AN INTRODUCTION TO THE LAPLACE TRANSFORM AND THE z TRANSFORM A.C. GROVE The rapid growth in computing power has meant that discrete mathematics and hence the z transform have become increasingly important. This basic introductory text covers the techniques and applications of Laplace and z transforms, which is prerequisite knowledge for courses in all engineering disciplines, IT, applied maths, computer science and the physical sciences. Bias rele) contains numerous worked examples to support the theoretical material, and end-of-chapter exercises is oriented towards engineering applications includes the necessary theory, but without going into in-depth mathematical detail. Tony Grove, senior lecturer in mathematics, Nottingham Polytechnic, UK, has based this book on his course of lectures given to engineering students at the Polytechnic, COVER BY RACHAEL ADAMS DESIGN ISBN O-13-488933-9 vii ND | | 1 | | 9 "78013: DocentAn Introduction to the Laplace Transform and the z Transform A. C. Grove Senior Lecturer in Mathematics, Nottingham Polytechnic Prentice Hall New York « London - Toronto - Sydney - Tokyo - SingaporeFirst published 1991 by Prentice Hall international (UK) Ltd 66 Wood Lane End, Hemel Hempstead Hertfordshire HP2 4RG A division of Simon & Schuster International Group © Prentice Hall International (UK) Ltd, 1991 All rights reserved. No part of this publication may be reproduced, stored in a retrieval syste OF transmitted, in any form, or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior permission, in writing from the publisher. For permission within the United States of America contact Prentice Hall Inc., Englewood Cliffs, NJ 07632. ‘Typeset in 10 on 12 point Times by MCS Ltd, Salisbury, Wiltshire, England Printed and bound in Great Britain by BPCC Wheatons Ltd, Exeter Library of Congress Cataloging-in-Publication Data Grove, A. C. (Anthony C.), 1937— ‘An introduction to the Laplace transform and the z transform / by AC, Grove. Pp. cm. Includes bibliographical references and index. ISBN 0-13-488933-9 1. Laplace transformation. 2. Z transformation. 1. Title. QA432.G76 1991 $15',723—de20 90-22912 CIP British Library Cataloguing in Publication Data Grove, A. C. (Anthony C.) 1937— An introduction to the Laplace transform and the z transform. 1. Mathematics. Transformations 1. Title ‘$15.723 ISBN 0-13-488933-9 12345 95 94 93 92 91Contents Preface 10 The Laplace transform and the inverse Laplace transform The transforms of derivatives and the application to differential equations Some useful theorems The Heaviside step function The impulse function Transfer functions, block diagrams and stability Convolution Sampled-data control systems and the z transform The inverse z transform Solution of difference equations Discrete transfer function vii 33 46 53 65 3 89 95vi Contents Appendix 1 An introduction to the definition of the Laplace transform Appendix 2 A table of Laplace transforms Appendix 3 A table of z transforms Appendix 4 Types of sampling Answers to exercises Bibliography Index 99 103 cae 114 M7 127Preface This book has arisen from a course of lectures given to engineering students and is intended to assist students in the use of the Laplace transform and the z transform. The background mathematical theory is kept to a minimum. Laplace transforms and z transforms are convenient methods for solving differential and difference equations respectively; both are amenable to computer implementation. Some indication is given of possible origins of these equations but the main purpose is to discuss methods of solution. However, readers who wish for a deeper mathematical approach should consult one of the more extensive texts. 1 would like to thank my colleagues Mr P. W. Moore and Dr A. Sackfield for their helpful criticism and encouragement, Prentice Hall for their sympathetic cooperation, Mrs A. Fullerton and Mrs S. Mohamedali for conscientiously typing the original manuscript and my family May, Allison and David for all the support and interest. vil‘This page intentionally left blank1 The Laplace transform and the inverse Laplace transform Introduction ‘One of the most efficient methods for solving certain ordinary and partial differential equations is the use of Laplace transforms. The effec- tiveness of the transform is its ability to convert many differential equations into algebraic equations. Although other methods provide sol- utions of equations where the input or forcing function is an exponential and/or sine function the Laplace transform is able to deal with other inputs such as pulses, square waves, point loads, etc., in exactly the same way. The differential equations that arise in the analysis of control systems are particularly suitable for solution by this means. The Laplace transform is defined in order to develop the necessary techniques but a more detailed introduction is given in Appendix 1. DEFINITION Multiply a given function of time f(r) by e~™ (s is a parameter) and integrate the product between zero and infinity. The result, if it exists, is denoted by ¥[/(t)} = F(s) and is called the Laplace transform of f(¢), i.e. Fos) = efi) = J" espe at Some examples are as follows. 1 fO=1 a= \° eo dt= (S\r-2 o -sjo $s provided that s> so that e~" +O as t+ 2,2 Laplace transform and z transform 1 sta 2 wemyn Tete are [7 et are provided that s+a> 0. 3 #ith= \, te" dt -[-fe"]}+ [bev ar s ° oS using integration by parts. Hence gin=|-4e* = provided that s> 0. = os 4. H{sin(at)} = \ e°" sin(az) dt 0 ee ; « = (for {-s sin(at)- a costar | " a Sta These transforms, and others, constitute a table of transforms which normally consist of a combination of general theorems and the trans- forms of specific functions, Some of the more common results are given in Table 1.1 but a more comprehensive table is presented in Appendix 2. Note that, since the Laplace transform is defined as an integral, the normal rules of integration apply; also the Laplace transform of a function may not exist. Thus, if K is a constant, aan = | Kem sea ae =K{ e-" sit) dr = K2f()) = KF(s)Laplace and inverse Laplace transforms 3 Table 1.1 Some Laplace transform pairs Ft) Zs = j ef) dt = F(s) ° K Kis ' Ys? e an =a 1 é —o Sta sin(wr) cos(wr) sinh(wr) cosh(wr) Kf) KF(s) f+ 8) F(s) + GOs) Also ALO + gC = ALO) + Ala) which is the linearity property, but ASO RO) 4 ALOR} Functions such as ~' and tan(‘) do not have Laplace transforms since the integrals diverge. The following example illustrates the use of Table 1.1. EXAMPLE 1.1 Determine the Laplace transform of (ay 42643 (b) 4 sin(3r) + cosh(Sr) fc) cos*(r)4 Laplace transform and z transform a) Bie + 24 3] = FI} + 2710 + 413) 2 1 Sat2st so st 4 s (by ¥14 sinGr) + cosh(S1)} = 42{sin(r)| + #{cosh(S1)} ©) Fie0s*(} = LE [1 + cos(2e)]} = LIS) + #1} cos(2s)} 1 s a 2s As? +4) EXERCISE 1.1 Use standard formulae to derive the Laplace transform of the following functions: 1 51-2 2 atbi+er 3 P+8e 4! 4 2? 6 5 2+4e" 6 etteh 7 asin(at) +b sin(br) 8 cos(at) + cos(br) 9 51744 cos(3r) 10 simh(3r) 11 cosh(3/) 12 fF cos(2r) — cosh(4t) 13 e~* cos(2t) 14. fe —e-* sin(e) 15 (e'—e°')? and hence sinh7(1) 16 &*! cosh(t) 17 cos(at— a) 18 sin(or+ a) 19 sin(r)cos(3/) W cos(2r) cos(¢) 2 sin2(1) where a,b,c and @ are constant. In questions 19, 20 and 21 use a suitable trigonometric identity.Laplace and inverse Laplace transforms 5 The inverse Laplace transform In practice it is important to be able to recover f(r) from its Laplace transform. It can be shown that f(r) is uniquely determined by F(s), If F(s) = %(f()} then f(r) = %"'[F(s)] where #7" denotes the inverse Laplace transform. For example 7 i a £ fata =} sin(2r) Rarely do the functions of s appear in standard form. In general F(s) is a ratio of two polynomials in s which can be expanded in partial frac- tions provided that the degree of the numerator is less than the degree of the denominator (a proper rational function). EXAMPLE 1.2 Determine the inverse Laplace transform of 25+3 &) s-6 © Fyaste &6-d ©) 45-9 6-5-2 wif 2543 )_ vif 2543 w frees 2 ical on completing the square. Then = 25+3 -1f_2ls + 2) = 1 6) 8 ee (5 “ (233) ar} ( = 2e"™ cos(3s) — fe™2" sin(3r) @ eft yews (s- 1I)(s- 2), s-l s-2 by partial fractions. Thus va : | =e" — ae (s— Is 2) 4s-9 5 gertf ASF) ay © aaa # (5 G1? by partial fractions. Thus a5) =e + Ste’ e6 Laplace transform and z transform EXERCISE 1.2 Determine the inverse Laplace transform of each of the following functions of s: 1 * 2 = 3 $e 4 A mea s 1 wy 8 + aes . " GaE-5 2 TP 13 am “4 aoa 15 16 aE ” 7d 18 rarer ® pigEe * Bes a , (s? + Dts? + 4)2 The transforms of derivatives and the application to differential equations Introduction In the process of formulating practical problems by mathematical equations, i.e. mathematical modeling, many kinds of differential equations can arise. Examination of such equations has led to the con- clusion that there are certain definite methods by which many of them can be solved. It is not the intention to provide detailed information on mathematical modeling here but to emphasize the similarity between many different applications and to discuss a method of solving some of the equations. A number of these applications will be used in subsequent sections. In many instances the mathematical model is obtained from scientific laws. Many of the general laws of nature find their most natural expres- sion in the language of differential equations. These equations have to be solved to give expressions relating the variables involved without the use of derivatives. It is important, however, to realize that in obtaining the mathematical model it may be necessary to make certain assump- tions. For example, it is sometimes assumed that resistance to motion is proportional to velocity — a valid assumption under certain conditions The derived equation and its solution must therefore be used under the same conditions. It is also important to note that different statements (laws) give rise to similar types of differential equations. The equations of some typical examples are given on p8.8 Laplace transform and z transform Electrical circuits The application of Kirchhoff’s law to an LCR circuit yields the differen- tial equations di Oy Rit de i= Lat Rit Gre i dq dt which on elimination of é gives i d L a4 R ct = =E or on elimination of g gives i aE dt Circuits with several loops generate simultaneous differential equations qi di Latta eabeRn=E Lp dis 1 p=rtheo cP a h=atis Particle dynamics If a particle of mass m is suspended from a fixed point by a spring of constant & then the equation of motion is 2 m ae = -kx where x is the displacement from the equilibrium position. This equation is derived by the application of Newton's second law assuming that the force exerted by the spring is proportional to the extension (Hooke's law). The inclusion of damping, assumed to be pro- portional to the velocity, and external forces generate further terms in the equation. Similarly the motion of several connected particles can be described by a set of simultaneous equations. The motion of a particle projected from the surface of the earth can be described, relative to suitable three-demensional axes, by theThe transforms of derivatives 9 equations atx rr 2w singe) 92 tye —w sine) 9% 2 costo) SE ve 2w costo) 2 — g Servomechanisms The error between a required angle @ and an actual angle @ in a radar control system is e=O-¢ If the correcting torque is assumed to be proportional to the error then the differential equation is of the form a6 dr The negative sign is required because the torque must oppose the error. - k(@-9%) Biology and chemistry Differential equations describe, for example, a mixture of chemicals, radioactive decay, population growth, epidemics, predator-prey relationships and Newton's law of cooling. Typical equations are as fol- lows: ST _ K(r-t) 6 T= KO-e) dx dy_ dt 3x-—2y ar ax+3y10 Laplace transform and z transform Structures The bending of a beam can be modeled by the use of one of the following equations, depending on the loading. d*y dty_ EITM ere a a’ EIT 3+ Py=0 EI at ky =f) Economics Commodity prices depend on quantities such as supply and inflation factors. The resulting differential equations are of the form dP ds a7 7 FO KS So) ar 7 hKP— Po) A linear second-order differential equation with constant coefficients is thus of the form ax dx ato ag taaso The transforms of derivatives The transforms of the first, second, third and fourth derivatives are given in Table 2.1 and are numbered 2, 3, 4 and 5 respectively in Table 2.1 Laplace transforms of derivatives fey F(s) 2 yo) sF(s) = f() s YW) S*F(s)~ sf) -F' (0) 4 Yi) s' F(s) — s*f(0) — 5f*(0) — f*(0) “ ap YO s*F(s)— s*f0) — s*f'0) — sf") — Ff" ()The transforms of derivatives 11 Appendix 2. Note that F(s)={f(f)] as usual and that (0), £0), f"(0), ... are the numerical values of f(t), f'(0), f(s. when f=0, where a prime denotes differentiation with respect to /. These results are obtained from the definition of the Laplace transform and applying the rule for integration by parts. For example a -(Tend og rw] = Joe" § Yor a = [enn] \-- sept) dt 0 o = =H) +5 |” s0 ae a provided that e~“f(t)+ 0 as f+ oo. Thus al ise] = = f(0) + s#F0) = sF(s)- f(0) Similarly : - «lf iso} = [oe 5 or ar [e“r'e| vt sf or 4 LA) de " i d - $'0)+ a(t noi} provided that e~“J'(1)— 0 as 1 «0, Hence 2 [Se tren} =~ s°@)+ sis21s4 £0) = 8 LUFC — sfO)- f'O) = 8" F(s) = sf(0) - f'(0) If f(¢) is a known function of ¢ then the results may be used to yield further standard formulae. In the solution of a differential equation such as dx, dx it a 3 a Sx = sin(21)42 Laplace transform and z transform x is an unknown function of ¢ which we are attempting to find. Its Laplace transform is therefore an unknown function of s, denoted by ¥ or X(s). Hence Function Laplace transform x For X(s) & sox ar ‘0 d?x = SR —sX0 — xd where xo and x¢ denote the initial values of x and da/dr respectively. These expressions are algebraic expressions in s; the transformed equation will thus be an algebraic equation in s which can be solved for &. The solution x in terms of ¢ is then obtained by inversion. Method 1. Transform the differential equation using all the necessary rules and standard forms. 2. Solve the resulting equation (usually algebraic) for the unknown as a function of s. 3. Invert the function of s to obtain the required solution. EXAMPLE 2.1 Solve the equation dx, dx t3—4+2x=0 att ag EO given that x= 4, dxjd¢= —3 when s=0. Transforming the equation ax ¥ ay 2¥{x\ = £10) {a} (3) txt = #10) (sR — 8x9 — xd) + (SR — Xo) +2¥=0 and since xo = 4, x/= —3 the equation can be solved for ¥ Hs? 43s4¢2=454+9 4s+9 4 (s+ 2)(s+1) (s+) (s+2)The transforms of derivatives 13 by partial fractions. On inversion this gives xs Se"'- EXAMPLE 2.2 Solve the equation ax dx at a taxee™ Since no initial conditions are given it is necessary to retain xo and xi as unknown constants. Transforming the equation (°F sq) + sk — 19) +28 = st+2 1 SXo +6 + 3X0 (s+ 2)(s? +3542) st43s42 Lt Xos(s + 2) + (00 + 3.xo)(s + 2) R= (s+ s+ 2)? since s? +35 +2=(s+1)(s+2). In partial fractions this is eA B c 2=—— + + S+i s#2 (s+2)' from which A=14+2X0 + x6 B= -x§-xo-1 C=-1 Inverting, x= Aen'+ Be?! — te" If non-initial conditions are known such as x= 1, dx/d¢=3 when ¢=4, then the constants A and B can be determined by substitution. Simultaneous differential equations The solution of a set of simultaneous differential equations is achieved using the same procedure. The transformed equations are again nor- mally algebraic and can be solved by elimination or even matrix methods,14 Laplace transform and z transform EXAMPLE 2.3 Solve the simultaneous equations dx dr dy = -3y-Sx=0 2.2) anor St 2.2) given that x=0, y=0 when f=0. +3x+5y Qn The transformed equations are SE+3R+5P= + stl s¥-37-SX=0 F(s43)45p=—— Q3) s#l —SE+(s—3)P=0 Qa Eliminating ¥ by the operation 5 x equation (2.3) + (s + 3) x equation (2.4) gives which on inversion gives y= Sen's § sims) — § costar) Equations (2.3) and (2.4) can now be solved for ¥ and a further inver- sion yields an expression for x in terms of #. A more convenient method is to note that the second of the original differential equations gives sx -3y 2 which on substitution for y and dy/dr from above gives in(4e) x= — Se '+ § costar) + & sin(4r) Sx=—Be'+ Hf cosa) +B This method of substituting into one of the original differential equations in order to avoid the necessity of inverting two separate functions is possible provided that one of the equations contains only one derivative, not two (i.e. either dx/dr or dy/dr but not both). EXAMPLE 2.4 Solve the simultaneous differential equations (2.5)The transforms of derivatives 15 dx, dy _ stg tayte=0 26 given that x=0, y= 0 when s=0, Note that both equations contain dx/d/ and dy/dé. In this case generate a third differential equation by cli ing either dx/dr or yd (free choice), For example, eliminating dyfdr by multiplying equation (2.6) by 2 and subtracting equation (2.5) gives 5s 6y—x=-e! an dt which will give y directly provided that x can be found. Returning to the original differential equations (2.5) and (2.6) and transforming we obtain (SX — xo) + As) —yo) +32 =— s+l 3(S¥— Xo) + (SP — yo) + 3F + ¥=0 Rearranging we have (5439842572 2.8) s+ G54 1)F+ (5+ NP=0 a» Now eliminate F (we need £) by multiplying equation (2.8) by 5 +3 and subtract equation (2.9) multiplied by 2s: Rls +9)? 2505+ 9) = 522 sel ee (s+ (Ss? + 45+ 9) a —(8+3) “(s+ I(5s— 9s +1) 30 6 1 =-—> + + 49Ss—9) 49541) 7s +1)? Inverting, we obtail wah! s setts et This result can then be used to obtain y by substitution into equation @7) ‘aici awe — Bett! te! >16 Laplace transform and z transform EXAMPLE 2.5 If an electron is projected into a uniform magnetic field perpendicular to its direction of motion its path is given by my. Hedy —, d'x Hedy as? e dt "Gee dt where m,c,H and e are constants. If x=0, dx/dt =, y=0 and dy/dt=0 at #=0, determine x and y in terms of ¢. The transformed equations are amis*9 ~ sv» —siy= ~ BE gx —xe) 2.19) amis? ~ sx xy =H (s¥— Yo) QD Inserting the initial conditions and rearranging gives He spa sx=0 (2.12 em i sp-s*k¥=-u (2.13) cm Now eliminate ¥ 2 ria spe () ¥=0 (2.14) om om, y= (2.18) == # s? + (Hejcm)* Therefore ate in (# ‘) “He *"\em Similarly ja 1~ coo He ) He cm EXERCISE 2 Solve the following equations:The transforms of derivatives 17 dy dr given that y=2 when 1=0. +4y=1 given that y= 2, dy/d/=0 when #=0. 2 Gara Ze ye sina given that y=3, dy/d¢=1 when ¢=0. ot _4r=0 ag” given that 7'= Ty when 6=0 and p and To are constant. 5 oe x samo given that r= 1, dr/d@ = 2 when @=0. 2 2X 6x=32 dr given that x=3, dafd?=2 when ¢=9. ax. dx sz+5 +4x=0" dr de given that x=0, dx/d¢=—2 when 1=0. dx dy dx dy S+2x+—=0 +2 3s ae ara given that x= —1, y=2 when ¢=0, dx dy : dy =,o =3 sin(2i oS 4y=0 arate? 3 sin(2r) aot Ly given that x=0, y= 0 when s=0. dx _ dy sista) os sta (s+a)?" (s+a)? in partial fractions. The constants A, B,C and D are evaluated and each term is then inverted.Some useful theorems 23 EXAMPLE 3.7 Solve the differential equation eo eRiet ire for the current (in a circuit if L,C,R and E are constants and /=0 when r=0. Transforming the equation Losin) + Rie hE i(us+ +4) == cs) 5 j-—£ __ (Ls? + Rs + WIC) wefel (s+ 2V (i se) 2L, i a) If we assume that R? < 4L/C, a= R/2L and p* = ILC = R7/4L? then 87> 0 and the solution on inversion becomes ms e- sin( ar) However, if R? > 4L/C, a= R/2L, y?= R7/4L? — ILC then 47 > 0. Hence soa Ll(s+a)?~ 77] and the solution becomes i ee == 1 i la e-* sinh(yt) Differentiation of transforms If F(s)= £1) then d 2(-SO = 5 LFS and more generally "ft)) = 2 = @U- OPO) =a FGM Wa24 Laplace transform and z transform Note also therefore that eren=-t 2" re} If (1) fle) has 2ft ro} = { F(s) ds or in inverse form limit as ¢— 0 from the right, then 2 FS neff Fs) as] EXAMPLE 3.8 Find #{é? sin(2e. 2 218 sina =(—1)* 4 (2tsin2n}) 2 (2 2 Ns? 44, (+4)? EXAMPLE 3.9 Find » if 1 2 = == (e7!-¢') == sinh 1 e) 7 taht)Some useful theorems 25 EXAMPLE 3.10 Find toe a ot pu -="() from Figure 3.1, Figure 3.1 EXAMPLE 3.11 Find y if ¥=sf(s? — 1)?. ree (aa el) gf) we ‘es sinh(r) t 426 Laplace transform and z transform EXAMPLE 3.12 Solve the differential equation aa Bian GD ar a : given that x=0, dx/d¢= A when £=0. Note that all differential equations considered previously have involved constant coefficients. In this case the method remains as follows: transform the equation, solve the resulting equation and then invert. Hence, transforming equation (3.1), d d =F (SF — ste - ad) — a — (S¥- xa) + aF=0 ds ds Since x =0, xd= A, de 4 = gy (PEA AY~ a 5 (sk) + oF 0 The terms s*¥ and sX are products of functions of s and must be differentiated as such. —23%-s? Go(res =) +a¥=0 SF (Ws? as)- 25-0 ds S (s+m+28=0 G.2) ‘The resulting equation (3.2) is therefore another differential equation but in ¥ and s which may or may not be easier to solve than the orig- inal equation (3.1). In this case the variables ¥ and s can be separated. dx_ lds RB sta In(X) = — 2 In(s+a)+K ka —— (s+eP where C= In(X). Inverting, the solution is x= Cte"Some useful theorems 27 Periodic functions (9) Suppose that f(t) is a periodic function of period T so that S(e+ T) = f(t) for all values of t. Then ein = J" ene ae T 2r = \ er) drt j ee" f(t) dit a Tr ar + j en" f(t) dt + -- os ¥ ¥ =| emp) ars et en p(n) dt lo 0 r sett ° Tr =Ute Tey] j en "f(t) dt ° Since 1+e7*7 +e°%7 +... is an infinite geometric series, its sum to infinity is 1/(1 — e-*") provided that e~*7 < 1. The transform of f(t) is then given by PU = paar J eH ar EXAMPLE 3.13 Determine the Laplace transform of the square wave defined by E Ogr< 7/2 t= mw {et T2T i.e. the Heaviside function shifted by 7 units. The function is defined such that the equality appears once and once Figure 4.1. The unit step function y= u(t)#4 = Laplace transform and z transform Figure 4.2. The delayed step function y=u(t—T) (b) Figure 4.3. (a) The function y=f(1); (b) the function yaf(u(r=T)The Heaviside step function 35 only, From the graph it can be seen that an alternative name is the ‘unit step function’ which may be interpreted as a model of a switch that is switched on at time f= T. The combination of the step function and any other function /(1), i.e. SWule-T) is clearly a function f(¢) multiplied by 0 if ¢< 7,1 if ¢> T, a fune- tion which is equal to 0 if ¢< T and f(t) if ¢ > T. This combination is shown in Figure 4.3. For T > 0, the Laplace transform of the unit step function is Lult- T= ‘ e“u(t—T) dt = {.° dr+ (. en" de aT s This result is listed as 11 in Appendix 2. EXAMPLE 4.1 Sketch the function y =3u(t— 1) — u(t - 2) and determine its Laplace transform. Figure 4.4 The function y = 3u(r— 1) — 2u(t—2)36 Laplace transform and z transform The function represents a step of magnitude 3 at ¢= 1 followed by a Step of magnitude —2 at f= 2. The sketch is therefore that shown in Figure 4.4. # (Bult — 1) — u(t — 2)) = 3H felt — 1) ~ 24 tule — 2) et de?! “os s using the above result. EXAMPLE 4.2 Determine an expression for the function $ agr2 can be written as e~*[1 — w(t -2)] and hence find #[/()}- 13 Express the following in terms of Heaviside’s unit step function: ft O i elsewhere40 Laplace transform and z transform 19 A square wave is given in the form E o«i0. 3. u(—t)=1-— u(t) (Figure 4.5). (b) Figure 4.5 (a) The step function y =w(s); (b) the function youl-t)44° Laplace transform and z transform 4. u(T-it)=u[-(@-T)] 5. Scaling: u(ar) = w(t) i.e. the step at ¢=0 is independent of a; u(at—T)=u(t— T]a) i.e. step occurs when af = T, ¢= T]a. t-u(t- 7). It may be possible to sum an infinite series after transforming each term. For example in transforming a square wave function or a staircase function the resulting series is a geometric series as in Chapter 3. The sum to infinity of a geometric series is Ssa+ar+ar+ar+--=a/(l—r) provided that |r| < 1. EXERCISE 4.2 Find the inverse of the following transforms; 1 2 te. +9 (s—2)(s?+4) 3 Sat a ‘ SS os “2 5 Ten: 6 SS an876 as: aS + = 3)* is See s(s? +1) Solve the following differential equations: ae . <5 -45~si=30u—1 aed wo given that /=0, difdr=6 at 1=0. d*x dx mn 843 pve ue-2 ith g texeue-2 given that x=0, dxfd/=0 at 1=0.4 The Heaviside step function 45 dx a Sexasn where f(t) = 2, 0<4<3, and x=1, dx/d¢=0 at 1=0. dty ae? given that y= —1, dy/dr=1 when r=0. +4 Rady Be" Malt = 1) The voltage E applied to an electrical circuit is given by pul O 0. 10 A light beam of length / rests horizontally on supports at each end and carries a load W at a point }/ from the left-hand end. Solve the equation dty ( ‘) EI —* = W5(x-= dt “3 subject to the conditions y=0 at x=0 and at x=/ and at each free end d*y/dx? (the bending moment) is zero. ILA circuit is subjected to an infinite series of impulsive voltages of magnitude £o at f=0,1,2,.... If the current in the circuit is given by ai Gud +3i= determine i in terms of f if /=0, difdt=0 when s=0.6 Transfer functions, block diagrams and stability Transfer functions In the study of equations representing physical systems, it is necessary to consider the relationships between the input and output variables of the system. Consider a system governed by the linear differential equation 2 ase +o Ee cx= se) 6.1) where a,b,c are constants and all the initial conditions are zero. In a stable system the roots of the auxiliary equation must have negative real parts so that the ratio 5/a is positive. Taking Laplace transforms of both sides produces the algebraic equation (as? + bs + c)¥(s) =f(s) where s is the Laplace transform variable and X(s),7(s) are the Laplace transforms of x and f(t) respectively. The function as? + bs + ¢ is called the characteristic function and the equation as? +bs+c=0 is the characteristic equation. Since the transformed equation is algebraic, then X(s) _ 1 Js) as? + bs +e54 Laplace transform and z transform 1 Yis)« Je) ("Ty bste ats) Figure 6.1 A block diagram The function on the right-hand side is the rransfer function relating ¥(s) and f(s) and is the ratio Laplace transform of output Laplace transform of input The usual notation for a transfer function is ¥(s),G(s). A block diagram (Figure 6.1) can be used to represent such a relationship. A transfer function can also be obtained by one of the following methods. Trial solution method Assume a particular integral x= Ae’*‘. By substitution into equation (6.1) we obtain an 1 SO) alja)* + b(ja) + € The function on the right-hand side is called the transfer function of the system and is the ratio of output to input. This form will be encountered in the discussion of the Fourier trans- form. D operator If D=d/dé and D? = d?*/de* then equation (6.1) can be written (aD? + 6D+e)x= fi) so that = el SW) aD? +bD¥e Note the similarity of the transfer function using s, jw or D.Transfer functions, block diagrams, stability 55 Simple block diagrams A system governed by a differential equation can be represented dia- grammatically by a block, or combination of blocks, each consisting of a transfer function ¥(s) with input f(s) and output X(s). Consider the system of an amplifier, a d.c. motor and a load shown in Figure 6.2. To facilitate the writing of the equations the system can be subdivided into smaller parts as illustrated. The equivalent block diagram can then be obtained: a a ris) (6.2) ze Y2(s) (6.3) F~ ys) (6.4) i Bo. Yas) (6.5) T Figure 6.2. A typical system56 Laplace transform and z transform Cr—>— Vs) = Yul s) ¥2(S)¥3(s)¥a(s) an Figure 6.3 Figure 6.4 The error sensor By substituting among these equations the intermediate variables can be eliminated and the relationship between the Laplace transform of the output (6) and the Laplace transform of the input (8) can be obtained. G5= ¥a(s)T= Vs(s)¥3(s)ir= = Therefore Se ¥4(s)¥a(s)¥a(O)¥i (5) 6.6 1 Hence the transfer functions of blocks in series are multiplied and the system can be represented by a block diagram as shown in Figure 6.3. In practice it may be necessary to compare the output @ to a par- ticular value or set of values, say 4i, so that suitable correction voltages € are applied to the amplifier, The equation for an error sensor is then &=6;- 65 (6.7) the block diagram of which is shown in Figure 6.4. Typical examples 1. Consider the system shown in Figure 6.5. From (6.6) and (6.7) above the equations are fm YuV¥u6s) B=. — BoTransfer functions, block diagrams, stability 57 Figure 6.5 Substituting for & we have 8, = Yi(s)¥2(s)18i — G0} Bo(1 + ¥i(s)¥2(s) = Yi(s)¥2(s)Bi Therefore do __ Vils)¥a(s) i b+ ¥i(s)¥2(s) 2. A system (Figure 6.6) may also have a transfer function in the feed- back loop. The equations are by 7" ¥i(s)¥2(s) = ¥3(s) é=6,-x PtH By substitution the overall transfer function becomes 65 = ¥i(s) ¥2(s) 6 1+ ¥i(s)¥a(s)¥s(s) Figure 6.658 Laplace transform and z transform & Figure 6.7 A CR network ea Figure 6.8 An operational amplifier For purposes of analysis transfer functions can be represented and studied by means of CR networks (Figure 6.7) EXAMPLE 6.1 Determine the steady state gain of the system whose transfer function is 100(1 + 25) (s? + 85 + 6s +3) when the input is u(f), the unit step function. If the output is y(r) then 100(1 + 2s) (s? +85 + 6s + 3) _ 1001 +28) (s? + 85+ (s+ 3) 5 ¥(s)= # lucyTransfer functions, block diagrams, stability 59 The steady state gain for unit input is defined by limit y(1) = limit s¥(s) (by the final value theorem) ie 0 seus 100(1 + 2s) = limit ——————_ sso (s°+85+6)(s+ 3) 300 18 Steady state error The prime reason for using feedback is to minimize the error between the actual system output and the desired system output. The magnitude of the steady state error, the value to which the error signal tends as the transient disturbance from any change of input dies away, is a measure of the accuracy of the system. Consider the system defined in Figure 6.9. The equations are Go= Vils)é £=5)— ¥2(s}00 and by eliminating 9, we obtain 6; “T+ Wi(s)¥05) By definition é steady state error = limit e(f)= limit SE ime 7 by the final value theorem. Therefore ii sj steady state error = limit TyKwne Figure 6.960 Laplace transform and z transform Poles, zeros and stability Let us consider the general form of a transfer function G(s) given by bo + bis + bis? ++ Nis) @o + a5 +038" te Dis) If N(s) and D(s) are of degree k and degree m respectively then they can be factorized into A factors and m factors respectively. The constants involved may be real or complex. The roots of M(s)=0 are called the zeros of the transfer function G(s), whilst the roots of D(s)=0 are called the poles. The power of each factor determines the order of the pole or zero. The stability and corresponding response of a system can be determined from the location of the poles. Gs) = EXAMPLE 6.2 s(s= I)(s— 0.5) (s+? +541 The zeros are at s=0, s=1, s=0.5. The poles are at s= —1 and s= —1/2 + jJ3/2. Since G(s) 0 as 5 +0, G(s) is also considered to have a zero at infinity. Since s may be complex all finite values of poles and zeros can then be plotted on an Argand diagram (Figure 6.10) where x and © mark the poles and zeros respectively, G(s)= Im(s) 1 Figure 6.10Transfer functions, black diagrams, stability 61 An output function defined by ¥(s) is a combination of two func- tions, i.e. ¥(s) = G(s) X(s) where G(s) is the transfer function of the system and X(s) is dependent on the input to that system. Hence the poles of Y(s) can arise from I, the poles of G(s) and 2. the poles of X(s). The portion due to the poles of G(s) is defined as the natural response and that due to the poles of X(s) the forced response. If the system has all its poles in the left-hand half of the s plane these values of s have negative real parts so that the natural response is a combination of nega- tive exponential terms which tend to zero as f°, i.e. a transient response, Stability considerations By a stable system we mean a system that yields a finite output for every finite input. Stability considerations can be represented by the transfer function G(s). A system is stable if all the poles of G(s) lie in the left-hand half of the s plane since again the response will contain only negative exponential terms, e.g. Go)=, aro represents a stable system, pole at s= —a. A system is unstable if one or more poles lie in the right-hand half of the s plane or if multiple-order pole(s) lie on the imaginary axis. In this case the response will contain terms such as e”, @ > 0 or f cos(wr), For example, 1 CSy=srr a>0 represents an unstable system, pole at s= a. If G(s) defines a function which approaches a non-zero value or a bounded oscillation then the system is marginally stable, ¢.g. 1 CO" ara62 Laplace transform and z transform The first-order complex conjugate pair s = +ja defines an oscillatory system. Note that the coefficients in the numerator which determine the zeros only affect the constants in the calculation of partial fractions and not the form of those fractions. The zeros therefore contribute to amplitude and phase but do not influence the form of the time function and hence stability. Sinusoidal steady state response If the driving function (input) is sinusoidal, e.g. A sin(wf) or Im{A exp(jw/)}, then the steady state (particular integral) is also sinu- soidal of the same frequency but with different amplitude and phase, i.e. E sin(wt) + F cos(wt) = R sin(wt + >) where R and @ depend on w and of course the constants of the system under consideration. This dependence is conveniently obtained by replacing s by jw ¥(jo) = Ge) X (ie) and discussion confined to the function G(jw) which, in general, will be complex having magnitude | G(jw)|, argument @(jw): G(jw) steady state transfer function | GGe)| magnitude (amplitude) response B(jw) phase response Hence given G(s) we can evaluate | G(jw) | and @(jw) as functions of w. Note 1 Gain M is defined as M=20 logio|G(jw)| and a Bode plot is graphs of M and 8 against w. Such information is helpful in determining the necessary additions to stabilize a system or improve its response. Note 2 Output amplitude = system magnitude x input ampli- tude since | ¥(jw)| = | G(jo) X(jo) | = |GGw) | | Xo) | and output phase = system phase + input phaseTransfer functions, block diagrams, stability 63 since arg[¥(jw)] = arg(G(jw)X(jo)) = arg[G(jo)] + arg[X(jo)] (the multiplication of complex expressions in polar form). EXAMPLE 6.3 Determine the magnitude response and phase response of l+s G(s)-—+*, One G(s)= l+s l+s+s? ‘ 1+ (jw) ole te +e _ die © =e) + ja magnitude response = | G(jw)| =| 1+ jw 7 la = 07) + ju 1+ je! “Td -a") + je] _ +0? © (da)? 40 phase response = arg [G(jw)] = arg| —tle _ (=a?) + jo = arg(1 + ja) — arg ((1—w*) + ja) =tan“(w)— aan'(5 +) EXERCISE 6 Determine the poles, zeros, magnitude response and phase response of the following functions of s:64 Laplace transform and z transform s#l ras 2 ! 4 2s+1 s ‘ 57 Convolution The concept of convolution is inherent in almost every field of the physical sciences and engineering. For example, in mechanics it is known as the superposition or Duhamel integral; in systems theory it plays a crucial role as the impulse response integral, and in optics as the point spread or smearing function. Let f(t) and g(f) be defined for ¢ > 0 and piecewise continuous. Then the convolution of f and g, written f+g, is defined by fre= [i e-ma du OK t<@ aay which is another function of f defined for positive 1. EXAMPLE 7.1 Determine the convolution of fi‘) and g(t) if fli)=e', gi)=P. fen | wPe'-" dus 2e'- 1 — 21-2 ° obtained by integrating by parts twice. EXAMPLE 7.2 Determine fog if 0) ores n= bs { 1>3 and f(n) = sin(r)66 Laplace transform and z transform Slt u) = sin(t - uv) g(u)=2 u>3 so that Sre-| 2 sind —u) dw for3 0, by evaluating the convolution integral. EXAMPLE 7.3 Determine 1 ellen fe (s+ 7 ee - 1 ¥"'e"} = 6- a) & tal =tConvolution 67 a on \" -ayi-u) du= {9 1S 4 £ [- zl ee ay(t—u) du f=o laa which may be written (¢—@)H( —a). EXAMPLE 7.4 Using Laplace transforms and the convolution integral solve the differential equation where /= diJd¢=0 when ¢=0 and f(t) is an unspecified function. Transforming the equation gives (S77 — Sip — ig) + °F = ¥LS(O} = F(s) Hence F(s) 1 Ste Inverting using Appendix 2, number 10, we obtain ' is j fu) * singe = ay) dw ° « or ij 1. \, Se=u) = singou) du since f*g=g +f. Hence, given (1), the current may be found by evaluating one of these integrals analytically, graphically or numeri- cally. EXAMPLE 7.5 Determine the convolution of the functions 7 or20 1 Osr 1 (Figure 7,1(f)) gives y=] Ixw duv=te-3(r-1)* 21 Hence combining equations (7.2) and (7.3) the result is ya$e=40-1PHO-) Some texts refer to the convolution integral as the Faltung integral from the German word for folding, 6970 Laplace transform and z transform Volterra integral equation In 1931 the Italian mathematician Vito Volterra published a model of population growth. An important equation called a Volterra integral equation occurred in this analysis and is of the form x(t) = ft) + f. h(t u)x(u) due o4 where x(r) is an unknown function and f(r), 4(t) are known functions. Subsequently many problems in thermodynamics, electrical systems, nuclear reactor theory and chemotherapy have been modeled with this type of equation. Note that the integral in equation (7.4) is a convo- lution integral so that taking Laplace transforms and using number 10 of Appendix 2 gives X(s) = F(s) + H(s)X(s) Solving for X(s) we obtain F(s) X= HG) which, in theory, can be inverted to give x(r). EXAMPLE 7.6 Solve the Volterra integral equation “ x(nee'-4 j cos [2(¢ — u)x](u) du © Taking Laplace transforms 1 s X(s)=—— - 4X(s) ‘) stl (ea 4s 1 xe) (ees) (s+2?_ 1 44 sel X(s) sed (s+) ot 4 2,8. “S4l (42) (42 by partial fractions. Inverting we obtain x(/) = Se~' — 4e"™ — 8re~™, X(s)=Convolution 74 EXERCISE 7 1 Determine the convolution f'*g and verify that # (fegl = [/1¥ tel for each of the following: (@) f()=5, 1720 g(=21, 620 (bo) f(=1-1 420 a= 30,120 (©) fle) = cost), 120 g(t)=sin(3), 120 (d) fryse"', 130 giy=r+2, 620 _ft O<@2 af) = sin(@2r), £20 2) Write the inverse Laplace transforms of the following functions as a convolution: 3 Ss ” (s+ 3(s+1) ® (s? + 9)(s—4) ot eu © a1 Sas 38) © eé : Vas 3° Evaluate the conyolutions (a) 5(1) * cos(4s) (b) 8 — 2)ee7* us) 4 Solve the following differential equations giving the result as a series integral: dy S24 ag =f(0} (a) 4 oe +a y=f() given that y=0, dy/d¢=0 when f= 0, a is constant and f(s) is an unspecified function of ¢. (b) at, +(a+ »$ < + abi=f(0) given that j= 2 dijdt=0 when f= 0, and a, 6 are constants. 9) 345% eee dr +t given that x=0, dx/d¢=2 when ¢=0. Do not attempt to determine YUsA+ oh.72 Laplace transform and z transform S The actual angle 6 in a control mechanism is given by the differential equation &e ar If ¢ is a function of ¢ and @=0, dé/d¢=0 when ¢=0, determine @ as a convolution integral in terms of ¢. =-K'(0-9) K constant 6 Solve the following equations: (a) x) =142 i sin(r—u)x(u) du ° (b) x(t) = 30? +t 5 (—u)?x(u) due °8 Sampled-data control systems and the z transform In certain types of feedback control systems the signal appears as a set of sequence values. This signal might be a discrete-time signal in its own right or derived from a continuous-time function by a sampling oper- ation. In either case the signal can be considered as a train of impulses equally spaced at discrete moments of time. Such systems are known as sampled-data control systems or discrete-time systems. The sampling operation introduces a time delay and a loss of information between samples which may lead to instability in the control system. The analysis of such systems can be undertaken using difference equations and the z transform in the same way that continuous-time systems were represented by differential equations and the Laplace transform. The impact of the digital computer, the transmission of digitized data with greater accuracy than continuous data and the use of time-shared com- munication have all contributed to the interest in sampled-data theory. Consider a continuous function f(r). If this function is sampled at equal intervals of time 7, the sampled function is a sequence of numbers FO), AT), f2T), fBT), -- MAT), --- This series of numbers gives a limited description of f(t). Its values at other times can only be approximated by means of interpolation and extrapolation (Figure 8.1). The sampled information will be designated J? (t), the graph of which will be a bar chart or simply a set of impulses of appropriate magnitude at ‘= 0, T, 27,37, .... Data predictors can be used to construct f(t) from the information f(t) (Shinners, 1964). Note that discontinuous functions must be defined precisely at the 774 Laplace transform and z transform y Figure 8.1 Sampling of f(r) points of discontinuity, e.g. u(t — k) at = k is defined, in Chapter 4, to be equal to +1. The sampled function f*(s) may be represented, as in Chapter 5, Property 7, by S70) = FO) (0) + S(T) 5(t— T) + fT) 80-27) + = 3 snr) on) = It is assumed that the sampling process is ideal and that no electronic switching time is considered. The Laplace transform of this function can be written as F*(s)= #1f*@)) = (, vo 81) + f(T) BT) + QT) B(t 27)... Je“ dt = fO + f(T + FAT)... - = faTye"™ If we now define a new variable z such that geetSampled-data control systems 75 or s=4 ing) T and denote F*(s) by F(z) then PO) = 2 ST )e" = 0) + MT + fOT|Z* + BN) ae which is a power series in z~' or 1/z. F(z) is called the z transform of the sequence or discrete-time signal /("7) and the relationship is written as FQ) = FUL@T)} In general, if any continuous function possesses a Laplace transform then the corresponding sampled function has a z transform. EXAMPLE 8.1 Determine the z transform of the sampled function f(#7) if f(r) is the unit step function u(r). Sampling this function gives /(nT) =1, 2 > 0. FQ@)= 3 Wem deg ez. = from equation (8.1). Hence (this is the sum of a geometric progression that is convergent for |z| > 1). Therefore Flu(aT)| = EXAMPLE 8.2 Determine the z transform of the sampled function /(nT) if /(t) is the ramp function f() =". Sampling gives f(n7) = 0, 7,27,37, From equation (8.1) Fz) = Tz" + 27278 +3723 +76 Laplace transform and z transform which is convergent for |z| > 1. Therefore Win TE ae F EXAMPLE 8.3 Determine the z transform of the sampled function f(m) if f(r) =¢7 where @ is a constant. Sampling gives f(n)=e"*". FR) = Dente eZ eg + ee 1-e z-e* which is convergent for |z| >¢~*. Therefore ster z Fle = z-e In the same way that a table of Laplace transforms is derived from basic definitions, so also is a table of z transforms (Appendix 3). Note the following. 1. zmay be complex so that conditions for convergence of the infinite series involve the modulus. 2. The z transform is a unique transformation. 3. It can be shown that the z transform is independent of T so that S(mT) can be replaced by f(r), the preferred notation. The sampling interval 7 is simply a scale factor. Some important theorems These results can be proved from the definition (Doetsch, 1971; Karni and Byatt, 1982). 1, Addition: Fin) + fla) = FIA) + FtAW))Sampled-data control systems 77 Multiplication by a constant: Flafryh = aF (fr) Translation — shift theorem: FU - myu(n— m)} =" F1F)) mt FUse+ m= 2"| ZV) ~ 5 sade] These are shown in the table of standard forms (23 and 19-22 respectively). Initial value theorem: limit [f*(0)] = limit [F(z)) ‘0 roo Final value theorem: tims 7*¢0 = mic [= ra] ie ret | Zz Periodic sequence theorem: if f(n) = (0), (1), Ss). + LON = 1) and the sequence of N values is repeated periodically then Fiz)= fe! where F(z) is the z transform of the periodic function and Fi(z) is the z transform of the sequence f(0), ..., {(N— 1). For example if JS(n)=1, —1, —1, 1, and these four values are repeated periodically then Fiz) =1- 27 and78 Laplace transform and z transform EXAMPLE 8.4 Using the table of standard forms determine the z transforms of the following sampled functions: f@) fiiy=n (b) fOr) = Ke“*" = 2 sin(bn) if a, b and K are constants. ©) Sony =e Mu(n — 2) A) fen) = 2"? — 2) fa) (by «) (d) S(ny=n: z @-1" using number 4. S(n) = Keo" = 2 sin(bn): Ke 2z sin(b) ze" 2-2 cob) +1 using numbers 6 and 8. Sin) =e" Du(n — 2): FQ)= F@)= F=z7 Fe? 2 z z-e using numbers 23 and 6. J(n)=2"(n? =n). From number 27 it is necessary first to determine the transform of n? =n: 2@+D_ @- The required transform is 22/2) _ 8 (3-0? @-2r =z 3 Fin EXERCISE 8 Determine the z transform of each of the following: lan 2 3 (-)" 4 1-(-59" 5S ne*" 6 243°"