220 > ‘Transform Solutions of Linear Models In the last example, notice thatthe current i(¢) through the inductor is felt) eo] = te" fort > 0 ‘The current jumps from zero to 1/L at ¢ = 0, thereby instantly inserting energy into the inductor. Fort > 0, when ¢((t) is again zero, this energy is dissipated in the resistor as the ‘current gradually returns to zero. AS a general rule, the current through an inductor cannot ‘change instantaneously unless there is an impulse of voltage. Similarly, the voltage across ‘a capacitor cannot change instantaneously unless there is an impulse of current. > 7.6 TRANSFORM INVERSION ‘When we use Laplace transforms to solve for the response of a system, we find the trans- form F(s) of a particular variable, such as the output, first. The final step in the process, known as transform inversion, isto determine the corresponding time function f(t) where L(0) = £""(F(s)], which is read “f(¢) is the inverse transform of F(s).” For the types of problems encountered in this book, we do not need a completely gen- eral method of uausfon inversion. In dis section, we present and illustrate & method that uses a partal-fraction expansion of F(s). An extension of this method that permits handling an additional category of transformed functions appears in Section 7.7. ‘Assume that we can write the transform F(s) as the ratio of two polynomials V(s) and D(s), such that NG) byl ney © Dis) ~ Functions that can be written inthe form of (64) are called rational functions. A proper rational funetion is one for which m 0, J) = Are Ane get Saee 8) 4 We can find the n poles s; by factoring the denominator of F(s) or, more generally, by finding the roots of D(s) = 0. We shall now develop a procedure for evaluating the n coefficients A; so that we can write f(¢) as a sum of exponential time functions by using 8). Multiptying both sides of (66) by the term (s—s1) yields (s—s1) (-s) set An (s—m) (s~5n) Because this equation must be a mathematical identity forall values of the variable s, ‘we can set s equal to si throughout the equation. The poles are distinct, so si 4 5) for j=2,3, ...;m. Thus each term on the right side will vanish except the term Ay, and we can write (~51)F(3) = Ai + Az At (s—si)F(s) (69) ‘To make it clear why (69) does not give a value of zero, we write F(s) as N(s)/D(s) with D(s) in factored form. Then {ssi N(s) Beate= 2m) The term (s—5s1) in the numerator will be canceled by the comesponding term in D(s) before s is replaced by 5, Repeating the above process with sy replaced by s2, we have Aa = (8—52)F (sles ‘The general expression for the coefficients is =(s-s)F(lmn In a numerical problem we can check the calculation of Ar, Az, ..., An by combining the terms on the right side of (66) over a common denominator, which should give the original function F (3). Ai penny «70222, P Transform Solutions of Linear Models > EXAMPLE 79 Find the inverse transform of =s+5 FO = Taye SOLUTION Comparing F (s) to (64), we see that N(s) = —s+5 so m = 1, and D(s) = (s+1)(s-+4) son =2. Because D(s) is already in factored form, we see by inspection that the poles are s1 = —1 and sp = 4. Thus we can rewrite the transform F (5) in the form of, (66) as Ar sea Using (70) with i= 1 and 2, we find the coefficients of the partial-fraction expansion to be Ls+t{-s+5) *Bertis+ a) At and, from (68), the time function f(0) is fl) =2e1= 3" fore > 0 Repeated Poles 1 (wo oF more of the n roots of L{s) = 0 are identical, these roots, Which are poles of F(s), ate said tobe repeated. When Fs) contains repeated poles, (66) and (68) no longer hold, Ifs; = so, for example, the first two terms in (68) become Aye" and Aze™, which are identical except for the multiplying constant and (68) is not valid. If) => and ifthe remaining poles are distinc, then (66) must be modified to be m) Referring to Appendix E, we see thatthe first term on the right-hand side of (71) is the transform of Ait", whereas the second term has the same form as the remaining ones. ‘Thus the inverse transform is F(t) = Anite! + Arne + Age pet Age (2d) Note that the repeated pole at s = s1 introduces into the time function a term of the formtcs". In order to evaluate any of the n ~ 2 coefficients As, ..., An We can use the procedure given by (70). This formula cannot be used to find A1y and Ay2 because of the term (s—s1)® in the denominator of F (3). However, multiplying both sides of (71) by (s~ si)? gives (=? 54 (0? ae Gay 1A ay 11 Ana(3~ 51) +a! 4 m7.6 Transform Inversion <4 223 Setting s equal to 5; throughout (73) results in PF ew may ‘To find Ay2, we note that the right-hand side of (73) has the terms Ati + Ara(s~ st) and that all the remaining terms contain (s— si)? in their numerators. Thus if we differentiate both sides of (73) with respect tos, we have Ana Gl) os) a 6 (=) — Aug lle nO At ‘where G(s) is a rational function without a pole at s = s1. Note that in (75) the coefii- cient A11 is not present, Aiz stands alone, and the function G(s) that contains all the other Coefficients is multiplied by the quantity s~ sy. Hence setting s equal tos) in (75) gives an= {Fleer} 16) ‘where the differentiation must be performed before sis set equal t0 51 F (3) has two or more pairs of identical poles, then each pair of poles contributes terms ofthe form e and re, and we can evaluate the coefficients of these terms by using (74) aad (76) with the appropriate indives. If F(s) has dhree ur more identical poles, (71) rust be modified further. For example, if si = s2 = 53 and if the remaining poles are distinct, the partal-fraction expansion has the form i-m ay By a procedure similar to that used in the derivation of (74) and (76), we find that Ant = (8-1) F (Neos, 4 is- s1)3F(s) an= {Fle-aprol}| 1p@ 5 au= a {gele—arcol}| s) ond Ap —S)F(s)\e=s, fori=4,5,...,n. ‘The time function corresponding to (77) and (78) is An? +Arst+A19) Mb Age bot Ane’ (79) > EXAMPLE 7.10 Find the inverse Laplace transform of 58416 P= GaP EF) SOLUTION The denominator of F(s) is given in factored form, so we note that the poles are 5; = s) = —2 and 3 = —5. Because of the repeated pole, the partial-fraction224 > ‘Transform Solutions of Linear Models expansion of F(s) has the form Au, Aw, As FQ) = At, 4 Ait, AD = Geapt saat is ‘Using (74), (76), and (70) in order, we find that the coefficients are ; 55416 6+2)F()| = a seis) asst lea As= (S+S)F (eas ‘Using numerical values forthe coefficients gives the partial-fraction expansion of the trans- form as FQ)= ‘The corresponding time function is Fi) eM ee Complex Poles ‘The form of f(t) given by (68) is valid for complex poles as well as for real poles. When. (68) is used directly with complex poles, however, ithas the disadvantage that the functions, ‘1 andthe coefficients A, are complex. Hence, further developments are necessary in order to write f(?) directly in terms of real functions and real coefficients. We shall present two ‘methods for finding f(¢) that lead to slightly different but equivalent forms. Fist we show hhow mathematical identities are used to combine any complex terms inthe partisl-fraction expansion of F(s) into real functions of time. Then we present the alternative method knowin as completing the square. Basic operations with complex numbers are reviewed in Appendix C. For simplicity, we shall frst assume that F(s) has only two complex poles and that the order of its numerator is less than the order of its denominator. Any other pole, whether real or complex, will ead to additional terms in the partial-fraction expansion. With these restrictions, we can write the transform as Bote F< Brat jayerar io) (80) which has poles at s1 = -a+ jv and s2 = — fo. Partial-Fraction Expansion AAs indicated in (80), the complex poles of F(s) always occur in complex conjugate pairs. Using the partial-fraction expansion of (66) for distinct poles, we can write F(s) as k bs FQ)= Soe tia e je" stat fe 1)7.6 Transform Inversion {225 where Ki = (044 je) O)lemnasio «er Ky = (5404 jus)F(s)m-o- je (826) ‘Because (82a) is identical to (82b) except for the sign of the imaginary term ja wherever it appears, the coefficient Ke isthe complex conjugate of Kis tha is, Kp = K3. Hence we ‘can write K, and K in polar form as Ki = Ke! and Ky = Ke~/*, where K and @ are the ‘magnitude and angle, respectively of the complex number Ky. Thus both K and are ral quantities, and K > 0. Rewriting (81) with K and Ky in polar fom, we have ie - y= SO, KO (3) eas bs‘ rath ‘which has the form of (67), although the coefficients and poles are complex. From (68) with n = 2 and with the appropriate values subsitted for A and s, the complex form of the time function is fle) = Keibeloiel 4 Renita iah To obtain f(r) as a real function of time, we factor the term 2Ke~* out of each term on the right-hand side of this equation and combine the remaining complex exponentials. This. yields ft) =2ke* [ - cil 48) cn) Recognizing from Table 1 in Appendix D thatthe term in the brackets is cos(ut + 4), we can write fl) =2KeMeos(wt+g) — fors>0 (84) Where the parameters a,w,K. and ¢ are all real. Note that (83) and (84) constitute one of the entries in Appendix E. Completing the Square ‘Todevelop the second ofthe two forms of f(t) when F(s) has complex poles, we multiply the denominator factors in (80) to get BstC FO) = Fiesta? ‘Next we write the denominator as the sum of the perfect square (s-+ a)? and the constant . «23, so that 5) ‘Now we rearrange the numerator of F (5) to have the term (s-+.a) appear: B(s-+a)+(C-aB) FO) = arte! a Tal + =) [eae226 > Transform Solutions of Linear Models Referring to Appendix E, we see thatthe quantities within the brackets are the transforms of €-* casi and e* sinwt, respectively. Thus sinwt (86) Hi) = Ber ost + (28) Equations (85) and (86) form another one of the entries in Appendix E. Although (84) and. (86) look somewhat different, we can always use Table 1 in Appendix D to show that they are equivalent functions of time. > EXAMPLE 7.11 Use each of the two methods that we have described to find f(t) when 448 FO) = sy 9545 SOLUTION The poles of F(s) are the rots of s? +2545 = 0, namely s1 ands) = —1~ 2. Hence a = 1, = 2, and the paria-fraction expansion of F(s) is 1472 Ki Ke FO ss1-pss1+R Solving for Kj according to (82a), we find (s-1—JV(4s +8) = Ger aeri lesen =2-)1 = Ve 46 ‘Tous K = VF and 4 = 0.4636 rad, Substituting the known values of aK, and ¢ into (84) yields J() =2VBe"cos(2r-0.4636) fort >0 en Having found f(t) from the partial-fraction expansion of F (s), we repeat the problem by the method of completing the square. Because the denominator of F(s) is s? +2s-+5, ‘we can obtain a perfect-square term by adding and subtracting I to give (242541)4(5-1)= (641) +4 448 PO) = Gyypee By comparison with (85), we see that B= 4,C = 8,@ = 1, and w = 2, With these values, (86) gives “eos2t-+2<"sin2e fore > 0 f= which we can convert to the form given by (87) by using the appropriate entry in Table 1 in Appendix D.76 Transform Inversion << 227 ‘Comparison of the two methods for finding f(¢) indicates thatthe partisl-raction ex- pansion requires the manipulation of complex numbers. Iti frequently employed, how ever, and we shall need it for a derivation inthe next chapter. The method of completing the square has the advantage of avoiding the use of complex numbers. The following example illustrates the two methods when F (s) has more than the two poles specified by (80) and 5). > EXAMPLE 7.12 Find the inverse transfom of 52 +85, FO) = air tast5) 52 485-5 = 3eti-Aerl+ SOLUTION In order to use (84), we write the partial-fraction expansion as, Ki K sap te a where AU = 42545 lao Aw (4 [228525 24605450) _ 4 BS Yas | P42545 | Slo (FHITTF hu 524805 -28- jd -7- 1 TS SCT let4p 1-72 4-8 “Although itis possible to evaluate the quotient of two complex numbers directly on many hhand calculators, an alternative procedure is to rationalize the fraction by multiplying both halves by the complex conjugate of the denominator. Then -T=jl 44/31 B= OF ER = 55-25-29 H1- j= vie P36 Inserting the values of Ar, Ara) and Kj into (88) and then esng (1), (3), and (84), we obtain f(0) = 1424 2VIec0s(21 2.356) fore>0 6%) In ordero avoid complex numbers, we may choose not o factor the quadratic s?+25-+ 5 that appears in the denominator of F(s). For the corresponding term in a partial-fraction expansion, however, we must then assume a numerator ofthe form Bs +C. Thus, for an altemative solution to this example, we write 524805 PP +2s+5) An An, BSC 3) 42543 Fis) =228 > Transform Solutions of Linear Models ‘Recombining the terms on the right side of this equation over a common denominator gives (Ai2 +B) + (Ain +2412 +O}? + (24+ SAra)s + SAN MEETS) By equating corresponding coefficients in the numerators, we have F(s) An+B=0 Aut 2A + 2Au + SAr Su = and C = 2, If we also complete the square in the factor ?-+2s+5, we may wrt the patal-fraction expansion as H1.2, 2042 Bt From (11), (3), and (86), the corresponding function of time is f(t) =—t+2—2ecos2t + 2€~ sin2r forr > 0 (90) It is easy to show by Table 1 in Appendix D that (89) and (90) are equivalent. iminary Step of Long Division Remember that the techniques discussed so far are subject to the restriction that F(s) is 4 stritly proper rational funetion—in other words, that m, the degree of the numerator polynomial.V(s) is less than n, the degree of the denominator polynomial D(s). Otherwise, the partal-fraction expansion given for distinct poles by (67) or fora pair of repeated poles by (71) is not valid. In order to find the inverse transform of F(s) when m= n, we must first write F(s) asthe sum of a constant” and a fraction whose numerator is of degree n— 1 or less. We can accomplish this by dividing the numerator by the denominator so that the remainder is of degree n ~ 1 or less. Then we may write F() -A+P'(s) ey where A is a constant ands the transform of A6(t). The function F"(s isa ratio of polyno- mials having the same denominator as F (s) but a numerator of degree less than n. We can find the inverse transform of F'(s) by using the techniques described previously. Thus SO = Ad) + Le] 2) » EXAMPLE 7.13 Find f() when 224148 FO)= 43542 When m > n, we can write Fs) 25 che sum of «polynomial ins and sey proper rational function, but we shal not encounter such eases hte,7.7 Additional Transform Properties 4 229 SOLUTION Both the numerator and the denominator of F(s) are quadratic in , so we have m= n= 2. Before carrying out the partial-fraction expansion, we must rewrite F (s) in the form of (91) by dividing its numerator by its denominator, as follows: 2 P4394 13274 TSFB 22 +6544 errs) Because the poles of Fs) ae sj = —1 and sy = ~2, we can write on td FQ)=2+ GanGaD where Lsrtis+4) A= EE! Lg Ber?) Lme+4)| CNET Thus 32 FQ) =24 5-5 and Flt) =25() 431-26 fore >0 Which is the sum of an impulse at. (0+ and two decaying exponentials. > 7.7 ADDITIONAL TRANSFORM PROPERTIES “We now introduce several useful transform properties that were not needed for the examples in Section 7.4. The first two of these concern functions that are shifted in time. The others provide certain information about the time function directly from its transform, without our having to carry out the transform inversion, ‘Time Delay the function f(t) is delayed by a units of time, we denote the delayed function as f(‘—a), where a > 0, In order to develop a general expression for the transform of the delayed function /(¢ — a) in terms of the transform of f(t), we must ensure that any part of f(t) that is nonzero for £ < 0 does not fall within the range 0 0, at least part of it falls within the interval 0 <1 <=, as shown in Figure 7.24(b).