You are on page 1of 60

Mooterey, Clitonji.

LINEAR CONTROL SYSTEM DESIGN

USING ALGEBRAIC METHODS

by

Albert Bernard Lemanski


Lieutenant, Unitecr States Navy
University of Illinois, 1958
B.S.
,

Submitted in partial fulfillment


for the degree of

MASTER OF SCIENCE IN ELECTRICAL ENGINEERING


from the

UNITED STATES NAVAL POSTGRADUATE SCHOOL


May, 1966

c.|

ABSTRACT

A method

is

presented for the analytic design of linear

Design

control systems.

is

accomplished by the solution

of

an algebraic set of simultaneous equations derived from the


design specifications.

Specifications considered include the

system damping ratio, undamped natural frequency, bandwidth,


steady state error coefficients and root sensitivity.

The method

is

shownWcfT5e

fa's|,

accurate, specific and

capable of working with any number of parameters

In

addition, the analytic approach minimizes dependence on

graphical techniques and trial and error solutions.

ry

SeEool

TABLE OF CONTENTS

Page

Section
Introduction

2.

The Algebraic Method; Concept and Development

3.

Application of the Algebraic Method

19

4.

Sensitivity as a Specification

P.S

5.

Conclusions

HOI

LIST

OF ILLUSTRATIONS

Page

Figure
2. 1

Block Diagram for Example 2.3

18

2.2

Block Diagram for Example 2.4

18

2.3

Block Diagram for Example 2.5

18

4.

Graph of Sensitivity vs. Gain


Example 4

39

for

SYMBOLS AND ABBREVIATIONS

CE

Characteristic equation

E ss

Steady state error

G(s)

Forward transfer function

kQ

Acceleration error constant

Accelerometer gain

Ka

Gain

or k

or k

ICy

Position error constant


Velocity error constant

Kt

Tachometer gain

M pw

Reasonance peak

Superscript or
subscript "o"

Specified or numerical value

Angular frequency

(omega)

wn

Undamped

natural frequency

wb

Bandwidth

wr

Rea sonant frequency

tf{f

-3 db)

Polynomial in

General sensitivity parameter

Complex variable

Sensitivity

SP

of the form cr +

Sensitivity of P with respect to

jw

gw n
S

jr

**

component

of root sensitivity

wn

component

of root sensitivity

Total root sensitivity


(zeta)

Damping

ratio

Introduction

Linear control system design can be defined as "a set of

procedures for determining the value of system parameters

which will cause a physical system to operate


manner.

The set of procedures noted above

"

One may choose

the definition.

in a prescribed

as broad as

is

to perform the design effort in

the time domain by applying modern state space concepts or by


direct solution of the system differential equations.

the complex frequency domain

Alternatively,

may be selected, employing any

of the universally accepted procedures such as Nichols

and

Bode plots, root locus, Mitrovic's Method, etc.


Notwithstanding the achievements which have resulted
from direct application of the above procedures,

admitted that they have inherent limitations.

it

must be

With the

exception of simple systems, time domain analysis virtually

demands the assistance


this is a disadvantage

of a digital computer.

because computers are neithe

to all people at all times nor is

employ a computer

In a sense,

for all

it

available

economically practical to

computation.

The complex frequency

schemes are considered imperfect because they are time


consuming, i.e.
based on

trial

variables.

and

essentially graphical in nature and often


error,

and normally limited to one or two

In

need

view

of the

above discussion,

for still another

new procedure

design tool.

it

appears that there

However,

to even merit consideration,

it

is

in order for a

must be

relatively free from the faults and limitations cited above.

Specifically, the method should be analytically simple, i.e.

extend the range of problems which can be solved without


resorting to graphical techniques or the exclusive use of a

computer.

It

should also be rapid, capable of handling several

variables, specific and accurate.


is

The remainder of this paper

devoted to the development and discussion of a design

procedure which is believed to have these attributes.

The Algebraic Method: Concept and Development

Concept
The proposed design procedure

is

based on the premise that

system design specifications can be transformed into algebraic


functions of the system parameters.

independent,

If

the specifications are

specifications will produce

equations which

Control system design

simultaneously constrain the system.

would thus revert to the relatively simple problem

of solving a

set of simultaneous algebraic equations.

The problem of transforming specifications into equations


is not

as broad and difficult as one might initially assume.

The most common and practical complex frequency specifications


are:

undamped natural frequency (w n )

state error

(E

specified.

In

ratio (")/ steady

darning

reasonance peak (M

many cases, both $ and

wn

ss ), bandwidth (w b)

reasonant frequency (wr)

and
are

This is equivalent to specifying the location of a

complex pole pair and generally implies that these poles should
dominate the system response.

If

attention is restricted to this

situation, then the specifications of primary interest are

wn

W]-,

and E ss because

and

wr

can be at least

approximately expressed in terms of jr and


of this section is therefore

The remainder

devoted to the development of methods

which can be used to obtain system constraint equations from


specifications

f or

wn

w, and E gs

Root Location ( and

wn

The coefficient equations associated with the specification


of a

complex pole pair can be developed

method

is

in

two ways

The

first

based on the root -coefficient relations of an

algebraic equation; the second is based on Mitrovic's Method.

Assume

the systems' characteristic equation (CE)

obtained

by setting the denominator of the closed loop transfer function


equal to zero, is the polynomial.

<2<

S<

(2-1)

where

s is

a complex variable and defined as

c*(2-2)

and the a^'s represent the real coefficients which are functions
of the

system parameters.

Since equation (2-1)

is of

which satisfy the equation.

order n, there must be n roots

Therefore, an equivalent expression

for equation (2~1) is

(2-3)

C~,

where

R^ represents the ith root of the

polynomial.

By expanding

equation (2-3) and equating the coefficients of like powers of

10

s in

equations (2-1) and (2-3)

n equations can be obtained which

relate the roots and coefficient of the original polynomial.


If Jr*

pole pair,

anc*
it

is

define the location of the specified complex

permissible to set

Rz -

F *^> -

(2-4)

ft

ft*'
(2-5)

Substituting these values of R, and R2 into the n root -coefficient


relations obtained previously gives n equations in terms of the

remaining n-2 roots and the coefficients of equation (2-1).


algebraic manipulation,

it

is

By

then possible to successively

eliminate each of the remaining n-2 roots and finally obtain

two equations involving only the coefficients.

Since the

coefficients are explicit functions of the system parameters

any combination

of the parameters

which satisfy these equations

will generate the desired pole pair.

Example

2.

Let the system characteristic equation be

S3

-h

qz s2 + G*S +
-

ae>

Find the relations which must exist between the coefficients in


order to locate a pole pair withj^ = 1/2 and

Denote the three roots

of the

11

CE as

R]_,

w Q =10.
R2 and R3.

Then

and

cr,

a& =

f%,

Ra.

*>

= S~

*-

R*.

*-

Rz )

-hRs
/

- j to

/> (q z -/ q ) -

Then

/?* /? 3

az = R,
Setting

+ R^ (&

/?,>?*

/
/

3/y.

(1)

=Qo
/OORs * ^7o

/0(Q,-/0)

(2)

Equations
equation

(1)

(3)

and

are the desired coefficient relations;

(2)

provides information about the third pole.

While the root-coefficient method works well with low


ordered systems (and even provides information about the

remaining n-2 roots),

it

requires an inordinate amount of

computation when applied to systems

of higher order

because

the coefficient relations must be extracted from a set of nonlinear equations.

This difficulty can be overcome

the expense

(at

of losing information about the remaining n-2 roots) by using

Mitrovic's Method.

With

this method, a kth ordered

be handled as easily as one of third order.

developed

in (3)

and

(5).

system can

The method

Only a brief review

is

is fully

given here.

The argument begins with equations (2-1) and (2-2) defined


previously.

Two independent equations

in a^,

wR

and

;^rcan

be

obtained by substituting equation (2-2) into equation (2-1) and


setting the real and imaginary portions of the resulting

12

expressions equal to zero.


the introduction of a

new

Through additional manipulation and

variable, the equations take the form

G< UJ* 4<

*r-o

oj$

ci K

cj>

(?)

(2-6)

(f)-O

=P

(2-7)

AC-O

where

with

(?)

(f) -

= - * F

Appendix

(f)

-f

4 tfJ - "I

Other values of 2^(3^


in

a*"*

4>.

(2-8)

(y)

for various values of

<W - 4<

Cf)

are tabulated

Equations (2-6) and (2-7) are known as the generalized


Mitrovic equations and can be used to obtain the coefficient
relations required to locate a pole pair as specified by JT

and

wQ

Example

Work example

2.1 using Mitrovic'

Method.

For a third order system, the generalized Mitrovic equations


are

o4* (f) + ** oj^4o

+ ** u>i

4>t

(?)

+ *9 *>i

Substituting

wn

Appendix

the above equations simplify to

I,

Pi =

tO

wQ

cf)

<j>

z cr)~

= 10 and using values for ^(1/2) from

7o

13

=.

/oo q z - iOOO

The results are equivalent to those obtained previously.


Although only a single pair of complex poles
considered here,

it

is

being

should be noted that either method

is

capable of handling any number of real or complex poles


(providing, of course, that the total

number considered

or equal than the order of the system)

is less

For example, assume

two complex poles pairs and two real roots

of a sixth order

The root -coefficient method will provide

system are specified.

six constraining equations by direct substitution of the given

values.

If

Mitrovic's Method is used, four constraint equations

dependent oh the complex poles can be obtained by repeated


use of equations (2-6) and (2-7).

The remaining two equations

which are dependent on the real roots are produced by repeated


use of

C-')*^*<7 *

<-=o

(2-8)

obtained by letting

-gr* in

equation (2-1)

Steady State Error

(E

ss )

There are three different steady state error constants which

may be

specified.

They are

position error constant

= k = lim G(s)
p

velocity error constant

= k

(2-9)

s-*-o

= lim sG(s)

(2-10)

s-o

acceleration error constant = k a = lim s^G(s)


s**o

14

(2-11)

The above definitions are only valid for systems which have

all

their roots in the left-half of the s-plane (because the

definitions are merely restatement of the final value theorem


for

complex variables) and unity feedback around the forward


To satisfy the last requirement,

loop transfer function, G(s).


it

may be necessary

to manipulate the block diagram into an

equivalent G(s) block with the required unity feedback.

The constraint equation

is

obtained by equating the specified

value of E ss with the algebraic expression resulting from


application of equation (2-9), (2-10) or (2-11).
is

demonstrated

in the following

The procedure

examples.

Example 2.3
For the system shown in figure 2-1, obtain an expression

between K, Z and P f or k

6.

For the given system

<U*

(S)

**<***>
52

(*+s)(s+C)(*+R)

Applying equation (2-11)

Ji =

which, for k a =

.^
6,

=
** Q^cs) .

reduces

to:

<z

18 OP = KZ

Example 2.4
For the system shown in figure 2-2, derive an equation in

terms of K,

K 2 and K

which will give K v =

15

Since the system does not have unity feedback, insert


fictitious positive

and negative feedback paths around the

G e q(s) can

system.

then be found by including the positive

feedback branch

S x (H-K K3 )-t-sfa+K,K2.)
t

-K,

Applying equation (2-10)

Therefore the specification cannot be satisfied.

Bandwidth (wij

Bandwidth will be defined as the angular frequency, w,,


at

which the magnitude

a value of 0.707, i.e.

of the closed loop transfer function has

- 3 db bandwidth.

The transformation

of the specification into the desired constraint equation

can be

accomplished by following the usual procedures:


= jw.

A.

Find the closed loop transfer function and set

B.

Find the magnitude of the resulting expression and

equate

C.

it

to 0.707.

Substitute

w^ and

simplify.

Example 2.5
For the system shown in figure 2-3, derive a relation in

terms of K and IC which assures the system has a bandwidth


of 10 radians/second.

16

Following the prescribed procedure:

Ccs)

iK-co*)* T io*(-+KKt)Z

RCju>)

For

wb

= 10

O (juS)

__

<2

= /0*-ZOOK +/00(S-+KKt) Z

'

The constraint equation obtained


for a

II

/z

\Rtju>)l

second order system

in the

is typical of the

preceeding example

expressions which

are developed from bandwidth specifications.

Since the

expressions increase in complexity for higher ordered systems


or for systems having closed loop zeroes,

most design problems

involving bandwidth specifications will usually require the

assistance of a computer or a graphical technique.

(An excellent

graphical procedure based on Mitrovic's Method is given in (4).)


If

such aids are not available, a convenient, and relatively fast,

alternative solution is to initially ignore the bandwidth

specification and design the system to meet the remaining

The results can then be inserted into the system

specifications.

bandwidth equation to determine


requirement.

If

if

they also satisfy the bandwidth

the specification is not satisfied, the magnitude

of the deficiency

may suggest

next attempt.

17

a more reasonable structure for the

4CS)
6>g

-~&*-

<3c CS)

^ St

S* (*+L)(S+S)

FIGURE 2.1

BLOCK DIAGRAM FOR EXAMPLE 2.3

gj (s)

^c

K.

\
)
.

s (s + s)

Tach FB

Ka S
FB
K3 s2

AcceL

FIGURE 2.2

BLOCK DIAGRAM FOR EXAMPLE

Q-s)

FIGURE 2.3

BLOCK DIAGRAM FOR EXAMPLE


18

Application of the Algebraic Method

As stated previously, the basis of the proposed design

method
of

is the transformation of

specifications into a set

simultaneous constraint equations that are explicit functions

of the

system parameters.

Since the specifications are

independent, a specific design will generally require that the

system have N parameters.

The insertion of sufficient

parameters, usually compensating elements, is an integral part


of the design procedure.
It

should be noted that the mere selection of sufficient

parameters does not guarantee a satisfactory solution.

converse of this statement

is

also true.

An unsatisfactory

solution does not imply that a solution does not exist; one

may

possibly be obtained by employing a different set of parameters.

These statements suggest that there


associated with the procedure.

amount

of trial

is

some

trial

This is correct.

and error

However, the

and error required can be minimized through

experience and application of control theory fundamentals.

In

this respect, root locus sketches are particularly valuable in

obtaining clues to the nature of the required parameters.

The algebraic method has the inherent capability to handle

any number

of specifications or parameters.

Therefore a

parametric solution can generally be obtained by simply inserting

more parameters into the system than there are design

19

specifications.

If

parameters are selected for use in the

design of a system subject to

specifications, for

M>N,

results will be parametric and expres sable in terms of any

the

parameters.

Example
Given G(s) = K/s(s +

4)

3.

design* a system which will

produce a pair of complex poles defined by


with Kv = 4
It

Jp

= 1/2 and

wQ

= 4

will be stated without proof that K = 16 satisfies all

specifications.

This seemingly trivial example has major significance.


Specifically, one parameter successfully satisfies three

specifications which is a direct contradiction of the previous

discussion.
of

It

would appear that adherence to the general rule

parameters for

specifications could possibly result in

the incorporation of redundant material into the system.

does not seem likely because

This

of the following observation:

In all cases studied, parameters

which did not assist

in

the solution tended to be eliminated.

Examples 3.2 and 3.3 will be used as illustrations

of this

observation.

Example

3. 2

Insert cascade compensation,

G c (s)

= (s+Z)/(s+P), into

the system given in example 3.1 and redesign the system to

*Note:
input

employ unity feedback between output and


addition to any other specified feedback)

All designs will


(in

20

b !i e

eet

r,

ene

s p c 1

cnt

on

_K*JLS

^69 fs) *

Applying equation (2-10):


(1)

-4*
^P

K* System CE:
S 3 +&(++/>)
Mitrovic equations:

A simultaneous solution

= V -r K
~ t&(*+P)-m

^/Vv-Pj
/CfL

K=

& (4P+K )-*.<*

-/-

/=

of equations

(1)

(2)

and

(2)

(3)

(3)

gives:

P = Z

16

Since the pdle and zero of the filter are coincident, the

filter

performs no function and can therefore be eliminated.

Example
For G(s) = 50/s(s + 5),
-5

(JT

- 0.707 and

tachometer feedback
P = 10)

In

(Kj.

it

wQ

is

possible to locate poles at

5 (2)

0. 1) or a

each case, K =
v

l/^ 2
)

by using either

cascade compensator

(Z

= 5,

Design a system using both

tachometer feedback and a series compensator to satisfy the

same specifications.
For combined tachometer feedback and cascade compensation:

^(s, - S 3
Applying equation (2-10):

-**<S+*}
*-

P+S + SOKr) +s(5P+SOKt2)

5*

<v

= Soz/(TP + <>OKtI)

21

(1)

System CE:

Mitrovic equations:

F>

i-

So 2 = SO (P+"+ SOKr)"Soo

&OKrlk

+SO /6
Z =

Simplifying:

(P-^'S '+ "0

First of all,

it

"

&C

(3)

(4)

P = 10(1 -

The result obtained

(2)

example

in this

(5)

5KJ.)

is

doubly significant.

indicates that the procedure is capable of

detecting conditional as well as complete redundancy.

example 3.2, the

filter

was redundant

equation
Kj.

= 0.1.

(5)

shows that the

filter is

Similarly, the tachometer

The second point of interest

may

not exist.

In this case,

which satisfy equation


solution is best?

(5)

for all values of the

In this example,

compensator pole and zero.

however,

redundant only when

may be discarded when

is the fact that a

also satisfy the specifications.

At this point,

it

for determining the best

P = 10.

unique solution

any positive values of P and

Kj.

Which

is a matter of conjecture.

However, this example will be rediscussed

means

In

in Section 4 after a

design has been developed.

Example 3.4
Given the forward loop transfer function, G(s)
design series compensation to give roots with

wQ

6.

The minimum acceptable value of K v


22


=
s (j^'/i;(s + 7;
/^

~ 1/2 and

is 4. 15

r-z

For cascade compensation:

Geo

Cs)

S*+

7>Sf>3 3 -*-

Applying equation (2-10):

(s-hlt)

S*(7.&P+3.) +

Kv -

>-f.l

**
0~ S f~

3-S~P>s

(1)

System CE:

Mitrovic equations:

<it m &+

p-

i+o-

K = HhP
Solving equations

(l)

(2)

and

(3)

gives:

K=

(2)

/9-T

(3)

1475

P = 37.9

Z = 0.373

Example :3
The bandwidth
radians/second.

for the

system

in

example 3.4

is

about 6.5

Redesign the system so that the bandwidth

is

reduced to 4 radians/second.

The introduction of another specification requires the


insertion of another parameter.

Use tachometer feedback around

the original plant and series compensator.

Ge

(s)

Then

gjifrgj

Applying equation (2-10):

/<v

**/./"

3*&P -t-K.IC.TT.

(l)

System CE:

S 4 + s 3 (y. S -,.P)
Mitrovic equations:

t-

Sx

(3 .

S + ?.Sp-t- KK-r) +S (s-SP+KKYt+K) +-Ki~0

Km . A*jb.f*><^l*J-^rt

{2)

System bandwidth equation

^^-

1/2 =

Solving equations

(1)

w^

for

4:

11^2

through

(4):

K=

252

Kt = 0.16

24

P = 5.93

Z = 1.015

(z/;

4.

Sensitivity as a Specification.

now

Attention is

directed to the question posed during the

discussion of example 3.3.

If

more than one design satisfies

the specifications, which is the best design?

The fact that a

was not obtained indicates the need

specific solution

for another

equation which, for the system being developed, implies the need
for

Since the design specifications have

another specification.

already been incorporated into the solution,

it

follows that the

additional specification should be general in nature and

applicable to all systems.

One choice

for a general control

system specification

is

sensitivity which will be defined as

C*

where Sq

is

(Jtrn

it

dP/r

(Sit)

(4-1)

read as "the sensitivity of a function P with respect

to a parameter Q.

because

P)

"

(This definition is preferred

by the author

imparts more physical meaning than most other

definitions of sensitivity found in the literature, i.e.

it

that sensitivity is essentially the ratio of the percentage


in the function to the percentage

reason for this choice

is

change

states

change

in the parameter.)

The

the fact that a sensitivity specification

will force the final design to be realistic.

Specifically,

it

requires

the consideration of changes in the parameters of the physical

system which will result because of tolerances, aging and


environment.

On

this basis, the best design is the one

25

which not

only satisfies the design specifications, but is also able to

maintain the desired performance in spite of undesired parameter


variations.

Root Sensitivity-

Use

of equation (4-1) requires definition of the function P.

Since sensitivity is being introduced as a means of insuring

compliance with the design specifications, P should ideally


be a function of all the design specifications.

However,

this presentation, only the root specification, i.e.

wQ

will be considered.

j? Q

in

and

Hence, the discussion will refer to

the root sensitivity of the system.

(Hereafter, root sensitivity

and pole sensitivity will be used interchangeably.)


Restricting attention to root sensitivity does not completely

preclude the determination or discussion of a best design.


fact, there are

In

two arguments which partially justify (and might

even suggest) this simplification.

One argument

is that root

specifications are more stringent than either the bandwidth or


error coefficient specifications.

The latter two are frequently

specified in terms of inequalities, i.e.


a certain

can be

minimum

or

built into the

specifications.

maximum value.

greater or smaller than

Therefore, some tolerance

design by working on the "safe side" of the

Another argument

is

based on the root-

coefficient relations of the system characteristic equation.

the mobility of two of the systems' closed loop poles are

26

If

restricted by

means

of a sensitivity constraint, then the mobility

of the remaining closed loop poles is also restricted.

This

statement stems from the fact that the specified poles have

corresponding roots which must be factors of the characteristic


equation.

Now,

in order for these roots to be perfect factors of

the characteristic equation, the coefficients of the characteristic

equation must have remained reasonably constant for small

parameter changes.

This implies that the remaining roots, and

hence the remaining poles, also remained reasonably constant.


Since both bandwidth and the error coefficients are partially

dependent (and, in fact, totally dependent for low order systems


without zeroes) on the characteristic equation,

it

follows that

their sensitivity functions bear an approximate relation to the


root sensitivity functions.

Root Sensitivity Equations

Although the practical importance of sensitivity has long

been recognized,

was expended

it

in the

has only been

in recent years that effort

development of sensitivity as a useful tool.

The work of Kokotovic and Siljak,

(2)

major advancement towards this goal.

is

considered to be a

In essence, they used

the definition of sensitivity, equation (4-1), in conjunction with


the generalized Mitrovic equations, equations (2-6) and (2-7),
to obtain algebraic expressions for (what the author considers

to be) the

components

of root sensitivity.

27

Their results are not

only particularly suited for use in the design of control systems

based on a specified pole pair, but also simplify sensitivity


computations because they permit the exclusive use of real
variables in calculating the sensitivity of a function containing

complex variables.
Since the Kokotovic and Siljak sensitivity equations will

be used in subsequent examples, a brief review of their develop-

ment will now be presented.

Differentiating the generalized

Mitrovic equations, equations (2-6) and (2-7), with respect to


a parameter

OJm

gives

^" j=^
+

-'

f61{

^k-l( J^

^ 4s? *>
and 2^-1

(4

4><

(f)

= o

are the derivatives of

'

" 2)

jz

and

respectively and are related by the general equation

4K (?) = -z[s

&

Numerical values for

Appendix

<3

$?4+**irf*o

+
where

II.

jzf,"

+*-,$

+ t^ <C* wj

for various values of

(4

. 4)

are given in

Manipulation of equations (4-2) and (4-3) and

application of equation (4-1) gives

(4-5)

28

where

Bi

= ^wf

(f ;

&

/%?<**&* 4 K <r)

(4- 7)

e2 =j

(4-8)

w$ <?^j

^ -&*>
Q

C -f ^ "* ^.^

Solving equations (4-5) and (4-6) for S

<?

Q and S^n

(4-9)

gives

(4-10)

if

(4-1 1L

5q ^~<3 D
where

4,

5;

a*

Be,

(4-12)

c,
^2.

(4-13)
|

A,

(4-14)
E

C*

Equations (4-10) and (4-11) are the desired sensitivity design


equations.

(Hereafter,

Sq and

SqI\ will be referred to as the

"component sensitivity functions.


simplicity is

number

masked by

their

dependence on the solution

of auxilliary equations

of the auxilliary equations

At this time, their true

")

of a

Suffice to state that the solutions

become exceedingly simple when the

location of a pole pair is specified

fQ

and

wQ

coefficients of the characteristic equation are

and the

known

will be clearly demonstrated in subsequent examples.


2-9

(aP)

This

Intrepretation of Sensitivity Values

Equations (4-10) and (4-11) may be used in two ways.


Substitution of a given set of parameter values into the right

hand side

of the

equations gives a definite sensitivity value

parameter values.

for a specific set of

Alternatively, the

assignment of a sensitivity value to the component sensitivity


functions give the parameter relations which must be satisfied in
order to obtain a system of the specified sensitivity.
In either case, the final design will be dependent

accepted/assigned value to sensitivity.

It

upon the

follows that some

guidelines for reasonable sensitivity values must be established.

Since the literature

is

lacking in this area, the following guides

(which are consistent with the original interpretation of equation

proposed by the author:

(4-1)) are

A.

Signs

(+ or -) of

only. Therefore, the

sensitivity values indicate sense

minimum value

implying that a function

is totally

of sensitivity is zero,

insensitive to variations of a

parameter.
B.

SJJ=

in Q.i.e.
is

if

indicates a linear variation in P due to changes

5% change

in

results in a

5% change

in P.

This

considered to be the maximum permissible value of sensitivity;


possible, systems should have sensitivity values

than one.

30

much less

c.

variation.

>1 indicates amplification of the parameter


This

generally considered undesireable for a

is

control system (but

may have application

in

some detection

schemes)

D.

As indicated

properties.

in

A above,

sensitivity values have vector

Therefore, the total root sensitivity, Si^", is

appropriately expressed as
(S<RT)2

(S

F)2 +

(S j^n)

(4-15)

Sensitivity as a Specification

The concept of sensitivity has been developed to the extent


that

it

can now be used as an additional system constraint

subject to analysis by the algebraic design method.

The

remainder of this section will be devoted to the design and

discussion of specific systems which must satisfy sensitivity,

as well as other, specifications.

Example 4

Reconsider example 3.3 in view of a specification requiring

minimum sensitivity

of the specified poles with respect to

variations in system gain, K.

Considering the system gain, K, as a parameter, the system


characteristic equation is

(1)

The sensitivity equations will now be devloped

in detail to

demonstrate the solution of the auxilliary equations.


equations (4-7) and (4-8) gives

3D

Applying

SOO Oj /OO Oz

/?/

/? 2

= /oo /z

Bz.

wn

after substituting

-^t%

m /oo oz
=

wQ

o,

?se> i/2 o3

2 oco

= 5(2) 1 / 2 and values for

and sz^_^(0. 707) from Appendices

and

II.

jz

(0.707)

Although specific

values of P and K^ are unknown, inspection of equation

(1)

in

conjunction with the results of example 3.3 show that the


coefficients of the characteristic equation must have the

following values for all solutions

QZ=Z0

Q,

=/>

<* *

/,

which, when substituted into equation

(2),

gives
(3)

Taking the partial of each coefficient in equation

(1)

with

respect to K gives

Applying equations (4-9) and substituting, as above,

and

f or

k (0.707) gives

jzf

which, for Z =
c, * s-

as obtained in example 3.3, reduces to

- o r

Cz

* $

>"

lJT ~

Applying equation (4-12) and substituting equation

(3)

gives

D -ZSo x to 3
Applying equation (4-13) and substituting from equations

and

wn

(5)

gives
32

(6)

(3)

D** =
1

ZSoo

(7)

Applying equation (4-14) and substituting from equations

and

(5)

gives

>k
Substituting
(7)

and

(3)

(8)

= K = 50,

**-'*-&*/& ('-'*#*)

fQ =

into equations (4-10)

^
(9)

and

and equations (6)

and (4-11) gives

-M/**r-/J

3**Equations

0. 707

(8)

(9)

do)

Vz

(10) are the

desired design equations.

was needed, they

Since only one additional constraint equation

would normally be squared and added to obtain an expression


In this case, however,

for the total root sensitivity.

S^nis

Therefore, the minimum pole

observed to be a constant.

sensitivity with respect to K can be found by setting equation


(10)

equal to zero.

This gives

Kj.

= 0.1 which corresponds to

the solution of tachometer feedback only.

For

Kj.

= 0.1, S w n = 1/2 and S = 0.


k

Therefore, the analysis

predicts that a ten per cent change in K should produce a five

per cent change in

wR

while

remains constant.

predictions can be mathematically verified.

The system

characteristic equation for the specified value of

ForK=50:

w Q =7.07

ForK=55:

w =7.42
33

These

Kj.

is

= 0.707

Jf =0.708

Au>^

Therefore

^^ 9 <r%

Aj_

0m/

+%

Parameter Selection

The preceeding example raises several questions which


have yet to be discussed.
a parameter on

The

deals with the selection of

first

which to base the sensitivity analysis.

choice might be the parameter which

One

considered to have, the

is

widest variations during expected conditions of operation, e.g.


the gain of an electronic amplifier.

Another choice might be

similar to the first but restricted to plant parameters.

assumes the designer

is

guarantee that either is correct.


is naturally

This

able to select and/or control the

sensitivity of the required compensating elements.

system

It

may be possible

There is no
that the

insensitive to the parameter likely to have

the widest variations in value while

it

could be highly sensitive

to a parameter that is expected to have only minute variations


in value.

It is

also possible that, by minimizing the sensitivity

due to one parameter, the effect of another is amplified* The


>

answer, then, requires investigation of


It is

of

all

possible parameters.

believed that a thorough investigation of a particular type

system could lead to the establishment of general guides

for

the selection of the appropriate parameter(s) in other similar

systems.

Since such an investigation

is not

consistent with the

purpose of this paper, the author will be content to arbitrarily

34

select any parameter which will facilitate demonstration of the

algebraic method.
It

should also be noted that there are no restrictions as to

the number of parameters which


If

may be similtaneously considered.

the sensitivity of pole P with respect to parameters

and

Q3

is desired, the

component sensitivity functions

each parameter may be individually calculated

component sensitivity function due to

all three

each component must be considered, there

that individual

circumstance

components may cancel.

is that the

Aj, A2,

for

parameters is
,

Since the sense


is a possibility

Another fortunate

parameters do not increase

This is because equations (4-7), (4-8) and (4-12), i.e.,


B]_,

B2 and D, remain the same for all parameters.

Minimum

Sensitivity

Another point of interest in example 4.1

is the

ease with

which the condition of minimum sensitivity was obtained.


is not a

unique situation because

it

was possible

seems quite likely that

relations for

for

This

to obtain

results by inspection in other systems studied also.


it

Q2

computations for a situation considering

the root sensitivity with respect to

N-fold.

The total

the algebraic sum of the individual components, e.g.

of

Q^

However,

many systems the parameter

minimum sensitivity must be obtained by setting

35

the derivative of the sensitivity function equal to zero.


the problem

parameter

if

the sensitivity with respect to more than one

being considered.)

is

(Note

In such cases, care

taken to insure that the function to be differentiated


form which will indicate minima near zero.

must be

is in a

This is necessary

because minimum sensitivity was defined to be zero.

The above

stipulation can be conveniently accomplished by considering


either the absolute value or the square of the

sensitivity function, i.e.,

Use

/S^^j

or

component

(Sq n )

vice

S<?

**

of the total sensitivity function, equation (4-15), is also

appropriate.

Discussions thus
sensitivity."

far

have repeatedly used the term "minimum

This does not necessarily mean that minimum

sensitivity must be attained by, or even an objective of, every

design.

The specified sensitivity value, whether

it

be zero or

two, should be consistent with the other specifications. For

example,

if

the pole locations are only required to be accurate

within five per cent for satisfactory system response and the

parameter of interest

is

also expected to have five per cent

variations, a sensitivity value of one is sufficient.


if

Of course,

analysis shows that a lower value is attainable, the design

should be based on the lower value provided

it

can be achieved

without additional cost or degradation of some other aspect of

system performance.

36

The nature of the sensitivity function


design value

is

also of concern.

If

in the vicinity of the

the design value is

associated with a point on the sensitivity function curve where


the curve is very steep or irregular, the system is likely to be

more sensitive than an alternative design based on a (reasonably)


higher value which corresponds to a smooth portion of the
sensitivity function curve.

Example 4.2

portion of a control system has the forward transfer

function, G(s) = K/s(s +

with

= 1/2 and

jT

Overshoot

is of

2)

(s

+4).

Dominant closed loop poles

are required for the desired response.

prime importance; therefore

Q should be

maintained within 1% accuracy for [v anticipated gain variations


of

5%.

in the

Velocity and acceleration signals are required elsewhere

system and are therefore available for use as feedback

signals.

The problem as stated gives


and SjJ and

parameters.

specifications (pole pair

(Note that an additional specification

concerning S w n would normally require the insertion of an


additional parameter whereas a specification regarding the total

pole sensitivity would not.


permit

wQ

to vary,

if

Also note that the specifications

necessary, in order to maintain

approximately constant.)

37

fQ

s3

System CE:

2* (k + <K<x} ts(ftKkT)t K =

Mitrovic equations:

K * zs

zs K Ka

*"

(1)

(2)

Applying equations (4-7) thru (4-9):

At - 31S- Sao*.

/?

SOqz. - So,

a.

(3)

C^i-ZSKKv

C*,

=S(Ka-Kr)

Unlike example 4.1, the coefficients of the characteristic


equation do not have fixed numerical values.
solution of the determinants for

D and

D,

Therefore, the

equations (4-12) and

(4-14), will be

hampered by the presence

At this point

it

will be convenient to solve equations

Then equation

for

and

K.

(3)

of all 3 parameters.

can be expressed

(1)

and

(2)

in terms of

only.

I2y

Kr=

^#

(4 )

Therefore
/),

C,

IZS-2K

Az

=-

*-

IZS

C z * -I/o/k

25/k

D and

7
0^:

Using equation

(5)

to solve for

Using equation

(6)

to solve equation (4-13:


(7)

^<
Inspection of equation

K*
(7)

IZSK ? I2SZ

shows that s =
38

for

K = 86.5.

Since

the only requirement is that

S,*

$:

0.2, this gain setting should

be considered as a possible design value.


Kj.

397 and K

0284

For K = 86.5,

For these values

loop pole is located at d" = -3.4.


are not completely dominant.

the third closed

Therefore, the complex poles

In addition, figure 4-1

sensitivity function is very steep at the point of


sensitivity.

and Ka = 0.0384.

The other design values are:

without proof that

becomes very

Kj-

= 0.238

The complex poles are assured of being

dominant because the third pole

a value <

minimum

Again using figure 4-1, the system will be designed

on the basis of K = 600.

is stated

shows the

large.

It

is

is

now located

at <T~= -24

It

w
also approaches zero as K
S^n

f
also less sensitive than S/

0.05 at the design gain setting.

39

having

*-K

*5/js/7-/

vry

(S$)

vs

x^^^/.s

/-or.

40

G,*/n (k)
4--

5.

Conclusions

The purpose of this paper was to develop a design technique


based on the premise that design can be accomplished by solving
a set of algebraic simultaneous equations derived from the

system design specifications.

In addition to demonstrating the

feasibility of this concept, the examples in the preceeding

sections also showed that the method developed has certain


desirable characteristics.
is analytically simple,

Specifically, the algebraic method

capable of handling any number of

parameters, rapid, specific and accurate.

The method

is in

no way restricted to the exclusive use of

the design specifications considered in this paper.

The

specifications considered were simply selected to enhance the


In general, a system can be designed to satisfy

presentation.

any set

of independent specifications providing the specifications

can be expressed as algebraic equations.

It is

obvious, of

course, that the solution of the resulting equations becomes

more difficult as the number


reason,

it

is

of specifications increase.

For this

suggested that analysis be normally limited to

specifications which are considered to be prime, i.e.

which are exactly specified as opposed to others stated


of inequalities or approximations.

can then be used to determine


also satisfied.
41

if

The results

those
in

terms

of this analysis

the secondary specifications are

The inclusion

of sensitivity equations into the set of

was considered especially

constraint equations

again, the use of sensitivity equations is in no

upon the assumptions made


of root sensitivity to

in Section 4

demonstrated in example 4

was

of particular

way dependent

This

was

use

of root

logically

where a root sensitivity component

concern while the system sensitivity had no

restrictions whatsoever.

discussed, there

Here

significant.

concerning the relation

system sensitivity, i.e.

sensitivity as a performance index.

system

is

Although only root sensitivity was

no apparent reason why similar restrictions

cannot be imposed on other design specifications.

As previously indicated

in Section 4, there is a definite

need

for the study of specification/parameter sensitivity relationships

before sensitivity can become a useful design tool.

At the

present, the sensitivity effects of all possible parameters must

be evaluated with respect to each specification and to each other


before a meaningful design can be achieved.

Analysis of even

a simple system on this basis would require a prohibitive amount


of computation

because

of the large possible

number

of

parameter/

specification combinations available.

"Seek Simplicity and Distrust

It"

Alfred North Whitehead

42

BIBLIOGRAPHY

1.

Clark, R. N. Introduction to Automatic Control Systems,


Wiley and Sons, 1955.

2.

Kokotovic, P.

and D. D. Siljak, "The Sensitivity Problem


Continuous and Sampled-Data Linear Control Systems by
Generalized Mitrovic's Method" IEEE Transactions
(Applications and Industry) v. 83, Sept. 1964: 321-324.
,

in

3.

Mitrovic, D.
"Graphical Analysis and Synthesis of Feedback
Control Systems", AIEE Transactions (Applications and
,

Industry)

v. 77, Jan. 1959:

476-503.

4.

Thaler, G. J.
U. S. Naval Postgraduate School, Monterey,
California, "Algebraic Methods for Dynamic Systems
(Extension and Generalization of Mitrovic's Method)",
unpublished notes.

Thaler and Brown


Systems, 2nd ed

6.

Truxal,

J.

G.

Analysis and Design of Feedback Control


McGraw-Hill, 1960.

Control System Synthesis

43

McGraw-Hill, 1955

"^

o 1CD ^r
^ LO ?
00 CD CD C^ 00
O
O O CD CD
00
CO 00 o
^ O O
i CM CO
LO
o
o
O o o. CM CO CD ^ CD
O o o o o O O r i 1 o o O O

o CD
CD
o "^
o
o O
o ^
o CD
r-H

co r "*r
CO t^
CD r>. CD
t>- CD CO
CD CD CD
r> "^ i

LO CD CD <* l-H
CO ID "^ ^r *?
i-H CD it r-H CD
CO CD t>
LO LO LO i
i-H
CM LO CD

ll

<3*

r-H

ii

ii

>Sl

i-H

t^.

i-H
i-H

i-H

t>-

r>.

i-H

i-H

O
O
^ o
o
O o

CD LO ^r CO 00 CD
c^ CM CM "^ CD CD
t^ CD 00 CO t^
CD
CD LO CD CM
CO
CO CD

i-H

CD

il

t^.

<tf

ii

CO
LO CD
o CM
00 CD o CD 00 r^
CO ^r CM CM t^ o CD "^ o ^ CD o
o o
o CD 00 ^
CO CD CD ^ o CO CO LO
CD
o CD CD CD ^r
LO CO
o CM r^
o CM LO CD o O 00 CD CO o "^ 00
o O O oo
O O o o o O
i-H

r>

is.

il

o
o
o
o
o
^
o o
o CM r^

ii CD i-H
CO t"^ 00
CD t* co
CM LO LO
00 CO
LO
CM

O
^ O
i-H

i-H

o O

CO CD

Oi-HCDi-HCDLOCDCD'^CDOCD^rCD^LO^rrHCDi-HO

OOi-HOOLOCslCDCDOCOOLOCDCOCOr^COCMt^CMO
ir^OCD^i CO 00 CM 00
O O

OMDM^H^Nr
OOlCOMnnONiflSOHNNOlOHlDNNO
O O O
O oooooooo
t>*

r-H
I

>&

Il

<

CO LO
I

00

o o oo
O o

o o o o o o o

ll

p
w
Oh

OOlNCOlDlO^NCOHO CMCOCDlOCDCOCMCDO
O050)N()NXmC0NO01NOl/)N0)HrtmO
Oi-HC0L0r^00CDOOOOCDCD^rOC000L0CMOO

CO

i-H

CI

r-H

CM CO ^T

ooiiCH^m^HioaiOH^oiiowiooi^HO
OCDCDCDCOr^COLOCOi-HOCM^CDCDCMLOCOCMCDO
CO
"Si

i-HOOOOOOOOOOOOOO
I

CM CM CO
I

Or-HCMC0^TL0CDr>.00CDOi-HCMC0^riOCDr^00CDO
CM

oooooooooo
I

CM

>Si

ooooooooooooooooooooo

OlOOlOOlOOlOOlOOloOlOOlOOlOOloO
OOHHNN(0()^^a)miClDNN00C0(J)(J)O

OOOOOOOOOOOOOOOOOOOOr-H

cooswiotDHOiHOOcoNN^room<joo
OOlONOO^'NS'JNOCONHN^roOOSO
iNoionNnHoaioooooo
o noiooNMnorcotDinroNinoiinHHO^oiNOitoooioocoo
>&

ro

s m O ^
I

00
I

o o
r-H

i-H

c-

O
O O LO
t^ O
HCO
II

ID CM ID CO CO
l

r-H

CO
CM CO

i-H

llDLO^OO)COCO(COOO)Sr- Ol H N tv Ol o
O
Or-HCDOCOOCOCDOiiOCO^CDS(31S(D^HO
i

0-)

"^

omcDiDiooinioNcooNcomNcncoNmNO
O^OCni
ILOiICOCMLOOtOtOiit^TffvLOiIOO
ocosrooiN^oiflcjiNHcoocJiaicO'j'Oioo

i CM i CM ^ CO
H CO
II
I

CDi-Hi-HOOCOt^i-HOCDCMOt^.
Or- CO O CO O CO O
cDOrpi/)HcocoHco^oa)HNO)HincnioO)0
tD^CnSNtDO)^0)NOl/)(MO^COCOCT)HCDO
CD
CD O rI^CDCD^CDOLOr ^ ^ CM S r-It^CMO
CM
id in co OCOOOOO)CO(DH
CM N CM N CO
O lO N
o
i
-Hill
CO O to
^T rH
I

-co

-*

i-H

CO >-iCDOi 00 >-1.iO>-HrHOi O i CO
CD O
O
OOCOCT)CMLOCOCDCv]^OmOCMi-HLOCT>CMCX)i-HO
I

O^CO^CONOIOIONOHCO^NNOOCOCOSO
OH^HHCOlOOtOtOlO^HlDOOCOtDCOCOO
CM
ON^CONNlOlOCMO^NOCM N O ^ COCO CO
CD

co.

-)>-<

a,
a,

<

S3
{D
Ph

O
a
p

i-H

.-H

OHlDHCDifi'JOl^cnoCO^OM'miOHlDHO

OCDinCMi-HCM^miDt^OOCMi N N O) CD H CO O
ONOmiOCOCOCONCOO^COOlOlOOHNlDO
I

_ CD

' *

ooLnmcocMOiicolocooocomcnnoocooio
I

CM CO

^ S

is

in

ONCOHCOONlD^OOOLOCOlONOCD^tOCSIO
ocnco<Moooc^LoocDomt^cvLoocDomr^o
OHcooiNO^iOLoooiomcnrtomcDooo
Oi-HCMco^inmmm^^cMOi in en co co CM
o
m n
CO
i

i
I

r}<

^lDO
oor^^omoooticoocMCDi-Ht^mcoco^rcDO
MTOCOCOCINHHOONCO^lONCnHCOlONO
i-Hr-li-li-HCM
0<5 COCD^O'JlOCOtDOCDM 'T'OOCOM
|

O^COCMIDO^COCMCDO'J'CONCOO^OOCMIOO
_ CO

OOOHHNCSKMCOCO^^ifLOlOlDlOCOSNCO
A
I

-CM

tf

CMCMNCMNNNCMCMCMCMCMCMCMNNCMCMNCMCM

oinomomomomomomoinoinomo
K OOHHCMCMCOCO'J^LOLOlOlOSNCOCOCncnO
OOOOOOOOOOOOOOOOOOOOi-h

'-7

"SI

INITIAL DISTRIBUTION LIST

No. Copies
1

Defense Documentation Center

Cameron Station
Alexandria, Virginia
2

4.

22314

Library

U.
3

S.

Naval Postgraduate School, Monterey, Calif.

Naval Material Command


Department of the Navy
Washington, D. C. 20360
Dr. G.

Thaler

J.

10

Department of Electrica 1
U. S. Naval Postgraduate School, Monterey, Calif.
5.

LT Albert

B.

Lemanski,

USN

USS NORTON SOUND (AVM-1)

FPO San Francisco 96601


6.

LT H. H. Choe

Naval Academy
Chinhae, Korea 96669
7.

LT C. H. Hyon

Naval Academy
Chinhae, Korea

96669

8.

Dr. D. D. Siljak
University of Santa Clara
Santa Clara, California

9.

Dr. Z. Slawsky

NOL, White Oaks


Silver Springs, Maryland
10.

CDRR.

L.

20910

Ashford

NOL, White Oaks


Silver Springs, Maryland
11.

20910

Mr. R. Kogge

NOL, White Oaks


Silver Springs
12.

20

Maryland 20910

Mr. Walt Blum burg


U.S.N. Marine Engineering Laboratory
Annapolis, Maryland 21402
i\9

13.

Mr. Seltzer

Code R-ASTR-NG
NASA George Marshall Space
Huntsville, Alabama 35812
14.

Mr. D.

Flight Center

R. Towill

Welch College

of Advanced Technology
Cathays Park
Cardiff, Wales, Great Britain

to

UNCLASSIFIED
Security Classification

DOCUMENT CONTROL DATA


(Security elmaaitieatlon of till; body of abatraet
1.

ORIGINATING ACTIVITY

R&D

and indexing annotation rnuK be entered whan


2a.

(Corporate author)

2b GROUP

REPORT TITLE

LINEAR CONTROL SYSTEM DESIGN USING ALGEBRAIC


DESCRIPTIVE NOTES (Type

4-

Thesis

AUTHOR(S)

LT,

ol report and inetueive datee)

(Laat name,

li rat

name,

initial)

B.

USN

REPORT DATE

May
6a.

b.

METHODS

MSBE

LEMANSKI, Albert
6-

L ASSIFIC A TIOH

UNCLASSIFIED

U. S. Naval Postgraduate School


Monterey, California
3.

the overall report la clae allied)

REPORT SECURITY C

7a.

TOTAL

9a.

ORIGINATOR'S REPORT NUMBERfSJ

NO.

1966

CONTRACT OR GRANT

OF PASES

7b. NO.

49
NO.

OF REFS

PROJECT NO.
9b.

OTHER REPORT

NOfSJ (A ny other numbers that may be aeeifnod

thia report)

d.

10.

AVAILABILITY/LIMITATION NOTICES

11.

SUPPLEMENTARY NOTES

12-

SPONSORING MILITARY ACTIVITY

Nayal Material- Corrim&nd


Department of the Navy
Washington P C
2Q36Q
,

13.

ABSTRACT

A method

presented for the analytic design of linear control


systems. Design is accomplished by the solution of an algebraic
set of simultaneous equations derived from the design specifications;
Specifications considered include the system damping ratio, undamped
natural frequency, bandwidth, steady state error coefficients and root
is

sensitivity.
is shown to be fast, accurate, specific and capable
any number of parameters. In addition, the analytic
approach minimizes dependence on graphical techniques and trial
and error solutions.

The method

of working with

DD ,^,1473

UNCLASSIFIED
Security Classification

"

UNCLASSIFIED
Security Classification

LINK B

LINK A

14.

KEY WORDS

LINK C

WT

ROLE

ROLE

*T

INSTRUCTIONS
ORIGINATING ACTIVITY:

Enter the name and address


of the contractor, subcontractor, grantee, Department of Defense activity or other organization (corporate author) issuing
1.

imposed by security classification, using standard statements


such as:
(1)

the report.

REPORT SECURITY CLASSIFICATION:

2a.
all

security classification of the report.

Enter the overIndicate whether

"Restricted Data" is included. Marking is


ance with appropriate security regulations.

to

be

(2)

in accord(3)

(4)

ized.

REPORT

DDC"

"Foreign announcement and dissemination of this


report by DDC is not authorized.
"U.

S.

Government agencies may obtain copies of


Other qualified DDC

this report directly from DDC.


users shall request through

26. GROUP: Automatic downgrading is specified in DoD Directive 5200. 10 and Armed Forces Industrial Manual. Enter
the group number. Also, when applicable, show that optional
markings have been used for Group 3 and Group 4 as author-

TITLE:

Enter the complete report title in all


Titles in all cases should be unclassified.
If a nv.-aninfcful title cannot be selected without classification, show title classification in all capitals in parenthesis
immediately following the title.
3.

"Qualified requesters may obtain copies of this


report from

"U. S. military agencies


report directly from
shall request through

DDC

may obtaimcepies

of this,
*

Other qualified users

capital letters.

4.

DESCRIPTIVE NOTES:

If

(5)

appropriate, enter the type of

report, e.g., interim, progress, summary, annual, or final.


Give the inclusive dates when a specific reporting period is

If the report has been furnished to the Office of Technical


Services, Department of Commerce, for sale to the public, indicate this fact and enter the price, if known.

covered.
5.
AUTHOR(S): Enter the name(s) of authors) as shown on
or in the report. Enter last name, first name, middle initial.
The name of
If military, show rank and branch of service.
the principal author is an absolute minimum requirement.

11.

TOTAL NUMBER OF PAGES: The

total

NUMBER OF REFERENCES

13-

It is

CONTRACT OR GRANT NUMBER: If appropriate, enter


the applicable number of the contract or grant under which
the report was written.

14.

subproject number, system numbers, task number, etc.


offi-

OTHER REPORT NUMBER(S): If the report has been


assigned any other report numbers (either by the originator

96..

by

KEY WORDS: Key

words are technically meaningful terms

the sponsor), also enter this number(s).

AVAILABILITY/LIMITATION NOTICES: Enter any

limitations on further dissemination of the report, other than those

10.

DD

FORM
1

JAN

64

How-

or short phrases that characterize a report and may be used as


index entries for cataloging the report. Key words must be
selected so that no security classification is required. Identifiers, such as equipment model designation, trade name, military
project code name, geographic location, may be used as key
words but will be followed by an indication of technical context. The assignment of links, roles, and weights is optional.

9a. ORIGINATOR'S REPORT


cial report number by which the document will be identified
and controlled by the originating activity. This number must
be unique to this report.

or

highly desirable that the abstract of classified reports

There is no limitation on the length of the abstract.


ever, the suggested length is from 150 to 225 words.

PROJECT

NUMBER: Enter the appropriate


8d.
86, 8c,
military department identification, such as project number,
Enter the

Enter an abstract giving a brief and factual

be unclassified. Each paragraph of the abstract shall end with


an indication of the military security classification of the information in the paragraph, represented as ( TS), (S), (C), or (V).

8a.

NUMBER(S):

ABSTRACT:

be attached.

Enter the total number of

references cited in the report.

&

explana-

summary of the document indicative of the report, even though


it may also appear elsewhere in the body of the technical report. If additional space is required, a continuation sheet shall

page count

should follow normal pagination procedures, Le., enter the


number of pages containing information.
76.

for additional

12. SPONSORING MILITARY ACTIVITY:


Enter the name of
the departmental project office or laboratory sponsoring (paying for) the research and development. Include address.

REPORT DATE:

7a.

SUPPLEMENTARY NOTES: Use

tory notes.

Enter the date of the report as day,


month, year, or month, year. If more than one date appears
on the report, use date of publication.

6.

"All distribution of this report is controlled. Qualified DDC users shall request through

1473 (BACK)

UNCLASSIFIED
Security Classification

%d

><M

DUDLEY KNOX

mm

5
2768 00416522

IBiiiiilSIH

Dudley knox library

I'.*.!/. ':,..'.

^!&PSW8ffl8wa

.v-'.M.i

.'
:

'

'.v>:'!

; :''.

:,
;

^v!-: '-;v
:

'..'

BHIMffflffl
;',',-'.
v, ,.';;, K
'.

'.'

'

.-'.

'..

Kwalfflfl Baal SB
'uW.
1

>

^iiiSwi
IP'

LIMY

v>

i,>

HH

ilm^^^^M