2-4 Examples Silustrating Evaluation of the Convolution Integral ‘The convolution of two signals, x(1) and (7), is a new function of time, y(1), which is given by (2-5). A useful symbolic notation often employed to denote (2-Sa) and (2-Sb}, respectively. is Y(t) = A(t) * x(2) (2-48) y(t) = x(0) * h(1) (2-49) Itis possible to prove several properties of the convolution integral. These are listed below, but their proofs are left to the problems. - h(t) * x(t) = x(t) * hie). h(t) * Lax(0)] = a h(t) * x60)], where a is a constant A(t) * [x,(1) + x3(9)] = (2) * xy(0) + hie) * ay(0) h(t) * [x,(0) x30] = [aCa) * xy(9) * 2,00. If A(2) is time-limited to (a,b) and x(¢) is time-limited to (c,d), then (2) * (0) is time-limited to (a+c.b+d). 6. If A, is the area under (1) and A, is the area under x(1), then the area under f(s) = x(4) is A\Ay Peeps In these relations, h(1), xy(1), and x,(¢) are signals. The parentheses and brackets show the order of car- rying out the operations. The integrand of (2-5a) is found by three operations: (1) reversal in time, or folding, to obtain x(— A); (2) shifting, to obtain x(t ~ A); and (3) multiplication of h(A) and x(f — A), to obtain the integrand. A similar series of operations is required for (2-5b). Four examples will be given to illustrate these three operations and the subsequent evaluation of the integral to form y(1). EXAMPLE 2.7 Consider the convolution of the two rectangular-pulse signals (2-50)62 Ch. 2 / System Modeling ana Analysis in the Time Domain |" | | | These signals are sketched in Figures 2-Sa and b. We base our evaluation uf the convolution of sie) and ir} on (2-Sa). Thus x¢) is reversed and the variable changed to A, which results in x(—A), also shown in Figure 2-5. Finaily. x(—) is shifted so that what was the origin now sppews at A=. ft is sometimes helpful cw think af «it — A) as xf -C ~ dD}: that is, ¢ in. At is replaced by A — 1. with A thought of gs the independent variable, uy hue by wou tema) (ay FIGURE 2-5. Signals to be convolved in Example 2-7.2-4 / Examples ilustrating Evaluation of the Convolution integral 63 whereupon the function x(A ~ #) is reversed or folded (its mirror image about the ordinate is taken) ‘The result of these operations is x(t ~ A), which is shown in Figure 2-5d for t = “The product of x(r~ A) and (A) is formed, point by point, and this product integrated to form y(t). Clearly, the product will be zero if r < 4 in this example. If r= 4, the rectangles forming A(A) and x(t — A) overlap until ¢> 10, whereupon the product is again zero. ‘After a little thought, we see that three different cases of overlap result. These are illustrated in Figure 2-6. In each case, the area of integration is shaded; because the height of h(t) is unity, the area is simply the width of the overlap times the height of x(¢— A). Thus the area for Case I is 2(¢ ~ 4); l con sarc Case Ml: 8<1<10 vn) @ FIGURE 2-6. Steps in the convolution of Example 2-7 and the final result.@4 Ch.2 / System Modeling and Analysis in the Time Domain that is, itis linearly increasing with t. For Case Il, itis 2{(1 ~ 4) — (¢ - 6)] = 4. Finally, for Case III, it is 2[4 — (¢ - 6)] = 2(10 — 9); that is, the area linearly decreases with 1. The resute for y(¢) is sketched in Figure 2-6d. EXAMPLE 2-8 As a second example of the convolution operation, we consider the signals shown in Figure 2-7a. We use the convolution integral (2-5b) in obtaining the result. This example differs from Example x) mo =I of | ° 2 too i ! | ' ' -2 "2 @ ° x0) me~Wi ! 1 } — » a fm §, OF 1 ' ' | FIGURE 2-7 Wavetorms pertinent to the convolution of Example 2-8.2-4 / Examples illustrating Evaluation of the Convolution Integral 65, 2-7 in that part of the area under the product x(A)h(i ~ A) is negative. Our first step is to express h(i — A) mathematically. Since 5. Osrs2 Ait) = (2-52) 0, otherwise we obtain 0<1-AS2 or t-25Ast Ale H= (2-53) 0, otherwise ‘The signals x(A) and A(r — A) are sketched in Figure 2-7b. Some thought on the part of the student should result in the four cases of nonzero overlap illustrated in Figures 2-7c through f. The resulting convolutions for these cases are expressed mathematically by the following equations: 0 6<-1 or 1>3 [@-ya, -1=1<0 a q (= ayaa [ (ayaa, Ost<1 WO = L, 6 (2-54) 0 1 f aaa [Gad tsr<2 2 o 1 - Ada, 251<3 2 Integration of these expressions results in the following for y(1): 0, t<-1 or 1>3 4+ -1sr<0 yneq cP tat, Osr c2ex8 % MATLAB plot of Example86 Ch.2 / System Modeling and Analysis in the Time Domain L=length(t) tpe[2te(1) del. x=2*(pls_fn(t+.5)-pls_fn(t-.5)): % Defined in Chapter 1 S*rmp_fn(t) .*stp_ % Defined in Chapter 1 lel_t*conv(x.h) : % Multiply by step size to approximate % rectangular rule integration subplot(3,2,1). plot(t.x), xlabel(‘t'), ylabel ('x(t)") axis({e(1) t(L) -3 31) subplot(3,1,2), plot(t.h), xlabel('t’), ylabel(‘h(t)"). axis((t(1) t(L) 0 21) subplot(3,1,3)., plot(tp.y), xlabel('t'). ylabel(ty(t)'). axis((t(1) t(L) -2 2]) y ‘The plot output of the program, given in Figure 2-8, is seen to be identical to Figure 2-7g. EXAMPLE 2.9 In this example, we consider the convolution of a ramp and the decaying exponential signal of the form x(t) = exp(—az)u(t). The convolution integral in this case is no y A ° 1 2 3 4 1 FIGURE 2-8. Maras verification of the result of Example 2-8,2-5 / Impulse Response of a Fixed, Linear System 67 >t) f ” nj = 2) de (2-56) Note that the first term increases linearly like the ramp, but with a slope of 1/. The second term ap~ proaches ~1/a? as tc. The student should sketch the result of the convolution along with the [_somponents of the integrand for different values of 10 provide justification for the result