I

AZ
§ m
--t
ffl ~
80
6 ::0
m"l>
0
-i
~ 2
>
~
-
en
en
>
2
C
en
-<
2
-t
:c
m
-
en

WILEY
TOPPAN
 PETER CLARKE
SECOND HAND BOOKS
Tel. Rochdale 50514

Network
Analysis and
Synthesis

Second Edition
 Wiley International Edition

Network
Analysis and
Synthesis
Second Edition
by Franklin F. Kuo
Bell Telephone Laboratories, Inc.

John Wiley & Sons, Inc., New York I London I Sydney

Toppan Company.Ltd., Tokyo, Japan

Copyright© 1962, 1966 by John Wiley & Sons. Inc.
All Rights Reserved

This book or any part thereof

must not be reproduced in any form
without the written permission of the publisher

Wiley International Edition

This book is not to be sold outside the country

to which it is consigned by the publisher.

Ubray of Congress Catalog Card Number : 66-16127

Printed in Singapore by Toppan Printing Co. (S) Pte. Ltd.
To My Father and Mother
 Preface

In the second edition, I have tried to keep the organization of the first
edition. Most of the new material are additions aimed at strengthening
the weaknesses of the original edition. Some specific changes deserve
mention. The most important of these is a new chapter on computer
applications (Chapter 15). In the past five years, digital computers have
brought about many significant changes in the content of engineering
subject matter concerned with both analysis and design. In analysis,
computation has become an important adjunct to theory. Theory estab-
lishes the foundation of the subject matter; computation provides clarity,
depth, and insight. In design, the computer has not only contributed
precision and speed to existing procedures but has made practicable
design methods that employ iteration and simulation. The importance of
computer-aided design cannot be overemphasized. In Chapter 15 I have
attempted to survey some digital computer applications in the areas of
network analysis and design. I strongly encourage all students to read
this chapter for cultural interest, if not for survival.
Another new section contains a rigorous treatment of the unit impulse.
It was difficult to decide whether to incorporate this m3~rial in Chapter 2
in the discussion of signals or in a separate appendix. By putting general-
ized functions in an appendix, I have left the decision of whether to teach
the rigorous treatment up to the individual instructor.
Other changes worth mentioning are: (1) two new sections on the
Fourier integral in Chapter 3; (2) a section on initial and final conditions
in Chapter 5; (3) a section on Bode plots in Chapter 8; (4) revised
material on two-port parameters in Chapter 9; and (5) new sections on
frequency and transient responses of filters in Chapter 13. Major or
minor changes may be found in every chapter, with the exception of
Chapters 11 and 12. In addition, many new problems are included at the
end of each chapter.
vii
viii Preface
A brief description of the subject matter follows. Chapters 1 and 2
deal with signal representation and certain general characteristics of
linear systems. Chapter 3 deals with Fourier analysis, and includes the
impulse method for evaluating Fourier coefficients. Chapters 4 and 5
discuss solutions of network differential equations in the time domain.
In Chapters 6 and 7, the goals are the same as those of the two preceding
chapters, except that the viewpoint here is that of the frequency domain.
Chapter 8 deals with the amplitude, phase, and delay of a system function.
The final seven chapters are concerned with network synthesis. Chapter
9 deals with two-ports. In Chapter 10, the elements of realizability theory
are presented. Chapters 11 and 12 are concerned with elementary driving-
point and transfer function synthesis procedures. In Chapter 13, some
fundamental concepts in modem filter design are introduced. Chapter 14
deals with the use of scattering matrices in network analysis and design.
And, as mentioned earlier, Chapter 15 contains a brief survey of digital
computer techniques in system analysis. In addition, there are five
appendices covering the rudiments of matrix algebra, generalized functions,
complex variables, proofs of Brune's theorems, and a visual aid to filter
approximation.
The book is intended for a two-semester course in network theory.
Chapters 1 through 8 may be used in a one-semester undergraduate or
beginning graduate course in transient analysis or linear system analysis.
Chapters 9 and 15 are to be used in a subsequent course on network
synthesis.
The second edition is largely a result of the feedback from the professors
who have used this book and from their students, who have discovered
errors and weaknesses in the original edition. I wish to express my sincere
appreciation to those who provided this feedback.
Special thanks are due to the following people: Robert Barnard of
Wayne State, Charles Belove and Peter Dorato of the Polytechnic
Institute of Brooklyn, James Kaiser and Philip Sherman of the Bell
Telephone Laboratories, Evan Moustakas of San Jose State College,
A. J. Welch of the University of Texas, and David Landgrebe of Purdue.
I am particularly indebted to Mac Van Valkenburg of Princeton University
and Robert Tracey of Illinois for editorial advice and to Donald Ford
of Wiley for help and encouragement.
In addition, I wish to thank Elizabeth Jenkins, Lynn Zicchino, and
Joanne Mangione of the Bell Telephone Laboratories for their efficient
and careful typing of the manuscript.
Berkeley Heights, F. F. Kuo
New Jersey,
December, 1965
 •

Preface to the First Ed it ion

This book is an introduction to the study of electric networks based upon

a system theoretic approach. In contrast to many present textbooks,
the emphasis is not on the form and structure of a network but rather on
its excitation-response properties. In other words, the major theme is
concerned with how a linear network behaves as a signal processor.
Special emphasis is given to the descriptions of a linear network by its
system function in the frequency domain and its impulse response in the
time domain. With the use of the system function as a unifying link, the
transition from network analysis to synthesis can be accomplished with
relative ease.
The book was originally conceived as a set of notes for a second course
in network analysis at the Polytechnic Institute of Brooklyn. It assumes
that the student has already had a course in steady-state circuit analysis.
He should be familiar with Kirchhoff's laws, mesh and node equations,
standard network theorems, and, preferably, he should have an elementary
understanding of network topology.
A brief description of the subject matter follows. Chapters 1 and 2
deal with signal representation and certain characteristics of linear net-
works. Chapters 3, 4, 5, and 6 discuss transient analysis from both a time
domain viewpoint, i.e., in terms of differential equations and the impulse
response, and a frequency domain viewpoint using Fourier and Laplace
transforms. Chapter 7 is concerned with the use of poles and zeros in
both transient and steady-state analysis. Chapter 8 contains a classical
treatment of network functions.
The final five chapters deal with network synthesis. In Chapter 9, the
elements of realizability theory are presented. Chapters 10 and 11 are
concerned with elementary driving-point and transfer function synthesis
procedures. In Chapter 12, some fundamental concepts in modern filter
design are introduced. Chapter 13 deals with the use of scattering matrices
Ix
x Preface to the First Edition
•
in network analysis and synthesis. In addition, there are three appendices
covering the rudiments of matrix algebra, complex variables, and proofs
of Brune's realizability theorems. ·
The book is intended for a two-semester cow;se in network theory.
Chapters 1 through 7 can be used in a one-semester undergraduate or
beginning graduate course in transient analysis or linear system analysis.
Chapters 8 through 13 are to be used in a subsequent course on network
synthesis.
It was my very good fortune to have studied under Professor M. E. Van
Valkenburg at the University of Illinois. I have been profoundly influenced
by bis philosophy of teaching and writing, which places strong emphasis
upon clarity of exposition. In keeping with this philosophy, I have tried
to present complicated material from a simple viewpoint, and I have
included a large number of illustrative examples and exercises. In addition,
I have tried to take a middle ground between mathematical rigor and
intuitive understanding. Unless a proof contributes materially to the
understanding of a theorem, it is omitted in favor of an intuitive argument.
For example, in the treatment of unit impulses, a development in terms
of a generalized function is first introduced. It is stressed that the unit
impulse is not r$lly a function but actually a sequence of functions whose
limit point is undefined. Then, the less rigorous, intuitive notion of an
impulse "function" is presented. The treatment then proceeds along the
nonrigorous path.
There are a number of topics which have been omitted. One of these
is network topology, which seems to be in vogue at present. I have pur-
posely omitted topology because it seems out of place in a book that
de-emphasizes the form and structure approach to network analysis.
In an expository book of this nature, it is almost impossible to reference
adequately all the original contributors in the vast and fertile field of
network theory. I apologize to those whose names were omitted either
through oversight or ignorance. At the end of the book, some supple-
mentary textbooks are listed for the student who either wishes to fill in
some gaps in his training or wants to obtain a different point of view.
I acknowledge with gratitude the help and advice given to me by my
colleagues at the Bell Telephone Laboratories and by my former colleagues
at the Polytechnic Institute of Brooklyn. I wish to express my sincere
appreciation to the many reviewers whose advice and criticism were
invaluable in revising preliminary drafts of the manuscript. Professors
R. D. Barnard of Wayne State University and R. W. Newcomb of Stanford
University deserve specific thanks for their critical reading of the entire
manuscript and numerous helpful suggestions and comments.
In addition, I wish to thank Mrs. Elizabeth Jenkins and Miss Elizabeth
 Preface to the First Edition xr
La Jeunesse of the Bell Telephone Laboratories for their efficient and
careful typing of the manuscript.
Finally, to my wife Dora, I owe a special debt of gratitude. Her
encouragement and cooperation made the writing of this book an
enjoyable undertaking.
Mu"ay Hill, New Jersey, F. F. Kuo
JflnlllJry, 1962
 Contents

Cho,cer I: Sipals and Systems I

1.1 Signal Analysis 1

1.2 Complex Frequency 4
1.3 Network Analysis 7
1.4 Network Synthesis u

Chof)ter 2: Si1nals and Waveforms 20

2.1 General Characteristics of Signals 20

2.2 General Descriptions of Signals 24
2.3 The Step Function and Associated Waveforms 28
2.4 The Unit Impulse 33

Cho,cer l: The Frequency Domain: Fourier Analysis 46

3.1 Introduction 46
3.2 Orthogonal Functions 47
3.3 Approximation Using Orthogonal Functions 48
3.4 Fourier Series 50
3.5 Evaluation of Fourier Coefficients 52
3.6 Evaluation of Fourier Coefficients Using Unit
Impulses 58
3.7 The Fourier Integral 63
3.8 Properties of Fourier Transforms 67
xiii
xiv Contents

Chapter 4: Differential Equations 75

4.1 Introduction 75
4.2 Homogeneous Linear Differential Equations 76
4.3 Nonhomogeneous Equations 82
4.4 Step and Impulse Response 85
4.5 lntegrodifferential Equations 91
4.6 Simultaneous Differential Equations 93

Chapter 5: Network Analysis: 100

5.1 Introduction 100

5.2 Network Elements 103
5.3 Initial and Final Conditions 106
5.4 Step and Impulse Response 111
5.5 Solution of Network Equations 114
5.6 Analysis of Transformers 122

Chapter 6: The Laplace Transform l:U

6.1 The Philosophy of Transform Methods 134

6.2 The Laplace Transform 135
6.3 Properties of Laplace Transforms 137
6.4 Uses of Laplace Transforms 144
6.5 Partial-Fraction Expansions 148
6.6 Poles and .zeros 155
6.7 Evaluation of Residues 162
6.8 The Initial and Final Value Theorems 165

Chapter 7: Transform Methods In Network Analysis 175

7.1 The Transformed Circuit 175

7.2 Thevenin's and Norton's Theorems 180
7.3 The System Function 187
7.4 The Step and Impulse Responses 194
7.5 The Convolution Integral 197
7.6 The Duhamel Superposition Integral 201
 Contents xv

Chapter I: Amplitude, Phase, and Delay 212

8.1 Amplitude and Phase Response 212

8.2 Bode Plots 221
8.3 Single-Tuned Circuits 229
8.4 Double-Tuned Circuits 238
8.5 On Poles and Zeros and Time Delay 245

Chapter 9: Network Analysls: II 153

9.1 Network Functions 253

9.2 Relationships Between Two-Port Parameters 264
9.3 Transfer Functions Using Two-Port Parameters 266
9.4 Interconnection of Two-Ports 271
9.5 Incidental Dissipation 276
9.6 Analysis of Ladder Networks 279

Chapter 10: Elements of Realizability Theory 290

10.1 Causality and Stability 29(J

10.2 Hurwitz Polynomials 294
10.3 Positive Real Functions 299
10.4 Elementary Synthesis Procedures 308

Cho,.,. II: Synthesis of One-Port Networks with Two Kinds

of Elements 315

11.l Properties of L-C Immittance Functions 315

11.2 Synthesis of L-C Driving-Point lmmittances 319
11.3 Properties of R-C Driving-Point Impedances 325
11.4 Synthesis of R-C Impedances or R-L. Admittances 329
11.S Properties of R-L Impedances and R-C Admittances 331
11.6 Synthesis of Certain R-L-C Functions 333
xvi Contents

Chaj,te, 12: Elements of Transfer Function Synthesis :UI

12.1 Properties of Tran.sfer Functions 341

12.2 2.eros of Transmission 345
12.3 Synthesis of Y11 and Z 11 with a 1-0 Termination 347
12.4 Synthesis of Constant-Resistance Networks 352

Chapter 13: Topics in FIiter Design 365

13.1 The Filter Design Problem 365

13.2 The Approximation Problem in Network Theory 365
13.3 The Maximally Flat Low-Pass Filter Approximation 368
13.4 Other Low-Pass Filter Approximations 373
13.5 Transient Response of Low-Pass Filters 388
13.6 A Method to Reduce Overshoot in Filters 392
13.7 A Maximally Flat Delay and Controllable Magnitude
Approximation 395
13.8 Synthesis of Low-Pass Filters 397
13.9 Magnitude and Frequency Normalization 402
13.10 Frequency Transformations 404

Chapter 14: The Scattering Matrix 413

14.1 Incident and Reflected Power Flow 413

14.2 The Scattering Parameters for a One-Port Network 415
14.3 The Scattering Matrix for a Two-Port Network 419
14.4 Properties of the Scattering Matrix 426
14.5 Insertion Loss 429
14.6 Darlington's Insertion Loss Filter Synthesis 431

Chapter 15: Computer Techniques in Circuit Analysis 438

15.1 The Uses of Digital Computers in Circuit Analysis 438

15.2 Amplitude and Phase Subroutine 450
15.3 A Fortran Program .for the Analysis of Ladder
Networks 453
15.4 Programs that Aid in Darlington Filter Synthesis 457
 Contents xvii

Appendix A: Introduction to Matrix Algebra 461

A. I Fundamental Operations 461

A.2 Elementary Concepts 462
A.3 Operations on Matrices 464
A.4 Solutions of Linear Equations 468
A.5 References on Matrix Algebra 469

Appendix 8: Generalized Functions and the Unit Impulse 470

B.1 Generalized Functions 470

B.2 Properties of the Unit Impulse 476

Appendix C: Elements of Complex Variables 481

C. l Elementary Definitions and Operations 481

C.2 Analysis 483
C.3 Singularities and Residues 486
C.4 Contour Integration 487

Appendix D: Proofs of Some Theorems on Positive Real

Functions 490

Appendix E: An Aid to the Improvement of Filter Approxi-

mation 493

E.1 Introduction 493

E.2 Constant Logarithmic Gain Contours 494
E.3 Constant Phase Contours 495
E.4 Contour Drawings 496
E.5 Correction Procedure 498
E.6 Correction Network Design 502·
E. 7 Conclusion 504

Bibliography sos
Name Index 509

Subject Index SIi

 I chapter
Signals and systems

This book is an introduction to electric network theory. The first half

of the book is devoted to network analysis and the remainder to network
synthesis and design. What are network analysis and synthesis? In a
generally accepted definition of network analysis and synthesis, there are
three key words: the excitation, the network, and the response as depicted
in Fig. 1.1. Network analysis is concerned with determining the response,
given the excitation and the network. .In network synthesis, the problem
is to design the network given the excitation and the desired response.
In this chapter we will outline some of the problems to be encountered
in this book without going into the actual details of the problems. We
will also discuss some basic definitions.

I.I SIGNAL ANALYSIS

For electric networks, the excitation and response are given in terms of
voltages and currents which are functions of time, t. In general, these
functions of time are called signals. In describing signals, we use the two
universal languages of electrical engineering-time and frequency. Strictly
speaking, a signal is a function of time. However, the signal can be
described equally well in terms of spectral ot frequency information. As
between any two languages, such as French and German, translation is
needed to render information given in one language comprehensible in the

- •l. .__ N_etwork

___ :_Res--"°:►se
FIG. I.I. The objects of our concern.
I
2 Network analysis and synthesis

I
I

,,L,___ FIG, 1.2. Sinusoidal signal.

e-

other. Between time and frequency, the translation is effected by the

Fourier series, the Fourier integral, and the Laplace transform. We shall
have ample opportunity to define and study these terms later in the book.
At the moment, let us examine how a signal can be described in terms of
both frequency and time. Consider the sinusoidal signal

(1.1)

where A 0 is the amplitude, 00 is the phase shift, and co0 is the angular
frequency as given by the equation
21r
Wo=- (1.2)
T
where T is the period of the sinusoid. The signal is plotted ag:.inst time
in Fig. 1.2. An equally complete description of the signal is obtained ifwe

0 WO
Angular frequency
FIG. I.la. Plot of amplitude A versus angular frequency Ci>.

WO
Angular frequency
FIG. 1,3&. Plot of phase 6 versus angular frequency Ci>.
 Signals and systems 3
A

FIG. Ua. Discrete amplitude spectrum.

'

FIG. 1.4b. Discrete phase spectrum.

let the angular frequency robe the independent variable. In this case, the
signal is described in terms of A 0, ro 0 , and 00 , as shown in Fig. 1.3a, where
amplitude is plotted against frequency, and in Fig. 1.3b, where phase shift
is plotted.
Now suppose that the signal is made up of2n + 1 sinusoidal components
,.
s(t) I A, sin (ro,t + 0,)
=,_,. (1.3)

The spectral description of the signal would then contain 2n + I lines at

± ro1 , ± ro 2,
••• , ± ro,., as given in Figs. 1.4a and b. These discrete spectra
of amplitude A versus ro and phase shift 0 versus ro are sometimes called
line spectra. Consider the case when the number of these spectral lines
become infinite and the intervals roi+ 1 - ro, between the lines approach
zero. Then there is no longer any discrimination between one frequency
and another, so that the discrete line spectra fuse into a continuous spectra,
as shown by the example in Figs. I.Sa and b. In the continuous case, the
sum in Eq. 1.3 becomes an integral

·s(t) = L: A(w) sin [wt+ 0(w)J dw (1.4)

where A(ro) is known as the amplitude spectrum and 0(w) as the phase
spectrum.
As we shall see later, periodic signals such as the sine wave in Fig. 1.2
can be described in terms of discrete spectra through the use of Fourier
series. On the other hand, a nonperiodic signal such as the triangular
◄ Network analysis and synthesis
(w)

-w
FIG, I.Sa. Continuous amplitude spectrum.

6(w)

-w

FIG, I.Sb, Continuous phase spectrum.

pulse in Fig. 1.6 can only be described in terms of continuous spectra

through the Fourier integral transform.

1.2 COMPLEX FREQUENCY

In this section, we will consider the concept of complex frequency. As

we shall see, the complex frequency variable

s == <1 + jw (1.5)

is a generalized frequency variable whose real part <1 describes growth and
decay of the amplitudes of signals, and whose imaginary partjw is angular
frequency in the usual sense. The idea of complex frequency is developed
by examining the cisoidal signal
S(t) = Ae1o>' (1.6)
l(t)

0 T t
FIG, 1.6. Triangular signal.
 Signals and systems 5
lm8

Re&

FIG. 1.7. Rotating phasor.

when S(t) is represented as a rotating phasor, 1 as shown in Fig. 1.7.

The angular frequency w of the phasor can then be thought of as a velocity
at 'the end of the phasor. In particular the velocity w is always at right
angles to the phasor, as shown in Fig. 1.7. However, consider the general
case when the velocity is inclined at any arbitrary angle V' as given in
Figs. I.Sa and I.Sb. In this case, if the velocity is given by the symbol s,
we see that sis composed of a component w at right angle to the phasor S
as well as a component u, which is parallel to S. In Fig. I.Sa, s has a
component -u toward the origin. As the pbasor S spins in a counter-
clockwise fashion, the phasor decreases in amplitude. The resulting wave
for the real and imaginary parts of S(t) are damped sinusoids as given by
Re S(t) = Ae-0-1 cos wt
(1.7)
Im S(t) = Ae--' sin wt

'"' (b)
FIG. 1.8. (a) Rotating phasor with exponentially decreasing amplitude. (b) Rotating
phasor with exponentially increasing amplitude.
1
A phasor S is a complex number characteriz.ed by a magnitude and a phase angle
(see Appendix C).
6 Network analysis and synthesis

-. L Envelope = Ae-o-t

----
0 ---

FIG, 1.9. Damped sinusoids.

lmSW· /
/ ' Envelope •Aeo-t

"'
''
''
FIG, 1.10. Exponentially increasing sinusoid.
 Signals and systems 7
8(t)

FIG. I.II. Exponential signals.

which are shown in Fig. 1.9. Note that the damped sinusoid has an
exponential envelope decay, Ac"'. In Fig. I.Sb, the phasor is shown with
a positive real component of velocity +O'. Therefore, as the phasor spins,
the amplitudes of the real and imaginary parts increase exponentially
with an envelope Ae+••, as shown by Im S(t) in Fig. 1.10.
From this discussion, it is apparent that the generaliz.ed cisoidal signal

S( t) = Ae"' = Ae'"+1"'u (1.8)

describes the growth and decay of the amplitudes in addition to angular
frequency in the usual sense. When O' = 0, the sinusoid is undamped, and
when jro = 0, the signal is an exponential signal

S(t) = Ar"' (1.9)

as shown in Fig. 1.11. Finally, if O' = jro = 0, then the signal is a constant
A. Thus we see the versatility of a complex frequency description.

1.3 NETWORK ANALYSIS

As mentioned before, the characterization of the excitation and response

signals in time and frequency makes up only part of the analysis problem.
The other part consists of characterizing the network itself in .terms of
time and frequency, and determining how the network behaves as a signal
processer. Let us turn our attention now to a brief study of the properties
of linear networks and the general characteristics of signal processing by
a linear system.
8 Network analysis and synthesis

BASIC DEFINITIONS
Linear
A system (network) is linear if (a) the principle of superposition and
(b) the principle of proportionality hold.
By the superposition principle, if, for a given network, [e1(t), r1 (t)J and
[e1(t), r 1(t)] are excitation-response pairs, then if the excitation were
e(t) = e1(t) + e1(t), the response would be r(t) = r1(t) + r.(t). By the
proportionality principle, if the excitation were C1ei(t), where C1 is a
constant, then the response would be C1ri(t), i.e., the constant of propor-
tionality C1 is preserved by the linear network. The two conditions of
linearity are summarized in Fig. 1.12.
Another definition of a linear network is that the excitation and response
of the network are related by a linear differential equation. We shall
discuss this definition in Chapter 4 on differential equations.
Passive
A linear network is passive8 if (a) the energy delivered to the network is
nonnegative for any arbitrary excitation, and (b) if no voltages or currents
appear between any two terminals before an excitation is applied.
Reciprocal
A network is said to be reciprocal if when the points of excitation and
measurement of response are interchanged, the relationship between

C1ei(t)
System

C1112(t) _ System Csrs(t)

Ciei(t) +C,,ea(t). C11'J.(I) + C,,ll(t)

- - - - ~ System -

FIG. 1.12. Linear system.

1 G. Raisbeck, "A Definition of Passive Linear Networks in Terms of Time and

Energy;' J. Appl. Phys., 25 (Dec. 1954), 1510-1514.
 Signals and systems 9
e(t)
T(t)

FIG. 1.13. Time-invariant system.

excitation and response remains the same. Thus must be true for any
choice of points of excitation and response.
Causal
We say a system is causal if its response is nonanticipatory, i.e., if
e(t) =0 t< T
(1.10)
then r(t) = 0 t< T
In other words, a system is causal if before. an excitation is applied at
t = T, the response is zero for - oo < t < T.
Time invariant
A system is time invariant if e(t)-+ r(t) implies that e(t :I: T)-+ r(t :I: T),
where the symbol-+ means "gives rise to." To understand the concept
of time invariance in a linear system, let us suppose that initially the
excitation is introduced at t = 0, which gives rise to a response r(t). If
the excitation were introduced at t = T, and if the shape of the response
waveform were the same as in the first case, but delayed by a time T
(Fig. 1.13), then we could say the system is time invariant. Another way
of looking at this concept is through the fact that time-invariant systems
contain only elements that do not vary with time. It should be mentioned
here that linear systems need not be time invariant.
Derivative property
From the time-invariant property we can show that, if e(t) at the input
gives rise to r(t) at the output (Fig. 1.14), then, if the input were e'(t),
IO Network analysis and synthesis

e(t)----,>>-11-1----;)l,...,r(t)

";:'>---·11,,11- 1-1--.. . . ":1''

FIG. 1,14. Some implications of linear time-invariant systems.

i.e., the derivative of e(t), the response would be r'(t). The proof is quite
simple. Consider an excitation e(t + E) where Eis a real quantity. By the
time-invariant property, the response would be r(t + E). Now suppose the
excitation were
1
ei(t) = - (e(t + E) - e(t)] (1.11)
E

then according to the linearity and time-invariant properties, the response

would be
1
ri(t) = - [r(t + E) - r(t)J (1.12)
E

Taking the limit as E-+ 0, we see that

lim e1 (t) = E.. e(t)
,➔ o dt
(1.13)
lim r 1(t) = E.. r(t)
,➔ o dt
We can extend this idea to higher derivatives as well as for the integrals of
e(t) and r(t), as shown in Fig. 1.14.
Ideal models
Let us now examine some idealized models of linear systems. The
systems given in the following all have properties which make them very
useful in signal processing.
 Signals and systems 11

e<
__,,____,.MI Am- I x•• •
FIG. I.IS. Amplifier.

I__K_£_ __,----►~K",/,'''
_.fi_n._,--Ji!M...

FIG. 1.16. Differentiator.

FIG. 1.17. Integrator.

__t<_t>____JMI..._ _D<_T'_ __.I IMI,

FIG. I.II. Time-delay network.

t
FIG. I.19. Excitation function.

1. Amplifier: An amplifier scales up the magnitude of the input, i.e.,

r(t) == Ke(t), where Kis a constant (Fig. 1.15).
2. Differentiator: The input signal is differentiated and possibly scaled
up or down (Fig. 1.16).
3. Integrator: The output is the integral of the input, as shown in
Fig. 1.17.
4. Time delayer: The output is delayed by an amount T, but retains the
same wave shape as the input (Fig. 1.18).
Suppose we take the triangular pulse in Fig. 1.19 as the input signal.
Then the outputs for each of the four systems just described are shown
in Figs. 1.20a-l.20d.
12 Network analysis and synthesis
r(t)
r(t)

1---
0 ;1 12 r

t
-1 ....
(a) (b)
r(t)
1

r(t)

(cJ (dJ
FIG. 1.20. ,(a) Amplifier output. (b) Differentiator output. (c) Integrator output. (d)
Delayed output.

Ideal elements
In the analysis of electric networks, we use idealized linear mathematical
models of physical circuit elements. The elements most often encountered
are the resistor R, given in ohms, the capacitor C, given in farads, and
the inductor L, expressed in henrys. The endpoints of the elements are
called terminals. A port is defined as any pair of two terminals into which
energy is supplied or withdrawn or where network variables may be
measured or observed. In Fig. 1.21 we have an example of a two-port
network.
The energy sources that make up the excitation functions are ideal
current or voltage sources, as shown in Figs. 1.22a and b. The polarities

FIG. 1.21. Two-port network.

 Signals and systems 13

FIG. l.22o. Voltage source. FIG. 1.22b. Current source.

indicated for the voltage source and the direction of tlow for the current
source are arbitrarily assumed for reference purposes only. An ideal
voltage source is an energy source that provides, at a given port, a voltage
signal that is independent of the current at that port. If we interchange the
words "current" and ''voltage" in the last definition, we then define an
ideal current source.
In network analysis, the principal problem is to find the relationships
that exist between the currents and voltages at the ports of the network.
Certain simple voltage-current relationships for the network elements also
serve as defining equations for the elements themselves. For example,
when the currents and voltages are expressed as functions of time, then the
R, L, and C elements, shown in Fig. 1.23, are defined by the equations

v(t) == Ri(t) or i(t) == !. v(t)

R

v(t) == L di(t)
dt
or i(t) == !
L
f.'
o
v(z) dz + i(O) (1.14)

v(t) =!
C
f.'
o
i(z) dz + v(O) or i(t) = C dv(t)
dt
where the constants of integratJ.on i(O) and v(O) are initial conditions to be
discussed in detail later.
Expressed as a function of the complex frequency variables, the equations

i(.t) i(t) i(t)

"'' R
+
11(1)
L ~:7c
~
(a) (b) (c)
FIG. 1.23. (a) Resistor. (b) Inductor. (c) Capacitor.
14 Network analysis and synthesis
I(s) I(s)
+ +

V(s) V(s) sL

(a) (b) (c)

FIG. 1.24. (a) Resistor. (b) Inductor. (c) Capacitor.

defining the R, L, and C elements, shown in Fig. 1.24, are (ignoring

initial conditions for the moment)

V(s) = Rl(s) or I(s) = -1 V(s)

R

V(s) = sLI(s) or l(s) = -1 V(s) (1.15)

sL
1
V(s) = -I(s) or I(s) = sCV(s)
sC
We see that in the time domain, i.e., where the independent variable is t,
the voltage-current relationships are given in terms of differential equations.
On the other hand, in the complexfrequency domain, the voltage-current
relationships for the elements are expressed in algebraic equations.
Algebraic equations are, in most cases, more easily solved than differential
equations. Herein lies the raison d'etre for describing signals and networks
in the frequency domain as well as in the time domain.
When a network is made up of an interconnection of linear circuit
elements, the network is described by its system or transfer function H(s).
The response R(s) and the excitation E(s) are related by the equation

R(s) = H(s) E(s). (1.16)

In network analysis, we are given E(s), and we can obtain H(s) directly
from the network. Our task is to determine R(s).

1.4 NETWORK SYNTHESIS

We will now briefly introduce some of the problems germane to network

synthesis. In network synthesis, we are given the response R(s) and the
excitation E(s), and we are required to synthesize the network from the
 Signals and systems IS

+
v,,, R

FIG. 1.25. Driving-point impedance FIG. 1.26. Black box.

Z(s) = R.

system function
H(s) = R(s) (1.17)
E(s)
Since R(s) and E(s) are voltages or currents, then H(s) is denoted generally
as an immittance if R(s) is a voltage and E(s) is a current, or vice versa.
A driving-point immittance8 is defined to be a function for which the
variables are measured at the same port. Thus a driving-point impedance
Z(a) at a given port is the function
Z(s) = V(s) (1.18)
l(s)
where the excitation is a current /(s) and the response is a voltage V(s), as
shown in Fig. 1.25. When we interchange the words "current" and
''voltage" in the last definition, we then have a driving-point admittance.
An example of a driving-point impedance is the network in Fig. 1.25,
where
Z(s) = V(s) = R (1.19)
l(s)
Now suppose the resistor in Fig. 1.25 were enclosed in a "black box."
We have no access to this black box, except at the terminals 1-1' in Fig.
1.26. Our task is to determine the network in the black box. Suppose we
are given the information that, for a given excitation I(s), the voltage
response V(s) is proportional to /(s) by the equation
V(s) = K I(s) (1.20)
An obvious solution, though not unique, is that the network consists of a
resistor of value R = K 0. Suppose next that the excitation is a voltage
V(s), the response is a current /(s), and that

Y(s) = I(s) = 3 + 4s (1.21)

V(s)
• IRE Standards on Circuits "Linear Passive Networks," Proc. IRE, 48, No. 9
(Sept. 1960), 1608-1610.
16 Network analysis and synthesis
r-----,

:~-,~ I+. ~tl +·

I1(B)

Two-port
network

L-----...J
FIG. 1.27. Network realiza. FIG. 1.21. Two-port network.
tion for Y(s).

Our task is to synthesi7.C a network equivalent to the network in the black

box. From a close scrutiny of the driving-point admittance Y(s), we see
n
that a possible solution might consist of a resistor of l in parallel with
a capacitor of 4 farads, as seen in Fig. 1.27.
The problem of driving-point synthesis, as shown from the examples
just given, consists of decomposing a given immittance function into
basic recognizable parts (such as 3 + 4s). Before we proceed with the
mechanics of decomposition, we must first determine whether the function
is realizable, i.e., can it be synthesized in terms of positive resistances,
inductances, and capacitances? It will be shown that realizable driving-
point immittances belong to a class of functions known as positive real or,
simply, p.r. functions. From the properties ofp.r. functions, we can test a
given driving-point function for realizability. (The Appendices present a
short introduction to complex variables as well as the proofs of some
theorems on positive real functions.) With a knowledge of p.r. functions,
we then go on to examine special driving-point functions. These include
functions which can be realized with two kinds of elements only-the L-C,
R-C, and R-L immittances.
Next we proceed to the synthesis of transfer functions. According to
the IRE Standards on passive linear networks,' a transfer Junction or
transmittance is a system function for which the variables are measured at
different ports. There are many different forms which a transfer function
might take. For example, consider the two-port network in Fig. 1.28.
If the excitation is / 1(s) and the response V.J..s), the transfer function is a
transfer impedance
(1.22)

On the other hand, if V1(s) were the excitation and V.(s) the response, then
we would have a voltage-ratio transfer function

(1.23)

'Loe. cit.
 Signals and systems 17
IH(.i<,,,)I

Ai--------.

0 "'c "'
FIG. 1.29. Ideal amplitude spectrum for low-pass filter.

As for driving-point functions, there are certain properties which a

transfer function must satisfy in order to be realizable. We shall study
these realizability conditions and then proceed to the synthesis of some
simple transfer functions.
The most important aspect of transfer function synthesis is filter design.
A filter is defined as a network which passes a certain portion of a fre-
quency spectrum and blocks the remainder of the spectrum. By the term
''blocking," we imply that the magnitude response IH(jw)I of the filter is
approximately zero for that frequency range. Thus, an ideal low-pass
filter is a network which passes all frequencies up to a cutoff frequency
we, and blocks all frequencies above we, as shown in Fig. 1.29.
One aspect of filter design is to synthesize the network from the transfer
function H(s). The other aspect deals with the problem of obtaining a .
realizable transmittance H(s) given the specification of, for example, the
magnitude characteristic in Fig. l.29. This part of the synthesis is generally
referred to as the approximation problem. Why the word "approxi-
mation?" Because frequency response characteristics of the R, L, and C
elements are continuous (with the exception of isolated points called
resonance points), a network containing these elements cannot be made to
cut off abruptly at we in Fig. 1.29. Instead, we can realize low-pass filters
which have the magnitude characteristics of Fig. 1.30. In connection with
IH(jw)I IH(jw)I

--:,

0 "'c W O We
FIG. 1.30. Realizable low-pass filter characteristics.
18 Network analysis and synthesis
the filter design problems, we will discuss certain problems in magnitude
and frequency normalization so that, in designing a filter, we deal with
element values such as R - 0.5 ohm and C - 2 farads instead of"practical"
element values of, for example, R - 500,000 ohms and C - 2 picofarads
(pico= 10-11). Also we will study a method whereby low-pass filter
designs might be transformed into high-pass, band-pass, and band-
elimination filters. The mathematical basis of this method is called
frequency transformation.
We next discuss some aspects of analysis and synthesis in which the
excitation and response functions are given in terms of power rather than
of voltage and current. We will examine the power-transfer properties of
linear networks, using scattering parameters, which describe the incident
and reflected power of the network at its ports.
Finally, in Chapter 15, we will examine some of the many uses ofhigh-
speed digital computers in circuit analysis and design. In addition to a
general survey of the field, we will also study some specific computer
programs in circuit analysis.

Problems
1.1 Draw the line spectra for the signal

.!(t) - 3 sin ( t + -i) + 4 sin ( 2t - i) + 6 sin 3t

1.2 Find the response to the excitation sin t into a sampler that closes every
K,r/4 seconds where K - 0, 1, 2, . . . . Draw the response for O :S: t :S: 2,,,..
1.3 Find the response to the excitation shown in the figure when the network
is (a) an ideal differentiator; (b) an ideal integrator.

w(t)

2
PROB. 1.3

1.4 If the system function of a network is given as

1
H(s) - (s + 2)(s + 3)
 Signals and systems 19
find the response R(s) if the excitation is
3
E(s) =-
s
1.5 Given the driving-point functions find their simplest network realiz-
ations.
1
(a) Z(s) = 3 + 2s + Js
3s
(b) Y(s) =2s + - -
s+2

(c) Z(s) =3+ r s+ 2

l 2s
(d)
Y(s) = 3s + 2 + s1 + 4

1.6 For the network shown, write the mesh equation in terms of (a) differential
equations and (b) the complex-frequency variables.
R L

+
II C

PROB. 1.6

1.7 For the network shown, write the node equation in terms of (a) differential
equations and (b) complex-frequency form.
II

C R

PROB. 1.7

1.8 S!TF°:; the response of a linear system to an excitation e(_t) were

r(t) = 3e . What would the response be to an excitation of e(t - 2)?
 chapter 2
Signals and waveforms

Our main concern in this chapter is the characterization of signals as

functions of time. In previous studies we have dealt with d-c signals
that were constant with time, or a-c signals which were sinusoids of
constant amplitude, such as s(t) = A sin (rot + 0). In engineering practice,
the class of signals encountered is substantially broader in scope than
simple a-c or d-c signals. To attempt to characterize each member of
the class is foolhardy in view of the almost infinite variety of signals
encountered. Instead, we will deal only with those signals that can be
characterized in simple mathematical terms and which serve as building
blocks for a large number of other signals. We will concentrate on formu-
lating analytical tools to aid us in describing signals, rather than deal with
the representation of specific signals. Because of time and space limita-
tions, we will cover only signals which do not exhibit random behavior,
i.e., signals which can be explicitly characterized as functions of time.
These signals are often referred to as deterministic signals. Let us first
discuss certain qualitative aspects of signals in general.

2.1 GENERAL CHARACTERISTICS OF SIGNALS

In this section we will examine certain behavior patterns of signals.

Once these patterns are established, signals can be classified accordingly,
and some simplifications result. The adjectives which give a general
qualitative description of a signal are periodic, symmetrical, and continuous.
Let us discuss these terms in the given order.
First, signals are either periodic or aperiodic. If a signal is periodic, then
it is described by the equation
s(t) = s(t ± kT) k = 0, 1, 2, . . • (2.1)
20
 Signals and waveforms 21

'"' 1

-2T -T O T 2T 3T 4T ST
-1

FIG. 2. I. Square wave.

where Tis the period of the signal. The sine wave, sin t, is periodic with
period T == 2,r. Another example of a periodic signal is the square wave
given in Fig. 2.1. On the other hand, the signals given in Fig. 2.2 are
aperiodic, because the pulse patterns do not repeat after a certain finite
interval T. Alternatively, these signals may be considered "periodic" with
an infinite period.
Next, consider the symmetry properties of a signal. The key adjectives
here a,e even and odd. A signal function can be even or odd or neither.
An even function obeys the relation

s(t) == s(-t) (2.2)

For an odd function s(t) == -s(-t) (2.3)

For example, the function sin tis odd, whereas cost is even. The square
pulse in Fig. 2.2a is even, whereas the triangular pulse is odd (Fig. 2.2b).
Observe that a signal need not be even or odd. Two examples of signals
of this type are shown in Figs. 2.3a and 2.4a. It is significant to note,
however, that any signal s(t) can be resolved into an even component s,(t)
and an odd component so(t) such that

s(t) == s,(t) + s0(t) (2.4)

For example, the signals in Figs. 2.3a and 2.4a can be decomposed into
odd and even components, as indicated in Figs. 2.3b, 2.3c, 2.4b, and 2.4c.

l(t)
l(t)

0 -f

'"'
(b)
FIG. 2.2. (a) Even function. (b) Odd function.
22 · Network analysis and synthesis
s(t)
s(t) 11----
1

½ 0
1
-1

(a)

(a)
se(t)

½
0

-1 0 1 t

(b) (b)

ao(t)

½
1
½----
-1 0 t 0 t
-----'-½
-½
(c) (c)

FIG. 2.3. Decomposition into FIG. 2.4. Decomposition into

odd and even components. even and odd components.
(a) Original function. (b) (a) Unit step function. (b)
Even part. (c) Odd part. Even part or unit step. (c)
Odd part or unit step.
From Eq. 2.4 we observe that
s(-t) = s.(-t) + s0(-t)
(2.5)
- s.(t) - s0(t)
Consequently, the odd and even parts of the signal can be expressed as
s.(t) == l[s(t) + s(-t))
(2.6j
sJ.t) == l[s(t) - s(-t))
Consider the signal s(t), shown in Fig. 2.5a. The functions( -t) is equal
to s(t) reflected about the t == 0 axis and is given in Fig. 2.Sb. We then
obtain s.(t) and s0(t) as shown in Figs. 2.Sc and d, respectively.
 Signals and waveforms 13

-''' at-,,
1.,___ _
-----11

0
,., 1
' -1 0
,,,,

··"' Slit)

t t
-1
t
-1 0 1
' 0
-t
1

(c> (tl)
FIG. 2.5•. Decomposition into odd and even components from .s(t) and .s(-t).

Now let us tum our attention to the continuity property of signals.

Consider the signal shown in Fig. 2.6. At t - T, the signal i~ discontinuous.
The height of the discontinuity is
f(T+) - f(T-) - A (2.7)

where f(T +) - ....

lim/(T + «)
(2.8)
....
f(T-) - limf(T- c)

and • is a real positive quantity. In particular, we are concerned with

discontinuities in the neighborhood oft - 0. From Eq. 2.8, the points
IW

0 T t
FIG. 2.6. Signal with discontinuity.
l◄ Network analysis and synthesis
a(t)

K -----·--·-------.

0 T1

FIG. 2.7. Signal with two discontinuities.

/(0+) and /(0-) are

f(O+) = lim/(E)
E ➔O

(2.9)
/(0-) = lim/(-E)
E ➔O

For example, the square pulse in Fig. 2. 7 has two discontinuities, at T1 and
T1• The height of the discontinuity at T1 is

(2.10)

Similarly, the height of the discontinuity at T1 is -K.

2.2 GENERAL DESCRIPTIONS OF SIGNALS

In this section we consider various time domain descriptions of signals.

In particular, we examine the meanings of the following terms: time
constant, rms value, d-c value, duty cycle, and crest factor. The term, time
constant, refers only to exponential waveforms; the remaining four
terms describe only periodic waveforms.

Time constant
In many physical problems, it is important to know how quickly a
waveform decays. A useful measure of the decay of an exponential is the
time constant T. Consider the exponential waveform described by

r(t) = KrlT u(t) (2.11)

From a plot of r(t) in Fig. 2.8, we see that when t = T,

r(T) = 0.37r(0) (2.12)

Also r(4T) = 0.02,(0) (2.13)

 Signals and waveforms 25
1.00

0.75
r(IJ

0.50
0.37
0.25

0.02
0 1 2 3 4 5

'
FIG. 2.1. Normalized curve for time constant T- 1.

Observe that the larger the time constant, the longer it requires for the
waveform to reach 37 % of its peak value. In circuit analysis, common
time constants are the factors RC and RfL.
RMS Value
The rms or root mean square value of a periodic waveform e(t) is defined
as
erm, - [~ LTe-Ct) dt] ~ (2.14)

where Tis the period. If the waveform is not periodic, the term rms does
not apply. AB an example, let us calculate the rms voltage for the periodic
waveform in Fig. 2.9.

erms -{½[LT/l(2: dt +s:,. A d,]f' ,r 1

· _{![4AI
T r•
~IT/I+ A•tlTTtJ}~
3 o

- (~· + ~-r, (2.15)

-.J2/3A v
D-C Value
The d-c value of a waveform has meaning only when the waveform is
periodic. It is the average value of the waveform over one period

ed•c - 1STe(t) dt
- (2.16)
T •
26 Network analysts and synthesis
e(t)

-A

FIG. 2.9. Periodic waveform.

The square wave in Fig. 2.1 has zero d-c value, whereas the waveform in
Fig. 2.9 has a d-c value of

e4.c =![AT - AT] = - !1 v (2.17)

T 4 2 4
Duty cycle
The term duty cycle, D, is defined as the ratio of the time duration of the
positive cycle t, of a periodic waveform to the period, T, that is,

D = !!T (2.18)

The duty cycle of a pulse train becomes important in dealing with wave-
forms of the type shown in Fig. 2.10, where most of the energy is con-
centrated in a narrow pulse of width t,.
The rms voltage of the waveform
in Fig. 2.10 is
erma == (½iqA dt)~ 9

(2.19)
= A.jt,/T
= A.JD

_ _fr_A-"!'°.__·
I•
----*-!'
T-...J
°.____ t
FIG. 2.10. Paiodic waveform with small duty cycle.
 Signals and waveforms 27

FIG.1.11. Periodic wawform with zero d-c and small duty cycle.

We see that the smaller the duty cycle, the smaller therms voltage. The
square wave in Fig. 2.1 has a 50 % duty cycle.
Crest factor
Crest factor is defined as the ratio of the peak voltage of a periodic
waveform to therms value (with the d-c component removed). Explicidy,
for any waveform with zero d-c such as the one shown in Fig. 2.11-crest
factor, CF, is defined as
CF=~ or ..!L (2.20)
erma erma

whichever is greater. For the waveform in Fig. 2.11, the peak-to-peak

voltage is defined as
(2.21)

Since the waveform has zero d-c value

(2.22)
Also, (2.23)

and e0 = e,.i,(1 - D) (2.24)

The rms value of the waveform is

e,.,.1(1 - D)1 t0 + e.,,.1 D''(T- 10))'°'

erma == ( T (2.2.S)
= e,.,.,J D(l - D)

1 G. Justice, ''The Significance of Crest Factor," Hewlett-Packard Journal, 15, No. 5

(Jan., 1964), 4-5.

28 Network analysis and synthesis
Since crest factor CF = eJerD18, we have

CF = e,.(1 - D)
e,,.,.jD(l - D) (2.26)
= ,Jl/D - 1
For example, if D = 1~ ,

CF =~ (2.27)
= ,J100 - 1 ~ 10
1
If D = 10,000 '
CF= .J 10,000 - 1 ~ 100 (2.28)
A voltmeter with high crest factor is able to read accurately rms values of
signals whose waveforms differ from sinusoids, in particular, signals with
low duty factor. Note that the smallest value of crest factor occurs for
the maximum value of D, that is, I>max = 0.5,
CFm1n = .Jl/Dmax - 1
(2.29)
=l

2.3 THE STEP FUNCTION AND ASSOCIATED WAVEFORMS

The unit step function u(t) shown in Fig. 2.12 is defined as

u(t) =0 t <0
(2.30)
=1 t ~ 0
The physical analogy of a unit step excitation corresponds to a switch S,
which closes at t = 0 and connects a d-c battery of 1 volt to a given circuit,
as shown in Fig. 2.13. Note that the unit step is zero whenever the
u(t)

--·.=-=-==. "~7._·_etWOl'k________.,
FIG. 2.12. Unit step function. FIG. 2.1 J. Network analog of unit step.
 Signals and waveforms 29
B(t)
ll(t)
4 ______ ,.....___
u(t-a)
1 -

0 G t 0 1 2

FIG. 2.14. Shifted step function. FIG. 2.15. Square pulse.

argument (t) within the parentheses is negative, and is unity when the
argument (t) is greater than uro. Thus the function u(t - a), where
a > 0, is defined by
u(t- a)== 0 t< a (2.31)
-1 t~ a
and is shown in Fig. 2.14. Note that the jump discontinuity of the step
occurs when the argument within the parentheses is 7.ero. This forms the
basis of the shifting property of the step function. Also, the height of the
jump discontinuity of the step can be scaled up or down by the multiplica-
tion of a constant K.
With the use of the change of amplitude and the shifting properties of
the step function, we can proceed to construct a family of pulse waveforms.
For example, the square pulse in Fig. 2.15 can be constructed by the sum
of two step functions

s(t) == 4u(t - I) + (-4) u(t - 2) (2.32)

as given in Fig. 2.16. The "staircase" function, shown in Fig. 2.17, is

characteri7.Cd by the equation
I
s(t) -Iu(t - kn (2.33)
t-o

"''
4 ,_ _____________

0 1 2 t

-4--------.____
FIG. 2.16. Construction of square pulse by step function.
30 Network analysis and synthesis
B(t)

0 T 2T
FIG, 2.17. Staircase function.

Finally, let us construct the square wave in Fig. 2.1. Using the shifting
property, we see that the square wave is given by (fort~ 0)
s(t) == u(t) - 2u(t - T) + 2u(t :.... 2T) - 2u(t - 3T) + ··· (2.34)
A simpler way to represent the square wave is by using the property
that the step function is zero whenever its argument is negative. Restricting
ourselves to the interval t ~ 0, the function

s( t) == u (sin~) (2.35)

is zero whenever sin (wt/T) is negative, as seen by the waveform in Fig.

2. 18. It is now apparent that the square wave in Fig. 2.1 can be represented
as
s(t) == u( sin~) - u(-sin ~) (2.36)

Another method of describing the square wave is to consider a generaliz-

ation of the step function known as the sgnfunction (pronounced signum).
The sgn function is defined as
sgn 1/(t)] == I f(t) > 0
== 0 f(t) == 0 (2.37)
== - I f(t) < 0
B(t)

-
0 T 2T 3T 4T t
FIG, 1.11. The signal u(sin Trt/T).
 Signals and waveforms 31

e(t)

'
nus the square wave in Fig. 2.1 is simply expressed as

s(t) - sgn (sin;) (2.38)

Returning to the shifting property of the step function, we see that the
single sine pulse in Fig. 2.19 can be represented as

s(t) - sin!! [u(t - 2n - u(t - 3n] (2.39)

T
The step function is also extremely useful in representing the shifted or
delayed version of any given signal. For example, consider the unit ramp
function
p(t) - t u(t) (2.40)

shown in Fig. 2.20. Suppose the ramp is delayed by an amount t a, as =

shown in Fig. 2.21. How do we represent the delayed version of ramp?

e(t)

1
'
PIG. 2.21. Ramp fuoction with zero time shift.
32 Network analysis and synthesis
s(t)

FIG, 1.11. Ramp function with time shift= a.

First, let us replace the variable t by a new variable t' =t- a. Then
p(t') = t' u(t') (2.41)
When p(t') is plotted against t', the resulting curve is identical to the plot
of s(t) versus tin Fig. 2.20. If, however, we substitute t - a = t' in p(t'),
we then have
p(t') = (t - a) u(t - a) (2.42)

When we plot p(t') against t, we have the delayed version of p(t) shown in
Fig. 2.21.
From the preceding discussion, it is clear that if any signal f(t) u(t) is
delayed by a time T, the delayed or shifted signal is given by
/(t') = f(t - T) u(t - T) (2.43)
For example, let us delay the function (sin 1rt/T) u(t) by a period T. Then
the delayed function s(t'), shown in Fig. 2.22, is

s(t') = [sin ~(t - T)]u(t _- T) (2.44)

s(t')

-1

FIG, 1.12. Shifted sine wave.

 Signals and waveforms 33

s(t)
2

0 2 t
FIG. 2.23. Triangular pulse.

As a final example, consider the waveform in Fig. 2.23, whose com-

ponent parts are given in Fig. 2.24. For increasing t, the first nonzero
component is the function 2(t - 1) u(t - 1), which represents the straight
line of slope 2 at t = 1. At t = 2, the rise of the straight line is to be
arrested,soweaddtothefirstcomponenta term equal to -2(t - 2)u(t - 2)
with a slope of -2. The sum is then a constant equal to 2. We then add a
term -2u(t - 2) to bring the level down to zero. Thus,

s(t) = 2(t - 1) u(t - 1) - 2(t - 2) u(t - 2) - 2u(t - 2) (2.45)

2.4 THE UNIT IMPULSE

The unit impulse, or delta function, is a mathematical anomaly. P.A. M.

Dirac first used it in his writings on quantum mechanics. 2 He defined the
I
I
s(t)

-2u(t-2)

FIG. 2.24. Decomposition of the triangular pulse in Fig. 2.23.

1 P.A. M. Dirac, The Principles of Quantum Mechanics Oxford University Press, 1930.
3-4 Network analysis and synthesis
delta function 6(t) by the equations

L: 6(t)dt =1 (2.46)
6(t) = 0for t ~ 0 (2.47)
Its most important property is the sifting property, expressed by

L: f(t) 6(t) dt == /(0) (2.48)

In this section we will examine the unit impulse from a nonrigorous

approach. Those who prefer a rigorous treatment should refer to
Appendix B for development of this discussion. The material in that
appendix is based on the theory of generalized functions originated by
G. Temple.3 In Appendix B it is shown that the unit impulse is the
derivative of the unit step
6(t) = u'(t) (2.49)
At first glance this statement is doubtful. After all, the derivative of the
unit step is zero everywhere except at the jump discontinuity, and it does
not even exist at that point! However, consider the function g,(t) in Fig.
2.25. It is clear that as E goes to zero, g.(t) approaches a unit step, that is,
lim g,(t) == u(t) (2.50)
,➔ o

Taking the derivative of g.(t), we obtain g',(t), which is defined by the

equations
1
g'.(t) == - ;
E (2.51)
= O; t < 0, t > E
as shown in Fig. 2.26. Now let E take on a sequence of values E1 such that
E1 > E,+1• Consider the sequence of functions {g'.,(t)} for decreasing

t 0 t
FIG. 2.25. Unit step when • -+ O. FIG. 2.26. Derivative or g/..t) in
Fig. 2.25.

1 G. Temple, "The Theory or Generalized Functions." Proc. Royal Socilty, ..4, 228,

1955, 175-190.
 Signals and waveforms 3S

1
El' ......;io,

la
1
- H- 6i1 (t)

li
I
1;1 m
1 1;/t}
ii

I
Ei E3E2Et t
FIG. 2.27. The sequence {K,,(t)}.

values of £ 1, as shown in Fig. 2.27. The sequence has the following

property:
q>O
Jim
c, ➔ O i
h<O
g' ,,(t) dt =l
where t1 and t 1 are arbitrary real numbers. For every nonzero value of£,
(2.52)

there corresponds a well-behaved function (i.e., it does not ..blow up")

g' ,,(t). As £ 1 approaches zero,

(2.53)
so that the limit of the sequence is not defined in the classical sense.
Another sequence of functions which obeys the property given in Eq.
2.52 is the sequence {/.,(t)} in Fig. 2.28. We now define the unit impulse
~t) as the class of all sequences of functions which obey Eq. 2.52. In
particular, we define

i is>O
'1 <O
~t) dt
4
= Jim
,, ➔ o
iq>O
h <o
g',,(t) dt
(2.54)
36 Network analysis and synthesis

------1.
En

____ J.
Ei

----!

-E1 -E; -En En E; +Et t

FIG. 2.28. The sequence {f.p)}

It should be stressed that this is not a rigorous definition (which, as

stated previously, is found in Appendix B) but merely a heuristic one.
From the previous definition we can think of the delta "function" as
having the additional properties,
15(0) = 00
(2.55)
'5(t) = 0 for t 'F 0
 Signals and waveforms 37
Continuing with this heuristic treatment, we say that the area "under"
the impulse is unity, and, since the impulse is zero for t =,f,. 0, we have

L: b(t) dt = J: b(t) dt =1 (2.56)

Thus the entire area of the impulse is "concentrated" at t = 0. Con-

sequently, any integral that does not integrate through t = 0 is zero,

as seen by J: b(t) dt = J:"' b(t) dt = 0

(2.57)
s(t)
s(t) =A u(t - a)

The change of scale and time shift properties A-----

discussed earlier also apply for the impulse
function. The derivative of a step function 0 a
· s(t) = A u(t - a) (2.58)

r
yields an impulse function
s'(t) = A t5(t - a) (2.59) s'(.to)I
which is shown in Fig. 2.29. Graphically, we
represent an impulse function by an arrow- a
head pointing upward, with the constant
s'(t) = A 8(t - a)
multiplier A written next to the arrowhead.
FIG. 2.29
Note that A is the area under the impulse
A b(t - a).
Consider the implications of Eqs. 2.58 and 2.59. From these equations
we see that the derivative of the step at the jump discontinuity of height A
yields an impulse of area A at that same point t = T. Generalizing on this
argument, consider any function/(t) with a jump discontinuity at t = T.
Then the derivative,f'(t) must have an impulse at t = T. As an example,
consider f(t) in Fig. 2.30. At t = T,f(t) has a discontinuity of height A,

f(t)

-L:,
ft(t) = f(t)
0 T

FIG. 2.30. Function with discontinuity at T.

38 Network analysis and synthesis
which is given as
A ==/(T+)-/(T-) (2.60)

Let us define / 1(t) as being equal to /(t) for t < T, and having the
same shape as /(t), but without the discontinuity for t > T, that is,
/ 1(t) == /(t) - A u(t - T) (2.61)

The derivative/'(t) is then

/'(t) == /'i(t) + A 6(.t - T) (2.62)

The following example illustrates this point more clearly. In Fig. 2.31a,
the function/(t) is
/(t) == A u(t - a) - A u(t - b) (2.63)

Its derivative is /'(t) == A 6(.t - a) - A 6(t - b) (2.64)

and is shown in Fig. 2.31b. Since /(t) has two discontinuities, at t == a and
t == b, its derivative must have impulses at those points. The coefficient of
the impulse at t == b is negative because

/(b+)-/(b-) = -A (2.65)

f(t)

A g(t)
4

II b t
(11)

f(t) t
(11)

A g'(t)
2
b 2
II t

0 1 t
-A
(b) (b)

FIG. 2.31. (a) Square pulse. FIG. 2.32. (a) Signal.

(6) Derivative of square pulse. (b) Derivative.
 Signals and waveforms 39
As a second example, consider the function g(t) shown in· Fig. 2.32a.
We obtain g'(t) by inspection, and note that the discontinuity at t-= 1
produces the impulse in g'(1) of area
g(l +) - g(l -,-) - 2, (2.66)
as given in Fig. 2.32b.
Another interesting property of the impulse function is expressed by the
integral
+ao
i -ao f(t) "<_t - T) dt -f(T)

This integral is easily evaluated if we consider that 6(1 - T) - 0 for all

(2.67)

t =,I= T. Therefore, the product

/(1) "(_1 - T) =0 all t 9' T (2.68)
If/(1) is single-valued at t - T,f(T) can be factored from the integral so
that we obtain
f(T) L:
"(_t - T) dt = f(T) (2.69)

Figure 2.33 shows /(t) and "(_t - T), where /(t) is co~tinuous at t == T.
If/(t) has a discontinuity at t = T, the integral

f_+aoi(t) 6(t - T) dt

is not defined because the value of/(T) is not uniquely given. Consider
the following examples.
Example 2.1
/(t) -e1-
+ao (2.70)
l-ao el_,d(t - T)dt ,.. eJm'L'

/(t) ==sin t
(2.71)

i +aosintd ( t - -
-ao
1
"') dt---=
4 V2
f(t)
a(t-T)

0 T
FIG.2.33
-40 Network analysis and synthesis
f(t)

f(t)

0 T---

FIG. 2.34. Impulse scanning.

Consider next the case where/ (t) is continuous for - oo < t < oo. Let
us direct our attention to the integral

L: f(t) t5(t - T) dt = f(T) (2.72)

which holds for all t in this case. If T were varied from - oo to + oo,
then/(t) would be reproduced in its entirety. An operation of this sort
corresponds to scanning the function/(t) by moving a sheet of paper with
a thin slit across a plot of the function, as shown in Fig. 2.34.
Let us now examine higher order derivatives of the unit step function.
Here we represent the unit impulse by the function f.(t) in Fig. 2.28,
which, as e--+ 0, becomes the unit impulse. The derivative of/,(t) is given
in Fig. 2.35. As e approaches zero,/' ,(t) approaches the derivative of the
unit impulse t5'(t), which consists of a pair of impulses as seen in Fig. 2.36.
The area under b'(t), which is sometimes called a doublet, is equal to zero.
Thus,

L: b'(t) dt =0 (2.73)

The other significant property of the doublet is

L: f(t) ~'(t - T) dt ~ -J'(T) (2.74)

1
;i

E
0 0

FIG. 2.35. Unit doubles as ~ -+ O. FIG. 2.36. The doublet

<)'(t).
 Signals and waveforms 41
where f'(T) is the derivative of f(t) evaluated at t = T where, again, we
assume that/(t) is continuous. Equation 2.74can be proved by integration
by parts. Thus,

L: f(t) t5'(t - T) dt = f(t) t5(t - T) 1:00 -L: f'(t) t5(t - T) dt

(2.75)
= -f'(T).

It can be shown in general that

(2.76)

where t5<n> and/ <n> denote nth derivatives. The higher order derivatives of
~(t) can be evaluated in similar fashion.

Problems
2.1 Resolve the waveforms in the figure into odd and even components.

s(tJ
s(tJ
1

-1 0 1 t +l t

(a)
(bJ

s(tJ f(tJ
2~ 2

I
0 1 2 t -1

(c) (cl)
PROB. 2.1
42 Network analysis and synthesis
2.2 Write the equation for the waveforms in the figure using shifted step
functions.
f(t)

f(tJ 2

2 t

(a) (b)

f(tJ f(tJ

2 2
1.5

0 1 2 0 2

'"'
(c)

PROB. 2.1

2.3 Find the derivative of the waveforms in Prob. 2.2 and write the equations
for the derivatives, using shifted step and/or impulse functions.
2.4 For the waveform f(t) given in the figure, plot carefully

for a value of t > T.

f(t)
f(t)

2a(t)
E6(t) It------.
E
T
0

-E6(t-T> K6(t-T)
PROB,2.4 PROB. 2.5
 Signals and waveforms 43
2.5 For the waveform /(t), shown in the figure, determine what value K
must be so that
(a) f_f(t)dt -o
(b) f:f(t)dt -o
2.6 For the waveforms shown in the figure, express in terms of elementary
l+
time functions, that is. ,.., ll(t - lfJ, d(_t - lfJ, (a) /(t) (b) f'(t)
Sketch the waveforms for (b) and (c) neatly.
(c)
f
-ao /(,r) d-r

f(t)

f(t)

20

0 0 1 2 3 t

(i) (ii) -5

PllOB.2.6
2.7 Prove that
(a) 6'(z) - -6'( -z)
(b) -d(.z) = z d'(z)
(c) L: z d(_z). _ o
2.8 The waveform/(t) in the figure is defined as
3
f(t) - ';i (t - •>'. 0 :S:: t :S:: •

- 0, elsewhere
Show that as • - 0, /(t) becomes a unit impulse.

0 e
PllOB.2.1
44 Network analysis and synthesis
1.9 Plot
(a) <5(cos t)
(b) t sgn (cost); 0 ~ t ~ 2 ..
2.10 Evaluate the following integrals

(a) L: 6(t - T 1) u(t - T 2) dt;

(b) J_: 6( w - w 0) cos wt dw

(c) 1:[6(t) - A 6(t - T 1) + 2A 6(t - T 2)]e-1nrot dt

2.11 Evaluate the following integral

L:sin ,[6 (, - i) + 15' (, -i) + 6*(t - ,r)] dt

2.12 The response from an impulse sampler is given by the equation

r(t) = L: sin; 6(1 - K ~) dt; K = 0, 1, 2, 3, ..•

Plot r(t) for O ~ t ~ 2T.

2.13 If the step response of a linear, time-invariant system is r ,(t) = 2e-t u(t),
determine the impulse response h(t), and plot.
2.14 For the system in Prob. 2.13 determine the response due to a staircase
excitation
3
s(t) =I u(t - kT)
k=O

Plot both excitation and response functions.

2.15 If the impulse response of a time-invariant system is h(t) = e-t u(t),
determine the response due to an excitation
e(t) = 26(t - 1) - 26(t - 2)
Plot e(t).
f(.t)

11.-------.

0 3

PROB.1.16
 Signals and waveforms 45
2.16 The unit step response of a linear system is

ae,> = c2e-2t - •> u(1>

(a) Find the response r(t) to the input f(t).
(b) Make a reasonably accurate sketch of the response. Show all pertinent
dimensions.
 chapter 3
The frequency domain:
Fourier analysis

3.1 INTRODUCTION

One of the most common classes of signals encountered are periodic

signals. If T is the period of the signal, then
s(t) = s(t ± nT) n = 0, 1, 2, •.• (3.1)
In addition to being periodic, if s(t) has only a finite number of discon-
tinuities in any finite period and if the integral

f'+Tls(t)I dt

is finite (where 0t is an arbitrary real number), then s(t) can be expanded

into the infinite trigonometric series

s(t) = ~2 + a 1 coswt + a1 cos2wt + · · · Q~

+ b1 sin wt + b sin 2wt + ••·
1

where a,= 2,,,/T. This trigonometric series is generally referred to as the

Fourier series. In compact form, the Fourier series is
Q GO
s(t) == ~ +I (a,. cos nwt + b,. sin nwt) (3.3)
2 -1

It is apparent from Eqs. 3.2 and 3.3 that, when s(t) is expanded in a
Fourier series, we can describe s(t) completely in terms of the coefficients

""
 L
The frequency domain: Fourier analysis ',7
of its harmonic terms, 0o, ai, a 1, ••• , b1, b1, • • • • These coefficients con-
stitute a frequency domain description of the signal. Our task now is to
derive the equations for the coefficients a,, b, in terms of the given signal
function s(t). Let us first discuss the mathematical basis of Fourier series,
the theory of orthogonal sets.

l.2 ORTHOGONAL FUNCTIONS

Consider any two functions/1(t) and/.J..t) that are not identically zero.
Then if
(3.4)

we say that fi(t) and /.J..t) are orthogonal over the interval [T1, T .]. For
example, the functions sin t and cos t are orthogonal over the interval
n21r ~ t ~ (n + 1)21r. Consider next a set of real functions {th(t),
Mt), ... , t/>,.(t)}. If the functions obey the condition

2'1
(t/>,, tf,1) a
i 2'1
tf,,(t) t/>l..t) dt = 0, i ,&j (3.S)

then the set {t/>,} forms an orthogonal set over the interval [Ti, T .]. In
Eq. 3.S the integral is denoted by the inner product (tf,,, t/> 1). For conven-
• ience here, we use the inner product notation in our discussions.
The set {tf,,} is orthonormal over [T1, T1] if

(tf,,, t/>,) =0 i -,6}

(3.6)
=l i=J
The norm of an element tf,,. in the set {t/>,} is defined as

(3.7)

We can normali7.e any orthogonal set {"'1, tf,., ••• , tf,,.} by dividing each
term t/>-. by its norm llt/>-.11.
Evwple ·u.
The Laguene set,1 which bas been shown to be very meful in
time domam approximation, ia orthogonal over (0, oo ]. The first four terms

1 W. H. Kautz, ..Transient Synthesis in the Time Domain.,. Thln.f. IRE°" Circtdt

'I'Mory, cr-1, No. 3 (Sept. 1954), 29-39.

-48 Network analysis and synthesis
of the Laguerre set are
+i_(t) = e-1
,f,,1..t) == e-1[1 - 2(at)]

t/,a(t) = e-1[1 - 4(at) + 2(at)8] (3.8) _,,

,f,..(t) = e-1[1 - 6at + 6(atY, - -t(at)8]

To show that the set is orthogonal, let us consider the integral

L«> +i_(t) t/,a(t) dt = L«>e-lat[l - 4(at) + 2(at}8] dt (3.9)

Letting -r = at, we have

("'1, t/>a) = -1 i«>e-a..o - 4-r + 2,-1) d-r
a o
(3.10) ;,
=~ u- 4 (~) + 2(nJ =o
The norms of ,f,1(t) and #.,t) are

f«> )~ 1
11"'111 = ( Jo 1
e-la dt = V2a (3.11) '

11</>tll = Uo«>e-lat[l - 4(at) + 4(at)1] dtr

(3.12)
1
= v2a
It is not difficult to verify that the norms of all the elements in the set are also
v
equal to 1/ 2a. Therefore, ·to render the Laguerre set orthonormal, we divide
each element ,f,t by l/V2a.

3.3 APPROXIMATION USING ORTHOGONAL FUNCTIONS

In this section we explore some of the uses of orthogonal functions in

the linear approximation of functions. The principal problem is that of
approximating a function/(t) by a sequence offunctions/,.(t) such that the
mean squared·error
=!.~ i
T1
E T1 [f(t) - f,.(t)) dt
1
=0 (3.13)
 The frequency domain: Fourier analysis -49
When Eq. 3.13 is satisfied, we say that {/tl(t)} converges in the mean to
f(t).
To examine the concept of convergence in the mean more closely, we
must first consider the following definitions:
Definition 3.1 Given a function/(t) and constant p > 0 for which
T1

LTi
lf(t)IP dt < oo,
we say tha:t f(t) is integrable LP in [Ti, T.], and we write J(t) e LP in
[Ti, T1].
Definition 3.2 Hf(t) e LP in [Ti, T1], and {/t1(t)} is a sequence of func-
tions integrable LP in [T1 , T.], we say that if
T1
lim.
t1-+«> L
Ti
lf(t) - ft1(tW dt =o
then {/.(t)} converges in the mean of order p to f(t). Specifically, when
p =
2 we say that {ftl(t)} converges in the mean to f(t).
The principle of least squares
Now let us consider the case when/tl(t) consists of a linear combination
of orthonormal functions t/,i, t/,1, ••• , f>t1•
ti

J.<t> = I a, 4>,<t> (3.14)

,-1
Our problem is to determine the constants a, such that the integral
squared error

i
T1
Ill - ft1ll 1 = [f(t) - ft1(t)J1 dt (3.15)
Ti

is a minimum. The principle of least squares states that in order to attain

minimum squared error, the constants a, must have the values
T1
c, =
LTi
f(t) cf,,(t) dt (3.16)

Proof. We shall show that in order for II/- ft1ll 1 to be minimum, we

must set a, = c, for every i = I, 2; .•. , n.

U/- ft1ll 1 == (/,/) - 2(/,/J + (Jtl,/J

t1 t1 (3.17)
== 11/11 1 - 2Ia,(/, t/>J + Ialllt/>,11 1
,-1 •-1
SO Network analysis and synthesis

Since the set {+,} is orthonormal, llt/,,11 1 - 1, and by definition, c, -

(/, tf,J.
We thus have
ti ti

11/- ft111 1 - 11/11 1 - 2 I a,c, +Ia,•

,-1 ,-1
(3.18)

I c,• gives
ti
Adding and subtracting
,-1
(3.19)

We see that in order to attain minimum integral squared error, we must

set a, - c,. The coefficients c,, defined in Eq. 3.16, are called the Fourier
coefficients of/(t) with respect to the orthonormal set {+,(t)}.
Parseval's equality
Consider/tl(t) given in Eq. 3.14. We see that
T1 t1
i Ti
Vt1(t))' dt - I c,'
1-1
(3.20)

since ti,, are orthonormal functions. This result is known as Parseval's

equality, and is important in determining the energy of a periodic signal.
>

3.4 FOURIER SERIES

Let us return to the Fourier series as defined earlier in this chapter,

a
s(t) == JI + Ioo (atl cos nwt + bt1 sin nwt) (3.21)
2 -1

From our discussion of approximation by orthonormal functions, we can

see that the periodic function s(t) with period T can be approximated by a
Fourier series stl(t) such that stl(t) converges in the mean to s(t), that is,
.+T
~ •
f. [s(t) - stl(t)'f dt - 0 (3.22)

where ot is any real number. We know, moreover, that if n is finite, the

mean squared error Us(t) - stl(t)ll 1 is minimi7.Cd when the constants a,, b,
are the Fourier coefficients of s(t) with respect to the orthonormal set
cos kwt sin kwt}
{ (T/2)"" , (T/2)"" ' k = 0, 1, 2, ..•
 The frequency domain: Fourier analysis 51
.,,,
A

-3r -2,r -r O 1r 2r 3r t
FIG. 3.1. Rectified sine waw.

in explicit form the Fourier coefficients, according to the definition given

earlier, are obtained from the equations

a0 = -2 f..+2's(t)dt (3.23)
T •
a,,= -2 f..+2's(t) cos k<.ot dt (3.24)
T •
b11 -
2 f..+2's(t) sin ko,t dt
_; (3.25)
T •
We should note that because the Fourier series s,.(t) onJy converges to
s(t) in the mean, when s(t) contains a jump discontinuity, for example,
at t0
s (t) = s(t.+)
fl O 2
s(t.-) + (3.26)

At any point t1 that s(t) is differentiable (thus naturally continuous)

s,.(tJ converges to s(tJ.1
As an example, let us determine the Fourier coefficients of the fully
rectified sine wave in Fig. 3.1. As we observe, the period is T = .,,. so
that the fundamental frequency is w = 2. The signal is given as
s(t) = A. lsin ti (3.27)
Let us take at = 0 and evaluate between O and .,,., Using the formula just
derived, we have
b,. = -2i"s(t) sin 2nt dt = 0 (3.28)
.,,. 0

a0 = -2A.i".sm t dt = -4A.
.,,. 0 .,,.
(3.29)

a,. = -2.,,.
i"0
s(t) cos 2nt dt

- 1-4n1
I 4A.
.,,.
1 For a proof see H. F. Davis Fourier &ries and Ort"°Konal Functions, Allyn and
(3.30)

Bacon, Boston, 1963, pp. 92-95.

52 Network analysis and synthesis
Thus the Fourier series of the rectified sine wave is

s(t) = -4A(
21r
1 + Im 2
n-11 - 4n I
cos 2nt) (3.31)

3.5 EVALUATION OF FOURIER COEFFICIENTS

In this section we will consider two other useful forms of Fourier series.
In addition, we will discuss a number of methods to simplify the evaluation
of Fourier coefficients. First, Jet us examine how the evaluation of
coefficients is simplified by symmetry considerations. From Eqs. 3.23-3.2S
which give the general formulas for the Fourier coefficients, let us take
or. == -T/2 and represent the integrals as the sum of two separate parts,
that is,

a. == -2
T
[iT/1s(t) cos no,t dt +JO
O -T/1
]
s(t) cos no,t dt
(3.32)
b. == ~[ fTl•s(t) sin no,t dt
T Jo
+f 0

-T/8
s(t) sin nwt dt]

Since the variable (t) in the above integrals is a dummy variable, let us
substitute z == t in the integrals with limits (0; T/2), and let z t in the =-
integrals with limits (,;-T/2; 0). Then we have

a. == -2 iT/1[s(z) + s( -z)] cos no,z dz

T o
(3.33)
b"
2
== - iT/1 [s(z) - s(-z)] sin nroz dz
T o

Suppose now the function is odd, that is, s(z) == -s( -z), then we see that
a. == O·for all n, and
b.
4
= -T iT/1
s(z)sin no,z dz
o
(3.34)

This implies that, if a function is odd, its Fourier series will contain only
sine terms. On the other hand, suppose the function is even, that is,
s(z) == s(-z), then b,. == 0 and

a. == -4 iT/1s(z) cos no,z dz (3.35)

T o

Consequently, the Fourier series of an even function will contain only

cosine terms.
 The frequency domain: Fourier analysis 53
B(~)

FIG. 3.2

Suppose next, the function s(t) obeys the condition

s(t ± f) = -s(t) (3.36)

as given by the example in Fig. 3.2. Then we can show that s(t) contains
~ only odd harmonic terms, that is,
a,.= b,. = O; neven

and a,. = -T4 iT/1

o
s(t) cos nwt dt
(3.37)
b,. = -
iT/1s(t) sin nrut dt,
4
n odd
T o
With this knowledge of symmetry conditions, let us examine how we
can approximate an arbitrary time function s(t) by a Fourier series within
an interval (0, T]. Outside this interval, the Fourier series s,.(t) is not
required to fit s(t). Consider the signal s(t) in Fig. 3.3. We can approxi-
mate s(t) by any of the periodic functions shown in Fig. 3.4. Observe
that each periodic waveform exhibits some sort of symmetry.
Now let us consider two other useful forms of Fourier series. The first
is the Fourier cosine series, which is based upon the trigonometric identity,
C,. cos (nrut + 0,.) = C,. cos nwt cos 0,. - C,. sin nwt sin 0,. (3.38)

0 T t
FIG. 3.3. SignaUo be approllimatcd.
54 Network analysis and synthesis
B(t)
A

-T T
--- (a)
-- t

B(t)

-- -T--...
A

T ...........

-- lb)
-A

B(t)

... __ A

(c)
FIG. 3.4, (a) Even function cosine terms only. (b) Odd function sine terms only. (c)
Odd harmonics only with both sine and cosine terms.

We can derive ~ form of the Fourier cosine series by setting

an = Cn cos (:)n (3.39)
and bn = -C,. sin (:)n (3.40)
We then obtain Cn and 8,. in terms of an and bn, as

Cn = (anI + bn"'"'
J

Co= ao
2 (3.41)

(:),. = tan-1(- a,.b,.)

If we combine the cosine and sine terms of each harmonic in the original
series, we readily obtain from Eqs. 3.38-3.41 the Fourier cosine series
f(t) = C0 + C1 cos (wt + OJ + C1 cos (2wt + 6.)
+ C1 cos (3<.ot + 6.) + · · · + c. cos (nwt + OJ + ••· (3.42)
 The frequency domain: Fourier analysis 55
It should be noted that the coefficients Cft arc usually taken to be positive.
If however, a term such as -3 cos 2wt carries a negative sign, then we can
use the equivalent form
-3 cos 2wt = 3 cos (2wt + ,r) (3.43)
For example, the Fourier series of the fully rectified sine wave in Fig. 3.1
was shown to be

s(t) =- 4A( 1 + I m 2 cos2nt) (3.44)

2,r ft-11 - 4n1
Expressed as a Fourier cosine series, s(t) is

s(t) = -4A[1 + Im 2 cos(2nt + ,r)] (3.4S)

2,r -14nI - 1
Next we consider the complex form of a Fourier series. If we express
cos nwt and sin nwt in terms of complex exponentials, then the Fourier
series can be written as

s(t)
a
=-!+I
m ( e1""''
aft--"------
+ e_,_, + bft -
e'""" -
----
e_,..,)
2 ~ 2 ~ ~~
= _!!, +
2
I (aft -2
-1
jbft e'""'' + aft + Jbft
2
e-,--)
lfwe define
{J _ aft - jbft
ft - 2 '
R
,,. = ~2 (3.47)

then the complex form of the Fourier series is

m
s(t) = /J, + I
-1
({Jfte 1""'' + {J_fte- 1-->
(3.48)

We can readily express the coefficient /Jft as a function of ,(t), since

{J _ aft - Jbft
ft - 2

= ! J.Ts(tXcos nwt - j sin nwt) dt (3.49)

T •
= ! J.Ts(t)e_,_ dt
T •
Equation 3.49 is sometimes called the discrete Fourier transform of s(t)
and Eq. 3.48 is the inverse transform of /Jft(nw) - /Jft.
56 Network analysis and synthesis

Co

-4<al -3'.1 -2w -w

FIG. 3.5. Amplitude spectrum.

-2T
-¥ -f0fTf2T t

-A

FIG. 3.6. Square wave.

 The frequency domain: Fourier analysts 57
Observe that {J,. is usually complex and can be represented as
{J,. - Re /J,. + j Im /J,. (3.50)
The real part of {J,., Re /J,., is obtained from Eq. 3.49 as

Re {J,. = -T1 iT o
s(t) cos nrut dt (3.51)

and the imaginary part of {J,. is

J Im {J,. = -J•1Ts(t) sin nrut dt (3.Sl)

T o
It is clear that Re /J,. is an even function inn, whereas Im {J,. is an odd
function in n. The amplitude spectrum of the Fourier series is defined as
1/J,.1 = (Rel /J.. + lm1 /JJ1A (3.53)
and the phase spectrum is defined as
'1,,. - arctan Im /J,. (3.54)
Re {J,.
It is easily seen that the amplitude spectrum is an even function and the
phase spectrum is an odd function in n. The amplitude spectrum provides
us with valuable insight as to where to truncate the infinite series and still
maintain a good approximation to the original waveform. From a plot of
the amplitude spectrum, we can almost pick out by inspection the non-
trivial terms in the series. For the amplitude spectrum in Fig. 3.5, we see
that a good approximation can be obtained if we disregard any harmonic
above the third.
As an example, let us obtain the complex Fourier coefficients for the
square wave in Fig. 3.6. Let us also find the amplitude and phase spectra
of the square wave. From Fig. 3.6, we note that s(t) is an odd function.
Moreover, since s(_t - T/2) - -s(_t), the series has only odd harmonics.
From Eq. 3.49 we obtain the coefficients of the complex Fourier series as

P.. - ! fTt•.A.e_,_, dt - ! fT .A.e_,_, dt

T Jo T JT/t ·
(3.55)
- ~ (1 _ 2e-(I-T/I) + e-,-~
JnruT
Since nruT == n2,r, {J,. can be simplified to

p,. = ~ (1 _ 2e-1ra11 + e-11"") (3.56)

j2n,r
58 Network analysis and synthesis
Amplitude

-5 -3 -1 1 3 5 ""
(a)

Phase
+!
2

0 1 3 5
-5 -3 -1
""
-.!:
2

(b)
FIG. l.7. Discrete spectra of square wave. (a) Amplitude. (b) Phase.

Simplifying /J. one step further, we obtain

{J. = ~A n odd
(3.57)
Jn'ff'
=0 neven
The amplitude and phase spectra of the square wave are given in Fig. 3.7.

J.6 EVALUATION OF FOURIER COEFFICIENTS USING

UNIT IMPULSES

In this section we make use of a basic property of impulse functions to

simplify the calculation of complex Fourier coefficients. This method is
restricted to functions which are made up of straight-line components only.
Thus the method applies for the square wave in Fig. 3.6. The method is
based on the relation
L: f(t) 6(t - TJ dt =J(TJ (3.58)
Let us use this equation to evaluate the complex Fourier coefficients for
the impulse train in Fig. 3.8. Using Eq. 3.58 with /(t) - e-1- 1, we have

fJ. = i f.2'a(, - i)e_,_, -i dt e-11-r1• 1 (3.59)

 The frequency domain: Fourier analysis 59

A
"'' A~ A

I I
-f O f T f 2T \t '
FIG. J.8. Impulse train,

We see that the complex Fourier coefficients for impulse functions are
obtained by simply substituting the time at which the impulses occur into
the expression, e-1-•.
In the evaluation of Fourier coefficients, we must remember that the
limits for the /J,. integral are taken over one period only, i.e., we consider
only a single period of the signal in the analysis. Consider, as an example,
the square wave in Fig. 3.6. To evaluate /J,., we consider only a single
period of the square wave, say, from t = 0 to
t = T, as shown in Fig. 3.9a. Since the square
Arr---""'
.,,,
wave is not made up of impulses, let us
dift"erentiate the single period of the square
wave to give a'(t), as shown in Fig. 3.9b. We
10 T
can now evaluate the complex Fourier coeffi- f
cients for the derivative a'(t), which clearly is
made up of impulses alone. Analytically, if -A -
a(t) is given as (G)
GO

I
a(t) -

then the derivative of a(t) is

/J,.e 1_, (3.60)
. ,,,
A A
GO
s'(t) = I jnw {J,.e1_, (3.61)
0 T
Here, we define a new complex coefficient f
y,. = jnw {J,. (3.62)
or /J= y,. (3.63)
" jnw
-2A
If the. derivative a'(t) is a function which (6)
consists of impulse components alone, then
FIG. 3.9. (a) Square wave
we simply evaluate y,. first and then obtain over period [0, T]. (b) Deriv-
{J,. from Eq. 3.63. For example, the derivative ative or square wave over
of the square wave yields the impulse train period [O, T].
60 Network analysis and synthesis
in Fig. 3.9b. In the interval (0, T], the signal s'(t) is given as

s'(t) = A 6(t) - 2A a( t - f) + A 6(t - T) (3.64)

Then the complex coefficients are

y,. = i LTs'(t)e- 1""'' dt

(3.65)
= ~ (1 - 2e-<l""'Tt•> + e-,-~
T
The Fourier coefficients of the square wave are

.
[J =b
. Jnw
(3.66)
= ~ (1 - 2e-11""'Tt•> + e-,-~
JnwT
which checks with the solution obtained in the standard way in Eq. 3.55.
If the first derivative, s'(t), does not contain impulses, then we must
differentiate again to yield
00

s"(t) == I .t,.e1""'' (3.67)

where .t,. = jnw-y,. == (jnw)1 [J,. (3.68)

For the triangular pulse in Fig. 3.10, the second derivative over the
period (0, T] is
s"(t) ==
2:[6{t) - 2a(, -f) + 6(t - T)] (3.69)

The coefficients .t,. are now obtained as

.t,. = .! fT s"(t)e- 1""'1 dt

rJo
(3.70)
= 2A (1 - 2e-<l""'Tt•> + e-l""'T)
r•
which simplifies to give
nodd
(3.71)
== 0 n even
 The frequency domain: Fourier analysis 61

~
A -

0 ~
t
(a)

M a'(t)
T

0 T

-2:- (b)
••(t)
¥ 2A
T

0 T

_.u
T
(c)
FIG. 3.10. The triangular wave and its derivatives.

From A" we obtain

(3.72)
nodd
=0 neven
A slight difficulty arises if the expression for s'(t) contains an impulse
in addition to other straight-line terms. Because of these straight-line
terms we must differentiate once more. However, from this additional
differentiation, we obtain the derivative of the impulse as well. This
presents no difficulty, however, because we know that

L:s(t) «5'(t - T) = -s'(T) (3.73)

so that L: «5'(t - T)e_,_, dt = jnwe- 1"°'2' (3.74)

62 Network analysis and synthesis

We can therefore tolerate doublets or even higher derivatives of impulses

in the analysis. Consider the signal s(t) given in Fig. 3.1 la. Its derivative
s'(t), shown in Fig. 3.1 lb, can be expressed as

s'(t) - i[u(t) - u( t - f)] + 6(t) - 26( t - f) (3.75)

The second derivative s•(t) consists of a pair of impulses and a pair of

doublets as given by

s"(t) - i["(t) - «'(t -f)] + 6'(t) - 26'(t -f) (3.76)

as shown in Fig. 3.llc. We therefore evaluate l,. as

(3.77)

2
.,,, •'(t)
i(t)

1 fr--
0
'
(a) (b)

. ,,, ,.,,,
,Jam
-21·(,-J)

f
0 t

(c)
FIG. 3.11
 The frequency domain: Fourier analysis 63
The complex coefficients {J., are now obtained as

A.,
{J
"= (jron)1
(3.78)
= 2 (1 _ e-(inwT/2)) + _1_ (1 _ 2e-(iflwT/ll))
(jwnT) 1 jwnT
Simplifying, we have
. 1 3
{J=- -+-
11 11
nodd
" n '1T j2'1Tn
(3.79)
1
=--- n even
j2'1Tn
In conclusion, it must be pointed out that the method of using impulses
to evaluate Fourier coefficients does not give the d-c coefficient, a0 /2 or {J0 •
We obtain this coefficient through standard methods as given by Eq. 3.23.

3.7 THE FOURIER INTEGRAL

In this section we extend our analysis of signals to the aperiodic case.

We show through a plausibility argument that generally, aperiodic signals
have continuous amplitude and phase spectra. In our discussion of Fourier
series, the complex coefficient {J., for periodic signals was also called the
discrete Fourier transform
{J(nfo) = -1 fT/2 s(t)e-i"S.fot dt (3.80)
T -T/2
and the inverse (discrete) transform was
co
s(t) = 1 {J(nfo)ei" 9•tot (3.81)

From the discrete Fourier transform we obtain amplitude and phase

spectra which consist of discrete lines. The spacing between adjacent lines
in the spectrum is
·1
!if= (n + l)fo - nfo = - (3.82)
T
As the period Tbecomes larger, the spacing between the harmonic lines in
the spectrum becomes smaller. For aperiodic signals, we let T approach
infinity so that, in the limit, the discrete spectrum becomes continuous.
64 Network analysis and synthesis

We now define the Fourier integral or transform as

S(f) = lim ()(nfo)

T-+oo lo
=f 00

-OD
s(t)e- 11" 11 dt (3.83)
4.f-+O

The inverse transform is

s(t) = L: S(f)e1•11tt df (3.84)

Equations 3.83 and 3.84 are sometimes called the Fourier transform pair.
Ifwe let :J' denote the operation of Fourier transformation and :,--1 deaote
inverse transformation, then
S(J) = :J' • s(t)
(3.85)
s(t) = :J'-1 • S(/)
In general, the Fourier transform S(/) is complex and can be denoted as
S(/) = Re S(/) + j Im S(/) (3.86)
The real part of S(/) is obtained through the formula
Re S(f) = ½[S(f) + S(-f)]
= L: s(t) cos 2'1Tjt dt
(3.87)

and the imaginary part through

1
Im S(f) = j [S(f) - S(-f)]
2

=- L: s(t) sin 2'1Tf t dt

(3.88)

The amplitude spectrum of S(/) is defined as

A(/) = [Re S(J)B + Im S(/)1 ]1A (3.89)
and the phase spectrum is
cf,(f) = arctan Im S(f)
_ ___,.'-'- (3.90)
Re S(f)
Using the amplitude and phase definition of the Fourier transform, the
inverse transform can be expressed as

s(t) = L:A(f) cos [2'1Tft - cf,(/)] df (3.91)

Let us examine some examples.

 The frequency domain: Fourier analysis 65
q,(n

-f 0 +f

(a) (b)
FIG. 3.12. Amplitude and phase spectrum of A "C,t - t 0).

s(_t) ... A d(_t - to)

S(f) - L:
_ Ae-llrfto
A d(t - to)e-llrft dt

(3.92)
Its amplitude spectrum is
A(f) == A (3.93)
while its phase spectrum is
AA- -2,,,fto (3.94)
as shown in Fig. 3.12.
Example 3.3. Next consider the rectangular function plotted in Fig. 3.13.
Formally, we define the function as the rect function.
w
=-1 lzl s: 2
rectz (3.95)
w
{ =-0 lzl > 2
The inverse transform of rect/ is defined as sine t (pronounced sink),
S--1 (rect /] - sine t
W/2
=
i-W/2
sin
e1a.ttdJ
,,,w,
(3.96)

,,,,
=---
1 It should be noted here that the Fourier transform of a generali7.ed function is also

a generali7.ed function. In other words, if ,f, e C, (3'" • t/>) E C. For example, 3'" • <Xt) =
1, where 1 is described by a generali7.ed function 1,.(/). We will not go into the formal
details of Fourier transforms of generali7.ed functions here. For an excellent treatment
of the subject see M. J. Lighthill, Fourier Analysis and Generalized Functions, England,
Cambridge University Press, 1955.
66 Network analysis and synthesis
rect X

1.0

+~
2
X

FIG. 3.13. Plot of rect function.

1.0
0.8
I I\1
7
0.6
J \
sine t o.4 I \
0.2
0.0
-L

....... i..,...,
I / t---..
I\. j
7 l
I\. j
I/ r--.
r-- I--' -
...._

-0.2
-0.3
-w5 _j_
w
0 1
w
2
w w
3 4
w
5
w
t
FIG. 3.14. The sine t curve.

eW rW
Excitation System

----1--.l....J------- I BW - ~: I ~---li---L--...__T_1

-IT1I--
Narrow pulse Wide band Narrow pulse

e(t) r(t)
Excitation System Response

I BW-~ I l0T1

Wide pulse Narrower band Wide pulse

FIG. 3.15. Illustration of the reciprocity relationships between time duration and
bandwidth.
 The frequency domain: Fourier analysis 67
From the plot of sine tin Fig. 3.14 we see that sine t falls as does lt1-1 ,
with zeros at t = n/W, n = 1, 2, 3, ... We also note that most of the
energy of the signal is concentrated between the points -1/W < t < 1/W.
Let us define the time duration of a signal as that point, t0 , beyond which
the amplitude is never greater than a specified value, for example, E0 • We
can effectively regar4 the time duration of the sine function as t 0 = ± 1/ W.
The value W, as we see from Fig. 3.13, is the spectral bandwidth of the
rect function. We see that if W increases, t0 decreases. The preceding
example illustrates the reciprocal relationship between the time duration
of a signal and the spectral bandwidth of its Fourier transform. This
concept is quite fundamental. It illustrates why in pulse transmission,
narrow pulses, i.e., those with small time durations, can only be trans-
mitted through filters with large bandwidths; whereas pulses with longer
time durations do not require such wide bandwidths, as illustrated in
Fig. 3.15.

3.8 PROPERTIES OF FOURIER TRANSFORMS

In this section we consider some important properties of Fourier

transforms.
Linearity
The linearity property of Fourier transforms states that the Fourier
transform of a sum of two signals is the sum of their individual Fourier
transforms, that is,

Differentiation
This property states that the Fourier transform of the derivative of a
signal is j2TTJ times the Fourier transform of the signal itself:

:T • s'(t) = j2TTJ S(f) (3.98)

or more generally,
:T • s<n>(t) = (j2TT/t S(/) (3.99)

The proof is obtained by taking the derivative of both sides of the inverse
transform definition,

s'(t) = ~ Joo S(f)eilfrtt df

dt -oo (3.100)
68 Network analysis and synthesis
Similarly, it is easily shown that the transform of the integral of s(t) is

:F [f' -00
s(,r) d-r] ~ SU)
= J2-rrf (3.101)

Consider the following example

s(t) = r' u(t) (3.102)
Its Fourier transform is

SU) = L: e-,,.' u(t)e- 1•rt1 dt

(3.103)
00
1-

The derivative of s(t) is

=
i
0
e-'e-ilrfl dt = --
· a+ j2-rrf
s'(t) = 6(t) - ar' u(t) (3.104)
Its Fourier transform is

:F[s'(t)] =1- a = j2-rrf

a + j2-rrf a + j2-rrf (3.10S)
=j2-rrf SU)
Symmetry
The symmetry property of Fourier transforms states that if
= X(/)
:F • z(t) (3.106)
then :F • X(t) = z(-f) (3.107)
This property follows directly from the symmetrical nature of the Fourier
transform pair in Eqs. 3.83 and 3.84.
Example 3.4. From the preceding section, we know that
:F • sine t = rcctf (3.108)
It is then simple to show that
:F•rcctt =sine(-/) =sine/ (3,109)
which conforms to the statement of the symmetry property. Consider next
the Fourier transform of the unit impulse, :F • d(_t) = 1. From the symmetry
property we can show that
(3,110)
as shown in Fig. 3.16. The foregoing example is also an extreme illustration
of the time-duration and bandwidth reciprocity relationship. It says that zero
time duration, 6(t), gives rise to infinite bandwidth in the frequency domain;
while zero bandwidth, d(_f) corresponds to infinite time duration.
 The frequency domain: Fourier analysis 69
f(t)

1.

-t -f 0 f

FIG. 3.16. Fourier transform of/(t) =:= 1.0.

Scale chan1e
The scale-change property describes the time-duration and bandwidth
reciprocity relationship. It states that

:F [s(!)] = lal S(afl (3.111)

Proof. We prove this property most easily through the inverse trans-
form
1
:;-- [1al S(af)] = la!J_: S(afle'1" 1' df (3.112)

Let f' = af; then

:F-1[1al SU')] = lalf 00 SU')e'1"l'U/a> df'

-ao a
(3.113)
= s(;)
As an example, consider

:F[e......, u(t)] = -.- 1 - (3.114)

J2wf + a
then
:F[e_, u(t)] = lal
j2waf + a
(3.115)
1
=---
j2wf + 1
if a> 0.
Folding
The folding property states that
:F[s(-t)] = S(-/) (3,116)
70 Network analysis and synthesis
The proof follows directly from the definition of the Fourier transform.
An example is
1
.1'[et u(-t)] = . (3.117)
1 - J2'1Tj
Delay
If a signal is delayed by an amount t0 in the time domain, the corre-
sponding effect in the frequency domain is to multiply the transform of the
undelayed signal by e-izrtto, that is,
.1'[s(t - to)] = e-iBrtto S(f) (3.118)
For example,
-ibfto
.1'[e-a(t-1o> u(t - t0)] = e . (3.119)
a+ J27rf
Modulation
The modulation or frequency shift property of Fourier transforms
states that if a Fourier transform is shifted in frequency by an amount Jo,
the corresponding effect in time is described by multiplying the original
signal by e1""101, that is,
.1°-1 [S(/ - fo)] = eibfot s(t) (3.120)
Example 3.5. Given S(f) in Fig. 3.17a, let us find the inverse transform of
S 1(f) in Fig. 3.17b in terms of s(_t) = :,--1 S(f). We know that
S1(f) = S(f - fo) + S(f + fo) (3.121)

S(f)
A

-f 0 +f

(a)
S1(f)

-f -to 0 +f
(b)
FIG.' 3.17. Demonstration of amplitude modulation.
 ------------

The frequency domain: Fourier analysis 71

Then :,--1 S1(/) = ellttlel s(t) + e--11•1,1 s(t) = 2s(t) cos 2,,,/ol (3.122)

Thus we see that multiplying a signal by a cosine or sine wave in the time domain
corresponds to shifting its spectrum by an amount ±fo. In transmission.termi-
nology / 0 is the carrier frequency, and the process of multiplying s(t) by cos 2,,,/ol
is called amplitude modulation.

Parseval's theorem
An important theorem which relates energy in the time and frequency
domains is Parseval's theorem, which states that

L: s1(t) s.(t) dt == L: Si(f) S.(-f) df (3.123)

The proof is obtained very simply as follows:

L: si(t) s.(t) dt = L: s1(t) dt L: Si(f)e'a,,11 df

= L: S1U) dfL: s1(t)e'

1 1
• ' dt (3.124)

==L: S1U) s.(-f) df

In particular, whens1(t) = sJ.t), we have a corollary of Parseval's theorem

known as P/ancheral's theorem.

L: s8(t) dt = L: 1
IS(/)1 df (3.125)

If s(t) is equal to the current through. or the voltage across a I-ohm

resistor, the total energy is

We see from Eq. 3.125 that the total energy is also equal to the area under
the curve of IS(f)j 1 • Thus IS(f)P11 is sometimes called an energy density
or energy spectrum.

Problems
3.1 Show that the set {1, sin n'frl/T, cos n,,,t/T}, n = 1, 2, 3, ... , forms an
orthogonal set over an interval [at, 0t + 2T], where 0t is any real number. Find
the nonns for the members of the set and normaliz.e the set.
n Network analysis and synthesis
3.2 Given the functions / 1(1) and /,1..1) expressed in terms of complex Fourier
series
00

/1(1) =I at,, e1n•I

--00
00

/,1..t) -= I Pm eima>I
---oo
where both/1(1) and/,1..1) have the same period T, and
at,, = l«.al el'1n, Pm - IPml el""'
show that P ""' j.{l' / (1)/,t.,1) di
1
00

- a.Jlo + 2 I l«..Pnl COS (On - !f,n)

-1

Note that rl'eiln+m>••

Jo
di = {T,0, m = -n
m ,.s -n
3.3 For the periodic signals in the figure, determine the Fourier coefficients
On, bn.
f(t)

(a)

ft)

-T
(b)
f(t)

(c)
PROB.3.3
 The frequency domain: Fourier analysis 73
3.4 For the wamorms in Prob. 3.3, find the disciete amplitude and phase
spectra and plot.
3.5 For the wamorms in Prob. 3.3 determine the complex Fourier coefficients
using the impuJse function method.
3.6 Find the complex ·Fourier coefficients for the function shown in the
figure.
f, ,,

2A

-~ -~
PllOB.3.6

3.7 Find the Fourier transform for the functions shown in the figure.

"''
1

0 ♦T t
(G) (6)
'

"'' "''
At-----.

0 T
' (d)
'
PROB. 3.7

3.8 Find the Fourier transform for

(a) f(I) .. A d(t)
(b) f(t) ,_ A sin OJo1
74 Network analysis and synthesis
3.9 Prove that (a) if f(t) is even, its Fourier transform F(jw) is also an even
function; (b) if f(t) is odd, its Fourier transform is odd and pure imaginary.
F(j11J)

B B

-wo
PROB. 3.10

3.10 Find the inverse transform of

F(jw) = {J «5(w - wo) + {J «5(w + w0)
as shown in the figure. What can you say about line spectra in the frequency
domain?
 chapter 4
Differential equations

4.1 INTRODUCTION

This chapter is devoted to a brief study of ordinary linear differential

equations. We will concentrate on the mathematical aspects of differential
equations and leave the physical applications for Chapter 5. The differen-
tial equations considered herein have the general form
.F[x(t), :r:'(t), ••. , :r:<">(t), t] =0 (4.1)
where tis the independent variable and :r:(t) is a function dependent upon
t. The superscripted terms :r:<'>(t) indicate the ith derivative of :r:(t) with
respect to t, namely,
:r:w(t) = tf'' x(t) (4.2)
dt'
The solution of F = 0 in Eq. 4.1 is :r:(t) and must be obtained as an
explicit function oft. When we substitute the explicit solution x(t) into F,
the equation must equal zero. If Fin Eq. 4.1 is an ordinary linear differ-
ential equation, it is given by the general equation
a,. :r:<"'(t) + a_1:r:1- 1
>(t) + ··· + a 1 :r:'(t) + Oo :r:(t) == f(t) (4.3)
The order of the equation is n, the order of the highest derivative term.
The term/(t) on the right-hand side of the equation is the forcing function
or driver, and is independent of :r:(t). When/(t) is identically zero, the
equation is said to be homogeneous; otherwise, the equation is non-
homogeneous.
In this chapter we will restrict our study to ordinary, linear differential
equations with constant coefficients. Let us now examine the meanings of
these terms.
75
76 Network analysis and synthesis
Ordinary. An ordinary differential equation is one in which there is only
one independent variable (in our case, t). As a result there is no need for
partial derivatives.
Constant coefficients. The coefficients a,., a,._1, ••• , a1, ai, "o are con-
stant, independent of the variable t.
Linear. A differential equation is linear if it contains only terms of the
first degree in z(t) and all its higher derivatives, as given by Eq. 4.3. For
example, the equation
3z'(t) + 2:z:(t) == sin t (4.4)

is a linear differential equation. On the other hand,

3[z'(t)]1 + 2:z:(t) z'(t) + 4:z:(t) = St (4.5)

is nonlinear, because the terms [z'(t)]• and z(t) z'(t) are nonlinear by the
definition just given.
An important implication of the linearity property is the superposition
property. According to the superposition property, if z 1(t) and z.(t) are
solutions of a given differential equation for forcing functions/1 (t) and/.(t),
respectively, then, if the forcing function were any linear combination of
/i(t) and / 1(t) such as
/(t) = a/ 1(t) + b/ 1(t} (4.6)
the solution would be
z(t) = a z 1(t) + b z.(t) (4.7)

where a and b are arbitrary constants. It should be emphasized that the

superposition property is extremely important and should be kept in
mind in any discussion of linear differential equations.

4.2 HOMOGENEOUS LINEAR DIFFERENTIAL EQUATIONS

This section deals with some methods for the solution of homogeneous,
linear differential equations with constant coefficients. First, let us find
the solution to the equation
z'(t) - 2:z:(t) = 0 (4.8)
Now, with a little prestidigitation, we assume the solution to be of the form
z(t) = Ce9' (4.9)

where C is any arbitrary constant. Let us check to see whether z(t) = Ce9'
is truly a solution of Eq. 4.8. Substituting the assumed solution in Eq. 4.8,
we obtain
2Ce9' - 2Ce9' = 0 (4.10)
 Dlfferentlal equations 77
It can be shown, in general, that the solutions of homogeneous, linear
differential equations consist of exponential terms of the form c~••'· To
obtain the solution of any differential equation, we substitute Ce•' for z(t)
in the equation and determine those values of p for which the equation is
zero. In other words, given the general equation
0 11 x<"'(t) + · · · + a 1 x'(t) + a. x(t) == 0 (4.11)
we let x(t) == Ce•', so that Eq. 4.11 becomes
(4.12)
Since e•' cannot be zero except at p == - oo, the only nontrivial solutions
for Eq. 4.12 occur when the polynomial

(4.13)
Equation 4.13 is often referred to .as the characteristic equation, and is
denoted symbolically in this discussion as H(p). The characteristic equation
is zero only at its roots. Therefore, let us factor H(p) to give
== a (p - pJ(p - pJ · · · (p - P-J
H(p) 11
(4.14)
From Eq. 4.14, we note that C.,e••', C1e•1', ..• , c_1e••-1' are all solutions
of Eq. 4.11. By the superposition principle, the total solution is a linear
combination of all the individual solutions. Therefore, the total solution
of the differential equation is
x(t) == C,,e••' + C1e•1• + · · · + c_1e•--1' (4.15)
where C0, Ci, ... , C 11_ 1 are generally complex. The solution x(t) in Eq.
4.15 is not unique unless the constants C0 , C 1, ••• , C-1 are uniquely
specified. In order to determine the constants C,, we need n additional
pieces of information about the equation. These pieces of information are
usually specified in terms of values of x(t) and its derivatives at t == O+,
and are therefore referred to as initial conditions. To obtain n coefficients,
we must be given the values x(O+ ), x'(O+ ), ..• , x«-11(0+ ). In a number
of special cases, the values at t == 0- are not equal to the values at
t == o+. If the initial specifications are given in terms of x(O-),
x'(O-), ..• , x«-1 1(0-), we must determine the values at t == o+ in order
to solve for the constants C,. This problem arises when the forcing
function /(t) is an impulse function or any of its derivatives. We will
discuss this problem in detail in Section 4.4.
For example, in Eq. 4.9 if we are given that x(O+) == 4, then we obtain
the constant from the equation

x(O+) == Ce' == C (4.16)

78 Network analysis and synthesis
so that x(t) is uniquely determined to be
x(t) = 4e1t
Example 4.1 Find the solution for
x'(t) + Sx'(t) + 4:r:(t) = 0 (4.17)
given the initial conditions
x(O+) =2 x'(O+) = -1
Solution. From the given equation, we first obtain the characteristic equation
H(p) = p" + Sp + 4 = 0 (4.18)
which factors into (p + 4)(p + 1) = 0 (4.19)
The roots of the characteristic equation (referred to here as characteristic values)
are p = - l; p = -4. Then x(t) takes the form
= C1e-t + C,tt-4t
x(t} (4.20)
From the initial condition x(O +) = 2, we obtain the equation
x(O +) = 2 = C1 + C11 (4.21)
In order to solve for C1 and C1 explicitly, we need the additional initial condition
x'(O +) = -1. Taking the derivative of x(t) in Eq. 4.20, we have
(4.22)
At t = O+, x'(t) is
x'(O+) = -1 = -Ci - 4Ca (4.23)
Solving Eqs. 4.21 and 4.23 simultaneously, we find that
C1 =½ C11 = -¼
Thus the final solution is (4.24)
Next, we examine the case when the characteristic equation H(p) has
multiple roots. Specifically, let us consider the case where H(p) has a
root p = p 0 of multiplicity k as given by
H(p) = a,.(p - p 0)1'(p - pJ · · · (p - p,.) (4.25)
It will be left to the reader to show that the solution must then contain k
terms involving e1101 of the form
x(t) = Coc,ePot + Co1tePot + Co,_t9ePot + .. • + Coit-1tlt-lePot
+ C1e" + C,tt 11•t + · · · + C,.e"•'
1
' (4.26)
where the double-scripted terms in Eq. 4.26 denote the terms in the solution
due to the multiple root, (p - Po}".
 Differential equations 79
Eumple 4.2 Solve the equation
z'(t) - 8:i;'(t) + 16z(t) = 0 (4.27)
with x(O+) =- 2 and z'(O+) == 4
Solution. The characteristic equation is
H(p)-= p'- - 8p + 16 == (p - 4)1 (4.28)
Since H(p) has a double root at p = 4, the solution must take the form
x(t) - C1 e'' + C1te'i (4.29)
In order to determine C1 and c., we evaluate x(t) and z'(t) at t == 0 + to give

x(O+) = C1 = 2
(4.30)
z'(O+) = 4C1 + C1 = 4
Thus the final solution is x(t) = 2e'' - 4te'' (4.31)
Another interesting case arises when H(p) has complex conjugate roots.
Consider the equation
H(p) == a1(p - pJ(p - pJ (4.32)
where p 1 and p. are complex conjugate roots, that is,
pi,pa == <I ±jw (4.33)
The solution z(t) then takes the form
z(t) == C1e10-+1•ll + C.,.elo--1•>• (4.34)
Expanding the term ei"'' by Euler's equation, z(t) can be expressed as
z(t) == C1eo-'(cos wt+ jsin wt)+ C~'(cos wt - jsin wt) (4.35)
which reduces to
z(t) == {C1 + C~' cos wt+ j(C1 - c~a• sin wt (4.36)
Let us introduce two new constants, M 1 and M 1 , so that z(t) may be
expressed in the more convenient form
z(t) == M 1ea• cos wt + M_eo-' sin wt (4.37)
where M 1 and M 1 are related to the constants C1 and C1 by the equations
M1 == C1 + C1
(4.38)
M 1 == j{C1 - C1)
The constants M 1 and M 1 are determined in the usual manner from
initial conditions.
80 Network analysis and synthesis
Another convenient form for the solution :t(t) can be obtained if we
introduce still another pair of constants, Mand ,f,, defined by the equations
M1 == M sin ,f,
(4.39)
M,. == Mcos,f,
With the constants M and ,f, we obtain another form of :t{t), namely,
:t(t) == Me•• sin (wt + ,f,) (4.40)
Example 4.3 Solve the equation
z"(t) + 2:r:'(t) + Sz(t) = 0 (4.41)
with the initial conditions
z(O+) = 1 z'(O+) == 0
Solution. The characteristic equation H(p) is
H(p) = p" + 2p + S = (p + 1 + j2)(p + 1 - j2) (4.42)
so that, assuming the form of solution in Eq. 4.37, we have a == -l and w = 2.
Then z{t) is
z(t) == M 1e' cos 2t + Mar' sin 2t (4.43)
At t =0+ z(O+) =-1 =M1 (4.44)
The derivative of z(t) is
z'(t) = M 1( -e-• cos 2t - 2e-1 sin 2t) + M,.( -e' sin 2t + 2e-t cos 2t) (4.45)
At t = 0 +, we obtain the equation •
z'(O+) =0 = -M1 + 2M1 (4.46)
Solving Eqs. 4.44 and 4.46 simultaneously, we find M 1 = 1 and M 1 ... +!.
Thus the final solution is
z(t) = e'(cos 2t + l sin 2t) (4.47)
If we had used the form of z(t) given in Eq. 4.40, we would have obtained the
solution
z(t) == v'le-t sin [2t + tan-1 (2)) (4.48)
Now let us consider a differential equation that illustrates everything we
have discussed concerning characteristic values.
Example 4.4 The differential equation is
:r:<6 >(t) + 9:z:l'l(t) + 32:z:(ll(t) + 58:r:"(t) + 56:r:'(t) + 24:r:(t) =0 (4.49)
The initial conditions are
z<&l(O+) =0 :r:<31(0+) =1
z"(O+) = -1 z'(O+) =0 z(O+) -1
 Differential equations 81
Solution. The characteristic equation is

H(p) - p5 + 9p' + 32p' + S8p' + S6p + 24 ,. 0 (4.50)

which factors into

H(p) =- (p + 1 + jl)(p + 1 - jl)(p + 2')'(p + 3) === 0 (4.51)

From H(p) we immediately write :r:(t) as

:r:(t) - M1e-• cos t + M,,r' sin t + coe-11 + C1te-11 + Cae--a, (4.52)

Since there are five coefficients. we need a corresponding number of equations

to evaluate the unknowns. These are

:r:(O+) =- M 1 + C0 + C8 = 1
z'(O+) - -M1 + M 1 - 2C0 - 3C1 + C1 -= 0
z'(O+)-= -2M1 + 4C0 - 4C1 + 9C1 == -1 (4.53)
z 111(0+) = 2M1 + 2M1 - 8Co + 12C1 - 27C1 = 1
z 1' 1(0+) = -4M1 + 16C0 + 81C1 - 32C1 = 0
Solving these five equations simultaneously, we obtain

M1 ===0 Ma =-f C0 =1
so that the final solution is

:r:(t) = fe-• sin t + e-11 + ite-11 (4.54)

We have seen that the solution of a homogeneous, differential equation

may take different forms depending upon the roots of its characteristic
equation. Table 4.1 should be useful in determining the particular form of
solution.
TABLE 4.1
Roots of H(p) Forms of Solution

1. Single real root, p = Po e11o•

2. Root of multiplicity, k, (p - Pi)" CoeP1I + C1teP1t + • • • + C,._1t1'-1eP1'
3. Complex roots at P•.•= a ± jw M 1e"' cos wt + Mae"' sin <»t
or
Me"' sin (wt + •>
4. Complex roots of multiplicity k Moe"' cos wt + M 1te"1 cos wt + · · ·
atp,,1 = a ±jw + M,._1 t1'-1e"' cos wt + Noe"' sin wt
+ N 1te" 1 sin wt+···
+ N,._1t1'-1e"1 sin wt
82 Network analysis and synthesis

4.3 NONHOMOGENEOUS EQUATIONS

As we mentioned in the introduction to this chapter, a nonhomogeneous

differential equation is one in which the forcing function f(t) is not
identically zero for all t. In this section, we will discuss methods for
obtaining the solution z(t) of an equation with constant coefficients
a11 z< 11>(t) + a_1 z<•H>(t) + · · · + a 0 z(t) = f(t) (4.55)
Let zp(t) be a particular solution for Eq. 4.55, and let zc(t) be the solution
of the homogeneous equation obtained by letting f(t) = 0 in Eq. 4.55.
It is readily seen that
z(t) = z,,(t) + zo(t) (4.56)
is also a solution of Eq. 4.55. According to the uniqueness theorem, the
solution z(t) in Eq. 4.56 is the unique solution for the nonhomogeneous
differential equation if it satisfies the specified initial conditions at
t = o+.1 In Eq. 4.56, z,,(t) is the particular integral; zo(t) is the comple-
mentary function; and z(t) is the total solution.
Since we already know how to find the complementary function zo(t),
we now have to find the particular integral z,,(t). In solving for z,,(t), a
very reliable rule of thumb is that z,,(t) usually takes the same form as the
forcing function iff(i) can be expressed as a sum of exponential functions.
Specifically, z,,(t) assumes the form of f(t) plus all its derivatives. For
example, if f(t) = « sin rot, then z,,(t) takes the form
z,,(t) = A sin rot + B cos rot
The only unknowns that must be determined are the coefficients A and B
of the terms in z,,(t). The method for obtaining z,,(t) is appropriately
called the method of undetermined coefficients or unknown coefficients.
In illustrating the method of unknown coefficients, let us takef(t) to be
f(t) = «e'' (4.57)
where ac and pare arbitrary constants. We then assume z,,(t) to have a
similar form, that is,
z,,(t) = Ae'' (4.58)
and A is the unknown coefficient. To determine A, we simply substitute
the assumed solution z,,(t) into the differential equation. Thus,
Ae''(ataP" + a 11_iP"-1 + · · · + a1p + a 0) == ate'' (4.59)

1 See, for example, C. R. Wylie, Advanlld Engineering Mathematics (2nd ed.),

McGraw-Hill Book Company, New York, 1960, pp. 83-84.

 Differential equations 83
We see that the polynomial within the parentheses is the characteristic
equation H(p) with p == p. Consequently, the unknown coefficient is
obtained as
A=~ (4.60)
H(P)
provided that H(/f) #- 0.
Example 4.5 Determine the solution of the equation
z'(t) + 3z'(t) + 2z(t) - 4e' (4.61)

with the initial conditions, z(O +) = 1, z'(O +) = -1.

Solution. The characteristic equation is

H(p) .,. ,r + 3p + 2 = (p + 2)(p + 1)

so that the. complementary function is

z,,(t) = C 1e-t + Cse-•

For the forcing function/(t) == 4e', the constants in Eq. 4.60 are ix = 4, {J = 1.
4 2
Then A= H(l) =3
Thus we obtain zJ.t) == fe'
The total solution is

To evaluate the constants C 1 and C1 , we substitute the given initial conditions,

namely,
z(O +) = 1 = C1 + Ca + f
(4.63)
z'(O+) = -1 = -C1 - 2C1 + f

Solving Eq. 4.63, we find that C1 ... -1, C1 = f. Consequently,

z(t) =- -e-• + te- 11

+ fe' (4.64)

It should be pointed out that we solve for the constants C 1 and C1 from the
initial conditions for the total solution. This is because initial conditions are
not given for z,,(t) or z,,(t), but for the total solution.
Next, let us consider an example of a constant forcing function/(t) == «.
We may use Eq. 4.60 if we resort to the artifice
/(t) == at == ace'° (4.65)
84 Network analysis and synthesis
that is, fJ =
0. For the differential equation in Example 4.5 with/(t) 4, =
we see that
4
zp(t) = A = - =2 (4.66)
H(O)
and z(t) == C1e-1 + C,r11 + 2 (4.67)
When the forcing function is a sine or cosine function, we can still
consider the forcing function to be of exponential form and make use of
the method of undetermined coefficients and Eq. 4.60. Suppose
f(t) = r,.e 1"'' == r,.(cos rot + j sin rot) (4.68)
then the particular integral z 111(t) can be written as
z 111(t) == Re z 11i(t) + jlm z 111(t) (4.69)

From the superposition principle, we can show that

if f(t)= r,. cos rot then zp(t) == Re z 111{t)
if /(t) == r,. sin rot then zit) == Im z 111(t)
Consequently, whether the excitation is a cosine function r,. cos rot or a
sine function r,. sin rot, we can use an exponential driver /{t) == u'"'';
then we take the real or imaginary part of the resulting particular integral.
Example 4.6 Find the particular integral for the equation

z•(t) + Sz'(I) + 4z(t) = 2 sin 31 (4.70)

Solution. First, let us take the excitation to be

/1(1) = 2e~t (4.71)

so that the particular integral z 111(t) takes the form

z 111(t) = Ae~ (4.72)

From the characteristic equation

H(p) =r + Sp + 4
we determine the coefficient A to be

2 2 2
A - --
H(J3)
=- - - ... ---=eJCtau-1<
-S + JlS S-V 10
3>-•l (4.73)

2
z (I) - eJ[tall-1 (8)+8'-•J (4.74)
pl 5-VlO
 Dlfferentlal equations 85
and the particuJar integral z.(t) for the original driver /(t) =- 2 sin 3t is
2
z.(t) - Im z 111(t) == • , - sin [3t + tan-1 (3) - ,r) (4.75)
5-v 10

There are certain limitations to the applicability of the method of

undetermined coefficients. If /(t) were, for example, a Bessel function
J 0(t), we could not assume z.(t) to be a Bessel function of the same form
(if it is a Bessel function at all). However, we may apply the method to
forcing functions of the following types:
=
1. /(t) A.; constant.
2. J(t) = A.(t" + b_1 t-1 + · · · + b1 t + b0); n, integer.
3. /(t) = e••; p real or complex.
4. Any function formed by multiplying terms of type 1, 2, or 3.
For the purposes of linear network analysis, the method is more than
adequate.
Suppose the forcing function were
/(t) = A.t11:e•• p = a + jw
The particular integral can be written as
z.(t) = (A.~11: + A.i---1 tlt-1 + · · · + A.1 t + A.0)e•• (4.76)
where the coefficients A.11:, A.i---i, ••• , A.i, A. 0 are to be determined.

4.4 STEP AND IMPULSE RESPONSE

In this section we will discuss solutions of differential equations with

step or impulse forcing functions. In physical applications these solutions
are called, respectively, step responses and impulse responses. As physical
quantities, the step and impulse responses of a linear system are highly
significant measures of system performance. In Chapter 7 it will be shown
that a precise mathematical description of a linear system is given by
its impulse response. Moreover, a reliable measure of the transient
behavior of the system is given by its step and impulse response. In this
section, we will be concerned with the mathematical problem of solving
for the impulse and step response, given a linear differential equation with
initial conditions at t = 0-.
From Chapter 2, recall that the definition of the unit step function was
u(t) =1 t ~ 0
=0 t<O
86 Network analysis and synthesis
and the unit impulse was shown to have the properties:
6(t) = oo t=0
=0 t ;,&!: 0

and J:6(t)dt =1
In addition, we have the relationship
6(t) = d u(t)
dt
As the definitions of 6(t) and u(t) indicate, both functions have dis-
continuities at t = 0. In dealing with initial conditions for step and impulse
drivers, we must then recognize that the solution x(t) and its derivatives
x'(t), x"(t), etc., may not be contir,uous at t == 0. In other words, it may be
that
x 1.. 1(0-) ;,&!: x<n>(o+)
x<-ll(0-) ;,&!: x<n-ll(0+)

x(0-) ;,&!: x(0+)

In many physical problems, the initial conditions are given at t = 0-.
However, to evaluate the unknown constants of the total solution, we must
have the initial conditions at t = o+. Our task, then, is to determine the
conditions at t == 0+, given the initial conditions at t = 0-. The method
discussed here is borrowed from electromagnetic theory and is often
referred to as "integrating through a Green's function."•
Consider the differential equation with an impulse forcing function
a .. x<n>(t) + a .._1 x<-1(t) + · · · + a 0 x(t) = A 6(t) (4.77)
To insure that the right-hand side of Eq. 4.77 will equal the left-hand
side, one of the terms x<n>(t), x<-u(t), ... , x(t) must contain an impulse.
The question is, "Which term contains the impulse?" A close examination
shows that the highest derivative term x<n>(t) must contain the impulse,
because if x 1- 11 (t) contained the impulse, x<n>(t) would contain a doublet
C d'(t). This argument holds, similarly, for all lower derivative terms of
x(t). If the term x<n>(t) contains the impulse, then x<-ll(t) would contain a
step and x<-•>(t), a ramp. We conclude therefore that, for an impulse
forcing function, the two highest derivative terms are discontinuous at t = 0:

• The Green's function is another name for impulse response; see, for example,
Morse and Feshbach, Methods of Theoretical Physics, McGraw-Hill Book Company,
New York, 1952, Chapter 7.
 Differential equations 87
For a step forcing function, only the highest derivative term is discontinuous
at t == 0.
Since initial conditions are usually given at t == 0-, our task is to
determine the values z 1">(o+) and z 1"-ll(0+) for an impulse forcing
function. Referring to Eq. 4.77, let us integrate the equation between
t == 0- and t == 0+, namely,

a,. J: 1 1
z " (t) dt + a,._1 f: 1 11
x "- (t) dt + · · · + a0 f: ==AI:
~t)dt "(t)dt
(4.78)
After integrating, we obtain

a,.[z1"-ll(0+) - :i:1- 1
>(0-)] + a,._1 [:i:1"-11 (0+) - :.e<-1 >(0-)] + · · · =A
(4.79)
We know that all derivative terms below (n - 1) are continuous at t == 0.
Consequently, Eq. 4.79 simplifies to
a,.[:e<-ll(0+) - :.e<-ll(0-)] == A (4.80)

so that :,:<n-11(0+) == ~ + :,:<n-11(0-) (4.81)

a,.
We must next determine :i:1"'(0+ ). At t == o+, the differential equation
in Eq. 4.77 is

a,. :i:1" 1(0+) + a,._1 z 1- 11 (0+) + · · · + flo :.e(0+) =0 (4.82)

Since all derivative terms below (n - I) are continuous, and since we
have already solved for :i:1-ll(0+ ), we find that

(4.83)
·For a step forcing function A u(t), all derivative terms except :e'"'(t), are
continuous at t == 0. To determine z 1" 1(0+ ), we derive in a manner
similar to Eq. 4.83, the expression

The process of determining initial conditions when the forcing function

is an impulse or one of its higher.derivatives can be simplified by the visual
88 Network analysis and synthesis
process shown in Eqs. 4.85 and 4.86. Above each derivative term we draw
its associated highest-order singularity. Note that we need only go as low
as a step in this visual aid.

+ J_ _f
a,.z< 11 >(t) + a,._1z(n-l>(t) + a,._r:< 11- 1>(t) + • •· + aoz{,t) =
+ d'(t) (4.85)

_L_f J_
a,.zl">(t) + a,._1zln-l>(t) + a,._r:<-•>(t) + · · · + a.,,;{t) = d(t) (4.86)

It should be noted that if a derivative term contains a certain singularity-

for example, a doublet-it also contains all lower derivative terms. For
example, in the equation

+_l_S+
z'(t) + 3z'(t) + 2z(t) 4d'(t) (4.87)

we assume the following forms for the derivative terms at t == 0:

x"(t) = A <5'(t) + B <5(t) + C u(t)

x'(t) = A <5(t) + B u(t) (4.88)
x(t) = A u(t)

Substituting Eq. 4.88 into Eq. 4.87, we obtain

A <5'(t) + B <5(t) + C u(t) + 3A <5(t) + 3B u(t) + 2A u(t) = 4"'(t) (4.89)

Or in a more convenient form, we have

'A <5'(t) + (B + 3A) <5(t) + (C + 3B + 2A) u(t) == 4"'(t) (4.90)

Equating like coefficients on both sides of the Eq. 3.90 gives

A== 4
+ 3A == 0
B (4.91)
C + 3B + 2A == 0
 Differential equations 89
from which we obtain B - -12 and C - 28. Therefore, at t - 0, it is
true that
z"(t) - 46'(t) - 12«'(t) + 28u(t)
z'(t) = 46(t) - 12u(t) (4.92)
z(t) = 4u(t)
The u(t) terms in Eq. 4.92 gives rise to the discontinuities in the initial
=
conditions at t 0. We are given the initial conditions at t - 0-. Once
we evaluate the A, B, C coefficients in Eq. 4.88, we can obtain the initial
conditions at t = o+ by referring to the coefficients of the step terms. For
example, if
z(0-) = -2
z'(0-) = -1
z"(0-) - 7
Then from Eq. 4.92 we obtain

z(O+) == -2 + 4 =2
z'(0+)= -1 - 12 = -13 (4.93)
z"(0+) = 7 + 28 = 35
The total solution ofEq. 4.87 is obtained as though it were a homogeneous
equation, since «''(t) = 0 fort,' 0. The only influence the doublet driver
=
has is to produce discontinuities in the initial conditions at t 0. Having
=
evaluated the initial conditions at t o+, we can obtain the total
solution with ease. Thus,

z(t) = C1e-' + C.e--1 ' (4.94)

From Eq. 4.93 we readily obtain

z(t) = (-9e-' + 1le-1 ') u(t) (4.95)

The total solution of a differential equation with a step or impulse forcing

function is obtained in an equally straightforward manner. For a step
forcing function, only the highest derivative term has a discontinuity at
=
t 0. Since we do not need the initial condition of the highest derivative
term for our solution, we proceed as if we were solving a standard non-
homogeneous equation with a constant forcing function. For an impulse
driver, once we determine the initial conditions at t - o+, the equation is
solved in the same manner as a homogeneous equation.
90 Network analysis and synthesis
EU1Dple 4.7 Find the step and impulse response for the equation
2z..(t) + 41:'(t) + lOx(t) = f(t)
where /(t) = "(t) and /(t) = u(_1), respectively. The initial conditions at t = O-
are
x(O-) = x'(O-) = x..(0-) = 0
Solution. Let us first find the impulse response. We note that the x .. term
contains an impulse and a step; the x' term contains a step; the x term contains
a ramp, and is therefore continuous at t = 0. Thus x(O +) = x(O-) = 0. To
obtain x'(O+), we use Eq. 4.81
K l 1
x'(O+) = - + x'(O-) = - + 0 = - (4.96)
all 2 2
Note that we actually need only x(O +) and x'(O +) to evaluate the constants
for the second-order differential equation. Next, we proceed to the comple-
mentary function x0(1). The characteristic equation is
H(p) = 2{p2 + 2p + 5) = 2(p + 1 + j2)(p + 1 - j2) (4.97)
Since H(p) has a pair of complex conjugate roots, we use a standard form for
x 0(t):
x0(1) = Me-t sin (21 +- 4') (4.98)
Substituting the initial conditions at t = O+, we obtain
x(O+)= 0 = Msin 4'
(4.99)
x'(O+) = ! = 2Mcos 4' - Msin 4'
from which we find 4' = 0 and M = ¼- Thus the impulse response, which we
denote here as x,(t), is
xl,..t) = le-• sin 2t u(t) (4.100)
Next we must solve for the step response x,.(I). For convenience, let us write the
complementary function as
xa(t) = e-'(A1 sin 2t + A1 cos 21) (4.101)
The particular integral is evaluated by considering the forcing function a
constant f(t) = 1 so that
(4.102)

The total solution is then

x(t) == (A1 sin 2t + A1 cos 2t)e_, + lo (4.103)
Since x'(t) and x(t) must be continuous for a step forcing function,
a:(O+) == a:(0-) = 0
(4.104)
x'(O+) - x'(O-) == 0
 Differential equations 91
Substituting these initial conditions into x(t) d/dt d/dt
and x'(t), we find that A1 = -0.05 A1 = -0.1. xpCt)---+ x,..(t)---+ x,(t)
Therefore, the step response is
x,..(t) = 0.1 [l - e-'(0.5 sin 2t
cos 2t)J u(t)
+
(4.105)
f~ f~
x 1(t)---+ x,..(t)---+ xp(t)
FIG. 4.1.
Note that the impulse response and the
step response are related by the equation

x, (t) = -d x,..(t) (4.106)

dt
We can demonstrate Eq. 4.106 by the following procedure. Let us substitute
x,..(t) into the original equation

d'
2 --:i x,..(t) + 4 -d x,..(t) + 10x.(t) = u(t) (4.107)
di- dt
Differentiating both sides, we have

2 ~[it x,..(t)] +4 :,[1, xu(t)] + 10(1t x,..(t)] = 6(t) (4.108)

from which Eq. 4.106 follows.

Generalizing, we see that, if we have the step response for a differential
equation, we can obtain the impulse response by differentiating the step
response. We can also obtain the response to a ramp function f(t) =
A p(_t) (where A is the height of the step) by integrating the step response.
The relationships discussed here are summarized in Fig. 4.1.

4.5 INTEGRODIFFERENTIAL EQUATIONS

In this section, we will consider an integrodifferential equation of the

form

a,.x"(t) + a,._1 x-1(t) + · · · + a 0 :,;(t) + a_1 f x(T)dT =J(t) (4.109)

where the coefficients {a,., a_1, ••• , a_1} are constants. In solving an
equation of the form of Eq. 4;109 we use two very similar methods. The
first method is to differentiate both sides of Eq. 4.109 to give
a,. :,:<n+i>(t) + a_1 x'"1(t) + · · · + a 0 x'(t) + a_1 x(t) = f'(t) (4.110)
The second method consists of a change of variables. We let y'(t) = x(t);
Eq. 4.109 then becomes
a,. y<n+l>(t) + a,._1 y'"1(t) + · · · + a 0 y'(t) + a_1 y(t) = J(t) (4.111)
92 Network analysis and synthesis
Note that from Eq. 4.110 we obtain x(t) directly. From Eq. 4.111, we
obtain y(t), which we must then differentiate to obtain x(t). An important
point to keep in mind is that we might have to derive some additional
initial conditions in order to have a sufficient number to evaluate the
unknown constants.
Example 4.8 Solve the integrodifferential equation

z'(t) + 3x(t) + 2 f~ x(T) dT = Su(t) (4.112)

The initial condition is x(0 - ) == 1.

Solution. Since the characteristic equation of Eq. 4.112 is of second degree,
we need an additional initial condition z'(0 + ). We obtain z'(0 +) from the
given equation at t - 0 +:
z'(0+) + 3x{0+) + 2 s: x(T)dT =S (4.113)
Since x{t) is continuous at t "" 0,

and
s: z{T) dT == 0

x(0+) ""x(O-) =- 1
(4.114)

(4.115)
Therefore, z'(0+) =S - 3x(0+) =2 (4.116)
• METHOD 1. Differentiating both sides of Eq. 4.112, we obtain
z"(t) + 3z'(t) + 2x{t) ... Sd(t) (4.117)
The complementary function is then
x,(1) = C1e-• + Cae-1 1 (4.118)
Using the initial conditions for x(0 +) and z'(0 + ), we obtain the total solution
x(t) ""'4e-, - 3e-• t (4.119)
METHOD 2. Letting y'(t) = x(t), the original differential equation then
becomes
y"(t) + 3y'(t) + 2y(t) = Su(t) (4.120)
We know that y'(0+) -= x(0+) =- 1
(4.121)
y"(0+) == z'(0+) =2
From Eq. 4.120, at t = o+, we obtain
y(0+) =- l[S -y"(0+) - 3y'(0+)) =0 (4.122)

Without going into details, the total solution can be determined as

y(t) - -4c' + fe-•• + t (4.123)

 Differential equations 93
Differentiating y(t), we obtain

:i:(t) = y'(t) = 4e-, - Je-1 ' (4.124)

4.6 SIMULTANEOUS DIFFERENTIAL EQUATIONS

Up to this point, we have considered only differential equations with a

single dependent variable z(t). In this section, we will discuss equations
with more than one dependent variable. We shall limit our discussion to
equations with two unknowns, z(t) and y(t). The methods described here,
however, are applicable to any number of unknowns. Consider first the
system of homogeneous equations

«1 z'(t) + «o z(t) + Pi y'(t) + Po y(t) = 0

(4.125)
y1 z'(t) + Yo z(t) + 61 y'(t) + 60 y(t) = 0

where «,, Pi• y i• 6, are arbitrary constants. The complementary function is

obtained by assuming that

y(t) = C.e"''
so that the characteristic equation is given by the determinant

H(p)=
(«J]J + «o) {PJ]J + po) I
I (YiP + Yo) (6J]J + 6.)
(4.126)

The roots of H(p) are found by setting the determinant equal to zero,
that is,
(«J/' + oto)(6J]J + 6.) - <PiP + Pol(YiP + Yo)= O (4.127)

It is seen that a nontrivial solution of H(p) = 0 exists only if

(4.128)

Assuming that the preceding condition holds, we see that H(p) is a second-
degree polynomial in p and can be expressed in factored form as

H(p) = C(p - po)(p - pJ (4.129)

where C is a constant multiplier. The complementary functions are

z(t) = K1ePo' + K,.e"'1'

(4.130)
y(t) = K1e""' + K,e"'i'
9◄ Network analysis and synthesis
and the constants Ki, K,., Ka, K, are determined from initial conditions.
As in the case of a single unknown, if H(p) has a pair of double roots; i.e.,
if Po = P1, then
x(t) = (K1 + K1t)ePot
(4.131)
y(t) = (Ka + K,t)ePo'

If H(p) has a pair of conjugate roots,

then x(t) = M 1f!'I' sin (wt + tf,J

(4.132)
y(t) = M.e"' sin (wt + tf,,.)
Eumple 4.9. Consider the system of equations

lx'(t) + ~t) + y'(t) - y(t) =0

(4.133)
x'(t) + lx{t) + y'(t) + y(t) =0
with the initial conditions

:i:'(0+) =2 y'(0+) = -3
(4.134)
:i:(0+) =0 y(0+) =1
Solution. The characteristic equation is

H(p) =12p +4 p-1, =0 (4.135)

p+2p+l
Evaluating the determinant, we find that

H(p) =p 1 + Sp + 6 = (p + 2)(p + 3) (4.136)

so that y(t) = K 1e-"' + x,.e-3'

(4.137)
z(t) == Kae-t.t + K,e-at
With the initial conditions, z'(0 +) ... 2, z(0 +) = 0, we obtain K 8 = 2, K, = -2.
From the conditions y'(0+) - -3, y(0+) = 1, we obtain K 1 = 0, K 1 = 1.
Thus the final solutions are
z(t) - 2e-1t - 2e-a,
(4.138)
y(t) - e-B'
 Differential equations 95
Next, let us determine the solutions for a set of nonhomogeneous
differential equations. We use the method of undetermined coefficients
here. Consider first an exponential forcing function given by the set of
equations
CX1X' + y + /J1Y' + /JoY = Ne"
(4.139)
Y1z' + Yr + 61y' + 6oY == 0
We first assume that xp(t) = .Ae"
(4.140)
yp(t) == Be"
Then Eq. 4.139 becomes
(cx10 + OCo).A + (/J10 + {JJB = N
(4.141)
(y10 + yJ.A + (610 + 6o)B == 0
The determinant for the set of equations 4.141 is

H(O) == A(O) = Icx10

Y10
+ CXo p.o + /Jo
+ Yo 610 + 60
I (4.142)

where H(O) is the characteristic equation with p = 0. We now determine

.•. ,;;the undetermined coefficients .A and B from A(O) and its cofactors, namely,

.A= N A11(0)
A(O)
(4.143)
B = N A 11(0)
A(O)
where A., is the ijth cofactor of A(O).
Eumple 4.10. Solve the set of equations
2z' +~ +y' - y =-3e"
(4.144)
+ 2z + y' + y = 0
z'
given the conditions z'(0 +) = 1, z(0 +) = 0, y'(0 +)
= 0, y(0 +) = -1.
Solution. The complementary functions zc(t) and y 0(t), as well as the charac-
teristic equation H(p), were determined in Example 4.9. Now we must find
A and Bin the equations
zp(t) - Ae"
(4.145)
yp(t) == Be"
The characteristic equation with p = 4 is
2(4) +4 (4) _ 1 I
H(4)- -42 (4.146)
1 (4) + 2 (4) +1
96 Network analysis and synthesis
Then we obtain from Eq. 4.143 the constants

A--h, B=-f
The incomplete solutions are
z(t) = K1e-" + Kr' +-he''
(4.147)
y(t) =- Kae-at + K,e-at - fe"
Substituting for the initial conditions, we finally obtain

y(t) = -t( _4e-ae - 3e41)

(4.148)
z(t) = -h(. -6e-" + e-at + Se")
Example 4.11. Solve the system of equations

2x' + 4x + y' + 1y = Su(_t)

(4.149)
x' + x + y' + 3y = S«5(t)
given the initial conditions
z(0-) = x'{0-) = y(0-) = y'(0-) = 0
Solution. First we find the characteristic equation

2p+4 p+11
H(p) = t:i.(p) = (4.150)
1 p+l p+3
which simplifies to give
H(p) = (r + 2p + S) = (p + 1 + j2)(p + 1 - j2) (4.151)
The complementary functions x 0 (t) and y 0(t) are then
x.(t) = A 1e-t cos 2t + Aae-t sin 2t
(4.152)
y.(t) = B 1e-t cos 2t + B-,-t sin 2t
The particular solutions are obtained for the set of equations with t > 0,
namely,
2x' + 4x + y' + 1y = S
(4.153)
x' +x +y' +3y =-0

Using the method of undetermined coefficients, we assume that x,, and y,, are
constants: x,, = C1 ; y,, = C1• Since the forcing function Scan be regarded
as an exponential term with :r.ero exponent, that is, S = Sell', we can solve for
C1 and C1 with the use of the characteristic equation H(p) with p = 0. Thus,

H(0) = t:i.(0) = I: :I= s (4.154)

 Differential equations 97
5A11(0) 5(3)
and C1 - A(O) = 5 ,.. 3
(4.155)
(5)1
c. - -s =- -1
The general solution is then
a:(1) .. .A.1e-• cos 21 + A,p-' sin 21 + 3
(4.156)
1/(I) ... B 1c 1 cos 21 + B,ie-• sin 21 - 1
In order to find .A.1, .A.1, B1, Ba, we need the values a:(0 + ), z'(0 + ), y(0 + ),
y'(0 + ). The. values for a:(0 +) and y(0 +) are first obtained by integrating the
original differential equations between I .. 0 - and I =- 0 +. Thus,

J:(2.z' + 4z + y' + 7y) di - J: 5u(I) di

(4.157)
J:(z' + z + y' + 31A dt - J: 5d(I) di

We know that only the highest derivative terms in both equations contain
impulses at I .. 0. Moreover, both a:(1) and 1/(I) contain, at most, step dis-
continuities at I - 0. Therefore, in the integration

J:(4.z + 7y) di =0
(4.158)
J:(z + 3y)dl -o
After integrating, we obtain
la:(0+) + y(0+)-= 0
(4.159)
a:(0+) + y(0+) =- 5
Solving, we find
a:(0+) = -5 11(0+) - 10 (4.160)
To find z'(0 +) and y'(0 + ), we substitute the values for a:(0 +) and y(0 +) into
the original equations at I == 0 +. Thus, ·
2.z'(0+) - 20 + y'(0+) + 70 =5
(4.161)
z'(0+) - 5 + y'(0+) + 30 =0
so that z'(0+) = -20
(4.162)
y'(0+) - -5
Substituting these values into Eq. 4.156, we eventually obtain the final solutions
a:(1) - ( -Se-• cos 21 - 14e-• sin 21 + 3) u(1)
(4.163)
11(1) ,_ (lie-• cos 21 + 3e-• sin 2t - 1) u(t)
98 Network analysis and synthesis

Problems
4.1 Show that
zt(I) = M1e- 1 cos, 2t
and zl_t) = Mae-' sin 2t
are solutions for the equation
x*(t) + lz'(t) + Sz(t) = 0
Show that x1 + x 1 is also a solution.
4.2 Determine only the form of the solution for the equation
(a) x'(t) + 4z'(t) + 3x(t) = 0
(b) x'(t) + Sz'(t) + Sx(t) = 0
(c) x'(t) - Sz(t) = 0
(d) x'(t) + Sz(t) = 0
(e) x'(t) + 6:r:'(t) + 2Sx(t) =0
(f) x'(t) + 6:r:'(t) + 9z(t) =0
4.3 Given the initial conditions z(O +) = 1, x'(O +) = -1, determine the
solutions for
(a) x*(t) + 6:r:'(t) + 2Sz(t) =0
(b) x"(t) + Sz'(t) + 16:r:(t) =0
(c) x*(t) + 4.8lx'(t) + S.16:r:(t) =0
4.4 Find only the particular integrals for the equations
(a) x*(t) + 1x'(t) + llz(t) = e-3 1
(b) x'(t) + 3x'(t) + lz(t) = 2 sin 3t
(c) x*(t) + 2z'(t) + Sx(t) = e-1 sin 2t
(d) x'(t) + 2z'(t) + Sz(t) = e-51/2
(e) x"(t) + S.Oz'(t) + 6.2Sz(t) =6
(f) x"(t) + 6:r:'(t) + Sz(t) = 2e-1 + 3e-31
4.5 Given the initial conditions z(O +) = 1, x'(O +) = 0, determine the
solutions for
(a) x*(t) + 4z'(t) + 3x(t) = se- 1 sin 21
(b) x*(t) + 6:r:'(t) + 2Sz(t) = 2 cos t
(c) x*(t) + Sz'(t) + 16:r:(t) =2
4.6 (a) A system is described by the differential equation
y*(t)+ 3y'(t) + 2y(t) = c5(t)
The initial conditions are y(O-) = 1; y'(O-) = 2. Find y(O+) and y'(O+).
(b) Given the differential equation
x<B>(t) + 14.l:"(t) + 8x(t) =- 6c5(t)
 Differential equations 99
with the intial conditions x(O-) = 12, z'(O-) = 6, and z'(O-) = - 7, find
:i:111(0 +
), z'(O + ), z'(O + ), and z(O + ).
4~7 Given the initial conditions z'(O - ) = z'(O - ) = 0, find the solutions
for the equations
(a) z'(t) + 2z'(t) + 2z(t) = 3d(t)
(b) z'(t) + 7z'(t) + 12:i:(t) -= Su(t)
4.8 Given the initial condition x(O - ) = -2, solve the integrodifferential
equations

(a) z'(t) + 52:(t) + 4 f ze T) dT = 2 sin t

(b) z'(t) + 2z(t) + 2 f zeT)dT =; r'

(c) z'(t) + 6z(t) + 9 L'ze,..> dT = 2u(_t)
4.9 Given the set of equations
z'(t) + z(t) + y(t) == u(t)
z(t) + y'(t) + 21/(t) = d(t)
with x(O-) = z'(O-) = y(O-) = y'(O-) = 0, find z(O+), z'(O+), y(O+), and
y'(O+).
4.10 Solve the set of equations
2z'(t) + Jz(t) + y'(t) + 6y(t) = d(t)
z'(t) + z(t) + y'(t) + 6y(t) = u(t)
All initial conditions at t ""' 0 - are uro.
4.11 Solve the set of equations
2z'(t) + 2z(t) + y'(t) - y(t) = 2d(t)
z'(t) + z(t) + y'(t) + 2y(t) = Je- 1 u(t)
The initial conditions are
x(O-) = 1 y(O-) = 0
z'(O-) = -1 y'(O-) =0
 chapter 5
Network analysis: I

5.1 INTRODUCTION

In this chapter we will apply our knowledge of differential equations to

the analysis of linear, passive, time-invariant networks. We will assume
that the reader is already familiar with Kirchhoff's current and voltage
laws, and with methods for writing mesh and node equations for a-c or d-c
circuits. 1 We will, therefore, consider only briefly the problem of writing
mesh and node equations when the independent variable is time t. The
problems in this chapter have the following format: Given an excitation
signal from an energy source and the network, a specified response that
is a current or voltage in the network is to be determined. When relating
these problems to the mathematics in Chapter 4, we shall see that,
physically, the forcing function corresponds to the excitation; the network
is described by the differential equation; and the unknown variable x(t)
is the response.
The problems encountered will be twofold. First, we must write the
differential problems of the network using Kirchhoff's current and voltage
laws. Next, we must solve these equations for a specified current or
voltage in the network. Both problems are equally important. It is
useless, for example, to solve a differential equation which is set up in-
correctly, or whose initial conditions are incorrectly specified.
The usual type of problem presented in this chapter might generally be
described as follows. A switch is closed at t = 0, which connects an energy
(voltage or current) source to a network (Fig. 5.1). The analog of a switch
closing at t = 0 is the energy source whose output is e(t) u(t). Before the
1 For a comprehensive treatment, see H. H. Skilling, Electrical Engineering Circuits

(2nd Edition), John Wiley and Sons, New York. 1965.

100
 Network analysis: IOI

FIG. 5.1. Switching action.

switch is closed, the currents and voltages in the network have known
values. These values at t = 0- are the initial conditions. We must then
determine the values of the currents and voltages just after the switch
closes (at t = o+) to solve the network equations. If the excitation is not
an impulse function or any of its derivatives, the current and voltage
variables are continuous at t = 0. For an impulse driver the values at
t -= o+ can be determined from methods given in Chapter 4. Having
obtained the initial conditions, we then go on to solve the network
differential equations. Unless otherwise stated, all the solutions are valid
only fort~ o+.
Since we are dealing only with linear circuits, it is essential that we bear
in mind the all-important principle of superposition. According to the
superposition principle, the current through any element in a linear circuit
with n voltage and m current sources is equal to the algebraic sum of
currents through the same element resulting from the sources taken one
at a time, the other sources having been suppressed. Consider the linear
network depicted in Fig. 5.2a with n voltage and m current sources.

i,.
FIG. 5.2a
I02 Network analysis and synthesis

FIG. 5.2b

Suppose we are interested in the current iT(t) through a given element Z,

as shown. Let us open-circuit all the current sources and short-circuit
n - 1 voltage sources, leaving only v1(t) shown in Fig. 5.2b. By i.,,(t),
we denote the current through Z due to the voltage source v1(t) alone. In
similar fashion, by i0 .(t) we denote the current through Z due to the
current source i,.(t) alone, as depicted in Fig. 5.2c. By 'the superposition

i1,
FIG. 5.2c. Superposition in linear circuits.
 Network analysis: I 103
principle, the total current ip(t) due to all of the sources is equal to the
algebraic sum
ip - i91 + i + · · · + i.,. + i + i + · · •+ i
91 01 01 0• (5.1)

5.2 NETWORK ELEMENTS

In this section, we will discuss the voltage-current (v-i) relationships

that exist for the basic network elements. Before we examine these
relationships, it is important to assign first arbitrary reference polarities for
the voltage across an element, and a reference direction of flow for the
current through the element. For the purposes of our discussion, we
assume that the positive polarity for voltage is at the tail of the current
arrow, as shown by the resistor in Fig. 5.3. Now, let us review the voltage-
current relationships for the resistor, the inductor, and the capacitor,
which we first discussed in Chapter 1.
Resistor
The resistor shown in Fig. 5.3 defines a linear proportionality relation-
ship between v(t) and i(t), namely,
v(t) - R i(t)
1 (5.2)
i(t) - Gv(t) G--
R
where R is given in ohms and G in mhos.
Capacitor
For the capacitor shown in Fig. 5.4a the v-i relationships are

i(t) - Cd v(t)
dt

v(t) - -1
C o-
i'
i(T) dT + vo(O-)
(5.3)

+
i(t)

R v(t)

(a) (b)
FIG. S.l. R.esistor. FIG. 5.4. (a) Capacitor. (b) Capacitor
with initial voltage.
104 Network analysis and synthesis
i(t)
+ +

v(t) L i(O-)
v(t)

(G) (b)
FIG. S.S. (a) Inductor. (b) Inductor with initial current.

where C is given in farads. The initial value v0 (0-) is the voltage across
the capacitor just before the switching action. It can be regarded as an
independent voltage source, as shown in Fig. 5.4b. We should point out
also that vo(O-) = v0 (0+) for all excitations except impulses and
derivatives of impulses.
Inductor
The inductor in Fig. 5.5a describes a dual relationship between voltage
and current when compared to a capacitor. The v-i relationships are
di
v(t) = L-
dt

i(t) =!
L o-
i' v(r) d-r + iL(O-)
(5.4)

where Lis given in henrys. The initial current iL(O-) can be regarded as
an independent current source, as shown in Fig. 5.5b. As is true for the
voltage across the capacitor, the current through the inductor is similarly
continuous for all t, except in the case of impulse excitations.
When the network elements are interconnected, the resulting i-v equa-
tions are integrodifferential equations relating the excitation (voltage or
current sources) to the response (the voltages and currents of the elements).
There are basically two ways to write these network equations. The first
way is to use mesh equations and, the second, node equations.
Mesh equations are based upon Kirchhoff's voltage law. On th~ mesh
basis, we establish a fictitious set of loop currents with a given reference
direction, and write the equations for the sum of the voltages around the
loops. As the reader might recall from his previous studies, if the number
of branches in the network is B, and if the number of nodes is N, the
number of independent loop equations for the network is B - N + 1.1
1 See Skilling, op. cit.
 Network analysis: I 105
We must, in addition, choose the mesh currents such that at least one mesh
current passes through every element in the network.
Example 5.1. In Fig. 5.6. a network is given with seven branches and five nodes.
We therefore need 7 - 5 + 1 = 3 independent mesh equations. The directions
of the mesh currents i1o;.. i1 are chosen as indicated. We also note that the

+
111W

FIG. 5.6

capacitors in the circuit have associated initial voltages. These initial voltages
are assigned reference polarities. as shown in the figure. Now we proceed to
write the mesh equations.
Meshi1 :
11i(t) - "c (0-) .,; R1 i1(t) + cl ff i1('1') d.,. - cl ff i,I. '1') d'I'
I 1 Jir 1 J~
Mesh 11 :
"c1(0-) - "cI (0-)""' - -1
C1
i' ~
'1(.,.)d.,. -1
+ 4di
dt

+ (: 1 + :.) f ~1,1_ .,.) d.,. - :, f~aH d.,.

Mesh 11 :
-v,l.t) +v0 (0-)
I
= -C
-1
I
i'
~
i,l_.,.)d.,. + C1
I
i'
~
i,l...,.)d.,. +Rai,1..t) (5.5)

After we find the three unknowns. 11, 11, and ia. we can determine the branch
currents and the voltages across the elements. For example. if we were required
to find the branch currents i01 and 101 through the capacitors, we would use the
following relationships
fol - /1 - /I
(5.6)
ic. - '• - '•
Alternatively. if the voltage r,,1..t) in Fig. 5.6 is our objective. we see that
"a(t) .. i,1..t)Ra (5.7)
Network equations can also be written in terms of node equations,
which are based upon Kirchhoff's current law. If the number of nodes in
I06 Network analysis and synthesis

~ Datum
FIG. 5.7

the network is N, the number of independent node equations required

for the complete solution of the network is N - 1.3 We can therefore
select one datum node in the network. All the node voltages will be
positive with respect to this datum node.
Consider the network in Fig. 5.7. Let us write a set of node equations
for the network with the datum node shown. Since the number of nodes
in the network is N = 3, we need N - 1 = 2 independent node equations.
These are written for nodes v1 and v1 , as given below.
Node v1 :

i11(t) - iL(O-) = G1 v1(t) + .! f' vi(-,) d-r - ! f' v1(-r) d-r

LJI)- LJI)-
Node v1 :

iL(O-) = - -1
L
i'
1)-
V1(-r) d-r +-L1 i'
1)-
Vz(-r)d-r + c-•-
d v (t) + Ga v.(t)
dt
(S.8)

Further examples are given in the following sections.

5.3 INITIAL AND FINAL CONDITIONS

In this section, we consider some methods for obtaining initial conditions

for circuit differential equations. We also examine ways to obtain particular
integrals for networks with constant (d-c) or sinusoidal (a-c) excitations.
In the solution of network differential equations, the complementary
function is called the transient solution or free response. The particular
integral is known as the forced response. In the case of constant or periodic
excitations, the forced response at t = oo is the steady-state or final
solution.
1 See Skilling, op. cit.
 Network analysis: I 107
There are two ways to obtain initial conditions at t =
o+ for a network:
(a) through the differential equations describing the network, (b) through
knowledge of the physical behavior of the R, L, and C elements in the
network.
Initial conditions for a capacitor
For a capacitor, the voltage-current relationship at t = o+ is
vo(0+) = J: i('r) dT + vo(0-) (5.9)

If i(t) does not contain impulses or- derivatives of impulses, vo(0+) =

vo(0-). If q is the charge on the capacitor at t =
0-, the initial voltage is

vo(0+) = vo(0-) = !. (5.10)

C
When there is no initial charge on the capacitor, vo(0+) = 0. We con-
clude that when there is no stored energy on a capacitor, its equivalent
circuit at t = o+ is a short circuit. This analogy is confirmed by examining
the physical behavior of the capacitor. As a result of the conservation of
charge principle, an instantaneous change in voltage across a capacitor
implies instantaneous change in charge, which in tum means infinite
current through the capacitor. Since we never encounter infinite current
in physical situations, the voltage across a capacitor cannot change
instantaneously. Therefore at t = o+, we can replace the capacitor by a
voltage source if an initial charge exists, or by a short circuit if there is no
initial charge.
Example 5.2. Consider the R-C network in Fig. S.8a. The switch is closed at
t = 0, and we assume there is no initial charge in the capacitor. Let us find
the initial conditions i(O +) and i'(O +) for the differential equation of the circuit
1 f'
V u(_t) = R i(t) + CJo-i(-r) dT (5.11)

! I
R R

B-C nelwork Equivalent circuit at , • O+

(o) (b)
FIG. 5.1. (a) R-C network. (b) Equivalent cin:uit at t - o+.
108 Network analysis and synthesis
The equivalent circuit at t - 0 + is given in Fig. 5.8b, from which we obtain
V
i(O+) - R (5.12)

To obtain i'(O+), we must refer to the differential equation

V cl(t) - R i'(t) + i(t) (5.13)

C

At t = 0 + we have 0 - Ri'(O+) + i(O+)

C
(5.14)

i(O+) V
We then obtain i'(O +) ... - RC =- ~C (5.15)

The final condition, or steady-state solution, for the current in Fig. 5.8a is
obtained from our knowledge of d-c circuits. We know that for a d-c excitation,
a capacitor is an open circuit for d-c current. Thus the steady-state current is
i,(t) == i( oo) =- 0
Initial conditions for an inductor
For an inductor, the voltage-current relationship at t - O+ is

iL(O+)- ! (o+v(-r)d-r + iL(O-) (5.16)

LJ&-
If v(t) does not contain impulses, then iL(O+) - iL(O-). If there is no
initial current, iL(O+) - 0, which corresponds to an open circuit at
t - o+. This analogy can also be obtained from the fact that the current
through an inductor cannot change instantaneously due to the conserva-
tion of flux linkages.
Example 5.3 In Fig. 5.9a, the switch closes at t == 0. Let us find the initial
conditions i(O +) and i'(O +) for the differential equation
di(t)
Vu(t) = L t h + Ri(t) (5.17)

L o.c.

R R

(a) (b)
FIG. 5.9, (a) R-L network. (b) Equivalent cm:uit at t ,_ O+.
 Network analysis: I 109
From the equivalent circuit at t - o+, shown in Fig. 5.9b, we see that
i(O+) =0 (5.18)
We then refer to the differential equation to obtain i'(O +).
V = Li'(O+) + Ri(O+) (5.19)
V R
Thus i'(O+) - L - L i(O+)
(5.20)
V
=L
The steady-state solution for the circuit in Fig. 5.9a is obtained through the
knowledge that for a d-c source, an indicator is a short circuit.

(5.21)

Example 5.4. In this example we consider the two-loop network of Fig. 5.10.
As in Examples 5.2 and 5.3, we use the equivalent circuit models at t ... 0 +
and t = oo to obtain the initial conditions and steady-state solutions. At
t == 0 the switch closes. The equivalent circuits at t = 0 + and t ... oo are
shown in Fig. 5.lla and b, respectively. The initial currents are
V
i.i(O+) = R1 (5.22)
i.(O+) =0
The steady-state solutions are
V
i1(oo) - - - - =- i.(oo) (5.23)
R1 +Rt
Final conditions for sinusoidal excitations
When the excitation is a pure sinusoid, the steady-state currents and
voltages in the circuit are also sinusoids of the same frequency as the
excitation. If the unknown is a voltage, for example, v1(t), the steady-state
solution would take the form
(5.24)
where w0 is the frequency of the excitation, and IVUwo)I and "1{w0)
represent the magnitude and phase of v1 p{t). A similar expression would
hold if the unknown were a current.
To obtain the magnitude and phase, we follo.w standard procedures in
a-c circuit analysis. For example, consider the R-C circuit in Fig. 5.12.
The current generator is
i,(t) = (/
0 sin wot) u(t) (5.25)
110 Network analysis and synthesis
Rt R2

v~3 3 C L

FIG. 5.10

R1 ll2

3 3
Equivalent circuit at-t =o+
o.c.

(a)
Rt ll2

:J :J o.c.

Equivalent circuit at t =oo

(b)

FIG. 5.11. (a) Equivalent circuit at t = o+. (h) Equivalent circuit at t = oo.

v(t)

R C

FIG. 5,12
 Network analysis: 111
If the steady-state voltage takes the form shown in Eq. 5.24,

IVUwo)I == IYi:o)I (5.26)

Io
== (G 1 + Wo1C8)1A
and t/>{roo) = tan-1 woC (5.27)
G
so that Vp(t) = (G• + /w0 •C,l\'A Sill
• (
Wot -
-t
tan
woC)
- (5.28)
0 G
We refer to this problem in Section 5.5.

5.4 STEP AND IMPULSE RESPONSE

As an introduction to the topic of solution of network differential

equations, let us consider the important problem of obtaining the step and
impulse responses for any voltage or

3
current in the network. As we shall see in R
Chapter 7, the step and impulse responses .------''\/\JI\/\,----,

are precise time-domain characterizations

of the network. The problem of obtaining i(t) c
the step and impulse response is stated as -
follows: Given a network with zero initial
energy, we are required to solve for a
specified response {current or voltage) due FIG. 5.13
to a given excitation function u(t) or "(t),
which either can be a current or a voltage source. 1If the excitation is
a step of voltage, the physical analogy is that of a switch-closing at
time t = ~which connects a 1-v battery to a circuit. The physical
analogy of an impulse excitation is that of a very short (compared to the
time constants of the circuit) pulse with large amplitude.
The problems involved can best be illustrated by means of examples.
Consider the series R-C circuit in Fig. 5.13. The differential equation of
the circuit is
v(t) = "(t) = R i(t) + ! f' i(-r) d-r (5.29)
cJ....
We assume vo(O-) = 0. Since Eq. 5.29 contains an integral, we substitute
z'(t) for i(t) in the equation, giving us
«'(t) = R z'(t) + ! z(t) (5.30)
C
112 Network analysis and synthesis
Integrating both sides between 0- and o+ gives

z(O+)=! (5.31)
R
The characteristic equation is

H(p) = Rp +-C1 (5.32)

and with little effort we have
z(t) =! e-•IBC u(t) (5.33)
R
1
so that i(t) = !.[c,ct)
R
- - - e-•IBC u(t>]
RC
(5.34)

which is shown in Fig. 5.14a. We thus arrive at the current impulse

response i(t) as the result of an impulse voltage excitation. In the process
we have also obtained the step response z(t)in Fig. 5.14b since, by definition,
the derivative of the step response is the impulse response. The reader
should check this result using a step excitation of voltage.
In the second example consider the parallel R-C circuit in Fig. 5.12
driven by a step current source i(t) = 10 u(t), where 10 is a constant.

i(O

*"''

(a)

s(t)

1
1f

t
(b)
FIG. 5.14. (a) Impulao respome or R-C cin:uit. (b) ~tep respome or R-C cin:uit.
 Network analysis: I 113
Da(I)

IoR - - - - - - - - - - -

0
(a)

(b)
FIG. 5.15. (a) Step response of parallel R-C circuit. (b) lmpulle nspome of parallel
R-C circuit.

Assuming zero initial conditions, the differential equation is

10 u(t) == G v(t) + Cd v(t) (5.35)

dt
from which we obtain the characteristic equation
H(p) == Cp +G (5.36)
The steady-state value of v(t) is
v.(t) == .JL == !J == loR (5.37)
H(O} G

1.---
i(t)

0 T t

FIG. S.16. Pulse excitation.

11-4 Network analysis and synthesis
v(t)

IoR
--------~....,._,._
0 t
',, ........
-IoR -------- -------
FIG. 5.17. Response to pulse excitation.

Thus the complete solution for the voltage step response is

v(t) = (Ke-t/RO + IoR) u(t) (5.38)
From the initial condition v(O+) = 0, we obtain K = -IoR so that
v(t) = IoR(l - e-t/R°) u(t) (5.39)
Differentiating Eq. 5.39 gives us the voltage impulse response

v(t) = ~ e-•tRO u(t) (5.40)

C
The step and impulse responses are plotted on Figs. S.lSa and b. Suppose
the excitation in Fig. 5.12 were a pulse
i(t) = 1 [u(t) -
0 u(t - T)] (5.41)
shown in Fig. 5.16. Then, according to the superposition and time-
invariance postulates of linear systems, the response would be
v(t) = IoR[(l - e-llR°) u(t) - (1 - e-(t-T)/RO)u(t - T)] (5.42)
which is shown in Fig. 5.17.

5.5 SOLUTION OF NETWORK EQUATIONS

In this section we will apply our knowledge of differential equations

to the analysis of linear networks. There are two important points in
network analysis: the writing of network equations and the solution of
these same equations. Network equations can be written on a mesh, node,
or mixed basis. The choice between mesh and node equations depends
largely upon the unknown quantities for which we must solve. For
 Network analysis: I 115
instance, if the unknown quantity is a branch current, it is preferable to
write mesh equations. On the other hand, if we wish to find a voltage
across a certain element, then node equations are better. In many cases
the choice is quite arbitrary. If, for example, we wish to find the voltage v
across a resistor R. we can either find v directly by node equations or find
the branch current through the resistor and then multiply by R.
Example S.S. Find the current i(t) for the
. B•lO
network in Fig. 5.18, when the voltage
source is e(_t) = 2e-e.5, u(_t) and vc(.0-) = O.
Solution. The differential equation is

e(_t) = Ri(t) + ~ f ~i(T) d-r + vc(.0-) +t

(5.43) -
Or, in terms of the numerical values, we
have
U--0· 6' u(_t) = i(t) +2 J~(-r) d-r (5.44) FIG. 5.18

Differentiating both sides of Eq. 5.44, we obtain

di(t)
2c5(t) - e➔· 11' u(_t) = di + 2i(t) (5.45)

To obtain the initial condition i(O +), we must integrate Eq. 5.45 between
the limits t = 0- and t = 0 + to give i(0 +) = 2. From the characteristic •
equation
H(p) = p + 2 = 0 (5.46)
we obtain the complementary function as
ic(.t) = Ke-11 (5.47)
H we assume the particular integral to be i,,,(t) = Ae.....·11', then we obtain
1 2
A =- H( -0.S) = -· 3 (5.48)
The incomplete solution is
= Ke-11 - fe--O.&t
i(t) (5.49)
From the initial condition i(0 +) = 2, we obtain the final solution,
;c,> = cie-11 - 1e-e·0'> u<.1> cs.so>
As we already noted in Section 5.3, in the solution of network differential
equations, the complementary function is called the free response, whereas
the forced response is a particular integral, and in -the case of constant or
periodic excitation, the forced response at t = oo is the steady-state
solution. Note that the free response is a function of the network elements
 116 Network analysis and synthesis
alone and is independent of excitation. On the other hand, the forced
response depends on both the network and the excitation.
It is significant that, for networks which have only positive elements,
the free response is made up of only damped exponential and/or sinusoids
with constant peak amplitudes. In other words, the roots of the charac-
teristic equation H(p) all have negative or zero real parts. For example,
if Pi is a root of H(p) written as Pi - a ± jw, then Re (pJ - a ~ 0.
This.fact is intuitively reasonable because, if a bounded excitation produces
a response that is exponentially increasing, then conservation of energy is
not preserved. This is one of the most important properties of a passive
network. If a characteristic equation contains only roots whose real parts
are zero or negative, and if the jw axis roots are simple, then the network
which it describes is said to be stable; otherwise, the network is unstable.•
Stability is an important property of passive networks and will be discussed
in greater detail later.
Evmple 5.6. For the R-C network in Fig. 5.19 with the excitation given by
Eq. 5.25, find the voltage v(I) across the capacitor; it is given that v(0-) -
vo(0-) - 0.
Solution. We have already obtained the particular integral in Eq. 5.28.
Now let us find the couiplementary function. The differential equation on a
node basis is
dv .
C di + Gv == 10 SID OJI u(.I) (5.51)

froui which we obtain the characteristic equation as

H(p)-Cp +G (5.52)
so that vo(t) - K£G'IC u(_t) (5.53)
and the incomplete solution is
v(1) ... x.e-G41°u<_1) + (GI :•o}C')1A sin ( OJI - tan-1 OJi) 11(,1) (5.54)

From the initial condition v(0-) - 0, we obtain

1
v co+> == v<o->- x - < GI + !,.C'>~
sin ( tan-
1
';,c) == o (5.55)

Froui the argand diagram in Fig. 5.20 we see that,

1 (5.56)
sin ( tan- ~) - (GI +OJ~C')1A
Consequently,
/0 11(,t) [ OJCe-GlfC • ( _ 1 OJC)]
v(t) =- (GI + o}C')1A (GI + o}C')~ + SID OJI - tan G (5.57)

'This is not a formal definition of stability, but it sufllces for the moment.
 Network analysts: I 117
.,,,,
Im

B C

Be
Dllurn
FIG. I.It FIG.Ut

From the complementary function in Eq. 5.53, we aee that the time constant of
the ciralit is T - C/G - RC.
Next, let us examine an example of the diff'erent kinds of free responses
of a second-order network equation that depend on relative values of the
network elements. Suppose we are given the network in Fig. 5.21; let us
find the free response vc(t) for the diff'erential equation

i.(t)-= cdv +Go+! f1 v(:r)dT + iL(O-) (5.58)

dt LJe-
Differentiating both sides of Eq. 5.58, we have
i'.(t) =- C v"(t) + G v'(t) + !L v(t). (5.59)
The cbaracteristic equation is then
H(p) ""' cp' + Gp + ! = c(p' + Q, + .!.) (5.60)
L C LC
In factored form, H(p) is
H(p) - C(p - p-J(p - p~ (5.61)
.,,,,

is(I) C G L

PIG. 5.21
118 Network analysis and synthesis
v(t)

FIG. 5.22. Overdamped response.

where Pi} = _ _Q_ ± ![(~t'f _...£]~ £ -A ± B (5.62)

Ps 2C 2 cJ LC
There are three different kinds of responses depending upon whether B is
real, zero, or imaginary.
CASE l. B is real, that is,

then the free response is

v 0 (t) = K1e-l.4.-B>t + K1e-<.4.+B>t (5.63)
which is a sum of damped exponentials. In this case, the response is said
to be overdamped. An example of an overdamped response is shown in
Fig. 5.22.
CASE 2. B = 0, that is,

then P1 = P1 = -A
so that va(t) = (K1 + K 8t)e-.4.t · (5.64)
When B = 0, the response is critically damped, as shown in Fig. 5.23.
v(t)

FIG. 5.23. Critically damped response.

 Network analysis: I 119
v(t)

FIG. 5.24. Underdamped response.

CASE 3. Bis imaginary, that is,

(Q'f < ..!.

cJ LC
Letting B == jp, we have

vo(t) == e-..4'(K1 sin Pt + K~ cos Pt) (5.65)

In this. case, the response is said to be underdamped, and is shown by the
damped oscillatory curve in Fig. 5.24.
Example 5.7. In this example we discuss the solution of a set of simultaneous
network equations. As in the previous examples, we rely upon physical reasoning
rather than formal mathematical operations to obtain the initial currents and
voltages as well as the steady-state solutions. In the network of Fig. 5.25, the
switch Sis thrown from position 1 to position 2 at t = 0. It is known that
prior to t = 0, the circuit had been in steady state. We make the idealized
assumption that the switch closes instantaneously at t = 0. Our task is to
find i 1(t) and i 1(t) after the switch position changes. The values of the batteries
V1 and V1 are V1 = 2 v, V1 = 3 v; and the element values are given as

L = lh, R1 = O.S n
C = if, Ra= 2.on
The mesh equations for i 1(t) and i.(t) after t = 0 are
V1 == L i\(t) + R1 i 1(t) - R 1 i 1(t) (S.66)

(S.67)
120 Network analysis and synthesis

R1

FIG.5.25

Since Eq. 5.67 contains an integral. we differentiate it to give

1
0 = -R1 1'1(t) + CiJ..t) + (R1 + R,Ji',f.t) (5.68)

Using Eqs. 5.66 and 5.68 as our system of equations, we obtain the character-
istic equation
l.p + R1
H(p)-
-RiJI

(S.69)

Substituting the element values into H(p), we have

H(p) - 2.5r + 4p + 1.5 == 2.S(p + l)(p + 0.6) (S.70)

1be free responses are then

i10(t) - (K1e--e.tt + Kr') 1(11)
(S.71)
i,c(t) - (K,--0·" + X.,r') 11(.t)

The steady-state solutions for the mesh currents are obtained at t - ao by

considering the circuit from a d-c viewpoint. The inductor is then a short
circuit and the capacitor is an open circuit; thus we have

v.
11.(t) - Ri .. 6 amp
(S.72)
,•.c,) - 0
Now let us determine the initial currents and voltages, which, incidentally, have
the same values at t .. 0- and t - 0 + because the voltage sources· are not
 Network analysis: I 121
impulses. Before the switch is thrown at t - 0, the circuit with Y1 u the voltage
IIOU1'Ce waa at steady state. Consequently,

oc(O-) - Y1 -2v

(5.73)

la(O-) - 0
We next find i'1(0+) from Eq. 5.66 at t =- o+.
Y1 - Li'1(0+) + R1 i1(0+) - R1 1a(O+) (5.74)
Substituting numeric:al values into Eq. 5.74, we find
i\(0 +) - 1 amp/llOC
From Eq. 5.68 at t - o+, we obtain similarly
R
i'J..O +) - Ri ;. Rs i\(O +) - 0.2 amp/llOC (5.75)

With these initial values of i1(t) and la(t), we can quickly arme at the ftnal
solutions
i1(t) .. (0.5r1 - 2.5r°·" + 6) 11(,t)
(5.76)
iJ..t> - <-0.se-1 + o.,e➔-") ,(t)
which are plotted in Fig. 5.26.

,_ 3.5 4.0 4.5 5.0

FIG. 5.26
122 Network analysis and synthesis

5.6 ANALYSIS OF TRANSFORMERS

According to Faraday's law of induction, a current i1 flowing in a coil L 1

may induce a current i1 in a closed loop containing a second coil L 1• The
sufficient conditions for inducing the
current i1 are: (a) part of the flux c1>1 in
+ + the coil L 1 must be coupled magnetically
• • to the coil L 8 ; (b) the flux cf>1 must be
changing with time.
112
In this section we will analyze circuits
containing a device made up of two
_ magnetically coupled coils known as a
.___ _ _-a transformer. In Fig. 5.27, the schematic
FIG. 5.27. Transformer. of a transformer is given. The L 1 side
of the transformer is usually referred to
as the primary coil and the L 1 side as the secondary coil. The only distinction
between primary and secondary is that the energy source is generally at
the primary side.
The transformer in Fig. 5.27 is described mathematically by the
equations
, ( )
vt==&J}-+
T di1 M -di.
1
dt dt
(5.11)
( ) M di1 + L di 1
Vat== - •-
dt dt
where Mis the mutual inductance associated with the flux linking Li to La,
and is related .to Li and L 8 by the relationship
M == K -J £iL1 (5.78)
The constant Kin Eq. 5.78 is called the coefficient of coupling. It is
bounded by the limits O ~ IKI ~ 1. If IKI = l, then all of the flux cf>1 in
coil Li is linked magnetically to L 1 • In this case, the transformer is a unity-
coupled transformer. If K = 0, the coils L 1 and L 1 may be regarded as
two separate coils having no effect upon one another.
For circuits with transformers, we must establish reference polarities
for the mutually induced voltages M di/dt Usually, the references are
given by small dots painted on the input and output leads of a transformer,
as shown by the dots on the schematic in Fig. 5.28. The reference dots are
placed at the time of manufacture according to the procedure outlined
here. A voltage source v is connected to the primary Li side of the
transformer, as shown in Fig. 5.28. A voitmeter is attached on the second-
ary. At the primary side, the terminal is assigned the dot reference to
 Network analysis: 123

Voltage+
source

FIG. 5.21. An experiment to determine dot references.

which we connect the positive lead of the voltage source. The dot reference
is placed on the secondary terminal at which the voltmeter indicates a
positive voltage. In terms of the primary current ii, the positive voltage at
the secondary dot is due to the current i1 flowing into the dot on the
primary side. Since the positive voltage at the secondary dot corresponds
to the current i1 flowing into that dot, we can think of the dot references
in the following way. If both currents are flowing into the dots or away
from the dots, then the sign of the mutual voltage term M di/dt is positive.
When one current flows into a dot and the other away from the second dot,
the sign of M di/dt is negative.
If N 1 and N1 are the number of turns of coils Li and L 1, then the flux
linkages of Li and La are given by N1<f>1 and N 1<f>1, respectively. If both i1
and i1 flow into the dots, the sum of flux linkages of the transformer is
.I <f> linkages = N1<f>1 + N1<f>1 (5.79)
If, however, one of the currents, for example, i1, flows into a dot, and the
other i1 flows out of the other dot, then
(5.80)
An important rule governing the behavior of a transformer is that the
sum of flux linkages is continuous with time.
The differe'1,tial equations for the transformer in Fig. 5.29 are
V u(t)= L1 i'1(t) + R1 i1(t) + Mi' .(t)
(5.81)
0 = Mi' i(t) + Ra i 1(t) + Lai'.(t)

FIG, 5.29
124 Network analysis and synthesis
Integrating this set of equations between t = 0- and t = o+ results in
the determinant

I 1.ic,1(0+) _ ,1(0-)] M[;.(o+) - ,.co-)]

Mlii(O+) - 11(0-)] l.s[;.(O+) - i.(0-)]
I- - 0 (5.82)

By evaluating this determinant, we obtain

(LiL. - .W)[i1(0+) - 11(0-)][t.(0+) - ,.co-)] = 0 (5.83)
If Li£. > .W, that is, K < l, then the currents must be continuous at
t =- 0 in order for the determinant in Eq. 5.82 to be equal to zero. Thus
11(0+) = 11(0-), K <l .(5.84)

and i.(O+) = ;.(O-), K < l (5.85)

Example 5.8. For the transformer circuit in Fig. S.29, L1 - th. L 1 ... 2h,
R1 -= 30, Rs ,. 80, and M =- th. The excitation is V =- 6u(_t). Let us find
i1(t) and is(t), assuming that 11(0 - ) - i,1..0 - ) ... 0.
Solution. The differential equations for the circuit are
6u(t) == I'1(t) + 3i1(t) + I'.(t)
(5.86)
0 == I'1(t) + 21'.(t) + 8i1(t)
The characteristic equation is given by the determinant

H(p) - Ipp + 3 2pp+ 8 I

- <P + 3)(2p + 8) - r
= (pl + 14p + 24)
= (p + 2)(p + 12) (5.87)
Thus the complementary functions are

i1 0(t) == Kie•' + Ksr"'

(5.88)
i1o(t) = K.e- 11
+ K,,cm
To obtain the particular integral, or steady-state solution, we rely upon
physical reasoning to arrive at

11.(t) = -V = lamp (5.89)

R1
i1.(t) == 0 amp
 Network analysis: I 125
Since the excitation does not contain an impulse (and since LiL. -
M' p 0, as we shall see later), we can assume that '1.(0+) and '9(0+) are
also 7.ero. Then using Eq. 5.86 we can find i'1(0+) and i'.(0+).
6 - i'1(0+) + i'.(0+)
(5.90)
0 =-·i't(O+) + 21'.(0+)
Solving, we find i'1(0+) =- 12 and I'.(O+) == 6. With these initial con-
ditions, we obtain the final solutions of i1(t) and i.(t)
'1.(t) .. (2 - tc•' - te-111) u(t)
(5.91)

Suppose, now, LiL. - M', that is, K =

l, then i1 (t) and la(t) need not be
continuous at t - 0. In fact, we will show that the currents are discon-
tinuou8 at t .. 0 for a unity-coupled transformer. Assuming that K l, =
consider the mesh equation for the secondary at t o+ =
R1 i.(O+) == -M i'1(0+) - L. i'.(O+) (5.92)

=- M [Li i'i(O+) + M i'.(O+)] (5.93)

Li
The mesh equation of the primary side then becomes

V == R1 i1(0+) + [Li i'1(0+) + M i'.(O+ )] == R1 i1(0+) - Li R1 i.(O+)

M
(S.94)
We need an additional equation to. solve for i1(0+) and '9(0+ ). This is
provided by the equation
Li[i1(0+) - i1(0-)] + M[i,(O+) - i.(0-)] == 0 (5.95)
which we obtained from Eq. 5.82. Since i1(0-) == i,(0-) = 0, we solve
Eqs. S.94 and S.9S directly to give
.1(0+) VL.
' == R1.l.a + R.Li
(S.96)
i.(O+) = -VM
R1.l.a + R.Li
Consider the following example. For the transformer in Fig. S.29
the element values are
126 Network analysis and synthesis
Assuming that the circuit is at steady state before the switch is closed at
t = 0, let us find i 1(t) and i 1(t). The differential equations written on mesh
basis are
10 = 4 di1 + 8i1 + 2 di1
dt dt
(5.97)
0 = 2 di1 + di1 + 3i1
dt dt
The characteristic equation is

H(p) = I4(p + 2)
2p· (p
2p
+ 3)
I- 0 (5.98)

which yields H(p) = 20p + 24 = 20(p + f) (5.99)

so that the complementary functions are

i1c(t) = K1e-1-1'
(5.100)
iac(t) = K.,r1.1t

The particular integrals that we obtain by inspection are

. V 10 5
' 1s,(') = R1 = 8 = 4 (5.101)
i1,it) =0
The initial conditions are

(5.102)
i1(0+) = -VM = -1.0
R1L 1 + R1Li
We then find K1 = -0.75 and K 1 == - 1.0 so that
ii(t) == (-0.75cl-lt + 1.25) u(t)
(5.103)
i,.(t) = -e--1•1' u(t)
We see that as t approaches infinity i1(t)--+ 0, while i1(t) goes its steady-
state value of 1.25.
 Network analysis: I 127

Problems
S.1 Write the mesh equations for the network shown.

PROB. 5.1

S.2 Write a set of node equations to solve for the voltage v.(.t) shown in the
figure.

i(t)u(t) t Ro
+
Co "2(t)

PROB. 5.2

S.3 The network shown has reached steady state before the switch S is
opened at t = 0. Determine the initial conditions for the currents i1(t) and
i.(.t) and their derivatives.

PROB.5.3

S.4 The network shown has reached steady state before the switch moves
from a to b. Determine the initial conditions for h(t) and Oc(t) and their first
derivatives. Determine also the final values for iii.t) and vo<.t).
118 Network analysis and synthesis

1D

PROB. 5.4

5.5 The network shown has reached steady state before the switch moves
from a to b. Determine the initial conditions for the voltages v1(t) and v1(t)
and their first derivatives. Determine also the final values for v1(t) and v,1..t).

PROB, 5,5

5,6 For the network shown h(0-) = 0.

(a) Find v(t); 0- < t < oo.
(b) Show that v(t) approaches an impulse as G -+ 0.
(c) Find the strength (area) of the impulse.

11(t)

PROB. 5.6

5.7 For the network shown. i(t) -= d(t) - c' u(_t) and v(O-) - 4. Find
and sketch ic(t); 0- < t < oo.
 Network analysis: I 129
10 v(t)

i(t) 10

PROB.5.7

5.8 For the network shown. before the switch moves from a to b, steady-
state conditions prevailed. Find the current i(t).

+
v.=.
L

PROB. 5.8

5.9 For the circuit shown, switch S is opened at t = 0 after the circuit had
been in steady state. Find i(t); 0- <t < oo.

PROB. 5.9

5.10 Find the current i(t) in the network shown when the voltage source is a
unit impulse. Discuss the three different kinds of impulse response waveforms
possible depending upon the relative values of R, L, and C. All initial conditions
are zero at t - 0-.
130 Network analysis and synthesis

+
e(t) R

PROB.S.10

5.11 An R-C differentiator circuit is shown in the figure. Find the require-
ment& for the R-C time constant, such that the output voltage vJ..t) is approxi-
mately the derivative of the input voltage.

C
-+-

+
. e(t) R vo(t)

PROB.5.11

5.12 An R-C integrator circuit is shown in the figure. Find the requirements
for the time constant such that the output v0(t) is approximately the integral of
the input voltage.
R
--+-

~
e(t) C vo(t)

PROB.5.12

5.13 Find the free response for i(t) in the figure shown for (a) a current
source; (b) a voltage source.
 Network analysis: 131

20

Source ¼t ~
") lh

PROB.5.13

5.14 At t = 0, the switch goes from position 1 to 2. Find i(t), given that
e(_t) = e-' sin 2t. Assume the circuit had been in steady state for t < 0.

2h

6f 40

PROB.5.14

5.15 For the circuit shown, switch S closes at t = 0. Solve for i(t) when the
initial conditions are zero and the voltage source is e(_t) = sin ( wt + 6). What
should 6 be in terms of R, C, and w so that the coefficient of the free response
term is zero?
R

PROB. 5.15

5.16 For the circuit shown, the switch S moves from a to b at t = 0. Find
and sketch v1(t) for O - < t < oo. The circuit is in a steady state at t = 0.
132 Network analysis and synthesis

~ l,. nt=O
vi r r-·+~
,:i..

PROB.5.16

5.17 For the circuit shown, i(t) = 4c1t u(t). Find v{t); 0 - <t < oo.
u(t)

-i (t)

1f

PROB. 5.17

5.18 Find the complete solution for IL(t) and vo(t) in Prob. 5.4.
5.19 For the circuit shown, the switch is closed at t = 0. Find i1(t) and
i.J_t) for O - < t < oo. Assume zero initial energy.
100 1h 1h

st
100-=-
100~ 100

PROB.5.19

5.20 For the transformer circuit shown, the switch closes at t -= O. Find
i 1(t) and i1(t). Assume zero initial energy.
S 10

,.l/2.w\ lh
•
4h~ 10

M=lh
PROB.5.20
 Network analysis: I 133

PROB. 5.21

5.21 For the transformer circuit shown, the s~tch S opens at t = 0. Find
the voltage (v1 t) for O - < t < co. Assume zero initial energy.
 chapter 6
The Laplace transform

6.1 THE PHILOSOPHY OF TRANSFORM METHODS

In Chapters 4 and 5, we discussed classical methods for solving differ-

ential equations. The solutions were obtained directly in the time domain
since, in the process of solving the differential equation, we deal with
functions of time at every step. In this chapter, we will use Laplace
transforms to transform the differential equation to the frequency domain,
where the independent variable is complex frequency s. It will be shown
that differentiation and integration in the time domain are transformed
into algebraic operations. Thus, the solution is obtained by simple
algebraic operations in the frequency domain. .
There is a striking analogy between the use of transform methods to
solve differential equations and the use of logarithms for arithmetic
operations. Suppose we are given two real numbers a and b. Let us find
the product
C=axb (6.1)
by means of logarithms. Since the logarithm of a product is the sum of the
logarithms of the individual terms, we have
log C = log a x b = log a + log b (6.2)
so that C = log-1 (log a + log b) (6.3)
If a and b were two six-digit numbers, the use oflogarithms would probably
facilitate the calculations, because logarithms transform multiplication
into addition.
An analogous process is the use of transform methods to solve integro-
differential equations. Consider the linear differential equation
y(x(t)) =/(t) (6.4)
134
 The Laplace transform 135

Differential .Algebraic
equation equation X(s) x(t)
Transform Solve Invert

FIG. 6.1. Philosophy of transform methods.

where /(t) is the forcing function, x(t) is the unknown, and y(x(t)) is the
differential equation. Let us denote the transformation process by T(·),
and let s be the frequency variable. When we transform both sides of
Eq. 6.4, we have
T[y(x(t))) == T[/(t)] (6.5)
Since frequency domain functions are given by capital letters, let us write
Eq. 6.5 as
Y(X(s), s) = F(s) (6.6)
where X(s) = 7lx(t)], F(s) = T[/(t)], and Y(X(s), s) is an algebraic
equation ins. The essence of the transformation process is that differential
equations in time are changed into algebraic equations in frequency. We
can then solve Eq. 6.6 algebraically to
obtain X(s). As a final step, we perform
Differential
an inverse transformation to obtain e(t) - equation r(t)
x(t) = T-1 [X(s)] (6.7) E(B) System R(•) -
function
In effecting the transition between the
time and frequency domains, a table of FIG. 6.2. Linear system.
transform pairs {x(t), X(s)} can be very
helpful. A diagram outlining the use of transform methods is given
in Fig. 6.1. Figure 6.2 shows the relationship between excitation and
response in both the time and frequency domains.

6.2 THE LAPLACE TRANSFORM

The Laplace transform of a function of time /(t) is defined as

C[/(t)] = F(s) = I: J(t)e-•' dt (6.8)

where s is the complex frequency variable

s==a+~ ~~
This definition of the Laplace transform is different from the definition
given in most standard texts,1 in that the lower limit of integration is
1
M. E. Van Valkenburg, Network Analysis, 2nd Ed. Prentice-Hall, Englewood Cliffs,
N:ew Jersey, 1964.
136 Network analysis and synthesis
t == 0- instead oft== o+. We thus take into account the possibility
that/(t) may be an impulse or one of its higher derivatives. It is clear that
C[6(t)] == 0 for the o+ definition, whereas for the 0- definition, C['5(t)] == I.
In the case when no impulses or higher derivatives of impulses are involved,
it was shown in Chapter 4 that /(0-) = f (0+ ). Therefore, all of the
"strong results" resulting from a rigorous treatment of Laplace trans-
forms• obtained by using t == 0+ as a lower limit also apply for the 0-
definition.
In order for a function to possess a Laplace transform, it must obey the
condition
J: lf(t)I e-a' dt < oo (6.10)

for a real, positive a. Note that, for a function to have a Fourier trans-
form, it must obey the condition

L: lf(t)I dt < <X) (6.11)

As a result, a ramp function or a step function will not possess a Fourier

transform,• but will have a Laplace transform because of the added
1
convergence factor e-"'· However, the function e' will not even have a
Laplace transform. In transient problems, the Laplace transform is
preferred to the Fourier transform, not only because a larger class of
waveforms have Laplace transforms but also because the Laplace trans-
form takes directly into account initial conditions at t == 0- because of
the lower limit of integration in the Laplace transform. In contrast, the
Fourier transform has limits of integration ( - oo, oo), and, in order to take
into account initial conditions due to a switch closing at t == 0, the forcing
function must take a form as f(t) u(t) + f(0-) '5(t), where f(0-) repre-
sents the initial condition.
The inverse transform c-1 [F(s)] is

f(t)
1 J."1+.100 F(s)e" ds
== -. (6.12)
2'1TJ 111-.1«>

where a 1 is a real positive quantity that is greater than the a convergence

factor in Eq. 6.10. Note that the inverse transform, as defined, involves a
1 D. V; Widder, Tire Laplace Transform, Princeton University Press, Princeton, 1941.
• These functions do not possess Fourier transforms in the strict sense, but many
possess a generalized Fourier transform containing impulses in frequency; see M. J.
Lighthill, Fourier Analysis and Generalized Functions, Cambridge University Press, New
York, 1959.
 The Laplace transform 137
complex integration known as a contour integration.' Since it is beyond
the intended scope of this book to cover contour integration, we will use a
partial fraction expansion procedure to obtain the inverse transform. To
find inverse transforms by recognition, we must remember certain basic
transform pairs and also use a table of Laplace transforms. Two of the
most basic transform pairs are discussed here. Consider first the transform
of a bnit step function u(t).
Euaple 6.1. /(t) - u(t).

J.....
00
F(s)-f u(t)e-ddt ..,·_ e-d,oo =-0
s 1)-
-(-!)s _!s (~.13)

Next, let us find the transform of an exponential function of time.

Example 6.2. /(t) .. e'"u(t).

F(s) -
L
oo
1)-
ene-d dt
c<e-o)I
=- - -
s-a
t =-
l
-
s-a
(6.14)

With these two transform pain, and with the use of the properties of Laplace
transforms, which we discuss in Section 6.3, we can build up an extensive table
of transform pain.

6.3 PROPERTIES OF LAPLACE TRANSFORMS

In this section we will discuss a number of important properties of

Laplace transforms. Using these properties we will build up a table of
transforms. To facilitate this task, each property is illustrated by con-
sidering the transforms of important signal waveforms. First let us
discuss the linearity property.
Linearity
The transform of a finite sum of time functions is the sum of the trans-
forms of the individual functions, that is

c[t .f.(t>] = +t[/Jt)J (6.15)

This property follows readily from the definition of the Laplace transform .
.Example 6.3. /(t) ... ~in o,t. Expanding sin o,t by Euler's identity, we have

(6.16)

' For a lucid treatment, see S. Goldman, Transformation Calculu and &ctrica/
Trans~nts, Prentice-Hall, Englewood Cliffs, New Jersey, 1949, Chapter 7.
138 Network analysis and synthesis
The Laplace transform of /(t) is the sum of the transforms of the individual
cisoidal e±ion terms. Thus
C(sin wt] == .!.(-1
-
1
- - - ) = _w_
2j s - jw s + jw + wl r (6.17)

R•I differentiation
Given that C(/(t)] == F(s), then

c[t] == s F(s) - f(0-) (6.18)

where/(0-) is the value of/(t) at t == 0-.

Proof. By definition,
C[J'(t)] f
== ~ e-•t j'(t) dt (6.19)

Integrating Eq. 6.19 by parts, we have

C[f'(t)] = e-•tj(t) I:+ sf~J(t)e-•t dt (6.20)

Since e-•t -+ 0 as t -+ oo, and because the integral on the right-hand side is
C(/(t)) = F(s), we have
C(/'(t)] = s F(s) - f(0-) (6.21)
Similarly, we can show for the nth derivative

c[d"f(t)] = s" F(s) - s"-1 f(0-) - s"-z f'(0-) - · · · - /"-11(0-)

dt"
(6.22)
where pn-1 > is the (n - l)st derivative of f(t) at t = 0-. We thus see
that differentiating by t in the time domain is equivalent to multiplying
by s in the complex frequency domain. In addition, the initial conditions
are taken into account by the terms f'i>(O-). It is this property that
transforms differential equations in the time domain to algebraic equations
in the frequency domain.
Eulllple 6.4a. /(t) = sin wt. Let us find

£(cos wt]= c[~ ~sin wt] (6.23)

By the real differentiation property, we have

{~sin wt] = s(r: w•) (6.24)

 The Laplace transform 139

so that C[cos cot] s(

= -co - co- ) ""' - -
s' + col s' + col
s (6.25)

In this example, note that sin co(0 - ) = 0.

Example 6.4b. f(t) - ,(t). Let us find t(f'(t)], which is the transform of the
unit impulse. We know that
1
t[u(t)] =- (6.26)
s

Then t[6(,)) = s(D = 1· (6.27)

since u(0-) = 0.
R•I int..ration
If £.[j(t)] = .F(s), then the Laplace transform of the integral of /(t) is
F(s) divided by s, that is,

c[f~ /(-r) d-r] = F~s) (6.28)

Proof. By definition,

(6.29)

(6.30)

Since e-•' - 0 as t - oo, and since

f' /(-r)d-rl
Jo- t-0--
=0 (6.31)
we then have
(6.32)

EuJDple 6.S. Let us find the transform of the unit ramp function, p(_t) - t u(_t).

We know that J:u(T)dT = p(t)

1
(6.33)

Since t[u(t)] =- (6.34)

s

C[u(t)] 1
then t[p(t)] ... - s- - ii (6.35)
I.fO Network analysis and synthesis
Differentiation by •
Differentiation by s in the complex frequency domain corresponds to
multiplication by t in the domain, that is,

C[tf(t)] == - d F(s) (6.36)

ds
Proof. From the definition of the Laplace transform, we see that

dF(s) == f"'f(t)~e-•'dt = _ f"' tf(t)e-•'dt = -t[tf(t)] (6.37)

ds Jo- ds Jo-
Euaple 6.6. Given/(t) = e-..,, whose transform is
1
F(s) ... - - (6.38)
s + (X

let us find C[te-..']. By the preceding theorem, we find that

C[te-«'] d ( -I- ) = -1-

= - -ds (6.39)
s + °' (s + 01)1
Similarly, we can show that
n!
C[t"e-«'] - - - ~ (6.40)
(s + 01)fl-tl
where n is a positive integer.
Complex translation
By the complex translation property, if F(s) = 1:(/(t)], then
F(_s - a) == t[e"' /(t)] (6.41}
where a is a complex number.
Proof. By definition,

l:[e"'J(t)] f
= ~ e''f(t)e-" dt = f: e-<a-a>'J(t) dt = F(s - a) (6.42)

Eumple 6.7. Given /(t) =- sin wt, find C[r' sin wt]. Since
w
. t[sin wt) - sl + o,I (6.43)

by the preceding theorem, we find that

w
1
t [ r sin wt] = .(s + a)I + w• (6.44)

Similarly, we can show that

s+a
C[e-o• cos wt] - (s + a)I + o,I (6.45)
 The Laplace transform 141
Real translation (shiftin1 theorem)
Here we consider the very important concept of the transform of a
shifted or delayed function of time. If t[/(t)] = F(s), then the transform
of the function delayed by time a is
t[/(t - a) u(t - a)] = r• F(s) (6.46)
Proof. By definition,

t[/(t - a) u(t - a)] = i<Xl e-"J(t - a) dt (6.47)

Introducing a new dummy variable -r =t- a, we have

C[/(-r) u(-r)] = J: e-•lr+al f(-r) d-r = e--f: J(-r)e-n d-r = e-a F(s) (6.48)

In Eq. 6.48 /(-r) u(-r) is the shifted or delayed time function; therefore
the theorem is proved.
It is important to recognize that the term r• is a time-delay operator.
If we are given the function r• G(s), and are required to find
c-1[r• G(s)] = K1(t),
we can discard r• for the moment, find the inverse transform
C-l(G(s)] - g(t),
and then take into account the time delay by setting
g(t - a) u(t - a) = g1 (t) (6.49)
Euaple 6.8a. Given the square pulse /(t) in Fig. 6.3, let us first find its trans-
form F(s). Then let us determine the inverse transform of the square of F(s),
i.e., let us find
/1(t) = C-1[.F'(s)] (6.50)
Solution. The square pulse is given in terms of step functions as
f(t) - u(t) - u(t - a) (6.51)
Its Laplace transform is then
1
F(s) .. -(I -e--)
s
(6.52) "''
Squaring F(s), we obtain 1----
1
F'(s) = ~(I - 2e-a + e-111') (6.53)

To find the inverse transform of F'(s), we 0 G

need only to determine the inverse transform FIG. 6.3. Square pulse.
 I◄2 Network analysis and synthesis
of the term with zero delay, which is

c-1[~] - t u(t) (6.54)

Then c-i[F'(s)] = t u(t) - 2(t - a) u(t - a) + (t - 2a) u(t - 2a) (6.55)

so that the resulting waveform is shown in Fig. 6.4. From this example, we see
that the square of a transformed function does not correspond to the square
of its inverse transform.

f(t)

Ko Ks

0 a 2a 0 Ti 2Ti 3Ti 4T1 5T1 6T1 7T1

FIG, 6.4. Triangular pulse. FIG. 6.5. Impulse train.

· Example 6.8b. In Fig. 6.S, the output of an ideal sampler is shown. It consists
of a train of impulses
f(t) ... Ko «5(t) + Ki «5(t - Ti) + · · · + Kn «5(t - nTi) (6.56)
The Laplace transform of this impulse train is
C[/(t)] "" K0 + K1e--''1 + K<;1r2•T1 + · · · + x_e-.Ti (6.57)
In dealing with sampled signals, the substitution z == e•T1 is often used. Then
we can represent the transform of the impulse train as
Ki Ka Kn
C[/(t)] - K0 + -;- + z• + · · · + zti (6.58)

The transform in Eq. 6.58 is called the z-transform of /(t). This transform is
widely used in connection with sampled-data control systems.

PERIODIC WAVEFORMS

The Laplace transform of a periodic waveform can be obtained in two

ways: (a) through summation of an infinite series as illustrated in Example
6.8d, and (b) through the formula derived below.
T f8T
£[/(t)]
i
= ~ f(t)e- st dt +JT J(t)e-•' dt + ···
(6.59)
J
(n+i)T
+ nT
f(t)e_., dt + ···
 The Laplace transform 143
Since/(t) is.periodic, Eq. 6.59 reduces to

£.[J(t)] = J:J(t)e-" dt + e-•TJ:J(t)e_., dt + · · ·

+ e-•nTJ: f(t)e- st dt + · · ·
(6.60)
= (1 + e-•T + e-••T + · · ·)J:J(t)e-•' dt

= 1 iT f(t)e-st dt
1...:. e-•T 0--
EuJDple 6.8c. Given the periodic pulse train in Fig. 6.6 let us use Eq. 6.60 to
determine its Laplace transform.

F(s) =1 i"
1 -,T o-e_., dt
-e

- ---,----=
-1
- s(l - e-•T)
e-d
o-
I" (6.61)

1 1 - e-a•
=; 1 - e-•T

f(t}

0 a T T+a 2T 2T+a
FIG. 6.6. Periodic pulse train.

Example 6.8d. In this example we calculate the Laplace transform of /(t) in

Fig. 6.6 using summation of an infinite series. The periodic pulse train can be
represented as

f(t) = u(_t) - u(_t - a) + u(_t - n - u[t - (T + a)] ( • )

2n - u[t - (2T + a)] + .. ·
+ u(_t - 6 62
Its Laplace transform is
1
F(s) =- (1 - e-aa + e-•'l' - e-<T+a>, + •· •)
s
(6.63)
 I# Network analysis and synthesis
which simplifies further to give
1
F(s) = - (1
s
- e.,...')(1 + e-•'1' + e-'u'l' + · · ·) (6.64)

We then see that F(s) can be given in the closed form

11 -e-•
F(s) = -s l
-e-•'1' (6.65)

Other periodic pulse trains also can be given in closed form. The reader is
referred to the problems at the end of this chapter.
At the end of the chapter is a table of Laplace transforms. Most of
the entries are obtained through simple applications of the properties just
discussed. It is important to keep those properties in mind, because many
transform pairs that are not given in the table can be obtained by using
these properties. For example, let us find the inverse transform of

F(s) = ws (6.66)
. (s + a)8 + w1
Since the s in the numerator implies differentiation in the time domain, we
can write
F(s) =s w1 (6.67)
(s + a) + w 1

Using the differentiation property, we obtain

c.-1[F(s)] = !!. (e-cn sin wt)

dt (6.68)
= (- a sin wt + <.o cos wt)e-cn
Note that e-"' sin wt at t = 0- is zero.
6.4 USES OF LAPLACE TRANSFORMS

Evaluation of definite int91rals

The Laplace transform is often useful in the evaluation of definite
integrals. An obvious example occurs in the evaluation of

I = i 00
e-11 sin St dt (6.69)

Ifwe replace e-1' bye-•', the integral I then becomes the Laplace transform
of sin St, which is
5
£.[sin St] = (6.70)
s1
+ 25
 The Laplace transform 145
Replacing s by 2, we have
I- I= 5 =2_ (6.71)
21 + 25 29
Perhaps a more subtle example is the evaluation of

I =f -1
+1 t•e-•ltl dt (6.72)

First, we note that t•e-•l•I is an even function; therefore

I =2 i 1
t•e-•• dt (6.73)

From the table of Laplace transforms we see that

2
t[t1e-a,] = (6.74)
(s + 2)1
and the transform of
f(t) = J: -r•e-lT d-r (6.75)

is C[/(t)] = s(s : 2)a (6.76)

Later we will see that the partial-fraction expansion ofC[f(t)] in Eq. 6.76
consists of three terms for the multiple root (s + 2)3, as given by
C(/(t)] == 0.25 _ 0.25 _ 0.5 1 (6 _77)
s s + 2 (s + 2)1 (s + 2)1

Taking the inverse transform of Eq. 6.77, we obtain

ft.t) == 0.25(1 - e-1 ' - 2te-•• - 2t1r') u(t) (6.78)
Now observe that the definite integral in Eq. 6.73 is equal to 2/(t) at
t = I, that is,
t =
2/(1) == -2.Se-8 + 0.5 = 0.162 (6.79)
Solution of lnt91roclifferential equations
In Section 6.3 we said that the real differentiation and real integration
properties of the Laplace transform change differential equations in time
to algebraic equations iri frequency. Let us consider some examples
using Laplace transforms in solving differential equations.
Example 6.9. Let us solve the differential equation
:,;'(t) + 3:1:'(t) + 2:,;(t) - 4e' (6.80)
given the initial conditionu(O-) = 1, z'(O - ) = -1.
1"46 Network analysis and synthesis
Solution. We first proceed by taking the Laplace transform of the differential
equation, which then becomes
4
[r X(s) - s x(0-) - x'(0-)] + 3[s X(s) - x(0-)] + 2X(s) = s _ (6.81)
1
Substituting the initial conditions into Eq. 6.81 and simplifying, we have
4
(r + 3s + 2) X(s) = s_
1
+ s+ 2 (6.82)
We then obtain X(s) explicitly as
r +s +2
X(s) = -(s---l)(_s_+_l_)(_s_+_2_) (6.83)

To find the inverse transform x(t) = c-i(X(s)], we expand X(s) into partial
fractious
K._i Ki Ks
X(s) =s - 1 +s +1 +s +2 (6.84)

Solving for K._i, Ki, and K1 algebraically, we obtain

K._i =i Ki = -1 Ka =t
The final solution is tht> inverse transform of X(s) or
x(t) = fe 1 - e-1 + te"-11 (6.85)
In order to compare the Laplace transform method to the classical method of
solving differential equations, the reader is referred to the example in Chapter 4,
Eq. 4.64, where the differential equation in Eq. 6.80 is solved classically.
Example 6.10. Given the set of simultaneous differential equations
2x'(t) + 4x(t) + y'(t) + 7y(t) = Su(t)
(6.86)
x'(t) + x(t) + y'(t) + 3y(t) = S6(t)
with the initial conditions x(0 - ) = y(0 - ) = 0, let us find x(t) and y(t).
Solution. Transforming the set of equations, we obtain
s
2(s + 2) X(s) + (s + 7) Y(s) = -s
(6.87)
(s + 1) X(s) + (s + 3) Y(s) = S
Solving for X(s) and Y(s) simultaneously, we have
(S/s)A11 + Sliu
X(s) = A
(6.88)
(5/s)Au + SAa
Y(s) = A
 The Laplace transform 147
where '1 is the determinant of the set of equations in Eq. 6.87, and '1" is the
ijth cofactor of '1. More explicitly, X(s) is

1
_ -Sr - 30s + 15 (6.89)
X,s) - s(r + 2s + 5)
Expanding X(s) in partial fractions, we have

K1 K-.f + Ka
X(s) = 7 + r + 2s + S (6.90)

Multiplying both sides of Eq. 6.90 by s and letting s = 0, we find

K1 == s X(s) 1,-o = 3.
K1 and Ka are then obtained from the equation
3 -8s - 36
X(s) - s = r + 2s + S (6.91)

A further simplification occurs by completing the square of the denominator

of X(s), that is,
r
+ 2s + S = (s + 1)1 + 4 (6.92)

As a result of Eq. 6.92, we can rewrite X(s) as

3 8(s + 1) + 14(2)
X(s) = s- (s + 1)2 + (2)9 (6.93)
so that the inverse transform is
:i:(t) = (3 - se-1 cos 2t - 14e-t sin 2t) u(t) (6.94)

In similar fashion, we obtain Y(s) as

1 lls + 17 1 ll(s + 1) + 3(2)

Y(s) == - s + r + 2s + S = - s + (s + 1) 2 + (2)9
(6.95)

The inverse transform is then seen to be

y(t) = ( -1 + 11e-1 cos 2t + 3e-t sin 2t) u(t) (6.96)

This example is also solved by classical methods in Chapter 4, Eq. 4.163.

We note one sharp point of contrast. While we had to find the initial
conditions at t = o+ in order to solve the differential equations directly
in the time domain, the Laplace transform method works directly with
the initial conditions at t = 0-. In addition, we obtain both the comple-
mentary function and the particular integral in a single operation when we
use Laplace transforms. These are the reasons why the Laplace transform
method is so effective in the solution of differential equations.
1'48 Network analysis and synthesis

6.5 PARTIAL-FRACTION EXPANSIONS

As we have seen, the ease with which we use transform methods depends
upon how quickly we are able to obtain the partial-fraction expansion of a
given transform function. In this section we will elaborate on some simple
and effective methods for partial-fraction expansions, and we discuss
procedures for (a) simple roots, (b) complex conjugate roots, and (c)
multiple roots.
It should be recalled that if the degree of the numerator is greater or
equal to the degree of the denominator, we can divide the numerator by
the denominator such that the remainder can be expanded more easily
into partial fractions. Consider the following example:

F(s) = N(s) = s1 +1 3s1 + 3s + 2 (6.97)

D(s) s + 2s + 2
Since the degree of N(s) is greater than the degree of D(s), we divide D(s)
into N(s) to give
s
F(s) = s + 1 - 1 2s (6.98)
s + +2
Here we see the remainder term can be easily expanded into partial
fractions. However, there is no real need at this point because the
denominator s• +,• 2s + 2 cah be written as

s• + 2s + 2 = (s + 1)1 + 1 (6.99)

We can then write F(s) as

F(s) = s+1- (s + l) - 1 (6.100)
(s + 1)1 + 1
so that the inverse transform can be obtained directly from the transform
tables, namely,
C-1 [F(s)] = «5'(t) + «'(t) + e-'(sin t - cost) (6.101)
From this example, we see that intuition and a knowledge of the trans-
form table can often save considerable work. Consider some further
examples in which intuition plays a dominant role.
Example 6.11. Find the partial-fraction expansion of

2s +3 (6.102)
F(s) = (s + l)(s + 2)
 The Laplace transform 149
If we see that F(s) can also be written as
Ffs' _ (s + 1) + (s + 2)
" 1 (s + l)(s + l) (6.103)

then the partial-fraction expansion is trivially

. 1 1
F(s) =s +1 +s +2 (6.104)

Eumple 6..11. Find the partial-fraction expansion of

s+S
F(s) = (s + 2Y, (6.105)

We see that .F(s) can be rewritten as

F(s) == (s + 2) + 3 == _1_ + 3 (6.106)

(s + 2t (s + 2) (s + 2)1
Real roots
Now let us discuss some formal methods for partial-fraction expansions.
First we examine a method for simple real roots. Consider the function

F(s) == N(s) (6.107)

(s - s0)(s - s1 )(s - s.)
where so, si, and s 1 are distinct, real roots, and the degree of N(s) < 3.
Expanding .F(s) we have

F(s) == _!.L + _&_ + _..!!._ (6.108)

S-So S-S1 S-S1

Let us first obtain the constant Ko- We proceed by multiplying both sides
of the equation by (a - so) to give

(s - so)F(s) == Ka + (s - so)K1 + (s - so)~• . (6.109)

S-Si S-S1

Ifwe let a=="• in Eq. 6.109, we obtain

X. = (s - so) .F(s) I-•• (6.110)
Similarly, we see that the other constants can be evaluated through the
general relation
~ == (s - sJ .F(s) I-•, (6.111)
Ev.,.. ~13. Let us find the partial-fraction expansion for
al + 2s - 2 Ko K1 Ks
F(s) - a(_s + 2Xs - 3) - s +s +2 +s - 3
(6.112)
ISO Network analysis and synthesis
Using Eq. 6.111, we find
K0 = sF(s)l-
sl + 2s -2
= (s + 2Xs - 3)
I
- =3
I

Ki =
r+2s-21 s(s -
3) •- = -
I
S
(6,113)
2

K, = al + 2s - 21 = 13
2
+ 2) -a
s(s 1S
Complex roots
Equation 6.111 is also applicable to a function with complex roots in
its denominator. Suppose F(s) is given by

F(s) = N(s)
Di(s)(s - <X - jp)(s - <X + jp)
= K1 + K2 + Ni(s) (6.114)
s - <X - jp s - <X + jp Di(s)
where N1 /D1 is the remainder term. Using Eq. 6.111, we have

K - N(<X + jp)
2jpDi(<X + jp)
1
-
(6.115)
K - N(<X -jp)
I - -2jPD1(<X - jp)

where we assume that s = <X ± iP are not zeros of D1{s).

It can be shown that the constants K1 and Ka associated with conjugate
roots are themselves conjugate. Therefore, if we denote K1 as
K1 =A+ jB (6,116)

then K,.=A-jB=K1* (6.117)

If we denote the inverse transform of the complex conjugate terms as

/i(t), we see that
fi(t) = c-ir_ K1 . + K1* . ]
[.; - (X - JP s- (X + JP
= e'1(K1e1' ' + K 1*e- 1'')
= 2t!"(A cos {)t - B sin Pt) (6.118)
 The Laplace transform 151
A more convenient way to express the inverse transform fi(t) is to intro-
duce the variables M and Cl> defined by the equations
Msinel> = 2A
(6.119)
M cos Cl> = -2B
where A and Bare the real and imaginary parts o.f K1 in Eq. 6.116. In
terms of Mand Cl>, the inverse transform is
/ 1 (t) = Me"-t sin ({Jt + Cl>) (6.120)
To obtain Mand Cl> from Ki, we note that
Me 1t1> = M cos Cl> + jM sin Cl> = -2B + j2A = 2jK1 (6.121)
When related to the original function F(s), we see from Eq. 6.115 that

Me 1t1> = N(a. + j/J) (6.122)

{JDi(a. + j/J)
Example 6.14. Let us find the inverse transform of
s8 +3 (6.123)
F(s) = (s8 + 2s + SXs + 2)

For the simple roots= -2, the constant Kis

K = (s + 2)F(s)j,_1 = t (6.124)

For the complex conjugate roots

s8 + 2s + S = (s + 1 + j2Xs + 1 - j2) (6.125)

We see that rx = -1, {J = 2, thus

Me'tl> s8+31
= ----,--- = -2 e-i<~- 1 1+r/1) (6.126)
2(s + 2) •-IHI VS
The inverse transform is then

t-1[F(s)] = ~ e-st + ~
S vs e-t sin (21 - ~2 - tan-1 2)
(6.127)
7 2
=5 e-st - vs e- cos (21 -
1 1
tan- 2)

Multiple roots
Next let us consider the case in which the partial fraction involves
repeated or multiple roots. We will examine two methods. The first
requires differentiation; the second does not.
152 Network analysis and synthesis
Method A
Suppose we are given the function
F(s) = N(s) (6.128)
(s - s0)n D1{s)
with multiple roots of degree n at s = s0 • The partial fraction expansion
of F(s) is

F(s) = Ko + K1 + Ka + ... + Kn-i + Ni(s)

(s - So)n (s - So)''-1 (s - So)-1 S - So D1(S)
(6.129)
where N1 (s)/D1 (s) represents the remaining terms of the expansion. The
problem is to obtain Ko, K1, ••• , x_1• For the Ko term, we use the method
cited earlier for simple roots, that is,

Ko = (s - s0 t F(s) I-•• (6.130)

However, if we were to use the same formula to obtain the factors Ki,
K1, ••• , K_i, we would invariably arrive at the indeterminate 0/0 con-
dition. Instead, let us multiply both sides of Eq. 6.129 by (s - s0}n and
define
(6.131)
Thus

F1(s) = Ko + K1(s - s0) + · · · + Kn-1(s - s0),._1 + R(s}(s - StJn

(6.132)

where R(s) indicates the remaining terms. Ifwe differentiate Eq. 6.132 by
s, we obtain
d
ds F1(s) = K1 + 2K.(s - s0 ) + · · · + Kn_1(n - l)(s - s0),._
1
+ ···
(6.133)

It is evident that K1 =~Fi(s)I {6.134)

ds _,.

On the same basis K ~! d8 F{s)I

1 (6.135)
I 2 di - ..
and in general

K , =1.,
-d dF
-
J. s
1
, i(s) I_,.,.
j = 0, 1, 2, ... , n - 1 (6.136)
 The Laplace transform 153
Eumple 6.15. Consider the function
s-2
F(s) == s(s + 1)8 (U37)

which we represen~in expanded form as

Ko K1 X. A
(6.138)
F(s) ""'(s + 1)8 + (s + 1)1 + s + 1 +-;
The constant A for the simple root at s = 0 is
A = s F(s) ls-o = -2 (6.139)
To obtajp the constants for the multiple roots we first find F 1(s).
s-2
F 1(s) = (s + 1)8 F(s) = -s- (6.140)

Using the general formula for the multiple root expansion. we obtain

Ko=,-1d (s -2) I
O.dt>
-s-
0

-1
=3 · (6.141)

K1 · 1 d(s-2)1
=,- -- 21s' =- =2 (6.142)
1.ds s -1 -1

Ka == ~2.t1sd (:.)
r I-1
= ( - s1
:.) I-
-1
2 (6.143)

so that F(s)
3 2
= (s + 1)8 + (s + 1)1 + s + 1 -
2 2s (6.144)

Method 8
The second method for arriving at the partial-fraction expansion for
multiple roots requires no differentiation. It involves a modified power
series expansion.6 Let us consider F(s) and F1(s), defined in Eqs. 6.128
and 6.131 in Method A. We define a new variable p such that p 8 - &0 • =
Then we can write F1(s) as
F 1(p + s0) = N(p + So) (6.14S)
D1(p +s 0)

Dividing N(p + so) by D1(p +s 0), with both polynomials written in

ascending powers of p, we obtain
K •
F1(p + s0) = K0 + KtP + KBP1 + · · · + K-tP-1 + .p
D1(p + So)
. (6.146)
1 I am indebted to the late Professor Leonard O. Goldstone of the Polytechnic

Institute of Brooklyn for showing me this method.

I54 Network analysis and synthesis
The original function F(s) is related \o Fi(p +s 0) by the equation

F(p + so) = Fi(p + So) = Ko + K1 + ... + Kn-I + Kn

p" p" P-1 p Di(p + so)
(6.147)
Substituting s - s0 = p in Eq. 6.147, we obtain
F(s) = Ko + K1 + ... + Kn-1 + Kn (6.148)
(s - Sor (s - So)n-l S - So Di(s)
We have thus found the partial-fraction expansion for the multiple-root
terms. The remaining terms K,./[D1(s)] still must be expanded into partial
fractions. Consider Example 6.16.
Example 6.16
2
F(s) = (s + 1)3(s + 2) (6.149)

Using the method just given, F 1(s) is

. 2
F 1(s) = (s + 1)3 F(s) = s + 2 (6.150)
Settingp =s + 1, we then have
2
F 1(p -1) = p-+-1 (6.151)

The expansion of Fi(p - 1) into a series as given in Eq. 6.146 requires dividing
the numerator 2 by the denominator 1 + p, with both numerator and denomi-
nator arranged in ascending power of p. The division here is
2 -2p +2p1
1 +p)2
2 +2p
-2p
-2p-2p8
2p2
2p8 + 2p3
-2p3
Since the multiplicity of the root is N = 3, we terminate the division after
we have three terms in the quotient. We then have

F1(p - 1) = 2 - 2p + 2p8 - p
2
l -r
(6.152)
The original function F(p - 1) is
F (p - 1) 2 2 2 2
F(p - 1) = --------=---
1
p3
=- - -
p3 p"
+ - - --
p p+l
(6.153)
 The Laplace transform I55
Su~tituting s +l == p, we have
2 2 2 2
F(s) = (s + 1)3 - (s + 1)1 + (s + 1) - s +2 <6· 154)
Example 6.17. As a second example, consider the function
s+2
F(s) = ~s + l)9 (6.155)

Since we have two sets of multiple roots here (at s = 0, and s = -1) we have a
choice of expanding F 1(s) about s = 0 or s = -1. Let us arbitrarily choose to
expand abouts = O; since p = s here, we do not have to make any substitutions.
F 1(s) is then s +2 2 s +
F 1(s) = (s + l)9 = 1 + 2s + s9 (6.156)
Expanding F 1(s), we obtain
~3s +4)
F 1(s) =2 - 3s + (s + l)I (6.157)

2 3 3s +4
F(s) is then F(s) = ~ - ; + (s + 1)9 (6.158)

We must now repeat this process for the term

3s +4
(s + 1)1
Fortunately we see that the term can be written as
3s + 4 3(s + 1) + 1 3 1
- - = - -(s-+=
(s + 1)1
--=
1)9
--+
s+l
--
(s + 1)1
(6.159)

The final answer is then

2 3 3 1
F(s) = ~ - ; + s + l + (s + 1)9 (6.160)

6.6 POLES AND ZEROS

In this section we will discuss the many implications of a pole-zero

description of a given rational function with real coefficients F(s). We
define the poles of F(s) to be the roots of the denominator of F(s). The
zeros of F(s) are defined as the roots of the numerator. In the complex s
plane, a pole is denoted by a small cross, and a zero by a small circle. Thus,
for the function
F( ) s(s - 1 + jl)(s - 1 - jl)
(6.161)
s == (s + l}'(s + j2)(s - j2)
the poles are at s == - I (double)
s == -j2
s == +j2
 156 Network analysis and synthesis

j"' I
j2 x • plane

~b:e po~~J-'1---<J--~o-.__.....___.,
-u -3 -2 -1 0 1 2 3 (T

-jl 0

-j2 X

-j"' I
FIG. 6.7. Pole-:zero diagram of F(s).

and the zeros are at s=O

s = 1 + jl
s = 1 -jl

S = 00

The poles and zeros of F(s) are shown in Fig. 6.7.

Now let us consider some pole-zero diagrams corresponding to standard
signals. For example, the unit step function is given in the complex
· frequency domain as ·

C[u(t)] = -1 (6.162)
s
and has a pole at the origin, as shown in Fig. 6.8a. The exponential
signal e-•ot, where o-0 > 0, has a transform

qe-°'] = _1_ (6.163)

s + O'o
which has a single pole at s = -ao, as indicated in Fig. 6.8b. The cosine
function cos ro 0t~ whose transform is
s
C[cos ro0 t] = (6.164)
s 1
+ roo1
has a zero at the origin and a pair of conjugate poles at s = ±jro0,
as depicted in Fig. 6.8c. Figure 6.8d shows the pole-zero diagram
corresponding to a damped cosine wave, whose transform is

(6.165)
 The Laplace transform 157

Ip F(•:=¾ . --.--¾
--x-- -x
-f.?
+- . . _: ;., :-. . . . . .
1 (a)

"
Ip F(•J==-;rb ft,, F() •+",

X -JWO X -iwo
I (c) (d)

FIG. 6.1. Poles and uros of various functions.

From these four pole-zero diagrams,. we note that the poles corre-
sponding to decaying exponential waves are on the -a axis and have zero
imaginary parts. The poles and zeros corresponding to undamped
sinusoids are on the jro axis, and have zero real parts. Consequently, the
poles and zeros for damped sinusoids must have real and imaginary
parts that are both nonzero.
Nowletusconsidertwoexponential waves/1 (t) - e-111' andfs(t) == r•',
where a 1 > a 1 > 0, so that/.(t) decays faster thanfi(t), as shown in Fig.
6.9a. The transforms of the two functions are
1
F 1(s)==--
s+ U1
(6.166)
1
F.(s)=--
s + a1
as depicted by the pole-zero diagram in Fig. 6.9b. Note that the further
the pole is from the origin on the -a axis, the more rapid the exponential
decay. Now consider- two sine waves, sin ro1 t and sin ro1 t, where ro1 >
ro1 > 0. Their corresponding poles are shown in Fig. 6.10. We note here
f(t)
jw
• plane

Fz{•~, Fi(•)'"'\
t -0'2 -0'1 0 O'

(a) (b)

FIG. 6.9. Eft'ect of pole location upon exponential decay.

I58 Network analysis and synthesis

I~"'
r"'2
XJWl
X
s2
jc.,

(J'
0 (J'

FIG. 6.10. Pole locations corre- FIG. 6.11

sponding to sin <»st and sin w 1 t.

that the distance from the origin on the jro axis represents frequency of
oscillation; the greater the distance, the higher the frequency.
Using these rules of thumb, let us compare the time responses / 1(t) and
J.(t} corresponding to the pole pairs {si, s 1 *} and {s1 , s2 *} shown in Fig.
6.11. We see that both pairs of poles corresponds to damped sinusoids.
The damped sinusoid/1(t) has a smaller frequency of oscillation than/1(t)
because the imaginary part of s1 is less than the imaginary part of s1 • Also,
/ 2(t) decays more rapidly than / 1(t} because Re s1 > Re s2 • The time
responses / 1 ( t) and / 1( t) are shown in Fig. 6.12.
(2

(a)

,,,- --- (b)

FIG, 6.12. Time responses for poles in Fig. 6.11.

 The Laplace transform 159

j"'
s plane
X

<Ti (T

(a) (b)
FIG. 6.13. Effect of right-half plane poles upon time response.

Let us examine more closely the effect of the positions of the poles in
the complex frequency plane upon transient response. We denote a
pole A as a complex number A= <1i + jw,. For a given function F(s)
with only first-order poles, consider the partial-fraction expansion

F(s) = -Ko- + -Ki- + · · · + -Kn- (6.167)

s - Po s - Pi s - Pn
The inverse transform of F(s) is
f(t) = Koe'•'+ Kie"1' + ... + Kne,,.,
(6.168)
= Koe"•'el<»ot + Kie"1'ei<»1t + ... + Kne"•'el<»,.t
ln/(t), we see that if the real part of a pole is positive, that is, <1, > 0, then
the corresponding term in the partial-fraction expansion

is an exponentially increasing sinusoid, as shown in Fig. 6.13a. We thus

see that poles in the right half of the s plane (Fig. 6.13b) give rise to
exponentially increasing transient responses. A system function that has
poles in the right-half plane is, therefore, unstable. Another unstable
situation arises if there is a pair of double poles on the jw axis, such as for
the function
F(s) - s (6.169)
- (s• + wo•)a
whose pole-zero diagram is shown in Fig. 6.14a. The inverse transform of
F(s) is
J(t) = _t_ sin w t 0 (6.170)
2w0
160 Network analysis and synthesis
f(t)

tT

•-iwo
I
(a) (b)
FIG. 6.14. Effect of double zeros on the jw axis upon time response.

which is shown in Fig. 6.14b. It is apparent that a stable system function

cannot also have multiple poles on the jw axis.
Consider the system function H(s) == N(s)/ D(s). If we factor the
numerator and denominator polynomials, we obtain
H 0(s - z.)(s - Zi) • • • (s - z )
H(s) == " (6.171)
(s - Po)(s - Pi) • · · (s - p,J
It is clear that H(s) is completely specified in terms of its poles and zeros
and an additional constant multiplier H 0• From the pole-zero plot of H(s),
we can obtain a substantial amount of information concerning the
behavior of the system. As we have seen, we c.an determine whether the
system is stable by checking the right-half plane for poles and the jw axis
for multiple poles. We obtain information concerning its transient
behavior from the positions of its poles and zeros; and, as we will discuss
in Section 8.1, the pole-zero diagram also gives us significant.information
concerning its steady-state Uw) amplitude and phase response. We thus
see the importance of a pole-zero description.
Suppose we are given the poles and zeros of an excitation E(s) and the
system function H(s). It is clear that the pole-zero diagram of the response
function R(s) is the superposition of the pole-zero diagrams of H(s) and
E(s). Consider the system function
H(s) == Ho(s - Zo) (6.172)
(s - p0)(s - pJ
and the excitation E(s) = Eo(s - zJ (6.173)
s - Pa
 The Laplace transform 161
Then the response function is
R(s) = HoEo(s - Zo)(s - Zt) ( 6 .174)
(s - p 0)(s - pJ(s - p.)
It is clear that R(s) contains the poles and zeros of both H(s) and E(s),
except in the case where a pole-zero cancellation occurs. As an example,
let us take the system function
H(s) = · l(s + 1) (6.175)
(s + 2 + j4)(s + 2 - j4)
whose pole-zero diagram is shown in Fig. 6.15a. It is readily seen from the
pole-zero diagrams of the excitation signals in Fig. 6.18a that the step
response of the system has the pole-zero diagram indicated in Fig. 6.1 Sb.
The response to an excitation signal 3 cos 2t has the pole-zero representa-
tion of Fig. 6.15c. To specify completely the given function F(s) on a pole-
zero plot, we indicate the constant multiplier of the numerator on the
plot itself.
Let us examine the significance of a pole or zero at the origin. _We
know that dividing a given function H(s) bys corresponds to integrating
the inverse transform h(t) = c-1 (H(s)J. Since the division bys corresponds
to a pole at the origin, we see that a pole at the origin implies an inte-
gration in the time domain. Because the inverse transform of H(s) in
Fig. 6.15a is the impulse response, placing a pole at the origin ·must
correspond to the step response of the system. In similar manner, we
deduce that a zero at the origin corresponds to a differentiation in the
time domain. Suppose we consider the pole-zero diagram in Fig. 6.1 Sc
with the zero at the origin removed; then the resulting pole-zero plot is
the response of the system to an excitation E0 sin 2t. Placing the zero at
the origin must then give the response to an excitation E 1 cos 2t, which, of

X j4
·w

K•2
X

r r ·4
J K•2
X j4
K•&

-2 -

X -j4
(T

-~-1 =~
X -j4
(T

X
I •
-2 -1
/2! (T

(a) (b) (c)

FIG. 6.15. (a) System function. (b) Response to unit step excitation. (c) Response to
excitation e(t) =- 3 cos 2t.
162 Network analysis and synthesis
course, is true. We therefore conclude that the system function of a
differentiator must have a zero at the origin; that of an integrator must
have a pole at the origin.

6.7 EVALUATION OF RESIDUES

Poles and zeros give a powerful graphical description of the behavior

of a system. We have seen that the poles are the complex frequencies of
the associated time responses. What role do the zeros play? To answer
this question, consider the partial fraction expansion
N K
F{s)=I-'- (6.176)
, s - s,
where we shall assume that F(s) has only simple poles and no poles at
S= 00,
The inverse transform is
(6.177)

It is clear that the time response f(t) not only depends on the complex
frequencies s, but also on the constant multipliers K,. These constants K,
are called residues when they are associated with first-order poles. We will
show that the zeros as well as the poles play an important part in the
determination of the residues K,.
Earlier we discussed a number of different methods for obtaining
the residues by partial-fraction expansion. Now we consider a graphical
method whereby the residues are obtained directly from a pole-zero
diagram. Suppose we are given

F(s) = A 0(s - z0 )(s - Zi) • • • (s - z.,) (6.178)

(s - Po)(s - Pi) · · · (s - Pm)
where m > n and all the poles are simple. Let us expand F(s) as

F(s) = -Ko- + -K1- + .. · + -Km- {6.179)

s - Po s - P1 s - Pm
Our task is to determine the residues Kt- We know that
_ (
K , - s - P·
) F( )
•
s
I•-P _
-
Ao(A - Zo)(p, - Zi) · · · (p, - z.,)
(p, - Po) · · • (p, - P,-J(p, - P,+1) · · · (p, - Pm)
{6.180)
 The Laplace transform 163
When we interpret the Eq. 6.180 from a jw
complex-plane viewpoint, we see that each one
ozo
PlX
oftheterms(p, - z1) represents a vector drawn
from a zero z1 to the pole in question, p,.
Similarly, the terms (p, - p.,J, where i -:;I= k, -x----+----
Po
represent vectors from the other poles to the
pole p,. In other words, the residue K, of any
pole p, is equal to the ratio of the product of
X 0
the vectors from the zeros to p,, to the prod- Pl* z1
uct of the vectors from the-other poles top,.
FIG. 6.16. Poles and zeros
To illustrate this idea, let us consider the pole- of F(s).
zero plot of
F(s) = Ao(s - Zo)(s - zJ (6.181)
(s - Po)(s - P1)(s - P1*)
given in Fig. 6.16. The partial-fraction expansion of F(s) is

F ( s ) = ~ + ~ + K1* (6.182)
s-po s-p1 s-p1*
where the asterisk denotes complex conjugate. Let us find the residues K0
and K1 by means of the graphical method described. First we evaluate K1
by drawing vectors from the poles and zeros to Pi, as shown in Fig. 6.17a.
The residue K1 is then
K _ A.,AB (6.183)
i - CD

where symbols in boldface represent vectors. We know that the residue

of the conjugate pole Pi* is simply the conjugate of K1 in Eq. 6.183. Next,

jw

(a) (b)
FIG. 6.17. Determining residues by vector method.
164 Network analysis and synthesis
jw. to evaluate Ko, we draw vectors from the
C•3 poles and zeros to p 0, as indicated in Fig.
6.17b. We see that
X jl
K0 = A.,RL (6.184)
MN
-x With the use of a ruler and a protractor,
-tT-2 -1
we determine .the lengths and the angles of
the vectors so that the residues can be deter-
X -jl mined quickly and easily. Consider the fol-
lowing example. The pole-zero plot of

F(s) = 3s
FIG, 6.18. Pole-:zero diagram of. (s + 2)(s + 1 - jl)(s + 1 + jl)
F(s).
(6.185)
is shown in Fig. 6.18. The partial-fraction expansion of F(s) is

F(s)= Ki + Ki* +_& (6.186)

s + 1 - jl s+1 + jl s+2
First let us evaluate Ki, The phasors from the poles and zeros to the pole
at -1 + jl are sho'Y]l in Fig. 6.19a. We see that Ki is

/2/135° 3
K =3 X -=-:........;==-- = -
i ✓2 / 45° X 2/ 90° 2

jw jw
c-3 C•3

(G) (b)

FIG, 6.19. Evaluation of residues of F(s).

 The Laplace tnll'lsform 165
From Fig. 6.19b we· find the value of the residue X. to be

K1 =_ 3 x2 == -3
,J2/-135° X ,J2/ +135°
Therefore, the partial-fraction expansion of F(s) is

F(s) == f + f __3_ (6.187)

s + 1 - jl ,s + 1 + jl s+2

6.1 THE INITIAL AND FINAL VALUE THEOREMS

In this section we will discuss two very useful theorems of Laplace

transforms. The first is the initial value theorem. It relates the initial
value of /(t) at t = o+ to the limiting value of s }ts) as s approaches
infinity, that is,
lim f(t) == lim s F(s) (6.188)
f-+o+ 1-+00

The only restriction is that /(t) must be continuous or contain, at most, a

step discontinuity at i == 0. In terms of the transform, }ts)== C[/(t)];
this restriction implies that }ts) must be a proper fraction, i.e., the degree
of the denominator polynomial of }ts) must be greater than the degree of
the numerator of }ts). Now consider the proof of the initial value theorem,
which we give in two parts.
(a) The function /(t) is continuous at t = 0, that is, /(0-) == /(0+ ).
From the relationship

C[/'(t)J == I:f'(t)e_., dt = s F(s) -f(0-) (6.189)

we obtain lim C[/'(t)] == lim s F(s) - f(0-) == 0 (6.190)

Therefore lim s F(s) == f(0-) == f(0+) (6.191)

1-+00

(b) The function/(t) has a step discontinuity at t =

0. Let us represent
/(t) in terms of a continuous part / 1(t) and a step discontinuity D u(t),
as shown in Fig. 6.20. We can then write /(t) as
/(t) == fi(t) + D u(t) (6.192)

where D =/(0+) - /(0-). The derivative of/(t) is

/'(t) == /'i(t) + D a(t) (6.193)

166 Network analysis and synthesis
Du(t)
f(t)

D Dr....---

0
= 0 + 0

FIG. 6.l0. Decomposition of a discontinuous function into a continuous function

plus a step function.

Since f 1(t) is continuous at t = 0, we know from part (a) that

lim s F 1(s) = /1(0-) = /(0-) (6.194)
,➔ 00

Taking the Laplace transform of both sides of Eq. 6.193, we have

s F(s) - J(0-) = s F1(s) - fi(0-) + D (6.195)
which simplifies to s F(s) = s F 1(s) + D (6.196)
Now, ifwe take the limit ofEq. 6.196 ass - oo and letf(0+) - f(O-) =
D, we have
lim s F(s) = lim s F 1(s) + f(0+) - /(0-) (6.197)

By Eq. 6.194, we then obtain

lim s F(s) = f(0+) (6.198)

Example 6.18. Given the function

2(s + 1)
F(s) = ii + 2s + S (6.199)

let us find/(0 +). Since the degree of the denominator is greater than the degree
of the numerator of F(s), the initial value theorem applies. Thus
. • 2(s + l)s
hm s F(s) = hm JI 2s S =2 (6.200)
8-+00 8-+00 ;r + +
Since C-1 [F(s)] = 2e-t cos 2t (6.201)
we see that f(O +) = 2.
Example 6.19. Now let us consider a case where the initial value theorem does
not apply. Given the function
/(t) = 6(t) + 3e' (6.202)
we see that /(0 +) = 3. The transform of f(t) is
3
F(s)=l+s+l (6,203)
 The Laplace transform 167

so that funsF(s) = fun s(l +~=co) (6.204)

......., •-00 s +
Next we consider the final value theorem, which states that
lim f(t) == lim s F(s) (6.205)
t ➔ ao a➔O

provided the poles of the denominator of F(s) have negative or .zero real
parts, i.e., the poles of F(s) must not be in the right half of the complex-
frequency plane. The proof is quite simple. First

f :f'(t)e--' dt == s F(s) - f(0-) (6.206)

Taking the limit ass- 0 in the Eq. 6.206, we have

i.,o-
Evaluating the integral, we obtain
f'(t) dt == lim s F(s) -
•➔o
f(O-) (6.207)

f( oo) - f(0-) == lim s F(s) - f(0-) (6.208)

•➔ o

Consequently, f( oo) == 1im s F(s) (6.209)

a➔O

Example 6.20. Given the function

/(t) = 3u(t) + 2e-t (6.210)
which has the transform
3 2 Ss+3
F(s) = s+ s + l = s(s + 1) (6.211)

let us find the final value/( co). Since the poles of F(s) are at s = 0 ands = -l,
we see the final value theorem applies. We find that
funs F(s) =3 (6.212)

which is the final value of/(t) as seen from Eq. 6.210.

Example 6.21. Given = 2e'
/(t) (6.213)
we see that fun /(t) = 00 (6.214)

2s
But from sF(s) =s _
1
(6.215)

we have funs F(s) =0 (6.216)

We see that the final value theorem does not apply in this case, because the pole
s == l is in the right half of the complex-frequency plane.
168 Network analysis and synthesis
TABLE 6.1
Laplace Transforms
/(t) F(s)

1. /(t) F(s) - J:1 (t)e-n dt

2. aif1(t) + a,.f,f.t) a 1 F 1(s) + a 1 F,f.s)

d
3. dtf(t) s F(s) - /(0-)

d" ft
4. dt"f(t) s" F(s) - I sn-lJH(0-)
,-1

1
s. f~ f(-r)d-r -F(s)
s

6. s: f~f(-r)d-rda
1
}F(s)

d"
7. (-t)" /(t) ds" F(s)

8. f(t - a) u(t - a) e-•F(s)

9. tP'f(t) F(s - a)
10. c5(t) 1
d"
11. dt" c5(t) s"
1
12. u(t)
s
1
13. t }
t" 1
14. ,rfl+l
n!
1
15. e-•,
s + C(

1 1
16. - - ( C " ' _ e-1')
fJ - at (s + «Xs + fJ)
Q)
17. sin wt .,. + ml
 The Laplace transform 169
TABLE 6.1 (cont.)
8
18. cos wt
.. + o,I
a
19. sinb at
8'-a'
8
20. coshat
.. -a'
Cl>
21. e-atsin wt
(8 + cxS) + o,I
(8 + at)
22. e-td COS o,t
(8 +at..,+ o,I
e-.,,- 1
23. -;;r- (8 + at}tt+1
t 8
24. -sin"''
2w (s' + o,1)1
1 1
25. -;J,.(att); n - 0, 1, 2, 3, ••.
at (s' + at'}'-'[(.,S + cxS)U - s]-
(Bessel function of first kind,
nth order)
26. (trtJ'-' .r'-'
r(k + 1)
27. t" (k need not be an integer) ;+"1

Problems
6.1 Find the Laplace transforms of
(a) /(t) - sin (act + {J)
(b) f(t) ... rlt-t«l cos (wt + {J)
(c) /(t) - (,a + l)r''
(d) /(t) - K.'; K is a real constant > 1.
(e) /(t) - (ti + ac~'
(/) /(t) - I -''(I)
(g) /(t) "" u[t(t' - 4))
6.2 Find the Laplace transforms of
(a) /(t) - cos (t - tr/4) ll(_t - tr/4)
(b) /(t) - cos (t - tr/4) u(t)
170 Network analysis and synthesis
6.3 Find the Laplace transforms for the waveforms shown.
f(t) f(t)
2 1

0 2 4 t 0 1 2 3 t

-1

(a) (b)

B(t) B(t)
1 E

2r t 0 T t

(c) (d)

T t

(e)
PROB.6.3

6.4 Find the Laplace transforms for the derivatives of the waveforms in
Prob. 6.3.
6.5 Find the inverse transforms for
2s +9
(a) F(s) = (s + 3)(s + 4)
Ss - 12
(b) F(s) = s1 + 4s + 13

(c) F(s) = + 13
4s
s1 +4s-S
Note that all the inverse transforms may be obtained without resorting to
normal procedures for partial-fraction expansions.
 The Laplace transform 171
6.6 Find the Laplace transform of the waveforms shown. Use (a) the infinite
series method and (b) the Laplace transform formuJa for periodic waveforms.
Both answen should be in closed form and agree.
f(t)
f(t)
A

0 T 2T 3T

-A

(a) (b)

a(t)
+1

-1
0 r 2r 3rt
(c) PROB. 6.6 (d)

6.7 Evaluate the definite integrals

(a) L. t sin 2t dt

(b) L. ,.,. cos 3t dt

6.8 Solve the following differential equations using Laplace transforms

(a) :r:'(t) + 6:1:'(t) + 9:r:(t) = cos 2t
(b) :r:'(t) + 3:r:'(t) + 2:r:(t) = «5(t)
(c) :r:'(t) + 2:r:'(t) + 5:r:(t) = u(t)
(d) z'(t) + 5:r:'(t) + 4a(t) -= (e- 1 + r'') u(t)
(e) z'(t) + 2:r:'(t) = u(t) + e- 1 u(t)
It is given that :r:(0-) = :r:'(0-) = 0 for all the equations.
6.9 Using Laplace transforms, solve the following sets of simultaneous
equations:
(a) 2:r:'(t) + 4a(t) + y'(t) + 2y(t) = «5(t)
:r:'(t) + 3:r:(t) + y'(t) + 2y(t) = 0
(b) 2:r:'(t) + :r:(t) + y'(t) + 4y(t) = 2e1
:r:'(t) + y'(t) + 8y(t) = «5(t)
The initial conditions are all 2'.CI'o at t = 0-.
1n Network analysis and synthesis
6.10 Find the inverse transforms for
s
(0)
F(s) - (s' + 9)(s + 3)

(b) F() s+«

s -s-+-
{J
(c) F(s) =- (s ; 1)'

s+l
(d) F(s) ... s' + 2s + 2
s'+3s+1
(e) F(s)-
s8 + s
s+S
(f) F(s) .. (s + l)l(s + 2)1
s8
(g) F(s) - (s + 1)(s + 2)'
3s'-s9-3s+2
(h) F(s) -
sl(s - 1)1
6.11 Find the inverse transforms for
1 +cl•
(a) F(s) -= sl(s + 2)
se-1• - e 11 •
1
b) Fi(s) ... - --
" s1 +3s+2
se-«•
(c) F(s) '"'" s + {J
6.12 Given the pole-zero diagrams shown in (a) and (b) of the figure, write
the rational functions F 1(s) and F,l..s) as quotients of polynomials. What can
you say about the relationship between the poles in the right-half plane and the
signs of the coefficients of the denominator polynomials?

jw F1(•J jw F2(•J
X

i jl
j0.6

(I
jl
j0.6

l-j0.6
+l
)(

'1

X
7-jl

I (G) (b)
PROB.6.12
 The Laplace transform 173
6.l3 For the pole-7.el'O plots in Prob. 6.12, find the residues of the poles by
the vector method. .
6.14 For the pole-7.ero plots shown in the figure, find the residues of the
poles.

j2r"'
jlf
-·---+--o-~:,.--
-2 -1 1 2 ,r
-ilf tT

-i2r
(a) (b)

lj"'
r-7jl
---X--O---->t---
-~,.._ I_,,
-1 0 +l

(c) (d)
PROL6.14
6.15 Two response transforms, R1(s) and R,/..s), have the pole-7.ero plots
shown in (a) and (b) of the figures, respectively. In addition, it is known that
ri(O +) - 4 and r,.(t) - 4 for very large t. Find r 1(t) and r,.(t).
jw

-x J
-• -2 -1 0
1(----- _,, -1,
~-- -jl
I
0

(a) (b)
PROL6.15
174 Network analysis and synthesis
6.16 It is given that
Ff)= ar+bs+c
,s s8+2sl+s+l
Find a, b, and c such that /(0 +) =f'(O +) = f"(O +) = 1.
6.17 Using the initial and final value theorems where they apply, find/(0+)
and/( oo) for

(a) F(s) = (s + l)(s + 2)

(s + 3)(s + 4)
s
(b) F(s) = (s + l)(s - 1)
s1 +3s+2
(c) F(s) = s8 + 3s2 + 3s + 1

(d) F(s) = s(s + 4)(s + 8)

(s + l)(s + 6)
 chapter 7
Transform methods in
network analysis

7.1 THE TRANSFORMED Cl~CUIT

In Chapter 5 we discussed the voltage-current relationships of network

elements in the time domain. These basic relationships may also be
represented in the complex-frequency domain. Ideal energy sources, for
example, which were given in time domain as v(t) and i(t), may now be
represented by their transforms V(s) = C[v(t)] and /(s) = C[i(t)]. The
resistor, defined by the v-i relationship
v(t) = R i(t) (7.1)
is defined in the frequency domain by the transform of Eq. 7.1, or
V(s) = R l(s) {7.2)
For an inductor, the defining v-i relationships are
di
v(t) = L dt
(7.3)
i(t) =L!.f.'
o-
v(-r) d-r + i(O-)
Transforming both equations, we obtain
V(s) = sLI(s) - L i(O;.....)
(7.4)
I(s) == !_ V(s) + i(O-)
sL s
175
 176 Network analysis and synthesis
l(s)

+ l(s) +

V(s) v,.,
Li(O-)

FIG. 7.1. Inductor.

The transformed circuit representation for an inductor is depicted in

Fig. 7.1. For a capacitor, the defining equations are

v(t) = -1
C o-
i' i(-r) d-r + v(O-)
(7.5)
i(t) = Cdv
. dt
The frequency domain counterparts of these equations are then

V(s) = ..!.1(s) + v(O-)

sC s (7.6)
l(s) = sC V(s) - C v(O-)
as depicted in Fig. 7.2.
· From this analysis we see that in the complex-frequency representation
the network elements can be represented as impedances and admittances in
series or parallel with energy sources. For example, from Eq. 7.4, we see
that the complex-frequency impedance representation of an inductor is sL,
and its associated admittance is 1/sL. Similarly, the impedance of a
capacitor is 1/sC and its admittance is sC. This fact is very useful in circuit
analysis. Working from a transformed circuit diagram, we can write
mesh and node equations on an impedance or admittance basis directly.

1 ..1.. l(sJ
o+ -+1,.,
v,., .c t v,.,
-.-
r.(0-)
C11(0-)

~
FIG. 7.2. Capacitor.
 Transform methods In network analysis 177
iB(t)
+ +
•

FIG.7.3

The process of solving network differential equations with the use of

transform methods has been given in Chapter 6. To analyze the circuit
on a transform basis, the only additional step required is to represent all
the network elements in terms of complex impedances or admittances with
associated initial energy sources.
Consider the example of the transformer in Fig. 7.3. If we write the
defining equations of the transformer directly in the time domain, we have

vi(t) = Li di1 + Mdi1

dt dt
(7.7)
vi(t) = M di + La di
1 1
dt dt
Transforming this set of equations, we obtain
Vi(s) = s£i / 1(s) - Li i1(0-) + sM Jt,s) - M i.(0-) (7.8)
V1(s) = sM 11(s) - M i1(0-) + sLa 1.(s) - La is(O-)
This set of transform equations could also hav~ been obtained by repre-
senting the circuit in Fig. 7.3 by its transformed equivalent given in Fig. 7.4.
In general, the use of transformed equivalent circuits is considered an
easier way to solve the problem.
I.!i2(0-)
+
+ +

MiB(O-)

FIG. 7.4
178 Network analysis and synthesis

Ii'
Li L-J

.- .-
'£1(0-) 'Lz{O-)

~ ~
v-=-
111(t) + R
- v~(O-)

FIG. 7.5

Example 7.1. In Fig. 7.5, the switch is thrown from position 1 to 2 at t = 0.

Assuming there is no coupling between Li and Lt, let us write the mesh equations
from the transformed equivalent circuit in Fig. 7.6. The mesh equations are

V1(s) + Li iL1(0-) - vo(~-) = (sLi + s~) / 1(s) - s~l,l..s)

(7.9)
vo<~-) + La iL1(0-) == - s~/1(s) + (s~ + sLa + R) 1,1..s)
sL1 "'2

---~
11c<O-)

FIG. 7.6

Example 7.2. In Fig. 7.7, the switch is thrown from position 1 to 2 at time
t = 0. Just before the switch is thrown, the initial conditions are iL(0-) = 2
amp, vo(0-) = 2 v. Let us find the current i(t) after the switching action.
s

-
1 30 lh

+I
_v(t)

5.=.vi(t) ~ i'
i £(0-)

Ve
(0
-)
+_½f

FIG. 7.7
 Transform methods In network analysis 179
2

T¾ , J·
2

FIG. 7.8
Since the switch is closed at t = 0, we can regard the S-v battery as an equiva-
lent transformed source S/s. The circuit is now redrawn in Fig. 7.8 as a trans-
formed circuit. The mesh equation for the circuit in Fig. 7.8 is

s
2
-S - - + 2
s
(
= 3+ s + -
s
2) I(s) (7.10)
Solving for I(s), we have
2s+3 1 1
(7.11)
I(s) - (s + l)(s + 2) =s + l + s + 2
Therefore, i(t) = C--1 [/(s)J = e-• + e-SI (7.12)
Example 7.3. Consider the network in Fig. 7.9. At t = 0, the switch is opened.
Let us find the node voltages v1(t) and v.<_t) for the circuit. It is given that
L=lh C=lf
G = lmho V= Iv
111

[
v-=- C
+11
0 (0-) G

FIG. 7.9
Before we substitute. element values, let us write the node equations for the
transformed circuit in Fig. 7.10. These are
Node V1 :
iL(0.J)
-- s
- + Cvc(0-) = ( sC +-1)
sL
1
V1(s) - - V.<_s)
sL
(7.13)
iL(O-) 1 ( 1 )
- - '"" - - Vi(.r)
s sL
+ -sL +G V,l..s)
180 Network analysis and synthesis

tCvc(O-) G

FIG. 7.10

If we assume that prior to the switch opening the circuit had been in steady
state, then we have vo(O-) == 1 v, iL(O-) = 1 amp. Substituting numerical
values into the set of node equations, we have

1- -
s
1= ( + -2) 2
s
s
V1(s) - - V.(s)
s
(7.14)
--1
s
= - 2
-s + (2-s + )
V1(s) 1 V,l..s)

Simplifying these equations, we obtain

s - 1 == (r + 2)V1(s) - 2V.(s)
(7.15)
1 .. -2V1(s) + (s + 2)V,l..s)
Solving these equations simultaneously, we have
Y,(~ s+l s+l
1 s,_,,. s8 + 2s + 2 ... (s + 1)' + 1
(7.16)
v.o s+2 s+2
...,s, == s8 + 2s + 2 = (s + 1)' + 1
so that the inverse transforms are
v,l..t) .. r'(cos t + sin t) (7.17)

7.2 THEVENIN'S AND NORTON'S THEOREMS

In network analysis, the objective of a problem is often to determine a

single branch current through a given element or a single voltage across
an element. In problems of this kind, it is generally not practicable to
write a complete set of mesh or node equations and to solve a system of
equations for this one current or voltage. It is then convenient to use two
very important theorems on equivalent circuits, known as Thlvenin's
and Norton's theorems.
 Transform methods In network analysis 181

NetworkN
n voltage sources
m cunent sources

FIG. 7,11

Th6venin's theorem
From the standpoint of determining the current J(s) through an element
of impedance Zi,(s), shown in Fig. 7.11, the rest of the network N can be
replaced by an equivalent impedance z.(s) in series with an equivalent
voltage source v.(s), as depicted in Fig. 7.12. The equivalent impedance
z.(s) is the impedance "looking into" N from the terminals of Z 1(s) when
all voltage sources in N are short circuited and all current sources are
open circuited. The equivalent voltage source V.(s) is the voltage which
appears between the terminals 1 and 2 in Fig. 7.11, when the elementZ1(s)
is removed or open circuited. The only requirement for Thevenin's
theorem is that the elements in Z 1 must not be magnetically coupled to any
element in N.
The proof follows. The network in Fig. 7.11 contains n voltage and m
current sources. We are to find the current J(s) through an element that
is not magnetically coupled to the rest of the circuit, and whose impedance
is Z 1(s). According to the compensation theorem,1 we can replace Z 1(s)
by a voltage source V(s), as shown in Fig. 7.13. Then by the superposition
principle, we can think of the current J(s) as the sum of two separate parts
· l(s) == Ji(,) + J.(_s) (7.18)

Let the current / 1(s) be the current due to the n voltage and m current

+ 1
V.(•J I
I
I
Z.(•J I Zi(•J

~ I

FIG. 7.11. Thcvenin's equivalent cin:uit.

------
See H. H. Skilling, Electrical &clneering Circllits, 2nd Ed. John Wiley and Sons,
1
New York, 196S.
182 Network analysis and. synthesis
sources alone; i.e., we short ci,-cuit the source V(s), as shown in Fig. 7.14a.
Therefore l,.(s) is equal to the short-circuit current / 80 • Let / 1(s) be the
current due to the voltage source V(s) alone, with the rest of the voltage
sources short circuited and current sources open circuited (Fig. 7.14b).
With the m current and n voltage
sources removed in Fig. 7.14b, we see
Network that the network N is passive so that
N V(s) 19(s) is related to the source V(s) by the
n voltage sources
m current sources relation

FIG. 7.13 ls(s) =- V(s) (7.19)

Z1n(s)
where Z1n(s) is the input impedance of the circuit at the terminals of the
source V(s). We can now write l(s) in Eq. 7.18 as

l(s) = lsc(s) - V(s) (7.20)

Z1n(s)
Since Eq. 7.20 must be satisfied in all cases, consider the particular case
when we open circuit the branch containing Z 1(s). Then l(s) = 0 and V(s)
is the open-circuit voltage Voc(s). From Eq. 7.20 we have
_ Yoc(s)
I scS
() - - - (7.21)
Z1n(s)

1,.
FIG. 7,l<lcr
 Transform methods In network analysis 183

FIG. 7.14b

so that we can rewrite Eq. 7.20 as

Z1n(s)/(s) - Voc(s) = -V(s) (7.22)
In order to obtain the current l(s) through Z 1(s), the rest of the network
N can be replaced by an equivalent source v.(s) = V oc(s) in series,
with an equivalent impedance z.(s) = Z1n(s), as shown in Fig. 7.12.
Example 7.4. Let us determine by Thevenin's theorem the current /i(s) flowing
through the capacitor in the network shown in Fig. 7.15. First, let us obtain

Li(<i-> LB

uc(O-)

~
•
B

FIG. 7.15

z.(s) by opening all current sources and short circuiting all voltage sources.
Then, we have in the network in Fig. 7.16, where
z.(s) = R + sL (7.23)
Next we find v.(s) by removing the capacitor so that the open-circuit voltage
between the terminals 1 and 2 is v.(s), as shown in Fig. 7.17. We readily
184 Network analysis and synthesis

FIG. 7.16

determine from Fig. 7.17 that

vo(O-)
v.(s) """l(s)R + Li(O-) - --
s
(7.24)
By Thevenin's theorem we then have
v.(s) I(s)R + Li(O - ) -vo(O- )/s
(7.25)
Ii(s) = z.(s) + Z(s) = K + sL + 1/sC

~
R
•

FIG. 7.17

Example 7.5. For the network in Fig. 7.18, let us determine the voltage vo(t)
across the resistor by Thevenin's theorem. The switch closes at t "" 0, and we
assume that all initial conditions are zero at t = 0.
8

+
v(t) vo(t)

FIG. 7.11

First let us redraw the circuit in terms of its transformed representation,

which is given in Fig. 7.19. We can almost determine by inspection that the
Thevenin equivalent voltage source of the network to the left of Nin Fig. 7.19 is
V(s)(.1/sC)
V.(s) -= Li_s + 1/sC (7.26)
 Transform methods In network analysis 185
Li•

R Vo(a)

.,....J
FIG, 7,19

and the input impedance to the left of N is

sLi(l/sC)
Z.(s) = 8~ + sLi + 1/sC (7.27)

RV.(s)
We know that VJ.s) = Z.(s) + R (7.28)

RV(s)(l/sC)
Therefore (7.'1.9)
VJ.s) ""' (R + s~sLi + 1/sC) + LifC
Finally v,l..t) ... C-1 (Vo(s)] (7.30)

Norton's theorem
When it is required to find the voltage across an element whose admit-
tance is Y1(s), the rest of the network can be represented as an equivalent
admittance Y.(s) in parallel with an equivalent current source I.(s), as
shown in Fig. 7.20. The admittance Y.(s) is the reciprocal of the Thevenin
impedance. The current J.(s) is that current which flows through a short
circuit across Y1(s). From Fig. 7.20,

(7.31)

The element whose admittance is Y1 must not be magnetically coupled to

any element in the rest of the network.
Consider the network in Fig. 7.21. Let us find the voltage across the
capacitor by Norton's theorem. First, the short-circuit current source
1.(s) is found by placing a short circuit across the terminals 1 and 2 of the

FIG, 7.20
I86 Network analysis and synthesis
BL BL
1

R R Ie(s)

.____..___ _ _ _---<>2

FIG. 7.21 FIG. 7.22

capacitor, as shown in Fig. 7.22. From Fig. 7.22 I,(s) is

1 (s) = I(s)R (7.32)

• sL+ R
The admittance Y.(s) is the reciprocal of the Thevenin impedance, or
1
Y,(s) = sL+ R (7.33)

Then the voltage across the capacitor can be given as

V(s) = l.(s) _ I(s)R (7.34)

Y.(s) + Y0 (s) (sL + R)sC +1
Rumple 7.6. In the network in Fig. 7.23, the switch closes at t = 0. It is given
that v(t) ..., 0.le-6' and all initial currents and voltages are :zero. Let us find the
current i,l..t) by Norton's theorem.
The transformed circuit is given in Fig. 1.24. To find the Norton equivalent
current source, we short circuit points 1 and 2 in the network shown. Then
l.(s) is the current flowing in the short circuit, or
V(s) 0.1 1
(7.35)
l.(s) = R 1 + sL = L(_s + R/LXs + S) = (s + SXs + 10)
The equivalent admittance of the circuit as viewed from points 1 and 2 is
1 rLC + sR1C +1 0.Sr + Ss + 10 . ( _ 6)
( =sC
Y,s) + Rl +s
L = R1 +sL
= s + 10 73

L=O.l R1= 1
1
il(t)

C=f R2=2

2
FIG. 7.23
 Transform methods In network analysis I~

FIG. 7.24

I.(s)
I,J..s) is then I,J..s) = Rt[ Y.(s) + GJ (7.37)

1
=----- (7.38)
(s + 5)1(s + 6)
By inspection, we see that I,J..s) can be written as

(s + 6) - (s + 5) 1 1
(7.39)
I,l..s) - (s + 5)1(s + 6) ""' (s + 5)1 - (s + 6)(s + 5)
Repeating this procedure, we then obtain

1 1 1
IJ..s) = (s + 5)1 - s +5 +s +6 (7.40)

Taking the inverse transform of I,J..s), we finally obtain

i,J..t) = (te-51 - e-51 + e-411)u(t) (7.41)

7.3 THE SYSTEM FUNCTION

As we discussed earlier, a linear system is one in which the excitation

e(t) is related to the response r(t) by a linear differential equation. When
the Laplace transform is used in describing the system, the relation between
the excitation E(s) and the response R(s) is an algebraic one. In particular,
when we discuss initially inert systems, the excitation and response are
related by the system function H(s) as given the relation
R(s) = E(s)H(s) (7.42)
We will discuss how a system function is obtained for a given network,
and how this function can be used in determining the system response.
As mentioned in Chapter 1, the system function may assume many forms
and may have special names such as driving-point admittance, transfer
188 Network analysis and synthesis

R T+
Vo(B)
1
'al!

FIG. 7.25
BL
1
impedance, voltage or current-ratio transfer function. This is because the
form of the system function depends on whether the excitation is a voltage
or current source, and whether the response is a specified current or
voltage. We now discuss some specific forms of system functions.
Impedance
When the excitation is a current source and the response is a voltage,
then the system function is an impedance. When both excitation and
response are measured between the same pair of terminals, we have a
driving-point impedance. An example of a driving-point impedance is
given in Fig. 7.25, where
H(s) = Vo(s) = R + (1/sC)sL (7.43)
Ig(s) sL + 1/sC
Admittance
When the excitation is a voltage source and the response is a current,
H(s) is an admittance. In Fig. 7.26, the transfer admittance /JV,, is
obtained from the network as
H(s) = I 1(s) = 1 (7.44)
V,(s) R1 + sL+ l/sC1
Voltage-ratio transfer function
When the excitation is a voltage source and the response is also a voltage,
then H(s) is a voltage-ratio transfer function. In Fig. 7.27, the voltage-
ratio transfer function V0(s)/V,,(s) is obtained as follows. We first find the
1
BL .C1

+
~(B) lz(B)
1
.c;

FIG. 7.26
 Transform methods In network analysis 189

T
~<•> ~ Vo<•>

FIG, 7:rt
l
current
/() V.,(s) (7.45)
I
s = Zi(s) + Z 1(s)
Since V0(s) == Z 1(s) l(s) (7.46)
Vo(s) Z.(s)
then --= (7.47)
V.(s) Zi(s) + Z1(s)
Current-ratio transfer function
When the excitation is a current source and the response is another
current in the network, then H(s) is called a current-ratio transfer function.
As an example, let us find the ratio 10/111 for the network given in Fig. 7.28.
Referring to the depicted network, we know that
1
l.,(s) = l 1(s) + Io(s), Ii(s)- = Io(s)(R + sL) (7.48, 7.49)
sC
Eliminating the variable Ii, we find
I.,(s) = I 0(s)(t + R + sL) (7.50)
1/sC
so that the current-ratio transfer function is
Io(s) 1/sC
-= (7.51)
I.,(s) R + sL + 1/sC

1 R
.c

FIG. 7.28
190 Network analysis and synthesis

C L G
Tv(t)

FIG. 7.29
l
From the preceding examples, we have seen that the system function is a
function of the elements of the network alone, and is obtained from the
network by a straightforward application of Kirchhoff's laws. Now let us
obtain the response transform R(s), given the excitation and the system
function. Consider the network in Fig. 7.29, where the excitation is the
current source i,,(t) and the response is the voltage v(t). We assume that
the network is initially inert when the switch is closed at t = 0. Let us find
the response V(s) for the excitations:
1. i,,(t) = (sin w0 t) u(t).
2. i,,(t) is the square pulse in Fig. 7.30.
3. i,,(t) has the waveform in Fig. 7.31.
First, we obtain the system function as
s
H(s) - l
sC + 1/sL+ G
=--------
1
C[s + s(G/C) + 1/LC]
(7.52)

1. i,,(t) = (sin w0t) u(t). The transform of i,,(t) is

mo
I,,(s) = s• + ro • (7.53)
0

so that V(s) = I,,(s)H(s) = Wo s (7.54)

s1 + ro0I C[s1 + s(G/C) + 1/LC]

ig(t)
ig(t)

1--- 2

0 a t
0 a
FIG. 7.30 FIG. 7.31
 Transform methods In network analysis 191
2. For the square pulse in Fig. 7.30, i,,(t) can be written as

i,,(t) = u(t) - u(t - a) (7.55)

Its transform is I,,(s) = ! (1 - e--M) (7.56)

s
The response V(s) is therefore given as
1 - e-{U
(7.57)
= C[s1 + s(G/C) + 1/LC]
V(s)

Note that in obtaining the inverse transform c-1 [V(s)], the factor e- 0
•

must be regarded as only a delay factor in the time domain. Suppose we

rewrite V(s) in Eq. 7.57 as

V(s) = H(s) (1 - e~ (7.58)

s
Then if we denote the inverse transform by v1(t),

v1(t) = c-i[H~s>] (7.59)

we obtain the time response

v(t) = vi(t) - v1(t - a) u(t - a) (7.60)

Observe that v1(t) in Eq. 7.59 is the response of the system to a unit step
excitation.
3. The waveform in Fig. 7.31 can be represented as

i,,(t) = u(t) + !. u(t) - (t - a) u(t - a) (7.61)

a a

and its transform is / 11(s) = -1 ( 1 + -1 - -e-o•) (7.62)

s as as

V(s) is then V(s) = (1 + .!_ - _e-oa_) _ _ _ _l _ _ __ (7.63)

as as C[s 8 + s(G/C) + 1/LC]
If we denote by vs(t) the response of the system to a ramp excitation, we
see that
c-1(V(s)] = vi(t) + ! V1(t) - ! Vz{t - a) u(t - a) (7.64)
a a
where vi(t) is the step response in Eq. 7.59.
192 Network analysis and synthesis
Let us now discuss some further ramifications of the equation for the
response R(s) == H(s) E(s). Consider the partial-fraction expansion of
R(s):

R(s)-I~+I_!!_L__ (7.65)
, S - s, I S - S1

where s, represent poles of H(s), and s1 represent poles of E(s). Taking

the inverse transform of R(s), we obtain

(7.66)

The terms A/!''' are associated with the system H(s) and are called free
.response terms. The terms B,e''' are du~ to the excitation and are known
as forced response terms. The frequencies s, are the natural frequencies of
the system andsI are the forced frequencies. It is seen from our discussion
of system stability in Chapter S that the natural frequencies of a passive
network have real parts which are zero or negative. In other words, if we
+
denote s, as s, == a, jw,, then a, ~ 0.
~pie 7.7;· For the initially inert network in Fig. 7.32, the excitation is
v.(t) =- l cost u(t). Let us find the response vJ..t) and determine the free and
forced response parts of vJ..t). The system function is
V0(s) 1/sC 2
(7.67)
V.(s) .. R +.1/sC =s +2
Since v.(s) is v.<s> -Hr: 1) (7.68)

the response is then

s 0.4 0.4s + 0.2
VJ..s) - V.(s)H(s) .. (sa + l)(s + 2) .. - s +2 + r +l (7.69)

We next obtain the inverse transform

vJ..t) - -0.4c" + 0.4 cos t + 0.2 sin t (7.70)

FIG. 7.D
 Transform methods In network analysis 193
It is apparent that the free response is -o.4e-•, and the forced response is
0.4 cos t + 0.2 sin t.
As a final topic in our discussion, let us consider the basjs of operation
for the R-C differentiator and integrator shown in Figs. 7.33a and 7.33b.
We use the Fourier transform in our analysis here, so that the system
function is given as H(jro), where "' is the ordinary radian frequency
variable. Consider first the system function of the differentiator in Fig.
7.33a.
V.(ja,) R jroRC
--== = (7.71)
V.Uro) R + 1/jroC jroRC +1
Let us impose the condition that

roRC «1 (7.72)
We have, approximately,
__ ,..,}"'
Yo(jro) . RC
YoUro)-
(7.73)

Then the response VIJro) can be expressed as

VIJro) ~ RCLJro V.(jro)] (7.74)

Taking the inverse transform of Vo(jro),.we obtain

vo(t) ~ RC ~ v.(t) (7.75)

dt
Note that the derivation of Eq. 7. 75 depends upon the assumption that the
R-C time constant is much less than unity. This is a necessary condition.
Next, for the R-C integrator in Fig. 7.33b, the voltage-ratio transfer
function is
Yo(jro) 1/jroC 1
--= ---- (7.76)
v.Uro) 1/jroC + R jroCR + 1
C R

volt)

,., (b)

PIG. 7.33. (a) R-C difl'crmtiator. (b) R-C integrator.

194 Network analysis and synthesis
If we assume that c.oRC » 1, then

V0(jro) ~ -.- -V,,(iw)

1
(7.77)
Jc.oRC
Under these conditions, the inverse transform is

v0(t),.., - 1
RC
l'
o
v,,(-r) d-r (7.78)

so that the R-C circuit in Fig. 7.33b is approximately an integrating circuit.

7.4 THE STEP AND IMPULSE RESPONSES

In this section we will show that the impulse response h(t) and the system
function H(s) constitute a transform pair, so that we can obtain step and
impulse responses directly from the system function.
We know, first of all, that the transform of a unit impulse <5(t) is unity,
i.e., C[<5(t)] = 1. Suppose the system excitation were a unit impulse, then
the response R(s) would be
R(s) = E(s) H(s) = H(s) (7.79)
We thus see that the impulse response h(t) and the system function H(s)
constitute a transform pair, that is,
= H(s)
C[h(t)]
(7.80)
c-1 (H(s)] = h(t)
Since the system function is usually easy to obtain, it is apparent that we
can find the impulse response of a system by taking the inverse transform
of H(s).
Example 7.8. Let us find the impulse response of the current i(t) in the R-C
circuit in Fig. 7.34. The system function is
J(s) 1 s
(7.81)
H(s) = V,,(s) = R + l/sC = R(s + l/RC)
Simplifying H(s) further, we have
1( 1/RC ) (7.82)
H(s) =R 1- s + l/RC
The impulse response is then
h(_t) = r.-1 [H(s)] = .!_
R
["<,) - -RC1 e-t/RO] u(t) (7.83)
which is shown in Fig. 7.35.
 Transform methods in network analysis 195
1
1i

i(t)
C

_..!._
R1 C
FIG. 7.34 FIG. 7.35. Current impulse
response of R-C network in
Fig. 7.34.

Since the step response is the integral of the impulse response, we can
use the integral property of Laplace transforms to obtain the step response
as

(7.84)

where cx(t) denotes the step response. Similarly, we obtain the unit ramp
response from the equation
y(t) = c-1 [~~s)] (7.85)

where y(t) denotes the ramp response. From this discussion, it is clear
that a knowledge of the system function provides sufficient information to
obtain all the transient response data that are needed to characterize
the system.
Example 7.9. Let us find the current step response of the R-L circuit in Fig.
7.36. Since l(s) is the response and V,,(s) is the excitation, the system function
is
/(s) 1
H(s) = Vo(s) = R + sL (7.86)
Therefore H(s)/s is
H(s) 1 1 (1 1 ) (7.87)
-s- = s(_R + sL) = R ; - s + R/L
The step response at(t) is now obtained as
1
at(t) =- (1 - e-(R/L'>t) u(_t) (7.88)
R
To check this result, let us consider the impulse response of the R-L circuit,
196 Network analysis and synthesis
which we found in Chapter S.
1
h(t) - L e-<RfL>t u(t) (7.89)

The step response is the integral of the impulse response, or

cx(t) - f'h(T) dT = .!. (1 - e-<RfL"') u(_t) (7.90)

Jo R
It is readily seen that if we know the impulse response of an initially inert
linear system, we can determine the response of the system due to any other
excitation. In other words, the impulse response alone is sufficient to charac-
terize the system from the standpoint of excitation and response.
R

Inert
network
~L

FIG, 7.36 FIG, 7,37

EulllpJe 7.10, In Fig. 7.37, the only information we possess about the system
in the black box is: (1) it is an initially inert linear system; (2) when vl.,t) = d<_t),
theri
(7.91)
With this information, let us determine what the excitation vl.,t) must be in order
to produce a response vo(t) - te-11 u(_t). First, we determine the system function
tobe·
Vo(s) 1 1 2s + S
(7.92)
H(s) = V ,<s) - s + 2 + s + 3 .. (s + 2)(s +3)

We next find the transform of t r11 u(_t).

1
VJ..s) -= t[tr"] - (s + 21 (7.93)

The unknown excitation is theri found from the equation

VJ..s) (s + 2)(s + 3)
V,<s) = H(s) - 2(s + 2.S)(s + 2)' (7,94)

Simplifying and expanding V,<s) into partial fractions, we have

(s + 3) 1 O.S
V,(s) ... 2(s + 2.S)(s + 2) - s +2 - s + 2.S (7 ,9S)

The system excitation is then

vt..t) - (e-11 - o.se-u') u(t) (7.96)
 Transform methods In network analysis 197

7.5 THE CONVOLUTION INTEGRAL

In this section we will explore some further ramifications of the use of

the impulse response h(t) to determine the system response ,(t). Our
discussion is based upon the important convolution theorem of Laplace
(or Fourier) transforms.. Given two functions fi(t) and f.Jt), which are
zero for t < 0, the convolution theorem states that if the transform of
/ 1(t) is F1(s), and if the transform of /.J.t) is F.J.s), the transform of the
convolution of /i.(t) and J.(t) is the product of the individual transforms,
F1(s) F.(.s), that is,

r.[ f.'fi(t - -r) f.(.-r) d-r] == Fi(s) F.(.s) (7.97)

where the integral

is the convolution integral or/olding integral, and is denoted operationally

as

(7.98)

Proof. Let us prove that t[fi*/.] = F 1F1• We begin by writing

t[/i(t)• /.(.t)] = f. e-•[ I:/

00

1(t - -r)/.(.-r) d-r] dt (7.99)

From the definition of the shifted step function

u(t - -r) =1 -r ~ t
(7.100)

we have the identity

f.'1 (t -
1 -r)/.J.-r) d-r = f. 00

J1(t - -r) u(t - -r)JJ.:r) d-r (7.101)

Then Eq. 7.99 can be written as

t[/1(t)• /.J.t)] = f. f.
00

e-•
00

J1(t - -r) u(t - -r)/.J.-r) d-r dt (7.102)

If we let z =t- -r so that

(7.103)
198 Network analysis and synthesis
then Eq. 7.102 becomes
C£f1(t)* fs(t)] = i"' i"'fi(x) u(x)f1(-r)e- ste-•z d-r dx

= i"'11(x) u(x)e-•z dxi"'J,.(-r)e_., d-r

= Fi(s)F,.(s) (7.104)
The separation of the double integral in Eq. 7.104 into a product of two
integrals is based upon a property of integrals known as the separability
property. 1
Example 7.11. Let us evaluate the convolution of the functions/1(t) = e-2t u(t)
and / 1(t) = t u(t), and then compare the result with the inverse transform of
Fi(s) F1(s), where
1
Fi(s) = l'.[fi.(t)] = - -2
s+
(7.105)
1
F8(s) =~
The convolution of / 1(t) and [a(t) is obtained by first substituting the dummy
variable t - 'T fort in/i(t), so that
Ji.(t - T) = e-ll(t-.) u(t - T) (7.106)
Then / 1(t)• fs(t) is

L/1(t - T)/1(T) dT = L'-re-•U-.) dT = e-•eL\e•• dT (7.107)

Integrating by parts, we obtain

/ 1(t)• / 1(t) = G-¼+ ¼e-Bt) u(t) (7.108)

Next let us evaluate the inverse transform of F 1(s) F1(s). From Eq. 7.105, we
have
(7.109)

so that c-1 [F1(s) F 2(s)] = ( -t - -1 + -1 e-•t ) u(t) (7.110)

2 4 4
An important property of the convolution integral is expressed by the
equation
(7.111)

• See, for example, P. Franklin, A Treatise on Advanced Calculus, John Wiley and
Sons, New York, 1940.
 Transform methods In network analysis 199
This is readily seen from the relationships
C[/1(t)* J.(t)J = Fis) Fl._s) (7.112)
and C[fs(t)* / 1(t)J = F,.(s) F1(s) (7.113)
To give the convolution integral a more intuitive meaning, let us examine
the convolution or folding process from a graphical standpoint. Suppose
we take the functions
/i(-r} = u(-r)
(7.114)
/z(-r} = T u(-r}
as shown in Figs. 7.38a and 7.39a. In Fig. 7.38, the various steps for
obtaining the integral ,
Lfi(t - -r}/1(-r} d-r
are depicted. Part (b) of the figure shows/i(--r} = u(--r}. The function
f1(t - -r) = u(t - -r) (7.115)
in part (c) merely advances / 1(--r} by a variable amount t. Next, the
product ·
f 1(t - -r}/1(-r} = u(t - -r)-r u(-r} (7.116)
is shown in part (d). We see that the convolution integral is the area under
the curve, as indicated by the cross-hatched area in part (d). Since the
convolution integral has a variable upper limit, we must obtain the area
under the curve of/1(t - -r)J.(-r} for all t. With t considered a variable in
Fig. 7.38d, the area under the curve is
,,.
Ji* / 1 = f(t) = (7.117)
2 u(t)
as plotted in part (e) of the figure.
In Fig. 7.39, we see that by folding/1(-r} about a point t, we obtain the
same result as in Fig. 7.38. The result in Eq. 7.117 can be checked by
taking the inverse transform of F 1(s) F.(s}, which is seen to be

c- [F (s)F (s)J = c-1[~] = ~ u(t)

1
1 1 (7.118)

Let us next examine the role of the convolution integral in system

analysis. From the familiar equation
R(s) = E(s) H(s) (7.119)
we obtain the time response as
r(t) = c- [E(s) H(s)J = J: e(-r) h(t -
1
-r} d-r (7.120)

where e(-r) is the excitation and h(-r) is the impulse response of the system.
200 Network analysis and synthesis
(1(7')

0 7' 0 7'
(G) (a)

ft(-7')

-T 0 T 7'
(b)

f1(t- T)
1

-T 0 t T 0 t 7'
(c) (c)

f1(t- T) h(T) ft (T) f,.(t - T)

-7' 0 7'

fi•h ,~,1
'"'
0 t t
(e) le)
FIG. 7.31 FIG.Ut

Using Eq. 7.120, we obtain the response of a system directly in the time
domain. The only information we need about the system is its impulse
response.
Euaple 7.12a. Let us find the responsei(t) of the R-L network in Fig. 7.40 due
to the excitation v(_t) _ 2r u(.t) (7•121)

The impulse response for the current is

1
h(_t) - L e-<RIL'/1 u(.t) (7.122)

Therefore, for the R-L circuit under discussion

h<.t) - ell u(.t) (7.123)
 Transform methods In networlt. analysis 20t
Using the convolution integral. we obtain the respome l(t) u

i(t) - f.'v<., - T) /i(T) dT - 2 J.'rtt-r>,-ar dT

- 2r' J.'r dT - 2(c' - rt') M(t)
(7.124)

+ "'' ~ 1
+
e(t)
Idell
amplifier
+
r(t)

FIG.7M FIG. 7.41

Evmple '7.lllt. The ideal amplifier in Fig. 7.41 bas asystem function H(8) - K,
where K is a constant. The impulse response of the ideal amplifier is then
h(t) - K "(_t) (7.125)
Let us show by means of the convolution integral that the response ,(t) is related
to the excitation e(t) by the equation
,(t) - K e(t) (7.126)
Using the convolution integral. we have

,(t) - J.'e<T) h(t - T) dT - K f.' e(T) "'-' - T) dT - K e(t) (7.127)

Since t is a variable in the expression for ,(t), we see that the ideal amplifier in
the time domain is an impllbe-sCfllUlinK device which scans the input e(t) from
t - Oto t - ao. Thus, the response of an ideal amplifier is a replica of the
input e(t) multiplied by the Kain K of the amplifier.

7.6 THE DUHAMEL SUPERPOSITION INTEGRAL

In the Section 1.S, we discussed the role that the impulse response plays
in determining the response of a system to an arbitrary excitation. In
this section we will study the Duhamel superposition integral, which also
describes an input-output relationship for a system. The superposition
integral requires the step response at(t) to characterize the system behavior.
We plan to derive the superposition integral in two different ways.
202 Network analysis and synthesis
The simplest is examined first. We begin with the excitation-response
relationship
R(s) == E(s) H(s) (7.128)
Multiplying and dividing by s gives

R(s) = H(s) · s E(s) (7.129)

s
Taking inverse transforms of both sides gives

i:-1[R(s)] = i:-1[H~s) · s E(s)]

(7.130)
= 1:-1 [H~s)] •1:-1[s E(s)]
which then yields
r(t) = 0t(t)*[e'(t) + e(0-) <S(t)]
= e(0-) 0t(t) + J: e'('r) 0t(t - -r) d-r (7.131)

where e'(t) is the derivative of e(t), e(0-) is the value of e(t) at t 0-, =
and 0t(t) is the step response of the system. Equation 7.131 is usually
referred to as the Duhamel superposition integral.
Example 7.13. Let us find the current i(t) in the R-C circuit in Fig. 7.42 when
the voltage source is v,,(t) = (Ai + Aat) u(_t) (7.132)

as shown in Fig. 7.4l, The system function of the R-C circuit is

I(s) s
(7.133)
H(s) = V,,(s) = R(s + 1/RC)
Theref(!re the transform of the step response is
H(s) I
-= (7.134)
s R(s + l/RC)
Vg(t)

"•'''~ ,~
iR A1

SL Tc 0

FIG.7,42 FIG. 7.43

 Transform methods in network analysis 203
Taking the inverse transform of H(s)/s, we obtain the step response

a.(t) == R.1 e-tlRO u(_t) (7.135)

From the excitation in Eq. 7.132, we see that e(O-) = 0 and

v',,(t) = A1 i5(t) + A 8 u(,t) (7.136)
The response is then
i(t) =J~v',,(T) a.(t - T)dT

-J~A1i5(T)a.(t- T)dT + f~A.a.(t-T)dT

= A1 et{t) + A1i'
. R o--
e-<t-r>JRO dT (7.137)

= _A1 e-1/ROu<,t) + As
R
_ e-tlRO
R
i',
o--
e1IRO dT

= [!1 e-tlRO + As C(l - e-tlRG)] u(_t)

=
As we see from the Example 7.13, e(O-) 0, which is the case in many
transient problems. Another method of deriving the Duhamel integral
avoids the problem of discontinuities at the origin by assuming the lower
limit of integration to bet= o+. Consider the excitation e(t) shown by
the dotted curve in Fig. 7.44. Let us approximate e(t) by a series of step
functions, as indicated in the figure. We can write the staircase approxi-
mation of e(t) as
e(t) = e(O+) u(t) + /:;.£1 u(t - a-r)
7 3
+ /:;.£1 u(t - 2d-r) + · · · + /:;.En u(t - nd-r) ( .I S)

.,,,
Signal°'v }
/ .U:3

FIG. 7.44. Staircase approximation to a signal.

204 Network analysis and synthesis
wherell.E,.istheheightofthe step increment at t - k4T. Since we assume
the system to be linear and time invariant, we know that if the response to a
unit step is «(t), the response to a step ~ u(t - T) is ~ «(t - T). There-
fore we can write the response to the step approximation in Fig. 7 .44 as

r(t) = e(O+) «(t) + M1 «(t - AT)

+ fl.Es «(t-- 2AT) + ... + M" «(t - nat) <7·139>
If AT is small, r(t) can be given as

r(t) =- e(O+) «(t) + AEi «(t - AT) AT

AT
AB1 · AE (7.140)
+ -AT «(t - 2AT) AT + · · · + -"
AT
«(t - nAT) AT

which, in the limit, becomes

r(t) .. e(O+) «(t) + 1im I" _,

AB «(t -
.,...o,-o+ AT
iAT) AT

-= e(O+) «(t) + f: e'(T) «(t - T) dT

(7.141)

Problems
7,1 In the circuit shown v(,t) - 2-(t) and iL(O-) - 2 amps. Find and
sketch i,.(t).

1000

I it<O I
+iz.w ♦
+
r,(O 5h 1000

PllOB,7.1

7.2 The switch is thrown from position 1 to 2 at t - 0 after having been at 1

for a long time. The source voltage is ol,.t) - V1r-t sin {Jt.
(a) Find the transform of the output voltage o,1..t).
(b) Find the initial and final values of v.(t).
(c) Sketch one possible set of locations for the critical frequencies in the s p1ane
and write the form of the response v.(t). (Do not take the inverse transform.)
 Transform methods In network analysis 205
2

llf
.±L
+ C R
7 llfit)

'-----'---v____-i.___ ___,J
PROL7.2

73 In the circuit shown, all initial currents and voltages are :zero. Find
i(t) for t > O using Thevenin's theorem.
11(t)

Sl(t) t it
lh

PROL7.3

7.4 In the circuit shown in (a), the excitation is the voltage source e(t)
desat"bed in (b). Determine the response i(t) assurning :zero initial conditioos.

e(t)

10

2 t

(a) (b)

PROL7.4

7.S Determine the expression for r,J.t) when i(t). - cl(t), assuming :zero initial
conditions. Use transform methods.
206 Network analysis and synthesis

i(t)

lf

PROB. 7.5

7.6 The circuit shown has zero initial energy. At t = 0 the switch S is
opened. Find the value of the resistor X such that the response is v(_t) =
0.5 sin V2t u(t). The excitation is i(t) = tr "It u(t).

s X

PROB.7.6

7.7 Use transform methods to determine the expressions for ii(t) and i.(t)
in the circuit shown. The excitation is v(_t) = 100e-10•. Assume zero initial
energy.
1000

~ .
•. . )lOh

PROB. 7.7

7.8 For the circuit shown. the switch Sis opened at t = 0. Use Thevenin's
or Norton's theorem to determine the output voltage v,l..t). Assume zero initial
energy.
 T~nsform methods In network analysis 207

0.10

lf lf
7 "2(t)

1..---:___--'-o-.1-o_--'----.J--0-.1-0_,J
PROB.7,8

7.9 For the transformer shown, find ii(t) and IJ.t). It is given that e(t) =-
6u(t), and that prior to the switching action all initial energy was zero; also
M == th.

30

~ lh 80

PROB. 7.9

7.10 For the circuit shown, find i,l..t), given that the circuit had been in steady
state prior to the switch closing at t = 0.

100 400

•
+ .
-=-40v
Sh~

PROB.7.IO

7.11 Find i,,(_t) using Thevenin's theorem. The excitation is e(t) == 100
cos 20 u(_t). Asstµne zero initial energy.
208 Network analysis and synthesis

100 2h
300
+
e ~ 7.50
6h

PROB. 7.11

7.12 Determine the transfer function H(s) == V,l..s)/V1(s). When v1(t) ==

u(t), find va(.t). Assume z:ero initial conditions.

10

V1 !h
2
lt
3
T
Va 3f lo
3
7
Va

~
PROB. 7.12

7.13 Using (a) standard transform methods and (b) the convolution integral,
find v(t) when i(t) ... :ze-• u(t). Assume z:ero initial energy.

- i (t)

¼D If
u(t)

PROB. 7.13

7.14 The impulse response of a linear system is shown in the figure. If the
excitation were e(t) 11=1 3e-u u(t), determine the response values r(l) and r(4)
using graphical convolution.
h(t)
2.....,_ _ _ ___

0 3

PROB. 7.14
 Transform methods In network analysis 209
· 7.15 The system function is given as H(s) =- 1/(s' + 9)1. H the excitation
were e(t) -= 3d'(t), determine the response r(t). (Hint: Use the convolution
integral to break up the response transform R(s).)
7.16 Solve the following integral equations for x(t).

(a) :r(t) + J.'

(t - T) x( T) dT == 1

(b) sin t == L'x(T)e-«c-r1 dT

(c) x(t) +L'xe:, - T) e-r dT == t

7.17 Using the convolution integral find the inverse transform of
K
(a) .fts) ,. (s + a)(s + b)
(b) Ff)=- 2(s + 2)
,s (s1 + 4)1
s
(c) F(s) = (s• + t)•
7.18 By graphical means, determine the convolution of /(t) shown in the
figure with itself (i.e., determine /(t)*/(1)).

f(t)

11-----

T t
PROB. 7.18

7.19 Using (a) the convolution integral and (b) the Duhamel superposition
integral, find v(t) for e(t) = 4e-1, ·u(t). Assume 7.el'o initial conditions.

lh

PROB.7.19
210 Network analysis and synthesis
7.20 Using (a) the convolution integral and (b) the Duhamel superposition
integral, find v(t) for e(t) = 2e-a1 u(t). Assume zero initial conditions.
30

lh
+
e(t)

10

PROB, 7.20

7.21 For the circuit in (a);- the system function H(s) == V(s)/I(s) has the
poles shown in (c). Find the element values for Rand C. If the excitation i(t)
has the form shown in (b), use the convolution integral to find v(t).
i(t)

C lh

(a) (b)

H(s) C•0.5 i<.1

-i><-:><,---+-----
- u -2 -1 0

(c)

PROB. 7.21

7.22 The excitation of a linear system is :i:(t), shown in (a). The system
impulse response is h(t), shown in (b). Sketch the system response to x(t).
(No equations need be written. A neat, carefully dimensioned sketch will
suffice.)
 Transform methods In network analysis 211
z(t)

2---
0 2 t
(a)
h(t)

t
(b)
PROB. 7.22

7.23 A unit step of voltage is applied to the network and the resulting current
is i(t) ... 0.0lc' + 0.02 amps.
(a) Determine the admittance Y(s) for this network.
(b) Find a network that will yield this admittance function.

i(t)

Network
Y(s)
~

PROB.7.23

8 v(t) Network

Z<•i
2
PROB.7.M

7.24 The current generator delivers a constant current of 4 amps. At

t = 0 the switch Sis opened and the resulting voltage across the terminals 1, 2 is
v(t) = 6e-4f + 12 v.
(a) Find Z(s) looking into terminals 1, 2.
(b) Find a network realimtion for Z(s).
 8 chapter
Amplitude, phase, and delay

8.1 AMPLITUDE AND PHASE RESPONSE

In this section we will study the relationship between the poles and
zeros of a system function and its steady-state sinusoidal response. In
other words, we will investigate the effect of pole and zero positions upon
the behavior of H(s) along the jro axis. The steady-state response of a
system function is given by the equation
HUro) ==M (ro)e 1• 1"'> (8.1)
where M(ro) is the amplitude or magnitude response function, and is an
even function in ro. t/,(.ro) represents the phase response, and is an odd
function of ro.
The amplitude and phase response of a system provides valuable
information in the analysis and design of transmission circuits. Consider
the amplitude and phase characteristics of a low-pass filter shown in
Figs. 8. la and 8. lb. The cutoff frequency of the filter is indicated on the
amplitude response curves as ro 0 • It is generally taken to be the "half-
power" frequency at which the system function IH(Jroa)I is equal to 0.707
of the maximum amplitude IH(j00max)I. In terms of decibels, the half-
power point is that frequency at which 20 log IH(Jroa)I is down 3 db from
20 log IH(j00max)I. The system described by the amplitude and phase
characteristics in Fig. 8.1 shows that the system will not "pass" frequencies
that are greater than ro0 • Suppose we consider a pulse train whose
amplitude spectrum contains significant harmonics above w 0 • We know
that the system will pass the harmonics below ro 0 , but will block all
harmonics above ro0 • Therefore the output pulse train will be distorted
when compared to the original pulse train, because many higher harmonic
terms will be missing. It will be shown in Chapter 13 that if the phase
212
 Amplitude, phase, and delay 213
M("')
l
Amplitude
--0.707 response

-"' 0
(a)
"'c "'

"""'' Phase response

-"' -"'c "'c "'

(b)
FIG. 8.1. Amplitude and phase response oflow-pass filter.

response 4'(.w) is linear, then minimum pulse distortion will result. We see
from the phase response </,(.w) in Fig. 8.1 that the phase is approxi-
mately linear over the range -w0 ~ w ~ +w0 • If all the significant
harmonic terms are less than ru0 , then the system will produce
minimum phase distortion. With this
example, we see the importance of an
amplitude-phase description of a system.
+
In the remaining part of this chapter,
we will concentrate on methods to
V1(•J
obtain amplitude and phase response
curves, both analytically and graph-
ically. 0

To obtain amplitude and phase FIG. 8.2

curves, we let a == jru in the system
function, and express HUw) in polar form. For example, for the amplitude
and phase response of the voltage ratio VJV1 of the R-C network shown
in Fig. 8.2, the system function is
H(s) _ V.(s) _ 1/RC
(8.2)
Y1(s) s 1/RC +
Letting s - jw, we see that HUru) is
1/RC
HUw) (8.3)
jru + 1/RC
214 Network analysis and synthesis
M(w)

0 w
q,(w)
-45·

-90·

FIG. 8.3. Amplitude and phase response of R-C network.

In polar form H(jw) becomes

)
H(j w= 1/RC -lt&n-la,RO
= M( we
} 14,(a,)
(w9 + 1/R2C8)1A e (8.4)

The amplitude and phase curves are plotted in Fig. 8.3. At the point
w = 0, the amplitude is unity and the phase is zero degrees. As w increases,
the amplitude and phase decrease monotonically. When w =
l/RC, the
amplitude is 0.707 and the phase is -45°. This point is the half-power
point of the amplitude response. Finally as ru-+- oo, M(w) approaches
zero and cf,(w) approaches -90°.
Now let us tum to a method to obtain the amplitude and phase response
from the pole-zero diagram of a system function. Suppose we have the
system function
H(s) = Ao(s - Zo)(s - Zi) (8.5)
(s - p0 )(s - pJ(s - P1)
H(jw) can be written as
H(jw) = Ao(jw - z0)(jw - zJ (8.6)
(jw - Po)Uw - P1)Uw - Pa)
Each one of the factors jw - z, or jw - p I corresponds to a vector
from the zero z, or pole p I directed to any point jw on the imaginary axis.
Therefore, if we express the factors in polar forw,
jw - z, =
N,e 1"'', j w - p1 = M 1e,,,, (8.7)
then H(jw) can be given as
H(jw) = AoNoN1 e'(,po+,p1-to-f1-etl (8.8)
MoM1M1
as shown in Fig. 8.4, where we note that 01 is negative.
 Amplitude, phase, and delay 215
jw

(T

FIG. 8.4. Evaluation of amplitude and phase from pole-zero diagram.

In general, we can express the amplitude response M(ro) in terms of the

following equation.

II" vector magnitudes from the zeros to the point on the jro axis
0
M(ro) = -•=- ----------------------

II"' vector magnitudes from the poles to the point on the jro axis
i-0

Similarly, the phase response is given as

tf,(w) = I" angles of the vectors from the zeros to the jro axis
i-0

I"' angles of the vectors from the poles to the jro axis
- i-0
It is important to note that these relationships for amplitude and phase
are point-by-point relationships only. In other words, we must draw
vectors from the poles and zeros to every point on the jro axis for which we
wish to determine amplitude and phase. Consider the following example.

F(s) _ 4s 4s (8.9)
- s1 + 2s + 2 - (s + 1 + jl)(s + 1 - jl)
Let us find the amplitude and phase for F(j2). From the poles and zeros
of .fts), we draw vectors to the point ro = 2, as shown in Fig. 8.5. From
216 Network analysis and synthesis
jw

O'

FIG. 8.5. Evaluation of amplitude and phase from pole-7.ero diagram.

the pole-zero diagram, it is clear that

M(j2) = 4(/l : ✓Io) = 1.78

and cf,(j2) = 90° - 45° - 71.8° = -26.8°
With the values M(j2) and cf,(j2) and the amplitude and phase at three or
four additional points, we have enough information for a rough estimate

jw
s~ "'"
jl

90• 90•

1.0 O'

&.
)( -jw

(a) (bl
FIG. 8.6. Determining amplitude and phase at uro and very high frequencies.
 Amplitude, phase, and delay 217
of the amplitude and phase response. At co == 0, we see that the vector
magnitude from the zero at the origin to co == 0, is of course, zero.
Consequently, M(j0) -= 0. From Eq. 8.9 for F(_s), F(_j0) is
4
lim FUco) == UO) (8.10)
a,➔O (1 + jl)(l - jl)
a,>O

Fro,n this equation, we see that the zero at the origin still contributes a
90° phase shift even though the vector magnitude is zero. From Fig.
8.6a we see that the net phase at co == 0 is 4'(0) == 90° - 45° 45° == 90°. +
Next, at a very high frequency coA, where coA 1, all the vectors are »
approximately equal to coAe'"°, as seen in Fig. 8.6b. Then

M(coA) ~ - 1 ==
4a,A
-4
COA COA

and 4'(coJ ~ 90° - 90° - 90° == -90°

Extending this analysis for the frequencies listed in Table 8.1, we obtain
values for amplitude and phase as given in the table. From this table we
can sketch the amplitude and phase curves shown in Fig. 8.7.
Next let us examine the effect of poles and zeros on the jco axis upon
frequency response. Consider the function

F(s) == s1 + 1.03 == (s + j1.015)(s - jl.015) (8.11)

s8 + 1.23 (s + jl.109)(s - jl.109)

2.5
2.0 .......
/~
j 1.5
' -.... ........ _
f 1.0
< 0.5
I
I
I
---
00 1 2 3 4 5 6 7 8 9 10

+90

11-ao "' '

+60
+30
0

If _60
-90
' r-
--
·o 1 2 3 4 5 6 7 8 9 10
Frequency, w -
FIG. 1.7. Amplitude and phase response for F(s) in Eq. 8.9.
 218 Network analysis and synthesis
TABLE 8.1
Frequency, o, Amplitude Phase, degrees
1.0 1.8 2S.8
1.S 2.0 -S.3
3.0 1.3 -so.o
s.o 0.8 -66.0
10.0 0.4 -18.S

whose pole-zero diagram is shown in Fig. 8.8. At ro = 1.015, the vector

from zero to that frequency is of zero magnitude. Therefore at a zero
on the jw axis, the amplitude response is zero. At ro = 1.109 the vector
from the pole to that frequency is of zero magnitude. The amplitude
response is therefore infinite at a pole as seen from Eq. 8.11. Next,
consider the phase response. When
jw w < 1.015, it is apparent from the pole-
jl.109 zero plot that the phase is zero. When
jl.015 w > 1.015 and ro < 1.109, the vector
from the zero at ro = 1.015 is now
(T
pointing upward, while the vectors from
the other poles and zeros are oriented
-jl.015 in the same direction as for ro < 1.015.
r-jl.109 We see that at a zero on thejro axis, the
phase response has a step discontinuity
FIG, 8.8 of + 180° for increasing frequency.
Similarly, at a pole on the jw axis, the
phase response is discontinuous by -180°. These observations are
illustrated by the amplitude and phase plot for F(s) in Eq. 8.11 for the
frequency range 0.9 ~ ro ~ 1.3, shown in Fig. 8.9.
With a simple extension of these ideas, we see that if we have a zero at
z = -a ± jw,, where a is very small as compared to ro,, then we will have
a dip in the amplitude characteristic and a rapid change of phase near
w = w,, as seen in Fig. 8.10. Similarly, if there is a pole atp = -a ±jw1,
, with <1 very small, then the amplitude will be peaked and the phase will
decrease rapidly near ro = w1, as seen in Fig. 8.11. A contrasting situation
occurs when we have poles and zeros far away from the jro axis, i.e., a is
large when compared to the frequency range of interest. Then we see that
these poles and zeros contribute little to the shaping of the amplitude and
phase response curves. Their only effect is to scale up or down the overall
amplitude response.
From stability considerations we know that there must be .no poles in
the right half of the s plane. However, transfer functions may have zeros
 Amplitude, phase, and delay 219

10
1

"'-
FIG. 1.9. Amplitude and phase for F(s) in Fig. 8.8.

in the right-half plane. Consider the pole-zero diagrams in Figs. 8.12a

and 8.12b. Both pole-zero configurations have the same poles; the only
difference is that the zeros in (a) are in the left-half plane at s == -1 ± jl,
while the zeros in (b) are the mirror images of the zeros in (a), and are
located at s == +I ±jl. Observe that the amplitude responses of the
two configurations are the same because the lengths of the vectors corre-
spond for both situations. We see that the absolute magnitude of the

.
"Cl
:e"ii.
E
C:
-....

w I .,..-
j
-a.
E
C:
A
0

_r-- w;
"' 0 "'i

51
if"'
0
-- "'i
"'
I
if"'
0
L "'i

FIG. 8.10. Effect of zero very near FIG. 8.11. Effect of pole very near
theJw axis. the/w axis.
220 Network analysis and synthesis
jw jw
X j2 X j2

0 jl jl 0

fT fT
-2 -1 -2 1
o -jl -jl 0

X -j2 X -j2

(a) (b)

FIG. 1.12. (a) Minimum phase function. (b) Nonminimum phase function.

phase of (b) is greater than the phase of (a) for all frequencies. This is
because the zeros in the right-half plane contribute more phase shift
(on an absolute magnitude basis) than their counterparts in the left-half
plane. From this reasoning, we have the following definitions. A system
function with zeros in the left-half plane, or on thejro axis only, is called a
minimum phase function. If the function has one or more zeros in the
right-half plane, it is a nonminimum phase function. In Fig. 8.13, we see
the phase responses of the minimum and nonminimum phase functions
in Figs. 8.12a and 8.12b.
Let us next consider the pole-zero diagram in Fig. 8.14. Observe that
the zeros in the right-half plane are mirror images of the poles in the
100

0
/
\
--- ✓I
M'.r
1mmum phase

J-100 \ \
I -200
\
\.

-300
-360
0 2
' ...... ~Nonminimum phase

4
I

6
-
8
I

10 12
"'-
FIG. 8.13. Comparison of minimum and nonminimum phase functions.
 Amplitude, phase, and delay 221
left-half plane. Consequently, the vector
drawn from a pole to any point w1 on j2
the jw axis is identical in magnitude with IC• 1
the vector drawn from its mirror image x jl 0
to w 1 • It is apparent that the amplitude
response must be constant for all frequen-
-2 -1 1 2
cies. The phase response, however, is
anything but constant, as seen from the ><-jl 0

amplitude and phase response curves

given in Fig. 8.15 for the pole-zero config- FIG. 1.14. All-pass function.
uration in Fig. 8.14.
A system function whose poles are only in the left-half plane and whose
zeros are mirror images of the poles about the jw axis is called an all-pass
function. The networks which have all-pass response characteristics are
often used to correct for phase distortion in a transmission system.

8.2 BODE PLOTS

In this section we will turn our attention to semilogarithmic plots of

amplitude and phase versus frequency. These plots are commonly known
as Bode plots. Consider the system function

H(s) = N(s) (8.12)

D(s)

I :IL.__ o
. .11_ ____._1_ _.....1...__ __._,_ __._(
2 4 6 a ro
"'-
+180

\
Ii 0

\~
if. -180
~ r--...__
-360
0 2 4 6 8 10 12
"'--
FIG. I.IS. Amplitude and phase of all-pass function in Fig. 8.14.
 222 Network analysis and synthesis

IKl>I
Ki---------------
~ _ _ _ _ _ _ _ _ _ _ __
)!
J logw-
~ K1 ..,__ _ _ _ _ _ _ __
IKl<I
Magnitude
(a)

0
flt
C
K positive
.!!
~ -~
2
Phase
'i
'-
-e- Kneptive
-r
logw-
(b)

FIG. 8.16. Magnitude and phase of cons&nt.

We know that the amplitude response is

M(w) = IHUw)I = INUc.o)I (8.13)

IDUc.o)I
If we express the amplitude in terms of decibels, we have
20 log M(c.o) = 20 log IN(jc.o)I - 20 log ID(jw)I (8.14)
In factored form both N(s) and D(s) are made up of four kinds of factors:
(a) a constant, K
(b) a root at the origin, s
(c) a simple real root, s + ex
· (d) a complex pair of roots, s2 + 2cxs + cx + /JI
2

To understand the nature ~flog-amplitude plots, we need only examine the

amplitude response of the four kinds of factors just cited. If these factors
are in the numerator, their magnitudes in decibels carry positive signs.
If these factors belong to the denominator, their magnitudes in decibels
carry negative signs. Let us begin with case (a).
(a) The factor K. For the constant K, the db loss (or gain) is

20logK= K,. (8.15)

 Amplitude, phase, an~ delay 223
The constant K1 is either negative if IKI < 1, or positive if IKI > 1. The
phase response is either 7.erO or 180° depending on whether K is positive or
negative. The Bode plots showing the magnitude and phase of a constant
are given in Fig. 8.16.
(b) The factor s. The loss (gain) in decibels associated with a pole
(7.ero) at the origin is ±20 log co. Thus the plot of magnitude in decibels
versus frequency in semilog coordinates is a straight line with slope of
±20 db/decade or ±6 db/octave. From the Bode plots in Fig. 8.17, we
see that the. zero loss point (in decibels) is at co - 1, and the phase is
constant for all co.
(c) The factor s + ot. For convenience, let us set ot - 1. Then the
magnitude is ±20log ljco + 11 - ±20log(co• + 1)~ (8.16)
as shown in Fig. 8.18a. The phase is
Arg(jco + 1)±1 - ±tan-1 co (8.17)
as shown in Fig. 8.18b.
A straight-line approximation of the actual magnitude versus frequency
curve can be obtained from examining the asymptotic behavior of the
factorjco + I. For co« l, the low-frequency asymptote is
20 log jjco + 11 l... ,u"' 20 log 1 - 0 db (8.18)
For co» 1, the high-frequency asymptote is
20 log jjco + 111 ► 1 "' 20 log co db (8.19)

.0
+20
~+10
•
J
-----. -------- ------------
¼!..
..s - ----
-----
0
..,
i-10
:IE
-20 ----
+.!:
2 • ..J
"'C
.!!
I
:Ii 1
T:V.
if -t
0.1 0.2 03 0.5 0.7 1.0 2.0 3.0 5.0 7.0 10.0
Frequency, "1
FIG. 1.17. Magnitude and phase of pole or :r.ero at s = 0.
22◄ Network analysis and synthesis

@ 31----+------:aof""--+--...--+------I
.g
J Or---_1
-31----t---~+..:,----'l\o:--t--------i

-12..__ _ _..,___ _ _~ - - - - - ' , , - - - - - -

0.25 0.5 1 2
"'-(a)

~r-----"T"""----.------,------,

"'C
101-----+-----+------+------t
l
f

-.!:0._.2_5_ _ _0
....5_ _ _.......,l_ _ _ _.._2_ _ ___,4
2

"'--
(b)

FIG. 8.18. Magnitude and phase ofaunple real pole or :zero.

 Amplitude, phase, and delay 225
TABLE 8.2
Actual Straight-Line
Magnitude, Approximation, Error,
Frequency db db db

w -l 2 octaves below ::1::0.3 0 ::1::0.3

w =l octave below ::1::1 0 ::1::1
(I)= 1 break frequency ::1::3 0 ::1::3
(I) =2 octave above ::1::7 ::1::6 ::1::1
(I) -4 2 octaves above ::1::12.3 ::1::12 ::1::0.3

which, as we saw in (b), has a slope of 20 db/decade or 6 db/octave. The

low- and high-frequency asymptotes meet at w == 1, which we designate
as the breakfrequency or cutofffrequency. The straight-line approximation
is shown by the dashed lines in Fig. 8.18a. Table 8.2 shows the comparison
between the actual magnitude versus the straight-line approximation. We
see that the maximum error is at the break: frequency w == 1, or in un-
normalized form: w == ac.
For quick estimates of magnitude response, the straight-line approxi-
mation is an invaluable visual aid. An important example of the use of
these straight-line approximations is in the design of linear control
systems.
(d) Complex conjugate roots. For complex conjugate roots, it is con-
venient to adopt standard symbols so that we can use the universal curves
that result therefrom. We describe the conjugate pole (zero) pair in terms
of a magnitude w0 and an angle 8 measured from the negative real axis,
as shown in Fig. 8.19. Explicitly, the parameters that describe the pole
(zero) positions are w., which we call the undampedfrequency ofoscillation,
and , == cos 8, known as the damping
factor. If the pole (zero) pair is P
given in terms of its real and imaginary
parts,
P1,1 = -ac ±j{J (8.20)
ac and {J are related to , and co0 by the
following:
-r(IIQ 0
at == w0 cos 8 == w0 C "
(8.21) )(

p == w0 sin 8 == w0 .J1 -'9

Returning to the definition of the FIG.I.It. Polelocation in terms of
damping factor, { == cos 6, we see that , and m,.
226 Network analysts and synthesis
the closer the angle Ois to 'IT/2, the smaller is the damping factor. When
the angle O is nearly zero degrees, the damping factor is nearly unity.
To examine the Bode plots for the conjugate pole (zero) pair, let us set
w0 == 1 for convenience. The magnitude is then
±20 log 11 - w1 + /2lwl == ±20 log [(1 - w8) 1 + 4'8w1 ]½' (8.22)

and the phase is ,1.1

'l'\w
)
== tan-1 -2tw
- -8 (8.23)
1 - ru
If we examine the low- and high-frequency asymptotes of the magnitude,
we see that the low-frequency asymptote is 0 decibels; the high-frequency
asymptote (for ru » 1) is ±40 log ru, which is a straight line of 40
db/decade or 12 db/octave slope. The damping factor l plays a significant
part in the closeness of the straight-line approximation, however. In
Fig. 8.20 the asymptotic approximation for a pair of conjugate poles
(ru 0 == 1) is indicated by the dashed line. Curves showing the magnitude
for l == 0.1, l = 0.6, and l = 1.0 are given by the solid lines. We see
that only for l !:::= 0.6 is the straight-line approximation a close one.
Universal curves for magnitude and phase are plotted in Figs. 8.21 and
8.22 for the frequency normalized function

G(s) = l (8.24)
(s/ru0) 8 + 2l(s/ru0) + 1
We see that the phase response, as viewed from a semilog scale, is an odd
function about ru/ru0 == 1. The phase at ru = ru0 is -90° or -'IT/2 radians.
+15

+10

+5
..Q
"O

l 0

I -5

-10

-15
0.1 0.2 0.5 1.0 2.0 5.0 10.0
Frequency, "' -
FIG. 8.20. Magnitude versus frequency for second,-order pole.
 Amplitude, phase, and delay 221

~
... , - :g.....I
..,-o-

~~
d I
....d ~ I ._
~ .....I
0

.....I .....I \ L
\ L
-
lj
,--- y Y?
........__

-----
~, ~

~
'l
I

'\. \ ~
I L /
IX
I \ I ;._
' I/ 9
J I

~\\ I A ~I
~\ I~ ....
l / !
I/

....
!ii ....on ~
,n
+
0 on
I ~ ....
,nd

+ + + qp , ll P>/'IOI. I I
 ~
0
z
la
Phase angle vs. frequency of !.,
~
-40 I · -·1 I .......,-...:::17 Of'-..::: 1"v+..: 'I. U I G(a)• 1
<•l"'oJ2+ 2t<•/"'GJ + .1 !:
e..
~
iii

I -80
~
A.
~
,:,
-:
a:r
5 I
~ -120 I I I I I I I I I I',\~'< '"< I I I I I I I I I
iii

- 160 I I I I I I I I I I \.: ' - ,......._.,,..__ ---==-- -L._---..., -k.::: I =-; --4...::.:1 I I

-200,___ _ _ ___.__ _ _...__ _,__ _.__.__......._............._.__ _ _ _ _...__ __.__ _,_____._ _,.____.____.__,,_,J

0.1 0.2 0.5 1.0 2.0 5.0 10.0
"'l"'o
FIG. 8.22. Phase versus frequency for second-order pole.
 Amplitude, phase, and delay 229
For a conjugate pair of zeros, we need only reverse the signs on the scales
of the magnitude and phase curves.
Evmple &1. Using Bode plot asymptotes, let us construct the magnitude
versus frequency curve for

~+l
(;(,) - ( ) ':";'' ) (8.2S)
(16 X 1()& + 1~ + l
We see there are two first-order break freq~ at cu - 0 and cu • 50. In
addition, there is a second-order break frequency at "' - 400. With a quick
calculation we find that C - 0.2 for the second-order factor. The asymptotes
are shown in Fig. 8.23. The magnitude and phase plots are given in Fig. 8.24
through a microfiJm plot computer program.

=8

I '°'""-

FIG. 8.23. Asymptotes for G(.r) in Eq. 8.25.

L3 SINGLE-TUNED CIRCUITS

We will now study a class of circuits whose system functions can be

described by a pair of conjugate poles. These circuits are called single-
tuned circuits because they only need two reactive elements-an inductor
and a capacitor._ The undamped frequency of oscillation of the circuit is
then ro0 == (LC)-l-i. An example of a single-tuned circuit is the R-1,C
circuit in Fig. 8.25, whose voltage-ratio transfer function is

H(s) = Vo(s) == 1/sC = 1/LC (8.26)

Vi(s) R + sL + 1/sC s1 + (R/L)s + 1/LC
230 Network analysis and synthesis

FIG. 1.24a. Magnitude of G(s) in Eq. 8.25.

FIG. 8.246. Phase of G(s) in Eq. 8.25.

+
1
•C

FIG. 1.25. Single-tuned circuit.

 Amplitude, phase, and delay 231
jw

K•a2+/J2
j{J

-a "

)(------ -jfJ

FIG.1.26

The poles of H(s) are

Pu = - .!._
' 2L
± i.(...i.
2LC lJ
- R')~ = -ex :I: JP (8.27)

where we assume that (RS/Lr,< (4/LC). In terms of ex.and Pin Eq. 8.27,
H(s) is
H(s) = ex•+ pt (8.28)
(s + ex + J/J)(s + ex - J/J)
From the pole-zero diagram of H(s) shown in Fig. 8.26, we will determine ·
the amplitude response IH(jw)I. Let us denote the vectors from the poles
to the Jw axis as IM11and IM11as seen in Fig. 8.26. We can then write

(8.29)
where K == ex• + pt and
IM1I == [ex1 + (w + fl)']~
(8.30)
IM,I == [ex•+ (w - /J)1J~
In characterizing the amplitude response, the point w == a>max, at which
IH(jw)I is maximum, is highly significant from both the analysis and design
aspects. Since IH(jw)I is always positive, the point at which IH(jw)l1 is
maximum corresponds exactly to the point at which IHUw)t is maximum.
Since IHU<»)l1 can be written as
232 Network analysis and synthesis

IHUw)l 1 - (at•+ (J8)•

- [«1 + (w + P)1][at1 + (w - P>9J
(8.31)
(at•+ fJ8>1
- ------''------''----'-----
w' + 2w8(«1 - fJ8) + (at1 + {J8)1
we can find Wmax by taking the derivative of IH(jw)l 1 with respect tow• and
setting the result equal to zero. Thus we have

dlHOw)l1 (at1 + {J8)1 [2ro1 + 2(at1 - {J8)]

= - (8.32)
dw 1 [ro' + 2w8{at1 - fJ8) + (at1 + {J8)1 ] 1

From the equation d IHUw)l

1
=0 (8.33)
dw 1
we determine ~ = {l8-at• (8.34)
Expressed in terms of the natural frequency of oscillation o,0 and the
damping factor C, ro!ax is

~ = (w0✓ 1 - r)1 - (Cw0) 1 = ro01(1 - 2r) (8.35)

Since Wmax must always be real, the condition for Wmax to exist, i.e., the
condition for IH(jw)I to possess a maximum, is given by the equation
2{1 ~· 1 (8.36)
so that C~ 0.707. Since C= cos 8, Wmax does not exist for 8 < 45°.
When 8 = 45°, we have the limiting case for which Wmax exists. In this
case, C== 0.707 and the real and imaginary parts of the poles have the same
magnitude, i.e., at = p, or
(8.37)

We see from Eq. 8.34 that, when at = /J, then Wmax = 0. This is the lowest
frequency at which Wmax may be located. For C> 0.707, or

(8.38)
Wmax 1s imaginary; it therefore does not exist. To summarize, the key
point in this analysis is that the imaginary part of the pole must be greater
or equal to the real part of the pole in order for Wmax to exist. Interpreted
graphically, if we draw a circle in the s plane with the center at -at and the
radius equal to fl, then the circle must intersect the jro ax.is in order for Wmax
to exist, as seen in Fig. 8.27. Moreover, the point at which the circle inter-
sects the positive jw ax.is is a>maz. This is readily seen from the triangle
 Amplitude, phase, and delay 233
jw

tT

FIG. 1.27. Peaking circle.

with sides«, {J, <omax in Fig. 8.27. By the Pythagorean theorem, we find
that
=
ro:iax p•-«· (8.39)
The circle described in Fig. 8.27 is called the peaking circle. When « = /J,
the peaking circle intersects the jw axis at w = 0, as seen in Fig. 8.28a.
When « > /J, the circle does not intersect the jw axis at all (Fig. 8.28b);
therefore, <omax cannot exist. When the imaginary part of the pole is
much greater than the real part, i.e., when /J »ex, then the circle intersects
the jw axis at approximately w = w0 , the natural frequency of oscillation
of the circuit (Fig. 8.28c).
A figure of merit often used in describing the "peaking" of a tuned
circuit is the circuit Q, which is defined in pole-zero notation as

4 1 1
(8.40)
Q=2,=2cos6

From this definition, we see that poles near the jw axis ( t small) represent
high-Q systems, as given in Fig. 8.28c, and poles far removed from the jw
axis represent low-Q circuits (Fig. 8.28a}. Although the Q of the circuit
given by the pole-zero plot of Fig. 8.28b is theoretically defined, it has no
practical significance because the circuit does not possess a maximum
point in its amplitude.
By means of the peaking circle, we can also determine the half-power
point, which is the frequency w 0 at which the amplitude response is
IH(jwa)I = 0.707 IH(jt:omax}I.
23-4 Network analysis and synthesis

j"'
j"' j"'

'I'
I i j{J j{J f
I I I
I
I I I
I I I
I I I
-a, I
"'max -a,
I
-a1
I
I I I
I I
I I
I I I

* Jc -j{J -j{J~

(al (bl
FIG. 8.21. Examples of peaking circles. (a) ex = P, Wmax = O. (b) IX > P, Wmax
»
undetermined. (c) p ex, Wmax ~ p.

We will now describe a method to obtain w 0 by geometrical construction.

Consider the triangle in Fig. 8.29, whose vertices are the poles {pi, Pi*}
and a point w, on the jw axis. The area of the triangle is

Area (&PJ/11 •wJ == P« (8.41)

In terms of the vectors IM11and IM1 1from the poles tow,, the area can also
be expressed as

(8.42)

where V' is the angle at w,, as seen in the figure. From Eqs. 8.41 and 8.42,
we see that the product IM1I IM1l is equal
j"' to

r
IM1I IM1I == 2_P« (8.43)
SID V'
Since the amplitude response is

(T
(8.44)

where K is a constant, then

(8.45)
FIG. 1.19
 Amplitude, phase, and delay 235
For a given pole pair {pi, Pt•} the parameters p, «, and Kare prespecified.
Therefore, we have derived IH(jw)I in terms of a single variable parameter,
the angle tp. When the angle tp == 1r/2 rad, then sin tp == 1, w, == Wmax, and,

IHUeomu)I == K (8.46)
2p«
When tp == 1r/4 rad, then sin tp == 0.707 and
IH(jwJI == 0. 707 IH(jromax)I
so that w, = w0 • Let us consider now a geometric construction to obtain
w 0 • Let us first draw the peaking circle as shown in Fig. 8.30. We will
denote by A the point at which the peaking circle intersects the positive
real axis. Now we draw a second circle with its center at A, and its radius
equal to AB, the distance from A to either one of the poles, as seen in
Fig. 8.30. The point where this second circle intersects the Jw axis is w0 •
Th~ reason is that, at this point, the inscribed angle is 'Pt == 1r/4 because it
is equal to one-half the intercepted arc, which, by construction, is 1r/2.
When Wmax = 0, the half-power point we is also called the half-power
bandwidth of the tuned circuit. In Fig. 8.31a the half-power point :s given
when Wmax == 0. For a high-Q circuit, where Wmax ~ w0 , the amplitude is
highly peaked at w == Wmax, as shown in Fig. 8.31b. In this case, if
IH(j0)I < 0.707 IH(ja>max}I, there are two half-power points we1 and we,
jc.J

FIG. LJO. Geometric construction to obtain half-power point.

236 Network analysis and synthesis

0.707 IBmul

IBol
(II

(a} (b}
FIG. 8.31. (a) Low-Q circuit response. (b) High-Q circuit response.

about the point a>max, as seen in Fig. 8.31b. By the construction process
just described, we obtain the upper half-power point w 01 • It can be shown1
that the point a>max is the geometric mean of w01 and w01, that is,

(8.47)
As a result, the lower half-power point is

(8.48)

The bandwidth of the system for a high-Q circuit of this type is described by

BW - <»oI - <»o1 (8.49)

In design applications these high-Q circuits are used as narrow bandpass

filters.
Finally, there are certain aspects of the phase response ofhigh-Q circuits
that are readily apparent. In Fig. 8.32 we see several steps in the process
of obtaining the phase response. The phase shift at w =;: 0 is 0, as seen
from Fig. 8.32a. At w - oo, the phase shift is -11 rad, as shown in part (c)
of the figure. Finally, in th~ neighborhood of w = w0 ~ a>max, the phase
shift resulting from the pole in the lower half plane is approximately
-81 - -11/2 (Fig. 8.32b). The change in phase in this region is controlled
in large by the pole p 1 • It is readily seen that the phase response in the
region of w0 has the greatest negative slope, as seen from a typical phase
response of a high-Q circuit shown in Fig. 8.33.

1 Seo for example, F. E. Terman. Electronic and Radio £nKlnffring, McGraw-Hill

Book Company, New York, 1953.

 Amplitude, phase, and delay ll7
jw

-tT 0

~ ~ w
FIG. -.n. Several steps in obtaining phase
(b) m, ~ m.. (c) m, - ex.
response for high-Q circuit. (a) m, - O.

Finally, as an example to illustrate our discussion of single-tuned

circuits, let us find the amplitude response for the system function

(8.50)

Now we determine the maximum and half-power points a>mas: and m0 ,

and also the amplitudes IH(jeomax)I and IH(jm 0 )1. In factored form, H(s) is
34
H(s) - (8.51)
(s + 3 + jS)(s + 3 - jS)
and the poles of H(s) are shown in Fig. 8.34. We next draw the peaking
circle with the center at s == -3 and the radius equal to S. At the point
where the circle intersects the j.a, axis, we see that a>mas: - 4. To check this
result, the equation m:.S: - (JI - a.• gives
a>mas: - (S1 - 31)'-' - 4 (8.52)
0
"'-
J -~·
-lSO-

FIG. L33. Phase ~ of high-Q c:m:uit.

238 Network analysis and synthesis
The amplitude IH(jromax)I is then

4
IHU )I = I (3 + j9~;3 - jl) I
= 34 = 1.133 (8.53)
30
The point A at which the peaking circle
--+---+-----=-+------'1'--'A'-_ intersects the positive real axis is located
2.0 <T at s = 2.0. With the center at A, we
draw a circle of radius AB (equal to
s✓2 in this case). At the point C where
this new circle intersects the jro axis, we
have ro 0 • By measurement, we find
Wo ~ 6.78 (8.54)
Let us check this result. Referring to
FIG. 8.34. Peaking circle construe-
tion example. Fig. 8.34, we know that the line segment
AB is of length 5./i,; it follows that AC
is also s✓2 units long. The line segment AO is of length

Then w 0 is given as
AO =5- 3 = 2 units (8.55)

w 0 == ✓(AC)1 - (AO)1 = .J'46 = 6.782 (8.56)

Finally, we obtain jH(jro0 )l as
34
IH(j6.782)1 =✓ = 0.802 (8.57)
(34 - 46)1 + (6vr.;;
46)1
which is precisely 0.707 IH(jromax)I.

8.4 DOUBLE-TUNED CIRCUITS

In Section 8.3, we studied the frequency response for a pair of conjugate

poles. Now we turn our attention to the amplitude response of two pairs
of conjugate poles in a high-Q situation. The circuit we analyze here is the
double-tuned or stagger-tuned circuit given in Fig. 8.35. We will consider
R C C
+

FIG. 8.35. Double-tuned circuit.

 Amplitude, phase, and delay 239
the special case when the R, L, and C elements in the primary circuit are
equal in value to their counterparts in the secondary. Since the primary
and secondary inductances are equal, the mutual inductance is
M=KL (8.58)
In this analysis we assume the coefficient of coupling K to be a variable
parameter. Let us determine the amplitude response for the voltage-ratio
transfer function V,l._s)/V1(s). From the mesh equations

V1(s) = ( R + sL + s~) Ji(s) - sM I.(s)

(8.59)
O = -sM Ji(s) + (R + sL + s~) J.(s)
we readily determine
1 1
H(s) _ V.(s) _ s RM/L (8.60)
- Vi(s) - [s1 + (R/L)s + 1/LC]1 - s'K1
Using tuned-circuit notation, we set
2{ru0 = -R
. L
(8.61)
1
=-
LC
H(s) can then be written as
H( ) = s'RM/1!
s (1 - K')(s• + 2{ruo s + _!!!£_) (s• + 2{ruo s + _!!!£_)
1+K 1+K 1-K 1-K
(8.62)

Ifwe set A= RM (8.63)

/!(1 - K8)

(8.64)

where
(8.65)

Let us restrict our analysis to a high-Q circuit so that r«

1. Furthermore,
let us assume that the circuit is loosely coupled so that K «
1. Under
l-40 NetwQrk analysis and synthesis
these assumptions, we can approximate the pole locations by discarding
the terms involving t' under the radicals in Eq.. 8.6S. Then the poles
{si, s1 *} can be given approximately as

{si, s1 *} ~ -tw. :I: Jw0 ( 1 - !) (8.66)

Similarly, {s1 , s1 *} can be given as

{s1, s1*} ~ -tw0 :I: Jw0 ( 1 + !) (8.67)

The pole-zero diagram of H(s) is given in Fig. 8.36. The real part of the
poles -{w0 is greatly enlarged in comparison to the imaginary parts for

llo

_,,.

i
FIG. 8.:16. Poles and zeros of a double-tuned circuit.
 Amplitude, phase, and delay 2-41
clarity purposes. Note that we have a triple zero at the origin. In terms
of the vectors in Fig. 8.36, the amplitude response is
1
Hu"·')I _ 1Mol
A (8 68
I .., - IM1I IMal IMal IM.1 " )
Since the circuit is high-Qin the vicinity of ro == ro0 , we have
IM,I ~ IM.I ~ 2 IMol ~ 2ro. (8.69)
so that in the neighborhood of ro0
."1

(8.70)

It is evident that the amplitude response

of IH(jro)I in the neighborhood of ro0
depends only upon the pair of vectors
IMil and IMal- The double-tuned prob-
lem has thus been reduced to a single-
tuned problem in the neighborhood of
ro0• Consequently, we can use all the
results on the peaking circle that were
derived in Section 8.3. Let us draw a
peaking circle with the center at
s =;= -{ro0 + jro0 (8.71)
and with a radius equal to roef(/2, as ::!t
~~ circle for cloublo-
shown in Fig. 8.37. The inscribed
angle 'I' then determines the location of the maxima and half-power points
of the response. Without going into the derivation, the amplitude response
can be expressed as a function of 'I' according to the equation

IHUro)I ==
Aro0 sin 'I' == sin tp (S.7l)
4roo K(tro0) 2(1 - K')
Referring to the peaking circle in Fig. 8.37, let us consider the following
situations:
1. ro0 { > roef(/2: In this case, the peaking circle never intersects the jw
axis; 'I' is always less than 11/2 (Fig. 8.38a), and the amplitude response
never attains the theoretical maximum

Hmax == 2(l ~ K') (8.73)

as seen by the curve labeled (a) in Fig. 8.39. In this case, K < 2{, and the
circuit is said to be ""'1ercoupled.
242 Network analysis and synthesis
jw

(a) (b) (c)

FIG. 8.38. (a) Undercoupling. (b) Critical coupling. (c) Overcoupling.

2. ru,, = w,,K/2: Here the peaking circle intersects the jro axis at a
single point ro = ro0 (Fig. 8.38b). At ro0 , the amplitude is equal to HDJAs:
in Eq. 8.73. In this case t = K/2, and we have critical coupling.
3. ro0 , < ro,,K/2: The peaking circle intersects the jro axis at two points.
ro1 and ro1, as seen in Fig. 8.38c. The intersecting points are given by the
equation
(8.74)

Consequently, the amplitude response attains the theoretical maximum

HDJAs: at two points, as shown by curve (c) in Fig. 8.39. In this situation,
the circuit is said to be overcoupled.
Note that in Fig. 8.39 the undercoupled and critically coupled curves
have their maximum points at ro0 • The overcoupled curve, however, is

IHI

FIG. 8.39. (a) Unden:oupled case. (b) Critically coupled case. (c) Overcoupled case.
 Amplitude, phase, and delay 243

"'
FIG. 1.40. Half-power points of oven:oupled circuit.

maximum at w 1 and w 1 • In the case of overcoupling and critical coupling,

we can determine the half-power points by using the geometrical con-
struction method given in Section 8.3. Observe that there are two
half-power points w01 and w01, as shown in the overcoupled curve in
Fig. 8.40. The bandwidth of the circuit is then
(8.75)
Example 11.2. The voltage-ratio transfer function of a double-tuned circuit is
given as
As'
H(s) = (s + 2 + jlOCPh + 2 - jlOOXs + 2 + jl06)(s + 2 - jl06) (3,76)

From H(s), let us determine the following: (a) the maximum points w1 max and
w1 max; (b) the 3 db bandwidth BW; (c)thedampingfactor C; (d) the coefficient
of coupling K; (e) the gain constant A; and (/) the maximum of the amplitude
response Hmu.
Solution. (a) The natural frequency of oscillation w0 is taken to be approxi-
mately halfway between the two poles, that is, w0 = 103 radians. In the neigh-
borhood of mo, we draw the poles s = -2 + jlOO and s = -2 + jl06, as
shown in Fig. 8.41. From the peaking circle centered atthe points = -2 + jw.,
shown in Fig. 8.41, we obtain
w1 max - w0 = V 31 - 21 = 2.236 radians (8.77)

so that w 1 max= w0 + 2.236 ... 105.236 radians

(8.78)
w 1 max = w0 - 2.236 = 100.764 radians
(b) Next we draw a circle centered at s = I + jruo with radius 3V2. Where
this circle intersects the jw axis, we have w01 so that

w01 - w0 =- ✓(3Vi)I - 1 =- 4.123 radians (8.79)

2-44 Network an_alysls and synthesis

FIG. 8.41. Peaking circle for Example 8.2.

The 3 db bandwidth is then

BW - 2(co01 - co.)• 8.246 radians (8.80)

(c) The damping factor Cis obtained from the real part of the poles Cco0 =- 2,
from which we obtain
2
C- 103 - 0.0194 (8.81)

(d) The coefficient of coupling K is obtained from the radius of the peaking
circle, which is ·
cool(
-=3 (8.82)
2
6
We thus have K =- =0.0582 (8.83)
coo
(e) The gain constant A is equal to
2Ccoof( 2(2)(0.0582)
0 2328 (8.84)
A - 1 - K1 - 1 - (0.0582)8 ... ·
(f) Finally, the maximum amplitude H...,. is
1
H • .,, = 2(l _ Kl) - 0.5009 (8.85)
 Amplitude, phase, and delay 2◄5

LS ON POLES AND ZEROS AND TIME DELAY

What is time delay? How do we relate it to frequency response? We

will attempt to answer these questions in this section. First consider the
transfer function of pure delay
H(s) = c•'l! (8.86)
For a system described by Eq. 8.86, any excitation e(t} produces an
identical response signal e(t - T), which is delayed by time Twith respect
to the excitation. This is shown by the Laplace transform relationship,
R(_s} = t[e(t - T)] = c•'l! t[e(t)] (8.87)
Let us examine the amplitude and phase response of the pure delay.
From the equation
HUw) = e-1er (8.88)
we obtain the amplitude response
IH(Jw)I =1 (8.89)
and the phase response tf,(w) = -wT (8.90)
We see that the delay T is equal to minus the derivative of the phase
response, that is,
T =- d4,(w) (8.91)
dw
The magnitude, phase and delay characteristics of H(jw) = c 1•'1! are
given in Fig. 8.42a, b, and c.
If we define delay as in Eq. 8.91, we can readily deduce that for the
response to be nearly identical to the excitation, the system amplitude
response should be constant, and its phase response should be linear over
the frequency range of interest. If the phase is not linear, we have what is
known as delay distortion. To visualize delay distortion more clearly,
A(w)

0 ·w 0 (IJ

00 00
FIG. LG. Amplitude, phase, and delay of ideal delay function. (a) Amplitude.
(b) Phase. (c) Delay.
2-46 Network analysis and synthesis
we recall from Fourier analysis that any signal is made up of different
frequency components. An ideal transmission system should delay each
frequency component equally. If the frequency components are delayed
by different amounts, the reconstruction of the output signal from its
Fourier components would produce a signal of different shape as the input.
For pulse applications, delay distortion is an essential design considera-
tion.
Let us next examine how we relate delay, or envelope delay (as it is
sometimes called) to the poles and zeros of a transfer function. For any
transfer function
...
II (s - zi)
H(s) = •,.;: ·-1"'-- - - (8.92)
n
II (s - P,>
i=l

with zeros at zi= -a, ± jwi and poles at p 1 = -a1 ± jw 1, the phase for
real frequencies is
~
M) =ktan
'f'\w
~w±~ ~ta~w±~
--~-k n --- (8.93)
,-1 a, 1-1 a1
Envelope delay is

- drf,(.w)
dw
=- I a/+ (wa, ±
i-1 w,)1
+!
J=1
a,
a/+ (w ± w1)
1
(8.94)

We see that the shapes of the delay versus frequency characteristic are
the same for all poles and zeros. The zeros contribute "negative" delay;
the poles, positive delay. However, linear physical systems do not have
transfer functions with zeros alone. The inductor H(s) = Ls is the only
=
exception. Its phase is t/,(_w) 1r/2; thus the delay is zero.
Now let us consider the delay due to one singularity, for example, a
pole at p 0 = - <10 + jw0• The delay due to the one pole is

ll. (w) - <1o (8.95)

210 -
<1o •+(w - Wo)lll

The following points are pertinent:

1. The maximum delay due to this pole is

ll..,. = -1 (8.96)
<1o

and occurs at w == w 0• The delay versus frequency curve is symmetric

about w = w 0 •
 Amplitude, phase, and delay 2-47

jw

-IT -uo

FIG. I.G. Graphic construction to obtain delay bandwidth.

2. The frequency at which the delay is half the maximum, or I/2a0 , is

(8.97)
3. The "effective delay bandwidth" is then ro0 - <10 < w < w0 + <10 or
simply 2a0 • The upper and lower half-bandwidth points can be obtained
graphically by drawing a circle with center at w = w0 and radius <10 •
The intersections of the circle with the jw axis are the half-bandwidth
points, as shown in Fig. 8.43.
4. The product of the maximum delay and delay bandwidth is always
2. Thus, if we wish to obtain large delay by placing zeros or poles near
the jw axis, the effective delay bandwidth is then very narrow.
5. The delay of an all-pass function is twice the delay due to the poles
alone.
The delay versus frequency curve for the pole is shown in Fig. 8.44.
We see that the delay-bandwidth concept is only useful for a rough
approximation, since the delay versus frequency characteristic only falls
as 1/w. To calculate the delay versus frequency characteristic for a
transfer function with a number of poles and zeros, it is convenient to
obtain the delay curves for the individual singularities and then to add the
separate delays. It is not as desirable to obtain the total phase response
and then differentiate numerically.
Finally, it should be pointed out that envelope delay only has meaning
when the phase response goes through the origin. If it does not, there is
a frequency-shift component in addition to the delay that is hard to
account for analytically.
248 Network analysis and synthesis

A.(111)

"'

FIG. 1.44. Plot of delay versus frequency.

Problems
8.1 Find the poles and zeros of the impedances of the following networks

~1
and plot on a scaled s plane.

4h to
Z(•)-
o -
(aJ

~
Z(•J::i
(c)
PROB.I.I

8.2 The circuit shown in the figure is a shunt peakinK circuit often used in
video amplifiers.
(a) Show that the admittance Y(s) is of the form
Y(s) = K(s - si)(s - sJ
(s -sJ
Express s 1, s., and s 8 in terms of R, L, and C.
 Amplitude, phase, and delay 249
(b) When .r1 =- -10 +JlO', s1 - -10 -JIO', and Y(JO) - 10- mhos, find 1

the values of R. L, and C and determine the numerical value of .r1 •

L
Y(•>- C

PltOB.11.2

8.3 Find the amplitude and phase response for the following functions and
sketch.
K s
(a) F(s) - - - (b) F(s) - - -
s +K s+K
s
(c) F(s) == .,.
+mo1
Note that K and t»o are positive quantities.
8.4 Given the function
G(Jw) = A(w) + JB(w)
C(m) + JD(m)

determine the amplitude and phase of G(Jm) in terms of A, B, C, D. Show that

the amplitude function is even and the phase function is odd.
8.5 By means of the vector method, sketch the amplitude and phase response
for
s +o.s s-1
(a) F(s) =- s(.r + to) (c) F(s) ==--
s +1

s s
(b) F(.r) = .,a + 2s + 2 (d) F(s) = .,a _ 2s +2
s+l .r8-2s+S
(e) F(s) =-- (g) F(s) =- (s + 2)(s + 1)
s -1

s1 +2s+S
(h) F(s) = (.r + 2)(.r + l)

8.6 Plot on semilog paper the Bode plots of magnitude and phase for

(a) F( ) = 100(1 + 0.Ss)

s s(s + 2)
(b) F(s) ... SO(l + 0.02Ss)(l + 0.ls)
(1 + 0.0Ss)(l + 0.0ls)
250 Network analysis and synthesis
8.7 Plot on semilog paper the Bode plots of magnitude and phase for
1000s
(a)
F(s) = (1 + 0.002sX1 + S · 10-5s + 10-8s2)
200(1 + 0.05s)
(b)
F(s) = (1 + 0.02sXl + 4 · 10-1s + lo-'s9)
8.8 For the function ,
s
F(s) = s• + 2s + S
determine c0mu:, IF(jcomu:)I, the half-power point co0 , and IF(ico0 )1. Sketch the
amplitude and phase response.
8.9 For the circuit shown, determine the current ratio Ir,/111 and find: (a)
the point '°mm where its amplitude is maximum; (b) the half-power point coo;
(c) the point '°v where lh(cov)/J,,(cov)I = 1. Use geometric construction.

PROB. 8.9

8.10 A network function consists of two poles at Pu = r~H<r -•> = -a, ±

Jco,, as given in the figure. Show that the square of the amplitude response
Afll{co) is maximum at co.,.1 = r/~ lcos 281.

(I'

PROB.I.ID

8.11 In connection with Prob. 8.2 plot the poles and zeros of the impedance
function Z(s) = 1/Y(s). Find, approximately, the maximum point of the
amplitude response. In addition, find the bandwidth at the half-power points
and the circuit Q.
 Amplitude, phase, and delay 251
8.12 The pole configuration for a system function H(s) is given in the figure.
From the plot, calculate:
(a) The undamped frequency of oscillation w0
(b) The bandwidth and Q.
jw

r- jlOO
I
I
I
I
I
I
I
I
-51 O fT
I
I
I
I
I
I
I
i!c -jlOO

PROB, 8.12

8.13 In connection with Prob. 8.10 determine the ratio M'{_wau.)/Ml(O).

8.14 Determine the amplitude and phase response for the admittance Y(s)
of the circuit shown. Is the peaking circle applicable here? What can you say
about the shape of the amplitude response curve in a high-Q situation? Deter-
mine the bandwidth of the circuit and the circuit Q.
20 lh

J*'
'

}'::o•-)_-
__________
PROB, 8,14

8.15 For the overcoupled case of a double-tuned circuit, derive an expression

for the peak-to-valley ratio that is, M(wm.._)/M(wo), where M(·) denotes ampli-
tude. Use the notation in Section 8.4. (Hint: see Prob. 8.13.)
8,16 For the voltage ratio of a double-tuned circuit
As1
H(s) = (s + 4 + jSO)(s + 4 - jSO)(s + 4 + jOO)(s + 4 - jOO)
252 Network analysis and synthesis
Use the peaking circle to determine the maximum and half-power points and
the circuit Q. Find the gain constant A and the coefficient of coupling K.
8.17 For the double-tuned circuit shown, determine the maximum and half-
power points and the circuit Q. Find the gain constant A and the coefficient of
couplingK.

3
l
M .__lo---h___l_O_.J
M== 5 X 10-6 h
PROB. 8.17

8.18 Determine the delay at co =- 0, 1, and 2 for

(a) F(s) = ., +1 2
s-3
(b) F(s) =s + 3
(c) Ff) 3s
\s = (s + l)(s + 2)
(d) F(s) ... .,1 +s+l
2s + S
s(s + 1)
(e)
F(s) == (s + 2)(.,S + 2s + 2)
 chapter 9
Network analysis II

9.1 NETWORK FUNCTIONS

In electric network theory, the word port has a special meaning. A

port may be regarded as a pair of terminals in which the current into
one terminal equals the current out of the other. For the one-port network
shown in Fig. 9.1, I =I'. A one-port network is completely specified
when the voltage-current relationship at the terminals of the port is given.
For example, if V = 10 v and J = 2 amp, then we know that the input
or driving-point impedance of the one-port is
Zin= V = 50 (9.1)
I
Whether the one-port is actually a single 5-0 resistor, two 2.5-0 resistors
in series, or two 10-0 resistors in parallel, is of little importance because
the primary concern is the current-voltage relationship at the port.
Consider the example in which J = 2s + 3 and V = 1 ; then the input
admittance of the one-port is
~=!=2s+3
y ~~
which corresponds to a 2-f capacitor in parallel with a i-O resistor in
its simplest case (Fig. 9.2).

-
I

~ 1=1
~
l'in 2f io
•
I'
FIG. 9.1 FIG. 9.2
253
25-f Network analysis and synthesis

1 2

Two-port
network
2'
FIG. 9.3

Two-port parameters
A general two-port network, shown in Fig. 9.3, has two pairs ofvoltage-
current relationships. The variables are V1 , V 2, Ii, 12 • Two of these are
dependent variables; the other two are independent variables. The number
of possible combinations generated by four variables taken two at a time
is six. Thus there are six possible sets of equations describing a two-port
network. We will discuss the four most useful descriptions here.
The z parameters
A particular set of equations that describe a two-port network are the
z-parameter equations
Vl = z1/1 + zl'l/a
(9.3)
V1 = z2/ 1 +z./1
In these equations the variables V1 and V~ are dependent, and /1' / 2 are
independent. The individual z parameters are defined by

(9.4)

It is observed that all the z parameters have the dimensions of impedance.

Moreover, the individual parameters are specified only when the current in
one of the ports is zero. This corresponds to one of ports being open
circuited, from which the z parameters also derive the name open-circuit

2'
FIG. 9.4
 Network analysis II 255
parameters. Note that z11 relates the current and voltage in the 1-1' port
only; whereas z11 gives the current-voltage relationship for the 2-2' port.
Such parameters are called open-circuit driving-point impedances. On
the other hand, the parameters z111 and z81 relate the voltage in one port to
the current in the other. These are known as (open-circuit) transfer
impedances.
As an example, let us find the open-circuit parameters for the T circuit
in Fig. 9.4. We obtain the z parameters by inspection

(9.5)

Observe that z111 = z81 • When the open-circuit transfer impedances of a

two-port network are equal, the network is reciprocal. It will be shown
later that most passive time-invariant networks are reciprocal.1
Most two-port networks, whether passive or active, can be characterized
by a set of open-circuited parameters. Usually, the network is sufficiently
complicated so that we cannot obtain the z parameters by inspection, as
we did for the T circuit in Fig. 9.4. The question is now, "How do we
obtain the z parameters for any circuit in general?" The procedure is as
follows. We write a set of node equations with the voltages at the ports
V1 and V,i, and other node voltages within the two-port V8 , V4 , ••• , Vi,
as the dependent variables. The independent variables are the currents 11
and la, which we will take to be current sources. We then proceed to
write a set of node equations.

la= nuV1 + nuVi1 + ...

0 = n81 V1 + · · · + · · · (9.6)

1
One important exception is the gyrator discussed later in this chapter.
256 Network analysis and synthesis
where n11 represents the admittance between the ith and jth nodes, that is,

1
n11 = G" + sC" + - (9.7)
sL"
If the circuit is made up of R-L-C elements only, then it is clear that
n 11 = n1,. As a result, the ijth cofactor of the determinant of the node
equations, !:;.11 , must be equal to the jith cofactor, du, that is, /:;,,.11 = I:;.,_.
This result leads directly to the reciprocity condition z11 =
Zill!, as we
shall see.
Returning to the set of node equations in Eq. 9.6, let us solve for V1
and V1 • We obtain

(9.8)

In relating this last set of equations to the defining equations for the z
parameters, it is clear that
/:;,,.11 /:;,,.llll
Z11=- Zi1=-
I:;. I:;.
(9.9)
I:;. I:;.
Zn=_!! .=.....n
I:;. I:;.

Since for a passive network ~ = d 11, it follows that Zn = z11, the

network is then reciprocal.
As an example, let us find the z parameters of the Pi circuit in Fig. 9.5.
First, the node equations are

11 = (Y...t + Y0 )V1 - Y0 V1
(9.10)
11 = -Y0 V1 +(YB+ Y0)V1

1 2

l' 2'
FIG. 9.5
 Network analysis II '157
The determinant for this set of equations is
AY= Y..tYB + Y..tYo + YBYo (9.11)
In terms of A Y, the open-circuit parameters for the Pi circuit are
YB+ Yo
Zi1 = AY
(9.12)
Yo Y..t+ Yo
Zia = AY Zaa = AY
Now let us perform a delta-wye transformation for the circuits in Figs.
9.4 and 9.5. In other words, let us find relationships between the im-
mittances of the two circuits so that they both have the same z parameters.
We readily obtain
Yo
Zia=Zb=-
AY

Zn -
-zb +z• -- Y..t+
AY
Yo (9.13)

We then find z =
0
YB
AY
(9.14)
z = Y..t
C AY
The y parameters
Suppose we were to write a set of mesh equations for the two port in
Fig. 9.3. Then the voltages V1 and V1 would become independent sources,
and the currents / 1 and / 1 would be just two of the dependent mesh
currents. Consider the general set of mesh equations
V1 = m11/1 + m1.I1 + · · · + mu/-.
V1 = m11/ 1 + m,,.11 + · · · + mu,l,:
... + m~J1: (9.15)

... + m1:J1:
where m" represents the sum of the impedances in the ith mesh and m11 is
the common impedance between mesh i and meshj. We note here again
258 Network analysis and synthesis
that for an R-L-C network, m., m,,
= for all i and j. Thus reciprocity
holds.
Solving the set of mesh equations for / 1 and / 2 , we obtain the following
equations.

(9.16)
I = A111 V. + Aaa v;
I Al All
The equations of 9.16 define the short-circuit admittance parameters as

(9.17)
ls= Y11V1 + Y111V11
where y., = A1iA for all i andj.
Let us find they parameters for the bridged-T circuit given in Fig. 9.6.
The mesh equations for the circuit are

V1 = (~ + 1 )11 + la - ~ 1 8

Va=I1 +(~+1)11 +;Ia (9.18)

0 = - ! / 1 +!Ia + 2(! + 1)13

s s s
In straightforward fashion we obtain
A= 2(2s + 1)
s•
A1 1_- An_- 2s11 + 4s
a
+1 (9.19)
5

1 2
.!.
•
Ii) 0
1' 2'
FIG. 9.6
 Network analysis II 259
The short-circuit parameters are then
2s8 +4s+1
1/11 = 1/11 = 2(2s + 1)
(9.20)
2s8 +2s+l
Yn = Y11 = - 2(2s + l)
When y11 = y11 or z11 = z11 , the network is symmetrical.1
Returning to Eq. 9.17, which defines they parameters, we see that they
parameters are expressed explicitly as

(9.21)
y Ia
--
11 - Vi
I
V1-o

y Ia
--
II -Vs
I
V1-0

The reason that the y parameters are also called short-circuit admittance
parameters is now apparent. In obtaining y11 and y11, the 2-2' port must
be short circuited, and when we find y11 and Ym the 1-1' port must be
short circuited, as shown in Figs. 9.1a and 9.1b.
As a second example, let us obtain the y parameters of the Pi circuit in
Fig. 9.5. To obtain y11 and y11, we short circuit terminals 2-2'. We then
have
Yu= YA+ Yo
(9.22)
1/11 = -Ya

(aJ (bJ
FIG. 9.7

1
A symmetrical network is easily recogni7.ed because by interchanging the 1-1'
and 2-2' port designations, the network remains unchanged.
260 Network analysis and synthesis
We next short-circuit terminals 1-1' to obtain
y.. =YB+Yo
(9.23)
Yu= -Yo
The h parameters
A set of parameters that are extremely useful in describing transistor
circuits are the h parameters given by the equations
V1 = h11l 1 + h18 V1 (9.24)
11 = h11l1 + h11Vi11
The individual parameters are defined by the relationships

h12 = Vi
V1
I
I1-0
(9.25)
hn = !J
11
I
V1-0
h11 = lz
Va
I
I1-0

We see that h11 and h11 are short-circuit type parameters, and h11 and hu
are open-circuit type parameters. The parameter h11 can be interpreted
as the input impedance at port 1 with port 2 short circuited. It is easily
seen that h11 is merely the reciprocal of y11 •
1
h11 =- (9.26)
Yn
The parameter h28 is an open-circuit admittance parameter and is related
to z111 by
(9.27)

Both the remaining h parameters are transfer functions; h21 is a short-

circuit current ratio, and h12 is an open-circuit voltage ratio. Their
relationships to the z and y parameters is discussed later in this chapter.
For the Pi circuit in Fig. 9.5, the h parameters are

h - 1
n - YA+ Yo
h - Yo
n - YA+ Yo
(9.28)
Yo
h21= - -~-
YA+ Ye
YAYo
hn = YB + --=-~
YA+ Yo
 Network analysis II 261

FIG. 9.8. Negative impedance converter with load impedance.

Observe that for the Pi circuit, h81 = -h11 • This is the reciprocity
condition for the h parameters and can be derived from their relationships
to either the z or y parameters.
Next let us consider the h parameters of an ideal device called the
negative impedance converter (NIC), which converts a positive load
impedance into a negative impedance at its input port.1 Consider the
NIC with a load impedance ZL shown in Fig. 9.8. Its input impedance is
(9.29)
which can be rewritten as
V V
-1= -1 (9.30)
I1 11
The following voltage-current relationships hold for the NIC.
V1 = kV1
(9.31)
/ 1 = k/1

If we interpret Eq. 9.31 using h parameters, we arrive at the following

conditions.
h11 =ha= 0
1 (9.32)
h11=-=k
h11
We see that since h11 ~ -h11, the NIC is nonreciprocal.
In matrix notation, the h matrix of the NIC is

11
h h11] = [~ k] (9.33)
[ h h11
81 kO
The NIC is a convenient device in the modeling of active circuits. It
is not, however, a device that exists only· in the imagination. Practical

• For a lucid discussion of the properties of the NIC, see L. P. Huelsman, Circuits,
Matrices, and Linear Vector Spaces, McGraw-Hill Company, New York, 1963,
Chapter 4.
262 Network analysis and synthesis
realizations of NIC's have been achieved using transistors. Some of
these are described in an article by Larky.'
The ABCD parameters
Let us take as the dependent variables the voltage and current at the
port 1, and define the following equation.

[~] = [; ;J[~;J (9.34)

This matrix equation defines the A, B, C, D parameters, whose matrix is

known as the transmission matrix because it relates the voltage and current
at the input port to their corresponding quantities at the output. The
reason the current / 1 carries a negative sign is that most ttansmission
engineers like to regard their output current as coming out of the output
port instead of going into the port, as per standard usage.
In explicit form, the ABCD parameters can be expressed as

A-!'.!I
Va Ia=O
(9.35)
C - 11 I
Vi1 I1-0
From these relations we see that A represents an open-circuit voltage
transfer function; B is a short-circuit transfer impedance; C is an open-
circuit transfer admittance; and D is a short-circuit current ratio. Note
that all four parameters are transfer functions so that the term trans-
mission matrix is a very appropriate one. Let us describe the short-circuit
transfer functions B and D in terms of y parameters, and the open-circuit
transfer functions A and C in terms of z parameters. Using straight-
forward algebraic operations, we obtain
1
B=--
Yn
(9.36)
C = ..!_ D = - Yu
Z11 Yn
For the ABCD parameters, the reciprocity condition is expressed by the
equation

(9.37)

• A. I. Larlcy, "Negative-Impedance Converten," Trans. IRE on Circuit Theory,

CT-4, No. 3 (September 1957), 124-131.
 Network analysis II 263
Let us find, as an example, the
A.BCD parameter for the ideal trans-
former in Fig. 9.9, whose defining
equations are
V1 - nV1
1 (9.38)
11 - -(-1.) FIG. 9.9. Ideal transformer.
n
lfwe express Eq. 9.38 in matrix form, we have

(9.39)

so that the transmission matrix of the ideal transformer is

[: :]-[: i] (9.40)

Note, incidentally, that the ideal transformer does not possess an im- ·
pedance or admittance matrix because the self- and mutual inductances
are infinite.11
For the ideal transformer terminated in a load impedance shown in
Fig. 9.10a, the following set of equations apply.
Yi- nv.
(9.41)
11-- v.
nZL
Taking the ratio of V1 to 11 , we find the input impedance at port I to be
V, 1
Z1 - J - n ZL (9.42)
11
Thus we see that an ideal transformer is an impedance transformer. If
the load element were an inductor L (Fig. 9.10b), at port I we would see
an equivalent inductor of value n1L. Similarly, a capacitor Cat the load
would appear as a capacitor of value C/n1 at port I (Fig. 9.10c).
As a second example indicating the use of the transmission matrix in

• For a detailed discussion concerning ideal transformers, see M. E. Van Valkcnburg,

Introduction ta Modem Network Synthesis, John Wiley and Sons, New York, 1960.
264 Network analysis and synthesis

~·□
12 ~
~ ZL:_]
(a)

~iOL~1□ Ideal
(b)
Ideal
(c)
FIG. 9J0. Ideal transformer as an impedance transformer.

network analysis, consider the ABCD parameters of the Pi circuit in

Fig. 9.5.

(9.43)

D = Y.4. + Yo
Yo
If we check for reciprocity from Eq_. 9.43, we see that
AD_ BC= (Y.4. + Yo)(YB + Yo)- (YAYB + YBYo + Y.4.Yo)
Yo•
= Yo'' == 1 (9.44)
Yo•

9.2 RELATIONSHIPS BETWEEN TWO-PORT PARAMETERS

The relationships between two-port parameters are quite easily obtained

because of the simple algebraic nature of the two-port equations. For
example, we have seen that h11 = 1/y11 and h21 = 1/z21• To derive h11 in
terms of open-circuit parameters, consider the z parameter equations
when port 1 is open circuited: / 1 = 0.
V1 = Zial1 (9.45)
V1 = Z1zl1
 Network analysis II 265

Therefore we have h11 = .!'.'i I

= Zia
Y1 I1-0 Zn
(9.46)

Similarly, since h11 is defined as a short-circuit type parameter, we can

derive h,.1 in terms of 11 parameters as

hn = Y11 (9.47)
1111
We can express all the h parameters as functions of the z parameters or 11
parameters alone. An easy way to accomplish this task is by finding out
what the relationships are between the z and 11 parameters themselves.
Certainly, by their very nature, the z and 11 parameters are not simply
reciprocals of each other (as the novice might guess), since one set of
parameters is defined for open-circuit conditions and the other for short-
circuit.
The z and 11 relationships can be obtained very easily by using matrix
notation. If we define the z matrix as

[ZJ = ell :j
Z11
(9.48)

and the y matrix as

[Y] = [1111 'U11] (9.49)

'U11 11n
In simplified notation we can write the two sets of equations as
[VJ= [Z][J] (9.50)
and [J] = [Y][VJ (9.51)
Replacing [J] in Eq. 9.50 by [Y][VJ, we obtain
[VJ = [Z][ Y][VJ (9.52)
so that the product [Z][Y] must yield the unit matrix [U]. The matrices
[Y] and [Z] must therefore be inverses of each other, that is,
[ZJ-1 = [Y] and [YJ-1 = [Z] (9.53)
From the relationship, we can find the telations between the individual
z and 11 parameters.

(9.54)
266 Network analysis and synthesis
TABLE 9.1
Matrix Conversion Table
A. = zi:1zn - z1aa:11
[z] [y] [h] [T]

Yu Y111 A,. ha A Ap
Zn Z11 -A,,, - A,,, hu hu C C
[z]
Y11 Yu hu 1 1 D
Zn Zn - A,,, -
A,,, hu hu
-
C C
1 ha Ap
Zn
A• - A.
Z1a
Yu Y11
hu hu
-DB B
[y]
A,.
Zn
- A•
-Zi1
A.
Y11 Yn
hu
hu
-hu - -B1 A
B
1 Ap
A.
-Zn Zia
Zn
-
Yu
Y11
--
Yu
hu h11 -BD D
[h]
1
_Zn
Zn
-
Zn
Y11
-Yu .6.,,,
Yu
hu hu --D1 -CD
1 A,. hu
-Zn -A. Yu
-- -- A B
Zn Zn Yu Y11 hu - hu
[T]
-Zn1 -Zn
Zn
A,,,
Y11
Yu
Y11
hu
hu
--1
h11
C D

Zii = Y11 Yt.1 = Yu

A,,, A,,, (9.55)
Zia = _ Yu Zn = _ Ya1
A,,, A,,,
where A,,,= y11y 82 - y11y11 • Using these identities we can derive the h
or ABCD parameters in terms of either the z or y parameters. Table 9.1
provides a conversion table to facilitate the process. Note that in the
table (9.56)
Ap=AD-BC

9.3 TRANSFER FUNCTIONS USING TWO-PORT PARAMETERS

In this section we will examine how to determine driving-point and

transfer functions of a two-port by use of two-port parameters. These
functions fall into two broad categories. The first applies to two-ports
 Network analysis II 267
without load and source impedances. These transfer functions can be
described by means of z or y parameters alone. For example, let us derive
the expressions for the open-circuit voltage ratio VJ V1 by using z param-
eters first and y parameters next. Consider the z parameter equations
for the two-port when port 2 is open circuited.
v. = Z1111 (9.57)
V1 = Z1111
If we take the ratio of V1 to Vi, we obtain

(9.58)

By letting 11 of the second y parameter equation go to zero, we derive

the open-circuit voltage ratio as ·

VaY11=- (9.59)
V1Y11
In similar manner we can derive the short-circuit current ratio of a
two-port as
11 =Y11
- - (9.60)
11Yu
11 Zn
and -=-- (9.61)
11 Zn

The open- and short-circuit transfer functions are not those we usually
deal with in practice, since there are frequently source and load im-
pedances to account for. The secon,d category of two-port transfer
functions are those including source or load impedances. These transfer
functions are functions of the two-port parameters z, h, or y and the
source and/or load impedance. For example, let us derive the transfer
admittance 1JV1 of a two-port network that is terminated in a resistor
of R ohms, as given in Fig. 9.1 I. For this two-port network, the following
equations apply.
(9.62)
V1 = -11R
I1 12
+ +"""
Vi Two-port
network
- -
FIG.9.11
268 Network analysis and synthesis

l'-o---+----_.
IL _________________ JI 2'

FIG. 9.12. Two-port equivalent.

By eliminating the variable V11, we obtain

V - 1. - Y11/R (9.63)
.111- -
Y1 Y11+ 1/R
Note that Y11 and y 11 are not the same. Y11 is the transfer admittance
of the two-port network terminated in a resistor R, and y11 is the transfer
admittance when port 2 is short circuited. We must be careful to make
this distinction in other cases of a similar nature.
In order to solve for transfer functions of two-ports terminated at
either port by an impeoance ZL, it is convenient to use the equivalent
circuit of the two-port network given in terms of its z parameters (Fig.
9.12) or y parameters (Fig. 9.13). The equivalent voltage sources z11/ 1
and z11/ 1 in Fig. 9.12 are called controlled sources because they depend
upon a current or voltage somewhere in the network.• Similarly, the
current sources y 11 V1 and y 11 V1 are controlled sources. For the circuit
in Fig. 9.1,2, let us find the transfer impedance Z.1 = Va/Ii, with port 2
r------------7
1 I . 2
+ I I
I I
: I
V1 t I
I I
I I
I I
l' IL ___________ _JI 2'

FIG. 9.13. Two-port equivalent.

• For a lucid treatment of controlled sources, see E. J. Angelo, Electronic Circuits,

2nd &I., McGraw-Hill Book Company, New York, 1964.
 Network analysis II 269
terminated in a load impedance ZL. If we write the mesh equation for
the 11 mesh, we have
-z1111 = (z111 + Z_cl 11 (9.64)
Since V1 = -11ZL, we readily obtain

Zn = -v. = -----=-
z.tZL
(9.65)
11 Z.11 + ZL

It also is clear that the current-ratio transfer function for the terminated
two-port network is
~ = -Zn (9.66)
11 Z11 + ZL
In similar fashion, we obtain the voltage-ratio transfer function for the
circuit represented in Fig. 9.13 as
Va= -
- Yn (9.67)
Yi Ya+ Yn
Next, suppose we are required to find the transfer function VJ V,, for
the two-port network terminated at both ends, as shown in Fig. 9.14.
We first write the two mesh equations
v,, = (R1 + Z1J 11 + Z1af1 (9.68)
0 = z1111 + (z11 + R.)11
Next, we solve for 11 to give

1. = _____V.=-,;Zn-=----- (9.69)
(R1 + Zi1)(R1 + z..) - Zin
From the equation V1 =- R111, we may now arrive at the following
solution.
(9.70)

-+-

FIG.9.14
270 Network analysis and synthesis

t
FIG. 9.15. Two-port equivalent circuit with one controlled-voltage source.

Note that the equivalent circuits of the two-ports in Figs. 9.12 and
9.13 are not unique. Two other examples are given in Figs. 9.15 and 9.16.
Observe that the controlled sources are nonzero in these equivalent
circuits only if the circuit is nonreciprocal.
Finally, let us consider the hybrid equivalent circuit shown in Fig. 9.17.
Observe the voltage-controlled source h11 V2 at port 1 and the current-
controlled source h 11/ 1 at port 2. Let us find the input impedance Zin·
The pertinent equations are
(9.71)

and (9.72)

Solving Eq. 9.72 for V2 , we find

v. = - h11ZLI1 (9.73)
1 + huZL
Substituting V,1 in Eq. 9.73 into Eq. 9.71, we have

v,
1
= (h _1h111h11ZL
11 + h11ZL
) I
1
(9.74)

so that

YU h

Yll + Y12 y 22 + YU

FIG. 9.16. Two-port equivalent circuit with one controlled-current source.

 Network analysis II 271
,-------------,
h I ~ •'
I
I
I
~,
1
/In
I

IL _____________ _JI

FIG. 9.17. Hybrid equivalent circuit.

We can easily check to see that Eq. 9.75 is dimensionally correct since h11
has the dimensions of impedance, h22 is an admittance, and h12, h11 are
dimensionless since they represent voltage and current ratios, respectively.

9.4 INTERCONNECTION OF TWO-PORTS

In this section we will consider various interconnections of"two-ports.

We will see that when a pair of two-ports are cascaded, the overall trans-
mission matrix is equal to the product of the individual transmission
matrices of the two-ports. When two two-ports are connected in series,
their z matrices add; when they are connected in parallel, their y matrices
add. First let us consider the case in which we connect a pair of two-
ports Na and N.,, in cascade or in tandem, as shown in Fig. 9.18. We see
that
(9.76)

The transmission matrix equation for Na is

(9.77)

~
11
)I

Na
,~: ~, Nb

FIG. 9.11. Cascade connection of two ports.

1i
~
2n Network analysis and synthesis

1
1
r------,
1
I
I
~ ------- --
I
I
I
12
.......
11·

V1 I I V2

I !
o----' ! I
L-1--....L.-___.L....-_ _ _ __,__ _ _ _-O
L--~ator___ J
FIG. 9.19. Gyrator in tandem with T circuit.

Correspondingly, for Nb we have

(9.78)

Substituting the second matrix into the first, we obtain

(9.79)
[~] = [:: ;:] . [:: ;:][ ~;.]
We see that the transmission matrix of the overall two-port network is
simply the product of the transmission matrices of the individual two-
ports.
As an example, let us calculate the overall transmission matrix of a
gyrator 7 in tandem with a T network shown in Fig. 9.19. The ideal
gyrator is an impedance inversion device whose input impedance Zin is
related to its load impedance ZL by

a• (9.80)
=-
ZL
The constant a in Eq. 9.80 is defined as the gyration resistance. If we
regard the gyrator as a two-port, its defining equations are

V1 = a(-1.)
1 (9.81)
11 = -v.
a
7 B. D. H. Tellegen, "The Gyrator, a New Electric Network Element," Phillips

Research Repts., 3 (April 1948), 81-101, see also Huelsman, op. cit., pp. 140-148.
 Network analysis II 273
so that the transmission matrix of the gyrator is

We see that for the gyrator

A.D-BC== -1 (9.82)
Therefore the gyrator is a nonreciprocal device, although it is passive.•
The overall transmission matrix of the configuration in Fig. 9.19 is
obtained by the product of the individual transmission matrices

[: :J-[: :rr (9.83)

a a(z,, + ze) ]
z,, z,,
==
[ z. + z11 z_a:11 + z,,ze + z.ze
az11 az,,
If we check the configuration in Fig. 9.19 for reciprocity, we see that for
transmission matrix in Eq. 9.83
A.D-BC== -1 (9.84)
We thus see that any reciprocal network connected in tandem with a
gyrator yields a configuration that is nonreciprocal.
Next consider the situation in which a pair of two-ports N. and
N,, are connected in parallel, as shown in Fig. 9.20. Let us find the
Ii.

lib

FIG. 9.20. Parallel connection of two ports.

• Most gyrators are microwave elev.ices that depend upon the Hall effect in ferrites.
274 Network analysis and synthesis

FIG.9.ll

y parameters for the overall two-port network. The matrix equations

for the individual two-ports are

[;j = [::: ::J[;j (9.85)

and
[;::] = [:::: :::j[;::]
From Fig. 9.20, we see that the following equations must hold.
(9.86)

Vi= Vi,.= V11, V1 = Vi1a = Vlb

(9.87)
/1 = 11,. + 111, / 1 = /111 + /1 b

In connecting two-ports in series or in parallel, we must be careful

that the individual character of a two-port network is not altered when
connected in series or parallel with another two-port. For example, when
we connect the two-ports in Fig. 9.21 in parallel, the impedances Z., and
Zs will be short circuited. Therefore, to insure that a two-port network
does not interfere with the internal affairs of the other, ideal transformers
are used to provide the necessary isolation. In matrix notation, the sum
of the individual [/,] matrices of the two-ports in parallel must equal the
[/] matrix of the overall two-port network. Thus we have

[;~ = [;j + [;:~ (9.88)

= {[Yna Y11a] + [Y111, Y11b]}[V1l
Y11a Yu.. Yan Ynb VJ
 Network analysis II 275
l1a lz,,
+ + +
V1a Na

V1 V2

+ +
V16 Nb v26

FIG. 9.21. Series connection of two ports.

so that the y parameters of the overall two-port network can be expressed

in terms of the y parameters of the individual two-ports as

1/11] = [Yna + 1/111, 1/11a + 1/111>]

1/11 ( 9 .89)
[
1/11 1/n 1/11a + 1/t.tb 1/na + Ynb
If we connect two-ports in series, as shown in Fig. 9.22, we can express
the z parameters of the overall two-port network in terms of the z param-
eters of the individual two-ports as

Zit Zij = [Zita + Znb Zi1a + Ztlb] (9.90)

[
Z :l:uZ.ta + Z1tb Zna + Znb
We may summarize by the following three points.

1. When two-ports are connected in parallel, find the y parameters

first, and, from the y parameters, derive the other two-port parameters.
2. When two-ports are connected in series, it is usually easiest to find
the z parameters.
3. When two-ports are connected in tandem, the transmission matrix
is generally easier to obtain.
As a final example, let us find the y parameters of the bridged-T circuit
in Fig. 9.6. We see that the bridged-T circuit would be decomposed into
a parallel connection of two-ports, as shown in Fig. 9.23. Our task is to
first find the y parameters of the two-ports N,. and Nb. The y parameters
of Nb are obtained by inspection and are
Ytlb = Yan = -½ (9.91)
1/ub = Ynb = ½
Na is a T circuit so that the z parameters can be obtained by inspection.
276 Network analysis and synthesis

L ___ ...J
Nb
FIG. 9.23

These are
s+l
Z1111 =Zna=--
S (9.92)
z11,.=z11,.=1
We then find they parameters from the equations
Zna s(s 1) +
Yua = Yna = liz,. = ·2s + 1
(9.93)
Z19a s1
Y11a = Yua = - = - --
liz,. 2s + 1
Since both N,. and Nb are symmetrical two-ports, we know that Yu = Y••
for the overall bridged-T circuit. The y parameters for the bridged-T.
circuit are now obtained as

Yu= Y1111 + Yin

= s(s + 1) + ! = 2s1 + 4s + 1
2s + 1 2 2(2s + 1)
(9.94)
Ylll = Y11a + Yu&
= - -s8- ·--=
1
-
2s1 +2s+l
2s +1 2 2(2s + 1)

9.5 INCIDENTAL DISSIPATION

As we have seen, the system function H(s) of an R-L-C network consists

of a ratio of polynomials whose coefficients are functions of the resistances,
inductances, and capacitances of the network. We have considered, up
to this point, that the inductors and capacitors are dissipationless; i.e.,
 Network analysis II 277
there are no parasitic resistances associated with the L and C. Since, at
high frequencies, parasitic losses do play an important role in governing
system performance, we must account for this incidental dissipation
somehow. An effective way 9f accounting for parasitic resistances is to
"load down" the pure inductances and capacitances with incidental
dissipation by associating a resistance r, in series with every inductor L,,
and, for every capacitor c,, we associate a resistor whose admittance is g,,
as depicted in Fig. 9.24. Suppose we call the system function of the
network without parasitic dissipation (Fig. 9.24a) H(s), and the system
function of the ''loaded" network H 1(s) (Fig. 9.24b). Let us consider the
the relationship between H(s) and H 1(s) when the dissipation is uniform,
i.e., in a manner such that
(9.95)

where the constant 0t is real and positive.

When a network has uniform dissipation or is uniformly loaded, the
sum of impedances in any mesh of the unloaded network
1
m11 ==R,+sL,+- (9.96)
sC,
becomes, after loading,
1
m'11 == R, + sL, + r, +
sC,+ g,
(9.97)
1
== R, + L;(s + at) + - - -
C;(s + at)

R;
-- ~ -
L;
o----l'ffl'-<>
C; C;

~
-
o--1 l----<>
g;
H(a)

Hi(•) • H(a + aJ
(a) (b)

FIG. 9.24. (a) Original network. (b) Loaded network.

278 Network analysis and synthesis

FIG.9.15

Similarly, on node basis, if the admittance between any two nodes of the
original (unloaded) network is
1
nii = G, + sC, + - (9.98)
sL.
then the same node admittance after loading is
1
n'., = G, + sC, + g, +
+ r, sL,
(9.99)
1
= G. + Ch + 0t) + - - -
L.(s + at)
Since any system function can be obtained through mesh or node equa-
tions, it is readily seen that the original system function H(s) becomes
H(s + 0t) after the network has been uniformly loaded, that is H 1(s) =
H(s + 0t).
Consider the following example. Let us first find the y parameters for
the unloaded network in Fig. 9.25. By inspection we have
1 s s 1 7s
Yn = 2+ 3 + 4= 2+ 12
1 ·s s1 +2
Y11 = 1 + -2s + -4 = 1 + -4s- (9.100)

s
1/11 = 1/11 = - -
4
Then, for a loading constant 0t = l, the loaded network is shown in Fig.
9.26. In parallel with the capacitor C1 = l f, we have an admittance
(9.101)
In parallel with the capacitor C1 = l f, the associated admittance is
g1 = 0tC1 = ¼mho (9.102)
 Network analysis II 'X/9
80

¼t 10

20 60 10

2h

FIG. 9.26. Loaded network ex= l-

ln series with the inductor L = 2 h, we have a resistor
r1 = rx.L= lil (9.103)
Now we determine they parameters for the loaded network to be
, 1 s 1 s 1
=2+3+6+4+8
1111

= ½+ l(s + l) + ¼(s + U
= ½+ lis + U
1 s 1
Y'n= 1 +--+-+-
2s+ 1 4 8
= 1 + (s + ½)1 + 2
4(s + ½)
11•11 =- (! + i) = - Hs + ~) (9.104)

We see that the y parameters of the loaded network could have been
obtained from the y parameters of the unloaded network by the relation-
ship Hi(s) = H(s + cx).
We will make use of the uniform loading concept to prove an important
theorem concerning network realizability in Chapter 10.

9.6 ANALYSIS OF LADDER NETWORKS

In this section we will consider a simple method of obtaining the

network functions of a ladder network in a single operation.• This
method depends only upon relationships that exist between the branch
• F. F. Kuo and G. H. Leichner, "An Iterative Method for Determining Ladder
Network Functions," Proc. IRE, 47, No. 10 (Oct. 1959), 1782-1783.
280 Network analysis and synthesis

FIG. 9.27. Ladder network.

currents and node voltages of the ladder. Consider the network shown
in Fig. 9.27, where all the series branches are given as impedances and
all the parallel branches are given as admittances. If the v denote node
voltages and i denote branch currents, then the following relationships
apply.
v,+1 = i1+.Z.1+1 + v,+a
(9.105)
i1 = V1+1 yl+l + il+B
These equations form the basis of the method we discuss here.
To illustrate this method, consider the network in Fig. 9.28, for which
the following node voltage and branch current relationships apply.
v.(s) = v.(s)
= Y.(s) Va(s)
/ 1 (s)

Va(s) = / (s) Za(s) + V (s) = [l + Za(s) Y.(s)] V.(s)

1 1

/ 1(s)= Y (s) Va(s) + I,.(s)

1 (9,106)
= {Y1(s)[l + Za(s) Y.(s)J + Y.(s)} V.(s)
V1(s) = l1(s) Zi(s) + Va(s)
= {Zi[Y8(1 + ZaYJ + Y,J + (1 + ZaYJ} V1(s)
Upon examining the set of equations given in Eq. 9.106, we see that
each equation depends upon the two previous equations only. The first
equation, Vii= V1 is, of course, unnecessary. But, as we shall see later,
it is helpful as a starting point. In writing these equations, we begin at
2

1' 2'
FIG.9.28
 Network analysis II 281
the 2-2' port of the ladder and work towards the 1-1' port. Each suc-
ceeding equation takes into account one new immittance. We see,
further, that with the exception of the first two equations, each subsequent
equation is obtained by multiplying the equation just preceding it by tho
immittance that is next down the line, and then adding. ta this product
the equation twice preceding it. For example, we see that /.(s) is obtained
by multiplying the preceding equation V.(_s) by the admittance Yc(s).
The next immittance is ZJ...s). We obtain V11(s) by multiplying the previous
equation / 1(s) by ZJ...s) to obtain I.ZS; then we add to this product the
equation twice preceding it, V.(s), to obtain V11 = /1 Z. + V1 • The process
is then easily mechanized according to the following rules: (1) alternate
writing node voltage and branch current equations; (2) the next equation
is obtained by multiplying the present equation by the next immittance
(as we work from one port to the other), and adding to this product the
results of the previous equation.
Using this set of equations, we obtain the input impedance Z 1.{s) by
dividing the equation for V.(s) by the equation for / 1(s). We obtain the
voltage-ratio transfer function V1(s)/V1 (s) by dividing the first equation
V.(s) by the last equation V1(s). We obtain other network functions such
as transfer immittances and current ratios in similar manner. Note that
every equation contains Vi1(s} as a factor. In taking ratios of these equa-
tions, the V1 (s) term is canceled. Therefore, our analysis can be simplified
if we let V1 (s) = 1.
If the first equation of the set were a current variable /Js) instead of the
voltage V1(s), the subsequent equations would contain the current variable
Ih) as a factor, which we could also normalize to /Js) = 1. An example
in which the first equation is a current rather than a voltage equation
may be seen by determining the y parameters of a two-port network.
Before we embark upon some numerical examples, it is important to
note that we must represent the series branches as impedances and· the
shunt branches as admittances. Suppose the series branch consisted of a
resistor R = 1 n in parallel with a capacitor C = l f. Then the impedance
of the branch is
Z(s) = R(1/sC) = _2_ (9.107)
1/sC + R s+2
and must be considered as a single entity in writing the equations for the
ladder. Similarly, if a shunt branch consists of a resistor R 1 = 2 n and
an inductor Li = 1 h, then the admittance of the branch is

1
Y(s)=-- (9.108)
2+s
282 Network analysis and synthesis
The key point in this discussion is that we must use the total impedance
or admittance of a branch in writing the equations for the ladder.
Eumple 9.1. Let us find the voltage ratio V1 /V1 , the current ratio Iaf/1, the
input impedance Z 1 = VJli, and the transfer impedance Z 11 = VJ/1 for the
network in Fig. 9.29. First, we must represent the series branches as impedances
and the shunt branches as admittances, as shown• in Fig. 9.30. The branch
current and node voltage equations for the network are

q [, F
v. , ~II v. ~·
:: J -~......-½-h-3 :• :
Ii-( {1----c~,_f____. : , ~•

FIG. 9.29 FIG. 9.30

V.(s) =1
/ 1(s) = 2s
Va(s) = 3s(2s) + 1 = 6s8 + l
2 2 (9.109)
/ 1(s) = -(6s8 + 1) + 2s = l4s + -
s s

V1(s) =; 4( 14s 2)
+ ; + 6s8 + 1 = 6s8 + 57 + ~ 8
The various network functions are then obtained.

(a)
V1 6s' + S7s2 + 8
Zin = 11 = 14s3 + 2s

(b)
v, s2
V1 = 6s' + 57s2 +8
(9.110)
(c)
la 2r
Ii= 14s2 + 2
(d)
v. s
Zn = / 1 = 14s2 +2
Eumple 9.2. Find the short-circuit admittance functions y 11 and y 11 for the
network in Fig. 9.31. To obtain the short-circuit functions, we must short
circuit the 2-2' port. Then we represent the series branches as impedances and
the shunt branches as admittances so that the resulting network is given as is
 Network analysis II 283

lh 2f 2f ½t

l' 2'
FIG. 9.31

illustrated in Fig. 9.32. The pertinent equations are

11 =1 V11 =2
la = is(2) +1= s +1
1 5
V1 = 2(s + 1) (s + 1) + 2 = 2 (9.111)

11 = 2s9 + 1(5)
- s - 2 + (s + 1) = 6s + 1 + 7
5/2
We now obtain our short-circuit functions as
11 12 2 1 1. 2
Y =-=-s+-+- Yn=-= -- (9.112)
11 V 5 5 s'1 Vi s

FIG.9.32

A number of other contributions on ladder networks have appeared

in the recent literature. To name but a few, there are the works of
Bashkow.10 O'Meara,11 Bubnicki,11 Walker,11 and Dutta Roy.H
10 T. R. Bashkow, "A Note on Ladder Network Analysis," IRE Trans. on Circuit

Theory, CT-I, (June 1961), 168.

11
T. R. O'Meara, "Generating Arrays for Ladder Network Transfer Functions,"
IEEE Trans. on Circuit Theory, CT-10, (June, 1963), 285.
11
Z. Bubnicki, "Input Impedance and Transfer Function of a Ladder Network,"
IEEE Trans. on Circuit Theory, CT-10, (June, 1963), 286.
11 F. Walker, "The Topological Analysis of Non-Recurrent Ladder Networks,"

Proc. IEEE, 52, No. 7, (July, 1964), 860.

HS. C. Dutta Roy, "Formulas for the Terminal Impedances and Transfer FunctiODI
of General Multimesh Ladder Networks," Proc. IEEE, 52, No. 6, (June, 1964), 738.
284 Network analysis and synthesis
Because of the recursive nature of the equations involved, the digital
computer is an ideal tool for the analysis of ladder networks. In Chapter
15 a detailed description is given of a digital computer program based
upon the algorithm described in this section for evaluating ladder network
functions.

Problems
9.1 Find the z, 11, h, and T parameters of the networks shown in the figures.
Some of the parameters may not be defined for particular circuit configurations.

<>--0----o

,., (bJ
PROB. 9.1
(c}

9.2 Find the z, 11, h, and Tparameters for the networks shown in the figures.
20 10
1 2

10 30

1
,., 3h
2

it

0 0
(bJ
2f 2f

(cJ
PROB.9.2
 Network analysis II 285
9.3 Find the z parameters for the lattice and bridge circuits in the figures
shown. (The results should be identical.)

Lattice network
(a)

Bridge network
(b)

PROB.9,3

9.4 For the lattice and bridge circuits in Prob. 9.3 find the 11 parameters in
terms of the admittances Y11 ... l/Z11 and Y• =- lfz•.
9.5 For the circuit shown, find the voltage-ratio transfer function VJV1 and
the input impedance Z 1 ... VJ/1 in terms of the z parameters of the two-port
network N and the load resistor RL•
1 I1 1,. 2
+
V1

':'
N BL ◄ :.
◄►

l' 2'

PROB. 9.5
286 Network analysis and synthesis
9.6 For the cascade connection of two-ports depicted in the figure, show that
the transfer impedance Zia of the overall circuit is given in terms of the z param-
eters of the individual two-ports by the equation
Zia = Zi1,hllb
Zub + Z92a
In addition, show that the short-circuit admittance y11 is given by
Yu = _ Y1IGY11b
Yub + Y11a

+ +
I1 Ii
+
- +
-

-
V1 N,. Nb V:2
-- "'.'."

PROB.9.6 PROB.9.7
9.7 Find the z and y parameters of the transformer (nonideal) shown in the
figure. Determine the T- and ,r-equivalent circuits for the transformer i~ terms
of L1, L 1 , and M. (Hint: Use the z parameters for the T-equivalent circuit,
and the y parameters for the ,r-equivalent circuit.)
9.8 The inverse hybrid parameters of a two-port network are defined by the

equation [:~ = [;: ::J [:~

Express the g parameters in terms of either the z or y parameters and give a
physical interpretation of the meaning of these parameters; i.e., say whether a
parameter is an open- or short-circuit parameter, and whether it is a driving
point or transfer function. Finally, derive the conditions of reciprocity for the
g parameters.
9.9 Prove that for a passive reciprocal network
AD -BC= 1
where A, B, C, D are the elements of the transmission matrix.
9.10 Find the z and h parameters of the common emitter transistor repre-
sented by its T-circuit model.

♦-
Ii

PROB. 9.IO
 Network analysis II 287
9.11 The circuit in part (a) of the figure is to be described by an equivalent
input circuit shown in part (b). Determine Zeq in (b) as a function of the elements
and voltages in (a).
R,

v,
7
._____.__h22--R-L.....,J
(a)

v.

(b)

PROB. 9.11

9.12 Find the Tparameters of the configurations shown in parts (a) and (b)
of the figure.

l'
C· (a)
Gyrator 2'

:Jor[ Gyrator

(b)
Ideal
transformer

PROB. 9.12

9.13 Find the 11 parameters of the twin-T circuit in Prob. 9.2c by considering
the circuit to be made up of two T circuits in parallel.
9.14 Find the z parameters of the circuits shown.
288 Network analysis and synthesis
1 I1 I2 2
+ +

V1 V2

l' Ideal 2'

transformer (a)
I2 2
+

V2

l' Ideal 2'

(b) transformer PROB. 9.14

Ii
+ +

Ideal
transformer PROB. 9.15

PROB.9.16
 Network analysis II 289
9.15 Find the z parameters of the circuit shown.
9.16 Find the z parameters of the circuit shown.
9.17 For the circuit in Prob. 9.2b determine they parameters of the uniformly
loaded circuit derived from the original circuit with the dissipation a. = 0.1.
Plot the poles and zeros of both cases.
9.18 Find the transfer impedance VJ/1 for the circuit in part (a) and the
voltage ratio VJV1 for the circuit in part (b). Plot the poles and zeros for the
transfer functions obtained.
11

=--•
- --0-;~_o_f
. . . . . _*_fo-~o-ho_,____.~•• :,
(a)

(b) PROB. 9.11

9.19 Find the short-circuit parameters for the ladder network utilizing the
method in Section 9.6.
10

20 ½t
PROB. 9.19

30
v..
-+
½t 20
lf V2

½t
PROB. 9.20
9.20 Determine the voltage ratio VJV1 , the current ratio JJ/1, the transfer
impedance VJ/1, and the driving point impedance VJ/1 for the network sho~.
 chapter 10
Elements ff realizability
I theory
I

10.1 CAUSALITY AND STABILITY j

In the preceding chapters we have beep. primarily concerned with the

problem of determining the response, give* the excitation and the network;
this problem lies in the domain of neti·ork analysis. In the next five
chapters we will be dealing with the pro lem of synthesizing a network
given the excitation E(s) and the respon e R(s). The starting point for
any synthesis problem is the system func~on

H(s) = R(t> (10.1)

E()
Our task is to synthesize a network from ja given system function.
The first step in a synthesis procedur~ is to determine whether H(s)
can be realized as a physical passive net~ork. There are two important
considerations-causality and stability. ) By causality we mean that a
voltage cannot appear between any pafr of terminals in the network
before a current is impressed, or vice ver~a. In other words, the impulse
response of the network must be zero fo~ t < 0, that is,
I
!

h(t) =0 ti< 0 (10.2)

As an example, the impulse response I

I

h(t) = ca u(t) (10.3a)

is causal, whereas h(t) = ca ti (10.3b)

290
 Elements of realizability theory 291
h(t) h(t-T)

0 0
(a) (b)

FIG. 10.1. (a) Nonrealizable impulse response. (b) Realizable impqlse response.

is not causal. In certain cases, the impulse response could be made

realizable (causal) by delaying it appropriately. For example, the impulse
response in Fig. IO.la is not realizable. If we delay the response by T
seconds, we find that the delayed response h(t - T) is realizable (Fig.
IO.lb).
In the frequency domain, causality is implied when the Paley-Wiener
criterion1 is satisfied for the amplitude function IH(jw)I. The Paley-Wiener
criterion states that a necessary and sufficient condition for an amplitude
function IH(jw)I to be realizable (causal) is that
jlog IHUw)l jdW <
f-00
00

1 +w
1
00 (10.4)

The following conditions must be satisfied before the Paley-Wiener

criterion is valid: h(t) must possess a Fourier transform H(jw); the
square magnitude function IH(jw)P1 must be integrable, that is,

L:IHUw)l 1dw < oo (10.5)

The physical implication of the Paley-Wiener criterion is that the amplitude

IH(jw)I of a realizable network must not be zero over a finite band of
frequencies. Another way of looking at the Paley-Wiener criterion is that
the amplitude function cannot fall off to zero faster than exponential
order. For example, the ideal low-pass filter in Fig. 10.2 is not realizable
because beyond w 0 the amplitude is zero. The Gaussian shaped curve
IHUw)I = e_,,,• (10.6)
shown in Fig. 10.3 is not realizable because
Ilog IH(jw)I I = w 1 (10.7)
1
R. E. A. C. Paley and N. Wiener, "Fourier Transforms in the Complex Domain,"
Am. Math. Soc. Colloq. Pub., 19 (1934), 16-17.
292 Network analysis and synthesis
IHUw>I
IHUw)I

Ao

FIG. 10.2. Ideal filter char- FIGj' 10.3. Gaussian filter character-
acteristic istic.
'

so that the integral

f
a)

-a)
~dw
1 + °'I
(10.8)

is not finite. On the other hand, the amplitude function

I
. : 1
IHUw)I - ✓~ (10.9)
✓J + w•
does represent a realizable network. I,;1 fact, the voltage-ratio transfer
function of the R-C network in Fig. 10.f 'bas an amplitude characteristic
given by IH(jw)I in Eq. 10.9. •
For the ideal filter in Fig. 10.2, the inyerse transform h(t) has the form

h(t) _ A 0 s!f, wot (10.10)

I

where A0 is a constant. From the sin*':,: curve in Fig. 3.f4, we see that
h(t) is nonzero fort less than zero. In ct, in order to make h(t) causal,
it must be delayed by an infinite amo t. In practice, however, if we
delay h(t) by a large but finite amount t,. ~uch that for t < 0 the magnitude
of h(t - tJ is less than a very small q~ntity E, that is,
i
lh(t - ,.)I <E I
I
t <0
10 I

+ +

V;(B} lf Vo<•>

FIG. 10.
 Elements of realizability theory 293
we then can approximate h(t - tiJ by a causal response lzi(t) which is
zeto for t < 0. (For a more detailed discussion of the Paley-Wiener
criterion, the reader is referred to an excellent treatment by Wallman.I)
H a network is stable, then for a bounded excitation e(t) the response
(t) is also bounded. In other words, if

le(t)I < C1
then lr(t)I < c.
where C1 and C1 are real, positive, finite quantities. If a linear system is
stable, then from the convolution integral we obtain

(10.11)

Equation 10.11 requires that the impulse response be absolutely integrable,

or
(10.12)

One important requirement for h(t) to be absolutely integrable is that the

impulse response approach zero as t approaches infinity, that is,

lim h(t)-0

Generally, it can be said that with the exception of isolated impulses, the
impulse rest>Qnse must be bounded for all t, that is,

lh(t)I <C all t (10.13)

where C is a real, positive, finite number.

Observe that our definition of stability precludes such terms as sin ro0t
from the impulse response because sin ro0 t is not absolutely integrable.
These undamped sinusoidal terms are associated with simple poles on
the jro axis. Since pure L-C networks have system functions with simple
poles on the fa, axis, and since we do not wish to call these networks
unstable, we say that a system is marginally stable if its impulse response
is bounded according to Eq. 10.13, but does not approach zero as t
approaches infinity.
In the frequency domain, the stability criterion requires that the system

1
G. B. Valley Jr. and H. Wallman, Vacuum Tube Amplifiers, McGraw-Hill Book
Company, New York, 1948, Appendix A, pp. 721-727.
 I
294 Network analysis and synthesis I

function possess poles in the left-half plaq.e or on the jw axis only. More-
over, the poles on the jw axis must be sin}ple. 3 As a result of the require-
ment of simple poles on the jw axis, if ~s) is given as

H(s) = 11
a 11 s +·a,._1s 1 + ·I · · + a1s + a 0
11
-
(10.14)
bms"' + b,,._1Sm-l + !· ' ' + bis + bo
then the order of the numerator n can~ot exceed the order of the de-
nominator m by more than unity, that is, n - m::;;; 1. If n exceeded m
by more than unity, this would imply that at s = jw = oo, and there
would be a multiple pole. To summari~e, in order for a network to be
stable, the following three conditions orl. its system function H(s) must
be satisfied:

1. H(s) cannot have poles in the right-half plane.

2. H(s) cannot have multiple poles in tihe jw axis.
3. The degree of the numerator of H(s) cannot exceed the degree of
the denominator by more than unity. ·

Finally, it should be pointed out that a rational function H(s) with poles
in the left-half plane only has an inverse transform h(t), which is zero for
t < O.' In this respect, stability implies c~usality. Since system funct;ons
of passive linear networks with lumped !elements are rational functions
with poles in the left-half plane or jw ax~s only, causality ceases to be a
problem when we deal with system functions of this type. We are only
concerned with the problem of causality jwhen we have to design a filter
for a given amplitude characteristic sue~ as the ideal filter in Fig. 10.2.
We know we could never hope to realtze exactly a filter of this type
because the impulse response would no( be causal. To this extent the
Paley-Wiener criterion is helpful in defi~ng the limits of our capability.

IG.2 HURWITZ POLYNOMIALS

In Section 10.1 we saw that in order for a system function to be stable,

its poles must be restricted to the left-halflplane or the jw axis. Moreover,
the poles on the jw axis must be simple. jrhe denominator polynomial of
the system function belongs to a class o~ polynomials known as Hurwitz

1
In Chapter 6 it was shown that multiple
Aot sin w,.t.
polt
,
on the jw axis gave rise to terms as
!
'G. Raisbeck, "A Definition of Passive Lin,ar Networks in Terms of Time and
Energy," J. Appl. Phys., 25 (Dec., 1954), ISIP--1514. The proof follows straight-
forwardly from the properties of the Laplace ~orm.
 Elements of realizability theory 295
polynomials. A polynomial P(s) is said to be Hurwitz if the following
conditions are satisfied:

1. P(s) is real when s is real.

2. The roots of P(s) have real parts which are zero or negative.
As a result of these conditions, if P(s) is a Hurwitz polynomial given by
(10.15)

then all the coefficients a, must be real; ifs, = ix, + j{J is a root of P(s),
then ix, must be negative. The polynomial

P(s) = (s + l)(s + 1 + j✓2 )(s + 1 - jJ2) (10.16)

is Hurwitz because all of its roots have negative real parts. On the other
hand,
G(s) = (s - l)(s + 2)(s + 3) (10.17)

is not Hurwitz because of the roots= 1, which has a positive real part.
Hurwitz polynomials have the following properties:

l. All the coefficients a, are nonnegative. This is readily seen by

examining the three types of roots that a Hurwitz polynomial might
have. These are
s = -r, y, real and positive
s = ±jw, w, real
s = -ix, ±j/J, ix, real and positive
The polynomial P(s) which contains these roots can be written as

P(s) = (s + r,)(s11 + wl)[(s + ixJ1 + /Jl] • · • (10.18)

Since P(s) is the product of terms with only positive coefficients, it follows
that the coefficients of P(s) must be positive. A corollary is that between
the highest order term in s and the lowest order term, none of the coeffi-
cients may be zero unless the polynomial is even or odd. In other words,
an-i, an-2, ... , a 1, a 1 must not be zero if the polynomial is neither even
nor odd. This is readily seen because the absence of a term a, implies
cancellation brought about by a root s - r, with a positive real part.
2. Both the odd and even parts of a Hurwitz polynomial P(s) have
roots on the jw axis only. If we denote the odd part of P(s) as n(s) and
the even part as m(s), so that
P(s) = n(s) + m(s) (10.19)
296 Network analysis and synthesis
then m(s) and n(s) both have roots on uie jw axis only. The reader is
referred to a proof of this property by Guillemin. 11
3. As a result of property 2, ifP(s) is either even or odd, all its roots are
on the jw axis.
4. The continued fraction expansion of the ratio of the odd to even
parts or the even to odd parts of a Hurwitz polynomial yields all positive
quotient terms. Suppose we denote the ratios as 'l'(s) = n(s)/m(s) or
tp(s) = m(s)/n(s), then the continued fraction expansion of tp(s) can be
written as
1
tp(s) = q1s + - - - - - - - (10.20)
1
q.s+-----
1
q.s+---

+_!_
q,.s
where the quotients qi, q1, ••. , q,. must be positive if the polynomial
P(s) = n(s) + m(s) is Hurwitz.• To obtain the continued fraction expan-
sion, we must perform a series of long divisions. Suppose tp(s) is

tp(s) = m(s) (10.21)

n(s)
where m(s) is of one higher degree than n(s). Then if we divide n(s) into
m(_s), we obtain a single quotient and a remainder

tp(s) = q1s + Ri(s) (10.22)

n(s)
The degree of the term R 1(s) is one lower than the degree of n(s). Therefore
if we invert the remainder term and divide, we have
n(s) = q.s + R.(s) (10.23)
Ri(s) Ri(s)
Inverting and dividing again, we obtain

Ri(s) = qa5 + R1(s) (10.24)

R1 (s) R1 (s)
' E. Guillemin, The Mathematics of Circuit Analysis, John Wiley and Sons, New York,
1949. An exciellent treatment of Hurwitz polynomials is given here.
• A proof can be undertaken in connection with L-C driving-point functions; see
M. E. Van Valkenburg, Introduction to Modern Network Synthesis, John Wiley and
Sons, New York, 1960.
 - - - - - - -

Elements of realizablllty theory 297

We see that the process of obtaining the continued fraction expansion
of 'P(s) simply involves division and inversion. At each step we obtain a
quotient term q,s and a remainder term, R1+1(s)/Ri_s) • .We then invert the
remainder term and divide RH1{s) into Rh) to obtain a new quotient.
There is a theorem in the theory of continued fraetions which states that
the continued fraction expansion of the even to oddu.odd to even parts
of a polynomial must be finite in length. 7 Another theorem states that, if
the continued fraction expansion of the odd to even or even to odd parts
of a polynomial yields positive quotient terms, then the polynomial must
be Hurwitz to within a multiplicative factor W(s). 8 That is, if we write
F(s) = W(s) F 1 (s) {10.25)
then F(s) is Hurwitz, if W(s) and F 1(s) are Hurwitz. For example, let us
test whether the polynomial
F(s) = s' + s8 + Sal + 3s + 4 (10.26)
is Hurwitz. The even and odd parts of F(s) are
m(s) = s' + ss• + 4
(10.27)
n(s) = r + 3s
We now perform a continued fraction expansion of 'P(s) = m(s)/n(s) by
dividing n(s) by m(s), and then inverting and dividing again, as given by
the operation
s8 + 3s)s' +Sr+ 4(s
s' + 3s8
2s8 + 4 )r + 3s(s/2
as+2s
s)2s8+4(2s
2s8
4 )s(s/4
s
=
so that the continued fraction expansion of 'P(s) is

'P(s) = m(s) = s + 1 _ (10.28)

n(s) !.. + 1
2 2s+...!..
s/4
7
See Van Valkenburg. loc. cit.
• W(s) is a common factor in m(s) and n(s).
298 Network analysis and synthesis
Since all the quotient terms of the continued fraction expansion are
positive, F(s) is Hurwitz.
Example 10.1. Let us test whether the polynomial

G(s) = s8 + 2s8 + 3s + 6 (10.29)

is Hurwitz. The continued fraction expansion of n(s)/m(s) is obtained from the

division
2s8 + 6)s8 + 3s (s/2
s8 + 3s
0
=
We see that the division has been terminated abruptly by a common factor
s8 + 3s. The polynomial can then be written as

G(s) = (s8 + 3s)( 1 + n (10.30)

We know that the term 1 + 2/s is Hurwitz. Since the multiplicative factor
s8 + 3s is also Hurwitz, then G(s) is Hurwitz. The term s8 + 3s is the mul-
tiplicative factor W(s), which we referred to earlier.

Example 10.2. Next consider a case where W(s) is non-Hurwitz.

F(s) = s1 + 2s8 + 2s6 + s' + 4s8 + Sr + 8s + 4 (10.31)

The continued fraction expansion of F(s) is now obtained.

n(s) s 1
- =-
m(s)
+ - - -1- -
2 ~
as+f~ (10.32)

We thus see that W(s) = s' + 4, which can be factored into

W(s) = (r + 2s + 2)(r - 2s + 2) (10.33)

It is clear that F(s) is not Hurwitz.

Example 10.3. Let us consider a more obvious non-Hurwitz polynomial

F(s) = s' + s8 + 2s8 + 3s + 2 (10.34)

 Elements of realizability theory 299
The continued fraction expansion is

s8 + 3s)s' + 2t2 + 2 (s
s' +3r
-} + 2)s3 + 3s ( -s
s8 -2s
Ss)-} + 2 (-s/S
-sl
2)Ss (ts
Ss

We see that F(s) is not Hurwitz because of the negative quotients.

Example 10.4. Consider the case where F(s) is an odd or even function. It is
impossible to perform a continued fraction expansion on the function as it
stands. However, we can test the ratio of F(s) to its derivative, F'(s). 8 If the
ratio F(s)/F'(s) gives a continued fraction expansion with all positive coefficients,
then F(s) is Hurwitz. For example, if F(s) is given as
F(s) = s7 + 3s5 + 2s8 + s (10.35)
then F'(s) is F'(s) = 7s8 + 15.r' + 6r + 1 (10.36)

Without going into the details, it can be shown that the continued fraction
expansion of F(s)/F'(s) does not yield all positive quotients. Therefore F(s) is
not Hurwitz.

10.3 POSITIVE REAL FUNCTIONS

In this section we will study the properties of a class of functions

known as positive real functions. These functions are important because
they represent physically realizable passive driving-point immittances. A
function F(s) is positive real (p.r.) if the following conditions are satisfied:
1. F(s) is real for reals; that is, F(a) is real.
2. The real part of F(s) is greater than or equal to zero when the real
part of s is greater than or equal to zero, that is,
Re [F(s)J ~ 0 for Res~O
Let us consider a complex plane interpretation of a p.r. function.
Consider the s plane and the F(s) plane in Fig. 10.5. If F(s) is p.r., then a
point o-0 on the positive real axis of the s plane would correspond to, or
map onto, a point F(ao) which must be on the positive real axis of the
F(s) plane. In addition, a point si in the right half of the s plane would
• See Guillemin, loc. cit.
300 Network analysis and synthesis
jlmF
ft,>
• plane Ff•) plane
0 0
Si Fl•il

O'O O' F(a-oJ ReF

FIG. 10.5. Mapping of s plane onto F(s) plane.

map onto a point F(si) in the right half of the F(s) plane. In other words,
for a positive real function, the right half of the s plane maps onto the
right half of the F(s) plane. The real axis of the s plane maps onto the
real axis of the F(s) plane.
A further restriction we will impose is that F(s) be rational. Consider
the following examples ofp.r. functions:
1. F(s) =Ls (where Lis a real, positive number) is p.r. by definition.
If F(s) is an impedance function, then L is an inductance.
2. F(s) =R (where R is real and positive) is p.r. by definition. If F(s)
is an impedance function, R is a resistance.
3. F(s) = K/s (K reai and positive) is p.r. because, when s is real, F(s)
is real. In addition, when the real part of s is greater than zero, Re (s) =
(1 > o.
Then Re (K)s = ~
Ka
+ w•
>0 (10.37)

Therefore, F(s) is p.r. If F(s) is an impedance function, then the corre-

sponding element is a capacitor of 1/K farads.
We thus see that the basic passive impedances are p.r. functions.
Similarly, it is clear that the admittances
Y(s) == K
Y(s) = Ks
(10.38)
Y(s) =K
s
are positive real if K is real and positive. We now show that all driving-
point immittances of passive networks must be p.r. The proof depends
upon the following assertion: for a sinusoidal input, the average power
dissipated by a passive network is nonnegative. For the passive network
 Elements of realizability theory 301
in Fig. 10.6, the average power dissipated by the network is
Average power == l Re [Zin(jw)] 111 1 ~ 0 (10.39)
We then conclude that, for any passive network
Re [Z1n(jw)J ~ 0 (10.40)
We can now prove that for Res == a ~ 0, Re Zin(a + jw) ~ 0. Consider
the network in Fig. 10.6, whose driving-
point impedance is Zin(s). Let us load I
)I
the network with incidental dissipation :
such that if the driving-point impedance Passive

___
network
of the uniformly loaded network is Z1n<•J
Zi(s), then o----------f.__ .....
Zi(s) == Zin(s + at) (10.41) FIG. 10.6

where at, the dissipation constant, is real and positive. Since Zi(s) is the
impedance of a passive network,
ReZi(jw) ~ 0 (10.42)
so that Re Zin(0t + jw) ~ 0 (10.43)
Since 0t is an arbitrary real positive quantity, it can he taken to be a.
Thus the theorem is proved.
Next let us consider some useful properties of p.r. functions. The
proofs of these properties are given in Appendix D.
1. If F(s) is p.r., then 1/F(s) is also p.r. This property implies that if a
driving-point impedance is p.r., then its reciprocal, the driving-point
admittance, is also p.r.
2. The sum of p.r. functions is p.r. From an impedance standpoint,
we see that if two impedances are connected in series, the sum of the
impedances is p.r. An analogous situation holds for two admittances in
parallel. Note that the difference of two p.r. functions is not necessarily
p.r.; fot example, F(s) == s - 1/s is not p.r.
3. The poles and zeros of a p.r. function cannot have positive real
parts, i.e., they cannot be in the right half of the s plane.
4. Only simple poles with real positive residues can exist on the jw axis.
5. The poles and zeros of a p.r. function are real or occur in conjugate
pairs. We know that the poles and zeros of a network function are
functions of the elements in the network. Since the elements themselves
are real, there cannot be complex poles or zeros without conjugates
because this would imply imaginary elements.
302 Network analysis and synthesis
6. The highest powers of the numerator and denominator polynomials
may differ at most by unity. This condition prohibits multiple poles and
zeros at s = oo.
7. The lowest powers of the denominator and numerator polynomials
may differ by at most unity. This condition prevents the possibility of
multiple poles or zeros at s = 0.
8. The necessary and sufficient conditions for a rational function with
real coefficients F(s) to be p.r. are
(a) F(s) must have no poles in the right-half plane.
(b) F(s) may have only simple poles on the jw axis with real and positive
residues.
(c) Re F(jw) ~ 0 for all w.
Let us compare this new definition with the original one which requires
the two conditions.
1. F(s) is real when s is real.
2. Re F(s) ~ 0, when Res ~ 0.
In order to test condition 2 of the original definition, we must test
every single point in the right-half plane. In the alternate definition,
condition (c) merely requires that we test the behavior of F(s) along the
jw axis. It is apparent that testing a function for the three conditions
given by the alternate definition represents a considerable saving of
effort, except in simple cases as F(s) = 1/s.
Let us examine the implications of each criterion of the second definition.
Condition (a) requires that we test the denominator of F(s) for roots in
the right-half plane, i.e., we must determine whether the denominator of
F(s) is Hurwitz. This is readily accomplished through a continued fraction
expansion of the odd to even or even to odd parts of the denominator.
The second requirement-condition (b)-is tested by making a partial
fraction expansion of F(s) and checking whether the residues of the poles
on the jw axis are positive and real. Thus, if F(s) has a pair of poles at
s = ±jw., a partial fraction expansion gives terms of the form shown.

K1 K •
--=--+ 1
s -jw1 s + jw1
The residues of complex conjugate poles are themselves conjugates. If
the residues are real-as they must be in order for F(s) to be p.r.-then
K1 == K1 • so that
K1 + K1* - (10.44)
s - jw1 s + jw1 s1 + w11
 Elements of realizability theory 303
If K1 is found to be positive, then F(s) -satisfies the second of the three
conditions.
In order to test for the third condition for positive realness, we must
first find the real part of F(jro) from the original function F(s). To do this,
let us consider a function F(s) given as a quotient of two polynomials

F(s) = P(s) (10.45)

Q(s)
We can separate the even parts from the odd parts of P(s) and Q(s) so
that F(s) is
(10.46)

where Ml..s) is an even function and Nl..s) is an odd function. F(s) is now
decomposed into its even and odd parts by multiplying both P(s) and
Q(s) by M 1 - N 1 so that
F(s) = M1 + N1 Ms - N1
M1 + N 1 M 1 -N1
(10.47)
= M 1 M 1 - N1 N1 + M 1 N1 - M 1 N1
M.1 - N,.1 M11 - N,l
We see that the products M 1 M 1 and N 1 N 1 are even functions, while
Mi N 8 and M 1 Ni are odd functions. Therefore, the even part of F(s) is

Ev [F(s)] = Mi M: - Ni N1 (10.48)
M1 - N 11
and the odd part of F(s) is
Odd [F(s)J = M,. N! - M\ Ni (10.49)
M 1 -N1
If we let s = jro, we see that the even part of any polynomial is real,
while the odd part of the polynomial is imaginary, so that if F(jro) is
written as
F(jro) = Re [F(jro)] + jlm [F(jro)] (10.50)

it is clear that Re [F(jro)] = Ev [F(s)] 1,-,.., (10.51)

and jlm [F(jro)] = Odd [F(s)] 1-,.., (10.52)

Therefore, to test for the third condition for positive realness, we

determine the real part of F(jw) by finding the even part of F(s) and then
lettings= jw. We then check to see whether Re F(jw) ~ 0 for all ro.
304 Network analysis and synthesis
A(IAI}

A(IAI}

"'
FIG. 10.8

The denominator of Re F(jw) is always a positive quantity because

M 1(jro)1 - N.(jro)1 = M 1(ro)1 + N 1(ro)1 ~ 0 (10.53)
That is, there is an extra j or imaginary term in NJ..jw), which, when
squared, gives -1, so. that the denominator of Re F(jw) is the sum of
two squared numbers and is always positive. Therefore, our task resolves
into the problem of determining whether
A(w) ~ Mi(jro) M.(jro) - Ni(jro) N.(jro) ~ 0 (10.54)
Ifwe call the preceding function A(ro), we see that A(ro) must not have
positive, real roots of the type shown in Fig. 10.7; i.e., A(ro) must never
have single, real roots of w. However, A(ro) may have double roots
(Fig. 10.8), because A(ro) need not become negative in this case.
As an example, consider the requirements for

F(s) = s+a (10.55)

1
s + bs + c
to be p.r. First, we know that, in order for the poles and zeros to be in
the left-half plane or on the jw axis, the coefficients a, b, c must be greater
or equal to zero. Second, if b = 0, then F(s) will possess poles on the
jw axis. We can then write F(s) as
F(s) = -1 s - +-0 - (10.56)
s +c s•+c
We will show later that the coefficient a must also be zero when b = 0.
Let us proceed with the third requirement, namely, Re F(jro) ~ 0. From
the equation
M 1(jro) M 1(jro) - N 1(fa,) Nz(jro) ~ 0 (10.57)
we have a( -ro1 + c) + brol ~ 0 (10.58a)
which simplifies to A(ro) = (b - a)wl + ac ~ 0 (10.58b)
It is evident that in order to prevent A(ro) from having positive real roots
of w, b must be greater than or equal to a, that is, b ~ a. As a result,
 Elements of realizability theory 305
when b == 0, then a == 0. To summarize, the conditions that must be
fulfilled in order for F(s) to be positive real are
1. a, b, c ~ 0.
2. b~a.
We see that F 1(s) - s+2 (10.59)
- s1 + 3s + 2
is p.r., while the functions
F(s)==s+l (10.60)
s +2
1 1

F(s)-
1
s+4 (10.61)
- s1 +2s+l
arc not p.r. As a second example, let us determine the conditions for the
biquadratic function
1
F(s) == s + a1s + ao (10.62)
s1 + b1s + b0
to be p.r. We will assume that the coefficients ai, a., bi, b0 are all real,
positive constants. Let us test whether F(s) is p.r. by testing each require-
ment of the second definition. ·
First, if the coefficients of the denominator b1 and b0 arc positive, the
denominator must be Hurwitz. Second, if b1 is positive, we have no poles
on the jro axis. Therefore we can ignore the second condition.
The third condition can be checked by first finding the even part of
F(s), which is
Ev [F(s)] == (s
1
+ a 0)(s8 +1 b0) - 1 a1 1 b1s1
(s1 + b0) - b1 s

(10.63)

The real part of F(fa,) is then

1
Re [FUw)] == ro' - [(ao + bo) -· a1b1Jro + a 0 b0 (10.64)
(-ro• + b0) 1 + b11 w1
We see that the denominator of Re [FUro)] is truly always positive
so it remains for us to determine whether the numerator of Re [F(jw)]
ever goes negative. Factoring the numerator, we obtain
306 Network analysis and synthesis
There are two situations in which Re [F(jw)] does not have a simple real
root.
1. When the quantity under the radical sign of Eq. 10.65 is zero (double,
real root) or negative (complex roots). In other words,
[(ao+ ho) - 01h1]11 - 4aoho ~ 0 (10.66)
or [(ao + ho) - a1h1J2 ~ 4aoh0 (10.67)
If (a0 + h0) - a1h1 ~ 0 (10.68)

then (a 0 + h0 ) - a1h1 ~ 2Jaoho (10.69)

or a1h1 ~ (J;;, - ✓hr,) 2 (10.70)

If + h0) - a1h1 < 0
(Oo (10.71)
then a1h1 - (a0 + ho) ~ 2Jaoho (10.72)
but (a0 + h0) - a 1h1 < 0 < a 1h1 - (a0 + h0) (10.73)

so again a 1h1 ~ (Ja0 - Jho) 2

(10.74)
2. The second situation in which Re [F(jw)] does not have a simple
real root is when 2
wt
in Eq. 10.65 is negative so that the roots are imagi-
nary. This situation occurs when
[(a0 + h0) - a 1h1]2 - 4aoh0 > 0 (10.75)
and (a0 + h0) - a1h1 < 0 (10.76)
From Eq. 10.75 we have
a1h~ - (a0 + h0) > 2Jaoho > (a0 + h0) - a1h1 (10.77)

Thus a 1h1 > (Ja Jh0 ) 2

0 - (10.78)
We thus see that Eq. 10.70 is a necessary and sufficient condition for a
biquadratic function to be positive real. If we have

a1h1 = (J;;,- .Jho)11 (10.79)

then we will have double zeros for Re [F(jw)].
Consider the following example:
1
F(s) = s + 2s + 25 (10.80)
s2 + 5s + 16
We see that
D1h1 =2X5~ (Jao -JhoY1 = (J25 - ✓16) 1 (10.81)
so that F(s) is p.r.
 Elements of reallzablllty theory 307
The examples just given are, of course, special cases. But they do
illustrate the procedure by which functions are tested for the p.r. property.
Let us consider a number of other helpful points by which a function
might be tested quickly. First, if F(s) has poles on the jw axis, a partial
fraction expansion will show if the residues of these poles are positive
and real. For example, 1
F(s) = 3s + 5 (10.82)
s(s + 1)
1

has a pair of poles at s = ±jl. The partial fraction expansion of F(s),

-2s
F(s) =- - + -5 (10.83)
1
S +1 S

shows that the residue of the poles at s = ±j is negative. Therefore F(s)

is not p.r.
Since impedances and admittances of passive time-invariant networks
are p.r. functions, we can make use of our knowledge of impedances
connected in series or parallel in our testing for the p.r. property. For
example, if Zi_(s) and Z,,(s) are passive impedances, then Zi connected in
parallel with Z,, gives an overall impedance

Z(s) = Z1(s) Za(s) (10.84)

Z.(s) + Z1 (s)
Since the connecting of the two impedances in parallel has not affected
the passivity of the network, we know that Z(s) must also be p.r. We see
that if F1(s) and F.(s) are p.r. functions, then

F(s) = F1(s) F.(s) (10.85)

F 1 (s) + F 1(s)
must also be p.r. Consequently, the functions
Ks
F(s)=-- (10.86)
s + 0t
K
and F(s)=-- (10.87)
s + 0t
where ot and Kare real and positive quantities, must be p.r. We then
observe that functions of the type

F(s) = s + {J ot, {J ~ 0
s + ot
= _s- + _{J_ (10.88)
s+ot s+ot
must be p.r. also.
308 Network analysts and synthesis
Finally, let us determine whether
Ks
F(s) =- - at, K ~ 0 (10.89)
1
s + at
is p.r. If we write F(s) as
F(s) = l (10.90)
s/K + a./Ks
we see that the terms s/K and a./Ks are p.r. Therefore, the sum of the
two terms must be p.r. Since the reciprocal of a p.r. function is also p.r.,
we conclude that F(s) is p.r.

10.4 ELEMENTARY SYNTHESIS PROCEDURES

The basic philosophy behind the synthesis of driving-point functions

is to break up a p.r. function Z(s) into a sum of simpler p.r. functions
Zi(s), Z,.(s), ... , Z;,(s), and then to synthesize these individual Z(s) as
elements of the overall network whose driving-point impedance is Z(s).
Z(s) == Zi(s) + Zs(s) + · · · + Z,.(s) (10.91)
First, consider the "breaking-up" process of the function Z(s) into the
sum of functions Z,(s). One important restriction is that all Z.(_s) must be
p.r. Certainly, if all Zh) were given to us, we could synthesize a network
whose driving-point impedance is Z(s) by simply connecting all the Z.(_s)
in series. However, if we were to start with Z(s) alone, how would we
decompose Z(s) to give us the individual Z.(_s)? Suppose Z(s) is given in
general as
a.s" + a,._1s- + · · · + a1s + a 0 == P(s)
1
Z(s) = 00_92)
bms"' + b,,._1, - - + · · · + b1s + b0 Q(s)
1

Consider the case where Z(s) has a pole at s = 0 (that is, b0 = 0). Let us
divide P(s) by Q(s) to give a quotient D/s and a remainder R(s), which
we can denote as Zi(s) and Z,.(s).

Z(s) = D + R(s) D ~0
s
= Z1(s) + Zrf..s) (10.93)
Are Zi and Z,. p.r.? From previous discussions, we know that Zi = DJs
is p.r. Is Z,.(s) p.r.? Consider the p.r. criteria given previously.
1. Z,,(s) must have no poles in the right-half plane.
2. Poles of Zs(s) on the imaginary axis must be simple, and their
residues must be real and positive.
3. Re [Z.(jw)] ~ 0 for all w.
 Elements of realizability theory 309
Let us examine these cases one by one. Criterion 1 is satisfied because
the poles of ZJ..s) arc also poles of Z(s). Criterion 2 is satisfied by this
same argument. A simple partial fraction expansion does not affect the
residues of the other poles. When s == jm, Re [Z(jm) == D/jm] == 0.
Therefore we have
Re z.(jm) == Re Z(jw) ~ 0 (10.94)
From the foregoing discussion, it is seen that if Z(s) has a pole at s == 0,
a-partial fraction expansion can be made such that one of the terms is of
the form K/s and the other terms combined still remain p.r.
A similar argument shows that if Z(s) has a pole at s == oo (that is,
n - m == 1), we can divide the numerator by the denominator to give a
quotient Ls and a remainder term R(s), again denoted as Zi(s) and Zs(s).
Z(s) = Ls + R(s) == Zi(s) + Zs(s). (10.95)
Here Zs(s) is also p.r. If Z(s) has a pair of conjugate imaginary poles on
the imaginary axis, for example, poles at s == ±ja>i, then Z(s) can be
expanded into partial fractions so that
2Ks
Z(s) == + Z.(s) (10.96)
S
1
+ W11
Here Re ( 2Ks I ) == Re ( j2mK ) == O (10.97)
S
I
+ W1 •-I• -W
I
+ W1 I

so that Zs(s) is p.r.

Finally, if Re (Z(jm)] is minimum at some point m,, and if the value of
Re 'Z(jmJ == K, as shown in Fig. 10.9, we can remove a constant K ~ X.
from Re [Z(jw)] so that the remainder is still p.r. This is because Re [Z(jm)]
will still be greater than or equal to zero for all values of m.
Suppose we have a p.r. function Z(s), which is a driving-point im-
pedance function. Let Z(s) be decomposed, as before, so that
Z(s) == Z,.(s) + Zs(s) (10.98)
ReZ(jr.,J

FIG. lo.t
310 Network analysis and synthesis

Z(s)).

FIG.10.10

where both Zi and Zt are p.r. Now let us "remove" Zi(s) from Z(s) to
give us a remainder z,,(s). This removal process is illustrated in Fig.
10.10 and shows that removal corresponds to synthesis of Zi(s).
Example 10.5. Consider the following p.r. function
s2+2s+6
Z(s) = s<..s + 3) (10.99)

We see that Z(s) has a pole at s = 0. A partial fraction expansion of Z(s) yields
2 s
Z(s) = s+ s + 3
= Z 1(s) + Z 1(s) (10.100)
H we remove Z 1(s) from Z(s), we obtain Z 1(s), which can be shown by a resistor
in parallel with an inductor, as illustrated in Fig. 10.11.
tf ,-
o------ilt------+-....----,

10

L__
FIG. 10.11

Example 10.6
7s +2
Y(s) = - - (10.101)
2s +4
where Y(s) is a p.r. function.
Let us synthesize the network by first removing min [Re Y(jw)]. The real
part of Y(jw) can be easily obtained as
8 + 14w1
Re [Y(iw)] = 16 + 4w• (10.102)

We see that the minimum of Re [Y(iw)] oocurs at w = 0, and is equal to

min [Re Y(jw)] = ½. Let us then remove Y1 -= ½mho from Y(s) and denote
 Elements of realizability theory 311
the remainder as Y,l_s), as shown in Fig. 10.12. The remainder function Y1(s) is
p.r. because we have removed only the minimum real part of Y,l_jw). Y.(s)
is obtained as
1 3s
Y,l_s) = Y(s) - - = - - (10.103)
2 s+2
It is readily seen that Y,l_s) is made up of a l-'1 resistor in series with a f-farad
capacitor. Thus the final network is that shown in Fig. 10.13.

20

-
20
Y(s) YJ!!..

FIG. 10.12 FIG. 18.13

Example 10.7. Consider the p.r. impedance

6s3+3r+3s+l
Z(s) = 6s3 + 3s (10.104)

The real part of the function is a constant, equal to unity. Removing a constant
of 1 n, we obtain (Fig. 10.14)
3r +1
Zi{s) == Z(s) - 1 = 6s3 + ls (10.10S)
The reciprocal of Z 1(s) is an admittance

6s3 + 3s (10.106)
Y1(s) = 3r + 1
which has a pole at s == co. This pole is removed by finding the partial fraction
expansion of Y1(s);
(10.107)

10

;;J______I FIG. 18.14

312 Network analysis and synthesis
10

2f
Z!!!_. !!._.

FIG. l0.15

and then by removing the term with the pole at s = co to give a capacitor of
2 farads in parallel with Y1(s) below (Fig. 10.15). Y.(s) is now obtained as
s
Y.(s) = Y1(s) - 2s = -_- -
3.r11 + 1
(10,108)
The reciprocal of Y.(s) is
1
Z.(s) = 3s +-s (10.109)

which is, clearly, an inductor of 3 h in series with a capacitor of 1 farad. The

final network is shown in Fig. 10.16.

-Z(s)

FIG. 10.16

These examples are, of course, special cases of the driving-point syn-

thesis problem. However, they do illustrate the basic techniques involved.
In the next chapter, we will discuss the problem of synthesizing a network
with two kinds of elements, either L-C, R-C, or R-L networks. The syn-
thesis techniques involved, however, will be the same.

Problems
10.1 Test the following polynomials for the Hurwitz property.
(a) s1+s1+2s+2
(b) s'+r+s+l
(c) s7+s> +s3 +s
(d) s3+4s3+5.r+2
(e) s>+2s3+.r
(/) .r7 +2t' +ls' +s' +4.fl +8.fl +8.r +4
 Elements of reallzablllty theory 313
18.2 Determine whether the following functions are p.r. For the functions
with the denominator already factored, perform a partial fraction expansion
first.
s2 + 1
(a) F(s) = s3 + 4a

2fl+2a+4
(b) F(s) = (s + l)(r + 2)

(c) F(s) = (s + 2)(s + 4)

(s + l)(s + 3)

r1 +4
(d) F(s) = s8 + 3.r + 3s + 1

(e) F(s) _Sr+ s

s2 + 1
10.3 Suppose F 1(s) and F.<.s) are both p.r. Discuss the conditions such that
F(s) = F 1(s) - F,l..s) is also p.r.
10.4 Show that the product of two p.r. functions need not be p.r. Also show
that the ratio of one p.r. function to another may not be p.r. (Give one example
of each.)
. r + Xs
10.S Given Z(s) - al + Ss + 4 :

(a) What are the restrictions on X for Z(s) to be a p.r. function?

(b) Find Xfor Re [Z(Jcu)] to have a second-order z:ero at cu - 0.

(c) Cioosc a numerical value for X and synthesir.e Z(s).
10.6 Prove that if Z 1(s) and Z.(.s) are both p.r.,

Z(s) _ Z 1(s)Z,1..s)
Z 1(s) +Z,l..s)
must also be positive real.

10.7 Z(s) - ',: +s

+s+
+: is p.r. Determine min [ReZ(jw)] and synthesir.e
Z(s) by first removing min [ReZ(jw)].
10.8 Perform a continued fraction expansion on the ratio

Yf'=s3+2r+3s+l
,s, s3 + s2 + 2a + 1

What does the continued expansion imply if Y(s) is the driving-point admittance
of a passive network? Draw the network from the continued fraction.
314 Network analysis and synthesis
10.9 The following functions are impedance functions. Synthesi7.e the
impedances by successive removals ofjw axis poles or by removing min [Re (jw)].
s8 + 4s
(a) s" +2
s+l
(b)
s(s + 2)
2s +4
(c)
2s + 3
s" + 3s + 1
(d)
s" + 1
 chapter 11
Synthesis of one-port networks
I , with two kinds of elements

In this chapter we will study methods for synthesizing one-port net-

works with two kinds of elements. Since we have three elements to choose
from, the networks to be synthesized are either R-C, R-L, or L-C networks.
We will proceed according to the following plan. First we will discuss the
properties of a particular type of one-port network, and then we will
synthesize it. Let us first examine some properties of L-C driving-point
functions.

I I.I PROPERTIES OF L-C IMMITTANCE FUNCTIONS

Consider the impedance Z(s) of a passive one-port network. Let us

represent Z(s) as
Z(s) = Mi(s) + N1(s) (11.1)
M 1(s) + NJ..s)
where M1 , M1 are even parts of the numerator and denominator, and
Ni, N 1 are odd parts. The average power dissipated by the one-port is
Average power=½ Re [Z(jw)] 111 1 (11.2)
where I is the input current. For a pure reactive network, it is known that
the power dissipated is zero. We therefore conclude that the real part of
Z(jw) is zero; that is
Re Z(jw) = Ev Z(jw) = 0 (11.3)

where Ev Z(s) = Mi(s) M:(s) - N1!s) N1(s) (11.4) .

M 1 (s) - N 1 (s)
31S
316 Network analysis and synthesis
In order for Ev Z{jro) == 0, that is,
Mi(jw) M,.(jro) - N.{jw) N,J.jro) - 0 (11.5)
either of the following cases must hold:
(a) M1 == 0 == N1
(11.6)
(b) M 1 == 0 = N1
In case (a), Z(a) is
Z(s) == N1 (11.7)
M,

and in case (b) Z(s) == M1 (11.8)

N,
We see from this development the following two properties of L-C
functions:
1. Z :w(a) or Y:w(s) is the ratio of even to odd or odd to even poly-
nomials.
2. Since both M.(s) and N ,(s) are Hurwitz, they have only imaginary
roots, and it follows that the poles and zeros of Z:w(s) or YLo(s) are on
the imaginary axis.
Consider the example of an L-C immittance function given by

Z(s) == at9' + a.s• + a 0 (11.9)

b,.s1 + b.s1 + b1s
Let us examine the constraints on the coefficients a, and b1• We know,
first of all, that in order for the impedance to be positive real, the coeffi-
cients must be real and positive. We also know that an impedance
function cannot have multiple poles or zeros on the jw axis. Since co is
defined to be on the jro axis, the highest powers of the numerator and the
denominator polynomials can differ by, at most, unity. For example, if
the highest order of the numerator is 2n, then the highest order of the
denominator can either be 2n - 1 so that there is a simple pole at s == co,
or the order can be 2n + 1 so that there is a simple zero at s == co.
Similarly, the lowest orders of numerator and denominator can differ by
at most unity, or else there would be multiple poles or zeros of Z(s) at
s == 0.
Another property of the numerator and denominator polynomials is
that if the highest power of the polynomial is 2n, for example, the next
highest order term must be 2n - 2, and the succeeding powers must
differ by two orders all the way through. There cannot be any missing
 Synthesis of one-port networks 317
terms, i.e., no two adjacent terms of either polynomial may differ by more
than two powers. For example, for Z(s) given in Eq. 11.9, if b1 - 0,
then Z(s) will have poles when
br,S" + b1s = 0 (11.10)
so that the poles will be at s = 0 and at

Src = (::r ef(lk-1) ../t k = 0, 1, 2, 3 (11.11)

It is clearly seen that none of the poles s,. are even on the jw axis, thus
violating one of the basic properties of an L-C immittance function.
From the properties given in Eq. 11.11, we can write a general L-C
impedance or admittance as

Z(s) = K(s' +
wi')(s' + o,a') · · · (s' + wh • • • (11.12)
s(s' + wl)(s' + w/) · · · (s' + w 1') • • •
Expanding Z(s) into partial fractions, we obtain

Z(s) = K0 + 2Kas + 2K.s + ... + K.,,s (11.13)

s s
1
+ w11 s' + w/
where the K, are the residues of the poles. Since these poles are all on
the jw axis, the residues must be real and positive in order for Z(s) to be
positive real. Letting s = jw, we see that Z(jw) has 7.Cl'O real part, and
can thus be written as a pure reactance jX(w). Thus we have

Z(jw) =j(- Ko+

w w
2K■w
w 1
1
-
1
+ ... + K.,,w)

=jX(w) (11.14)
Differentiating X(w) with respect tow, we have
1
dX(w) = K 0 + K.,, + 2KJ.,.w + w1') + ... (11.15)
dw w• ( w11 - w')1
Since all the residues K, are positive, it is seen that for an L-C function,

dX(w) ~ O (11.16)
dw
A similar development shows that the derivative of Im [Y(Jw)] == .B(co)
is also positive, that is,
dB(w) ~ O (11.17)
dco
318 Network analysis and synthesis
Consider the following example. Z(s) is given as

Z(s) = Ks(s' + wl) (11.18)

(s1 + w18}(s1 + w,8)
Lettings= jw, we obtain X(w) as

ZUw) =j X(w) = +i Kw(-w• + wa8> (11.19)

(-w1 + w,.'')(-w1 + w,8)
Let us draw a curve of X(w) versus w. Beginning with the zero at w = 0,
let us examine the sequence of critical frequencies encountered as w
increases. Since the slope of the X(w) curve is always positive, the next
critical frequency we encounter is when X( w) becomes infinitely large or
the pole is at w1 • As we pass w 1, X(w) changes sign and goes from +
to -. In general, whenever we pass through any critical flequency,
there is always a change of sign, as seen from the way j X(w) is written in
the last equation. After we pass through w 2, with the slope of X( w)
always positive, it is easy to see that the next critical frequency is the zero
at w 8• Thus, if an impedance function is an L-C immittance, the poles
and zeros of the function must alternate. The particular X( w) under
discussion takes the form shown in Fig. 11 .1. Since the highest powers
of the numerator and the denominator always differ by unity, and the
lowest powers also differ by one, we observe that at s = 0 and at s = oo,
there is always a critical frequency, whether a zero or a pole.
For the example just discussed, there is a zero at s = 0 and a zero at
s == oo. The critical frequencies at s = 0 ands= oo are called external
critical frequencies, whereas the remaining finite critical frequencies are
referred to as internal. Thus, in the previous example, w 11, w 3, and w,
are internal critical frequencies. _
Finally\ let us summarize the properties of L-C impedance or admittance
functions.
1. ZL<;(s) or YLo(s) is the ratio of odd to even or even to odd poly-
nomials.

00

"'

FIG. II.I
 Synthesis of one-port networks 319

X(w)

-x--o---x--o------x-
0 +1 +2 +3 00
w

FIG. 11.2

2. The poles and zeros are simple and lie on the jw axis.
3. The poles and zeros. interlace on the jw axis.
4. The highest powers of numerator and denominator must differ by
unity; the lowest powers also differ by unity.
5. There must be either a zero or a pole at the origin and infinity.
The following functions are not L-C for the reasons listed at the left.

Ks(s + 4)
2
Z s _
3.
( ) - (s 2 + l)(s 2 + 3)

= s + 4s +1 5s
5 3
2. Z(s) (11.20)
3s + 6s 4

Z(s) = K(s + l)(s + 9)

1
2
1.
(s 2 + 2)(s 2 + 10)
On the other hand, the function Z(s) in Eq. 11.21, whose pole-zero
diagram is shown in Fig. 11.2, is an L-C immittance.

Z(s) = 2(s2 + l)(s2 + 9) (11.21)

s(s + 4) 2

I 1.2 SYNTHESIS OF L-C DRIVING-POINT IMMITTANCES

We saw in Section 11.1 that an L-C immittance is a positive real function

with poles and zeros on the jw axis only. The partial fraction expansion
of an L-C function is expressed in general terms as

F(s) = Ko + 2Ks5 + ... + Krr,s (11.22)

s s2 + w/·
The synthesis is accomplished directly from the partial fraction expansion
by associating the individual terms in the expansion with network elements.
If F(s) is an impedance Z(s), then the term Ko/s represents a capacitor of
1/Ko farads; the term Krr,s is an inductance of Ker, henrys, and the term
320 Network analysis and synthesis

FIG. 11.3

2K,s/(s2 + ro;") is a parallel tank circuit that consists of a capacitor of

l/2K, farads in parallel with an inductance of 2Kiwl'. Thus a partial
fraction expansion of a general L-C impedance would yield the network
shown in Fig. 11.3. For example, consider the following L-C function.

Z(s) = 2(s2 + l)(s2 + 9) (11.23)

s(s + 4)
8

A partial fraction expansion of Z(s) gives

Z(s) = 2s + -I + -lf-s
- (11.24)
s s1 + 4
We then obtain the synthesized network in Fig. 11.4.
The partial fraction expansion method is based upon the elementary
synthesis procedure of removing poles on the jro axis. The advantage
with L-C functions is that all the poles of the function lie on the jw axis
so that we can remove all the poles simultaneously. Suppose F(s) in
Eq. 11.22 is an admittance Y(s). Then the partial fraction expansion of
Y(s) gives us a circuit consisting of parallel branches shown in Fig. 11.5.
For example,
Y(s) = s(s1 + 2)(s1 + 4) (11.25)
(s1 + l)(s1 + 3)

z~

FIG. 11.4
 Synthesis of one-port networks 321

K..,f

FIG. 11.5

The partial fraction expansion of Y(s) is

Y(s) = s +-- ½s +-- is (11.26)

1
s1 + 3 s +1
from which we synthesize the network shown in Fig. 11.6. The L-C
networks synthesized by partial fraction expansions are sometimes called
Foster-type networks. 1 The impedance form is sometimes called a Foster
series network and the admittance form is a Foster parallel network.
A useful property of L-C immittances is that the numerator and the
denominator always differ in degree by unity. Therefore, there is always
a zero or a pole at s = oo. Suppose we consider the case of an L-C
impedance Z(s), whose numerator is of degree 2n and denominator
is of degree 2n - 1, giving Z(s) a pole at s = oo. We can remove this
pole by removing an impedance L 1s so that the remainder function Z,.(s)
is still L-C:
(11.27)
The degree of the denominator of Zz(s) is 2n - 1, but the numerator is of
degree 2n - 2, because the numerator and denominator must differ in
degree by 1. Therefore, we see that Z 1(s) has a zero at s = oo. If we
invert Z 1(s) to give Y1(s) = 1/Z,.(s), Y1(s) will have a pole at s = oo,
which we can again remove to give a capacitor Cz9 and a remainder
Ya(s), which is
(I 1.28)

J
0 I FIG. 11.6

1
~. M. Foster, "A Reactance Theorem," Bell System Tech.J., No. 3 (1924), 2S9-261.
322 Network analysis and synthesis

FIG. 11.7 FIG. 11.8

We readily see that Y3(s) has a zero at s = oo, which we can invert and
remove. This process continues until the remainder is zero. Each time
we remove a pole, we remove an inductor or a capacitor depending upon
whether the function is an impedance or an admittance. Note that the
final structure of the network synthesized is a ladder whose series arms
are inductors and whose shunt arms are capacitors, as shown in Fig. 1I. 7.
Consider the following example.
6
Z(s) = 2s + 12s'1 + 16s (ll. 29)
+ 4s + 3
s'
We see that Z(s) has a pole at s = oo, which we can remove by first
dividing the denominator into the numerator to give a quotient 2s and a
remainder Zs(s), as shown in Fig. 11.8. Then we have

4s3 + 10s
Z 1(s) = Z(s) - 2s = - - - - (11.30)
s' + 4s + 3
1

Observe that Zs(s) has a zero at s = oo. Inverting Zs(s), we again remove
the pole at infinity. Then we realize a capacitor of! farad and a remainder
Y3(s), as may be seen in Fig. 11.9.
1
Y3(s) = Y2(s) - - s = ----'--
is1 + 3 (11.31)
4 3
4s + 10s
Removing the pole at s = oo of Zs(s) = 1/ Y3(s), gives a series inductor
of lh and
8
Zis) = Za(s) - - s = -.l-12s
-- (11.32)
3 IS +3

FIG.11.9
 Synthesis of one-port networks 323

Remainder

FIG. 11.10

as shown in Fig. 11.10. The admittance Y,(s) = 1/Z.(s) has a pole at

s = oo, which we remove to give a capacitor of f farad and a remainder
Y&(s) = 3/2s, which represents an inductor of f h. Removing this in-
ductor gives us zero remainder. Our synthesis is therefore complete and
the final network is shown in Fig. 11.11.
Since we always remove a pole at s = oo by inverting the remainder
and dividing, we conclude that we can synthesize an L-C ladder network
by a continued fraction expansion. The quotients represent the poles at
s = oo, which we remove, and we invert the remainder successively until
the remainder is zero. For the previous example, the continued fraction
expansion is

s' + 4s2 + 3)2s6 + 12s3 + l6s(2s +--+ Z

2s5 + 8s3+ 6s
4s3 + lOs)s' + 4s2 + 3(¼s +--+ Y
s' + !s2
fs 2 + 3}4s3 + lOs(ts +--+ Z
4s3 + 8s
2s~s2 + 3(ls +--+ Y
fs 2
3)2s(is+--+ Z
2s

We see that the quotients of the continued fraction expansion give the
elements of the ladder network. Because the continued fraction expansion

FIG. II.II
324 Network analysis an synthesis
always inverts each remainder and divides, the successive quotients alter-
nate between Z and Y and then Z again, as shown in the preceding
expansion. If the initial function is an impedance, the first quotient must
necessarily be an impedance. When the first function is an admittance,
the first quotient is an admittance.
Since the lowest degrees of numerator and denominator of an L-C
admittance must differ by unity, it follows that there must be a zero or a
pole at s = 0. If we follow the same procedure we have just outlined,
and remove successively poles at s = 0, we will have an alternate real-
ization in a ladder structure. To do this by continued fractions, we
arrange both numerator and denominator in ascending order and divide
the lowest power of the denominator into the lowest power of the numer-
ator; then we invert the remainder and divide again. For example, in
the case of the impedance we have

Z(s) = (ss + l)(s• + 3) (11.33)

s(s1 + 2)
The continued fraction expansion to give the alternate realization is
2s + s3)3 + 4s1 + s'(3/2s +-+ Z
3 + Js2

fs + s'}2s + s3(4/5s +-+ Y

2

2s + fs3
ls3)ts + st(2s/2s +-+ z
2

ts 2

s'}ls3(I/5s +-+ Y
¼s3

The final synthesized network is shown in Fig. 11.12. The ladder networks
realized are called Cauer ladder networks because W. Cauer2 discovered
the continued fraction method for synthesis of a passive network.
Note that for both the Foster and the
Cauer-form realizations, the number of
elements is one greater than the number of
internal critical frequencies, which we defined
previously as being all the poles and zeros
of the function, excluding those at s = 0 and
FIG. 11.12 s = co. Without going into the proof of the
• Wilhelm Cauer, "The Realization of Impedances with Prescribed Frequency
Dependence," Arch. Electrotech., IS (1926), 355-388.
 Synthesis of one-port networks 325
statement, it can be said that both the Foster and the Cauer forms give
the minimum number of elements for a specified L-C driving-point
function. These realizations are sometimes known as canonical forms.

11.3 PROPERTIES OF R-C DRIVING-POINT IMPEDANCES

The properties of R-C driving-point impedances can be derived from

known properties of L-C functions by a process of mapping the jw axis
onto the -a axis. 3 We will not resort to this formalism here. Instead, ·
we will assume that all driving-point functions that can be realized with
two kinds of elements can be realized in a Foster form. Based upon this
assumption we can derive all the pertinent properties of R-C or R-L
driving-point functions. Let us consider first the properties of R-C
driving-point impedance functions.
Referring to the series Foster form for an L-C impedance given in
Fig. 11.3, we can obtain a Foster realization of an R-C impedance by
simply replacing all the inductances by resistances so that a general R-C
impedance could be represented as in Fig. 11.13. The R-C impedance,
as seen from Fig. 11.13, is

Z(s) = Ko + Kcx, + ~ + ~ + · · · (11.34)

s s + <11 s + <11
where C0 = 1/Ko. R<XJ = K 00 , C1 = 1/Ki, R1 = K 1/ai,
and so on. In
order for Eq. 11.34 to represent an R-C driving-point impedance, the
constants K, and a, must be positive and real. From this development,
two major properties of R-C impedances are obtained, and are listed in
the following.

R..,

FIG. 11.13

• M. E. Van Vallcenburg, I11troductio11 to Modern Network Synthesis, John Wiley and

Sons, New York, 1960, pp. 140-145.
326 Network analysis and synthesis
1. The poles of an R-C driving-point impedance are on the negative
real (-a) axis. It can be shown from a parallel Foster form that the
poles of an R-C admittance function are also on the axis. We can thus
conclude that the zeros of an R-C impedance are also on the -a axis.
2. The residues of the poles, K,, are real and positive. We shall see later
that this property does not apply to R-C admittances.
Since the poles and zeros of R-C impedances are on the - a axis, let
us examine the slope of Z(a) along the -a axis.- To find the slope,
dZ(a)/da, we first let s = a in Z(s), and then we take the derivative of
Z(a) with respect to a. Thus we have

Z(a) = -Ko + K + -Ki- + -Ka- + · · · (11.35)

a + 0'1 O' + 0'2
00
a
and dZ(a) = _ K + -K1 + -K2 + ...
0
(11.36)
da a2 (a + o-1) 2 (a+ o-2) 2
It is clear that dZ(a) ::5: 0 (11.37)
da
Let us now look at the behavior of Z(s) at the two points where the
real axis and the imaginary axis intersect, namely, at a = w = 0 and at
a = w = oo. This is readily done by examining the general R-C network
in Fig. 11.13 at these two frequencies. At a= 0, (d-c), if the capacitor C0
is in the circuit, it is an open circuit and there is a pole of Z(s) at a = 0.
If C0 is not in the circuit, then Z(0) is simply the sum of all the resistances
in the circuit.
Z(0) = R 1 + R 2 + · · · + R 00 (11.38)
because all of the capacitors are open circuits at a = 0.
At a = oo, all the capacitors are short circuits. Thus, if R 00 is in the
circuit, Z(oo) = R 00 • If R 00 is missing, then Z(oo) = 0. To summarize
these last two statements, we have
oo, C0 present
Z(0) = m
( IR,, C0 missing
i-1

0, R 00 missing
Z(oo) = (R
00 , R 00 present
If we examine the two cases for Z(0) and Z( oo), we see that
Z(0) :2:: Z(oo) (11.39)
 Synthesis of one-port networks 327
Next, let us see whether the poles and zeros of an R-C impedance
function alternate. We have already established that the critical frequency
nearest the origin must be a pole and the critical frequency nearest a = oo
must be a zero. Therefore, if Z(s) is given as

Z(s) = (s + a 8)(s + a,) (11.40)

(s + '11)(s + aa)
Then, if Z(s) is R-C, the singularity nearest the origin must be a pole
which we will assume to be at s = -111 ; the singularity furthest from
the origin must be a zero, which we will take to bes= -a,. Let us plot

versus -a, beginning at a= 0 and extending to a= -oo. At a= 0,

Z(O) is equal to a positive constant
Z(O) = 11111' (11.42)
'11'11
Since the slope of Z(a) is always positive as - a increases, Z(a) must
increase until the pole s = -111 is reached (Fig. 11.14). At a= -111 ,
Z(a) changes sign, and is negative until the next critical frequency is
reached. We see that this next critical frequency must be the zero,
s = -112 . Since Z(a) increases for increasing -a, the third critical
frequency must be the poles= -113 • Because Z(a) changes sign at -113,
the final critical frequency must be the zero, s = -a,. Beyond a= -a,,
the curve becomes asymptotic to Z( oo) = 1. From this analysis we see
that the poles and zeros of an R-C impedance must alternate so that for

Z(uJ
I
I I
I I
I I
I I
I I
Z(oo)-------1-- -----+----
1 I
-~---x--~c,----X---+---
1113 1111
I I
I I
I I
I I
I I
I I
FIG. 11.14
328 Network analysis and synthesis
the case being considered
oo > a, > <1s > <11 > <11 ~ 0. (11.43)

In addition, we see that (11.44)

which shows that Z(O) > Z( oo).

To summarize, the three properties we need to recognize an R-C
impedance are:
1. Poles and zeros lie on the negative real axis, and they alternate.
2. The singularity nearest to (or at) the origin must be a pole whereas
the singularity nearest to (or at) a= -oo must be a zero.
3. The residues of the poles must be real and positive.
An example of an R-C impedance is:
Z(s) = (s + l)(s + 4)(s + 8) (11.45)
s(s + 2)(s + 6)
The following impedances are not R-C.
Z(s) = (s + l)(s + 8)
(s + 2)(s + 4)
Z(s) = (s + 2)(s + 4) (11.46)
(s + 1)
Z(s) = (s + l)(s + 2)
s(s + 3)
Let us reexamine. the partial fraction expansion of a general R-C
impedance.
K0 K,
F(s) = - + K 00 + - - + · · · (11.47)
s s + a,
Instead of letting F(s) represent an impedance, consider the case where
F(s) is an admittance Y(s). If we associate the individual terms in the
expansion to network elements, we then obtain the network shown in

FIG. II.IS
 Synthesis of one-port networks 329
Fig. 11.15. We see that an R-C impedance, ZRo(s), also can be realized
as an R-L admittance Y RL(s). All the properties of R-L admittances are
the same as the properties of R-C impedances. It is therefore important
to specify whe~er a function is to be realized as an R-C impedance or an
R-L admittance.

11.4 SYNTHESIS OF R-C IMPEDANCES OR R-L ADMITTANCES

We postulated in Section 11.3 that the Foster form realization exists

for an R-C impedance or an R-L admittance. Since Foster networks are
synthesized by partial fraction expansions, the synthesis is accomplished
with ease. An important point to remember is that we must remove the
minimum real part of Z(jro) in the partial fraction expansion. It can be
shown' that min [Re Z(jro)] = Z( oo), so that we have to remove Z( oo)
as a resistor in the partial fraction expansion. In cases where the numer-
ator is of lower degree than the denominator, Z( oo) = 0. When the
numerator and the denominator are of the same degree, then Z( oo) can
be obtained by dividing the denominator into the numerator. The
quotient is then Z( oo). Consider the following example.

F(s) = 3(s + 2)(s + 4) (11.48)

s(s + 3)
The partial fraction expansion of the remainder function is obtained as

F(s) = ~ + - 1- + 3 (11.49)
s s+3
where F( oo) = 3. If F(s) is an impedance Z(s), it must be an R-C im-
pedance and it is realized in the series Foster form in Fig. 11.16. On the
other hand, if F(s) represents an admittance, we realize Y(s) as an R-L
network in the parallel Foster form (Fig. 11.17).

1f
z~

FIG. 11.16

'Van Valkcnburg, foe cit.

330 Network analysis and synthesis

30

¼o
lh

FIG. 11.17

An alternate method of synthesis is based on the following fact. If

we remove min Re [Z(jw)] = Z(oo) from Z(s), we create a zero at s = oo
for the remainder Z 1(s). If we invert Z 1(s), we then have a pole at s = oo,
which we can remove to give Z 2(s). Since min Re [Ys(jw)] = Y1(oo), if
we remove Y1( oo), we would have a zero at s = oo again, which we
again invert and remove. The process of extracting Z( oo) or Y( oo) and
the removal of a pole of the reciprocal of the remainder involve dividing
the numerator by the denominator. Consequently, we see that the whole
synthesis process can be resolved by a continued fraction expansion.
The quotients represent the elements of a ladder network.
For example, the continued fraction expansion of F(s) in Eq. 11.48 is

s2 + 3s)3s1 + 18s + 24(3

3s + 9s
2

9s + 24)s1 + 3s(½s
s" + fs
½s)9s + 24(27
9s
24)½s(s/72
ls
If F(s) is an impedance Z(s), the resulting network is shown in Fig.
11.18. If F(s) is an admittance Y(s), we have the R-L network of Fig. 11.19.

to

FIG. 11.18 FIG. 11.19

 Synthesis of one-port networks 331

11.5 PROPERTIES OF R-l IMPEDANCES AND

R-C ADMITTANCES

The immittance that represents a series Foster R-L impedance or a

parallel Foster R-C admittance is given as

K,s
F(s) = KJX)s + K0 + - - + · · · (11.50)
s + <1,
The significant difference between an R-C impedance and an R-L im-
pedance is that the partial fraction expansion term for the R-C "tank"
circuit is K'1(s + u,); whereas, for the R-L impedance, the corresponding
term must be multiplied by an s in order to give an R-L tank circuit
consisting of a resistor in parallel with an inductor.
The properties of R-L impedance or R-C admittance functions can be
derived in much the same manner as the properties of R-C impedance
functions. Without going into the derivation of the properties, the more
significant ones are given in the following:

1. Poles and zeros of an R-L impedance or R-C admittance are located

on the negative real axis, and they alternate.
2. The singularity nearest to (or at) the origin is a zero. The singularity
nearest to (or at) s = oo must be a pole.
3. The residues of the poles must be real and negative.

Because of the third property, a partial fraction expansion of an R-L

impedance function would yield terms as

_ _&_ (11.51)
s + u,.
This does not present any trouble, however, because the term above does
not represent an R-L impedance at all. To obtain the Foster form of an
R-L impedance, we will resort to the following artifice. Let us first expand
ZC.s)/s into partial fractions. If ZC.s) is an R-L impedance, we will state
without proof here that the partial fraction expansion of ZC.s)/s yields
positive residues. 6 Thus, we have ·
Z(s)0 K K,
-=-+K«>+--+··· (U.52)
s s s + <1,
• Actually, ZIU,(s)/s has the properties of an R-C impedance; see Van Valkenburg
loc. cit.
332 Network analysis and synthesis

i-t lit
th ,Ah 20
40
Z(1)
Y(1)

FIG. 11.20 FIG. 11.21

where Ko, Ki, ... , K,,; ~ 0. If we multiply both sides by s, we obtain

Z(s) in the desired form for synthesis. Consider the following function:

F(s) = 2(s + l)(s + 3) (11.53)

(s + 2)(s + 6)
F(s) represents an R-L impedance or an R-C · admittance because it
satisfies the first two criteria cited. The partial fraction expansion of
F(s) is 1. u
F{s) = 2 - - - · - - •-
8
(11.54)
s+2 s+6
so we see that the residues are negative. The partial fraction expansion of
F(s)/s, on the other hand, is
F(s) = 2(s + l){s _:_ 3) = ! + __J_ + _L (11.55)
s s(s + 2){s + 6) s s+ 2 s+ 6
If we multiply both sides by s, we obtain
1 ¼s ¾s
F(s) =-+--+-- (11.56)
2 s+2 s+6
If F(s) represents an impedance Z(s), it is synthesized in series Foster
form, giving the R-L network in Fig. 11.20. If F(s) is an admittance Y(s),
then the resulting network is the R-C network shown in Fig. 11.21.
To synthesize an R-L impedance in ladder form, we make use of the
fact that min Re [Z(jw)] = Z(0). If we remove Z(0) from Z(s), the
remainder function Zi(s) will have a zero at s = 0. After inverting Zi(s),

FIG. 11.22
 Synthesis of one-port networks 333

-
Y(a) 20

FIG. 11.23

we can then remove the pole at s = 0. Since the value Z(O) is obtained
by dividing the lowest power of the denominator into the lowest power
term of the numerator, the synthesis could be carried out by a continued
fraction expansion by arranging the numerator and denominator poly-
nomials in ascending order and then dividing. For example, the following
function is e~ther an R-C impedance or an R-C admittance.
1
F(s) = 2(s +l}(s + 3) =6+ 8s + 2s (ll.S?)
(s + 2)(s + 6) 12 + 8s + s1
The continued fraction expansion of F(s) is

12 + 8s + s8)6 + 8s + 2sl(l
6 + 4s + ½s1
4s + is1}12 + 8s + r{.3/s
12 + -ls
· ls + s8)4s + fsl(f-
4s + -fsl
f-4 s1)½s + s1(49/5s
is

If F(s) is an impedance function, the resulting network is the R-L network

shown in Fig. 11.22. If, on the other hand, F(s) is an R-C admittance
Y(s), the network is synthesized as in Fig. 11.23.

11.6 SYNTHESIS OF CERTAIN R-L-C FUNCTIONS

Under certain conditions, R-L-C driving-point functions may be syn-

thesized with the use of either partial fractions or continued fractions.
For example, the function
Z(s) = s•+ 2s + 2 (11.58)
s1 +s+1
334 Network analysis and synthesis
is neither L-C, R-C, nor R-L. Nevertheless, the function can be synthesized
by continued fractions as shown.

s 2 + s + l)s2 + 2s + 2(1 +- Z
s2 + s+l
s + l)s2 + s + l(s +- Y
S
2
+S
l)s + l(s + 1 +- Z
s+l

The network derived from this expansion is given in Fig. 11.24.

In another case, the poles and zeros of the following admittance are all
on the negative real axis, but they do not alternate.

Y(s) = (s + 2)(s + 3) (11.59)

(s + l)(s + 4)
lh The partial fraction expansion for Y(s) is

FIG. 11.24
Y(s) = 1 + _f_ + -f (11.60)
s+l s+4
Since one of the residues is negative, we cannot use this expansion for
synthesis. An alternate method would be to expand Y(s)/s and then
multiply the whole expansion by s.

Y(s) ½ f ¼
-s= -s - - - +s+4
s+1
-- (11.61)

When we multiply by s, we obtain,

fs
Y(s) = -3 - -. - + -¼s- (11.62)
2 s+1 s+4
Note that Y(s) also has a negative term. If we divide the denominator of
this negative term into the numerator, we can rid ourselves of any terms
with negative signs.

3 (2
Y(s)=-- - - -f -) ¼s
+--
2 3 s+1 .__ s+4
5 f ¼s (11.63)
=-+--+--
6 s+1 s+4
 Synthesis of one-port networks 335

FIG.11.15

The network that is realized from the expanded function is given in Fig.
11.25.
If we try to expand Y(s) by continued fractions, we see that negative
quotients result. However, we can expand Z(s) = 1/Y(s) by continued
fractions, although the expansion is not as simple or straightforward as
in the case of an R-C function, because we sometimes have to reverse the
order of division to make the quotients all positive. The continued
fraction expansion of Z(s) is

6 + Ss + s9)4 + Ss + sl(i
4 + .1/-s + isl
ts + lr)6 + Ss + s (18/Ss 2

6 + ls
V-s + s•)ls• + ts(l
ls1 + H-s

sl)hs(6/1Ss

As we see, the division process giving the quotient of 1/3 involves a

reversal of the order of the polynomials involved. The resulting ladder
network is given in Fig. 11.26.

FIG, 11,26
336 Network analysis and synthesis
In the beginning of this section. it was stated that only under special
conditions can an R-L-C driving-point function be synthesized with the
use of a ladder form or the Foster forms. These conditions are not
given here because they are rather involved. Instead, when a positive
real function is given, and it is found that the function is not synthesizable
by using two kinds of elements only, it is suggested that a continued
fraction expansion or a partial fraction expansion be tried first.

Problems
11.1 (a) Which of the following functions are L-C driving point impedances?
Why?
s<..r + 4)(r + 16) z-" <r
> _ + 1xs1 + s>
Zi(s) "" (I + 9)(r + 25) ' s s<..r + 4)
(b) Synthesize the realizable impedances in a Foster and a Cauer form.
11.2 Indicate the general form of the two Foster and the two Cauer networks
that could be used to synthesize the following L-C impedance.

zo = <r + 1xs1 + 9)(r + 25)

s s<..r + 4)(r + 16)
There is no need to calculate the element values of the four networks.

11.3 Synthesize the L-C driving~point impedance

Z(s) = 6s' + 42r + 48

s5 + 18r + 48s

in the form shown in the figure, i.e., determine the element values of the network
in henrys and farads.

Ca

PROB.11.J

11.4 There exists an L-C network with the same driving-point impedance as
the network shown in the figure. This alternate network should contain .>nly
two elements. Find this network.
 Synthesis of one-port networks 337

PROB. 11.4

11.5 The input impedance for the network shown is

2s8 +2
Z1n .. s" + 2s8 + 2s + 2

If Z 0 is an L-C network: (a) Find the expression for Z 0 • (b) Synthesi7.e Z 0 in a

Foster series form.

PROB. 11.5

11.6 Indicate which of the following functions are either R-C, R-L, or L-C
Impedance functions.

;a+ 2s
(a) Z(s) = s' + 4s' + 3

(b)
.r+61+s
Z(s) = r + 4s + 3

(c)
r+4s+3
Z(s) = r + 6s + g

(d)
r+Ss+6
Z(s) =
,. + of

(e)
s'+Sr+6
Z(s) - ,a+ 8
338 Network analysis and synthesis
11.7 An impedance function has the po1~7.Cl'o pattern shown in the figure.
HZ( -2) = 3, synthesi7.e the impedance in a Foster form and a Cauer form.
jw

_x,--o---x--
-<T -5 -3 -1 10
PROB. 11.7

11.8 From the following functions, pick out the ones which are R-C admit-
tances and synthesi7.C in one Foster and one Cauer form.
Y(s) = 2(s + l)(s + 3) Y(s) = 4(s + + 3)
l)(s
(s + 2)(s + 4) s(s + 2)
Y(s) = s(s + 4)(s + 8) Y() = (s + l)(s + 4)
(s + l)(s + 6) s s(s + 2)
11.9 Find the networks for the following functions. Both Foster and ladder
forms are required.
( ) Z(s)
= (s + l)(s + 4)
a' s(s + 2)

(b) Z(s) .... 3(s + l)(s + 4)

s+3
11.10 For the network shown, find Y when
Vi, 1
V0 =2+Y
s<r + 3)
=2r+r+6s+l
Synthesi7.C Y as an L-C admittance.

PROB. 11.10

11.11 Synthesi7.C by continued fractions the function

Y(s) = s3 + 2s3 + 3s + 1
s3+r+2s+l
 Synthesis of one-port networks 339
11.12 Find the networks for the following functions in one Foster and one
Caucrform.
Y(s) .,. (s + lXs + 3)
(s + 2Xs + 4)
Z( ) = 2(s + O.S)(s + 4)
s s(s + 2)
11.13 Synthesize the following functions in Caucr form.
Z(s) == r + 2r + s + 1
r +r +s
Z() .r+.r+2s+l
s =s'+r+3r+s+l
Z(s) = 4s8 + ~ + 4s + 2
+s
11.14 Synthesi7.e Z(s) = ~; ! :~; ! :~ into the form shown in the figure.

PROB. 11.14
11.15 Of the three pole-7.el'O diagrams shown, pick the diagram that repre-
sents an R-L impedance function and synthesi7.e in a series Foster form.

--4 -3 -2..-1T"r~
___ '"_r
-4 -1T'
-3 -2

jw

--x--o-x---o-~-
-4 -3 -2 -1 (T 10
PROB. II.IS
 3<f0 Network analysis and synthesis
. rattern
11.16 Synthesize a driving-point impedance with the pole-:zero shown
in the figure in any form you choose. (Hint: Use uniform loading concepts.)

I i'-"
❖I
f
-,f!f 1
,/2
1 I
I
❖I
X 1
4
PROB. 11.16

11.17 Following are four successive approximations of tanh s.

(a) ~ (b) r + 15s

s2 +3 6s8+ 15
10s2 + 105s (d) s5 + 105s8 + 945s
(c) s' + 45s2 + 105 15s' + 420r + 945
Synthesi7.e networlcs for the functions above whose input impedances approxi-
mate tanh s.
 chapter 12
Elements of transfer
function synthesis

12.1 PROPERTIES OF TRANSFER FUNCTIONS

A transfer function is a function which relates the current or voltage

at one port to the current or voltage at another port. In Chapter 9 we
discussed various descriptions of two-port networks in terms of the open-
circuit parameters z11 and the short-circuit parameters Yu• Recall that for
the two-port network given in Fig. 12.1, the open-circuit transfer im-
pedances Zi1 and zt1 were defined as

Y11
Zi.--
1. 11-0

Zn-- v.,
Ii 1.-0
(12.1)

In terms of the open-circuit transfer impedances, the voltage-ratio transfer

function is given· as
v. =Zai
- - (12.2)
Y1 Zit
In terms of the short-circuit parameters, the voltage ratio is shown to be
Y1 1/n
-==-- (12.3)

When the network is terminated at port two by a resistor R, as shown in

Fig. 12.2, the transfer impedance of th«? overall network is

Zn - Y1 - ZnR (12.4)
11 z. + R
341
342 Network analysis and synthesis

:1
+ I1
ill

l:=I
I2
E
+
:2 ; l=I I1
ill
I2
E +
R~V2

FIG.1:2.1 FIG.12.2

The transfer admittance of the overall structure in Fig. 12.2 is

_ Ia_ Y11G
Y.u - - (12.5)
V1 YH +G
where G = I/R. When both ports are terminated in resistors, as shown in
Fig. 12.3, the voltage-ratio transfer function V1 /V11 is
v.
-=
Z11R1
(12.6)
(z11 + R1)(z11 + RJ - ZuZis
V,,
Other transfer functions such as current-ratio transfer functions can
also be described in terms of the open- and short-circuit parameters. In
Chapter 10 we discussed the various properties of driving-point imped-
ances such as z11 and z118 • This chapter deals with the properties of the
transfer im.mittances z11 and y11 for a
passive reciprocal network. First, let
+ us discuss certain . properties which
Two-port apply to all transfer functions of passive
network V2 linear networks with lumped elements.
V, We denote a transfer function as T(s).
1. T(s) is real forreal s. This property
FIG.12.3 is satisfied when T(s) is a rational
function with real coefficients.
2. T(s) has no poles in the right-half plane and no multiple poles on
the jru axis. If T(s) is given as T(s) = P(s)/ Q(s), the degree of P(s) cannot
exceed the degree of Q(s) by more than unity. In addition, Q(s) must be
a Hurwitz polynomial.
3. Suppose P(s) and Q(s) are given in terms of even and odd parts,
that is,
T(s) = P(s) = Mi(s) + N1 (s) (12.7)
Q(s) M.(s) + N.(s)
where M.(s) is even and N.(s) is odd. Then TUw) is
TUw) = M 1Uw) + N 1~ru) (12.8)
M 1 Uw) + N 1(Jru)
The amplitude response of T(jru)js
ITUw)I = rM18Uru) + NlUw>]~ (12.9)
Ul18Uru) + N19Uw)
 Elements of transfer function synthesis 3-43
and is an even function in w. The phase response is

Arg T(jw) = arctan rEiCw)] - arctan rLMEa(w)] (12.10)

LM (w)
1 1 (w)
If arg T(jO) = 0, we see that the phase response is an odd function in w.
Now let us discuss some specific properties of the open-circuit and
short-circuit parameters.
1. The poles of z11(s) are also the poles of z11(s) and z1a(s). However,
not all the poles of z11(s) and zH(s) are the poles of z11(s). Recall that in
Chapter 9 we defined the z parameters in terms of a set of node equations
as
dn
zu(s) = -
d

If there is no cancellation between each numerator and denominator of z11,

z11., and zu, then the poles are the roots of the determinant d, and all three
functions have the same poles. Consider the two-port network described
by the black box in Fig. 12.4a. Let z' 11, z' 111, and z'11 be the z parameters
of the network. Let us examine the case when we attach the impedances
Zi and Z 1 to ports one and two, as shown in Fig. 12.4b. The z parameters
for the two-port network in Fig. 12.4b are

zu = z'u + Zi
Z11 = z'11 + Za

It is clear that the poles of z11 include the poles of Zi; the poles of z.1
include the poles of Za. However, the poles of z11 include neither the
poles of Zi nor Za- Consequently, we see that all the poles of z11 are also
poles of z11 and z.1 • The reverse is not necessarily true.

I1 ~
- .
+ z'11., z'22 +
V1 V:2
- z'12
-
(a) (b)
FIG. 12.4
344 Network analysis and synthesis

11
+ +
.Y'11,.Y'22
Vi y'12
V:a

FIG. 12.5 FIG. 12.6

2. The poles of y11(s) are also the poles of y11(s) and y 11(s). However,
not all of the poles of y11(s) and y11(s) are the poles of y11(s). This property
is readily seen when we examine the two-port network in Fig. 12.5.
The y parameters are
Yu= Y 11 + Y1
1

Y11 = Y 11 +1
Ya
1
Yu== Y 11
Clearly, the poles of y11(s) do not include the poles of either Y1 and Y1 •
Consider the network in Fig. 12.6. The y parameters are

Yu(s) = -2 + 3s
s

y.,.(s) = -4 + 3s
s
Y11(s) =
-3s
Observe that y11(s) and y11(s) have poles at s = 0 and s = oo, whereas
y11(s) only has a pole at s = oo.
3. Suppose y11(s), y11(s), and y11(s) all have poles at s = s1 • Let us
denote by k11 the residue of the pole at s1 of the function y11(s). ,:he
residue of the pole s = s 1 of Yn(s) will be denoted as k11, and the residue
of the same pole of y11(s) will be denoted as k11 • Without going into the
proof,1 a general property of L-C, R-C, or R-L two-port networks is that
(12.11)
This equation is known as the residue condition. For example, for the
L-C network in Fig. 12.6, the residue condition applied to the pole at
s = oo gives 3 x 3 - 31 = 0; whereas for the pole at s = 0, we have
2 x 4 - oa = 8 > 0. Thus we see that the residue condition is fulfilled
for both poles.
1 For a general discussion. see M. E. Van Valkenburg. Introduetlon to Modern

Network Synthesis, John Wiley and Sons, New York, 1960, pp. 305-313.
 Elements of transfer function synthesis 345
C
+ +
V1 V2

FIG. 12.7

I1 12
+ +
V1 C V2

FIG. 12.8

12.2 ZEROS OF TRANSMISSION

A zero of transmission is a zero of a transfer function. At a zero of

transmission, there is zero output for an input of the same frequency.
For the network in Fig. 12.7, the capacitor is an open circuit at s = 0,
so there is a zero of transmission at s = 0. For the networks in Figs.
12.8 and 12.9, the zero of transmission occurs at s = ±j/JLC. For
the network in Fig. 12.10, the zero of transmission occurs at s = -1/RC.
In general, all the transfer functions of a given network have the 8J!U]le
zeros of transmission, except in certain special cases. For example if
iu(s) has a zero of transmission at s = s1, than y11(s), V .(,s)/V1(s), etc.,

+ +

FIG. 12.9

+
V1

PIG. 12.lt
346 Network analysis and synthesis

FIG. 12.11

will also have a zero at s = s1 • This fact is clearly seen when we examine
the relationships between the transfer functions. For example, we have

Z11= - (12.12)
Y11Y11 - Y11Y11
and Y11 =- Zi1Zu - Z1a%11
(12.13)

In addition, the voltage- and current-ratio transfer functions can be

expressed in terms of the z and y parameters as
Vs Z11 ls =Yu
- -
-=- (12.14)
Yi Zn 11 Yn
In Chapter 8 we saw that transfer functions that have zeros of trans-
mission only on the jw axis or in the left-half plane are called minimum
phase functions. If the function has one or more zeros in the right-half
plane, then the function is nonmimimum phase. It will be shown now that
any transfer function of a passive reciprocal ladder network must be
minimum phase. Consider the ladder network in Fig. 12.11. The zeros of
transmission of the ladder occur at the poles of the series branch im-
pedances or at the zeros of the shunt branch impedances. Since these
branch impedances are themselves positive real, the poles and zeros of
these impedances cannot be in the right-half plane. Consequently, the
transfer functions of ladder networks must be minimum phase. For the
network in Fig. 12.12, a transfer function would have two zeros of

FIG. 12.12
 Elements of transfer function synthesis 347

L-C
ladder

FIG. 11.13

transmission at s = co due to the elements Li and C3 • It would also have a

zero of transmission at s = 0 due to Ci, a zero at s = -1/R,.C1 due to
the parallel R-C branch, and a zero of transmission at s = j/(L,.CJ'A
due to the L-C tank circuit. It is seen that none of the transmission
zeros are in the right-half plane. We also see that a transfer function may
possess multiple zeros on the jw axis.
In Section 12.4 it may be seen that lattice and bridge circuits can easily
be nonminimum phase. It can also be demonstrated that when two ladder
networks are connected in parallel, the resulting structure may have
right-half-plane zeros.•

12.3 SYNTHESIS OF Y11 AND Z 11 WITH A 1-0 TERMINATION

In this section we consider the synthesis of an L-C ladder network

with a 1-0 resistive termination to meet a specified transfer impedance Z.1
or transfer admittance Y11• In terms of the open- and short-circuit
parameters of the L-C circuit, Z.1(s) can be expressed as

-~
Z u- (12.15)
Zn + 1

and Y11(s) is Y11 = Yit (12.16)

1/n +1
as depicted in Figs. 12.13 and 12.14.
Before we proceed with the actual details of the synthesis, it is necessary
to discuss two important points. The first deals with the ratio of the odd
to even or even to odd parts of a Hurwitz polynomial Q(s). Suppose

12

Qv1 ~~ I E

I ~10

FIG. 11.14

• Van Valkenburg, op. cit., Chapter 11.

3-48 Network analysis and synthesis
Q(s) is given as
Q(s) = M(s) + N(s) (12.17)

where M(s) is the even part of Q(s), and N(s) is the odd part. We know
that the continued fraction expansion of M(s)/N(s) or N(s)/M(s) should
yield all positive quotients. These quotients can, in turn, be associated
with reactances. Therefore it is clear that the ratio of the even to odd or
the odd to even parts of a Hurwitz polynomial is an L-C driving-point
function. ·
The second point to be discussed is the fact that the open-circuit transfer
impedance z11 or the short-circuit transfer admittance y11 of an L-C circuit
is an odd function. To show this, we must remember that in an L-C
circuit with steady-state input, the currents are 90° out of phase with the
voltages. Thus the phase shifts between the input currents and output
voltages or input voltages and output currents must be 90° out of phase,
or
v..(jw)
Arctan - - = ± -7T rad (12.18)
l 1(jw) 2

and Arctan li(jw) = ± !! rad (12.19)

Yivw) 2
so that Re z11(jw) = 0 = Re y11(jw) for an L-C network. In order for the
real parts to be equal to zero, the functions z11 and Yu of an L-C two-port
network must be odd.
Suppose, now, that the transfer admittance Y11 is given as the quotient
of two polynomials
y = P(s) = P(s) (12.20)
M(s) + N(s)
11
Q(s)
where P(s) is either even or odd. Now, how do we determine the short-
circuit parameters y11 and y11 from the Eq. 12.20 to get it into the form

_ Yu (12.21)
Y.11 -
1 + Y11
The answer is quite simple. We divide both the numerator P(s) and the
denominator Q(s) by M(s) or N(s), the even or the odd part of Q(s).
Since y11 must be odd, if P(s) is even, we divide by·N(s) so that

Y. _ P(s)/ N(s) (12.22)

11
- 1 + [M(s)/ N(s)]
 Elements of transfer function synthesis 349
P(s)
From this we obtain Yu= N(s)
(12.23)
M(s)
Yn = N(s)

On the other hand, if P(_s) is odd, we divide by M(s) so that

Y. P(s)/M(s)
11
= 1 +[N(s)/M(s)] (12.24)

P(s)
and Yu= M(s)
(12.25)
N(s)
Yu= M(s)

We assume that P(_s), M(s), and N(s) do not possess common roots. For
our purposes, we will consider only the synthesis of Y11 or Z.1 with zeros
of transmission either at s = 0 or s = oo. In a ladder network, a zero of
transmission at s = 0 corresponds to a single capacitor in a series branch
or a single inductor in a shunt branch. On the other hand, a zero of
transmission at s = oo corresponds to an inductor in a series branch or
a capacitor in a shunt branch. In terms of the ttansfer impedance

Zu(s) _ P(s) = K(s" + a_1s-1 + · · · + a1s + a 0) (l 2.26)

Q(s) s"' + b,_1 s,_1 + · · · + b1s + b0
the presence of n zeros of Z.1(s) at s = 0 implies that the coefficients a_1 ,
a_1, ••• , ai, a. are all zero. The number of zeros of Z.1(s) at s = oo
is given by the difference between the highest powers of the denominator
and the numerator, m - n. We know that n can exceed m by at most
unity, while m can be greater than n by more than one. For example,
if m - n = 2, and n = 3 with a"_.. ... , a 1 , a. = 0, we know that the
transfer function has three zeros of transmission at s = 0 and two zeros
of transmission at s = oo.
We can now proceed with the matter of synthesis. Consider the
following example.
2
Z11(s)- 1
(12.27)
-s8+3s +4s+2
We see that all three zeros of transmission are at s = oo. Since the
numerator P(_s) is a constant, it must be even, so we divide by the odd
350 Network analysis and synthesis
part of the denominator s3 + 4s. We then obtain
2
Zz1 =---
S8 + 4s
(12.28)
3s8 + 2
z111= sa+4s

We see that both z.1 and z.1 have the same poles. Our task is thus simpli-
fied to the point where we must synthesize z811 so that the resulting network
has the transmission zeros of z81 • This requires that we first examine the
possible structures of the networks which have the required zeros of
transmission and see if we can synthesize z22 in one of those forms. For
the example that we are considering, a network which gives us three zeros
of transmission at s = oo is shown in Fig. 12.15. We can synthesize z12
to give us this structure by the following continued fraction expansion
of l/z21•
3s2 + 2}s3 + 4s(ls - Y
s3 + is
lfs)3s + 2(/os- Z
8

3s1
2}1/s(ts- Y
lfs

Since z81 is synthesized from the 1-0 termination toward the input end,
the final network takes the form shown in Fig. 12.16. Examining the
network more closely, we see that it takes the form of a low-pass filter.
Thus the specification of all zeros at s = oo is equivalent to the specification
of a low-pass filter.
As a second example, consider the transfer impedance
s•
Z (s) - -
8
--8
--- (12.29)
11 -s +3s +4s+2
Since the numerator of Z 8i(s) is an odd function, we have to divide both

11 ff
~h

½t rn
T
V2

0
T T
FIG. 12.15
0

FIG. 12.16
1
 Elements of transfer function synthesis 351

10
T
V2

FIG. 12.17
1
numerator and denominator by the even part of the denominator so that
s
1
+ 4s (12.30)
z. = 3s' + 2
=
The network that gives three zeros of transmission at s 0 is a high-pass
structure which is realized by a continued fraction expansion of z..
The final realization is shown in Fig. 12.17.
Finally, consider the transfer admittance
s•
Yu(s) = s• + 3s• + 4s + 2 (12.31)
=
which has two zeros of transmission at s 0, and one :zero at s = oo.
Since the numerator is even, we divide by s' + 4s so that

(12.32)

The question remains as to how we synthesi:r.e fin to give a zero of trans-

mission at s= oo and two :zeros at s = 0. First, remember that a parallel
=
inductor gives us a zero of transmission at s 0. We can remove this
parallel inductor by removing the pole at s = 0 of fin to give
1 Ss/2
1/1 = 1/11 - - = -
1 -
(12.33)
2s s +4
If we invert 1/i, we see that we have a series L-C combination, which gives
us another transmission :zero at s = 0, as represented by the I-farad
capacitor, and we have the zero of trans~ssion at s = oo also when we
remove the inductor off h. The final realization is shown in Fig. 12.18.

FIG. 12.11
352 Network analysis and synthesis

(a,

Network 10

(b)
FIG. 12.19

An important point to note in this synthesis pr~dure is that we must

place the last element in series with a voltage source or in parallel with a
current source in order for the element to have any effect upon the transfer
function. If the last element is denoted as Zn, then the proper connection
of Zn should be as shown in Fig. 12.19.

12.4 SYNTHESIS OF CONSTANT-RESISTANCE NETWORKS

In this section we will consider the synthesis of constant-resistance two-

port networks. They derive their name from the fact that the impedance
looking in at either port is a constant-
resistance R when the other port is termi-
Two-port nated in the same resistance R, as depicted
network R in Fig. 12.20. Constant-resistance net-
works are particularly useful in transfer
function synthesis, because when two
FIG. 11.10. Constant-resistance
network. constant-resistance networks with the
same R are connected in tandem, as shown
in Fig. 12.21, neither network loads down the other. As a result, if the
voltage-ratio transfer function of N,,. is V1 / V1 and that of Nb is Val V8 , the
voltage-ratio transfer function of the total network is
V V1 Va
-8 = - - (12.34)
V1 V1 Va
Equation 12.34 implies that, if a voltage-ratio transfer function is to be
realized in terms of constant-resistance networks, the voltage ratio could
 Elemenu of transfer function synthesis 353

FIG. 12.21. Constant-resistance networks in tandem.

be decomposed into a product of simpler voltage ratios, which could be

realized as constant-resistance networks, and then connected in tandem.
For example, suppose our objective is to realize

Y11 == K(s - Zo)(s - Zt)(s - z.) (l 2.35)

V.. (s - P0)(s - Pi)(s - P1)
in terms of constant-resistance networks. We can first synthesize the
individual voltage ratios
Yi Ko(s-z0)
-==
Y,. s - Po
Y, Ki(s- Zi)
-== (12.36)
Yi s - Pi
v,,
-==
K 8(s - z.)
v. s - Pa
as constant-resistance networks and then connect the three networks in
tandem to realize V11/ V,..
Although there are many different types of constant-resistance net-
works, we will restrict ourselves to networks of the bridge- and lattice-type
structures as shown in Figs. 12.22a and 12.22b. These networks are
balanced structures; i.e., the input and output ports do not possess com-
mon terminals. Upon a close examination, we see that the bridge and
lattice circuits in Figs. 12.22 are identical circuits. The bridge circuit is
merely the unfolded version of the lattice. Consider the open-circuit
parameters of the bridge circuit in Fig. 12.23. First we determine the
impedance z11 as
(12.37)

Next we determine the transfer impedance z11, which can be expressed as

Y, - Y,-
Z91 == (12.38)
Ii
354 Network analysis and synthesis
11
+

(a}

(bJ
FIG. 12.22. (a) Bridge circuit. (b) Constant-resistance lattice.

and is obtained as follows. We first obtain the current I' as

I'= Yi = !J, (12.39)

z,.+z,, 2

Next we find that V1 - V1• = (Z,, - Z,.)I'

= (Z,, - Z,.) !J (12.40)

2
11
+

FIG. 12.23. Analysis of bridge circuit.

 Elemenu of transfer function synthesis 355

so that Zn=
z.-z II (12.41)
2
From the lattice equivalent of the bridge circuit we note that z11 = z11•
Now let us consider the lattice circuit that is terminated in a resistance
R, as shown in Fig. 12.22b. What are the conditions on the open-circuit
parameters such that the lattice is a constant-resistance network? In
other words, what are the conditions upon z11 and z11 such that the input
impedance of the lattice terminated in the resistor R is also equal to R 1
In Chapter 9 we found that the input impedance could be expressed as
Znl
Z11 = Zi1 - --=-- (12.42)
z.. + R
Since zn = z11 for a symmetrical network, we have
1
Z 11 -_ Zi1R + Zii' - Zsi.
(12.43)
z11 +R
In order for Zi1 - R, the following condition must hold.

(12.44)
For the lattice network, we then have
l[(Z,. + Z,,)1 - (Z,. - Z,,)I] - R1 (12.45)

which simplifies to give z..z. = RI (12.46)

Therefore, in order for a lattice to be a constant-resistance network,

Eq. 12.46 must hold.
Next, let us examine the voltage ratio VJ v. of a constant-resistance
lattice whose source and load impedonces are equal to R (Fig. 12.24).
From Chapter 9 we can write

--
V1
v.
Zm.R
(Zi1 + R)(zn + R) - ZnZi1
(12.47)

FIG. 12.24. Double-terminated lattic:ic.

356 Network analysis and synthesis

which simplifies to Vii 11R

- = - - -z= ---
v,, + R) - z1i8
1
(zu
=--.__-
Z11R
(12.48)
2Rzu + 2R
8

In terms of the element values of the lattice, we have

V1 ½(Zb - Z 0 )R
-= (12.49)
R(Zb + Z 0 ) + 2R
1
V,,
From the constant-resistance condition in Eq. 12.46, we obtain

v.
-=
½[Zb - (R 1 /Zb)]R
v,, R[Zb + (R1/Zb)] + 2R8
½(Zb1 - R2)
- (Z/• + R2) + 2RZb
½(Zbll - RI)
=
(Zb + R)1
½(Zb - R)
- zb+R
(12.50)

In Eq. 12.50, the constant multiplier½ comes about from the fact that
the source resistance R acts as a voltage divider. If we let

G(s) ~ 2Va (12.51)

v,,
we can express Zb in Eq. 12.50 in terms of Gas

zb = R[l + G(s)J (12.52)

1 - G(s)
In terms of Z 0 , the voltage ratio can be given as
V1 1R - Z
-=---~ 0
(12.53)
V,, 2 R + Z 0

In the following examples, we will usually let R be normalized to unity.

Example 12.1. The voltage ratio is given as
Vi1 ls-1
(12.54)
V11 =2.s + 1
 Elements of transfer function synthesis 357
which, as we recall, is an all-pass transfer function. By associating Eq. 12.54
with Eq. 12.50, we have
R =1 (12.55)
1
Since Z,,Za = 1, we then obtain Za =- (12.56)
s

We see that Zb is a 1-h inductor ~d Zais a I-farad capacitor. The final network
is shown in Fig. 12.25.
lf

1
~
V2

~
FIG. 12.25

Example 12.2. Let us synthesize the all-pass function

V1 ls-1 s2-2s+2
(12.57)
V11 = 2 s + l . s8 + 2s + 2
whose pole-zero diagram is shown in Fig. 12.26. Since the portion
V1 ls-1
V11 = 2s + 1
has already been synthesized, let us concentrate on synthesizing the function
V1 r-2s+2
(12.58)
Va =s1+2s+2
First, we separate the numerator and denominator function into odd and
even parts. Thus we have
v1 <r + 2> - 2s (12.59)
Va = (r + 2) + 2s
jw

X jl 0

--x;----+----o--
-u -1 +l <T

X -jl 0

FIG. 12.16
358 Network analysis and synthesis
H we divide both numerator and denominator by the odd part 2s, we obtain
v. [(al + 2)/2s] - 1
(12.60)
V0 .. [(al+ 2)/2s] + 1
al +2
We then see that z,, - ~
s 1
=2 +; (12.61)

which consists of a i-h inductor in series with a I-farad capacitor. The im-
pedance Z 0 is then
2s
Za=...51-- (12.62)
.r + 2
and is recogniz.ed as a !-farad capacitor in parallel with a 1-h inductor. The
voltage ratio VJV0 is thus reali7.ed as shown in Fig. 12.27. The structure that
lh

10
l
V2

FIG, 12.27
l
realizes the transfer function VJV, in Eq. 12.57 is formed by connecting the
networks in Figs. 12.25 and 12.27 in tandem, as shown in Fig. 12.28. Finally,
it should be pointed out that constant-resistance lattices can be used to realize
other than all-pass networks.

FIG, 12.21

Next, let us consider the constant-resistance bridged-T network in Fig.

12.29. If the resistances in the network are all equal to R ohms, the
network has constant-resistance if
(12.63)
 Elements of transfer function synthesis 359

I 7
l Bo--i,__________R_,j
FIG. 12.29. Constant-resistance bridged-T cin:uit.

Under the constant-resistance assumption, the voltage-ratio transfer

function can be given as
V R z.
- 1 = - - = ------ (12.64)
Y1 R+z.. z.+R
Eu_,... 12.3. Let us synthesi7.e the voltage ratio
v. r +1 (12.65)
V1 ..,r+2s+l
as. a constant-resistance bridged-T network terminated in a 1-ll Raistor. First
let us write VJY1 as
v.
-=
1
(12.66)
Y1 1 + c2s1c.s1 + •>1
2s
so that Z,, - r +1 (12.67)

.,. + 1
and z.,,---2s
(12.68)

We recogniz.e Z,, as a parallel L-C tank circuit and z.,, as a aeries L-C tank
circuit. The final network is shown in Fig. 12.30.
2h

FIG. 12.31
360 Network analysis and synthesis
Example 12.4. Let us synthesize the voltage ratio

V1 (s + 2Xs + 4) (12.69)
Yi = (s + 3X3s + 4)
in terms of two constant-resistance bridged-T circuits connected in tandem.
At first, we break up the voltage ratio in Eq. 12.69 into two separate voltage
ratios
v,. s+2
-V1
=s+3-- (12.70)

-
v. s +4
and
v,. =3s-+4
- (12.71)

For the voltage ratio V,./Vi, we have

s +2 Z111
(12.72)
s + 3 = Z 111 + 1

so that z.1 =- s + 2 and (12.73)

For the voltage ratio VJ V,. we have

s+4 1
(12.74)
3s + 4 = 1 +z,..
2s
from which we find z ... ---
- s +4
(12.75)

to 20

FIG. 12.31
 Elements of transfer function synthesis 361
s+4
and Zbl =--t- (12.76)

The final synthesm,ci network is shown in Fig. 12.31.

Problems
12.1 Give an example of a network where: (a) a transfer function has
multiple :zeros on the jw axis; (b) the residue of a pole of a transfer function on
the jw axis is negative.
12.2 Show that the residue condition holds for the networks shown in the
figure.

lh
10 3h
1 2

1'
10

(a)
½t
2'

PROB. 12.2
:.It··::. (b)

12.3 For the network shown, find by inspection the 7.el'OS of transmission
and plot on a complex plane.

20

lh

PROB. 12.3

12.4 For the networks in the figure show that the driving point i1w\pedaoces
Ztn are equal to R when z.z,, -
Iii.
362 Network analysis and synthesis
B

(a)

V Z1n•B
1 ~
B
1· V2

(b)
J
PROB. 12.4

12.5 For the networks in Prob. 12.4, find the voltage-ratio transfer functions
YJY1 •
12.6 For the network shown in the figure (a) show that
v. 1
V0 -2 +Y
(b) Syntbesi7.e Y when
V1 O.S(sl + 2)
Vo -s8+2s+2

7
L------L---1-0_,J
PROB. 12.6

12.7 (a) For the constant-resistance bridged-T circuit, show that if z.z. - 1,
then
 Elements of transfer function synthesis 363
(b) Syntheme z. and z. if
v. r+3s+2
Y1 -r+4r+Ss+2
12.8 Synthesi7.e the following voltage ratios in one of the forms of the
netWorks in Prob. 12.4
(a)
v. s +2
Y1 -s+3
v1 2(r + 3)
Y1 -2sl+2s+6

(c)
Y1 3(s + 0.5)
Y1 - 4s + 1.5
12.9 Synthesize N. with termination resistors Rs = 4 o, R1 - 1 O to give
v. 1281
Y., - lSr + 1s + 2

PROB. 12.t

12.10 For the network in Prob. 12.9, reali7.e network N. to give

Y1 l 1
v.,-22s+3
(a) Syntheme N. as a constant-resistance lattice. (R =- l O.)
(b) Synthesize N. as a constant-resistance ladder as in Prob. 12.4. (R - 1 0.)
(c) Synthesi7.e N. as a constant-resistance bridged-T circuit. (R =- 1 O.)
12.11 Synthesize the following functions into the form shown in the figure
1
(a) Zn - s8 + 3r + 3s + 2
s
(b)
Zn•s8+3r+3s+2

(c) Y. r
n'""s8+3s8+3s+2

(d) Y. r
n - s8 +3s8 +3s +2
(e) Y.
n == (s
r
+ 2)'
364 Network analysis and synthesis

10-----1
L-C
network 10
1'0----1

PROB. 12.11

12.12 Synthesi7.e as a constant-resistance lattice terminated in a 1-0 resistor.

V1 r-s+l
(a)
V1 =r+s+l
(b)
11 r - 3s + 2
Yi=r+3s+2
(c)
V1(s) s8 - 20r + Ss - 20
V1(s) = s8 + 20r + Ss + 20
12.13 Synthesiz.e the functions in Prob. 12.8 as constant-resistance bridged-T
circuits.
 chapter 13
Topics in filter design

13.1 THE FILTER DESIGN PROBLEM

In the preceding chapters we examined different methods for synthesizing

a driving point or transfer function H(s). Most problems have as their
initial specification an amplitude or phase characteristic, or an impulse
response characteristic instead of the system function H(s). Our problem
is to obtain a realizable system function from the given amplitude or
phase characteristic. For example, a typical design problem might be to
synthesize a network to meet a given low-pass filter characteristic. The
specifications might consist of the cutoff frequency w 0 , the maximum
allowed deviation from a prescribed amplitude within the pass band, and
the rate of/all off in the stop band. We must then construct the system
function from the amplitude specification. After we obtain H(s), we
proceed with the actual synthesis as described in the Chapter 12. Another
problem might consist of designing a low-pass filter with a linear phase
characteristic within the pass band. Here, both amplitude and phase are
specified. We must constructH(s) to meet both specifications. Problems
of this nature fall within the domain of approximation theory. In this
chapter we will consider selected topics in approximation theory and then
present examples of filter design where both the approximation and the
synthesis problems must be solved.

13.2 THE APPROXIMATION PROBLEM IN

NETWORK THEORY

The essence of the problem is the approximation of a given func-

tion/(z) by another function/..(z; exi, ... , ex,.) in an interval z 1 ~ z ~ z1 •
The parameters exi, ... , ex,. in the approximating function are fixed
365
366 Network analysis and synthesis
by the particular error criterion chosen. When we let E == /(z) -
/J.z; «i, ... , ot,.), the following error criteria are most common:
1. Least squares. The value of I(ot1 , ••• , otJ is minimized where

l(ot1, ••• , otJ ==f .. IEl 1 w(z) dz

Ill

and w(z) is a weighting function which stresses the error in certain sub-
intervals.
2. Maximally flat. The first n - 1 derivatives of E are made to vanish
at z == z..
3. Chebyshev. The value ofµ is minimized in the interval :r:1 :S;; z ~ :r:1
whereµ== IElmu:•
4. Interpolation. The value of E is made to vanish at a set of n points
in the interval :r:1 ~ z :S;; z 1 •
After an error criterion is chosen, we must determine the particular
form of the approximating function. This depends upon whether we
choose to approximate in the time or frequency domain. Suppose /(z)
represents a magnitude function in the frequency domain and the approxi-
mating function is to be rational in ro1 ; then

ot1 + ac,z + «1:r:' + · · · (13.1)

/,.(z; oti, • • • , ot,.) - --•
ota + «.Z + IXeW + • ••
where z == ro1 • In addition, the values of «1: must be restricted to insure
thatf.(z; ot1 , ••• , otJ ~ 0. In the time domain,/(z) might represent an
impulse response of a system to be synthesized. In the case of an R-C
transfer function, we have

J.<.:r:; oti, ••• , ot,.) == otii'S"' + «. ._.,.. + · · · (13.2)

where z == t. Since an R-C transfer function must have its poles on the
negative real axis, the values of «1:, k even, are restricted to negative real
numbers.
The keystone of any approximation problem lies in the choice of a
suitable error criterion subject to reali7.ability restrictions. The problem
can be simplified when some of the ot's are assigned before applying the
error criteria. All the error criteria cited, except the Chebyshev, can then
be reduced to a set of linear algebraic equations for the unknowns
oti, •.. ' ot,..
Time-domain approximation
The principal problem of time-domain approximation consists of
approximating an impulse response h(t) by an approximating function
 Topics in filter design 367
h*(t) such that the squared error

E = L" [h(t) - h*(t)]1 dt

is minimum.
A generally effective procedure in time-domain approximation utilizes
orthonormal functions ,f,t(t).1 The approximating function h*(t) takes
the form

h*(t)
" ,f,it)
= !ext (13.3)
1t-1
so that the error

is minimized when

exit = iCX) h(t) ,f,1t(t) dt k = 1, 2, ... , n (13.4)

as we saw in Chapter 3. If the orthonormal set is made up of a sum

of exponentials eai', then the approximate impulse response

n
h*(t) = !ex1te"i' (13.4)
1t=l

" ex
has a transform H*(s)=!-1t- (13.5)
1t=1 S - S1:

Realizability is insured if in the orthonormal set {exteai'}, Re sk s; O;

k = 1, 2, ... , n. Synthesis then proceeds from the system function H*(s)
obtained in Eq. 13.5.
jH(jw)I
Frequency-domain approximation 1
In frequency-domain approximation
the principal problem is to find a
rational function H(s) whose magnitude
IH(jw)I approximates the ideal low-pass -1 0 +1 w
characteristic in Fig. 13.1 according to
a predetermined error criterion. In the FIG. 13. I. Ideal low-pass filter char-
next few sections we examine several acteristic.

1
W. H. Kautz, "Transient Synthesis in the Time Domain," IRE Trans. on Circuit
TMOry, CT-1, No. 3, September 1954, 29-39.
368 Network analysis and synthesis
different ways to approximate the ideal low-pass: the maximally flat or
Butterworth approximation, the equal-ripple or Chebyshev approximation,
and the optimal or Legendre approximation. Another major problem is
that of obtaining a transfer function H 1(s), whose phase is approximately
linear or whose delay is approximately flat over a given range of frequencies.
Here again, there are two different methods: the maximally flat or the
equal-ripple methods. Our discussion will center around the maximally
flat method. The joint problem of approximating both magnitude and
phase over a given frequency range is possible, but will not be discussed
here.

13.3 THE MAXIMALLY FLAT LOW-PASS FILTER

APPROXIMATION

In Chapter 10 we saw that the ideal low-pass filter in Fig. 13.1 is not
realizable because its associated impulse response is not zero for t < 0.
However, if we use a rational function approximation to this low-pass
filter characteristic, the Paley-Wiener criterion will be automatically
satisfied. We will therefore restrict ourselves to rational function approxi-
mations.
In low-p11ss filter design, if we assume that all the zeros of the system
function are at infinity, the magnitude function takes the general form

M( ) Ko (13.6)
w = [1 + J(w1)]½
where Ko is the d-c gain constant and/(w 1) is the polynomial to be selected
to give the desired amplitude response. For example, if
/(w1) = w1" (13.7)
then the amplitude function can be written as

M( ) Ko (13.8)
w = (1 + w8")½
We see that M(0) = Ko, and that M(w) is monotonically decreasing with
w. In addition, the 0.707 or 3-decibel point is at w = 1 for all n, that is,

M(l) = ;i all n (13.9)

The cutoff frequency is thus seen to be w = 1. The parameter n con-

trols the closeness of approximation in both the pass band and the stop
band. Curves of M(w) for different n are shown in Fig. 13.2. Observe
 Topics In filter design 369

Radian frequency, w
0.1 0.2 0.3 0.4 0.6 0.8 1 2 3

0
r-.;:
~~
-2
-4
-6
--n=3
".
,~\\\
-8 - - - n=5
- - - n=7 \\ \
\\
I \
\
\\
'\
\
-16
\\ \
-18
\\ \
-20
\\ \

-22
I
\
\
I \
~
-24 \
\
\ \
-26 \\ \
FIG, 13.1. Amplitude response of Butterworth low-pass filters.

that the higher n is, the better the approximation. The amplitude approxi-
mation of the type in Eq. 13.8 is called a Butterworth or maximally flat
response. The reason for the term "maximally flat" is that when we
expand M(w) in a power series about w = 0, we have
M(w) = Ko(l - ½w1 n + lw'n - f-ew•n + Naw8n + · •·) (13.10)

We see that the first 2n - 1 derivatives of M(w) are equal to zero at

w = 0. For w » I, the amplitude response of a Butterworth function
can be written as (with Ko normalized to be unity)

M(w)~ ...!_
wn
w» 1 (13.11)

We observe that asymptotically, M(w) falls off as w-n for a Butterworth

response. In terms of decibels, the asymptotic slope is obtained as
20logM(w) = -20nlogw (13.12)
370 Network analysis and synthesis
Consequently, the amplitude response falls asymptotically at a rate of
6n db/octave or 20n db/decade.
One question remains. How do we obtain a transfer function H(s)
from only the amplitude characteristics M(ro)? The procedure is as
follows. We first note that the amplitude response M(w) and the complex
system function H(jro) are related by
M2(w) = H(jw) H(-jw) (13.13)

If we define a new functioti h(s8) such that

h(s8) = H(s) H(-s) (13.14)

we see that M 1(ro) = h(-ro8) (13.15)

From h(-ro 8) all we need do is to substitute s 2 = -w2 to give h(s2).

Then we factor h(s2) into the product H(s) H( -s). Since the poles and
zeros of H(s) are the mirror images of the poles and zeros of H( -s), we
simply choose the Hurwitz factors of h(s1) as H(s). An example will serve
to clarify this discussion. Consider the third-order (n = 3) Butterworth
response given by
(13.16)

= - -1 - -
2
(13.17)
1 - (-ro )3
1
We see that h(sl) is h(s2) =1- (s2)a (13.18)
Factoring h(s8), we obtain
h(s2) = 1 1
1 + 2s + 2s1 + s3 1 - 2s + 2s8 - s3
= H(s)H(-s) (13.19)
We then have
H(s) = ___l_ __
s8 + 2s2 + 2s + 1
1
= (s + l)(s + l + jJ3/2)(s + l - jJ'J/2) (13.20)

The poles of H(s) and H( -s) a~e shown in Fig. 13.3. Observe that the
poles of H( -s) are mirror images of the poles of H(s), as given by the
theorem on Hurwitz polynomials in the Chapter 2.
 Topics In filter design 371

jw

FIG. 13.3. Poles of H(s) H(-s) for an n = 3 Butterworth filter.

For a Butterworth response, the poles of H(s) H( -s) are the roots of
(-lrs•n = -1
(13.21)
= ei<llk-U.- k = 0, 1, 2, ... , 2n - 1
The poles s,. are then given by
S1, = ei[(llk-ll/lln].- neven
(13.22)
= ei(1:fn).- nodd
or simply by s,. = ei[(lt+n-ll/lln].- k = 0, 1, 2, ... , 2n (13.23)
Expressing s,. ass,.= C11, + jw,., the real and imaginary parts are given
by
Cl1, = COS 2k + n - 1 . (2k - l)1r
- - - - - 1 T = Sin - - - -
2n n 2
(13.24)
W1, = . 2k
Sin
+n- 1= 1T COS
(2k -
--- -
1)
1r

2n n 2
It is seen from Eqs. 13.22 and 13.23 that all the poles of H(s) H(-s) are
located on the unit circle in the s plane, and are symmetrical about both
the Cl and the jw axes. To satisfy realizability conditions, we associate the
poles in the right-half plane with H( -s), and the poles in the left-half
plane with H(s).
As an example, consider the construction of an H(s) that gives an n = 4
Butterworth response. From Eq. 13.23, it is seen that the poles are
given by
S1, = ei[(llk+3)/8]1' (13.25)
3n Network analysis and synthesis
H(s) is then given as

H(s) = (s + ei!'-""')(s + ei<'½>•);s + ei<'-'"')(s + ei<1¼>•) (13.26)

If we express s,. in complex form and expand, we obtain

1
H(s) - -2 - - - - - - - = ------ (13.27)
- (s + 0.76536s + l)(s1 + l.84776s + 1)
To simplify the use of Butterworth functions, H(s) is given in Tables
13.1 and 13.2 for n = 1 to n = 8, in factored form as in Eq. 13.27, or
multiplied out as
1
H(s) = - - - - - - ------ (13.28)
a,.s" + a,._1s-1 + · · · + a 1s + 1

TABLE 13.1
Butterworth Polynomials (Factored Form)
n

1 s +1
2 s2 + v2s + 1
3 (s" + s + l)(s + 1)
4 (r + 0.76536s + l)(s2 + 1.84776s + 1)
S (s + l)(s2 + 0.6180s + l)(s" + 1.6180s + 1)
6 (s2 + 0.5176s + l)(sl + v2s + l)(sl + 1.9318s + 1)
7 (s + l)(sl + 0.4450s + l)(s" + 1.2456s + l)(sl + 1.8022s + 1)
8 (s" + 0.3986s + l)(sl + 1.1110s + l)(s" + 1.6630s + l)(s" + 1.9622s + 1)

TABLE 13.2
Butterworth Polynomials•
n al a2 aa a, 05 as 07 as

1 1
2 v2 1
3 2 2 1
4 2.613 3.414 2.613 1
s 3.236 S.236 S.236 3.236 1
6 3.864 7.464 9.141 7.464 3.864 1
7 4.494 10.103 14.606 14.606 10.103 4.494 1
8 S.126 13.138 21.848 2S.691 21.848 13.138 S.126 1

•a.=l.
 Topics In filter design 373

13.4 OTHER LOW-PASS FILTER APPROXIMATIONS

In Section 13.3, we examined the maximally fl.at approximation to a

low-pass filter characteristic. We will consider other low-pass filter
approximants in this section.
The Chebyshev or equal-ripple approximation
We have seen that the maximally fl.at approximation to the ideal low-
pass filter is best at w == 0, whereas, as we approach the cutoff frequency
ro == 1, the approximation becomes progressively poorer. We now
consider an approximation which "ripples" about unity in the pass band
and falls off rapidly beyond the cutoff ro == 1. The approximation is
equally good at ro == 0 and ro == 1, and, as a result is called an equal-
ripple approximation. The equal-ripple property is brought about by the
use of Chebyshev cosine polynomials defined as
C,.(ro) == cos (n cos-1 ro) lrol ~ 1
== cosh (n cosh-1 ro) lrol >1 (13.29)

For n == 0 we see that C0(ro) == 1 (13.30)

and for n == 1, we have Ci{ro) == ro (13.31)
Higher order Chebyshev polynomials are obtained through the recursive
formula
C,.(ro) == 2ro c.._1(ro) - c_1(ro) (13.32)
Thus for n == 2, we obtain C.(ro) as
C.(ro) == 2ro(ro) - 1 (13.33)
== 2ro8 - 1
In Table 13.3, Chebyshev polynomials of orders up ton== 10 are given.
TABLE 13.3
n Cbebyshev polynomials c..(co) = cos (n cos-1 co)

0 1
1 co '
2 2col - 1
3 4co3 - Jco
4 Seo' - 8eo1 + l
s 16eoli - 20eo3 + Seo
6 32eo1 - 48eo' + 18w8 - 1
7 64eo7 - 112w11 + S6w3 - 7w
8 128w8 - 256w8 + 160eo' - 32col + 1
9 256w8 - S76w7 + 432w11 - 120w8 + 9w
10 Sl2w10 - 1280w8 + l 120w1 - 400w' + SOcol - 1
374 Network analysis and synthesis
C3(W)
tJ__ _

FIG. 13.4. C1(w) and Ca(w) Chebyshev polynomials.

The pertinent properties of Chebyshev polynomials used in the low-pass

filter approximation are:

1. The zeros of the polynomials are located in the interval lwl ~ 1,

as seen by the plots of C3(ro) and Ciro) in Fig. 13.4.
2. Within the interval lwl ~ 1, the absolute value of C,.(ro) never
exceeds unity; that is, IC,.(ro)I ~ 1 for lwl ~ 1.
3. Beyond the interval lwl ~ 1, ICn(ro)I increases rapidly for increasing
values of lwl.

Now, how do we apply the Chebyshev polynomials to the low-pass

filter approximation? Consider the function e2 C,.2(ro), where e is real and
small compared to 1. It is clear that e2 C,. 2( w) will vary between O and e2
in the interval lrol ~ 1. Now we add 1 to this function making it 1 +
e2 C,. 2(ro). This new function varies between 1 and 1 + e2, a quantity
slightly greater than unity, for lwl ~ 1. Inverting this function, we obtain
the function which we will associate with IH(jro)l 2 ; thus

(13.34)

Within the interval lwl ~ 1, IH(jw)l 2 oscillates about unity such that the
maximum value is 1 and the minimum is 1/(1 + e2). Outside this interval,
C,.2(ro) becomes very large so that, as ro increases, a point will be reached
where e1 C,.2(w) » 1 and IH(jro)l 2 approaches zero very rapidly with
further increase in w. Thus, we see that IH(jw)l 2 in Eq. 13.34 is indeed a
suitable approximant for the ideal low-pass filter characteristic.
Figure 13.5 shows a Chebyshev approximation to the ideal low-pass
filter. We see that within the pass band O ~ w ~ 1, IH(jro)I ripples
between the value l and (1 + e2)-½. The ripple height or distance between
 Topics In filter design 375

IHOw>I

"'
FIG. 13.5. Chebyshev approximation to low-pass filter.

maximum and minimum in the pass band is given as

=
+l E~~
Ripple 1- (13.35)
(1

= ({ + E~~
1
At w = 1, IH(jw)I is IHUl)I (13.36)

because C,.1(1) = 1.
In the stop band, that is, for lwl ~ 1, as w increases, we reach a point
w 1 , where E1 C,.l(w)» 1 so that
1
IHUw)l,.....,-- (13.37)
ECn{w)
The loss in decibels is ~ven as

Loss = -20 log10 IH(jw)I

(13.38)
~ 20 log E + 20 log C,.(w)

But for large w, C,.(w) can be approximated by its leading term 2-1w",
so that
Loss = 20 log E + 20 log 2-1 w"
(13.39)
= 20logE + 6(n -1) + 20nlogw
376 Network analysis and synthesis
We see that the Chebyshev response also falls off at the rate of 20n
db/decade after an initial drop of 20 log E +
6(n - 1) decibels. However,
in most applications, E is a very small number so that the 20 log E term is
actually negative. It is necessary, therefore,
to compensate for this decrease in loss in
iw the stop band by choosing a sufficiently
large n.
From the preceding discussion, we see
that a Chebyshev approximation depends
upon two variables, E and n, which can be
determined from the specifications directly.
The maximum permissible ripple puts a
sinh 13,. bound on E. Once E is determined, any
--+-------- (1 desired value of attenuation in the stop
band fixes n.
The derivation of the system function
H(s) from a Chebyshev amplitude approxi-
mation IH(jw)I is somewhat involved and
will not be given here. s Instead, we will
simply give the results of such a derivation.
First we introduce a design parameter.
FIG. 13.6. Locus of poles of
Chebshev filter. 1 . h_1 1 (13.40)
/31:=-s1n
n
-
E

where n is the degree of the Chebyshev polynomial and E is the factor

controlling ripple width. The poles, s1: = <11: + jw1:, of the equal-ripple
approximant H(s) are located on an ellipse in the s plane, given by
I I
<11: + <.01: =1 (13.41)
sinh1 {J.,. cosh1 {J-,.
The major semiaxis of the ellipse is on the jw axis and has a value w =
±cosh /J1:• The minor semiaxis has a value e1 = ±sinh /31:, and the foci
are at w = ± 1 (Fig. 13.6). The half-power point of the equal-ripple
amplitude response occurs at the point where the ellipse intersects the jw
axis, i.e., at w = cosh {J.,.. Recall that for the Butterworth response, the
half-power point occurs at w =
1. Let us normalize the Chebyshev poles
s1: such that the half-power point also falls at w = 1 instead of at w =
cosh /31:; i.e., let us choose a normalizing factor, cosh {J.,., such that the

1
Interested parties are referred to M. E. Van Valkenburg, Introduction to Modern
Network Synthesis, John Wiley and Sons, New York, 1960, Chapter 13.
 Topics In filter design 3"n
normalized pole locations s' 1: are given by

, S1,
S1:=--
COSh Pt
=_!!};_ + jw1,
(13.42)
cosh P1: cosh /J1:

=~ (1k +JW1,
I I

The normalized pole locations can be derived as

, = tanhp1:Stn. (2k
<11: -n - -1).,,.
2
(13.43)
w,., = cos (2k
-n - -1).,,.
2
Comparing the normalized Chebyshev pole locations with the Butterworth
pole locations in Eq. 13.24, we see that the imaginary parts are the same,
while the real part a',. of the Chebyshev pole location is equal to the real
part of the Butterworth poles times the factor tanh {J,.. For example, with
n = 3 and tanh {J,. = 0.444, the Butterworth poles are

=
S1 -1 + j0
S2,s = -0.5 ± j0.866
so that the normalized Chebyshev poles are given by

= -1(0.444) + j0
S1

= -0.444 + j0
S2,S = -0.5(0.444) ±j0.866

= -0.222 ± j0.866
Finally, to obtain the denormalized Chebyshev poles, we simply multiply
s',. by cosh {J,., that is,
s,. = (a',.+ jw' J cosh {J,. (13.44)
There is an easier geometrical method to obtain the Chebyshev poles,
given only the semiaxis information and the degree n. First we draw two
circles, the smaller of radius sinh p,. and the larger of radius cosh {)1 , as
378 Network analysis and synthesis

'
''
'\ \
\
I
I (T

I
I
I

shown in Fig. 13.7. Next, we draw radial lines according to the angles of
the Butterworth poles (Eq. 13.22) as shown. Finally, we draw vertical
dashed lines from the intersections of the smaller circle and the radial lines,
and horizontal dashed lines from the intersections of the large circle and
the radial lines. The Chebyshev poles are located at the intersection of
the vertical and horizontal dashed lines, as shown in Fig. 13.7.
Consider the following example. We would like to obtain a system
function H(s) that exhibits a Chebyshev characteristic with not more than
I-decibel ripple in the pass band and is down at least 20 decibels at co= 2.
When we design for I-decibel ripple, we know that at co= I, IH(jl)I is
down 1 decibel so that
1
20 log IHUl)I = 20 log ½ = -1 (13.45)
(1 + £)2

1
We then obtain - - ---:1A,..., = 0.891 (13.46)
(1 + £)
1

and £ = 0.509 (13.47)

Our next task is to find n from the 20 decibels at co = 2 specification.

From Eq. 13.39 the loss can be given as approximately

20 ,..._, 20 log 0.509 + 6(n - 1) + 20n log 2 (13.48)

Solving for n, we obtain n = 2.65. Since n must be an integer, we let n = 3.

With the specification of n and £, the pole locations are completely speci-
fied. Our next task is to determine these pole locations. First we must
 Topics In filter design 379
find Pt· From Eq. 13.40 we have

R
flt=
1 . h-11-
-sm
n E (13.49)
= i sinh- 1
1.965 = 0.476
In order to find the normalized Chebyshev poles from the Butterworth
poles, we must first determine tanh Pt· Here we have
tanh Pt= tanh 0.476 = 0.443 (13.50)
From Table 13.1, then= 3 Butterworth poles are
s1 = - 1.0, s 2, 3 = -0.5 ± j0.866 (13.51)
Multiplying the real parts of these poles by 0.443, we obtain the normalized
Chebyshev poles.
s'1 = -0.443, s' 2,3 = -0.222 ± j0.866
Finally, the denormalized Chebyshev poles are obtained by multiplying
the normalized ones by cosh Pt= 1.1155 so that the denormalized poles
are s1 = -0.494 and s2,3 = -0.249 ±j0.972.
H(s) is then
H(s) = 0.502
(s + 0.494)(s + 0.249 - j0.972)(s + 0.249 + j0.972)
(13.52)
0.502
= ----- ------
s31
+ 0.992s + 1.255s + 0.502
In Fig. 13.8, the amplitude responses of the Chebyshev and an n = 3
Butterworth filter are shown.
Monotonic filters with optimum cutoff
In comparing Butterworth filters with Chebyshev filters, the following
can be said. The Butterworth response is a maximally fiat, monotonic
response, whereas the Chebyshev response is equal ripple in the pass band.
In the stop band, the Chebyshev response falls otr more rapidly than the
Butterworth (except when Eis very, very small). In this respect, the Cheby-
shev filter is a better filter than the Butterworth. However, as we shall see
in Section 13.5, the transient response of the Chebyshev filter is very poor.
If we require sharp cutotr characteristics for a given degree n, however,
the Butterworth filter is quite unsatisfactory. In 1958, Papoulis3 proposed

• A. Papoulis, "Optimum Filters with Monotonic Response," Proc. IRE, 46, No. 3,
March 1958, pp. 606-609.
380 Network analysis and synthesis
Radian frequency, w
0.1 02 0.3 0.4 0.6 08 1 2 3

-1

-2
--- i--..__
- .......
~ k' ..-
\
"
\
\
\
-3
-4 \\ \I
I
-5
--
Amplitude response of n • 3
- - - Chebyshev filter with
1.0-db ripple in pass band
- - - - - Butterworth response (n = 31 ii
\
-9 ,,
-10 \
\
-11 I
\
-12 I
\
I
-13
I
-14 I I

FIG. 13.8. Amplitude response of n = 3 Chebyshev filter with LO-decibel ripple in pass
band and Butterworth response (n = 3).

a class of filters called Optimum or "L,. filters, which have the following
properties:
1. The amplitude response is monotonic.
2. The fall-off rate at w cutoff is the greatest possible, if monotonicity is
assumed.
3. The zeros of the system function of the L filter are all at infinity.
·Recall that the magnitude response of a low-pass filter with all zeros at
infinity can be expressed as

M(ro) = [1 + :C:,•)]H (13.53)

Let us denote the polynomial generating the L filter by

/(w1) = L,.(w1) (13.54)
 Topics In filter design 381
The polynomial Ln(w~ has the following properties:
(a) L,.(O) = 0
(b) L,.(1) = 1

~O
1
dL,.(w )
(c)
dw

(d) ~ W w=l
=M (Mmaximum)

Properties a, b and c are the same as for the Butterworth generating poly-
nomial/(w~ = w1 ". Property c insures that the response M(w) is mono-
tonic and property d requires that the slope of L,.(w~ at w == 1 be the
steepest to insure sharpest cutoff.
Papoulis originally derived the generating equation for the polynomials
L,. (for n odd) to be
1-1[ JI
where n = 2k +1
L,.(w") = _
f lw

1
le
~ a, Ph:) dx
0

and the Pix) are the Legendre polynomials of the

(13.55)

first kind'
P0(x) =1
P1(x) = x
Ps(x) = ½(3x2 - 1)
(13.56)
P 3(x) = ½(5x3 - 3x)

and the constants a, are given by

a1 a 1 ale 1
(13.57)
ao - 3 - 5 - - 2k +1- .[i(k + 1)
Later Papoulis5 and, independently, Fukada,8 showed that the even-
ordered L,. polynomials can be given by

i a, Pix)JIdx
1

La+,.(w") = _lw1
n=2k+2
i -1
(x + 1) [ le

0
(13.58)

• E. Jahnke and F. Emde, Tables of Functions, Dover Publications, New York, 1945.
1
A. Papoulis, "On Monotonic Response Filters," Proc. IRE, 41, February 1959,
332-333.
• M. Fukada, "Optimum Filters of Even Orders with Monotonic Response,"
Trans. IRE, CT-6, No. 3, September 1959, 277-281.
382 Network analysis and synthesis
where the constants a; are given by:
Case 1 (k even):

(13.59)

01 = 03 = ' ' ' = D.t----1 = 0

Case 2 (k odd):
a1 a8 a,. 1
3- 7 - ... - -2k_+_l - -;J==(k=+=l)=(k=+=2) (13.60)
a0 = a2 = · · · = a 1,. = 0
Fukada tabulated the L,.(w 2) polynomials up to n = 1 together with
dL,.(w'l.)/dw evaluated at w =
1 to give an indication of the steepness of
the cutoff. This is shown in Table 13.4.
To obtain the system function H(s) for the L filter, we must factor the
equation for h(s2) and choose the Hurwitz factors as H(s).
1
= H(s) H(-s) = (13.61)
h(s2}
1 + L,.(-s2)
For example, for n = 3, the magnitude response squared is
M2(w) - 1
- 1 + La(w 1
)
(13.62)
1
= - - 1- - -4- - -8
1 +w - 3w + 3w
Substituting -w 2 = s we obtain
2,

1
h(s2} = H(s) H(-s) = , (13.63)
1 - s - 3s - 3s8
1

TABLE 13.4
L,.( co2) Polynomials
dL,.(1)
n
~

2 co4 4
3 3co8 - 3co4 + co2 8
4 6w8 - 6w6 + 3co4 12
S 20co10 - 40co8 + 28co6 - 8w4 + wl 18
6 S0co12 - 120co10 + 10Sw8 - 40w8 + 6w' 24
7 175~1' - S2Sco11 + 61Sco10 - 35Sw8 + 10Sco8 - lSw' + co1 32
 Topics In filter design 383
Frequency, rad/sec
0.1 0.2 0.3 0.4 0.6 0.8 1.0 2 3

-2
Amplitude response of
- - -- - .... ~
"I
\
-4 - - - - Optimum versus

-6
- Butterworth filters \
-8
\\
....
.a
ai-10
\\
J-a. -12 0.4
0
0.5 0.6
Frequency

........
0.8 1.0 1.2 , \\
~
-14 -0.5 - - - - ~ '
\\
\\
-16

-18
-1.0

1-1.5
,,,
,,
\\
~ \ \
-20 -2.0

-22

-24
-2.5

-3.0
Expanded
scale
,'
11 ', \
\\
-26 \\
FIG. 13.9. Amplitude response of Optimum versus Butterworth filters.
After we factor h(r), we obtain
H( ) 0.577 (l3 64)
s = sa + 1.31s2 + 1.359s + 0.577 ·
where the numerator factor, 0.577, is chosen to let the d-c gain be unity.
The poles of H(s) are s 1 = -0.62; su = -0.345 ±j0.901. The ampli-
tude response of third-order Optimum (L) and Butterworth filters are
compared in Fig. 13.9 Note that the amplitude response of the Optimum
filter is not maximally flat, although still monotonic. However, the cutoff
characteristic of the Optimum filter is sharper than the cutoff of the
Butterworth filter.
Linear phase filters
Suppose a system function is given by
H(s) = Ke-•T (13.65)
where K is a positive real constant. Then the frequency response of the
system can be expressed as
HUw) = Ke-i<»T (13.66)
384 Network analysis and synthesis
so that the amplitude response M(w) is a constant K, and the phase re-
sponse
,f,(w) = -wT (13.67)
is linear in w. The response of such a system to an excitation denoted by
the transform pair {e(t), E(s)} is
R(s) = K E(s)e-•T (13.68)
so that the inverse transform r(t) can be written as
r(t) = c-1 [R(s)]
(13.69)
= K e(t - T)u(t - T)
We see that the response r(t) is simply the excitation delayed by a time T,
and multiplied by a constant. Thus no signal distortion results from
transmission through a system described by H(s) in Eq. 13.65. We note
further that the delay T can be obtained by differentiating the phase
response ,f,(w), by w; that is,
Delay= - d,f,(w) = T (13.70)
dw
Consequently, in a system with linear phase, the delay of the system is
obtained by differentiating the phase response ,f,(w).
A system with linear phase and constant amplitude is obviously desirable
from a pulse transmission viewpoint. However, the system function H(s)
in Eq. 13.65 is only realizable in terms of a lossless transmission line called
a delay line. If we require that the transmission network be made up of
lumped elements, then we must approximate H(s) = Ke-•T by a rational
function in s. The approximation method we shall describe here is
due to Thomson. 7 We can write H(s) as

H(s) = e'T
Ko
(13.71)
Ko
=sinh
-- - ---
sT + cosh sT
where K0 is chosen such that H(O) = 1. Let the delay T be normalized to
unity and let us divide both numerator and denominator of H(s) by sinh s
to obtain
H(s) = K 0/sinh s (13.72)
coth s + 1

' W. E. Thomson, "Network with Maximally Flat Delay," Wireless Engrg., 29, Oct.
1952, 256-263.
 Topics in filter design 385
If sinh s and cosh s are expanded in power series, we have
s1 s' s'
coshs = 1 + - + - +-. + · · ·
2! 4! 6!
(13.73)
s3 s6 s7
sinhs=s+-+-+-+ ···
3! 5! 7!
From these series expansions, we then obtain a continued fraction expan-
sion of coth s as
1
coth s = ! + - - -- - -
s ~+ 1
s ?+ 1 (13.74)
s 1
-+
s
···
If the continued fraction is terminated in n terms, then H(s) can be
written as
H(s) = Ko (13.75)
B,.(s)
where B,.(s) are Bessel polynomials defined by the formulas
B0 =1
B1 =s+l (13.76)

B,. = (2n - l)B,._1 + s1B,._1

From these formulas, we obtain
B_1 =s1 +3s+3
(13.77)
B1 = s3 + 6s1 + 15s + 15
Higher order Bessel polynomials are given in Table 13.5, and the roots of
Bessel polynomials are given in Table 13.6. Note that the roots are all in
the left-half plane. A more extensive table of roots of Bessel polynomi-
als is given by Orchard.8
The amplitude and phase response of a system function employing an
unnormaliz.ed third-order Bessel polynomial

H(s) = 15 (13.78)
8
s + 6s1 + 15s + 15
•H.J. Orchard, "The Roots of Maximally Flat Delay Polynomials" IEEE Trans.
on Circuit T/,eory, CT-11 No. 3, September 1965, 452-454.
386 Network analysis and synthesis
TABLE 13.5
Coefficients of Bessel Polynomials
n bo bi ba ba b, b1 bs b7
0 1
1 1 1
2 3 3 1
3 15 15 6 1
4 105 105 45 10 1
5 945 945 420 105 15 1
6 10,395 10,395 4,725 1,260 210 21 1
7 135,135 135,135 62,370 17,325 3,150 378 28 1

are given by the solid lines in Figs. 13.10 and 13.11. These are compared
with the amplitude and phase of an unnormalized third-order Butterworth
function given by the dotted lines. Note that the phase response of the
constant-delay function is more linear than the phase of the Butterworth
function. Also, the amplitude cutoff of the constant-delay curve is more
gradual than that of the Butterworth.

TABLE 13.6
Roots of Bessel Polynomials

n Roots of Bessel Polynomials

1 -1.0 +j0
2 - l.S ± j0.866025
-2.32219 + j0
3 { -1.83891 ± jl. 75438
-2.89621 ±j0.86723
4 { -2.10379 ±j2.65742
-3.64674 + j0
5 -3.35196 ±jl.74266
( -2.32467 ±j3.57102
-4.24836 ±j0.86751
6 -3.73571 ±j2.62627
( -2.51593 ±j4.49267
-4.97179 + j0
-4.75829 ±jl.73929
7
{ -4.07014 ±j3.51717
-2.68568 ±j5.42069
 Topics In filter design 387
Frequency, rad/sec
0.1 0.2 0.4 0.6 0.8 1 2 3 4 5 6 7

-2
0
"'-, -...........
:'-
-4
-6
I\

\
\
"" '¥-\ .-Bessel

-8 \
.a -10
"O
:g-12
Butterworth~
\

'\
'\
.9 -14 \
\
-16 \ \
\
-18 ' \
-20
\. \1
'\
-22
-24 \ \
\\
-26
FIG. 13.10. Amplitude response of n = 3 Bessel and Butterworth filters.
'
Frequency, rad/sec
00 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

-15
-30
' ' "' ' , ...
r---.
r---.
~ ........ .....
-45
-60 ~
- .. .. , / -Bessel n-= 3

-
~..,c~

I
-75
-90 "" .....
,..._ ,_
1-105
i-120 " I'- 1-

~ -135
-150
-165
" f' . / Butterworth n • 3

""' .... ... ,

-180
~r----. ....
-195
-210
-....... i--

FIG. 13.11. Phase responses of low-pass f;ilters.

388 Network analysis and synthesis

13.5 TRANSIENT RESPONSE OF LOW-PASS FILTERS

In this section we will compare the transient response of the filters

discussed in Section 13.4. In particular, we will compare the step response
of the filters according to the following figures of merit:
1. Rise time tR· The rise time of the step response is defined here as
the time required for the step response to rise from 10% to 90% of its
final value as depicted in Fig. 13.12.
2. Ringing. Ringing is an oscillatory transient occurring in the response
of a filter as a result of a sudden change iil input (such as a step). A
quantitative measure of the ringing in a step response is given by its
settling time.
3. Settling time. The settling time is that time t, beyond which the
step response does not differ from the final value by more than ±2%,
as depicted in Fig. 13.12.
4. Delay time, t». Delay time is the time which the step response
requires to reach 50% of its final value as shown in Fig. 13.12.
5. Overshoot. The overshoot of the step response is defined as the
difference between the peak value and the final value of the step response
(see Fig. 13.12) expressed as a percentage of the final value.

4 10 12 14 16 18 20
Time t,·sec
FIG. 13.12. Figures of merit for step response. t 11 = rise time; 11 = setting time;
tD = delay time.
 Topics In filter design 389
Most of the foregoing figures of merit are related to frequency response,
particularly bandwidth and phase linearity. Some of the quantities, such
as rise time and delay time are intimately related to each other but have
rather tenuous ties with overshoot. Let us examine qualitatively the
relationships between the transient response criteria just cited and
frequency response.
Rise time and bandwidth have an inverse relationship in a filter. The
wider the bandwidth, the smaller the rise time; the narrower the band-
width, the longer the rise time. Physically, the inverse relationship could
be explained by noting that the limited performance of the filter at high
frequencies slows down the abrupt rise in voltage of the step and prolongs
the rise time. Thus we have
TR X BW= Constant (13.79)
Rise time is a particularly important criterion in pulse transmission. In
an article on data transmission, 11 it was shown that in transmitting a pulse
of width T1 through a system with adjustable bandwidths, the following
results were obtained:

Bandwidth Rise Time

( /0 = 1/TJ (milliseconds)

/c o.s
2/c 0.25
3/0 0.16
4/c 0.12
S/0 0.10

The table shows a definite inverse relationship between rise time and
bandwidth.
A definition of time delay is given by Elmore as the first moment or
centroid of the impulse response
TD = i"° t h(t) dt (13.80)

provided the step response has little or no overshoot. Elmore's definitio

of rise time is given as the second moment

TR= [211 i"° (t - TD)1 h(t) dt] !4 (13.81)

• R. T. James, "Data Transmission-The Art of Moving Information," IEEE

Spectrum, January 196S, 65-83.
390 Network analysis and synthesis
These definitions are useful because we

t can obtain rise time and delay time directly

l(s)
from the coefficients of the system function
V(s) C R H(s). Without going into the proof, which
is in Elmore and Sands, 10 if H(s) is given as
I H(s) = 1 + a1s + a8s2 + · · · + ans"
FIG. 13.13. R-C network. 1 + bis + bs5 + · · · + bmsm
2

(13.82)
the time delay Tn.is Tn = h1 - a1 (13.83)
and the rise time is
TR = {2'1T[h12 - al + 2(a 2 - h2))}¼ (13.84)
For the R-C network in Fig. 13.13, H(s) is
H(s) _ V(s) _ R (13.85)
- I(s) - 1 sRC +
so that Tv=RC
(13.86)
TR= J2'1TRC
It should be emphasized that Elmore's definitions are restricted to step
responses without overshoot because of the moment definition. The more
general definition of rise time is the 10--90% one cited earlier, which has
no formal mathematical definition.
Overshoot is generally caused by "excess" gain at high frequencies. By
excess gain we normally mean a magnitude characteristic with a peak
such as the shunt peaked response shown by the dashed curve in Fig.
13.14. A magnitude characteristic with no overshoot is the magnitude
characteristic of an R-C interstage shown by the solid curve in Fig. 13.14.
,,
I \
---- .,../ \
\
\
\
\.,-shunt peaked response
\
\
', ~

Log frequency
FIG. 13.14. Comparison of shunt-peaked and simple R-C magnitudes.

10 W. C. Elmore and M. Sands, Electronics, National Nuclear Energy Series, Div.

V, 1, McGraw-Hill Book Company, New York, 1949, pp. 137-138.

 Topia In filter design 391
1.2
,r
1.1
1.0
0.9 I'
I
r
-- 'r--1..
I
~ r---....

I
I
' -
-
'
'\.
LX"
~

- /
- -~

j0.8
J j

N
~ 0.7
J I I
E
~ 0.6 I I I
.s I
GI 0.5
-g .
io.4
n=3 In.= 1 /.,,, = 10
~ 0.3 J I I
0.2
I J

0.1
I I I
0
,) l./ ~V
0 2 4 6 8 10 12 14 16 18 20
Time t, sec
FIG. 13.15. Step response of normalized Butterworth low-pass filters.

The step responses of then= 3, n = 1, and n = 10 Butterworth filters

are shown in Fig. 13.15. Note ti1.at as n increases, the overshoot increases.
This is because the higher order Butterworth filters have flatter magnitude
characteristics (i.e., there is more gain at frequencies just below the cutoff).
Ringing is due to sharp cutoff in the filter magnitude response, and is
accentuated by a rising gain characteristic preceding the discontinuity.
The step response of an n = 3 Bessel (linear phase) filter is compared
to the response of an n = 3 Chebyshev filter with I-decibel ripple in
Fig. 13.16. We cannot compare their rise times since the bandwidths of
the two filters ·have not been adjusted to be equal. However, we can
compare their ringing and settling times. The Chebyshev filter has a
sharper cutoff, and therefore has more ringing and longer settling time
than the Bessel filter. Note also the negligible overshoot of the Bessel
filter that is charactedstic of the entire class of Bessel filters.
The decision as to which filter is best depends upon die particular
situation. In certain applications, such as for transmission of music,
phase is not important. In these cases, the sharpness of cutoff may be the
dominant factor so that the Chebyshev or the Optimum filter is better
than the others. Suppose we were dealing with a pulse transmission
system with the requirement that the output sequence have approximately
the same shape as the input sequence, except for a time delay of T =
T 1 - T1 , as shown in Fig. 13.17a. It is clear that a filter with a long rise
time is not suitable, because the pulses would "smear" over each other
392 Network analysts and synthesis
1.1
I/
1.0
0.9 !
,..
I "' ."- i....,;'
......
0.8
f I
I
0.7 ,I ----Step response of n • 3 Bessel filter
.g 0.6 J --Step response of n • 3 Chebyshev ~

{o.s I filter with 1-db ripple

~ 0.4 I
0.3
0.2
.I' J
0.1
i I
0
&
0 2 4 6 8 W ~ ~ ~ 18 20
Time. sec
FIG. 13.16. Comparison of filter transient responses.

,on n
T1
Input pulse train .
-+-
System
with short
rise and
settling
times
- J\I\ [\
T2
Output pulse train
(a)

Ti
,nn n-Input pulse train
System
with long
rise and
settling
times
-~ T2
Output pulse train
(b)

FIG. 13.17. Smearing of pulses in systems with long rise and setting times.

as seen in fig. 13.17b. The same can be said for long settling times.
Since a pulse transmission system must have linear phase to insure un-
distorted harmonic reconstruction at the receiver, the best filter for the
system is a linear phase filter with small rise and settling times.

13.6 A METHOD TO REDUCE OVERSHOOT IN FILTERS

We present here a method to reduce the overshoot and ringing of a

filter step response. The step response of a tenth-order Butterworth filter
is shown in Fig. 13.18. It is seen that the overshoot is about 18 %. We
 Topia In filter design 393
12

I~
1.0

0.8
7
~V
~~

--
/ Smp response

I 0.6
ci.
E
< 0.4
II
02

0
,.J"'· Second derivative

7
·' ......... t-.
·,. _.,, ----
-1.2
0 2 4 6
"' .........
8
~
W
/
I

U ~ ~ IB ~
Time t, sec
FIG. 13.18

note that after the first peak, the ringing of the step response bas an
approximate sinusoidal waveshape. Let us now consider the second
derivative of the step response shown by the dashed curve in Fig. 13.18.
Beyond the first peak of the step response, the second derivative is also
(approximately) sinusoidal, and is negative when the step response is
greater than unity, and positive when the step response is less than unity.
If we add the second derivative to the step response, we reduce the over-
shoot and ringing.11
Suppose «(t) is the step response and H(s) is the system function of the
filter. The corrected step response can be written as

«1(t) = «(t) + K -d' «(t)

-
1
(13.87)
dt
where K is a real, positive constant. Taking the Laplace transform of
Eq. 13.87, we have
CC«1(t)] = H!s) + Ks•[H!s)]
(13.88)
= (Ks 1 + 1) H(s)
s
11
F. F. Kuo, "A Method to Reduce Overshoot in Filters," Trans. IRE on Circuit
Theory, CT-9, No. 4, December 1962, 413-414.
3CU Network analysis and synthesis
1.2
,~ ....
1.0 -'~" ! ~
~-
... _.
.... --... I'-.._-:_•

0.8
j V
~
~

j
----Tenth order Butterworth-

0.4 ,,I -Zerosatw= :!: 1.0

-Zeros at w = :!: 1.5 -
0.2 /A
(~i
.,,
~
0
0 2 4 6 8 10 12 14 16 18 20
Time t, sec
FIG. 13.19

From Eq. 13.88 we see that by adding a pair of zeros on the jw axis at
s = ±j/✓K, the overshoot and ringing are reduced.
For low-pass filters, the factor 1/✓K must in general be greater than
the bandwidth of the system. For normalized Butterworth filters the
bandwidth is w = 1 so that K ~ 1. The factor K also controls the amount
of overshoot reductian. If K is too small, adding the zeros on the jw axis
-o
-2 --
·= ----::::::.~
-- ''•

"
·~

-4 \
\ ' '\_
-6
\ \ \
-8 1 '\ \
-10
.0 -12 \
---- Tenth order Butterworth

' \
i::,

c -14 - - - Zeros at w = :!: 1.0 \I

,I
~ -16 _ - - - Zeros at w = :!: 1.5 I I I
- - Zeros at w = ± 1.8: with
-18 original bandwidth also
\I i
-20
equal to 1.8 I~ 'I
-22 ~i !
-24
i, I/'
-26 /ft I; \
-28 !\ 1, \
0.1 0.2 0.3 0.4 0.6 0.8 1.0 1.5 2.0 3.0
Radian frequency, w
FIG. 13.20
 Topics In filter design 395
will have negligible effect. Therefore the zeros should be added some-
where near the band edge. Figure 13.19 shows the effects of adding
zeros at ro == ± 1 (i.e., right at the band edge), and at ro == 1.5. We see
that the further away the zeros are placed from the band edge, the less
effect they will have. The addition of the zeros will decrease the 3-decibel
bandwidth of the filter, however, as seen in Fig. 13.20. Therefore, a
compromise must be reached between reduction in bandwidth and
reduction in overshoot.
An effective way to overcome this difficulty is to scale up the bandwidth
by a factor of, for example, 1.8. Then the zeros are placed at ro == ± 1.8,
which will reduce the 3-decibel bandwidth to approximately its original
figure ro == 1.0, as shown in Fig. 13.20. The overshoot, however, will be
reduced as though the zeros were at the band edge.

13.7 A MAXIMALLY FLAT DELAY AND CONTROLLABLE

MAGNITUDE APPROXIMATION

In this section we will examine an interesting result, which is due to

Budak. 11 The result deals with linear phase approximation with con-
trollable magnitude. In Section 13.4 we discussed approximation of a
flat delay using Bessel polynomials. The resulting rational approximant
was an all-pole function (all transmission zeros at a== oo), whose de-
nominator was a Bessel polynomial. There was no control of the magni-
tude using the all-pole approximant. In Budak's method the magnitude
ia controllable, while the phaae ia as linear as the standard Bessel approxi-
mation.
Budak's approximation is obtained by introducing the parameter k to
split r into two parts such that

(13.89)

and then approximate independently e-'" and e-0•-ll• with all-pole Bessel
polynomial approximations. Thus the resulting approximation for r
will have Bessel polynomials for both numerator and denominator. The
poles of the c<i:-ll• approximant will be the zeros in the final approximant,
while the poles of the c'" approximant remain as poles in the final
approximant. For realizability, the degree of the e-<i:-ll• approximant
should be less than the degree of the e-'" approximant.

11
A. Budak, "A Maximally Flat Phase and Controllable Magnitude Approximation,"
Trans. of IEEE on Circul't 77teory, CT-12, No. 2, June 1965, 279.
396 Network analysis and synthesis

M (w) M, .----.---.--.---.----.----.

0.6

0.4 0.4 I---+--+-

0.2 t-----t----f

1 2 3 4 5 1 2 3 4 5
(a) (b)
FIG. 13.21

As an example, consider the approximation with three zeros and four

poles.
-Te• 105 (13.90)
e ~-----3 ----
1 ------
= (ks)' + 10(ks) + 45(ks) + 105(ks) + 105
-(►l)a 15 (13 91)
e ~ [(k - l)s]3 + 6[(k - l)s]1 + 15[(k - l)s] + 15 .

We then perform the operations as indicated by Eq. 13.89 to obtain

e-• !::.! 7{[(k - l)s] 3 + 6[(k - l)s]1 + 15[(k - l)s] + 15} (l3_ )
92
- (ks)' + 10(ks) + 45(ks)8 + 105(ks) + 105
8

In Fig. 13.21a the magnitude characteristic of Eq. 13.92 is plotted with

k as a parameter. The phase characteristic is given in terms of deviation
of phase from linearity liq,= w - q,(w) and is shown in Fig. 13.21b.
The improvement in phase linearity over the all-pole Bessel approximation
k == 1 is shown by these curves. Note as the bandwidth is increased
(k decreasing), phase linearity is improved. Figure 13.22 shows the step
response of Eq. 13.92 also with k as a parameter. Since the effect of
decreasing k increases bandwidth, the corresponding effect in the time
domain is to decrease rise time.
Budak also observes that ask decreases from unity, the poles and zeros
migrate to keep the phase linear. The zeros move inward from infinity
along radial lines, while the poles move outward along radial lines.
 Topia In filter design 3W

1.1

1.0
k•0.6-

0.9
~ //' / '
0.8
J~ 1/
0.7
k•0.1-
1W ~ -k• 0.8

0.6
fl
0.5
j rr---- _k• 1.0
(Bessel)
~
'8'
0.4
I
I. ,
0.3

0.2 /2'
0.1 J 'l/i
o.o I ~..... /./ 1/I
u _,/

-0.2
-0.1 \
-V
0 ~ M ~ M ll U U U ~ W U ~
t
FIG.13.22

13.8 SYNTHESIS OF LOW-PASS FILTERS

Given the system function of the low-pass filter as derived by the

methods described in Section 13.4, we can proceed with the synthesis of
the filter network. If we consider. the class of filters terminated in a 1-0
load, and if we let the system function be a transfer impedance,

Z1 i(s) = -Zn- (13.93)

1 + Zn
or a transfer admittance Y11(s)
Yn = (13.94)
1 + Yn
we can synthesize the low-pass filter according to methods given in
=
Chapter 12. For example, consider then 3 Optimum (L) filter function
398 Network analysis and synthesis
given as a transfer impedance
zst() 0.577 (13.95)
s = s8 +
1.3ls8 + 1.359s + 0.577
We see that the zeros of transmission are all at infinity. Since the numer-
ator of Z 11 is even, we divide both numerator and denominator by the
odd part of the denominator s3 + 1.359s. Thus we have

0.577
z -- ---
111 -8s + 1.359s
(13.96)
1.31s1 + 0.577
z - -8- - - -
BB - s + 1.359s
The structure of the low-pass filter with three zeros of transmission at
infinity is given in Chapter 12. We must synthesize z11 to give the Tr
reactance structure. This we accomplish through the following continued
fraction expansion of 1/za,.:
l.3ls2 + 0.577) s3 + 1.359s ( 0.763s
s8 + 0.440s
0.919s) 1.3ls1 + 0.577 ( 1.415s
l.3ls1
0.577) 0.919s ( 1.593s
0.919s

The optimum filter is shown in Fig. 13.23. For then = 3 Butterworth filter
given by the transfer impedance

1
Z (s) - - - -- - - (13.97)
11 -s8 +2s1 +2s+l
1 2s + 11
we have z2iCs) = --- ; Zsa(s) = -a---- (13.98)
s
8
+ 2s s + 2s
We then synthesize z111(s) by a continued fraction expansion to give the
filter shown in Fig. 13.23.

(a) Optimum filter (b) Butterworth filler

FIG. 13.23
 Topics in filter design 399
Lr,
-----
Ca C1
T
1D V2

l
FIG. 13.24. Canonical form for filters described in Tables 13.7, 13.8, and 13.9.

In Tables 13.7, 13.8, 13.9 are listed element values (up to n = 7) for
single-terminated Butterworth, Chebyshev (I-decibel ripple), and Bessel
filters, respectively. 13 These apply to the canonical realization for a
transfer impedance Z 21(s) shown in Fig. 13.24. If a Y21(s) realization is
desired, we simply replace all shunt capacitors by series inductors and
vice versa. The element values all carry over.
In Chapter 14, we will consider some examples of synthesis of double
terminated filters. To stimulate the curiosity of the reader, note that the
voltage-ratio transfer function V2 / V0 of the network in Fig. 13.25 is
precisely the n = 3 Butterworth function.
Recall that in Chapter 12, when we cascaded two constant-resistance
networks, the overall system function Ho(s) was the product of the in-
dividual system functions Hi(s} H 2(s). We can apply this property to
networks which are ~not constant-resistance if we place an isolation ampli-
fier between the networks, as shown in Fig. 13.26. Since pentodes provide
the necessary isolation, our task is simplified to the design of the individual
structures H 1(s), H,.(s), ... , H,.(s}, which we call interstage networks.
Some common interstage structures are shown in Fig. 13.27. In Fig.
13.27a a structure known as the shunt-peaked network is shown. The trans-
fer impedance of the shunt-peaked network is

z (s) _ _! s + Rf L (13.99)
u - C s2 + sR/L+ (1/LC}

FIG. 13.25. n = 3 double-terminated Butterworth filter.

11
More extensive tables are given in L. Weinberg's excellent book, Network Analysis
and Synthesis, Chapter 13, McGraw-Hill Book Company, 1962.
'400 Network analysis and synthesis

TABLE 13.7
Normalized Element Values for a Single Terminated
Butterworth FIiter

n C1 Ls Ca L, Ci; La C7
1 1.000
2 0.707 1.414
3 0.500 1.333 1.500
4 0.383 1.082 1.577 1.531
5 0.309 0.894 1.382 1.694 1.545
6 0.259 0.758 1.202 1.553 1.759 1.553
7 0.222 0.656 1.055 1.397 1.659 1.799 1.558

TABLE 13.8
Normalized Element Values for a Single Terminated
Chebyshev FIiter with I-db Ripple

n C1 Ls Ca L, C5 Le C7
1 0.509
2 0.911 0.996
3 1.012 1.333 1.509
4 1.050 1.413 1.909 1.282
5 1.067 1.444 1.994 1.591 1.665
6 1.077 1.460 2.027 1.651 2.049 1.346
7 1.083 1.496 2.044 1.674 2.119 1.649 1.712

TABLE 13.9
Normalized Element Values for a Single Terminated
Bessel FIiter

n C1 Ls Ca L, C11 La C7
1 1.000
2 0.333 1.000
3 0.167 0.480 0.833
4 0.100 0.290 0.463 0.710
5 0.067 0.195 0.310 0.422 0.623
6 0.048 0.140 0.225 0.301 0.382 0.560
7 0.036 0.106 0.170 0.229 0.283 0.349 0.511
 Topics in filter design -401

Interstage Interstage Interstage

network network network
FIG. I 3.26. Pentodes used as isolation amplifiers.

We sec that Z 11(s) has a real zero and a pair of poles which may be com-
plex depending upon the values of R, L, and C. In Fig. 13.27b, a simple
R-C interstage is shown, whose transfer impedance is

Z (s) -
11
.! __l _ (13.100)
- Cs+ 1/RC
Observe that all the filter transfer functions considered up to this point
are made up of pairs of conjugate poles and simple poles on the - a axis.
It is clear that if we cascade shunt-peaked stages and R-C stages, we can
adjust the R, L, and C elements to give the desired response characteristic.
The only problem is to cancel the finite zero of the shunt-peaked stage.
For example, if we wish to design an amplifier with an n = 3 low-pass
Butterworth characteristic, we first break up the system function into
complex pole pairs and real pole terms, as given by

Z(s)- l
11
- (s1 + s + l)(s + 1)
(13.101)
s+l 1- 1
= s1 +s+ls+ls+l
We then associate the individual factors with shunt-peaked or simple
R-C stages and solve for the element values. The n = 3 Butterworth
amplifier is given in Fig. 13.28.

L
C

(a) (b)

FIG. 13.27•. (a) Shunt-peaked interstage. (b) R-C interstage.

402 Network analysis and synthesis

FIG. 13,28. Butterworth amplifier.

IJ.9 MAGNITUDE AND FREQUENCY NORMALIZATION

In Section 13.8, we discussed the synthesis of low-pass filters with a

cutoff frequency of 1 rad/sec and a load impedance of 1 Q. Filters designed
with these restrictions are considered to be normalized in both cutoff
frequency and impedance level. We will now discuss methods whereby
the normalized filters can be converted into filters which meet arbitrary
cutoff frequency and impedance level specifications. Let us denote by a
subscript n the normalized frequency variable s.. and the normalized
c...
element values L .., R .. , and The normalized frequency variable s.. is
related to the actual frequency s by the relation
s
s.. = - (13.102)
Wo

where ro 0 , the normalizing constant, is dimensionless and is often taken to

be the actual cutoff frequency.
Since the impedance of an element remains invariant under frequency
normalization, we obtain the actual element values from the normalized
values by setting the impedances in the two cases equal to each other. For
example, for an inductor, we have
s..L .. = sL = woS,,.L (13.103)
From this equation we then obtain the denormalized value of inductance
as
L=L" (13.104)
Wo

Similarly, from the impedance 1/s..c.. of a frequency normalized capacitor

c.., we obtain the denormalized value of capacitance through the equation
1 1
--=- (13.105)
s..c.. sC
 Topics In filter design 403
so that the actual value of the capacitance is

C= C., (13.106)

Since resistances, ideally, are independent of frequency, they are unaf-

fected by frequency normalization.
Consider, next, impedance denormalization. Suppose the actual
impedance level should be Ro ohms instead of 1 n. Then a denormalized
impedance Z is related to a normalized impedance Z,. by
Z = Rr,Z., (13.107)
where Ro is taken to be dimensionless here. Thus, for a normalized resistor
R.., the denormalized (actual) resistance is
R = R.,R., (13.108)
For an inductance, the corresponding relationship is
sL = R.,(sL.,) (13.109)
so that the actual inductance value is
L=R0 L., (13.110)
Similarly, for a capacitor we have
1 R0
-=- (13.111)
sC sC,.
so that the actual capacitance is
C= C,. (13.112)
Ro
For combined frequency and magnitude denormalization, we simply
combine the two sets of equations to give
R = RoR,.
C= C,.
RoWo (13.113)
L= RoL,.
Wo
Let us consider an actual example in design. In Section 13.8, we
synthesized a transfer impedance Zt1 with an n = 3 Butterworth amplitude
characteristic with a cutoff frequency of 1 rad/sec and a load impedance of
l n. Let us redesign this filter for a cutoff frequency of lOt rad/sec to work
404 Network analysis and synthesis

66-7mh l
_ _ _ _....__0._3-µf_ _ _ ___._O_._lµ_f_500_0___,J

FIG. 13.29. Denormalized low-pass filter.

into a load of 500 n. From the original network in Fig. 13 .23, we take the
element values and denormalize with the normalizing factors, ru0 = 1()&
and R0 = 500.
Then the denormalized element values are

R = 500RL = 500
½
C1 = 500(l0') = 0.1 µf
(13.114)
L = f( 500) = 0.0067 h
10,000
t
C1 = 500(l0') = 0.3 µf
The final design is shown in Fig. 13.29.

13.10 FREQUENCY TRANSFORMATIONS

Up to this point, we have discussed only the design of low-pass filters,

while neglecting the equally important designs of high-pass, band-pass,
and band-elimination filters. We will remedy this situation here, not by
introducing new design procedures but through a technique known as a
frequency transformation, whereby, beginning from a normalized low-pass
filter, we can generate any other form of filter. Using frequency trans-
formations, the elements of the normalized low-pass filter are changed
into elements of a high-pass, band-pass, or band-elimination filter.
Analytically, a frequency transformation simply changes one L-C
driving-point function into another L-C driving-point function. Therefore,
the transformation equations must be L-C functions themselves. Also,
since we proceed from normalized low-pass filters, the transformation
equations include built-in frequency denormalization factors so that the
resulting networks need only be scaled for impedance level. Consider the
 Topia in filter design -405
simplest transformation equation, that of low-pass to high-pass, which is

s - roo (13.115)
s,.
where s,. represents the normalized low-pass frequency variable, s is the
regular frequency variable, and ro0 is the cutoff frequency of the high-pass
filter. In terms of real and imaginary parts, we have
• a>o
a+Jro---.:<.-
a,. + jro,.
(13.116)
ro0(a,. - jro,.)
- a
" +w
1
.1

Since we are interested principally in how the jro,. axis maps into the jro
axis, we let a,. - 0 so that
ro __ roo
{13.117)
ro,.
which is the equation that transforms normalized low-pass filters to
denormalized high-pass filters. From Eq. 13.117 we see that the point
ro,. - ± 1 corresponds to the point ro - ±ro0 • It is also clear that the
transformation maps the segment lro,.I ~ 1 on to the segments defined by
ro0 ~ lrol ~ oo, as shown in Fig. 13.30.
Now let us see how the frequency transformations change the network
elements. For convenience, let us denote the normalized low-pass network
elements with a subscript n, the high-pass elements with a subscript h, the
band-pass elements with a subscript b, and the band-elimination elements
with a subscript e. For the low-pass to high-pass case, let us first consider
the changes for the capacitor C,.. The transformation is given by the

jw,. jw

+i a,. plane iwo • plane

0 v,. 0 (T

-i -iwo

FIG. 13.30. Low-pass to high-pass transformation.

406 Network analysis and synthesis
equation
1 S A
- - =--=Lhs (13.118)
C,.s,. w0 C,.
For the inductor L,., we have
Wo A 1
L,.s,.=L,.-=- (13.119)
s Chs
We observe that a capacitor changes into an inductor and an inductor
changes into a capacitor in a low-pass to high-pass transformation (Fig.
13.31). The element values of the high-pass filter are given in terms of the
normalized low-pass filter elements as

Lh = -1- (13.120)
woe ..
1
and Ch=-- (13.121)
WoL ..
Consider the following example. From the normalized third-order
Butterworth filter given in Fig. 13.23, let us design a corresponding high-
pass filter with its cutoff frequency w 0 = 108 rad/sec and the impedance
level of 500 n. From the low-pass filter, we can draw by inspection the

Low-pass High-pass Band-elimination Band-pass

L,.BW
Lei= wo2

L,.
~ o-------1 ~
C h =1--
woLn
-0. Cel -- L,.BW
1
Lb1
~(--<,
.!!n..
BW
Cb1

BW
.,02L,.

Cb2=-jw

FIG. 13.31. Element changes resulting from frequency transformations.

 Topics In filter design -407
Ct,= 1.5 X 10-9 f

-8
Ltl.•1\-h

FIG. 13.32. Transformation of low-pass filter in Fig. 13.23 into high-pass filter.

high-pass-filter circuit shown in Fig. 13.32. Its element values are:

RL = 5000
Lu= 500
8
= 10-ah
10 (½)
(13.122)
C = l = 1.5 x H,'f
" (500)108(!)
5
L.,. = 00 = 0.333 X 10-3 h
1<>8(t)
Next, let us examine the low-pass to band-pass transformation (also
an L-C function):
0
Sn= ; ~ ( ;
0
+: ) (13.123)

where, if wa1 and wm denote the upper and lower cutoff frequencies of the
band-pass filter, BW is the bandwidth
BW= waa- wm (13.124)
and w 0 is the geometric mean of wa1 and wm

Wo = ✓waawm (13.125)
The low-pass to band-pass transformation maps the segment lwnl ~ 1 to
the segments waa ~ lool ~ w 01 , shown in Fig. 13.33. The normalized
low-pass elements are then modified according to the following equations:

L S = Ln S + WoI Ln
n n BW BWs
(13.126)
.f08 Network analysis and synthesis

jw
iwn c.Ja

WO
+j "'ct
•n plane • plane

0 (In 0 (T

-j

FIG. 13.33. Low-pass to band-pass transformation.

We note that the inductor L,. is transformed into a series-tuned tank,

shown in Fig. 13.31, whose elements are given as

Lbl = L,.
BW
(13.127)
BW
Cbl=-.-
OOo L,.

The capacitor C,. is transformed into a parallel-tuned tank (Fig. 13.31),

whose eiements are
BW
Lbll=-1-
roo C,.
(13.128)
CH= c,.
BW
Let us transform the third-order Butterworth low-pass filter in Fig. 13.23
into a band-pass filter with a 1-0 impedance level, whose bandwidth is
BW = 6 X 104 rad/sec, and its band-pass is "centered" at ro0 = 4 x 104
rad/sec. We draw the band-pass filter shown in Fig. 13.34 by the rules

FIG. 13.34. Band-pass filter transformed from low-pass filter in Fig. 13.23.
 Toplcs In filter design -409
given above. The element values of the band-pass filter are given in the
following equations:

Li_ = 6 X lO' = 0.75 X lo-4 h

(4 X 10')1(!)

C1
½ = -1 X
= ---=e.-- lo-'f
6 X 10' 12

Ls = t =~ X lo-4 h
6 X 10' 9
(13.129)
C1 = 6 X lO' = _! X lo-4 f
(4 X 10')8(t) 32

6
La= X lO' = 0.25 X 1o-4b
(4 X 10')8(-f)
t
C1 =6 X l0' = 0.25 X U,' f

Finally, the band-elimination filter is obtained through the transfor-

mation
BW
s,.= (13.130)
coo(_!_+ coo)
COg S

where BW and ro0 are defined in a manner similar to that for the band-pass
filter. The transformation maps the segment ofthejro,. axis in Fig. 13.35a
onto the segments shown on the jro axis in Fig. 13.35b. For the low-pass
jw

0 "" 0

-j

(a) (b)
FIG. 1:us. Low-pass to band-elimination transformation.
410 Network analysis and synthesis
to band-elimination transformation we, therefore, have the following
element changes:
1
Lnsn = •
(s/L,.BW) + (w0 /L,.BWs)
A 1
Ceis + (1/L.is)
(13.131)
_1__ ~ (~ + Wo)
C,.s,. C,.BW w 0 s
A 1
= L •.s +-
c•.s
Observe that the normalized low-pass inductor goes into a parallel-
tuned circuit and the capacitor C,. goes into a series-tuned circuit, as
shown in Fig. 13.31. In Table 13.10, we have a composite summary of the
various transformations.
- TABLE 13.10
Table of Various Frequency Transformations
Transformation
Low-Pass to Equation

High-pass

a>o ( s a>o)
Band-pass s =-
n BW -+-
w0 s

BW
Band-elimination

Problems
13.1 Find the transfer impedance Z 11 = VJ/1 for the filter shown in the
figure. What should L be in order for IZu(jw)I to be maximally flat?

PROB.13.1
 Topics In filter design ◄I I
13.2 Find the poles of system functions with n = 3, n = 4, and n == S
Butterworth characteristics. (Do not use the tables.)
13.3 Show that the half-power point of a Chebyshev low-pass amplitude
response is at "' = cosh {J,. for • « I.
13.4 Determine the system function for the following filter specifications:
(a) Ripple of l db in band lwl ~ I;
(b) at"' = 3, amplitude is down 30 db.
13.5 Compare the slopes at "' = 1 of the following polynomials (for n = 3):
(a) /(w8) = w•"
(b) /(w1) = !C,.(2w1 - 1) +l = C,.l(w)
(c) /(w8) = L,.(w1).
13.6 Determine the polynomials L.( w8) and L,.( a,8).
13.7 Expand cosh s and sinh s into power series and find the first four terms
of the continued fraction expansion of cosh s/sinh s. Truncate the expansion at
n = 4 and show that H(s) = Ko/B.(s).
13.8 Synthesi7.e the n = 3 linear phase filter as a transfer impedance termin-
ated in a 1-!l load.
13.9 Synthesi7.e the low-pass filter, which, when terminated in a 1-!l resistor,
will have a transfer admittance whose poles are shown in the figure.

jw

PROB. 13.9.

13.10 Determine the asymptotic rate of falloff in the stop band of: (a)
optimum filters; (b) linear phase filters.
13.11 Synthesi7.e the n = 3 and n = 4 Butterworth responses as transfer
impedances terminated in a load of 600 n with a cutoff frequency of I 08 rad/sec.
13.12 Synthesi7.e a Chebyshev low-pass filter to meet the following speci-
fications:
(a) load resistor, RL = 600 n
(b) !-db ripple within pass band
(c) cutoff frequency = S x 105 rad/sec
(d) at 1.5 x 108 rad/sec, the magnitude must be down 30 db.
-412 Network analysis and synthesis
13.13 Synthesi7.e n "" 3 Optimum and linear phase filters to meet the following
specifications:
(a) load resistor - 108 n
(b) cutoff frequency ,. 10' rad/sec.
13.14 Design an n = 4 Butterworth amplifier with the following specifi-
cations:
(a) impedance level = 500 n
(b) cutoff frequency = 108 rad/sec.
13.15 Synthesi7.e a high-pass filter for a given transfer admittance terminated
in a 108-0 load, whose amplitude characteristic is Optimum (L), with a cutoff
frequency of roe =- 10' rad/sec.
13.16 Synthesi7.e: (a) a band-pass filter; (b) a band-elimination filter, with
maximally flat (n = 4) amplitude response with 0,171 = 8 x 10' and °'oi =
2 X 10'.
 chapter 14
The scattering matrix

14.1 INCIDENT AND REFLECTED POWER FLOW

In this chapter, we will devote our attention to certain power relation-

ships in one- and two-port networks. The characterization of a network in
terms of power instead of the conventional voltage-current description is a
helpful analytical tool used by transmission engineers. It is especially
important in microwave transmission problems where circuits can no
longer be given in terms of lumped R, L, and C elements. In the power-
flow description, we are concerned with the power into the network, which
we call the incident power and the power reflected back from the load,
which is the reflected power. A convenient description of the network in
terms of incident and reflected power is given by the scattering matrix,
which is the main topic of discussion in this chapter.
It is convenient to think of incident and reflected power when dealing
with transmission lines. Therefore, we will briefly review some concepts in
transmission line theory. For a more comprehensive treatment of trans-
mission lines, the reader is referred to any standard text on wave propa-
gation.1 Consider the transmission line shown in Fig. 14.1. The voltage
at any point down the line is a function of z, the distance from the source.
The parameters which describe the transmission line ~ given in the
following:
R = resistance per unit length
G = conductance per unit length
L = inductance per unit length
C = capacitance per unit length
1 See, for example, E. C. Jordan, Electromagnetk Wave.r and Radiating Sy.rtem.r,

Prentice-Hall. Englewood Cliff's, New Jersey, 1950.

413
414 Network analysis and synthesis

~----L-----

l' 1--x- 2'

FIG. 14.1. Transmission line.

Given these parameters, we can now define the impedance per unit
length as
Z=R+jwL (14.1)
and the admitt~ce per unit length as
Y=G+jwC (14.2)
The characteristic impedance 2 0 of the line is given in terms of Z and Y as
20 = Jz/Y (14.3)
and the propagation constant is
r= ✓zi (14.4)
With these definitions in mind, let us turn to the general equations for
the current and voltage at any point x down the line

V(x) = "'e- + V,.e7

"'
7
"'

I(x) = I,e- Ire 7

"' -
7
"'
(14.5)
V, _ ., Yr .,
=-e 1 --e7
Zo Zo
The terms with the subscript i refer to the incident wave at point x and the
terms with subscript r refer to the reflected wave at x. Solving Eqs. 14.5
simultaneously, we obtain explicit expressions for the incident and reflected
waves
Yte- 1"' = ½[V(x) + Z0 I(x)]
(14.6)
V,.e 1"' = ½[V(x) - Z0 I(x)]
Consider the case when a transmission line of length L is terminated in
its characteristic impedance, that is, ·
V(L) _ z (14.7)
0
I(L) -
 The scattering matrix ◄ IS

Then we sec that the reflected wave is zero.

V,.e1L - 0 (14.8)
Since e1L cannot be zero, we see that the coefficient Yr is identically zero
for this case. As a result, the reflected wave at any point z is zero. Also,
the impedance at any point z down the line is equal to Z 0 as seen from
Eq. 14.S with Yr - 0. With these brief thoughts of transmission lines in
mind, let us turn our attention to the main topic of this chapter, namely,
the scattering parameters.

14.2 THE SCATTERING PARAMETERS FOR A

ONE-PORT NETWORK

For the one-port network shown in Fig. 14.2a, consider the following
definitions. The incident parameter a is defined as

a- !(....!:... + ✓Roi) (14.9)

2 ✓Ro
and the reflected parameter b, is defined as

(14.10)

where Ro is an arbitrary, positive, dimensionless constant called the

reference impedance factor. For the transmission line described in
Section 14.1, if the characteristic impedance Z 0 - Ro, then we can describe
the incident parameter in terms of the incident voltage as
Yte-1•
a - ---=-
vZo
(14.11)

Similarly, b can be expressed in terms of the reflected wave as

V,.e1•
b--=- (14.12)
✓zo

I
+ One-port
V
·• network

(6)
FIG. 14.2. Scattering parameters ot a one-port network.
,.,6 Network analysis and synthesis
Thus we see that the parameters a and b do indeed describe an incident-
reflected wave relationship in a one-port network, as depicted in Fig. 14.2b.
To give further meaning to a and b, consider the power dissipated by the
one-port network
P == !Re VI* (14.13)

where I* denotes the complex conjugate of I. From Eqs. 14.9 and 14.10
we solve for V and I in terms of the incident and reflected parameters to
give
V == (a + b)JR0
(14.14)
l=a-b
JRo
Then the power dissipated by the one-port network is

P = l(aa* - bb*)
(14.15)
= !(lal 1
- 1h11)
where, again, the asterisk denotes complex conjugate. The term laa*
can be interpreted as the power incident, while !bb* can be regarded as the
power reflected. The difference yields the power dissipated by the one-port
network.
The incident parameter a and the reflected parameter b are related by the
equation
b=Sa (14.16)

where Sis called the scattering element or, more commonly, the reflection
coefficient. From the definitions of a and b we can make the following
substitution:
(14.17)

Solving for S, we obtain S == _Z_-_R_,.o (14.18)

Z+Ro
where Z is the impedance of the one-port network

Z == V (14.19)
I
A further useful result is that when the impedance Ro is set equal to the
impedance Z, the reflected parameter b = 0.
For the one-port network excited by a voltage source V, with a source
 The scattering matrix 417
resistor R,,, as depicted in Fig. 14.3, we
will show that when we choose the I
reference impedance Ro to be equal to +
R,, V
V.
a=~ (14.20)
2JR,,
The proof follows. By definition, the
FIG. 14.3
incident parameter a is given as

(14.21)

From Fig. 14.3, we have V,,-IR,,= V {14.22)

and I= v, (14.23)
R,,+z
Substituting these equations for V and / into the expression for a in Eq.
14.21, we obtain
a =_!_[(v. _ V.R,, ) + R,,R +Y,Z] ·
2.JRo " R,, + Z
0

= V,, (t + R R,,) 0 -
{14.24)
2JRo R,, +Z
v.
= 2JRo I
&-R.
Consider once more the expression for the power dissipated in a one-
port network:
P = ½(aa* .... bb*) (14.25)
Factoring the term aa* = lal1 from within the parentheses, P becomes
1 1
p = lal ( 1 _ 1h11)
2 lal
(14.26)
= lal• (1 - 1S19>
2
When we choose the reference impedance to be equal to the source resist-
ance, i.e., when Ro = R,,, then S = 0 and
1 1
lal IV,1
P.,4=-=-- (14.27)
2 SR,,
418 Network analysis and synthesis
where PA represents the available gain or
available power of a voltage source v. with
a source resistance R,,. For the case of a
one-port network, the available gain is
defined as the power dissipated in the one-
port network when the impedance of the
network Z is equal to the resistance of the
FIG. 14.4 source R,,. As a result of this definition, we
see that for the one-port network shown in
Fig. 14.4, the power dissipated in Z with Z = R,, is

PA= 1¥pll (14.28)

8R11
The available gain thus represents the maximum available power at the
terminals of the voltage source.
From this discussion, it is apparent that the value of the reference im-
pedance Ro should be chosen equal to the source impedance R,,. A
standard procedure is to assume a 1-'2 source impedance and denormalize
when necessary, that is, let
z-1
S=-- (14.29)
z+t

where z=-
z (14.30)
Ro
Next, let us briefly consider some of the important properties of the
scattering parameter S for a one-port network.

1. The magnitude of S along the jw axis is always less or equal to unity

for a passive network, that is,
IS(jw)t ~ 1 (14.31)
This property follows from the fact that the power dissipated in a passive
network is always greater or equal to zero. Since the power can be
expressed as

~0
1
P = lal (1 - ISi') (14.32)
2
we see that IS(jw)l 1 ~ 1 (14.33)

2. For a reactive network IS(jw)I =

1. This property follows from the
fact that the power dissipated in a purely reactive network is zero.
 The scattering matrix ,.19

3. For an open circuit S =

1, and for a short circuit S = -1. This is
shown to be true from the equation
z-1
S - -- (14.34)
z+l
For an open circuit z = oo, so that S = 1. For a short circuit z - O;
therefore S =-1.
Before we proceed to the next property, let us consider the following
definition.
DEFINITION. A bounded real function F(s) is defined by the con-
ditions
(a) IF(s)I ~ K for Res ~ 0
(b) .F(s) is real whens is real. ·
In (a), K denotes any positive real constant.
4. If z - Z/Re is a positive real function, then S is a bounded real
function.
The proof follows from the equation,
S - (z - 1)/(z + 1).
From the positive real condition (a) Re z(s) ~ 0, when Re s ~ 0, we see
that
S =
j Im z(s) - (1 - Re z(s)]
(14.35)
j Im z(s) + [1 + Re z(s)]

so that IS(s)I - {Im• z(s) + (1 - Re z(s)J•}!-i ~ 1 (14.36)

lm1 z(s) + (1 + Re z(s)f
when Res ~ 0. (b) Wheres is real, z(s) is real. Then
S - (z - 1)/(z + 1)
must be real. Thus the scattering parameter for a passive network is a
bounded real function.

14.3 THE SCATTERING MATRIX FOR A

TWO-PORT NETWORK

In this section we will .extend the concepts developed for one-port

networks discussed in Section 14.2 to two-port networks. In the two-port
network shown in Fig. 14.5, we are concerned with two sets of incident and
reflected parameters {ai, b1 } at the 1-1' port, and {a., b1 } at the 2-2' port.
These parameters are defined in similar manner as for the one-port
"420 Network analysis and synthesis

R1[ '•, T= : • .: JR
2L
l' -
1

:,

...__ _ _..., -
..E'4.2'
2

FIG. 14.5. Scattering parameters for a two-port network.

network, that is,

(14.37)
a1 =
2
t• + .JRosla)
!(vR01
b. = H✓~OI - ✓Rosi.)
where R01 and R08 are the reference impedances at the input and output
ports respectively.
The scattering parameters s., for the two-port network are given by the
equations
b1 = Sua1 + S1.a1
(14.38)
b8 = S11a1 + S1.a1
In matrix form the set of equations of Eqs. 14.38 becomes

(14.39)

where the matrix [SJ= [Su S11]

S11 Sn
(14.40)

is called the scattering matrix of the two-port network. From Eqs. 14.38,
we see that the scattering parameters of the two-port network can be
expressed in terms of the incident and reflected parameters as

(14.41)
 The scattering matrix -421

r----7
I
I
I
I
·~ I
I
I
L ____ _JI
8s
FIG. 14.6

In Eqs. 14.41, the parameter S 11 is called the input reflection coefficient;

S 11 is ·the forward transmission coefficient; S11 is the reverse transmission
coefficient; and Sn is the output reflection coefficient. Observe that all
four scattering parameters are expressed as ratios of reflected to incident
parameters.
Now let us examine the physical meaning of these scattering parameters.
First, consider the implications of setting the incident parameters a 1 and as
to zero in the defining relations in Eqs. 14.41. Let us see what the con-
dition a,. = 0 implies in the definition for the forward reflection coefficient

Sn= b1l
a1 -o
Figure 14.6 shows the terminating section of the two-port network of Fig.
14.5 with the parameters a1 and b,. of the 2-2' port shown. If we treat
the load resistor R1 as a on~port network with scattering parameter

R,. -Ro.
s. = --=---== (14.42)
R,.+ROI
where Ru is the reference impedance of port 2, then as and b1 are related
by•
(14.43)
When the reference impedance Ru is set equal to the load impedance Rs,
then S1 becomes

S .-
_ROI-ROl_
ROl+ROI
0 - (14.44)

so that a 1 = 0 under this condition. Similarly, we can show that, when

a 1 = 0, the reference impedance of port 1 is equal to the terminating

1
From the viewpoint of the load resistor R 1 , the incident parameter is b 1 and the
reflected parameter is a 1 •
422 Network analysis and synthesis
=
impedance {i.e., R 01 RJ. We see as a result of this discussion that the
conditions o1 = 0 and o1 = 0 merely imply that the reference impedances
R01 and Roa are chosen to be equal to the terminating resistors R1and R1,
respectively.
Next, let us consider the relationship between the driving-point imped-
ances at ports 1 and 2 and the reflection coefficients S11 and s.. Let us
denote the driving-point impedances at ports 1 and 2 as

z.= v. {14.45)
1.

From the equation (14.46)

We can write
s11 _ llCV1/.Ji;> - .Jli;111 (14.47)
- UCV1/.JR01) + .JRoJ.111

which reduces easily to S11 == Zi - Roi

Z1 + Roi
I
Ra-RGI
(14.48)

'
Similarly, we have SII== Z1-Ro11 {14.49)
z. +
ROI R1-Ro1
These expressions tell us if we choose the reference impedance at a given
port to- equal the dri~ing-point impedance at that port, the reflection
coefficient of that port will be zero, provided the other port is terminated
in its reference impedance.
Next, let us derive some physically meaningful expressions for the
forward and reverse transmission coefficients S11 and S11• Consider the
definition for S11
S11 == b. I
01 aa-o
(14.50)

As we have just seen, the condition o1 = 0 implies that the reference im-
pedance Roa is set equal to the load impedance R 1, as seen in Fig. 14.7.
If we connect a voltage source V.,1 with source impedance Roi =
Ri,
,----------,
Rot• R1 I ------ I
,---.J\N\r-...llto-1~ 1-.0,2
_ _ _ __

I Two-port
.!!.I network
1 I .___ _ ___. I
L ________ ...J

FIG. 14.7
 The scattering matrix 423
then we- can express "1 as
a -....!'.L
i - 2JR1 (14.51)

Since as = 0, we have the equation

(14.52)

from which we obtain Y, = -JR 1 1 1 (14.53)

JR,
Consequently, b1 = H~. - JR 111 )

v. (14.54)
= JR,
Finally, we can express the forward transmission coefficient as

(14.55)

In similar fashion, we find that when port 1 is terminated in Roi = R1 and

when a voltage source Y.a with source impedance R 1 is connected to port 2,
then

{14.56)

We see that both S11 and S11 have the dimensions of a voltage-ratio transfer
function. Indeed, if Roi = Ro., then S 11 and S 11 are simple voltage ratios.
It is seen that for a passive reciprocal network, S 11 = S 11•
Now let us consider as an example the scattering matrix of then: 1 ratio
ideal transformer in Fig. 14.8a. Recall that for an ideal transformer

1
11 =- -1.
n
(14.57)

Assuming first that Roi~= Ro. = 1, let us find S11 by terminating the
2-2' port in a 1-0 resistor, as shown in Fig. 14.8b. Then

(14.58)
42.f Network analysis and synthesis

1 I1 ~

~t_;
l' Ideal transformer
(a)
2'

~GJ·
1' ,,,,

(cJ

~,C7+
~IL:f
... :· (d)
1
,N.f,

~·~ ··E·
(e)
FIG. 14.8. Determination of scattering parameters for an ideal transformer.

so that Z1 = Y1 = n• (14.59)
Ii
n 1 -1
From Eq. 14.48, we have s11 - --
- n1 + 1
(14.60)

Next we terminate the 1-1' port in a 1-0 resistor (Fii,.14.Sc). We obtain,

as we did for S11, _-0-
1
S = (1/n') - 1 = 1 - n (14.61)
11
(1/n') +1 1 + n1
 The scattering matrix -425
We obtain Sn by connecting a voltage source v.1 with a source impedance
Roi= n
1 at the 1-1' port and terminating the 2-2' port with a resistance
Ra= l 0, as seen in Fig. 14.8cl. Since V1 /11 =
n1 , the equivalent cil'cuit
of the ideal transformer as seen from the voltage source as a 1-0 imped-
ance in series with an nl..ohm resistance (Fig. 14.Se). Then V1 can be
expressed in terms of v.i, as
V, - V,,1nl (14.62)
1
- n1 + 1
Since V1 = V1 /n, we have
V. _ V,,1n (14.63)
1
- n1 +1
Since Roi = Ro. = 1 0, Sn is
Sn_
-
2v._~
- (14.64)
¥,1 n• + 1
We can show in similar fashion that

s11 - ~
- n• + 1
(14.65)

Therefore the scattering matrix for the ideal transformer is given as

S == [:: +
1 - n1
2n
! n.2: 1] (14.66)
-- --
n1 + 1 n1 + 1
As a second example, let us find the scattering matrix for a lossless
transmission line of length L terminated in its characteristic impedance, as
shown in Fig. 14.9. If we assume that Rn= Ro.= Z., then the reflection
coefficients are
Z 0 -Z....0 _ = Sn
Su = _..___ 0 (14.67)
z0 +z0
1 M-----L: 2

1' 2'
FIG. 14.9. 1AJsalcss transmilliou line. .
◄26 Network analysis and synthesis
This result is not implausible, because a transmission line terminated in
its characteristic impedance has zero reflected energy. To determine S1i,
we terminate the line in Z 0 at both ends and connect at the 1-1' port a
voltage source V,,i, as depicted in Fig. 14.9. Since the transmission line
has zero reflected energy, that is,
b1 == a1 == 0
then V.I == V.i,le-yL (14.68)

From the equation S1 1_- 2 (Ro1)¾V1

- - (14.69)
Roa V.1
we obtain S11 == 2e-1L (14.70)
In similar fashion, we find that
S11 == 2e-7L (14.71)
Therefore the scattering matrix for the lossless transmission line is

(14.72)

14.4 PROPERTIES OF THE SCATTERING MATRIX

Having defined the scattering matrix of a two-port network in

Section 14.3, let us consider some important properties of the scattering
matrix. From the general restriction for a passive network that the net
power delivered to all ports must be positive, we obtain the condition
P == ½(a1a1* + a.01* - b1b1* - bJ,1*) ~ 0 (14.73)
Equation 14.73 follows from the fact that the power delivered to the 1-1'
port is
P 1 == ½(a1a1* - b1b1*) (14.74)
and the power delivered to the 2-2' port is
P1 = ½(aaa1* - bab1*) (14.7S)
The total power delivered to the network is then
p ==Pi+ P1 (14.76)

which is exactly the expression in Eq. 14.73. In matrix notation, the power
delivered to the network is
(14.77)
 The scattering matrix 427
where T denotes the transpose operation, and

[a)= [::]
(14.78)

[b] = [::]
Since [b] = [S][a], then
[b*]T = [a*]T[S*]T (14.79)
Equation 14.77 can now be rewritten as

2P = {[a*JT[a] - [a*JT(S*JT[S][a]}
(14.80)
= [a*]T[[u] - [S*]T[S]J[aJ ~ 0
This then implies that the determinant of the matrix [[u] - [S*JT[S]]
must be greater or equal to zero, that is,

Det ([u] - (S*JT[S]] ~ 0 (14.81)

Consider the special but, nevertheless, important case of a lossless
network. In this case P = 0, so that

(14.82)
A matrix satisfying the condition in Eq. (14.82) is unitary. For a lossless
two-port network

[S*JT[S] = [S11* 11
S *] [S11 S
S11* S11• S11 S
1
:J = [1 0]
O 1
(l4.83)
From this equation, we have the following conditions for the scattering
matrix
S11*S11 + S 11 *S11 = 1 (14.84)

S 0 *S11 + S11 *S11 = 0 (14.85)

S11 *S11 + S11 *S11 '= 0 (14.86)

S11*S11 + s..•s. = 1 (14.87)

Note that Eqs. 14.85 and 14.86 are conjugates of each other. If the
network is reciprocal, then S11 =·
S11 and
IS11(/w)l1 + IS11(/w)l1 = 1
(14.88)
IS..(jco)l1 + ISn(/w)l 1 -= I
"428 Network analysis and synthesis
R1

Two-port,
network

FIG. 14.10

from which it follows that for a lossless reciprocal network IS11(iw)I =

ISn(jw)I ~ 1 and IS11(jw)I ~ 1. Also it is clear that when IS11(iw)I = 0
(i.e., when there is a zero of transmission), then IS11(iw)I = 1. This
condition states that all the power that has been delivered to the network
from port 1-1' is reflected back to port 1-1 '.
At this point, it might be profitable to discuss why we use scattering
matrices. What are the advantages of the scattering description over
conventional descriptions? Let us discuss three major reasons for the
scattering formalism.
1. Many networks do not possess an impedance or admittance matrix.
For example, an ideal transformer has no Z or Y matrix because its
elements are not finite. However, as we have seen, the ideal transformer
can be described by a scattering matrix. Carlin states8 that all passive
networks possess scattering matrices.
2. At high frequencies, incident and reflected parameters play dominant
roles in problems of transmission, while voltage-current descriptions are
relegated to the background. Then the scattering matrix is necessarily the
more powerful description of the system. Note that the voltage standing-
wave ratio (VSWR) is given in terms of a reflection coefficient Sas

VSWR = 1 + ISi (14.89)

1- ISi
3. In networks where power flow is a prime consideration (e.g., filters),
the scattering matrix is very useful. For example, in the network given in
Fig. 14.10, if P4 represents the available power from the generator and P 1
is the power dissipated in the load R1 , then we can.show that the magnitude-
squared forward transmission coefficient is

IS11(jw)l 1 = Pi (14.90)
PA
We will discuss this point in more detail in Section 14.5.

• H. J. Carlin, "The Scattering Matrix in Network Theory," Trans. IRE, CT-3,

No. 2, June 19S6, 88-96; see his extensive bibliography.
 The scattering matrix -429

14.S INSERTION LOSS

In Section 14.4, we described the forward and reverse transmission

coefficients in terms of voltage ratios. Perhaps a more appropriate
description of a transmission coefficient is in terms of a power ratio rather
than a voltage ratio. In this section we will show that IS11(jw)l 1 and IS11
(jw)l 1 can be expressed in terms of power ratios. We will then introduce
the very important concept of insertion loss and finally relate IS11(jw)l 1 to
the insertion power ratio.
Consider the equation for S11 in the two-port network: shown in Fig.
14.5,
(14.91)

From the equation IS11(jw)l 1 = S11(jw)S11 *(jw) (14.92)

1
we obtain ISnUw)I• = IV1Uw)l /2R1
IV,1Uw)r1/8R1
P,
=- (14.93)
p.J.l

where P I is the power dissipated by the load R1, and P.J.l is the available
gain of the generator Vg1. Similarly, we have

(14.94)

We see that both IS11(ja,,)l 1 and IS1.(jw)l 1 are power transfer ratios which
relate the power dissipated at a given port to the available power in the
other port.
Now let us examine the idea of insertion loss. Consider the network
shown in Fig. 14.lla. Between the terminals 1-1' and 2-2' we will insert a
two-port network, as shown in Fig. 14.llb. Let us denote by VIII the
voltage across the load resistor R1 before inserting the two-port network,
and by V1 the voltage across R1 after inserting the two-port network. A
measure of- the effect of inserting the two-port network is given by the
insertion voltage ratio IVR, which is defined as

IVR A v.. (14.95)

V.
Another method of gaging the effect of inserting the two-port network
is to measure the power dissipated at the load before and after inserting the
430 Network analysis and synthesis

E :~E- Rt :

(G)
l'
2

2'

R1 2
+
Inserted
V.1 two-port R2 V2
network

1' 2'
(b)
FIG. 14.11

two-port network. If Pao is the power dissipated at the load before the
two-port network is inserted, and if P1 is the power dissipated after inser-
tion, then the insertion power ratio of the two-port network is defined as
Pao le A
e =- (14.96)
Pa
If we take the logarithm of both sides, we obtain

IX = 10 log Pao (14.97)

P1
where IX is the insertion loss of the two-port network. In terms of the
circuit given in Fig. 14.11, we can calculate Pao from the relation

V. - V,,1 R (14.98)
ID - R1 + R1 I

1
p _ IV111l
Then P• is so - 2Ra
1
- R1 IV,11 (14.99)
2(R1 + R.)1
The power dissipated by the load after inserting the two-port network is
given by iv.ii•
Pa = 2R (14.100)
1
 The scattering matrix 431
The insertion power ratio can then be expressed as
2
e
2« Pzo IV,,11
= - = - -2
Rl (14.101)
P,. Iv.1 (R1 + Ra) 1
In the special case when the source and load impedances are equal, that
is,
(14.102)

the reciprocal of the squared magnitude of the forward transmission

coefficient in Eq. 14.93 is equal to the insertion power ratio

1 Pzo
=- (14.103)
When R1 "F' R 1 , then
Pao 4R 1R 2 1
-= (14.104)
P2 (R1 + R,)2 IS21Uro)l 8

In any event, we see that the magnitude-squared transmission coefficients

1
1IS.(jw)l 8 and IS18(jw)l can be regarded physically as equivalent insertion
power ratios. In Section 14.6, we will use this relationship in the synthesis
of double-terminated filter networks.

14.6 DARLINGTON'S INSERTION LOSS FILTER SYNTHESIS

In this section, we will consider a filter synthesis procedure first proposed

by Darlington in a classic paper in 1939.4 We will use scattering matrix
notation to describe the essence of Darlington's original work. Our
coverage will be restricted to the class of low-pass filters which are termi-
nated in equal source and load impedances, Roi= R 08 = R 0 , as shown in
Fig. 14.12. For normalizing purposes we will let Ro be equal to l n.

Ro

Filter
network Ro

FIG, 14.12

' S. Darlington, "Synthesis of Reactance 4-Poles which Produce Prescribed Insertion

Loss Characteristics," J. Math. Phys., 18, 1939, 257-353.
432 Network analysis and synthesis
Recall that when the source and load
impedances are equal, then the insertion
power ratio is equal to the reciprocal of
1Sa1(jro}P1, that is,
PIO 1
(a) (14.105)
Ps = IS21(j6>}l 1
! Or----
Expressed as a loss function, the insertion
power ratio is ·
r§ -3
J db A =10 log PIO
~ P2
0 1 = -10 log IS11CiwW1 db (14.106)
(b)
"'
In circuit design, the specification of an
FIG. 14.13 insertion loss A (Fig. 14.13a) is equivalent
to the specification of the amplitude-
squared transmission coefficient shown in Fig. 14.13b. One of the most
ingenious techniques given in Darlington's synthesis procedure is the re-
duction of insertion loss synthesis to an equivalent L-C driving-point
synthesis problem. This technique can be developed in terms of scattering
parameters. Our initial specification is in terms of IS11(jro)I. For an L-C
two-port network
IS11(jw)l 1 = 1 - 1Su(jro)l1 (14.107)
Next, S 11(s) is obtained from the magnitude-squared function

S11(s}S11( -ls) =1- IS11Uw)l 2 1,...-, (14.108)

Then from the equation S _Z1-Ro (14.109)

u - Z1 + Ro
we obtain the driving-point impedance

Zi(s) = Ro 1 + S11(s) (14.110)

1 - S11(s)
shown in Fig. 14.12. We then synthesize the network from Z 1 (s).
We will restrict our discussion here to low-pass filters given by the
lossless ladder structure terminated at both ports by 1-Q resistors in
Fig. 14.14. These low-pass filters can take the form of a Butterworth or
Chebyshev specification for IS11(jro)l 1 , that is,

(14.111)
 The scattering matrix 433

FIG. 14.14. Canonical form for double-terminated low-pass filters.

or ISn(jw)I• = 1 + E.;C,.8(w) (14.112)

where C,.(w) represents an nth-order Chebyshev polynomial.

Example 14.1. Let us synthesiz.e a low-pass filter for the specification

(14.113)

which represents a third-order Butterworth amplitude characteristic. The load

and source impedances are Ros= Ro1 = 1 n. First we find ISn(jw)l 2 as

1
ISu(jw)l 2 = 1- 1 + w8
a,8
(14.114)
= 1 + a,8
Lettingjw = sin 1Sn(jw)l 1, we obtain
s8
Sn(s)S11( -s) = - 1 _ s8 (14.115)
which factors into
s3{-s8)
Sn(s) Sn( -s) = (l + 29 + 2s• + s8}(l _ 2s + 2s• _ s3) (14.116)

s8
so that S11(s) is (14.117)
Su(S) = s3 + 2s3 + 2s + l
Next, Zi(s) is obtained from the equation

_ 1 + S11(s)
Zi(s) - 1 - S11(s)

2s3+2r+2s+l (14.118)
2r+2s+l
434 Network analysis and synthesis

CTflO l'
FIG. 14.15
2'

We next perform a Cauer ladder expansion for Z 1(s).

2r + 2s + l)i? + 2s' + 2s + l(s

2.s3+2r+s
s + 1)2s' + 2s + 1(2s
2r +2s
l)s + l(s
s
1)1(1
1
=
The low-pass filter is thus synthesi7.ed in the structure shown in Fig. 14.15.
An equivalent realization for the double-terminated filter is obtained if
we use the equation
S11(s) = Yi(s) - Go (14.119)
Y1(s) + G0
Then, assuming G0 = 1 mho
Yi(s) = 1 + Su(s) (14.120)
1 --- S11(s)
The canonical realization for Yi(s) is shown in Fig. 14.16. Tables 14.1,
14.2, and 14.3 list element values (up to n = 1) for double-terminated
Butterworth, Chebyshev (1-db ripple) and Bessel filters, respectively.
These apply to the canonical realization for Y1(s) given in Fig. 14.16. If
a Z 1(s) realization in Fig. 14.14 is desired, we simply replace all shunt
capacitors by series inductors and vice versa.

Ls

V1 C1 Cs C1 10
I V2

J_
FIG, 14.16. Canonical form for filters in Tables 14.1, 14.2, and 14.3.
 The scattering matrix 435
TABLE 14.1
Normalized Element Values for a Double-Terminated
Butterworth FIiter (Equal Terminations)
n C1 La Ca L,. C11 La C1

1 2.000
2 1.414 1.414
3 1.000 2.000 1.000
4 0.765 1.848 1.848 0.765
5 0.618 1.618 2.000 1.618 0.618
6 0.518 1.414 1.932 1.932 1.414 0.518
7 0.445 1.248 1.802 2.000 1.802 1.248 0.445

TABLE 14.2
Normalized Element Values for a Double-Terminated
Chebyshev Filter with I-decibel Ripple (Equal Terminations)
n C1 La Ca L,. Cs Ls C1

1 1.018
3 2.024 0.994 · 2.024
5 2.135 1.091 3.001 1.091 2.135
7 2.167 1.112 3.094 1.174 3.094 1.112 2.167

TABLE 14.3
Normalized Element Values for a Double-Terminated
Bessel Filter (Equal Terminations)
n C1 La Ca L, Co La C1

1 2.000
2 1.577 0.423
3 1.255 0.553 0.192
4 1.060 0.512 0.318 0.110
5 0.930 0.458 0.331 0.209 0.072
6 0.838 0.412 0.316 0.236 0.148 0.051
7 0.768 0.374 0.294 0.238 0.178 0.110 0-038

Note that the even orders for the double-terminated Chebyshev filters
are not given. This is because the even-ordered Chebyshev filters do not
meet realizability conditions for minimum insertion loss at s = 0. 5 We
have only given tables for equal source and load terminations. For other
possible realizations, the reader should consult L. Weinberg's excellent
book. 6 '

• L. Weinberg, Network Analysis and Synthesis, McGraw-Hill Book Company,

New York, 1962, p. 589.
'Ibid., Chapter 13.
436 Network analysis and synthesis

Problems
14.1 Determine the reflection coefficient S for the one-port networks shown
in the figure.

½t lh

(a) (b) (c)

PROB. 14.1

14.2 For the one-port network in Fig. 14.3, let Ro = R11 • If the incident
parameter is a = V11/2 v' Ro, find the reflected parameter b.
14.3 For the network in Prob. 14.1, determine IS(jc.o)I. Show that the
scattering elements S for the networks in Prob. 14.1 are bounded real functions.
14.4 For each of the networks shown, find the scattering matrix for Roi =
Roa= 1.
20

10 10

l'o-------'----~~
(a) (b)

;Q C.;-
Ii I2

2 -.!i=
12 11 aa
Vt
- Gyrator
(c)

I,
I1 I2 + +

l=I
Jt f
: V1 =nV2
v~ Vi V2
_V2 I1 =¼Iz

(d) (e)
PROB. 14.4
 The scattering matrix ◄37

14.5 Find the insertion voltage ratio and insertion power ratios for each of
the networks shown. These networks· are to be inserted between a source
impedance R. = 2 n and a load impedance Rr, = 1 n. From the insertion
power ratio, find IS11(jco)l 1 •

2h
40

0
(cJ

PROB. 14.5

14.6 Synthesi7.e low-pass filters for the specifications

(a) ISu(J'w)l 1 = -1 +1 -co'

(b) IS11(jw)l 1 =1 1 o,8

14.7 Synthesi7.e an equal-ripple low-pass filter such that 20 log IS11(Jco)I has
at most }-db ripple in the pass band and an asymptotic falloff of 12 db/octave in
the stop band.
 chapter 15
Computer techniques in
circuit analysis

IS.I THE USES OF DIGITAL COMPUTERS IN

CIRCUIT ANALYSIS

The advent of the high-speed computer has made routine many of the
formerly tedious and difficult computational aspects of circuit theory.
Digital computers have become widely used in circuit analysis, time and
frequency-domain an:alysis, circuit (filter) design, and optimization or
iterative design. We will discuss these aspects in general in this section.
In succeeding sections, we will discuss some specific circuit-analysis
computer programs.
Circuit analysis
The primary objective of a linear circuit-analysis program is to obtain
responses to prescribed excitation signals. These programs are based on
many different methods: nodal analysis, mesh analysis, topological
formulas, a:nd state variables. Most of them can handle active elements
such as transistors and diode~ by means of equivalent circuit models.
The state-variable programs based upon Bashkow's A matrix formula-
tion1 perform their calculations directly in the time domain via numerical
integration and matrix inversion. The outputs of these programs provide
impulse and step response in tabular form. If the excitation signal were
given in data form, the state-variable programs would calculate the
response directly in the time domain.

1
T. R. Bashkow, "The A Matrix-New Network Description," IRE Trans. on Circuit
Theory, CT-4, No. 3, September 1957, 117-119.
438
 Computer techniques in circuit analysis 439
The majority of circuit-analysis programs, however, perform their
calculations in the frequency domain. The program user is only required
to specify the topology of the network, the element values, and what
transfer functions he wishes to obtain. The computer does the rest. It
calculates the specified transfer functions in polynomial form, calculates
the poles and zeros of these functions, and can also provide transient
response and steady-state response, if desired. With versatile input-
output equipment, the output can also provide a schematic of the original
network, as well as plots of time- and frequency-response characteristics.
Time- and frequency-domain analysis
The time- and frequency-domain analysis programs can be used in
conjunction with the circuit-analysis programs or independently. The
time-domain programs depend upon solving the convolution integral

FIG. IS.I. Frequency response of fifth-order Butterworth filter.

FIG. 15.2. Phase response of fifth-order Butterworth filter.
-440 Network analysis and synthesis

FIG. 15.3. Impulse response of fifth-order Butterworth filter evaluation by Laplace

transform.

numerically. This approach obviates the necessity of finding roots of

high-order polynomials. It has the advantage that the excitation signal
need not be specified analytically, but merely in numerical form.
The frequency-domain programs· usually consist of finding transient
and steady-state responses, given the transfer function in factored or
unfactored form. The program user must specify the numerator and
denominator polynomials of the transfer functions, the types of transient
response he wishes (i.e., impulse or step response), and the types of
steady-state responses he wishes (amplitude, amplitude in decibels, phase,
delay, etc.). In addition, he must specify the frequency and time data
points at which the calculations are to be performed. This may be done
in two ways. If he requires evenly spaced data, he need only specify the
minimum point, the increment, and the number of points. If he wishes
to obtain data at certain points, he must supply the lis~ of data points at
the input.
Examples of outputs of a steady-state and transient analysis computer
program are shown in Figs. IS.I, 15.2, and 15.3. In Fig. 15.1, the magni-
tude of a fifth-order Butterworth filter is plotted via a microfilm plotting
subroutine. Figure 15.2 shows the phase of the filter, while Fig. 15.3
shows its impulse response.
 Computer techniques In circuit analysis -441
In Section 15.2 we will examine further details of a typical steady-state
analysis program.
Circuit (filter) design
The filter design programs are probably the most convincing argument
for the use of computers in circuit design. Designing insertion loss filters•
to meet certain amplitude requirements requires considerable numerical
calculation even in the simplest cases. The use of digital computers in
insertion loss filter design is clearly a logical alternative. The amount of
programming time for a general filter synthesis program is considerable.
However, the ends certainly justify the means when large numbers of
filters must be designed to meet different specifications.
An outstanding example of a digital computer program for filter design
is the one written by Dr. George Szentirmai and his associates. 8 The
program is complete in that it handles the approximation as well as the
synthesis problem. It is capable of dealing with low-, high-, and band-pass
filters with prescribed zeros of transmission (also called attenuation poles
or loss peaks). There are provisions for either equal-ripple or ma,cimally
flat-type pass-band behavior, for arbitrary ratios of load to source im-
pedances, and for predistortion and incidental dissipation.
In addition, Dr. Szentirmai has a modified program that synthesizes
low- and band-pass filters with maximally flat or equal-ripple-type delay
in their pass band, and monotonic or equal-ripple-type loss in the stop
bands.
In the specifications, the designer could specify both the zeros of
transmission (loss peaks) and the network configuration desired.' If
his specifications include neither, the computer is free to pick both con-
figuration and zeros of transmission. The computer's choice is one in
which the inductance values are kept at a minimum. The program was
written so that the same network could be synthesized from both ends.
Finally, the computer prints out the network configuration, its dual,
the normalized element values, and the denormalized· ones. It also
provides information such as amplitude and phase response, as well as
plots of these responses obtained from a microfilm printer.
Figures 15.4, 15.5, 15.6, 15.7, 15.8, and 15.9 show the results of a
band-pass filter synthesis using Dr. Szentirmai's program. Figure 15.4

1 R. Saal and E. Ulbrich, "On the Design of Filters by Synthesis," Trans. IRE on

Circuit Theory, CT-5, December 1958, 287-327.

1 G. S7.Clltirmai, "Theoretical Basis of a Digital Computer Program Package for

Filter Synthesis," Proceedings of the First Allerton Conference on Circuit and System
T1teory, November 1963, University of Illinois.
• Saal and Ulbrich, op. cit.
 442 Network analysis and synthesis
IAIIID ,ass FILTER SYNTHESIS

CASE NUMlfR 3. l
DEGREE IF FILTER • 13
MULTIPLICITY IF PEAK AT ZERI .. 3
MULTIPLICITY IF PEAK AT INFINITY • 2
NUNBER IF FINITE PEAKS BELIW THE BAND• l
NUNBER IF FINITf PEAKS ABIVE THE BAND• 3

EQUAL RIPPLE PASS BAND REQUESTED

PASS BAND RIPPLE MAGNITUDE • 0.05000 DB.
LIWER PASS BAND EDGE FREQUENCY • CPS.
l ■ OOOOOOOE+04
NIO-BAND FREQUENCY • l.34l640BE+04 CPS.
UPPER PASS BAND EDGE FREQUENCY • l.8000000E+04 CPS.
•ARBITRARY• STIP BAND REQUESTED
NUNBER IF SPECIAL PILES• 1 MA• 1.oooooooE+oo
TERNINATIINS
INPUT TERNINATIIN • 6.0000000E+02 IHNS
IUTPUT TERNINATIIN • O. fllHMS
LIWER FINITE STIP BAND PEAKS
FREQUENCY CPS NIRNALIZED VALUE IF M
6.ZOOOOOOE+03 4.62t2011E-Ol 0.6229266 1
UPPER FINITE STIP BAND PEAKS
FREQUENCY CPS NIRMALIZED VALUE IF M
Z.0100000E+04 lo5428869E+OO 2.3188111 1
2. 3100000E+04 lo 1211723E+OO 1.9296560 2
3.5200000E+04 2.6236531E+OO 1.4968156 3

CIMPUTER WILL SPECIFY CINFIGURATIIN

LAST INDUCTIR IS A SERIES BRANCH
FIG. 15.4

gives the specification of the problem. The pass-band magnitude is to be

equal ripple with ripple magnitude of 0.05 db, and the degree of the filter
is to be 13. As we indicated in Chapter 14, odd-degree, equal-ripple
filters are nonrealizable. In this example the designer utilized an ingenious
device-an extra pole-to accomplish the synthesis. The program logic
then provided two extra zeros: one to cancel the extra pole and the other
to provide the odd degree. The extra pole is called a special pole in Fig.
15.4, and is located at s = -1.0.
Further specifications call for the lower band-edge frequency to be 104
. cycles; the upper, 1.8 x 104 cycles; and the midband frequency to be
.J1.8 x 104 = 1.3416408 x 104 cycles. In the stop band, there are to be
zeros of transmission at/= 0 (three),/= oo (two), and four finite zeros
of transmission: one below the pass band and three above the pass band.
The positions of these finite zeros of transmission are chosen by the
designer as indicated by the notation "arbitrary" stop band requested.
 Computer techniques In circuit analysis 443

CASI 1111-• Sol F•u•o •EALIZATIIN , ... A SHIIIT cracurT AOIIITTANCE

DHAH ., •ru1• • u
IIULTIN.IClff IP PIM AT Ual • I
IIULTIPI.ICITY _, PIAK AT INFINlff • Z
IGUAL_RIPPI.I PASS IAND • O.O!MIO DI •UIITRUY• STIP IANO
L-R·PASS-UIID IDGE FaHUINCY • loODDOOOOE•M CPS
UPPIR.PASS-IANII EDIE FHQUENCY • l.lODDODDE•M CPS
IIID-1..., FIIHUlflCY • loJ4l64Dll.04 CPS
CINFIIURATIIN SPKIFIID IY CMPUTER
IOI 501 101 lDO ID2 400 ZOO 100 JOO
NIIIIIALIZID UNNHIIALIZEO
loOODODDDE•OO
So01D4JJllE-0l
.....•.... .
•••• c ••••
6.DOOOO00l•0Z
6.0l0610ZE-09
TERIIINATIIN
103

,.O21n11-01 L C ,.H92116E-OJ PEAK FREQUENCY

lo14ff7691-0l
. S.6576UIE-09 J.5ZD00DOE+D4
z.nZS015E+00
loOOIUIDE•DD
. .
••• L.C •••
C 5o01.?1440E-0I
lol711955E-03 PEAK FIIIQUENCY
!Mil

4o64l91D6E•0O
lo J64l491UOO
.••••c••••. 9.17912121-0I
Z.6912703£-0I
6.2000000009
101

5. J765ZlZIE-01 L C lol261107E-Oll PEAK FREQUENCY

1.11121111-01
. lo5447656E-08 2.0,00000004
lo'62H94E-0l
lo4OM4H•0D
.••••c ••••.
C lo49517:17E-0I
Z.15465141-01
100
102

Zo lll2146E-Ol L C 1.92971371-03 PIH FREQUENCY

loZ441566E+OO
.••••c••••. Zo459144'E-01 z. 3l0000OE+D4
1.ounn1+01
9o57l0UIIE-OI
.••••l••••. Z.00226651-07
•• II 65641E-04
400

200

lo63ll719E+OO
. C ,.21121ooe-01 100
6.29631561-01
SNIIIT
.....•.•...L 4. 41l 41l 6E-0J
S. .IIT TEIIIIINATIIN
300

FIG. 15.5

The terminations are: input == 60 0, output = 0 n which means the

filter terminates in a short circuit. In this example, the filter configuration
is chosen by the computer, and is a minimum inductance configuration.11
In Fig. 15.4 there are, in addition, listings of the finite zeros of trans-
mission (loss peaks), which the designer specified.
Figure 15.5 is a printout of the configuration of the filter as shown by
the dotted lines flanked by the associated element values, both normalized
(left column) and unnormalized (right column). Since there are four
finite zeros of transmission, there must be associated four L-C tank circuits.

'W. Saraga, "Minimum Inductance or Capacitance Filters," Wireless Engineer,

30, July 19S3, 163-17S.
444 Network analysis and synthesis

FIIEIIUENCY VILTAGE RA PHASE REAL l IN IIIAG l IN REAL Y IN IIIAG YIN

IN CYCLES Tit IN 01 IN DEGREES IN I ...S IN IHIIS IN IIILIIHIS IN IIILIIHIS
16000 !10711532 H!l.!1 ♦ 9711 134.691 !16.911 6.2973 -2.6641
16250 5.104319 1280951181 l:S0.664 61. 193 602166 -2.9:,95
16500 5.701644 111.554102 us.en 64.162 602814 -J.2315
16150 s.129113 92.934801 119.931 61.928 6.3129 -3.5756
11000 50150516 72.163152 112.61.7 71.154 6.3445 -4.0061
11250 !1.743140 !I0.659800 104.489 74.696 6.3337 -4.5211
11!100 5.708979 26.219204 95.117 77.919 6.2835 -5.1151
11750 s. 720993 358.194819 150165 79.138 6.3011 -5.8552
11000 5.702947 322.320795 67.628 18.161 6.2148 -,.,on
11250 4oUl556 272.913485 40.032 81.140 4.3533 -9.4H9
11500 -1.119811 225.689589 16.316 111.450 1.2913 -8.1821
18150 -7.753694 195.922981 5.462 139.624 .2797 -7.1517
19000 -14.148566 177.085862 1.725 164.004 .0641 -6.0967
19250 -20.210264 163.74865♦ .541 184.583 .ou8 -5.4176
19500 -26.118735 153.507803 .167 202.605 .0040 -4.9357
19750 -32.100133 145.221029 .049 211.952 .0010 -4.5672
20000 -38.473164 138.296917 .013 234.147 .0002 -4.2708
20250 -♦ 5.857195 132.354561 .002 248.501 .0000 -4.0241
20500 -56.325030 127.166241 .ooo 262.211 .0000 -3.8137
20750 -71.701981 302.572268 .ooo 275.407 .0000 -3.6310
21000 -59.444119 298.458582 .ooo 288.171 .0000 -J.4701
21250 -57.4139 ♦9 294.740796 .ooo 300.592 .0000 -3.3268
21500 -57.587501 291.354706 .ooo 312.698 .0000 -3.1980
211,0 -58.703992 211.250341 .ooo 324.536 .0000 -3.0113
22000 -60.546362 215.388013 .ooo 336.138 .0000 -2.975()
22250 -63.101161 282.735723 .ooo 347.527 .0000 -2.1775
22500 -660602572 210.267304 .ooo 358.726 .0000 -2.7876
22150 -71.167♦60 211.961066 .ooo 369.753 .0000 -2.704!5
23000 -13.40J113 275. 791866 .ooo 310.621 .0000 -2.6273
23250 -10.597811 93.165362 .ooo 391.3♦ 5 .0000 -2.555:,

FIG. 15.6

20

r
0 '
1
-20 J
I \
I
:@ -40 \
B
I!
I

,_
1-60 V
' '
.....
~

> /
-IKI
j i/ r--....
I
-100 "'
' I
.I
)'

r
-120
0.000 0.400 D.800 1.200 1.600 2.000 2.400 2.IKIO 3.200 3.600 4.000
Add l.OOOOOE + 03 to abscissa
Frequency, cycles x 10 4
FIG.15.7
 . Computer techniques In circuit analysis -445

i 4
Sl
)(
3
.IE
:>,' 2
j

-1.__.._..1-...,_............................__.___..._..__.._..1---...,_.....................__.___.___.~
0.000 D.400 0.800 1.200 1.600 2.000 2.400 2.800 3.200 3.600 4.000
Ml l.OOOOOE + 03 to abscissa frequency, cycles X 10 4
FIG.15.8

0.800
•
Jl
0.600

0.400 J 1
iSl I/
0.200 17 ~
X
.§
E
0.000

::,.;- -0.200
1...-""'
' I'\.
~

\. i-
z::- -, 1,,/'

J-o.400 I
-0.600
\
7

-0.800
li
-1.000
0.000 0.400 0.800 1.200 1.600 2.000 2.400 2.800 3.200 3.600 4.000
Ml l.OOOOOE + 03 to abscissa Frequency, cycles x 104
FIG. 15.9
446 Network analysis and synthesis
The peak frequencies are listed under the column entitled "termination,"
and the associated L-C tank circuits are on the same line two columns to
the left.
Figure 15.6 is a listing of a small portion of the frequency response,
which was calculated after the filter had been synthesized. Figure 15.7
is a plot of voltage ratio in decibels versus frequency. Note that 1000
cycles must be added to every frequency value in the abscissa. This
merely reflects an idiosyncrasy of the plot routine.
Figures 15.8 and 15.9 show the real and imaginary parts of the input
admittance of the filter. Note particularly the shapes of these character-
istics. If a number of these band-pass filters were connected in parallel
and if the pass bands of the filters were adjacent but nonoverlapping,
then the input conductance of the filter system would be essentially
constant over the entire pass band, while the input susceptance would
cancel out, as shown in Figs. 15.lOa and b. This then means that the
input admittance of the filter system would be real and could then be
driven by any arbitrary source impedance without fear of reflections.
Filters obtained using computer programs of the type we have just
described are rapidly supplanting conventional hand- and handbook-
designed filters. The filter synthesis programs provide filters that are
orders of magnitude more sophisticated than filters designed by the

_g
E
::,.;-
m
a::

Frequency
(a)

en
0
.c
E
::,.;-
~
ftl Frequency
.
C
'So
.5

(b)

FIG. 15.10. Effects of paralleling several band-pass filters with adjacent pass bands.
 Computer techniques In circuit analysis '447
conventional manner. Moreover, the designs are completed in minutes
rather than days and at a typical cost of twenty dollars rather than two
thousand (not including initial programming costs, of course).
Optimization
Network design cannot always be accomplished by way of analytical
means. Quite often networks are designed through a trial-and-error
process. The network designer begins with a set of specifications. He
then selects a network configuration, and makes an initial guess about
the element values. Next he calculates or measures the desired responses
and compares them with the specifications. If the measured responses
differ by a wide margin from the specified responses, the designer changes
the values of the elements and compares again. He does this a number
of times until (hopefully) the measured responses agree with the specified
responses to within a preset tolerance.
Thi~ process of cut and try can be made to converge, sometimes quite
rapidly, if one uses a method of steepest descent.• To get a rough idea
of the steepest descent method, suppose there are n parameters in the
network. Let us regard each parameter x, as a dimension in an n-
dimensional euclidean space. We define a function /(xi, x1 , x1 , ••• , x,.)
that assigns a functional value to each point of the euclidean space. We
then ask, at point p,, what direction of motion decreases the value of
f(x 1, x1, x3 , ••• , x,.) most rapidly? The function /(x,) may be defined as
a least squared error, or as an absolute error between calculated response
and specified response. The direction of steepest descent is the direction
defined by the gradient
grad/= - x1
of of x + · · · + -of x,.
+ -0:1: (15.1)
0:1:1 1
1
ox,.
Therefore, incremental value of change for each parameter is

l,z, = - Cof- (15.2)

oz,
where C is a constant. After all parameters have been changed by the
incremental value in Eq. 15.2, a new gradient and new incremental values
are obtained. The process continues until the optimum is obtained.
An excellent paper by C. L. Semmelman7 describes a steepest descent

• Charles B. Tompkins, "Methods of Steep Descent,'' in Modern Mathematics for the

Engineer (Edwin F. Beckenbach, ed.), Chapter 18, McGraw-Hill Book Company, New
York, 1956.
7 C. L. Sernmelman, "Experience with a Steepest Descent Computer Program for

Designing Delay Networks,'' IRE International Convention Rec., Part 2, 1962, 206-210.
448 Network analysis and synthesis
Fortran program used for designing delay networks. The specifications
are the delay values R, given at the frequency data points Ji. The program
successively changes the parameters x, so that the squared error
ft

E(x,) =I [R, - T(x,,J,)J"' (15.3)

i-1

is minimized. In :eq. 15.3, T(x,,jj) represents the delay, at frequency Ji,

of the network with parameters x,.
In order to use the program, the network designer must first select the
initial values for the parameters x,. He must also provide the specified
delays R, and the frequency data points Ji. The program provides for 128
match points and 64 parameter values. It is capable of meeting require-
ments simultaneously in the tiine and frequency domains. The designer
is not restricted to equal-ripple approximations or infinite Q requirements.
He is free to impose requirements such as nonuniform dissipation and
range of available element values on the design. For related methods of
optimization, the reader should refer to a tutorial paper by M. R. Aaron. 8
Machine-aided design
The concept of real-time interaction between 'man and computer holds
much promise in the field of network design. Multiple-access computing
systems, such as the project MAC of Massachusetts Institute of Tech-
nology, make cut and try design procedures practicable. The initials
MAC could describe either the term multiple-access computer or the term
machine-aided cognition. In an article describing the MAC computer
system, 9 Professor R. M. Fano states "The notion of machine-aided cog-
nition implies an intimate collaboration between a human user and a
computer in a real-time dialogue on the solution of a problem, in which
the two parties contribute their best capabilities."
A simple multiple-access computer system is shown in Fig. 15.11.
There are n data links and n terminals connected to the central processor.
Located at each terminal is input-output equipment, such as teletype-
writers, teletypes, and oscillographic displays. The sequence in which
the central processor accepts programs from the terminals is controlled by
a built-in queueing logic.
The main reason a multiple-access computer is effective is that the
central processor computes thousands of times faster than the user's
reaction time. When a user feeds in a program, it seems only a "moment"

• M. R. Aaron, "The Use of Least Squares in System Design," IRE Trans. on Circuit
Theory, CT-3, No. 4, December 1956, 224-231.
• R. M. Fano, "The MAC System: The Computer Utility Approach," IEEE
Spectrum, 2, No. 1, January 1965, 56-64.
 Computer techniques In circuit analysis #9

Memory Data Remote

bank terminal
link 1

Buffer Data Remote

memory terminal
link 2
Memory Central
bank processor

Queueing Data Remote

logic terminal
link 3

Memory
bank

Memory
bank
Data Remote
terminal
link n

FIG. IS.I I. Block diagram showing typical multiple-access computing system.

before he gets the results through the teletypewriter or oscilloscope

display. He examines the results, changes some parameters, and feeds the
program into the central processor again. The queueing time and proc-
essing time on the multiple-access computer may amount to only a
minute or two, which, for the user, is probably not signifi.cantly long.
Dr. H. C. So has written a paper on a hybrid description of a linear
n-port network10 in which he has shown that the hybrid matrix is ideally
suited for such problems such as multielement variation studies and
iterative design. Suppose there are n variable elements in the network.
These can be "extracted" and the rest-the bulk of the network-can be
described by the n-port hybrid matrix.
10 H. C. So, "On the Hybrid Description of a Linear n-Port Resulting from the

Extraction of Arbitrarily Specified Elements," Trans. IRE on Circuit Theory, CT-12,

No. 3, September 1965, pp. 381-387.
-450 Network analysis and synthesis

Command Initialize

Frequency
range

Engineer Control Computer

Parameter
adjust

Display

FIG. 15.12. Iterative design of network using machine-aided cognition.

Dr. So has written a computer program to formulate the hybrid matrix

for the n-port network automatically. 11 The inputs to this program are
(l) the node connections specifying the network topology; (2) the im-
pedance functions of the elements; and (3) the specifications of the
special elements to be extracted. This program was written with man-
machine interaction in mind. The process is described in Fig. 15.12.
First the computer reads in the n-port program with initial specifications.
It performs the calculations and feeds information, such as transient or
steady-state responses, back to the engineer via a visual display console.
The engineer assesses the data and then changes the parameter values of
the extracted elements, and, in some instances, the frequency range over
which the calculations are made. The process is repeated a number of
times until the engineer has obtained the desired results. Such a program
could also be controlled by a steepest descent steering program if the
engineer wished to obtain his results with less ..eyeballing."
Now let us examine some details of specific network analysis programs.

15.2 AMPLITUDE AND PHASE SUBROUTINE

Purpose
Our purpose is to compute the amplitude and phase of a rational
function
H(s) = Cn E(s) (15.4)
Cd F(s)
11
H. C. So, unpublished memorandum.
 Computer techniques In circuit analysis 45 I
over a set of frequencies w;, In Eq. 15.4, Cn and C11 are real constants,
and E(s) and F(s) are polynomials ins= <1 + jw. The program described
here computes the amplitudes M(wk) and phase ¢,(wk) over a set of
frequencies wk, where
jw
(15.5)
Wi
and
z j{J
E 1:wk) F-(jwk)
¢,(wk) = arctan -"-'-- - arctan -•-- I
El,iwi) Fr(jwk) I
I
I
(15.6) I
I
where in Eq. 15.6 the r and i subscripts
-IT a 0
indicate real and imaginary parts,
respectively. Note that in Eq. 15.6,
¢,(wk) is limited to the range -1r ~
cf, ~ 1r radians. In order to circumvent
this restriction on the phase, we use the
-jw
following method to compute amplitude
FIG. 15.13. Calculation of magni-
and phase. tude and phase at w, due to zero z.
Method used
Let H(s) be factored into poles and zeros such that

H(s) = CnCs - z1)(s - z1) • • • (s - zn)

(15.7)
Cis - Pi)(s - p 1) · • • (s - p.,.)

Consider the amplitude and phase due to any pole or zero, for example,
z = -ot + j{J, shown in Fig. 15.13. The magnitude is

M.(w;) = [ot2 + (w, - p)•]l4 (15.8)

and the phase is c/>.(w;) = arctan ( W;: P) (15.9)

The amplitude and phase of the overall function is

ft

C,.JIM•.,.(w,)
M(w,) = ~k~=~1_ __
ffl
(15.10)
C11 JI M,,1(w,)
l=l

and tf,(w,) =I "

c/,•.,.(w;) - I
ffl
c/,,,,(w,) (15.11)
k=l 1-1
452 Network analysis and synthesis
where the subscripts p and z indicate the contribution due to a pole and
zero, respectively.
Input
l. The numerator E(s) and denominator F(s) can be read in either as
polynomials (high degree to low) or in terms of their roots. If numerator
and/or denominator are read in as polynomials, their roots will be printed.
If they are given in terms of their roots, the program could generate
polynomials from these roots.
2. The data points can be read in if unequally spaced; if equally spaced,
only the minimum point and the spacing need be read in.

Read Nl, ND, NN

KK, Ll, L2, L3

Read CN, CD

KK=O KK= 1

Read W(I) Read WMIN, OW

Ll =O L1 = 1

Read A(I), B(I) Read E(I)

L2=0 L2 = 1

Read C (I), D (I) Read F(I)

To main program
FIG. 15.14. Flow chart of input instructions for magnitude and phase program.
 •
Computer techniques In circuit analysis -453
A flow chart listing the input instructions is given in Fig. 15.14. Def-
initions of the symbols. in the flow chart follow.

Nl = number of frequency points

ND = degree of denominator
NN = degree of numerator
KK = 0 indicates equally spaced data points. We then read in only
"°KIN, dro, and number of points.
= 1 indicates unequally spaced data points. We then read in all
co,.
Ll =0Read in zeros (if complex conjugate, both zeros must be
read in).
= 1 Read in numerator polynomial, high degree to low.
L2 = 0 Read in poles (if complex conjugate, both poles must be
read in).
= 1 Read in denominator polynomial, high degree to low.
CN = numerator multiplier
CD = denominator multiplier
Output
The main output of the program is the frequency response consisting
of the following columns: frequency in radians W(I); amplitude,
M(W(I)); magnitude in decibels, 20 log10 M; and phase in degrees,
#_W(I)). A typical printout of an amplitude and phase program is given
in Fig. 15.6, which shows the magnitude and phase of the band-pass filter
designed using Szentirmai's program.

15.3 A FORTRAN PROGRAM FOR THE ANALYSIS OF

LADDER NETWORKS

In this section we will discuss the methods and organization for a

Fortran computer program for the analysis of linear ladder networks.
The analysis proceeds by computing the voltage and current at the input
of successive L sections, beginning with the terminating section.
The program calculates the branch impedances of the ladder by com-
bining R, L, C series impedance arms according to the instructions of the
individual programmer.
The. poles and zeros and the frequency responses of the input im-
pedance and voltage transfer ratio at the input of each L section are
found. In addition, the program provides for the analysis of short- and
open-circuited networks as well as for those with normal terminations.
Separate problems may be run consecutively if so desired.
454 Network analysis and synthesis
-

FIG. IS.IS

The network is initially decomposed into separate L sections as in

Fig. 15.15. These L sections are then added successively to the terminating
L section to form the complete network. With the addition of an L
section, the voltage and current at the input of the resulting network are
computed by the equations

I,. = V,._1 Y,. + l,._1

(15.12)
V,. = I,.Z,. + V,._1
which were originally discussed in Chapter 9 of this book. Thus the
program proceeds toward the front of the ladder requiring only the branch
immittances of the present L section and the voltage and current of the
previous L section to make its calculations.
Suitable assumptions for the initial voltage and current V0 and / 0 allow
for analysis of short- and open-circuited networks as well as for those
with normal terminations. These initial values of voltage and current are
determined by control instructions.
For the calculation of the voltage and current at an L section, the
branch impedances of the section are required. To simplify the prepara-
tion of the input data specifying these impedances, the following procedure
is used. Each branch impedance is formed by the addition in series or
parallel of basic R, L, C impedance arms. A basic R, L, C impedance
arm is a series combination of a resistance, an inductance, and a capac-
itor-any or all of which may be absent. The impedance arms are speci-
fied by the element values R, L, C and an instruction indicating how the
arm is to be combined.
The elements of the impedance arms are frequency and impedance
normalized to aid computation, and then the arms are combined as
specified to form the appropriate branch impedance.
The roots of the voltage and .current polynomials for each L section
are computed by a root-finding routine.
 Computer techniques In circuit analysis '455
For the normal load and open circuit termination, the frequency
responses of the input impedance, and voltage-transfer ratio are computed;
for short-circuit termination, the input impedance and current gain are
computed. Provision is made for either a linear or logarithmic increment
in frequency. The frequency boundaries and increments are read by the
program along with the input data.
According to instructions given by the programmer, the various cal-
culations are printed for each L section or only for the network as a whole.

Pro1ram operation
The ladder network-analysis program requires two sets of input data.
The first set consists of the control instructions. These tell the computer
which of the various options, such as information concerning polynomial
roots or frequency responses, are to be exercised.
In addition, the control instructions provide the computer with necessary
parameters such as normalizing factors, and the number of L sections in
the ladder. The second set of data determines the branch impedances of
the network by specifying the R, L, C impedance arms and their combining
instructions.
Impedance data
The data specifying the branch impedances consist of an ordered series
of instruction cards, each determining an impedance arm and/or a com-
bining instruction. The first arm of any branch impedance requires no
combining instruction. The branch impedance is set equal to the im-
pedance of this arm. Subsequent impedance arms are added in parallel
or series with the existing branch impedance, according to the combining
instructions of the arms. A blank card tells the computer that a complete
branch impedance has been formed and that the next arm begins a new
branch impedance. If blank cards are not properly inserted, the computer
will calculate a single giant branch impedance.
The order of the data determining the branch impedances is Z1 , 1/ Yi,
Z 1 , l/Y1 , ••••
The values of R, L, C and the combining instruction XIN are written
on one card in the order XIN, R, L, C.
The values of XIN and the associated operation are shown in the
table at the bottom of page 456.
Although the instructions XIN = 1.0, 2.0 will suffice for the con-
struction of many branch impedances, they are inadequate for other
impedances.
For example, suppose the series combination of two tank circuits (Fig.
15.16) is desired as a branch impedance, this impedance may only be
456 Network analysis and synthesis

2mh .05h

Tank circuit A Tank circuit B

XIN R L C

1.0 5.00000£ - 11
2.0 2.00000E - 03
3.0 .05
2.0 .000001
4.0
Blank Card

FIG. 15.16. Construction of branch impedances using instructions for XIN.

constructed with use of three more instructions (XIN =

3.0, 4.0, and 5.0)
and an additional set of computer storage locations.
Tank circuit A is formed in the usual manner, but must then be tem-
porarily stored during the calculation of tank circuit B. The two tank
circuits are then added to form the final branch impedance. Examples of
impedances that cannot be formed are shown in Fig. 15.17. Observe that
a series branch impedance of zero ohms often simplifies the formation of
branch impedances. Two blank cards will indicate a short-circuited
impedance.

XIN Operation

Blank card A complete branch impedance has been calculated, and

the next arm begins a new branch impedance.
1.0 This arm is added in series with the existing branch
impedance.
2.0 This arm is paralleled with the existing branch impedance.
3.0 This arm begins a new internal branch impedance.
Subsequent XIN = 1.0 and 2.0 refer to this new in-
ternal branch.
4.0 Add the two internal branches in series to form the final
branch i111~
s.o Parallel the internal branches.
 Computer techniques In circuit analysis ◄57

FIG. 15.17. Branch impedances that cannot be formed by program.

Output
The coefficients of the voltage and current polynomials are printed in
frequency normalized form. To calculate the denormalized coefficients,
divide by the normalizing frequency raised to the power of the corre-
sponding exponent.
The poles and zeros of the voltage and current polynomials are in
frequency normalized form to facilitate zero-pole plots.
Frequency responses appear in denormalized form. The frequency
variable is in radians per second. The impedance level is in decibels,
the phase in degrees. The gain and phase of the transfer ratios are
expressed similarly.

15.4 PROGRAMS THAT AID IN DARLINGTON

FILTER SYNTHESIS

Two double precision Fortran programs have been written that help
in the synthesis of double-terminated filters using the Darlington pro-
cedure. The mathematics of the programs is described here.
Given a forward transmission coefficient

S (s) _ KN(s) (15.13)

a1 D(s)
-458 Network analysis and synthesis
we generate an input reflection coefficient S11(s) from the equation

S11(s) S11(-s) =1- S11(s) S11(-s)

D(s) D(-s) - K 2 N(s) N(-s) (15.14)
= D(s) D(-s)
The denominator of S11(s) is D(s) because all the poles of S 11(s) must be
in the left-half plane. The zeros of S 11(s) are not so restricted. If S11(s)
S 11(-s) has zeros at -a ± jb and a ± jb, the zeros of S11(s) can be chosen
as either -a ±jb or a ±jb. The only difference is that certain choices
lead to filters with unity-coupled coils and others without. 11
Once the zeros of S 11(s) are chosen, the input impedance of the filter
is then
Zi(s) = 1 - S11(s) (15.15)
1 + S (s)11

Read CN Numerator multiplier

Read CD Denominator multiplier

Number of zeros
Read LN (complex conjugate zeros
----r---' count as one zero)

Read Read real and imaginary

parts of zeros and
A (I), B (I), N (I) multiplicity, N (I)

Number of poles
Read LD (conjugate poles
.___---,._ __. count as one pole)

Read Read real and imaginary

parts of poles and
H (I), G (I), M (I) multiplicity, P(I)

To program

FIG. 15.18. Input to S 11(s) S 11(-s) program.

11
T. Fujisawa, "Realizability T~rem for Mid-Series or Mid-Shunt Low Pass
Ladders without Mutual Induction," Trans. IRE on Circuit Theory, CT-2, December
1955, 320-325.
 Computer techniques in circuit analysis 459
Program descriptions
We next proceed with brief descriptions of the two programs. The
first program finds the roots of S 11(s) S1i(-s) given the roots of N(s),
N(-s), D(s), and D<.. -s). A flow chart for the input of the program is
given in Fig. 15.18. Note that we must read in both the zeros of N(s) and
N(-s), although only half of a complex conjugate pair need be read in.
The same applies for the roots of D(s) and D( -s). The constant CN is
K 1 in Eq. 15.14.
The second program performs the computations in Eq. 15.15 for
different combinations of zeros of S11(s), given the fact that the zeros of
S 11(s) need not be in the left-hand plane. The previous program finds all
the zeros of S 11(s) S 11(-s). We must choose only half the number of
zero pairs for S11(s). Certain combinations of zeros of S11(s) S11(-s)
lead to filters with coupled coils; others do not. If one wishes to try a
number of combinations of zeros for S 11(s), the program has the facility
to enable him to do so. He need only supply those zeros of S 11(s) S 11(-s)
in the right-hand plane, C(l) + jB(I), and a list of constants S(I), which
are either 1.0 or - 1.0 so that the zeros of S11(s) may be represented as
S(I) · C(I) + jB(I) == A(I) + jB(I). The routine forms numerator and
denominator polynomials of S11(s)
S11(s) == E(s) (15.16)
F(s)

Number of pole pairs of Su

Read LD
(real poles count as one pair)

Read Real and imaginary parts of poles

H(I), G(I), M(I) of Su and multiplicity, M (I)

Number of right-hand plane

Read LN
zeros of 811(1)811(-,)

Read Real and imaginary parts of right-hand

C (I), G (I), N (I) plane zeros and multiplicity, N (I)

Read S(I) Multiplicative constants ( :!: 1.0) (each

combination of zeros has LN cards)

To program
FIG. IS.It. Input to Z1n(s) program.
'460 Network analysis and synthesis
and then computes the numerator and denominator of Zi(s) as
_ 1 - S 11 _ F(s) - E(s)
Z1 ( S ) - - (15.17)
1 + S11 F(s) + E(s)
The input instructions are summarized on the fl.ow chart in Fig. 15.19.

Problems
15.1 Organize a flow-chart for computing magnitude and phase given only
the unfactored numerator and denominator polynomials. Do not use a root-
finding subroutine.
15.2 Write a program to calculate the delay of
H() = 3(s + 1)
s s2+2s+s
at the points w = O, 1, 2, ... , 10.
15.3 Repeat Prob. 15.2 for a general transfer function given in terms of its
unfactored numerator and denominator polynomials. The frequency points
are to be read in by the computer.
w,
15.4 Suppose you have calculated the phase of a transfer function at points
w = 0, 1, 2, ... , SO. Devise an algorithm to test for phase linearity. The
deviation from phase linearity is to be called phase runoff.
15.5 Write a program to analyze a two-mesh network made up of only
R, L, and C elements.
15.6 Repeat Prob. 15.S for nodal analysis.
15.7 Write a program to calculate the step and impulse response of a linear
system whose system function contains only simple, real poles and zeros.
15.8 Repeat Prob. 15.7 if simple complex poles and zeros are allowed.
15.9 Write a program to calculate the residues of a transfer function with
multiple as well as simple poles. The poles and zeros are to be real.
 appendix A
Introduction to matrix algebra

A.I FUNDAMENTAL OPERATIONS

Matrix notation is merely a shorthand method of algebraic symbolism

that enables one to carry out the algebraic operations more quickly.
The theory of matrices originated primarily from the need (1) to solve
simultaneous linear equations and (2) to have a compact notation for
linear transformations from one set of variables to another.
As an example of (2), consider the set of simultaneous linear equations
a11X1 + a1a:z:1 + ··· + a1,.x,. = Y1
a 11x 1 + a.,ri:1 + ··· + a 1 ,.x,. = y1 • (A.I)

a,,.1:z:1 + a,,.,ri:1 + · · · + a,,.,.x,. = y,,.

These may express a general linear transformation from the x, to the 1h·
In general, m ;;iii n. An example where the numbers of variables in the
two sets are unequal is that of representing a three-dimensional object in
two dimensions (in a perspective drawing). Here, m = 2 and n = 3.
Definition
A matrix is an ordered rectangular array of numbers, generally the
coefficients of a linear transformation. The matrix is defined by giving
all its elements, and the location of each.
The matrix of the equations in Eq. A.I written as
a 11 alll
... a1ft ]
a 11 a11 .•. aa,.
[
a,,.1 a,,. 8 a,,.,.
is of order m x n. (The first number here is the number of rows; the
-461
462 Network analysis and synthesis
second is the number of columns.) The matrix may be denoted by a
single capital letter A or by [a"].
A matrix is a single complete entity, like a position in chess. Two
matrices A and B are equal only if all corresponding elements are the
same: a" = b" for all i and j. A matrix may consist of a single row or

: :: a1n]
aan [ Xs = Ya
.. .
amn Xn
X1
...
Ym
l [l
single column. The complete matrix notation applied to Eq. A.I is

Y1
(A.2)

which says, if put crudely, that A operates on xi to yield y,. This

emphasizes the similarity between a matrix and a transformation. The
:x- and y-matrices are column matrices.
Row and column matrices are called vectors (specifically, row vectors
and column vectors) and their similarity to the more usual type of vector
is discussed later. Here, vectors will be written as small letters, such as
:x or y. Note that the elements of vectors need only one subscript, while
elements of matrices need two.

A.2 ELEMENTARY CONCEPTS

Square matrix
A square matrix has the same number of rows as columns (i.e., a
matrix of order nn). The Y matrix is an example of a square matrix

Y = [Yn Y11] (A.3)

Y11 Y11
Diagonal matrix
A diagonal matrix is a square matrix whose elements off the main
diagonal are zero (i.e., one in which ex" = 0 for i ¥: j). The following
matrix is diagonal.

(A.4)

Unit matrix
A unit matrix is a diagonal matrix for which a" == 1 for i = j, and is
denoted as U. For example, U is

u-[: !~ (A.~
 Appendix A -463

Equality
Two matrices are equivalent if they have the same number of rows and
columns and if the elements of corresponding orientation are equal
Suppose Y1 = 2, y1 = -3, and Ya = -6. If we write this set of equations
in matrix form, we have

(A.6)

Transpose
The transpose of a matrix A denoted as AT, is the matrix formed by
interchangin.g the rows and columns of A. Thus, if we have

A=[~: :] (A.7)
· 3 2

then AT= [1
0
-65 3]
2
(A.8)

Determinant of a matrix
The determinant of a matrix is defined only for square matrices and is
formed by taking the determinant of the elements of the matrix. For
example, we have

det [ 1
-5 4
2] = I-51 214 = 14 (A.9)

Note that the determinant of a matrix has a particular value, whereas the
matrix itself is merely an array of quantities.
Cofactor
The cofactor A# of a square matrix is the determinant formed by
deleting the ith row and jth column, and multiplying by (-1)'+1, For
example, the cofactor A 11 of the matrix

(A.IO)

is A11 = (-1)1+1 X 6 = -6 (A.11)

Adjoint matrix
The adjoint matrix of a square matrix A is formed by replacing each
clement of A by its cofactor and transposing. For example, for the
 -464 Network analysis and synthesis
matrix in Eq. A.10, we have

adjA = [_: -:JT

(A.12)

= [_; -:J
Sin1ular and nonsin1ular matrices
A singular matrix is a square matrix A for which det A = 0. A non-
singular matrix is one for which det A "F 0.

A.3 OPERATIONS ON MATRICES

Addition
Two matrices may be added if both matrices are of the same order.
Each element of the first matrix is added to the element of the second
matrix, whose row and column orientation is the same. An example of
matrix addition is shown in Eq. A.13.

[-2 4] + [ 6] = [
3 0 -7
3
-2
1
-4
10]
-2
(A.13)

Thus if matrices A, B, C, ... , K. are all of the same order, then

(A.14) ·

The associative and commutative laws apply.

Associative: A + (B + C) = (A + B) + C
(A.15)
Commutative: A +B =B +A
Multiplication by a scalar
We define multiplication by a scalar as

AA = A[ail] = [AD"] (A.16)

Thus to multiply a matrix by a scalar, multiply each of its elements by

the scalar.
· Example A.I

3[-: :]-[-: :]
-3 2 -9 6
(A.17)
 Appendix A -465
Linear combination of matrices
If the rules of addition and multiplication by a scalar are combined,
for two matrices of the same order we have

(A.18)
where at and {J are scalars.
Multiplication
In order for matrix multiplication AB to be possible, the number of
columns of the first matrix A must equal the number of rows of the second
matrix B. The product C will have the number of rows of the first and
the number of columns of the second matrix. In other words, if A has
m rows and n columns, and B has n rows and p columns, then the product
C will have m rows and p columns. The individual elements of C are
given by
(A.19)
Example A.2

2 -1
(A.20)
[ -1 0

Example A.3. The system of equations

Zu_/1+ zufa = V1 (A.21)
;i/1 + Znl1 = v.

can be written in matrix notation as

(A22)

We see that systems of equations can be very conveniently written in

matrix notation.
Matrix multiplication is not generally commutative, that is,
(A.23)
Observe that the product BA is not defined unless p == m. Even a product
of square matrices is generally not commutative, as may be seen in the
following example.

[ I OJ[-1 OJ [-1 OJ
-1 2 0 2
==
1 4
(A.24)
'466 Network analysis and synthesis
If we interchange the order of multiplication, we obtain

[-1 OJ[ 1 OJ= [-1 OJ

0 2 -1 2 -2 4
(A.25)

Because of the noncommutative nature of matrix multiplication, we

must distinguish between premultiplication and postmultiplication. In
BA, A is premultiplied by B; B is postmultiplied by A. For matrix
multiplication, the associative and distributive laws apply.
Associative: A(BC) = (AB)C = ABC
(A.26)
Distributive: A(B + C) = AB + AC
Transpose of a product
The transpose of a product AB is equal to the product, in reverse order,
of the transpose of the individual matrices A and B, that is,

(AB)= BTAT (A.27)

and (ABC)T = CT(AB)T = CTBTAT (A.28)

The product ,i:Tx, if xis a column vector, is a scalar number equal to
the sum of the squares of the elements of x. Thus we have

··•= [~•·· ... ·•-{:.] (A.29)

= :x:l + :x:.I + ... + :x:,.I

The product uT is a square matrix C such that C = CT.
Common expressions in simple notation
(a) The sum of products a 1b1 + a.h1 + · · · + a,.b,., which are the
typical element of a product matrix, may be written as a7b or bTa where
a and b are column vectors.
(b) The sum of squares :x:11 + :x:11 + · · · + :x:,.• is thus xTx (x, column
vector). It is also uT if x is a row vector.
(c) The expression a 11:x:11 + a 1s:i:11 + · · · + a,.,.:x:,.• is xTAx (x, column
vector), where A is a diagonal matrix with elements a11• Similarly, the
expression a11:x:i111 + a..:x:1111 + · · · + a,.,.:x:,JJ,. is xTAy or yTAx.
(d) An expression such as
 Appendix A -467
is a quadratic form. This is

(A.30)

+ · · · + a nnx n1 =• xTAx
Inverse
Division is not defined in matrix algebra. The analogous operation is
that of obtaining the inverse of a square matrix. The inverse A- 1 of a
matrix A is defined by the relation

A-1A = AA-1 =U (A.31)

To obtain A-1, we first obtain the adjoint of A, adj A. Then we obtain

the determinant of A. The inverse A-1 is equal to adj A divided by IAI,
that is,
1
A- 1 = - adj A (A.32)
IAI
Example A.4. Let A be given as

A= [ _; :] (A.33)

Its determinant is IAI == 3 (A.34)

and the cofactors are

A 11 = 1
(A.JS)
A11 = -1
The adjoint matrix is
l 17T
adj A= [ -1 2
J
(A.36)

= G -:J
so that A-1 is A-l=!G -:]
(A.37)

= [! -:]
""6.8 Network analysis and synthesis
As a check we see that

[ll -lJi [ lJ = [lo OJ

2
-1 1 1
(A.38)
2
[ -1
lJ [ll . -lJi = [lo OJ
1 1
If the determinant of the matrix is uro, then the inverse is not defined. In
other words, ~nly nonsingular square matrices have inverses.

A.4 SOLUTIONS OF LINEAR EQUATIONS

Consider a set of linear algebraic equations, to be solved simultaneously.
011X1 + 01ra + • •• + a1,.x,. = h1
(A.39)

a111x 1 + a,..z8 + · · · + a ,.x,. = h,.

11

The h, are constants. It is desired to solve for the x 1• In matrix notation

we have
(A.40)
Premultiplying by A- 1 gives (A.41)
In expanded form, this is

[j ! [ Aj [h~
Au A11 1

Xa = A1a A11 Aas ha (A.42)

1 1
A1a A13 A · h

Thus

h1 a11 a11
ha On 013

=ha
- -aa2
-- a13
-
IAI
and similarly for x1 and x8 • This is the familiar Cramer's rule for solving
such equations.
Example A.S. Solve for x, y, z.
x-y+z=2
2x + y = l (A.43)
-x + 3y +z == -1
 Appendix A -469
In matrix form, these equations are written as Ax = h where

(A.44)

Also, IAI =- 10. Thus, we have x = A-1h

(A.45)

Therefore z = 0.7, y = -0.4, z = 0.9

A.S REFERENCES ON MATRIX ALGEBRA

The following is a short list of books on matrices that the reader might wish to
examine.
A. C. Aitken, Determinants and Matrices, 9th Ed., Interscience Publishers, New York,
1956.
R. Bellman, Introduction to Matrix Analysis, McGraw-Hill Book Company, New
York, 1960.
R. L. Eisenman, Matrix Vector Analysis, McGraw-Hill Book Company, New York,
1963.
D. K. Faddeev and V. N. Faddeeva, Computational Methods of Linear Algebra, W. H. ·
Freeman and Company, San Francisco, 1963.
F. R. Gantmacher, Applications of the Theory of Matrices, Interscience Publishers,
New York, 1959.
F. E. Hohn, Elementary Matrix Algebra, The Macmillan Company, New York, 1964.
L. P. Huelsman, Circuits, Matrices, and Linear Vector Spaces, McGraw-Hill Book
Company, New York, 1963.
P. LeCorbeiller, Matrix Analysis of Electric Networks, John Wiley and Sons, New York,
1950.
M. Marcus and H. Mine, Survey of Matrix Theory and Matrix Inequalities, Allyn and
Bacon, Boston, 1964.
E. D. Nering, Linear Algebra and Matrix Theory, John Wiley and Sons, New York,
1964.
S. Perlis, Theory of Matrices, Addison-Wesley, Reading, Massachusetts, 1952.
L. A. Pipes, Matrix Methods for Engineering, Prentice-Hall, Englewood Cliffs, NJ.,
1963.
A. M. Tropper, Matrix Theory for Electrical Engineering Students, Harrop, Loadon,
1962.
A. von Weiss, Matrix Analysis for Electrical Engineers, D. Van Nostrand, Princeton,
NJ., 1964.
 appendix B
Generalized functions and
the unit impulse

8.1 GENERALIZED FUNCTIONS

The unit impulse, or delta function, is a mathematical anomaly. P.A.

M. Dirac, the physicist, first used it in his writings on quantum mechanics. 1
He defined the delta function d(x) by the equations

L: d(x) dx =1
(B.1)
d(x) =0 for x '¢ 0

Its most important property is

L: f(x) 6(x) dx = /(0) (B.2)

where /(x) is continuous at x = 0. Dirac called the delta function an

improper function, because there existed no rigorous mathematical justi-
fication for it at the time. In 1950 Laurent Schwartz• published a treatise
entitltd The Theory <ffDistributions, which provided, among other things,
a fully rigorous and satisfactory basis for the delta function. Distribution
theory, however, proved too abstract for applied mathematics and

1
P. A. M. Dirac, The Principles of Qlltllltum Mechanics, Oxford Univenity Press,
1930.
1 L. Schwartz, Theorie des Distributions, Vols. I and II, Hermann et Cie, Paris, 1950

and 1951.
470
 Appendix B 471
physicists. It was not until 1953, when George Temple produced a more
elementary (although no less rigorous) theory through the use of general-
ized functions,• that this new branch of analysis received the attention it
deserved. Our treatment of generalized functions will be limited to the
definition of the generalized step function and its derivative, the unit
impulse. The treatment of these functions follows closely the work of
Temple' and Lighthill.5
To get an idea of what a generalized function is, it is convenient to use
as an analogy the notion of an irrational number rL beng a sequence {rL,.}
of rational numbers rL,. such that

rL = limrL,.
n ➔ oo

where the limit indicates that the points rL,. on the real line converge to
the point representing rL. All arithmetic operations performed on the
irrational number rL are actually performed on the sequence {rL,.} defining
rL. 8 We can also think of a generalized function as being a sequence of
functions, which when multiplied by a test function and integrated over
( - oo, oo) yields a finite limit. Before we formally define a generalized
function, it is important to consider the definition of (1) a testing function
and (2) a regular sequence.
DEFINITION B.I A function <f,(t) of class C[<f,(t) € CJ is one that (1)
is differentiable everywhere, any number of times and that (2) when it or
any of its derivatives are multiplied by I raised to any power, the limit is

lim [tm cf,1t 1(t)]-. 0 for all m & k ~ 0 (B.3)

t ➔ ±oo

Any testing function is a function of class C.

Example 8.1. The Gaussian function e-1•1n• is a function of class C. It is
obvious that if a function is of class C then all of its derivatives belong to class C.

• G. Temple, "Theories and Applications of Generalized Functions," J. London

Math. Soc., 28, 1953, 134-148.
'G. Temple, "lbe Theory of Generalaed Functions," Proc. Royal Society, A, 228,
1955, 175-190.
11
M. J. Lighthill, Fourier Analysis and Generalized Functions, Cambridge University
Press, 1955. Lighthill dedicated his excellent book to "Paul Dirac, who saw it must be
true, Laurent Schwartz, who proved it, and George Temple, who showed how simple
It could be made."
• This defines an irrational number according to the Cantor definition. For a more
detailed account see any text on real variables such as E. W. Hobson, The Theory of
Functions of a Real Variable, Vol. I, third edition, Chapter 1, Cambridge University
Press, Cambridge, England, 1927.
4n Network analysis and synthesis
DEFINITION B.2 A sequence {/,.(t)} of functions of class C is said
to be regular if for any function tp{t) belonging to C, the limit

!~°!u... <1>> = !~ L: 1.. <1> 41..t> dt (B.4)

exists. Note that it is not necessary that the sequence converge pointwise.
1
For example, the sequence {e-"' (n/11)~} approaches infinity as n - oo at
the point t = 0. However, the limit lim (f,., </,) exists.
,.... 00

DEFINITION B.3 Two regular sequences {/,.} and {g,.} are equiv-
alent if for all </, € C
lim U,., </>) = lim (g,., </,) (B.5)
n-+00 n-+ao

Example B.2. The regular sequences {e-••(n/'11')~} and {e-t1/ 1111(1/v2'11'n)} are
equivalent.
DEFINITION B.4 A generalized function g is defined as a total, or
complete, class of equivalent regular sequences. The term total implies
here that there exists no other equivalent regular sequence not belonging
to this class. Any member of the class, for example, {g,.}, is sufficient to
represent both g and the total class of equivalent regular sequences
defining g. We denote this symbolically by the form g,..,, {g,.}.
Eumple B.3. All of the equivalent, regular sequences
[{e-•1/ 111}, {e-14111•}, {e-141111}, ••• , {e-•1 •/111•}1
represent the same generalized function g ~ {e- 11/ 111}.

DEFINITION B.5 The inner product (g, </,) of a generalized function,

g and a function tp(t) € C is defined as

(g, </,) = !~ L: g,.(t) 4/..t) dt (B.6)

The inner product is often given the following symbolic representation.

(g, </>) = L: g(t) 4/..t) dt (B.7)

Note that the integral here is used symbolically and does not imply
actual integration.
DEFINITION B.6 If g and h are two generalized functions rep-
resented by the sequences g,..,, {g,.} and h,..,, {h,.}, the sum g + h is
defined by the representation g + h ,..,, {g,. + h,.}.
Note that the set of sequences {g,. + h,.} represents a total class of
equivalent regular sequences made up of the sum of sequences defining g
and h; therefore g + h is a properly defined generalized function.
 Appendix B ◄73

DEFINITION 8.7 The product atg of a generali7.Cd functiong ,_, {g.}

and a constant is defined by the representation atg ,_, {atg11}.
at
DEFINITION a.a The derivative g' of a generali7.ed function g ,_,
{g11} is defined by the representation g' ,_, {g'11}.
Example B.4. For the generali7.ed function g 1 ~ {e-' 1 111
/ } the derivative is
represented by

g'1 ~ {- ~e-tt/,.a}

and (g'1, tf,) == fun =!

f ao ( - e-tt/11•)«1) dt (B.8)
,.... ao J-ao ~
In Definitions B.6, B.7, and B.8 we have defined the operations of addition,
multiplication by a scalar and differentiation. It must be pointed out that
the operation of multiplication between two generalized functions is not
defined in general.
We next consider an important theorem, whose proof is given in
Lighthill, 7 which will enable us to represent any ordinary function, such
as a step function by a generali:zed function equivalent.
Theorem B.1. Given any ordinary function f(t) satisfying the condition

f ao 1/(t)I dt < oo (B.9)

J_ao(l +to/1

for some N ~ 0, there exists a generali7.ed function• f ~ {/ (1)} such that

11

<J. ,> s_:,c,> «1> t1t

== (B.10)

for all tf, € C. In other words, an ordinary function satisfying Eq. B.9 is equiva-
lent in terms of inner products to a generali7.ed function. Symbolically, we
write / = /. H, in addition, / is continuous in an interval, then fun /,. =/
pointwise in that interval. ,....ao
Furthermore, it can be shown that all the operations of addition, sca1ar
multiplication, and differentiation performed on both/ and f yield equivalent
results, that is,
(«/1 + {Jf,i)' = (a./1 + {Jf,i)' (B.11)

when differentiation is permitted on the ordinary function.

7
Lighthill, op. cit., Section 2.3.
• Note that when we rep~t an ordinary function by generalized function equivalent,
we use a bold face italic letter to denote the generalized function.
◄7◄ Network analysis and synthesis
DEFINITION B.9 The generalized step function II is defined as the
total class of equivalent regular sequences {u,.(t)} such that

(11, tf,) = !~1! [: u,.(t) tf,(t) dt

= L: u(t) tf,(t) dt (B.12)

= La, tf,(t) dt
where u(t) is the unit step defined in Chapter 2. That {u,.(t)} exists is
guaranteed by the previous theorem allowing representations of ordinary
functions by generalized functions. Hence, we write 11 = u.
Example B.S. The sequence

u,.(t) = exp [ -1(~ + t9)] t>O (B.13)

=0 t S: 0
which is plotted in Fig. B.1, is one member of the class of equivalent regular
sequences which represents the generalized step function.

0.8

.:-0.6
'-
:f

0.4

0.2

1 2 3 4 5 6 7 8

FIG. 8.1. The generali7.ed step sequence, u.(t).

DEFINITION B.10 The unit impulse, or Dirac delta function ~(t),

is defined as the derivative of the generalized step function ~(t) ,..., {u' ,.(t)}.
 Appendix B -475
It should.be stressed that «'(t) is merely the symbolic representation for a
total class of equivalent regular sequences represented by {u',.(t)}. Thus
when we write the integral

we actually mean
L: «'(t) r/,(_t) dt

L: «'(t) r/,(_t) dt = («', cf,) = !~1! L: u',.(t) r/,(_t) dt (B.14)

Example B.6. The sequence

t>O
(B.15)

== 0 t :S: 0
in Fig. B.2 is one member of the class of equivalent regular sequences which
represents the unit impulse. Other members of the class are the sequences
{e-•"1 (n/1r)H} and {e11/lfll(l/Vhn)}.

FIG. 8.2. The generalized sequence, u'.(t).

476 Network analysis and synthesis

8.2 PROPERTIES OF THE UNIT IMPULSE

Sifting
The most important property of the unit impulse is the sifting property
represented symbolically by

J>o~t)f(t) dt = J(O); (lcxl, IPI < ex>) (B.16)

f.•<O
where/ is any function differentiable over [ex, p]. The left hand side of
Eq. B.16 is defined formally by
J>O f.J>O
6(t)J(t) dt = lim u',.(t)f(t) dt (B.17)
f.•<O ,a ➔ oo •<0 .

The proof of the sifting property is obtained by simply integrating by

parts, as follows.

J.!'! 1:0

u',a(t)/(t) dt = !~"! u,.(t)f(t)I: -!~"! f: u,a(t)f'(t) dt

= J(P) - i, J➔°! u..(t)f'(t) dt

(B.18)

= J(P) - f :f'(t) dt

= J(P) - [f(/1) - f(O)J = J(O)

Pictorially we represent 6(t) by a spike as shown in Fig. B.3. If the
impulse is centered at t = a, then the sifting property is given symbolically
as
l>a

f.
•<o
6(t - a)f(t) dt = f(a) (lcxl, IPI < oo) (B.19)

0 t

FIG. B.3. The unit impulse.

 Appendix B ~77
where f'(t) must exist over [ot, /1). Note that when the limits of integration
are infinite, we actually mean

L: t,(t)f(t) dt = 11~~00

11 ➔ 00
I:6(t)f(t) dt (B.20)

In the sifting property, if both ot, p > 0 or ot, p < 0, then

f:6(t)f(t) dt =0 (B.21)

The proof of this property is similar to the original proof of the sifting
property, and will be left as an exercise for the reader.
lntqration
The defining equations of the delta function according to Dirac are
fl>O

f.11<0
6(:r:) dz= 1
(B.22)
6(:r:) = O; z "F 0
These are actually properties of the delta function as viewed from the
generalized function standpoint. The proof can be obtained directly from
the sifting property. Suppose we have the integral
fl>O

f.11<0
6(t)f(t) dt = f(O) (B.23)

and we let Jtt) = 1. Then we have

fl>O

f.11<0
6(t) dt = /(0) = 1 (B.24)

If both ot, Pare greater than zero or both are less than zero, then

f:6(t)dt =0 (B.25)

This property is stated symbolically by the conditions 6(t) = 0 for t "F 0.

Differentiation across a discontinuity
Consider the function/(t) in Fig. B.4. We see that/(t) has a discon-
tinuity of A. at t = T. Ifwe let/1(t) = /(t) fort< T, andfi(t) = /(t) - A.
for t ~ T, then we have
/(t) = fi(t) + A. u(t - T) (B.26)
478 Network analysis and synthesis

f(t)

- - - -.... {i(t)

0 T

FIG. B.4. Function with discontinuity.

Since f(t), / 1(t), and u(t) satisfy the condition

< oo
i-oo
oo
(1
lg(t)I
+ {')N
dt

for some N; we can represent these ordinary functions by generalized

functions
(ft) = / 1(t) + u(t) (B.27)

Taking derivatives on both sides of Eq. B.27 yields

f'(t) = /'i(t) + A u'(t) (B.28)

which symbolically can be written as

f'(t) = f'i(t) + A b(t) (B.29)

We thus see that whenever we differentiate across a discontinuity, we

obtain a delta function times the height of the discontinuity.
Example B.7. The step response of an R-C network is given as

h(t) = Ae-lf'l' u(_t) (B.30)

shown in Fig. B.Sa. The impulse response is

A
h'(t) = A c5{t) - - e-l/T u(_t) (B.31)
T
and is shown in Fig. B.Sb.
 Appendix B -479

h(t)

(a)

/,,'(t)

A
'f

(b)
FIG, B.S. Differentiation across a discontinuity.

Differentiation
The derivative of a delta function, which we call a doublet, is defined
symbolically as 15'(t),....,, {u• ,.(t)}. It has the following property, where
t•(t) exists over [ex, p].
/1>0
f.«<O «J'(t)f(t) dt = -f'(O); · (lexl, IPI < oo) (B.32)

The proof is obtained through successive integration by parts.

/1>0 f.fl>O
f.«<O t5'(t)f(t) dt = lim a<O u",.(t)f(t) dt
,a ➔ ao (B.33)
= lim u',.(t)f(t) /1>0 - lim f./1>0u ',.(t)f'(t) dt
n ➔ ao 1a<O ,a ➔ ao «<O

We see that since lim u' ,.(P) = lim u' ,.(ex)= 0,

,._.co 1/1>0
lim u' ,.(t)f(t)
n ➔ co «<O
=0 (B.34)
-480 Network analysis and synthesis
a'(t)

FIG. B.6. The doublet 6'(t).

We then integrate by parts again so that

0
-lim i'> u'n(t)f'(t) dt =- lim u,.rl' + lim i'u,.(t)f"(t) dt
n ➔ 00 ar<O n ➔ ao « n ➔ ao 11.

(B.35)
= -f'((J) + f:u(t)f"(t) dt
= -f'({J) + f'((J) - f'(O) = -f'(O)
In general, the derivative-sifting property can be stated symbolically as

,>o«S"(t - a)f(t) dt = (-1)"/<" 1(a) (B.36)

i s<O
where J<"+i1(t) exists over [ct, /J].'1
The generalized function c}'(t) is sometimes called a doublet. The
pictorial representation of a doublet is given in Fig. B.6.
Other properties of the unit impulse
Dirac and others have obtained a host of identities concerning the unit
impulse. We will merely give these here without proof.
c}(-t) = "(t) (1)
c}'(-t) = -c}'(t) (2)
t "(t) = 0 (3)
t c}'(t) = -c}(t) (4)
c}(at) = lal-1 c}(t) (5)
"(t1 - a1) = l lal-1 {"(t - a) + "(t + a)} (6)
f(t) c}(t - a) = /(a) c}(t - a) (7)
The proofs of these properties are obtained through the inlier product
with a testing function 4,(t) E C.
• The condition on/"i+11 is sufficient, but not necessary.
 appendix C
Elements of complex variables

C.I ELEMENTARY DEFINITIONS AND OPERATIONS

A complex variable z is a pair of real variables (x, y) written as

z=x+jy (C.l)

where j can be thought of as J - l.

The variable x is called the real part of z, and y is the imaginary part of z.
Written in simpler notation, we have
x = Re(z), y = Im(z) (C.2)
The variable z can be plotted on a pair of rectangular coordinates. The
abscissa represents the x or real axis, and the ordinate represents th& y or
imaginary axis. The plane upon which x and y are plotted is called the
complex plane. Any point on the complex plane, such as z = 3 + j2, can
be represented in terms of its real and imaginary parts, as shown in Fig. C. l.
From the origin of the complex plane, let us draw a vector to any point z.
The distance from the origin to z is given by
lzl = (x1 + y1)~ (C.3)
jy

and is known as the modulus of z. The

angle which the vector subtends is known
as the argument of z or
arg z = tan-ill (C.4)
X

Letting O = arg z and r = lzl,we can

represent z in polar coordinates as -jy

z = re" (C.5) FIG. C.I

-481
482 Network analysis and synthesis
Expanding this last equation by Euler's formula, we obtain
z == r cos (J + jr sin 0, (C.6)
so that x == rcos (J
(C.7)
y==rsinO
The rule for addition for two complex numbers is given as
(a + jb) + (e + jd) == (a + e) + j(b + d) (C.8)
When two complex numbers are multiplied, we have
(a + jb)(e + jd) == (ae - bd) + j(ad + be) (C.9)
where j1 == -1. If we express the complex numbers in polar form, we
obtain
(a + jb) == r1eill1 (C.10)
and (e + jd) == r1e1111 (C.11)
When we multiply the two numbers in polar form, then

r1ei"1r1e'"■ == r1rae1<111+"■ > (C.12)

If we divide these two numbers in polar form, then

(C.13)

In r~tangular coordinates, the operation of division can be expressed as

a + jb (a + jb)(e - jd)
--==
e+jd (e+jd)(e-jd)
(C.14)
== ae + bd + j be - ad
e1 +d8 e1 +d8
In connection with the modulus of a complex number, it is useful to note
the following rules:
lz1z1I == Iz1I · lz1I
lz1z1 *I == lz1I · lz1 *I == lz111 (C.15)
z. z* == lzl•
where z* is the complex conjugate of z and is defined as
z* == X + jy == X - jy (C.16)
 Appendix C 483
The foil owing rules deal with operations involving the conjugate definition:

Zi+z.==z1*+z.*

ZiZa == z1* · z.* (C.17)

Zi/Za == Z1*/z1*
Finally, if z has a modulus of unity, then

1
z ==- (C.18)
z*
The operations of raising a complex number to the nth power, or taking
the nth root of a complex number, can be dealt with most readily by using
the polar form of the number. Thus, we have z" == (rel")" == r"el"',

and k, 0, 1, ... , n - 1 (C.19)

C.2 ANALYSIS

If to each z == x + jy, we assign a complex number w == u + jv, then w

is a function of z or
w ==/(z) (C.20)
The following are examples of complex functions, i.e., functions of a
complex variable:
w == 2z
w == log. z
w == 1/z (C.21)
w==z1 +4
w == lzl
We see that w may be complex, pure real, or pure imaginary, depending
upon the particular relationship with z. In general, the real and imaginary
parts of ware both functions of x and y. That is, if we let w == .1 + jv,
then
u == /(x, y) and v == J(x; y) (C.22)
As an example, let us find u and v for the function w == z• + 4.
w == z• + 4 == (x + jy)1 + 4 (C.23)
Simplifying, we obtain
w == (x1 - y1 + j2xy) + 4 (C.24)
484 Network analysis and synthesis

jy jy

/l:i
~~+.u ~z+.u
~
z /l:i z

-% 0 % -% 0 s

Path 1 Path 2

-jy -jy

FIG. C.2 FIG. C.3

so that u = x• - y• + 4 and v = 2xy (C.25)

The derivative of a complex function/(z) is defined as

f'(z) = lim f(z + Az) -

f(z) (C.26)
A• ➔O
Az
If one restricts the direction or path along which dz approaches zero, then
we have what is kno~ as a directional derivative. However, if a complex
function is to possess a derivative at all, the derivative must be the same
at any point regardless of the direction in which dz approaches zero. In
other words, in order for /(z) to be differentiable at z = Zo, we must have

df(z)
dz
I•-ao= constant (C.27)

for all directions of approach of dz.

Consider the two directions in which dz approaches zero in Figs. C.2
and C.3. For path l, we have
'( ) . . f(z + dz) - f(z)
/ z = 1Im 1Im ---------'"-...:-~ (C.28)
A""➔OA11➔0 dz
If we substitute
(C.29)
into Eq. C.28, we obtain
'() . . /[x +Ax+ j(y + Ay)] -f(x + jy) (C.30)
/ z = 1Im 1Im-------~--~'"-~---
"""➔o A11➔0 Ax + j Ay
Since /(z) = u + jv (C.31)
and f(z + dz) = u + Au + j(v + Av) (C.32)
 Appendix C 485
we finally arrive at

f'(z) - Jim Jim Au + J Av

+ JAy
Aa➔O 6v... o Az
(C.33)
-limAu +}Av_ au +J'Jv
Aa,➔O Az az az
For path 2, we have
f'(z) - fun lim Au + J Av
-'v ...o 6111--0 Az + J Ay

- fun Au + J Av (C.34)
6v➔o }Ay
av . au
---1-
ay_ au
Since we assume that the function /(z) is differentiable, the derivatives
must be independent of path. Thus, we have

av -Jau _ au +}av (C.35)

ay ay az az
From this last equation, we obtain the Cauchy-Riemann equations, which
are
av au
-=-
ay az
(C.36)
au
-=-
av
oy oz
We have just seen that in order for a function to have a derivative, the
Cauchy-Riemann equations must hold. A function which is single valued
and possesses a unique derivative is called an analytic function. A set of
sufficient conditions for analyticity is that the Cauchy-Riemann equations
are obeyed. For example, consider the function

/(z)- z1 +4 (C.37)
/(z) is analytic because
av
--2z--
au
oy oz
(C.38)
: - -2y- - :
486 Network analysis and synthesis
On the other hand,/(z) = z• is not analytic because
U=X and v= -y
(C.39)
OU= +1 and ov = -1
OX oy
C.3 SINGULARITIES AND RESIDUES

If/(z) is analytic within a region or domain in the complex plane except

at a point z0 , then/(z) has an isolated singularity at z0 • Suppose/(z) has a
singularity at z0 , then we can expand/(z) about z0 in a Laurent series

f(z) = a_n n + · · · + ~ + a 0(z - z0) 0 + a 1(z - z0 ) + ···

(z - z0) z - z0
+ am(z - zor + ...
(C.40)
In the expansion, if m is finite, then z0 is called a pole of order m. 1 The
term a_1 is called the residue of the singularity.
Example C.1. Consider the Laurent series for the function /(z) = e-/z about
the pole at the origin. We can expand e• in a power series to give

-e• == -1( 1 + z + -z
1 11 + -1 z3 + · · ·)
z z 2! 3!
(C.41)
1 1 1
=-+l+-z+-z1 +···
z 2! 3!
ACC()rding to the definition, the residue of the pole at z = 0 is equal to 1.
Example C.2. Expand the function/(z) = 1/z(z - 1)1 about the pole at z = 1,
and find the residue of the pole at z = 1.
1 1 1
z(z - 1)1 = (z - 1)1 1 + (z - 1)

= l
(z -1)8
[l - (z - 1) + (z - 1)1 - (z - 1)3 + · · ·] (C.42)

1 1
= -----,,1 - - - + 1 - (z - 1) + (z - 1)8 + · · ·
(z - 1) z - 1
for O < tz - 11 < 1. Here, the residue of the pole at z = 1 is equal to -1.
1 Note that if we have an infinite number of nonz.ero terms with negative exponents,

then 111 0 is an essential singularity.

 · Appendix C "187
Example C.3. Find the residues of the poles at s =0 and s =- -1 of the
function
s+2
/(s) = r(_s + 1)' (C.43)

To find the residues, we simply perform a partial fraction expansion

2 3 1 3
f(s) ""' ~ - s+ (s + 1)' + s + 1 (C.44)

Thus the residue of the pole at s =- 0 is -3, and the residue of the pole at
s = -1 is +3.

C.4 CONTOUR INTEGRATION

In complex integration the integral is taken over a piecewise smooth path

C and is defined as the limit of an infinite summation

f f(z) dz = lim ! f(z 1) liz1 (C.45)

Jc n➔ oo 1-1
where z1 lies on C. Unlike the process of differentiation, the path along
which we take the integral makes a difference as to the ultimate value of
the integral. Thus the integral

f.
•1
•f(z) dz

in general, has different values depending upon whether we choose to inte-

(C.46)

grate along path C1 or path C1, as shown in Fig. C.4. If we integrate along
a closed path, say from a to b and then to a again, we are integrating along
a closed contour. The path shown in Fig. C.5 is an example of a closed
contour. The following theorem, known as Cauchy's residue theorem gives
a method for rapid evaluation of integrals on closed paths.

jy

0
~Ll
jy

0
·O.._ contour

FIG. C.4 FIG. C.S

Tbeorem C.1. If C is a simple closed curve in a domain D, within which /(z) is

analytic except for isolated singularities at Zi, ;, ••• , z,., then the integral along
"488 Network analysis and synthesis
the closed path C is

£f(z)dz = 21rj(K1 + K! +···+IQ (C.47)

where K, represents the residue of the singularity z,.

EU1Dple C.4. Consider the integral
J, s+2 ds (C.48)
j s'(s + 1)8
along the circle Isl = 2, as given in Fig. C.6. Since there are two singularities
within the circle, at s = 0 and at s = -1, whose residues are respectively -3
and +3, then the integral along the circle is
J, s + 2 ds = 21rj( -3 + 3) = O (C.49)
j s'(s + 1)8
jy

----<i--x--x----+---
-1 0
"

-jy
FIG. C.6

EU1Dple C.S. Find the integral of f(s) along the closed contour shown in Fig.
C.7. The function/(s) is given as
3s +S (C.SO)
/(s) = (s + l)(s + 2)
jy

-2 -1 0

-b
FIG. C.7
 Appendix C -489
A partial fraction expansion of/(s) shows that

f(s) = _1_ + _2_ (C.51)

s+l s+2
so that the residues are 1 and 2. The value of the integral along the closed
path within which both the singularities lie is then

·£f(s) ds = 21rj(l + 2) = 61rj (C.52)

If /(z) is analytic in a domain with no singularities, then the integral

along any closed path is zero, that is,

f
0
J(z)dz =0 (C.53)

This result is known as Cauchy's integral theorem.

 appendix D
Proofs of some theorems on
positive real functions

Theorem D.1. If Z(.r) and W(.r) are both positive real, then Z{W(.r)) is also
positive real.
Proof. When Re .r ~ 0, both ReZ(.r) and Re W(.r) ~ 0, then Re Z(W(.r)) ~ 0,
also. When .r is real, both Z(.r) and W(.r) are real, hence Z(W(.r)) is real. Since
Z(W{.r)) satisfies both conditions of positive realness, it is positive real.
Theorem D.2. If Z(.r) is positive real, then Z{l/.r) is positive real.
Proof. W(s) = Ifs is positive real, hence Z(W{.r)) = Z{l/.r) is positive real.
Theorem D.3. If W(.r) is positive real, then 1/W(.r) is also positive real.
Proof. Z(s) = 1/s is positive real, henceZ(W(.r)) = 1/W(.r) is positive real by
Theorem D.l.
Theorem D.4. The sum of positive real functions is psitive real.
Proof. Suppose Z 1(.r) and Z,l...r) are both positive real. When Re .r ~ 0, then

ReZ1 ~ o· and ReZ1 ~ 0

so that ReZ1 + Rez. == ReZ ~ 0.

Also, when .r is real, both Z 1 and Z 1 are real. The sum of two real numbers
is a real number. Therefore, Z 1 + Z 1 is positive real.
Theorem D.S. The poles and zeros of Z(.r) cannot have positive real parts
(i.e., lie in the right half of the .r plane).
Proof. Suppose there is a pole .r0 in the right-half plane. Let us make a
Laurent series expansion about .r0 so that

Z(.r) == k_,.
(s - so),.
+ (s. k-n+l
- so)n-1
+ · · · + k 1(s - s)
o
+ · · · + k,{_s - so)r + ·· ·
490
 Appendix D 491
ReZ(a)

'
FIG. D.I

where n is real and finite. In the neighborhood of the pole so, :i(s) can be
approximated by
Z(s) !::>! k_n
(s - so)n

We can represent Z(s) in polar form by substituting each term by its polar form;
i.e., let (s - so)+" = ,ne,,.. and k_n = Kei• so that
K K
zr"s) - - e1<.-ne>
,n • ,
ReZ(s) ==-;; cos (,f, - n9)

which is represented in Fig. D. l. When 9 varies from Oto 2,r, the sign of Re Z(s)
will change 2n times. Since ReZ(s) ~ 0 when Res~ 0, it is seen that any
change of sign of Re Z(s) in the right-half plane will show that the function is not
positive real. Therefore, we cannot have a pole in the right-half plane. Since the
function 1/Z(s) is positive real if Z(s) is positive real, it is obvious that there
cannot be any zeros in the right-half plane also.
Theorem D.6. Only simple poles with positive real residues can exist on the
j<» axis.
Proof. As a consequence of the derivation of Theorem D.S, it is seen that
+
poles may exist on the j<» axis if n = 1, and = O. The condition n - 1 implies
that the pole is simple and the condition ,f, == 0 implies that the residue is positive
and real. It is readily seen that zeros on the j<» axis must also be simple.
Theorem D.7. The poles and zeros of Z(s) are real or occur in conjugate pairs.
Proof. If a complex pole or zero exists without its conjugate, Z(s) cannot be
real when s is real. As a result of this theorem and Theorem D.S, it is seen that
both the numerator and denominator polynomials of Z(s) must be Hurwitz.
Theorem D.8. The highest powers of the numerator polynomial and the
denominator polynomial of Z(s) may differ by at most unity.
Proof. Let Z(s) be written as
ansn + "-1,n-1 + • • • + a1s + Clo P(s)
Z(s) .. brnsm + bm- 1sm-l + · · · + bis + bo ... Q(s)
◄92 Network analysis and synthesis

(T

FIG. D.2

Ifm -n ~ 2,whens = oo,Z(s)willhaveazerooforder2orgreaterats = oo,

which is on the jw axis. Similarly, if n - m ~ 2, then, at s = oo, Z(s) will have a
pole of order 2 or more at s = oo. Since Z(s) cannot have multiple poles or
zeros on the jw axis, these situations cannot exist; therefore In - ml ~ 1.
Theorem D.9. The lowest powers of P(s) and Q(s) may differ by at most unity,
Proof. The proof is obtained as in Theorem 0.8 by simply substituting 1/s
for s and proceeding as described.
Theorem D.10. A rational function F(s) with real coefficients is positive real if:
(a) F(s) is analytic in the right-half plane.
(h) If F(s) has poles on the jw axis, they must be simple and have real, positive
residues.
(c) Re F(jw) ~ 0 for all w.
Proof. We need only show that these three conditions fulfill the same require-
ments as Re Z(s) ~ 0 for Res ~ 0. We will make use of the minimum modulus
theorem which states that if a function is analytic within a given region, the
minimum value of the real part of the function lies on the boundary of that
region. The region with which we are concerned is the right-half plane which is
bounded by a semicircle of infinite radius and the imaginary axis with small
indentations for the jw axis poles. If the minimum value on the jw axis is greater
than zero, then Re Z(s) must be positive over the entire right-half plane (Fig. D.2).
 appendix E
An aid to the improvement
of filter approximation

E.I INTRODUCTION

The introduction of an additional pole and zero in the second quadrant

of the complex frequency plane, and at their conjugate locations, can give
amplitude or phase corrections to a filter approximant over some desired
band of frequencies without significantly changing the approximant at
other frequencies. However, a cut-and-try procedure for finding the best
positions for such a pole-zero pair can be tedious. A visual aid is presented
herein which reduces the amount of labor required to make modest
corrections of this type.
Constant phase and constant logarithmic gain contours for the correc-
tion by a pole-zero pair1 ·are plotted on transparent overlays. One of these
may be placed over a suitably scaled sheet of graph paper representing the
complex frequency plane. Then the pair-shaped phase and gain correc-
tions along the jru axis are indicated by the intersections of the overlay
contours with this axis. Corrections which best reduce the errors in the
original approximant are ihen sought by variation of the overlay position
and orientation. Either phase or amplitude may be corrected. However,
it is not always possible to simultaneously improve both the phase and the
amplitude characteristics of an approximant by a single pair-shaped
correction.

F. F. Kuo and M. Karnaugh, reprinted from the IRE Transactions on Circuit Theory,
ct-9, No. 4, December 1962, pp. 400--404.
1
This will be called, hereafter, a pair-shaped correction.
◄93
~9-4 Network analysis and synthesis

E.2 CONSTANT LOGARITHMIC GAIN CONTOURS

Suppose we begin with a transfer function G(s) with certain deficiencies

in its amplitude or phase response. Let us consider the transfer function
of a corrective network G1(s) such that the product
(E.l)
will have better gain or phase characteristics. For the purposes of this
paper, we will restrict G1(s) to have the form

G (s)
1
= C (s - q)(s - q) (E.2)
(s - p)(s - p)
where C is a constant, and q and p are a zero and a pole, respectively, in
the second quadrant of the complex frequency plane. If the correction
is to be applied at sufficiently high frequencies such that s - q:::::::: s - p
then
Gi(s)"' C(s -q) (E.3)
s-p
Let us consider the effect when the pole-zero pair in Eq. E.3 is used to
augment any given rational function in the complex frequency plane.
The added gain, in decibels, due to this pole-zero pair is

.
D = 20 log10 I I
s - q
s -p
(E.4)

where we have neglected the effect of the constant C in Eq. E.3. If we let
k = (lO)D/IO (E.S)

then k = 1~1
s-p
(E.6)

For fixed k, this is the equation of a circle with inverse points• at q and
p. Its radius is
kip-qi
(E.7)
P =
11 - k1 I
and its center is at

(E.8)

• E. C. Titchmarsh, 1716 Th6ory of Functions, Oxford University Press, Oxford,

England, second edition 1939, pp. 191-192.
 Appendix E 495
E
Let p=so+-
2
(E.9)
E
q=so--
2

Then (E.10)

(E.11)

Here, s0 is the midpoint between the pole and zero, and their separation
is E. It is easy to see from Eq. E.11 that the center of each circle of constant
gain is externally collinear with the pole and zero. For the purpose of
drawing the family of constant gain contours, we may let s0 be the origin
of coordinates and the scale fa~tor E may be set equal to unity.
Furthermore, only half the pattern need be drawn because the function
D has negative symmetry with respect to a reflection about the perpen-
dicular bisector of the pole-zero pair.

E.3 CONSTANT PHASE CONTOURS

Figure E.la represents a pole at p, a zero at q and an arbitrary point

s in the complex frequency plane. When the pole and zero are used to
correct a given phase characteristic, the added phase at s is
tJ, = ()fl - O,, (E.12)
where ()fl and fJ,, are measured with respect to a single arbitrary reference.
In Fig. E.Ib, a circle has been drawn through p, q, ands. Angle ,J, is
s

I
I
I
I
I
I
I
\
\
\

'' '
s
(a) (b)
FIG. E.1. Derivation of constant phase contours.
496 Network analysis and synthesis
equal to one half the subtended arc ps'q. Therefore, the arc qsp is a
constant phase contour. The angle at the center of the circle between cp
and the perpendicular bisector of chord pq is also equal to cf,. All of the
circular contours of constant phase have their centers on this perpendicular
bisector.
Note that minor arc ps'q is also a contour of constant phase, but the
phase angle
(E.13)
is negative, as indicated by the clockwise rotation from s'p to s'q. The
convention for positive rotation is herein taken to be counterclockwise.

E.4 CONTOUR DRAWINGS

Sets of constant phase and constant logarithmic gain contours are

drawn in Figs. E.2 and E.3. The curves are symmetric about the zero-
decibel line, except that the gain curves are of opposite sign. Therefore,
only one half of each figure has been drawn.

15°
20° ~~~::=lsl~~H~H~k--ji<Y-Tt----t--t-t-~rl-t-\---i
30°

30°
200
A~~~~~A-J)<::-+~~~:-l--l-~---ll--+-~--+t4-l--l---l

FIG. E.2. Constant amplitude and phase contours.

 Appendix E ◄97

db

FIG. E.3. Constant amplitude and phase contours.

The zero-decibel line is the perpendicular bisector of the line segment

joining the pole and zero. It is a gain contour of infinite radius.
The line through the pole and zero is a zero phase contour. Together,
these perpendicular axes form a useful reference system. The signs of the
phase and gain in the four quadrants formed by these axes are shown in
Fig. E.4.
The corrections in Figs. E.2 and E.3 do not carry algebraic signs because
498 Network analysis and synthesis
Zero gain

G- G+
q,- 4>-

q
x-zero phase
p
G- G+
q,+ 4>+

FIG. E.4. Signs of the gain G and phase ,f, corrections.

only half of the symmetrical pattern has been drawn. The signs may be
obtained from Fig. E.4, and are important in selecting the orientation of
the pole-zero pair.

E.5 CORRECTION PROCEDURE

In correcting a given approximant, it is first necessary to plot the desired

magnitude and/or phase corrections versus frequency. When the contours
of Fig. E.2 or E.3 are overlaid on a second sheet representing the complex
frequency plane, the contour intersections with the jw axis indicate the
corrections that will actually be realized.
There are several variables at the designer's disposal. The first is the
distance between the correcting pole and zero. Since the scale of the
contours in Figs. E.2 and E.3 is not specified, the frequency scale of
the un9erlying complex plane determines the distance between the pole
and zero. The second design variable is the location of the center of the
pole-zero pair. The distance of the center from a given band on the jw
axis determines the magnitude of the correction.
The third variable is the orientation of the pole-zero pair. Figure E.4
shows how the orientation affects the gain and phase corrections. As a
simple example, suppose one wishes to have zero phase correction at
w = 1.0, negative phase correction above and positive correction below
that frequency. The attack would be to point the zero phase axis at
w = 1.0 with the pole nearest to the jw axis. If it is desired to have equal
phase correction above and below w = 1.0, the zero phase axis should be
oriented parallel to the real axis of the complex frequency plane. If one
wishes to have more phase correction above w = 1.0 and less below, the
 Appendix E -499
zero phase axis should be rotated clockwise with respect to the a (real)
axis.
We thus see that by varying the frequency scale of the complex frequency
plane, the position of the center of the pole-zero pair, and the orientation
of the pole-zero pair, the pair-shaped correction can be made to approxi-
mate the desired correction.
It must be emphasized that the method suggested herein is an aid to
cut-and-try correction. As such, it is easier to use the method than to
precisely set down rules for applying it. However, a few rather general
statements may be helpful.
Unless the pole and zero both lie on the real axis, one must remember
that another pole and zero are located at conjugate positions. The con-
tributions from both pole-zero pairs may be added algebraically. In most
practical cases, the desired correction will have a band-pass character.
Therefore, only one pole-zero pair will normally contribute significantly
at any frequency.
The shape and magnitude of the desired correction will dictate the way
in which the jw axis must intersect the correction contours. The_broadness
of the desired correction will dictate the proper scaling of the jw axis.
Usually, only a few trials are needed to fix the pole and zero locations for
the best fit.
It will be found that a worthwhile correction can be made in either the
phase or the gain characteristic. Only fortuitously can they be improved
simultaneously by a single pair-shaped correction.

Example E.1. The amplitude response of a third-order Butterworth filter is

given by the solid curve in Fig. E.5. It is desired to steepen the gain roll-off
near the cutoff frequency me = 1. This is done by increasing the gain just below
co = co0 and decreasing it above that frequency. Figure E.6 illustrates the type
of correction desired. Figure E.7 shows a pole-zero pair that achieves this type
of correction. The gain at co = 1 remains unchanged by this particular choice.
Other pole-zero pairs that "aim" the zero-decibel line at co == 1 but give asym-
metrical corrections about that point might also be used. The dashed curve in
Fig. E.5 shows the corrected gain.
A different pole-zero pair, also shown in Fig. E.7, has been chosen to minimize
the deviation of the slope of the phase response from its slope at co = 0. This
pair-shaped correction decreases the phase for m < 0.7 and increases it for
co> 0.7.
Figure E.8 shows the deviation of the phase responses from linear phase.
It is clear from Figs. E.5 and E.8 that the gain corrected approximant has a
poorer phase response than the original, while the phase corrected approximant
has a gentler gain roll-off than the original.
500 Network analysis and synthesis
2
-----
0

-2
-
.0
i::,
-4
ai
i::,
:E
Q.
E
< -6

-8

-10
\
\
-12
0.1 0.2 0.3 0.5 0.7 1.0 1.5 2.0
Frequency, w
FIG. E.S. Amplitude response of corrected and uncorrected Butterworth filters.

This will not surprise the experienced filter designer. It is possible, however,
to achieve moderate corrections in both gain and phase by using two pair-
shaped corrections. A useful approach to this objective lies in localizing the
gain correction further out of band, and the phase correction further in-band
than in the separate corrections just discussed. This can be done by shifting
the pole-zero centers to higher or lower frequencies and also by experimenting
with nonsymmetrical corrections.

w-
FIG. E.6. Amplitude correction to steepen fall-off of third-degree Butterworth filter.
 Appendix E SO I

loss 1.2
correction
0
1.0
X
Butterworth ___ _ .
poles o.s
I X O
I Phase 0.6
I correction
I
I 0.4
I
I
I 0.2
I
_ ___._ _,_ _..__ _.__.....__-10 jw
-1.0 -0.8 -0.6 -0.4 -0.2
\
\ -0.2
\
\
\ -OA
\
\
\ -0.6
' ', X O

'~ -0.8
0
-1.0
X

-1.2
FIG. E.7. Poles and z.eros of the original filter and correction equaliz.ers.

20.----r--~--,---t----r--.-----,.---.---F--,

C:
~
·i -10
"C

i - 20 r--,-i-,-4---"1';;;;;:::::::t:;;;;,,,,---,-7
-&.

,/
-~~_ _,__......__......__ ___._ _~~-'_ _.__....___......_ . . . . J
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Frequency, rad/sec

FIG; E.8. Phase deviation from slope at z.ero frequency.

502 Network analysis and synthesis

E.6 CORRECTION NETWORK DESIGN

A few words are in order concerning the synthesis of the equalizer from
the pole-zero pair obtained by the method described.
In order to provide optimum power transfer and to facilitate cascading
the correction network with the original filter, the correction network
should be of the constant-resistance type. We will restrict our discussion
to the bridged-T network in Fig. E.9, whose voltage transfer function is
given as
G(s) = V:a(s) = R0 (E.14)
Vi(s) R 0 + Z 1(s)
provided the network is of the constant resistance type as given by the
equation
Z 1(s) Z 1(s) = Ro'• (E.15)
For normalization purposes, we will let R 0 = 1 n.
Let the pole-zero correction be written in general as
2
G(s) = s + a1s + a0 (E.16)
k(s 2 + b1s + bo)
Since the correction has minimum phase, we know that {a,, b, ~ O}. In
addition, since the d-c gain cannot exceed unity, kb0 ~ a0 •
We can express the impedance Z 1(s) in terms of G(s) in Eq. E.14 as
1
Zi(s) = _1__ 1 = (k - 1)s + (kb1 - a 1)s + (kb 0 - a0) (E.l?)
G(s) s 8 + a 1s + a 0
Since Z 1(s) must be positive real, the coefficients must all be nonnegative
so that
k-1~0
kb1 - a1 ~ 0 (E.18)
kb0 - a0 ~ 0

----------1 Z1 - - - - - .

Ro Ro

Ro Ro

FIG. E.9. Bridged-T network.

 Appendix E 503

/x
,,,,.,---- --
• plane
I
/
I
I
I
I
I
I

---+----1----'---'-------"'10
.I
-u I I

I
\
\
\
\ \ ·,.
--
\
\

''
' 'x, ...._
-jw
--
FIG. E.10. Poles and zeros of Z 1(s).

Moreover, in order for a biquadratic driving-point immittance to be

positive real, the following condition must apply. 3

(E.19)

Details concerning the synthesis of Zi(s) are also given in Seshu's paper.
Let us plot the pole-zero pair of Zi(s) as in Fig. E.10. We can represent
the locations of the poles and zeros in terms of the polar angles </,f) and o.
and their distances from the origin, nf) and n.. Rack' has shown that in
order for in-band loss to be finite
nf) 2
-<
n.
(E.20)

In addition he has shown that for an unbalanced bridged-T circuit, if

ot = max [O •• </,f)] then the larger the angle ot, the larger the in-band loss.
In particular, ot should be less than 70° to restrict the in-band loss to
reasonable proportions. If one considers other network configurations
1
S. Seshu, Minimal Realizations of the Biquadratic Minimum Function, IRE
Trans. on Circuit Theory, CT-6, December 1959, 345-350.
'A. J. Rack, private communication.
504 Network analysis and synthesis
such as lattice networks, it is conceivable that Rack's restrictions might
be relaxed. For lattice network design, Weinberg's method is applicable, 5
although here again, the in-band loss restrictions are severe.

E.7 CONCLUSION

We have presented here a simple visual aid to the correction of the

amplitude or phase response of filters. The method has the advantage of
facilitating the commonly used cut-and-try approach to this problem. It
has the disadvantage, in many cases, of failing to provide simultaneous
amplitude and phase corrections in a single step.

'L. Weinberg, "RLC lattice networks," Proc. IRE, 41, September 1953, 1139-1144.
 Bibliography

SIGNAL ANALYSIS

Goldman, S., Frequency Analysis, Modulation and Noise, McGraw-Hill,

New York, 1948.
Javid, M. and E. Brenner, Analysis, Transmission and Filtering of Signals,
McGraw-Hill, New York, 1963.
Lathi, B. P., Signals, Systems and Communications, John Wiley and Sons,
New York, 1965.
Lighthill, M. J., Introduction to Fourier Analysis and Generalized Functions,
Cambridge University Press, New York, 1958.
Mason, S. J. and H. J. Zimmerman, Electronic Circuits, Signals and
Systems, John Wiley and Sons, New York, 1960.
Papoulis, A., The Fourier Integral and Its Applications, McGraw-Hill, New
York, 1962.
Rowe, H. E., Signals and Noise in Communications Systems, D. Van
Nostrand, Princeton, N.J. 1965.
Schwartz, M., Information Transmission, Modulation and Noise, McGraw-
Hill, New York, 1959.

NETWORK ANALYSIS

Balabanian, N., Fundamentals of Circuit Theory, Allyn and Bacon, Boston,

1961.
Bohn, E. V.; The Transform Analysis of Linear Systems, Addison-Wesley,
Reading, Massachusetts, 1963.
Brenner, E. and M. Javid, Analysis of Electric Circuits, McGraw-Hill, New
York, 1959.
505
506 Network analysis and synthesis
Brown, R. G. and J. W. Nilsson, Introduction to Linear Systems Analysis,
John Wiley and Sons, New York, 1962.
Brown, W. M., Analysis of Linear Time-Invariant Systems, McGraw-Hill,
New York, 1963.
Carlin, H.J. and A. Giordano, Network Theory, Prentice-Hall, Englewood
Cliffs, N.J., 1964.
Cassell, W. L., Linear Electric Circuits, John Wiley and Sons, New York,
1964.
Chen, W. H., The Analysis of Linear Systems, McGraw-Hill, New York,
1963.
dePian, L., Linear Active Network Theory, Prentice-Hall, Englewood
Cliffs, N.J. 1962.
Friedland, B., 0. Wing, and R. B. Ash, Principles of Linear Networks,
McGraw-Hill, New York, 1961.
Gardner, M. and J. L. Barnes, Transients in Linear Systems, Vol. 1, John
Wiley and Sons, New York, 1942.
Guillemin, E. A., Introductory Circuit Theory, John Wiley and Sons, New
York, 1953.
Guillemin, E. A., Theory ofLinear Physical Systems, John Wiley and Sons,
New York, 1963.
Harman, W. W. and D. W. Lytle, Electrical and Mechanical Networks,
McGraw-Hill, New York, 1962.
Hayt, W. H., Jr. and J. E. Kemmerly, Engineering Circuit Analysis,
McGraw-Hill, New York, 1962.
Huelsman, L. P., Circuits, Matrices and Linear Vector Spaces, McGraw-
Hill, New York, 1963.
Kim, W. H. and R. T. Chien, Topological Analysis and Synthesis of Com-
munication Networks, Columbia University Press, New York, 1962.
Ku, Y. H., Transient Circuit Analysis, D. Van Nostrand, Princeton, N.J.,
1961.
Legros, R. and A. V. J. Martin, Transform Calculus for Electrical Engineers,
Prentice-Hall, Inc., Englewood Cliffs, N.J., 1961.
LePage, W. R. and S. Seely, General Network Analysis, McGraw-Hill,
New York, 1952.
Ley, B. J., S.G. Lutz, and C. F. Rehberg,Linear Circuit Analysis, McGraw-
Hill, New York, 1959.
Lynch, W. A. and J. G. Truxal, Introductory System Analysis, McGraw-
Hill, New York, 1961.
Paskusz, G. F. and B. Bussell, Linear Circuit Analysis, Prenctice-Hall, Inc.
Englewood Cliffs, N.J., 1964.
Pearson, S. I. and G. J. Maler, Introductory Circuit Analysis, John Wiley
and Sons, New York, 1965.
 Bibliography 507
Pfeiffer, P. E., Linear System .Analysis, McGraw-Hill, New York, 1961.
Reza, F. M. and S. Seely, Modem Network .Analysis, McGraw-Hill, New
York, 1959.
Sanford, R. S., Physical Networks, Prentice-Hall, Englewood Cliffs, N.J.,
1965.
Schwarz, R. J. and B. Friedland, Linear Systems, McGraw-Hill, New York,
1965.
Scott, R. E., Elements of Linear Circuits, Addison-Wesley, Reading,
Massachusetts, 1965.
Seely, S., Dynamic Systems .Analysis, Reinhold, New York, 1964.
Seshu, S. and N. Balabanian, Linear Network .Analysis, John Wiley and
Sons, New York, 1959.
Skilling, H. H., Electrical Engineering Circuits, Second Edition, John
Wiley and Sons, New York, 1965.
Van Valkenburg, M. E., Network .Analysis, Second Edition, Prentice-Hall,
Englewood Cliffs, N.J., 1964.
Weber, E., Linear Transient .Analysis, John Wiley and Sons, New York,
1954.
Zadeh, L.A. and C. A. Desoer, Linear Systems Theory, McGraw-Hill, New
York, 1963.

NETWORK SYNTHESIS

Balabanian, N., Network Synthesis, Prentice-Hall, Englewood Cliffs, N.J.,

1958.
Bode, H. W., Network .Analysis and Feedback .Amplifier Design, D. Van
Nostrand, Princeton, N.J., 1945.
Calahan, D. A., Modem Network Synthesis, Hayden, New York, 1964.
Chen, W. H., Linear Network Design and Synthesis, McGraw-Hill, New
York, 1964.
Geffe, P.R., Simplified Modem Filter Design, John F. Rider, New York,
1963.
Guillemin, E. A., Synthesis of Passive Networks, John Wiley and Sons,
New York, 1957.
Guillemin, E. A., The Mathematics of Circuit .Analysis, John Wiley and
Sons, New York, 1949.
Hazony, D., Elements of Network Synthesis, Reinhold, New York, 1963.
Kuh, E. S. and D. 0. Pederson, Principles of Circuit Synthesis, McGraw-
Hill, New York, 1959.
Matthaei, G. L., L. Young, and E. M. T. Jones, Microwave Filters,
Impedance-Matching Networks and Coupling Structures, McGraw-Hill,
New York, 1964.
508 Network analysis and synthesis
Saal, R., Der Entwurf von Fi/tern mit Hilfe des Kataloges Normierter
Tiefpiisse, Telefunken GMBH, 1961.
Skwirzynski, J. K., Design Theory and Data for Electrical Filters, D. Van
Nostrand, Princeton, 1965.
Storer, J. E., Passive Network Synthesis, McGraw-Hill, New York, 1957.
Truxal, J. G., Control System Synthesis, McGraw-Hill, New York, 1955.
Tuttle, D. F., Network Synthesis, Vol. I, John Wiley and Sons, New York,
1958.
Van Valkenburg, M. E., Introduction to Modern Network Syntheses,
John Wiley and Sons, New York, 1960.
Weinberg, L., Network Analysis and Synthesis, McGraw-Hill, New York,
1962.
Yengst, W. C., Procedures of Modern Network Synthesis, Macmillan, New
York, 1964.
 Name Index

Aaron, M. R., 448 Eisenman, R. L., 469

Aitken, A. C., 469 Elmore, W. C., 390
Angelo, E. J., 268 Ende, F., 381
Ash, R. B., 506
Faddeeva, V. N., 469
Balabanian, N., SOS, 507 Faddev, D. K., 469
Barnes, J. L., 506 Pano, R. M., 448
Bashkow, T. R.., 283, 438 Feshbach, H., 86
Bellman, R., 469 Foster, R. M., 321
Bode, H. W., 221, 507 Franklin, P., 198
Bohn, E. V., SOS Friedland, B. 0., S06
Brenner, E., SOS Fujisawa, T., 458
Brown, R.. G., S06 Fukada, M., 381
Brown, W. M., S06
Bubnicki, Z., 283 Gantmacher, F. R., 469
Budak, A., 39S Hayt, W. H., Jr., S06
Bussell, B., S06 Geffe, P. R., 507
Giordano, A., S06
Goldman, S., 137, SOS
Calahan, D. A., S01
Goldstone, L. O., 1S3
Carlin, H. J., 428
Cassell, W. L., S06
Guillemin, E. A., 296, 299, 506, so1.
Cauer, W., 324 Harman, W.W., S06
Chen, W. H., 506 Hayt, W. H., Jr., S06
Chien, R.. T., 506 Hazony, D., 507
Hobson, E. W., 471
Darliftgton, S., 431, 457 Hohn, F. E., 469
Davis, H. F., Sl Huelsman, L. P., 261, 469, S06
de Pian, L., 506
Desoer, C. A., S07 Jahnke, E., 381
Dirac, P. A. M., 33, 470, 471 James, R.. T., 389
Dutta Roy, S. C., 283 Javid, M., SOS
509
510 Name Index

Jones, E. M. T., 507 Reza, F. M., 507

Jordan, E. C., 413 Rowe, H. E., 505
Justice, G., 27
Saal, R., 441
Karnaugb, M., 493 Sands, M., 390
Kautz, W. H., 47, 367 Sanford, R. S., 507
Kemmerly, J.E., 506 Saraga, W., 443
Kim, W. H., 506 Schwartz, L., 470, 471
Ku, Y. H., 506 Schwartz, M., SOS
Kub, E. S., 507 Schwarz, R. J., 507
Kuo, F. F., 279, 393, 493 Scott, R. E., 507
Seely, S., 506, 507
Larky, A. I., 262 Semmelman, C. L., 447
Lathi, B. P ., SOS Seshu, S., 503, 507
Le Corbeiller, P., 469 Skilling, H. H., 100, 104, 106, 181, 507
Legros, R., 506 Skwirzynski, J. K., 508
Leichner, G. H., 279 So, H. C., 449, 450
Le Page, W.R., 506 Storer, J. E., 508
Ley, B. J., 506 Szentirmai, G., 441, 453
Lighthill, M .., 65, 136, 471, 473, 505
Lutz, S. G., 506 Tellegen, B. D. H., 272
Lynch, W. A., 506 Temple, G., 34, 471
Lytle, D. W., 506 Terman, F. E., 236
Thomson, W. E., 384
Maler, G. J., 506 Titchmarsh, E. C., 494
Marcus, M., 469 Tompkins, C. B., 447
Martin, A. V. J., S06 Tropper, A. M., 469
Mason, S. J., SOS Truxal, J. G., 506, 508
Matthaei, G. L., 507 Tuttle, D. F., 508
Mine, H., 469
Morse, P. M., 86 Ulbrich, E., 441

Nering, E. D., 469 Valley, G. E., Jr., 293

Nilsson, J. W., 506 Van Valkenburg, M. E., 135, 263, 296,
297, 325, 329, 331, 344, 347, 376,
O'Meara, T. R., 283 507, 508
Orchard, H. J., 385 Von Weiss, H., 469

Paley, R. E. A. C., 291 Walker, F., 283

Papoulis, A., 379, 381, 505 Wallman, H., 293
Paskusz, G. F., S06 Weber, E., 507
Pearson, S. I. 506 Weinberg, L., 400, 435, 504
Pederson, D. 0., 507 Widder, D. V., 136
Perlis, S., 469 Wiener, N., 291
Pfeiffer, P. E., 507
Pipes, L. A., 469 Yengst, W. C., 508
Young, L., 507
Rack, A. J., 503
Raisbeck, G., 8, 294 Zadeh, L. A., 507
Rehberg, C. F ., 506 Zimmerman, H., SOS
 Subject Index

A.BCD parameters, 262 Cauchy-Riemann equations. 48S

Admittance, driving-point, lS, 187 Cauer ladder expansion, 324
All-pass network, 221, 3S7 Causality, 9, 290
Amplifier, 11 Characteristic equation, 77
impulse response of, 201 Characteristic impedance, 414
Amplitude response, 212, 342 Characteristic value, 78
computer program for, 4-SO Cbebyshev filter, 373
evaluation by vector method, 21S approximation, 366
Amplitude spectrum, 3 locus of poles, 376, 378
Analytic function, 48S tables of, 373, 400, 43S
Approximation problem, 17, 36S transient response, 392
Available gain, 418 Chebyshev polynomial, 373
Available power, 418 Circuit, bridged-T 2S8, 27S, S02
double-tuned, 238
Band-elimination filter transformation, reciprocal, 8, 2SS
409 single-tuned, 229
Band-pass filter transformation, 407 symmetrical, 2S9
Bandwidth, half-power, 23S Coefficient of coupling, 122
spectral, 67, 389 Compensation theorem, 181
Bessel filter, 383, 43S Complementary function, 82
phase response of, 387 Complex variables, 481
tables of, 400, 43S analysis, 483
transient response o~, 392 differentiation, 484
Bessel polynomial, 386, 39S- integration, 487
Biquadratic immittance, 305, S03 Computer programs, 18, 438, 439,
Black box, 1S 441, 447 448, 4SO 4S3, 4S7
Bode plots, 221 Constant-resistance network, 3S2, S02
Bounded real function, 419 Contour integration, 487
Break frequency, 22S Controlled source, 268
Bridge circuit, 28S, 3S4 Convergence in the mean, 49, S 1
Butterworth filter, 368, 434, SOO Convolution integral, 197
amplitude response of, 369, 387, 394, Crest factor, 27
soo Critical coupling, 242
pole locus of, 371 Current source, 12
step response of, 391, 394 Cutoff frequency, 17, 22S
tables of, 372, 400, 43S
Damping factor, 22S
. Canonical form, 32S, 400, 433 d-c value, 2S
Capacitor, 13, 103, 176 Delay, 32, 212, 24S, 384
Cauchy integral theorem, 489 distortion, 24S
Cauchy residue theorem, 487 time, 388
511
512 Subject Index

Delta function, see Unit impulse Fourier transform, phase spectrum, 64,
Delta-wye transformation, 257 6S
Differential equation, 75, 14S properties of, 67
forcing function of, 75 symmetry conditions, 68
homogeneous, 75, 76 Free response, 106
integrodifferential equation, 91, 14S Frequency, angular, 2
linear, 76 complex, 4
nonhomogeneous, 75 domain, 14, 134, 367
ordinary, 76 forced, 192
simultaneous, 93, 146 natural, 192
Differentiator, 11, 18, 193 Frequency normalization, 18, 402
Digital computer, 18, 438 Frequency transformation, 18, 404
Distributions, theory of, 470
Dot reference for transformer, 123 Gain contours, logarithmic, 494, 496
Doublet, 40, 479 Gaussian filter, 291
Duhamel superposition integral, 201 Generalized functions, 34, 64, 470
Duty cycle, 26 g parameters, 286
Green's function, 86
Energy density, 71 Gyrator, 272, 287
Energy spectrum, 71
Equal ripple approximation, see Cheb- Hall effect, 273
ysbev approximation High-pass filter transformation, 405
Essential singularity, 486 Hurwitz polynomial, 294, 316 347
Excitation, 1 Hybrid (h) parameters, 260
Hybrid matrix, 261, 449
Faraday's law of induction, 122
Filter approximation, 368, 373, 379, Ideal low-pass filter, 292, 367
383, 493 Ideal transformer, 263, 273, 275, 424
Filter design, 17, 36S, 397, 433, 457 Immittance, 15, 315
computer programs for, 441, 457 Impedance, 15, 176
Final conditions, 106, 109 driving-point, 15, 253, 31S
Final value theorem, 165 matrix, 254
Flux linkage, 122 transfer, 188
Forced response, 106 transformer, 263
Foster network, 321, 32S Impulse function, see Unit impulse
Fourier series, 1, 46, SO Impulse response, 44, 85, 111, 194
amplitude spectrum, S1 Incident power, 416
complex form, 55 Incidental dissipation, 276
cosine series, 53 Inductor, 13, 104, 176
evaluation of coefficients, 52, 58 Imtial conditions, 13, 77, 86 107, 176
orthogonality conditions, 49 Initial value theorem, 16S
phase spectrum, 57 Insertion loss, 429
symmetry conditions, 53 filter synthesis, 431
Fourier transform, 1, 63 Insertion power ratio, 430
amplitude spectrum, 64, 6S voltage ratio, 429
discrete, S6, 63 Integrator, 11, 18, 193
inverse, 64 Integrodifferential equation, 91, 14S
of unit impulse, 65
of unity, 68 Kirchhoff's laws, 100, 104
 Subject Index 513

Ladder network, 279, 346, 347, 453 Node equations, 106, 255
Laguerre polynomial, 48 Norm, 47
Laplace transform, 1, 134 Normaliz.ation, frequency, 402
definition of, 135 magnitude, 402
inverse, 136 Norton's theorem, 180, 185
properties of, 137
table of, 168 Open-circuit impedance parameters,
uses of, 144 254, 343
Lattice network, 285, 354, 503 Optimization techniques, 447
Laurent series, 486 Optimum (L) filter, 379
L-C immittance, 315 amplitude response of, 380
properties of, 315 polynomials, 382
synthesis of, 319 Orthogonal set, 47
Least squares, principle of, 49, 366 Orthonormal set, 47
Legendre polynomial, 381 Overcoupling, 242
Linear phase filter, see Bessel filter Overshoot, 388, 392
Linear system, 8
derivative property, 9 Paley-Wiener criterion, 291
ideal elements, 12 Parseval's equality, SO
ideal models, 10 theorem, 71
Partial fraction expansion, 148
MAC, Project, 448· conjugate poles, 150
Machine-aided design, 448 multiple poles, 151
Magnitude normali7.lltion, 18, 402 real poles, 149
Matrix algebra, 461 Particular integral, 82
definitions, 462 Passive network, 8
operations, 464 Peaking circle, 233, 241
references, 469 Peak-to-valley ratio, 251
Maximally flat response, see Butter- Phase contours, logarithmic, 495
worth filter Phase response, 212, 343
Mean squared error, 48 computer program for, 450
Minimum inductance transformation, evaluation by vector method, 215
443 linearity of, 213, 383
Minimum modulus theorem, 492 Phase shift, 1
Modulation, amplitude, 70 distortion, 213
Monotonic filter, see Optimum filter minimum, 220, 346
Multiple-access computer, 448 spectrum, 3, 57
Mutual inductance, 122 Phasor, S
Plancheral's theorem, 71
Negative impedance converter (NIC), Pole, definition of, lSS
261 Pole-zero diagram, 156
Network, analysis, 7, 100, 175, 253 Port, 12, 253
linear, 8, 100 Positive real (p.r.) function, 16, 299,
n-port, 449 490
passive, 8, 100 Power, average, 301, 315
reciprocal; 8, 100, 2SS available, 418
symmetrical, 259 Propagation constant, 414
synthesis, 290, 315, 341, 397, 431 Proportionality, principle of, 8
time-invariant, 9, 100 Pulse transmission, 389, 392
51.f Subject Index

Q, circuit, 233, 236 Spectra, continuous, 3, 63

line (discrete), 3, 57
Ramp function, 31 Stability, 116, 290
R-C admittance, properties of, 331 marginal, 293
R-C impedance, properties of, 325 Steady-state solution, 106, 109
synthesis of, 329 Steepest descent method, 447
Realizability conditions, 16, 290, 301, Step function, see Unit step
342, 490 Step response, 44, 45, 85, 111, 194
for driving-point functions, 301 Synthesis procedures elementary, 308
for transfer functions, 342 filter synthesis, 397
Reciprocal network, 8, 255 L-C functions, 315, 319
Rect function, 65 R-C functions, 325, 329, 331
Reference impedance factor, 415 R-L functions, 325, 329, 331
Reflected parameter, 415 R-L-C functions, 333
Reflected power, 416 transfer functions, 347, 352
Reflection coefficient, 416, 421, 458 System function, 14, 187, 194
Remainder function, 308
Residue, definition of, 162 Terminals, 12
evaluation by vector method, 162 Thevenin's theorem, 180
Residue condition for two-ports, 344 Time constant, 24
Resistor, 13, 103, 175 Time delay, 32, 212, 245, 384, 388
Response, critically damped, 118 Time delayer, 11
overdamped, 118 Time domain, 14, 134, 366
underdamped, 119 Time-invariant system, 8
Ringing, 388 Transfer function, 14, 16, 188, 266,
Rise time, 388 341, 352
R-L admittance, properties of, 325 admittance, 188, 347
synthesis of, 329 impedance, 16, 188, 347
R-L impedance, properties of, 331 Transformed circuit. 175
synthesis of, 331 Transformer. 122. 1'77
ideal, 263
Sampled-data system, 142 unity coupled, 125
Sampler, 18, 142 Transient response of low-pass filters,
Scattering parameters, 415, 419 388
element, 416
Transient solution, 106
matrix, 420
Transistor, common emitter, 286
Settling time, 388
Transmission coefficient, 421, 457
Short-circuit admittance parameters,
257, 344 Transmission line, 413, 425
Shunt peaked circuit, 248, 400 Transmission matrix, 262
Signal, continuous, 23 Transmittance, 16
decomposition into even and odd Two-port, 12, 16, 254, 264
parts, 21 cascade connection, 271
deterministic, 20 equivalent circuits of, 268, 271
periodic, 20, 46 matrix representation, 264
symmetrical, 21, 54 parallel connection of, 273
Signum (sgn) function, 30 series connection of, 275
Sine function, 65
Singularity, 486 Undercoupling, 241
 Subject Index 515

Undetermined coefficients, method of, Voltage ratio, 16, 188

82 Voltage source, 12
Uniform loading, 278 Voltage standing wave ratio, 428
Unit coupled transformer, 12:i
Unit impulse (delta) functions, 33, 43, y-parameters, 2S7, 344
470 see also Short-circuit parameters
derivative of, 40
properties of, 34, 36, 476, 480 z-parameters, 2S4, 343
Unit ramp function. 31 see also Open-circuit parameters
Unit step function, 28, 474 z-transform, 142
derivative of, 34, 37, 47S Zero, definition of, lSS
Zero of transmission, 34S
 A
C 2
0 m
This Wiley International Edition is part of a continuing pro- -I
gram of paperbound text books especially designed for stu-
dents and professional people overseas. It is an unabridged
(j)
m
0
0
:e
0
z
reprinting of the origin al hardbound edi tion , which is also 0 :::c
"
available from your bookseller. m
0
-f ):Iii
Wiley International Editions 0
z 2
include titles in the fields of:
>
Chemical Engineering Psychology
~
-
en
Civil Engineering Agricultural Engineering
Electrical Engineering and Agricul.ture
Mechanics Biological Science en
Metallurgy Chemistry
Mechanical tngineering
Business Management
Earth Sciences
Mathematics
>
z
Statistics Physics C
en
-<
2
-t
::J:
m
-
(/)
( /)

JOHN WILEY & SONS, INC., U.S.A.

WILEY
TOPPAN COMPANY, LTD., JAPAN TOPPAN
J \\ - l tt.i (I'- 10 )