Solution Manual for B. P. Lathi Signal Processing and Linear Systems Oxford University Press(Oxtord University Press (Oxford New York ‘Athens Auclland Bongkok Bogti Buenos Aes Calcutta ‘Cape Town Chennsi Dares Salam Dethi Florence Hong Kong. Istanbul Karachi Kuala Lumpur Madd Metboume Mexico City" Mumbsi NNaiobi Pais’ Sto Paulo Shanghai Singapore Taipei Tokyo Toro Waraw nd srsociated companies in Berlin Toadan Copyright © 2001 by Oxford University Press, nc. Published by Oxfoed Univesity Pres, Ine. 198 Madison Avenue, New York, New York, 10016 ‘tp wr oup-osa org Oxford isa registered iademask of Oxford University Press Allright reserved. No pat ofthis publication maybe reproduced, stored ina rticvl system, or transmted, in any form or by any means, lecronie, mechanical, photocopying reording, or alherwise, ‘without the price permission of Oxford University Press. Library of Congress Cataloging in-Publication Data ISBNOA9- Printing mmber: 987654321 Printd inthe United States of Ameri on cidfve paperChapter 1 11-1 For parts (a) and (b) By= | sinteae [ cos 2tdt =n +0= 5 B= [entrant [ad [owaaar tone co) ay [ cosnot anal a=} [con zea] = te 0)~ te see freoran Ey= [ (eae 11-3 (a) Ex = f(a = 2. By = fp(aPde + fe(-1at (tae =a, Beye f (aya =4 ‘Therefore Bray = Be + By (ayaatn By = [Cores [ cvrae [rae (fae = ae EL = [orw Bem [ara [Cores [C orante wwe can show that Ex-y = 4m Therefore Bxiy = Be + By a 5 © Ben [Cae [" capaewn 8px [ora tye [ras [orane Bae (“ora f Therefore, in genetal Evey # Be + By [ (eae = 64/7 nro fiwra [oerus swum rr} [ea 64c8/7Sign change of a signal docs not affect its power. Multiplication of a signal by a constant c increases the power by a factor © has Prem Ef" soreode= gm tf" Svea ‘The integrals of the cross-product terms (when k 4 r) are finite because the integrands are periodic signals (niade up of sinusoids). These terms, when divided by T — oo, yield zero. The remaining terms (k =r) yield mdf Simran Sos wim Py 1.1-6 (a) Power of a sinusoid of amplitude C is C2/2 [Bq, (1.6a)) regardless of its frequency (w # 0) and phase. ‘Therefore, in this case P = (10)%/2 = 50. (b) Power of a sum of sinusoids is equal to the sum of the powers of the sinusoids (Ea. (1.5b)). Therefore. in this ease P= OO" 4 OOF = 178, (c) (10 +2 sin 3t) cos 10t = 10cos 10¢ + sin 13t ~ sin St. Hence from Eq. (1.5b) P= OO" +} +} = 51 (4) W0con Secon 10¢= con 51+ 15 Hence trom Eq. (1.56) P= 2 + (6) 10sin Stcos 10t = S(sin 1 ~ in St. Hence from Eq. (1.58) P= SY + GBF = 25 (£) 2°" conot = } [e+ oX°"9)]. Using the result in Prob. 1-1-5, we obtain P = (1/4) + (1/4) = 1/2 aaa Salt) = HE- + AUC 1) $51) = FE-1)+ ALE +N) Lal) = fE- 05) + E+ 0.8) ‘The signal ft) can be obtained by (i) delaying f(t) by 1 second (replace ¢ with ¢~ 1, (ji) then by a factor 2 (replace ¢ with ¢/2), (ii) then multiply with 1.5. Thus fs(0) = 1.5/(4 ~ 1). 41,8-2. All the signals are shown in Fig. S132. Foe) 44 : { Flere) expanding =f 2] tt 3 64 3 o> “4 “a #G at aie? af 2 $) 1\/* a t> 24 $0 30 ete ot 74 Fig. 51.82 1.8-3 All the signals are shown in Fig. $1.3-3 44 $A) Cc “4 Gs) Pa a] $8) o oe 6 => Fata) wean Fig. S1.3-32 y ’ ce at G 4 eo] t>~ 6 7 ° 4 os 2 t> 14-1. All the signals are shown in Pig. $1.41 142 Salt) = A(C+ ule + 1) — ule} + (-2¢-4-4)fu(t) — ule —2)] = aCe + ut +1) — 6tult) + Bult) + (21 —A)u(e—2) J) = Blue) = ult = 2)] + Qt = 8)fult ~ 2) — u(t — 4)] = Pale) ~ (CF ~ 2¢+ 8)u(e ~ 2) ~ (Bt ~ s)ult— 4) 143 ey [ espa PWdte= Ey, Egy [iucore [ P(a)ds = By Bron = [ue-mPam [ Pe)de= Fr Bran= [” Wooran Frans [uer-opa=? f” Perde= y/o, Em =f Uweirane f” Peden ob, Baws f (os(ey? de =o! £ Plde= oy ded Using she fact that /(2)6(«) = f(0)6(z), we have (a) 0 (b) $560) (©) $60) (A) ~48(C= 1) (@) pAyHlw +3) (F) Kélw) (se L! Hpita's rule) 1.45 In these problems rememnber that impulse 6(z) is located at x = 0. Thus, an impulse 6(¢~r) is located at 7 = t, ‘and s0 0n (a) The impulse is located at r = ¢ and f(r) at 7 = tis f(0). Therefore [ S(r)6(t =n) dr = H(0) (b) The impulse 8(r) is at 7 =O and f(t—r) at r = Os f(t). Therefore [ S(r)flt = Ndr = 0) Using similar arguments, we obtain (1 (0 (ee (5 (w) s-1) (a) @ 4 Bes OC © 4 Aa ={ Oo 2 t= a @ > rents 34 a t7 @ Fig. S1.4-7 1.4-6 (a) Recall that the derivative of a fmetion at the jump discontinuity is equal to an impulse of strength equal to of discoutinity. Hence. df /dt contains impulses 46(¢ + 4) and 26(t ~ 2). In addition, the derivative the interval (4, 0), and is 1 over the interval (0, 2). ‘The derivative is zero for t <4 and £ > 2. ‘The result is sketched in Fig. S1.4-6a. (b)Using the procedure in part (a), we find d?f/a¢? for the signal in Fig, P1.4-2a as shown in Fig. S1.4-6b, 14-7 (m) Recall that the area under an impulse of strength k is k. Over the interval 0 < t < 1, we have woe [re (Over the interval 0.< <3, we have wow fades (Coes AUt=3, the impulse (of strength unity) yields an additional term of unity. Thus, ostst -t 1st<3 vos [ides fenecs ff senndentseayei= toa () (th = [1h ~ 6¢~ ~ 1) ~ 6(2 ~ 2) ~ 6(2 ~ 8) + --Jde = tule) — u(t ~ 1) ~ u(t ~2)~ u(t ~~ 41.4-8 Changing the variable ¢ to —2, we obtain [ (Ost) 4 [Ocoee [sos [onlay as = 60) ‘This shows that [ (0)6(-#) at = 6(0) Therefore 5(8) = 5(-1) 9 Letting at = z, we obtain (for « > 0) [0 s100 0 [22 tte)4e= L000 Stay fora < 0, we show that this integral is ~Le(0). Therefore [eps qe = afew ‘Therefore a) = te)14-10 (a) [sosa- swewrs,= [7 sonra =o fsnnae= 50) LALA (a) 12 = 493 (b) eM coe Bt = 0.5fe-87" 4 e-FIN], Therefore the frequencies are s12 = -3+)3 (e) Using the argument in (b), we find the frequencies si2= 2453 (4) 43 a. 3 ° and fa(t) + 0.5|cos wot u(t) — cos(—wot)u(—t)] = 0.5|coswot u(t) ~ coswot u(—t)) (e) felt) = 0.5|sinwot + sin(—wot}] = 0 and f.(t) = 0.5[sinwot — sin(—wot)) inwat pipettes enced cette oneal asia z Cay ft & ee 7 ete olt= be ll wy oltey lo te ‘o> 4 fo & Se , fe fy 7 7 |o [Ot Gy Ee a | G) Fig. S1.5-1 If f(¢) and y(t) are the input and outpnt, respectively, of an ideal integrator. ther u(t) = fontr= fi sovees (r)dr = x) +f f(r) dr L.7-1 Only (b),and (h) are linear, All the remaining are nonlinear. This can be verified by using the procedure ‘ligeussedin Exaniple 1.10 1.7.2 (a) ‘The system is time-invariant because the input f(¢) yields the output y(t) = f(t —2). Hence. if the input is F(t T), the output is /(¢~ T ~ 2) = ylt~T), which makes the system time-invariant. (b) The system is time-varying. The input /(t) vields the output u(t) = f(t). Thus. the output is obtained by changing the sign oft in f(t). Therefore. when the input is f(¢—T). the output is f(-¢—T) = /(-[t+ 7) = W(C+T). which represents the original output advanced by (not delayed by 7).ars Lr Lr (c) The system is time-varying. The input f(t) vielis the output y(t Input. Thus, the output is obtained by repla delayed by T), the output is flat —T) = f(alt ~ is time-varying (4) The system is time-varying. The input f(t) yields the output y(t) = f(t). For the input f(t — 7), the output is {f(¢—T). which ix not £/(@) delayed by T. Hence the system is time-varying (©) The system is time-varying. The output is a constant, given by the area under f(t) over the interval jt] < 5. Now, if f(t) is delayed by T, the output. which isthe area under the delayed J(), is another constant. But this ‘output is not the same as the original output delayed by 7 Hence the system is time-varying. (8) The system is time-invariant. The input f(e) yields the output y(t). which is the square of the second derivative of f(¢). If the input is delayed by T, the output is also delayed by T. Hence the system is time Flat), which is a scaled version of the in the input with at ‘Thus, if the input is f(t - 7) (/(®) £]), which is f(at) delayed by T/a (not T'). Hence the system: We construct the table below from the first three rows of data. Because of the linearity property of the system), ‘we can multiply any row by a constant. We can also add (or subtract) any two rows. Let rs denote the jth row. Row 10) (0) za(0) oft) n ° 1 -1 etu(t) 2 o 2 1 et(at + 2)u(0) “®) “1 -1 2u(t) o 1 ° (e+ Deu) qu) 0 - Gets nu fu) -1 (set + tet + 1) 4) ° ° (et + tem + 2)ule) Im our case, the input f(t) = u(t +5) ~ u(t — 5), From rr and the superposition and time-invariance property, wwe have ult) = ro(t-+5) = ro(e= 5) [er 4 206+ 5)eM49 4 2} ult + 5) — [eH 4 20 — 5)e- 9 4.2] ult — 5) I the input is K/(t), the new output y(t) is w= END) = HUA D) Hence the homogeneity is satisfied. Also fom (f/f) and fa + ve = (f2)*/(Fa) But fit fam (hhh fa) én tin From the hint itis clear that when 1.(0) = 0, the capacitor may be removed, and the circuit behaves as shown, in Fig. $1.7-5. It is clearly zero-state linear. To show that itis zero-input nonlinear, consider the circuit with J(t) = 0 (zero-input). The current y(¢) has the same direction (shown by arrow) regardless of the polarity of v (because the input branch is a short). Thus the system is zero-input nonlinear. =©® L = R Figure $1.7-5 6Lr6 Lint Le Let 182 o-dH VES + fH @) ye) Figure $1.7-6 ~ ‘The solution is trivil. 1 behaves as shown in le input is a enirent source, Hence, as far ax the output y(t) is concerned, the circuit ig. S17-8, The nonlinear elements are irrelewnnt in computing the output y(¢). Hence the ‘output u(t) satisfies the linearity conditions. Yet. the circuit is nonlinear becasue it contains nonlinear elements fand the ontpnts associated with nonlinear elements Zand C will not satify linearity couditions (a) vt) = f(¢—2). Thus, the output y(t) always starts after the input by 2 seconds (see Fig. SI.7-7a). Clently, the system ik causal (b) ult) = (8). The output y(t) is obtained by time inversion in the input. Thus, ifthe input starts at ¢ = the output starts before ¢ = 0 (see Fig. $1.7-7b). Heuce, the system is not causal. (©) ult) = f(at), a > 1. The output y(t) is obtained by time compression of the input by factor «. Hence, the ‘outpu: can start before the input (see Fig. $1.7-7e), and the system is not causal (A) ylt) = flat), «<1. The output y(t) is obtained by time expansion of the input by factor 1/a, Hence, the ‘outpu: can start before the input (see Fig, S1.7-7d), and the system is not causal ye) f08 YD say Figure $1.7-7 (a) Irvertible because the input can be obtained by taking the derivative of the output. Hence, the inverse system equation is y(t) = df/at (b) The system y(t) = f(8t—6) = f(3[¢—2)) represents an operation of signal comipression by factor 3, and then time delay by 2 seconds. Hence, the input can be obtained from the output by frst advancing the output by 2 seconds, and then time-expanding by factor 3. Hence, the inverse system equation is y(t) = f(§ +2). Although the system is invertible. itis not realizable because it involves the operation of signal compression and signal ‘advancing (which makes it noncausal). However, if we can accept time delay, we can realizes noncausal system (€) Not invertible for even values of n, because the sign information is lost. However, the system is invertible for odd values of n. The inverse system equation is vt) = [f(t]. (4) Not invertible because cosine is @ multiple valued function, and cos—'[4(0)] is not unique, he loop equation forthe circuit i Sun(t) + Delt) = f(t) or (D+ 3)n(t) = Fe) a Ako Dralt) = val) = (0) = Saal ®) Sutton of (2) in (2) yields Prd ne) = 10) of (D+3)nl0)= DIC) 3) ‘The currents in the resistor, capacitor and inductor are 2ya(t), Dya(¢) and (2/D)ya(t), respectively. Therefore (0424 Zr = 10) (D? + 2D + 2)alt) = Ds(t) a) isoLes init) = Dual) or lt) = boat) Suhatitting of (2) in (1) yields Dix2D42) Pee 2 y= DI) (D? +20 + 2)ni(t) = D*s(t) la(t) ~ go(t))At = AA ne) = Hlaae) ~ (0) But gol) = RAC) Diflerentaton of (2) yields 8 felt) = RAC) = Flee) ~ et0] and (0+ #) eto = Fat (D+ e)ett) = ante) an4 and . wl = hy substituting this in (1) yields y= 4 (01- ph) a= (D+ a)h(t) = @ @) w e@ a a)Chapter 2 22a 22a 2.24 = 0, Alo 1? +5046= ‘The characteristic polynomial is A? + 5A +6. The characteristic equation is A? + 54+ (4+2)(043). Therefore the characteristic roots are Ay = ~2 and Az = ~8. The characteristic anodes ate e aud e-". Therefore Setting and substituting initial conditions yo(0) = 2, io(0) = ~1 in this equation yields refore wo(t) = Se — se 0. Also AF 4 4A44= = aud "The characteristic polynomial is 4? +4444. The characteristic equation is 1? +4444 (442), so that the characteristic roots are ~2 and ~2 (repeated twice). The characteristic modes are e te, Therefore elt) = cre" + cate and siolt) = ~2eye™® ~ Deate™™ + ene Setting = 0 and substituting intial conditions yields Bea a3 d= Dey tee =? ‘Therefore w(t) ‘The characteristic polynomial is A(A+ 1) = A?4+A. The characteristic equation is A(A-+1) roots are O and —1. The characteristic modes are 1 and e~*. Therefore ‘The characteristic elt) = 04 + e207 and bolt) = ~eae Setting ¢ = 0, and substituting initial conditions yields ates ‘Therefore welt) characteristic polynomial is A? +9. The characteristic equation is A? +9 =0 or (A+8)(A~ 53) characteristic roots are +)3. The characteristic modes are ©" and e~?°". Therefore The welt) = eeas(3t +8) and in(t) = ~Sesin(3t +0) Setting ¢ = 0, and substituting initial conditions yields en2 o=-n/2 ‘Therefore225 227 2d = 2sin3t ‘The characteristic polynomial is A? 44413. The characteristic equation is A? +4A+13 = 0or (h+2-78)(A+ 2+ j8) = 0. The characteristic roots are ~2+ 33, The characteristic modes are cye'2*!9"" and ego?" Therefore ole) = re" e083" + 8) “ Holt) = ~20e-* cos(3t + 6) ~ See" sin(3t +6) Setting ¢ = 0, and substituting initial conditions yields 5 = econ = cos = 5 }= o=10 15.08 = -2rcos@ ~ Sesind. cain d = ~8.66 o=-n/3 ‘Therefore a(t) = 10€-* con(st ~ $) ‘The characteristic polynomial is \7(A + 1) or A* +27. The characteristic equation is M%(A 41) = 0. The characteristic roots are 0, 0 and —1 (0 is repeated twice). Therefore wit) =o teat tee" and Setting t = 0, and substituting initial conditions yields deate ass Bea-o hee ma -I=6 ae-l ‘Therefore lt) =542t- 6 ‘The characteristic polynomial is (X1)(A? +5A +6). The characteristic equation is (A-+1)(A? + 6-46) (A+1)Q+2)(443) =0. The characteristic roots are ~1, =2 and -3, The ch dor teristic modes are ef «#* and c™. Therefore and Setting ¢ = 0, and substituting initial conditions yields 2eatare -1= =e: ~ 20 ~ Se) => Sart deat 99 Therefore sole) = bet — Te“ + eH ‘The characteristic equation is 47+ 44+ Therefore = (A+ 1A +3) = 0. The characteristic modes are «* and «® walt) = e168 + exe Unt) = ee" ~ Beae™ Setting # = 0, aud substituting v(0) = 0. (0) 1, we obtain oan ter mei — Sea 10‘Therefore watt) = let = 94) A(O) = [PD)wa(tu(t) = [(D + 5)u] u(t) = lat) + Syn( tule) = (et ~ em yult) 2.3-2. The characteristic equation is P+ 5A.46=(A+2)(4+3) = 0. and alt) = ere + e207 Jinlt) = ~Beve™* ~ Bege™ Setting ¢ = 0, and substituting y(0) = 0, (0) = 1, we obtain O=nter Be ~ See Therefore alt) = and [P(Dyyalt)hu(t) = Wale) + Tinlt) + Mallat) = (em +e" u(t) Hence a(t) = buble) + [P(D)vn( tule) = S(t) + (em + elt) 2.8.3 The characteristic equation is 4+ 1=0 and walt) = oe In this case the initial condition is yi" (0) = y(0) = 1. Setting ¢ = 0, and using y7,(0) = 1, we obtain ¢ = 1, and a(t P(D)unt) = [-iiat) + yn (O]ult) = 2eFu(t) Hence ‘(t) = b(t) + [P(D)yn(t)]ule) = ~8(t) + 2e*u(t) ‘The characteristic equation is A? +6A+9= (243)? = 0. Therefore alt) = (er + eathe™ Gin(t) = [Blea + eat) + eale’ Setting ¢ = 0, and substituting yn(0) = 1, a(0) = 1, we obtain o=a = Be $02, and a(t) = te Hence A(t) = [P(D)va(u(0) = (inl) + Sunl@)]u) = (2 +3 u(t) fom [aoa = [[reom-nale wal reclaes = ay [atte = Asay 24-1 u‘This property cab be readily verified from Examples 2.7 and 2.8, For Example 2.6, we note that 5, and AyAn For example 28, Ay = 2. Ay = Leavtaa [2 ye ~per tas [Gears [em oye “4 sii ArAg 2a Het) aot) =f Henaott— ar #1 [7 towers wtee) a0 ‘When a <0. the limits of integration become from oo to —o0, which is equivalent to the limits from —o0 to 90 with a negative sign. Hence, f(at) «g(at) = |2le(at). 24-3 Let {(i)+9(t) = (2). Using the time scaling property in Prob. 2.4.2 with a = —1, we have f(—t)+a(-t) = e(—0) Now, if f(€) and g(t) are both even functions of t, then f(@) = f(t) and g(t) = g(t). Clearly «¢) Using a parallel argument, we can show that if both functions are odd, o(t) = e(-~), indi But if one is odd and the other is even, () = ~c(~t) indicating that c(t) is odd 2nd 245 wot arm fare fort co Therefore u(t) + u(t) = tule) (li), Because both functions are causal otal) eet) et dp and emt u(t) eMule) = ut) Because both funetions are eausal 122.4.6 247 28 eo onc f ru(r)u(r — tar Seer ee ee ee ee re wee [rue d ee and \ tu(t) ule) = FPuco © sintu(t) «ue = ¢ sin ru(r)u(e ~ner) ue Becaue rand {+ are both nonnegative (when 0 < + <4), u(r) = u(t~+) = 1, and sinew eo = (snr ar) w= = conte (4) Sinitcty conta) = ( [ara nate In this problem, we use Table 2.1 to find the desired convolution (=) we u(t) « ut) = (= eu(t) (b) ate mty(t) = temult) (e) ult =e ule) (4) ple) = sin 32) + e~Fu(t) Here we use pair 12 (Table 2.1) with a = 0, 8 = 3,@= .90° and A= —1. This yields ~1084" sod ty py = LOBIBAPIEN* — coslSt+ 18.4) sinseu(d)«e“tu(e) = <8 (e) ) H(t) = (20° — o™)u(t)« u(t) = 2e-Mult)« ult) — eMule) u(t “> =] Let) a(t) co) (2 = Yule) « eue) = 2eMule) + elt) — eMule) ee) -f |x =(e* ue) © u(t) = (267 = e-*jult) +e“ u(t) = 2e~Mu(t) » e ult) — [ut - a(t) eMule wo] 0) = [(2— He? — 26 Jule) 1324-9 a(t) = (1 26 Mai(t)eu(t) = eMule) #H(e) — He -Mult ult) -)] uo = tule) 24-10 (a) For y(t) = 4e-** cont u(t) + u(t), We use pair 12 with a = 2,8 =3,0=0,\=0. Therefore and | ue) = A [0.555 - «const + 56.51°)] u(t) (b) For y(t) = te-* costu(t) «e~*u(t), we use pair 12 with a = 2, 8 =5, @=0, and X= —1, Therefore ono and sag [cont 50"}er* ~ e°* conte + 71.56") wo =4 [ ay | ue) = pg [ost86n* — econ se + 71.507)] uC) 1 iam, . Tage conse + 7.56 | u(t) 24-11 (a) wit em * ult) (0) HM a(e) = Foul), and y(t) = o [e-Fu(t) «em u(e)] = ehlert = ents) (c) “u(t 3) = ee" u(t —3). Now from the result in part (a) and the shift property of the convolution (Fa. 234) (o-By(e) = oH] Ge — 3) i(t) ~ u(t ~ 1). Now y:(t), the system response to u(t) is given by (@ 0 alt) = elt) eult) = (1 =e Dult) ‘The system response to u(t ~ 1) is a(t~ 1) because of time-invariance property. Therefore the response y(t) to Jt) = u(t) ~ u(t = 1) is given by ut) = yale) ~ w(t 1) = (1 =e u(e) = [1 eMule = 1) response is shown in Fig, $24-11, ye) yay +> pik SP) Fig. $2.4-11 “-t é Yt “Ne lo t= ol t= Fie S242 ie) = fate) Tle} + efn(—t) ~ #0) set) + 20 u(t) etal —t) welu(—t) + [etu(t) + efu(-)] alo) nets 1 1 nwo = [7 pana Because u(t) =1 for r < t and is 0 for r > t, we need integrate only up tor 1 an =e “yt mae fi hyena tall a taeteed war n= f) ayy earl attest Figure $2.4.19 shows =7hy snd e(t) (the result of the convolution) HDs ol t tt Fig. 82.419, 2A-14 For t-< 2e (see Fig. $2.4-14) (0) = st) 90 sint dr cost OS <2 For ¢ > 2n, the area of one eyele is zero and M0 +9() =o. 122" and t2.AA1S For 02 © w= [ a toa a) I arated -1202-4 dy=0 te-4 @ 16(e) @ @) ae se eG t He tT 3S E 7d wal “77 cu) ; et ee Oo a4t to i ces oI +> eee anny cc) : : i—™ ey C $3 + a z > : vr ' Pai ect) syp-——_ (3) ot > ti 75 tS = z ozte-l -lote-2 f70 ete) a) “2 tele e wR met lO ite -2KE 2-1 4 Fig 524-16 1"2417 © © ans fe ao= fore (= 0 <0 (g) This problem is more conveniently solved by inverting fy(t) rather than f(t) ao [e-oeet exo a= ["e-ndr= fae) 0202 )=0 for t>0 ) AWw=e fle) Ale, flt-nae—, a= fi cemrarnen fl erar yee) open a aiy= fee t-ar emmy bean? tt ts-2 Indicating the input and corresponding response graphically by an arrow, we have FW) — vo) St-7) wt) $() - Ht- 1) wl) — eT) (by Timeinvariance) (by linearity) ‘Therefore Jim, Ef) ~ 40-79) — him, EIy(e) - eT} ‘The bft-hand side is f(t) and the right-hand side is (t). Therefore fo 0) Next we recognize that semen fi poe ® “This follows from the fact that integration is performed over the range —oc < 7 < &, where tz) =1, Now the response to Jt f(r) a is St Hence Ct) ¢ ule] Ae) = UH] © u(t) = V1) # a) 182.4018 2.4019 24-20 2.4.21 26-1 But as shown in Ea. (1). vt) «ut is f' u(r) dr. Therefore the response to input f',. f(r)dr is f, y(e) dr Using the hints, we obtain 1 - 1 = tim, 4 (e) = eft 79) = A AO = 9) = Bim, 7 CO ~ HO TY] + 90) = Fe) + Bim, [al ~ oft TI] = J Fle) ~ elt TH] = AO) Successive applications of the above procedure yields £8) ge) = MM (ED ‘The system response to u(t) is g(t) and the response to step u(t — 1) is a(t 7). The input (0) is made up of step components. The step component at r has a height Af which can be expressed as 4, AL new friar Blane = frye ‘The step component at ndr has a height /(ndr)Ar and it can be expressed as [f(ndr)A7]u(t —nAr). Its response A(t) is Ault) = [nar arlatt nd) ‘The total response due to all components is ut) = fim, Y> flndr)g(t—ndr\Or = [fate nar= fe e960) ‘An clement of length Ar at point mdr has a charge f(uAr)Ar (Pig. $2.4-20). The electric field due to this charge at point x is Sndr)ar rele — nr? ‘The total feld due to the charge along the entire length is AE= B()= im, x _linanan, free = nr)? «fo glia er = seve | tae re ae oO, AT Zz a 2-NbT-= Fig. 52.420 For an ideal delay of T secs., the impulse response is h(t) = 6(¢—T). Hence, from Eq, (2.48) (using the sampling property (Eq. (1.24b)] Hs) Or = Tye dr We can also obtain the same result using Bq. (2.49). Let the input to an ideal delay of T seconds be an everlasting. exponential e*, The output is eT), Hence, according to Eq. (2:49), H(s) = eT) jert = eT, METH ID=(AS3)(A44) 192.52 25-8 ‘The natural response is volt) = Kie™ + Kae (a) For f(t) = ult) =e"u(t), volt) = HO) = B= b Y= Kae + Kae + H(t) = Bre — 4Kae* Setting t = 0 and substituting initial conditions, we obtain Reto ~8Ks - 4K and 120 &) HO =u, = Ke + Kae + Be 8Kie"™ — 4Kae Setting t = 0, and substituting initial conditions yields On Kit hath 1=-3K; — 4k and u)= u(t), volt) = H(-2) © 10 ult) = Kie™ + Kae 0) = BK e™ = AK Setting ¢ = 0, and substituting initial conditions yields Kit ks } ul) 2? = 6425 = (A+3— j4)(A+3 + JA) characteristic roots are ~3:4 j4 alt) = Ke cos(at +8) Por fi) =u), volt) = 100) Esothat u(t) = Kem cos( at +0) + 3 i{t) = ~3Ke-™ cos(t +8) — 4Ke™™ eas( at + 0) Setting t= 0, and substituting intial conditions yields K=0477 0= Keos0+ & }- Koso 9 =~1003 BK cos 8 ~ 4K sind Keind and ult) = 0427 con(at 1088) 4+ 3 120 Charecteristic polynomial is 4° + 4A +4 = (A+ 2)*. The roots are ~2 repeated twice. lt) = (Ka + Kade 20(a) For f(0) = eMule). yolt) = H(-3) = 20 wt) = (Ki + Kathe" ~ 20° HO) = 2K + Kate™ + Kee + oe Setting ¢ = 0, and substituting initial conditions yields and v(t) = (2+ Boe 20 t>0 (b) FQ) = ertult), volt) = H(—1)en* Hy + Kathe 2K + Kate + Kaen? Setting t=, and substituting initial conditions yields gem Kad 2K + Ka Kak and uit) =(3+ Boe* — t20 Because (A? +22) = A(A+2), the characteristic roots are O and 2. alt) = Ki + Kae In this ease /(t) = u(t). The input itself is characteristic mode, Therefore volt) = Bt But ye) satisfied the system equation (D? + 2D)volt) = (D+ I)vlt) = elt) + Bolt) = HO + £0) Substituting f(t) = u(t) and yo(t) ‘Therefore yo(t) = # Setting ¢ = 0, and substituting init and ‘the natural response a(t) is found in Prob. 2.5-1 alt) = Rie + Kae" The input 4(0) Isa characteristic mode. Therefore elt) = te Also yo(t) satisfies the system equation (0? + 1D + 12)volt) = (D+2/(0) a26-1 2.62 2.64 Holt) + Tielt) + L2y9Kt) = Fle) + 240) Substituting f(t) =e and ye(t) = te in this equation yields (9Bt — 6a)e™* + 7(—39t-+ Ble“ + 12Bte* = —3e~ + De Be ‘Therefore lt) = Kiet ae = tet it) = ~BK xe ~ AK ae“ + 38 — Setting ¢ = 0, and substituting intial conditions yields and (a) 2 480412= (04246) Both ots are in LHP. The system is asymptotically stable. (b) MAP +3042) = A(A+ 1A +2) Roots are 0-1, ~2. One root on imaginary axis and none in RHP. The system is marginally stable (©) MQ? 42) = 804 iVDO - jv) Roots are 0 (repeated twice) and V2. Multiple roots on imaginary axis. The system is unstable (d) (A+ 1)? - 6A +5) = (44 1-1) = 5) Roots are =I, 1 and 5. Two roots in RHP. The system is unstable (@) (4104+ 2445)? = 4 A+1- 32704143277 Roots “1, ~1 2 (repeated twice) are all in LHP. The system is asymptotically stable () 408 +9) = + NA+ 59) — 33) Roots are ~1, +3. Two (simple) roots on imaginary axis, none in RHP. The system is marginally stable. (©) ANE! +97? = 04043970 - 33)? Roots are ~1 and +33 repeated twice. Multiple roots on imaginary axis. The system is unstable (A) (P41)? + 4)(AF +9) = (AF DIA = FDA + 52)1A ~ 92) + 59) ~ 53) ‘The roots are 4j1, 42 and 73. All roots are simple and on imaginary axis, None in RHP. The system is marginally stable (a) Because u(t) = eu(t), the characteristic root is 0 (b) Tae root lies on the imaginary axis, and the system is marginally stable fe) fone) dt = co ‘The sistem is BIBO unstable. (a) Tae integral of 61) is u(t). The systems response to 6(t is u(t). Clearly, the system isan ides! integrator ‘Assume that a system exists that violates Eq. (2.57) and yet produces @ bounded output for every bounded input. The response at ¢ =f is ute) = [ merstn = nar Consider a bounded input f(t) such that at some instant ty fp ita(r) > 0 ta-ne{t nee In this ease 2ara ara ara oa ~ Inn) and wery= [more This vlntes the assumption (a) The time-constant (rise-time) of the system is Ty = 10°* ‘The rate of pulse communication “3 2 Za 4 54> Fig. $2.41 Plt) ZiT war = ar I oo t= Fig. $8.42 ‘We can show that the corresponding signal pairs are also orthogonal [Lnonea= [ esvoisevoiee= [Untoncnar= [7 ty ~zxojent + 2eo)ae=0 [Lnonoa= [ne +2noiee0 +2900) In deriving these results, we used the fact that J zfde= J 23(t)de = 1 and f%, 21(thza(¢\d Bd Here Ty = 2, s0 that wp = 25/2 x, and W0)= 00+ Sroncosnnt+éqsinnet = -1StS1 where Psinnetdt = po=be 3S wisest Figure $3.4-1 shows /(t) = for all ¢ and the corresponding Fourier series representing f(t) over (—1, 1) 8.4.2 Here Ty = 2x, s0 that wo = 2n/2n = 1, and S(O) = 00+ encosnt+ basin nt areten where 3-18 2° 2" Tre . no saci Shene oases Fe $93.42 stows 0) =f aad th sreponding Fre si to pret J) (2 34-8 (a) To = 4, wo = HE = F. Because of even symmetry, all sine terms are zero. 6ao = 0 (by inspection) oun [foe (s) a [coe ) a] Aint ‘There‘ore, the Fourier series for f(2) is wo 4 (ogg tt — Veg Mt, Legg Skt 1 Tat (cmH — gore F + Seon — joo Bt 4) and we allow Cx to take negative values, Figure $3.4-3a shows the plot of Cy ‘}. Because of even syrimetr, all the’sine terms are zero. oon ae [oo (5) = se SG), ee (F) bem Ee [sin(Ze) deo Gntegand i an 04 fntion of ) for n = 0, 1, 2,3, --». Figure $3.43 nd we allow Cy to take negative values. Note that C, shows the plot of Cn (©) Te= 2, up = 1 F (0) = 29+ Jrancosnt +bxsinnt with ao=0.5 (by inspection) [gemnane af 5-1 (ines Janars einai + banaey--) and f= = 05+ [oos (t+) + feos (2t+$) + feos (a+ $) +] ‘The reason for vanishing of the cosines terms is that when 0.5 (the de component) is subtracted fom f(t), the remaining function has odd symmetry. Hence, the Fourier series would contain dc and sine terms only. Figure '$8.4-2¢ shows the plot of Cn amd By, (@)% ‘x, wo =2and f(t) = St, (by inspection). (50) because of odd symmetry. aan ae 1) + ewe (40-4) + glyco (se 3) + Leo (s+ 4) + igure $3.4.3d shows the plot of Cy and 0, Se) = Aysinze + Linge — A; singe — 2 singe + aTta af we ll @) ape ee ay. > can =I wor aa Tiga eet e t ate a x [| «) 12> +5 6 we 123456 wo qd) - (e) A) x F ad 5 w> oo Fig snes (e) Ty = 3, wo = 20/3. if 1 wed feet * Qne 3 Qen | 2an | Qen eos Bede gS en 2 4 2 in 22 aan 240 Bem gg 28) ‘Therefore Co = } and [Eee(8) To= 6, wo = 9/3, a0 = 05 (by inspection). Even symmetry; by se ‘Observe that even harmonics vanish. The reason is that if the de (0.5) is subtracted from J(t), the resulting function has hale-wave symmetry. (See Prob. 3.4-7). Figure $3.4-3f shows the plot of Cn. Bide (a) Here 7) =r, and wy = #8 = 2. Therefore S(O) = 00+ J an.cos 2nt + bn sin Int ‘To compute the coefficients, we shall use the interval x to 0 for integration. Thus wont [ear =0soe en meat 0504 (525) mead fen amar = -0504 (85) ‘Therefore 0504, C= VaR R= 0504 (Toe), Om =tan* (2) = tan tn 1) = 0504+ 05045 eBay cos (ant + tan An) (b) This Fourier series is identical to thet in Eq. (8.56a) with ¢ replaced by ~ (©) IEF(0) = Co + SCn cos(rwot +), then S(=t) = Co+ J Cn cos(—nwot +n) = Co+ > Cn cos(nwot ~ Bn) ‘Thus, time inversion of a signal merely changes the sign of the phase 0. Everything else remains unchanged. Comparison of the above results in part (a) with those in Example 3.3 confirms this conclusion, 84-5 (a) Here To = x/2, and wo = HE = 4. Therefore SO= 20+ J ancos dnt + basin dnt where = and bm ThereloreSM, Co= VFA =0506(sedo). b= —tan en (b) This Fourier series is identical to that in Eq. (3.56a) with ¢ replaced by 2 (©) HF) = Co+ S Ce cos(nant + 6m), then. Lat) = Co + JY Cn cos(n(auo)t + 8) ‘Thus, time scaliig by a factor a merely scales the fundamental frequency by the same factor o. Everything clse remains unchanged. If we time compress (or time expand) a periodic signal by a factor a, ite fundamental frequency increases by the same factor a (or decreases by the same factor a). Comparison of the results in part (a) with those in Example 3.3 confirms this conclusion. This retult applies equally well 3.4-6 (a) Here Ty = 2, and wo = JE = x. Also (¢) is an even function of ¢. Therefore S(t) = 00+ ancos net ‘where, by inspection ag = 0 and from Eq, (9.660) S Jf 20+ 100 mtd = sig own b= (Oe oo Therore 84 1 1 1 [ow nes Bente doco Set + Boen Tet] (b) This Fourier series is identical vo that in Eq, (8.63) with ¢ replaced by t+ 0.5. (€) IF =Co-+ SCacos(nsot +8), then Mt+T) = Co+ J Cn cosnsio(t + 7) + Aq] = Co+ S Cn cosinwot + (Om + nwoT)] ‘Thus, time shifting by T merely increases the phase of the nth harmonic by nuu?’. 8.4-7 (a) For half wave symmetry 4) =-1 (v4 B) sp 1 [Po reconcorars [" so coanvotae lyr wot aed [" tonnes te Let z= ¢~T9/2 in the second integral. This gives E[[ wrommatts [°° sor) mn e+) a] mt mar -2 if Hocornentat +f ~f(0l-cornsid] pa -4 if He conta] In a similar way we can show that 4 ne “Th 4, a0 =0 (by inspection). Half wave symmetry. Hence a-t[ffsomtgia] <3 fonttia] iis (cos SF + Bsn 3-1) (n oad) bn J(d)sin nupt dt () @ To=8, welt = sha (Fain -1) tn oaay 30Therefore Siaiaely ond fo= (li) 75 = 2x, wo = 1, a =0 (by inspection). Half wave symmetry. Hence HO= So eneaant + desinnt aon 2 [eM conneae <2 ee torent sans] 6 tt woo fe af ero A =? [owen Om on] ferro 0.05 ~ Tea FOOT) - +00 and oP sin ntdt etn . [ (-0 sin —nosn] (oad) +001 Bn (ano) = 14610 ore oon” =F e 001 3.4-8 (a) Here, we need only cosine terms and wo = $. Hence, we must construct a pulse such that it is an even function oft, has a value t over the interval 0 < t < 1, and repeats every 4 seconds as shown in Fig, $3.48a. We selected the pulse width W’ = 2 seconds. But it can be anywhere from 2 to 4, and still satisfy these conditions. Each value of W results in different series. Yet all of them converge to ¢ over 0 to 1, and satisfy the other requirements, Clearly, there are infinite number of Fourier series that will satisfy the given requirements. The present choice yields 1) = 40+ Yano (2S) By inspection, we find ay = 1/4. Because of symmetry ba = 0 and alo (F)+F@()-] (b) Here, we need only sine terms and wo ‘Hence, we must construct a pulse with odd symmetry, which has. ‘2 value ¢ over the interval 0 < t < 1, and repeats every + seconds as shown in Fig. S3.4-8b. As in the case (a), ‘the pulse width can be anywhere from 1 to x, For the present caseBecause of odd symmetry, an = 0 and dialoinan — 2ncor2n) (©) Here, we need both sine and cosine terms and wo = §. Hence, we must construct a pulse such that it has no symmetry of any kind, has a value ¢ over the interval O< ¢ <1, and repeats every 4 seconds as shown in Fig. ‘S3.d-Be, AS usual, the pulse width can be have any value in the range 1 to 4 10 = 00+ Sooneos (3) e+ basin ($e By inspection, a9 = 1/8 and a Fig. $3.48 (€) Here, we need only cosine terms with uy = 1 and odd harmonics only. Hence, we must construct a pulse such that it isan even function oft, has a value ¢ over the interval 0 < t < 1, repeats every 2x seconds and has half-wave symmetry as shown in Fig. $3.4-8d. Observe that the first half eycle (from 0 to x) and the second half eycle (from x to 2) are negatives of each other as required in half-wave symmetry. This will cause even hharmonics to vanish. The pulse has an even and half-wave symmetry. This yields M0) = 00+ SP an cosnt By inspection, ao = 0. Because of even symmetry by = 0. Because of half-wave symmetry (see Prob. 3.47), apy? 7 2 25.0% A comma- [te-ounca] = Zeon -r Ean not (c) Here, we need only sine terms with wo = + and odd harmonies only. Hence, we must construct a pulse such that itis an odd function oft, has a value ¢ over the interval 0 < ¢ < 1, repeats every 4 seconds and has half-wave symmetry as shown in Fig. $3.4-8e. Observe that the first half cycle (rom 0 to 2) and the second half cycle (from 2 to 4) are negatives of each other as required in half-wave symmetry. This will eause even harmonies to vanish. The pulse has an odd and half-wave symmetry. This yields 32n= Sm By inspection, ay = 0. Because of odd symmetry aq = 0. Because of half-wave symmetry (see Prob. 3.4-7), 4 ox oe seed [antars fees (f) Here, we need both sine and cosine terms with up = 1 and odd harmonies only. Hence, we must construct & puilbe such that it has half-wave symmetry, but ueither odd nor even symmetry, has a value ¢ over the interval OS t < 1, and repeats every 2x seconds as shown in Fig. S3.4-8f. Observe that the first half cycle from 0 to 17) and the second half cycle (from = to 2x) are negatives of each other as required in half-wave symmetry. By inspection, ap = 0. This yields ne, ae in rat = Basin" n odd Fit) = SO ancosnt + basinnt Because of half-wave symmetry (see Prob. 3.4-7), stxloentmsinn=1) bea gt f'tanntar= 26m neem) nal a bede f g hi periodic? yes yes no yes no yes yes yes yes o Dio bb. 2 period = Or Oe Mor arr 34:10 e a F(t) = 00+ San cos deat + basin ent (= 2) ae roa ff tae} ono f teos2entdt=0 n> 1. (ninteger) mone [esinzenea= Hence+ (seas Jones Lenars) 1S nae If E4(N) is the energy of the error signal in the approximation using first N’ terms, then From Eq, (3-40) sion {Poa [Qe (zy +-+( (Note that By = 1/2 for n ooras7s 3.411 L(t) = carole) + er2a(t) ++ erzr(t) sinees =f * eat 1 [tonto = 5 asf fonioe= ° a © = [tosoa= om [romtoa= Hence 1 Law jo0- fa - Zn 1 = Boat) [ Po# If E-(W) is the energy of the error signal in the approximation using first N’ terms, then From Eq. (8.40) Also‘The Walsh Fourier series gives small error than the trigonometric Fo ‘number of terms in the approximation. ier series (in prob. 3.4-10) for the same saa HO = cool) + came) ++ esl) [10d=0 eo amcence=on soae-t onLf wolfe-fae? Hence Also P(e)de=2 and using Bq. (3.40) ea = [Poa ea)= [Pow (b) This is a scaled version (time-expansion by factor 2n) of the signal f(t) in pair a wo-i(2)--20)-4@) 40] - 3.51 (a): Ty = 4,0 = #/2. Also Do = 0 (by inspection). 2 Lf! sone, = aint Dam x fl overnae- = Zein iniza (b) Te = 108, up = 25/107 = 1/5 =F Det, where a FO = Dre! wh m= © Ji) = Do+ Y> Dae’, where, by inspection Dy = 0.5 $ n>0 Da= hema= 2, sothat [Dyl= cb, and wn {i > Ps Im Ten neo (@) = 5,40 =2and Da =0 HO= SS Dae, where (©) Te= 3.00 = % wherero + jos Ph W qt “2 Fle ee Fig. $3.5-1 ‘Therefore Ea a aco. 3 wos aay (6) T= 6,09 = 4/3 Do = 05 roel mak [[erme tars fas [cone te 363.82 In compact trigonometric form, all terms are of cosine form and amplitudes are positive, We can express f(t) as S(t) = 3+ 200s (20 - F) + cos (st 3) + 3 con (se + 5 - x) 29.420 (ot £) ene (0-- 5) + ee (428) From this expression we sketch the trigonometric Fourier spectra as shown in Fig, $3.5-2a. By inspection of these spectra, we sketch the exponential Fourier spectra shown in Fig. $3.5-2b. Prom these exponential spectra, wwe cas now write the exponential Fourier series as a AB 4 ORD g War gg Hever gy y Leste aey 4g Usa sa=3 By L0r# Dy +t w> ¢ en Cn =m 7/2. 12345 Wo -27 Dal LDo t . t Tasts “35 we o> Fig. $3.5-2 3.5-3 (a) The exponential Fourier series can be expressed with coefficients in Polar form as LO) = VBE I 4 209 434 20 4 (BV FE IIe From this expression the exponential Spectra are sketched as shown in Fig. $3.5-8a (b) By inspection of the exponential spectra in Fig. $3.5-2a, we sketch the trigonometric spectra as shown in Fig. §3.5-3b, From these spectra, we can write the compact trigonometrie Fourier series at Slt) = 3+ 4eos (1 £) + 4vFcos (at - F) (c) The lowest frequency in the spectrum ix and the highest frequency is S. ‘Therefore the bandwidth is 3 rad/sor 3 He 3.544 (a) F pom 37 sl fe) = #@-7)£Dn Te enee "3 r2 or a =] t oes "Ia - 4 a 3 On 1423 w> TH: se 123 Os Fig. $3.5-3 Bn = Dye so that [Dal @) 3.5-5 (a) From Bxercise E3.6a no=behSPreont — -1ees1 ‘The power of f(¢) is 198 For N = 1, Pe = 0.1111; for N = 2, Py = 0.19823, For N = 3, Pz = 0.19837, which is greater than 0.198. Thus, N = 3, 8.5-6 (a) From Exercise E3.6b po Mew Sedan ansese 383.67 3.61 ‘The power of /(t) is Moreover, from Parseval’s theorem (Ea. (3.82)] +S ys (b) Ifthe V-term Fourier series is denoted by 2(2), then 2) = An SE sinnnt arstsn ‘The power Ps i required to he 0.904% = 0.342, Therefore For N = 1, Pe = 0.202647; for N = 2, Pe = 0.253347, for N = 5, Py = 0.296584? for N 0.302224, which is greater than 0.342, Thus, N= 6. ‘The power of a rectified sine wave is the same as that of a sine wave, that is, 1/2. Thus Py = 0.5. Let the 2N +1 term truncated Fourier series be denoted by f(t). The power Py is required to be 0.9975P; = 0.49875. Using the Fourier series coeficients in Exercise E3.10, we have Pe . 24 S 1 yo Sit 3 aap tae Dian eng sh sun se y= 2 irs at ae ae rt Bee ee eG AO 0a pommel wre. Te power of the error in the approximation of f(t) by /(t) is only 0.21% of the signal power Py. ee 4053 for N= 0 (only de), Py = 0.49535 for and wo = 2, and 0.504 a T+ 34m and from Eq. (3.74) De ——_ rT $ mot = 31.08n Therefore, v(t) = > Duff nwo)” D wet 39Chapter 4 4a2 413 rays [Treas [7 piocmnt-3 [7 nosnare If {(¢) is an even function of t, f(t) sinwt is an odd function of t, and the second integral vanishes. Moreover, f(0)coswt is an even function of t, and the first integral is twice the integral over the interval 0 to co. Thus when f(t) is even Flw) F(t) coswt dt q) Sines argument shows that when (ti odd Fe) =a) [ sosinueat @ If (2) is also real (in addition to being even), the integral (1) is real. Moreover from (1) [fF Hoccnstae= Foe) Hence Fs) is real and even function of w. Similar arguments can be used to prove the rest of the properties. [rite bf newerena, ie 1F(u)|ooslot + creaes [~ [F(e)I infor + 2F (0) Since |F()] is an even function and ZF (.s) is an odd function of w, the integrand in the second integral is an ‘odd function of w, and therefore vanishes. Moreover the integrand in the first integral is an even function of w. ‘and therefore ro=2 [stone + ere Because f(t) = falt) + fo(t) and e-™ = cos wt + jsin wt Ftay= foes sdniermar= [ito + toleonseat—5 J (0) + Jo(0)Jsinwt dt Because fu(t)coswt and fo(t)sinwt are even functions and fu(t)coswt and J.(t)sinut are odd functions of, these integrals (properties in Eqs. (B.43), p. 38] reduce to Fanaa f” seremtar25 [~ ferinatat a Also, trom the results of Prob. 4.1, we have FU} a2 [" pcjosseat and 71s) 25 [” nlosinara 2) From Eqs. (1) and (2), the desired result follows 4041-4 (a) 7 a ection an [core (b) 4.1-5 (a) rays [ernie [arma (b) ° mare Pefonsr tursinur = tare f t This result could also be derived by observing that (t) is an even function. Therefore from the result in Prob. 4a. Fig. S4.1-66 (b) The derivation can be simplified by observing that F() can be expressed as a sum of two gate functions Fi(w) and Fe(w) as shown in Fig. S4.1-6. Therefore w= f [Filw) + Fwyleaa = {f om aos f ona} sin 2ebeint ° 7 41-7 @) 1 . =f [ coswe™ ds el, Baa (ites + sind a x -a=m™ (3) © Lt pyetan= bf"? ca HO ae fg Oe FLU Peorcorstde ts [7 rapsnstas] Because F() is even function, the second integral on the right-hand side vanishes. Also the integrand of the first term is an even function, Therefore 4aBe (a Om sine( £)veet (&) al gluat + oatineat =H 4.2: Figure $421 shows he plots of vrion factions, The function npr () i ete fnsion cere atthe critnandof width 2 Te anton npr) cab epee x & (rf) hii tral alse centered ste rg and af with 10/8.‘ ftom i part (ea ene futon re) dele by 10. In ther ovis gio pln contend at f= 10 and of wth BT fonction art (@) i ne ple centred We onigin an the et eo octng at say a fe t= 3 he faction npr () ae ple Soc) dae ty Toe For the sie ps sin() the rt tr ners at gm, that at oS Therefore the faction a Sine ple comer at = 10r Sd rman spaced at er of Sas shown nthe Bg State ‘The fncon past (Fa prods of agate pu (ceed the on) of with Ter and ane pulse lected a the arin) with oon raced at interes ofS Thi een the ste pe reed Boj th lara (12) se ahown Pig ane “ak = sine (2) ans “a gigtemm ammo om _ inet = sine(rde 424 (@) (b)4g. 432 (@) u(t) = x6(0) + “ is 70 3 Application of duality property yields n8(t)-+ 4 os 2eu(—w) go 3 aioe 1 I feos; | = ul-w) Application of Eq. (4.35) yields 1 1 a [a-9- Al = ule) But 6(t) is an even function, that is 6(—t) = 6(@), and 1 i Fe + 2) o> lw) (e) £08 wot <=> x6(u + wo) + Ale — wo] we For jon of duality property yields 16(¢-+ 0) + 8(¢ — wo} e=# 2008 (wow) ee=———eEer Fo who Setting wo = T yields S(t4 7) +6(t~ 7) <=> Boos Tes © ginal => jrl6(w + wo) ~ blu — wo] wo For Application of duality property yields Srl6(t+ 0) — 6(¢ — wo)] 49 2 sin( wou) = —2e sin(wow) Re aw) Setting wo = T yields 8(¢+.7) — 6(¢~ 7) = sin TH Fig. (b) fale) = S(-#) and Flu) = (robe Be + jue" Fig. (¢) falt) = f(t 1) + file 1). Therefore Fle) (Fw) + Fi(wile™ = [F(w) + Fla) em SF (osu + wsinw —1) Fig. (d) fa(t) = (0-1) + flt-+) File) = F(w)e™ + F(-w)e*Fig. (0) flt) = s(t 4) + t+ Hane Fula) = Faye? + (we! oe tee pe Sem a jue =a) J posin = sine (2) Fe. (£) (can be obtained intneenepe: (1) time-expanding J) by factor 2 i) then delaying it by 2 Bree fai sd ntplyng by 13 we tay interchange the sequence for tee (i) and (i) The ft ep {Gnwecpansion bya factor) yllde 1(8) a 20a.) = bole — jnoe™ =) Second step of te delay of2 ses i 1 (S52) — ge™ — me ne JNiplying the resulting signal by 1.5 yields my 1 = Gilt - ia ‘The third step of sn =254 (452) em ght s2™) aN ; sint ule sae” . mNe ntu 7 : wo si nt) Ute? naar aoa (eT) wel) Fig. $4.3 43-5 (a) 10 = ( wa (22 T and (b) Fiom Fig. 843-30 we verify that 10) = sintu(t) + sin(t— x)utt — =) Note that sin(¢ ~ x)u(t~ x) is sin tu(t) delayed by x. Now, sintu(t) <=» $[6(w~1) ~ 6fw-+ 1)| + zby and sin(t — m)u(t — mn) =e {glo = 1) = 6+ + 4444 Therelore -m) Floto 1) -oo+ 1+ Ap Mate Recall that f(2)6(2 ~ 0) = J(zo)6(z ~ xo). ‘Therefore &(w + 1)(1 +e") = 0, and Fl) ail eo) (©) From Fig, $4.3-3¢ we verify that 10 = cot [= £)] a cortate conta (e- But sin(t — $) = -cosé. Therefore 1 =cortu() +n (1 5) (e= Flo) = Flow - 9) + 6+ N+ (oto) swan) hy ic Blak Ye"? = 6(w + eH? = 4j6(0 41) Also because {(2)6(2~ 20) = S(z0)8(x - 20), ‘Therelore alia te (a) F(t) = e*[ult) ~ u(t ~ T)) = e*u(t) ~ e-*u(t~ 7) etult) ee eM M y(t T) we er ig tot FO)= sya jona® janet! iy From time-shifting property HOLT) ee F(w)etT ‘Therefore SWE T) + (t= T) > Pluie + Flw)e™? = 2F(W) cos ‘We can use this result to derive transforms of signals in Fig. P4.3-4 (a) Here f(t) isa gate pulse as shown in Fig. S4.3-4a. $(0) = reer (5) ep 2sinc(u) Also 7 = 3. The signal in Fig. P43-da is f(t+3) + f(¢—9), and E43) + F(t 3) 9 Asine() cos (b) Here /(@) isa triangular pulse shown in Fig. S4.3-4b. From the Table 4.1 (pair 19) 19 -0() 0 (G) Also T = 3. The signal in Fig. P4.3-4b is f(t +3) + f(t — 3), and 1049) 10-2) = dein? (2) ede 45Fe | =6 ' (ay a Fig. $4.3-4 4.85 Frequency-shifting property states that Fle) JIN w) = s(et0" es Plu Fu0) ‘Therefore Liners ~s0t {()sinwot = FU + Le ‘Time-shifing property states that 1 1 Fife eo) + Fle — 20) SET) = Pls) Therefore HET) ~ S(t = 1) > F(T ~ Ful = 25 F(w) sin and 1 BUT) ~ 1 1)] > Flw)sin Tw ‘The signal in Fig. PAS is f(¢+3) ~ f(t— 3) where 110 = 10 ($) a ante) ‘Therefore (t+ 3) — f(t — 8) <> 2j[28ie(w) sin 3} ) = 4) sine(w) sin 3 on Fig. $4.3-6, 4643-6 4-7 48 Fig. (a) The signal f(2) in this case sa triangle pulse A(z) (Fig, 843-6) multiplied by cos 10. $0) = () cos 108 ‘Aso from Table 41 (pair 29) A(z) => wsinc®(3!) From the modulation property (441) it follows that 1) = 8 (5) eo 11 > Zin? 2] sine? Es} ‘The Fourier transform in this case isa real function and we need only the amplitude spectrum in this ease as shown in Fig. S4.3-6a, (b) The signal f(t) here is the same as the signal in Fig. (a) delayed by 2x. From time shifting property. its Fourier transform is the sane as in part (a) multiplied by e~™""). Therefore 5 fone git sone Sep} om ‘The Fourier transform in this case is the same as that in part (a) multiplied by e~™, This multiplying factor represents a linear phase spectrum ~2rw. Thus we have an amplitude spectrum [same asin part (2) a well as ‘linear phase spectrum ZF (w) = ~2rw as shown in Fig. $4.3-6b, the amplitude spectrum in this case as shown in Fig, $48.6, Note: In the above solution, we fist multiplied the triangle pulse A(z) by cos 108 and then delayed the result by 2x. This means the signal in Fig. (b) is expressed as (525) cos 10(¢ ~ 2x) ‘We could have interchanged the operation in this particular case, that is, the triangle pulse A(t) is ist delayed by 2x and then the result is multiplied by cos 10¢. In this alternate procedure, the signal in Fig (b) is expressed as A522) con 10 ‘This interchange of operation is permissible here only because the sinusoid cos 10t executes integeel number of| cycles in the interval 2x. Because of this both the expressions are equivalent since cos 10(t ~ 2x) = cos 10 Fig. (c) In this case the signal is identical to that in Fig. b, except that the basic pulse is ret (34) instead of ‘triangle pulse A(:4), Now Fl) wnt (sh) = arsine) ‘Using the same argument as for part (b), we obtain F(e) = n(sinee(u-+ 109] + sincfe (ea — 10))}e-77™ (a) A There (b) to Lancy a 2 (2) There 2 inc? F(t) = Fsinc*(t) cos ae fa) ult) and u(t) eo 26(0) + ie jaa eMu(t) u(t), then a7430 4310 “Taking the inverse transform ofthis equation yields Leet f= Lee yu) () ut) em AG and eMule) oe IE 7) = eMu(t) «cult, then 1 xy FW)" Gora)Ges me) "wah Jesh ‘Therefore s)= (ule) N= © hu) = and e™u(—1) <=> ~ p IE f(e) = eMule) «efu(—t), then FO)" GoTayGeo de) "jena ea ‘Therefore 1 400 [eu(e) + e*u(—0)) Note that hecause Ag > 0, the inverse transform of = is e*fu(t) and not ~e™"u(t), The Fourier transform of the latter does not exist because Aa > 0. @ Malate - and &u(=t) ee — jo-m joe If f(t) = eM u(t) eeu(—t), then ‘Therefore (et = )u(—2) = Yam ‘The remarks at the end of part (c) apply here also, From the frequency convolution property, we obtain Pa Lew) + Fe) Because of the width property of the convolution, the width of F(j)+ F(w) is twice the width of F(w). Repeated application of this argument shows that the bandwidth of f"(t) is mB Hz (n times the bandwidth of f(0)) @ ) rays [mae [oman Bi -coour= Ha 484311 Aad >) © 8(¢+7) - 26(0) + (tT) ‘The Fourier transform of this equation yields juFlw) = OT 240° HT = 211 ~ cos wT] ‘Therefore rays [Troma wa Eo kf poema ‘Changing the order of differentiation and integration yields < f fume) [Cisse “Thersore oe E © zt 1 ae jute angen ant ~~ 1 uo gig Hw) = Oe ord (a)() 1 joa 1 wore y(t) = teu) FG Y(w)= © ye 42 wT jer je ult) = Fertu(e) + Seu(—t) @ Fle) ae) + ) ou) + 2 ves ig fw 3] 1 = rb) + et aay ecamses (==) = 1082) (eu 44-2 (0) roach mH and 1 roe gramme el ‘Thereoce ule) = 3 [eult) + eu) @) Fees and Hw) = 54g and 1 Y= GG ‘There‘ore ft) = fe = eu(-8) 443 Fi(u) = sine( gig) and Faw) = 1 Figure $4.43 shows F(w), Fa(u), Hs(w) and Ha(w). Now Yaw) = Fi(w) Aw) Yalos) = Fa(w)Ha(w) ‘The spectra ¥;(a) and ¥a(u) are also shown in Fig. S4.4-3. Because y(t) = ya(t)ya(t), the frequency convolution property yields ¥(w) = Yi(w) + Yalu). From the width property of convolution, it follows that the bandwidth of 50ae 00% I] He 20ereT O-> Ate i 445 446 4 £ 0a) 4 (9? Ela? -aoev0n | 200meT wo> eT J One Fig. 54.43 (wis the sum of bandwidths of ¥i(w) and Ya(w). Because the bandwidths of Y;(e) and Ya(w) are 10 kHa, 5 kHz, tespectively, the bandwidth of ¥(w) is 15 kHz. H(w) = 10" sine (Hg) and P(e) = 0.5 x 10" sine (zs) ‘The two spectra are sketched in Fig. S4.4-4. It consider P(w) to be nearly constant of value P(0) ¥(w) = Plu)H(w) =» P(O)H(w) = 0.5 x 10-°H(w) lear that H(w) is much narrower then P(w), and we may 10-*/2 over the entire band of H(w). Hence, = u(t) = 05 x 10°%A(t) Recall that (¢) is the unit impulse response of the system. Hence, the output y(t) is equal t» the system response to an input 0.5 x 10°*6(w) = A6lw). Hiw) <_ aad an 221P 1 Fig. S44 Hw) = 10" sine (35) ‘The two spectra are sketched in Fig. S445. It WF" SRIO é sod P(e) = 05sine(3) Fwd clear that P(w) is much narrower than H(w), and we may consider H(w) to be neatly constant of value H (0) = 10 over the entire band of P(w). Hence, Y (e) = Plu)H(w) » P(w)H(0) = 10° Pw) = ult) = 10-*pi) Note that the de gain of the system is k = H(0) = 10". Hence, the output is nearly KP). Every signal can be expressed as a sum of even and odd components (see Sec. 1.5-2) expressed as a sum of its even and odd components a aC elt) + hel) 51 Hence, h(t) can be45-1 Wi) Pi a> & af 4 9 Gy es e o> 1020 Fig. 4.45, EUn(O)u(t) + A(—O(—H] and Aol) = Hh(Ou(t) ~A(-Ou(—)). From these equations, we make observation that het) = helt)san(t) and hel) = helt) sgn (0) a provided that A(t) has no impulse at the origin, Tis result applies only if h(t) is causal. The graphical proof of this result may be seen in Fig. 1.24, Moreover, we have proved in Prob. 4-1 that the Fourler transform of a real and even signal is a real and even, function of w, and the Fourier transform of a real and odd signal is an imaginary odd function of w. ‘Therefore if Plu) = Rw) +5X(W), then Ae(t) => lu) and — oft) = 5X(w) @ Applying the convolution property to Ea, (1), we obtain = Ajxwy Zot [7 Xo. rule gxeese [She and 1 f* Rw) = RW) on™ Using pair 22 (Table 4.1) and time-shifting property, we get enltto te Me) hee) Figure $4.5-1 This is noncausal. Hence the filter is unrealizable. Also Linea £ Mi wee4.52 45-3 46-1 Hence the filter is noncausal and therefore unrealizable. Since h(¢) is a Gaussian function delayed by to, it looks 1s shown in the adjacent figure. Choosing to = 3V2K. h(0) = c~*® = 0.011 or 1.1% ofits peak value. Hence \V/TE is @ reasonable choice to make the filter approximately realizable. 2x08 0 He) Fron pair 3, Table 4.1 and time-shifting property. we get ny = ett ‘The impulse response is noncausal, and the filter is wntealizable he & to Figure $4.52 4x10 ‘The exponential delays to 1.8% at 4 times constants. Hence to make this filter approximately realizable. 4y. 40ys is a reasonable choice ‘he uit impale responses the inves Four ranfn of H(o) Hence, we have w= (aseet (sf) (eysne*0,00%0) (28 [All the three systems are noncausal (and, therefore, unrealizable) because all the three impulse responses start before t = 0 For (a), the impulse response is a rectangular pulse starting at ¢ = ~10~®. Hence, delaying the h(t) by 1 usecond will make it realizable. This will not change anything in the system behavior except the time delay of ljsecond the system response. For (b), the impulse response is a sine square pulse, which extends all the way to ~oo. Clearly, this system ‘cannot be made realizable with a finite time delay. “The delay has to be infinite. However, because the sine ‘square pulse decays rapidly (see Fig. 4.24d), we may truncate it at t = 10-4, and then delay the resulting A(t) by 10°, This makes the filter approximately realizable by allowing a time delay of 100 jseconds in the system response, For (¢), the impulse response is 1, which never decays. Consequently, this filter cannot be realized with any amount of delay. y= fT rode gs [7 ‘$y and consequently dt = Jydz Lie [Px Vor B= tae [eae 1 Fat Vino ‘Also from pair 22 (Table 4.1) Letting Fu) =e L [lirwrre-2 [7+ Letting ow = Jy and consequently ds = hyde ria Vie 1 a aaa. Brov8 ~ Io4.6-2 Consider a signal HO =sine(kt) and F(w) = Bp= J snctunan 2 fF fe (S)] 463° IF /2(0) cm A(u), then the output ¥(u) = A(w)H(w), where H(w) is the lowpass iter transfer function (Fig. $4.46). Because this filter band AF — 0, we may express it as an impulse function of area 4rQF. Thus, H(w) = UA F]6(w) and ¥(u) = [br A(s)AF|6(w) = [4 AOAFA(W) Here we used the property f(2)6(2) = f(0)4(z) (Ea. (1.23a)]. This yields u(t) = 2A)AF Next, because f2(t) ¢=> A(w), we have aw = f Pleat so that A(0) [ Plat = By Hence, y(t) = 28 ,AF. try aad x AIA S12 D ras- $ o> aT Oe Fig. 46-3 4.64 Rec that HO = ze [area ana [pine at = F-w) Therfore [normed £ [001 fete aie HE peif nmetaiae= 2 [ n-ontra Interchanging the roles of f(t) and f2(t) in the above development, we can show that [Cnontos= % [7 renee 4.6-5 Application of duality property (Eq. (4.1)] to pair 3 (Table 4.1) yields‘The signal eneray Is given by Bb [ amemtaante fe Es ‘The enerzy contained within the band (010) w= [ae 2m 16 By = 0906, then e788 001 mo Wr = 2802 rays = 2888 4.7-1 (i) For m(t) = cos 10008 oem.se(#) = m(t) 608 19,000 = cos 1000¢ cos 10, 00t 1 ‘lgo5 9000s + cos 11,0004 (li) For m(t) = 2c0s 10008 + cos 2000¢ Pose-sc(t) = m(t) €0810,000t = [2.c05 1000¢ + cos 2000t| cos 10, 000% cos 9000¢ + c0511.000¢ + Zfcos 8000 + 0s 12, 000%] = coor + obo + ont, 00+ Hon, 0 — eee Se i) For m(t) = cos 1000¢ cos 30008 Poan.se(t) = m(t 0s 10,0008 = 3 {cos 2000¢ + cos 4000t] cos 10, 000¢ = Hwee +n 00 + e000 + cn ¥en + conn + or 200-0400 ‘This information is summarized in a table below. Figure S4.7-1 shows various spectra. Mio? Modulated signal speetwn ay w 1 o> f o> ars Sipe tov 1 1, ARO 8 ttt tt 14 > aK -oK- BR SLO BR aK TR thee att tt t ~ AK OT aK HK “iv 712K -3 ER e t ae aR s] \ aK Fig. $4721 55Tass | Baseband frequency | DSB Requency [LSB Frequency | USE frequency ] i 1000 9000 and 11,000 9000 | 11.000 000 2000 and 11-500] 9000 11.000 7000 3000 and 12,000 T8000 [12,000] i 72000 BOOT and 13.000 | 8000 12,000] 060 600 anc 14.000 | 6600 [14.000] 4.7.2 (a) The signal at point bis fat) = mt) 206? set m(t)[Reosuct+ $ coset] ‘The term 4m(t)coset is the desired modulated signal, whose spectrum is centered at wz. The remaining term }in(t} coset is the unwanted term, which represents the modulated signal with carrier frequency Suse ‘with spectrum centered at 43ue as shown in Fig, $4.7-2. The bandpass filter centered at ue allows to pass the desired term 4yn(t) coswt, but suppresses the unwanted term tm)cos wet. Hence, this system works as desired with the output 2m(t) coswet (b) Figure $4.7-2 shows the spectra at points b ard c. () The minimum usable value of wis 2x in orcer to avoid spectral folding at de @ mit) PO + cos zat] 1 1 Jm(e) + L(t) cost s(t) cos? wet ‘The signal at point b consists of the baseband sigral m(t) and a modulated signal $m(t) cos 2vct, which has a carrier frequency 2uet, not the desired value wet. Both the components will be suppressed by the filter, whose center center frequency is we. Hence, this system will not do the desired job. (c) The reader may verify that the ide ty for cos nusct contains a term coswt when n is odd. This is not true when n is even. Hence, the system works for a cartier cos" wet only when n is odd. Fig. 84.7.2 [f\ a&@ “— x Dm By a@ 4.7-3. This signal is identical to that in Fig. 3.88 with period Tp (instead of 2). We find the Fourier series for this signal as 1,2 focust Hew sts eo si) = 442 fonet- Love ants beset] Hence, y(t), the output of the multiplier is vi) m(t)a(e) 1 mio [S42 (covenants deena +--)] ‘The bandpass filter suppresses the signals m(¢) and m(t)cos nuct for all n # 1. Hence, the bandpass filter output is honest = Bron et 564.724 fult) = (A+ m(t)Jcosuet. Hence, {s(0) = [A + {eh} cos? act 1 1 = HA + min] + Hat m(e]eos duet ‘The first term is @ lowpass signal because its spectrum is centered at w= 0. The lowpass iter allows this term to pass, but suppresses the second term, whose spectrum is centered at +2u,. Hence the output of the lowpass filter is y(t) = A+ m(0) When this signal is passed through a de block, the de term A is suppressed yielding the output m(t). This shows that the system can demodulate AM signal regardless of the value of A. This is a synchronous or coherent demodulation, ars @ =2 asm O) 2 sa-0 © 2 «) z ‘This means that 4 = oo represents the DSB-SC case. Figure S4.7-5 shows various waveforms, Fig. $47.5, 4.7-6 To gonerate a DSB-SC signal from m(t), we multiply m(t) with cosust. However, to generate the SSB signals of the same relative magnitude, itis convenient to multiply m(¢) with 2coswet. This also avoids the nuisance of the fractions 1/2, and yields the DSB-SC spectrum M(w—w) + Af(w + we). We suppress the USB spectrum (above we and below we) to obtain the LSB spectrum. Similarly, to obtain the USB spectrum, we suppress the LSB spectrum (between —1e and «-) from the DSB-SC spectrum. Figures S4.7-6 a, b and c show the three (0)From Fig. a, we can express yjs9(t) = €08900t and yga(t) = cos 11008. (b)From Fig. b, we can express 29 (t) = 2c0s700t + cos 900¢ and spt) (c)From Fig. c, we can express ,<9(t) = $[c0s 400¢ + cos 600¢] and 59 (0) 0s 1100¢ + 2.08 1300 4Hleos 1400¢ + 2¢0s 16004 8?a tht tol ff ast ine tt * +t af st gtet, ee Fig. 4.7.6 med ; yooo -— nt? il t= ad ~ 800? en 10 age 10 Io } bes t= i] t= pence) Fig. S481 4.841 In this case F = 10 MHz, mp = 1 and ny For FM AF = kymp/2n = 2n x 108/2n = 10° Ha, Also Fe = 107. Hence, (Fi)max = 107 + 10 (Fimin = 10" — 10° = 9.9 Mitz. The carrer frequency increases linearly from 9.9 MHz to 10.1 MHz over a quarter (rising) cycle of duration a seconds, For the next a seconds, when m(t) = 1, the carrier frequency remains at 10.1 MHz. Over the next quarter (the falling) cycle of duration a, the carrer frequency decreases linearly from 10.1 MH to 9.9 MH2., and over the last quarter eyele, ‘when m(t) = ~1, the carrer frequency remains at 9.9 MHz, This cycles repeats periodically with the period 4a seconds as shown in Fig. S48-1a. For PM AF = kpm)/2n = 50 x 8000/27 = 2.x 10° Hz. Also F_ = 107. Hence, (F;)max = 107 +2 x 10° = 10.2 MHz. and (FiJmin = 107 ~2x 10° = 9.8 MHz, Figure S48-1b shows rn(t). We conclude that the frequency remains at 10.2 MHz aver the (rising) quarter cycle, where shit) = 8000. For the next a seconds, a(t) = 0, and the carrier frequency remains at 10 MHz, Over the next « seconds, where rn(t) = —8000, the carrier frequency remains at 9.8 MHz. Over the last quarter cycle sn(t) = 0 again, and the carrier frequency remains at 10 MHz. This cycles repeats periodically with the period 4a seconds as shown in Fig, S4.8-1 3000. 10.1 MHz. and 4.8-2 In this case Fe = 1 MHz, my = 1 and mn = 2000. For FM ar 20,000n/2x = 104 Ha. Also Fe = 1 MHz. Hence, (Fi)max = 10° + 104 = 1.01 Mitz, and 0.99 Miz, The carrer frequency rises linearly from 0.99 MHz to 1.01 MHz over the cycle (over the interval ~19=* < ¢ < 1952), ‘Then instartaneously, the carrer frequency falls to 0.99 MHz and starts rising linearly to 10.01 MH over the next cycle. The cycle repeats periodically with period 10~* as shown in Fig. $4.8-2a For PM Here, because m(t) has jump discontinuities, we shall use a direct approach. For convenience, we select the origin for m(t) as shown in Fig. S4.8-2. Over the interval 19* to 10%, we can express the message signal as ‘m(t) = 20008. Hence,1 mit) Wri Meat ; ; . i 40 dP ; i Fig. $4.8-2 Youll) cos [2x(10)%e + Fm(e)] on [2n(10)%t + 5.20004] 05 [25(10)% + 1000z¢] = cos [2x (108 + 500) ¢) A the amount of jump is mg = 2. Hence, the phase discontinuity is kymg = x. Therefore, the carrier frequency is constant throughout at 10° + 500 H2. But at the points of discontinuities, there is a phase discontinuity of x radians as shown in Fig. 34.8-2b. In this case, we must maintain kp < x because there is a discontinuity of the amount 2. For kp > , the phase discontinuity will be higher than 2r giving rise to ambiguity in demodulation, 4.8: In thie case ky = 10008 and ky = 1. For 200 sin 100¢ ~ 36,000 sin 2000xt 08 1004 +18 cos 200% and rat) 36,000% +200. Also the baseband signal bandwidth B = 2000x/2r = 1 kHz, 0,000, and Bene = (AF + B) = 2(20,000 + 1000) = 42 kHz 13806366 kHz 8,000 + 12 He, and Bras = 2(AF + B) = 2(18,031.88 + 1000) 1000%/2x = 1000 Hz. Also, 484 10 cos(wet + 0.1 sin 2000xt). Here, the baseband signal bandwidth B w(t) = we + 200% c08 2000 (AF + B) = 2(100-+ 1000) = 2.2 KH. ‘Therefore, Aw = 200r and AF = 100 He and Bw 48-5 veu(t) = 5 cos(uct +20 sin 1000xt + 10 sin 200022), Hece, the baseband signal bandwidth B = 2000r/2x = 1000 Hz. Also, Je + 20,0007 cos 1000x¢ + 20,000 cos 2000z¢ 20 kHz and Bey = 214F +B) walt) ‘Therefore, Aw = 20,000 +20, 000% 42 Kells, 2(20, 000 +1000) 40, 000r ard. 59Chapter 5 Sal-L_ ‘The bandwidths of f(t) sud fa(t) are 100 ktle and 150 kia, respectively. Therefore the Nyquist sampling rates for s(t) in 200 kite and for f(t) ix 300 ke Also f:2(0) => Fu) « Fy(w). and from the width property of convolution the bandwidth of f(t) is twice the bandwidth of f(t) and that of f(t) i three times the bandwidth of fat) (se also Prob, 43-10). Similarly the handwidth of f:(t)J(¢) isthe sum of the bandwidth of fx() and fa(t). Therefore the Nyquist rate for f(t) is 400 kia. for f24(0 i 900 KH, for fa(0)falt) is 500 Kia 52 (a) sinc*(100rt) 3 0.010 a3) ‘The bandwidth of this signal is 200 + rad/s or 100 Hz. The Nyquist rate is 200 Hz (samples/sec) (b) The Nyquist rate is 200 He, the same as in (a), because multiplieation of a signal by a constant does not change its bandwidth, © sine(1O0st) + 3sine*(6Oxt) <=> 0.01 ret si) + ACh ‘The bandwidth of rct( xz) is 50 Ha and that of A(z) is 60 Ha. The bandwidth of the sum is the higher of the two, that i, 60 Ha. ‘The Nyquist sampling rate i 130 Ha @ sinc(60xt) <=> 0.02eet(iie) sinc(100xt) <=> 0.01 ret( fe) ‘The two signals have bandwidths 25 Hz and 50 Hz respectively. The spectrum of the product of two signals is 1/2r times the convolution oftheir spectra. From width property of the convolution, the width of the convoluted signal is the sum of the widths of the signals convolved. Therefore, the bandwidth of sine(SOnt)sine(120r¢) is 25 +50= 75 Hz. The Nyquist rate is 150 Ha. 5.18 ‘The spectrum of f(t) = sine(200xt) is F(w) = 0.005 rect( zz). The bandwidth of this signal is 100 He (2007 rai/s). Consequently, the Nyquist rate is 200 He, that is, we must sample the signal at a rate no less than 200 samples/second Recall that the sampled signal spectrum consists of (1/T)F(w) = 2688 rect(aiiz) repeating periodicaly with period equal to the sampling frequency F, Ha. We present this information in the folowing Table for three sampling rates: F, = 150 He (undersampling), 200 He (Nyquist rate), and 300 Hz (oversampling). sampling fequency Fs [ sampling interval? | PFCs) [comments] 150 Hz 0.006667 0.75 reet (aif) | Undersampling | 200 He 0.005 rect (aiiz) | Nyquist Rave | “300 He 0.008334 TS rect (ai) | Oversampling | ‘The spectra of Fw) for the three cases are shown in Fig. $5.1-3. In the first case, we cannot recover f(t) from the sampled signal because of overlapping cycles, which makes it impossible to identify F(w) fom t corresponding Fw). Inthe second and the third case, the repeating spectra do not overlap, and its posible to recover F(4) from Fs) ating lowpass ltr of bandwith 10 Hef the last cae the spectrum F(a) = 0 ver the band between 100 and 200 Ha. Hence to recog Fe) we may se practical lowpva ter with gralcutoff between 100 and 200 Hz. The output in the second case is f(t), and in the third case ix 1.5 f(t). The ‘output spectra in the three cases are shown in Fig. $5.13. i 5.1-4 (a) When the input to this filter is (¢), the output of the summer is 6(t) — 6(¢ — T). This acts 8 the input to ‘the integrator. And, h(t), the output of the integrator is: et y a ne) = f(r) = 87 = PI] dr = ult) = alt= 7) =e ( slse response h(t) is shown in Fig. $5.14, Gy A tak? fini of this Crent = reine“) e771 H(s) = ( a ) and or wwtan=7 ine(F)| ‘The amplitude response ofthe flter is shown in Fig. $5.L-4. Observe that the fer is a lovpase fier of ‘Fit Napulse laponng of the crcl lar pulse. Wh ed signals applied a ‘4 ute rapes of tn et i a rectangular pute. When a sampled signals applied at te inp enc sande genrals rectanpilar pub ot tne oufput Beopontonal tp the eoresbonding sample tae. Hence the Sutpit a saireane approximation of the imput as'shown in Fig. S9.14e. St 17 we ar Tt Z z Figure $5.1-4 51-8 (a) Figure $5.1-50 shows the signal reconstruction from its samples using the first-order bol circuit, Each erate ike rte dt Ten cnet fete empl aa.‘ hig ofthe tase seal to the sample value. The resulting signal consists of straight line segments joining the sample to (b) The transfer fimetion of ths ere is eit ine ee n ns pie nors. Ha) = FUm0) = F {a (sh) = Paine (2 Becaase H(u) is positive for all wit aso represets the driplitude response Fig. $5.1-Sb shows the impulse response h(t) = A( ip). The corresponding amplitude response H(w) and the ideal amplitude response (lowpass) ged ft vigoal FLennsuston hows n Big 98.86 Peer Tea alee fs Mead tof 0) causal (realizable). Such delay would cme the reconstructed signal in Fig, $5.1-5a to be delayed by T' secs. (Gy Wie the eat ta the tea ie te SU then ab Shown in Prob, 5-4, its output i a rectangular pulse Plt) = u(t) ~ u(t 7) shown in Fig. S5.1-4a, ‘This pulse p(¢) is applied to the input of the second sential filter 61‘The output ofthe summer ofthe second filter spt) pt 7) = u(t)—2u(¢—7) + u(t 27), which isa pled to's imtegrtor” The output hy of se negra the aa unde (Ot) hic " em fle) Aube 1) + ale = 20) de = te) — 20 T)t =) #0 BCE 27) = shown in Fig. $5.1-5b. ney 10 FF Figure $5.1.5, Fle) ous7 os 1 | l + eo HES BO Sig Figure 5.1.6 5.1-6 Thesignal f(t) = sine(200xt) is sampled by a rectangular pulse sequence pr(¢) whose period is 4 ms so that the fundamental frequency (which is also the sampling frequency) is 250 Hz. Hence, w, = 500x. The Fourier series for pr(t) is given by Prlt) = Co+ Sn cosnunt Use of Eqs. (8.68) yielts Co= , Cu = 2 sin (ME), that is, Co=02, Cr= 0.374, Cz= 0808, Cy=0.202, Cy=0.003, Cs Consequently Fie) = sOpr(e) = 02 F(t) + 0.374f(e) cos 500Kt + 0.303 f(t) cos 1000%t + 0.202 f(t) cos 1800Rt + and F(w) =0.2F(u) + 0.187 (Fw ~ 500r) + F(u + 500%)] + 0.151 [Fs — 1000n) + F(w + 1000%)] + 0.101 [F Ge ~ 1500n) + Fw + 1800%)] + In the present case F(u) = 0.05 ect(zifz). ‘The spectrum F(x) is shown in Fig. 851-6. Observe that the spectrum consists of Fw) repeating periodically at the interval of 500r rad/s (250 Ha). Hence, there is no ‘overlap between cycles, and F(c) can be recovered by using an ideal lowpass Alter of bandwidth 100 Hz. An idea lowpass filter of unit gain (and bandwidth 100 Ha) will allow the fst term ou the he right-side of the above equation to pass fully and suppres all the other terms. Hence the output y(t) is u(t) = 0.240) 625.1-10 hit =pee) Ts 0.025 (a ~eT 4oT Om Figure 85.1-7 Because the spectrum Fl) has a 2ero valve in the band from 100 to 150 Hz, we cau use an ideal lowpass fiter of bandwidth B Hz where 100 < B < 150. But if B > 150 Hz, the filter will pick up the unwanted spectral components from the next eycle, and the ontput will be distorted, ‘The signal /(), when sampled by an impulse train, results in the sampled signal f(t)ér(¢) (as shows in ig. 5.1). If this rigual i transmitted through a filter (Fig. 8.1-7a) whose impute response is M(t) = p(t) = tect (pe) then each impulse in the input will generate a pulse p(t) resulting in the desiced sampled signal shown in Fig 51-7. Moreover, the spectrum of the impulse train /()6r(t) is $3.79, Fl ~ nas). Hence, the output of the fiter in Fig, $5.1-Ta is Fw) Hw) [: ¥ Fe ~] whers H(w) = P(w) = 0.025 sine), the Fourier transform of rect (gop). Figure S6.1-7b shows this spectrum consisting of the repeating spectrum F(w) multiplied by H(.) = 0.083sine (35). Thus, each cycle is somewhat distorted ‘To recover the signal (0) from the flat top samples, we reverse the process in Fig. $5.1, Fltst, we pase the sampled signal through a filter with transfer function 1/H()- Thia will yield the signal samped by impulse train. Now we pass this signal through an ideal lowpass filter of bandwidth B Hz to obtain f(t) (a) The bandwidth is 15 kHz, The Nyquist rate is $0 kia (b) 65536 = 2°, so that 16 binary digits are needed to encode each san (e) 20000 x 16 = 480000 bits/s. (A) 44100 x 16 = 705600 bits/s. (a) The Nyquist rate ie 2x 4.5 x 10° =9 MHz. The actual sampling rate = 1.2 x 9 = 10.8 MHz, (b) 1024 = 2", so that 10 bits or binary pulses are eeded to encode each sample. (@) 10.8 x 10° x 10 = 108 x 10° or 108 Mbits/s. ‘Assume a signal f(t) that is simultaneously timelimited and bandlimited. Let F(s) = Ofor |u| > 2B. Therefore F (o)ect( zl) = Fw) for BY > B. Therefore from the time-convolution property (4.42) Jd) = Je) + 2B'sine(2nB"0)| = 2B" f(t) sine(2xB't) Because f(t) is timelimited. f(t) = 0 for |e] > T. But f(t) is equal to convolution of f(t) with sino(2rB't) which is nor timelimited. It is impossible to obtain a time-limited signal from the convolution of a time-limited signal with non-timelimited signal. To= 20ms ef x B= 10000 Hence F, > 21Since No must be a power of 2, we choose Na = 512. Also T = S0ys, and Ty = NoT = 512 x Sts = 25.6ms, Fo =1/To = 39.0625 Hz. Since f(t) is of 10 ms duration, we need zero padding over 15.6 ins. Alternatively, we could also have used 5.2-2. For the signal f(¢), 6 Let us choose T = 1/8. Also Tp = 4. Therefore, No = To/T = 32. The signal f(t) repeats every 4 seconds with samples every 1/8 second. The samples are T (AT) = (1/8)f(k/8). Thus, the first sample is (at k = 0) 1x (1/8) = 1/8. The 32 samples are (starting at k = 0) Be Be Be a Be A 0 0.0, 0,0,0,0, 9,0,0,0,0,0,0,0,0, 25, 35. Boas Be ‘The samples of f(t) and 9(¢) are shown in Fig, $5.2-2. 5.23 1 jort (a) We take the folding frequency F, to be the frequency where |F()| is 1% of its peak value, which happens tobel (at w=0). Hence, £00) = e“*u(o) FG 01 su= 2B Fels? 00 ‘This yields B = 50/x, and T < 1/2B = x/100. Let us round T to 0.03125, resulting in 32 samples per second. ‘The time constant of e~' is 1. For To, a reasonable choice is 5 to 6 time constants or more. Value of To = 5 or 6 results in No = 160 or 192, neither of which is a power of 2. Hence, we choose Ty = & resulting in No =32 x 8 = 256, which is a power of 2 ©) WF) = We take the folding frequency F, to be the 99% energy frequency as explained in Bxample 4.16. From the results in Example 4.16, (with a = 1) we have SEF tant 5 W = 63.660 = 63.66 rad/s. ‘This yields 8 = HE = 10.13 Ha. Ato T < 1/28 = 0.04096. This reaults in the sampling rate } = 20.26 Hs ‘Also Tp = 8 as explained in part (a). This yields No = 20.26 x 8 = 162.08, which is not a power of 2. Hence, we ‘choose the next higher value, that is No = 256, which yields T = 0.03125 and To = 8, the same as in part (a). 5.24 a eT Application of duality property to pair 3 (Table 4.1) yields Wo 6Following the approsch of Prob. 5.2-2, we find that the peak value of |F(w)| = 2re~!+! is 2x (occurring at w = 0). Also, 2re~™! becomes 0.01 x 2 (1% of the peak value ) at w =n 100 = 4.605. Hence, B = 4605/2~ = 0.73% Ha, and T < 1/2B = 0.682, Also, £(0) tod 2 and sy = 2 Cheese To (the duration of f(t)) to be the instant where f(t) is 1% of £(0) 2 2 War P= ‘This results in No = To/T = 10/0.682 = 14.66. We choose No = 16, which is a power of 2. This sields T = 0.625 and To = 10. (b) The energy of this signal is S(T) = se [Perms ‘The energy within the band from w = 0 to W is given by ant ew= Ef ae = 2m But Bw = 0.996; = 0,99 « 2x. Hence, 0.99(2n) = 2x(1-e”) => W = Hence, B = W/2r = 0.366 Ha. Thus, T < 1/2B = 1.366. Also, Ty = 10 as found in part (a). Hence, No = To/T lect No = 8 (a power of 2), resulting in No = 8 and T= 1.25, Figure $5.2-5 5.2-5 The widths of f(t) and o(t) are 1 and 2 respectively. Hence the width of the convolved signal is L+2 = 3. This ‘means we need to 2ero-pad f(t) for 2 secs. and 9(¢) for 1 sec., making Tp = 3 for both signals. Since T = 0.125 _3 mr [No must be a power of 2. Choose No = 32. This permits us to adjust Ta to 4. Hence the final values are T=0.125 and Tp = 4, The samples of f(t) and 9(¢) are shown in Fig. $5.2-5. No 4