0% found this document useful (0 votes)
78 views50 pages

Solution Manual For Communication Systems Haykin 6

Uploaded by

Nam Lê
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
78 views50 pages

Solution Manual For Communication Systems Haykin 6

Uploaded by

Nam Lê
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 50
_ Kofay) expl-s2"f(ty=ty,)] 01 a, 2 1 a expl-Janf(ty, ~ ty,)I Since a << a4, we may approximate H,(f) as follows % ae BoA) = Zo expl-s2ntty — tyy)] (1 ~ Zo expl-s2at(tyg ~ ty] a, 2 2. 5 + ra expl=Jtaf(ty - ty,)]} Comparing Eqs. (1) and (2), we deduce that » We find that the tap weights of the equalizer are as follows 248 (2) Problem 4.35 The Fourier transform of the tapped-delay-line equalizer output is defined by oul = HO Xjq(®) @ where H(f) is the equalizer’s transfer function and X,,(f) is the Fourier transform of the input signal. The input signal consists of a uniform sequence of samples, denoted by {x(nT)). We may therefore write (see Eq. 6.2): x= 25 xe @ r where T is the sampling period and s(t) is the signal from which the sequence of samples is derived. For perfect equalization, we require that You) = 1 for all f From Eqs. (1) and (2) we therefore find that T H) = — 7 _ y XW) ® k (sequenced Let the impulse responsefof the equalizer be denoted by (w,). Assuming an infinite number of taps, we have Hf) = ry Wy exp(2nfT) nsv0 We now immediately see that H(f) is in the form of a complex Fourier series with real coefficients defined by the tap weights of the equalizer. The tap-weights are themselves defined by Wat $ [2h HOexp(-j2nt/1), n=6, +1, +2... 249 The transfer function H(f) is iteelf defined in terms of the input signal by Eq, (3). Accordingly, a. tapped-delay-line equalizer of infinite length can approximate any function in the frequency interval (-/2T, V/2T). Problem 4.36 (a) Asan example, consider the following single-parameter model of a noisy system: d{n] = wolnlxin] + ln) where x[v] is the input signal and v{n] is additive noise. To track variations in the parameter woln], we may use the LMS algorithm, which is described by Error signal n) stn (d| w[n]x[n]) = (1- px" [nwfn] + wxlnld{n] a) To simplify matters, we assume that W{n] is independent of x(n]. Hence, taking the expectation of both sides of Eq. (1): El[n+ 1] = (1 poeta] + wry. 2) where E is the statistical expectation operator, and = El ln]) rae = Eldtnlxtn}] Equation (2) represents a first-order difference equation in the mean value E[¥[n]]. For this difference equation to be convergent (i.e., for the system to be stable), we require that <1 |1 -n03| or equivalently @ — 1-por<1, ie, p>o (i) -1+por= i/(at1)); out (index) = (r(1) + i) * index, /index; Answer to Problem 4.38 Figure 4: Eye pattern for a=0 257 wie) Figure 2: Bye pattern for a=0.05 Figure 3: Eye pattern for a=0.1 Figure g : Bye pattern for a=0.2 Problem 4.39 Matlab codes % problen 4.39, cS: Haykin % oot raised-cosine and raised cosine sequences %M. Seliathurad, Datas[1 0 1100)"; 4% sample frequency 20 sample_freq=20; Ygenerate antipodal signal syms=PAM_nod(Data, sample_treq, 2); i, root raised cosine pulse rer = raisecos_sqrt(syns, sample freq ); 4% normal raised cosine pulse Fem raisecos_n(syms, sample_treq ); % plots telength(r_e_r)-1; figure; hold on 259 plot (0:4/20:+/20, x Plot (0:1/20:+/20, x xlabel (’time’) Legend('root raised-cosine’ , raised-cosine’) hold off 260 function osyns = raisecos_n(syms, sample. % function to generate raised-cosine si % used in Problem 4.39, CS: Haykin WM. SeLlathurai req ) 4% size of data Lisyms, w_syms] = size(syms); 4% data Reo We =[3, 343]; % Calculation of Raised cosine pulse W_T(1) = -abe(W_T(1)); time_T = [0 : 1/sanple_treq : max(W_T(2), abs(W_T(1)))]; time_TLR = R + tine_T; den = 1 - (2 * time indext = find(den’ index2 = find(den when denominator not equal to zero b(indext) = sinc(time T(indext)) .* cos(pi * time_T_R(indexi)) ./ den(index1); % when denominator equal to zero, (using L’Hopital rule) if “isempty(index2) b(index2) = 0; end; (o(sample_treq * abs(W_T(1))+1 : -1: 1), b(2 : sample.treq + ¥_T(2)+1)]; beb(:)"s % filter parameters order= floor(Length(b)/2); bbe Ol; for i= 1: order bb = [bb; b(1+i:order+i)]; end; fu, 4, v] = eva(op); d= diag(a); index = tind(d/a(1) < 0.01); if isempty(index) © = Length(ob) ; else 261 ona; aa ut Length(bb)-1, 1 : 0); vi Length(bb)-1, 1: 0); u2 Jength(bb), 1: 0); dd = sqrt (4(1:0)); vdd = 1 ./ aa; uu = ut? * uz; at = uu .* (vad * dd’); a2 = dd .# via, 3)": a3 = ui(i, :) .* dd’; (num, den] = ss2tt(at, a2, a3, a4, 1); fsyms = zeros(1_syns+S+sample_freq, ¥syns) for i= 1: sampletreq : l_syns toyns(i, syme(i, end; % tittering for i = t:vsyas fsyns(:, i) = filter(num, den, feyma(:, i)); end; osyns = fsyns(( (3 - 1) © sample_treq + 2):(size(feyme, 1) - (sample_treq - 1)), :); 262 function osyms = raisecos_sqrt(syns, sample_treq ) 4% function to generate root raised-cosine sequence % used in Problem 4.39, CS: Haykin 4. Sellathurai size of data Cisyms, wsyms] = size(syms); % rollott tactor R=0.3; % window WT=[3, 383]; 4% Calculation of Raised cosine pulse WT(1) = -abs(W_T(1)); timeT = [0 : 1/sample_treq : max(W_T(2), abs(W_T(4)))]; indext = tind(den den = 1 ~ (4 + tine TeR). index2 = find(den % when denominator not equal to zero b(indext)=( cos((1 + R) * pi * time_T(indext))+ (sinc((1-R)stime_T(indexi))*(1-R)*pi/4/R))./den(indert)*4#R/ pi; % when denominator equal to zero t=\pm T/4/alpha if “isempty(index2) (1+2/pi)*sin(pi/4/R)+(1-2/pi) #cos (pi/4/R))*R/eqrt(2) “RHMR/pi; Ye=0; b = (b(sample_treq * abs(W_T(1))+1 4), b(2 : sample_treq * ¥_T(2)+1)]; order= floor(1ength(b)/2); bb= tl; for i order bb = [bb; b(tti:order+i)]; end; tu, 4, v] = sva(op); a= diag(a); 263 index = find(d/a(1) < 0.01); if isempty (index) © = Length(bb) ; elee 0 = index(1)-1; end; ad = bb(1); ui vi 42 F Length(bb)-1, 1 : 0); Length(bb)-1, 1: 0); lengtn(bb), 4 : 0); dd = sqrt(4(1:0)); vad 7 aa; uu = ul?» 02; at = uu .* (vdd * dd’); a2 = dd .* vi(t, :)?; a3 = ul(1, :) . dd’; (num, den] = ss2tt(ai, a2, 03, a4, 1); ‘fsyns = zoros(1_syne+3+sample_freq, w_syns); for i= 41: sample_treq : 1syms feyms(i, :) = syms(i, end; % filtering for i= t:¥syms feyms(:, i) = filter(num, den, fayme(:, i); end; osyms = feyns(( (3 - 1) + sample troq + 2):(size(tsyns, 1) - (sanple_treq - 1)), 264 Answer to Problem 4.39 ro0t raised-cosine faiged-cosine Figure 1: Raised-cosine and root raised-cosine pulse for sequence {101100} 265 Problem 5.1 (a) Unipolar NRZ code. ‘The pair of signals s)(2) and so(f) used to represent binary symbols 1 and 0, respectively are defined by y(t) where E}, is the transmitted signal energy per bit and 7; is the bit duration, From the definitions of s1(f) and s)(0, itis clear that, in the case of unipolar NRZ signals, there is only one basis function of unit energy. The basis function is given by OsrsT, ‘Then, we may expand the transmitted signals s;(¢) and s9(¢) in terms of 6,(t) as follows: 5() = JE,O,(9, O N. The inner product of the pair of signal s,(0) and s,(2) is given by 1 Jsosnar 6) 0 By substituting (1) in (5), we get the following result for the inner product: Tew N [Ss9,0] [Zamofe i=l oljat non of =D Ysjsufolnotnae © 0 1 ie Since the 9,(1) form an orthonormal set, then, in accordance with the two conditions of Eq. (3) and (4), the inner product of s(t) and s4(t) reduces to r Ny Jsise@ar = Ysys4j 0 jst 280 T. = 8) (b) Consider next the squared Euclidean distance between s; and s;, which can be expressed as follows: IF = 6:50" (s.-s) Is- = 8/5, + 5,8, — 287 5, 7 T r Jstlanae + fsp(ende -2[5,(2)sae ° ° ° " r = f-5()"at ° Problem 5.9 Consider the pair of complex-valued signals s,() and s2(1), which are defined by S(t) = 44104(2) + ay202(1) ay S21) = a310)(1) + ay909(t) 2) The basis functions 6,(¢) and 03(¢) are real-valued and the coefficients a1, a1, a1 and ayy are complex-valued, We may denote the complex-valued coefficients as follows: x2 = O92 + B22 On this basis, we may represent the signals s,(1) and s3(t) by the following respective pair of vectors: | Oy ay = [Bul gy = {Be 2) O29] Bus Bn ‘The angle subtended between the vectors gy and gp is r 2182 cos = led Teo) To gat ap ACB + O12 + Biz + 4021 B2i + Op + Boy H, si'sy * Bl Ts a 2) where 8) = [ and sy = are complex vectors. Recognizing that 12 2. cos = I we may go on to write Jsi@synae < = 7S! re (aye | J ora iu ps00Fa) < = The complex form of the Schwarz unequality becomes J isynaa 2 = - £ J fsioPae f jsp(o/Par The equality holds when s;(t) and s2(1) are co-linear, that is, s(t) = ksa(t) where k is any real- valued constant. Problem 5.10 EQ WCE) = ELC, + WW CELDT Els, W(t) = 9,5 EWE] = We also note that N Wt) = MCL = EW oy (ty) ist We therefore have BOC W(t, )7 = ECW, WCET N 7 - 26 ECW W(t) me apt ELW,W,1 1 T B W. Wt, = t, Ww jdt) = ELW(t, )W(t)]dt ut EL 5 IC mE ELW( 1 - Ct) ot) J 45 oo) CWC m 1 T Ny Ny 2) + 2 acetate = 38 a $36 ) (tt ” it 5st Mo, tes 2 ELW,W,1 = oO i¥j Hence, we get the final result Ny x EX 4g WCE] = r P ayty 3 20. Problem 5.11 For the noiseless case, the received signal r(1) = s(),0<1. x2. On the other hand, if x > x), it decides in favor of s2(0). If x decision is made by tossing a fair coin = xp the (b) Energy of signal s,(t) is given by r or =Jo Pare f (- 1dr + far ‘di ° r ar Energy of signal s3(0) is 1” ara sr ar = Ie irae f (uyrare f Curae f arar 12 ar st =30=E The orthornomal basis functions for the signal-space diagram of these two orthogonal signals are given by a = 22 and st) 0 = BE The signal-space diagram of signals s1 and sy is as follows: 286 = 4 Br ‘The distance between the two signal points s(t) and s(t) is d= .2E = J6T The average probability of error is therefore G ae) a x a 1" oe s 4x 107 " Problem 5.13 Energy of binary symbol 1 represented by signal (0) is, 12 r = fGias f Caw = 7 ° 12 Energy of binary symbol 0 represented by signal s3(¢) is the same as shown by 287 The only basis function of the signal-space diagram is s(t) _ sy) oD = TE, WT ‘The signal-space diagram of the Manchester code using the doublet pulse is as follows: Hence, the distance between the two signal points is d = 2./T’. The average probability of error over an AWGN channel is given by a ent) sere [z-) HT) Problem 5.14 (a) Let Z denote the total observation space, which is divided into two parts % and z, Whenever an observation falls in zy we say Hos and whenever an observation falls in 2, we say He Thus, expressing the risk R in terms of the conditional Probability ante functions and the decision regions, we may write Ha (ait ox Cog Po Az f, 00 Po 4z.x1H, #49 Po Sy fyyy CEH Ax 1='o + Cy, By Fo Fa, tHe + oy Py So fa, EI a For an N-dimensional observation space, the integrals in Eq. (1) are N-fold integrals. To find the Bayes test, we must choose the decision regions Z, and Z, in such a manner that the risk R will be minimized. Because we require that a decision be made, this means that we must assign each point x in the observation space Z to Z, or Z,; thus Zety+2, Hence, we may rewrite Eq. (1) as R = PgCgg Fafayy MMorAx + Poli 9 Fan fxjn, Hoe + PyC,, SF, a2g ty, 2M PE + Piloy Foye, EM ae (2) We observe that FL xyy EModee Fagin, ela = 1 Hence, Eq. (2) reduces to {P9699 Fy SIO? + TPC y gy AMD, @) The first two terms in Eq. (3) represent the fixed cost. The integral represents the cost controlled by those points x that we assign to Z). Since Cy > Coy and Co, > C,,, we find that the two terms inside the square brackets are positive. Therefore, all values of x where the first term is larger than the second should be included in Z, because the contaibute a negative amount to the integral. Similarly, all values of x where the second term is larger than the first should be excluted from Z, (i.e., assigned to 2,) because they would contribute a positive amount to the integral. Values of x where the two terms are equal have no effect on the cost and may be assigned arbitrarily. Thus the decision regions are defined by the following statement: If 289 PCC, (XH) > P96, ot Fu, 10°00) fx 4(EIo) + assign x to Z, and consequently say that H, is true. If the reverse is true, assign x to 2, and say Hy is true. Alternatively, we may write fyyq CMH) x1, SMY Bt pec aby) PyCCy4-C 4) Fy lA) # > < 21Hy i 0 The quantity on the left is the likelihood ratio: Fan et? Ap Tq Zp) tH 2MHo Let accu Cre) Pip ,C 44) Thus, Bayes criterion yields a likelihood ratio test described by AGO (b) For the minimun probability of error criterion, the likelihood ratio test is described by #. inca) ea eee . H -Thus, we may view the minimum probability of error criterion as a special case of the Bayes criterion with the cost values defined as S00 = 811 =? “10 = Sor That is, the cost of a correct decision is zero, and the cost of an error of one kind is the same as the cost of an error of the other kind. 200 Problem 5.15 From the signal-space diagrams derived in the solution to Problem 5.1, we immediately observe the following 1. Unipolar NRZ and unipolar RZ codes are non-minimum energy signals. 2. Polar NRZ and Manchester codes are minimum energy signals. Problem 5.16 ‘The orthonormal matrix that transforms the signal constellation shown in Fig. 5.11(a) of the textbook into the one shown in Fig. 5.11(b) is 2 1 A ° fl Sl Ble To prove this statement, we note that the constellations of Fig. 5.11(a) is defined by the four points {(ct, &), (0, ), (ot, ~01), (ct, —ct) }. The new constellation is defined by S;, rotate = QS,, which for i= 1 yields ee nw[# BIg (a (0 Similarly, 5» josie = | 5) Hence, the transformation from Fig. 5.11(a) to Fig. 5.11(b) is given by Q, except for a scaling factor. 291 (a) The minimum distance between any two adjacent signal points in the constellation of Fig. PS5.17a of the textbook is d® = 20 ‘The minimum distance between any two adjacent signal points in the constellation of Fig. P5.17b of the textbook is de = A(fBay’ + (Bay which is the same as d‘“). Hence, the average probability of symbol error using the constellation of Fig. P5.17a is the same as that of Fig. P5.17b. 2a (b) The constellation of Fig. P5.17a has minimum energy, whereas that of Fig. P5.17b is of non- minimum energy. Applying the minimum energy translate to the constellation of Fig, P9.17, which involves translating it bodily to the left along the 6)-axis by the amount V/2o., we get the corresponding minimum energy configuration: Problem 5.18 Consider a set of three orthogonal signals denoted by{s;(1)}7_g, each with energy E,. The average of these three signals is a(t) Ls Applying the minimum energy translate to the signal set {sj(1)}7_y, we get a new signal set defined by 8 s() = s()-a(t), 1 = 0,1,2 on) The signal energy of the new set is B= f(a =f side 2” sacar + f° ated Fn - a@ny(s()-a(n)yat 5 FELL seos nd [° ancy sfondi [a nat) Q) Since s,(¢) and s(t) are orthogonal by choice, Eq. (2) reduces to ieee eh) 1738s * 5GED) which is the maximum negative correlation that characterizes a simplex signal with M = 3. thus, 2 the signal set {sj(t)};.9 defined in Eq. (1) is indeed a simplex signal. 293 2 To represent the signal set {sj(t)}ju9 in geometric terms, we use the Gram-Schmidt orthogonalization procedure. Specifically, we first set sol) G01) = a or equivalently so(t) = JEGo(0) The projection of s(t) unto go(t) is so =f si(oo(nar = Fpl ntna 5 AEE sicnsgtnar) “(4 ‘The second basis function is therefore 1) = Siobo() JE-Sio = 310) + CLE/2)(G0(0) JE = (E74) . FAloitn + Fon) Accordingly, we may express s}(t) in terms of the basis functions go(#) and 6(1) as 204 GB) s()) = ons FE on ) ‘The remaining signal s3(t) may be expressed in terms of @o(t) and (1) as 0 = Foon 6) Thus, using Eqs. (3) to (5), we may represent the simplex code by the following signal-space diagram: Problem 5.19 (a) An upper bound on the complementary error function is given by exp(-w’) anu Hence, we may bound the given P, as follows: erfe(u)< ox ( [E, Ny 1 E, N enfel [P| <= Sexo(—5t) x Je a (| 2fmE,Ny 2 PW) * ine, 7 For large positive w, we may further simplify the upper bound on the complementary error function as shown here: 295 2 erfe(u) < PCH) Correspondingly, we may bound P, as follows: exp(-E,/No) pele 2) re ae Q) (b) For By/No = 9, we get the following results: (i) The exact calculation of P, yields Jerfe(3) = 10x 105 (ii) Using the bound in (1), we have the approximate value: = 116 x10 Gi) Using the looser bound of (2), we have _ xp(-9) Wn = 3.48x 10° As expected, the first bound is more accurate than the second bound for calculating P,. Problem 5.20 According to Eq, (5.91) of the textbook, the probability of error is over-bounded as follows: P= 120,M w where dy is the distance between message points s; and sy. With the M transmitted messages assumed equally likely, the average probability of symbol error is overbounded as follows: Q) ‘The second line of Eq, (2) defines the union bound on the average probability of symbol error for any set of M equally likely signals in an AWGN channel. Equation (2) is particularly useful for the special case of a signal set that has a symmetric geometry, which is of common occurrence in practice. In such a case, the conditional error probability P,(m)) is the same for all i, and so we may simplify Eq. (2) as is G <4 5 o{ “| for all i (3) f P, = P(m)) ik 2JNo, ‘The complementary error function may be upper-bounded as follows: dy) 4 di) erfe < Lexy SH (5 Ny) Jn No} Hence, we may rewrite Eq. (3) as th ora P.<—= Y, exp)-—“] for all i @) aed oom) Kal 297

You might also like