You are on page 1of 61

CH 26 Design of Digital

Controllers
l The digital controller and its design
l Discrete PID controllers from the
continuous domain
l Discrete controllers based on discrete
domain strategies
l Digital multivariable controllers

Part IVC: Computer Process


Control
26.1 Design of Digital Controllers:
Preliminary Considerations

l We assume that the dynamics of the final


control element (actuator) and the measuring
device have all been combined with the
dynamics of the process into one overall
process transfer function to be designed as
g(s)
l The input and output of this overall process
shall be designed as u(t) and y(t) with
corresponding Laplace transforms u(s) and
y(s).
Part IVC: Computer Process
Control
Preliminary Considerations
l Under digital control the continuous input
u(t) is actually a discrete signal that has
been made continuous by the hold element
l With a Zero order hold-element u(t) will be
piecewise constant input signal generated
from the discrete sequence u(k)
l The controller itself is considered as
receiving discrete error information e(k)

Part IVC: Computer Process


Control
Preliminary Considerations

Part IVC: Computer Process


Control
26.2.1 The Digital Controller
l Digital Controller
• accepts a discrete sequence of error signals e(k)
and
• computes the discrete sequence of corrective
action, u(k),
• according to a discrete control law that may be
represented
• in the time domain or
• in the z-domain

Part IVC: Computer Process


Control
The Digital Controller: Z-domain
representation
l Control law

u( z ) = gc ( z )e ( z )

l Digital controller Transfer function in its most


general form( a ratio of two polynomials in z-1)

Q(z -1 ) æç θ0 + θ1z -1 + θ2 z -2 + ... + θr z -r ö÷


g c (z ) = =
Y(z ) çè 1 + y 1z -1 +y 2 z -2 + ... + y q z -q ÷ø
-1

Part IVC: Computer Process


Control
The Digital Controller: Poles and
Zeros of the Digital Controller
This image cannot currently be displayed.

l Roots of the polynomial equation (denominator)

Y( z -1 ) = 1 +y 1 z -1 +y 2 z -2 + ... +y q z - q = 0

are the poles of the digital controller

l Roots of the polynomial equation(numerator)

( )
Q z -1 = θ0 + θ1 z -1 + θ2 z -2 + ... + θr z - r = 0

are the zeros of the digital controller


Part IVC: Computer Process
Control
The Digital Controller: Realization of
the Digital Controller
This image cannot currently be displayed.

l Digital controllers must be implemented


physically in time domain
l Time-domain realization
y ( z -1 )u ( z ) = Q( z -1 )e ( z )
(1 -y 1 z -1 +y 2 z - 2 + .... + y q z - q )u ( z ) =
(q 0 + q1 z -1 + q 2 z - 2 + .... + q r z - r )e ( z )

l Taking inverse z-transform in the usual manner


u (k ) + y 1u (k - 1) + y 2u (k - 2) + ... + y qu (k - q) =
θ0e (k ) + θ1e (k - 1) + θ2e (k - 2) + ... + θr e (k - r )
Part IVC: Computer Process
Control
The Digital Controller: Realization of
the Digital Controller

l The time domain form of the discrete


controller consists of two distinct parts
• The autogressive part depends on past
values of the control signal and are thus
directly related to the poles of gc(z)
• The moving average part involves current
and past values of the error sequence thus
directly related to the zeros of the gc(z)

Part IVC: Computer Process


Control
The Digital Controller: Physical
Realizability
This image cannot currently be displayed.

l Digital control is physically realizable if and


only if the determination of u(k)(current
control action) is on the basis of only current
and past information
l In z-domain this means that:
The digital controller is physically realizable if
and only if the highest power of z in the
numerator (Q(z-1) is less than or equal to the
highest power of z in the denominator

Part IVC: Computer Process


Control
The Digital Controller: Physical
Realizability
This image cannot currently be displayed.

l Example 26.1 a)
The digital controller pulse transfer function :
3 + 2 z -1 + 3z -2
gc ( z ) =
1 + 4 z -1
the highest power of numerator : 0
the highest power of denumerator : 0

A time domain realization :

u (k ) = -4u (k - 1) + 3e (k ) + 2e (k - 1) + 3e (k - 2)

Part IVC: Computer Process


Control
The Digital Controller: Physical
Realizability
This image cannot currently be displayed.

l Example 26.1 b)
The digital controller pulse transfer function :
z + 3 + 2 z -1 + 3z -2
GC ( z ) =
1 + 4 z -1
the highest power of numerator : 1
the highest power of denumerator : 0

A Time domain realization :

u (k ) = -4u (k - 1) + e (k + 1) + 3e (k ) + 2e (k - 1) + 3e (k - 2)

unrealizable

Part IVC: Computer Process


Control
The Digital Controller: Physical
Realizability
This image cannot currently be displayed.

l Example 26.1 c)
The digital controller pulse transfer function :
2 z -1 + 3z -2
GC ( z ) =
1 + 4 z -1
the highest power of numerator : -1
the highest power of denumerator : 0

A Time domain realization :

u (k ) = -4u (k - 1) + 2e (k - 1) + 3e (k - 2)

Part IVC: Computer Process


Control
The Digital Controller: Physical
Realizability
This image cannot currently be displayed.

l Example 26.1 d)
The digital controller pulse transfer function :
z 2 + z + 3 + 2 z -1 + 3z -2
GC ( z ) =
z 2 + 1 + 4 z -1
the highest power of numerator : 2
the highest power of denumerator : 2
A Time domain realization :
u (k + 2) + u (k ) = -4u (k - 1) + e (k + 2) + e (k + 1) + 3e (k ) + 2e (k - 1) + 3e (k - 2)

By shifting the time index


u (k ) = -u (k - 2) + -4u (k - 3) + e (k ) + e (k - 1) + 3e (k - 2) + 2e (k - 3) + 3e (k - 4)

Part IVC: Computer Process


Control
26.2.2 Digital Controller: Two
Approaches to Digital Controller
Design

l Indirect
• Carry out the familiar continuous domain
design and then discretize the resulting
continuous controller
l Direct
• Carry out the design directly in discrete time
and obtain the discrete controller directly

Part IVC: Computer Process


Control
Digital Controller: Two
Approaches to Digital Controller
Design
l Indirect Design
• Primary focus is the continuous system, the discrete
nature of controller is only an implementation detail

+ advantage of relaying on controller design principles


which are familiar from continuous controllers design
- discretized controllers may not perform as well as the
continuous controllers from which they were obtained
- controllers are limited in form to discretized versions of
continuous controllers

Part IVC: Computer Process


Control
Digital Controller: Two
Approaches to Digital
Controller Design

l Direct Design
• Primary focus is the discrete controller
+ all the effects of sampling and discretization are
already directly incorporated into the design
+ resulting controllers are only limited in form by
realizability conditions
- fundamentally continuous nature of the process is
not easily taken into account

Part IVC: Computer Process


Control
26.2.3 From Continuous Time to
Discrete Time: The effect of Sampling
and Holding

l No inter-sample process information is


available to the controller.
• The sampling frequency is in important role!
l The effect of the sample-and-hold
element on the sampled data system
is approximately equal to adding a
time delay of Dt/2 to the process
dynamics
• the additional time delay adds wDt/2 radians to
the process phase lag reducing the phase margin
by w1Dt/2 radians (w1 freq when AR=1)
Part IVC: Computer Process
Control
From Continuous Time to
Discrete Time: The effect of
Sampling and Holding

Part IVC: Computer Process


Control
From Continuous Time to
Discrete Time: The Advantages
of Digital Computer Control

l NOTE!!!!Digital computer control allows the


implementation of control strategies that
are impossible to implement with
conventional analog controllers (because
of hardware limitations)!!!!

Part IVC: Computer Process


Control
26.3 Discrete PID Controllers
from the Continuous Domain

l The discrete PID controllers are the


simplest and most familiar digital
controllers
l Obtained by approximating classical
continuous PID controllers

Part IVC: Computer Process


Control
Digital approximation of the
analog PID controller

l The continuous analog PID controller in time


domain
æ 1
t
de ö
p(t ) = K c çç e (t ) + ò e (t )dt + t D ÷ + ps
÷
è t I 0 dt ø

l In s-domain
u( s ) = g c ( s )e ( s )
æ 1 ö
= KC çç1 + + t D s ÷÷ × e ( s )
è tIs ø

Part IVC: Computer Process


Control
Digital approximation of the
analog PID controller
l There are many ways to approximate the analog PID
controller in discrete form
• Approximating integral with the rectangular rule for
numerical integration
• Approximating the derivative with a finite difference
l Remember, there are also many forms of analog PID
controllers
t k
Integral : ò e (t )dt » å e (i)Dt
0 i =1

de e (k ) - e (k - 1)
Derivative : »
dt Dt

Part IVC: Computer Process


Control
Digital approximation of the
analog PID controller
l Discretizing the continuous analog PID
controller
l Depending on the result presentation the
digital controller can take one of the two
forms
• The position form
• The velocity form

Part IVC: Computer Process


Control
Digital approximation of the
analog PID controller: The
Position Form

l In time-domain
æ Dt k tD ö
u (k ) = K c ç e (k ) + å e (i) +
ç ( )
e (k ) - e (k - 1) ÷÷
è t I i =1 Dt ø

l In z-domain
æ Dt æ 1 ö t D -1 ö
u( z ) = Kc çç1 + ç -1 ÷
+ (1 - z )÷÷e ( z )
è t I è 1 - z ø Dt ø

Part IVC: Computer Process


Control
Digital approximation of the
analog PID controller: The
Velocity Form

l In time-domain
æ æ Dt t D ö æ 2t D ö tD ö
ç
Du (k ) = K c ç ç1 + + ÷e (k ) - ç
ç ÷ + 1÷e (k - 1) + e (k - 2) ÷÷
è è t I Dt ø è Dt ø Dt ø

l In z-domain
æ æ Dt t D ö æ 2t D ö -1 t D -2 ö
Du( z ) = Kc çç çç1 + + ÷÷ - ç + 1÷ z + z ÷÷e ( z )
è è t I Dt ø è Dt ø Dt ø

u(k) = u(k-1)+Δu(k) For u(k-1) the real implemented value


No windup problem! should be used in the formula!

Part IVC: Computer Process


Control
26.3.2 Digital PID Controller
Tuning
l In general, digital PID controllers are tuned
using the same classical techniques used
for tuning continuous PID controllers, but
with additional compensation for the effect
of sampling time delay
• The inclusion of the ZOH adds an
effective time delay of Dt/2 to the process
delay
• If Dt is not small, it is necessary to adjust
the process delay from a to a+D t/2
before carrying out the continuous design
Part IVC: Computer Process
Control
26.4 Other Digital Controllers
Based on Continuous Domain
Strategies

l The other continuous controllers could also


be converted to digital ones

l There are many ways to approximate


continuous transfer functions to discrete
ones and the obtained digital controllers
differ of each other

Part IVC: Computer Process


Control
26.4.1 Techniques for Digital Approximation
of Continuous Transfer Functions

l Relation between s and z:


e-sDt = z-1
l 1.A series expansion of e-sDt, backward difference
- sDt -1 1 - z -1
e » 1 - sDt z » 1 - sDt s»
Dt

l 2.An expansion of 1/esDt, forw. diff


e sDt » 1 + sDt -1 1 1 - z -1
z » s » -1
1 + sDt z Dt

l 3.Tustin (First-order Padé Approximation for e-sDt)


- sDt 2 - sDt -1 2 - sDt 2 æ 1 - z -1 ö
s » çç ÷
e » z » Dt è 1 + z -1 ÷ø
2 + sDt 2 + sDt

Part IVC: Computer Process


Control
Techniques for Digital Approximation of
Continuous Transfer Functions

l Tustin-approximation is most accurate


l Backward difference easy to use, stable g(s)
à stable g(z)
l Forward difference, easy to use, stable g(s)
can lead to unstable g(z)

Part IVC: Computer Process


Control
26.4.2 Translating Analog Controller
Designs to Digital Equivalents

l Next we will now translate some typical


continuous controllers
• PID
• Direct Synthesis Controller
• Feedforward Controller
• Smith Predictor Algorithm
into digital controllers using the backward
difference conversion formula

Part IVC: Computer Process


Control
Translating Analog Controller
Designs to Digital Equivalents:
PID

l Conventional Analog PID-controller s»


1 - z -1
Dt
æ 1 ö
gc ( s) = KC çç1 + + t D s ÷÷
è tIs ø

l Digital version is obtained by replacing s


æ Dt 1 tD -1 ö
gc ( z ) = KC çç1 + + (1 - z )÷÷
è tI 1- z
-1
Dt ø

Part IVC: Computer Process


Control
Translating Analog Controller
Designs to Digital Equivalents:
The Direct Synthesis Controller

l Continuous Direct Synthesis Controller s»


1 - z -1
Dt

1 1
gc ( s) =
t r s g ( s)

l Digital controller is simply obtained by


replacing s
in the continuous controller transfer function
Dt 1 æ1ö
gC ( z ) = ç ÷÷ 1- z -1
-1 ç
t r 1 - z è g ø s = Dt

Part IVC: Computer Process


Control
Translating Analog Controller
Designs to Digital Equivalents:
The Direct Synthesis Controller

l Example 26.2 g ( s) =
K

1 - z -1
ts + 1 Dt

KDt
g ( s) = t + Dt
æ 1- z -1 ö
s =çç ÷ æ t ö -1
è Dt ø
÷ 1- ç ÷z
è t + Dt ø

æ æ t ö -1 ö
ç1- ç ÷z ÷
t + Dt ç è t + Dt ø ÷
gc ( z ) = PI-controller
trK ç 1 - z -1 ÷
ç ÷
è ø

Part IVC: Computer Process


Control
Translating Analog Controller Designs
to Digital Equivalents: The Discrete
Smith Predictor Algorithm
l Continuous domain Smith Predictor(diff impl)

æ gC ö 1 1 - z -1
ç
u( s ) = ç ÷e ( s)
-as ÷ gg c = s»
è 1 + gc g (1 - e ) ø t rs Dt
l Digital form is simply obtained by replacing s
in the continuous controller transfer function,
gc PI controller
l For the system of type first-order-process-
plus-time-delay(D), controller would be
æ t r + Dt ö t r -1 (t + Dt ) t -1
÷u ( z ) - z u ( z ) - z u ( z ) = e (z ) - z e (z )
-D
ç
è Dt ø Dt (KDt ) KDt

Part IVC: Computer Process


Control
Translating Analog Controller Designs
to Digital Equivalents: The Discrete
Smith Predictor Algorithm

æ tr ö æ Dt ö é t + Dt ù
u (k ) = çç ÷÷u (k - 1) + çç ÷÷u (k - D) + ê úe (k )
è t r + Dt ø è t r + Dt ø ë K (t r + Dt ) û
é t ù
-ê úe (k - 1)
ë K (t r + Dt ) û

Part IVC: Computer Process


Control
26.5 Digital Controllers based on Discrete
domain strategies
26.5.1 Direct Synthesis in the z-domain

l Sampled-data control system(With feedback


control)
g ( z ) gc ( z )
y( z) = yd ( z )
1 + g ( z ) gc ( z )

l If we now represent the desired closed-loop


behavior by pulse transfer function
y( z)
q( z ) =
yd ( z )

1 q( z )
l The digital controller is obtained gc ( z) =
g ( z ) 1 - q( z )
Part IVC: Computer Process
Control
Direct Synthesis in the z-
domain: Deadbeat Controller

l Based on the requirement that , for any set-


point change, the closed loop response
should exhibit
• Minimum rise time
• Zero steady state offset
• Finite settling time

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller
This image cannot currently be displayed.

l Solution:
• The closed loop system is required to
follow the setpoint exactly at each
sampling point with unit sampling delay
y ( z ) = z -1 yd ( z )
y( z)
= q( z ) = z -1
yd ( z )

1 q( z ) 1 z -1
gc ( z ) = =
g ( z ) 1 - q( z ) g ( z ) 1 - z -1

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller [3]
This image cannot currently be displayed.

l If process has a time delay, this must be


taken to account

y ( z ) = z - D -1 yd ( z )
y( z)
= q( z ) = z - D -1
yd ( z )

1 q( z ) 1 z - D-1
gc ( z ) = =
g ( z ) 1 - q( z ) g ( z ) 1 - z - D-1

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller
This image cannot currently be displayed.

l Looks fine so far, but…


• Deadbeat controllers cannot be used for
processes with inverse response (Processes
which have zeros outside the unit circle)
• When the process have negative zeros inside
the unit circle, deadbeat controller causes so
called “ringing” in the controller output u(k) ,
which produces “rippling” in the continuous
process output y(t) between samples

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller [5]
This image cannot currently be displayed.

>> Gp=tf(1,[1 1 1]);

Example 26.3
>> Hp=c2d(Gp,0.3)

l Transfer function:

0.04052 z + 0.03665

1 ----------------------

g ( s) =
z^2 - 1.664 z + 0.7408

s2 + s + 1 Sampling time: 0.3

>>set(Hp,’Variable’,’z^-1’), Hp

Transfer function:
-1 -2
0.04052z + 0.03665z 0.04052 z^-1 + 0.03665 z^-2

g( z) = ----------------------------
1 - 1.664 z^-1 + 0.7408 z^-2
1 - 1.664 z -1 + 0.7408z -2 Sampling time: 0.3
>> pole(Hp)

ans =

poles 0.8318 ± 0.0221i 0.8318 + 0.2211i


0.8318 - 0.2211i

zero - 0.9045 >> zero(Hp)

ans =

-0.9046

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller
This image cannot currently be displayed.

>> Gc=inv(Hp)*tf([0 1],[1 -1],0.3,'Variable','z^-1')

l Example 26.3 Transfer function:


1 - 1.664 z^-1 + 0.7408 z^-2
--------------------------------------
0.04052 - 0.003864 z^-1 - 0.03665 z^-2
æ z -1 ö 1
g c ( z ) = çç ÷
-1 ÷ Sampling time: 0.3
è 1 - z ø g ( z) >>pole(Gc)

æ z öæ 0.04052z + 0.03665z
-1 -1 -2
ö ans =

= çç ÷ç
-1 ֍ -1 -2
÷÷ 1.0000
è 1 - z øè 1 - 1.664 z + 0.7408 z ø -0.9046

poles 1 Ù -0.9046

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Deadbeat Controller
This image cannot currently be displayed.

l Example 26.3
u

To Workspace2

Gc Gp y_continuous

Step LTI System2 LTI System1 To Workspace1

y_sampled

To Workspace

>>subplot(2,1,1),plot([0:0.3:10],y_sampled,'rx','MarkerSize',10)
>> hold on
>>plot(tout,y_continuous,'-b'),title('Intersample rippling')
>>plot([0 10],[1,1],'-k'),ylabel('y','Rotation',0),hold off
>>subplot(2,1,2),stairs([0:0.3:10],u,'-k')
>>title('Controller ringing')
>>ylabel('u','Rotation',0), xlabel('t')

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm

l How could we improve the performance of


the deadbeat controller ?
• Lets try to choose a less aggressive
reference trajectory

l With Dahlin’s algorithm, the closed loop


behavior is desired to be of the first-order-
plus-time-delay type

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm

l Choose the desired process output trajectory


of first-order
e -gs
q( s ) =
t rs + 1

l Discrete version of desired process output


trajectory with ZOH if g = MDt, n = 0,1,2,...
æ - ö
Dt
ç1 - e t r ÷ z - M -1
ç ÷
q( z ) = è ø
Dt
-
tr
1- e z -1

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm

l The Dahlin controller may now be obtained

æ æ - ö
Dt
ö
ç ç1 - e ÷ z
t r - M - 1 ÷
1 q( z ) 1 ç ç ÷ ÷
qc ( z ) = = ç è ø ÷
g ( z ) 1 - q( z ) g ( z ) ç -
Dt
æ - ö
Dt
t r ÷ - M -1 ÷
t r -1 ç
çç 1 - e z - ç1 - e ÷ z ÷÷
è è ø ø

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm
This image cannot currently be displayed.

l Example 26.4
0.04052z -1 + 0.03665z -2
g( z) =
1 - 1.664 z -1 + 0.7408z -2

æ - ö
Dt
æ 0.3
ö
ç1 - e t r ÷ z -n -1 -
çç1 - e 0.3 ÷÷ z -0-1
ç ÷
q( z ) = è ø = è ø
Dt 0.3
- -
0.3 -1
1 - e t r z -1 1- e z
(1 - e -1 ) z -1 0.6321z -1
= »
1 - e -1 z -1 1 - 0.3679 z -1

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm
This image cannot currently be displayed.

>>Gc=inv(Hp)*tf([0 1-exp(-1)],[1 -1],0.3,'Variable','z^-1')

l Example 26.4 Transfer function:


0.6321 - 1.052 z^-1 + 0.4683 z^-2
--------------------------------------
1 q( z ) 0.04052 - 0.003864 z^-1 - 0.03665 z^-2
gc ( z ) =
g ( z ) 1 - q( z ) Sampling time: 0.3
>> pole(Gc)
æ æ - ö
Dt
ö
ç ç1 - e t r ÷ z - M -1 ÷ ans =

1 ç ç ÷ ÷ 1.0000
= ç è ø ÷ -0.9046
g ( z) ç -
Dt
æ - ö
Dt
t r ÷ - M -1 ÷
1 - e t r -1
z - ç 1 - e z
çç ç ÷ ÷÷
è è ø ø
æ 1 - 1.664 z -1 + 0.7408 z -2 öæ 0.6321z -1 ö
= çç -1
֍
- 2 ֍ -1
÷÷
è 0.04052 z + 0.03665 z øè 1 - z ø

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm
This image cannot currently be displayed.

l Example 26.4
u

To Workspace2

Gc Gp y_continuous

Step LTI System2 LTI System1 To Workspace1

y_sampled

To Workspace

>>subplot(2,1,1),plot([0:0.3:10],y_sampled,'rx','MarkerSize',10)
>> hold on
>>plot(tout,y_continuous,'-b'),title('Intersample rippling')
>>plot([0 10],[1,1],'-k'),ylabel('y','Rotation',0),hold off
>>subplot(2,1,2),stairs([0:0.3:10],u,'-k')
>>title('Controller ringing')
>>ylabel('u','Rotation',0), xlabel('t')

Still rippling and ringing!

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Dahlin Algorithm
This image cannot currently be displayed.

l As we see, Dahlin algorithm doesn’t exclude


controller rippling and output ringing, the
effects of rippling are just a little bit weaker

l Dahlin Controller couldn’t handle processes


with inverse response

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Vogel-Edgar Controller

l Vogel and Edgar have proposed


modifications to Dahlin’s Algorithm by
removing the model numerator dynamics
from the controller
l Vogel and Edgar Controller does not try to
cancel the process zeros
• can handle processes with inverse response
• is not subjected to ringing with processes
which have negative zeros

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Vogel-Edgar Controller
l Modifications are added to the reference trajectory
æ - ö
Dt
ç1 - e t r ÷
ç ÷ B( z -1 )
q( z ) = è ø z - M -1
Dt
- B(1)
1 - e t r z -1

l Vogel-Edgar controller is now obtained


B( z -1 ) æç - ö
Dt

1 - e t r ÷ z - M -1
1 q( z ) 1 B(1) çè ÷
ø
gc ( z ) = =
g ( z ) 1 - q( z ) g ( z ) -
Dt
tr -1 B( z -1 ) æç - ö
Dt
t r ÷ - M -1
1- e z - 1- e z
B(1) çè ÷
ø

where B(z-1) is the numerator polynomial of the process model

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Vogel-Edgar Controller
This image cannot currently be displayed.

>>q1=tf([0 1-exp(-1)],[1 -exp(-1)],0.3, 'variable','z^-1')

Transfer function:
l Example 26.3 0.6321 z^-1
---------------
1 - 0.3679 z^-1

0.04052z -1 + 0.03665z -2 Sampling time: 0.3


g( z) = >> [num,den]=tfdata(Hp,'v');
1 - 1.664 z -1 + 0.7408z -2 >>q2=tf([num(2) num(3)],sum(num),0.3,'variable','z^-
1')

æ - ö
Dt Transfer function:
ç1 - e t r ÷ z - M -1 0.04052 + 0.03665 z^-1

B( z -1 ) çè ÷ ---------------------------

q( z ) = ø 0.07717
Dt
B(1) - Sampling time: 0.3
1 - e t r z -1 >> q=minreal(q1*q2)
Transfer function:
0.04052 + 0.03665 z -1 0.6321z -1 0.02561 z^-1 + 0.02317 z^-2
= ---------------------------
0.07717 1 - 0.3679 z -1 0.07717 - 0.02839 z^-1
Sampling time: 0.3

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Vogel-Edgar Controller
This image cannot currently be displayed.

>> Gc=minreal(inv(Hp)*feedback(q,-1))

l Example 26.3 Transfer function:


8.191 - 13.63 z^-1 + 6.068 z^-2
-------------------------------
1 - 0.6998 z^-1 - 0.3002 z^-2

Sampling time: 0.3


B( z ) æç - ö
Dt
-1
t r ÷ - M -1
1- e z
1 q( z ) 1 B (1) ç ÷
gc ( z ) = = è ø
g ( z ) 1 - q( z ) g ( z ) -
Dt
B ( z -1 æ
) - ö
Dt

1 - e t r z -1 - ç1 - e t r ÷ z - M -1
B(1) çè ÷
ø

=
A( z -1 ) (1 - e ) -1
=
8.191 - 13.63 z -1 + 6.068z -2
B( z -1 ) 1 - 0.6998 z -1 - 0.3002 z -2
B(1)
1- e z - -1 -1

B(1)
(1 - e -1 )z -1

Part IVC: Computer Process


Control
Direct Synthesis in the z-
domain: Vogel-Edgar Controller
This image cannot currently be displayed.

l Example 26.4
u

To Workspace2

Gc Gp y_continuous

Step LTI System2 LTI System1 To Workspace1

y_sampled

To Workspace

>>subplot(2,1,1),plot([0:0.3:10],y_sampled,'rx','MarkerSize',10)
>> hold on
>>plot(tout,y_continuous,'-b’)
>>plot([0 10],[1,1],'-k'),ylabel('y','Rotation',0),hold off
>>subplot(2,1,2),stairs([0:0.3:10],u,'-k')
>>ylabel('u','Rotation',0), xlabel('t')

No ringing!

Part IVC: Computer Process


Control
26.5.2 Time domain designs

l The discrete controller is designed


directly in the discrete time domain using
discrete-time difference equation model

l The control law is given in time-domain


too

Part IVC: Computer Process


Control
Time domain designs: Direct
Synthesis Design

l Choose a reference trajectory in the discrete-


time domain
l Choose a Controller parameters so, that the
closed loop behavior is the same as desired
reference trajectory

Part IVC: Computer Process


Control
Time domain designs: Pole
Placement Design

l Characteristic equation can be obtained


directly from the discrete time
• one need not calculate the pulse transfer
function to use pole placement design
l Choose the parameters of the controller
so that poles of the closed loop are in
desired locations

Part IVC: Computer Process


Control
Other Approaches to Digital
Controller Design

l Root Locus method


l Time domain techniques of optimal control
l Model based control algorithms(Dynamic Matrix
Control DMC, Model Algorithmic Control MAC)
l etc.

Part IVC: Computer Process


Control
26.6 Digital Multivariable
Controllers

l Multivariable systems are usually so


complex, that it’s impossible to implement
multivariable controllers with classical
analog controllers
l So, almost all multivariable controllers are
implemented in digital form!

Part IVC: Computer Process


Control

You might also like