) with the moments of inertia J I . (s)jT(s) and 8 2 (s)jT(s) for the systems in paJ1S (b). (b) Find the transfer function and statespace vari ables. h. which means the thrust and drag cancel out and the lift and weight balance out each other. Lift z'. (c) Write the state space equations of the system. (b) Find the transfer function. The moments of inertia of gears are lumped as J]. 412. 4P14. Write the torque equations of the geartrain system shown in Fig. 4P.
and h(t) represents a load torque. 4p·15 shows a motor· load system coupled through a gear train with gear ratio n = N.230 Chapter 4. (a) Find the optimum gear ratio n* such that the load acceleration (XL = d 2 eL/dt 2 is maximized. Figure 4p·15 . (b) Repeat part (a) when the load torque is zero. ! N2.I T(t) K Rigid shaft (a) """ T(t) Flexible shaft (b) [) \(I) K (h(t) J~ J1 T(t ) J'2 Flexible shaft (e) (d) (e) Figure 4p·1 3 N~ Figure 4p·14 4·15. Theoretical Foundation and Background Material: Modeling of Dynamic Systems [)(t) [l [)\(t) . Fig. The motor torque is TIIl(I) .
nt where R" L" AMPLlFIER 1( + en rt) ~d~ Figure 4P1 B . 4P17 shows the diagram of a printwheel system with belts and pulleys.nd the transfer function Y(s)/T". the motor im~Ttia. (b) Fi. (a) Write the differential equation of the system. the motor displacement. and X3 = Will = dem/dt as the state variables. The steel plate is fed through the rollers at a constant speed of V ft/s.:~. Fig. the mass of the printwheel. The following parameters and variables are defined : ~nCt) is the motor torque.. The distance between the rollers and the poi. M . =0 Figure 4p·17 418. 4P16 shows the simplified diagram of the printwheel control system of an old word processor.0:. Bm. (d) Find the transfer function Y(s)/Tm(s) . the motor viscousfriction coefficient.y.Problems • 231 416. Fig. Prinlwheel ~ 8m Motor 1 Jill' Bill . The printwheel is controlled by a dc motor through belts and pulleys. J m.(s). Assume that the belts are rigid. y(t). X2 = dy/dt.W . 8111 (/). (e) Find the characteristic equation of the system. (c) Draw a stateflow diagram for the system. Tm 8m Motor ~ 0T. The belts are modeled as linear springs with spring constants Kl and K 2 • (a) Write the differential equations of the system using em and y as the dependent variables. The schematic diagram of a steelrolling process is shown in Fig. 4P18. r. the pulley radius. Jill' 8". (b) Write the state equations using XI = rem . the linear displacement of the printwheel.) l0r~"'"'(\1) y ~~y Figure 4p·16 417.
. the torsional spring constant. The basic concem from the Rightcontrol standpoint is the lateral force of the air. show that Will Find the steadystate speeds W. (d) Let T. the motor di~'Placement .. as shown in Fig. K.Norma ll y. (a) Write the torque equations of the system. due to the side force. Let the angular acceleration of the missile about the point C. which tends to rotate the missile about its center of gravity. 419.. 4P. tJ".1' 8 m Flexible shaft B. 4P20. = WL = constant in the steady state. The schematic diagram of a motorload system is shown in Fig. The rotary displacement of the motor. The total force Fa may be considered to be applied at the center of pressure P. tJL(rj. a missile encounters aerodynamic forces that tend to cause instabi lity in the attitude of the missile. especially if the point P is in front ofthe cen ter of gravity C.132 Chapter 4. be a constant applied torque. the load velocity. (b) Derive the forwardpath transfer function Y(s)/E(s) and the closedloop transfer func tion Y(s) / R(s) . J m • the motor inertia: B"..(r) LOAD JL 1. and BL • the load viscollsfriction coefficient. Bm(t). wdt). (e) Repeat part (d) when the value of h is doubled . but Jm stays the same . where 11 is a positive constant in ftlrad. the motor velocity.Uthe missile centerline is not aligned with the direction in which the center of gravity C is traveling. The equivalent inertia of the load that is reflected to the motor shaft is J I •. (l) .(/) = T". 4P20. this side force has a tendency to cause the missile to tumble end over end.(r) K •I ') w{. and (VL. (a) Draw a functional block diagram for the system. is converted to the linear displacement y(t) by the gear box and linearactuator combination y( /) = ntJ". (s) / TI1l (s). be denoted by (XF. When traveling through the atmosphere. This problem deal s with the attitude control of a guided missile..(5) / T".(t). a side force is produced by the drag of the air through which the missile travels. (XI" is directly proportional to the angle of attack tJ and is give n by e. w".(t).(s) 1 ®". Figure 4P19 420. the load displacement. (b) Find the transfer functions 0. the motor viscollsfriction coefficient. As shown in Fig. with angle which is also called the angle of altack.. Theoretical Foundation and Background Material: Modeling of Dynamic Syste ms the thickness is measured is d ft.19. Figure 4P20 .. The following parameters and variables are defined: TI/I(I) is the motor torque.. md (c) Find the characteristic equation of the system . MOTOR I)'" TII/(t) w".
X2 = de / dl. 0. = forc e at broom base in vertical direction Atlt. Assume that oCt) is very small. Fig.. x(t) CAR (a) ' W fy + u(/) + x(l) J~ . X} = x . The freebody diagram of the system is shown in Fig.F should be changed to d2 . Find the transfer function 0(s)/ A(s) . the system is analogous to a onedimensional control problem of the balancing of a unicycle or a missile immediately after launching.. and J = missile moment of inertia about C d 1 = distance between C and P The main objective of the flightcontrol system is to provide the stabilizing action to counter the effect of the side force..v directions at the pivot point of the broom.Problems 233 where KF is a constant that depends on such parameters as dynamic pressure. . and so on. so that sin oCt) is apprOJdmated by oCt).''. The objective of the control system is to maintain the broom in the upright position by means of the force u(t) applied to the car as shown. In practical applications. (b) Express the equations obtained in part (a) as state equations by assigning the state variables as XI = B. where 0(s) and A(s) are the Laplace transforms of B(t) and o(t). (a) Write a torque differential equation to relate among Ts . velocity of the missile. Write the force equation of the car in the horizontal direction. (b) Meg Figure 4P21 fx = force at broom base in horizontal direction j. Write the torque equation about the center of gravity CG of the broom. 4P21(b). respectively. e. (b) Assume that Ts is a constant torque. 4P21(a) shows a wellknown " broombalancing" system in control systems. Simplify these equations for small II by making the approximations sinll ~ Band cosB ~ 1. The d l in the expression of Ci. 421. . where CG . and the system parameters given . . u(t) . air density. = moment of inertia of broom about center of gravity CG = M"L2/3 (a) Write the force equations in the x and the . Assume that 0 is very small. = mass of broom g = gravitationaL acceLeration Me = mass of car it. andx4 := dx/dt. (c) Repeat parts (a) and (b) with points C and P interchanged. One of the standard control means is to use gas injection at the tail of the missile to deflect the direction of the rocket engine thrust Ts ' as shown in the figure.
at the equilibrium point xoi ( 0 = 1. a n d ^04(1) = O.234 234 Chapter 4. (a) Draw the stateflow diagram of the system. X03(t) = 0.. of Guide ~~ y(t) //I Friction Free K Figure 4P22 423. 4P23 shows a simple vibration absorber. M yit) M x{t) Figure 4P23 Assuming the harmonic force F(t) =Asin(a>/) is the disturbance applied to the mass M: Assuming pet) = Asin (wt) M: (a) Derive the state space equations of the system. h X02(t) 422. which is called rotating unbalanced. stateflow (b) Find the transfer function. Assume that the frequency of of frequency of rotation of the machine is co. w. transfer function of . FCI) Asin{wt) lht: tlisturbanct: applied (he mass /vi : Assuming the harmonic force F(t) = Ai>u\(col)i~is the disturbanceapplied (0to the mass M: (a) Derive the state space equations of the system. Detennine transfer of th e system 424. step (c) Use MATLAB to obtain the time response of the system. Fig. Theoretical Foundation and Background Material: Modeling of Dynamic Systems (c) Obtain a smallsignal linearized stateequation model for the system in the form of smallsignal of dAx(t) dAx(t) = A' Ax(L) + B*Ar{t) = A*AxU) + B*Ar(r) dl dt point XOI (I) 0> *03(0 0' and X04(t) 0. transfer function. of (b) Determine the transfer function of the system. Most machines and devices have rotating parts. *02 W = 0. of (b) Determine the transfer function of the system. Fig. Fig. 4P24 represents a damping in the vibration absorption. Vibration absorbers are used to protect machines that work at the constant speed from steadysleadystate harmonic disturbance. Even a small irregularity in the mass distribution of rotating components can cause vibration. calise 4P22 represents the schematic of a rotating unbalanced mass of m.
V.:::. (e) Use MATLAB to plot the step response of the system. R L b A K M B e x fer) Figure 4P26 . 426. (b) Determine the tmnsfer function. c =~ (a) R . R o~~~~o 1 C2 (b) Figure 4P25 For each circuit: (a) Find th e dynamic equations and state variables.:::.l UI Vm . Consider the electrical circuits Cl2 II II ~ hown in Figs. Cl2 :12R ~ + 2R :. 4P26 represents a moveableplate capacity. 4P2S(a) and (b).Problems 235 1 Figure 4P24 yCt) PROBLEMS FOR SECTION 42 425. An electromechnnical system sbown ill Pig.
4P30. Theoretical Foundation and Background Material: Modeling of Dynamic Systems Assume that the plate a of the parallel capacitor is fixed and the plate b with mass M is moved by force! If C(x) l:A = d' when the s is the dielectric constant and A is the surface of the plates. Find the transfer function of the circuit for the simple opamp circuit given in Fig. 4P29. . Consider the electromechanical system shown in Fig.236 Chapter 4. An opamp circuit with connection to both tenllinals is shown in Fig. 4P27. (b) Determine X(s) / C(s). 4P28.ion of the plates. Figure 4P28 PROBLEMS FOR SECTION 43 429. Figure 4P27 (a) Draw the freebody diagram. Repeat Problem 427 for the electromechanical system shown in Fig. R r + Figure 4P29 430. 428. and it is related to the charge (q) across the plates: Ie = 2sA q2 (a) Find the differential equations of thi s system. (b) Find the differential equation that describes the operation of the system. then the electric field produces a force opposing the mot. 427. (c) Calculate the transfer function of the system.
. As shown. 431. depending on the difference between the temperature of the ambient air and the temperature of the actuator. and i + and L show the current of these terminals. 4P31.t   .. Find the transfer function for each circuit given in Fig. the actuator is a pure electric resistance and the heat flow is generated by the electric power input. A thermal lever is shown in Fig. The lever (at the top) moves up or down proportionally..v. 4P32.Problems Rill 237 R J + Vi" + RI .J when v+ and v_ represent the voltages of positive and negative terminals.. (a) Find the positive feedback ratio.. c Rin RI + c Rill + + + o . . (b) Find the negative feedback ratio. respectively. + VOllI Figure 4P30 The opamp can be modeled as VOU1 107 s+l i+ = L = 0 = [v+ .. Calculate V(s)/X(s). assuming zero initial conditions. (c) Determine when the circuit remains stable.." 1 0 Figure 4P31 PROBLEMS FOR SECTION 44 432.
r Insulator ___ Oil Figure 4P33 The radii shown in Fig. and the heat sink temperature is constant and equal to the atmospheric temperature. Assign any needed parameters. the frame is an ideal insulator. Theoretical Foundation and Background Material: Modeling of Dynamic Systems x 1n + R V ~ Figure 4P32 433. Q. /"2> and I"J from inside to outside. 4P33.238 Chapter 4. Hot oil forging in quenching vat with its crosssectional view is shown in Fig. A power supply within an enclosure is shown in Fig. Assuming: kl' = The thermal conductivity of the vat k i = T he thermal conductivity of the insulator c" = The specific heat of the oil do = The density of the oil c = The specific heat of the forging m = Mass of the forging A = The surface area of the forging Jz = The thickness of the bottom of the vat T" = The ambient temperature Determine the system model when the temperature of the oil is desired. 4P33 are /"[. 4P34. Enclosure Figure 4P34 . The heat is transferred to the atmosphere from the sides and bottom of the vat and also the surface of the oi I with ~ convective heat coefficient of k". detennine the model of the system that can give the temperature of the power supply during its operation. the heat transfers from the power supply to the enclosure by radiation and conduction. Assuming the rate of heat generation within the power supply is known and constant. Because the power supply generates lots of heat. a heat sink is usually attached to dissipate the generated heat. r. 434.
12.(t) / dt = rate of air leaving intake manifold through intake valves T(t) = engine torque T" = disturbance torque due to application of auto accessories := constant . PROBLEMS FOR SECTION 45 436. Fig .. and c shows the specific heat of fluid. Engine torque is developed from the buildup of manifold pressure due to air intake and the intake of the air/gas mixture into the cylinder.Problems 435. 4P36). .(t) = average air mass in manifold qA t) = amount of air leaving intake manifold through intake valves dq. The input of the system is the throttle position that controls the rate of air flow into the manifold (see Fig.T) = qgUillEd where Ih represents the mass flow. Assign any required parameters.{l)/dl = rate of air flow across throttle into manifold q". such as radii. then (rnc)(T2 . 4P35 shows a heat exchanger system . ~ TO MANIFOLD. The engine variations are as follows: AIR FLOW • + + + + I IDLESPEED / J \ / . ~ CONTROL MOTOR Figure 4P36 q i(l) = amount of air flow acrOss throltle into manifold dq . If the length of the heat exchanger cylinder is L. derive a model to give the temperature of Fluid B leaving the heat exchanger. The objective of this problem is to develop a linear analytical model of the automobile engine for idlespeed control system shown in Fig. thennal conductivity coefficients. TJ and T2 are the entering and leaving Auid temperature. and the thickness. 239 Fluid A Fluid B Figure 4P35 Assuming the simple material transport model represents the rate of heat energy gain for this system.
(d) Approximate the engine time delay by rv I . The engine drag is modeled by a viscousfriction torque Bro(t).TDS/2 e t:DS . The equation describing the mechanical components is T(t) = J tit + Btv(t) + Td dw(t) (a) Draw a functional block diagram of the system witha(t) as the input.. K3 = constant 3.. where B is the viscous~friction coefficient. Fig.240 ~ Chapter 4. s. (b) Find the transfer function fl(s)/a(s) of the system. the weight density is J1. )1t) + ... 4P38.. Vibration can also be exhibited in fluid systems.. . A long pipeline connects a water reservoir to a hydraulic generator system as shown in Fig. (c) Find the characteristic equation and show that it is not rational with constant coefficients. w(t) as the output. The average air mass qm(t) is determined from qm{t) = J (dq.. The rate of air flow into the manifold is linearly dependent on the throttle position: d~y) = Kla(t) engine speed: Kl = proportional constant 2..J  y(t) '_ _ _ _ L _ _ _ _ _.TD) K4 = constant 4. 4~38.. The rate of air flow leaving the manifold depends linearly on the air mass in the manifold and the dqo(t) ~ = K2qm(t) + K3 W (t) X2.dqo(t))dt dt dt 6. 4P37 shows a U tube manometer. A pure time delay of 1:0 seconds exists between the change in the manifold air mass and the engine torque: T(t) = K4qm(t ... and Tdas the disturbance input. (b) Calculate the natural frequency of oscillation of the fluid. Show the transfer function of each block. (a) Write the state equation of the system. Theoretical Foundation and Background Material: Modeling of Dynamic Systems wet) a(t) = engine speed = throttle position TD = time delay in engine Je = inertia of engine The following assumptions and mathematical relations between the engine variables are given: 1. J Figure 4P~37 Assume the length of fluid is L. and the crosssection area of the tube is A.. 4~37..(t) .1 + TDS/2 and repeat parts (b) and (c).
P. v__.. A simplified oil weU system is shown in Fig. v Figure IIP38 At the end of the pipeline.In Ie maL radius = R T Figure 4P39 .A I ~ /' . there is a valve controlled by a speed controller..Problems 241 Q. ~ 2. Walking Beam FLexible Rod No friction Sucker rod 111 . Consider the turbinegenerator is an energy converter. It may be closed quickly to stop the water flow if the generator loses its load.. TinCt) . Assign any required parameters.. Assuming the pressure in the surrounding rock is fixed at P and the walking beam moves through small angles. 4P39. In tbis figure. PzA mg ~ . determine a model for this system during the upstroke of tbe pumping rod. Determine the dynamic model for the level of the surge tank. the drive machinery is replaced by the input torque. Q. 439..
and Hz are steadystate heads. derive the statespace model of the system when h. 4P41 shows a twotank liquidlevel system. 4P40. . and H. Fig. so then the differential displacement of the spring determine~ the input to the hydraulic servomotor. 4P41 are supposed to be small . As shown in Fig. :________ ___ _________ ___ ___ ~. If the other quantities shown in Fig. The flyweight is moved by engine. Theoretical Foundation and Background Material: Modeling of Dynamic Systems 440. (a) Find the dynamic equation of mOlion . Assume that Q.net I Fuel (Steam) A I t I : Hydraulic I Servomotor I ~ Drain S\lpply t I Pressure Drain Figure 4P40 441. and Q2 are the steadystate inflow rates. 4P42.ow to Eng. A hydraulic servomotor usually is used for the speed control of engines. and hz are outputs of the system and qu and qi2 are the inputs. An accelerometer is a transducer as shown in Fig.242 Chapter 4. T Figure 4P41 T H2+h2 PROBLEMS FOR SECTION 46 442. Determine the state space model of the system. (b) Determine the transfer function. (c) Use MATLAB to plot its impulse response. the reference speed is controlled by thc throHle lever.
TEMPERATURE SENSOR (b) Figure 4P43 .Problems 243 + vOltag e x(t) 1 y(t) Figure 4P42 443. Fig. 4P43(a) shows the setup of the temperature control of an airnow system.EXCHANGER TA o T. r i TAl> Air intake TAJ Heat exchanger (a) Temperature sensor Ts OIl1PIll air temperature ~ + _ CONTROLLER Gc(s) Error ~ ELECTRICTOPNEUMATIC TRANSDUCER ~ dM" PNEUMATIC HEAT ACTUATORIVALVE f. The temperature sensor senses the air temperature TAo and sends it to be compared with the reference temperature TrValve HOTWATER RESERV01R p(l) ELECTRICTOPNEUMATIC TRANSDUCER u(/) CONTROLLER T. The hotwater reservoir supplies Ihe waler that flows into the heat exchanger for heating the air. __ _ _ Hemeda.
KJ )1 Tnt 1111 I) T".T ( ) 1m S Till S m S (d) Find the characteristic equation of the system. the output air temperature. Theoretical Foundation and Background Material: Modeling of Dynamic Systems The temperature error Te is sent to the controller. the load displacement. Jm . I) 01. the spring constants of the shafts. 8111' COm OmtWm 8f. the load velocity.() . 8. The schematic diagram of a control system containing a motor coupled to a tachometer and an inertial load is shown in Fig. 4P45. The output of the controller.. J"" the motor inertia. (V" the tachometer velocity. is the How rate of the heating fluid = KMu. the motor inertia.. The following parameters and variables are defined : BIIl(t) is the motor displacement. Find the transfer functions Eo(s)/ 1~. The motor torque T". WI . the potentiometer viscousfriction coefficient. the motor torque. K. B " the motor velocity. (b) Derive the transfer function TAO (S)/T. I=t===l=j TI/l(/) LOAD + B p Figure 4P44 The potentiometer has a maximum range of 10 turns (2071' rad). Fig. is converted to a pneumatic signal by a transducer. and T. The following parameters and variables are defined : elM . TmCt). 8z.'\O. The output of the actuator controls the waterflow rate through the threeway valve. .Ts Ts (a) Draw a functional block diagram that includes all the transfer functions of the system. and Bm. Leave out the initial states. is the input. WL. J" the tachometer inertia. the load n displacement. Bp. KM = 0. andwm as the state variables (in the listed order). 1. KR = 65° Clkg/s .. which is an electric signal .. The following parameters and variables are defined: Tm is the motor torque. K! 1. em. Heattransfer equation between water and air: Tc . el .(s) when Gc(s) = 1. at the left and ending with BL on the far right. 445. the torsional spring constant. BL(l).(s). BII " the motor viscousfriction coefficient. WL. BL. the water temperature = KRdM ".054 kgJsN. eQ(t). the motor viscousfriction coefficient. Ju the load inertia. the tachometer displacement. 4P43(b) shows the block diagram of the system. (b) Draw a signal flow diagram with T.. )1 wL Load Tachometer Motor Figure 4P45 . T". . An openloop motor control system is shown in Fig. ®ds) 8((s) ®m(s) (c) Fllld the folIowmg transfer functIOns: r() . and K2 .. and K. The state diagram should have a total of 10 nodes. (a) Write the state equations of the system using eL.1(£).. 4P44.. Potcmiollleter + MOTOR Ev. the output voltage.244 Chapter 4. 444. which has the transfer function Gc(s). dTAO dt = Tw  TAO Tc = 10 seconds = 2 seconds Temperature sensor equation: Ts dTs dt = TAO . .
(t ) + La . = = (b) Draw a functional block diagram of the system with gains or transfer functions in the blocks. The basic elements of a phaselocked loop system incorporating a de motor are shown in Fig. counter gain = I/N. ea(t ) = R. Ln. The idea of using the counter N is that./. the armature cun'ent. the motor velocity. encoder gain = Ke. (a) Derive the transfer function E. various desired output speeds can be attained by changing the value of N.. The digital encoder produces digital pulses that represent motor speed.(s) / E(s) of the filter shown in Fig.. PROBLEMS FOR SECTION 47 448. Find N if the desired output speed is 200 rpm. Find N if the desired output speed is 1800 rpm. The reference frequencY!f is fixed <\t 120 pulses/so Find K" in pulses/rad. (c) Derive the forwardpath transfer function n ll1 (s)/ £(s) when the feedback path is open. the armature inductance.. and dcmotor torque constant Kj • Assume zero inductance and zero friction for the motor. with!.lt) is the applied voltage. The phase detector compares the motor speed and the refere nce frequency and sends an error voltage to the filter (controller) that governs the dynamic response of the system. Phase detector gain Kp. (e) Repeat p311s (a). (f) The digital encoder has an output of 36 pulses per revolution. K b• the backe mf constant. 4P46(b). Assume that the filter sees infinite impedance at the output and zero impedance at the input. wm(t). An input pulse train represents the reference frequen cy or desired output speed. 4P46(a). Describe how an incremental encoder can be used as a frequency divider. (d) Find the closedloop transfer function !1m (s) / P. 4P46(c). and Wll ~ t).(t) .(s) .(t) + Kb Wm (t ) dt where e.. and (d) for the filter shown in Fig. fixed. The voltage equ ation of a de motor is written as di" . the reference input voltage. with zero initial .. tic motor R" III II Reference frequency N FILTER ec(t) AMPLIFIER K + en i" LOAD h Jill fw COUNTER IIIII Feedback pulses (a) K" R] + R2 eel) ec(t) e(t) + + + C 0 I 0 o~~o (b) (e) Figure 4P46 447. Phaselocked loops are control systems used for precision motorspeed control. Rm the armature resistance. Taking the Lapl ace tra nsform on both sides of the voltage equation. (cl. i.Problems 245 446.
(b) Find the transfer function between Om (S) and Or(S) (TL = 0) when H. (a) Write the causeandeffect equations of the system. (a) o 0. Show that you can obtain the solutions by taking the limit as KL approaches infinity in the results in part (c). The block diagram in Fig. 4P48 shows a dcmotor system. X4 = Wm . x 3 = em . K L .Molor and load ~.sec 2) .'m S  ('\ ( ) _ Ea(s) . The load in this case is the printwheel. K b . the motor velocity wm(t) is totally independent of the loaddisturbance torque TL . (b) Draw the signal flow diagram using the nodes shown in Fig. the geartrain ratio = e2 / e". for speed control.e. Jm .sec 2).. andX5 = i". o o o 0 0 Wm (b ) 0 8m 0 w. Rearrange these equations into the form of state equations with XI = eo . K. the backemf constant (V/radlsec). K. Voltage feedback ~I.sec). we get ·l.246 Chapter 4.{s) and His) are selected as in part (a). i. the torsional spring constant of the motor shaft (ozin. 4P49 shows the schematic diagranl of a dcmotor control system for the control of the printwheeI of a word processor. the motor viscousfriction coefficient (ozin.) 0 (I). (c) Derive the forwardpath transfer function (with the outer feedback path open): G(s) = E>o(s) / E>e{s). KL = 00. Find the closedloop transfer function M(s) = E>o{s) / E>r(s) .lA). Figure 4P49 . Show that when H. and JL the load inertia (ozin. n. the torque constant (ozin. (a) Let K1 be a very high gain amplifier. with voltage and current feedbacks.(Ra + Las)la(s) Kb which shows that the velocity information can be generated by feeding back the armature voltage and current. tu e K JII/' Bill GEAR TRAIN n LOAD Jl . which is directly coupled to the motor shaft. Bm . X2 = w o . Theoretical Fo undation and Background Material: Modeling of Dyna mic Systems conditions and solving for 0m(05) . 0 00 m"..(s) / He(s ) = (Ra + Las). = Tm/T2. 4P49(b).I Figure 4P48 449. The following parameters and variables are defined: Ks is the errordetector gain (V/rad). Fig. the motor inertia (ozin. (d) Repeat part (c) when the motor shaft is rigid.lrad). the amplifier gain (V IV).
s2 . The voice coil of tht: VCM in Fig. 4PSl(b). When current is sent througb the coil. the motor input voltage. 4PSI(a). Kb.(t) . the errordetector applied voltage = 2rr: V.. T1I1 . BIIJ. (h. la(s) . the back emf.000 ozi n. R".s2 . . MT . the mutual inductance between the primary and shortedturn coils. 1L. the load inertia =O. e. the amplifier gain. Kb. the primarycoil current. (. the total viscousfriction coefficient of the voice coil and load. the motor resistance = 1. the displacement of the voice coil. = (}3 . Primary tums '\ I \ Magnetic flux (a) Shorted tums [(. and Y(s) as variables.(t) is the applied coil voltage. K. A dcmotor positioncontrol system is shown in Fig. the force of the voice coil. f(t) = KiV (t). en the reference input. L" L. and B1'.s2 . ea . the back emf. 1"1> the motor inertia =0. the torsional spring constant =50. Ks. the primarycoil resistance. the backemf constant. causing the coil to move lillearly. v(t).. is shown in Fig.lA. the backemf constant = IS. E. KA . and variable~ = (}e X2 = (}r  (}L· (a) Write the state equations of the system using the following state variables: Xl = (h . the total mass of the voice coil and load . (b) Figure IlP5D 451. 4PSO(b) shows the equivalent circuit of the coik The following parameters and variables are defined: e. i. the magnetic field of the PM interacts with the currentcarrying conductor. the shortedturn coil current. 4PSO(a). The schematic diagram of a voicecoil motor (VCM).( /). The latter is installed for the purpose of effectively reducing the electric constant of the device. V(s). Is(s) . yet). (a) Write the differential equations of the system. L". (b) Draw a block diagram of the system with Ea(s). the motor torque. 4PSO(a) consists of a primary coil and a shortedturn coil. the motor viscousfriction coefficient =10 ozin. Fig.. Cc) Derive the transfer function YCs) / EaCs).S V/lOOO rpm. the force constant.lrad. the errordetector gain = E / 2rr:. i(I> the motor current. R(/.OS ozin. eb(t) = Kb v(t).. the velocity of the voice coil.Problems 247 450. KL .. the motor torque constant 21 ozin. The VCM consists of a cylindrical pemlanent magnet (PM) and a voice coil. The following parameters and are defined: e is the error voltage. the primarycoil inductance. andx4 = d(}m / dt = Wm' (b) Draw a signal flow diagram using the nodes shown in Fig.03 oZin.1511. X3 = d(}L/dt = WL. Las. K j . the load position. used as a linear actuator in a disk memorystorage system.
Show that the system is described .(t) where e(t) is the applied voltage.1'3 0 . Theoretical Foundation and Background Material: Modeling of Dynamic Systems (c) Derive the forwardpath transfer function G(s) =.k (lI) . the magnetic flux frum a separately excited field = "0'&(1).. (d) Deri ve the closedloop transfer function M(s) = ElL(s) / (><)c(s). e ~4 0 X::.g(x) dv(t) dl + f(t) where xU) = linear displacement of train v(t) = linear velocity of train k(v) = resistance force on train [odd function of I'. tj(t). the armature current.\"'2 0 x~ 0 XI 0 XI (b) Figure 4P51 452.. (a) Consider that the motor is a dc series motor with the armature and field windings connected in series. Find the poles of G(s). and Ril = O. and comment on the significance of these values of KA • + e + K" + =E (a) 0 0 0 0 0 X4 0.2738.(l) f(t) = Ki¢(I)i. Locate these poles in the splane. 0 . i(lU). k(v) = Bv(t). and K. Find the poles of M(s) when KA = 1..248 Chapter 4. B . the field current. The following differential equations describe the motion of an electric train in a haclion system: dx(t) dt = v(t) = . the force constant. with the properties k(O) = 0 and dk(v) / dv = OJ g(x) = gravitational force for a nonlevel track or due to curvature of tJack f(l) = tractive force The electric motor that provides the trac tive force is described by the following equations: e(t) = K. and 5476.8 L(s )!Ele(s) when the outer feedback path from eL is opened... cp(t). Rlj' the armature resistance. so that ia(t) = i J(I).¢ (t)v(t) + Rai. g(x) = 0.
= G(s) [ la{s) ] TL(s) Robot ann Figure 4P53 . TL(s). (e) Consider the same conditions as in part (a) but with ljJ(t) as the input. Derive the state equations. and 8ds) as node variables.n (t) and edt) as outputs. Derive the stale equations of the system.Problems by the fonowing nonlinear state equations: dx(t) = vet) dt ~ 249 dv(t) d = t  Bv(t) + Ki 2 2( ) KbKfV t e (t) 2 (b) Consider that. (e) Express the transfer~function relations as 0 nz {S)] [ ®[. The system parameters and variables are given as follows: DC Motor Robot Arm J L = inertia of arm TL = disturbance torque on ann Tm = motor torque = Kjia K. (b) Draw an SPG using la(s). = torque constant ill = armature current of motor Jm = motor inertia 8 m = motor viscousfriction coefficient B viscousfriction coefficient of shaft between the motor and arm BL = viscousfriction coefficient of the robot arm shaft = arm displacement K = torsional spring constant Bm = motorshaft displacement (h = (a) Write the differential equations of the system with ia(t) and Tdt) as input and 8. io(t) is the input of the system [instead of e(t)]. for the conditions stated in part (a).(s) Find G(s). 453. 4P53. The linearized model of a robot arm system driven by a de motor is shown in Fig. 8 m(s).
Theoretical Foundation and Background Material: Modeling of Dynamic Systems PROBLEMS FOR SECTION 48 454. Use. e(t) 1 T Mg Electromagnet Stee l ball Figure 4P57 . The praeticlIl application of this system \s the mugnetk levitation of trains or magnetic bearings in highprecision control systems. Find the nonlinear (c) Linearize the state equations about the equilibrium point and express the linearized state equations as dCix(t ) = A* ~x(t) + B*Cie(t) df The force generated by the electromagnet is KFU)/y(t) . (a) Let Eeq be a nominal value of E. e). magnet with the voltage e(t) .h ' G(s ) = A(s) = (tiS + 1)(t2 S + 1) where Ttl is the time delay. MATLAB to solve Problem 455 and plot the step response of the systems. The objective of the control is to keep the metal ball suspended at the nominal equilibrium position by controlling the current in the. The resistance of the coil is R . Find the polar plot of the following functions by using the approximation of delay function described in Section 2. (b) Use MATLAB to verify your answer in part (a). Find the nominal values of y(t) and dy (t) / dl at equilibrium. X2(1) = y( t).8.sL (a) G(s) = (Ts + 1) 2 + 2se . where L is a constant. Fig. The steel ball is suspended in the air by the electromagnetic force generated by the electromagnet. 4P57 shows the schematic diagram of a ballsuspension control system. PROBLEMS FOR SECTION 4. andx3(t) dx(t) state equations in the form of & = f (x.9 457. e. The applied voltage e(t) is a constant with amplitude E.2s (b)G(s)= s2+ 3s + 2 456.s + 4e. (b) Define the state variables at XI (I) = i(/). 455. where K is a proportional constant. The transfer function of the heat exchanger system is given by T(s ) Ke. = dy(t) / dt. (a) Plot the root~ and zeros of the system. and the inductance is L(y) = L/Y(t) . and the gravitational force on the steel ball is Mg.250 ~ Chapter 4.
2 Assume alI units are consistent. X2(1) = dY I(t)ldl. Fig. (a) Given YI = 1. 4PS8(b). and X4(t) = dY2(t) / dt . Fig. Y I (1) . where M 1 = mass of electromagnet = 2. . The state variables are defined as . X3(t) = Y2(t). The freebody diagram of the system is shown in Fig. respectively. (e) Develop an analogous electrical circuit for this system. When the system is at Iht: stablt: t:quilibrium point. 4P59 shows a typical grain scale. Let the stable equilibrium values of the variables. The objective of the control is to keep the metal ball suspended at the nominal position by controlling the current in the electromagnet. (a) Find the freebody diagram. The steel ball is suspended in the air by the electromagnetic force generated by the electromagnet. and Y2(t) be 1. Y" and Y2.0 g = gravitational acceleration = 32 . tind / and Y2' (b) Write the nonlinear state equations of the system in the form of dx(t)ldt = f(x. Assign any required parameters. any small perturbation of the ball position from its floating equilibrium position will cause the control to return the ball to the equilibrium position.Problems 251 458. 0.0 M2 = mass of steel baJJ = 1.0 B = viscousfriction coefficient of air = 0. (e) Find the state equations of the linearized system about the equilibrium state J. (b) Derive a model for the grain scale that determines the waiting time fo r the reading of the weight of grain after placing on the scale platform. 4P58(a) shows the schematic diagram of a ballsuspension system.\1(1) = YI(t) . Y" and Y2 in the form dx(t) = A *~x(t) df + B* tli(t) Fixed iron plate Freebody diagram Electromagnet (a) (b) Figure 4p· 58 PROBLEMS FOR SECTION 410 4·59.1 K = proportional constant of electromagnet = 1. i(t) .
Develop an analogous electrical circuit for the mechanical system shown in Figure 4P60.======== + P2 + Q Q Figure 4P61 PROBLEMS FOR SECTION 411 462._ _ _ _ _ f P ====. illustrated in Fig.252 Chapte r 4. Develop an analogous electrical circuit for the fluid hydraulic system shown in Fig. 463. The openloop excitation model of the car suspension system with IDOF. 484(c) . is given ill Example 411 3. . An active control designed for the car suspellsion !. Use MATLAB and the transfer function of the system given in Example 41 13 to plot the impulse response of the system. T h 1 F _ ' .ystem with \DOF is designed by using a dc motor and rack. Compare your result with the result of Problem 462. 4P61. Use MATLAB to find the impulse response of the system. Theoretical Foundation and Ba ckground Material: Modeling of Dynamic Systems Figure 4P59 460. m Figure 4P60 461.
using inverse Laplace transforms. simply. in most controlsystem problems. we depend on the background material discussed in Chapters 14 to arrive at the time response of simple control systems. we can find the overall transfer function of the system and. it is usually of interest to evaluate the state and output responses with respect to time or. then. Let y(t) denote the time response of a continuousdata system. Also in this chapter. in general. the time response.CHAPTER 5 TimeDomain Analysis of Control Systems In this chapter. For instance. Therefore. we first need to model the overall system dynamics and find its equation of motion. it can be written as y(t) = y/(I) + Yss(t) (51) 253 . and develop simple design criteria for manipUlating the time response. we can manipulate the system final response by adding feedback or poles and zeros to the system block diagram. We finally look at simple proportional. Next. Depending on our objectives. · 51 TIME RESPONSE OF CONTINUOUSDATA SYSTEMS : INTRODUCTION Because time is used as an independent variable in most control systems. using Laplace transforms. Each subsystem may have sensors and actuators to sense the environment and to interact with it. and the perfolll1ance of the system is evaluated by studying the system response in the time domain. or other subsystems. In the end. Throughout the chapter. initial condition. we utilize MATLAB in simple toolboxes to help with our development. if the objective of the control system is to have the output variable track the input signal. The system could be composed of mechanical. we look at more details of the time response analysis. delivative. and integral controller design concepts in time domain. it is necessary to compare the input and output responses as functions of time. we can find the transfer function of all the subcomponents and use the block diagram approach or signal flow diagrams to find the interactions among the system components. obtain the time response of the system to a test input. The time response of a control system is usually divided into two parts: the transient response and the steadystate response. we look at the effects of adding a simple gain or poles and zeros to the system transfer function and relate them to the concept of control. electrical. a reference input signal is applied to a system.nolll1ally a step input. In the analysis problem. discuss transient and steady state time response of a simple control system. Finally. In order to find the time response of a control system. the final evaluation of the perfolll1ance of the system is based on the time responses. starting at some initial time and.
not only is the mathematical treatment of the problem systematized. Thus. the response of a typical control system cannot follow sudden changes in the input instantaneously. the inputs to many practical control systems are not exactly known ahead of time. mass. By selecting these basic test signals properly. in a radartracking system for antiaircraft missiles. Time~Dornain Analysis of Control Systems where Yt (I) denotes the transient response and Y. it is necessary to assume some basic types of test inputs so that the perrormance of a system can be evaluated. It is possible to predict the timedomain behavior of the system from its frequencydomain characteristics. Because inertia.254' Chapter 5. the steadystate response when compared with the desired reference position gives an indication of the final accuracy of the system. and controllers are designed so that the specifications are all met by the designed system. the actual inputs of a control system may vary in random fashion with respect to time. transient response is defined as the part of the time response that goes to zero as time becomes very large. a sinusoidal input with variable frequency is used. and inductance are unavoidable in physical systems. For the purpose of analysis and design. the position and speed of the target to be tracked may vary in an unpredictable manner. In control systems. Therefore. and the deviation between the output response and the input or the desired response.. For instance. In many cases. since the response to complex signals can be determined by superposing those due to simple test signals. In general~ if the steadystate response of the output does not agree with the desired reference exactly. stable control systems exhibit transient phenomena to some extent before the steady state is reached. because it is a significant part of the dynamic behavior of the system. the control of the transient response is necessarily important. because it is difficult to design a control system so that it will perform satisfactorily to all possible forms ofinput signals. such as a sine wave.. before the steady state is reached.. must be closely controlled. and transients are usually observed. This approach is particularly useful for linear systems. Thus. In the design problem. curves in terms of the amplitude ratio and phase between the input and the output are drawn as functions of frequency. The study of a control system in the time domain essentially involves the evaluation of the transient and the steadystate responses of the system. because it indicates where the system output ends up when time becomes large. so that they cannot be predetennined. When the response of a linear timeinvariant system is analyzed in the frequency domain. In the design problem.s (1) denotes the steadystate response. When the input frequency is swept from zero to beyond the significant range of the system characteristics. The steadystate response of a control system is also very important. For a positioncontrol system. . but the response due to these inputs allows the prediction of the system's performance to other more complex inputs. . or a ramp function that increases with time. Yt (t) has the property 1·00 lim Yt{t) =0 (52) The steadystate response is simply the part of the total response that remains after the transient has died out. specifications are usually given in terms of the transient and the steadystate performances. All real. perfonnance criteria may be specified with respect to these test signals so that the system may be designed to meet the criteria.52 TYPICAL TEST SIGNALS FOR THE TIME RESPONSE OF CONTROL SYSTEMS Unlike electric networks and communication systems. This poses a problem for the designer. the steadystate response can still vary in a fixed pattern. the system is said to have a steadystate error.
the ramp input denotes the constantspeed r(t) r(t) R ~ o (a) o (b) r(t) o (c) Figure 51 Basic timedomain test signals for control systems. Mathematically. 51(b). it is equivalent to the application of numerous sinusoidal signals with a wide range of frequencies. The step function as a function of time is shown in Fig. Also. a step input represents the sudden rotation of the shaft. a wide band of frequencies in its spectrum. the following deterministic test signals are used. because the step function contains. a ramp function is represented by r(t) = Rtus(t) (55) where R is a real constant.52 Typical Test Signals for the Time Response of Control Systems ~ 255 To facilitate the timedomain analysis. if the input is an angular position of a mechanical shaft. For example. Ramp·Function Input: The ramp function is a signal that changes constantly with time. (a) Step function. (b) Ramp function. (c) Parabolic function. The ramp function is shown in Fig. Or (53) r(t) = RUs(t) (54) where us(t) is the unitstep function. StepFunction Input: The stepfunctiun input represents an instantaneous change in the reference input. 51(a). If the input variable represents the angular displacement of a shaft. in principle. as a result of the jump discontinuity. . The step function is very useful as a test signal because its initial instantaneous jump in amplitude reveals a great deal about a system '8 quickness in responding to inputs with abrupt changes. The mathematical representation of a step function or magnitude R is r(t)=R t~O =0 t<O where R is a real constant.
we seldom find it necessary or feasible to use a test signal faster than a parabolic function. 52 illustrates a typical unitstep response of a linear control system. and the recovery in speed should be made as quickly as possible. . However. the transient response of a control system is necessarily important. Let y(t) be the unitstep response.Ys:. 51 (c). in addition to striving for a desirable idle speed in the steady state. 53 THE UNITSTEP RESPONSE AND TIMEDOMAIN SPECIFICATIONS As defined earlier.SS' The maximum overshoot of y( t) is defined as maximum overshoot = Ymax . Yss' the steadystate value of y(t).256 . A system with a large overshoot is usually undesirable. These signals all have the common feature that they are simple to describe mathe~ matically. the transient portion of the time response is the part that goes to zero as time becomes large. because both the amplitude and the time duration of the transient response must be kept within tolerable or prescribed limits. Time~Domain Analysis of Control Systems rotation of the shaft. For example. Let Ymax denote the maximum value of yet). we can define signals with still higher rates. in the automobile idle~speed control system described in Chapter I. the system must have highorder integrations in the loop. This is because. Fig. For design . as we shall see later. Maximum overshoot. the characterization of the transient response is often done by use of the unitstep function us(t) as the input. the signals become progressively faster with respect to time. which is called the jerk junction. which usually leads to serious stability problems. For linear control systems. With reference to the unitstep response. Mathematically. Nevertheless. The ramp function has the abili ty to test how the system would respond to a signal that changes linearly with time. (57) The maximum overshoot is often represented as a percentage of the final value of the step response. In theory. in reality. the transient drop in engine speed must not be excessive. Chapter 5. and Ymax ?: Y. and so forth. percent maXImum overshoot = maximum overshoot Yss x 100% (58) The maximum overshoot is often used to measure the relative stability of a control system. such as t 3. From the step function to the parabolic function. that is. performance criteria commonly used for the characterization of linear control systems in the time domain are defined as follows: 1. The response of a control system when the input is a unit~step function is called the unitstep response. in order to track a highorder input accurately. it is represented as r(t) = ""2 us(t) RP (56) where R is a real constant and the factor 1/2 is added for mathematical convenience because the Laplace transform of r(t) is simply R/s3• The graphical representation of the parabolic function is shown in Fig. ParabolicFunction Input: The parabolic function represents a signal that is one order faster than the ramp function.
parabolicfunction. 52. and.f'L~.. 5.I!:. the maximum overshoot is often given as a timedomain specification.. although Fig. or even a sinusoidal input.1 Figure 52 Typical unitstep response of a control system illustrating the timedomain specifications. The settling time Is is defined as the time required for the step response to decrease and stay within a specified percentage of its final value.95 0. The rise time I. is defined as the time required for the step response to rise from 10 to 90% of its final value. Rise time. 3. Ir Setting timets .£~~=> 0. except for simple systems lower than the third order. This is shown in Fig.. 52 shows that the maximum overshoot occurs at the first overshoot. 52.10 O~~~+~~~. .05 1..00 0. 52. Steadystate error. the maximum overshoot may occur at a later peak. For some systems. as shown in Fig. The delay time td is defined as the time required for the step response to reach 50% of its final value.50 0. The steadystate error of a system response is defined as the discrepancy between the output and the reference input when the steady state (t ~ oo) is reached. Delay time. rampfunction. A frequently used figure is 5%.5·3 The Unit·Step Response and Time·Domain Specifications y(t) 257 1. 4. 51 (Problem 523).. 2. These timedomain specifications are relatively easy to measure when the step response is wen defined. An alternative measure is to represent the rise time as the reciprocal of the slope of the step response at the instant that the response is equal to 50% of its final value. purposes. It should be pointed out that the steadystate error may be defined for any test signal such as a stepfunction.90 ~. if the system transfer function has an odd number of zeros in the righthalf splane. a negative undershoot may even occur (4... Settling time.r. 52 only shows the error for a step inpul... these quantities are difficult to establish. The four quantities just defined give a direct measure of the transient characteristics of a control system in terms of the unitstep response. Analytically. The unitstep illustrated in Fig. as shown in Fig.
For instance. the steady state of the output response seldom agrees exactly with the reference. 53. and the error is simply e(t) The steadystate error is defined as = r(t). . The error of the system may be defined as e(t) = reference signal . the final position accuracy of an elevator would be far less stringent than the pointing accuracy on the control of the Large Space Telescope. Definition of the SteadyState Error with Respect to System Configuration Before embarking on the steady~state error analysis. and y(t). we can regard the error as a signal that should be quickly reduced to zero. we must first clarify what is meant by system error. Therefore. bet). When the system has unity feedback. Let us assume that the objective of the y(t) Y(s) Figure 53 Nonunity feedback control system. depending on the form and the purpose of H(s). H (s) = 1. then the input r(t) is the reference signal. that is. In a design problem. where r(t) is the input. steadystate errors in control systems are almost unavoidable. ifpossible. The accuracy of position control of such a system is often measured in rnicroradians. Let us refer to the closedloop system shown in Fig. 541 SteadyState Error of Linear ContinuousData Control Systems Linear control systems are subject to steadystate errors for somewhat different causes than nonlinear systems. the steadystate errors of linear control systems depend on the type of the reference signal and the type of the system.y(t) (510) ess = 1+00 e(t) lim (511) When H(s) is not unity. u(t). which is a telescope mounted onboard a space shuttle. the actuating signal. The accuracy requirement on control systems depends to a great extent on the control objectives of the system. and no corrective effort is exerted. and at the same time the transient response must satisfy a certain set of specifications. one of the objectives is to keep the steadystate error to a minimum. In general. because of friction and other imperfections and the natural composition of the system. In general. although the reason is still that the system no longer "sees" the error. TimeDomain Analysis of Control Systems 54 STEADYSTATE ERROR One of the objectives of most control systems is that the system output response follows a specific reference signal accurately in the steady state.the output. In the real world. 5#2 mayor may not be the error. the feedback signal.y(t) (59) where the reference signal is the signal that the output y(t) is to track.258 p Chapter 5. The difference between the output and the reference in the steady state was defined earlier as the steadystate error. or below a certain tolerable value. the actuating signal u(t) in Fig.
and the steadystate error would be zero. 5~15: 1 G(s) = s2(s + 12) H(s) = Kts %use Kt=lO 1~~~~~ 100 %Step input Kt=10. (59). The closedloop transfer function of the system is yes) M(s) = R(s) = 1 + G(s)H(s) = s(s2 + 12s + 10) Step Response G(s) 1 (516) Toolbox 541 For the system in Eq. and the closedloop system is unstable. where the reference signal is the desired velocity and not r(t). 53 is to have the outputy(t) track the input ret) as closely as possible. and K t is the tachometer constant.Kt) . let Kt = 10 volts/rad/sec. the output voltage of the tachometer would be 1 volt. This Ineans that.~) i SOO lime (sec) cloop=feedback(G. 1000 . because r(t) and y(t) are not of the same dimension. ylabel('Ampl~tude'). and the system transfer functions are G(s) We can show that. (512) stabilizes the system. it would be meaningless to define the error as in Eq.H) step(cloop) xlabel('Tirne(sec)'). To illustrate the system further.1 rad/sec. (510). the desired velocity in the steady state is 1/10 or 0.system in Fig. However. [] . In this case. the system error may still be defined as in Eq. for a unitstep input of 1 volt. if H(s) . and the characteristic equation becomes :l + 17s3 + 6082 + 5s + 5 = 0 (514) In this case. consider a velocity control system in which a step input is used to control the system output that contains a ramp in the steady state. The system error should be defined as in Eq. because wht!I1 this is achieved. the characteristic equation is 3 8 + 12s2 + 1 = 0 (513) which has roots in the righthalf splane.s2(s + 12) ~ 1 H(s) = 5(s + 1) (s + 5) (512) = 1. [0 0 12] G=tf(Gzpk) H=zpk(O. We can show that the H(s) given in Eq. Gzpk=zpk( []. The system transfer functions may be of the form 1 (515) G(s) = s2(s + 12) H(s) = Kts where H(s) is the transfer function of an electromechanical or electronic tachometer. (510).
More specifically. This number is known as the type of the control system or. e" depends on the characteristics of G(s) . sR(s) (519) 1 + G(s) Clearly.12 . Because of these reasons.901t t~0 (517) Because the exponential terms of y(t) in Eq. 2. 1.0. The steadystate error of the system is written ess = lim t + DC e(t) = lim sE(s) S>O = Illn '::'::s+O . Systems with unity feedback. Type of Control Systems: Unity Feedback Systems Consider that a control system with unity feedback can be represented by or simplified to the block diagram with H(s) = 1 in Fig.260 Chapter 5. The objective here is to establish a definition of the error with respect to one basic system configuration so that some fundamental relationships can be determined between the steadyslate error and the system parameters. Fig. Systems with nonunity feedback. Not all system errors are defined with respect to the response due to the input.0. Systems with nonunity feedback. the steadystate error of the system is ess = lim [ 0.11 .lt . = For a unitstep function input.0. we shall classify three types of systems and treat these separately. and H(s) has zeros at s = 0 of order N.lt .12 (518) More explanations on how to define the reference signal when H (s) =f 1 will be given later when the general discussion on the steadystate error of nonunity feedback systems is given. we can show that ess depends on the number of poles G(s) has at s = O. 53. in addition to the input r(t). 54 shows a system with a disturbance d(t). H(s) = 1. system type.OOO796e.l l.y(t)] / + 00 = 0. TimeDomain Analysis of Control Systems D(s) R(s) + + yes) Figure 54 System with disturbance input. R(s) l is.12. (517) all diminish as t > 00. The output due to d(t) acting alone may also be considered an error. To establish a systematic study of the steadystate error for 1inear systems. the definition of system error has not been unified in the literature. The output time response is y(t) = O. . 3. the steadystate part of yet) is O. but H(O) = K{j = constant.1208e O.l t + 0. Thus. simply.
If G(s) is described by Eq. for e. G(s) can be expressed for convenience as G(s) = K~l + T\s)(l + T2 S)'" (1 + TmIS+ Tm2S2) e. R lun = so I sR(s) = S~O 1 +R = . The total number of tenns in the numerator and the denominator and the values of the coefficients are not important to the system type. (520).'1)( 1 + 2s}( 1 + .Ii' + s2) G(s) = K(l s~ 2s) type 1 type 3 (522) Now let us investigate the effects of the types of inputs on the steadystate error.s( 1 + .. In general.jrnust be at least equal to unity. we can summarize the steadystate error due to a step function input as follows: = Type 0 system: Type I or higher system: ess = 1 + K p = constant R ess = 0 .TdS sl(l + Tas)(l + TbS)'" (1 + TnlS + Tn2S2) (520) where K and all the T'g are real constants. Therefore. We shall consider only the step. for Kp to be infinite. The system type refers to the order of the pole of G(s) at s = O. The fonowing example illustrates the system type with reference to the form of G(s). where j = 0. when the input is a step function~ K" must be infinite.5s) .. 2. as system type refers only to the number of poles G(s) has at s = O. 53 is a step function with magnitude R. we see that. Thus." 261 We can show that the steadystate error ess depends on the type of the control system. and parabolic inputs. Steady~State Error of System with a Step~Function Input When the input r(t) to the control system with R(s) = 1 of Fig. We see from Eq. G(s) must have at least one pole at s O. Then Eq.'1'0 (524) as the steperror constant. (520) is type j. 1. · EXAMPLE 541 G(s) _ K(l + O. (525) that. (519)~ ess ' .54 Steady~State Error '. ramp. 55. that is.hm + G(s) G(s) 1 + y~ O(s) (523) For convenience~ we define Kp = lim G(s) .fS to be zero. Let us fonnaJize the system type by referring to the form of the forwardpath transfer function G(s).. (5~23) becomes ess = 1+ Kp R (525) A typical ess due to a step input when K" is finite and nonzero is shown in Fig.. R(s) =Rls. the steadystate error is written from Eq. . the closedloop system having the forwardpath transfer function of Eq.
the Laplace transform of r(t) is (526) R(s) = 2" s R The steadystate error is written using Eq.~(t) r(t) y(t) R~~t~~~~~=*==== o Figure 5~5 t Typical steady~state error due to a step input. (519). . A typical e. . 56.S+'O limsG(s) ess Then. Time~Domain Analysis of Control Systems Reference input lit) = Ru. r(/) y{t) o t Figure 56 Typical steadystate error due to a rampfunction input. Eq. 53 is a ramp function with magnitude R.262' Chapter 5. (526) becomes =K v R (530) which is the steadystate enor when the input is a ramp function.$+'OS + sG(s) limOsG(s) s+ (528) We define the ramperror constant as Kll = . e H = I 1m •• R R = . r(t) = Rtus(t) where R is a real constant. SteadyState Error of System with a Ramp~Function Input When the input to the control system [H(s) = 1] of Fig.u due to a ramp input when Kl • is finite and nonzero is illustrated in Fig.
must be infinite. I.n = 00 = K .5~4 Steady~State Error . The following conclusions may be stated with regard to the steadystate error of a system with ramp input: Type 0 system: ess e. 263 Eq. (530) shows that.\0 IimsG(s) = lim==r so sJ K j = 0.. Using Eqs. for K to be infinite. 57. . . for ess to be zero when the input is a ramp function. or the system must be of type 2 or " higher. 53 with H(. (520) and (529). j must be at least equal to 2. we obtain K.". = . 2. (531) Thus. Defining the parabolicerror constant as Ka the steadystate error becomes = so lim iG(s) (535) R ess = Ka r(t) y(t) (536) o I Figure 57 Typical steadystate error due to a parabolicfunction input.. K •.. = constant R Type 1 system: Type 2 system: Steady~State ess = 0 Error of System with a ParabolicFunction Input When the input is described by the standard parabolic fann (532) the Laplace transform of r(t) is R(s) = 3" s R (533) The steadystate error of the system in Fig.....') = 1 is R A typical ess of a system with a nonzero and finite Ktl due to a parabolicfunction input is shown in Fig.
for these results to be valid. the types of systems with reference to Eq. = 0 We cannot emphasize often enough that. Ramp. the method is applicable to only the system configuration shown in Fig. depending on the individual situation.<. the closedloop system must be stable. the steadystate error due to the parabolic input is zero if the system is of type 3 or greater. Because the error constants are defined with respect to the forwardpath transfer function G(s). When the system configuration differs from that of Fig. 5. respectively. 4. (520)1 and the input types. and ParabolicFunction Inputs for UnityFeedback Systems SteadyState Error ess Type of System Error Constants Step Input Ramp Input Parabolic 11. TimeDomain Analysis of Control Systems TABLE 51 Summary of the SteadyState Errors Due to Step. Table 51 shows the relations among the error constants. The steadystate error properties summarized in Table 51 are for systems with unity feedback only. . it is important to check first to see if sEes) has any poles on the jwaxis or in the righthalf splane. The steadystate error of a system with an input that is a Hnear combination of the three basic types of inputs can be determined by superimposing the errors due to each input component. 5~3 with H(s) = 1. ramp function. The error constants defined here mayor may not apply. 3. or parabolicerror constants are significant for the error analysis only when the input signal is a step function. KG 00 00 R j Kp Kv 0 K 00 00 0 1 2 3 K 00 DC Ka 0 0 K 00 R l+Kp Ji K~ T+K 0 R 00 K R 00 0 0 0 0 K 0 Following the pattern set with the step and ramp inputs. ramp. 53 or establish the error signal and apply the finalvalue theorem. 2. As a summary~ the following points should be noted when applying the errorconstant analysis just presented.264 ::: Chapter 5. 53 with H(s) = 1. 1. or parabolic function. Because the error analysis relies on the use of the finalvalue theorem of the Laplace transform. the steady~state error of any linear closedloop system subject to an input with order higher than the parabolic function can also be derived if necessary. As a summary of the error analysis. By using the method described. we can either simplify the system to the form of Fig. The step. The following conclusions are made with regard to the steady~state error of a system with parabolic input: Type 0 system: Type 1 system: Type 2 system: Type 3 or higher system: ess = ess 00 = 00 Ii R ess = K = constant es.
EXAMPLE 542 Consider that the system shown in Fig. Gs _ ( ) 8 2 ($ 5(s + 1) + 12)(s + 5) R(s) = 1 Type 2 system We can show that the closedloop system is stable. The following examples illustrate the utility of the error constants and their values in the determination of the steadystate errors of linear control systems with unity feedback.5) % Step input K=1. which is 0 < K < 1. . G K(s + 3. c. b. K G(s) = s2(s + 12) H(s) =1 Type 2 system The closedloop system is unstable for all values of K.5)(s + 0.2K l • R R Parabolicerror constant Ka == 0 R ess = Ka = 00 These results are valid only if the value of K stays within the range that corresponds to a stable closedloop system.0 . since the finalvalue theorem cannot be applied. and error analysis is meaningless. the errorwconstant method does not indicate how the error varies with time. 5] 1) I G=tf(Gzpk). The error constants and steadywstate errors are calculated for the three basic types of inputs using the error constants.304. H=li clooptf=feedback(G. that is.15) (s) = s(s + 1. 53 with H(s) == 1 has the fonowing transfer functions. This is one of the disadvantages of the errorconstant method.15} (a) G(s) = ~'(s + 1.2K ess = K = 4.H) step(clooptf) xlabel('Time(sec)'). a.5){s + 0.5·4 Steady~State Error . % Use K~l H(s) =1 Type 1 system Gzpk=zpk ( [ . 5 . when the error increases continuously with time. 15] .3 . Toolbox 542 For the system in Example 542: K(s + 3. ylabel('Amplitude')..= 0 R 1 + Kp Ramp input: Parabolic input: Ramperror constant Kv = 4.5) H(s) =1 Type 1 system Step input: Steperror constant Kp = 00 e SI = .~: 265 When the steady~state error is infinite. The steadystate errors are calculated for the three basic types of inputs. [0 1. The errorconstant method also does not apply to systems with inputs that are sinusoidal. .
For the present discussion. let us impose the following condition. which is usually known. plot(t. 1. [y.y. plot(t. u = 0 .t.6 04 02 O~~~~~L~~~~ o 20 4D 60 80 100 120 140 160 180 Time(seC) (sec) Similarly you may obtain the ramp and parabolic responses %Ramp input t=O:O. Often. [y.r~. the closedloop transfer function is derived in the analysis process.t).u): title( 'Closedloop response for Parabolic Input') xlabelC'Time(sec') ylabel('Amplitude') Step input: Ramp input: Step~error constant: Kp = 00 Ramperror constant: Kv = 00 ess = . Ali it turns OU4 the closedloop transfer function can be used to find the steadystate error of systems with unity as well as nonunity feedback.266 Chapter 5.xJ=lsim(clooptf.x]=lsim(clooptf. 5 *t .8 1. *t .__.y.u)j title(' Closedloop response for Ramp Input') xlabel('Time(sec)') ylabel('Arnplitude') %Parabolic input t=O:O.= 0 Kv R Parabolic input: Parabolicerror constant: Ka = 1/12 R eu = Ka = 12R Relationship between SteadyState Error and ClosedLoop Transfer Function In the last section. Time~Domain Analysis of Control Systems step Response 2r.2 0.1:50. the steadystate error of a closedloop system was related to the forwardpath transfer function G(s) of the system.4 1. and it would be of interest to establish the relationships between the steadystate error and the coefficients of the closed1oop transfer function. u=t.u._~._.1:50.t).. s+O lim H(s) = H(O) = KH = constant (537) .t.u.
... in the transfonn domain. which means that the system is stable. Because the signal that is fed back to be compared with the input in the steady state is KH times the steadystate output. R(s) = R/s. bo = aD.hoKH)R ao (543) Thus. Y(s)/R(s). (540) into the last equation and simplifying. 1 (538) E(s) = KH R(s)  I Y(s) = KH [1  1 KHM(s)JR(s) (539) where M(s) is the closedloop transfer function.yet) or. For a stepfunction input. 1. when this feedback signal equals the input. We further require that all the poles of M(s) are in the lefthalf . the steadystate error in Eq..jor a unityjeedback system KH = 1. we get (542) We consider the three basic types of inputs for ret). the steadystate error due to a step input can be zero only if ao boKH =0 or (544) M(O) = bo =_1 ao KH (545) This means that. + bIs + bo sn + an_ISn1 + .. the steadystate error would be zero. for the steadystate error to he zero. the constant terms of the numerator and the denominator of M(s) must be equal.5·4 SteadyState Error ( 267 which means that H(s) cannot have poles at s = O. that is. . Step~function input. (542) becomes _ 1 ess ~ KH (ao .KHM(s)]sR(s) (541) Substituting Eq.'Iplane. Thus. Notice that the above development includes the unityfeedback case for which KH = 1. Let us aCOjsume that M(s) does not have any poles at s = 0 and is of the form M(s) = Y(s) R(s) = bm:/u + bm_1Sm.O KlH [1 lim lim .1 + . + atS + ao (540) where n > m. The steadystate error of the system is written ess = s·O sE(s) = S. we can define the reference signal as r(t)/ KH and the error signal as e(t) = KH ret) .
The steadystate errors due to the three basic types of inputs are evaluated as follows: Step input: Ramp input: Parabolic input: ess =0 since ao sinceao QOKH = bo( = 5) == bo(= 5) andal = hi (= 5) 5 .l + .. (542) becomes (550) The following values of ess are possible: ess = 0 ess = ess = a2 b2KH if aibiKH = 0 for for for i = 0.fO (548) (549) ess = 00 3. Gs ( ) . +ats+ao) + (ao  boKH) R (546) The following values of e. and thus KH = H(O) = 1. the same results are obtained with the errorconstant method. R(s) = R/s3 • For a parabolic input. (542) becomes e = _I_lim sn KHS+O ss + . _ bzKH R ~ 60 Because this is a type 2 system with unity feedback.12R R~ ess =0 _ a2  ess .boKH = 0 ao .biKH = 0 aj  ° and I and I (552) (553) biKH =10 i= 0 f'· EXAMPLE 543 The forwardpath and closedloop transfer functions of the system shown in Fig. the steadystate error in Eq. + (al s(sn +all_lsn. . Rampfunction input.boKH =10 and at  bIKH. Thus. the steadystate error in Eq. and2 i= (551) ao 00 K H R = constant if if ai .. The system is assumed to have unity feedback.btKH)S For a rampfunction input.~!'i are possible: ess =0 if ao . the system is stable.. . Parabolicfunction input. Time~Domain Analysis of Control Systems 2. so H(s) = I. R (s) = R/s2 . 53 are given next.268 i: Chapter 5.blKH aoKH R = constant if if ao ..s2(s + 12)(s + 5) 5(s + 1) Ms _ ( ) s4 + 1783 + 60s2 + 5s + 5 5(8 + 1) (554) The poles of M(s) are all in the lefthalf splane.boKH = 0 and a) b{KH =0 (547) ess = at . I.
1 + G(s)H(s) .t(t)/2~ sinceal .!!86t + t1.289Bt 1:. az = 60. .OO01381e4. 53 has the following transfer functions: O(s) Thus. = Unitstep input. and taking the inverse Laplace transform of yes). at = 5.8.0302t cosO.8eO.3.0.O. which has the following transfer functions: 1 G(s) = s2(s + 12) Then. and the steadystate error to a unit ramp is 0.11.8! + 11. and the steadystate error is zero.O.05t + 2.. r{t) = t[{.8861 .05t . bo = 5.0302t cosO.1 + G(s)H(s)  $4 s+5 + 17s3 + 60s2 + 5s + 5 (556) Comparing the last equation with Eq.:::0 (557) Thus. and unitparabolic inputs to the system described by Eq. Applying the unitstep.886t + 0. The closedloop transfer function is M s _ Yes) _ G(s) _ ( ) .step input: yet) = I . the steadystate error is O. the outputs are Unit.5t2 . hI = I. ret} = us(t): Unitramp input.O.blKH #0 It would be illuminating to calculate the steadystate errors of the system from the difference between the input and the output and compare them with the results just obtained.365eo. (556).5.682 x 1O5 e J2.2898t Unitparabolic input: 3.00056e12..0. unitramp.::: 0 (559) The steadystate portion of y(t) is 0.2898t t 2: 0 Thus.8t .\'4 (562) .05t .R(s) .8t . (540).2.2eO. EXAMPLE 545 Consider that the system shown in Fig.R(s) .0302t sinO.2898t t. yet) = 0.11.2&9&t . Thus.928ge0. the steadystate value of y( t) is unity.O. r(t) = tus(t): e tS = 00 Unitparabolic input. 269 EXAMPLE 544 Consider the system shown in Fig.784 x 1O6e~4. 53.8842 x 1O6 e 12.5t2 .9993eo.2898t + 3. Unitramp input: yet) =t .2 .8. KH H(s) = 5(s + 1) 8+5 (555) = H(O) = 1.2.O.0.8 + 4. the steadystate portion of y(t) is t . = s2(s+ 12) KH I H(s) = s+ 5 lO(s+ 1) (560) = s+(} H(s) = 2 lim s+5 + 17s3 + 6052 + lOs + 10 (561) The closedloop transfer function is M(~) _ Yes) _ G(s) _ .0302t cosO. The steadystate errors of the system are calculated for the three basic types of inputs..1301eo.0302t sinO.54 SteadyState Error :. we have ao = 5.0302t sinO. which becomes infinite as time goes to infinity.826 x 1O5 e 4. and b2 O.
270' Chapter 5.. In the real world.. = [0. and because KH = 2.4t . (540) will have a pole at s = O. (562).hoKH) KH ao y(t) = 0.2. as calculated in Eq. SteadyState Error of Nonunity Feedback: H(s) Has NthOrder Zero at s = 0 This case corresponds to desired output being proportional to the Nthorder derivative of the input in the steady state. + aIS btKH)S + (al . Unitramp input r(t) = tus(t): es. we get + transientterms (564) Thus.4t . the steadystate error due to a unitramp input is 0.4t + 2.transient terms (567) Because the transient terms will die out as t approaches infinity. the transfer function of M(s) in Eq. = KH _1 (a1 ao = 2 (10 10 x 2) = 04 b1KH) ~ 1 .4us (t) . using Eq. the steadystate error due to a unitstep input is zero.blKH:PO y(t) = [0..4]u. Thus..2. In this case.5t O.6]us(t) + transienttenns The error due to the unitparabolic input is e(t) = ~H r(t)  y(t) = (O.5 x 2) = 0 2 10 (563) Solving for the output using the M(s) in Eq. Unitparabolic input r(t) = tlus (t)/2: es. which increases with time. (538).4. this corresponds to applying a tachometer or rate feedback.transient tenns (569) Thus..5.6. + (a2 s" + an_lSn1 + . the steadystate error is O. (570) as e _ ss  1 r KH sl!!A [SII1 + . (566). TimeDomain Analysis of Control Systems The steadystate errors of the system due to the three basic types of inputs are calculated as foHows: Unitstep input r(t) = us{t): ess = _1 (ilO . the steadystate error is calculated as e(t) = :H r(t)  y(t) == O.s = 00 The unitparabolic input is since at . The steadystate error is written from Eq.. the steadystate value of yet) is 0.. the reference signal can be defined as R(s)/KHSN~ and the error signal in the transform domain may be defined as E(s) where 1 = KHSN R(s) · H(s) KH::::::: 1 Im:.5us (t) =! (10 .6)us{t) . for the steadystate error analysis. or ao = O.~(t) (565) The unitramp response of the system is written y(t} + transient terms (566) Thus.bOKH)] R( ) S S (572) .r50 y' yes) (570) (571) We shall derive only the results for N = 1 here.2512  O.
den) j sys_cl=£eedback(sys. because the objective is to control velocity with a step input. (575)~ is ess = == _1_ KH (a 2  ht at KH) = ~ (60 . the reference signal is considered to be tus(t)/KH = 0. although M(s) has a po1e at s O. The result is (5 .. [Yrt)=step(sys_cl). a2 = 60. bo = 5. we find the unitstep response using the closed~loop transfer function in Eq. the closedloop system would be unstable. and bl = 1. Of course. 81) y(t) = (0. The coefficients are identified to be ao = 0. the steadystate error. 53 has the following transfer functions: 1 G( s) . = .2.9)us (t) + transientterms From the discussion that leads to Eq. Toolbox 543 The corresponding responses for Eq. (579).6 Consider that the system shown in Fig. the derivative control in the feedback path also has a stabilizing effect... the steadystate error is 2.51 . al = 10. For a unitstep input. 5~79 are obtained by the following sequence of MATLAB functions t=0:O. (570).:. 271 For a step input of magnitude R.bIKH::/:O if al  eS $ = 00 boKH f:.boKH = 0 (574) (575) (5~ 76) bIKH K R = constant H if at boKH = 0 but a2 .s2(8 + 12) H(s) = lOs s+5 (577) Thus. sys tf(nurn. the steadystate error is ess ess =0 = a2 if at a2 blKH =0 and al . the last equation is written (573) Thus. it should be pointed out that if R(s) were a constant for this type 2 system. KH = so s hm=2 .R(s)  s4 + 17s3 + 60$2 + lOs s+5 (579) The velocity control system is stable. from Eq.9....9 1 2 (580) To verify this result.5tus(t) in the steady state. nurn== [1 5] .. den= [11760100].54 SteadyState Error . EXAMPLE 54. 0 We shall use the following example to i11ustrate these results.10 x 2) = 2 . thus.l).. H(s) (578) The closedloop transfer function is M s _ Y(s) _ ( ) . So.1:5Q.
.. Outputred 1. no magnetic flux. then: when the amplitude of the amplifier input signal falls within the dead zone. plot(t. The fluxtocurrent relation of the magnetic field of an electric motor may exhibit a similar characteristic. For instance..4 0.. and the system may generate an error in the output whose Output Input Figure 5~8 Typical inputoutput characteristics of an amplifier with dead zone and saturation. 58 are not limited to amplifiers.S J ._.. if an amplifier used in a control system has the inputoutput characteristics shown in Fig. and.6 0.. This property is illustrated by the quantization characteristics shown in Fig.y. 59. 1. When the input to the quantizer is within ±q/2..4 r....:~_. can take on only discrete or quantized levels..2 10 20 30 Time(secs) 50 60 542 SteadyState Error Caused by Nonlinear System Elements In many instances~ steadystate errors of control systems are attributed to some nonlinear system characteristics such as nonlinear friction or dead zone. The output signals of digital components used in control systems.2 . and the control would not be able to correct the error if any exists. g ') xlabel('TirneCsecs)') ylabelC'Aroplitude') title('Inputgreen. no torque will be produced by the motor to move the load. Time~Domain Analysis of Control Systems u=ones(size(t)). ..t. Deadzone nonlinearity characteristics shown in Fig. the output is zero.. As the current of the motor faUs below the dead zone D. such as a microprocessor.~ 0. thus. 58.u..rr___.g O. the output of the amplifier would be zero.272: Chapter 5. 'r' . Outputred') I Inputgreen...
4'1 . . When there is no friction. Coulomb friction is a common cause of steadystate position errors in control systems. Fig.4'1 . The torque on either side of point 0 represents a restoring torque that tends to return the output to the equilibrium point when some angulardisplacement disturbance takes place. When the control of physical objects is involved. because there is always a restoring torque so long as the position is not at the stable Torque Figure 5·10 Torqueangle curve of a motor or closedloop . as well as the other periodic intersecting points along the axis where the slope on the torque curve is negative.3'1 .54 SteadyState Error Ou tput 273 5'1 4" 3" 21} q 5'1 . This type of error is also known as the quantization error in digital control systems. The torque curve typically could be generated by a step motor or a switchedreluctance motor or from a closedloop system with a position encoder. the position error should be zero. Point 0 designates a stable equilibrium point on the torque curve. magnitude is related to ±q/2.3'1 2'1 q 2q 3'1 4'1 5'1 fnput .2q . friction is almost always present.5</ Figure 59 Typical inputoutput characteristics of a quantizer.yslem with Coulomb frictio n. 510 shows a restoringtorqueversusposition curve of a control system..
100 Fig. Thus. (582): y(t) = I .et/r: where 'f is the time for yet) toreach 63% of its final value of lim yet) = 1. Note that the input in Eq. 55 TIME RESPONSE OF A PROTOTYPE FIRSTORDER SYSTEM Consider the prototype firstorder system of form dy(t) dt + !y(t) =. (584).c(y(t)) = Y(s). For a unitwstep input t /(t) If yeO) = us(t) = {~: : ~~' (583) 1 =y(O) = 0. The step response will not have any overshoot for any combination of system parameters.274 liP Chapter 5.63 o Figure 5~ 11 Unit wstep response of a prototype firstorder system. then Y(s) = :. and. Therefore.c(us(t)) = s and . exact and detailed analysis of errors in nonlinear control systems can be carried out only by computer simulations. hecaU!~e all physical systems have nonlinear characteristics of one fonn or another. Usually. TimewDomain Analysis of Control Systems equilibrium point. .. as the motor torque falls below TF as the rotor position approaches the stable equilibrium point. it may stop at any position inside the error band bounded by ±8e. it is difficult to establish general and closedform solutions for nonlinear systems. Although it is relatively simple to comprehend the effects ofnonlinearities on errors and to establish maximum upper bounds on the error magnitudes. we must realize that there are no errorfree control systems in the real world. (582) is scaled by for cosmetic reasons. 510. we get the time response ofEq. steadystate errors can be reduced but never completely eliminated. 511 shows typical unitstep responses ofy(t) for a general value of T. as shown in Fig. If the rotor of the motor sees a Coulomb friction torque T p. . As the value of time constant r decreases. /(t) 'f r where'f is known as the time constant of the system. which is a measure of how fast the system responds to initial conditions of external excitations. yet) 0. then the motor torque must first overcome this frictional torque before producing any motion. the system response approaches faster to the final value.s + l/T 1 1/'f (584) Applying the inverse Laplace transform to Eq.!. : .
I'plane of the system transfer function. The openloop transfer function of the system is G(s) = yes) = w~ E(s) s(s + 2~(V1I) where ~ (586) and WII are real constants. 512 shows the location of the pole at s = .~ will always stay in the lefthalf splane. especially the ones that can be approximated by secondorder systems. 56 TRANSIENT RESPONSE OF A PROTOTYPE SECONDORDER SYSTEM Although true secondorder control systems are rare in practice. 275 jw splane I o Figure 512 Pole configuration of the transfer function of a prototype firstorder system. (587) to zero: il(s) = i + 2i. (586) and (587) is defined as the prototype secondorder system. and the system is always stable. The characteristic equation of the prototype secondorder system is obtained by setting tbe denominator of Eg. For positive T. 513 with the transfer functions given by Eqs. . the output response of the system is obtained by taking the inverse Laplace transform of the output transform: yes) = u} s(s2 + 2~wlIs + w~) II (589) e(t) Y(f) s(s E(s) + 2~{OIl) Y( s) Figure 513 Prototype second·order contm l system.~ in the .. r Fig.wlls + (V~ = 0 (588) For a unitstep function input. Consider that a secondorder control system with unity feedback is represented by the block diagram shown in Fig. their analysis generally helps to form a basis for the understanding of analysis and design of higherorder systems.. R(s) = 1/ s.50 Transient Response of a Prototype SecondOrd er System . 513. The closedloop transfer functio n of the system is Yes) R(s) (587) The system in Fig. the pole at s = .
(590) plotted as functions of the nonnalized time w"t for various values of ~. the step response does not exhibit any overshoot. and settling time but does not affect the overshoot..8 r~I"':""""1lr.::'~!"~~'r 1. 514 shows the unitstep responses ofEq.at 9 10 II 12 13 Figure 514 Unitstep responses of the prototype secondorder system with various damping ratios. When ~. As seen. y(t) never exceeds its final value during the transient. Time~Domain Analysis of Control Systems 1.2 1 I o 2 3 4 5 678 lU.lj\l~j+++jl+++I 10. The result is t~O Fig. that is. This can be done by referring to the Laplace transform table in Appendix C. The responses also show that W}l has a direct effect on the rise time.276 ~ Chapter 5. delay time.I++·.. .7 1.6 ff t.. the response becomes more oscillatory with larger overshoot as ~ decreases. =O.::: 1. These will be studied in more detail in the following sections.
e 1 .1:50.4 1." 2] .. e. As seen from Eqs. 4 0..Ol:2... 2 1.. (590) and (592). . 8 2] t=O:O. 2 O. fox 1= [0 .nd xlabelC1Time(secs)') ylabelC'Amplitude ' ) title( 'ClosedLoop Step' ) 1.. (588)..... den = [~ 2*1*w w.. nunt = [w.8 '2 The two roots can be expressed as SI.2 U.6 0.4 0.52 = SWn ± jw" ~ = a ± jw (591 ) where (592) and (593) The physical significance of ~ and a is now investigated.5~6 Transient Response of a Prototype SecondOrder System .2 0.B Time(secs:) (sec) 'I . 4 1. t) hold on.. 514 are obtained by the following sequence of MATLAB junctions clear all w=lO.~r~r.6 O.. Closed~LOop step 1.den... Toolbox 56.2 0."""'T""~T~.2 o.4 1 ... 8 1 1 .6 1.. a appears as the constant that is multiplied to t in the exponential term of y(t). step(num... 1 The corresponding time responses for Fig. A2].. 277 561 Damping Ratio and Damping Factor The effects of the system parameters ~ and Wn on the step response y(t) of the prototype second~order system can be studied by referring tu the roots of the characteristic equation in Eq.. t=O:O.e 'I .~~.6 .
a controls the rate of rise or decay of the unitstep response y(t). or the damping constant. When ~ '# 0. Fig. we see that critical damping occurs when ~ = 1. or ~ = cose (595) j(J) Root splane )(+r f) A cq. 1/ct. • CtJ is the imaginary part of the roots. (591). when t 0. From Eq. a controls the "damping" of the system and is called the damping factor. = w • ron is the radial distance from the roots to the origin of the splane. the imaginary part of the roots has the magnitude of w. (591). Under this condition.278 ~ Chapter 5. Time~Domain Analysis of Control Systems Therefore. When the two roots of the characteristic equation are real and equal. is proportional to the time constant of the system. Eq. As seen from Eq.2 Natural Undamped Frequency The parameter llJll is defined as the natural undamped frequency. Thus. For the purpose of reference. For the complexconjugate roots shown. ~.. that is. the damping factoris simply ct = Wn. W is sometimes defined as the conditional frequency. (590) shows that the unitstep response is purely sinusoidal.6. and 117 defined in Eq. .. we can regard ~ as the damping ratio. S = d ampmgratlo =  . In other words. . • a is the real part of the roots. 515 illustrates the relationships among the location of the characteristic equation roots and ct. or the damped frequency. (591) shows that. • ~ is the cosine of the angle between the radial line to the roots and the negative axis when the roots are in the lefthalf splane. (5 93) is not a frequency. ~~ Wm and llJ. Wn corresponds to the frequency ofthe undamped sinusoidal response. the roots of the characteristic equation are imaginary7 and Eq. (J X Root Figure 515 Relationships among the characteristicequation roots of the prototype second~order system and a. a Wn actual damping factor = ~~ damping factor at critical damping (594) 5. Therefore. the damping is zero. the response of yet) is not a periodic function. The inverse of a. we called the system critically damped. rum andw. when 0 < {< I.
Negative damping gives a response that grows in magnitude without bound~ and the system is unstable. plane corresponds to positive damping. and the system is marginally stable or marginally unstable. • The righthalf splane corresponds to negative damping. . Positive damping causes the unitstep response to settle to a constant final value in the steady state due to the negative exponent of exp( ~cvnt). (c) the constanta loci~ and (d) the constantcv loci.. that is. Thus. • The imaginary axis corresponds to zero damping (a = 0 or ~ = 0).. The system is stable. Fig. the damping factor or damping ratio is positive. Regions in the splane are identified with the system damping as follows: • The left~half s. (b) Constantdampingratio loci. splane jOJ splane Positive damping Negative damping ~{JJ2 a2 a) 0 a3 a . ~. (b) the constant~ loci.rut o Positive damping Negative damping a2> al>O a3<O (e) (d) Figure 516 (a) Constantnaturalundampedfrequency loci.2> 'I (b) (a) jm . 516 shows in the splane (a) the constantcon loci.. we have demonstrated with the help of the simple prototype secondorder system that the location of the characteristic equation roots plays an important role in the transient response of the system.56 Transient Response of a Prototype SecondOrder System j(J) ~ 279 splane Negative damping (J Positive damping (2 Negative damping . (d) Constantconditional"frequency loci. (c) Constantdampingfactor loci. Zero damping results in a sustained oscillation response.
Thus. that is.= I.2 '1'1.> I: 1. y( t) never exceeds its final value during the transient. The following classification of the system dynamics with respect to the value of ~ is made: 0«< J: 1. In practical applications.lVl1±jWIl~ = WII (l. 518 illustrates typical unitstep responses that correspond to the various root locations already shown.280 . only stable systems that correspond to > 0 are of interest. (J)1l is held constant while the damping ratio ~ is varied from 00 to +00..damped u1ldamped SI. t. (593) and (595). We can show that the quantity inside the square bracket in Eq. (596) can be reduced to sin wt.lV" <0) wzdel'damped critically damped 5). TimeMDomain Analysis of Control Systems splane (00 (f Figure 5 17 Locus of roots of thc characteristic equation of the prototype secondorder system.d = ~ ~sin(wt+8) t 1 . Thus~ dy(t) w esw"t [ . M The effect of the characteristic equation roots on the damping of the secondorder system is further illustrated by Fig. 514 gives the unitstep responses ofEq. Fig.lvil ± wn Jt. 517. (590) with respect to t and setting the result to zero.w" ± jw"Jl  t. respectively. S2 = t.2 (t.52 $). (590) plotted as functions of the normalized time wnt for various values of the damping ratio ~..=0: S2 52 = t. 518.1. the response becomes more oscillatory as ~ decreases in value. S2 = t.~2cos(wt +8) ~ ] t2:':O (596) where wand () are defined in Eqs. Chapter 5.<0: SI.2  1 ove. (596) is simplified to (597) . . When ~ ~ 1.>O) negatively damped Fig. In Fig. 517 and Fig. the step response does not exhibit any overshoot. Eq..so). = ±jw" 1.w. s 563 Maximum Overshoot The exact relation between the damping ratio and the amount of overshoot can be obtained by taking the derivative of Eq. As seen.
5·6 Transient Response of a Prototype SecondOrder System'" 281 jOJ s~plane y(t) o o jOJ splane o o y(t) o a x jro o y(t) splane o jOJ splane X 0 0> (>1 X jlJJ y(t) spJane s<l 0 C1 o Figure 5·18 Stepresponse comparison for various characteristicequationroot locations in the splane. .
.{ With reference to Fig. . the time at which it occurs is given by Eq. . that is. . (5101) 1 mr:r. 513. Whether the extremum is an overshoot or an undershoot.. The magnitudes of the overshoots and the undershoots can be determined by substituting Eq. n = 1. (599). It should be noted that. the time at which the maximum overshoot occurs is lmax= .... (5102) y(t) ~r.1 + (_1)11. 5. 2. .282 ~ Chapter 5.. nrr ~ n=O.{2 (599) The solution at t = 00 is the maximum of yet) only when t ~ 1. as shown in Fig..2.l y(tllmaxormin = I . Thus. Time~Domain Analysis of Control Systems Setting dy(t)/ dt to zero.2. . 1.. cunV 1 .e or ~ sin(lI1r + 9) II = I.e n = 1 2 ~ .I.. and the undershoots occur at even values of n.I. The result is mrt. we have the solutions: t = 00 and (598) CU1Z~t from which we get t= = lIn: n = 0.y (t )1 max ar mm .~ (5100) wnV I . although the unitstep response for t # 0 is not periodic. This corresponds to Il = 1 in Eq. (599).Ymax o Figure 519 Unitstep response illustrating that the maxima and minima occur at periodic intervals. . 519.. the first uvt:rshoot is the maximum overshoot. (590). . For the unitstep responses shown in Fig. 513. the overshoots occur at odd values of n. 3.JNi . the overshoots and the undershoots of the response do occur at periodic intervals./Vlt. (599) into Eq.
we would have to set yet) = 0. 1= e1Ct.Q '1 C 0 ~ 0 Q) \ > Cl) 60 40 20 ~ \ OJ) E.1 + 0. (5103) shows that the maximum overshoot of the step response of the prototype secondorder system is a function of only the damping ratio ~./JIi./"'.8 Dan1ping ratio. and then approximate the curve by a straight line or a curve over the range of 0 < ~ < I. the exact value can be determined directly from the responses of Fig. An easier way would be to plot cv"td versus ~. (5100) is a function of both ~ and W ll • 564 Delay Time and Rise Time It is more difficult to determine the exact analytical expressions of the delay time td. even for just the simple prototype secondorder system. The relationship between the percent maximum overshoot and the damping ratio given in Eq.5 in Eq. ~ \ ~ "" 0. The time t max in Eq.7~ ())" 0 < {< 1. for the delay time. The . From Fig. rise time I". (590) and solve for t.  1. and settling time (s·.125{ + 0.0 For the rise time I r.469{2 W II 0 < s< 1.6 0. For instance.2 (5104) Eq.0 1.~ (5103) percent maximum overshoot = 1OOe nl. (5102). which is the time for the step response to reach from 10 to 90% of its final value.2 Figure 520 Percent overshoot as a function of damping ratio for the step response of the prototype secondorder system. The maximum overshoot is obtained by letting n maximum overshoot and =: Ymax  =: 1 in Eq. as shown in Fig. Cl) I1J Q.0 We can obtain a better approximation by using a secondMorder equation for ttl: td~ 1.2 0. 514. 11.4 ~ ~ o 0. (5104) is plotted in Fig. 521. the delay time for the prototype secondorder system is approximated as td ~ 1 + O.5~6 Transient Response of a Prototype SecondOrder System ~ 283 100 ~ 80 . 520.. Therefore. 521..
0 r_~r_tti::Io!tF_tI 2...0 r_It~I7"'t+I 1.4 0....8 +2.5~ {J)n 0 <?:< 1 A better approximation can be obtained by using a secondorder equation: tr = 1 ._...~.0 FttActual W'ld+rr.2 1. 522...5 .8 1.0 I""~=+ o 0...2 0.4167~ + 2.5 I++I.4 0..8 { ~ 1. plot of wntr versus S is shown in Fig.II 2.6 0...~+_~~ 1.6 Figure 521 Nonnalized delay time versus for the prototype secondorder system.0 1.....0 rtj+t+h...... ..2 0...0 1..6 0.....0 r.. 4.2 ~ Figure 522 Normalized rise time versus s for the prototype secondMorder system....284 ~ Chapter 5.~I 1..0 Ifttti?I 3...4 1... TimeMDomain Analysis of Control Systems 2..___r____... In this case t the relation can again be approximated by a straight line over a limited range of C: tr = 0.0.___:r.917~2 Wn 0< ~< 1 (5108) 5.Il o 0.
95 and 1.69.95 and 1. when 0 < ~ < 0.< 0.69. and the response can enter the band between 0. 523(a) and (b) show the two different situations.05 for the last time from either the top or the bottom. the settling time has a 1 + _ 1_ ~ 1 + _ I _e(av / ~ O.05 1.69 1. the overshoot is less than 5%. will reduce (increase) 5~65 Settling Time From Fig.69 Figure 523 Settling time of the unitstep response. the following conclusions can be made on the rise time and delay time of the prototype secondorder system: • t. Thus. .' o~~L+ 1 _ _1 . and the response can enter the band between 0. Fig. (b) .. the unitstep response has a maximum overshoot greater than 5%.56 Transient Response of a Prototype SecondOrder System 285 From this discussion . we see that.05 only from the bottom.00 0.> 0... o t.JGfl (a) 0< . Wben ~ is greater than 0. 514. and td are proportional to ~ and inversely proportional to OJ" WI! ' • Increasing (decreasing) the natural undamped frequency tr and td.95 ..95 \:U~=~E===~~~~i~:=~~~~~~~~~=~=~ r .
We can obtain an approximation for (.286 ~ Chapter 5. as shown in Fig. t.~ In (c ts .... must satisfy the following condition: 1 ~ e ~Wnt.t={2) (5111 ) where C'S is the percentage set for the settling time. for 0 < {< 0. (5109) for wnt. Solving Eq. (5109) or Eq. 523(a) for a 5% requirement..05. the righthand side of Eq. (5110) would be 0...~ is difficult to obtain. = lower bound of unitstep response (5110) For the 5% requirement on settling time. for a 5percent settling time.95.69 by using the envelope of the damped sinusoid of y(t) .. if the threshold is 5 percent. Thus. In general~ when the settling time corresponds to an intersection with the upper envelope of y(t).. (5109) would be 1.. The exact analytical description of the settling time t.05. It is easily verified that the same result for t. the Cts = 0.. For example. and that of Eq. the following relation is obtained: 1+ R 1 I _ {2 e r. (5110). we have Wilts = . TimeDomain Analysis of Control Systems (c) Figure 523 (continued) discontinuity at { = 0.~ is obtained using either Eq.t.69. the righthand side of . = upper bound of unitstep response (5109) When the settling time corresponds to an intersection with the bottom envelope of y(t).
40 6.69.~.2 5..5 S 0040 0.64 0.10 1. We can approximate the settling time for the prototype secondorder system as t.33 30.35 2.69.90 1.21 8.2 (' 4.0 to.03 4.00 5.60 4.\ and the system parameters as follows~ • For < 0.5~ 6)fl ts =  s>O. (5112) and (5113) for their respective effective ranges.69 (5·1I2) {COil The approximation will be poor for small values of ( < 0. 4.15 3. > 0.10 0. the unitstep response will always enter the band between 0.3).80 0.00 1.69 0.69 (5113) Fig.7 13.4 5.92 4. along with the approximations using Eqs.71 4.~  3.0 7.75 4.2 O<{<O.62 0.20 0.05 from below.50 0.00 4.5·6 Transient Response of a Prototype SecondOrder System ~ 287 Eq.2 5.70 0.7 8. versus ~ for the prototype secondorder system described by Eq.5 5.50 4. We can summarize the relationshi ps between 1. is greater than 0. 514 that the value of wnts is almost directly proportional to ~. (587).64 3. (5111) varies between 3.7 10.69. The numerical values are shown in Table 52.16 5.. cunts S 0.05 4.73 5.16 5.69.71 4. the settling time is inversely proportional to ~ and (Vn' A practical way of reducing the settling time is to increase (On while holding constant. The following approximation is used for Is for c.95 5.86 3. 523(c) shows the actual values of W'lt.0 and 3.60 0.0 6.20 .50 28.32 as ~ varies from 0 to 0. Although s s TABLE 52 Comparison of Settling Times of Prototype SecondOrder System.30 Actual 3.00 4. When the dumping ratio c.65 0. We can show by observing the responses in Fig.68 0.2 16.20 1.3 5.50 6.95 and 1.
Keep in mind that while the definitions onYmax. the settling time is proportionallo { and inversely proportional ro {J)n' Again. [w. (587)..Ol:5. Amp/_ucla: 1.. More stringent design problems may require the system response to settle in any number less than 5%. 0. den = [12*1"'ww.288 ~ Chapter 5.4 ~ Time (sec): 0.5 2 2. and Is and.. However. For example clear all w=lO. the damping ratio ~ and the natural undamped frequency Wn strictly apply only to a secondorder system whose closedloop transfer function is given in Eq.. rise time. settling time should be used to measure how fast the step response settles to its final value. ts can be reduced by increasing (J)n' Toolbox 562 To find PO. the rise and delay times should be adequate to describe the response behavior.25 1. step(num. tmax .:::. It seems that.1\2]. ~ • For ~ > 0.6 ~ ~ 0. and Is apply to a system of any order.69 is truly a measure of how fast the step response rises to its final value.4..2 r~ xlabelC'Time(secs)') ylabel('Amplitude') title ( 'ClosedLoop Step') 4) .. t=O:O.5 3 Time(secs) (sec) 3. t r. 1\2] .2 o o I I I I 0.1=O.t) Closed.den.354 . these relationships can be used to measure the performance of higherorder systems that can be approximated by secondorder ones.6 r ..5 5 It should be commented that the settling time for ~ > 0. As the name implies. and settling time using MATLAB.5 1.69.Loop Step System: sys 1....5 4 4. the maximum overshoot depends only on and can be controlled independently.. under the stipulation that some of the higherorder poles can be neglected. TimeDomain Analysis of Control Systems the response will be more oscillatory. Idt tr. Naturally. It should also be pointed out that the 5% threshold is by no means a number cast in stone.4 0. the relationships among td.. point at a desired location on the graph and rightclick to display the x and y values.. nunt. t . and Wn are valid only for the same secondorder system model. i ~ g 0. for this case.
1 = h i n2 + 1m .57 Speed and Position Control of a DC Motor 289 57 SPEED AND POSITION CONTROL OF A DC MOTOR Servomechanisms are probably the most frequently encountered electrOInechanical control systems. refer to Chapter 4 for more details) speed sensor (usual1y a tachometer) gain it. . kgm? (refer to Chapter 4 for more details) gear ratio equivalent viscousfriction coefficient of the motor and load referred to the motor shaft. volt back emf. we discussed various issues associated with modeling of dc motors. The common characteristic of all such systems is that the variable to be controlled (usually position or velocity) is fed back to modify the command signal. Applications include robots (each joint in a robot requires a position servo). B must be scaled by n.tSellsor Figure 524 An armaturecontrolled de molor with a gear head and a load inertia h . and laser printers. Nm moment of inertia of the motor (motor shaft). h ] n B Gear head . ohm armature inductance.. Nm/rad/sec (in the presence of gear ratio. The servomechanism that win be used in the experiments in this chapter complises a dc motor and amplifier that are fed back the motor speed and position values. henry applied armature voltage. volt e T angular di splacement of the motor shaft. particularly the actuator. to name but a few. The system parameters include armature resistance. One of the key chaUenges in the design and implementation of a successful controller is obtaining an accurate model of the system components. where the field current is held constant in this system. 524. kg _m2 any external load torque considered as a disturbance. kgn? equivalent moment of inertia of the motor and load connected to the motorshaft. numerical control (NC) machines. In Chapter 4. We will briefly revisit the modeling aspects in this section. Nm moment of inertia of the load. 571 Speed Response and the Effects of Inductance and DisturbanceOpen Loop Response Consider the armaturecontrolled dc motor shown in Fig. radian torque developed by the motor.
. If the load inertia and the gear ratio arc incorporated into the system mode I. the load applied to a juice machine by the operator pushing in the fruit) as a disturbance torque TL . as discussed in Chapter 4.290 Chapter 5.) is the motor mechanical time constant. the speed of the motor shaft may be simplified to 'e Kill D(s) = or s+ KK+RBlI 1Il!J a RaJIIl R1I1m V (s)  s+ KKb + RBL a III 1m T (s) (5115) RaJI/I (5 116) where Kejf = KIII / (R aB + KmKb) is the motor gain constant.g. (5114) through (5116) is replaced with 1 (total inertia). we have included the effect of any possible external load (e. the armaturecontrolled dc motor is itself a feedback system. is neglected. The system may be arranged in inputoutput form such that Va(s) is the input and D(s) is the output: (5114) The ratio L ll / Ra is called the motor electrictime constant. where backemf voltage is proportional to the speed of the motor. and til/ = Ra1m /( RaB + KmKz. because La in the armature circuit is very small.(S) _1_ + Js+B Figure 525 Block diagram of an armaturecontrolled de motor. TimeDoma in Analysis of Control Systems V.Also. Thus. . the inertia 1m in Eqs. which makes the system speedresponse transfer function second order and is denoted by LC. 525. As shown in Fig. it introduces a zero to the disturbanceoutput transfer function. 525. However. resulting in the simplified transfer functions and the block diagram of the system. In Fig..
vu. Gear .Sens() r figure 526 Feedback control of an armaturecontrolled de motor with il load inertia.e.Ir. . A'Km / Ra > D. the final speed value becomes wf\' = A/Kb. If the load ineltia is incorporated into the system model . and hence for the motor to turn. we get (5117) To nnd the response w(t). the speed final value is w(t) = A! Kb. I \I a(t) r=::::1 rf ~ Feedback ~. from Eq. The block diagram of the system is also shown in Fig. 525)? 572 Speed Control of DC Motors: ClosedLoop Response As seen previously. note that the motor mechanical time constant rill is reflective of how fast the motor is capable of overcoming its own inertia 1m to reach a steady state or constant speed dictated by voltage Va. If we apply a constant load torque of magnitude D to the system (i. (5118). 527. the ou tput speed of the motor is highly dependant on the value of torque TL . the speed response from Eq. From D Eq. As rm increases. such that Va(s) = A/ s. (5119). at steady state. 526.57 Speed and Position Control of a DC Motor 291 Using superposition. Here the final value of w(t) is reduced by RaDI KIIlKb' A practical note is that the value of h = D may never exceed the motor stall torque. (5118) In this case. Does the stall torque of the motor affect the response and the steadystate response? In a realistic scenario. the approach to steady state takes longer. (5 118) will cbange to (5119) which clearly indicates that the disturbance h affects the final speed of the motor. The controller is composed of a sensor (usually a tachometer for speed applications) to sense the speed and an amplifier with gain K (propOltiona! control) in the configuration shown in Fig. you must measure motor speed using a sensor. which sets a limit on the magnitude of the torque TL .(t) I Amp. (5119) . How would the sensor affect the equations of the system (see Fig. the value of the stall torque can be fo und in the manufacturer's catalog.RKa111 ).. the speed of the motor is w'jiv = Klb (A . For a given motor. TL = Dis). head + ~~+~. For h = 0 (no disturbance and B = 0) and an applied voltage Va(t ) = A. we use superposition and nnd the response due to the individual inputs. From Eq. We can improve the speed perronnance of the motor by using a proportional feedback controller.
we have ®(s) = . The steadystate response in this case is (5122) where wf\' . assuming La = 0.fl(s) I s. speed control may reduce the effect of disturbance. armaturecontrolled dc motor. TimeDomain Analysis of Control Systems 1 Js+B Figure 527 Block diagram of a speedcontrol. How will that affect your equations? 5~7~3 Position Control The position response in the openloop case may be obtained by integrating the speed response. So. considering Fig. As in Section 571. the input to the control system is converted from voltage ViII to speed D.292 Chapter 5. as in Section 571. Then. 525. the output becomes TCD(l e . If the load inertia h is too large. For ease in comparison of input and output. The openloop transfer function is therefore 0(s) KI1l Va(s) = s(La1ms2 + (LaB + Ra1m)s + RaB + KIIlKb) (5123) .ill using the tachometer gain KtoHence.. we have KtKmK n(s) = s+ ( KmKb + RaB + KIKmK) Rolm 1 1m Rolm D. 00 . will the motor be able to turn? Again. A as K .. you will have to read the speedsensor voltage to measure speed. the reader should investigate what happens if the inertia h is included in this model.. Note that the speed at the motor shaft is sensed by the tachometer with a gain K/.tire) .e .i/!(S) (5120) s+ ( KmKh + RaB + KtKmK) Raim For a step input Tds) nin = W Al s and disturbance torque value h = Di s.lire) 1m (t) _ AKKmKl Tc (1 Ralm til  (5121) where Tc = Km Kb a t K K +~II~+K is the system mechanicaltime constant.
we set up numerical and experimental case studies to test and verify the preceding concepts and learn more about other practical issues. and.58 Time Domain Analysis of a PositionControl System ~ 293 _ 1_ fs+B ~~I+~ Figure 528 Block diagram of a positioncontrol. Tu cO/l{rol the position of the motor shaft. r e = (La / RlI) may be neglec ted for small LII • ® (s) ®I/I (s) (5126) Later. Recall from Chapter 4 that G(S) = 0 :.127) KsK\KjKN .(s) == ®e(S) s[LaJ/S2 + (Ra Jt + LaB/ + K 1K2Jt)S + RaBe + K.K2B/ + K. annaturecontrolled de motor. the simplest strategy is to use a proportional controller with gain K. The system is composed of an angular position sensor (usually an encoder or a potentiometer for position applications). The closedloop tran<.) (5. the time response in this case is (5124) which implies tbat the motor shaft is turning without stop at a constant steadystate speed. Alternatively. the output can be converted into voltage using the sensor gain value. where we have used the total inertia 1. as before.fer function in this case becomes (5125) where Ks is the sensor gain. Note that. we shall analyze the performance of a system using the timedomain criteria established in the preceding section. in Chapter 6. The purpose of the system considered here is to control the positions of the fins of an airplane as discussed in Example 411 . For small L". The block diagram of the closedloop system is shown in Fig. 528 . 58 TIMEDOMAIN ANALYSIS OF A POSITIONCONTROL SYSTEM In this section.Kb + KK1KtK. the input voltage can be scaled to a position input ®ill(S) so that the input and output have the same units and scale.l. for simplicity.
+ KiKb + KKIKiKt) s ( s+~~~~~~~~~ RaI.0002 (5129) Because the electrical time constant is much smaller than the mechanical time constant.2) Comparing Eq.0003 sec Jt La 0.::~:~~=:.003 (5 I28) M The mechanical time constant of the motorload system is 7:t = B. Later we will show that this is not the best way of approximating a highorder system by a low~order one. we get G s _ 4500K ( ) .015 = 0. we see that the natural undamped frequency run is proportional to the square root of the amplifier gain K.j4500K rad/sec RaJI + K}K2I . (586)" we have natural undamped frequency (... KsKIKiKN s[(RoJ.294 .:::=~=~~~ RaBt + KIK2B.KN = ±. TimeMDomain Analysis of Control Systems The system is of the third order. (5131) and (5132) with the prototype secondorder transfer function of Eq.Kb + KKIK.01333 sec 0. The forward~path transfer function is now G(s) :::. Chapter 5..s(s + 361. In this way. + KIK2 Jr)S + RoBt + Kr K 2B t + K.Un =± KsK1K.Kt 1 (5130) KsK} KiKN RaJI + KIK2 Jt = . it is informative to analyze the system petformance by applying the unitstep input with zero initial conditions. on account of the low inductance of the motor. The result is a secondorder approximation of the thirdorder system../K.. we can perform an initial approximation by neglecting the armature inductance La. whereas the damping ratio ~ is inversely proportional to .. = 0. (5132) Thus.. it is possible to characterize the . since the highestorder term in G(s) is? The electrical time constant of the amplifier~motor system is To = Ra + KIK2 = 5 + 5 = 0. + KIK21( Substituting the system parameters in the last equation. The closedloop transfer function of the unityfeedback control system is 8 y(s) Be(s) = 82 4500K + 361...2s + 4500K (5134) 581 UnitStep Transient Response For timedomain analysis.
. .9997e180. then @(s) = 1/ s.204Ie180.2s  4500K 4500K) ] (5135) The inverse Laplace transform of the righthand side ofthe last equation is carried out using the Laplace transform table in Appendix D. with zero initial conditions. Table 53 gives the comparison of the characteristics of the three unitstep responses for the three values of K used.00 1. is 1 [ ey(t) = [.6t . The output of the system.lsor + 150eI81. . if necessary.L _ _ _ r V r K=1.2): Oy(t) = (1 ~ eI80.03 0.60 1.181.2 K= 14.05 Figure 529 responses of the attitudecontrol system in Fig.707): = (1 e180. or using Eq.6t cos 884.O. Let the reference input be a unitstep function Br(t) = us(t) rad. delay time. (590) directly.02 Time (sec) G.2t)us(t) (5136) K = 14.6t sin 884. . such as rise time.:. The following results are obtained for the three values of K indicated.K.248 V 0.7t  O.5(~ = 0.0): Oy(t) = (1  I5le. G. 529.17(~ = 0. La = O.58 TimeDomain Analysis of a PositionControl System ~ 295 system performance in tenns of the maximum overshoot and some of the other measures. .00 0. When 2.01 Unit~step 0.5 0.:.20 f\ I.248«(~1.6t sin 180.04 4~18.6t cos Oy(t) K 180. and settling time. s(s2 + 361. V1\ I ~ .7t)us (t) (5138) The three responses are plotted as shown in Fig.40 {t V/ / )cV . K = 7.80 0.6t)us (t) (5137) =181.
.6 + '. the damping ratio is 0. limeDomain Analysis of Control Systems TABLE 53 Comparison of the Performance of the SecondOrder PositionControl System with lhe Gain K Values (f)n GainK 7.50 181. 139) (5140) S2 = 180. the roots are Sl = 180.5.6 + j884.2] num = [4500""k] .24808 14. is to establish methods so that the total reliance on computer simulation can be reduced. even with the aid of a computer. Indeed. (5135). The motivation behind this discussion is to show that the pelformance of some control systems can be predicted by investigating the roots of the characteristic equation of the system.2.2 0. it would be time consuming.3%.00929 0.3 52.0.00125 0. 14. den = [1 361. When K is set at 14.24808. one of the main objectives of studying control systems theory.707.jlSO.6 .6 = 180.0259 0.4500K Toolbox 581 The Fig.00 0 4.\"1 = SI S1 = 180. % Equation 5. end xlabelC 'Time(secs) ') ylabel('Amplitude') ti tle( I ClosedLoop Step t) For K = 7.0114 0.2.707 0.000 0.6 .7%. t is very close to 1.132616 . 248 14. and the system is almost critically damped.6 S2 S2 = 180.6 + j884.24808: K = 14. The unitstep response does not have any overshoot or oscillation.5 181.6 + jl 80. and the overshoot is 4.00369 K = 181.296 ~ Chapter 5. the roots of the characteristic equation are tabulated as follows: K = 7. When the val ue of K is set at 7. It should be pointed out that.2 4500*k] i step(num. the system is lightly damped.0150 0.00136 0.17 ~ (rad/sec) % Max overshoot ttl I J• (sec) (sec) l.62 255. using either the conventional or modern approach.248. in practice.01735 0.6 = 82 = 180.17.7 = 180. and the maximum overshoot is 52.44 903. and 181.00560 0.200 180. ~ = 0.136 % UnitStep Transient Response for k=[7.f (sec) lmax (sec) 1.0084 0. to compute the time response for each change of a system parameter for either analysis or design purposes.5: K 181.2: .5.32616  4500K (5. which is excessive. den) hold on.7 .0186 0. 529 responses may be obtained by the following sequence of MATLAB /Unctiolls. For the characteristic equation of Eq.
These roots are marked as shown in Fig.24808 K = 7. the natural undamped frequency will increase with When K is negative. 530 are summarized as follows: .2 j884.24808 7. (5134) as K varies. (5135) and are used extensively for the analysis and design of linear control systems.. we see that the two roots are real and negative for values of K between 0 and 7.24808.. The trajectories of the two characteristic equation roots when K varies continuously from 00 to 00 are also shown in Fig.75 ~ jIOOD 8 Figure 530 Root loci of the characteristic equation in Eq. 181.58 TimeDomain Analysis of a PositionControl System < 297 J{J) I ~ jlOOO K= 181. and the step response will have no overshoot for this range of K..24808. one of the roots is positive~ which corresponds to a time response that increases monotonically with time.K K<O j200 K"" a j40U j600 }800 K.. (5140) and (5141). 530. one positive and one negative «(' > 1) . For values of K greater than 7. These root trajectories are called the root loci (see Chapter 4) of Eq. The dynamic characteristics of the transient step response as determined from the root loci of Fig.6 ".JR. 530. This means that the system is overdamped.75 JROO sp)une j6(){) j400 j180.24808 < K < 00 oo<K<O Two negative distinct real roots Two negative equal real roots Two complexconjugate roots with negative real parts Two distinct real roots. From Eqs. and the system is unstable.2 j884. Amplifier Gain Dynamics Characteristic Equation Roots System Overdamped Critically damped (t = 1) Underdamped (s < I) Unstable system (~< 0) 0< K < 7.
6tsinI80. as given by Eq.(/) = tus(t}.0.(t . (5132) has a simple pole at s = 0. Coulomb friction is almost always present.18 + 4500K) 4500K ] (5142) which can be solved by using the Laplace transform table in Appendix C. = lim 4500K I . The output response of the system in Fig. 583 Time Response to a UnitRamp Input The control of position may be affected by the control of the profile of the output.18 1. respectively.. (5132) into Eq. rather than just by applying a step input. The unitstep responses in Fig. It may be necessary to investigate the ability of the positioncontrol system to follow a rampfunction input. In other words. Substituting Eq.248: 8y (t) = (t  0. the steadystate error of the system due to a step input.OO5536e. 479 is I [ s2(s2 8y (t) =£ + 361.838ge180t)us(t) (5145) K = 145: Oy(t) .005536 + O. the steperror constant is K . (5134) and (5133). (525).0. The ramp responses of the system for the three values of K are presented in the following equations. In the practical case. This means that the steadystate error of the system is zero for all positive values of K when the input is a step function.1RO. The result is (5143) where (5144) The values of ~ and Wn are given in Eqs. 9.2t + O. 529 verify this result.6t cus180. For a unitramp input. the system may be designed to follow a reference profile that represents the desired trajectory. the system is of type 1.5. The zerosteadystate condition is achieved because only viscous friction is considered in the simplified system model.6 t . so the steadystate positioning accuracy of the system can never be perfect.298 ~ Chapter 5. limeOomain Analvsis of Control Systems 582 The SteadyState Response Because the forwardpath transfer function in Eq.f+os(s + 361.01107 .6t)us(t) (5146) . (524). K = 7.8278e.467 X 107 e180. is zero.2) = oc (5141) Thus.
From Eq..02 1+7"~~~~ItI_I_I___I____1 o Figure 531 0.:..04 0.000443 + O..01 0.58 TimeDomain Analysis of a PositionControl System ell ~ 299 == 0.07 0.0.248) 0..... Kv = lim sG(s) = lim $ . A more direct method of determining the steadystate error due to a ramp input is to use the ramperror constant Kv.03 0. 478. (5144) is the transient response.09 0...2: Oy(t) = (t . 0.. The steadystate portion of the unitramp response is (5148) Thus.. K= 181. 0 $ ..2 = 12.. 531.0803 (5149) which is a constant. Notice that the steadystate error of the ramp response is not zero... 0 S 45~~~.' 0..00044 (K= 181. (5~31). the steadystate error of the system due to a unitramp input is ess==(Vn K 2?.02 0.6t sin884.04 !+1r+~~_r_.0.7t .5) 0.7t)us (t) (5147) These ramp responses are plotted as shown in Fig.05 Times (sec) 0.46K + (5150) .08 J+!+!+I+:JII~___:JII""'___:P""'__l e" = 0.000443e180..00104e180. The last tenn in Eq..1 Unitramp responses of the attitudecontrol system in Fig.0111 (K=7.08 0.6t cos884...06 0. La := O...2) QIO~~~~~~~~~~~~~~~~~~~~~~~~lJe"=QOO~ ) J\ (K= 14...
(5149). This phenomenon is rather typical in all control systems. plot(t. 5. is always stable for all positive values of K. if we attempt to improve the steadystate accuracy of the system due to ramp inputs by increasing the value of K.y. Chapter 5.000) s(s + 400. elden = [1 361. the transient and the steady~state error can be improved simultaneously. 248 14.300 ~.3s2 1.5 X 107 K 1. For higherorder systems. Let us investigate the performance of the positioncontrol system with the armature inductance La ..5 X 107 K = :~:::. the system can become unstable.3s + 1.204. 1 0.003 H. . (5128) becomes 1. 531 responses are obtained by the following sequence of MATLAB functions for k=[7.2] elnum = [4500~'tk] . hold on. which corresponds to a damping ratio of 0. more appropriately.5. 14.s(s2 + 3408.OOOl:0.707. =tj end title( 'Unitramp responses ') xlabelC'Time(see)') ylabel('Amplitude') The result in Eq. the steadystate error is 0.~ v which agrees with the result in Eq.5 181. 0. Thus.0803 (5151) Toolbox 5. t=O:O.204.5 X 107K + 1. all secondorder systems with positive coefficients in the characteristic equations are stable. (5~ 151) shows that the steadystate error is inversely proportional to K.2 4500*k] . if the loop gain of the system is too high. 82 The Fig.26)(8 + 3008) G(s}  (5152) The closedloop transfer function is 8 y(s) E>r(s) = s3 + 340S.55% of the rampinput magnitude. The forwardpath transfer function of Eq.u.: : : 0. in general.0055 rad or.. by using the controller in the system loop.3.OOOs+ 1. 84 Time Response of a ThirdOrder System In the preceding section. Apparently.t).elden. the transient step response will become more oscillatory and have a higher overshoot.5 x l07K (5153) .u). TimeDomain Analysis of Control Systems Thus. we have shown that the prototype secondorder system. the steadystate error is e. It is not difficult to prove that. U [Yrx]=lsirn(elnum. obtained by neglecting the armature inductance.t.u=K :::. For K :::.
33 == 186. . whereas the thirdorder system has three distinct real roots. in general. Furthermore. we see that the justification and accuracy of the secondorder approximation diminish as the value of K is increased.8.. which is much smaller than the value of 0.2 for the secondorder system.8 decays rapidly.8 3035.87 X 10 156.428.6 83 3021. the two roots that dominate the transient response correspond to a damping ratio of 0. The roots that are closer to the imaginary axis will dominate the transient response.0633. However. that the fact that the secondorder approximation is justified for K = 14.s(s + 10. 532 illustrates the root loci of the thirdorder characteristic equation of Eq.. Th~ root at 3021.5 does not mean that the approximation is valid for all values of K.53 . which is more than 13 times faster than the next fastest time constant because of the pole at 230.000s + 1.248: = 14.28e230.OO45e3021.8 corresponds to a time constant of 0. when K = 7.204. (5154) as K varies. we see that.5: K = 181. and these are defined as the dominant roots of the characteristic equation or of the system.33)(s + 3021. The output transient response is dominated by the two roots at 156.28e156.49  j906.33t  0. the secondorder system is critically damped.3 Comparing these results with those of the approximating secondorder system.3s2 + 1. the two complexconjugate roots of the thirdorder system again dominate the transient response. and the pole can be neglected from the transient standpoint.707. and the equivalent damping ratio due to the two roots is only 0. the secondorder system has a damping ratio of 0.2It + 2. decays to zero very rapidly.33. When K = 181.5.jl92 S3 = S3 = s] = 57.33 millisecond. This analysis is verified by writing the transformed output response as e (s) _ y .3 + 3408.~(t) (5156) The last term in Eq.697.49 + j906.2: 8) s] = 156. which is due to the root at 3021. It should be noted.6 == 230.81)u.5. the transient response due to the pole at 3021.33. recall that the damping ratio is strictly not defined. Thus.5 x 107 K = 0 (5154) Transient Response The roots of the characteristic equation are tabulated for the three val ues of K used earlier for the secondorder system: K K = 7. Thus.58 TimeDomain Analysis of a PositionControl System < 301 The system is now of the third order. for K = 14. the magnitude of the term at t = 0 is very small compared to the other two transient terms. Thus.8 is negligible.. For the thirdorder system. the secondorder approximation by setting La to zero is not a bad one. This simply demonstrates that.21 and 230.21)(8 + 230.2 = 3293. Fig.8) 7 (5155) Taking the inverse Laplace transform of the last equation. and the system is slightly overdamp~u.53 + j192 = 57.2. the contri bution of roots that lie relatively far to the left in the splane to the transient response will be small.21 S2 S2 S2 = 186. (5156). and the characteristic equation is . When K = 14. we get Oy(t) = (1  3. because the real and imaginary parts of the two characteristic equation roots are identical. because the effect on transient of the root at 3021. however.
andjl097.8 K = 7. the real root at 3293. 532.jI097. TimeDomain Analysis of Control Systems splane j600 186.53 + j192 K= 14.3)(s2 + 1.57. When K = 181.75.49 ± j906.57 jl400 Figure 532 Root loci of the thirdorder attitudecontrol system.}1097.3. .2.3. .3 K= 273.6 ± j884.248 j200 j400 j600 j800 jIOOa . By using the RouthHurwitz criterion.5 j400 J200 K=O 0 3293.3 still contributes little to the transient response.204 x 106 ) (5157) The roots of the characteristic equation areats = 3408. the closedloop transfer function becomes ey(s) 1.3 K = 181.302 ~ Chapter 5.3.6 are much closer to the jwaxis than those of the secondorder system for the same K. which are at 180. This explains why the thirdorder system is a great deal less stable than the secondorder system when K = 181.2. These points are shown on the root loci in Fig. With this critical value of K. the marginal value of K for stability is found to be 273. but the two complexconjugate roots at 57.0872 x 108 ares) = (s + 3408.2 K<O Koo a 3021.
den= [13408. ..5~8 Time·Domain Analysis of a Position. t=O:0. and the system is unstable.05 Figurs 5·33 Unitstep responses of the thirdorder attitudecontrol system.57 is (5158) By(t) = [1  O.2 273.3t + 72.<"'. = Toolbox 583 The root locus plot in Fig..5' r. 57 I J 4""K=rr r (I 1. we see that the thirdorder system is capable of being unstable.den) hold on.1. whereas the secondorder system obtained with La =: 0 is stable for all finite positive values of K. Ititttl+'' . and the system is said to be marginally stable.04 v 0.Ift v 0. Thus.05j rlocus(nwo. 529. Fig..02 v Time (sec) v 0.k.L/ It~+It+Itlll.20 ti11H\ 0.60 1+"'H+II~II+IIJI~HIllIlI1I+I+III Ir\K=IB~ 1. 16°)]us(t) The steadystate response is an undamped sinusoid with a frequency of 1097.H .57.V + \I.00 0..952sin(1097.2 are quite different. The unitstep response of the system when K = 273. When K is greater than 273. 5·32 is obtained by the following MATLAB commallciv fork=[7..ft 1/ .01 v 0.3t .5* (10"7) *k] .3 rad/sec.\f v\K(0248 li~ I \ J :\ l~ I '\ '\.tt+ff/~~++fY\ttfrit\t~tt!...248 14. K=.001:0... 5*(lO"7)*k]. I .0.. the two complexconjugate roots will have positive real parts~ the sinusoidal component of the time response will increase with time...80 1f1IlIf/f1.312040001. The responses for K = 7.248 and K 14.57] num = [1.5 are very close to those of the secondorder system with the same values of K that are shown in Fig. the two responses for K == 181. However.094e3408. I~ 1\'"'"rnr.00 Yn"f\""r"""''h. . Control System.5181.. 14. 303 2. 533 shows the unitstep responses of the thirdorder system for the three values of K used. ~ 1// 'v 1 v . end .. .03 \ 0. .
'. the inductance of the motor does not affect the steadystate performance of the system. (5152).~. (5150). the effects on the transient response relative to the location of the roots of the characteristic equation are demonstrated. The value of Kv is still the same as that given in Eq. A good engineer should always try to interpret the analytical results with the physical system. Specifically. amplifiers.. since La affects only the rate of change and not the final value of the motor current. Intuitively. which stands for proportional. Therefore. because the control signal at the output of the controller is simply related to the input of the controller by a proportional constant. Thus. we can consider a more general continuousdata controller to be one that contains such components as adders or summers (addition or subtraction).rr. 59 BASIC CONTROL SYSTEMS AND EFFECTS OF ADDING POLES AND ZEROS TO TRANSFER FUNCTIONS The positioncontrol system discussed in the preceding section reveals important properties of the time responses of typical second.304 Chapte r 5. TimeDoma in Analysis of Control Systems 4000 . attenuators.. we see that.see Section 433 and Chapter 9 for more details. in addition to the proportional operation. and integrators .' 3000 2000 1000 en 0' ~4~*~~~4 . In all previous examples of control systems we have discussed thus far. integral. For example. This type of control action is formally known as proportional control.and thirdorder closedloop systems. differentiators. the controller has been typically a simple amplifier with a constant gain K. when the inductance is restored. provided that the system is stable. ! Ii V i / '" E 1000 2000 3000 \i ~'" 4000~~~~~~~~~~:~~ 4000 3500 3000 2500 2000 1500 1000 500 0 5QO Real Axis SteadyState Response From Eq. one of the bcstknown controllers used in practice is the PID controller.. the thirdorder system is still of type 1. The integral and derivative components of the PID controller have .f. This is expected. and derivative. one should also be able to use the derivative or integral of the input signal .
(Un = 1. 1. so its effect is small except when the value of K is relatively large. particularly the stabilitythe zeros of the transfer function are also important.26. because of the specific value of the inductance chosen. (5131). and their applications require an understanding of the basics of these elements. (5160): ~ = 1..~ o 5 10 15 20 Time (sec) Figure 534 Unit~step responses of the system with the closedloop transfer function in Eq. . 2. we see that the effect of the motor inductance is equivalent to adding a pole at s = 3008 to the forwardpath transfer function of Eq.59 Basic Control Systems and Effects of Adding Poles and Zeros ~ 305 individual performance implications. When the motor inductance is restored.91 Addition of a Pole to the ForwardPath Transfer Function: UnityFeedback Systems For the positioncontrol system described in Section 58. 534. and the proportional constant is also increased. when the motor inductance is neglected. In this section. and 5. 5. (5131) and (5149). and T/1 = 0. the system is of the third order. (5149). what these controllers do is add additional poles and zeros to the open. We show thatalthough the roots of the characteristic equation of the system. As shown by the rootloci diagrams of Fig.26. (5131) while shifting the pole at 361. All in all. we show that the addition of poles and zeros to forwardpath and closedloop transfer functions has varying effects on the transient response of the closedloop system. and the forwardpath transfer function is given in Eq. The apparent effect of adding a pole to the forwardpath transfer function is that the thirdorder system can now become unstable if the value of the amplifier gain K exceeds 273. the system is of the second order. and the forwardpath transfer function is of the prototype given in Eq.0r~~_. Comparing the two transfer functions ofEqs. it is important to appreciate the effects of adding poles and zeros to a transfer function first.2 to 400. Actually. As a result.57. the additional pole of the thirdorder system is far to the left of the pole at 400. the addition of poles and zeros andlor cancellation of undesirable poles and zeros of the transfer function often are necessary in achieving satisfactory timedomain performance of control systems. Thus. 532 alld Fig.or closedloop transfer function of the overall system.2. which are the poles of the closedloop transfer function~ affect the transient response of linear timeinvariant control systems. the new pole of G(s) at s = 3008 essentially "pushes" and "bends" the complexconjugate portion of the root loci of the seconuorder system toward the righthalf splane.
These responses again show that the addition of a pole to the forwardpath transfer junction generally has the effect of increasing the maximum overs/zoot of the closedloop !Jystem. because the additional pole has the effect of reducing the bandwidth (see Chapter 8) of the system.001:20. t=0:0. step(num.2 0. and 5. 537 are obtained by the following sequence of MATLAB functions clear all w=l . Toolbox 59. thus cutting out the highfrequency components of the signal transmitted through the system. den = [Tp 1. the pole at 1 IT p moves closer to the origin in the . 1 The corresponding responses jar Fig. consider the transfer function (5159) The pole at s = 1 ITp is considered to be added to the prototype second~order transfer function.t) i hold on. TimeDomain Analysis of Control Systems To study the general effect of the addition of a pole. This is not surprising. As the value of ~) increases.den. for Tp=[O 1 2 5J .+2 *1 *w*Tp 2 *1 *w wA2 J . The closedloop transfer function is written M(s) = Y(s) R(s) = G(s) 1 + G(s) = Tps3 + (1 + 2~cvnTp)S2 + 2~(vns + WTt w~ (5160) Fig.y(t)') title( 'Unitstep responses of the system') The corresponding responses for Fig. 1=0 .vplane. .OOl:20. ~ = 1. These responses also show that the added pole increases the rise time of the step response. and its relative location. for Tp=[O 0.306 ~ Chapter 5. 25 . 1=1. t=O:0.6671]. num= [w]. and the maximum overshoot increases. den = [~p 1+2*1~\w*Tp 2*1*w wA2J . 2. 534 are obtained by the following sequence of MATLAB functions clear all w=l. 534 illustrates the unitstep responses of the closedloop system when CVIl = 1. and T p = O~ 1. end xlabel('Time(secs)') ylabel('apos. to a forwardpath transfer function of a unityfeedback system.. num= [w].
.0. and 1.2. 535. they control the transient response of the system directly. o 5 10 Time (sec) 15 20 Figure 5~35 Unitstep responses of the system with the closedloop transfer function in Eq. and Tp = 0. and Tp = 0. . In this case.0. Consider the closedloop transfer function (5161) where the term (1 + Tps) is added to a prototype secondorder transfer function..25. and the system is unstable. end xlabelC'TimeCsecs)1) ylabe1. Fig.. 307 step(num.( 'yet) ') title( 'Unitstep responses of the system') The same conclusion can be drawn from the unitstep responses of Fig. (5160): {= 0.5. 536 illustrates the unitstep response of the system with Wn = 1. t = 0.667 t the amplitude of the unitstep response increases with time. ::.0100. hold on.0~ and 4.O~ 2.2.0..0.25.5~ l.den. and 1.++#+. As the pole at s = 1 IT p is moved toward the origin 2.667. Cl)n = 1. and T p = O~ O.0. which are obtained with WIJ = 1.t).0. when ~J is greater than 0. 5~9~2 Addition of a Pole to the ClosedLoop Transfer Function Because the poles of the closedloop transfer function are roots of the characteristic equation.0.. S = 0..+1 .667.0.59 Basic Control Systems and Effects of Adding Poles and Zeros .
for Tp= [0 0 .80 Tp=O 1.50 12.Tp 0. Thus.00 13.00 10.00 Figure 5·36 Unitstep responses of the system with the closedloop transfer function in Eq.50 Time (sec) 9.00 'C" 1:: 0. 537 shows the unitstep responses of the c1osed~loop system with the transfer function (5162) . den = conv( [1 2*l*w wA 2] .50 6. Toolbox 592 The corresponding responses for Fig. adding a pole to the closedloop transfer function has just the opposite effect to that of adding a pole to the forwardpath transfer function.80 05 r.OOl:15.0 /'" Y. hold on. and 4.t).=4  ~ ~ 1.308 .. [Tp~]).l=O.2.40 J~ V 0. (rJn =: 1. == 0. r. end xlabel('Time(secs)') ylabel( 'yet) ') title( 'Unitstep responses of the system f) 5~9~3 Addition of a Zero to the Closed~Loop Transfer Function Fig.20 1.50 3. step(num.50 15. in the s~plane.0.1.den. Time~Domajn Analysis of Control Systems 1.5. ~~ ~ ~~ ~ ~.00 4. and T p == 0. Chapter 5.0.5.. num = [w"2 J .Tp =1 ."'/ II /T/ I //1 V / IIV// =l. the rise time increases and the maximum overshoot decreases.0. t=O:O. clear all 5~36 are obtained by the following sequence of MATLAB functions w=l.00 tJ Y /. as far as the overshoot is concerned. 5 1 2] .S. (5161): .00 7.0.
00 7..50 15.\ .80 l . .. according to Eq.) _ 6(1 + Tzs) ~ s(s + l)(s + 2) .2.50 6. In fact. we see that adding a zero to the closedloop transfer function decreases the rise time and increases the maximum overshoot of the step response. ~ = 0. 1.1.. (5~162): Tz = 0.+ ..6.4 ..r"""''r'""' 4... reduces the rise time and increases the maximum overshoot.3....00 Figure 537 Unitstep responses of the system with the closedloop transfer function in Eq.... (5164). the total unitstep response is dYI (t) = YI (t) + TZ"d[ y(t) (5164) Fig.. and 10..00 ~ 309 r~'". 594 Addition of a Zero to the ForwardPath Transfer Function: UnityFeedback Systems Let us consider that a zero at .50 Time (sec) 9..5. (5163) be Yl(t).60 111++++++++1++1 2..\ . the maximum overshoot also approaches infinity.I 3.00 4. In this case.00 1..50 12.4 . so G y (.59 Basic Control Systems and Effects of Adding Poles and Zeros 6. ..... Then.\ ..6.00 13. 1.. and 10.+ . and yet the system is still stable as long as the overshoot is finite and ~ is positive.:. (5162) as (5163) For a unitstep input.! f .00 to..2. where ron . 0..I ITz is added to the forwardpath transfer function of a thirdorder system. as T= approaches infinity..\ ..3.4 .40 0. let the output response that corresponds to the first tenn of the right side of Eq. We can analyze the general case by writing Eq..+ ... and Tz :::. 538 shows why the addition of the zero at s = 1 IT.50 3.
or reducing the maximum overshoot. When Tz 0. the closedloop system is on the verge of becoming unstable. but the denominator of M(s) also contains T.0.== 0. The term (1 + Tzs) in the numerator of M(s) increases the maximum overshoot. which has the effect of improving damping. and 10. Notice that. 0..0. Fig.0. .:. in the present case. 0.2. 0.5.. when T. 2.0. TimeDomain AnalysIs of Control Systems Time (sec) Figure 538 Unitstep responses showing the effect of adding a zero to the closedloop transfer function.5 Time (sec) 6 7 8 9 10 Figure 539 Unitstep responses of the system with the closedloop transfer function in Eq.0. 539 illustrates the unitstep responses when Tz = 0. but T: appears in the coefficient of the s term in the denominator. the maximum overshoots are reduced.2. 5.2 and 0.5. 2. (5166): T. The closedloop transfer function is (5166) The difference between this case and that of adding a zero to the closedloop transfer function is that. and 10. mainly = = o 2 3 4 . 0. 0. not only the term (1 + Tzs) appears in the numerator of M(s).310 ~ Chapter 5. 5.5.0.
den. For all practical purposes. We intentionally do not assign specific values to the coordinates. the (1 + Tzs) term in the numerator becomes more dominant. as shown in Fig. although the damping is still further improved. The question is: How large a pole is considered to be really large'? It has been recognized in practice and in the literature that if the magnitude of the real part of a pole is at least 5 to 10 times that of a dominant pole or a pair of complex . As T:: increases beyond 2. it would be useful to establish guidelines on the approximation of high~order systems by lowerorder ones insofar as the transient response is concemed. step(num. The poles that are close to the imaginary axis in the left~half splane give rise to transient responses that will decay relatively slowly. although the characteristic equation roots are generally used to study the relative damping and relative stability of linear control systems.0.Si for Tz= [0 0 . Toolbox 593 The corresponding responses for Fig. den= [13 2+6*Tz 6]. nurn= [6~\'Tz 6]. we can divide the splane intu regions in which the dominant and insignificant poles can lie. so the maximum overshoot actually becomes greater as T= is increased further.510 Oominant Poles al1d Zeros of Transfer Functions'" 311 because of the improved damping. hold on. In design~ we can use the dominant poles to control the dynamic performance of the system. it is important to sort out the poles that have a dominant effect on the transient response and call these the dominant poles_ Because most control systems in practice are of orders higher than two. the zeros of the transfer function should not be overlooked in their effects on the transient performance of the system.~ 510 DOMINANT POLES AND ZEROS OF TRANSFER FUNCTIONS From the discussions given in the preceding sections. end xlabel('Time(secs)') ylabel( 'y(t)') title( 'Unitstep responses of the system') . t=O:O_OOl:15.l=O.t) . For analysis and design purposes. since these are all relative to a given system. it becomes apparent that the location of the poles and zeros of a transfer function in the splane greatly affects the transient response of the system. 2 0 . The distance D between the dominant region and the least significant region shown in Fig_ 540 will be subject to discussion. whereas the insignificant poles are lIsed for the purpose of ensuring that the controller transfer function can be realized by physical components. An important finding from these discussions is that. 5 3 5] . 5~40. whereas the poles that are jar away from the axis (relative to the dominant poles) correspond to fast~decaying time responses. 539 are obtained by the following sequence of MATLAB functions clear all w=l.
540 are selected merely for the definitions of dominant and insignificant poles. the dominant poles and the insignificant poles should most likely be located in the timed regions in Fig. we do not show any absolute coordinates. . while designing. 541. such as in poleplacement design. then the pole may be regarded as insignificant insofar as the transient response is concemed. we cannot place the insignificant poles arbitrarily far to the left in the splane or these may require unrealistic system parameter values when the pencilandpaper design is implemented by physical components.312 Chapter 5.plane for design purposes.707. We must point out that the regions shown in Fig. whereas the zeros that ru. For design purposes. Again. dominant poles. e jar away from the axis (relative to the dominant poles) have a smaller effect on the time response. It should also be noted that. except that the desired region of the dominant poles is centered around the line that corresponds to S = 0 . TimeDomain Analysis of Control Systems jOJ splane Regiooof insignificant poles Region of dominant poles Unstable region o D (J Unstable region Figure 540 Regions of dominant and insignificant poles in the splane. The zeros that are close to the imaginary axis in the lefthalf splane affect the transient responses more significantly. jOJ sp lane Region of insignificant poles Unstable region "" / I+D~ 0 Unstable / region Figure 541 Regions of dominant and insignificant poles in the s.
which are at 1 ± jL We can refer to the relative damping ratio of the system as 0. 2.5~10 Dominant Poles and Zeros of Transfer Functions ·~'i 313 5~10~1 Summary of Effects of Poles and Zeros Based on previous observations.R(s) . the faster the 4.. The farther to the left in the splane the system's dominant poles are. When a pole and zero of a system transfer function nearly cancel each other. a sports car can accelerate faster. we can no longer strictly use the damping ratio ~ and the natural undamped frequency (J)lh which are defined for the prototype secondorder systems. 5. the portion of the system response associated with the pole will have a small magnitude. Transients due to those poles~ which are farther to the left. the steadystate performance must also be considered. loop transfer function lead to a step response that is underdamped.State Response Thus far. the more expensive it will be and the larger its internal signals will be. Let us consider the transfer function in Eq. it is obvious that striking a nail harder with a hammer drives the nail in faster but requires more energy per strike. However. (5167) as M(s) _ 20 .($ + 10)(82 + 2s + 2) (5167) The pole at s = 10 is 10 times the real part of the complex conjugate poles.n be justified analytically. we have provided guidelines for neglecting insignificant poles of a transfer function from the standpoint of the transient response. 5~10~2 The Relative Damping Ratio When a system is higher than the second order.707. While this ca. If all system poles are reat the step response is overdamped. we can summarize the following: 1. For example. we should first express Eq. decay faster. To do this. The response of a system is dominated by those poles closest to the origin in the splane. However.10(s/]0 + 1)(s2 + 2s + 2) (5168) . and the damping ralio in this case is referred to as the relative damping ratio of the system. The farther to the left in the splane the system's dominant poles are. then we can stilI use ~ and CUll to indicate the dynamics of the transient response. Similarly. Complexconjugate poles of the closed. the pole at 10 can be neglected from the transient standpoint.. However. zeros of the closedloop transfer function may cause overshoot even if the system is overdamped. going through with the mechanics. system will respond and the greater its bandwidth will be. but it uses more fuel than an average car. consider the closedloop transfer function M s _ Yes) _ 20 ( ) . if the system dynamics can be accurately represented by a pair of complexconjugate dominant poles. (5~167). 3. 5~10~3 The Proper Way of Neglecting the Insignificant Poles with Consideration of the Steady.
we look at controllers that include derivative or integral of the input signal in addition to the proportional operation. TimeDomain Analysis of Control Systems Then we reason that Is / I 01 « I when the absolute value of s is much smaller than 10. (5168) is approximated by M(s) £:<: 20 1O(s2 + 25 + 2) (5169) This way. (5167) and the secondorder system approximated by Eg.314 Chapter 5. So far in this chapter. . In other words. The term silO can be neglected when compared with 1. if we simply throwaway the term (s + 10) in Eq. the steadystate performance of the thirdorder system will not be affected by the approximation. (5 167). EXAMPLE 5111 Fig.(s) = (u~( Kp + Kos) .172) R(s) £ (s ) Kp U(s) W/~ s(s yes) + KoS + + 2sWII ) Gis) + GrCs) Figure 542 Control system with PD controller.171 ) In this case. f s(s + 2~w. In this section. (5] 69) all have a final value of unity when a unitstep input is applied. On the other hand. the approximating secondorder system will have a steadystate value of 5 when a unitstep input is applied. we have discussed the effect of adding a simple gain in the time responsei . ) (5. the forwardpath transfer func tion of the compensated system is G(s) = Y(s) E(s) = Gc(s)G . Then. 542 shows the block diagram of a feedback control system that arbitrarily has a secondorder prototype process with the transfer func tion G . because of the dominant nature of the complex poles. Eg.. 511 BASIC CONTROL SYSTEMS UTILIZING ADDITION OF POLES AND ZEROS In practice we can control the response of a system by adding poles and zeros or a simple amplifier with a constant gain K to its transfer function . (~) = I ' (j} s(s + 2~wll) ]I (5170) The series controller in this case is a proportionalderivative (PD) type with the transfer function GJ~) = Kp + KJ)s (5..rdorder system described by Eg. the thi.e. proportional control.
(5175) no longer represents a prototype s. = Kp / KD. The characteristic equation is written as = (5l77) Ignoring the zero of the transfer function in equation (S177) and comparing (5177) to prototype secondorder system characteristic equation 52 + 2?. It turns out that for this secondorder system. 2. and the closedloop system will not have any overshoot. and the settling time. 543 shows the unitstep responses of the closedloop system with Kp = 8 and KD = 3.s(s + 2) Rewriting the transfer function of the PD controller as 2 (5173) Gc(s) == (Kp + KDS) (5174) the forwardpath transfer function of the system becomes G(s) = Y(s) E(s) The closedloop transfer function is = 2(Kp + KDS) s(s + 2) Yes) R(s) 2{Kp + KDS) = s2 + (2 + 2KD)S + 2Kp (5176) Eq. In the present case. the zero at s = . Fig. as Kn increases. Consider the secondorder model G(s). Increasing the damping term.wns + w~ I +Ko J2Kp = 0 (5178) we get the damping I·atia and natural frequency values of {= (J)n = J2Kp which clearly show the positive effect of Kf) on damping.414. in either case. for ~ I.Kp / K[) to the closedloop transfer function.vely cancel the pole of G(s) at s O.Kp / K[) of the closedloop transfer function has a smaller impact on the response of the system. (5179) gives KD = 3. ~ = I. which implies the zero at s = . however. Adding a zero at s = . With the PO control. fior Kp = 8. We should quickly point out that Eq. In general.511 Basic Control Systems Utilizing Addition of Poles and Zeros ~ 315 which shows that the PD control is equivalent to adding a simple zero at s == Kp / KD to the forward~ path transfer function. since the transient response is also affected by the zero of the transfer function at :. although KD is chosen for critical damping. if we wish to have critical damping. (5176) shows that the effects of the PD controller are the following: 1. (5179) gives KD = 0. and the overall response is similar to that of a prototype second~order system. the transfer function in Eq. from 2 to 2 + 2KD. Eq. for higherorder systems. the rise time. Fig. . upon increasing K D• tlre genet·at conclusion is tltat the PD controller decreases the = rnaximum ol'ershoot. (5175) approaches that ofa firstorder system with the pole ats = 2.Kp I Kn may increase the overshoot when KIJ becomes very large. Upon selecting a smaller Kp = 1. as the value of Kn increases.Kp / Ko of the closedloop transfer function. Thus. the overshoot is due to the zero at s = .econd~order system. 5·43 shows a critically damped unitstep response in this case. the maximum overshoot is 2%. Eq. the zero will move very close to the origin and effecti. However. which is the coefficient ofthe s term in the denominator.
2 0.4r.t).6 0. ) I title( Unitstep responses of the system') I :'7.8 0. Because the PD controller does not change the system type. nurn = [2*.5 5 Figure 5 43 Unitstep response of Eq. it may not fulfill the compensation objectives in many situations involving steadystate error. step(num. % KP=4 and KD=3 den= [12+2*316]. The integral part of the PID controller produces a signal that is proportional to .~. an integral controller may be used. 543 are obtained by the following sequence of MATLAB junctions clear all t=O:O. TimeMDomain Analysis of Control Systems UnitMstep responses of the system 1.den.5 Til11e(secs) (sec) 3 3._~~~~~~~~ 1.5 4 4.112 We saw in the previous example that the PD controller can improve the damping and rise time of a control system. EXAMPLE 5.4142].5 M 1.4142] i % KP=1 and KD=O.316 ~ Chapter 5..2 0. Toolbox 5111 The corresponding responses for Fig. (5176) for two sets of KD and Kp values.5 2 2.den.t) : hold on. 414 den= [12+2*.4 0. For this purpose. step(num..OOl:5. xlabel('Time(secs)') ylabel( yet) . nurn = [2 * 3 16] .
the PI control reduces it to zero (provided that the compensated system remains stable) . ihat is. both the damping and the steadystate errol' can be improved. h may seem that PI control w111 improve the steadystate error at the expense of stability.WII (5. because the system is now of the third order. 544.181 ) Clearly.(s) is selected properly. ilmGy be less stable than the origi. 2.(s) = Kp + K. (5180) is shown in Fig.511 Basic Control Systems Utilizing Addition of Poles and Zeros 317 R(s) £(s) Kp + KI K) + V(s) s(s W. slrould be re/ativezv small.nal secondorder system or even become unstable if the parameters Kp and K.. the time integral of the input of the contro ller. Thus.~ + 2~WI/) GII(s) yes) s GJ'J Figure 544 Control system with PI controller. are not properly chosen. Because the PI controller is essentially a lowpass filter. Fig. In the case of a type 0 system with a PD control . the steadystate enor of the original system is improved by one order. Adding a pole at s = 0 to the forwardpath transfer function . A viahle method of de~'ignirtg the PI control is IV select the zeru at s = . with the forwardpath transfer function in Eq . the steadystate error due to rI step input is rllways zerO if the system is stable.('\') Op(s) = (o~(Kps + K[) 2( 2 ) s s+ I. The polezero configuration of the PI controller in Eg. When a type 0 system is converted to type 1 using a PI controller.. The transfer function of the PI controller is O. (5182). Adding a zero at s = .KJ/ Kp to the forwardpath lrnnsfer function. the values (~f Kp and K. we shall show that. Consider the secondorder model Op(s) = (s + 2 1)(s + 2) (5182) The system in Fig. However. s (5180) Using the circuit elements given in Table 44 in Chapter 4. However. 545. jf the steadystate error to a given input is constant. The problem is then to choose the proper combination of Kp and K. the magnitude ofthe steadystate error is inversely proportional to K p . At first glance. 544 illustrates the block diagnl. This means that the system type is increased by one. the immediate effects of the PI controller are the following: 1. . so that the transient response is satisfactory. if the location of the zero of O.m of the prototype secondorder system with a series PI controller. will now have a zero steadystate error when the reference input is a step [unction.. the forward path transfer function of the compensated system is G(s) = O.KI/Kp so {hat il is relatively close iO the origin and away from the most Sigllifical1l poles of tire process. the compensated system usually will have a slower rise time and longer settling time.
From the denominator of Eq. the system time response is slow and the desired steadystate error requirement is not met fast enough. 546. or the system will be unstable. Chapter 5. (5185) yields the result that the system is stable for 0 < KI / Kp < 13. Applying the PI controller of Eg. This should also be true for the thirdorder system with the PI controller if the value of KIfKp satisfies Eg. Kr / Kp should be chosen so that the following condition is satisfied: With the condition in Eg. the forwardpath transfer function of the system becomes O(s) == G(. If K/ is too small.'13 2Kp(s + KdKp) + 382 + 2(1 + Kp)s + 2Kl (5184) The characteristic equation of the closedloop system is (5185) Applying Routh's test to Eq. This means that the zero of G(s) at s = Kl / Kp cannot be placed too far to the left in the lefthalf splane. the desired response is met. Thus. For the present case.125. (5186). however. as shown in Fig.25. Let us place the zero at Kl / Kp relatively close to the origin. (5180). (5186) satisfied. As a design criterion.318 ~.5. we assume a desired percent maximum overshoot value of about 4. . The closedloop transfer function is Y(s) R(s) = . the most significant pole of Gp(s) is at 1. (5184) can be approximated by G( } s 9:! s2 + 3s + 2 + 2Kp 2Kp (5187) where the term KIf Kp in the numerator and K[ in the denominator are neglected. Eq. (5187) compared with a prototype second~order system. Thus.1213 rad/s and the required proportional gain of Kp = 1. Upon increasing K/ to 1.(s)Gp(s) = 2Kp(s ~ KdKp) = 2Kp(s + KdKp) s(s + 1)(8 + 2) s3 + 3s2 + 2s (5183) The steadystate error due to a step input uit) is zero. which utilizing expression (5104) results in a relative damping ratio of 0.3 for a unitstep input. Time~Domain Analysis of Control Systems j(J) splane o cr Figure 545 Polezero configuration of a PI controller. we get natural frequency value of W'I = 2.707. to achieve this. we pick a small K/.
4 0.625 den = [132+2*1. 319 1..t). \Ve also introduced the concepts of root contours and root locus and included MATLAB codes to draw them for simple control examples. (5185) for two sets of K.512 MATLAB Tools Unitstep responses of the system ~.25 andKl=O. %KP=1.6 0.2 0.....r. 512 MATLAB TOOLS In this chapter we provided MATLAB toolboxes for finding the time response of simple control systems.. .251.252*1.251.001:10j num= [2*1... s'tep(num.den. and Kp values.125 den= [1 3 2+2*1. 546 are obtained by thefollolVing sequence of MATLABfunctions clear all t=O:O.125]. hold on...4 r.t). 1.125].. num= [2*1.. xlabelC'Time(secs)') ylabelC 'y(t)') title( 'Unitstep responses of the system') .125]. step(numrden..8 0.2 2 3 4 5 Time(secs) (sec) 6 7 8 9 IO Figure 546 Unitstep response of Eq.25 2*1. %KP=1. In Chapters 6 and 9.25 andKI=1. Toolbox 5112 The conesponding responses/or Fig.125]...
and mechanical components . The reader is especially encouraged to explore the Control Lab software tools presented in Chapter 6 that simulate de motor speed and position control topics discussed earlier in this chapter. respectively. we will introduce the Automatic Control Systems software (ACSYS) that utilizes MATLAB and SIMULINK mfiles and aUIs (graphical user interface) for the analysis of more complex control engineering problems.. The transient and steadystate analyses were c8ITied out first by approximating the system as a secondorder system. The error constants are not defined for systems with nonunity feedback. Designs were carried out in the timedomain (and sdomain). electronic components. See Chapters 6 and 9 for more detail. and parabolic inputs. These simulation tools provide the user with virtual experiments and design projects using systems involving dc motors. as well as the system type. Through the GUI approach provided by ACSYS.. ramp. Give the definitions of the error constants Kp' Kv. settling time. The transient response is characterized by such criteria as the maximum overshoot. and Ka.. these programs are intended to create a userfriendly environment to reduce the complexity of control systems design. delay time. 2. sensors. and such parameters as damping ratio. simple controllersnamely the PD. The effect of varying the amplifier gain K on the transient and steadystate performance was demonstrated. maximum overshoot. The analytical expressions of these parameters can all be related to the system parameters simply if the transfer function is of the secondorder prototype. Timedomain analysis of a positioncontrol system was conducted. the analytical relationships between the transient parameters and the system constants are difficult to determine. the steadystate error is characterized by the error constants Kp.. the finalvalue theorem ofthe Laplace transfonn is the basis. K. MATLAB toolboxes and the Automatic Control System software tool are good tools to study the time response of control systems. and K llSpecify the type of input to which the error constant Kp is dedicated. rise time. and the system was then analyzed as a thirdorder system. The time response of control systems is divided into the transient and the steadystate responses. For non unityfeedback systems. The perfonnance is generally measured by the step response and the steadystate error. When applying the steadystate error analysis. PI. rise time. delay time. When the system has unity feedback for the step. The timedomain design may be characterized by specifications such as the relative damping ratio. TimeDomain Analysis of Control Systems where we address more complex controlsystem modeling and analysis. For secondorder systems that are not of the prototype and for higherorder systems.320 ~ Chapter 5. The effects of adding poles and zeros to the forwardpath and closedloop transfer functions were demonstrated. keeping in mind that the zeros of the system transfer function also affect the transient response. a method of detennining the steadystate error was introduced by using the closedloop transfer function. Later in the chapter. It was shown that the secondorder approximation was accurate only for low values of K. and time constant. it should be ascertained that the closedloop system is stable or the error analysis will be invalid. . or simply the location of the characteristicequation roots. The steadystate error is a measure of the accuracy of the system as time approaches infinity. ~ REVIEW QUESTIONS 1. . natural undamped frequency. and PIDwere introduced. and settling time. This established the significance of the location of the poles of the transfer function in the splane and under what conditions the insignificant poles (and zeros) could be neglected with regard to the transient response. The dominant poles of transfer functions were also discussed. The concept of the rootlocus technique was introduced.513 SUMMARY This chapter was devoted to the timedomain analysis of linear continuousdata control systems. Computer simulations are recommended for these systems.
(F) (F) If a unityfeedback control system type is 2. 4. W'l 11. 22. What is a PD controller'? Write its inputoutput transfer function. ~ 321 is dedicated. (T) . Specify the type of input to which the error constant Ka is dedicated. Adding a zero to the forwardpath transfer function will generally improve the system damping and thus will always reduce the maximum overshoot of the system. Specify the type of input to which the error constant K~. IncreaCiing the undamped natural frequency will generally reduce the rise time of the step response. (T) 7. ~. Linear and nonlinear frictions will generally degrade the steadystate error of a control system. the maximum overshoot of the output stays the same. (T) 14. 9. w~(1 (T) (F) 12. 6. 20. Give the effects of the PO control on rise time and settling time of a control system. 5. Y(s) R(s) + Ts) = s2 + 2~wlIs + w.::::: s(s + 1) (T) (F) 19. Given the following characteristic equation of a linear contml system. Increasing the undamped natural frequency will generally reduce the settling time of the step response. Give the definition of the system type of a linear timeinvariant system. The maximum overshoot of a unitstep response of the secondorder prototype system will never exceed 100% when the damping ratio and the natural undamped frequency ilJn are aU positive. For the characteristic equation given in question 15. (On. IS. The location of the roots of the characteristic equation in the splane will give a definite indication on the maximum overshoot of the transient response of the system. increasing the value of K will increase the frequency of oscillation of the system. How does the PD controller affect the bandwidth of a control system? (F) 23. s (T) (T) (F) (F) 10. s:' (F) (T) (F) + 3i + 5$ + K = 0 (T) (F) 16. increasing the coefficient of the i' term will generally improve the damping of the system. Once the value of KD of a PD controller is fixed. 10 10 (T) (F) G(s) = s(s + I)(s + 20) Gds). 18.Review Questions 3. 8. A PD controller has the constants KD and Kp • Give the effects of these constants on the steadystate error of the system. Does the PD control change the type of a system? 21. when the undamped natural frequency increases. The maximum overshoot of the following system will never exceed 100% when and T are all positive. (T) (F) 13.. then it is certain that the steadystate (T) error of the system to a step input or a ramp input will be zero. For the secondorder prototype system. Define an error constant if the input to a unityfeedback control system is described by r( t) = PI US (t) /6. (T) (F) 17. The following transfer function G(s) can be approximated by GL(s) because the pole at 20 is much larger than the dominant pole at s = 1. increasing the value of Kp will increase the phase margin monotonically.
Automatit.707 Wn (e) ~::. McGraw~HiU. 81.ans. "On the Zero NonRegular Control System:' J. Observability. Pole Allocation. Dec. 2nd Ed. AC31. 2002.707 Wn :5 5 radl sec 52. What is a PI controller? Write its inputoutput transfer function. Distefano. Norimatsu and M. Answers to these review questions can be found on this book's companion Web site: www.'. A. 0. K. ·'Design ofIndustrial Regulators:' Pmc. NJ. "Root Locus Design of a Robust Speed Control. pp. A pair of complexconjugate poles in the s~plane is required to meet the various specifications that follow. D. 1990. Meshkat. Davison. REFERENCES I. For each specification. 0. then the maximum overshoot of the system is always reduced. F. Determine the type of the following unityfeedback systems for which the forwardpath transfer functions are given. T. C. pp. M "" PROBLEMS In addition to using the conventional approaches~ use MA TLAB to solve the problems in this chapter. Incremental Motion COlllro/ S. pp. 4. If a PD controller is designed so that the characteristic~equation roots have better damping than the original system. Vol. Automatic Control. Smith and E. (positive damping) (positive damping) (positive damping) (positive and negative damping) (b) 0:5 ~ ::. 28. N. Vol. 440.. Vol. Prentice~Han.1) . M. What does it mean when a control system is described as being robust? 26. Franklin and 1. 3. A PI controller has the constants Kp and K/. and L J. Modern Control Engineering. (F) 27.. lOe. Willems and S. R. TimeDomain Analysis of Control Systems 24. Vol. Powell. 2006. June 1983. Schaum's Outline of Theory and Problems of Feedback and Control Systems. 121012]6. Aug. Ogata. 8. Ito. 1972. Give the effects of the PI control on the rise time and settling time of a control system.o.' Control. sketch the region in the s~plane in which the poles should be located. 2. Mitter. 567575. W. (e) G(s) = s(s (d) G(s) . K.. F.5 5 { :5 0. 582595. 5. Williams. May 1986. lEE (London).vmposium. 51. Give the effects of the PI controller on the steadystate error of the system. 119. 1971. pp. AC I6." Prot. J. 4954. Eng. p. (T) (F) 25. Japan.707 Wn 2:2rad/sec ~ Wn :::. O. H. 1961. (a) ~2 0.322 i. Does the PI control change the system type? 29. Vidyasngar. "Controllability.com/college/golnaraghi. Bailey and S.wiley.2s K (a) G(s) = (I + s)(1 + 10s)(1 + 20s) lO(s + 1) + 5)(s + 6) (b) G(s) = (1 + s)(1 + IOs)(1 + 20s) = s2(s + 5)(s + 6)"" loo(s . "On Undershoot nnd Nonminimum Phase Zeros. m. 6." IEEE T. J.. NJ. Feedback Control of Dynamic Systems. A system compensated with a PD controller is usually more robust than the system (T) compensated with a PI controller. J. Elec. and State Reconstruction:' IEEE Trans. Stubberud.• Prentice Hall." Chapter 5.5 1 S 2 radlsec 5 rad/sec (d) 0. 5th Ed. Insl. 4th Ed. 7. J.
SS. (t zJ2)u s (t). and a parabolic input.3 + 50s2 + lOs' s = s + 5 K(s + 5) . and a parabolic input. (a) M(s):. ramp. Find the poles and zeros of the closed loop system and the system type. a unitramp input. 5P8 has a transfer function YIX. find its value so that the answers are valid.nts of the following unityfeedback control systems. a unit·ramp input. . (t 2 /2)u s (t). Check the stability of the system before applying the finalvalue theorem. The forwardpath transfer functions are given. (t 2 J2)u s (t). and parabolic error consta. (b) M(s) 8+4 s4 + 16s·:1 + 48s2 + 4s+ 4' KH = 1 K(s+3} == . a unitramp input.55) 1000 K lOOO (b) G(s) = S(s2 + lOs + 100) = 82(s2 + IDs + 100) 100 100 (c) G(s) := 5(1 (d) G(s) (e) O(s) = s(s + lO)(s + 100) (f) G(s) = K(1 + 2s)(1 + 4s) 9 2(. Find the steady~state errors of the following singleloop control systems for a unitstep input. The output of the system shown in Fig.ls)(1 + lOs) + 0.Problems o4J 323 lO(s + 1) + 5s + 5) (e) G(s) = s3(s2 (g) (O G(s) (h) = 100 . Find the steadystate errors due to a unitstep input. (a) G(s) = (1 + O. (a) G(s) = ( 2 (b) G(s) = s(s 1 s +s+ 2) H(s) = (s+ 1) I + 5) 1 I 1 H(s) H(s) H(s) =5 =~ s+5 (c) G(s) = $2(s + 10) (d) G(s) = s2( s+ 12) = 5(s + 2) 56.~ + 17s3 + 60$2 + 5Ks + 5K' KH = 1 ( ) M( ) c s = s4 (d) M(s) = 57. determine the steadystate error for a unitstep input. 1 s +5 H( ) lOs + 15. Determine the step.\'3(8+ 2)2 O(s) = 5($ + 2) s2(s + 4) G(s) _ . For the unityfeedback control systems described in Problem 52.52+s+1) 54. For systems that include a parameter K. The following transfer functions are given for a singleloop nonunityfeedback control system. and a parabolic input.15)(1 + 0.(s2 + 2s + 3)(s + 1) 8(8 + 1) 53..~3 + 3s2 + (K + 2)s + 3K' KH :.
Repeat Problem 58 for the system given in Fig.k 1. Find the position. Assume that the system is stable..(t) R(s) . and acceleration error constants for the system given in Fig.with Ikl < I s. x~ Figure 5P8 s+ 1 s+3 5 ·1 .k 513. so that the answers are valid? Determine the minimum steadystate error that can be achieved with a unitramp input by varying the values of K and K. The block diagram of a control system is shown in Fig. when the following inputs are applied..... ramp. and (c) a unitparabolic input. 5P13.Dom ain Analysis of Control Systems 58. 534 Find the rise time of the following firstorder system: G(s) = . 510. x~ Figure 5P10 511. + . 100 Gp(s} = (1 +0. if any. velocity. Find the step.5s) What constraints must be made. . (b) a unitramp input. Repeat Problem 513 when the transfer function of the process is.5+1 5+ 2 ·1 4 1. s I • y I Find the steadystate error of the system given in Problem 510 when the input is X .324 Chapter 5. Find the steadystate elTor for Problem 58 for (a) a unitstep input.L 20s yes) Figure 5P13 5·14. 1 _ ..f(S + 2) I • y I 59.1s)(1 + 0. 5P1O. Time. The error signal is defined to be eel). 5P8. instead . and parabolicerror constants. (a) r(t) = uAt) (b) r(t) = n~s(t) (e) r(t) = (t 2 /2)u. on the values of K and K. Find the steadystate errors in terms of K and K .+2s s2 s3 512.
(b) Find the steadystate value of y(t) when 1'(1) = 0 and l1(t) = us(t). (a) Find the steadystate value of the error signal Be(t) in terms of the system parameters when the input is a unitstep function. Assume that the system is stable. where r(t) is the reference input and n(t) is the disturbance. (t). so that the answer is valid. Let r(l) = O. (a) Find the steadystate valLIe of e(t) when n(l) = 0 and r(l) = tU. Let n(t) = 0 for this part. (b) Repeat part (a) when the input is a unitramp function. 5P17 . (b) Find the steadystate value of y(t) when n(t) is a unitstep function. The block diagram of a linear control system is shown in Fig. Find the transfer function of a secondorder prototype system to model the system. Assume that the system is stable. For the positioncontrol system shown in Fig. 1. Give the constraints on the values of K and K. Find the conditions on the f values of a and K so that the solution is valid."ec) . 5P16. The unitstep response of a linear conlrol system is shown in Fig. N(s) R(s) ~ K(s+3) Y(s) + C011lrolter ~ Process Figure 5P17 518. N(s) R( s) ~ + Figure 5P16 5·]7. 3P7. (a) Find the steadystate error of the system in terms of K and K.Problems 325 51S. 5P18. The erJ'Or signal is defined to be e(t). determine the following. The block diagram of a feedback control system is shown in Fig.25 1. when the input is a unitramp function..0 o Figure 5P1 8 I(. 516.
Repeat Problem 5. 526. SP13. Repeat Problem 5... Use Eg. You may usc any rootfinding computer program to solve this problem . 529.05 sec.0 1 sec. Simulate the system with a computer program to check the accuracy of your solutions.5 and the rise time of the unitstep response is approximately 1 sec. (598) for the risetime relationship.326 ~ Chapter 5.. The forward path transfer function of a contrnl system with unity feedback is K G( s) = =s(s+'a). find the values of K and K. 523. For the control system shown in Fig. sO that the maximum overshoot of the output is approximately 4. Use Eg. 5P.(s+30"'7') where a and K are real constants. find the value of K so that the characteristic equation has two equal real roots and the system is stable. Use Eg.707 and a settling time of 0.l3. (5102) for the settling time relationship.1 sec.19 with a maximum overshoot of 20% and a rise time of 0. 527. Repeat Problem 522 with a maximum overshoot of 10% and a delay time of 0. Repeat Problem 525 with a damping ratio of 0. (a) By means of trial and error. find the steadystate errors of the system when the reference input is (i) a unitstep function and (ii) a unitramp function.1 sec. 522.) sec. With the values of a and K found. and what may have caused it? K(s . find the values of K and K.6 and the settling time of the unitstep response is 0. (c) Repeat part (b) when K = 1 . For the control system shown in Fig. find the values of K and K. 520.3% and the rise time tr is approximately 0. Use any computer simulation program for this. (b) Repeat Problem 525 with a maximum overshoot of 20% and a settling time of OJ») sec. (b) Find the unitstep response of the system when K has the value found in part (a) . Repeat Problem 522 with a maximum overshoot of 20% and a delay time of 0. 525.imately 0. Simulate the system with any timeresponse simulation program to check the accuracy of your solutions. 521. What is peculiar about the step response for small t.2 sec. determine the actual rise time usi ng computer simulation . The block diagram of a linear control system is shown in Fig. Use Eg. 5P13 .3% and the delay time ttl is approx. TimeDomain Analysis of Control Systems 519. Simulate the system with a computer program to check the accuracy of your results. (5103) for the settling time relationship. (a) Repeat Problem 525 with a maximum overshoot of 10% and a settling time of 0.05 sec. Use Eg.05 sec. 528. Set all the initial conditions to zero.l9 with a maximum overshoot of 10% and a li se time of 0. so that the max. For the control system shown in Fig. (598) as an approximation of the rise time. so that the damping ratio of the system is 0.1 sec.imum overshoot of the output is approximately 4.1) Y(s) s(s + 1)(s + 2) Figure 5P29 . . (596) for the delaytime relationship. 524. (a) Find the values of a and K so that the relative damping ratio of the complex roots of the characteristic equation is 0. (b) With the values of a and K found in part (a). SP29.
. 531. (b) Let r(t) = O. and k. Find the minimum steadystate value of yet) that can be obtained by varying K . (t) + 5X2(1) . Construct a parameter plane of Kp versus Kd (Kp is the vertical axis).01 . and show the following trajectories or regions in the plane. Determine how the values of K and K. determine how the values of K and K( affect the steadystate value of eCt).. (d) Assume that it is desired to operate the system with the value of K as selected in part (c). affect the steadystate value of y( l ) when the disturbance input Td (t) = us(t). (c) Find the values of kl and 1<2 such that ~ = 0 .tep input is zero.[t = 6xJ (/) yet) = u(t) XI (/) + U(l) The control is obtained through state feedback with = kj Xt (t) . Find the restrictions on K and K( so that the system is stable. (d) Let the error signal be defined as e(t) = r(l) . The block diagram of a linear control system is shown in Fig. 1. 5P3 1.Problems • 327 530. (b) Find the locus in the k l versusk 2 plane on which the overall system has a damping ratio 0[0. (a) When ret) = tU.5. + .707 and = 10 fad/sec. A controlled process is represented by the following dynamic equations: . 5P32. so that the complex roots of the characteristic equation will have a real part of .(t) is a unitstep function. are at the values found in part (c).. (c) Let K. The block diagram of a linear control system is shown in Fig.2 (D Trajectory on which the natural undamped frequency w" is 50 fad/sec (g) Trajectory on whieh the system is either ullcontrollable or unobservable (hint: look for polezero cancel lation) Y(s) R(s) . The fixed parameters of the system are given as T = 0.yet). Find the value of this K. = 0.707 .01 and r(t) = O.2. and K j = 10. Find the steadystate error when ret) and k.k2X2 (t) + ret) where kl and k2 are real constants.I(t) and Td(t) = 0. and r(t} is the reference input.Xl . when T. (a) Find the locus in the k l versusk2 plane (k l = vertical axis) on which the overall system has a Wit natural undamped frequency of 10 rad/sec. Find all three roots of the characteristic equation. (a) Unstable and stable regions (b) Trajectories on which the damping is critical (~ = 1) (c) Region in which the system is overdamped (~ > 1) (d) Region in which the system is underdamped (~ < 1) (e) Trajectory on which the parabolicerror constant K" is 1000 sec .t = dX2(t) dXJ (I) . would you operate the system at this value of K? Explain. Find the value of K.: ]00 Figure 5P31 532. From the transient standpoint. J = 0. = us(t) (e) Find the locus in the k) versusk 2 plane on which the steadystate error due to a uniH.
and settling time when a unitstep input is applied to the system.. If the maximum overshoot of the unitstep input and the peak time are 0.6 and WIJ = 5 rad/sec.\' ] 6.v) + Ii Figure 5P32 533. Assume J = 1 kgm 2 and B = 1 Nm/ rad/sec.. Fig.. peak time. respectively.328 Chapter 5.\. Consider a secondorder unity feedback system with ~ = 0. 534. x y Figure 5P·34 535. Also.  Ki 1 + Ts 1 KrS .. 5P34 shows the block diagram of a servomotor..2 and 0.r l[ '\.1 sec. TimeDomain Analysis of Control Systems 7d(s) + R(s) + £(s) K  +a. (a) Find its damping ratio and natural frequency.tl ] = [I 1 1 + [I1 01][''1] ][. maximum overshoot.2 J = [0I . (b) Find the gai n K and velocity feedback KJ.5 0 1 I[Y =[1 O][XI] +[O °][11 ] 0 1 0 0 112 Y2 _ x2 112 (b) . Find the unitstep response of the following systems assuming zero initial conditions: . Calculate the rise lime. calculate the rise time and settling time. 1 Y(.2 X2 (3) [.I][XI] + [0] 1 1 "2 II (c) [il l ::~ [0 ~1 ~l ~ :~ 0] [X I] [0] + ~ II y= [O 0 I l[:~] X3 .
Use MATLAB to solve Problem 535 .on the poles of O(s) and the roots of the characteristic equation . The characteristic equation of the closedloop system is (a) Let KL = 10. Write the forwardpath transfer function O(s) and find the poles of G(s). 5P39 shows a mechanical system. (b). K" =0.lrad._ = 1000 ozin. Fig. 538. and Bm S:' O. R" = 5fl. the motor shaft is rigid. (a) Find the differential equation of the system. (b) Repeat part (a) when K. The block diagram ofthe guidedmissile attiludecontrol system described in Problem 420 is shown in Fig. 541.001 ozin. A. The command input is ret).. (d) Compare the results of parts (a). Find the critical value of K for the closedloop system to be stab le. y R lIo/rictioll Figure 5P39 540.sec 2 ./rad. that is. and (c).Problems 536.)S2 + (n2RaKLJL + R"KLJm + BIIIKLL{l)s + Rl/BmKL + KiKhKL] where Ki = 9 ozin '. The objective of this problem is to study the effect of the controller G.636 V/rad/sec. (b) Use MATLAB to find the unitstep input response of the system. 537.Kbh + Rl/Bmh. Find the roots of the characteristic equation of the closedloop system when K is at marginal stability. and del) represents disturbance input. 5PAl. (c) Repeat part (a) when KL == 00.(s) on the steadystate and transient responses of the system. . and comment on the effects of the va lues of Kl. Find the impulse response of the given systems in Problem 535. Use MATLAB to solve Prohlem 537.000 Olin. 1(. 329 539. L" = I mH. The dcmotor control system for controlling a printwheel described in Problem 449 has the forwardpath transfer function where 6(s) = s [LaJllfhs4 2 + JL(RaJIII + BIIILa)sJ + (n KLLlIh + KLL"Jm + K. = 1 V/rad. JI/I =h= 0. n = 1/10.
5P42 represents a liquidlevel control system. r ::.(s) as given in part (e) when 1'(1) = 0 and dill = us{t) . and 500. 10.0.0.1) Controller Missi le dynamics Figure 5P41 542. 50 H(s) s(0. and N denotes the number of inlets. A~Slllne zero initial conditions.10. Approxi mate the system by a secondorder system by neglecting the pole of G(s) at s = . . t ::. Find the steadystate value of y(t) when d(t) = uA t). with G.(t) .5) Figure 5P42 543. Find the steadystate error of the system when r(t) is a unitstep function . (a) Because one of the p01es of the open100p transfer function is relatively far to the left on the real axis of the splane at s = 10. Find the steadystate error when ret) is a unitstep function. Use any computer simulation program that is available. (b) Obtain the unitstep response (with zero initial conditions) of the original Lhirdorder system with N = I and then with N = 10. Q' = 5. 50. (e) Let Gc(s) = (s + a)ls. Comment on the accuracy of the approximation as a function of N. The forwardpath transfer function of a unityfeedback control system is G () = s 1 + Tzs 2 sis + 1) CompuLe and plol the unilsLep responses of the closedloop system for T: = 0. (b) Let Gc(s) = (s + a) l s.5 sec with Ge(s) as given in part (b) and a = 5. Use any available computer simulation program. 50. TimeDomain Analysis of Control Systems (a) Let Gc{s) = 1. it is suggested that this pole can be neglected. Set dU) = O.0. The approximation should be val id for both the transient and the steadystate responses. 0.5 sec. Record the maximum overshoot of y(t) for each case. The liquid level is represented by hU). Comment on the effect of varying the value of a of the controller on the transient response . and 500. 1. (d) Set r(/) = 0 and G . D(s) £(s) IOO(s + 2) yes) (l. Use zero initial conditions. (c) Obtain the unitstep response of the system for 0 ::. and 50.5. Find the steadystate value of y (t) when d(l) = u.05s + 0. 0.330 Chapter 5. Compare the responses of the Oliginal system with those of the secondorder approximating system . (f) Obtain the output response for 0 ::. Comment 00 the effects of the various values of T= on the step response. The block diagram shown in Fig... (g) Comment on the effect of varying the value of a of the controller on the transient response ofy(l) and d(t). Apply the formulas for the maximum overshoot and the peak time tlll llX to the secondorder mode1 for N = I and N = 10. Assume zero initial conditions.(s) = 1. 0.
Comment oil the effects of the pole at s = .I / Tp in G(s) . Compare and plot the unitstep responses of the unity feedback closedloop systems with the forward path transfer functions given.20 T~ = 0.I') is a unitstep input and D(s) is a unit impu lse input.1') = . 5P46 shows the block diagram of a servomotor with tachometer feedback. K(s + 2. 20 2 (c) G(s) = (s2 + 2s + 2)(1 10 (d) G(. (a) G(s) = s(s (b) G(s) = 1 + Ts + 0.Problems 331 544...5s + 10) (i) For K (ii) =5 For K = 10 (iii) For K === 30 546... The forwardpath transfer function of a unityfeedback control system is G(s) = 1 s(s + 1).5) 1 + Ts {2 2 .5) s(s + l.0 (e) G(s) = s(s + 1.5. 0. (c) Use MATLAB to plot the response of the system for palt (b) . 1. Find the critical value of Tp so that the closedloop system is margina. 5 .55 )(~ + 1. x K + _ 1_ s y Figure 5P46 .5x + 10) (i) For K === 5 (ii) For K = 10 (iii) For K === 30 (f) G(s) = . (a) Find the error signal E(s) in the presence of the n~ference input X(s) and disturbance input D (s) .0 Tp s(s + 5)(1 + T"s) K J . = 0.~ + 2. 5.0. (d) Use MATLAB to plot the response of the system when X(. 1 . (b) Calculate the steadystate error of the system when Xes) is a unit ramp and D(s) is a unit step.. I .25 )(s + 2.25 )(.5. and 0. 1. Assume zero initial conclitions. 0.(1 + Tps) J Compute and plot the unitstep responses ofthe closedloop sys tem for T/I = 0.2) s + s+ For For For For T~= O .707. Assume zero initial conditions. Use any computer simulation program. + Tps) Tp = 0 .5. Fig .lly stable. 545.
C Figure 5P48 549.= +.. use MATLAB to plot the unitstep input response of the system and verify your controller design.. If the maximum overshoot and I % settling time of the unitstep response of the closedloop system shown in Fig. . 5P4~ are no more than 25% and 0.. Chapter 5.Rc (s) (b) Figure 5P5D .. find the region in the splane corresponding to the poles that meet this specification. R I. c ...332 . (Ails + 1) (BIs + I )(B2S + 1) . TimeDomain Analysis of Control Systems 5·47.. If the closedloop transfer function can be rewritten as Y(s) X(s) O(s) 1+ O(s) (Al s + 1)(A2S + I) .... (B/lls + t) (a) Find the steadystate error to a unitstep input... A unity feedback control system shown in Fig. Also. find the gain K and pole location P of the compensator.1 sec..+ .+. 5P50(a) is designed so that its closedloop poles lie within the region shown in Fig. (b) Calculate k = liD1'_~SC(S) ' 548. The feedforward transferfunction of a stable unity feedback system is O(s) . 5P50(b).... 550. If a given secondorder system is required to have a peak time less than I.
0. Find the transfer function of the controller Gc(s) so that the following specifications are satisfied: (a) The ramperror constant Kv is 5.. (a) Find the transfer function between the applied voltage and the motor speed. then find the values for K and K. (b) Calculate the steadystate speed of the motor after applying a voltage of 10 V. and A = 20 O. the plant transfer function is and the controller transfer function is H(s) = k{s + a) (s + b) Design the controller parameters so that the closedloop system has a 10% overshoot for a unit step input and a 1% settling time of 1. (b) If Kp = 2 and P = 2. The transfer function between pitch angle a and elevator angle /5 are given by a(s) P(s) 60(5+1)(5 + 2) (s2 + 6s + 40) (s2 + 0. 551. 5P48.5. obtain the transfer function between 6p and 0. Explain controller design difficulties for complex systems..07) The autopilot pitch controller uses the pitch error e to adjust the elevator as E(s) ' j + 1 0 a unity feedback configuration and utilize Use MATLAB to find K with an overshoot of less than 10% and a rise time faster than 0. B = 0002 Nmsec..5. 552. 553.1. and confirm the results. The block diagram of a control system with a series controller is shown in Fig. regardless of values Kp and P. where A is the feedback T gain.04^ + 0. and 2.. In the unity feedback closedloop system in a configuration similar to that in Fig. 333 (c) Show that.04 Vsec. (f) If the rise time is less than 3 sec.04 Nm/A. The motion equation of a dc motor is given by Assuming Jm = 0.. (c) Determine the transfer function between the applied voltage and the shaft angle 9. K\ = 0. (e) If the maximum overshoot is less than 25%.5 sec for a ^ unitstep input. K2 = 0. (b) The closedloop transfer function is of the form M(s) = Y(s) R(s) K (s + 2 0 5 + 200)(5 + 0) 2 where K and a are real constants. Find the rise time and overshoot. (g) Use MATLAB to plot the step response of the position servo system for K = 0.0.sec.. determine K.. the controller can be designed to place the poles anywhere in the left side of the 5plane.02 kgm2. . Use MATLAB to find the values of K and a. determine K. (d) Including a closedloop feedback to part (c) such that v = K (6p — 6m). 554. 5P54.Problems (a) Find the values for con and f.. An autopilot is designed to maintain the pitch attitude a of an airplane.
5. TimeDomain Analysis of Contro l Systems The design strategy is to place the closedloop poles at 10 + jlO and . (b) Find the values of Kp and KD so that the ramperror constant K. What is the maximum value of Kv that can be realized? Comment on the difficulties that may arise in attempting to realize a very large K. 5P59 shows the block diagram of the liquidlevel control system described in Problem 542. 5S6.2 to 1. R(s) £(s) _100 G. and then adjust the values of K and a to satisfy the steadystate requirement. Find the maximum overshoot of the designed system. The number of inlets is denoted by N.. ::.0.1~~f2)' Use MATLAB to design a series PD controller with the transfer function Gc(s) = Kp + Kos so that the following performance specifi cations are satisfied: Steadystate error due to a unitramp input ::.10 . 5P56.0 in increments of 0. set the value of Kp so that the ramperror constant is 1000.707. Consider the secondorder model of the aircraft attitude control system shown in Fig. 529 . £(s) Kp+Kns 1000 $(8+ 10) Y(s) Figure 5P56 S57. 5 % Risetime tr ::.005 sec 5S9. Set N = 20.005 sec Settling time t.2 to 1. (b) Vary the value of Kj) from 0. The transfer function of the process is GI'(s) = J(. (c) Find the values of Kp and Ko so that the ramperror constant Kv is 1000 and the damping ratio is 1.001 Maximum overshoot ::. 0.2 and determine the values of rise time and maximum overshoot of the system. is 1000 and the damping ratio is 0. The value of a is large so that it will not affect the transient response appreciably. Use MATLAB to (a) Vary the valne of KD from 0.(s) ? Y(s) + s(s. Repeat Problem 554 if the ramperror constant is to be 9.0 in increments of 0. Use MATLAB to (a) Find the values of Kp and KD so that the ramperror constant K. 0. . Fig. For the control system shown in Figure 5P56. A control system with a PD controller is shown in Fig.2 and find the value of KfJ so that the maximum overshoot is minimum. 0. is 1000 and the damping ratio is 0...)10. Use MATLAB to design the PD controller so that with a unitstep input the tank is filled to within 5% of the reference level in less than 3 sec without overshoot.+ lOs + 100) Figure 5P54 sss. 558.334 ~ Chapter 5.
.5 and determine the values of rise time and maximum overshoot of the system. Use MATLAB to design the controller parameters so that the following specifications are satisfied: Ramperror constant K" = I 00 Rise time t. (b) Find the value of Kp so that the magnitude of the imaginary parts of the complex roots of the characteristic equation of the system is 15 rad/!. Use the value of K. For the control system described in Problem 561. set KJ so that the ramperror constant is 10. Find the value of Kp when the maximum overshoot is a minimum. 5·65. find the critical value of Kp so that the system is stable. (b) With the value of K. use MATLAB to perform the following: (a) Find the value of K. Kp < oc. set Kp so that the ramperror constant is I . Sketch the root contours of the characteristic equation for 0 ::.01 sec. found in part (a). as determined in part (a) and for 0 ::5 Kp < 00 . Repeat Problem 563 for Kv = 10. What is the value of this maximum overshoot? 564. so that the ramperror constant K. A control system with a type 0 process Gp(s) and a PI controller is shown in Fig. found in part (a). Maximum overshoot < 2% Plot the unitstep response of the designed system. Y(s) Figure 5P65 . 5·61.Problems ~ 335 ION s(s +1)(s + 10) Yes) Figure 5P59 5·60.ec. For the liquidlevel control system described in Problem 559. 563. For the control system shown in Fig. is 10. Use MATLAB to vary Kp and determine the values of rise time and maximum overshoot of the system. Use MATLAB to (a) Find the value of K. Use MATLAB to vary Kf) from 0 to 0. (c) Show that the maximuln overshoot is high for both large and small values of Kp. so that the rampenor constant K" is 100. 5P61.. (c) Sketch the root contours of the characteristic equation with the value of K. Find the roots of the characteristic equation. A control system with a type 0 process and a PID controller is shown in Fig. s K G (s)= p 2 tOO s + lOs + 100 yes) Figure 5P61 562. < 0. E(5) K p +1. 5P61. 5P65.
m = 0.5 kg.1 Nlmlsec (friction of the cart).2 kg. k = 1 N/m.3 m.002 sec/m.05 sec and the overshoot is no more than 3%. 11. K = 20 N/m. Use MATLAB to design a proportional controller where the 1% settling time is less than 0. Chapter 5. (b) It is desired to design a PD controller. Time~Domain Analysis of Control Systems 5M66.336 . Use MATLAB to design a PO controller where the rise time is less than 0.8 mls . Assume m = 25 kg.wns + wn 2 (a) It is desired to design a proportional controller.05 sec and the overshoot is no more than 3%. ry(t) ~ ~fttl M B Figure 5P~67 5·68. Assume 2 M = 1 kg. B = 10 N. m = 0. and 1 = 0. Use MATLAB to design a proportional controller where the peak time is less than 0. Use MATLAB to design the controller parameters where the rise time is no more than 0. Plot the unit~step response of the designed system. Consider the inverted pendulum described in Problem 49. J 5 kg_m1. = 0. g = 9. Use MATLAB to design the controller parameters where the ris~ lilllt! is no more than 0. The Laplace transfoml between the base acceleration and displacement is given by Z(s) 1 s2 Y(s) = + 2t.05 sec and the overshoot is no more than 3%.. Consider the springmass system shown in Fig.= 0. Use MATLAB to design the controller parameters where the rise time is no more than 0.s/m. Plot the unitstep response of the designed system.. 567. 5P~67. (d) It is desired to design a PIO controller. = = 569. K 100 N/m. (c) It is desired to design a PI controller. Use MATLAB to design the controller parameters where the rise time is no more than 0.5 kg. and r = 0.05 sec and the overshoot is no more than 3%. Consider the train system described in Problem 46. Plot the unitstep response of the designed system.2 sec and the overshoot is no more than 10%.006 kgrn2. 570. and g = 9.8 m/s2. Plot the unitstep response of the designed system. '" 1 Its transfiiiunctton IS given b!i!l = Ms2+Bs+K' er y F(s) Repeat Problem 5 66 where M M = 1 kg. .l = 0.1 sec and the overshoot is no more than 2%.05 sec and the overshoot is no more than 4%. where M = 0.J. Consider the quartercar model of vehicle suspension systems in Example 4113.35 m. Plot the impulse response of the system. Consider the vehicle suspension system hitting a bump described in Problem 43. f.
It is a demanding task to develop software that provides the reader with practical appreciation and understanding of dc motors including modeling uncertainties. There are three classes of simulation experiments designed for this chapter: SIMLab. speed response. you must have completed the relevant background preparation in Chapters 4 and 5. To gain practical knowledge of the Quarter Car Sim software. 6. 2. and their accuracy has been verified. There experiments are intended to supplement the experimental exposure of the students in a traditional undergraduate control course. the focus on de motors in these experiments is intentional. To give a better feel for controller design through realistic examples. In this chapter. system identification. speed control. the first involving control of a simple robotic system and the last one investigating the response of an active suspension system. which exhibits many of the same nonidealized behaviors observed in an actual system. we have created a series of virtual lab experiments that are designed to help students understand the concepts discussed in Chapters 4 and 5. The main objectives of this chapter are: 1. To show bow different parameters and nonlinear effects slich as fri<. However. To provide preliminary instruction on how to identify the parameters of a system. In this chapter. The fOCllS of this chapter is therefore on these lab problemsnamely. using MATLAB and Simulink. and controller design amid these practical challenges. 3. Virtual Lab.6 The Control Lab ~ 61 INTRODUCTION The majority of undergraduate courses in control have labs dealing with time response and control of dc motors. position response. These virtual labs include experiments on speed and position control of dc motors followed by two controller design projects. speed control. 5. To get started using the SIMLab and Virtual Lab. and Quarter Car Sim. and position control of dc motors. and position control concepts. Before starring the lab. 4. we created a virtual dc motor in Virtual Lab. because of their relative simplicity and wide usage in numerous industrial applications. All the experiments presented here were compared with real systems in the lab environment. nonlinear effects. through the use of MATLAB and Simulink. 337 . To provide an indepth description of dc motor speed response. This chapter also contains two controller design experiments.:tiun and saturation affect the response of the motor.
. Gear head +~<~i ~~ ". The block diagram of the closedloop system is shown in Fig. The Control Lab 62 DESCRIPTION OF THE VIRTUAL EXPERIMENTAL SYSTEM The experiments that you will perform are intended to give you handson (virtually!) experience in analyzing the system components and experimenting with various feedback control schemes. 6. Note that for simplicity the input voltage can be scaled to a position input Tin(s) so that the input and output have the same units and scale. 1 V"u. the simplest strategy is to use a proportional controller with gain K. 63. To study the speed and position response of a dc motor.(s) Js +B ~. the controller may be composed of an analog circuit system) A simple speed control system is composed of a sensor to measure motor shaft speed and an amplifier with gain K (proportional control) in the configuration shown in Fig.L The setup components are as follows: • A dc motor with a position sensor (usually an encoder with incremental rotation measurement) or a speed sensor (normally a tachometer or a differencing operation performed on encoder readings) • A power supply and amplifier to power the motor • Interface cards to monitor the sensor and provide a command voltage to the amplifier input and a PC running MATLAB and Simulink to control the system and to record the response (alternatively. I Feedback ~. 62. The system is composed of an angular position sensor (usually an encoder or a potentiometer for position applications). armaturecontrolled de mOlor. a typical experimental test bed is shown in Fig. The block diagram of the system is also shown in Fig. To control the position of the motor shaft.(s) Figure 62 Block diagram of a speedconlrol.338 Chapter 6.Sensor Figure 61 Feedback control of an armaturecontrolled dc motor with load inertia. The components are described in the next sections. 61.1/(1) I Amp.
henry Va = applied armature voltage.3172 E005 sec A reduction gear head may be attached to the output disk of the motor shaft. volt e = angular displacement of the motor shaft. 622 Position Sensor or Speed Sensor For positioncontrol applications. 1 = h l n2 + 1m . ohm La = armature inductance.56 mH Armature moment of inertia 0. kgm2 (refer to Chapters 4 and 5 for more details) ratio = geM B = equivalent viscousfriction coefficient of the motor and load refelTed to the motor shaft. the gear head will scale the inertia of the load by I / n2 in the system model . B must be scaled by n. 621 Motor Recall from Chapter 5 that for the annaturecontrolled de motor shown in Fig. armaturecontrolled de motor. an incremental encoder or a potentiometer may be attached directly to the motor shaft to measure the rotation of the armature.35 ohm Armature inductance 0. T/1I = = Motor (torque) constant 0. 524. kgm 2 h J n = any extemal load torque considered as a disturbance. the system parameters include R" = armature resistance. kg_m 2 = equivalent moment of inertia of the motor and load connected to the motorshaft.10 V/radlsec Armature resistance 1. refer to Chapter 4 for more details) K/ = speed sensor (usually a tachometer) gain The motor used in this experiment is a permanent magnet de motor with the following parameters (as given by the manufacturer): K". radian T = torque developed by the motor.10 Nm/A Speed Constant 0. Nm J L = moment of inertia of the load.0019 kgm 2 Motor mechanical time constant 2. If the motor shaft's angular rotation is considered the output. volt v" = back emf.62 Description of The Virtual Experimental System ' 339 Figure 63 Block diagram of a positioncontrol. In speed . Nro 1m= moment of inertia of the motor (motor shaft). Nm/rad/sec (in the presence of gear ratio. where n is the gear ratio. = Kb = R" = La = 1".
it is customary to connect a tachometer to the motor shaft. 64.. there are five experiments. The output from each sensor is proportional to the variable it is measuring. SIMLab and Virtual Lab are series of MATLAB and Simulink files within the Automatic Control Systems (ACSYS) applet that makes up an educational tool for students learning about dc motors and control systems. The details of the amplifier design are somewhat complex and will not be discussed here. ~ 63 DESCRIPTION OF SIMLAB AND VIRTUAL LAB SOFTWARE As shown in Fig. a motor may not be modified but must be replaced by a new one! In both the SIMLab and the Virtual Lab. that is. Openloop step response of the motor appears in the third experiment. The Control Lab control. Real motors have issues such as gear backlash and saturation. K t = 1 (speed control). In the fourth experiment. We will assume a proportionality gain of 1. The output current from the interface should normally be limited. SIMLab was created to allow students to understand the basic simulation model of a de motor. The motor parameters cannot be modified in the Virtual Lab because. and K. A controller design project is the last experiment. . You would be required to attach all the experimental components and to connect the motor sensor and the amplifier to a computer equipped with MATLAB and Simulink (or some other realtime interface software). The maximum voltage that can be output by the amplifier is effectively limited to 20 V. in a realistic scenario. The sensor output would also have to go through an analogtodigital (AID) card to reach the computer. Sensorshaft inertia and damping are normally too small to be included in the system model. 623 Power Amplifier The purpose of the amplifier is to increase the current capacity of the voltage signal from the analog output interface card. interfacing is an important issue. feedback speed control and position control are explored. The maximum current that the amplifier can provide to the motor is limited to 8 A. Therefore. 1. Amp gain 2 VIV Amplifier input saturation limits ± 10 V Current saturation limits ±4 A 624 Interface In a realworld scenario.340 ~ Chapter 6. The parameters of the motor can be adjusted to see how they affect the system. = 1 (position control). But we should note two important points regarding the amplifier. the frequency response of the open~loop system can be examined by applying a sinusoidal input. Simulink would then provide a voltage output function that would be passed on to the amplifier via a digitaltoanalog (D!A) interface card. which may cause the motor response to deviate from expected behavior. you could avoid using a computer and an AID or DIA card by using an analog circuit for control. 2. The Virtual Lab was designed to exhibit some of the key behaviors of real de motor systems. Alternatively. In the first two experiments. whereas the motor can draw many times this current. Users should be able to cope with these problems.
The plots or animations that the experiment supports appear in the display panel on the right. The standard . as in Fig. Zoom control buttons allow you to view the response at greater detail . navigate to the appropriate directory in the MATLAB command window. the experiment control window will be displayed.'). Every experiment has a button to enter model parameters. Close and Exit Figure 6·4 The Automatic Control Systems (ACSYS) applet. The Experiment menu can be used to swiLch between different control experiments. you may directly call SIMLab or Virtual Lab from the MATLAB command window by navigating to the Vi11ualLab subdirectory and typing simlab or virtual/ab.':itl Transfer function Symbolic Controller Design Tool Sim Lab Virtual Lab Quarter Car Sim . The timeresponse plot is displayed. click on the Help Me button in the menu bar. Fig. and type Acsys at the command prompt. C (1#/ trill· . and the data cursor gives precise point values for graphed data. 111t 01 1II/t iI' . The SIMLab or Virtual Lab experiment windows can be called from the ACSYS applet by clicking on the appropriate button.<. and additional experimentspecific plot controls. The grey control panel on the left contains the control buttons for the experiment.'11.6·3 Description of SIMLab and Virtual Lab Software ~ 341 . Alternatively. respectively. 65. SIMLab allows you to display the motor transfer functions in various formats and to access other custom tools from the SIMLab Tools dropdown menu . and plot control buttons appear below the axes. When SIMLab or Virtual Lab is opened. For stepbystep instnIctions on using the experiment window. To launch the ACSYS applet. a field to enter simulation time. 66 shows a typical experiment control window in SIMLab.
. 'Opt I'l~'St'wI~ r r 0tt+4 :" . !I(s) = 7'$ · . .342 • Chapter 6."'~ to J .~ Cmtrd ~C~"" ~~~~================~==~=========================.... . The Control Lab .I ~ 0rit:1 I~ ~ I!~ i: Pmii:lncatTol • (h1+2 CtJff3 l:Open~Spool2 .~...' ~I W]I r RelJseAzal .s . Figure 65 The Experiment menu for SIMLab or Virtual Lab. I 5:Ct:ftJoI~~Cau.:<1 I OU1PU1~ pUl TIme [sec) Figure 66 Typical SIMLab experiment control window..: .
@ 1~9 1t1:. shown in Fig.:~d ~~ Figure 67 SlMLab Speed Control Simulink model. The Start Time and Stop Time settings in the Solver 0CllIlfigurahon Parameters: .I. 67. dit. a Simulink (. pt:. Selecting Simulation from the Simulink menu and next choosing Configuration Parameters allows access to these settings. shown in Fig.l Int ".63 Description of SIMLab and Virtual Lab Software 343 fie edt Y~_~tiOn Fomut Tools H:l.tlon DC mirlor. with a reference step input and mUltiple outputs.Ii!. .mdl) window containing the model for the experiment will be launched.t~d Coo"ol D OlJ bllt'tl l ~OI'I lh. Rlgr'l''''hdt(Hllh."".. . DC ."odt ' Microsoft Windows calculator and a unit conversion tool can be accessed from the top menu.oto'p.ttr$.! ' • e . 68..M lof blo&to mt. By selecting the Enter Model Parameters button.p D : Qi. The model..sJ ~>t. DC motor bl.hsY' 10 su· the Jimu. ctLook lJndtl t.dl:udhl. The model parameters must be set first in any experiment.§pcedctrl_slm/ ConOguralion • llK Figure 68 Configuration Parameters: Solver properties. ~~~::::~~ ~S. All the simulation parameters for the Simulink model are preset. contains a simple closedloop system using PID speed control.lmtnl 1  • ~ ltjo"naI Sp.
The Control Lab l:l Function Block Parameters: 5ub5ystem2 .0019 Viscous d. such as the resistance. is a torquedisturbance input into the 111otor.:::~"!t= . They allow you to modify the simulation running time. Selecting the Reuse Axes checkbox prints all plots to the same set of axes in an external figure. Parametet . Selecting Close Experiment in the control window exits the program.fig or image file for future reference and analysis. such as animation and calculation tools. load inertia. Armalure Inductance (l [HD \0. may be modified. backemf constant.000792 Back emf conslant (Kb IV Iradl~]) 1°·1 0 Motor conctant (Km (N. doubleclick on the appropriate model block to modify model parameters such as the PlD values. doubleclicking on the motor block brings up a window contai ning a list of adjustable motor parameters (see Fig. These are discussed in the following sections..m"2/sn )0. and damping coefficient. Again. . For SIMLab. All motor parameters. Other options in Configuration Parameters shou ld not be modified. you cannot change the system parameters. One other feature that STMLab has . To run the simulation. 69). click the zoom button and select the area of the timeresponse plot to view it closer. However. close the Simujink model and click the button labeled Run Simulation. which is useful for comparing the system response after changing a parameter in the Simulink model. Rightclicking on a SIMLab motor block and selecting Look under Mask makes the de motor model available. but PID values are available for modification. in the Virtual Lab. options are most impOliant as far as SIMLab and Virtual Lab examples are concerned. This can be used to investigate the stall torque and the effect of an integral controller. When the Simulink model is opened.00056 Armature Retistance (R [ohm)) 11.344 Chapter 6.:J! ~ancel Help ~ply Figure 69 Adjustable parameters for the SIMLab mOlor blocks. or on the left panel in the SIMLab/ Virtual Lab interface. the Virtual Lab motor blocks are completely opaque to the user since they model actual dc motors. and they can be manipulated in Configuration Parameters.mping Iridion (b(k9. as they may cause errors in the SIMLab and Vjrtual Lab software. Some of the experiments have additional features. The data cursor button allows the graphed values to be displayed as the cursor dot is moved around on the graph using the mouse or arrow keys.m/All 10. The Print to Figure button allows the current response plot to be sent to a separate MATLAB figure.~__. This figure can also be saved as a .m"2)) 10. which Virtual Lab does not.35 Motor and Load Inertia [J (kg.10 I C:~. For more detail.
The Simulink system model is shown in Fig. Then gradually increase its value and check the speed response. select 3: Open Loop Speed from the Experiment menu~ and perform the following tests: 1. The response of the actual system (in this case. Once the model of the Virtual Lab system is identified and confinned. 641 OpenLoop Speed The first step is to model the motor. we use the openloop model represented in Experiment 3: Open Loop Speed. Once satisfactory model performance has been achieved. Assuming that you do not know the overall inertia J for the system in Step 3. certain model parameter values would have to be revised or the model structure refined to reflect more closely the observed system behavior. various control schemes can be implemented. using additional load inertia at the output shaft of the gear head of 0. it will be useful to build a model using these values and to simulate the dynamic response for a step input. the controller that was originally designed using SIMLab should be tested on the Virtual Lab model. 5. Because most of the system parameters are available (see motor specifications in Section 621). (5118)1 (try dc motor alone with no gear head or load applied in this case). simulate the openloop velocity response of the motor to a step voltage applied to the armature. Using the parameter values in Section 621 for the model of the motor in Fig. confinn the values of motor and load inertia. 524. Apply step inputs of +5 V~ +15 V. Repeat Step 2 and connect the gear head with a gear ratio of 5. In this experiment.05 kg_m 2 (requires modification of J in the Simulink motor parameters). Derive and calculate the disturbance torque steady~state gain. 2. or they may be identified experimentally. 610. Introduce an appropriate stepdisturbance input TL and study its effect on the system in Step 3. First set B = 0 in the Simulink motor parameter window. Should the actual response to the test input be significantly different from the predicted response. SIMLab represents the simulation model with adjustable parameters. Start up SIMLab. Note that the steadystate speed should be approximately the applied armature voltage divided by Kb as in Eq. 6. Study the effect of viscous friction on the steadystate motor speed. can you use the speedresponse plot to estimate its value? If so. I w(t) = :b (I . and Virtual Lab represents the actual (virtual) system. and 10 V. Determine the viscous friction required at output shaft to reduce the motor speed by 50% from the speed it would rotate at if there were no viscous friction. representing a simple openloop model with a motor speed output.et!tm) (5IlS) . In this chapter. the virtual system) to the same test input will then verify the validity of the model.2: I. How about the viscousdamping coefficient? Can you use the time response to find other system parameters? 3. Try using the gear head calculator in the SIMLab Tools dropdown menu.64 Simulation and Virtual Experiments ~ 345 64 SIMULATION AND VIRTUAL EXPERIMENTS It is desired to design and test a controller offline by evaluating the system performance in the safety of the simulation environment. The simulation model can be based on available system parameters. 4.
. rightclick the dc motor block and select Look under Mask to obtain the motor model shown in Fig . .. assuming all other parameters are known.. ~u l .... The steadystate velocity and the time constant Tin can be found from the timerespom.. A typical openloop speed response is shown in Fig. ' tct "l. . 0' blotAt " ". ..1 ." ~ r IN"~ 3 1 i'~@) Ij).~ ii.lh . the time constant is the time to reach (le I) x 100. the voltage saturateS.. A10"' cI.. Step [)k:tulbl no. l ion OC moHol rn".j. The current within the motor may also be saturated. In a realistic scenario. 6.' T ~~rii In..O\IIt IlIt d lf "~ I o" .. or 63. the motor is connected to an amplifier that can output a voltage only in a certain range.~. For a unitstep input.eUdc on tn. for a firstorder system. To create these effects in software. Chapter 6.119)].. The Control Lab 0It1 ~~ 1l '* £dt_~_..12.2% of the final value for a step input [verify using Eq. DC If'IOl orl>IOclc t o " o4 i1y tt. If you do not wish to include saturation..ul . the openloop speed response is shown in Fig. Assuming a small electrictime constant. the disturbance torque default value is set to zero. 612. Run the above experiments again and compare the results. (5118) or (5.. simply change its final value. the motor inertia and the viscousdamping friction could be calculated with measurements of the mechanicaltime constant using different input magnitudes. 611. (N. Outside of this range. l1' ot Ol PoIIr .ntt t l'S.d("''' U o. . As a result. Doubleclick both the voltage and current blocks and adjust their values (default values of ± 10 volts and ±4 amps have already been set). After measuring the mechanicaltime constant of the system T"" you can find the inertia J.346 . In SIMLab.""" ~ e~(*.. To change an input value.m) OC)ub't.c: . Recall that.. you can set the limits very large (or delete these blocks altogether).e plot by using the cursor. QUI L<Kb~~ Molor feedoack Figure 611 DC motor model including voltage and current saturation. we may model the dc motor as a firstorder system.d".~l Figure 610 SIMLab openloop speed response of de motor experiment.
.. ·· · ·· ··· . ·t .........·.\...J Tom. For higher input magnitUdes. 0..tIifOW:1 '....... <.====~=========91 ...... •.......... __ ..del Par...... . Use the Virtual Lab (Q test the following: 7.... ..~ ••• ............. 1 60 . The friction effect is observed when the motor starts....•• ••••• . For both SIMLab and Virtual Lab.... .... I r RiUnAus ·1···· . ~ •••••••••••. ! : 1 ..·.. .••.. ! 20 ....... e. DistOited values may be obtained if the input to the motor is excessive and saturates... .•• . PIOIIO Ft9l..... and record both the motor velocity and .. it is time to apply your knowledge to test the virtual experiment...... ·· .:... Now that you have gained insight into the motor speed response.···········r······ · 70 •• •• • ••••• .... ·· ..... JO ..aLl~ ~C~~r5al tio'fp:':': "''f:=::=:''''''==:... .. your model may vary for different input values. .. Use the mechanical time constant and final value of the response in this case to confinn the system parameters defined in Section 621.. Recall that the motor and amplifier have builtin nonlinear effects due to noise.• ... 9. j._..1ft I OllIput Inp~ ······· ··· ·.. ...• ~.. ..) ~ ~~O~5~~1~7~25~~ 5 Figure 612 Speed response of the openloop system (SIMLab)..).. ·· ·i· .............. How different are the results from Step I? From the transient and steadystate responses.. : j j : 1 1 ~ 50 ~ 40 ...... apply a sine wave with a frequency of 1 rad/sec and amplitude of 1 V to the amplifier input...•..•.... . friction. ·... ...... Titn(l(... .. So.and positioncontrol tasks. : : : i..I 90 ~...•• •• .... and 10 V........ ...........64 Simulation and Virtual Experiments 347 &OttWrlf'lt _"""'~ _ _ _ _ _.•~ • •. · ........ 10 .. These parameters are needed to conduct the speed... (' ..... ··~ .....:.. . \.:. + 15 V... j····..·.~ .... Fig..i...... •.. Enw M. :" ....:. j. • .. •••••••••••• ~ ••• ........ :. Here you have no access to the system pllramt:ler values.. "". j .... ·· .... 642 OpenLoop Sine Input The objective of using openloop sine input is to investigate the frequency response of the motor using both SIMLab and Virtual Lab. .... ......)l'lY'ter~ ..·... S'm"~"o" (... . · .·.. Apply step inputs of + 5 V. and saturation. .•.. ·( · . .l .... in Step 8..... Caution must be taken to ensure that the motor input is low enough such that thi s does not happen. ·· ......_ Run !!oni~JI . The noise at steady state may also be observed......:.. +. • . identify the system model as closely as possible. 8.r...... ....•••••• ••••. the response will saturate.1:1 shows the Virtual Lab motor speed response to a small step input..• •... . .. ... . r .........•••• •• ~ •••• ....... r 'OM.... ..... . :..... .. ···i··· ... ·1.• 1....
. For the Viliual Lab version. 1 } :: •••. Figure 613 Speed response of the openloop system (Virtual Lab). and 50.. Change the input magnitude to 20 V and repeat Step 9. OC ... ···i ..•.....". the saturation values are adjustable to allow you practice with their effect... Doubleclick on the Sine Wave block to modify the properties of the input wave.....0.IJ4.. nd. . Repeat this experiment for frequencies of 0...••. t ..... j •.2...... . . Open Experiment 4: Open Loop Sine Input from the SIMLab or Virtual Lab Experiment menu. mOWf llo lod( t'O morOI plluml1_ S tep O~" rb .IOf2 fCl.. 10.·· .'$ 09 ...••••• :: ... \ ..:! .. ~ ~ 4> ~ :: .• .. · .. · · r. ...·· ... . The SlMLab response for sine input with ~cl"t D ~ iii 09 IIleW s.. •. ~r fjJ ~ 111 ~ ~" lai 'n..: . .. :· ........ <. ...1 M ~t<." .. ··1= ~~U\ [ """"~SO:l!!t....• .." (NII'I) DC mC4orm~4..0.. the amplitude should below to avoid amplifier or armature current saturation..... The Simulink model is shown in Fig. U". hHIld( o.•••!1: .·... 5.npur O.. !.0...• • • • \••. .. ··.t6!:on th.c) : 1). I .. WoO ~.... 0.. ·1"··. ··· . .tn'f....T. ..••••.....lt\~ Oc. @ . rI'I o t~f blt>dt" lld ~ I4'd"'looll . Amplitude of 1 is a low enough value to avoid saturation in this example. . The Control Lab Slmul.> ubt'. • • • :• ••• • • Trm .. "r" .•••.• • • • ' ..... °O~~OS~L~S~~~25~~ .. ·r ·.I' ..(~.. ....348 Chapter 6.• • j• • .·.. 614..... The input and disturbance blocks and the motor parameters are adjustable in the SIMLab model..... sine wave input signals.... . 10..... 2..!!!9'J Plolto Figure r ReuseAtt....1 Figure 614 Experiment 4: Simulink modeL .I... . In the SIMLab version..5. · ·~ . r. " ct "4~1'" To...0 rad/sec (keeping the sine wave amplitude at 1 V).... .fron flOl8 ( s t r s 1 • ..:ntAttiOn f~ Teds Ht:Ip ~:1>. ·'···········1···: '·· · · · · · 06 ·.... ... . ..
" r~Iar. · 100 . For a sine input of magnitude 20 Y.. briefly discllssed at the end of Section 65)..::.... ('11'1'0""""" [ EnlflrModel p..IrotC:::..I 11. .64 Simulation and Virtual Experiments 349 b:peri'Plri ~ I. the Virtual Lab system exhibits saturation as shown in Fig.$.:. ~ 10 Figure 615 SIMLab time response and gain and phase calculation for input = sin(t). magnitude and frequency of 1 is shown in Fig.038 ! ! 6 . 616. .::.h_iJl~r i I_rR8:_A______ "' I I p: tM J I ·60 . . 615 ... I ER«fl~rtdh).n:.:::::====r===============. You may also try adding dead zone and backlash to your motor block to test their effects (these functions are available in the Simulink Library Browser...':= . r.1 """ "" ""' """lC'""" _______d·~1 Si""'I"~n Tom.i······· ..)Ja~~ r """"'~ ~ PWt> loFlgtira rReuseAxn GIItIoN1IJFtIMe~_ _.. . P...:::.6002 ~13 . 10 Figure 616 Virtual Lab time response and gain and phase calculation for input = 20 sin(t).15{) ....':=O"".•.
4>~·./lU. 10. 62. you can record the gain and phase of the response . 65 will be displayed. There is a Gain and Phase Calculator in the Experiment 4 control window.ot"\. ~ _ O \ ~ liiI ol r t. ' '''t t60 ..t :. For proportional gains of 1.. Apply step inputs of + 5 V. Repeat with other input frequencies..t.. Doubleclicking on the PlD block displays the editable PlD values... " . s:l i:i~~~ m~ Input TO'WOMp . 13. and 10 V (try dc motor alone. .lo. we can now extend the model to include velocity feedback from the motor and use a proportional controller. To measure the magnitude and phase of the steadystate response. perform the following tests using SIMLab: 11. p... ~ Fa ..2: 1) of 0.. Repeat Step 11 using additional load inertia at the output shaft of the gear head (gear ratio 5.1I\. 643 Speed Control Having simulated the openloop motor characteristics in previous sections. 14.. 617...~"'n~u \IIld('tou..... ". and observe the effect of the closedloop control. .. and discuss any trends. Apply the same viscous friction to the output shaft as obtained in Step 4 in Section 641. select the Enter Model Parameters button to get the system Simulink model. Edt .05 kg_m 2 (requires adjustment of the J value in SIMLab motor parameter block). By how much does the speed change? Repeat Step 5 in Section 641.1!i11 '''''''' r Figure 617 Experiment 1: Simu1ink model. . Assume that the motor velocity is measured using a sensor that provides I V/rad/sec . Next.. 615). The Control Lab The frequency of the sine wave will dictate the gain and phase of the response curve. 12. ul 1>4101'.' DC'lno(It. Using the Gain and Phase Calculator." RiOtlhlld(utMOCl'Tlbi tl bfodcl n .... no gear head or load applied in this case).. This figure is a simple PID speedcontrol model.... .·''Ic:k. and 100. + 15 V.ditf 1I.. and compare the results.350 Chapter 6. enter a frequency of 1 rad/sec in the edit block.. The block diagram that you should be modeling is shown in Fig. . T ... A screen similar to Fig.~l ~ ·r~.. \~"'.. as shown in Fig. I .nmulJt!o ll ~". Entering the input frequency and clicking on Calculate displays the gain and phase of the system (see Fig. Open Experiment 1: Speed Control from the SIMLab menu window. The values of F.
This was done intentionally to illustrate the scaling (or cOtlVersion) that is pelformed by the sensor... because the de motor can utilize unlimited voltage and current levels. Doubleclick both the voltage and current blocks and adjust their values to very large (or delete their blocks). the stepinput and the disturbancetorque blocks may also be adjusted. To change an input value. simply change the number in the final value field. but assume that the velocity feedback signal is a voltage generated by a sensor . 611 . °O ~O~I~O~2~O~3~~O. p = ! P = 10 P = 100 1= 0 1= 0 I= O D =0 D = 0 D = 0 The input units used in these simulations are specified in volts.6~~O~7~O~8~O~9J ~ Tim. Increasing the prop0l1ional gai[l in lht: PID block will decrease the rise time. 618 displays a typical speed response from the SIMLab. For an unsaturated model..e".. To create this effect in software. a different response would have been obtained. Recall from Section 641 that the defaul t saturation limits are ± 10 V and ±4 A. For a given input to change the proportional gain values. Fig.~~05~~C. Had the velocity been specified in volts per radians per second. enter the following sets of PID values and print all three plots on the same figure (use the Print to Figure button and the Reuse Axes checkbox in the experiment control panel). The disturbancetorgue default value is set to zero . rightclick the dc motor block. respectively.64 Simulation and Virtual Experiments 351 """'''.. and select Look under Mask to obtain the motor model similar to Fig. while the feedback units at the motor are in radians per second. repeat the preceding experiments using a proportional gain of 10. To check the effect of the velocity feedback scaling... the SIMLab version of this experiment could exhibit extremely fast rise times at very high proportional gains..) BlC:I~1 Figure 61 8 Speedcontrol response in the STMLab control window.
For the motor alone. and compare the time response for a constant proportional gain. 18. Using the disturbance torque in Step 17. Choose several different integral gain values. Doubleclicking on the PID block allows you to edit the PID gain values. 644 Position Control Next. A screen similar to Fig. What is the maximum step input amplitude that will meet these ca1culalt=u requirements (i. 21. choose Experiment 2: Position Control from the Experiment menu. 20. Open Experiment 2: Position Control from the SIMLab Experiment menu. and 10 V. select Enter Model Parameters to get the system Simulink model. 22. + 15 V. The Control lab with a conversion factor of 0. 619. and 100 (requires modification of PID block parameters).2: 1 (requires modification of Jm and B in the motor parameters). In all previous cases..e. Examine the effect of voltage and current saturation blocks (requires modification of the saturation blocks in the motor model). comment on the validity of Eq. Chapter 6. respectively. This model represents a simple PID positioncontrol system. The Deg to Rad and Rad to Deg gain blocks convert the input and the output . (Note: in commonly used industry standards. not cause the amplifier to saturate). 10.) Next. test the following: 15. Next. for the Virtual Lab. and plot the different simulation results in an external figure for comparison. Estimate the disturbancetorque gain based on your observations. apply a 160 step input. Apply a step disturbance torque (0.352 ~. 65 will be displayed. How does an increase in J affect the system with a PI controller? Compare the transient and steadystate response. using additional load inertia at the output shaft of 0. given the default current and voltage saturation limits of ±4 A and ± 10 V. investigate the closedloop position response. examine the effect of integral control by modifying the Simulink PID block.2 V/rad/sec. (5126). perform the following tests using SIMLab: 16. How different are the results from the SIMLab? You may again confinn the system parameters obtained in Section 641.1 seconds. Apply step inputs of +5 V. as shown in Fig. \Vhat can be said about the effect of the increased load on the system perfonnance? 19. Select the Reuse Axes checkbox. Design a PI controller that will give a 30% overshoot and a rise time of 0.05 kgm2 and the gear ratio 5. For proportional gains of 1. How large is the error when the 0 system reaches steady state? 17.2 23. Eliminate the disturbance torque and repeat Step 16.1) and repeat Step 16. the tachometer gain is in volts per RPM.
'....... Fig. The values of the stepinput and the disturbancetorque blocks are also adjustable.'iVl JI~I ~I1'~ tI._..riCol'l th.. ::::m.t. To change an input value._ . .j • • r IN. Figure 619 Experiment 2: Position Control Simulink mode l. such that the user enters inputs and receives outputs in degrees only.:"!Jj ~ lli ~~ r~~~I~ r OwOllclp.\ d.hOll PI>!I to Figure 270 I Posril(ln Re spo~ e I 1 r Reus. _ :..lnOIO ' bJodi".. .j!"... ~ . ... hckonU•• (J C lI"tot. .OC.Q.I I r~I I S!opSomul. ..T.. : ...______ .. Modet Parin'@UHIl I 120 ____. (.._ cros~ E~ penm8 . The di sturbancetorque default val ue is set to zero...~.' ~ 15 Ttrne (9) 25 AICl ~ 1 Figure 620 Position response in the Experiment 2 control window. _. frrU:!I.ew S~ FonnJt lools He1p DI~ iii ~    . ..'Il .. ~.IQt.! . ~ .. th4 1U"ul.. doubleclick on the relevant block and change the number in the final value field ..·e1.0 60 ___ .._ ..64 Simulation and Virtual Experiments 4 353 f:'e Edt \I. : _..r 10 ' .. F. _ ..~..····· ·· ..ct ~lo(llr.. _. 1 H 'S lip OIrtU(blflOot (tIm) poub\4..5 · ~ ··.. ...~~ [ Enle..ut c.. .. "....._ ..)r bl ~dtto lh~ lfIodlfy n.lIl On OC molurno. Axes ." .... l . . . 620 displays a typical position response from the SIMLab. ..
A solid disk is attached to the end of the beam through a magnetic device (e. If the magnet is on. (25. The time. 10. You are encouraged to apply the methods that you have learned throughout this book. allows introduction of a disturbance function or changes of the system parameters if necessary. 24. The arm may not overshoot the desired position by more than 50. The system is composed of the de motor used throughout this chapter. the disk is released.. and outputangle values are displayed on the animation field.g. to design a controller for your system. 620. For proportional gains of 1. We connect a rigid beam to the motor shaft to create a simple robotic system conducting a pickandplace operation.ROBOTIC ARM The primary goal of this section is to help you gain experience in applying your control knowledge to a practical problem. The animation tools provided make this experience more realistic. in particular. 0 I Payload I / Figure 621 Control of a simple robotic arm and a payload. and when the magnet is turned off. The nonlinearities due to voltage and current limits cause the time response to saturate at a high enough proportional gain. particularly in Chapter 5 and later on in Chapter 9. The maximum speed and acceleration of the dc motor are dictated by the voltage and current saturation limits. Objective: The objective is to drop the disk into a hole as fast as possible. 65 DESIGN PROJECT l .4 mm) below the disk (see Fig. repeat Steps 16 and 19 using Virtual Lab. This is a useful tool that gives the user a physical sense of how a real motor turns. a solenoid). the disk will stick to the beam. The hole location may be anywhere within an angular range of 20 to 180 0 from the initial position. The SIMLab. Description of the Project: Consider the system in Fig. The Control Lab The position time response is also animated when the simulation is run. inputangle. The hole is 1 in. . Design Criteria: The arm is required to move in only one direction from the initial position. The project may be more exciting if it is conducted by teams on a competitive basis. 621. as shown in Fig. and 100 (requires modification of PID block parameters). The SIMLab and Vhtual Lab software are designed to provide enough flexibility to test various scenarios. 622).354 ~ Chapter 6.
As in Section 64.II ngll . this figure represents a Sit.lI lh . It is highly recommended that you do the design project only after fully appreciating the earlier experiments in this chapter and after understanding Chapter 5. ·5 1 .(I/) r>orit!oll JI /I~I. be tnOoto.""Ylldt· 1r. In.. you can also introduce a disturbance torque to alter the final value properties of the system.. The objective may be met by looking at the settling time as a key design criterion. 623. The idea of this experiment is to get a metal object attached to a robot arm by an electromagnet from position 0° to a specified angular position with a specified overshoot and minimum overall time. Olnuib. A screen similar to Fig.65 Design Project lRobotic Arm 50S lllm (20") 355 ~I Electromagnet Metal puck 25.. Select Experiment 5: Control System Design from the SIMLab Experiment menu.. In SlMLab. from earlier experiments. These criteria may easily be altered to create a new scenario. as shown in Fig. Have fun! This experiment is similar to the positioncontrol experiment in some respects.ltlO. However.l ll"tt II "'1I1I41..oltdcwtnDd lt>! l"O!'l11 ot o r p"{~"l llw ~EI.m) Double· elide (1 1'111'1.1. Next. nU'' I) .4 .d 1.. You may try to confiml the system model parameters first.. you may make the design challenge more interesting by introducing other design constraints such as the percent overshoot and rise time.~ t Cl th. The Virtual Lab system contains nonlinear effects that make the controller design more challenging.ub tion ' '.4 mm (I ")[. . mod • • 'h'i h~ olod! 011 thl DC mo lol b1odo. A tolerance of ± 2% is acceptable (settling time).oIJh ff)r d lo pp oritiOft ~tl .· It tho ..l AC.d Figure 623 Experiment 5: Simulink mode l. .:If.1 _______ Robot arm DC Motor Figure 622 Side view of the robot arm .lnd s:tltct "L.aIIM ". rtf.lo nt i:l'l ' II m.lim'''''''. 65 will be displayed.t tIl G U ·· !$ " )fbl. .u. mtlll \kick De ""ow... select Enter Model Parameters to get the system Simulink model. (fl.
311. the " Drop position angle" is the angle where the electromagnet turns off...······~ ·· ··········i·· ····· · · O~~O~5~~1~5~~~'5~~ O Figure 625 PosiLion response for Experiment 5." An important note to remember is that in the Virtual Lab the electromagnet will never drop the object exactly where it is specified... dropping the payload. 624.. " Start to wait for drop position at time" refers to the time where the position trigger starts to wait for the position specified by "Drop position angle.. respectively..356 Chapter 6...~ DrOll position angle 1115 S t~rt 10. This feature is particularly useflll if the response overshoots and passes through the target more than once. 11 6 deg : : : ······ . a parameter window pops up. Stop$'(fIl. which allows the user to adjust the dropoff payload location and the time delay (in seconds) to turn the magnet off after reaching the target.. the magnet holds on for a short time after the trigger is tripped. rRc u!le~$ J  ::KC 'TIme: 0 .4 1$ · 1_~:Ul ~~oropped j .. simple PID positioncontrol model with the same functionalities..05. .. A time response of the system for proportional gain and derivative gain of 3 and 0.f ~Smk Block Parameters: Electromagnet Conttol Sub~stem [mask] Parameters . . The added feature in this model is the electromagnetic control. 625 .·~···· . I I : : 50 ···· rr·· · · · ·:·· ~ r0s(uon.3 to wait for drop positron allim. in Fig.. Because any electromagnet has residual magnetism even after the CU1Tent stops flowing..". The Control Lab .. !:lelp Figure 624 Parameter window for the electromagnet control... . 624. By doubleclicking the Electromagnet Control block.itSlICt'I . .=:I l I _ I 1oF. is shown in Fig. So.~.. as in Fig.
as illustrated in Fig. 626(c). However. you are now well acquainted with the use of the ACSYS tools and Simulink and their applications in the study of controls. In this case. Choose Modify Puck Drop Setup from the SIMLab Tools menu to adjust the height of the drop and the length of the arm. we will assume a rigid wheel. I (6~1) where the base acceleration A(s) is the input and relative displacement. for simplicity. As discussed in Chapter 4. T is the torque produced by the motor with shaft position 8. 627. This feature makes the problem more realistic. in this case. Hence. is the output. and r is the radius of the motor drive gear. 626(a).wn s + w.. 626.e. . 626(c). where a two degree of freedom (2DoF) system in Fig. the magnet dropoff takes place prematurely. there are various representations of a quartercar system. it is customary to ignore tire dynamics and assume a lDoF model as shown in Fig. shown in Fig. and change your controller design accordingly.66 Design Project 2UuarterCar Model ~ 357 The model response is also animated. We further assume hereafter the following parameter values for the system illustrated in Fig. The objective here is to control the resulting displacement or acceleration of the mass of the systemwhich is reflective of the chassis of the car. Note that. Let us next consider the active control suspension system and use the same dc motor described ill Section 62 used in conjunction with a rack as shown in Fig. The drop angle and drop time are displayed on the timeresponse plot. 66 DESIGN PROJECT 2QUARTERCAR MODEL 661 Introduction to the QuarterCar Model After studying position and velocity control of the dc motor in the preceding sections of this chapter.9135 NrnIs· 1 m m m rnls2 Recall from Eq. 6~26 (c) is presented for studying base excitation response (i. as shown in Fig. The puck has overshot the hole in this case.7135 N/m 0. the payload has been released earlier and is not on target! In SIMLab. for the duration of this design project. ~ m k c x(t) y(t} =(1) a(t) Effective 114 car mass Effective stiffness Effective damping Absolute displacement of the mass m Absolute displacement of the base Relative displacement (x(t}y(t» Base acceleration 5'(1) 10 kg 2. This study follows the modeling exercise that was discussed in Example 4113. (4324) that the open loop response of the system to a base acceleration a(t) has a transfer function: Z(s) A(s) = s2 + 2l. 626(b) takes into account the damping and elastic properties of the tire. Z(s). it is possible to change the dimensions of the experiment setup. In this section a simple one degree of freedom quartercar model. As a result. road transmitted effects).
Using superposition. Figure 627 Active control of the IDOF model via a dc motor and rack. and K = kr2. (c) One degree of a) b) c) freedom . Z(s ) + Figure 628 Block diagram of an armaturecontrolled dc motor. .I I 3 that the block diagram in Fig. KK (~ + I R" Ra a S ( L) K K mrA s) ( . (a) Quarter car. this system is rearranged to the following form: Klltr Z(s) Ra L) (Js2 + Bs + K) + ~ s V (SJ. 628 represents the open loop system with no base excitation.3:. where J = mr2 + Jill ' B = cr2 + Bm.358 Chapter 6. (b) Two degrees of freedo m. s + 1 (Js2+ Bs+K)+ ~ s Rli Ra ( La s + Ra l)r (62) mrA(s) E(s) K". Recall from Example 4. The Control Lab 'rr' 1 Xs 1 x Figure 626 Quartercar model realization.
66 Design Project 20uarterCar Model 359 Vi. where I« is the sensor gain. 62 and is reduced to Z (s) = 2 S + Raj BRa + KmKb RcJ S +J K Va( s) 2 S J BRa + KmKb + Raj S +J K A(s) (63) For simplicity. 630: Z (s) J . where a robot arm is given the command signal to displace a metal puck. but rather to reject the socalled disturbance input. (63) is written as Z(s) = Geql(s)V. armaturecontrolled de motor. .Ks 2(s). as in previous sections.. we assume the motor electrictime constant is insig11ificant relative to the mechanicaltime constant. the sensor is a linear variable differential transformer (LVDT). Hence the command voltage. Again. Eq . 629 can be reduced to an inputoutput relation of Yes) and Z(s). Vin(s) = 0 V. or set point.Geqz(s)A(s). The posi tion control block diagram in Fig. 629 illustrates the feedback of relative position. Z(s) .A(s) (64) 662 Closedloop Acceleration Control Relative position control is a familiar way to introduce the control of the quartercar model. In this application. Eq. where the simplified closedloop system is represented in Fig. which transforms the displacement z(t) between the base yet) and mass x(t) to voltage. Setting £(s) = 0 .cept perhaps by comparing their height to fixed objects.M) . with units VIm. the block diagram represented in Fig. If you have ever driven a car too quickly over A (s) J + Z(s) Figure 6·30 Simplified block diagram of the qu artercar dc motor pOSition control. the vehicle operator cannot really sense displacement ex.(S) + Figure 629 Block diagram of a positioncontrol. however. The goal of position control in this scenario is not to create offset as in the previous lab.
(66). where the controller design topics are studied in more detail. slightly more complex system. two accelerometers with gain Ks are fixed to the mass and base to provide the relative acceleration feedback. 663 Description of Quarter Car Modeling Tool The Quarter Car Modeling Tool allows the students to implement the familiar dc motor and ampliher described in Section 621 and conduct experiments to observe its effect on a new.360 ~ Chapter 6. as demonstrated in Chapter 5. The block diagram in Fig. the position control system used an LVDT to provide the feedback.A(s) 2 S + BRa Raj + KmKb KmKampKsrO 2 S+ J + Raj cS J l mr . 632(a) is Simplified to the closedloop form in Fig. 632(b) to obtain the inputoutput relation X(s) A(s) (66) Note that. designing in the time domain may require that the systems be approximated by lowerorder systems. you can feel the effect of acceleration in your stomach. to control the relative acceleration of the mass. Thus. it is more desirable to control the acceleration because the ultimate goal of the suspension system is to improve ride and driving performance. 630 can be modified to control the relative acceleration of the system. The block diagram in Fig. The second derivative of the forwardpath transfer function yields the acceleration control system in Fig. two accelerometers can be used to measure both x(t) and y(t). a sharp rise and fall of the road. The inputoutput relation is as follows: t(s) = . !5... 631. It is also of interest to control the absolute acceleration of the mass /"1'1. sLlch as driving . (65) As described above. implementing a compensator will lead to a higherorder transfer functjon. Just as the LVDT measures relative displacement. The Control Lab A(s) 2(s) + Figure 631 Simplified block diagram of the quartercar relative acceleration control. (65) and Eq. Designing a controller for a vehicle suspension system requires studying its peIformance under the influence of different inputs. in the case of the systems represented by Eq. In this case. Thus.Y(t) . Also see Chapter 9. The closedloop system is determined by reconfiguring Fig. where z(t) = x(t) . 629 to yield absolute acceleration from the relation X(s) = Z(s) + A(s) where Z(s) and X(s) are the LapJace transforms of z(t) and x(t). respectively.
All of these features are available in one simple window which automatically controls the Simulink model.66 Design Project 2.. over a curb or speed bump.(S) irs) + (a) A (s) Xes) + (b) Figure 6·32 Block diagrnm of the absolute acceleration control system. 633) and MATLAB graphical user interface. To start the program. This launches both a Simulink model file (Fig.QuarterCa r Model ~ 361 Vj. such as backlash and saturation in the Virtual Lab component. to be used as Figure 6·33 Quarter Car Modeling Tool top· level Simulink model. This tool also incorporates nonlinear effects. . click on Quarter Car Sim on the ACSYS applet.
. or loaded from.'""'~"". the control panel (Fig.. v_ M'O'<l 113 ....'" I .."" I . ..:::... Model settings may be saved to. 634)...'."'" ~ r r Cortrclp:jfl"'li'tl:'~ f .. impulse.. The Control Lab eo. 635) from which you can modify the parameters of the dc motor.!f.'>'... I I ""'...F. I <>n.... I ~ 5.. sin. The compensator frame allows the user to select the sensor output to be used ...362 Cha pte r 6...)c~rve SuspcnCICli =t I 3  " "....JtJJiJ v_ ""'"'000>. There is very little need to access the Simulink model.." Psramei.J I d ~eIrotlb' . rounded step. rounded pulse. Clicking on Defaults assigns the default values to the parameters.mat file may be selected in the left IMPORT frame and then assigned to the selected model parameters in the right MODEL frame... There are a number of inputs to select from: step. Selecting Control Parameters calls a window (Fig. .. and of course the quartercar modeL Parameters from the workspace or a . ~ W! .....mat files..tlrk:lP'X'. .J """"..O. uon nm.. sensor gains. I """ .':".......rS... From the control window. f I ] ~ r". ~ ~IIICkI'i~ ~· 1 r I ~ (:' \ll.. R"P"" S1cP Slmll!lron I __ I¥...L~"".. Figure 634 Quarter Car Modeling Tool control window. amplifier. Click Apply to implement your changes or Close to cancel.. and random..... other than to reference the model or to modify the simulation parameters. 636) used to configure the compensator command signal in the left frame and the compensator in the right pane..mut.w· Ent."". I Figure 635 Mode l Parameters window. clicking on Model Parameters brings up a window (Fig.
Note that the control menus may be dragged off the axes by clicking and dragging the top bar or closed by clicking the X in the top right corner.66 Design Project 2QuarterCar Model 363 . PID gains are specified in array format.. The various transfer functions of the form displayed in Fig. which allows the user to toggle to different experiment modes. The passive suspension system operates as a spring and damper." Figure 637 Closedloop transfer functions . The zoom control and cursor buttons appear at the bottom right comer of the display panel.1 controller Parameters DISTURBANCE •~ COMPENSATOR [step __ _3 (m/s"2) Feedback. A(s) . Click Apply to implement your changes or Close to cancel. Z( s). click Print to Figure. ~c . using the SISOTool button. The MATLAB S1S0 Design Tool may be activated. +_ rr. To store the input/output plots on a new figure. without the added control of the dc motor. Below the progress bar is a popup menu. AmpJ~ude [PIDJ:I_ . " Compen'S8lor Tf: INone ::J Step Time (s) S1S0Too) Apply Close Figure 636 Control Parameters window. Also. with the appropriate system transfer functions automatically loaded. At the top right corner of both the upper and lower axes. and any existing transfer function object may be selected from the workspace for use as a controller via the dropdown menu. Click on Simulate Response to begin the simulation. ~i'4. . Once the model amI cuntruller parameters are specified. Click on Stop Simulation to stop the animation and simulation. the system is ready tor simulation. as seen in Fig. This will start the animation and plot the data on the upper and lower graphs. 638. 637 can be selected from the popup menu. lz ::J [275012) . as feedback. The active suspension system is the de motorcontrolled system from Fig. TIle closedloop transfer functions of the system are displayed in the top right comer.::Ic 'L2J Z(s) . 627. pressing the Setup Axes button will display a small control menu that is used to select which data are to be displayed on the graph.X(s) I ' cc .
.1~ ~ ...0 31882 Litle 3 ...~ !:.. Experiment with the stiffness and damping of the system by clicking Model Parameters and changing the stiffness.!j t. Click Simulate Response..· . • 'ieu."t't Un.• • I•...... 7 " I :~:+I:::::i:::::j/:= : .....:::::...... the result will appear similar to the window di splayed in Fig.. from the default 2....!."" ".. .• •. Click on Control Parameters. and select a step input with amplitude 0. 639.::. ..:::~: : t.. click on Setup Axes.. There is no need to configure the compensator si nce it is not used in a passive system .... How does the period of oscillation compare to the value that was observed in Step 1" Open the Setup Axes menu for the lower axes and click the Print to Figure button.. [!•~ 1..~ ..0 I m/s2 and step time a seconds.. When finished...t. This data can be accessed in the Setup Axes menu for either axis. The Control Lab ( ... 2. 2. Set the simulation mode to Passive Suspension and set up the top axes to display )i(t)... Note the shape ofthe road profile y(t) as well as z(t) and x(t).. ~~: ::J .. ~ ..:.364 Chapter 6. and the lower axes configured to display i( l ).F. . With a step input of 0....2.. 664 Passive Suspension The following experiments explore the response of the openloop qua1tercar model..01 mls2 .... choose Acceleration from the dropdown menu .•l• l:f•. what is the frequency of the oscillatory response? This is the damped frequency of the system using default parameters (Wd) .+ i. To accomplish this. Repeat this procedure for 0. Run the following tests in order: 1..• Figure 638 A view of the axes controls. 71 ~~0!! 1 :l I_ _ _ _ _ _ _ :::l ' __ 1 . + +." Using the same method. . This will plot the data on the axes to an .. and click the checkbox labeled "y...T T II I1 : 1k.. Studying the response of the passive system is essential to understanding the effect of various inputs on the quality of the vehicle's ride and is necessary for appreciating the effect of adding compensators to actively control the suspension system.1 and 1 m/s2 input. k ... ~. j. ~ . ...[:: :~m i j. ..17 NI111 to 10 N/m..F .. • • • • • • • • • #< • • • • "' ..o r _ 3l . configure the bottom axes to di splay z(t) ...
... • • • . gradually reducing the damping (variable c in the Model Parameters control window)... . and click Pri llL LU Figure to plot data from a new simulation in the same externat figure ..... i j l O'Olf:7:~'___. .1 rad/s.~ ....... 5... then Defau lts...... ~ ..01 mfs2 and vary the frequency from 5 rad/s to 0.... and observe the response for each case... _•• . . 3. Calculate values of c such that the system is underdamped (~ = 0.. This emulates driving over a pothole if given amplitude less than zero. critically damped (~= 1).. set the Feedback popup menu to z. .. Figure 639 Passive system response to a step input. Now try using tbe rounded step input with ampli tude 0. Repeat Step 4 using the unidirectional rounded pulse (URP) input (amplitude 0. Study the effect of a si nusoidal input (washboard bumps) on the response of the system . then Apply) .1: ~.tn~ 1 •••.y(t). ..••__ _••• ... Set the system parameters to the default settings (click Model Parameters..1 m and duration 0. Wd. . ~ . This is useful for comparison. Select an amplitude of 0. Repeat the simulation several more times. duration 0. This activates the feedback control system as defined in the Control Parameters window.. The frequency of the oscillation when c is reduced to 0 is the natural frequency of the system (w n )..01 seconds... This input function simulates driving the quartercar model over a curb. thus. • . __ .. .....1 m..~... external figure . Select the checkbox labeled Reuse Axes... .. and overdamped (~ = 1.•• . Relative position control is the control of z(t) = x(t) .5).. What happens to the amplitude of relative di splacement at the damped natural frequency.. The set point is . measured in Step 27 4..BB Design Project 2QuarterCar Model 365 o.01 st::t:unds) . . or a speed bump if the amplitude is positive.... .•••... 665 Closedloop Relative Position Control Now set the simulation mode to active suspension by selecting Active Suspension from the control panel dropdown menu or from the Experiment menu.1).
8 .._ .. _.. . ...5 ~ .. because it is the acceleration of the vehicle that affects the comfort of the ride. PID :::: [1 00].... _ t ..... ...: ~.. _•• _... . . Plot successive trials to an external figure for comparison..kf ... .:... 640. _:. . . . . .. ... •• .. 30... and set the feedback to in Control Parameters.. . ... • t ... _! ...:. ... What value of Kp (PlD = [Kp 0 0]) will yield a steadystate error less than 5 mm? This will require a gain much higher than 1.... ~ . Set the simulation mode to Active Suspension.. 10........L j .. '" . ~ ...."" .. This causes z .. : : : : : : : • • • • .._.. ... Simulate the response... 1·" • I * • :II I 4 • I • ~ . • t : : ! .. ... ~. .. ~. ! .. ~.. . .~ ... ~ .. : : : J . _.... I I • : : 1 : : l 1 : *... ~ _ _ to . ... . ......... ___ .... . Design four phaselead compensators with sufficient gain Kp to meet the steadystate error requirement as specified in Step 7 and with phase margins CPm...~ . Do the values correspond? Validate the observed overshoot and rise time using the timedomain analysis techniques introduced in Chapter 5.. both a PD and phaselag controller wiIl be implemented.L.:. . • I • I I . __ .. :.... ~ • • • • I 4 • I . ...366 ~ Chapter 6.:.~.. Vin(t) = 0 V..~ .. I : • 4 .. ·4 : : : : ! . ... . and 60°.. feedback z.. .. ~ :. ... ..... and 5.. ...~ ..§.L. }~~i~~t*~~t....... . .~ ~ 1 I I • 4 I : : : : : ~ _. Compare the optimized response with the PD response. a'i in Fig. Again......··..... 4 r. ___ ! .:..... validate these results using timedomain analysis techniques... ....... What happens to the system overshoot at this gain? Does this match your calculations? Increase the derivative gain in steps from 0 to 22~ keeping the proportional gain that was found in Step 7.. ... __ ..........:~. ... : . 9..r··KdO'·'1 . . .. Test your controller's response to the inputs applied in Steps 3...... In this section.. . = 1 VIm. .. What is the steadystate error? mTl Apply the final value theorem to Z(S()) A s = 2 ...e 1 m/s2 amI slep lime 0 seconds.____ ·7 . :r'·'· .... ~ • : • o Figure 6·40 Relative Position Time Response plot........ click SISOTool in the Control Parameters window and increase the gain while observing the LTI Viewer step response.. _____ . ........ ·~{ ~=5 : K . 1.~ =15 i .... ··r· _ •·r·· ···_·T ··'·····r···. 6. to .... . .. 40.. ·:·······:·······1! '~ ..... ""'""..... 1. .. ..... 7. .. .. it is preferable to control the acceleration of the mass m.... ... . t ~ I . : 1:1: : ~.... ... ~ ...J. __ . . of 10... where feedback gain is /(.9 .......::.10 J ~ I : .. To reduce the need for trial and error. 4.. _______ . Click on Control Parameters and select a slep wslurbam...... ...! ~ I 4 4 • I • • I • . •. ..~....l....... ·1'········r·······:····· ...6 \..... .....~: . 666 ClosedLoop Acceleration Control As mentioned previously.~...... ____ .. t. T·· . . ~ ... The Control Lab o 1 2 X 10'3 CarSlm Relative Position Time Response _····r·'·'·[" .K =20 : : .f + BRa + KnrKb Raj S J +J + K KmKampl(ff G Raj ...... :. . and observe the effect of adding derivative gain.._! ... . L.... 8. _..........to_ ~. ~......:_______ .... ..._.
This creates a transfer function object in the workspace. a dc motor) and mechanical (gears) and electrical components . For the phaselag compensator..1 I SISOTool i. implement it by using the following MATLAB script: K = 5.g. What gain achieves the smallest rise time? Test this gain on the actual controller. and rounded pulse. change the PID gains to PID = [5 00].1 rad/s . [T 1]) 12. 641). 13.1996 PhaseLag = tf([a*T 1]. with gain K = 5 and a>~ = 0. T = 231. In the Control Parameters window. Use the LTI Viewer to monitor the step response while tweaking the gain. Design a phaselag compensator. in a realistic system including an ac tuator (e. rounded step. Repeat Steps 11 and 12 using absolute acceleration as the feedback (Feedback x). a = 0. 67 SUMMARY In thi s chapter.0.COMPENSA TOR 367 _ CJ ~ I. try K = 5 and a>~ . K. clicking on Print to Figure in the Setup Axes menu to save the results for comparison. Close J Figure 641 The Controller Parameters window_ the relative acceleration.1 g I :4 II f . Once the compensator is designed.<= I Vlm/s2 to be compared with the set point 11. = Test the controllers designed in the last few steps with variolls inputs sllch as sinusoidal.1 rad/s. Compare the step response to the response in Step 11. Any transfer function object created in the MATLAB workspace is accessible in this menu and can be used in place of the PID controller Ge(s) . z(t) . We discussed that. we described the SIMLab and Virtual Lab software to improve your understanding of control and (0 provide a better practical appreciation of the subject. Now click on Control Parameters and select PhaseLag as the compensator in the Compensator TF dropdown and enter K as the proportional gain (Fig. with gain Vinet) = Ov. f step Time (s) ! Compensator TF: IphaseLag . Try designing various phaselag controllers as per Section 962 and compare the results. and click SrSOTool to apply the changes and launch the S1SO Design Tool.67 Summary ) [ontroller Patilmeters . : DISTURBANCE Istep (m/sN2)  _:::1 Feedback: Iz ::::::I [K 0 O J i I Ampl~ude [P J Dj: I i i . 14.1442.
You may wish to visit Chapter 9 to become more acquainted with these topiCS. and record the motor speed and current for 10 sec following the step input. What is the steadystate speed? (a) How long does it take the motor to reach 63% of its steadystate speed? (b) How long does it take the motor to reach 75% of its steadystate speed? (e) What is the max. 2008. Golnaraghi. friction in the motor. We pre. and Kb = O. Set the viscous friction B to zero in Problem 61.5 sec (requires change of the disturbance block parameters in SIMLab). PROBLEMS 61. aside from the speed and position control topics.008 H. Mechatronic Systems Engineering Program. or backlash in gears will seriously affect the controller design.001 Nmlrad/sec. "ENSC 383 Laboratory Experiment:' Simon Fraser University. Apply a SV step input to the motor.imum current drawn by the motor? (d) What is the steadystate speed when the applied voltage is 10 V? 63. because of their simplicity and wide use in industrial applications. Note that.368 . F. This chapter focused on problems involving dc motors including modeling. Apply a 5V step input to the motor. and record the motor speed and current drawn by the motor for 10 sec following the step input. how long does it take the motor to reach 75% of its steady~state speed following a 5V step voltage input? 64. Km =: O. British Columbia. issues such as saturation of the amplifier. Use the following parameter values: J m = 0. I VIradJ sec. (a) What is the steady~state speed? (b) How long does it take the motor to reach 63% of its steadystate speed? (e) How long does it take the motor to reach 75% of its steadystate speed? (d) What is the maximum current drawn by the motor? 62. Apply a 5V step input to the motor. 525. Lab Manual. Assume that the load torque TL is zero. I NmfA. The focus on dc motors in these experiments was intentional. B = 0. and controller design. Chapter 6. .1 Nm (don't forget the minus sign) starting after 0. other controllers such as PID and leadllag were also discussed. The Control lab (amplifiers).0004 kg_m 2 . in the design projects. and record the motor speed and the current drawn by the motor (requires modification of SIMLab blocks by making current the output) for 10 sec following the step input. REFERENCE I. La = 0. Canada. Set the armature inductance 41 to zero in Problem 62. system identification. ReI = 2 n. (a) How does the steadystate speed change once TL is added? (b) How long does it take the motor to reach 63% of its new steadystate speed? (c) How long does it take the motor to reach 75% of its new steadystate speed? (d) What is the maximum current drawn by the motor'? (e) Increase TL and further discuss its effect on the speed response. Repeat Problems 6~1 through 63. how long does it take the motor to reach 63% of its steadystate speed following a 5~V step voltage input? (f) If Jm is increased by a factor of 2. and assume the load torque TL = 0. followed by two controller design projects involving control of a simple robotic system and control of a single degree of freedom quartercar model. Create a model of the motor shown in Fig."ented experiment" on speed and position control of de motors. (a) What is the steadystate speed? (b) How long does it take the motor to reach 63% of its steadystate speed? (e) How long does it take the motor to reach 75% of its steadystate speed? (d) What is the maximum current drawn by the motor? (e) If Jm is increased by a factor of 2.
. (a) What is the steadystate speed? (b) How long does it take for the motor to reach 63% of its steadYftstate speed? (c) How long does it take for the motor to reach 75% of its steadystate speed'? (d) What is the maximum current drawn by the motor? (e) How does increasing Kp affect the response (with and without saturation effect in the SIMLab model)? 69.1 Nm starting after 0. select K{J = 1.0. apply a 10 rad/sec step input. (a) How does the steadystate speed change once TL is added? (b) How long does it take the motor to reach 63% of its new steadystate speed? (e) How long does it take the motor to reach 75% of its new steady~state speed? 6·11. What is the steadystate speed when the amplifier input voltage is 5 V? 6·7. apply a 10 rad/sec step input. Insert a velocity sensor lransfer function Ks in the feedback loop. state speed? (c) How long does it take the motor to reach 75% of its new steadystate speed? (d) What is the maximum current drawn by the motor? (e) Increase TL and further discuss its effect on the speed response.2 Nm starting after 1 sec (requires change of the disturbance block parameters in SIMLab). 6·6.5 sec following the step input. For the system in Fig. Repeat Problem 67. 6~ 1. (a) At what time after the step does the maximum occur? (b) What is the maximum rotation? . Use the same motor parameters as in Problem 6. 63. and record the motor position for 1 sec. use the parameters for Problem 6·] (but set Lu = 0) and an amplifier gain of 2 to drive the motor (ignore the amplifier voltage and current limitations for the time being).5 sec (requires change of the disturbance block parameters in SIMLab). 1.Problems . (a) How does the steaoy"state speed change once T1• is added? (b) How long does it take the motor to reach 63% of its new steadystate speed? (e) How long does it take the motor to reach 75% of its new steadystate speed? 610. it take the motor to reach 75% of its steadystate speed? (d) What is the maximum current drawn by the motor? 68.. 6·12. (a) What is the steadystate position? (b) What is the maximum rotation? (c) At what time after the step does the maximum occur? <i"13...2 V/radlsec (requires adjustment of the SIMLab model). apply a 1 rad step input.... For the system in Fig. (a) How does the steady"state speed change once TL is added? (b) How long does it take the motor to reach 63% of its Hew steady . and assume the load torque TL = 0. (a) What is the steadystate speed? (b) How long does it take the motor to reach 63% of its steadystate speed? (c) How long doe. and record the motor speed and current for 2 sec following the step input. Repeat Problem 67. Repeat Problems 6·1 through 63.1. Find the value of Kp that gives the same result as in Problem 67. Change Kp to 2.$ = 0. Change Kp to 1. 369 6·5. and record the motor speed and current for 0. Modify the model in Problem 66 by adding a proportional controller with a gain of Kp:::::: 0. Apply a 2 rad/sec step input. and assume the load torque TL = 0.0 in Problem 67. and record the motor speed and current for 2 sec following the step input. and assume the load torque TL = 0.0 in Problem 6~12. where K. and record the motor position for 1 sec.2 N·m (don't forget the minus sign) starting after 1 sec (requires change of the disturbance block parameters in SIMLab). apply a 1 rad step input.
and estimate what you consider to be the best value for the proportional gain. Select a step input with amplitude 0.tigate the closedloop position response using a PD controller. Plot the response. Chapter 6.c. measured in part (b)? . What happens to the amplitude of relative displacement at the damped natural frequency. 625. Repeat Problem 615 and assume a disturbance torque To = 0.5.0.5. investigate the closedloop motor speed response using a proportional controller. to 15 N/m. and record both the motor velocity and sine wave input signals. investigate the closedloop position response using a PD controller.0. 6. 623. (a) Use a PD controller in your investigation. 1. use proportional controller gains of 0. 620. 0. (e) Obtain the effect of washboard bumps with an amplitude of 0.02 m1s2 and the lower axes configured to display £(t). Use the SIMLab and parameter values of Problem 61 to design a PID controller that eliminates the effect of the disturbance torque. (e) Use a PID controller in your investigation. 616. Using the best value you obtained for Kpt try various values for KDt and record the step response in each case. Repeat this experiment for frequencies of 0. What is the effect of the gain on the steadystate velocity? 622.0.02 mls 2 on the response of the system. inve. use the CarSim tool to investigate the effects of controlling acceleration Xon relative motion (or bounce) Z and vice versa.2 Nm in addition to the step input of 1 rad (requires change of the disturbance block parameters in SIMLab). with a percent overshoot of 4.1. Using the SIMLab. Modify the controller used in Problem 615 by adding derivative action to the proportional controller. 0.2. Using the Virtual Lab Tool.). Record the closedloop response of the motor velocity to a step input of 2 rad/sec for proportional gains of 0. Use the same motor parameters as in Problem 61.0.2. Using the Quarter Car Modeling Tool controlling.5. investigate the closedloop position response using a proportional controller. 0.1 rad/s. UJd. and 50. and record the step response in each case.3.5 mls2inpllts~ Compare the results. Vary the frequency from lOradls to 0. 624. k. (b) Change the stiffness. For a positioncontrol case.1.4. Modify the controller used in Problem 614 by adding deri vative action to the proportional controller.. 619.0.1 Hz (don't forget: 1 Hz = 21T rad/sec) and amplitude of 1 V the amplifier input.8.370 ~.2. 0. gradually reducing the damping (variable c in the Model Parameters control window) to find the natural frequency of the system (w. Apply a sine wave with a frequency of 0. use proportional controller gains of 0. Using the Virtual Lab Tool.2 and 0. and 2.0. and estimate what you consider to be the best value for the proportional gain. 2. 0. In Design Project 2 in Section 67. 10. (b) Use a PI controller in your investigation. Using the best value you obtained for Kl" try various values for Kf). 1. 6~21. With a step input ()fO. 0. 617. Repeat this procedure for 0.1.0. 618.0 Hz (keeping the sine wave amplitude at 1 V).investigate the closedloop position response using a proportional controller.8. Using the Virtual Lab Tool.02 mls 2 and step time 0 seconds. record the step response for a 1 rad change at the output shaft.1 Nm in addition to the step input of 1 rad (requires change of the disturbance block parameters in SIMLab). Repeat Problem 615 and assume a disturbance torque Tn = 0. Use the SIMLab and parameter values of Problem 61 to design a PID controller that eliminates the effect of the disturbance torque. 1. with a percent overshoot of 2. For a positioncontrol case. what is the frequency of the oscillatory response? This is the damped frequency of the system using default parameters (COd)' How does the period of oscillation compare to the value that was observed in part (a)? Repeat the simulation several more times.. and 0. (a) Set the simulation mode to "Passive Suspension" and setup the top axes to display Y(t). 0. Using the SIMLab. The Control Lab 614.. record the step response for a 1 rad change at the output shaft. 5.0. Investigate the frequency response ofthe motor using the Virtual Lab Tool. and 2.2. 15.
Plot successive trials to an external figure for comparison. 1). and (e). roundedstep. (I) What value of K/ (PID [5 K10]) will yield a steadystate error less than 4 mm? What happens to the system overshoot and rise time at this gain? Does this match your calculations? = = (m) Test the controllers designed in the last few parts with sinusoidal. and 50 0 • Compare the optimized response with the PD response. Compare the results.e results using time domain analysis techniques. (g) What value of Kp (PID = [Kp 0]) will yield a steadystate error less than 4 mm? What happens to the system overshoot at this gain? Does this match your calculations? (h) Increase the derivative gain in steps from 0 to 50. Simulate the response.o. Explain what happens.2 m and duration 0. = (e) Repeat part (d) using the unidirectional rounded pulse (URP) input (amplitude 0. 25. keeping the proportional gain that was found in part (g).Problems .02 seconds).: 371 (d) Simulate driving the quartercar model over a curb by using the rounded step input with amplitude 0. PID = [1 0 0]. a (j) Test your controller's response to the inputs applied in parts (c). duration 0. validate the. 20. (k) Change the PID gains to PID [5 10] and click SISO Tool to apply the changes and launch the SISO Design Tool.. = 2. Again.25). . Find the steadystate error by simulation and by applying the finalvalue theorem. and observe the effect of adding derivative gain. of 15. (I) Add a step disturbance of 2 m/s2 and step time of 0 seconds. Calculate values of c such that the system is underdamped t?: = 0. and roundedpulse input.2 m.02 seconds. fPm. (i) Design four phaselead compensators with sufficient gain Kp to meet the steadystate error requirement as specified in part (g) and with phase margins. Validate the observed overshoot and rise time using the time domain analysis.5) and observe the response for each case. feedback z. critically damped (?. (d). and overdamped (?.
R. P(s ) = s" + all _ JS'I. Keep in mind that the transient properties of the system also depend on the zeros of the closedloop transfer function. we do not place any limitations on the relative magnitudes between II and m. K is a real constant that can vary from 00 to +00 . See Chapter 9 for examples. as well as how to interpret the data provided by the root loci for analysis and design purposes. s: F(s) = P(s) + KQ(s) = 0 (71) where pes) is an nthorder polynomial of s. determine the absolute and the relative stability of linear SISO systems. the MATLAB rootlocus tool in the Control Systems Toolbox component of ACSYS can be used. For plotting the root loci accurately. The rootlocus technique is not confined only to the study of control systems. details on the propelties and construction of the root loci are presented in Appendix E. nand m are positive integers. As a design engineer. (73) For the present. say. The basic properties and the systematic construction of the root loci are first due to W. The general rootlocus problem ean be formulated by referring to the following algebraic equation of the complex variable. The roots of the characteristic equation.. simply.CHAPTER 7 Root Locus Analysis 71 INTRODUCTION In the preceding chapters. root loci may be sketcbed by following some simple rules and properties. In general . In Chapter 5. However. which are the poles of the closedloop transfer function. several examples already illustrated the usefulness of tbe root loci of the characteristic equation in the study of linear control systems. In general.J + . the root lociwhen a certain system parameter varies. it is important to learn the basics of tbe root loci and their properties. The material in tbis text is prepared with these objectives in mind. An important study in linear control systems is the investigation of the trajectories of the roots of the characteristic equationor. . Evans [1. we have demonstrated the importance of the poles and zeros of the closedloop transfer function of a linear control system on the dynamic performance of the system. 372 . it may be sufficient for us to learn how to use these computer tools to generate the root loci for design purposes.3] . + (l j S + ao (72) and Q(s) is an mthorder polynomial of s. the metbod can be applied to study the behavior of roots of any algebraic equation with one or more variable paramctcrs.
Under unusual conditions. for most controlsystem applications~ the values of K are positive. When the variable parameter K does not appear as a multiplying factor of G(s)H(s). we can always condition the functions in the form of Eq. we define the following categories of root loci based on the values of K: 1. Eq. . Root contours (RC).. (7~7) is identical to Eq. Refers to the entire root loci for 00 < K < 00.112 + (3 + 2K)s + 5 = 0 (78) . we have identified the RL of a control system with the general rootlocus problem. hm are considered to be real and fixed. Thus..) (76) where pes) and Q(s) are polynomials as defined in Eq. hI. (71). let us consider the closedloop transfer function of a singleloop control system: yes) G(s) R(s) = 1 + G(s)H(s) (74) keeping in mind that the transfer function of multipleloop SISO systems can also be expressed in a similar fonn. The characteristic equation of the closedloop system is obtained by setting the denominator polynomial of Y(s)/R(s) to zero. the root loci of the characteristic equation of a linear discretedata system can be constructed in a similar fashion (Appendix E). consider that the characteristic equation of a control system is s(s + 1)(. (71) through (73). . Root loci (RL). (71). . Contour of roots when more than one parameter varies. 7·2 BASIC PROPERTIES OF THE ROOT LOCI (RL) Because our main interest is control systems. As an illustrative example. By replacing s with z in Eq.72 Basic Properties of the Root loci (Rl) . The resultant loci are called the root contours. Root loci of multiple variable parameters can be treated by varying one parameter at a time. then we have the situation that K is negative. we need to place the emphasis only on positive values of K in developing the rootlocus techniques. . In general. Thus. (75) is written (77) The numerator polynomial ofEq. (72) and (73). such that the rational function can be written as G(s)H(s) == Kit~. h2. when a system has positive feedback or the loop gain is negative. and the subject is treated in Section 75.•. 2. a2. (76). Although we should be aware of this possibility. an. by considering that the loop transfer function G(s)H(s) can be written in the form of Eq. 373 The coefficients aI.11 + 2) + . For the purpose of identification in this text. the roots of the characteristic equation must satisfy 1 + G(s)H(s) = 0 Suppose that G(s)H(s) contains a real variable parameter K as a mUltiplying factor. respectively.
(75) is written To satisfy Eq. We shall show that the RL of Eq. Eq. the rootlocus problem is another example in which the characteristics of the closedloop system. (75) can be constructed based on the properties of Q(s)/P(s). In practice. • Once the root loci are drawn. (714) or Eq. In the case where G(s)H(s) = KQ(s)/P(s). (75) or Eq. 0 0 = even multiples of 1t radians or 180 (715) where i = 0.374 . in this case represented by the roots of the characteristic equation. (77). Then. (715) are used to determine the trajectories of the root loci in the splane. are determined from the knowledge of the loop transfer function G(s)H(s). (77)... the values of K on the loci are determined by using the condition on magnitude in Eq. Chapter 7. (any integer). . we get Q(s) pes) = s3+4s2 +5s+5 (710) Now K is isolated as a multiplying factor to the function Q(s)/ pes). • The conditions on angles in Eq.. Now we are ready to investigate the conditions under which Eq. Root Locus Analysis ( To express the last equation in the form ofEq. (7~12). (77) is satisfied. (7~ 13). the conditions stated in Eq. ±2. ±1. we divide both sides of the equation by the terms that do not contain K. Let us express G(s)H(s) as G(s)H(s) = KGl (S)Hl (s) (711) where Gl (S)Hl (s) does not contain the variable parameter K. . and we get 1+ s(s + 1)(s + 2) 2Ks + 82 + 3s + 5 2s = 0 (79) Comparing the last equation with Eq. the following conditions must be satisfied simultaneously: Condition on magnitude lOt (S)Hl (s)1 = IKI Condition on angles 1 oo<K<oo (713) LGt(s)Hl(S) = (2i+ I)Jr K~O 0 = odd multiples of 1T radians or 180 LGI (s)Hl (s) = 2in (714) K::. (713) through (715) play different roles in the construction of the root loci.
although some of the properties are derived analytically.72 Basic Properties of the Root Loci (RL) ~ 375 The construction of the root loci is basically a graphical problem.EL(s+ Pj) j=l = (2i+ 1) x 180 0 (718) For 00 <K ~ 0: LGl (s)Ht (s) = EL(s+Zk) k=l m n EL(s+ pj) j=l = 2i X 180 0 (719) where i = 0. (718) is that any point SI on the RL that corresponds to a positive value of K must satisfy the following condition: The difference between the sums of the angles of the vectors drawn from the zeros and those from the poles of G(s)H(s) to S1 is an odd multiple of 180 degrees. The graphical construction of the RL is based on the knowledge of the poles and zeros of the function G(s)H(s). (s + Pn) + ZI)(S + Z2) .. (717) as IKI = _j:~l__ Ills + zd i::.• (s + Zm) (716) where the zeros and poles of O(s)H(s) are real or in complexconjugate pairs.. the numeratorofEq.. any point 81 on the RL must satisfy the following condition: The difference between the sums of the angles of the vectors drawn from the zeros and those from the poles of G(s)H(s) to S1 is an even multiple of 180 degrees. . Applying the conditions in Eqs. The graphical interpretation of Eq. 1 " Il~s+ Pi! (720) The value of K at any point 81 on the RL is obtained from Eq. = JKI IGl s Ills + Pkl k=l 00 < K < 00 (717) For 0 ~ K<oo: m k=l n LGI (S)Hl (s) == EL(s + Zk} .. ::1::1. In other words. G(s)H(s) must first be written as G(9)H(s) = KGI (s)HI (9) = K(s (s + Pl)(S + P2) . (713). (714). including zero degrees. (720) by substituting the value of 81 into the equation.. the values of K along the loci can be determined by writing Eq. For negative values of K. we have m Ills + ziI 1 ()HI (s)1 == i1 . . and the denominator represents the product of lengths of the vectors drawn from the zeros of G(s)H(s) to S1' . (720) represents the product of the lengths of the vectors drawn from the poles of G(s)H(s) to S1.. Once the root loci are constructed... and (715) to Eq. Graphically. (7~16). ±2.
the angles of the vectors are measured with the positive real axis as reference. If the RL for negative values of K are desired. Root Locus Analysis jro (f Figure 71 Polezero configuration of G(s)H(s) = K(s + ZI )/[s(s + P2) x (s + P3)]. it must satisfy Eq. . (718) for positive K. (722) (K ~ O)or Eq. (719) must be satisfied. (718). If Sl is indeed a point on the RL for positive K.Bp2 . (720) to find the magnitude of K on the RL. (718) to (720) for the construction of the root loci. Eq. 0).. 71. ±1. Chapter 7. If Sl is found to satisfy either Eq. (723)(K ::::.. C. as shown in Fig. if SI is a point on the RL for negative values of K. B.. 71.BpI . Use Eq.8 p l .L(SI + P2) 2i X L(SI 180 0 + P3) = (723) OZI . (719).L(si + P2)  L{SI + 1'3) 180 0 = 6:1 . As shown in Fig. 71.8 p2 . ±2 •. As shown in Fig. then Eq. Let us select an arbitrary trial point SI in the splane and draw vectors directing from the poles and zeros of G(s)H(s) to the point. (720) is used to find the magnitude of K at the point. the angles of the vectors shown in Fig.Bp3 = where i = 0. that is. that is y L(SI + Zl)  lsI .Ls] . Similarly. the lengths of the vectors are represented by A.. the construction of the RL of the zeros of 1 + G(s)H(s) involves the following two steps: 1. (722) or Eq.. To illustrate the use of Eqs. A search for all the SI points in the splane that satisfy Eq. given the function G(s)H(s) with K as a multiplying factor and the poles and zeros are known. 2. and D. (723).let us consider the function G(s)H(s) = K(s + Zl) s(s + P2)(S + P3) (721 ) The location of the poles and zero of G(s)H(s) are arbitrarily assigned... 71 must satisfy L(s{ + Zl) . ±1. it must satisfy Eq. The magnitude of K is IKI = IS 111 s 1 + pzllsl + P31 = BCD lSI + ztl A (724) The sign of K depends on whether Sf satisfies Eq. Thus.376 . ±2•.8 p3 = (2i + 1) x (722) where i = 0.
We shall limit the discussion only to the properties but leave the details of the proofs and the applications of the properties to the construction of the root loci in Appendix E. O. called the Spirule. Dividing both sides of Eq. the analyst should still have an understanding of the properties of the root loci to be able to manually sketch the root loci of simple and moderately complex systems. Years ago. if we were to use the trialanderror method just described. G 1(s)H. sos must approach the poles of G 1(S)Hl (s) or of G(s)H(s). the three roots of the equation are at s = 0. the search for all the rootlocus points in the splane that satisfy Eq. (719) and Eq. (719). However. Even with the Spirule. which can be used to assist in adding and subtracting angles of vectors quickly. for the device to be effective~ the user still has to first know the general proximity of the roots in the splane. which are the roots of the characteristic equation. the three roots of the equation are at s = 1. (720) would be a very tedious task. We can visualize that the finite s~plane is only a small portion of a sphere with an infinite radius.. and interpret the computer results correctly. and 00. The properties are developed based on the relation between the poles and zeros of G(s)H(s) and the zeros of 1 + G(s)H(s). Nevertheless. (718) or Eq. even with a highspeed computer and an effective rootlocus program. he had to devise a special tool. (718) or Eq. The reason for these properties are seen from the condition of the root loci given by Eq. (726) by the terms that do not contain K. The poles and zeros referred to here include those at infinity. we get I + G(s)H(s) = I + s(s+ 2) (s + 3) = . if any. K(s + 1) 0 (727) . When the magnitude of K is infinite. :.7~3 Properties of the Root Loci ~ 377 We have established the basic conditions on the construction of the rootlocus diagram. EXAMPLE 731 Consider the equation s(s + 2)(s + 3) + K(s + 1) = 0 (726) When K:::. if necessary.{s) approaches infinity.2] first invented the rootlocus technique. when Evans [1. Then. the Spirule and the trialanderror method have long become obsolete. and 3. infinity in the s~plane is a point on the opposite side of the sphere that we face. With the availability of digital computers and efficient rootfinding subroutines.2.. as the magnitude of K approaches infinity. when applying the root loci for analysis and design of control systems. (712). which is (725) As the magnitude of K approaches zero. 73 PROPERTIES OF THE ROOT lOCI The following properties of the root loci are useful for the purpose of constructing the root loci manually and for the understanding of the root loci. digital computer technology was still at its infancy. s must approach the zeros of G(s)H(s).00. The K= ± 00 points on the root loci are at the zeros of G(s )H(s). 7~31 K = 0 and K = ± 00 Points The K = 0 points on the root loci are at the polps of G(s)H{s). Similarly. according to Eq. It is useful to consider that infinity in the spIane is a point concept.
In general when n :F In. the order of P(s). the roots must be real or in complexconjugate pairs. The following property of the RL results. 732 Number of Branches on the Root Loci A branch of the RL is the locus of one root when K varies between 00 and 00. The three points on the root loci at which K = 0 and those at which K = ±oo are shown in Fig. in Eq. Keeping track of the individual branches and the total number of branches of the rootlocus diagram is important in making certain that the plot is done correctly. The reason behind this property is because for real coefficient. because unless each root locus branch is coded by a different color. (75) is equal to the order of the polynomial. K. there will be 21n . The angles of the asymptotes and their intersect with the real axis of the splane are described as follows. since the equation has three roots. since the number of branches of the RL must equal the number of roots of the equation. (726) when K ::= 0 are the same as the poles of the function G(s)H(s). . For example.3 Symmetry of the RL The RL are symmetrical with respect to the real axis of the splane. but there are two zeros at infinity. The three roots of Eq. it is up to the user to make the di s tinctions. one finite zero is at s = 1.. 7. the three roots of Eq. is not equal to m. 734 Angles of Asymptotes of the RL: Behavior of the RL at lsi =00 When n. (71) or Eq. The properties of the RL near infinity in the splane are described by the asymptotes of the loci when Isl+ 00. (726) when K = ±oo are at the three zeros of G(s)H(s). the RL are symmetrical with respect to the axes of symmetry of the pole~zero configuration of G(8)H(8).Chapter 7. The number of branches of the RL of Eq.378 '. which gives K(s+ 1) G(s)H(s) = s(s + 2)(s + 3) (728) Thus.3. some of the loci will approach infinity in the splane. the order of Q(s).. (71). Root Locus Analysis jfJJ splane K=O 3 K=O 2 K=±oo 1 K=O 0 Figure 72 Points at which K = 0 and K = ±oo on the RL of s(s + 2)(s + 3)+ K(s + 1) = O. (726) when Kvaries from 00 to 00 is three.ml asymptotes that describe the behavior of the RL at lsi = 00. In general. 72. In this case. the number of branches of the root loci ofEq. This is particularly true when the root~locus plot is done by a computer. including those at infinity.
73. In . respectively. The asymptotes of the root loci for K ~ 0 are simply the extensions of the asymptotes for K 2:: o.. K :5 00..J.• . 2. the RL for K given by ~ ~ 379 0 are asymptotic to asymptotes with angles . or 0'1= z::: real parts of poles ofG(s)H(s)  z:: rea] parts of zeros ofG(s)H(s) nm (731) The root loci and their asymptotes for Eq. (729) where i = 0. The intersect of the asymptotes (jI represents the center of gravity of the root loci and is always a real number.. nand m are the number of finite poles and zeros of G(s)H(s). More examples on root . ~ : I I jm s~plane I I l \ I ~ I Asymptote I I ~K=OK<O K=±O K>O K=O K<QO K~QO 0 Cf :2 I I I I I I I I I I 1 I I I I I I Figure 7~3 Root loci and asymptotes of s(s + 2)(s + 3) + K(s + I) = 0 fotx :::. •.73 Properties of the Root Loci For large values of s. (726) for 00 :5 K :5 00 are shown in Fig. 7~35 Intersect of the Asymptotes (Centroid) The intersect of the 21n ~ ml asymptotes of the RL lies on the real axis of the splane. at Ul =~~~~~~~~~~~ L finite poles of G(s)H(s) .eLI(2i + 1) 180 nm IX Q nrm .2: finite zeros of G(s)H(s) llm (730) where n is the number of finite poles and In is the number of finite zeros of G(s)H(s). . loci asymptotes and constructions are found in Appendix E. respectively. 1.m I .1.
all breakaway points on the root loci must satisfy Eq. For complex situations. can also be used for this purpose. associated with the frequency response. Chapter 7. den=conv(den. In other words. ti tIe ( Root loci for equation 7 . To be a breakaway point~ the solution ofEq. den=conv( [1 0] . I [k. RL for K ~ 0 are found in the section only if the total number of poles and zeros of G(s)H(s) to the right of the section is odd. that is. 73 num:::[11] .poles] = rlocfind(mysys) % rlocfind command in MATLAB can choose the desired poles on the locus axis([3 0 8 8]) . (732) is necessary but not sufficient. On a given section of the real axis. of G(s)H(s) denotes the angle of the tangent to the locus near the point. 739 Breakaway Points (Saddle Points) on the RL Breakaway points on the RL of an equation correspond to multipleorder roots of the equation. 737 Angles of Departure and Angles of Arrival of the RL The angle of departure or arrival of a root locus at a pole or zero. Complex poles and zeros of G(s)H(s) do not affect the type of RL found on the real axis. Note that the remaining sections of the real axis are occupied by the RL for K :5 O. rnysys=tf(nurn. 27' ) . [1 2]). The Bode diagram method in Chapters 2 and 8. when the RL have mUltiple numbers of intersections on the imaginary axis. (732) are breakaway points. but not all solutions of Eq. the intersects and the critical values of K can be determined with the help of the rootlocus computer program. must also be a point on the root loci for some real K.. [1 3] ) . (75). Root Locus Analysis 736 Root Loci on the Real Axis The entire real axis of the splane is occupied by the RL for all values K. 738 Intersection of the RL with the Imaginary Axis The points where the RL intersect the imaginary axis of the splane and the corresponding values of K may be detennined by means of the RouthHurwitz criterion. Toolbox 731 MATLAB statements for Fig.den)j rlocus(mysys). The breakaway points on the RL of 1 + KG} (s)H} (s) = 0 must satisfy dGl (S)HI (s) = 0 ds (732) It is important to point out that the condition for the breakaway point given in Eq. (732) must also satisfy Eq. (732).380 . respectively.
and m = number of finite zeros of G(s)H(s). For quick reference. except for extremely complex cases. including the plotting of the final loci. The total number of root loci is equal to the order of the equation 1 + KGl (S)Hl (s) O. (712) with respect to s. the important properties described are summarized in Table 71.1~ g. = . K Properties of the Root Loci of 1 + KG. Number of separate root loci 4. I 0 6. For large values of s. The K = 00 points are at the zeros of G(s)H(s). However. Intersection of the asymptotes of finite poles of G(s)H(s). including those at s = 00. the properties on the root loci just presented should be adequate for making a reasonably accurate sketch of the rootlocus diagram short of plotting it point by point. (a) The intersection of the asymptotes lies only on the real axis in the splane.LrealpartsofzerosofG{s)H(s) ~m 7. the condition in Eq. TABLE 71 1. Root loci on the real axis. 00. and the details are given in Appendix E. RL for K ~ 0 are found in a section of the real axis only if the total number of real poles and zeros of G(s)H(s) to the right of the section is odd. the RL are asymptotic to where i = 0. Symmetry of root loci The K = 0 points are at the poles of G(s)H(s)~ including those ats =::.73 Properties of the Root Loci ~ 381 If we take the derivatives on both sides of Eq. one cannot rely on the computer solution completely. n = number = jllml x 180 2i . RL for K :5 0 are found. If the total number of real poles and zeros to the right of a given section is even. Inmj For K<0.. the RL (K > 0) are asymptotic to asymptotes with angles given by g.• In . 1. = 5. since the user still has to decide on the range and resolution of K so that the rootlocus plot has a reasonable appearance. X 1800 . (732) is equivalent to dK=O ds (734) In summary. (Continued) .ml . = 0 = 0 points ±oo points 2. K = 3.. the breakaway points. (b) The point of intersection of the asymptotes is given by UI _ L realpartsofpolesofG(s)H(s). Asymptotes of root loci as s~oo The root loci are symmetrical about the axes of symmetry the of polezero configuration of G(s)H(s). (s) H. and some of the other specific details of the root loci.2i±l. we get dK dOl (S)Hl (s)/ds ds = [Gl (S)Hl (s)f (733) Thus.2. The computer program can be used to solve for the exact root locations.
Breakaway points 11. For instance. or zero. not only it is important to arrive at a system that has the desired characteristics.Y) can be detennined by assuming a point SI that is very close to the pole. From the rootsensitivity standpoint.... Root Locus Analysis TABLE 71 (Continued) The angle of depru1ure or arrival of the RL from a pole or a zero of G(s)H(.. In the design of control systems. and applying the equation III 111 8. but if it is very sensitive to the variation of K.LL(sl j=1 Pj) =2{i+l)180° K. but. Angles of departure /G(s))H(s)) = LL(SI k=1 zd . ~ EXAMPLE 7~3~2 Fig. (734) leads to the root sensitivity [17.itivity is infinite. The RL are computed and plotted digitally. At the breakaway point. Thus. The sensitivity of the foots of the characteristic equation when K varies is defined as the root sensitivity and is given by SK = dK / K ds/ s . 19] of the characteristic equation. 7~4 shows the root locus diagram of s(s + 1) +K = 0 (736) with K incremented uniformly over 100 values from 20 to 20. ± 1. Eq. 18. The breakaway points on the root loci are determined by finding the roots of dK/ds = D. a system that is insensitive to parameter variations is calIed a robust system. it may get into the undesirable perfonnance region or become unstable if K varies by only a small amount. Each dot on the root~locus plot represents one root for a distinct value of K. we should avoid selecting the value of K to operate at the breakaway points. These are necessary conditions only. Intersection of the root loci with the imaginary axis 0 K S0 iO. which correspond to multiple~order roots of the characteristic equation. In fonnal controlsystem terminology... on the root loci is on the 1'00t loci determined from the equation IKI = IGI (Sl )HJ (sl)1 1 7310 The Root Sensitivity The condition on the breakaway points on the RL in Eq. the rootlocus study of control systems must involve not only the shape of the root loci with respect to the variable parameter K but also how the roots along the loci vary with the variation of K. Thus. the movements of the roots become larger for the same incremental change in K. Fig. or dG(s)H(s)/{lv = O. 9. As the magnitude of K decreases.382 ~ Chapter 7. s = 0. we see that the root sensitivity is low when the magnitude of K is large. The absolute value of K at any point 8. the root senr..:::O = 2i x 180 where i = O.. the system should be insensitive to parameter variations. a system may perform satisfactorily at a certain K. (734) shows that the root sensitivity at the breakaway points is infinite.5.: K ds s dK (735) Thus. 75 shows the RL of . Calculation of the values of K The crossing points of the root loci on the imaginary axis and the corresponding values of K may be found by use of the RouthHurwitz cri terion. ±2 . just as important.
0 '1.0 K=O 4.0 0 ::. ....·O·..0 1..0 K=O K =0 1... ...... . K>(} 3.0 u ···........0 .0 1... ···• '...0 .." '\:~ ..:: 4...0 \.0 . ~. ..0 2..: i ~ 14·0 ...... K<O ...... K<O i 3 .. " .~~ K::::O 2..0 Figure 74 RL of .0 j(j) ~ splane 3.0 o :2..0 '..O 4..... <f \ K>O oo"K ..0 K .0 'to: .." I il ..0 2.. K>O . .0 3.····Kh..0 K:= 0 0.... .0 '.' 1....0 1.K 5.0 K:=O K<O Koo 3..73 Properties of the Root Loci 4 383 j{J) 8 t :. : .0 3... • 0.~(s + 1) + K = 0 showing the root sensitivity with respect to K..0 splane 3.___ 3.... 2....~ 2. ....0 Z.... Figure 75 RL of s2(s+ 1)2 + K (s + 2) = O.:: A 2. showing the root sensitivity with respect to K.....0 .0 1..0 1..
where ISKI\ denotes the root sensitivity of the first root. (740) indicates that the sensitivities of the two real roots are different for a given value of K. dK ::::: 2s1 (738) ds Toolbox 732 MATLAB Slalenlenls for Eqs. poles] = rlocfind(mysys2) From Eq.den2).poles] = rlocfind(mysysl) %rlocfind command inMATLAB can choose the desired poles on the locus. (736). [1 1]).25 ( + la+ll 20" 1 (740) \SKla=o. (739).543. den2:. [k. (736). Table 72 gives the magnitudes of the sensitivities of the two roots ofEq. title( 'Root loci for equation 737'). We can investigate the root sensitivity further by using the expression in Eq. (739) leads to ISKlw=o When the two roots are complex. mysysl=tf(num1. Root Locus Analysis with K incremented unifonnly over 200 values from 40 to 50. and ISK21 denotes that of the second root. For the secondorder equation in Eq. although the two real roots reach a = 0. K = s(s + 1).conv( [1 0 0]. and 2.5 for the same value of . [1 1]).denl). (739) gives 0.2) mysys2=tf(num2. since w appears only as w2 in the equation. num2=[1 2]. These values indicate that. it is apparent that the sensitivities of the pair of complexconjugate roots are the same. the root sensitivity becomes (739) where S = (J + j{J). rlocus(mysys2). Eq. (734). Eq.0. 0. the loci show that the root sensitivity increases as the roots approach the breakaway points ats = D. For the roots on the real axis.1. Thus. 736 and 737 numl=[l]. axis([3 0 8 8]) [k.s= 4m2 + (J)2) 1/2 (741) From Eq. Again.36' ) . (J) = O.384 ~ Chapter 7. den2=conv(den2. and s must take on the values of the roots ofEg.5 for all values of w. subplot(2. (736) for several values of K. [1 1]). Eq. (741).1)j rlocus(rnysys1)j title ( 'Root loci for equation 7. subplot(2. 1.1. denl=conv( [1 0].457. s = = 0.
817 0.5 . it is useful to learn the effects on the RL when poles and zeros of G(s)H(s) are added or moved around in the s~plane.527 0.454 0.937 0.5 + }1. phaselead.16 0.j1.00 00 1.5 + jO.000 00 ROOT 2 1. The design of the PI.333 2.937 0.25 0. the analyst or designer can obtain vital information on the perfonnance of the system by making a quick sketch of the RL using some or all of the properties of the root loci.333 3.jO.045 1.28 0.500 K = 0.500 0. and each root travels the same distance from u sensitivities of the two real roots are not the same.500 0. the ~.800 0.516 0. 74 DESIGN ASPECTS OF THE ROOT LOCI One of the important aspects of the rootlocus technique is that~ for most control systems with moderate complexity.042 0.jO.5 . Addition of Poles to G (8) H (s) Adding a pole to G(s)H(s) has the effect of pushing the root loci toward the righthalf splane.516 0.500 0.387 0.5 + jO.173 0.24 0.817 0.958 0. From the design standpoint.5 .5 .25.562 0.000 ISK21 0 0. 741 Effects of Adding Poles and Zeros to G(s) H(s) The general problem of controller design in control systems may be treated as an investigation of the effects to the root loci when poles and zeros are added to the loop transfer function G(s)H(s).7~4 Design Aspects of the Root loci '.jO. ' EXAMPLE 741 Consider the function G(s)H(s) = () s s+a K a> 0 (742) .20 4.600 0.527 0.400 0. .387 0.5 +}0. = 0 and s = 1.5+ joo jSKll 1.000 1. The effect of adding a zero to G(s)H(s) can be illustrated with several examples .562 0.joo 1.975 0. PID.40 1. phaselag~ and the leadlag controllers discussed in Chapter 9 all have implications of adding poles and zeros to the loop transfer function in the splane. respectively. 385 TABLE 72 K Root Sensitivity ROOT 1 0 0.200 0.173 0. Some of these properties are helpful in the construction of the rootlocus diagram.5 .975 0.000 00 0 0.04 0. It is of importance to understand all the properties of the RL even when the diagram is to be plotted with the help of a digital computer program.
386
Chapter 7. Root Locus Analysis
8 jOJ t splane
~
K=O
a a
K=O
0
co
K
2
cr
a
(a)
(b)
8 'Cv
sp1ane
b  c=
00
I J
~
+
K=O K=O
b
(l
(e)
(d)
Figure 76 Rootlocus diagrams that show the effects of adding poles to G(s)H(s).
The RLof 1 + G(s)H(s) :::: 0 are shown in Fig. 76(a). These RL are constructed based on the poles of G(s)H(s). which are at s = 0 anda. Now let us introduce a pole at s :::: b. with b > a. The function G(s)H(s) now becomes
G(s)H(s)
=
_,K _ _ s(s +a)(s + b)
(143)
Fig. 76(b) shows that the pole at s = b causes the cOIllpl~x purl uf the root loci to bend toward the righthalf splane. The angles of the asymptotes for the complex roots are changed from ±90° to ±60°. The intersect of the asymptotes is also moved from a/2 to (a + b)/2 on the real axis.
14 Design Aspects of the Root loci
: 387
Toolbox 741
klATLAB statements for Fig. 73
The results for Fig. 76 can be obtained by the following Matlab statements:
a=2;
b=3; c=5;
nurn4=(1] ;
den4=conv( [1 0] , [1 a]);
subplot (2 ,2,1.) rnysys4=tf(num4,den4);
rlocus(mysys4);
axis ( [  3 0  8 8] )
num3=(~];
den3=conv( [1 0] , convC [1 a] , [1 a/2] ) ) ;
subplot(2,2,2)
mysys3=tf(num3,den3); rlocus(mysys3); axis ( [  3 0  8 8]) num2=(1];
den2=conv( [1 OJ , conv( [1 a] , [1 b]));
subplotC2,2,3) mysys2=tf(num2,den2): rlocus(mysys2); axi s ( [  3 0  8 8])
numl=[l] ; denl=conv([1 0] ,conv([l a], [1 b])); den1=convCden1, [1 c]) ;
mysysl=tf(num1,denl); subplotC2,2,4)i rlocus Cmysysl) ;
If G(s)H(s) represents the loop transfer function of a control system, the system with the root loci in Fig. 7 6(b) may become unstable if the value of K exceeds the critical value for stability, whereas the system represented by the root loci in Fig. 76(a) is always stable for K > O. Fig. 7~6(c) shows the root loci when another pole is added to G{s}H(s} at s = c, c > b. The system is now of the fourth order, and the two complex root loci are bent farther to the right. The angles of asymptotes of these two complex loci are now ±45 D • The stability condition of the fourthorder system is even more acute than that of the thirdorder system. Fig. 76(d) illustrates that the addition of a pair of complexconjugate poles to the transfer function ofEq. (742) will result in a similar effect. Therefore. we may draw a general conclusion that the addition of poles to G(s)H(s) has the effect of moving the dominant portion of the root loci toward the righthalf splane.
Addition of Zeros to G(s)H(s) Adding lefthalf plane zeros to the function G(s)H(s) generally has the effect of moving and bending the root loci toward the lefthalf splane. The following example illustrates the effect of adding a zero and zeros to G(s)H(s) on the RL.
388 to Chapter 7. Root locus Analysis
;.;.. EXAMPLE 742 Fig. 77(a) shows the RL of the G(s)H(s) in Eg.
complex~conjugate
(7~42) with a zero added at s = b(b > a). The part of the RL of the original system is bent toward the left and forms a circle.
Thus, if G(s}H(s) is the loop transfer function of a control system, the relative stability of the system is improved by the addition of the zero. Fig. 77(b) shows that a similar effect will result if a pair of
complex~col1jugate zeros is added to the function ofEq. (742). Fig. 77(c) shows the RL when a zero at s = c is added to the transfer function of Eq. (743).
8 j{J)
splane
8
splane
~
j{J)
b=oo+ ~
t
t
K=oo
ooK
a/2
K::::O
K=O
a12
0
0
(1
(a)
(b)
splane
K=oo
c
K=O K=O
b
a
~
J 8
(c)
Figure 77 Rootlocus diagrams that show the effects of adding zeros to G(s)H(s).
74 Design Aspects of the Root Loci
.~
389
Toolbox 7..4~2
MATLAB statements for Fig. 77
a=2;
b=3j
d=6; c=20; num4=[1 d] i den4=conv( [10] , [1 a]) i subplot(2,2,1)
rnysys4=tf(num4,den4)j
rlocus(rnysys4)j num3=[1 c] ; den3=conv( [1 0] , [1 a]) ; subplotC2,2,2)
rnysys3=tf(num3,den3);
rlocus(mysys3); axi s ( [  6 0  8 8])
num2=[1 d]
i
den2=conv([1 0] ,conv([l a], [1 b])); subplot(2,2,3) mysys2=tfCnum2,den2)j
rlocus(mysys2)i
axi s ( [  6 0  8 8])
~
EXAMPLE
743 Consider the equation
S2(.t
+ a) + K(g + h) =
s(s + a)
0
(744)
Dividing both sides ofEq. (744) by the terms that do not contain K. we have the loop transfer function
G(s)H(s) =
~(s + b)
(745)
It can be shown that the nonzero breakaway points depend on the value of a and are
s = ±va2 lOa+9 4 4
a+3
I_I
(746)
Fig. 78 shows the RL ofEq. (744) with b = 1 and several values of a. The results are summarized as follows: Fig. 78(a): a = 10. Breakaway points: s
= 2.5 and
 4.0.
Fig. 78(b): a = 9. The two breakaway points given by Eq. (746) converge to one point at s = 3. Note the change in the RL when the pole at 0 is moved from  10 to ~ 9.
For values of a less than 9, the values of s as given by Eq. (746) no longer satisfy Eq. (744), which means that there are no finite, nonzero, breakaway points. Fig.
7~8(c):
a = 8. No breakaway point on RL.
As the pole at s = a is moved farther to the right along the real axis. the complex p0l1ion of the RL is pushed farther toward the righthalf plane. Fig. 78(d): a = 3.
Fig. 78(e): a = b = 1. The pole at s =  { l and the zero at b cancel each other out. and the RL degenerate into a secondorder case and lie entirely on the jwaxis.
390
~
Chapter 7. Root locus Analysis
8 t
jru
~
splane
K=O
10
4.5
K=O K=O
a
:t<
l 8
(a) a
=IO
s~plane
~
t
8
jm
splane
j{))
K=O
K=O
8
K=O
9
3.51
K=O a
~
!
8
(b)
a=9
(c) a=
8
Figure 78 Rootlocus diagrams that show the effects of moving a pole ofG(s)H{s). G(s)H(s) = K(s + 1)/ [82(s + a)}
(Continued ).
7~4
Design Aspects of the Root Loci <l 391
8 t
~
~
t
8
jOJ
jOJ
spIane
splane
K=O K=O
K=O
o
K=O
a
(d) a
=3
(e) a::: I
Figure 78 (Continued)
Toolbox 743
MATLAB statements for Fig.
num1=[1 b];
7~8
al=lO;a2=9;a3=8;a4=3;b=1; den1=conv( [1 0 0] , [1 al]) ; subplot(2,2,1) mysysl=tf(numl,den1); rlocus(mysysl);
num2=[1 b]; den2=conv( [1 0 0], [1 a2]); subplot(2,2,2) mysys2=tf(num2,den2); rlocus(mysys2) j
nwn3=[1 b]; den3=c onv ( [1 0 0] , [1 a3] ) ; subplot(2,2,3) rnysys3=tf(num3,den3); rlocus(mysys3); num4=[1 b] ;
den4=conv( [1 0 0] , [1 a4]) ;
5ubplot(2,2,4) rnysys4=tf(num4,den4); rlocus(mysys4);
392
Chapter 7. Root Locus Analysis
:., EXAMPLE 744 Consider the equation
oS"
(s2
+ 2s + a) + K (s + 2) =
K(s + 2) = s (') + 2s+a } s~
0
(747)
which leads to the equivalent G(s)H(s) as
G(s)H(s)
(748)
The objective is to study the RL for various values of a( > 0). The breakaway point equation of the RL is determined as
/' +4i +4s+a = 0
Fig. 79 shows the RL of Eq. (747) under the following conditions.
(749)
Fig. 79(a}: a ;:;: :; 1. Breakaway points: s = 0.38, 1.0, and  2.618. with the last point being on the RL for K ;;::: O. As the value of tl is increased from unity. the two double poles of G(s)H(s) at s = 1 will move vertically up and down with the real parts equal toI. The breakaway points at s = O.38ands = 2.618 will move to the left, whereas the breakaway point at s 1 will move to the right. Fig. 79(b): a = 1.12. Breakaway points: s = 0.493, 0.857, and2.65. Because the real parts of the poles and zeros of G(s)H(s) are not affected by the value of a, the intersect of the asymptotes is always at s = O. Fig. 79(c): a = 1.185. Breakaway points: s = 0.667, 0.667, and  2.667. The two breakaway points of the RL that lie between s = 0 and  1 converge to a point. Fig. 7(;J(d): a = 3. Breakaway point: s = 3. When a is greater than 1.185. Eq. (749) yields
=
only one solution for the breakaway point. The reader may investigate the difference between the RL in Figs. 79(c) and 79(d) and fill in the evolution of the loci when the value of a is gradually changed from 1.185 to 3 and beyond.
splane
~
t
8
j(tJ
Jplane
8
~
j{J)
t
K
K
K
2.65
2.618
(j
(a)
a= 1
(b) a = 1.12
Figure 19 Rootlocus diagrams that show the effects of moving a pole of G(s)H(s) = K(s + 2)/[s(s2 + 2s + a)] (Continued).
7~5
Root Contours (RC): Multiple~Parameter Variation .... 393
jCO splane
jco
8
s~plane
:.c:
t
jl.414
.(IQ
K
K=O Keo
2.667
3
o
j1.414
CT
K=O 1 j0.43
I  j1.414
(c) (l
= 1.185
K>O
K<O
(d) a
=3
Figure 7~9 (Continued)
75 ROOT CONTOURS (RC): MULTIPLEPARAMETER VARIATION
The rootlocus technique discussed thus far is limited to only one variable parameter in K. In many controlsystems problems, the effects of varying several parameters should be investigated. For example, when designing a controller that is represented by a transfer function with poles and zeros, it would be useful to investigate the effects on the characteristic equation roots when these poles and zeros take on various values. In Section 74, the root loci of equations with two variable parameters are studied by fixing one parameter and assigning different values to the other. In this section, the multiparameter problem is investigated through a more systematic method of embedding. When more than one parameter varies continuously from 00 to 00, the root loci are referred to as the root contours (RC). It will be shown that the root contours still possess the same properties as the singleparameter root loci. so that the methods of construction discussed thus far are all applicable. The principle of root contour can be described by considering the equation
(750)
where Kl and K2 are the variable parameters and pes), Ql(S), and Q2(S) are polynomials of s. The first step inVOlves setting the value of one of the parameters to zero. Let us set K2 to zero. Then, Eq. (750) becomes
(751)
394
~
Chapter 7. Root Locus Analysis
which now has only one variable parameter in Kl. The root loci of Eq. (751) may be detennined by dividing both sides of the equation by P(s). Thus,
I
+
K\QI(S) = 0 pes)
(752)
Eq. (752) is of the form of I + KI Gl (S)HI (s) = 0, so we can construct the RL of the equation based on the polezero configuration OfGI (8)Hl (s). Next, we restore the value of K2; while considering the value of K} fixed, and divide both sides of Eq. (750) by the terms that do not contain K2. We have
1+
pes)
KzQz(s) =0 + KI Ql (s)
(753)
which is of the form of 1 + K2G2(S)H2(S) = O. The root contours ofEq. (750) whenK2 varies (while KI is fixed) are constructed based on the polezero configuration of
Q2(S) G2{s)H2(S) = pes) + KIQI (s)
(754)
It is important to note that the poles of G2{s)H2(S) are identical to the roots of Eq. (751). Thus. the root contours of Eq. (750) when K2 varies must all start (K2 = 0) at the points that lie on the root loci of Eq.(751). This is the reason why one rootcontour problem is considered to be embedded in another. The same procedure may be extended to more than two variable parameters. The following examples illustrate the construction of Res when multiparW:Ui;!lervariation situations exist.
;,~
EXAMPLE 751 Consider the equation
S3 +K2S2 +KIs+ KI = 0
where KI and Kz are the variable parameters, which vary from 0 to As the first step. we let K2 = O. and Eq. (755) becomes
8 3 00.
(755)
+K}s+KI
=0
(756)
Dividing both sides of the last equation by s3, which is the term that does not contain KI, we have
(757)
The root contours of Eq. (756) are drawn based on the polezero configuration of
s+l Gl (s)HI (s) ==  3 
s·
(758)
as shown in Fig. 7IO(a). Next, we let K2 vary belween 0 and 00 while holding KI at a constant nonzero value. Dividing both sides of Eq. (755) by the tenns that do not contain K2, we have
1+
83
+K)s+KI
K2i
= 0
(7 59)

Thus, the root contours ofEq. (755) when K2 varies may be drawn from the polezero configuration of (760) The zeros of G2 (S)H2(S) are at s = 0, O. but the poles are at the zeros of I + KI GI (S)Hl (s), which are found on the RL of Fig. 71 O(a). Thus. for fixed KI. the RC when Kz varies must all emanate from the root contours of Eq. 710(a). Figure 7lO(b) shows the root contours of Eq. (755) when K2 varies from 0 to 00, for KJ = 0.0184, 0.25. and 2.56.
75 Root Contours (RC): MultipleParameter Variation
8
.~
395
t:< ~ K2 =O
K2 0(004 +j1.74) K] =2.56
t
=
splane
splane
K 2 = 0(0.25
+jO.66)
K)=O.25 K'2 =0(0.1665 + jO.3IS)
K] =0.0184
Ki=oo
I
(j
(a)
(b)
Figure 710 Root contours of? K 1 is a constant.
~
+ Kzs2 + KIS + KI
;:::: O. (a) K2 = O. (b) K2 varies and
EXAMPLE 752 Consider the loop transfer function
G(s)H(s)
= s(1 + T.r)(s2 + 2s + 2)
K
(761)
of a closedloop control system. It is desired to construct the root contours of the characteristic equation with K and T as variable parameters. The characteristic equation of the system is 3(1
+ Ts)(s2 + 2s + 2) + K = 0
s(s2 +2s+2) +K =
0
(162)
First, we set the value of T to zero. The characteristic equation becomes (763)
Toolbox 7..51
M..4.TIAB statements for Fig. 710
forkl=[O.0184 0.25 num=[l 0 0]; den=[l 0 kl kl]; 2.56];
mysys=tf(num,den); rlocus(mysys);
hold on;
end;
196 : Chapter 7. Root Locus Analysis
.vplane
.~·plane
K=O
I+j
K=O
T~ot)
ooK
(J
K= 10 T=O
(Y
lj
K=O
K=O T=oo
(a)
(b)
Figure 711 (a) RL for s(s2
G2(s)H2{S)
= Ts2(s2 + 28 + 2)/[s(s2 + 2s + 2) + KJ.
+ 28 + 2) + K = O.
(b) Polezero configuration of
The root contours of this equation when K varies are drawn based on the polezero configuration of 1 Gr (S)H1 (s) = (2 2 2) (764)
ss
+ s+
as shown in Fig. 711(a). Next. we let K be fixed and consider that T is the variable parameter. Dividing both sides of Eq. (762) by the terms that do not contain T. we get
1 + TG2(S)Hz(s)
= 1 + ss + 2s+ ?) + K = 0 (2 ...
Ts2(s2
+ 2s + 2)
(765)
The root contours when T varies are constructed based on the polezero configuration of G2(s)H2(S). When T 0, the points on the root contours are at the poles of G2 (s)H2 (s), which are on the root contours of Eq. (763). When T = 00. the roots of Eq. (762) are at the zeros of G2(s)H2(S), which are at s = 0, 0, 1 +j. and 1 j. Figure 711(b) shows the polezero configuration of G2(s)H2(S) for K = 10. Notice that G2(s}H2(S) has three finite poles and four finite zeros. The root contours for Eq. (762) when Tvaries are shown in Figs. 712, 713. and 714 for three different values of K.
=
OT
(J
Figure 712 Root contours for s(1
+ Ts)(s2 + 2s + 2) + K = 0. K > 4.
75 Root Contours (RC): MultipleParameter Variation
~ 397
OT
(J
$(1
Figure 713 Root contours for + Ts)(s2 + 2~ + 2) + K = O.
K=O.5 .
.rplane
O+T
s(1
Figure 7..14 Root contours for + TS)(S2 + 2$ + 2) + K = O. K < 0.5.
The root contours in Fig. 713 show that when K Eq. (762) has a quadruple foot at s = 1.
= 0.5 and T = 0.5, the characteristic equation in
Toolbox 752
M4TLAB statel1lelltsjor EXlll1"lple 752
%T= 0
nurn=[l] ; den=conv( [1 0], conv( [0 1], [1 2 2])); mysys=tf(num,den); subplot(2,2,1)irlocus(mysys);
%k>4 for k=4:10; num=conv( [1 0 0], [1 2 den=den+k; mysys=tf(num,den)j
end;
2]); den=conv( [1 0], [1 2 2]);
subplot(2,2,2)jrlocus(rnysys)j
k=O.5;
num=conv( [1 0 0], [1 2 den=den+kj
2J); den = conv( [1 0], [1 2 2]);
mysys=tf(num,den);
J98
~
Chapter 7. Root Locus Analysis
subplot(2,2,3) rlocus(mysys); %k<O.S for k=100:0. 5; num=conv( [1 0 0] I [1 2 2]); den=conv( [1 0], [1 2
2]);
den=den+k;
mysys~tf(num,den);
subplot(2,2,4);rlocus(mysys); end;
:. EXAMPLE 75..3 As an example to illustrate the effect of the variation of a zero of G(s)H(s). consider the function
K(l + Ts) G(s)H(s) = s(s + l)(s + 2)
The characteristic equation is
(766)
s(s + l)(s + 2)
+ K(l + Ts)
= 0
(767)
(7~68)
Let us first set T to zero and consider the effect of varying K. Eq. (767) becomes
s(s + l)(s + 2) + K == 0
This leads to
Gl (S)Hl (8)
= S (,t;+ 1)(s+ 2)
1
(769)
The root contours ofEq. (7~68) are drawn based on the po]e....zero configuration ofEq. (769). and are shown in Fig. 715. When the K is fixed and nonzero. we divide both sides of Eq. (767) by the tenns that do not contain T. and we get 1 + TG2(S)H2{S)
= 1 + SS+ 1)(s+ 2) + K = (
TKs
0
(770)
The points that correspond to T = 0 on the root contours are at the poles of 02 (s )H2(s) or the zeros of s(s + l)(s + 2) + K, whose root contours are sketched as shown in Fig. 715 when K varies. Ifwe choose K = 20 just as an illustratioIl, the polezero configuration of G2(s)H2(S) is shown in Fig. 716. The root contours of Eq. (767) for 0 S T < 00 are shown in Fig. 7 ~ 17 for three different values of K.
s~plane
ooK
K=O X
2
Figure 115 Root loci for
s(s + l)(s + 2) + K:=: O.
75 Root Contours (RC): MultipleParameter Variation
~
399
Figure 716 Polezero configuration
of G2(s)H2{S) ~ Ks/[s(s + I) (s + 2)+K]. K = 20.
splane
Asymptote of _ _+1 root contours
K=20 T=20
4 3.85
2
Root contour on real axis
K=6+H
Figure 1~17
Root contours of s(s + l)(s
+ 2) + K + KTs == O.
400
Chapter 7. Root locus Analysis
Toolbox 75..3
MATUB statements for
Fig~
717
Same results as Fig. 717 can be obtained by using the following MATLAB statements:
for k= [3 6 20]; 0]; den=conv( [1 0] r conv( [1 1], [1 2])); den=den+k; mysys=tf(num,den); rlocus(mysys);
num~[k
axis([4 4 10 10]);
hold on
end;
Because G2(s)H2(S) has three poles and one zero, the angles of the asymptotes of the root contours when T varies are at 90° and 900 • We can show that the intersection of the asymptotes is always at s = 1.5. This is because the sum of the poles of G2 (.Y)H2 (s), which is given by the negative of the coefficient of the $2 tenn in the denominator polynomial of Eq. (7~ 70), is 3; the sum of the zeros of G2(s)H2(S) is 0; and nm in Eq. (730) is 2. The root contours in Fig. 717 show that adding a zero to the loop transfer function generally improves the relative stability of the closedloop system by moving the characteristic equation roots toward the left in the splane. As shown in Fig. 7 17 ~ for K = 20, the system is stabilized for an values of T greater than 0.2333. However, the largest relative damping ratio that the system can have by increasing T is only approximately 30 percent.
76 MATLAB TOOLS AND CASE STUDIES
Apart from the MATLAB toolboxes appearing in this chapter, this chapter does not contain any software because of its focus on theoretical development. In Chapter 9, when we address more complex control system modeling and analysis, we will introduce the Automatic Control Systems MATLAB tools. The Automatic Control Systems software (ACSYS) consists of a number of mfiles and OUIs (graphical user interface) for the analysis of simple control engineering transfer functions. It can be invoked from the MATLAB command line by simply typing Acsys and then by clicking on the appropriate pushbutton. A specific MATLAB tool has been developed for most chapters of this textbook. Throughout this chapter, we have identified subjects that may be solved using ACSYS. with a box in the left margin of the text titled "MATLAB TOOL."
77 SUMMARY
In this chaptet"s. we iULfuuuced the rootlocus technique for linear continuous data control systems. The technique represents a graphical method of investigating the roots of the characteristic equation of a linear timeinvariant system when one or more parameters vary. In Chapter 9 the rootlocus method will be used heavily for the design of control systems. However. keep in mind that, although the characteristic equation roots give exact indications on the absolute stability of linear SISO systems. they give only qualitative information on the relative stability, since the zeros
Review Questions
401
of the closedloop transfer function, if any, play an important role in the dynamic performance of the system. The rootlocus technique can also be applied to discretedata systems with the characteristic equation expressed in the ztransform. As will be shown in Appendix H. the properties and construction of the root loci in the zplane are essentially the same as those of the continuousdata systems in the splane, except that the interpretation of the root location to system perfonnance must be made with respect to the unit circle lzl = 1 and the significance of the regions in the zplane. The majority of the material in this chapter is designed to provide the basics of constructing the root loci. Computer programs, such as the MATLAB Toolboxes used throughout this chapter, can be used to plot the root loci and provide details of the plot. The final section of Chapter 9 deals with the rootlocus tools of MATLAB. However. the authors believe that a computer program can be used only as a tool, and the intelligent investigator should have a thorough understanding of the fundamentals of the subject. The rootlocus technique can also be applied to linear systems with pure time delay in the system loop. The subject is not treated here~ since systems with pure time delays are more easily treated with the frequency~domain methods discussed in Chapter 8.
REVIEW QUESTIONS
The following questions and trueandfalse problems all refer to the equation P(s) where P(s) and Q(s) are polynomials of s with constant coefficients.
+ KQ(s) == O.
1.
Give the condition from which the root loci are constructed.
2. Detennine the points on the complete root loci at which K :::: 0, with reference to the poles and zeros of Q(s)/ P(s).
3. Determine the points on the root loci at which K = ±oo, with reference to the poles and zeros of Q(s)/P(s).
4. Give the significance of the breakaway points with respect to the roots of P(s)
5. Give the equation of intersect of the asymptotes.
+ KQ(s} == O.
6. The asymptotes of the root loci refer to the angles of the root loci
w~nK=±oo.
cr)
(T)
(T)
(~
7. 8. 9. 10.
There is only one intersect of the asymptotes of the complete root loci. The intersect of the asymptotes must always be on the real axis. The breakaway points of the root loci must always be on the real axis. Given the equation I
(F) (F) (F)
(T)
of.~ and does not contain
+ KGl (S)HI (s) = O. where GI (S)Hl (s) is a rational function K. the roots of dGI (S)Hl (s}jds are all breakaway points
(T)
on the root loci (00 < K
< (0).
(F)
11.
At the breakaway points on the root loci, the root sensitivity is infinite.
(T)
(F)
12. Without modification. all the rules and properties for the construction of root loci in the splane can be applied to the construction of root loci of discretedata systems in the zplane. (T) (F) 13. The detennination of the intersections of the root loci in the splane with the jwaxis can be (T) (F) made by solving the auxiliary equatioll of Routhts tabulation of the equation. 14. Adding a pole to Q(s) / P(s) has the general effect of pushing the root loci to the right, whereas (T) (F) adding a zero pushes the loci to the left. Answers to these trueand·Calse questions can be found on this book's companion Web site: www.wiley.comlcollege/golnaraghi.
102 > Chapter 7. Root locus Analysis
, REFERENCES
Jeneral Subjects
1. W. R. Evans. "Graphical Analysis of Control Systems." Trans. AlEE. Vol. 67. pp. 548551. 1948. 2. W. R. Evans. ··Control System Synthesis by Root Locus Method," Trans. AlEE. Vol. 69. pp. 6669. 1950. 3. W. R. Evans. Control System Dynamics. McGrawHill Book Company. New York. 1954.
:onstruction and Properties of Root Loci
4. C. C. MacDuff. Theory of Equations. pp. 29104. John Wiley & Sons. New York. 1954. 5. C. S. Lorens and R. C. Titsworth, "Properties of Root Locus Asymptotes." IRE Trans. Automatic Control. AC5, pp. 7172, Ian. 1960. 6. C. A. Stapleton. "On Root Locus Breakaway Points," IRE Trans. Automatic Control. Vol. AC7. pp. 8889. April 1962. 7. M. J. Remec, "Saddle~Points of a Complete Root Locus and an Algorithm for Their Easy Location in the Complex Frequency Plane," Prot. Natl. Electronics Conf., Vol. 21. pp. 605608. 1965. 8. C. F. Chen, 4'ANew Rule for Finding Breaking Points of Root Loci Involving Complex Roots." IEEE Trans. Automatic Control. AC10. pp. 373374. July ]965. 9. V. Krishran 1 "SemiAnalytic Approach to Root Locus." IEEE Trans. Automatic Control, Vol. ACIl, pp. 102108. Jan. 1966. 10. R. H. Labounty and C. H. Houpis, "Root Locus Analysis of a HighGrain Linear System with Variable Coefficients: Application of Horowitz's Method," IEEE Trans. Automatic Control. Vol. ACI1, pp. 255263. April 1966. 11. A. Fregosi and J. Feinstein. "Some Exclusive Properties of the Negative Root Locus," IEEE Trans. Automatic Control. Vol. AC14. pp. 304305. June 1969.
~nalytical
Representation of Root Loci
12. G. A. Bendrikov and K. F. Teodorchik, "The Analytic Theory of Constructing Root Loci. " Automation and Remote Control. pp. 340344, March 1959. 13. K. Sreiglitz. "Analytical Approach to Root Loci:' IRE Trans. Automatic Control. Vol. AC6, pp. 326332. Sept. 1961. 14. C. Wojcik, "Analytical Representation of Root Locus," Trans. ASME.J. Basic Engineering. Ser. D. Vol. 86. March 1964. 15. C. S. Chang. "An Analytical Method for Obcaining the Root Locus with Positive and Negative Gain," IEEE Trans. Automatic Control. Vol. ACIO. pp. 9294. Jan. 1965. 16. B. P. Bhattacharyya, "Root Locus Equations of the Fourth Degree:' lnterna. }. Control. Vol. I. No.6. pp. 533556. 1965.
loot Sensitivity
17. J. O. Truxal and M. Horowitz, "Sensitivity Consideration in Active Network Synthesis." Proc. Second Midwest Symposium on Circuit Theory, East Lansing. MI. 1956. 18. R. Y. Huang. «The Sensitivity of the Poles of Linear ClosedLoop Systems." IEEE Trans. Appl. Ind., Vol. 77. Part 2. pp. 182187, Sept. 1958. 19. H. Ur. "Root Locus Properties and Sensitivity Relations in Control Systems," IRE Trans. Automatic Control. Vol. AC5 t pp. 5865, Jan. 1960.
PROBLEMS
7.. 1. Find the angles of the asymptotes and the intersect of the asymptotes of the root loci of the following equations when K varies from _·x to 00. (a) 8 4 + 4s3 + 4s2 + (K + 8}s + K = 0 (b) 8 3 + 5s 2 + (K + l)s + K = 0 (c) s2 + K(s3 + 3; + 2s + 8) = 0 (d) s3 + 28 2 + If + K (s2  l)(s + 3) = 0 (e) s5 + 2s4 + 3$'; + K(s2 + 3s + 5) = 0 (f) .f4 + 2s2 + 10 + K(s + 5) = 0
Problems
72. 73. Use MATLAB to solveProb1em 71. Show that the asymptotes angles are
403
74.
Prove that the asymptotes center is
u,=
7 5.
2: finite poles ofG(s)H(s)  L tinite zeros ofG(s)H(s)
nm
Plot the asymptotes for K
> 0 and K < 0
GH=
for
K ') s(s + 2)(s" + 2s + 2)
76. For the loop transfer functions that follow. find the angle of departure or arrival of the root loci at the designated pole or zero.
(a) G(s)H(s)
= ($ + 1)($2 + 1)
= (s _
Ks
1)(s2
Ks
Angle of arrival (K < 0) and angle of departure (K > 0) at s = j.
(b) G(s)H(s)
+ 1) + 2s + 2)
=
I
Angle of arrival (K < 0) and angle of departure (K > 0) at s = j.
(c) G(s)H(s) = s(s + 2)($2
K
Angle of departure (K > 0) at s
+ j.
K (d) G(s)H(s) = s2(s2 + 2s + 2)
Angle of departure (K > 0) at s
(e) G(s)H{s) "(
S'"
= 1 + j.
+ j.
. _ K (S2
+ 2s + 2)
s+ 2)( s+3 )
Angle of arrival (K > 0) at s = 1
77. Prove that:
argGH; where argGH' is the phase angle of GH at the complex pole t ignoring the effect of that pole. (b) the arrival angle of the root locus at the complex zero is (}D = 1800  argGH'[ where argGH" is the phase angle of GH at the complex zero, ignoring the contribution of that particular zero.
(a) the departure angle of the root locus from a complex po]e is (}n = 1800
78. Find the angles of departure and arrival for all complex poles and zeros of the openloop transfer function of
79. Mark the K:::; 0 and K = ±oo points and the RL and complementory root loci (CRL) on the real axis for the polezero configumtions shown in Fig. 7P9. Add arrows on the root loci on the real axis in the direction of increasing K.
zeros at 4. no finite zeros (i) Poles at 0. 6. zeros at 1. no finite zeros (g) Poles at 0* 0. Find all the breakaway points of the root loci of the systems described by the polezero configurations shown in Fig. (a) $'''' + 3i + (K + 2)s + 5K 0 = (b) .j. Construct the root loci for K ~ o. 2 (I) Poles atj. 0. 1 + j.5 0 4 j 3 2 1 0 j (j 0 (e) (d) Figure 7P9 710. zero at 1 (e) Poles at O~ ° 713. 8. zero at 4 (d) Poles at 0. 712. zero at 8 (b) Poles at 0. j~J2. j2. zeros at 4 + j2. 2. 2. 4 . 2. 8. 5. 8. 8. 1 .. 8. 0. 7P9. j2. 4 (h) Poles at 0.j.. 40 (0) Poles at 0. 3 (n) Poles at 0. 2.104 . zeros at 2. The characteristic equation is obtained by equating the numerator of I + G(s)H(s) to zero. zeros at 5. zero at 5 (f) Poles at 0. Use MATLAB to solve Problem 712. 0. 8. I. 0. no finite zeros (e) Poles at O~ 0" 2. i=l i= 1 n 1 m I 711. 100. (k) Poles at j. 4. j. I~ 3. 1 .' + 8 + (K + 2)s + 3K = 2 0 . splane splane X I j2.f' Chapter 7.j2 (j) Poles at 2. 1 + j. zero at2 1 + j. 1. (a) Poles at 0.j. 0. 0. zeros at l~ 1 (m) Poles at 0. 200. The characteristic equations of linear control systems are given as follows. I . zeros at 0. j2. 714. Root locus Analvsis splane splane o (a) (J (b) jm . Prove that a breakaway a satisfies the fol1owing: L=La + Pi a + Z.. Construct the rootlocus diagram for each of the following control systems for which the poles and zeros of G(s)H(s) are given.
+ 2)s + 100 = 0 Construct the root loci of the equation for Kt ~ o.1) K(s+3")(. . with 721. Use MATLAB to solve Problem 714. 7. = 2~ (c) n = 3. 5P16 when K = 100 is . Use MATLAB to solve Problem 720. 717. . if such a solution exists.2s+K(s+4)(s+ 1) (g) =0 + K(s2 + 4s + 5) = 0 + 18 + 28 + K(s2 . (d) n 4.i 405 + 5Ks2 + 10 = 0 (d) 84 + (K + 3 )83 + {K + 1)82 + (2K + 5) (e) 83 + 282 + 2s + K (82 . The characteristic equation of the control system shown in Fig.(s+4)" Construct the root loci of the characteristic equation of the closedloop system for K ~ (a) n:::.o:82+4s+5:'") = = K(s + 2)(s + 3) s(s + 1) +~s + 5) Construct the root loci for K ~ O. Use MATLAB to verify your answer to Problem 718. 718. 22.2)(s + 4) = 0 (i) s(r .5. Find the value of K that makes the relative damping ratio of the closedloop system (measured by the dominant complex characteristic equation roots) equal to 0. Construct the root locus diagram for K 2:: O. Use MATLAB to verify your answer to Problem 716. The forward~path transfer function of a unityfeedback control system is G(s) _ K . A unityfeedback control system has the forward~path transfer functions given in the following. 716.I) + K(s + 2)(8 + 0.:__ (b) G( 8 ) = .l}2 (e) G(s) (d) G(s) = (.r2 + 2s + 2) K(s + 2) (g) G(s) = (s + 1)(82 + 68 + 10) (i) O(s) = s($2 (0 G(8 ) (h) 0(8) = ( 8+1)(. = 00.0.5) = ==1. and (e) Il .707..~+5) .1)(s + 2) = 0 (c) 8 3 + 10 = 0 (1) .3 + 25s2 + (lOOK.\"(8+1)(s+~3 )(.~ + O.S)(s _ K K 1. The fOlwardpath tmnsfer functions of a unityfeedback control system are given in the following: ( a) G(s)  .Problems .s(s + 2)(8 .5)(.0.\" + 1) 1 (0 G(s) = s(s2 + 6s + 25) 719.5) (s . Find the values of K at all the breakaway points. K K (a) G( s) = :.8(s2 + 4s + 4)(8 + 5)(8 + 6) K(8 + 3) (b) G(s) (d) G(s) = = K 8(S + 2)(s + 4)(s + 10) (e) G(s} = K(s2 + 2s + 8) s(s + 5)(8 + 10) (s + 2)2(s + 5)(8 + 6) K K(s2 + 4) K(s + 10) ( ) G( ) e s == s2(s + 2.: 1 ~ (b) n.v(s + 10)(s + 20) (e) G(s) = K(s . 720.5) = 0 (j) s4 + 2s3 + 28 2 + 2Ks + 5K = 0 (k) $5 + 2s4 + 383 + 2s2 + $ + K = 0 68 3 i + + 98 2 (h) 8 3 2 715.
Usc MATLAB to solve Problem 731. construct the root loci of the characteristic equation for A:::: O.25h)i + 116.6) 731.s Figure 7P24 725. (b) Construct the root loci for IX :::: 0 with K = 10. 7P24 is G(s) = K(s + ~) (s + 3) s(s~ . Construct the root loci of the characteristic equation for KI :::: O. = 100 (b) N = 20 and Ko = 50 (c) A = 100 and N (a) A = Ko = 20 Use MATLAB to verify your answer to Problem 733. The characteristic equation of the liquidlevel control system described in Problem 542 is written 0. (b) For N = 10 and K" = 50. 734.06s(s + 12 . 726. 727. The block diagram of a control system with tachometer feedback is shown in Fig. (b) Use MATLAB to verify your answer to part (a). s _ K(s + 0.s2(s + 3. construct thc roOl loci for K" 2: 0. The forwardpath transfer function of a control system is G (a) Construct the root loci for K :::: O. 732. (b) Set K = lO. a variable parameter. construct the root loci of the characteristic equat ion as N varies from 0 to oc . (c) For A = 50 and N = 20. The forwardpath transfer function of the control system shown in Fig. Use MATLAB to solve Problem 724. Root Locus Analysis 723. R(s) £(. Use MATLAB to verify your answer to Problem 726. 729. 728. The characteristic equation of the dcmotor control system described in Problems 449 and 540 can be approximated as 2. Use MATLAB to verify your answer to Problem 722. Construct the root loci of the characteristic equation for h :::: O.5 )(As + Ko) + 250N = 0 (a) For A = K " = 50. 733. 724.4) ( ) .84s + 1843 = 0 when KL = ex:.1) K + 1 Y(s) +   i(s + 5) K.05JLS 3 + (l + 1O. and the load inertia h is considered a<. . 730.1) (a) Construct the root loci for K ~ 0 with IX = 5. (a) Construct the root loci of the characteristic equation for K :::: 0 when Kf = O. 7P24. Repeat Problem 731 for the following cases.106 Chapter 7. Use MATLAB to solve Problem 728.
41. 7 .1. Let K = I" Jm = 0. 8 m :::::.". (b) Find the range of K value for which the system is stable. The characteristic equation of the dcmotor control system described in Problems 449 and 540 is given in the following when the motor shaft is considered to be rigid (KL = 00). ~ 0 to show the effects of variation of the load inel1ia on system pelfonnance.n G(s) = . it is desired to investigate the root loci of this equation for 00 < K < 00 and for various values of a.49 and 5 . (b) Repeat part (a) with K = 1000.r (1m + n2ids"!. The forwardpath transfer function of a unityfeedback system is 407 G(s) = (S2 K(s + 2)2 ') + 4)(s + 5t (a) Construct the root loci for K = 25. (a) Let K = 1. as the gain factor. and K. (b) Repeat part (a) when a = 4. For the dcmotor control system described in Problems 449 and 540.. (e) Determine the value of 0'. (a) Construct the root loci for 00 < K <: DC when a = 12.40. 737. L. n = 0.001. The transfer functions of a singlefeedbackIoop control system are G(s) = S""~S+ 1)( s+ 5) H(s) = ')( K 1 (b) Repeat (a) Construct the loci of the zeros of 1 + G(s) for K . 742. 743. Find an equivalent G(s)H(s) with Kl. 7 . (e) Use MATLAB to verify your answer to part (a). (e) Use MATLAB to verify your answer to part (a). 738.Problems 735.· = 1. 736. Use MATLAB to solve Problem 736. 0. La = 0. The forwardpath transfer function of a unityfeedback system is Ke. s+t (a) Construct the root loci for T = 1 sec and K> O. with the other system parameters as given in Problems 4 . Use MATLAB to verify your answer to Problem 740. (b) Use MATLAB to verify your answer to part (a). . part Ca) when H(s) = 1 + 5s. The transfer functions of a singlefeedbackIoop control system are 10 G(s) = 2( 1)( 5) H(s) = 1 + Td S s s+ s+ (a) Construct the root loci of the characteristic equation for ~J 2:: O. it is of interest to study the effects of the motorshaft compliance KL on the system performance.. Kb = 0.::: O. Construct the root loci of the characteristic equation for KL ~ O. Kj = 9. 739. so that there is only one nonzero breakaway point on the entire root )oci for .no < K < 00. The system can be approximated as a fourthorder system by canceling the large negative pole and zero of G(s)H(s) that are very dose to each other. Given the eqnation sJ + as2 + Ks + K = 0.+ (RaJm + n2RoJL + BmLa)s2 + (RaBm + KiKb)S + IlK.0636. 40.. (b) Use MATLAB to verify your answer to part (a).001. (b) Find the values of K where the system is stable.KiK = 0 (a) Construct the root loci for Jr.. Construct the root loci. Rll = 5..
746. one. The polezero configuration of G(s)H(s ) of a si nglefeedbackloop control system is shown in Fig. splane j(f) splane j (f) splane K== K=oo K=O jlO K=O o jlO j 12 (J o K=O K== (b) (J o K=O K=oo (c) (J (a) Polezero configuration splane j(f) splane j(f) splane K= oo K=oo K=O oo<K K=O K=O K=O oo K K=O 0 o a o a a K=O K=oo • Cd) (e) (f) Figure 7P47 . Design a proper controller H(s) for the system. Root Locus Analysis 7·44. 7P47(a).Ts ~ y I I Figure 7P46 747. and two breakaway points. 745. Construct the root loci for 00 < K < ()C for all three cases. x :9 jl2 ·1 H(s) ~I G(s)e. apply the angleofdeparture (and arrival) property of the root loci to determine which rootlocus diagram shown is the correct one. Use MATLAB to solve Problem 743 . Without actually plotting. Fig. The forwardpath transfer function of a unityfeedback control system is _ K(s + a) ( ) GS.s2(s + 3) Determine the values of ex so that the root loci (00 < K < 00) will have zero. 7P46 shows the block diagram of a unityfeedback control system. respectively.J08 Chapter 7. not including the one at s = O.
in the frequency domain. We leamed in Chapter 5 that the time response of a control system is usually more difficult to determine analytically. then the Laplace transforms of the input and the output are related through Y(s) = M(s) R(s) (83) 409 . will be a sinusoid with the same frequency but possibly with different amplitude and phase. that is. Therefore. r(t) = RsinUJot (81) w() the steadystate output of the system. It is well known from linear system theory that when the input to a linear timeinvariant system is sinusoidal with amplitude R and frequency wo. delay time. y(t) = Ysin((lJot + rjJ) (82) where Y is the amplitude of the output sine wave and ¢ is the phase shift in degrees or radians. yet). Let the transfer function of a linear SISO system be M(s). settling time. there are no unified methods of aniving at a designed system that meets the timedomain perfol1nance specifications.osting analytical tools. In design problems. It may never be. Rather. The reason is that the pelfol1nance of most control systems is judged based on the time responses due to certain test signals. rise time. so the timedomain properties of the system can be predicted based on the frequencydomain characteristics. especially for highorder systems. With these concepts in mind. such as maximum overshoot. The frequency domain is also more convenient for measurements of system sensitivity to noise and parameter variations. The starting point for frequencydomain analysis of a linear system is its transfer function . there is a wealth of graphical methods available that are not limited to loworder systems. the perfonnance of a control system is more realistically measured by its timedomain characteristics. This is in contrast to the analysis and design of communication systems for which the frequency response is of more impoltance. and so on. Another reason is that it presents an alternative point of view to controlsystem problems. It is important to realize that there are correlating relations between the frequencydomain and the timedomain performances in a linear system. In practice. we consider the primary motivation for conducting control systems analysis and design in the frequency domain to be convenience and the availability of the e. to conduct a frequencydomain analysis of a linear control system does not imply that the system will only be subject to a sinusoidal input. CHAPTER ·8 FrequencyDomain Analysis 81 INTRODUCTION The basic concepts and background material for this subject appear in Chapter 2. we will be able to project the timedomain performance of the system.. from the frequencyresponse studies. On the other hand. since most of the signals to be processed are either sinusoidal or composed of sinusoidal components. which often provides valuable or crucial information in the complex analysis and design of control systems.
Eq. and the phase characteristic~ LM(jw). The crux offrequencydomain analysis is that the amplitude and phase characteristics of a closedloop system can be used to predict both timedomain transient and steadystate system performances. Loop Systems For the singleloop controlsystem configuration studied in the preceding chapters.. 1. respectively. Eq. M(jw) = IMUw)[iM(jw) .. Y(jw) G(jw) M (Jw) = RUm) = 1 + G(jw)H ( jw) (811 ) The sinusoidal steadystate transfer function M(jm) may be expressed in terms of its magnitude and phase.. by knowing the transfer function M(s) of a linear system. (810) becomes . for the input and output signals described by Eqs. IM(jw)I. 8. s = jw. 1 Frequency Response of Closed . we replace s by jw.R(s) . FrequencyDomain Analysis For sinusoidal steadystate analysis. that is. ) . the amplitude of the output sinusoid is Y = R!M(jwo) [ and the phase of the output is (88) ¢ = LM (jwo) (89) Thus.410 Chapter B.I + G(s)H(s) G(s) (810) Under the sinusoidal steady state. (81) and (82). and the last equation becomes Y(jw) = M(jw)R(jw) By writing the function Y(jw) as (84) Y(jw) = IY(jw)ILY(jw) (85) with similar definitions for M(jw) and R(jw). completely describe the steadywstate performance when the input is a sinusoid. the closedloop transfer function is M ~ _ Y(s) _ (. the magnitude characteristic. (84) leads to the magnitude relation between the input and the output: [Y(jw)1 = [M(jw) 1 [R(jw)1 and the phase relation: (86) LY(jw) = LM(jw) + LR(jw) (87) Thus.
frequency noise. Or M(jw) can be expressed in terms of its real and imaginary palts: MUw) = Re[M(jw)] + jIm[MUw)] The magnitude of M(jw) is (813) G(jw) !M(Jw)! = 1 + G(jw)H( jw) = 11 and the phase of M(jw) is . the output Y(jev) would be identical to the input R(jw) for all frequencies. as we shall see that they affect the stability of the system. since all signals would be passed without distortion below the frequency We. 81 shows the gain and phase characteristics of an ideal lowpass filter that has a sharp cutoff frequency at We. (814). and the control system is regarded as a signal processor. impossible to achieve in practice~ nor would it be desirable. of course. if the ideal lowpassfilter characteristics shown in Fig.tics on the input signal. An infinite magnitude of G(jaJ) is.. and completely eliminated at frequencies above W(' where noise may lie. they would be highly desirable for a control system. Thus. we see that. Furthermore. since most control systems may become unstable when their loop gains become very high. I I + G(jw)H{ jw)1 [GUw)1 (814) LM(jw) == Q>M( jw) = LG(jev) . In many ways. rr Wc is increased indefinitely. in addition to responding to the input signal. the design of control systems is quite similar to filter design.i[l + G(jw)H( jw)] If M(s) represents the inputoutput transfer function of an electric filter. Fig. for IM(jw)J to be unity at all frequencies. From Eq. such as airframe vibration of an aircraft. For control systems with high. In fact. the magnitude of G(jm) must be infinite. It is well known that an ideal filter characteristic is physical1y unrealizable. all control systems are subject to noise during operation. 81 were physicaIly realizable.81 Introduction IM(jco)1 J 411 Figure 81 Gainphase characteristics of an ideal lowpass filter. the frequency response should have a finite cutoff frequency WeThe phase characteristics of the frequency response of a control system are also of importance. then the magnitude and phase of M(jw) indicate the filtering characteris. the system should be able to reject and suppress noise and unwanted signals. Such a system would follow a stepfunction input in the time domain exactly. .
the magnitude of M. it is necessary to define a set of specifications so that the performance of the system can be identified. In practice. a large M. (814) and (815). Nonnally.. • BW gives an indication of the transient response properties of a control !iystem. For most control systems. 812 FrequencyDomain Specifications In the design of linear control systems using the frequencydomain methods.1 and 1. The following frequencydomain specifications are often used in practice.> Chapter B. the frequency responses of G(s) and H(s) can often be determined by applying sinewave inputs to the system and sweeping the frequency from 0 to a value beyond the frequency range of the system. As shown by Eqs.412 . corresponds to a large maximum overshoot of the step response. A large bandwidth corresponds to a faster rise time. • M]" indicates the relative stability of a stable c1osedloop system. its zerofrequency value. the bandwidth of a control system gives indication on the transient~ robustness of the system. the gain and phase of a closed~loop system can be determined from the forwardpath and loop transfer functions. response properties in the time domain. 82 illustrates typical gain and phase characteristics of a control system.707 I ________ 1I ______ _ 1 I I 1 1 BW Figure 82 Typical gainphase characteristics of a feedback control system. Specifications such as the maximum overshoot~ damping ratio. . it is generally accepted in practice that the desirable value of Mr should be between 1. the noisefiltering charac teristics and In general.7% of. and the like used in the time domain can no longer be used directly in the frequency domain. Resonant Frequency Cdr The resonantfrequency Cdr is the frequency at which the peak resonance Mr occurs. Bandwidth BW The bandwidth BW is the frequency at which IM(j&J)1 drops to 70.5. gives indication on the relative stability of a stable closedloop system. In general. Resonant Peak Mr The resonant peak Mr is the maximum value of IM(jw) I. Fig. or 3 dB • BW gives an indication of down from. Frequency~DDmain Analysis I I 1 0.
a 2 MI'l w" w and Bandwidth of the Prototype SecondOrder System ~. The perfonnance criteria for the frequencydomain defined above are illustrated in Fig... Apparently. A robust system is one that is insensitive to parameter variations. 413 since higherfrequency signals are more easily passed through the system. Eq. which is called the cutoff rate of the frequency response. 1 M(Ju) = . The robustness represents a measure of the sensitivity of a system to parameter variations. but the cutoff rates may be different. and the bandwidth BWare all uniquely related to the damping ratio ~ and the natural undamped frequency (J)1l of the system.: = 2 1 + j2u~ . if the bandwidth is small. Sometimes it may be necessary to look at the slope of IM(jw) I. s = jev.. at high frequencies.. (816) At sinusoidal steady state. the resonant peak Mro the resonant frequency w. (816) becomes M(jw) = y (~w) R(jw) = (j()))2+2~wt!(jw) +w~ 1 0. (817) We can simplify Eq.= Yes) R(s) s 2 w~ + 2~(J)lls + wir . (8 20) w . Eq. .. Then. 82. Bandwidth also indicates the noisefiltering characteristics and the robustness of the system. Cutoff Rate Often. (817) by letting l( = co/con..1 lw' 2~tl. Conversely. only signals of relatively low frequencies are passed. two systems can have the same bandwidth. Consider the closedloop transfer function of the prototype secondworder system Resonant Peak and Resonant Frequency M{s) = . Other important criteria for the frequency domain will be defined in later sections of this chapter. 8~2 Mrl 821 {UrI AND BANDWIDTH OF THE PROTOTYPE SECONDORDER SYSTEM For the prototype secondorder system defined in Section 56. (8 17) becomes w . bandwidth alone is inadequate to indicate the ability of a system in distinguishing signals from noise. and the time response will be slow and sluggish..u (818) The magnitude and phase of MUu) are (819) and !M( ju) == "'M( ju) = tan.
damping ratio ~ only. MCz. MCz.707 1 ~. (820) for u and simplifying. Thus. 001: 3 z=lj MCz.707.40. (821) from which we get (822) In nonnalized frequency. z=z+l. This means simply that. the roots of Eq. MR is a function of ~ only. Eq. (824) is meaningful only for 2~2 :5 1.z=z+1: M(z.Z=Z+li M(z.707 (825) • For the prototype secondIt is important to note that. (822) are Ur =0 and The solution of Ur = 0 merely indicates that the slope of the IM( ju) Iversusw curve is zero at w 0. Mr is a function of the order system.2 0. and (J)r is a function of both { and ron' Furthermore.i)=abs(1/Cl+(j*2*zeta(z)*u)(uA 2))).i)=abs(1/(1+(j*2*zeta(z)*u)(uA 2))). (8~23) gives the resonant frequency = (824) Because frequency is a real quantity.i)=abs(1/(1+(j*2*zeta(z)*u)(uA2))). zeta = [00. or ~ ::.higherorder systems.i)=abs(1/(1+(j*2*zeta(z)*u)(uA2))).z=z+1. this analytical method is quite tedious and is not recommended.707. i)= abs(1/(1+(j*2*zeta(z) ~fU)_(UA2»).707. . it is not a true maximum if { is less than 0. Eq.i)=abs(1/(1+(j*2*zeta(z)*u)(u A2))).. Toolbox 821 MATLAB statements for Fig.i)=abs(1/(1+Cj*2*zeta(z)*u)(uA 2))). although taking the derivative of IM( ju)1 with respect to u is a valid method of detennining Mr and W r• for • For the prototype second. for all values of ~ greater than 0. the resonant frequency is (J)r = 0 and Mr = 1. highorder systems. Frequency~Domain Analysis respectively.Z=Z+1.0] for u=O = 0 . we get Mr = 2~~ ~:5 0. Graphical methods to be discussed and computer methods are much more efficient for = () when l. (823) into Eq. Substituting Eq.z=Z+1.414 ~ Chapter 8. The resonant frequency is determined by setting the derivative of IM( ju)1 with respect to u to zero.707. 83 i=l. ~ 0. MCz. 5 2. for the prototype secondorder system.z=z+li M(z. order system. 0.10.60. M" = ) and (1).
. 84 and 85 illustrate the relationship between My and ~.:) . hold on. respectively.82 Mr . for i = 1 ~ length ( zeta) plot(u. end xlabel( \mu = \omega/\omega_n'). Wr ::::: (J)1l' Figs. and ur (::::: WI'/wn) and r. (824).001:3. as indicated by Eq.~ ! 3rff+1~~ 2r~r+1+_4_1 o u= (1)/co" 2 3 Figure 83 Magnification versus nonnalized frequency of the prototype secondorder control system. (819) versus u for various values of Notice that.i)=abs(l/(1+(j*2*zeta(z)*u)(uA 2))). ax. grid I Fig.is ( [0 3 0 6] ) . 83 illustrates the plots of IM( ju)1 of Eq. ylabel(' IM(j\omega)I ').z=z+1. When = O. s· s 6~~~~~ 5~~~~1 4~f4~~~1 ~ .i)=abs(1/(l+(j*2*zeta(z)*u)(uA 2)). 6)" and Bandwidth of the Prototype SecondOrder System ~ 415 M(z.z=z+lj M(z. i=i+1j end u:::::O:O. the value of lVr = UraJn would increase when ~ decreases.M(i. if the frequency scale were unnonnalized.
822 Bandwidth In accordance with the definition of bandwidth.u) 2 +(2("ut 1/2 = v2 (827) 1.707 ~ 1.s 0.707 1.0 0.~2 ~ ~ : I I I o 0... we set the value of jM(ju)1 to 1/..2 o o. Chapter 8. (826) Thus~ [ (1 ') "] ~ . 0.0 Damping ratio .5 2..2~"'... II ::so.0 Figure 85 Normalized resonant frequency versus damping ratioJor the ~rototype system.707. 0. u.. 0..416 . FrequencywDomain Analysis 5 4 \ 3 2 \ \ Mr= 1 2(J1.6 S.8 ~ .J2 9i.. second~order Damping ralio . = 1 . 1.4 0.0 Figure 84 My versus damping ratio for the prototype second~order system.5 0.
M. The resonant peak. 1\2+2)) . zeta(i)=zetai i=i+1. that is. (825)]. However. Mr is infinite. BW increases and decreases linearly with W Il • BWalso decreases with an increase in t for a fixed Wn 2. When { is negative. M. the system is unstable. 84). since u must be a positive real quantity for any (. end TMP_COLOR. (5~ 103) also depends only on ~. xIabel( I \zeta'). The summary of these relationships is as follows.1\44*zetai.M) .. the bandwidth of the prototype secondorder system is detelmined from Eq. the maximum overshoot is zero when ~ 2::: 1. Bandwidth is directly proportional to W 17 [Eq. no h)nger has any meaning.. 3. axis ( [0 1. pIot(zeta.. 85). and WI' = 0 (see Fig. ceases to have any meaning. (829)]. Mr decreases.2 M(iJ = sqrt( (12*zetai~ 1\2. • When a system is unstable. Notice that. and Bandwidth of the Prototype SecondOrder System·: 417 • BW/w" decreases monotonically as the damping mtio ~ decreases. the maximum overshoot in Eq. and the value of M. Even more important. for zetai=O: sqrt (1/2)/100: 1.. as ~ increases. ylabel('BW/\omega_n').8~2 M. . Eq. For ~? 0.707. grid The plus sign should be chosen in the last equation.. (iJr. As ~ increases.::: 1 . Mr of the closedloop frequency response depends on ~ only [Eq.)+sqrt(4*zetai . 8~6 shows a plot of BW /w n as a function of ~. 1. 2 0 2]) . BW /wn decreases monotonically. which leads to (828) Toolbox 822 kfATLAB statenlents for Fig. Therefore. 8·6 clear all i=l. (828) as (829) • BW is directly proportional to (tJn' Fig. When ~ is zero. = 1 (see Fig. (829) shows that BW is directly proportional to W il • We have established some simple relationships between the timedomain response and the frequencydomain characteristics of the prototype secondorder system. In comparison with the unitstep time response.
0 1. (816). 521. rise time increases as (JJn decreases. Therefore. FrequencywDomain Analysis 2. When other secondorder or higherorder systems are involved. BWand rise time are inversely proportional to each other.8 wJ(l.2 0. ~ 83 EFFECTS OF ADDING A ZERO TO THE FORWARDPATH TRANSFER FUNCTION The relationships between the timedomain and the frequencydomain responses arrived at in the preceding section apply only to the prototype secondorder system described by Eq. it is more realistic from a design standpoint to modify the forwardpath transfer function. (see Fig. fJ2 0. (816) may be considered as that of a unityfeedback control system with the prototype secondorder forwardpath transfer function G(s)  (J)2 n .4 0.. 4. (5108) and Fig. as demonstrated in Eq.2 ~ 0.1 1.s(s + 2?. It is of interest to consider the effects on the frequencydomain response when poles and zeros are added to the prototype secondorder transfer function.6 ~ 1. unitstep response. the relationships are different and may be more complex.8 BW = 0. 87."(JJn) (830) . however.2 ~ o 0. The correlations among pole locations.4  4l.4 ~. Figure 86 Bandwidthlwn versus damping ratio for the prototype secondorder system. Chapter 8. 86). • Bandwidth and rise time are inversely proportional to each other.2+ 2 ~ 1.6 1. The closedloop transfer function of Eq.418 . and the magnitude of the frequency response for the prototype secondorder system are summarized in Fig. For the unitstep response.707.0 Damping ratio .2~2} + v'4l. It is simpler to study the effects of adding poles and zeros to the c1osedloop transfer function. Bandwidth and Mr are proportional to each other for 0 ( ~ 0.
gets larger. As 1.3 dB Frequency response o I+lrOO As 1 0.917'(. liS add a zero at s = . 419 Pole locations As ~. gets larger.83 Effects of Adding a Zero to the ForwardPath Transfer Function . the larger the bandwidth is. As (Un gets larger.4167~ + 2.0 I /~ 0. r gets smaller and the system r responds faster. angular distance from negafive real axis gets smaller.. jOJ sPlane Protolype secondorder system ret) Y(l) As ~ gets larger. the faster the sY5tem will respond.Unitstep response . Therefore. unitstep response.. y(l) IM(jm) I OdB~ 1.9 . As r.gets larger and the system [llIijlonds slower. BW gets smaller. gets larger. Increasing (Uti itlcreases BW and decreases IT' Increasing '(. Let. BW gets larger. wn 1 + Ts) 2 S2 + ( 2l. pole distance from on gin gets larger. decreases BW and increases IT' Figure 87 Correlation among pole locations.wn ) (831) The closedloop transfer function is = M( s) 2( .wll + TW"2) s + (VII (832) . and the magnitude of frequency response of the prototype secondorder system. t. Bandwidth and rise time are inversely proportional.2 m tl o (Uti gets larger.1IT to the transfer function so that Eg . (830) becomes G(s) = w~(l + Ts) s(s + 2l.
over a range of small values of T. Notice that the general effect of adding a z8TO to the forwardpath transfer function is to increase the bandwidth of the closedloop system. Frequency. (833) affects the bandwidth. However.in the previous section. Chapter 8. . except for a range of small values of T. ulT2 Il (834) While it is difficult to see how each of the parameters in Eq.2ai n n . ~ = 0. Il + 4"aiT . 88..420 .2.... for which BW is actually decreased. the bandwidth of the system is found closedloop system. and T takes on various values. as shown in Fig... 89(a) and 89(b) give the plots of [M{jw)] of the closedloop system that has the G(s) of Eq. 88 shows the relationship between BW and Tfor ~ = 0.jb2 + 4w~ ) l~ (833) b = 4. However. After a length derivation. because there are now three parameters in ~.707 and 0. Figs. the bandwidth is actually decreased.707 and (Vn 1. Mr . ll)m and T. Oomain Analysis • The general effect of In principle.414)1. and BWare difficult to obtain analytically even though the increase the BW of the system is still second order. = 6 5 4 / V / 2 ~ _v 2 / V 3 T /V 4 /7 o 5 6 7 Figure 8~8 Bandwidth of a secondorder system with openloop transfer function 0(5) = (1 + T.2a1 . Fig. respectively. to be BW = where ( b + ~. These curves verify that the bandwidth generally increases with the increase ofTby the addition of a zero to G(s). Wn and BWofthe system can all be derived using the same steps used adding a zero to the forward. (831) as its forwardpath transfer function: WI! = 1. path transfer function is to the exact expression for M. (Vr.y)/[s(s + 1..
(s 1 +Ts + 1.6 0. (a) {lJ1l = It t = 0.0 G(s) 0.707 '§ =.707 (b) WIJ = 1.83 Effects of Adding a Zero to the ForwardPath Transfer Function \.0 1'=:5 0.4 2..8 0.2 ~ 421 1.0 G {s) = s(s + 0.4 o 2 (j)(radlsec) (b) 3 4 Figure 89 Magnification curves for the secondorder system with the forwardpath transfer function G(s) in Eq.8 0.4 0.414) r=S ~  ~ 0.4) l+T$ 1.2 0 2 lU (rad/sec) 3 4 (a) 2.~ ~ 1.6 i .707 0. . (832). ~ ~ 0.2 1.2..8 2.
ylabel(' IM(j\omega)I '). moves very close to the origin. t=t+l. abs( (l+(T(t) *8) )/(s"2+(2*zeta+T(t)) *s+l)).10]. the zero of the closedloop transfer function.Me! I :)) .2. for W=O : 0 . ylabel(' IM(j\ornega)[').t=t+l. but the larg(~ time constant of the zero near the origin of the splane causes the time response to drag out in reaching the final steady state (Le. zeta=O. 2] ) . abs( (l+(T(t) *s) )/(s"2+(2*zeta+T(t) )*s+1)) . ab8 CCl+(T(t) *8) )/(s"2+(2*zeta+T(t)) *8+1)) .M(i. t=t+l.i) = abs«1+(T(t)*s))/Cs"2+C2*zeta+T(t))*s+1)) . These curves show that a high bandwidth corresponds to a faster rise time.t=t+l.0:0. Met.T=[S 1. zeta=O. B9(b) clear all i=l. 810 illustrates the situation that the rise time is fast. for i = 1: length(T) plotCw . t::::t+l. Thus. = = = = w=0:O. M(t Ii) = abs«1+(T(t)*s))/(sA2+(2*zeta+T(t))i~s+1)) . causing the system to have a larg(~ time constant. t=t+l.. i) M(t. end w::. 707. 89(a) clear all i=l . s=j*Wj MCt I i) M(t t i) M(t . as T become very large. i) M(t.:)) . Fig. axi s ( [0 4 0 1. 5] .. 8] ) . . 810 and 811 show the corresponding unitstep responses of the closedloop system.t=t+li Met Ii) :i. i) = abse el+(T(t) *s) )/es"2+(2*zeta+T(t)) *s+1)) . t=t+1. for w=O: 0. M(t. hold on. grid Figs. xlabel('\omega(rad/sec)')..t=t+l. Met.'ts) )/(s"2+(2*zeta+T(t) )*s+1)) . s=j*w.422 ~ Chapter 8.t=t+1.i) = abs(e1+eT(t)*s))/(s"2+(2*zeta+T(t))*s+1)) . end xlabel('\omegaCrad/sec)')..=i+1: = abs«1+(T(t)*s))/(s"2+e2*zeta+T(t))*s+1)) . T= [0 0 . 2 5 2 1 0 . However. FrequencyDomain Analysis Toolbox 831 MATLAB statements for Fig. grid MATLAB statements for Fig. 001 : 4 t=l. axis ( [0 4 0 2 .01:4: hold on: end for i = 1: length(T) plot(w.OOl:4: TMP_COLOR= 1.i) i=i+1.41410. abs( (l+(T(t) .i) M(t. t=t+l. = absCC1+(T(t)*s))/(s"2+(2*zeta+TCt))*s+1)) . abs( (1+(T(t) *s) )/(s"2+(2*zeta+T(t)) *8+1)) . end . the settling time will be longer). which is at s = 1fT.01: 4 t=l.
8 ~ 0.J ~ ~~ .4 G( s) = s(s + Q.2 T.2 'f~() ~ 1.0 T= 0.2 o V II J' 2 3 4 5 6 7 8 9 Time (sec) Figure 8~10 Unitstep responses of a secondorder system with a forwardpath transfer function G(s).0 . 1.*5.0 1.Ol 5) = s(s + 1.6 ( ~~r.0 I / 0.4) 1 + Ts 0. ~ P" io"""" G( I""" 0.8 ~ 0.414) 1 + Ts 0.2 o N 2 3 4 5 6 7 8 Time (sec) Figure 8 11 Unitstep responses of a secondorder system with a forwardpath transfer function G(s).4 (/:1' I"T=O..2 T=2.8~3 Effects of Adding a Zero to the Forward~Path Transfer Function III 423 1. .6 IJ1 1 If ~ I~ 1.Q":~ 0.
. for i=1 : length(T) num=[T(i) 1] .707 t M (835) 1M . t:=O:O. ylabel( I yet) ') . end xlabel( I Time ').2].2i for i=l: length(T) num=[T(i) 1] .M(i. close all. zeta= 0.10.use clear all. hold on. M(i.den. den = [1 2*zeta+T(i) 1].:)=step(num. FrequencyDomain Analysis Toolbox 832 MATLAB statements for Fig.t).0101.424 ~ Chapter 8.4140. ) .t). t==O:O. loop system less stable and (835) is quite tedious. grid I Toolbox 8·3·3 MATLAR statements for Fig. I xlabel ( Time' ) . for i = 1: length(T) plot(t . grid ~ 8~4 EFFECTS OF ADDING A POLE TO THE FORWARDPATH TRANSFER FUNCTION Adding a pole at s = 1 IT to the forwardpath transfer function of Eq. (mel de T==(5 1. 707. :)). 8/ / .use clear ali.01:9. 8 12~ which shows the plots of (jw) I versus w for W" = I. We can obtain a qualitative indication on the bandwidth properties decreases the bandwidth. 810 . hold on. ylabel( yet) . den = [1 2*zeta+T(i) 1].01:9j zeta = O.M(i. end lIlld cle if necessar)' for i end = 1: length(T) plot(t . close all. :)).den. by referring to Fig. (830) leads to • Adding a pole to the forwardpath transfer function makes the closedThe derivation of the bandwidth of the closedloop system with G(s) given in Eq. M(i.. T=[l 50. = 0. end if necessary TMP_COLOR = 1 .:)=step(num.
0 0. When Tbecomes large. o 0.5 1.0 1.8 1. T 1 and T = 5. The unitstep responses of Fig. Thus.707 0.5 2. 813 show that. the following relations are observed: = = 1. 2. it can be unstable for a certain set of system parameters. the bandwidth of the system is slightly increased by the addition of the pole.5 Figure 812 Magnification curves for a thirdorder system with a forwardpath transfer function G(s).414)( l I + Ta) 1. The rise time increases with the decrease of the bandwidth.0 ~ 425 2.. the pole added to G(s) has the effect of decreasing the bandwidth but increasing Mr. 812 show that.0 ~ i..5 2. for (J)n = 1 and ~ O. for small values of T. of a thirdorder system with a forwardpath transfer function G(s). 707 ~ the system is stable for all positive values of T. Because the system is now of the third order. forlargervalues ofT.0 0. the effect of adding a pole to the forwardpath transfer function is to make the closedloop system less stable while decreasing the bandwidth. 0 2 4 6 8 12 14 16 18 Time (sec) .5 1.41~)(1 + Ts) 10 Figure 813 Unitstep 0... The larger values of M.2 response. The 1M (jw) Iversusw curves of Fig. ' G(s) = S(S + 1. also correspond to a larger maximum overshoot in the unit~step responses. 1. butMris also increased. It can be shown that. in general.0 w(rad!sec) and various values of T.o:.4 ~ 1. we can conclude that.84 Effects of Adding a Pole to the ForwardPath Transfer Function 3..6 G(s) = s(s + 1.
t=linspace(O. The objective of these last two sections is to demonstrate the simple relationships between BW. On the other hand. title ( .. . SpecificallY7 the Nyquist plot of l1. Typical effects on BW of adding a pole and a zero to the forward~path transfer function are investigated. and the closedloop system is marginally stable. the loop transfer function G(s)H(s)~ or L(s). %step response for basic system plot(t .:::. Of course. In addition to providing the absolute stability. 707. %timevector y=step(nc I de. if the coefficients of the characteristic equation are all known. zeta. when the system is unstable.1). " 85 NYQUIST STABILITY CRITERION: FUNDAMENTALS • The Nyquist plot of L(jw) Thus far we have presented two methods of determining the stability of linear SISO is done in polar coordinates systems: the RouthHurwitz criterion and the rootlocus method of detennining stability by as (JJ varies from 0 to 00.use clear all. but it no longer has any significance.y). M no " longer has any meaning. like the RouthHurwitz criterion.den.x. den=conv(den.1001).:::. FrequencyDomain Analysis The correlation between M I . for T= [0 O. [1 2*zeta*wn]). This is another example of using the properties of the loop transfer function to find the perfonnance of the closedloop system. if needed. num=[wnA 2]j den=conv([l 0]. and clc ifnecessary wn.426 ? Chapter 8. the Nyquist plot. of stability. t) . ylabel('Amplitude'). The Nyquist criterion has the following features that make it an alternative method that is attractive for the analysis and design of control systems.1.dc]=feedback(num. hold on end xlabelC 'w (rad/s)'). It also gives an indication of how the system stability may be improved. locating the roots of the characteristic equation in the splane. the val ue of IM(jw) I is analytically finite.s) is a plot ofL (jro) in the polar coordinates of 1m [L(jw)] versus Re[L(jw)] as w varies from 0 to 00.4. 1. 812 . Toolbox 841 MATIAB statements/or Fig. close all. [nc. the closedloop system is marginally stable.l.O. the Nyquist criterion also gives infonnation on the relative stability of a stable system and the degree of instability of an unstable system. we can solve for the roots of the equation by use of MATLAB. [T 1]). and the maximum overshoot of the step response is meaningful only when the system is stable. 5 1 5] .1. No attempt is made to include all general cases. Mn and the timedomain response. • When Mr = . When the system is unstable. • The Nyquist criterion also The Nyquist criterion is a semi graphical method that determines the stability of a closedgives indication on relative loop system by investigating the properties of the frequencydomain plot. When G(jw) = 1~ IM(jw) I is infinite. Step Response' ) .
Or.Loop Stability: A system is said to be openloop stable if the poles of the loop transfer function L(s) are all in the lefthalf splane. Unlike the rootlocus method. The Nyquist plot is useful for systems with pure time delay that cannot be treated with the RouthHurwitz criterion and are difficult to analyze with the root~locus method.ls sP(l + Tas)(1 + TbS) . + L(s) = roots of the characteristic Stability Conditions We define two types of stability with respect to the system configuration.. the characteristic equation roots must satisfy 6(s) = I + G(s)H(s) = 0 (838) In general. Because the characteristic equation is obtained by setting the denominator polynomial of M(s) to zero. Before embarking on the details of the Nyquist criterion. and Ttl is a real time delay. Wn BW. The Nyquist plot of G(s)H(s) or of L(s) is very easy to obtain. (I + Tns) (837) where the T's are real or complex~conjugate coefficients. 427 2. and others with ease.. This subject is also treated in Appendix F for the general case where the loop transfer function is of nomninimumphase type. the roots of the characteristic equation are also the zeros of 1 + G(s)H(s).I + G(s)H(s) G(s) (836) where G(s)H(s) can assume the following form: G(s)H(s) = K(l + T1s)(1 + Tzs) .85 Nyquist Stability Criterion: Fundamentals . The Nyquist plot of G(s)H(s) gives information on the frequencydomain characteristics such as M". especially with the aid of a computer. For a singleloop . 851 Stability Problem The Nyquist criterion represents a method of determining the location of the characteristic equation roots with respect to the left half and the right half of the splane.. 4. for a system with multiple number of loops. • Open. it is useful to summarize the polezero relationships of the various system transfer functions.. 3. (837). the Nyquist criterion does not give the exact location of the characteristic equation roots. Identification of Poles and Zeros Loop transfer function zeros: zeros of L(s) Loop transfer function poles: poles of L(s) Closedloop transfer function poles: zeros of I equation poles of I + L(s) = poles of L(s). (1 + ~ns) eT.. the denominator of M(s) can be written as (839) A(s) = I + L( s) = 0 where L(s) is the loop transfer function and is of the form of Eq. Let us consider that the closedloop transfer function of a SISO system is Ms _ ( ).
. because A is inside the closed path. For example.428 ". 815(a) and (b) are considered to be enclosedby the closed path r. Enclosed A point or region is said to be enclosed by a closed path if it is encircled in the CCW direction or the point or region lies to the left ofthe path when the path is traversed in the prescribed direction. 8IS(b) is not. this is equivalent to the system being stable when the loop is opened at any point. or simply stable. the encirclement. point A in Fig.. if the poles of the closedloop transfer function or the zeros of 1 + L(s) are all in the left. Furthermore. but point A in Fig. Figure 814 Definition of encirclement. The concept of enclosure is particularly useful if only a portion of the closed path is shown. 814 is encircled by the closed path r. Point B is not encircled by the closed path f.half splane. point A is encircled by r in the CCW direction. we need to establish the concepts of encircled and enclosed. We can say that the region inside the path is encircled in the prescribed direction. In other words. 815(b) are enclosed. point A in Fig. 52 Definition of Encircled and Enclosed Because the Nyquist criterion is a graphical method. when the closed path r has a direction assigned to it. 814. which are used for the interpretation of the Nyquist plots for stability. • ClosedLoop Stability: A system is said to be closedloop stable. because it is outside the path. can be in the clockwise (CW) or the counterclockwise (CCW) direction. Exceptions to the above definitions are systems with poles or zeros intentionally placed at s = O. FrequencyDomain Analysis system.. the shaded regions in Figs. 8. if made. 815(a) is enclosed by r. point B and all the points in the shaded region outside r in Fig. Encircled A point or region in a complex function plane is said to be encircled by a closed path For example. However. and the region outside the path is not encircled. if it is found inside the path. Chapter 8. As shown in Fig.
854 Principles of the Argument The Nyquist criterion was originated as an engineering application of the wellknown "principle of the argument" concept in complexvariable theory. Suppose that a continuous closed path r s is arbitrarily chosen in the splane. and point B is encircled twice or 4n: radians. For example. infinity in the complex plane is interpreted as a point. Starting from a point S1> the f s locus is traversed in the arbitrarily chosen direction (CW in the illustrated (a) (b) Figure 816 Definition of the number of encirclements and enclosures. then the trajectory f.l mapped by A(s) into the Ll(s)plane is also a closed one. As defined in Chapter 7. . 817(a). (a) Point A is enclosed by r. as shown in Fig. in the complex Ll(s)plane. N is positive for CCW encirclement and negative for CW encirclement. In Fig. 817(b). 816(b). The total net number of revolutions traversed by this arrow is N. or the net angle is 2nN radians. a number N can be assigned to the number of times it is encircled. 816(a) is encircled once or 2n: radians by r . including infinity.in Fig. point A in Fig. there is one and only one corresponding po. Let Ll(S) be a singlevalued function of the form of the righthand side of Eq. for each point in the splane. and point B is enclosed twice by f. all in the CW direction. If f s does not go through any poles of Ll(S).int. Single valued means that. 853 Number of Encirclements and Enclosures When a point is encircled by a closed path f. as shown . The magnitude of N can be determined by drawing an arrow from the point to any arbitrary point SI on the closed path r and then letting SI follow the path in the prescribed direction until it returns to the starting point.85 Nyquist Stability Criterion : Fundamentals 429 (a) (b) Figure 815 Definition of enclosed points and regions . but B is enclosed by the locus r . By definition. The principle is stated in the following in a heuristic manner. which has a finite number of poles in the splane. (b) Point A is not enclosed. (837). point A is enclosed once.
rf. and S3. and finally return to the starting point. . S2. for a given value of K.~ in the a(s)plane. consider the function the same.) and go through points ~(S2) and . The reader should recognize the parallel of this situation to the rootlocus diagram that essentially represents the mapping of a(s) = 1 + jO onto the loci of roots ofthe characteristic equation in the . the reverse il(. (840) as K s(s + l)(s+2) . The principle of the argument can be stated: Let ~(s) be a singlevalued function that has afillite llumber of poles in the splane. Ll(SI)' The direction of traverse of r A can be either CW or CCW. FrequencyDomain Analysis jOJ . The simplest way to illustrate this is to write Eq. and then returning to S I after going through all the points on the r~ locus. the root loci of Eq.d(S3). Thus. there is only one corresponding point in the . which represents a point on the real axis in the ~(s)plane. and . However. the function maps into three corresponding points in the splane. respectively. (b) Corresponding locus r.($)plane splane a Rea (a) (b) Figure 817 (a) Arbitrarily chosen closed path in the splane. the direction of a is arbitrarily assigned. (841) gives three roots in the .~) with L(s). r K A(s) = s(s + l)(s + 2) (840) which has poles s = 0. in the same direction or the opposite direction as that of depending on the function Ll(s). through the points S2 and S3.I 7(b).d(s)~plane. (840) have three individual branches in the splane. the thirdorder equation in Eq.~·plane. In Fig. for each point in the . 8. • Do not attempt to relate Although the mapping from the splane to the ~(s)plane is singlevalued. For each point in the splane. corresponding to S].d(s)pIane. as shown in Fig. The corresponding r a locus will start from the point ~(s.2 in the splane.6.'1plane. the corresponding A locus mapped ill the ~(s )plane will encircle the origin as many times as the difference between the number o/zeros and poles of ~(s) that are encircled by the spume locus fso r r . 817(a).A(s) =0 (841) If ~(s) is a real constant. case). They are not process is not a singlevalued mapping. that is. Suppose that an arbitrary closed path s is chosen in the splane so that the path does not go through anyone ofthe poles or zeros of.1. for illustration purposes~ to be CCW.430 Chapter S.6(s). For example.
N > 0(2 > P). N can be positive (Z> P). Fig. If the splane locus encircles more poles than zeros of A(s) in a certain direction. the principle of the argument is stated as N=ZP where N Z P (842) = number of encirclements of the origin made by the A(s)plane locus f ~. In general. 2. the number of net intersections of this line with the A(S) locus gives the magnitude of N. ReA Figure 818 Examples of the determination of N in the 8(s)plane. N is a positive integer.85 NVquist Stability Criterion: Fundamentals 431 In equation form. If the splane locus encircles more zeros than poles of ~(s) in a certain prescribed direction (CW or CCW). N = O(Z = P).\'plane locus r. or negative (Z < P). If the splane locus encircles as many poles as zeros.~ in the splane. 818 gives A(s)plane A(s)plane N=O ReA ReA jlmA A(s)plane N=3 jIm A A(s)plane N=O N=O Ret. These three situations are described in more detail as follows. of ~(s). zero (Z = P). the ~(s)plane locus f ~ will encircle the origin N times in the opposite direction as that of fso A convenient way of determining N with respect to the origin (or any point) of the a(s)· plane is to draw a line from the point in any direction to a point as far as necessary. = = number of zeros of ~(s) encircled by the splane locus r s in the splane. 3. the A(s)plane locus f ~ will encircle the origin of the ~(s)plane N times in the same direction as that of f". the ~(s)plane locus f ~ will not encircle the origin of the A(s)plane. .\. number of poles of Ll(s) encircled by the . or no poles and zeros. N is a negative integer. In this case. l. In this case. N < O(Z < P).
dependent on the way the Nyquist criterion is applied.L(s + pIJ. rat r . A rigorous proof of the principle of the argument is not given here. In these illustrated cases. Chapter 8. with the arbitrary does not pass through any of the poles and the zeros of A(s).1(s) in Eq. (b) . Thus. Let us consider the function A(s) is of the fonn As _ K(s+zt} ( ) .. The function A(s) can be written as A(s) = IA(s)JLA(s) = Is+Kls ~s+ P2 I[L(s +zIJ. FrequencyDoma in Analysis several examples of this method of detennining N. respectively. 819(a) shows an arbitrarily chosen trajectory in the splane. we shall designate other points in the complexfunction plane as critical points.6.432 . Critical Point For convenience. we shall designate the origin of the A(s)plane as the critical point from which the value of N is determined. Similar vectors can be drawn for (SI + PI) and (s + P2).(s+ pt)(s+ P2) (843) where K is a positive real number. 819(a). of .1(s)plane locus which corresponds to the s locus of (a) through the mapping of Eq. The poles and zeros of A(s) are assumed to be as shown in Fig. and The function . The following illustrative example may be considered a heuristic explanation of the principle. (844). it is assumed that the r s locus has a CCW sense.(s) evaluated at s = 81 is (845) • Z and P refer to only the zeros and poles.L(s + P2)] zd PI rs rs (844) Fig.1(s) that are encircled by C~ The term (SI +zr) can be represented graphically by the vector drawn from Zl to Sr. Later. point SI on the path. (844) and the splane trajectory f s . A(st} is represented by L1(s)plane C1 ReA (a) (b) Figure 819 (a) Polezero configuration of .
6o(s).. Z = 1 and P = o. 8 55 Nyquist Path a Years ago when Nyquist was faced with solving the stability problem. of ~(s) that are encircled by r s and not the total number of zeros and poles of A(s). which means that the corresponding ~(s) plot must go around the origin 2it radians. A summary of all the possible outcomes of the principle of the argument is given in Table 8. if the point Sl is moved along the locus r s in the prescribed CCW direction until it returns to the starting point. or one revolution. Thus. as shown in Fig. Now. N in Eq. and the A(s)plane locus must encircle the origin N times in the opposite direction to that of r s. as the splane locus is traversed once in any direction.P) = 21rN radians (847) This equation implies that if there are N more zeros than poles of . whereas the vector (St + Zl) drawn from the zero at Zl.d(s) that are inside the r s trajectory in the sMplane will contribute to the value of N ofEq. 819(a). in a prescribed direction. respectively. Because the poles of A(s) contribute to a negative phase. as shown in Fig. the net angle traversed by the A(s)plane locus. In general. which are encircled by the splane locus r s. the value of N depends only on the difference between Z and P. the A(s)plane locus will encircle the origin N times in the same direction as that of r s. the angles generated by the vectors drawn from the two poles that are not encircled by r s when Sl completes one roundtrip are zero. in the CCW direction. is equal to r 21Z'(Z .6o(s) = 1 + L(s) has zeros in the righthalf splane. (847) will be negative. which is encircled by generates a positive angle (CCW) of21r radians. and zeros contribute to a positive phase. 819 (a). 819(b). For the case illustrated in Fig. which involves determining if the function . if N more poles than zeros are encircled by fs in a given direction. (842).85 Nyquist Stability Criterion: Fundamentals c 433 TABLE 81 Direction of N=ZP Encirclement Summary Df All Possible Outcomes Df tbe Principle of the Argument A(s)Plane Locus Sense of the Locus CW CCW CW CCW s~plane Number of Encirclements of the Origin Direction of Encirclement CW CCW CCW CW N>O N<O N=O CW CCW N N N N 0 0 No encirclement No encirclement the vectors drawn from the finite poles and zeros of ~(s) to the point s. Conversely. he . N=ZP=l (846) which means that the ~(s)plane locus a should encircle the origin once in the same direction as that of the s~plane locus r s. It should be kept in mind that Z and P refer only to the zeros and poles. This is why only the poles and zeros of .1. rs.
1. 820. the stability of a closedloop system can be determined by plotting the A(s) = 1 + L(s) locus when s takes on values along the Nyquist path and investigating the behavior of the A(s) plot with respect to the critical point. and the same conclusion on the stability of the closedloop system call be obtained by observing the behavior of the L(s) plot with respect to the (1. 820 are used to indicate that righthalf . jO) point in the L(s)plane. It is apparent that. Fig. For singleloop systems.~ for the Nyquist criterion. This path is chosen to be the splane trajectory r.I·pla ne. Because the function L(s) is generally known. )0) point of the G(s)H(s)plane. once the Nyquist path is specified. the Nyquist stability . jO) point in the L(s)plane. FrequencyDomain Analysis jw . Thus. if any pole or zero of A(s) lies inside the righthalf splane. the previous development leads to the detennination of the closedloop stability by investigating the behavior of the G(s)H(s) plot with respect to the (. the small semicircles shown along thejwaxis in Fig. In principle. 856 Nyquist Criterion and the L(5) or the G(5)H(S) Plot The Nyquist criterion is a direct application of the principle of the argument when the . jO ) point in the L(s)plane becomes the critical point for the determination of closedloop stability. JOO splane Poles of ~(s ) joo Figure 820 Nyquist path. which in this case is the origin of the A(s)plane. the path should go around these poles and zeros if they fall on the jwaxis. as an alternative. apparently discovered that the principle of the argument could be applied to solve the stability problem if the splane locus f s is taken to be one that encircles the entire right half of the splane. sinee in mathematics. as the solution is a relative one. it would be simpler to construct the L(s) plot that con'espollds to the Nyquist path. L(s) = G(s)H(s). it will be encircled by the Nyquist path f s . This is because the origin ofthe A(s) = 1 + L(s) plane corresponds to the (. The path fs shown in Fig. Because the Nyquist path must not pass through any poles and zeros of • The Nyqu ist path is detined to encircle the e nt ire A(s) . 820 is defined to be the Nyquist path. Of course. f s can be chosen to encircle the entire lefthalf splane. 820 illustrates a fs locus with a CCW sense that encircles the entire right half ofthe splane.1.I'plane locus is the Nyquist path of Fig. CCW is traditionalJy defined to be the positive sense. Thus the (1.434 Chapter 8.
excluding the poles at s = 0. when s = jev and as ill varies from 00 to 0. if any. 86 NYQUIST CRITERION FOR SYSTEMS WITH MINIMUM~PHASE TRANSFER FUNCTIONS We shall first apply the Nyquist criterion to systems with L(s) that are minimumphase transfer functions. .jO)pointas many times as the number ofpoles of L(s) that are in the righthalf splane. Thus. number of poles of 1 + L(s) that are inside the Nyquist path. number of zeros of 1 + L(s) that are inside the Nyquist path. A minimumphase transfer function does not have poles or zeros in the righthalf splane or on the jillaxis. Z must equal zero. For a minimumphase transfer function L(s) with m zeros and n poles. and the encirclement. For openloop stability. The stability requirements for the two types of stability defined earlier are interpreted in terms of Z and P. the application of the Nyquist criterion to the stability problem involves the following steps. 2. 2. that is. jO) point made by the L(s) plot. The value of N. P must equal zero. jO) point made by the L(s) plot. The Nyquist path rs is defined in the splane. excluding the origin. 1. given a control system that has the characteristic equation given by equating the numerator polynomial of 1 + L(s) to zero. The properties of the minimumphase transfer functions are described in Chapter 2 and are summarized as follows: 1.. must be made in the clockwise direction (ifr. Notice that the poles of 1 + L(s) are the same as that of P = L(s). the righthalf splane. that is. the condition of stability according to the Nyquist criterion is stated as N=P (849) That is. the total phase variation of L(jw) is (n . 4.m)1r/2 radians. The Nyquist criterion follows from Eq. Thus. is observed. (842). as shown in Fig. the L(s)plot must encircle the (l. N=ZP where (848) N Z = = number of encirclements of the (1. fora closedloop system to be stable. 3. the righthalf splane. For closedloop stability.8·6 Nyquist Criterion for Systems with Minimum~Phase Transfer Functions 435 criterion is another example of using the loop transfer function properties to find the behavior of closedloop systems. The L(s) plot corresponding to the Nyquist path is constructed in the L(s)plane. where L(s) is the loop transfer function. 820. the number of encirclement of the (1. is defined in the CCW sense)• .
Nyquist when the trajectory is traversed in the prescribed direction. points on the by plotting the segment of positive jwQxis. that is. if the system is unstable. the system is unstable. jeouxis. not how many times. (850) would be a positive integer. and the poles of . As if turns out. jO) point is enclosed by the Nyquist plot. Because a majority of the loop transfer functions encountered in the real world satisfy • A minimumphase condition 1 and are of the minimumphase type. we shall define the L(jw) plot that corresponds to the positivejevaxis of the splane as the Nyquist plot of L(s). the Nyquist criterion for a system with L(s) being a minimumphase transfer function is simplified to = = N=O (850) Thus. A nonminimumphase transfer function will always have a more positive phase shift as w varies from 00 to O. From this point on. the Nyquist criterion can be stated: For a closedloop system with loop transfer function L(s) that is of minimumphase type. Thus.436 . it would be prudent to investigate the transfer function does not have poles or zeros in the application of the Nyquist criterion to this class of systems. if it is. jO) point. which means that the critical point ( 1 ~ jO) is enclosed N times (corresponding to the direction of the Nyquist path defined here). Chapter 8. in practice.e (1. except at s = O. FrequencyDomain Analysis 3. the Nyquist criterion of stability for systems with minimumphase loop transfer functions can be further simplified: For a closedloop system with loop transfer function L(s) that is of minimumphase type. it is often necessary in design to create an equivalent loop transfer function Leq(s) so that a variable parameter K will appear as a multiplying factor in Leq(s). Thus. • For L(s) that is minimum. Because a minimumphase L(s) does not have any poles or zeros in the righthalf splane or on thejevaxis (except at s = 0) P 0.Because the region that is enclosed by a trajectory is defined as the region that lies to the left pha'ie type. the system is closedloop stable if the L(s) plot that corresponds to the Nyquist path does 1Iot ellclose ti. or. this information is not vital. 861 Application of the Nyquist Criterion to MinimumPhase Tranter Functions That Are Not Strictly Proper Just as in the case of the root locus. L(s) = KLeq(s). Thus.. The value of a minimumphase transfer function cannot become zero or infinity at any finite nonzero frequency. the Nyquist criterion can be criterion can be checked checked simply by plotting the segment of L(j(J)) from (J) = 00 to 0. If the (1. if the system is found to be unstable. The only drawback to this method is that the Nyquist plot that corresponds to the jwaxis tells only whether the critical point is enclosed or not and. This simplifies the procedure considerably. easily on a computer. the system is closedloop stable if the plot of L(9) that corresponds to the Nyquist path does not encircle the critical point (1. the enclosure property does not give infonnation on how many roots of the characteristic equation are in the righthalf splane. it may not have more . Furthermore. N in Eq. However.d(s) 1 + L(s) also have the same properties. since the plot can be made LUw) from (J) = 00 to O. Because the equivalent loop transfer function does not correspond to any physical entity. equally true. Z # 0. this is quite righthalf splane or on the simple. 4. Or. jO) in the L(s )plane. it will always have a more negative phase shift as (J) varies from 0 to 00.
when L(s) has poles orland zeros in the righthalf splane. ±2.. . (851) as 1+ 1 KLeq(s) =0 (852) by dividing both sides of the equation by KLetj(s). 87 RELATION BETWEEN THE ROOT LOCI AND THE NYQUIST PLOT Because both the root locus analysis and the Nyquist criterion deal with the location of the roots of the characteristic equation of a linear SISO system. The variable parameter on the Nyquist plot is now 1/K. and the Nyquist criterion can be applied for stability studies without any complications. and. some computer programs may not be prepared for handling improper transfer functions. for the RL K ::. and the critical point is still (1. Exploring the relationship between the two methods will enhance the understanding of both methods. the mapping points are on the positive real axis of the L(s)plane. for the RL K 2:: 0. and it may be necessary to reformulate the equation for compatibility with the computer program. the Nyquist criterion can still be applied. 822 as a = . It was pointed out earlier that the mapping from the splane to the function plane for a rational function is single valued. but the reverse process is multivalued. The root loci for the same system are shown in Fig. 0. 821. However. . and the transfer function is not strictly proper. with this minor adjustment. the Nyquist plot of a typel thirdorder transfer function G(s)H(s) that corresponds to points on the jwaxis of the splane is shown in Fig. To examine this case. The Nyquist criterion presented here is cumbersome when the loop transfer function is of the nonminimumphase type. . ± I. (853) must satisfy the conditions LKG1 (S)Hl (s) = (2j + l)n K ~0 (854) 1KG1 (S)H1 (s) = 2jrr K::. Now we can plot the Nyquist plot of 1/Leq(s). Because the root loci of Eq. the two analyses are closely related. there is no difficulty in constructing the Nyquist plot of a transfer function that is not strictly proper. as defined in Chapter 2. jO) for K > O. A generalized Nyquist criterion that will take care of transfer functions of all types is presented in Appendix F.87 Relation between the Root Loci and the Nyquist Plot ~ 437 poles than zeros. for example. we can rewrite Eq. . In principle. Thus. Given the characteristic equation 1 +L{s) = 1 + KG1 (s)H] (8) = 0 (853) the Nyquist plot of L(s) in the L(s)plane is the mapping of the Nyquist path in the splane. As a simple illustration. the mapping points are on the negative real axis of the L(s)plane. In fact.KLeq(s) =0 (851) If Lcq(s) does not have more poles than zeros. • the root loci simply represent a mapping of the real axis of the L(s)plane or the G(s)H(s)plane onto the splane. consider that the characteristic equation of a system with a variable parameter K is conditioned to 1 I. ° (855) for j 0.
jO) point of the G(s)H(s)plane corresponds to the two points where the root loci intersect the jwaxis and a point on the real axis. 823 illustrates the G(8)H(s) plots that correspond to different constantdampingratio lines in the splane. and the corresponding trajectory in the splane passes through the root of the characteristic equation. As shown by curve (3) in Fig. when the G(s)H(s) curve passes through the ( 1. it would be useful to consider the mapping of points other than those on the jwaxis of the splane and on the real axis of the G(s)H(s)plane. Fig. For instance~ we may use the mapping of the constantdampingratio lines in the splane onto the G(s)H(s)plane for the purpose of determining relative stability of the closedloop system. Similarly. we can construct the root loci that correspond to the straight lines . it means that Eq. each point of the G(s)H(s)plane corresponds to three points in the splane. Chapter 8. 823. jlmGH G(s)H(s)pIane Figure 8·22 Root. jO) point. The (1.. (852) is satisfied.438 . in this case. Note that. mapping of the real axis of the G(s)H(s)plane onto the splane. The Nyquist plot and the root loci each represent the mapping of only a very limited portion of one domain to the other.. FreQuency~Domain Analysis JlmGH s~plane jllJ G(s)H(s)plane _<IIt___ 1 jOJI o C1 Figure 8·21 Polar plot of G(s)H (s) = K J[s(s + a) (s + b)] interpreted as a mapping of the jwaxis of the splane onto the G(s)H(s)~plane. In general.locus diagram of G{s)H(s) ::::: KJ[s(s + a)(s + b)] interpreted as a mapping of the real axis of the G(s)H(s)plane onto the splane.
=0 (857) jlmGH G(s)H(s)~plane (I) ReGH Figure 824 Root loci that correspond to different phaseangle loci in the G(s)H(s)plane. 824.8 7 Relation between the Root Loci and the Nyquist Plot 4 439 w jlmGH G(s)H(s)plane jOJ ReGH Figure 823 G(s)H(s) plots that correspond to constantdampingratio lines in the splane. 824 must satisfy the equation 1 + G(s)H(s)ej(J for the various values of e indicated. as shown in Fig. Notice that these root loci now satisfy the condition of LKGl (s)HJ (s) = (2j+ l)n(.} K~O (856) Or the root loci of Fig. . rotated at various angles from the real axis in the G(s)H(s)plane.
information on the intersection on the real axis and the properties of L(jw)/K at w = 00 and w = 0 would allow the sketching of the Nyquist plot without actual plotting. to O. 1.v(s + 2)(s + 10) for w = 00 to w = O. + 2)(jw + to) 1. traversed in the direction that corresponds to £t) = 00 to 0 on the Nyquist path. because we are interested only in whether the critical point is enclosed. jO) point. Frequency·Oomain Analysis 88 ILLUSTRATIVE EXAMPLES: NYQUIST CRITERION FOR MINIMUMPHASE TRANSFER FUNCTIONS The following examples serve to illustrate the application of the Nyquist criterion to systems with minimumphase loop transfer functions. However.. 8·25 shows the Nyquist plot of LUw)/K for w = .x. it is not necessary ~o produce an accurate Nyquist plot. L( jO)/ K = 00/ .90 0 jbnL L(jm)plane m=oo o ReL Figure 825 Nyquist plot of 1 o a L(s)jK = . we get the zerofrequency property of L(jw). (8·58). In many cases. The Nyquist plot of LUw)/K may be plotted using freqtooL Fig. I . jw{jw (859) Substituting w = 0 in the last equation.440 Chapter 8. we get L(jw)/ K 2. We can use the following steps to obtain a sketch of the Nyquist plot of L(jw)/K. EXAMPLE 881 Consider that a singleloop feedback control system has the loop transfer function L(s) := G(s)H(s) = s(s + 2)(s + 10) K (8·58) which is of minimum·phase type. Because the area that is enclosed by the Nyquist plot is to the left of the curve. Substitute s = jw in L(s). All examples in this chapter may also be solved using the ACSYS (see Chapter 9) or MATLAB Toolboxes incorporated in this chapter. all that is necessary to determine stability is to find the point or points at which the Nyquist plot crosses the real axis in the L(jw)!Kplane. in general. The stability of the c1osed·loop system can be conducted by investigating whether the Nyquist plot of LUw) / K for w = DC to 0 encloses the (1. Setting s = jw in Eq. =.
The result is (863) The solutions of the last equation are w = and w 00. the intersect would be to the left of the . which is known to be a solution at L(jm)/ K . 59) becomes L(jw)/K = . [1. the correct answer is U) = J20 rad/sec. Eq. if K is less than 240. In this case. When K is negative.den.. under this condition.is. and the system is stable. = oc in Eq. we set the imaginary part of L(jw)/ K to zero. den = conv(conv( [110] . (8 . we rationalize L(jw)/ K by multiplying the numerator and the denominator of the equation by the complex conjugate of the denominator. (859). Thus. If the gain is increased to a value beyond 240. 825 w=O.:.w] =nyquist(num. To find the possible intersects on the real ax.01 0. Substituting CtJ established. nurn = [1] .. the Nyquist plot of L(jw) would intersect the real axis at the I point. and the system would be unstable. Substituting this frequency into Eq. the +1 point on the real axis would be enclosed for all negative .w). we see that. we have the intersect on the real axis of the L(jw)plane at L( j. (862). 825. the intersect of the L(jw) locus on the real axis would be to the right of the critical point (1.irn.(V2)) 5. plot(re. and the system would be marginally stable.s( [O. Fig.2]) [1 0]) . rad/sec (864) = ±v'2D Because w is positive..1:0. the characteristic equation would have two roots on the jcvaxis in the splane at s = ±JJ20..60.J. Ere . ax1. 825 shows that.al)J = CtJ[144cv2 + (20 . O.. these results are verified by the plot shown in Fig. [12w2 [12m2 joo(20  + jcv{20  ( 2 )Jt12m2 . Nyquist Criterion for MinimumPhase Transfer Functions • 441 Toolbox 881 MATLAB statements for Fig.j(20 .1:1000.20)/K =  2~~O = 0.01]) 1 grid 3. if any.004167 The last five steps should lead to an adequate sketch of the Nyquist plot of L(jw) / K short of plotting it. jO)..im). we can use the ( +1~ jO) point in the L(jw)~plal1e as the critical point. 4.88 lIIustrative Examples. the latter is not enclosed. If K = 240. the property of the Nyquist plot at infinite frequency is L(joo)JK =OL  270 Q Apparently. Thus.I point on the real axis. To find the intersed(s) of the Nyquist plot with the real axis. 0.jm(20  ' oo2)J cv2)] [12w .
. and the system would always be unstable. ~ EXAMPLE 882 Consider the characteristic equation Ks:' + (2K + 1)s2 + (2K + 5}s + 1 = Ks(il. [110J).0 Figure 8~26 RL of L(s) = s{s + 2)(s + lOY K values of K. (858). rlocus(mysys).442 . Frequency~Domain Analysis splane Kmm= 0..00 Kmax = 500. Thus. Note that application of the RouthHurwitz stability criterion leads to this same result. Fig. mysys=tf(.0 K=O 10. 826 shows the root loci of the characteristic equation of the system described by the loop transfer function in Eq. .00 I~~~~~~~~~~~~~a ooK 14. Chapter 8. The correlation between the Nyquist criterion and the root loci is easily observed.0001.+ 2s + 2) 0 Dividing both sides of the last equation by the terms that do not contain K. title(' Root loci of the system' ) . 8~26 den=conv( [1 2 0] . we have 1 + KLeq(s) == 1 + s 2 + 5s+ 1 = 0 (868) Thus.den). the Nyquist criterion leads to the conclusion that the system is stable in the range of 0 < K < 240.0 2. Toolbox 882 MATlAB statements for Fig.
the Nyquist plot of Leq(j{J))/K is sketched as shown in Fig. (871) to zero. (872) are imaginary. Notice that this plot is sketched without any detailed data computed on Leq ( j(})) / K and. axi s ( [ . Setting s = jw in Eq. the sketch is adequate to determine the stability of the system.(1w2) + 5joo From the last equation. we obtain the two end points of the Nyquist plot: Leq(jO) = OL900 and Leq(joo) = ooL90° (870) Rationalizing Eq.t)plane Area enclosed 1 o 1 ReL Figure 827 Nyquist plot of Le~S) = S(32 +28+2) 2 5 1 S + s+ for w = 00 to w = O.1:1000. jImL L(j(. 827.( 2 )(1 . grid We can show that all the four roots of Eq. in fact. (868). We get (V = 0 and (Ji + 7002 + 2 = 0 (872) Toolbox 883 MATLAB statements for Fig. could be grossly inaccurate.oo2)2+ 25w2 (871) To find the possible intersects of the Leq( jw) / K plot on the real axis.1:O. 827 w=O. However. We can obtain the information to manually sketch the Nyquist plot of L~q(s) to determine the stability of the system. . plot(re. (870) and the fact that there are no other intersections on the real axis than at lJ) = 0. (8~69) by multiplying its numerator and denominator by the complex conjugate of the denominator.( 2 )1 K. Using the information given by Eq. we get Leq(jw) w[2w+ j(2 .88 Illustrative Examples: Nyquist Criterion for MinimumPhase Transfer Functions ~ 443 which is an improper function.im). num = [1 2 2 0] . den = [1 5 1].2 1 1 5] ) .w] =nyquist(num. [re.= ai[5(2  w2 )  2(1  2 00 )] + jW[10w2 + (2 . we get Leq{jOJ) K. which indicates that the Leq( j{J)) / K locus intersects the real axis only at (J) = O.im.( 2 )] (1 .w).den. we set the imaginary part of Eq.
827 does not enclose the ( 1. Because the RL stays in the lefthalf splane for all positive values of K. (867) for K > 0.(J)2)] (873) for w = 00 to O. which agrees with the conclusion obtained with the Nyquist criterion. + 5s+ 1 Because the Nyquist plot in Fig.= ') 5 lorw = K s~ + 8 + 1 w=O. 828 shows the Nyquist plot ofEq.1.444 . FrequencyDomain Analysis K L (j{J)) . Chapter 8. . and the results confirm the Nyquist criterion results on system stability. the system is stable for all finite positive values of K. the system is stable for 0 < K < 00.plane ~+~+ReK L Area enclosed Figure 8·28 Nyquist plot of K/Lt'q(jw) e o J ~ Leq(s) s(s2 + 2s + 2) ~ l o r . 9 EFFECTS OF ADDING POLES AND ZEROS TO L(5) ON THE SHAPE OF THE NYQUIST PLOT Because the performance of a control system is often affected by adding and moving poles and zeros of the loop transfer function. it is important to investigate how the Nyquist plot is affected when poles and zeros are added to L(s).. jO) point.. jru 00 to splane OK Figure 829 RL of L(s) == Ks(s2 s2 + 28 + 2). Notice that the RL stays in the lefthalf splane for all positive values of K. ~ 8. (868). based on the poles and zeros of Leq(s)/ Kin Eq. and the system is again stable for all positive values of K by interpreting the Nyquist plot of K/Leq(jw). The plot again does not enclose the ( . 829 shows the RL of Eq.. Fig. Fig.( 2 ) + 5 jw Leq( jw) . (866). (868). using the polezero configuration of Leq(s) of Eq.[2cv2 + jW(2 . K _ (1 . jO) point as (J) varies from 00 to 0.
In addition.) 0)=00 o C ritical point Critical point ReL oo> K>O t no\ enclosed) Closedloop stable K<O (enclosed) Closedloop unstable :3 Enclosed area o I Nyquist plOL of K L(f)= .table L(s) K Let us begin with a firstorder transfer func tion L (~) . P = 0. then L(~) K . the phase of L(jw) is reduced by 90" at both zero and infinite frequencies. (874). (875) and the closedloop stability interpretations with respect to the critical points for 00 < K < 00. In j Im L LUOJ)plane Pw= I . 830. Addition of Poles at s = 0 Consider that a pole at 05 = 0 is added to the transfer function of Eq. the magnitude of L(jw) at w = 0 becomes infinite . as shown in Fig.. Fig.s( \ + Tp ) Figure 831 .05(1 + TI S) (875) Because adding a pole at s = 0 is equivalent to di. WI\ =90 0 for stability (Refer fO Appen. The figure also shows the interpretation of the closedloop stability with respect to the critical point for all values of K between 00 and 00 . The Nyquist plot of L(jw) for 0 :S w :S 00 is a semicircle. . P = 0. (Refer co Appe/ldix F. K =1 + TIS (874) where T\ is a positive real constant.1 (not enclosed) Closed loop stable Figure 830 Nyquist plot of Closedh>op . 4>11 =' 0" for stability.89 Effects of Adding Poles and Zeros to L(s) on the Shape of the NVquist Plot j 1m L 445 L(jOJ)plane Pw= 0.dix F.viding L(s) by jw. 831 illustrates the Nyquist plot of L(jw) in Eq.) OJ=oo I Critical point 00 > K> 0 (not enclosed) o Re L Critical point 1 >K enclosed Closedloop unstabl e Critieal point 0> K > .
. (874).446 110 Chapter 8. adding a pole of multiplicity p at 5 = 0 to the transfer function of Eg. (874) will give the following properties to the Nyquist plot of L(jw): w>oo 1im LL(jw) = (p w>O + 1)90° (876) (877) lim LL(Jw) = .\' = 0 to a loop transfer function will rcduce stability of the closedloop system. A system that has a loop transfer function with more than one pole at s = 0 (type 2 or higher) is likely to be unstable or difficult to stabilize. FrequencyDomain Analysis general. 832 shows the Nyquist plot of L(s) = s2(l + h~) K K (880) and the critical points. with stability interpretations.P=O cP) ) = 180° for stability (Ref er to Appendix F.25 ( +. 833 illustrates tbe same for L( s) • Adding poles at .+ T "'7') (882) .( +==)s r :) '1 (881) The conclusion from these illustrations is that the addition of poles at s = 0 to a loop transfer function will affect the stability of the closedloop syslt:rn auyt:rst:!y.) Oill (j)=oo I o Critical point ReL Enclosed area Critical poi at oo>K>O (enclosed) Closedl oop unstable K<O (enclosed) Closedloop unstable Figure 832 Nyquist plot of L(s) = 2( K T )' s 1 + IS Addition of Finite Nonzero Poles When a pole at s = 1 / T2(T2 > 0) is added to the function L(s) of Eg . )('. jlm L LUoJ)plane P w =2. = s ~3. Fig.p x 90° w> oc lim IL(jw)I = 0 ClI>O (878) (879) lim IL(jw) I = oc The following example illustrates the effects of adding multipleorder poles to L(5). EXAMPLE 891 Fig. we have K L( s) = :1 r s:: 1.
Curve (1): K L(s) = ( \ + T.::::=7" + T]s )" The Nyquist plot of L(jw) at w = 0 is not affected by the addition of the pole.) I Critical point o w=oo Critical point Re L OO> X>O (enclosed) ClosedIo()p unstable K<O (enclosed) Closedloop unstable Figure 833 Nyquist plot of L( s) = K s3(I ". The system represented by Eg. . stability when poles are added to the loop transfer function . the effect of adding a pole at s = . as shown in Fig.1/ T2 to the transfer function of Eq. j Jm L L(j(tJ)plane ReL Figure 834 Nyquist plots.K 2 OL. 834. (875) is to shift the phase of the Nyquist plot by 900 at w = 00.s)(J + T2S )' Curve (2): (2) L(.180° (884) Thus. the Nyquist plot at w = 00 is rotated clockwise by another the loop transfer function also reduces stability of the 90° from that of Eq. (882).89 Effects of Adding Poles and Zeros to L(sl on the Shape of the Nyquist Plot jlmL L(jw)plane 447 Pm= 3. The closedloop systems with the loop transfer functions of Eqs.f) = ( \ + hI') K (I + T2S)( 1 + J:~s). > 0) . The figure also shows the Nyquist plot of L(s) = (1 + h~)( l + T2S) (1 + T3S ) K (885) • Adding nonzero poles to where two nonzero poles have been added to [he rransfer function of Eq.P=0 <1>11=270° for stability (Refer to Appendix F. jO) point when K is positive. (885) is unstable if the intersect of the Nyquist plot 011 the negative real axis is to the left of the (1. (874) (T] . since lim L(jw) W ' O =K (883) The value of L(jw) at w = 00 is u)>OC lim L(jw) = lim W > OO T] 2W . These examples show the adverse effects Ott closed· loop dosed loop system . (874) and (882) are all stable as long as K is positive. T2 ) T3. In this case.
\' 1 + T. [llJ) [re. im. this stabilization effect is easi ly demonstrated. s I + T2S ) K( I + T<is) L(s) = . (886) .448 Cha pter 8. 834 w=0:0. (886) at w = 00 w hile not jlmL L(j (o)pl a\\e ReL Figure 835 Nyquist plots.\' (1 + T )( 1 + T2S ). wJ = nyquist (num. [1. den = conv([ll J . Td < T.1IT" to a loop transfer function.s)(l + h5) (888) The Nyquist plots of the two transfer functions of Eqs. The following example illustrates the effect on stability of adding a zero at .:100.J. num = [1 J .s .TJ11. w) . 'r'). im. 'b' ) .. EXAMPLE 892 Consider that the loop transfer function of a closedloop control system is K L(s) = s( 1 + Tp)(l + T2S) (886) It can be shown that the closedloop system is stable for TJ + T.wJ =nyquist( num. h K . Suppose that a l. since the multiplication of the term (1 + Trls) to the loop transfer function increases the phase of L(s) by 90° at Ct) = 00 .den. then. FrequencyDoma in Analysis Toolbox 891 MATLAB statements for Fig. (887) > 0) is added to the transfer function of Eq. axis ( [12 11]) grid Addition of Zeros It was demonstrated in Chapter 5 that adding zeros to the loop transfer function has the effect of reducing the overshoot and the general effect of stabilization.loop system . im. (886) and (888) are shown in Fig. den. to the loop transfer fUllction has the e ffect of stabilizing the closed. plot (re.w).Ol. hold on den = conv(conv([l1. O< K < . Ctllve (1): L(s) = ( )( . plot(re . [11]) [ re. 835. L(s) = K(l + Td S ) 05( 1 + T. (888) is to add 90° to the phase of the L(jw) in Eq. i m.1J). The effect of the zero in Eg. Curve (2): .ero at s = l / T(/ (Td • Addin g zero!. . In telms of the Nyquist criterion.
axis ( [1 2 11]) grid To demonstrate the concept of relative stability in the frequency domain.TdTI . we are interested not only in the absolute stability of a system but also how stable it is.Td(TJ + T2) (8 89)  1:. w] = nyquist (num I den I w) . im. relative stability is measured by paralllel~rs such as the maximum overshoot and the damping ratio. 836 for four different values of loop gain K.8~ 10 Relative Stability: Gain Margin and Phase Margin ~ 449 affecting the value at w == O. I r') . [11]) . In the time domain. irn r w] nyquist (num r den. and K. In general. hold on 1 num= [11]. the system with the loop transfer function in Eq. The four cases are evaluated as follows. (888) is stable for 0< K < which. The latter is often called relative stability. plot(re. den= conv(conv([11] [11]) [1 0]) [re. w) . jO) point. w] :::. the Nyquist plots and the corresponding step responses and frequency responses of a typical thirdorder system are shown in Fig. plot(re.TdTZ)/(TI + T2)' Thus.01:100. im. den = conv(conv( [11] . for positive Tl + T2 TIT2 . Comparisons and correlations between frequencydomain and timedomain parameters such as these are useful in the prediction of the performance of control systems. Another way of measuring relative stability in the frequency domain is by how close the Nyquist plot of L(jw) is to the (1. [11]) [re. has a higher upper bound than that of Eq. so that the enclosure of the (. • Relative stability is used to indicate how stable a system is. plot (re I im I I r ' ) . The intersect on the negative real axis of the L(jw)plane is moved from KII T2/(T J + T2) to K(TIT2 . the resonance peak Mp can be used to indicate relative stability. [1 0]) [re.1. . (887) . Toolbox 8101 MATlAB statements for Fig. num= [1]. = 1 1 axis ( [2 2 11]) grid hold on den = conv(conv( [11] . jO) point is sufficient for stability analysis. 835 w=0:0. im. nyquist (num den. In the frequency domain. 'b') . w) . It is assumed that the function L(jw) is of minimumphase type. im. • 810 RELATIVE STABILITY: GAIN MARGIN AND PHASE MARGIN We have demonstrated in Sections 82 through 84 the general relationship between the resonance peak Mp of the frequency response and the maximum overshoot of the time response. [11]).
the loop gain K is low: The Nyquist plot of L(jm) intersects the negative real axis at a point that is quite far to the right of the ( 1.... and frequency responses. Fig..._ _ _ _ _.. . Frequency~Domain Analysis jlmL L(jro)·plane Cdco y(t) IM(jlO)J o ReL 0 " ' . step responses... 8~36(a).. Oloc:::. The corresponding step response is quite well damped~ and the value of Mr of the frequency response is low..450 ::.. 0100::+ tjJ(jm) 270c (b) Stable but oscillatory system y(1) lM(jm) I 0 0 tjJ(jm) ~ w 270° (c) Marginally unstable system y(t) lM(jCd) [ Figure 8~36 Correlation among Nyquist plots. 270a tjJ(jm) (a) Stable and welldamped system y(/) [M(jw)! o IL... jO) point.Chapter 8... {) 1'1~'" 01+ tjJ (jm) (d) Unstable system 1.
let us first define the phase crossover on the Nyquist plot and the phasecrossover frequency. Notice that the negative slope of the phase curve becomes steeper as the relative stability decreases. PhaseCrossover Frequency: The phasecrossover frequency w p is the frequency at the phase crossover. Before giving the definition of gain margin. jO) point. Fig. is denoted as wI" and the magnitude of jImL 14~fJ'I o ReL Figure 837 Definition of the gain margin in the polar coordinates. The Nyquist plot of a loop transfer function L(jw) that is of minimumphase type is shown in Fig. and the system is marginally stable. . but the step response has a larger maximum overshoot. Fig. When the system is unstable. the phase curve rjJ( jw) of the closedloop frequency response also gives qualitative infonnation about stability. The step response becomes unbounded. the phase characteristics of the closedloop system are seldom used for analysis and design ptrrposes. the system is still stable. 3. In the frequency domain. and Ml' becomes infinite. B36(b). K is increased further: The Nyquist plot now passes through the (1. for the unstable system. 837. 4.810 Relative Stability: Gain Margin and Phase Margin 451 2. the slope beyond the resonant frequency becomes positive. • M. 836(d). 836(c). jO) point. jO) point. and Mr is also larger. The step response becomes oscillatory with constant amplitude. the value of Mr is still finite! In all the above analysis. Fig. In fact. and the system is unstable. ceases to have any meaning when the closedloop system is unstable. In practice. K is relatively very large: The Nyquist plot now encloses the ( I. The phasecrossoverfrequenc). or where (S90) • Gain margin is measured at the phase crossover. 8101 Gain Margin (GM) Gain Margin (GM) is one of the most frequently used criteria for measuring relative stability of control systems. The magnitude curve of IM(jw) Iversusw ceases to have any significance. K is increased: The intersect is moved closer to the (1. gain margin is used to indicate the closeness of the intersection of the negative real axis made by the Nyquist plot of L(jw) to the (1. Phase Crossover: A phasecrossover on the L(jw) plot is a point at which the plot intersects the negative real axis. • The definition of gain margin given here is for minimUlnphase loop transfer functions. jO) point. because the critical point is not enclosed.
452 :. and thus a negative gain margin may still correspond to a stable system. The L(jw) plot encloses (phase crossover is to the left of) the (1. we can draw the following conclusions about the gain margin of the system shown in Fig. which implies that the loop gain can no longer be increased. the gain margin is 0 dB. • When the Nyquist plot does not intersect the negative real axis at any finite nonzero frequency. jO) point. 0< ILUw p ) 1< 1 GM >OdB (893) 3. jO) point. the physical significance of gain margin can be of gain in dB that can be summarized as: added to the loop before the Gain margin is the amount of gain in decibels (dB) that can be added to the loop closedloop system becomes before the closed~loop system becomes unstable. 1. (894) 4. The L(jw) plot intersects the negative real axis between (phase crossover lies between) 0 and the 1 point.. jO) point.~ Chapter 8. the gain margin of the closedloop p) system that has L(s) as its loop transfer function is defined as = ILUcv I· gain margin = GM = 201og lO 1 (? )1 L JWI' = 20 log 10 (891) !LUcv p ) IdB On the basis of this definition. 837. FrequencyDomain Analysis L(jw) at w wp is designated as Then. Gain Margin of Nonminimum~Phase Systems Care must be taken when attempting to extend gain margin as a measure of relative stability to systems with nonminimumphase loop transfer functions. • When the phasecrossover is to the left of the (1. . • When the Nyquist plot of L(jw) passes through the (1. jO) point. unstable. depending on the properties of the Nyquist plot. and the loop gain must be reduced by the gain margin to achieve stability. as the system is at the margin of instability. The L(jw) plot does not intersect the negative real axis (no finite nonzero phase crossover). ILUw p ) I = 0 GM = 00 dB (892) 2. this means that. theoretically. The L(jw) plot passes through (phase crossover is at) the (1. jO). For such systems. a system may be stable even when the phasecrossover point is to the left of (1. the value of the loop gain can be increased to infinity before instability occurs. the phase margin is negative in dB. (895) • Gain margin is the amount Based on the foregoing discussions. the gain margin is infinite in dB.
indicates the effect on system stability due to changes in system parameter. than system A. . which jImL L(jw)·plane 1 o ReL Figure 838 Nyquist plots showing systems with the same gain margin but different degrees of relative stability. In contrast to the gain margin. we introduce the phase margin. the closeness of the phasecrossover to the (1. and the phase margin given here is for a margin is shown as the angle between the line that passes through the gain crossover and system with a minimumphase loop transfer function. 838 apparently have the same gain margin. locus B may easily be altered to enclose the (1. because with any change in the system parameters that affect the phase of L{jw). is the frequency of L(jw) at the gain crossover. gain margin indicates system stability with respect to the variation in loop gain only. we can show that the system B actually has a larger M. locus A actually corresponds to a more stable system than locus B.8~ 10 Relative Stability: Gain Margin and Phase Margin ~ 453 Nevertheless. phase margin jO) point. To include the effect of phase shift on stability. one would believe a system with a large gain margin should always be relatively more stable than one with a smaller gain margin. In principle. However. GainCrossover Frequency: The gain~crossover frequency. For instance. Furthermore. Wg. which requires that we first make the following definitions: Gain Crossover: The gain crossover is a point on the L(jw) plot at which the magnitude of L(jU) is equal to 1. jO) point. Unfortunately. As the name implies. 8~10~2 Phase Margin (PM) The gain margin is only a onedimensional representation of the relative stability of a closedloop system. which is determined by loop gain. • The definition of phase Fig. the origin. jO) point still gives an indication of relative stability.. the two systems represented by the L(jU) plots in Fig. 839 shows the Nyquist plot of a typical minimumphase L(jw) plot. Or where jL(jwg ) I = I (896) The definition of phase margin is stated as: Phase margin (PM) is defined as the angle in degrees through which the L(jw) plot must be rotated about the origin so that the gain crossover passes through the (1. gain margin alone is inadequate to indicate relative stability when system parameters other than the loop gain are subject to variation.
80dB p I The phase margin is measured at the gain crossover. the phase margin is obtained from Eq. the analytical expression of the phase at the gai n crossover. (891): t 1 GM = 20log lO JL(jw ) = 20 log 10 0. consider that the loop transfer function of a control system is L s _ 2500 ( ) . Phase margin is the amount of pure phase delay that can be added to the loop before the closedloop system becomes unstable.72°. Thus.s(s + 5)(s + 50) (898) The Nyquist plot of L(jw) is shown in Fig. (8~97) is no longer valid. system becomes unstable.88 rdd/sec The gain margin is measured at the phase crossover. The magnitude of L(jwp ) is 0.182. Care should be taken when interpreting the phase margin from the Nyquist plot of a • Phase margin is the nonminimumphase transfer function.72° . EXAMPLE 8101 As an illustrative example on gain and phase margins. The phase of L(jCIJg ) is 211. the gain margin is obtained from Eq. 840. and that can be added before the the definition of phase margin given in Eq. FrequencyDomain Analysis jlmL Figure 839 Phase margin defined in the L(jm)plane. . as seen from Fig.180° = 31. margin.22 rad/sec Phase crossoverwp = 15. 839. Thus. • Phase margin is measured When the system is of the minimumphase type.72° (8100) .182 = 14. the gain crossover can occur in any quadrant of the L(jw)plane. When the loop transfer function is of the nonamount of pure phase delay minimum~phase type. theoretically alter the phase of L(jw) by an equal amount at all frequencies. The following results are obtained from the Nyquist plot: Gain crossover Wg = 6. can be expressed as phase margin (PM) = LL(jwg )  1800 (897) where Wg is the gaincrossover frequency. (897): PM = lL(jwg ) 180° = 211.454 Chapter 8.
it would be beneficial to summarize advantages and disadvantages of the Nyquist plot. phase crossover. In the absence of a computer. 2. gain margin. Bode plots were often called the "asymptotic plots. Gain crossover. ·811 STABILITY ANALYSIS WITH THE BODE PLOT The Bode plot of a transfer function described in Chapter 2 is a very useful graphical tool for the analysis and design of linear control systems in the frequency domain. For design purposes.811 Stability Analysis with the Bode Plot jlmL 455 1 ReL Figure 8~4D Nyquist plot of L(s) = s(s + ~~~~+ 50)' Before embarking on the Bode plot technique of stability study. Before the inception of computers. Disadvantage of the Nyquist Plot 1. jO) point once the plot is made. Modern applications of the Bode plot for control systems should be identified with the following advantages and disadvantages: Advantages of the Bode Plot 1. It's not so easy to CatTY out the design of the controller by referring to the Nyquist plot. Advantages of the Nyquist Plot 1. The Nyquist plot can be used for the study of stability of systems with nonminimumphase transfer functions. The stability analysis of a closedloop system can be easily investigated by examining the Nyquist plot of the loop transfer function with reference to the ( 1. 3. a Bode diagram can be sketched by approximating the magnitude and phase with straightline segments." because the magnitude and phase curves can be sketched from their asymptotic properties without detailed plotting. . 2. and phase margin are more easily determined on the Bode plot than from the Nyquist plot. the effects of adding controllers and their parameters are more easily visualized on the Bode plot than on the Nyquist plot.
That is.180° at the gain crossover. The following results are observed easily from the magnitude and phase plots. FrequencyDomain Analysis 40 (DB' gain crossover 20 Unstable region for intersect on magnitude curve aI phase crossover.. there is no way of telling what the stability criterion is on the Bode plot. The phase margin is mea~ured at the gain crossover. 1. the gain margin is negative. That is. 842. positi ve ga in margin ~ co (rad/sec) :. respectively. the phase margin is negative.456 Chapter 8. phase crossover ! Figure 841 Determination of gain margin and phase margin on the Bode plot. If the gain margin is measured above the O axis. the phase margin is measured above the . the gain margin is measured below dB the OdB axis. The gaincrossover freq uency wfi is 6. For instance. EXAMPLE 8111 Consider the loop transfer function given in Eq.urve:: aI phase crossover. 841 for a typical minimumphase loop transfer function. the interpretation of these parameters from the Bode diagram is illustrated in Fig. The gain crossover is the point where the magnitude curve intersects the OdB axis. Disadvantage of the Bode Plot 1. and the system is unstable. The gain margin is positive and the system is stable if the magnitude of L(jw) at the phase crossover is negative in dB. (898). The phase margin is positive and the system is stable if the phase of L(jw) is greater than . 2.J11lullphase loop transfer functions. negative phase margin cop. positive ph~sc margin ~ ~ '' ~ t Unstable region for intersect on phase curve at phase crossover. and the system is unstable. If the phase margin is measured below the . Absolute and relative stability of only minimumphase systems can be determined from the Bode plot. negative gain margin CQ ~ 0 20 40 Stable region for intersect 011 magnitude curve at ph(lse crossover.:r '' 60 0 Phase margin ~ .22 rad/sec.90 180 270 360 Stable region for imersect on phas~ r.180°axis. The following observations can be made on system stability with respect to the properties of the Bode plot: • Bode plots nre llseful only for stability studies of systems with mini. 837 and 839.180°axis. The phase . the Bode plot of the function is drawn as shown in Fig. With reference to the definitions of gain margin and phase margin given in Figs.
22 10 15... 240 270 1 6..88 OJ (rad/sec) 100 .... 842 G = zpk( []...72 0 E '§" .... . 840 with the Bode plot of Fig. r.... 1M" v Phase crossover "r. 842.. ~ i ...8 dB.... and the interpretation of wg • wI" OM.... ~ ...... Toolbox 8111 MATLAB statements for Fig. ~ain margin 20 1_I++1. the phase margin is positive.. The phase crossover is the point where the phase curve intersects the 180° axis. and the system is stable. .. ""os Phase margin! 31.. The phasecrossover frequency is wp = 15. r '!'.....2500) rnargin(G) grid The reader should compare the Nyquist plot of Fig.... I 1000 2500 figure 842 Bode plot of L(s) = s(s + 5){s + 50)' margin is measured from the 180°axis and is 31. the gain margin is positive...Sg OJ (rad/sec) 100 1000 90 120 150 ..88 rad/sec...... Because the gain margin is measured below the OdBaxis.. • . and the system is stable....... ~ 40 6{) '30 100 120 1 6.82 dBH+++IIrHHl t I J1 .JWIW=4S~+_14..... ~ """"r... 180 210 " " .. The gain margin is measured at the phase crossover and is 14.. Because the phase margin is measured above the 180° ~axis..811 Stability Analysis with the Bode Plot 20 0 "'~ ~ 457 vGain cros~()ve~ .. [0 1 1] ..72°..........22 10 15.. and PM on these plots....
. K= I. Td= 1 sec " \ . roo 10 ~ ..... i"..TdS Figure 843 Bode plot of L(s) = ( )( ).66 1..... ~ .... \ 360 OJ}! 0....458 ~ Chapter 8.. the system with the present parameters is stable. i""1.446 rad/sec Phasecrossover frequency = I . ... LL(jlt) f+~ K<=>I. """ .. lO 90 135 rt r::::::: :.. ~ .Td=O II .4° Gain margin = 15. ... FrequencyDomain Analysis 8111 Bode Plots of Systems with Pure Time Delays The stability analysis of a closedloop system with a pure time delay in the loop can be cumllll::ted easily with the Bode plot. 53~4° + ISO " v ... I LUm) I.0 1\ 10 Ke...416 rad/sec 40 30 2() ~ ~ r.. ~ 20 30 4() ~ 10 II"... ~ EXAMPLE 8112 Consider that the loop transfer function of a closedloop system is Ke.. ~ " .. to 'S §' .... + dBJ 4.1 tQ(rad/scc) 0..57 dB Thus....446 It)(rad/sec) 1..5 50 {iO '" "..TdS L(s)::::s(s + l)(s + 2) (8101) Fig... = 0.. rr." I I 0.... ~I\ " 225 270 315 ~ LL(jOJ) ..1 0.0 'r... .~J ... Example 8112 illustrates the standard procedure. 1" ...01 0. K= I ~ () r........ ss+1 s+2 ... .... 843 shows the Bode plot of LUm) with K = 1 and Td:::: O. The following results are obtained: Gaincrossover frequency Phase margin = 53...
Fig. because the negative phase shift caused by the time delay increases rapidly with the increase in w. and Mp as relative stability measures.8~12 Relative Stability Related to the Slope of the Magnitude Curve of the Bode Plot . 104.09 seconds. the corresponding phase is also negative. the more negative the phase. Because the negative slope of the magnitude curve is a result of having more poles than zeros in the transfer function. if the loop gain of the system is decreased from the nominal value. Continuing with the example. if the loop gain is increased. the critical value of K for stability is "'''. For example. The adverse effect of the time delay on stability is apparent. the gaincrossover frequency is increased. The reason behind these stability evaluations is quite simple.4 0 1:00 = 0. 8~43 shows the Bode plot of L(jw) with this new time delay. we set TdWt: = 53. the relation between its magnitude and phase is unique.5/20 :=:: 1. the magnitude curve is unchanged. 844 for K:= 1. while the phase curve is unchanged. it is likely that the phase margin will be small or negative.5 dB (8~91). The phase curve droops with the increase in (J). Thus. Thus. PM.66 rad/sec Gain margin = 4. The following example illustrates a conditionally stable system that is capable of going through stable/unstable conditions as the loop gain varies. and the following results are obtained: Phase~crossover frequency = 0. in Fig. and the system is less stable. 812 RELATIVE STABILITY RELATED TO THE SLOPE OF THE MAGNITUDE CURVE OF THE BODE PLOT In addition to GM. On the other hand. To find the critical value of the time delay for stability. EXAMPLE 8121 Consider that the loop transfer function of a cJosedloop system is L(s) = lOOK(s + 5)(s + 40) 83(S+ 100)(8+200) (8103) The Bode plot of L{jw) is shown in Fig. With K still equal to 1. This causes the gaincrossover frequency to be lower. ~. we get the critical value of Tel to be 2. the slope of the magnitude curve of the Bode plot of the loop transfer function at the gain crossover also gives a qualitative indication on the relative stability of a closedloop system.68. and the slope of the magnitude curve is more negative. In general. the magnitude curve is shifted downward. This corresponds to a smaller phase margin. using the definition of gain margin of Eq. The following resuhs on the system stability are obtained: Gaincrossover frequency Phase margin = 1 rad/sec = _78 0 . if the gain crossover is at a point where the slope of the magnitude curve is steep. the steeper the slope of the magnitude curve.~ 459 The effect of the pure time delay is to add a phase of TdW radians to the phase curve while not affecting the magnitude curve. the corresponding phase margin is increased. For a minimumphase transfer function. 8121 Conditionally Stable System The illustrative examples given thus far are uncomplicated in the sense that the slopes of the magnitude and phase curves are monotonically decreasing as w increases. 842.932 radians (8102) Solving for ~f from the last equation. and the slope of the magnitude curve at this frequency is less negative. we set Tdarbitrarily at 1 second and find the critical value of Kfor stability.
. . vY~ ~ .. 25.. V ./ ..... If the gain crossover falls in the region in which the magnitude curve has a slope of ....5 dB. ..... and the system is unstable...... if the gain crossover lies in this range. . 100K(s + 5)(s + 40) There are lwo phase crossovers: one at 25. 8~44..... r.. .. Fig.. Phase crossovers l. 40 dB/decade / i ~ .'"" 60 dB/decade . 8~46......5 /V 69 ""'~ 20 dB/decade '/ :::r r.. For values of K above and below this range... ll2 77..8 rad/sec and the other at 77.. FrequencyDomain Analysis 0 '" Gain crosso~er......... III 28 t".. """  r. It is of interest to compare the results on stability derived from the Bode plot and the Nyquist plot. ... ~ 180 210 240 270 ". From the magnitude curve... ..7 ..... .. the phase of L(jw) is less than 180°. . The number of crossings of the root loci on the jwaxis of the s~plane equals the number of crossings ...7 rad/sec. As observed from Fig...8 )00 200 '"1000 co (radlsec) 90 120 150 i :s. . r.. ~ ~/ ~T".8 40 200 1000 w(radlsec) Figure 844 Bode plot of L(s) = s3(s + 100)(8 + 200)' K = 1.JOO /' 1 T I 77. ..... 845 shows the Nyquist plot of L(jw). .460 • Chapter 8... the range of K for stable operation is found to be between 69 and 85... the system would be stable.. This example serves as a good example of the relation between relative stability and the slope of the magnitude curve at the gain crossover.. . ~ l.. and the system is unstable.. I I r. at both very low and very high frequencies... the slope of the magnitude curve is 60 dB/decade... if the gain crossover falls in either one of these two regions.. the system is stable only if the gain crossover falls in about half of these regions.. The phase characteristics between these two frequencies indicate that.... .. the system is stable.... ~V 40 dB/decade ~~ / ~ 60 dB/decade . " 140 1 5 10 25. .. In the two sections of the magnitude curve that have a slope of 40 dB/decade. 5 10 II " '.. The root.~ 56 70 84 85.. locus diagram of the system is shown in Fig.. l.. but even then the phase margin is small. t'rI'o.. a . I I i' 40 ~~ I. The root loci give a clear picture on the stability condition of the system with respect to K....20 dB/decade.7 .. the phase margin is negative.
............:...:..... ... s3 (s + 100)(s + 200) splune K=OK>O K=~ KQCl 40 Not to scale lOOK(s + 5)(s + 40) " OCI s Frgure 846 Root I ' 0 f G() ..~ + IOO)(s + 200)" .8~12 Relative Stability Related to the Slope of the Magnitude Curve of the Bode Plot" 461 jlmL L(jeo)plane o ReL Figure 845 Nyquist plot of Les) = 100K{s + 5)(s + 40) • K = 1..:... .s3(...
r. 813 STABILITY ANALYSIS WITH THE MAGNITUDEPHASE PLOT The magnitudephase plot described in Chapter 2 is another form of the frequencydomain plot that has certain advantages fot analysis and design in the frequency domain .462 Chapter 8. FrequencyDomain Analysis of the phase curve of L(jw) of the .135 . The magnitudephase plot of a transfer fu nction L(jw) i. Ga in positive gain nlnrgin 20 i...02s .+ .c Unstable region . • The phase crossover is where the locus intersects the .~ Unstable region.1 8 j 8 . (898) is constructed in Fig. positive phase margin I . s 1 + . • The critical point is the intersect of the OdBaxis and the 180°axis.100 ':'''' 270 225 . s 1 + 0. negative pha. 10 Figure 847 Gamphase plot of L(s) = ( 02 )( ).s done in IL(jw) l(dB) versus L L(jw) (degrees) . 842.".180 . The gain and phase crossovers and the gain and phase margins are clearly indicated on the magnitudephase plot of L(jw) . The magnitudephase plot of the transfer function in Eq. negative gai n margin Cri~cal pOInt ...60 r+r~lr~ 300 80 H~~~. The reader should check the gain margins obtained from the Bode plot and the coordinates of the crossover points on the negative real axis of the Nyquist plot with the values of K at the jwaxis crossings on the root loci. 20r.180°axis..1800 axis of the Bode plot and the number of crossings of the Nyquist plot of L(jw) with the negative real axis. 847 by use of the data from the Bode plot of Fig.e margin t +++ Stable region .:.I o 11~~~j(Stable region.90 Pha~e (deg) .
. Consider that G(s) is the forwardpath transfer function of a unityfeedback system. The regions in which the gain and phase crossovers should be located for stability are also indicated. The closedloop transfer function is M(s) = 1 + G(s) G(s) (8104) For sinusoidal steady state. It is necessary to develop a graphical method for the detennination of Mr. the effect of the time delay is to add a phase equal to WTd x 1800 /7r along the curve. the resonant peak M" and bandwidth BWare difficult to obtain for highorder systems. Similarly. when the loop gain of L(jw) changes. Because the vertical axis for IL(jw)1 is in dB. and BW using the forwardpath transfer function G(jUJ). W r• and BW can all be determined from the plot with the help of the constantM loci. the method is directly applicable only to unityfeedback systems. although with some modification it can also be applied to nonunityfeedback systems. (8106) leads to MV(1 +x)2+y2 = vx2 + y2 Squaring both sides uf Eq. let M denote lM(jw)\.. OJ. Another advantage of using the magnitudephase plot is that. the locus is shifted horizontally withuut distortion to the curve. closedloop system parameters such as My. then Eq. >·814 CONSTANTM lOCI IN THE MAGNITUDEPHASE PLANE: THE NICHOLS CHART It was pointed out earlier that~ analytically. As we shall see in the following development. (8~107) (8107) gives (8~I08) M2[Cl +x)2+y2] =X2 +y2 . for simplicity. x denotes ReG(jUJ) and y denotes ImG(jUJ). G(s) becomes G(jw) = ReG(jw) + jlmG(jw) =x+jy (8105) where. If L(jw) contains a pure time delay Td . These closedloop performance parameters are not represented on the Bode plot of the forward~path transfer function of a unityfeedback system. for unitY1eedback systems. and the Bode plot provides infonnation on the closedloop system only in the form of gain margin and phase margin. the locus is simply shifted up and down along the vertical axis.814 ConstantM Loci in the MagnitudePhase Plane: The Nichols Chart • The gain crossover is where the locus intersects the OdBaxis. The magnitude of the closedloop transfer function is written IM(jw)l = I G(jW? = 1 + G(Jw) I J(1 /x 2 + y2 (8106) +x)2+y2 For simplicity of notation. ~1 463 • The gain margin is the vertical distance in dB measured from the phase crossover to the critical point.. we replace s with jw. when a constant phase is added to L(jw). • The phase margin is the horizontal distance measured in degrees from the gain crossover to the critical point.
M2) and adding the term [M2/ (1 . (8111) describes in the G(jw)plane a family of circles that are called the constantMloci. Frequency~Domain Analysis Rearranging Eq.2M2x = M2 (8109) This equation is conditioned by dividing through by (1 .833 ReG 2 1 Figure 848 ConstantM circles in polar coordinates. These circles are symmetrical with jlmG G(jto)·plane M= 1.M2)x 2 + (1 . (8108) yields (1 . (8111) represents a circle with the center at x The radius of the circle is = ReG(jw) = M1 y=o (8112) (8113) When Mtakes on different values. 848 illustrates a typical set of constantM circles in the G(jw)plane. Eq. Eq. . Fig.M2 (8110) which is finally simplified to (X  M2)2 1 _ M2 +y2 = (M)2 1 _ M2 M2 1_ (8111) For a given value of M. We have x 2 + y2  2M2 I _ M2 X + (M2) 2 M2 (M2) 2 1 _ M2 = 1 _ M2 + 1 .464 Chapter B.M2) ]2 on both sides.M2)y2 .2 M=O. or the constant·M circles.
(b) Corre. Graphically. The circles to the left of the M = 1 locus correspond to values of M greater than 1~ and those to the right of the M 1Une are for M less than 1. If we want to keep the value of M" less than a certain value. jO). For a given loop gain K = Kl. 849(a) illustrates the Nyquist plot of G(jlf)) for a unity·feedback control system together with several constant·Mloci..jO) point. when M becomes infinite. the G(jw) curve must not intersect the corresponding M circle at any point and at the same time must not enclose the (I. the intersects between the G(jw) curve and the constantM loci give the points on the IM(jw)lversusw curve. Eqs. . (811. the circle degenerates to a point at ( 1.707 II++"""""'"""'~.. Fig. The constantM circle with the smallest radius that is tangent to the G(jw) curve gives the value of MI" and the resonant frequency Wr is read off at the tangent point on the G(jw) curve. The resonant peak Mr is found by locating the smallest circle that is tangent to the GUto) = GU colplane ReG K=K3 K= K2 (a) K= KI M(DJ) 1 0."~ ro (b) Figure 849 (a) Polar plots of G(s) and constant~M loci.~ponding magnification curves.1) and (8112) show that. the intersection of the G(jw) curve and the constantM circle gives the value of M at the corresponding frequency on the G(jw) curve.814 ConstantM loci in the MagnitudePhase Plane: The Nichols Chart ~ 465 respect to the M = 1 line and the real axis.
A major disadvantage in working in the polar coordinates of the Nyquist plot of G(jw) • When the system is is that the curve no longer retains its original shape when a simple modification such as the unstable. and thus the resonant peak will be larger. as shown in Fig. and if the system is still stable. The resonant frequency is shown to be W r 2. This requires the complete reconstruction of the Nyquist plot of the modified G(jw). and the constantM loci and Mr no longer have any meaning. If the loop gain is increased to K 2 . the magnitudephase plot is affected only in the horizontal direction. and the resulting loci are called the Nichols chart. it i1) more convenient to work with the magnitudephase plot of G(jw).1 lO~~~~~~ 20~________~____~____~~~~~~ 360 315 270 225 180 Phuse(deg) 135 90 45 o Figure 850 Nichols chart.no longer have must the ]oop gain be altered. the entire G(jw) curve is shifted up or down vertically without distortion. 850.707 locus. the system is unstable. Once the G(jw) curve of the system intersects the M = 3 dB is constructed in the Nichols chart~ the intersects between the constantM loci and the G locus of the Nichols chart. a constantM circle with a smaller radius that corresponds to a larger M is found tangent to the G(jw) curve. and MI' is infinite. FrequencyDomain Analysis curve. any meaning. the system is marginally stable. . because when the loop gain is altered. Wp3 is now the same as the resonant frequency WI" When enough points of intersection between the G(jl. For design work involving Mr and BWas specifications. The resonant 30r~~r~r_r_~~ M= 1. (jw) trajectory give the value of M at the corresponding frequencies of G(jw). Frequent1y~ in design situations.tJ) curve and the constantM loci are obtained~ the magnification curves of 1M(jw ) Jversusw are plotted. jO) point. the constantM loci in the polar coordinates are plotted in magnitude• BW is the frequency phase coordinates. For that reason. The bandwidth of the closedloop system is found at the intersect of the G(jw) curve and the M = 0. The resonant frequency is found at the point of tangency and is designated as l. A typical Nichols where the GUw) curve chart of selected constantM loci is shown in Fig. the constantM change of the loop gain is made to the system. 849(b). without affecting the gain.466 lilt Chapter 8.tJrl.0 20~=_r_H_~~~~~~~~+__=~ M= 1. not only loci and M. For values of K beyond K 3 .. which is closer to the phasecrossover frequency (J)p than Wrl' When K is increased to K3• so that the G(jw) curve now passes through the (1. but a series controller may havc to be added to the system. When the phase properties of G(jO) are changed independently.
....14..5..1'0 't.. 80 1...... i'...: 40 r20 r.. l / Phase crossover t5' pivt ~ og ... .... i"1""""":. '~ 1'...2....248 ... &i' r.... ~ ~~61 dB . ......000 Figure 851 Bode diagrams of the system in Example 8141.... and 273."".. 'l 2lO 1'\r'\.25<> ......5 II I  tr. ~ .... The bandwidth of the closedloop system is the frequency at which the GGw) curve intersects the M = 0..5 X 10 K 7 (8114) The Bode plots forG(jw) are shown in Fig.. .. (5153). ~ K=273.....t~  IT r.707 or M = 3 dB locus....0 10 100 OJ (rad/sec) 1000 10. PM::. The following example illustrates the relationship among the analysis methods using the Bode plot and the Nichols chart..." I 240 270 V 11 07j 56 {ar'ier 10 1..tr   I::::r..814 ConstantM Loci in the MagnitudePhase Plane: The Nichols Chart .57 dB.26}(s + 3008) 1. ~ .. EXAMPLE 8...:: 1::::"""" I"'" K~ 181...248. Gain crossovers \ l I I '~tGM.: GM..2 1 I ~ 100...000 90 120 150 ~ ~U 'PM ='64...57 J 7" 0 K= 14.. l' .... .:: ?5 20 40 60 K=7.. The forwardpath transfer function of the unityfeedback system 1S given by Eq. Gain crossovers It ~ "t'.. 467 peak Mr is found by locating the smallest of the constantM locus (M ~ I) that is tangent to the G(jw) curve from above..... The gain and phase margins of the closedloop system for these values of K are determined and shown on 70 60 ~ ~ r.o ~ ..1 Consider the positioncontrol system of the control surfaces of the airship analyzed in Section 58. OM =j 1.14.. ~~ GM = 25.. ~ r...~ I' ~ '§' 180 . "...0 100 OJ (rad/sec) lOoo 10. r""'1"0 If 'S '§' ... The resonant frequency is the frequency of G(jw) at the point of tangency.181. .. ....55 dB .::: 7. 0 dB + r. r.. and is repeated here: G( ) S = s(s + 400. 851 for K. r0.....57.""'~ r"'" I ~ r r... r ~~ r . ..=778~ I t\....
Table 82 summarizes the results of the frequencydomain analysis for the four different values of K together with the timedomain maximum overshoots determined in Section 58. and the bandwidth BW. The gain and phase margins are also clearly marked on the magnitudephase plots.~~~_+_r_r~__. Fig. 852. The magnitudephase plots of G(jw) corresponding to the Bode plots are shown in Fig. give information on the resonant peak M TO resonant frequency W.a eI!I 0 ~ 20 OJ ~ 1107. These magnitudephase plots. together with the Nichols chart. 60~·~4~~+~+~4+4 80~~~~~~~~~~~ 270 225 180 135 90 Phase (deg) Figure 852 Gainphase plots and Nichols chart of the system in Example 8141. . the Bode plots. Frequency~Domain Analysis o a 60~~~44+_+rr~ r 40H_4_+~+_~_4_+~~ill 20 :8"0 6 u ..468 ~ Chapter 8. 853 shows the closedloop frequency responses.56 rad/sec 40r~1~~_.~ ..
0 (rad/sec) 1. K= 181.0 43.. ~ o 1.. ~~ 10.3 15.8·15 Nichols Chart Applied to Nonunity·Feedback Systems 10 9 ~ 469 K = 273.57 .00 Gain Margin (dB) 31.2 273.33 900.5 Mr= 1. the closedloop transfer function of the system is expressed as M( ) G(s) s = 1 + G(s)H(s) (8115) where H(s) =f 1. since the numerator of M(s) does not contain H(jw).0 to (0 100 (md/sec)  \\ 1000 ~ r..78 0 • 815 NICHOLS CHART APPLIED TO NONUNITYFEEDBACK SYSTEMS The constantM loci and the Nichols chart presented in the preceding sections are limited to dosedloop systems with unity feedback whose transfer function is given by Eq..25 BW (rad/sec) 119.0 Mr= 1.9 64. TABLE 8·2 Summary of FrequencyDomain Analysis K 7.2 M.55 3.5 181.5 1402..5 41! 14. the constantM loci and Nichols chart can stilI be applied to a nonunityfeedback system.6 ce 7. When a system has nonunity feedback..0 1661..2 100. Let us consider the function G(s)H(s) P(s) = H(s)M(s) = 1 + G(s)H(s) (8116) ..61 0 Phase Margin (deg) 75.f Mr=oo I I I  8 7 6 '§ .=7..000 Figure 853 C1osed·loop frequency response of the system in Example 8·141..0 7.0 1. The constantM loci and the Nichols chart cannot be applied directly to obtain the closedloop frequency response by plotting G(jw)H(jw).0 270.25 Maximum Overshoot (%) 0 WI' Mr 1.0 / I I I ~~ J\ V~/ ~ . By proper modification. r..00 1000.25_ K= 14.. (8104).6 rt I ~ 5 4 3 2 K=7.57 5.57 4..
We shall show how the Nyquist plot and the Nichols chart can be utilized for the analysis and design of control systems based on sensitivity considerations. Once this is done... which is a multiplying factor in O(s). (8116) is of the same fonn as Eq.00 0..) ~ I '" i IM(jm)1 l~ If... (8121) and simplifying. the system with respect to parameter variations can be easily interpreted using frequencydomain plots.90 0. is defined as dM(s) SM ( ) = M(s) = dM{s) G(s) G s dG(s) dG(s) M(s) G(s) (8121) Taking the derivative of M(s) with respect to G(s) and substituting the result into Eq.80 ::!!\.70 2.60 \ i\ ~ .. the frequencyresponse infonnation for M(jw) is obtained as follows.20 1. .. (8104).470 ~ Chapter 8.1 1St! (jm) I.8 16 SENSITIVITY STUDIES IN THE FREQUENCY DOMAIN • Sensitivity study is easily The advantage of using the frequency domain in linear control systems is that higherorder carried out in the frequency systems can be handled more easily than in the time domain.. I "" '. ~V V" ~ (D (l'adfsec) \ lO ~ 100 . the sensitivity of domain. The frequency response of P(jw) can be determined by plotting the function G(jw)H(jw) in the amplitudephase coordinates along with the Nichols chart. I 1.30 0.50 1. in terms of dB. Figure 854 IM(jw) I and IsM (jw) I versus w for G 2500 (8) = s(s + 5)(s + 2500) . 0.... Eq.01 0. ~ / / I V .00 2.00 0. we have SG{s) = 1 + G{s) 3...10 1.40 M 1 I/G(s) 1 + I/G(s) (8122) '§' 100') 2. FrequencyDomain Analysis Apparently. Consider a linear control system with unity feedback described by the transfer function M(s) = 1 + G(s) G{s) (8120) The sensitivity of M(s) with respect to the loop gain K. Furthermore. (8117) IM(jw)l(dB) = IP( jw)l(dB)  IH(jw)l(dB) (8118) (8119) tPm(jw) M = !MUw) = LP(jw)  LH(jw) . IP(jw)1 IM{Jw) I = IH{ jw)1 or. .
20 10 1\ )' r:::::: ~ L . with G(jm) replaced by I/G(jm)..2'::::: ~M=r4.. Thus. / / f 0 e I _L S $' G. the sensitivity function Stj (8) is a function of the complex variable $..::. (8 98). Fig... phase plots of GUw) and IIGVw) for 2500 60 270 t7 &/ ~ 225 180 Phase (deg) 135 90 G(s} =s(s + 5}(s + 50} ../ /J L V V 20 30 40 / 1(" ):=0.7017 ::...~ ~  ~ /' /1 .. (8123) can be detennined by plotting l/G(jw) in the magnitudephase coordinates with the Nichols chart.M = 2. _ + G(jw)Ill JI/GUw)[ < k (}w)I11 + I/G(jw)1 G where k is a positive real number. 854 shows the magnitude plot of S~ (s) when G(s) is the transfer function given in Eq..0. .81ocus from below.'... given in Eq.... ]M(jw)l. Notice that G(jw) is tangent to the M = 1. Fig.a 0 10 ~ ~  '" I II a \ 0 \ a / / LM=1.8 tad/sec to the openloop system whose sensitivity to the variation of K is always unity.I 't:I '8 . 855 shows the magnitudephase plots of G(jw) and IjG(jw) ofEq.816 Sensitivity Studies in the Frequency Domain 4 471 Clearly. (8106).:\ t:.2  A G(jDJ) 50 Figure 855 Magnitude. (8~123) is analogous to the magnitude of the closedloop transfer function. This sensitivity criterion is in addition to the regular performance criteria on the steadystate error and the relative stability. In general.O~ . it is desirable to formulate a design criterion on sensitivity in the following manner: w 1 _ I ~."':': / M=l. (898). the sensitivity function of Eq.... 60 50 40 1 G(jm) 30 \.M=O. It is interesting to note that the sensitivity of the closedloop system is inferior at frequencies greater than 4..8 ~ . Eq.
It was shown that the frequency~response plots of G(jm) and l/G{jw) can be readily used for sensitivity studies. The relationships among these parameters of a secondorder prototype system were derived analytically. system. List the advantages and disadvantages of carrying out frequencydomain analysis with the Bode plot. is the maximum value of IS~ (s) I. described in detail in Chapter 9. Explain why it is important to conduct frequency~domain analyses of linear control systems. this chapter does not contain any software. Frequency~Domain Analysis which means that M t . the reader may practice all the concepts discussed here. 5.. Thus. List the advantages and disadvantages of studying stability with the Nyquist plot. the condition of stability is simplified so that the Nyquist plot will not enclose the critical point. 2. The Automatic Control Systems software (ACSYS) consists of a number of mfiles and GUIs (graphical user interface) for the analysis of simple control engineering transfer functions. resonant frequency (ll. of the closedloop system is 1. The design of robust control systems (low sensitivity) with the frequencydomain methods is discussed in Chapter 9. the loop gain ofG(jro) must be high. If G(s)H(s) is a minimum~phase transfer function. Performance specifications such as the resonance peak M". Eq. Define bandwidth BW of a 4. The discussion should add more perspective to the understanding of both subjects. The stability of a singleloop control system can be investigated by studying the behavior of the Nyquist plot of the loop transfer function G(s)H(s) for w = 0 to w 00 with respect to the critical point. . but it is known that. Define resonance peak Mr of a closedloop control system. closed~loop 3. and bandwidth BW were defined in the frequency domain. The relationship between the root loci and the Nyquist plot was described in Section 8~ 7.2 curve from above and. ~ 817 MATLAB TOOLS AND CASE STUDIES Apart from the MATLAB toolboxes in this chapter. for low sensitivity. 818 SUMMARY The chapter began by describing typical relationships between the open~loop and closed~loop frequency responses of linear systems. All the frequency response topics may also be solved utilizing ACSYS. These quantities were defined in the polar coordinates as well as on the Bode diagram. Sensitivity function S~ (jw) was defined as a measure of the variation of M(jw) due to variations in G(jw). the designer is again challenged by the task of designing a system with both a high degree of stability and low sensitivity. The stability of systems with pure time delay is analyzed by use of the Bode plot. The gainphase plot allows the Nichols chart to be constructed for closedloop analysis. in general.8. The values of Mr and BW can be easily found by plotting the G(jw) locus on the Nichols chart. (8123) shows that. 854. The 1/G(jw) curve is tangent to the M = 2. Relative stability was defined in terms of gain margin and phase margin. The Nyquist criterion for stability analysis of linear control systems was developed. The effects of adding simple poles and zeros to the loop transfer function on My and BW were discussed..472 "'~ Chapter 8. high gain could cause instability. using the MATLAB toolboxes developed in this chapter or the ACSYS software. Finally. = REVIEW QUESTIONS 1. In Chapter 9 we will introduce the Automatic Control Systems MATLAB tools. according to Fig.
Tech. For a pl·ototype secondorder system. Jan. T. E 28. Bode plot can be used for stability analysis for minimum.comJeoUege/golnaragbi• ." IEEE Trans.."Frequency Domain Stability CriteriaPart I. Feb. A closedloop system with a pure time delay in the loop is usually less stable than one without a time delay. Willems. Answers to these review questions can be found on this book's companion Web site: www. J. K. Gain margin is measured at the phasecrossover frequency. 12. 22. Automatic Control. (T) (T) (T) (T) (F) (F) (F) (F) 12. Nichols chart can be used to find BW and Ml' information of a closedloop system. Willems.l26147. Give all the properties of a minimumphase transfer function. R. 1932. the steadystate output of the system will also be of the same frequency." IEEE Tmns. Vol. pp. 407413. "A Supplement to the Note on the Generalized Nyquist Criterion. 14. pp." IEEE Trans. List the advantages and disadvantages of carrying out frequencydomain analysis with the magnitudephase plot. 10. "Regeneration Theory:'Bell System.wiley. (T) (T) (F) (F) 20. (T) (T) (F) (F) 23. Oct. 2. What condition must be satisfied by the function L(jw) so that the Nyquist criterion is simplified to investigating whether the (1.• Vol 11. Automatic Control. (T) (F) (F) 21. Brockett and J. 1965. L. 0 • (T) {T} (T) (F) (F) (F) Gaincrossover frequency is the frequency at which the gain of L(jw) is 0 dB. number of zeros of L(s) that are in the righthalf splane P = number of poles of L(s) that are in the righthalf splane p(r) = number of poles of L(s) that are on the jwaxis Give the conditions on these parameters for the system to be (a) openloop stable and (b) closedloop stable. Vol. pp. Educ. "A Reformulation of Nyquist's Criterion. W. H. The following quantities are defined: Z =:. 8. if the phase margin is negative. Brockett and l. For a minimumphase loop transfer function L(jfJ. July 1965.. jO) point is enclosed by the Nyquist plot? 9. Phase marght is measured at the gaincrossover frequency. 15. By applying a sinusoidal signal of frequency l. Yeung. REFERENCES Nyquist Criterion of ContinuousData Systems J. Natesan. R. pp. 5.as well as nonminimumphase transfer functions. 19. the value of Ml' depends solely on the damping ratio t. L. "Frequency Domain Stability CriteriaPart II. w. 7. 11. Automatic Control. AClO. 4. M . 16. then the closedloop system is always unstable. R. S. 215216. 18. Vol. Nyquist. 255261. AC.. 13. The general effect of adding a pole to the loop transfer function is to make the closed~loop system less stable while decreasing the bandwidth. Give the definitions of gain margin and phase margin. 5960. 1985. 3. AClO.. Phasecrossover frequency is the frequency at which the phase of L{jw) is 0 17. Adding a zero to the loop transfer function will always increase the bandwidth of the closedloop system. April 1967. Vol.t)o to a linear system. The slope of the magnitude curve of the Bode plot of L(jw) at the gain crossover (T) usually gives indication on the relative stability of the closedloop system.J)." IEEE Trans. pp.References ~ 473 6.
30% and If ::. The transfer function of a system is G(s) =/ s+ A2 s+A" Determine when the system is a leadnetwork and lagnetwork.375s + 1) 1.(s + 1.58)(1 + O.. 5758. .1s) (b) G(s) = s(l + 0. AC7.:::. resonant frequency loop system for the following values of K: (a) K = 5 (b) K (J)" and bandwidth BW of the closed = 21. 84. FrequencyDomain Analysis Sensitivity Function 6. Automatic Control.1 sec.5K (a) Analytically (ind K such that the closedloop bandwidth is. Vol.24 Hz). Consider the forwardpath transfer function of a unityfeedback control system given by G() s . 8M6.5 500 (d) G(s) = s(s + 2)(s + 10) (£) G(s) 10(s + I) lOOe.58)(1 + O. O. Find the corresponding limiting values of M.. find the resonance peak Mr.474 > Chapter 8. 0. pp. The forwardpath transfer function of a unityfeedback control system is K G(s). Do not attempt to obtain the solutions analytically. A. "Graphical Evaluation of the Sensitivity Function Using the Nichols Chart.s (e) O(s) = 8($2 (g) G(s} +...\'(s2 + 5s + 5) 85.8.2 sec.s(s + 6. s(0.1. 82.2 sec.' + 50) lO(s + 5) (h) O(s) = . 87. The forwardpath transfer functions of unityfeedback control systems are given in the following equations.) 5 ]0 (a) G(s) = s{1 + 0.2)(s + 4)(. about (b) Use MATLAB to verify your answer to part (a).25s2 + O. 1 (c)K=100 Use the formulas for the secondorder prototype system given in the text. PROBLEMS 81. Gclb. and BWanalytically.It + 1) 100es = S(82 + lOs + 100) = $($2 + 10. (Reminder: Make certain that the system is stable. 83.\' + 10) 0.5 rad!s (0.Is) _ (c) G(s) . July 1962. resonant frequency lU.39 Use MATLAB to verify your answer to Prob1em 81. Find the resonance peak Mr. 8. Use MATLAB to solve the following problems.54) Analytically. and bandwidth BW of the closedloop systems. Repeat Problem 85 for maximum overshoot :5 0. Repeat Problem 85 for maximum overshoot::." IRE Trans. The specifications on a secondorder unityfeedback control system with the closedloop transfer function M(s) = Yes) R(s) = s2 + 2~£t)ns + w~ :5 20% and tr £t)~ are that the maximum overshoot must not exceed 10% and the rise time must be less than 0.
0. 4.ls){l + O.2s)(l +O. Solve for the intersect of L(jw) on the negative real axis of the L(jw)p\ane analytically. 1. overshoot for step BW according to the result of part (a).. find the number of poles of the closedloop transfer function that are in the righthalf splane. The forwardpath transfer function of a unity~feedback control system is G(s)  . 2. Repeat Problem 8·8 with a resonance peak of 2..Problems 8·9. 3. peak time. Determine the stability of the closedloop system. 15. und 5.3. indicate the values of the maximum overshoot. Use MATLAB to find the solutions. If the system is unstable. (b) If BW == 1. 8.. find K and the damping ratio. 814. If a loop transfer function of a system is given by 0.05. pha. The forwardpath transfer function of a unityfeedback control system is 1 +Ts G(s).18)(1 +O.5s) 1 () () GsHs= O. (c) Use MATLAB to verify your answer to part (b).2s(s2 .5K 3 =.ls){l + O.4. 13.2s)(1 +0. 3 rad/sec The forwardpath transfer function of a system with an integral control H(s) = K s is 1 G(s) = lOs + I input~ (a) Find K when the closedloop resonance peak is 1.5s} s( \ + O. 8. 2.'ie margin~ and closedloop 8...9 o Figure 8P10 8~11. 1.5. i '§' 0.. Sketch the corresponding unitst~p response of the system..::~~S 1)( 1 T. 20 _ 10 (a) L(s) = (b) L(s) s{ 1 + O. Sketch the Nyquist plot of L(jw) for w = 0 to w = 00.2. The closedloop frequency response IM(jw)Jversusfrequency of a secondorder prototype system is shown in Fig 8P10.. and the steadystate error due to a unitstep input. You may construct the Nyquist plot of L(jw) using MATLAB..5s) .= (e) L(s) = toO ( I + s) s(1 +0.375s" + S + 0. and 5.2s( s2 + s + I) Use MATLAB to find the values of BWand Mr of the closedloop system for T = 0. (b) Determine the frequency at resonance.5radls.5k (a) Use the secondorder approximation to find the BW and the damping ratio.~) 1 + + + Use MATLAB to find the values of BW and Mr of the closedloop system for T = 0.25s· + O.5s) (d) L(s) 2 10 s (I +0. 4. The loop transfer functions L(s) of singlefeedbackIoop systems are given below. 12. 475 8·10.
The forwardpath transfer function of a unityfeedback control system is G(s) = (s+5yr Determine by means of the Nyquist criterion..s(s + 2)(8 + 5)(8 + 15) K K (d) L(s) = .. (b) Check the answers by means of the RouthHurwitz criterion... You may use a computer program to plot the :Kyquist plots. FrequencyDomain Analysis (e) L(s) = 3(8 + 2) s(s3 + 3s + I) 100 0. SPIS. The loop transfer functions of singlefeedbackloop control systems are given in the following equations. Kn8) Select the value of Kp so that the parabolicerror constant Xu is 100. K(s + 1) (a) L(s) = s(s + 2)(s + 10) . The forwardpath transfer function of a unityfeedback control system with a PD (proportionalderivative) controller is G(s) = 10(Kp. x Figure 8P18 Determine by means of the Nyquist criterion. (b) Check the answer obtained in part (a) with the RouthHurwitz criterion. Sketch the Nyquist plot of L(jw) with K = I for lIJ = 0 to w = 00. The block diagram of a feedback control system is shown in Fig. K (b) L(s) ..476 ~ Chapter 8. Determine the range of Kv for stability by the Nyquist criterion. 819.] (I) L(s) = S(8 + l)(s2 + s + I) (g) L(s) = s(s + l)(s2 + 2} (b) L{s) = s(s + l)(s + 100) 10(8 + 10) 8~ 16. s(s3 +2s2 +s+ 1) +K(s2+s+ I) =0 (a) Apply the Nyquist criterion to determine the values of K for system stability. . (e) L(s) = 8 2(S + 2}(s + 10) (s + 5}(s + 2)2 (e) L(8) = K(s + 5)(8 + 1) (s + 50)(s + 2)3 K 8·17. Repeat Problem 819 for 83 + 3s2 + 3s + 1 + K = O... (a) Apply the Nyquist criterion to determine the range of K for stability.loop system to be stable. (a) n = 2 (b) n (e) =3 Sketch the Nyquist plot for the controlled system shown in Fig.. Find the equivalent forwardpath transfer function Geq(s) for (V = 0 to w = 00. K n= 4 818. Sketch the Nyquist plot of G(jw) for (V = 0 to w = 00.".. The characteristic equation of a linear control system is given in the following equation. 8P22. 820.. 821. Apply the Nyquist criterion and determine the values of K for the system to be stable. 822. . the range of K ( 00 < K < 00) for the closedloop system to be stable. the range of K( 00 < K < co) for the closed.
006. sketch the Nyquist plot of an equivalent transfer function Geq ( )0)) that has Ko as a mUltiplying factor. has the following values: (a) Kt = 0 (b) Kl = G(s) = om (c) K/ = 0. (a) For A = 50 and Ko = 100.ls 1 0. 8P24. 8P24.0706. The values of A. The block diagram of a dcmotor control system is shown in Fig. 824. For the system shown in Fig. (b) LetN = 10 and Ko = 100. Kj = 50. 8P26 shows the block diagram of a se·rvomotor. The forwardpath transfer function of the liquidlevel control system in Problem 542 is KaKjnKjN s(RaJs + KiKb)(As + Ko) The following system parameters are given: Ka = 50. N. Sketch the Nyquist plot of an equivalent transfer function Geq(jO)) that has A as a multiplying factor. sketch the Nyquist plot of G(jO)) for 0) = 0 to 00 with N as a variable parameter. let K = 10.011 Y(s) KzS Figure 8P24 825. and Ra = 10. and Ko are variable. Find the critical value of Ko for stability. Fig.Problems ~ 477 G(s) Y(s) ~ G(s)= (s+4~s+5) Figure 8P22 823. x y Figure 8P26 . 826. Kb == 0. Find the maximum integer value of N so that the closedloop system is stable. J = 0.01. Kj = 10. (c) For A = 50 and N = 10.1 R(s) E(s) K +  h(}+  10 1 +O. Find the critical value of Ko for stability. Determine the range of K for stability using the Nyquist criterion when K. n = 0. Find the range of K/ for stability with the Nyquist criterion.
draw the Nyquist plot and apply the Nyquist criterion to determine the range of K for stability and determine the number of roots in the righthalf splane for the values of K where the system is unstable.. The openloop transfer function of a system is given by G(s)H(s) K = s(r]s+ 1 r2s + 1) )( Study the stability of the system for the following: (a) K is small. FrequencyDomain Analysis Assume J::::: 1 kg_m2 and B = I Nm/radlsec. of [he valve motion changes the output concentration from zero to maximum.~:::~ s(:~ + lOs + 100) (a) When K = 1. and 0. SP32 is devised to control the concentration of a chemical solution by mixing water and concentrated solution in appropriate proportions. values: (a) Kf = 0 (b) Kf = 0.. (a) When K O. 8~32.Tds 829. The valve ports can be assumed to be shaped so that the output . when it is viewing concentrated solution. 828. The steelrolling control system shown in Fig. find the maximum value of K for system stability. let K = 10.lt determine the maximum time delay Ttl in seconds for the to be stable. 8P26. find the maximum value of K for system stability... For the system shown in Fig. ea = 10 V. (b) When the time delay Td is 1 sec. determine the maximum time delay Tel in seconds for the closedloop system to be stable. = closed~loop system (b) When the time delay T. 830. The system schematic shown in Fig. Figure 8P28 (a) G(s) = (s _ 1)2 s+1 (b) G{s) ~ (s s1 + 1)2 lOOKe.) is X(s) Eo(s) K = s2 + lOs + 100 When the sensor is viewing pure water... Repeat Problem 829 with the following conditions. Chapter 8. 831. 4P18 has the forwardpath transfer function G(s) = :. the amplifier output voltage ell is zero.f is 0.1 sec.478 .. Determine the range of K for stability using the Nyquist criterion when Kf has the following. x ~L.2 827.1 (e) Kf ~O. ____ /(I_G_(_S)_:_ ~ to y .1 in. Find the range of Kf for stability with the Nyquist criterion. SP2S. (b) K is large. For the controlled system shown in Fig. The transfer fUllction of the system components between the amplifier output ea (V) and the valve position x (in..
833. 835. find the max imum distance D (in.1 in?.1'+1 1. it is desirable to place it at some disulllce D in . and BW of the closed. WI'. from the valve. ea = 1 Y. x K . The forwardpath transfer function of a unityfeedback conLrol system is G(s) = s(s2 + lOS.) so that the system is stable. For the mixing system described in Problem 832. and 0. (b) When K = 10. Figure 8P34 shows the block diagram of a conu'ol system. of the valve motion changes the output concentration fro m zero to maximum. (b) Find the parameters of the secondorder system with the openloop transfer function . Repeat the three parts of Problem 832. The rest of the system characteristics are the same as in Problem 832. and the rate of tlow is 103 inJsec regardless of the valve position. Find the maximum value of K for system stability. the following system parameters are given: When the sensor is viewing pure water.1 in. (c) Let D = 10 in. (b) Determine the number of roots in the righthalf splane for the values of K where the system is unstable. 834. WaleI' Concentrated solution Scn~or Solenoid Amplifier Figure 8P32 (a) Derive the loop tnll1sfer function of the system .+ Y Figure 8P34 (a) Dntw the Nyquist plot and appl y the Nyquist criterion to determine the range of K for stability. when it is viewing concentrated solution. (c) Use Routh's criteri on TO determine the range of K for stability. the amplifier output voltage es = 0 V. To make sure the senSor views a homogeneous solution. The output tube has a crosssectional area of 0.loop system. Use the Nyquist stability criterion.Problems 479 concentration varies linearly with the valve position.\' + 600) 1000 (a) Find the values of M r .
2s)(1 + 0.1s)(1 + 0. Find the gain. 41. a) G(s) _ ( .. (b) Plot the root locus. Plot G(jw)/K in the gainphase coordinates of the Nichols chart.s (e) G(s) = s(1 + 0.. (e) Find out how much the loop gain must be changed from its nominal value if the phase margin is 45°.2.s(1 + 0. The forward .. (e) Find the gain and frequency where instability occurs.1s)(1 + 0. Chapter 8.path transfer functions of unityfeedback control systems are given in the following equations. 836.18 + 0. The forwardpath transfer functions of unityfeedback control systems are given in the following equations. . Plot the Bode diagram of GUw) / K.ls)(1 + 0. (b) Repeat part (a) if the gain is doubled from its nominal value.58) K K (d) G(a) = .5s) 10Ke.s(1 + 0. .. and Wr as the thirdorder system.480 . phase margin. and do the following.ls + 0.58) () () . 839. and determine the stable range of K by using phase margin and gain margin.s + O. (d) Find out how much the gain must be changed from its nominal value if the gain margin is 40 dB. and do the following: (1) Find the value of K so that the gain margin of the system is 20 dB.2s}(1 + 0. (d) Find the gain at the phase margin of 20°. The Bode diagram of the forwardpath transfer function of a unityfeedback control system is obtained experimentally (as shown in Fig... SP34 when K = 1..( ) _ = sCI + O.5s) 10K 10K . (d) G(s) = s(1 5K(s + I) + O.5s} s(s'l +3+ 1) 8 . The forwardpath of a unityfeedback system is given by = s2(s K(s + 1)(s + 2) + 3)(sZ + 2s + 25) (a) Plot the Bode diagram. (c) Repeat part (a) if the gain is 10 times its nominal value. (2) Find the value of K so that the phase margin of the system is 45°.. Sketch or plot the Bode diagrams of the forwardpath transfer functions given in Problem 84.3 = (s+ 3) (s+ 3) Ke. K b Gs _ K(s+l) (a) G(s) = s(1 + 0. 8P42) when the forward gain K is set at its nominal value. 838. (2) Find the value of K so that the phase margin of the system is 45°. The loop transfer function of a system is given by 25(8+ I) ( )H{) Gs s =8(s+2)(82 +2s+16) Use MATLAB to plot the Bode diagrams of the system andfmd the phase margin and gain margin of the system. gaincrossover frequency. (a) Fmd the gain and phase margins of the system from the diagram as best you can read them.01s2) 8. and phasecrossover frequencies. Find the gain margin. ( e) Find the gain margin when the phase margin is 200 • 8~42. Use MATLAB to plot the Bode diagrams of the system shown in Fig... FrequencyDomain Analysis that will give the same values for M. (1) Find the value of K so that the gain margin of the system is 10 dB.ls + o. 40. (e) G(s) . (3) Find the value of K so that Mr = 1.s{1 . 837.. and the phasecrossoverJrequency for each system.4 (e) G{s) . Compare the values ofBW of the two systems.18)(1 + 0.ols2) G{s)H(s) . (b) G s .0Is2) (f) G(s) = K(I + 0.
1'~ 80 100 120 0.. Find the gain. (c) Find the marginal value of the forwardpath gain for system stability.. (a) Fino the gain and phase margins if [he gain is four times its nominal value. The gain is set at nominal. so that the forwardpath transfer function is mUltiplied by e.1 1... ~ 0 40 60 "'r.. 60 40 .01 ~l'o "" ~ r 1. 8P42 for the fonowing parts. 481 (f) Find the steadystate error of the system if the reference input to the system is a unitstep function.... 20 @ 0 r" I..i"'10 r.0 ro (rad/sec) 10 100 1000 45 :. (h) With the gain set at nominal... (g) The forward path now has a pure time delay of Ttl sec.01 0 ~ 'i'o "to. 135 ~ 180 1\[\ 0.0 ru(rad/sec) 10 100  1000 Figure 8P42 843..r.... (d) Find out how much the gain must be changed from its nominal value if the phase margin is 60°. (b) Find out how much the gain must be changed from its nominal value if the gain margin is 20 dB.rgin for Ttl = 0.... Repeat Problem 842 using Fig. 0.. 1' i""' i'~ I' '§' .Problems . (e) Find the steadystate error of the system if the reference input is a unitstep function and the gain is twice its nominal value.. ~ 90  ~ _. (0 Find the steadystate error of the system if the reference input is a unitstep function and the gain is 20 times its nominal value...1 225 270 0.E.. find the maximum time delay Td the system can tolerate without going into instability.rri""1' :s 20 '§' . ..1 sec..and phasecrossover frequencies.TdS • Find the gain margin and the phase ma.. .
482 Chapter B. + (BL + B".01 . = 0 . 845..(Bm + B) + J".6. B = 0.2.25.6667 x 1065 + 4. (h) With the gain set at 10 times its nominal .1 sec. Motorrobotann Figure 8P49 .2s)(1 +O . 849.087 X 10 8 K (s) = s(s3 +423. as shown in Fig. using the information found from the Bode plot.4252+ 2. s(s + 4)(s + 10) (b) Plot the Bode diagram of the system. 844.(BL + B)J s2 +[BLB". The gain is set at its nominal value. 8PA9.81. = 16000(s + l)(s+5) s(s + O. find the maximum time delay 1:1 the system can tolerate without going into instability. The forwardpath transfer function of a unityfeedback control system is O 1s G(s)H(s) = (a) Draw the Nyquist plot of the system.. 3Pll is G 6.4. = K(s + 1.1". 846. K = 100. FrequencyDomain Analysis (g) The system now has a pure time delay so that the forwardpath transfer function is multiplied by e. and B". The system parameters are K" = 65. The transfer function between the output position 0ds ) and the motor current f . Determine the values of K and w at the points where the root loci cross the jwaxis. and determine the gain margin and phase margin of the system.2342 x 108) Plot the Bode diagram of G(jw) with K = 1. Repeat Problem 845 for the following transfer fu nc tion: G(s) 847.5)(s + 2) 2? of (s· + 2s + 2) Repeat Problem 845 for the following transfer function: G s)H s ( () 848. = 0. = 0. (b) Construct the root loci of the system for K ~ O.l)(s + 8)(s + 20)(s + 50) The forwardpath transfer function of the dcmotor control system described in Fig. 4P53 is _ ) G I' (S  etCs) _ K.0 Is )2 (a) Construct the Bode and Nyquist plots of G(jw)/K and determine the· range of K for system stability.(Bs + K) () fa S A u() where ~II(S) = s{ hims' + [ir.(s) of the robot arm modeled in Fig. = 0. (c) Find the phase margin and gain margin of the system. h = 0.T(/s. The forwardpath transfer function of a unityfeedback control system is G(s) = K(l +0.ls) 5 2 (1 + s)(1 + 0. K. )B + (im + h )K)s + K(BL + Bm)} The arm is controlled by a closedloop system.2. 80e. Find the critical value of K for stability. Find the gain and phase margins when Ttl = 0.
Find the gain and phase margins of the system. 8P50. 851. (a) Find the transfer function from the genT nngle (9) to the ball position (P). (I) Resonant frequency Wr (rod/sec) when K = 1.Problems (a) Derive the forwardpath transfer function G(s) = 8ds)/E(s). (h) The value of K so that the gain margin is 20 dB.015 d = O. (g) BWof the closedloop system when K = 1. (c) Find the range of K for stability. The gainphase plot of the forwardpath transfer function of G(jw) / K of a uni tyfeedback control system is shown in Fig. (b) Phasecrossover frequency (rad/sec) when K = 1. and BW. W". where M(s) is the closedloop transfer function.Om Figure 8P50 If the system is controlled by a proportional controller in a unityfeedback control system. where M(s) is the closedloop transfer function. (a) Gaincrossover frequency (rad/sec) when K = 1.8 m/s2 mass of the ball radius of the ball lever arm offset gravitational acceleration length of the beam I = 9. (d) Phase margin (deg) when K = 1.99 p X 10. Find M". (d) Plot the Bode diagram for the system for K = 1. (e) Draw IM(jw)1 versus (I). 0) Steadystate error when the reference input is a unitstep function. 8P51. (i) The value of K so that the system is marginally stable. Find the frequency of sustained oscillation in rad/sec.03m g = 9. .6 kgm 2 ball's moment of inertia ball position coordinate beam angle coordinate servo gear angle L = l. 850. (b) Find the cIosed~loop transfer function. . (b) Draw the Bode diagram of G(j'w). assume the following: III = 0. For the ballandbeam system described in Problem 411 and shown in Fig.11 kg r = 0. 483 (e) Draw IM(jcu)1 versus w. Find the following performance characteristics of the system. Find Mr. (c) Gain margin (dB) when K = 1. and BW. CUr. and find the gain and phase margins of the system. (e) Resonance peak Mr when K = 1.
. Then plot the Bode diagram of the closedloop system._____""J 270..Ols) (b) G(s) s(1 +0..5(s + 1) (a) G(s) = s(s + 1)(1 + O.O. polar plots.5s2) 50 ...I.::*""::I+:IH 10.ls 0.2s)(1 +s+0..Ols) = 82 + 2$ + 2 tOe.+. and gainphase plots.5s) = = (e) O(s) (g) G(s) = s(s + 1)(1 + 0.0 202.~)eTdS.. Chapter 8.00 H/+.0 157.r~__r_rr"T""""..:::.5 225.5s2) 1 (s + 1) (d) G(s) s(l + 8)(1 + O.ls)e. so that the forwardpath transfer function becomes G(.I_ _...._ _1. For the system in Problem 844. plot the Nichols chart and find magnitudes and phase angles of the c1osedIoop frequency response.. gain margin.0 Phase (deg) Figure BP51 852.. and compute the phase margin. the system now has a pure time delay of 1:1 in the forward path.00 .00 to.5 135.L..2s){1 + 0. . 20.....__ o 4 20..".0 247..00 L. 1 + O.5 90..00 $ S 4) 'Q 'i ~ ~ a 0..._ _. 8p·51..0 112....... 8·54._I.. For the gainphase plot of G(jw)/K shown in Fig.5 180. Use ACSYS or MATLAB to analyze the frequency response of the following unityfeedback control systems..5s) (e) G(5) = s(1 + O.b (f) G(S) ::::: s(s + 1)(1 + O. FrequencyDomain Analysis 30...484 . 853.O•1s 8·55... Plot the Bode diagrams.. and BW._L. Mr.. Repeat parts (a) through (g) of Problem 851 when K == 10. (1 + O.OO k~+==+===~=~:r. Repeat part (h) for gain margin = 40 dB..' _ _.
I 1 + TdS + TdS2 / 2 and repeat part (a). find Ttf so that the phase margin is 40°. 8P58. 8P61 shows the pitch controller system for an aircraft. 857. 856. (4224)] eT. Repeat Problem 858 wilh Td = 1 sec. lind the maximum value of Ttf so that the system will remain stable. R(s) VI s) + Figure BP5B (a) Plot the Bode diagram of G(s) = Y(s) / E(s ). Repeat Problem 855 with K = 10. as described in Problem 412 . Plot the IS~ ( jw ) Iversusw plot for the system described in Problem 849 for K = 1.Td ·1. The transfer function of the process is I Gp(s) = (1 + 10$)(1 + 25s) The time delay Ttl is 2 sec. 861.Is ~ 1  . Find the gain margin and the phase margin. Fig.1 + T"s/2 Td S / 2 859. The block diagram of a furnacecontrol system is shown in Fig. Figure BP61 . Comment on the accuracy of the approxi mation.Problems (a) With K = 1. Repeat Problem 855 so that the gain margin is 5 dB when K = I. 860. Find the frequency at which the sensitivity is maximum and the value of the maximum sensitivity. 485 858. (4223)] e. and find the gaincrossover and phasecrossover frequencies . (b) Approximate the time delay by [Eq. (b) With K = I. What is the maximum frequency below which the polynomial approximation is accurate') (c) Repeal part (b) for approximating the time delay term by [Eq.
Find Mr.. . and find the gain and phase margins of the (e) Draw IM(jw)l versus w. Chapter 8. w" and BW. where M(s) is the closedloop transfer function.486 . FrequencyDomain Analysis If the system is controlled by a proportional controller in a unity~feedback control system. (c) Find the range of K for stability. (a) Find the transfer function between pitch angle and elevator deflection angle.. (b) Find the closedloop transfer function. = 1.. (d) Plot the Bode diagram for the system for K system.
Other specifications such as maximum overshoot. relative to how it is connected to the controlled process. transientresponse characteristics. are used specifically for timedomain design. 91. Determine the controller or compensator configuration. Starting with the controlled process such as that shown by the block diagram in Fig. For instance. These design tasks are explored further in the following sections. and the design to meet a certain requirement is more conveniently carried out in the time domain. there are simple analytical relationships between some of these timedomain and frequencydomain specifications. therefore. and Nichols chart. 3. control system design involves the following three steps: 1. 2.· CHAPTER 9 Design of Control Systems 91 INTRODUCTION All the foundations of analysis that we have laid in the preceding chapters led to the ultimate goal of design of control systems. steadystate accuracy is often specified with respect to a step input. a ramp input. and frequencyresponse characteristics. or a parabolic input. rise time. We have shown that. Determine what the system should do and how to do it (design. As pointed out earlier. for higherorder systems. which should be used in conjunction with such tools as the Bode pIOl. and settling time are all defined for a unitstep input and. robustness. These are typical frequencydomain specifications. In some applications there may be additional specifications on sensitivity to parameter variations. correlations between timedomain and frequencydomain specifications are difficult to establish. or disturbance rejection. that is. These specifications are unique to each individual application and often include specifications about relative stability. 911 Design Specifications We often use design specifications to describe what the system should do and how it is done. the analysis and ____ u_(t)__~ Control vector y(t) CONTROLLED PROCESS G p Controlled variables '' (output vector) Figure 91 Controlled process. Determine the parameter values of the controller to achieve the design goals. and Mr. phase margin. The design of linear control systems can be carried out in either the time domain or the frequency domain. steadystate accuracy (error). for a secondorder prototype system. However. 487 . pular plul. gainphase plot. We have learned that relative stability is also measured in terms of gain margin. specifications).
we present the Automatic Control Systems software package (ACSYS)it is easy to use and fully graphics based to eliminate the user's need to write code. For example. such as MATLAB. To an inexperienced designer... phase margin. Nyquist plot. Therefore.1 may bring in system performance. For instance. Now with MATLAB. General design procedures using timedomain specifications are difficult to establish for systems with an order higher than the second. the design of linear control systems was developed with a wealth of graphical tools such as the Bode plot. we have incorporated small MATLAB toolboxes to help your understanding of the examples. Design in the time domain using such performance specifications as rise time. at the end of the chapter.01 sec.488. however. delay time. timedomain specifications such as maximum overshoot. rise time. 2. The development and availability of highpowered and userfriendly computer software. it is generally difficult (except for an experienced designer) to select a meaningful set of frequencydomain speCifications that will correspond to the desired timedomain performance requirements. Finally. 1. specifying a phase margin of 60° would be meaningless unless we know that it cOlTesponds to a certain maximum overshoot. M" and the like. does a gain margin of 20 dB guarantee a maximum overshoot of less than 10%? To a designer it. the designer can carry out designs using frequencydomain specifications such as gain margin. design procedures in the frequency domain are available to reduce the trialanderror effort to a minimum. the choice of whether the design should be conducted in the time domain or the frequency domain depends often on the preference of the designer. It is less obvious what. a phase margin of 60° and an Mr of less than 1. is rapidly changing the practice of control system design. The advantage of these tools is that they can all be sketched by following approximation methods without detailed plotting. establishing an intelligent set of frequencydomain specifications becomes a trial~anderror process that precedes the actual design. As it turns out. This diminishes considerably the historical edge of the frequencydomain design. settling time. and Nichols chart. which are all carried out in the frequency domain. makes more sense to specify. for example. the designer can go through a large numher of design runs using the timedomain specifications within a matter of minutes. Throughout the chapter. We should be quick to point out. which until recently had been dictated by historical development. gainphase plot. that the maximum overshoot should be less than 5% and a settling time less than 0. and the like is possible analytically only for secondorder systems or for systems that can be approximated by secondorder systems. for example. usually one has to specify at least phase margin and M. and settling time are usually used as the final measure of system perfonnance. Highorder systems do not generally pose any particular problem. it is difficult to comprehend the physical connection between frequencydomain specifications such as gain and phase margins and resonance peak to actual system performance. Historically. Design of Control Systems design of control systems is pretty much an exercise of selecting from several alternative methods for solving the same problem. The following outline will hopefully clarify and explain the choices and reasons for using timedomain versus frequencydomain specifications. Thus. and.. which often is also a . For certain types of controllers. Eventually. Chapter 9. that in most cases. maximum overshoot. to control maximum overshoot. which is based on the convenience of performing graphical design manually.
and the scheme is known as state feedback. • Seriesfeedback compensation: Fig. The compensation schemes shown in Figs. the maximum overshoot of the step response may still be excessive because of the zeros of the closedloop transfer function. The design objective is to have the controlled variables. . the large number of state variables involved would require a large number of transducers to sense the state variables for feedback. and an observer or estimator may be necessary to create the estimated state variables from measurements of the output variables. behave in certain desirable ways. Or if the roots of the characteristic equation are selected to provide a certain amount of relative damping. The disadvantage with a onedegreeoffreedom controller is that the performance criteria that can be realized are limited. and. • Feedback compensation: In Fig. Most of the conventional design methods in control systems rely on the socalled fixedconfiguration design in that the designer at the outset decides the basic configuration of the overall designed system and decides where the controller is to be positioned relative to the controlled process. the actual implementation of the statefeedback control scheme may be costly or impractical. • Series (cascade) compensation: Fig. for highorder systems. 92(b). These are described briefly as follows. and (c) all have one degree of freedom in that there is only one controller in each system. Therefore. the dynamics of a linear controlled process can be represented by the block diagram shown in Fig. and (0 all have two degrees of freedom. 92 illustrates several commonly used system configurations with controller compensation. so that the methods can be easily compared. The problem then involves the design of the elements of the controller. 92( d) shows the seriesfeedback compensation for which a series controller and a feedback controller are used. The problem essentially involves the determination of the control signal u(t) over the prescribed time interval so that the design objectives are all satisfied. Even for loworder systems. 92(d). Because most control efforts involve the modification or compensation of the systemperfonnance characteristics. 92(a). if a system is to be designed to achieve a certain amount of relative stability. it may have poor sensitivity to parameter variations. most important. (e). 92(c) shows a system that generates the control signal by feeding back the state variables through constant real gains. The problem with statefeedback control is that. they offer another perspective to the design process. The compensation schemes shown in Figs. the controller is placed in the minor feedback path. Thus. even though the controller may have more than one parameter that can be varied. and the scheme is called feedback compensation. represented by the output vector y(t). 9~12 Controller Configurations In general. For example. Fig. 91. 92(a) shows the most commonly used system configuration with the controller placed in series with the controlled process. frequencydomain methods are still valuable in interpreting noise rejection and sensitivity properties of the system. However. the general design using fixed configuration is also called compensation. and the configuration is referred to as series or cascade compensation. • Statefeedback compensation: Fig. (b).91 Introduction 489 trialanderror effort. in this chapter the design techniques in the time domain and the frequency domain are treated side by side. often not all the state variables are directly accessible.
. (e) FOlward compensation with series compensation (two degrees of freedom). the feedforward controller Gel (s) is placed in parallel with the forward path. (f) Feedforward compensation (two degrees of freedom)./.. (c) Statefeedback control.490 Chapter 9. (b) Feedback compensation. 92(f). 92(e).. the feedforward controller Gel (s) is placed in series with the closedloop system. G. which has a controller Gc (s) in the forward path. (d) Seriesfeedback compensation (two degrees of freedo m). CONTROLLER Gp(s) x(t) yet ) C K S~ate feedback (c) CONTROLLER GeCs) CONTROLLED PROCESS Gp(s) y(t) CONTROLLER G. In Fig. (a) Series or cascade compensation.+1 CONTROLLER 1 . • Feedforward compensation: Figs. The key to the feedforward compensation is that the controller G</ (s) is not in . Design of Control Systems CONTROLLER Ge(s) u(r) CONTROLLED PROCESS Gp(s) y(t) CONTROLLED PROCESS Gp(s) y(r) CONTROLLER (aJ Ge(s) (b) State variables ._.s) (d) r(t) CONTROLLER GejCS) CONTROLLER Ge(s) U(I) CONTROLLED PROCESS Gp(s) yet) (el . In Fig. 92(e) and Cf) show the socalled feedforward compensation./s) y(r) ret) CONTROLLER Ge(s) CONTROLLED PROCESS Gp(s) (t) Figure 92 Various controller configurations in controlsystem compensation.
: 491 the loop of the system. The names of these controllers come from properties of their respective frequencydomain characteristics. and it entails as much art as it does science. In carrying out the design either in the time domain or the frequency domain.. Based on this information. It should be pointed out that these compensation schemes are by no means exhaustive. For example. in which case the controllers are all digital~ with the necessary interfacings and signal converters. Because these signal components are easily realized and visualized in the time domain. with proper selection of its element values. The types of controllers available for controlsystem design are bounded only by one's imagination. In most cases. By understanding that confidence comes only through experience.and frequencydomain viewpoints. Frequencydomain . The details of these compensation schemes will be discussed in later sections of this chapter. These parameter values are typically the coefficients of one or more transfer functions making up the controller. controller parameters are selected so that all design specifications are met. these controllers are often designed using frequencydomain concepts. will satisfy all the design specifications. Keep in mind that timedomain design usually relies heavily on the splane and the root loci. this chapter provides guided experiences that illustrate the basic elements of control system designs. but in the process of varying another parameter value in an attempt to meet the risetime requirement. One of the commonly used controllers in the compensation schemes just described is a PID controller. more often than not it involves many design iterations since controller parameters usually internet with each other and influence design specifications in conflicting ways. you may initially find it difficult to make intelligent choices of controllers with confidence. the more complex a controller is. Although the systems illustrated in Fig. As a novice. which applies a signal to the process that is proportional to the actuating signal in addition to adding integral and derivative of the actuating signal.. In addition to the PIDtype controllers. After a controller is chosen. The poles and zeros of Gel (s) may be selected to add or cancel the poles and zeros of the closedloop system transfer function. While this process is sometimes straightforward. lag. the more it costs. the more complicated the design process becomes. 9~13 Fundamental Principles of Design After a controller configuration is chosen. The basic design approach is to use the analysis tools discussed in the previous chapters to determine how individual parameter values influence the design specifications and. it is important to establish some basic guidelines or design rules. leadlag. Thus. and the more difficult it is to design. the less reliable it is. Choosing a specific controller for a specific application is often based on the designer's past experience and sometimes intuition. however. lead. and notch controllers are also frequently used. Despite these design tendencies. the same configurations can be applied to discretedata control. the more design specifications there are and the more controller parameters there are.. the designer must choose a controller type that. the next task is to choose controller parameter values. so it does not affect the roots of the characteristic equation of the original system. both methods will be used extensively in this chapter. all control system designs will benefit by viewing the resulting design from both time.91 Introduction . a particular parameter value may be chosen so that the maximum overshoot is satisfied. 9~2 are all for continuousdata control. PID controllers are commonly designed using timedomain methods. system performance. As a result. finally. the maximum overshoot specification may no longer be met! Clearly. Engineering practice usually dictates that one choose the simplest controller that meets all the design specifications.
and their applications require an understanding of the basics of these elements. 1. Therefore. ~. 2. Design of Control Systems design is based on manipulating the gain and phase of the loop transfer function so that the specifications are met. zeros of the closedloop transfer function may cause overshoot even if the system is overdamped. one of the bestknown controllers used in practice is the PID controller. Transients due to those poles farther to the left decay faster. Complexconjugate poles of the closedloop transfer function lead to a step response that is underdamped. attenuators. we can consider a more general continuous. While this can be justified analytically. where the letters stand for proportional. However. Rise time and bandwidth are inversely proportional. Similarly. 6. When a pole and zero of a system transfer function nearly cancel each other. Intuitively. but it uses more fuel than an average car. If all system poles are real. and how they are connected. differentiators.. the step response is overdamped. in addition to the proportional operation. amplifiers. 93 shows the block diagram of a feedback control system that arbitrarily has a secondorder prototype process with the transfer function (91) . To gain an understanding of this con troller. Fig. in what proportion. 4. 92 DESIGN WITH THE PD CONTROLLER In all the examples of control systems we have discussed thus far. one should also be able to use the derivative or integral of the input signal. 3. This type of control action is formally known as proportional control. and integrators. and lower Mr will improve damping. In general~ it is useful to summarize the timedomain and frequencydomain characteristics so that they can be used as guidelines for design purposes. a sports car can accelerate faster. because the control signal at the output of the controller is simply related to the input of the controller by a proportional constant. The response of a system is dominated by those poles closest to the origin in the splane. Timedomain and frequencydomain specifications are loosely assodaled with each other. it is obvious that striking a nail harder with a hammer drives the nail in faster but requires more energy per strike. The farther to the left in the splane the system's dominant poles are. the controller has been typically a simple amplifier with a constant gain K. we consider just the PD portion of the controller first.492 . the more expensive it will be and the larger its internal signals will be. 5. For example. the faster the system will respond and the greater its bandwidth will be. the portion of the system response associated with the pole will have a small magnitude. integral~ and derivative. The farther to the left in the splane the system's dominant poles are.data controller to be one that contains such components as adders (addition or subtraction). larger gain margin. Chapter 9. The integral and derivative components of the PID controller have individual performance implications. Larger phase margin. The designer's task is to determine which of these components should be used..
two electroniccircuit realizations of the PD controller are shown in Fig.. respectively.92 Design with the PO Controller 493 R(s) E(s) Kp U(s) llJII S(S 2 Y(S) + + + 2tw. The series controller is a proportionalderivative (PD) type with the transfer function (92) Thus. the conu'ol signal applied to the process is u(t) = Kp e(t) + KDdt de Ct) (93) where Kp and KD are the proportional and derivative constants. The transfer func tion of the circuit in Fig.) Cp(s) + ' KDs G c(s) Figure 93 Control system with PO controller. 94. Using the components given in Table 44. 94(a) is (94) E" (a) r< R R + (b) Figure 94 Opamp circuit realization of the PD controller. .
The large overshoot and subsequent oscillations in the output y(t) are due to the excessive amount of torque developed by the motor and the lack of damping during this time interval. 9.5(b) and (c). respectively. 2. = Kp/Kf) to the forwardpath transfer G(s) = yes) = Gc(s)G E(s) P (s) = (J)~(Kp + KDS) s(s + 2~wn) (98) function. 94(b) is (96) Comparing Eq_ (92) with Eq. the circuit does not allow the independent selection of Kp and KD because they are commonly dependent on R2 • An important concern of the PD controller is that. For the sake of illustration. we have (97) The advantage with the circuit in Fig. we assume that the system contains a motor of some kind with its torque proportional to e(t). l~ During the time interval 0 < t < tl: The error signal e(t) is positive. (9~6). 921 TimeDomain Interpretation of PO Control The effect of the PD control on the transient response of a control system can be investigated by referring to the time responses shown in Fig. a large capacitor C) would be required. and its time derivative de(t)/dt are shown in Figs. these issues are of the utmost importance in practice. which is the difference between the unitstep input and the outputy(t). The forwardpath transfer function of the compensated system is • PO control adds a simple zero at . thus resulting in a realistic value for Cli' Although the scope of this text does not include all the practical issues involved in controller transfer function implementation. and the corresponding motor torque is negative. Design of Control Systems Comparing Eq. However. 95(a)~ which has a relatively high maximum overshoot and is rather oscillatory. The circuit in Fig. . The motor torque is positive and rising rapidly. 94(b) allows Kp and KD to be independently controlled.494 ~ Chapter 9. The overshoot and oscillation characteristics are also reflected in e( t) and de( t) / dt. 95. (9~4). if the value of KD is large. which shows that the PD control is equivalent to adding a simple zero at s = KpJ KD to the forwardpath transfer function. 94(a) is that only two opamps are used. This negative torque tends to slow down the output acceleration and eventually causes the direction of the output y(t) to reverse and undershoot. The performance of the system with proportional control is analyzed as follows. (92) with Eq. The corresponding error signal.r. we have Kp = Rz/R1 Kn = RZCl (95) The transfer function of the circuit in Fig. During the time interval t1 < t < t3: The error signal e(t) is negative. A large KD can be compensated by choosing a large value for Rd. Let us assume that the unitstep response of a stable system with only proportional control is as shown in Fig.
showing the effect of derivative control. Because the control signal of the PD control is given by Eq..~~~~~~~~~~ (e) figure 9·5 Waveforms of y(t). the error amplitude is reduced with each oscillation. Considering the above analysis of the system time response. and the output eventually settles to its final value. which is now in the positive direction. 2. Therefore. (9·3). The positive correcting torque in the interval 0 < t < t} is too large. Fig. 95(c) shows the following effects provided by the PD controller: . The PD control described by Eq. a logical approach would be to 1. we can say that the contributing factors to the high overshoot are as follows: 1. Dw"illg the time interval t3 < t < ts: The motor torque is again positive~ thus tending to reduce the undershoot in the response caused by the negative torque in the previous lime interval. without significantly increasing the rise time. The retarding torque in the time interval tt < t < t2 is inadequate. (9·2) gives precisely the compensation effect required. should be increased to improve the undershoot of y(t). 495 (a) e(t) 1 o~~~~+~~~~~~~~~=~~ e(l) I (b) de(t) Tt O~~~~~~. (b) Error signal. 3. e(t). and de(t)ldt. the negative corrective torque in < t < t3 should be reduced. Because the system is assumed to be stable. t2 Increase the retarding torque during tl < t < t2. during the time interval.92 Design with the PO Controller . (c) Time rate of change of the error signal. Similarly. to reduce the overshoot in the step response. (a) Unitstep response. Decrease the amount of positive correcting torque during 0 < t < tl' 2. and the retarding torque during t3 < t < t4. f2 < t < '4.
the accentuates highfrequency magnitude characteristics of the PD controller push the gaincrossover frequency to a nOIse. in linear systems. If the steadystate error of a system is constant with steadystate error only if the respect to time. 96.fthe phase margin is realized at the new gaincrossoverfrequency. = Kp +KDS = Kp( 1 + ~~ s) (99) so that it is more easily interpreted on the Bode plot. Thus. all these effects will result in smaller overshoots and undershoots in y(t). That is. a torque is again developed in proportion to de(t)/ dt. both e(t) and de(t) / dt are negative. the PD control is essentially an anticipatorycontroI. Unfortunately. The practical disadvantage of tht: PD controller is that the differentiator portion is a highpass filter. Thus. derivative control affects the steadystate error of a system only if the will have an effect on a steadystate error varies with time. so that the zerofrequency gain of the PD controller can be regarded as unity. control in the frequency domain are that t due to its highpass characteristics. The derivative control measures the instantaneous slope of e(t). (99) is shown in Fig. which usually accentuates any highfrequency noise that enters at the input. Other apparent effects of the PO the step response. e(t) and de(t)/ cit have opposite signs. which means that the negative retarding torque developed will be greater than that with only proportional control. which reduces the magnitude of the error. . The highpass filter characteristics of the PO • The PD controller has the controller are clearly shown by the Bode plot in Fig. The phaselead property may disadvantage that it be utilized to improve the phase margin of a control system. 2. Another way oflooking at the derivative control is that since de(t)/ cIt represents the slope of e(t). For 12 < t < t3. if the slope of e(t) or y(t) due to a step input is large~ a high overshoot will subsequently occur. the negative torque that originally contributes to the undershoot is reduced also. higher value. For 0 < t < tIt de(O/tit is negative. the transfer function of the PO controller is written Gc(s) • The PD controller is a highpass filter. controller provides no input to the process. w = Kp/KD. In general. predicts the large overshoot ahead of time~ and makes a proper corrective effort before the excessive overshoot actually occurs. 96 with Kp = 1. Therefore. For a given • The PD controller will system. Design of Control Systems 1. this will reduce the original torque developed due to e(l) alone. the controller can anticipate direction of the error and use it to better control the process. and the derivative portion of the error varies with time. 922 FrequencyDomain Interpretation of PO Control For frequencydomain design. Eq. For tl < t < t2. • PD is essentially an anticipatory control. such that an effective improvement o. the proportionalcontrol gain Kp can be combined with a series gain of the system. the de:~ign principle of the PD contrc)ller involves the placing of the corner frequency of the controller.496 Chapter 9. there is a range of values of Kp/KD that is optimal for improving the damping generally increase the BW of the system. Normally. in most cases it will increase the BW of the system and reduce the rise time of the step response. by knowing the slope. (98) also clearly shows that the PD control does not alter the system type that govems the steadystate error of a unityfeedback system. But if the steadystate error increases with time. the time derivative of this error is zero. Another practical consideration in selecting the values of Kp and Kn is and reduce the rise time of in the physical implementation of the PD controller. • Derivative or PD control Intuitively. The Bode plot of Eq. 3.
a properly designed PD controller can affect the performance of a control system in the following ways: 1.IKplKD o ~  . ... 2. Possibly requiring a relatively large capacitor in circuit implementation.lKplKD o  .. v . Improving GM... PM.Kp = 1.. Improving damping and reducing maximum overshoot..... Kp 923 Summary of Effects of PO Control Though it is not effective with lightly damped or initially unstable systems.I  ~ 'bO Gl 60 ~ .:> 45 \!:)'" / ..  ~. ~V V' ro(radlsec) Figure 96 Bode diagram of 1 + KDS . and frequency~domain The following example illustrates the effects of the PD controller on the timedomain responses of a secondorder system. Reducing rise time and settling time. 3. ...... 4.. Increasing BW.0 (radlsec) 90 75 ".92 Design with the PD Controller 20 ~ 497 yV 18 16 14 ~ ~ V' // / . Possibly accentuating noise at higher frequencies. // '§ :.. and Mr..r 30 15 : /" ~ O.."..../ S 12 10 . 1' ~~ V /" KplKo (.V /~ ~ ~'" S 6 /' 4 2 O..... 6. S.
99. KD= 1. and the maximum overshoot is 52. 815. (99) and K = 181. (5132) and is repeated here: G{s)  .2. with this value of K. the forwardpath transfer function of the system becomes G(s) = ®y(s) E>e (s) The closed~loop = 815. The forward~path transfer function of the system is given in Eq. K should be set at 181. 265KD)S + 815. However. ~ TimeDomain Design With the PD controller ofEq.498 ~ Chapter 9. bode (tf(num. 529.2) 4500K (9~10) Let us set the performance specifications as follows: Steadystate error due to unitramp input Maximum overshoot Rise time tr ~ :5 0. Design of Control Systems Toolbox 921 The Bode diagram for Fig. 96 is obtained by the following sequence of MATLAB functions %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% close all: clear all. as shown by the unit~step response in Fig. den =[1].. 531 and again in Fig. = so sG(s) = lIm .2) (911) transfer function is E>\'(s) E)I'(s) = 815. num = [KD KP] . the damping ratio of the system is 0. clef %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\ KP= 1. 265Kp 3 1. 265(Kp + KDS) s2 + (361. EXAMPLE 921 Let us reconsider the secondorder model of the aircraft attitude control system shown in Fig..17.2 6 = 2257.s(s + 361.2 + 815.005 sec To satisfy the maximum value of the specified steadystate error requirement.17. 265Kp (912) The ramperror constant is K~. den)) grid .7%.000443.000443 :5 5% 0. 265(Kp + KDS) s(s + 361.005 sec Settling time Is :5 0. Let us consider inserting a PD controller in the forward path of the system so that the damping and the maximum overshoot of the system are improved while maintaining the steady~state error due to the unitramp input at 0.1Kp (913) .
As Kv increases in value.. the roots are real and equal at 902. the characteristic equation roots are at 180.Kp / KD.001772. term in the denominator. however. 98 for Kp 0. the two characteristic equation roots move toward the real axis along a circular arc. = 361. 99 shows the unitstep responses of the closedloop system without PD control and with Kp = 1 and KD = 0. the characteristic equation in Eq. which is the coefficient of the.67.76. as the value of KD increases. ~ = 1. Eq. 265KD)S + 815. since the transient response is also affected by the zero of the transfer function at s = ... (912) no longer represents a prototype secondorder system.2 + 815." 1805. (915) gives KD 0. and the damping ratio of the closedloop system is 0. the zero will move very close to the origin and effectively cancel the pole of G(s) at s = O. Fig.2 + 361. 97.6 . by setting KD to zero. and the damping is critical.9·2 Design with the PO Controller· 499 The steadystate error due to a unitramp input is ess = l/Kv = O.2 + 815. When Kp is 0. for higherorder systems.j884. When KD is increased beyond 0.. With the PD = . from 361. the zero at s =Kp!KD may increase the overshoot when KlJ becomes very large. as KD increases.00177. (914) to examine the effect of varying Kp and KlJ.6 + j413.84 .2.2. We should quickly point out that Eq. the root contours again show the improved damping due to the PD controller. 265Kp = 815.92. 265KD The characteristic equation is written . + 451. When the value of Kv is increased.2s + 815. the transfer function in Eq. It turns out that.76 and 180. (914) with Kp = constant and KDvarying are constructed based on the polezero configuration of Gcq(s) and are shown in Fig.2 to 361. (914) is conditioned as 1 + Geq{S}= 1 + :. the two roots become real and unequal. First. The damping ratio of the system is .00177.~.2s + 815.OOO443jKp.67 and 180. and the closedloop system will not have any overshoot. for this secondorder system. (912) approaches that of a firstorder system with the pole at s = 36l. Thus.25 and KD = 0.46K D (915) which clearly shows the positive effect of KD on damping.6 + j884. When Kv is increased to 0. When KD ~ 0. If we wish to have critical damping.2 + 361.~ = Kp/KD to the closedloop transfer function Increasing the damping term. (914) becomes = . Eq. We see that. which is acceptable from the steadystate error requirement.. when Kp = 1 and KD = 0.2 + 815.00177. 265Kp = 0 (916) The root loci of the last equation as Kp varies between 0 and 00 are shown in Fig.6 . Adding a zero at .. 265KD :::.25 and Kp = 1. Eq. (9 I2) shows that the effects of the PD controller are as follows: J 1. In general. 0. the two characteristic equation roots are at 180. and the system is overdamped.2 + (361.j413. 2. We can apply the rootcontour method to the characteristic equation in Eq. 265Kp = 0 (914) We can set Kp = 1. 265Kos 0 (917) The root contours of Eq.
0076 0. and 0.00177 0..2%. which may cause highfrequency noise problems. although KD is chosen for critical damping.00177.0049 0.2 25.0025.67 Kp=O Kp=O (1 361.7 4. Table 91 gives the results on maximum overshoot. w control. Large KD corresponds to large BW.0005. the maximum overshoot is 4. the overshoot is due to the zero at s = Kp/KD of the closedloop transfer function.7 .00177.6 j884.2 0. (9 16). rise time. 0. (sec) Maximum Overshoot (%) 0. and there is also the concern of the capacitor value in the opampcircuit implementation.2 180.0151 0. (sec) 0. TABLE 91 Attributes of the UnitStep Responses of the System in Example 921 with PO Controller KD 0 0.67 Figure 97 Root loci of Eq.00125 0. In the present case. The results in Table 91 show that the perfonnance requirements are all satisfied with KD 2': 0..0025 t. and settling time for Kp = 1 and KD = 0.0013 52. Design of Control Systems jm 8 t Kp= 1 lI<~ j884.500 ~ Chapter 9.00119 0.00103 t. It should be kept in mind that KD should only be large enough to satisfy the performance requirements.0005 0.0076 0. 0.
.l PDco ~. K= 81. I 1. 1.rol. Ko varies.018 0..004 0.0.plane j884.:o.17.036 0.00 0.00 = 0.60 1. Kp=i KV".024 Time (sec) 0. rc= 181. (914) when Kp 2.012 0.40 0.001772 cr j413.032 0.008 0.76 j884. r~1 .25 and 1.92 Design with the PO Controller. 501 Kp=1 KD=O s.67 j413.76 KD =: 0.040 Figure 99 Unit· step response of the attitude control system in Fig.O.20 '=' Q.. 529 with and without PO controller./Vr  NoP ) cOlltrol.020 0.028 0.OI 177 '\ I '' [\ ~ '""  0.67 Figure 98 Root contours of Eq.80 1 I j r j.00 ~ 0.
and the settling time.. rlocus(num. (914). T = tf(numCL I denCL) .2 815265*KP] . i+l) = .2 + 451. Another analytic way of studying the effects of the parameters Kp and KD is to evaluate the performance characteristics in the parameter plane of Kp and KD • From the characteristic equation of Eq.:)). num = [1] . (914) shown in Fig.sqrt( (real(PoleData(l. PoleData(:.000443 Toolbox 923 Root contours oj Eq. den = [1 36J. . KD=l.imag(PoleData(2 .den).46KD yKp IVC ~= (918) Applying the stability requirement to Eq. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\ KP = 0. [nurnCL.0. hold on KD = le6. den) hold on KP= 0.502 Chapter 9. plot(real(PoleData(1. for system stability.25. we find that.denCL]=cloop(nurn. 98 are obtained by thefolfowing sequence ofMATLABfunctions KP=l. KD = KD +3e5. i) )"2)) . tf(num. 2+815265*KD 815265'f:KP] . 2+815265'itKD 815265*KP] .den) The general conclusion is that the PD controller decreases the maximum overshoot. rlocus (num . 97 are obtained by the following seque1lce of MATLAB functions KP= 1. we have 0. end i=60. num = [1] .real(PoleData(2.sqrt( (realCPoleDataC2. i) )"2)) .:)). KD=l. :))). (9/6) shown in Fig. i) )"2)+(imag(PoleData(1. Design of Control Systems Toolbox 922 Root loci of Eq. the rise time. i) )"2)+(imag(PoleDataC2. den = [1 361. num = [1] . rlocus(num.2 815265*KP] . num= [1].i)=pole(T). %%% for continuation of graph PoleData(l. (914). den = [1 361.:)).den).imag(PoleData(1. Kp > 0 and KD > . PoleData(2. i+1) = . den) for i = 1:1:260 den = [1 361.
i+1.2+815265*KD 815265*KP]. The phase margin of . i+1) = .den) hold on KD = le6. i) )"2)) . num= [1). ~ft%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%\ axis ( [3000 1000 1000 1000]) xaxis1 == 181. 911 shows the Bode plot of G(s) in Eq. %%% for continuation of graph FoleData(l.:)).) = . 2+815265*KD 815265*KP] . = 1000: 20: 10001 grid The boundaries of stability in the KpversusKD parameter plane are shown in Fig. The figure gives a clear picture as to how the values of Kp and K{) affect the variolls performance criteria of the system. which describes a horizontal1ine in the parameter plane. if Kv is set at 2257. for i = 1: 1: 260 den= [1 361.:))). as shown in Fig. 910.. 503 tf(num.sqrt( (real(PoleData(l.sqrt( (real(PoleData(2 ~ i) )"2)+(imag(PoleData(2.den))j hold on. [numCL.denCL]=cloop(num. For instance. and ?. and 1.imag(Poleoata(1.2 = [81526S*KD(i) 815265] .den). PoleData(:. i) )"2)+(imag(PoleData(1. KD KD +3e5. the constal1t~ luci show that the damping is increased monotonically with the increase in K D • The intersection between the constantK v locus and the constant~ locus gives the value of KD for the desired K~. Toolbox 924 Bode plot of Fig.:)). The ramperror constant Kv is given by Eq. (911) with Kp = 1 and KD = O. for i = 1: length(KD) nurn den~[1361.92 Design with the PO Controller den = (1 361. The constantdampingratio trajectory is described by Eq. grid FrequencyDomain Design Now let us carry out the design of the PD controller in the frequency domain. rlocus(nurn. 911 is obtained by the following sequence of MATLAB functions KD= [00.imag(PoleData(2. (913). PoleData(2. Fig. Fig.00050.yaxis1). 0. (918) and is a parabola.00250.5.:)). yaxisl plot(xaxisl. = end i==23. 910 illustrates the constant?. which corresponds to Kp ~ I.denCL). end axis ( [110000 180 90]).100) .00177]. 0]. bode(t£(nurn. i) )"2)) . nurn = [1] . trajectories for ~ = 0. T = t£(numCL. 910.707.xeal(PoleData(2.0076 *ones (1.den). plot(real(PoleData(1.1.0.i)=pole(T).
the uncompensated system is 22.002 0..522. and thus the relative stability is measured by the phase margin.00443 Phase margin ~ 80° Resonant peak Mr $ 1. Chapter 9. The results in Table 92 show that the gain margin is always infinite. Constatlt Kv.5 ++. This is one example where the gain margin is not an effective measure of the relative stability of the system. .05 BW ~ 2000 rad/sec The Bode plots of G(s) for Kp = 1 and KD = 0..92°. 0..00177. 9. and 0. andBWis 1669 rad/sec...003 0.~+ t> 1 5 4 3 2 r~r+T~~~. The phasecrossover frequency is always infinite in this case. and the resonant peak Mr is 2.0025 are shown in Fig. 0. Other effects of the PD control are that the BW and the gaincrossover frequency are increased. Use ACSYS component controls to reproduce the results in Table 92.5 6 +#+ t > 0.504 .1 Kp= 1 o 0. Design of Control Systems Kp '=0.. The Bode plots as well as the performance data were easily generated by using MATLAB tools.004 0.005 KD Figure 910 KrversusKD parameter plane for the attitude control system with a PD controller. t > 0. along with the timedomain attributes for comparison. When KD = 0.68°. the resonantpeakMr is 1.005. The performance measures in the frequency domain for the compensated system with these controller parameters are tabulated in Table 92.707 t< 1..001 0.. The performance requirements in the frequency domain are all satisfied. Let us give the following performance criteria: Steadystate error due to a unitramp input::.. Kv = 2257..5 7 t < 0. These values correspond to a lightly damped system.. 11.025. 0.00177. which corresponds to critical damping.001 0. the phase margin is 82..000443 0.
OO25 r.. I'~ 20 I 111111 (l)(radlsec) 1000 ~ f'\ ~ 10000 I~ 90 100 110 120 \30 ~ ~ ~ r..000 0..2 25.381 (sec) (sec) 00 00 00 22.00119 1.0013 52.265(1 + KDS) s(s + 361.X2 I. .. F.ro.22.95 2046 1.~ ~ .522 1. I r"" 1"1' "1\ " J I I J J I r r..0076 0.92° II I ..I'..5 1502 0...~ K D = .. . .2 82. & 20 10 {) "'~ ~~ ~ ~Ioo.~ TABLE 92 Frequency~Domain GM PM (deg) KD 0 0..0151 88.92 868 913.0076 0. I "" PM =82.0005 0...0005 ~ ~ ~ ~ b.00177 0. . I"':~~ ~ .00103 0.. t/K 0.r.. ~ i 140 150 160 170 r\ V'" KD =0. K!~'\ II ==O.0025 (dB) 00 Gain CO (rad/sec) MaxiJnum Overshoot (%) Mr 2.2) Characteristics of the System in Example 921 with PO Controller BW (rad/sec) 1370 1326 1669 2083 tT t... 50 "I" s fE 40 30 ~ "" ~ 'r.J I'V KD :: O PM.7 ... Kn =0..0049 0.00125 0. r..OOO5 L [\~ II I '§' C ~ ..7 4.025 0.2 0.9~2 80 Design with the PO Controller • 505 70 60 ~ ..68 46.00 77 )= 10 K[)::: KD=O 1 100 0...OOI77 D ~[( . 'I ~ \ VV ~~ \ V I'.6~~ I 180 1 10 100 co (rad/sec) lJ I' r l"'tt 1000 10000 Figure 911 Bode plot of G(s) = 815.
7~8*10"9*kd 0] . or clear all before running the following % Root contours kd=0. when K = 181.49  j906.718 X 109 KD)S + 2.3i + (1. while the two complex roots start out toward the left and eventually approach the vertical asymptotes that intersect at s = 1704.6) Toolbox 925 Root contours of Fig.005. based on the polezero configuration of Geq(s).49 + j906. OOOs + 2.506 Chapter 9.88%. The root contours of Fig. 912 are obtained by the following sequence of MATIAB functions Yvu may wish to use cic. It was shown in Chapter 5 that. the maximum overshoot of the system is 78.718 x 109 (Kp + Kv s ) (s) = s(s2 + 3408.5 X 10 K (919) The same set of timedomain specifications given in Example 921 is to be used. 7 G(s) = 2 s(s + 3408. The root contours ofEq.17 is G .6)(s + 57. one root of the characteristic equation moves from 3293.718 X 109 =0 (921) To apply the rootcontour method.204.3 toward the origin. den) = • If a system has very low damping or is unstable. nurn = [2. rlocus(num. Design of Control Systems EXAMPLE 922 Consider the thirdorder aircraft attitude control system discussed in Chapter 5 with the forwardpath transfer function given in Eq.718 x 109 KDS 1 + Cell(S) = 1 + $3 + 3408. as the value of Kf) increases. (92).000) (920) You may also use our MATLAB toolbox ACSYS to solve this problem. (5153).17.000) 1.204. The immediate assessment of the situation is that.000 + 2.718*10"9] .2. close all.3)(s + 57. the two complex roots will actual(v have reduced .718 X 109KDS (923) (922) Geq(s) = ($ + 3293.2 1204000 2.2.3s + 1. Let us attempt to meet the transient performance requirements by use of a PD controller with the transfer function given in Eq. the characteristic equation of the dosedloop system is written S3 + 3408. den [1 3408. 204.718 x 109 = 0 where 2. 912 reveal the effectiveness of the PD controller for the improvement on the relative stability of the system. 912. Notice that. The forwardpath transfer function of the system with the PD controller and K = 181. (921) as 2. we condition Eq.\'2 + 1.3$ + 1. TimeDomain Design Setting Kp = 1 arbitrarily. See Section 919.204. (921) are plotted as shown in Fig. the PO control may not be etTective in improving the stability of the system. if the value of KD is too large.
9~12 clearly show that~ if the original system has low damping or is unstable.906.49 . is near the bend of the foot contour.718 X 109 = o.204..D05 Figure 912 Root contours of s'3 + 3408.000 +2.49 +j906.3s2 + (1. where the relative damping ratio is approximately 0. .707 j2oo0 JIDDa 57. The root contours of Fig. damping while increasing the natural frequency of the system.6 11000 j2000 1'3000 KD =O. It appears that the ideal location for the two complex characteristic equation roots. the zero introduced by the PD controller may not be able to add sufficient damping or even stabilize the system.718 x 109 KD)S + 2.6 3293 3000 cr ~57 .D05 j3000 t=O. from the standpoint of relative stability.707.92 Design with the PO Controller jOJ ~ 507 K[)=O.
20 ± j1744. 531.62 ± j936. Fig.59 1438. The following conclusions are drawn on the effects of the PD controller on the thirdorder system. 805.14 61.00157 0. Rise time is improved (reduced) with the increase of K D • 3.00125 0. 2843.00093 0.0005 0. 57.0 =.01 ~ ~\ 1\ 1\ \ V f\ .00120 0. !(\ / V 'V ~ I II 0.00255 0.02 942. \f . The minimum value of the maximum overshoot. 1 f\ \ ~ ~ '\.0.00010 t. rise time..~ (sec) 0.0495 0.OO{ I VI \ I I \J \ I v 0.02 Time (sec) 0.80 tr (sec) 0.00337 0.00042 0. 11. 191. and the roots of the characteristic equation as functions of the parameter K D . .03 0.00 1608.48 ± j1296.0 KD=O '~ ! {\ fD= "KD 0.. occurs when KD is approximately 0. 1.00144 Characteristic Equation Roots 3293. 96.89.11. it does not meet the maximumovershoot requirement.37 17.0106 0.97 14.88 41.07. while the PD control does improve the damping of the system.04 Figure 913 Unit~step responses of the system in Example 92~2 with PO controller.00127 0.00026 0. settling time.85.05000 Overshoot 78.97 31. 2.508 • Chapter 9.3.05 11.0005 01 7 {~ 1. 19. Too high a value of KD will actually increase the maximum overshoot and the settling time substantially.52 1655. r KD=O.00500 0.29 ± j3404.01000 0. 913 shows the unitstep responses of the system with the PD controller for several values of K D • The conclusion is that.58 1301. 1523.00091 0.31 17.002..83.30 ± jI1583 Table 93 gives the results of maximum overshoot.33.72 ± j5032 1694.37%.00130 0.00398 0. The latter is because the damping is reduced as KD is increased indefinitely.00200 0. 2.71.6 282. Design of Control Systems tABLE 93 TimeDomain Attributes of the ThirdOrder System in Example 922 % Maximum with PO Controller KD 0 0.60 ± j946.00100 0.49 ± j906.00080 0.
Mr 80 = 7.... 914 shows the Bode plot for Kp = 1 and KD = O..~I\ (KD=O) ~I' P ( K =O. I I I '" ~ I'i' 1'0....~ IIII \ ~ I V l . 40 20 ~~ I 111111 r (KD =O.05 S $ ~ 0 I' r. 3..~ V ....62 60 ~ ...~" 250 " ~~ 102 ru (rad/sec) )031621 lOS Figure 9. ~ .. ~V 1'1' I.....t i> 150 180 '§' t.1 ....05) crossover ~ 20 40 60 : Gain crossovir (Kn=O) N 1 S2.. . 'r"....14 Bode diagram of G(s) of the system in Example 9~22 with PD controller.I" ". r.6 dB Phase margin = 7../10 I I 1 (KD =O. it'1' ~i' Kf)"" ?OO2 I I I '" [.i' I"/ ...~~ i. r.J 'I ~ PM =7..OO2) D M=58~~1'o N~ 200 ~I ~ ~~ KD=O ~ .oKD =O. r ~ r0....9·2 Design with the PO Controller ~ 509 FrequencyDomain Design The Bode plot of Eq..100 ~ ~D=O.... Gain crossover 1/ / "'r...... Fig.OO2) ilill .OO2 100 S r 1'\.)''\ ""'I' ~ 1'"" r 80 100 120 10° 0 v~1 I I [ KD""O ~ I I I I I\~ ~ 102 ro (rad/sec) 103 1621 50 ~ V r.r.77° Resonant Peak.....770......05 J I . 1". The following perfonnance data are obtained for the uncompensated system: Gain margin::::. (920) is used to conduct the frequencydomain design of the PD controller. )1 r. KI}~O..
and the results are shown in Table 94.75 3118. 9~6. Notice that the gain margin becomes infinite when the PD controller is added.62 35.30 1620.91 1210.62 1.300]. and the phase crossover is at infinity.42 47.83 58. This is because the phase curve of the POcompensated system stays above the 180°axis. the PD controller must provide an additional phase of 72. The frequencyMdomain performance data of the compensated system with the values of KD used in Table 93 are obtained from the Bode plots for each case. and the damping would actually be decreased. . in which case the larger KD would simply push the roots to a higher frequency..03 (rad/sec) 888.69 00 00 00 00 0 0.99 4980.00500 0.89 1. 912.002]. As a result.94 53.63 (fad/sec) 1408.34 Bandwidth BW = 1408. Because the uncompensated system with the gain set to meet the steadystate requirement is only 7.83 4789. we may run into the problem of diminishing returns. The logical way to approach this problem is to first examine how much additional phase is needed to realize a phase margin of 80°. we see that the additional phase is always accompanied by a gain in the magnitude curve.0500 00 00 00 00 00 00 00 00 00 3.00127 0. The Bode plots of some of these cases are shown in Fig.77 30.6 00 PM (deg) 7. Referring to the Bode plot of the PD controller in Fig.89 17989.00050.01000 0.23 0 • This additional phase must be placed at the gain crossover of the compensated system in order to realize a PM of 80°.94radlsec = 1103.01270.83 2604.00 Phase CO (fad/sec) 1103. for i col: length(KD) num =[2.77°. Toolbox 926 Fig. den = [13408. Thus.69 BW Gain CO M.07 1.98 1939.19 1. and the phase margin becomes the dominant measure of relative stability. 7. This symptom is parallel to the situation illustrated by the rootcontour plot in Fig. Design of Control Systems TABLE 9M4 FrequencYMDomain Characteristics of the ThirdMOrder System in Example 9M 2M2 with PO Controller Kv OM (dB) 3.den).42 11521.74 1372.00157 0.32 56. 9 13 is obtained by the following sequence of MATLAB functions M KP=lj KD= [0.12 1.83 1485. 718e9*KD(i) 2.00200 0.71 16. 914.24 1.83 rad/sec Gain crossover (OeO) = 888.0005 0.94 935.510 • Chapter 9. the gain crossover of the compensated system will be pushed to a higher frequency at which the phase of the uncompensated system would correspond to an even smaller PM. tf(num.34 7565. 718e9*KP] .21 2198.69 rad/sec Phase crossover (peO) Let us use the same set of frequencyMdomain performance requirements listed in Example 921.
denCL]=c1oop(num. den = [1 3408. end " 93 DESIGN WITH THE PI CONTROLLER We see from Section 92 that the PO controller can improve the damping and rise time of a control system at the expense of higher bandwidth and resonant frequency.0020. 04 0 2]) When KD = 0. Just as in the timedomain design. which agrees with the findings from the timedomain design that large values of KD actually decrease damping.42°.den).002. which is typically not the case for stepfunction inputs.. and the steadystate error is not affected unless it varies with time.den). The transfer function of the PI controller is (924) . bode(num. or is unstable. Toolbox 927 Bode diagram of G(s) in Example 92 in Fig. for i = 1 : length(KD) num =[2. KP=l. we have demonstrated that if the original system has very low damping. 511 [numCL.3 1204000 0] . 718e9*KD(i) 2. step(numCL. and Mr is also minimum at 1. When the value of KD is increased beyond 0. Another situation under which PD control may be ineffective is if the slope of the phase curve near the gaincrossover frequency is steep.denCL) hold on end axis ( [0 0 . the BWand the gain crossover increase continuously with the increase in KD • The frequencydomain design again shows that the PD control falls short in meeting the performance requirements imposed on the system.002.. 915 illustrates the block diagram of a prototype secondorder system with a series PI controller. 914 is obtained by the following sequence of MATLAB junctions KD= [00. hold on. Thus. However. Fig.93 Design with the PI Controller . the phase margin decreases. 718e9*KP] .05]. in which case the rapid decrease of the phase margin due to the increase of the gain crossover from the added gain of the PD controller may render the additional phase ineffective. the phase margin is at a max. The integral part ofthe PID controller produces a signal that is proportional to the time integral of the input of the controller.07. PO control may not be effective in improving the stability of the system. the PD controller may not fulfill the compensation Objectives in many situations.imum of 58. which happens to agree with the optimal value obtained in the timedomain design summarized in Table 93..
we have (926) The transfer function of the threeopamp circuit in Fig.(s) = Kp +. 916(a) is (925) Comparing Eq. (925). (a) Twoopamp circuit.512 Chapte r 9..(s) U( s) S(S W ll 2 Y(S) + 2~cu. 916(b) is (927) R R (a) RI I? EI/I CI R Hi (b) Figure 916 Opampcircuit realization of the PI K. (924) with Eq.. (924) are shown in Fi g. ~ + G. Design of Control Systems R (s) £(s) Kp + K.J) GpCs) s Figure 915 Control system with PI conlroller. G. two opampcircui t reali zations of Eq. controller. Using the circuit eleme nts given in Table 44. S (b) Threeopamp circuit. 916 . . The transfer fUllction of the twoopamp circuit in Fi g.
Adding a zero at s = KIfKp to the forwardpath transfer function. Because the PI controller is essentially a lowpass filter. 931 Time. and the latter is always zero for a stable system with a ramp input. 2. the value of Kp is important because the ramperror constant K~. The system in Fig. 513 (928) The advantage with the circuit in Fig. will now have a zero steadystate error when the reference input is a ramp function. Adding a pole at s . the steadystate error of the original system is improved by one order. in either circuit. that is. 917. it may be less stable than the original secondorder system or even become unstable if the parameters Kp and Kl are not properly chosen. Unfortunately. . should be relatively small. On the other hand. Kp no longer affects the steadystate error. for a type 0 system. In the case of a type 1 system with a PD control. The forwardpath transfer function of the compensated system is (929) Clearly. However. However.:::: 0 to the forwardpath transfer function. Similarly. effective PIcontrol designs usually result in small values of K" and thus we must again watch out for unrealistically large capacitor values. K[ is inversely proportional to the value of the capacitor..93 Design with the PI Controller Thus. and thus the magnitude of the steadystate error is inversely proportional to Kp when the input is a ramp. with the forwardpath transfer function in Eq. At first glance. 916(b) is that the values of Kp and KI are independently related to the circuit parameters. so that the transient response is satisfactory. the steadystate error due to a step input will be inversely proportional to Kp.zero configuration of the PI controller in Eq. the system may become unstable.. the immediate effects of the PI (.. Thus. the compensated system usually will have a slower rise time and longer settling time. we shall show that. it may seem that PI control will improve the steadystate error at the expense of stability. (929).~.· the values ofKp and K. A viable method of designing the PI control is to select the zero at s = KdKp so that it is relatively close to the origin and away from the most significant poles of the process. the parameters of the PI controller are related to the circuit parameters as . However. 915.. This means that the system type is increased by 1 to a type 2 system.. The problem is then to choose the proper combination of Kp and K.ontroller are as follows: 1. if the location of the zero of Gc(s) is selected properly. because the system is now of the third order. Domain Interpretation and Design of PI Control The pole. the PI control reduces it to zero (provided that the compensated system remains stable). if Kp is too large. (924) is shown in Fig. if the steadystate error to a given input is constant. When a type 1 system is con¥erted to type 2 by the PI controller. is directly proportional to Kp. both the damping and the steadystate error can be improved.
For a specified phase margin requirement. the transfer function of the PI controller is written The Bode plot of GcCjw) is shown in Fig. 2. which is detrimental to stability. The phase margin and the gain margin of the uncompensated system are determined from the Bode plot. In other words. The phase of G~M(J) is always negative.514 • Chapter 9. The Bode plot of the forwardpath transfer function Gp(s) of the uncompensated system is made with the loop gain set according to the steadystate performance requirement. set from which we have (932) . the new gaincrossover frequency (J)~ corresponding to this phase margin is found on the Bode plot. the PI controller must provide the amount of attenuation equal to the gain of the magnitude curve at the new gaincrossover frequency. The frequencydomain design procedure for the PI control to realize a given phase margin is outlined as foIIows: 1. Notice that the magnitude of GcUw) at W = 00 is 20 lOglOKp dB. The magnitude plot of the compensated transfer function must pass through the OdBaxis at this new gaincrossover frequency in order to realize the desired phase margin. (J) = K[ / Kp. 3. This attenuation may be utilized to improve the stability of the system. we should place the corner frequency of the controller. 932 FrequencyDomain Interpretation and Design of PI Control For frequencydomain design. Design of Control Systems jro splane o Figure 917 Polezero configuration of a PI controller. which represents an attenuation if the value of Kp is less than 1. To bring the magnitude curve of the uncompensated transfer function down to 0 dB at the new gaincrossover frequency w~. as far to the left as the bandwidth requirement allows. so the phaselag properties of Gc(j(J) do not degrade the achieved phase margin of the system. 918. Thus.
..9·3 Design with the PI Controller" 515 ~ ~ '" ~ "' " "'r. it is necessary only to select the proper value of K/ to complete the design.. On the other hand~ the value of K1/Kp should not be too small or the bandwidth of the system will be too low... the attenuation property of the PI controller is accompanied with a phase lag that is detrimental to the phase margin. Gc(s) = Kp + KI. causing the rise time and settling time to be too long...  ~~ :'f :s '§' .radlsec Kp 10 J ... the phase lag of the PI controller will have a negligible effect on the phase of the compensated system near Q)~. Up to this point.. . / / IOKIKp 0 15 ~ 0 30 45 60 V ". below w~.!ltl ~ . ~"'" 1'"./ V ~~ /' IOKIKp 75 90 /' V /' ~ ~ K/lOKp K/Kp ro (rad/sec) Figure 918 Bode diagram of the Pl controller.= . if the comet frequency (J) = KI/ Kp is placed far below w~.. KIf Kp should correspond to a frequency that is at least one decade.::::::t f.. although the gaincrossover frequency is altered by attenuating the magnitude of GcCjw) at w~. we have assumed that. we set Kl Wg .::.. S Once the value of Kp is determined.. K/Kp CQ(radlsec) .. as shown in Fig. This is not possible. however... 918.. the original phase is not affected by the PI controller. sometimes as much as two decades.. As a general guideline. It is apparent that.. since.. .. That is.
(930) to give the desired transfer function of the PI controller. Fus (t)/2 ~ 0.( )0p (s) (s = 4500KKp(s+ KJ/Kp) s2(s + 361. Improving damping and reducing maximum overshoot. so that the capacitor in the circuit implementation of the controller is not excessively large. 5. is more difficult than in the case of the PD controller. By computing the phase margin. 4. the selection of the value of KJ/ Kp is pretty much at the discretion of the designer. to determine. selection of a proper combination of K/ and Kp . 3. and Mr.2) You may use ACSYS to solve this problem. 2. Filtering out highfrequency noise.0 I sec ~ Settling time Is 0. The values of K/ and Kp are substituted in Eq. K p . (924). TimeDomain Design Let the timedomain performance requirements be Steadystate error due to parabolic input Maximum overshoot :$ 5% Rise time tr :::. the best value for Kp can be easily selected. Decreasing BW. Based on the preceding discussions. the value of K/ may be selected based on the ramperrorconstant requirement. Improving gain margin. gain margin. phase margin. . and then there would only be one parameter. Applying the PI controller of Eq.02 sec We have to relax the rise time and settling time requirements from those in Example 921 so that we will have a meaningful design for this system. we can summarize the advantages and disadvantages of a properly designed PI controller as the following: 1. 4.516: Chapter 9. Increasing rise time. . Mro and BW of the closedloop system with a range of values of K p . The significance of the requirement on the steadystate error due to a parabolic input is that it indirectly places a minimum requirement on the speed of the transient response. The Bode plot of the compensated system is investigated to see if the performance specifications are all met. It should be noted that in the PI controller design process.2 0. If the controlled process Gp(s) is type 0. EXAMPLE 931 Consider the secondorder attitude~control system discussed in Example 921. the forwardpath transfer function of the system becomes G( s) = G. s. The following examples will illustrate how the PI control is designed and what its effects are. who should be mindful of its effect on BW and its practical implementation by an opamp circuit. Design of Control Systems Within the general guideline.
0. With the condition in Eq.0. Apparently.2.175. (934) can be approximated by G(s) ~ 815. (938) satisfied. .2s2 + 81S. Eq. KJ/Kp should be chosen so that the following condition is satisfied: :~ ~361.00221S. the most significant pole of Gp(s).4.j17S. 97.7 < ~ < 1. Substituting K := 181.2 (938) The root loci ofEq. This should also be true for the thirdorder system with the PI controller if the value of Kd Kp satisfies Eq. Notice that.17.2.605. the value of K can be adjusted later.17 in Eq. which takes on valu~s along the operating points on the complex portion of the root loci that correspond to. the smaller K[can be. KK[ (936) Let us set K == 181.08. In fact. With K = 181. the three characteristic equation roots are at w s = 10.2 4500~Kp(s + K/ / Kp} s~(s + 361.46KKI 361. (9 39).. simply because this was the value used in Example 921.3 + j17S.. (937) with KJ / Kp ::::: 10 are shown in Fig. besides the pole at s = O.08. (936) and solving Klfor the minimum steadystate error requirement of 0. 919. If necessary. Let us place the zero at .K[ / Kp relatively close to the origin. to satisfy a given steadystate error requirement for a parabolic input. the required value of Kp for this damping ratio is 0.93 Design with the PI Controller . 175. 26SKI = 0 (937) Applying Routh's test to Eq. (938).707.2) (939) where the term KJ/ Kl' in the numerator is neglected when compared with the magnitude of s. which are for Eq. a relative damping ratio in the range of 0.) _ 1 _ 0.17. is at 361.3 .4 . 265Kps + 815. Let us assume that we wish to have a relative damping ratio of 0. and ...: 517 The parabolicerror constant is Kll = lim s2G(s) = lim s>O s>O . the root loci in Fig.707.. or the system will be unstable. For the present case. otherthan the small loop around the zero at s = 10.s(t)/2 is e. 265Kp s(s + 361. with Kp = 0. these root loci for the most part are very similar to those shown in Fig. Kc == 0. This means that the zero of G(s) at s = Kl/Kp cannot be placed too far to the left in the lefthalf s~plane. the characteristic equation of the closedloop system is S3 + 361. say.2 The steadystate error due to the parabolic input t2 u. Thus.8. Thus.2} (935) = 4500KKf = 12.Ka . we get the minimum value of K[ to be 0. (937) yields the result that the system is stable for 0< Kl/Kp < 361. (916). the larger the K.. From Eq.2. 919 show that the relative damping ratio of the two complex roots is approximately 0.08026 ( 02) < .
let us select K/ /Kp 5. we see that the zero at s = 10 is relatively close to the origin and.4 and Hlook" toward the neighborhood near the origin.145. For example.3 + jl 75. In fact.03 and the relative damping ratio is still 0.03)(s + 178.03  j17S.2($ + 5) = (s + 5. we can show that.03 .707. Design of Control Systems H ~ t jm splane Kp=O.518 ~ Chapter 9.2 Kp=O o Kp=O Kp=O.221. Although the real pole of the closedloop transfer function is moved. practically cancels one of the poles at s = O.j178. and . the relative damping ratio of the complex roots will be very close to 0. as long as Kp = O.4 Kp=O 361. For example. Kp varies.145 is very close to the zero at s = 5. and the system dynamics are essentially dominated by the two complex poles.K/ / Kp so that the transient due to the real pole is negligible. when Kp = 0. 178.03 + jI7S.03) (940) Because the pole at s = 5. thus. the three characteristic equation roots are at = s = 5. (938) is satisfied.03 + jI78.03.08 175. the c1osedloop transfer function of the compensated system is Ely{S) ares) 65.OS and the value of K/ is chosen such that Eq.3 + j17S.4. . it is very close to the zero at s = .707.178.08 and K/ = 0. The reason for this is that when we Hstand" at the root at 175. (937) with KI! Kp = 10.145)(s + 178. the transient response due to this pole is negligible.08 Figure 9·19 Root loci of Eq.
16 0.93 (sec) (sec) 0.1 '7.08 0.049 0. unless Kp is also reduced.47 4.16 9. % start with a very small KP see Figure 919 KI=lO*KP. is also plotted in the same figure as a comparison.08 0.60 0.04 and KI 0.38 0.08 0.000001.0074 0. and the settling time is 0.08.0080 0.08 Overshoot (%) 52.0084 Table 95 gives the attributes of the unit~step responses of the system with PI control for various values of KIf Kp .08~ which corresponds to a relative damping ratio of 0. since the maximum overshoots are less than 5%. num:= [KP KI] .0114 0..04 0. 920 shows the unitstep responses of the attitudecontrol system with PI control. As mentioned earlier. for practical reasons. The unitstep response of the same system with the PD control1er designed in Example 921.9·3 Design with the PI Controller .0194 0. the settling times in Table 95 actually show a sharp reduction. 1 Root loci of Eq.00 0. the rise time and settling time will be excessive.7 15.95 and 1. G=tf(num.2 815265*KP 815265*KI].023 0.0115 5 2 0. the value of the capacitor C2 is inversely proportional to K/. there is a lower limit on the value of K/.015 0.1 '7 5.[1361. which is misleading.024 seconds. = .0084 0. However.1 %. This is because the settling times for these cases are measured at the points where the response enters the band between 0. den) rlocus(G) TABLE 95 Attributes of the UnitStep Responses of the System in Example 9~31 with PI Conlroller Maximum Ir Is KJ/Kp 0 20 10 K/ D 1. Fig.61 4. 519 Toolbox 93.08 and several values of Kp.00.008 Kp 1. den=. with Kp = I and KD = 0.5 0.00135 0. The maximum overshoot of the system can still be reduced further than those shown in Table 95 by using smaller values of Kp than 0. For Kr :S I.0114 0.0182 seconds.0078 0. Thus.08 0.0084 0.0083 0.707.08 0.08 0. For the system considered. with Kp = 0. but the rise time is increased to 0. (937) in Fig.00177.08.0294 0..40 0.80 0. The results in Table 95 verify the fact that PI control reduces the overshoot but at the expense of longer rise time. with Kp = 0. the maximum overshoot is 1.08 0.04.89 4. 919 are obtained by the following sequence of MATlAB [unctions KP = 0. For example. improvement on the maximum overshoot slows down for K/less than 0. with Kp = 0.
O f I ..20] i tf(nuDl.00177 . (934). KP=KI/S i K=100j for :l = 3. Chapter 9. r. The phase margin is 22. 921. 920 is obtained by the following sequence of MATIAB functions K=10. FrequencyDomain Design The forwardpath transfer function of the uncompensated system is obtained by setting Kp = 1 and K/ 0 in the O(s) in Eq.4r. step(numCL.05 Time (sec) Figure 920 Unitstep responses of the system in Example 931 with PI control.08 . den =[1 361.I.KJ=O.denCL]=cloop(num.Kr =0. : length(KI) num = [4500*K*KP(i) 4500*K*KI(i)/KP(i)] .200].05 02]) .JKI= 1.520 .den). unitstep response of the system in Example 921 with PD controller.L 0. tf(nwu.. o o / .OO8 /.denCL1=cloop(num. [numCL. 8 ] .0 With PO controJ~ Kp / =1.03 0.o I " 1 .denCL) hold on KI = [ 1. 6 0 . = Toolbox 932 Fig. /' ~ ~. step(numCL.:.S t I ~. 1 I 'KJ = 0.04 0.:. and the Bode plot is shown in Fig. Design of Control Systems 2.68°.6 .denCL) hold on end axis([O 0.den).den). KD =0.den)i [numCL. Also.02 0. and the gaincrossover frequency is 868 rad/sec. num ::::[4500~rK] .01 0.. den= [1361.
....OO8 V~ V ~ I' I I I ~~ V ~ 1/ vI """ I K/=O.... i'. .r. ::s '§' 50 K/=O... V 10.~I'o "8~8 r.!' K[= 1.II 140 160 ISO 200 102 M V1/1I / I/' "".... G(s) = 815.6 ~~ •... K/=O PM=65" K1 .9 3 Design with the PI Controller <t1 521 M 200 150 ~ ~ ~ ~ ~r to V Kl = 1.Io~ ... r.I K/=0. J I I 100 0 102 101 101 m(radlsec) 102 170 I I 20 I I 40 60 I 1 I ( ~ C SO ::s 100 120 .I"Io r.. 21..1'1"102 M 10° 10 1 ltJ (rad/sec) 103 104 Figure 9 21 Bode plots of the control system in Example 9 31 with PI controller. ~ i ~ roe I 10 ......... 0 1I crossover K/=O 50 I Gain crossovers I I t..... ~l' " V K/=0.6 \ r.S II"" ". ....~ ~ I". l""'.. r.OS/ K/=O......• ~... 265Kp(s + KI/ Kp) s2(s + 361..::: 1..J K[::::O i' :=:> K[=0. r.... ~ ~ ~ I:"'!'~~ \!l ~ ~ i"'ooo 1'0" ..68" . i'o 100 "'r....1 I"" t'..OO 81'.5 dB Gain ~r N... ~~ PM =22.... ~ ~ r.6 .8 .08 .2) . ~ r.~ ii' ~ j......
5 dB.08. in the timedomain design conducted earlier. for i nurn = [815265*KP 8~5265*KI(i)] . This could not be just a coincidence. and 1. 921. 0. the value of K/ is not rigid~ as long as the ratio KJ/Kp is sufficiently smaller than the magnitude of the pole of G(s) at ~ 361. 0.den) hold on = 1: lengthCKI) end grid Let us specify that the required phase margin should be at least 65°. Design of Control Systems Toolbox 9·3·3 Bode plots of the control system in Example 931. 921 is obtained by the following sequence of MATLAB functions KI = [ 0 1 . Kp was selected to be 0. Can you believe that we have had no prior knowledge that. As it turns out. (Perhaps we have cheated a little by selecting the desired phase margin to be 65°.707. den =[1 361. (932). 008] . bode(num. 08 O. (933) gives the general guideline of finding K/ once Kp is determined.2. (942) As pointed out earlier.200].707 corresponds to PM = 65°?) 2. a .6 are shown in Fig. Table 96 shows the frequencydomain . Following the procedure outlined earlier in Eqs.5 dB into Eq. 921. The Bode plots of the forwardpath transfer function with Kp = 0. given by Eq.08 and K/ = 0.08 = 10 10 1 36 . (931) through (933) on the design of the PI controller. .522 . (942) is not sufficiently small for this system. From Fig. 8 o. Thus. w~ is found to be 170 rad/sec.5 dB at w~ = 170radlsec. Let us choose Kp = 0. the PI controller should provide an attenuation of 21. Look for the new gaincrossover frequency w~ at which the phase margin of 65° is realized. and this is to be achieved with the PI controller of Eq.008. Thus. K / = (J)~Kp = 170 x 0.Chapter 9. 0. Fig. in this case~ ~ = 0.8. we conduct the following steps: 1.08. we get IG( I= (941) Notice that. and solving for Kp . Eq. Substituting jW~) 21. (930). 6 KP=0. so that we can compare the design results of the frequency domain with those of the timedomain design obtained earlier.08 so that the relative damping ratio of the complex characteristic equation roots will be approximately 0.08. K=l. the value of K. The magnitude of G(jw) at this frequency is 21.
17 is (944) .26)(s + 3008) (943) We can show that the steadystate error of the system due to the parabolic input is again given by Eq. 0.08 0.01~ but BW is reduced to 117. and gaincrossover frequency all vary little.204. M". for values of KJ/ Kp that are sufficiently small. You may use ACSYS to solve this problem. 0.96 164. third~order attitude control system described by First.4 0.08 0. TimeDomain Design Let the timedomain specifications be as follows: Steadystate error due to the parabolic input t 2us (t)/2 :::.6 58.8 0.95 256. = = EXAMPLE 932 Now let us consider llsing the PI control for the Eq.73 164.3s + 1.s2(s2 + 3408.01 1.00 0. Applying the PI controller of Eq.02 sec These are identical to the specifications given for the secondorder system in Example 931. BW. and~ arbitrarily setting K = 181. the bandwidth of the system will be further 0.2 Maximum overshoot Settling time ts ~ :s..08.002215.45 61.5 x 109KKp(s + KJ/Kp) .00 BW (rad/sec) 1390.92 262. (924)~ the forwardpath transfer function of the system becomes G(s) = Gc(s)Gp{s) _ 1.38 258.13 255.70 Phase CO (rad/sec) 00 0 20 00 10 5 1 0.15 65. and M}' = 1.08 0. For example. the minimum value of K/ is 0. the timedomain design is carried out as follows. (936). for Kp 75.47 M. (9~ 19).75 65. The characteristic equation of the closedloop system with K = 181.17. Notice that.93 Design with the PI Controller ·523 TABLE 9~6 FrequencyDomain Performance Data of the System in Example 931 with PI Controller OM KdKp KJ 0 1.55 1.6 0. 5% Rise time t .3 rad/sec.87 268.77 164.03 1. the phase margin.04 and K/ 0. 2.08 (dB) oc 00 PM (deg) 22.49 OainCO (rad/sec) 868 165. However.06 1.04.5 x 109 KKp(s + Kd Kp) s2(s + 400.7°.08 0. :::.08 0. the phase margin is increased to reduced. .98 63.12 1.01 sec 0. It should be noted that the phase margin of the system can be improved further by reducing the value of Kp below 0.71 164. OOO} 1.008 Kp 1.1 00 00 00 ex:: (Xl 00 OC DC properties of the uncompensated system and the compensated system with various val ues of K ..
718 x 109 KI 3408.'1 sl 1 1.S24 .255Kp . (944) are plotted using the polezero configuration ofEq. Fig. (943). relative to the nearest pole of G(s) to the origin which is at 400.OOOKp .3408.797465K~ .000 797465Kp 0 0 2.204.718 x 0 0 109Kl sO 2. Chapter 9.718 x 109Kp 2.233. Design of Control Systems The Routhts tabulation of the last equation is perfonned as follows: S4 s3 2 .204.204.26 3008 Figure 922 (a) Root loci of the control system in Example 932 with PI controller K/ / Kp = 2: 0 :::. The root loci of Eq.718 x 109KI The stability requirements are KI>O Kp< 1.3KI 1.5098 K/ (945) < 353.204.000 797465Kp 1.3 1.26.98K~ t The design of the PI controller calls for the selection of a small value for KI/ Kp. . 922(a) shows the root loci as Kp splane ocKp Kp=O Kp =0 400.000 2. Kp < 00..
.1"..::..5 ~B r. Kp = 0..788° I"[\ K1 =O. Kp= 1 ~....15" PM:=6S.J I 1. K[ = 0. 534. .. it may be possible to satisfy the perfonnance specifications .. varies for K/ / Kp = 2. fIIoo . '.. " :.. [ [ [ "'' 100 I 101 102 101 102 163 lIJ(radlsec) o 50 K/=O...ISo I IIIII 200 I 1/ I_~ I K/=O.. i' I"~...l!tU IIIllII I I /" . By selecting the value of Kp properly along the root loci.02 I" P~=7.0 l)'" ~~~ yrrfll... ~ ~.04 = 0. 0 ~~ ~~~ ~t% ~!'< ~"'" ~ I" Gain Iii J crossovers IIJ2.6 K1 ~~ II '..UJ... "' f'.. Kp= 1 r"' ~ II N ... I~ I I '~ I I I' 50 ~~ " t\ . 250 300 102 lO1 10 1 co (rad/sec) (b) Figure 922 (b) Bode plots of the control system in Example 932 with PI control. VI f'. r:::I I I .28 Kp=O..Q. ~ ~PM=65.74· ~~ ~ ~ ~ PM ::::66.Kp= 1 100 ... ~i""1' IK. The root loci near the origin due to the pole and zero of the PI controller again fonn a small loop.. ~ r..075 ~~ i"" I'.93 Design with the PI Controller 150 001II 525 100 r..tK[= Kp ~~ = 0. s ff 150 180 K1 = Kp =0.02 I S $ '§' ~ 50 t. ~ ~ V Kp =0. which are shown in Fig..O7~ . and the root loci at a distance away from the origin will be very similar to those of the uncompensated system... K1=0.."" ~~ i I l ~ II' ~ t::"rj'.28.
7 259 5. (931) through (933) on the design of the PI controller.8 5. £v~ is found to be 163 rad/sec.1 234.0202 0. we carry out the following steps.5 dB at w~ = 163radlsec. The suggested value of K/ is found from Eq. we get JG( Kp = 10IG(jw~nlB/20 = 1022.imum t1' ls KlKp 0 20 20 5 2 5 2 2 2 K{ 0 1.0487 0.5 dB into Eq.0881 0.00158 0.572 BW = 1378rad/sec Let us require that the compensated system has a phase margin of at least 65° • and this is to be achieved with the PI controller of Eq. or KI/Kp = 2. 1.0077 0.01616 0. Substituting jW~) = 22.9 141.4 ±j178 ±jI80.2 ±j906.01212 3293.2 2 5. and the magnitude of G(jw) at this frequency is 22. the PI controllyr ShOUldlProvide an attenuation of 22. Look for the new gaincrossover frequency (J)~ at which the phase margin of 65° is realized. Mr = 6.788° FrequencyDomain Design The Bode plot ofEq.6 3035 3021.3 28 ±jI89.7 6.0085 0. 920. From Fig.5 3031.075 and K/ = 0. Kp 9~22(b).16 0.14 Kp 1 Overshoot (%) (sec) (sec) Roots of Characteristic Equation 0.2 187. the one with Kp = 0.5 dB. To minimize the rise time and settling time.9% when KI = 0.2 3033.6 240 185. Notice that.2 4.04 0.7 3021.5 22.0202 0. (943) for K = 181.6 175.075 = 1.15 0. although several combinations of these parameters correspond to systems that satisfy the perfonnance specifications.075 0.04 0.3 3035 3021.04 0.9 4.08 0.02202 0.8 ±j19I. = 1.6 0.2 0.15.0471 0. Table 97 gives the performance attributes of severa1 combinations of K.01818 0. and K/ = 0 is shown in Fig. and solving for Kp.17.222 10 (947) .0101 0.3 2 1 1 1 ±jl64 given above.8 0.08 0. (932). we should select Kp so that the dominant roots are complex conjugate.070 76.1 4.526 IL Chapter 9.5/20 = 0.0 0.5 186 187.2 ±j190.00792 0.6 15.8 5.578 dB Phase margin = 7. The performance data of the uncompensated system are as follows: Gain margin = 3.0134 0.3 99 184 149. Thus.8 57.4 0.2 3035.075 (946) This is exactly the same result that was selected for the timedomain design that resulted in a system with a maximum overshoot of 4. 2.3 2.2 15. Following the procedure outlined in Eqs.08 0. (930).00883 0.01515 0.j Kp and Kp.08 0.00787 0.08 0.15 gives the best rise and settling times among those shown. (933): K/ = w~Kp = 10 163 x 0.08 0.01796 0.7 3035. Design of Control Systems TABLE 97 Attributes of the UnitSlep Responses of the System in Example 932 with PI Conlroller Max.00917 0.
4 134. 922(b) is obtained by the following sequence of KI = [ O.270 0] ) . Reducing KJ would increase the capacitor value in the implementing circuit. TABLE 98 KdKp 0 16.~ (sec) 0.74 BW Mr 6. 5e9*KI(i)] . 01 10000 .075 case gives the best allaround performance in both the frequency domain and the time domain.04 0. only the K[ = Kp = 0. However. However.::::.020 (dB) PM (deg) 7. den) hold on end grid axis ( [ 0 .. 075 OJ . 923 shows the unitstep responses of the uncompensated system and several systems with PI control. No details are explored further.222 0.02 0.4 253. the design ramifications show the following: Reducing Kp would reduce BWand increase Af.040 0. Fig. Notice that the last three designs in the table all satisfy the PM requirements.578 25.15 66.209 (rad/sec) 1378 264. K=l.4 17.788 59..0490 0. This is one example that shows the inadequacy of specifying phase margin only.572 1.0085 0. for i = 1:1ength(KI) num = [1.15 65.52 65.4 0.2 t.1 4.3 1.67 26.006 1.600 0. Thus. bode (num .098 1.0142 0.0478 0.075 1] .0970 0. [0.16 37.3 I 15 14 0 Performance Summary of the System in Example 93~2 with PI Controller OM K/ Kp 1 0.3 12. 28 0 .06 31. den =[13408. KP.0015 0.3. the phase margin of the system with these design parameters is only 59.1616 . K/ / Kp = 16.0268 (sec) 1.0086 0.20 13.52. In fact. t.075 0.204000 0 OJ.9~3 Design with the PI Controller ~ 527 Toolbox 9·3·4 Bode plots of the control system in Example A1ATLAB junctions 93~2. To realize the desired PM of 65 0 • we can reduce the value of Kp or Kl • Table 98 gives the results of several designs with various combinations of Kp and K/. 6 0 . The purpose of this example is to bring out the properties of the PI controller and the important considerations in its design. Fig. In attempting to increase Kb the maximum overshoot becomes excessive..280 3. 5e9~'rKP(i) 1.133 1.0116 0.34 Maximum Overshoot (%) 77.075 0.6 66.075 0.
This leads to the motivation of using a PID controller so that the best features of each of the PI and PD controllers are utilized. . Equating both sides ofEq. Design of Control Systems 1. we have Kp = Kp2 + KDIK/2 KD = KDIKp2 (9~49) (950) (951) K/ = K12 .J. ..02 VI 0. 1.04 (\I\/\A __ I V 1.528 . Chapter 9. We can outline the following procedure for the design of the PID controller.K[=O.Kp2 (948) The proportional constant of the PD portion is set to unity.Kp =1 1.~ KI 0.6 . The transfer function of the PID controller is written as Gc(s) K( = Kp + KDS +"8 = (1 + KD1S) (Kl1) +. ~ J. we see that the PD controller could add damping to a system. The PI controller could improve the relative stability and improve the steadystate error at the same time. Kp = 0. Consider that the PID controller consists of a PI portion connected in cascade with a PD portion. (948).4 0.. but the steadystate response is not affected. Kp =0.o75 • K.2 Figure 923 Unitstep response of system with PI controller in Example 9~3~2 . 94 DESIGN WITH THE PID CONTROLLER From the preceding discussions. .2 V ~( .i = .8 1.. nM V \.4 LK'~ Kp~o. since we need only three parameters in the PID controller.15 0.6 ~V \ v V' "" 'iY II ~1 I v / 0.28.05 0.8 0.1 Time (sec) 0.2 1/ 0. but the rise time is increased.
(919). step(numCL. 2 .02 0.075] .718 x 10 (1 + KDIS) s(s + 400.005 sec We realize from the previous examples that these requirements cannot be fulfilled by either the PI or PD control acting alone. see Section 9~19. 5e9*KICi)].3 1204000 0 0] . the PI portion of the controller can be designed first for a portion of the requirement on relative stability. the PD portion is designed.17. ~ EXAM PLE 9~41 Consider the third~order attitude control system represented by the forwardpath transfer function given in Eq. The following example illustrates how the PIn controller is designed in the time domain and the frequency domain. for i = 1:1ength(KI) num = [1.718 X 10 G ( ) p s == s(s + 400. Se9*KPCi) 1. 0001: 0 . KP = [10.26) (s + 3008) You may use ACSYS to solve this problem. K=1. and in the frequency domain it is the phase margin. [numCL. 5% Rise time tr ~ 0.denCL]=cloop(nurn. 9 (9 52) • TimeDomain Design Let the timedomain performance specifications be as follows: Steadystate error due to a ramp input t2 us (t)/2 ~ 0.075]. t = 0 : 0 . Select the parameters KI2 and Kn so that the total requirement on relative stability Toolbox 941 Fig. den =[1 3408.denCL.den). the transfer function is 2. is satisfied. and. 8] ) As an alternative. 923 is obtained by the following sequence of MATLAB jimctions KI= [00. tf(num.den). 3. Let us apply the PD control with the transfer function (I + KDIS). The forwardpath transfer function becomes G(s) = 2. 2 0 1. ~ 529 Consider that the PD portion only is in effect. Select the value of KDl so that a portion of the desired relative stability is achieved.2 Maximum overshoot :::.040.005 sec ~ Settling time ts 0. this relative stability may be measured by the maximum overshoot.9·4 Design with the PID Controller 2. In the time domain. With K = 181.26)(s + 3008) 9 (953) .t) hold on end grid axis ( [0 0 .280.60. finally.
4 0.Apparently.4. Eq.1 16..002 and K12 = 15Kp2 = 4.5 Kp = Kp2 KD + KDIK12 = 0. (949) through (951). Chapter 9.3 . when designed properly.37%.: 0.1 4.26)(s + 3008) (954) Following the guideline of choosing a relatively small value for KI2/ Kn.1 15.1 15. Selecting Kn = 0.2 15.309 (956) = KDIKp2 = 0. Fig.1 10. (954) becomes G(s) = 5.5 0.3 94.3 0.00088 0. Next.002 x 0.9 215.5 1571. respectively.8 4.0006 Notice that the PlD design resulted in a smaller KD and a larger KI .436 x 106Kn(s + SOO)(s + 15) s2(s + 400.00747 0. t.6 0. and the forwardpath transfer function becomes G(s) = 5.2 and 0.4 263.5 533.3 15.00400 0.9 5.00111 0.5 5.00303 0. K/ = K12 = 4.00155 0. The rise time and settling time are well within the required values.5 558.5 0.. the following results are obtained for the parameters of the PID controller using Eqs.4 579. captures the advantages of both the PD and the PI controls. which correspond to smaller capacitors in the implementing circuit.00172 0.530 .8 0.8 1385.5 16.26)(s + 3008) (955) Table 99 gives the timedomain performance characteristics along with the roots of the characleristic equation for various values of K p2 .3 + 0.00404 0.3 629 1993 2355 2613 284 286.7 897.3 15.00271 0.00202 0. as well as those with PD and PI controls designed in Examples 922 and 932.436 x 106KP2(s + 500Hs + K12/ K p2) s2(s + 400.1 Table 93 shows that the best PD controller that can be obtained from the maximum overshoot standpoint is with KDI = 0. we add the PI controller.1 0.1 221.00895 0.6 6.00747 0. and the maximum overshoot is 11.3 8..002. Design of Control Systems TABLE 99 TimeDomain Performance Characteristics of ThirdOrder Attitude Control System wilh PID Controller Designed in Example 941 KP2 1.04545 15.7 0.002 x 4.00130 0. we let K12/Kp2 = 15.6 1168.2 15.00505 0.5.00303 0.00404 0.8 9.3 1430 ± j 1427 ± j 1423 ± j 1417±j 1406+ j 1382± j 700±j 519 ± j 390±j 284±j 266 ± j 1717..00303 0.1 15.s (sec) (sec) Roots of Characteristic Equation 11.0025 0.2 6.2 0. . 924 shows the unitstep responses of the system with the PlD controller.6 5.1 15. the optimal value of K p2 is in the neighborhood of between 0.08 Maximum Overshoot (%) I. and with KDJ = 0.3.2 538.00127 0.7 546.00303 0.0 0.2 4.1 470..5 = 0. Notice that the PlD control.00214 0.9 0.
925. Using this PD controller. PI. the forwardpath transfer function of the system is G(s) = 2. KD = 0.15 f 0. domain criteria corresponds to the timedomain specifications given in this problem. Phase margin ~ 70° Mr ::. K12/Kn = 15~ and .26)(s + 3008) (957) and its Bode plot is shown in Fig.0~5'K. (932). The result given in Eq.94 Design with the PID Controller <Ill 531 1. 925. and the results were tabulated in Table 93.4 I 0.2 I Y 0. KD =0.002. the value of Kp2 is calculated to be Kp2 = 107 / 20 = 0.5.718 x 109 (1 + 0. K I.2 to 0.002. Thus. but this is the best that the PD control could off