Network Analysis and Synthesis Second Edition by Franklin F. Kuo Bell Telephone Laboratories, Inc. John Wiley & Sons, Inc., New York | London | Sydney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 material 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. vil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 domait 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 modern 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. Jam 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, T 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Chapter I: Chapter 2: 21 2.2 2.3 24 Chapter 3: 31 3.2 3.3 34 3.5 3.6 37 3.8 Contents Signals and Systems Signal Analysis Complex Frequency Network Analysis Network Synthesis Signals and Waveforms General Characteristics of Signals General Descriptions of Signals The Step Function and Associated Waveforms The Unit Impulse The Frequency Domain: Fourier Analysis Introduction Orthogonal Functions Approximation Using Orthogonal Functions Fourier Series Evaluation of Fourier Coefficients Evaluation of Fourier Coefficients Using Unit Impulses The Fourier Integral Properties of Fourier Transforms Ruan 20 28 33 46 47 48 50 52 58 63 67xiv Contents Chapter 4: 41 4.2 43 44 45 46 Chapter 5: 5.1 5.2 Ree! 5.4 5.5 5.6 Chapter 6: 61 Oe cel 6.4 6.5 6.6 CE 6.8 Chapter 7: Al 72 i. 74 7S 76 Differential Equations Introduction Homogeneous Linear Differential Equations Nonhomogeneous Equations Step and Impulse Response Integrodifferential Equations Simultaneous Differential Equations Network Analysis: | Introduction Network Elements Initial and Final Conditions Step and Impulse Response Solution of Network Equations Analysis of Transformers The Laplace Transform The Philosophy of Transform Methods The Laplace Transform Properties of Laplace Transforms Uses of Laplace Transforms Partial-Fraction Expansions Poles and Zeros Evaluation of Residues The Initial and Final Value Theorems Transform Methods in Network Analysis The Transformed Circuit Thévenin’s and Norton’s Theorems The System Function The Step and Impulse Responses The Convolution Integral The Duhamel Superposition Integral 15 75 76 82 85 91 93 100 103 106 Wd 114 122 134 134 135 137 144 148 15S 162 165 175 175 180 187 194 197 201Chapter 8: 8.1 8.2 8.3 8.4 ce Chapter 9: 91 9.2 _ 9.4 _ 9.6 Chapter 10: 10.1 10.2 10.3 10.4 Chapter II: 77 11.2 7 11.4 11.5 11.6 Contents Amplitude, Phase, and Delay Amplitude and Phase Response Bode Plots Single-Tuned Circuits Double-Tuned Circuits On Poles and Zeros and Time Delay Network Analysis: II Network Functions Relationships Between Two-Port Parameters Transfer Functions Using Two-Port Parameters Interconnection of Two-Ports Incidental Dissipation Analysis of Ladder Networks Elements of Realizability Theory Causality and Stability Hurwitz Polynomials Positive Real Functions Elementary Synthesis Procedures Synthesis of One-Port Networks with Two Kinds of Elements Properties of L-C Immittance Functions Synthesis of L-C Driving-Point Immittances Properties of R-C Driving-Point Impedances Synthesis of R-C Impedances or R-L Admittances Properties of R-L Impedances and R-C Admittances Synthesis of Certain R-L-C Functions xv 212 212 221 229 238 245 253 _. 264 266 271 276 279 290 290 = 308 3s J. 319 325 eg 331 333xvi Contents Chapter 12: 121 12.2 12.3 124 Chapter 13: 13.1 sk 7 13.4 13.5 13.6 13.7 13.8 7 13,10 Chapter 14: 14.1 77 14.3 14,4 14.5 14.6 Chapter 15: 7 15.2 15.3 15.4 Elements of Transfer Function Synthesis Properties of Transfer Functions Zeros of Transmission Synthesis of Yq, and Z,, with a 1-Q Termination Synthesis of Constant-Resistance Networks Topics in Filter Design The Filter Design Problem The Approximation Problem in Network Theory ‘The Maximally Flat Low-Pass Filter Approximation Other Low-Pass Filter Approximations Transient Response of Low-Pass Filters ‘A Method to Reduce Overshoot in Filters A Maximally Flat Delay and Controllable Magnitude Approximation Synthesis of Low-Pass Filters Magnitude and Frequency Normalization Frequency Transformations The Scattering Matrix Incident and Reflected Power Flow The Scattering Parameters for a One-Port Network The Scattering Matrix for a Two-Port Network Properties of the Scattering Matrix Insertion Loss Darlington’s Insertion Loss Filter Synthesis Computer Techniques in Circuit Analysis The Uses of Digital Computers in Circuit Analysis Amplitude and Phase Subroutine A Fortran Program for the Analysis of Ladder Networks Programs that Aid in Darlington Filter Synthesis, 341 341 345 347 =e 365 = 365 368 a 388 392 oy = 402 404 413 413 415 419 426 429 431 438 438 450 453 457Appendix A: Al A2d A3 A4 — Appendix B: “B.A B2 Appendix C: _ 7 _ _ Appendix D: Appendix E: El E.2 = ES E6 ET Bibliography Name Index Subject Index Contents Introduction to Matrix Algebra Fundamental Operations Elementary Concepts Operations on Matrices Solutions of Linear Equations References on Matrix Algebra Generalized Functions and the Unit Impulse Generalized Functions Properties of the Unit Impulse Elements of Complex Variables Elementary Definitions and Operations Analysis Singularities and Residues Contour Integration Proofs of Some Theorems on Positive Real Functions An Aid to the Improvement of Filter Approxi- mation Introduction Constant Logarithmic Gain Contours Constant Phase Contours Contour Drawings Correction Procedure Correction Network Design Conclusion xvii 461 461 462 464 468 469 470 470 476 481 481 483 486 487 490 493 493 494 495 498 502 504 505 509 Silchapter I 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. 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 or 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 Excitation Response ——>|_ Network {+—> FIG. 1.1. The objects of our concern. 1Signals and systems 3 =@ =w4 wy 02-01 fog Wy wy wy FO FIG. 1.4a. Discrete amplitude spectrum. wy wn ws % =e 04-03 502 01 +o FIG. 1.4b. Discrete phase spectrum. let the angular frequency w be the independent variable. In this case, the signal is described in terms of Ag, @, and 9, as shown in Fig, 1.3a, where amplitude is plotted against frequency, and in Fig. 1.30, where phase shift is plotted. Now suppose that the signal is made up of 2n + I sinusoidal components s(t) = ¥ Asin (ot + 6) (13) The spectral description of the signal would then contain 2n + 1 lines at £0, £0... , -£0,, as given in Figs. 1.4a and b. These discrete spectra of amplitude A versus « and phase shift 6 versus « are sometimes called line spectra. Consider the case when the number of these spectral lines become infinite and the intervals «,,, — «, 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. 1.5a and 5. In the continuous case, the sum in Eq. 1.3 becomes an integral (0) -f A(o) sin [ot + ()] do (1.4) where A(o) is known as the amplitude spectrum and 6(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4 Network analysis and synthesis A(w) -o 0 o FIG. 1.5, Continuous amplitude spectrum. 9.9) FIG. 1.5b. Continuous phase spectrum, pulse in Fig. 1.6 can only be described in terms of continuous spectra through the Fourier integral transform. 12 COMPLEX FREQUENCY In this section, we will consider the concept of complex frequency. As we shall see, the complex frequency variable s=o+jo (5) is a generalized frequency variable whose real part ¢ describes growth and decay of the amplitudes of signals, and whose imaginary part jo is angular frequency in the usual sense. The idea of complex frequency is developed by examining the cisoidal signal S(p) = Ae (1.6) s(t) 0 T t FIG. 1.6, Triangular signal.Signals and systems 5 Ims| jo FIG. 1.7. Rotating phasor. when S(t) is represented as a rotating phasor,’ as shown in Fig. 1.7. The angular frequency « of the phasor can then be thought of as a velocity at the end of the phasor. In particular the velocity @ 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 p as given in Figs. 1.8a and 1.8). In this case, if the velocity is given by the symbol s, we see that sis composed of a component « at right angle to the phasor S as well as a component o, which is parallel to S. In Fig. 1.8a, s has a component —o toward the origin. As the phasor 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) Im (1) le cos wt je sin wt (7) FIG. 1.8. (a) Rotating phasor with exponentially decreasing amplitude. (6) Rotating phasor with exponentially increasing amplitude. +A phasor S is a complex number characterized by a magnitude and a phase angle (sce Appendix C).6 Network analysis and synthesis Re S| A FIG. 1.9. Damped sinusoids. Im s(t) / /~ Envelope = Ast N. \ FIG. 1.10. Exponentially increasing sinusoid.Signals and systems 7 0 FIG. 1.11. Exponential signals. which are shown in Fig. 1.9. Note that the damped sinusoid has an exponential envelope decay, Ae~**. In Fig. 1.8, the phasor is shown with a positive real component of velocity +. Therefore, as the phasor spins, the amplitudes of the real and imaginary parts increase exponentially with an envelope Ae**, as shown by Im S(¢) in Fig. 1.10. From this discussion, it is apparent that the generalized cisoidal signal S(t) = Act = Aes" (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 joo = 0, the signal is an exponential signal S(t) = Ae** (1.9) as shown in Fig. 1.11. Finally, if ¢ = jo = 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 (8) the principle of proportionality hold. By the superposition principle, if, for a given network, [e,(), r,(t)] and [e,(t), ra(t)] are excitation-response pairs, then if the excitation were e(t) = e(t) + e,(t), the response would be r(t) = n(f) + r,(t). By the proportionality principle, if the excitation were Cye,(t), where C, is a constant, then the response would be C,r,(0), i.c., the constant of propor- tionality C, 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 passive? if (a) the energy delivered to the network is nonnegative for any arbitrary excitation, and (6) 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 Creitt) Cars(t a4W4| pa System Crealt) Carett) ‘System n(t) + Ci Crest) + Cael | ayy, [ETA Cora) FIG. 1.12, Linear system. *G, Raisbeck, “A Definition of Passive Linear Networks in Terms of Time and Energy,” J. Appl. Phys., 25 (Dec. 1954), 1510-1514,