Inthe preceding chapter, we considered digital communication over an AWGN chan- nel and evaluated the probability of error Performance of the optimum receiver for several different types of baseband digital modulation techniques. In this chapter, -we treat digital communication over a channel that is modeled as a linear filter with a bandwidth limitation. The bandlimited channels most frequently encountered in prac- tice are telephone channels, microwave line-of-sight (LOS) radio channels, satellite channels, and underwater acoustic channels. In general, a linear filter channel imposes more stringent requirements on the design of modulation signals. Specifically, the transmitted signals must be designed to satisfy the bandwidth constraint imposed by the channel. The bandwidth constraint generally precludes the use of rectangular pulses at the output of the modulator. Instead, the transmitted signals must be shaped to restrict their bandwidth fo that available on the channel. The design of bandlimited signals is one of the topics treated in this chapter. We will see that a linear filter channel distorts the transmitted signal. The channel distortion. results4n intersymbol interference at the output of the demodulator and leads to an increase in the probability of error at the detector. Devices or methods for correcting or undoing the channel distortion, called channel equalizers, are then described. DIGITAL TRANSMISSION THROUGH BANDLIMITED CHANNELS A bandlimited channel such as a telephone wireline is characterized as a linear filter With impulse response c(t) and frequency response C(f), where cy) = in e(tye PF" dt. 1.1 es 475476 Digital Transmission through Bandlimited AWGN Channels Chap, erg leh RaW 7 on Figure 9.1 Magnitude and phase responses of bandlimited channel. If the channel is a baseband channel that is bandlimited to B, Hz, then C(f) = 0 for If| = Be. Any frequency components at the input to the channel that are higher than B, Hz will not be passed by the channel. For this reason, we consider the design of signals for transmission through the channel that are bandlimited to W = Be Hz as shown in Figure 9.1. Henceforth, W will denote the bandwidth limitation of the signal and the channel. Now, suppose that the input to a bandlimited channel is a signal waveform @7r(t), where the subscript T denotes that the signal waveform, is the output of the transmitter. Then, the response of the channel is the convolution of gr(t) with c(t), ie. hw = f _ e(2)ar(e — r)dt = c(t) * ar). eid 4 or, when expressed in the frequency domain, we have HP) = CNGrf). G12) where Gr(f) is the spectrum (Fourier transform) of the signal g7(t) and a the spectrum of /1(¢). Thus, the channel alters or distorts the transmitted signal s7(° Let us assume that the signal at the output of the channel is corrupted PY AWGN. Then, the signal at the input to the demodulator is of the form ht n(@), where n(t) denotes the AWGN. The linear filter channel model is show? ure 9.2.section 9.1 Digital Transmission through Bandlimited Channels 477 —————_—__, ert) Linear filter i no Gi | eect) ri) = (+00 no) Figure 9.2 Linear filter model for a bandlimited channel, From the preceding chapter, we recall that in the presence of AWGN, a demod- ulator that employs a filter that is matched to the signal h(t) maximizes the SNR at jts output. Therefore, let us pass the received signal r(r) = h() + n(t) through a filter that has a frequency response Galf) = Ht fe, (9.14) where fo is some nominal time delay at which we sample the filter output. The subscript R denotes that the matched filter is at the receiver. The signal component at the output of the matched filter at the sampling instant t= Ip is ys(to) = [ lH(APaf =n. (9.1.5) which is the energy in the channel output waveform h(t). The noise component at the output of the matched filter has a zero mean and a power-spectral density ‘ N Sf) = SEP. (9.1.6) Hence, the noise power at the output of the matched filter has a variance 00 No [® NoB a= [sid = > [Lucra =" ean 00 00 Then the SNR at the output of the matched filter is Si Gi 2€n -\=ye77 9.1.8) (i), Noéx/2 No 0.18) This is the result for the SNR at the output of the matched filter that was obtained in Chapter 8, except the received signal energy %j, has replaced the transmitted signal energy €,. Compared to the previous result, the major difference in this development is that the filter impulse response is matched to the received signal h(t) instead of the transmitted signal. Note that the implementation of the matched filter at the receiver Tequires that h(t) or, equivalently, the channel impulse response c(t) must be known to the receiver.“1 Digital PAM Transmission through Bandlimited Baseband Channels Let us consider the baseband PAM communication system illustrated by the fun tional block diagram in Figure 9.4. The system consists of a transmitting filter havin an impulse response gr(t), the linear filter channel with AWGN, a receiving file, with an impulse response ga(t), a sampler that periodically samples the Output of the receiving filter, and a symbol detector. The sampler requires the extraction ofa timing signal from the received signal as described in Section 8.6. This timing signal serves as a clock that specifies the appropriate time instants for sampling the output of the receiving filter. Let us consider digital communication by means of M-ary PAM. Hence, the input binary data sequence is subdivided into k-bit symbols, and each symbol is mapped into a corresponding amplitude level that amplitude modulates the output of the transmitting filter. The baseband signal at the output of the transmitting filter (the input to the channel) may be expressed as a vt) = Yo angr(t—nT), G19) n==00 where T = k/Rp is the symbol interval (1/T = Rp/k is the symbol rate), Ry is the bit rate, and {a,} is a sequence of amplitude levels corresponding to the sequence of k-bit blocks of information bits. The channel output, which is the received signal at the demodulator, may be expressed as co r= YP agh(t nT) +-n(0), 9.1.10) n==00 Input Pr ansmitt = Ouipst data, ™B10(0) | Channel r(1) | Receiving : dat ata anne E 0) ly(KT) Gif) cy) Gai Sampler >> Detector} t Noise Symbol ao) timing estimator Figure 9.4 Block diagram of a digital PAM system.gection9.1 Digital Transmission through Bandlimited Channels aot wate Seen one of the cascade of the transmitting, filter and the channer. cents me gr(t), c(t) is the i Fe: . ee fd) eepresents the AWGN. he impulse response of the channel, and ig edit is passed through a fineur receiving filter with the impulse response 82 (0) tnaximum atthe proner Ga(f). If ga(t) is matched to (1), then its Guiput SNR is a maximum at the proper sampling instant. ‘The output of the receiving, filter may be expressed as 00. VO) = SP agar = nT) + wld, 1) n==00 where x(1) = h() * galt) = gr(t) # (0) * galt) and WD) = n(t) * ge(t) denotes the additive noise at the output of the receiving filter. To recover the information symbols {aq}, the output of the receiving filter is sampled periodically, every T seconds. Thus, the sampler produces eS y(mT) = Yo anx(mT — nT) + wm), (9.1.12) n=—00 or equivalently, 2 Ym = YS Ankman + Wm =o = Xodm + AnXman + Wms (9.1.13) ngm where xp = x(mT), Wm = wimT), and m = 0, £1, £2, 0200 The first term on the right-hand side (RHS) of Equation (9.1.13) is the desired symbol am, scaled by the gain parameter Xo. ‘When the receiving filter is matched to the received signal h(t), the scale factor is e e w= ff ends = [HP as w =| IGr( MPC? af = En, (9.1.14) -W as indicated by the development of Equations (9.1.4) and (9.1.5). The second term on the RHS of Equation (9.1.13) represents the effect of the other symbols at the sampling instant ¢ = mT, called the intersymbol interference (ISI). In general, IST causes a degradation in the performance of the digital communication system. Finally, the third term, wm, which represents the additive noise, is a zero-mean Gaussian random variable with variance o2, = Noth /2, previously given by Equation (9.1.7). By appropriately designing the transmitting and receiving filters, we can satisfy the condition x, = 0 for n # 0, 80 that the ISI term vanishes. In this case, the only\ 482 Digital Transmission through Bandlimited AWGN Channels ch ‘Ate, : 9 term that can cause errors in the received digital sequence is the additive noi design of transmitting and receiving filters is considered in the next Section,” Me ge ‘SIGNAL DESIGN FOR BANDLIMITED CHANNELS In this section, we consider the problem of designing a bandlimited ansminy filter. First, the design will be done under the condition that there is no ¢ ant distortion. Later, we consider the problem of filter design when the channe| dis, 4 the transmitted signal. Since H(f) = C(f)Gr(f), the condition for distortion fae transmission is that the frequency response characteristic cf) of the channe} mis, have a constant magnitude and a linear phase over the bandwidth of the ransmiteg signal, i.e, ~j2nfto, anx(mT — nT) + w(mT), (9.22) nim or more simply, Ym = 04m + YD dnXnn + Wn, (9.23) nem where x(t) = gr(t) * gp(t) and w(t) is the output Tesponse of the matched filter to the input AWGN process n(t). The middle term on the RHS of. Equation (9.2.3) represents the ISI. The amount of ISI and noise that is present in the Teceived signal can be viewed on an oscillo coe: Specifically, we may display the received signal on the vertical input wih ft horizontal Sweep rate set at 1/7. The resulting oscilloscope display is called an ¢y¢ pattern because of its resemblance to the human eye. Examples of two eye pattem one for binary PAM and the other for quaternary (M = 4) PAM, are illustrated in Figure 9.5(a). ae , itive pauses a reduction in the eye opening. As a consequence, the system is more sensit! to a synchronization error and exhibits a smaller margin against additive noise-section 9.2 Signal Design for Bandlimited Channels = BINARY, QUATERNARY @) Optimum sampling time Sensitivity Distortion of to timing zero crossings error ieee ‘Noise margin Peak distortion (b) Figure 9.5 Eye patterns. (a) Examples of eye pattems for binary and quaternary PAM and (b) the effect of ISI on eye opening. Example 9.2.1 ‘ Consider a binary PAM system that transmits data at a rate of 1/7 bits/sec through an ideal channel of bandwidth W. The sampled output from the matched filter at the receiver is Ym = Am + 0.2dm—1 = 0.3¢m—2 + Wan Where a,, = -:1, with equal probability. Determine the peak value of the ISI and the noise margin, as defined in Figure 9.5(b). ; ; Solution If we compare the matched filter output yin pa eats Sie by i it ii arent that xo = 1, x1 = 0.2, x2 = —0.3, and Xm = 0, Equation (02.3). Hi appa ISL oor WHEN n= = ayn 80 thatthe IL termed AWGN Channels Digital Transmission through Bandlimit Chapter 484 ' i = 1 and am = +1, the ISI ¢, ait take the peak value of +0.5. Since xo = 1 — aa ea in the eye opening at the sampling times ¢ = m7. m = 0, oan reset noise margin is reduced by 30% to a value of 0.5. Thus, compared Hence which there is no ISI, a noise component that is 50% smaller wit ca error at the detector. 3 the problem of signal design under two conditions, Name} Next, we conside! fe: and that a controlled amount of Is} i that there is no ISI at the sampling instants allowed. 921 Design of Bandlimited Signals for Zero ISI — The Nyquist Criterion ystem as previously described. It transmits throug he bandwidth of gr(t) is less than or equal to hich is the Fourier transform of the signa Consider a digital communication s; an ideal bandlimited channel, when W. Thus, the bandwidth of the channel, w! at the output of the receiving filter, is given as X(f) = Gr NCNGRA) Gr(f)GR(f)Coe PF” (9.24) = Gr(f)Gr(f), where Gr(f) and Gr(f) denote the frequency responses of the transmitter and receiver filters and C(f) = Co exp(—j27fto), If] < W denotes the frequency response of the channel. For convenience, we set Co = 1 and fo = 0. We have also seen that the output of the receiving filter, when sampled periodically at t = m7, m ,—-2-1,0, 1,2.... yields the expression given by Equation (9.2.3). In this equation, the first term on the right-hand side of the equation is the desired symbol, the second term constitutes the ISI, and the third term is the additive noise. To remove the effect of ISI, it is necessary and sufficient that x(m7 —nT) =0 for n # m and x(0) # 0, where without loss of generality, we can assume x(0) This means that the overall communication system has to be designed such that 1, n=0 =}. .2.5) x(nT) Hoes ioe (9.2.5) In this section, we derive the necessary and sufficient condition for X(f) so *(0) & satisfy the preceding relation. This condition is known as the Nyquist pulse-shaping criterion or Nyquist condition for zero ISI. ) can Nyquist Condition for Zero ISI. A necessary and sufficient condition fot x(t) to satisfy x(nT) = { i ao 9.28) is that its Fourier transform X(f) must satisfy D2 x(f4+2) =r. 2.) agion 92 Signal Design for Bandlimited Channels 485 roof. In general, Soe pr Beneral, x(t) is the inverse Fourier transform of X(f). Hence, xQ=[" - [ “Xie ap ‘aim ‘atthe sampling instants ¢ = n7, this relation becomes 00 X(nT) = E X (felt ap. 29) break the integral i wT. Thus ea in Equation (9.2.9) into integrals covering the finite range ~ (2m+1)/2T- x(nT) = x | X(pollT af m=—oo% Qm=1)/2T a = i M) stray mance! WAT -[" CJ ni a eile el af i = I Zell" at, (9.2.10) -12T where we have defined Z(f) by Lf= ¥ (149): (92.11) Obviously, Z(f) is @ periodic function with period 4; therefore, it can be expanded in terms of its Fourier series coefficients {zn} as nn= Ye (9.2.12) where. i eet | Fy per ian df. (2.13) tr Comparing Equations (9.213) and (92.10), we obiin tee TH(-nT). 214486 Digital Transmission through Bandlimited AWGN Channels Chapter g ‘Therefore, the necessary and sufficient conditions for Equation (9.2.6) to be Satisfieg is that _ _frn=0 sn = 10, n #0" (9.2.15) which, when substituted into Equation (9.2.12), yields Zf)=T, 0.2.16 or equivalently, yx (f+ 3) =7 (2 Tee 2.1) m==00 This concludes the proof for the condition that X ( ‘f) must satisfy to obtain zero ISI, Now, suppose that the channel has a bandwidth of W. Then, C(f) = 0 for \fl > W; consequently, X(f) = 0 for Lfl > W. We distinguish three cases: 1. In this case, T < zy, or equivalently, 1 > 2W. Since Z(f) = tee ai X (f +4) consists of nonoverlapping replicas of X(f), which are separated by + as shown in Figure 9.6, there is no choice for X(f) to ensure Z(f) = T in this case. Thus, there is no way that we can design a system with no ISI. 2. In this case, T = zp, or equivalently, 4 = 2W (the Nyquist rate). The repli- cations of X(f), separated by +, are about to overlap, as shown in Figure 9.7. Figure 9.6 Plot of Z(f) for the case T < 3 an-Sx(+9) Figure 9.7 Plot of Z(f) for the case T = sy.section 9.2 Signal Design for Bandlimited Channels 487 It is clear that there exists only one X(f) that results in Z(f) = T, namely, JT Ifl ly, Z(f) consists of overlapping replications of X(f) separated by +, as shown in Figure 9.8. In this case, there exists an infinite number of choices for X(f), such that Z(f) = T. y For the T > shy case, a particular pulse spectrum that has desirable spectral properties and has been widely-used in practice is the raised cosine spéctrum. The raised cosine frequency characteristic (see Problem 9.5) is given as T, O or Figure 9.8 Plot of Z(/) for the case T > 317‘ad AWGN Channols 488 Digital Transmission through Bandlimit Chopra g where @ is called the roll-off factor, which takes values in the range 0 < 0 = 1,4) handwidth occupied by the signal beyond the Nyquist frequency yp in call ed ty excess bandwidth and is usually expressed as percentage of the Nygui: ieauenet For example, when « = 1, the excess bandwidth is 50%; when o = 1, the excey bandwidth is 100%. The pulse x(t) having the raised cosine spectrum is sinzt/T cos(rat/T) xt/T 1 — 4012/7? . cos(rat/T') = sinc(t/T) —qq7;2/T?" x(t) = (2a) | | a Note that x(t) is normalized so that x(0) = 1, Figure 9.9 illustrates the raised cosing s for a = 0, 1/2, 1. We note tha spectral characteristics and the corresponding pulses for a = 0, the pulse reduces to x(t) = sinc(t/T) and the symbol rate is 1/7 = 2W, When a = 1, the symbol rate is 1/T = W. In general, the tails of x(t) decay a 1/13 for a > 0. Consequently, a mistiming error in sampling leads to a series of intersymbol interference components that converges to a finite value. Due to the smooth characteristics of the raised cosine spectrum, it is possible to design practical filters for the transmitter and the receiver that approximate the overall desired frequency response. In the special case where the channel is ideal x0 Figure 9.9 Pulses having a raised cosine spectrum.9. 20 Design of Bandlimited Signals with Controlled ISI — Partial Response Signals ‘As we have observed from our discussion of signal design for zero ISI, it is neces to reduce the symbol rate 1/7 below the Nyquist rate of 2W symbols/sec in Orde, to realize practical transmitting and receiving filters. On the other hand, suppose ye choose to relax the condition of zero ISI and, thus, achieve a symbol transmission rate of 2W symbols/sec. By allowing for a controlled amount of ISI, we can achieve this symbol rate. We have already seen that the condition of zero ISI is x(n7) = 0 for n 40, However, suppose that we design the bandlimited signal to have controlled IS] at one time instant. This means that we allow one additional nonzero value in the samples {x(nT)). The ISI that we introduce is deterministic or “controlled”; hence, it can te taken into account at the receiver. We will discuss this case next. One special case that leads to (approximately) physically realizable transmitting and receiving filters is specified by the samples! 1,n=0,1 9.22) x07) = { 0, otherwise * Now, using Equation (9.2.14), we obtain aft 1=0-1 (9.2.25) a 0, otherwise ° which, when substituted into Equation (9.2.12), yields Z(f) =T +7 eF2I7, e238 0,1. 1 ; Lis convenient to deal with samples of x(t) that are normalized to unity for =r f t jon 9.2 ‘Signal Desi x section sign for Bandlimited Channels 491 in the preceding section, it is impossi A owever, for T= sip. we obtain possible to satisfy this equation for T < ay- 1 : xp= [alten new , otherwise a = | WeIPOW cos (FE w {i (3): lW This pulse and its magnitude spectrum are illustrated in Figure 9.12. It is calleda modified duobinary signal pulse. It is interesting to note that the spectrum of this signal has a zero at f = 0, making it suitable for transmission over a channel tht does not pass DC. We can obtain other interesting and physically realizable filter characteristics by selecting different values for the samples {x(n7)} and by selecting more that two nonzero samples. However, as we select more nonzero samples, the problem of unraveling the controlled ISI becomes more cumbersome and impractical. In general, the class of bandlimited signals pulses that have the form = = 1) sin2W(t — n/2W) x(t) 2 (Gy) 2n W(t —n/2W) and their corresponding spectra 1 00 — x (fh) em snns/W, Wsection 9.3 Probability of Error for Detection of Digital PAM 493 are called partial response signals when controlled ISI is purposely introduced by selecting two or more nonzero samples from the set (x(n/2W)}. The resulting signal pulses allow us to transmit information symbols at the Nyquist rate of 2W sym- bols/sec. The detection of the received symbols in the presence of controlled ISI is described in Section 9.3.2. 9g PROBABILITY OF ERROR FOR DETECTION OF DIGITAL PAM In this section, we evaluate the performance of the receiver for demodulating and detecting an M-ary PAM signal in the presence of additive white Gaussian noise at its input. First, we consider the case in which the transmitter and receiver filters Gr(f) and Ga(f) are designed for zero ISI. Then, we consider the case in which Gr(f) and Gr(f) are designed such that x(t) = gr(t) * ga(0) is either a duobinary signal or a modified duobinary signal. pa Probability of Error for Detection of Digital PAM with Zero ISI In the absence of ISI, the received signal sample at the output of the receiving matched filter has the form Yn = XOAm + Wy (9.3.1) where, from Equation (9.1.5) with H(f) = Gr(f), w x= [ Iorinras=8, (93.2) _w and wm is the additive Gaussian noise which has zero mean and has variance (Equation (9.1.7) oy = €_No/2- (9.3.3) In general, am takes one of M possible equally spaced amplitude values with equal Probability. Given a particular amplitude level, the problem is to determine the prob- ability of error. In the absence of ISI, the problem of evaluating the probability of error for digital PAM in a bandlimited, additive white Gaussian noise channel is identical to the evaluation of the error probability for M-ary PAM, as given in Section 8.5.1. The final result that is obtained from the derivation is 2(M — 1) 2€, Pu == ( 5) : (93.4)issi dlimited AWGN Channels 494 Digital Transmission through Ban Chota = ke i ‘erage energy/symbol and ¢. j But &, = 3€v/(M? — 1), €av = kEbay is the averag 10 Eh, ig wl 4 Hence, average energy/bit. | 2M = 1) 9 ( [900082 Mierw ) Pu = 2 (Gt = Do ay a form for the probability of error of M-ary PAM derives , seuion Se treatment of PAM in this chapter, we imposed the addi constraint that the transmitted signal is bandlimited to the band width allocated fory channel. Consequently, the transmitted signal pulses were designed to be banding ve zero ISI. wm rateonna no bandwidth constraint was imposed on the PAM signals con: sidered in Section 8.5.1. Nevertheless, the receivers (demodulators and detectors) in both cases are optimum (matched filters) for the corresponding transmitted Signals, no loss in error-rate performance results from the bandwidth Constraint for zero ISI and the channel does not distor the Consequently, when the signal pulse is designed transmitted signal. 9,32 Symbol-by-Symbol Detection of Data with Controlled ISI In this section, we describe a symbol-by-symbol method for detecting the information symbols at the demodulator when the received signal contains controlled ISI. This symbol detection method is relatively easy to implement. A second method, based on the maximum-likelihood criterion for detecting a sequence of symbols, is described in Section 9.3.4. This second method minimizes the probability of error but is a litle more complex to implement. In particular, we consider the detection of the duobinary and the modified duobinary partial response signals. In both cases, we assume that the desired spectral characteristic X(f) for the partial response signal is split evenly between the transmitting and receiving filters, ite., |Gr(f)| = |Ga(f)| = IX(/)I! For the duobinary signal pulse, x(n7) = 1, for n = 0, 1 and zero otherwise. Hence, the samples at the output of the receiving filter have the form Am + Am—1 + Wm, (9.36) Ym = bm + Wm where (dm} is the transmitted sequence of amplitudes and {wm} is the sequent of additive Gaussian noise samples. Let us ignore the noise for the moment a consider the binary case where am = -1 with equal probability. Then,.bm takes one of three possible values, namely, by = —2, 0, 2, with corresponding probabilit® 1 4 1/2, 1/4. If amy is the detected symbol from the signaling interval beginning iL Or — 1), its effect on bm, the received signal in the mth signaling interval °™ eliminated by subtraction, thus allowing a, to be detected. This process oS Tepeated sequentially for every received symbol. : noise eng ntior Problem with this procedure is that errors arising from the adit Gm-1, its re Propagate. For example, if the detector makes an error in detecting Gust. its effect on by is not eliminated; in fact, it is reinforced by the inc ‘action. Consequently. the detection Of am is also likely to be in error.gection 9.3 Probability of Error for Detection of Digital PAM 485 Error propagation can be avoided by precoding the data at the transmitter instead of eliminating the controlled ISI by subtraction at the receiver. The precoding is performed on the binary data sequence prior to modulation, From the data sequence {dy} of 1’s and 0's that is to be transmitted, a new sequence (pq) called the precoded sequence is generated. For the duobinary signal, the precoded sequence is defined as Pm = dm Pm, M=1, 2... y (9.3.7) where the symbol © denotes modulo-2 subtraction? Then, we set dm = —1 if pm =0, and am = Vif Pm = 1, i.e, ay = 2pm — 1. The noise-free samples at the output of the receiving filter are given as Bin = Am + Am—1 = (2pm — 1) + (2pm—1 — (9.3.8) = 2(Pm + Pm-1 — 1)- Consequently, bm Pm + Pm = + I (9.3.9) Since dm = Pm ® Pm-1, it follows that the data sequence dj, is obtained from bm by using the relation b, din = or +1 (mod 2). (9.3.10) Consequently, if bm = £2, dm = 0 and if bn = 0, dm = 1. The effect of precoding is clear from Equation (9.3.10). The received level at the mth transmission bm is directly related dj, the data at the same transmission time. Therefore, an error in reception of bm, only affects the corresponding data d,,, and no error propagation occurs. Example 9.3.1 For the binary data sequence {d,} given as 111010010001101, determine the precoded sequence {pn}, the transmitted sequence {a,}, the received sequence {b,)}, and the decoded sequence {du}. Solution By using the Equations (9.3.7), (9.3.8), and (9.3.10), we obtain the desired sequences, which are given in Table 9.1. . In the preceding derivation, we neglected the effect of the additive noise on the detection method. In the presence of additive noise, the sampled outputs from the receiving filter are given by Equation (9.3.6). In this case, Ym = Bm + Wm is 2 Although this is identical to modulo-2 addition, it is convenient to view the precoding operation for duobinary in terms of modulo-2 subtraction. In the Mary case, modulo-M addition and subtraction are clearly different.gh Bandlimited AWGN Channels, 496 _ Digital Transmission throu Chat TABLE 9.1 BINARY SIGNALING WITH DUOBINARY PULSES 1 0 01000 Data sequence da Herel tae, hare + Precoded sequence Pr or ort oO 09 OLFTE O1y Transmitted sequence an{ 1 1-1 1 1-1! -1 -1 0 002 0-2 20222 003% Tcl 0) og tie On 08 100" Ome alenaG Received sequence by Decoded sequence dy compared with the two thresholds set at +1 and —1. The data sequence (ay) ig | obtained according to the detection rule | 1, if —1 0, O1 > 1,10 — 2, and 11 + 3. Determine the precoded sequence {pn}, the transmitted sequence {dn}, the received sequence {bq}, and the decoded sequence (dy). Solution By using Equation (9.3.12) through (9.3.17), we obtain the desired sequences, which are given in Table 9.2. . In the presence of noise, the received signal-plus-noise is quantized to the nearest possible signal level and the preceding rule is used on the quantized values to recover the data sequence. In the case of the modified duobinary pulse, the controlled ISI is specified by the values x(n/2W) = —1 forn = 1, x(n/2W) = | for n = —1, and zero otherwise. Consequently, the noise-free sampled output from the receiving filter is given as bm = Am ~ Am=2s (9.3.18) where the M-level sequence {an} is obtained by mapping a precoded sequence according to the relation Equation (9.3.13) and Pm = 4m ® Pm-2 (mod M). (9.3.19) From these relations, it is easy to show that the detection rule for receiving the data sequence {dim} from {Dm} in the absence of noise is bm hg ry CH De (9.3.20) As demonstrated, the precoding of the data at the transmitter makes it possible to detect the received data on a symbol-by-symbol basis without having to look back at previously detected symbols. Thus, error propagation is avoided. The probability of error of the symbol-by-symbol detector previously described is determined in the following section. z=Chemnet 2 retron: ™~ GN Channels oy, 508 Digital Transmission through Bandlimited AW‘ hte, om , 4 the channel are known, Our objective was to design these filters for yop, the sampling instants. This design methodology is appropriate when the ¢ 0 1S ge with tin halt precisely known and its characteris ; f° often encounter channels whose frequency response chy, In practice, w tics are either unknown or change with time. For example, in data tr he communication channel will be di the dial-up telephone network, ¢ time we dial a number because the channel route will be diffe is made, however, the channel will be time-invariant for a re time. This is an example of a channel whose chara Examples of time-varying channels are radio channels, tion channels, These channels are characterized by time-varying frequency resp Spon characteristics. These types of channels demonstrate where the optimization of y transmitting and receiving filters, as described in Section 9.4.1, is not possible he Under these circumstances, we may design the transmitting filter to have square-root raised cosine frequency response, i.e., ~s au { VXefye Fl, [fl W = . ifl> WwW ete Mission o\. ifFerent gy nt. Once a CONNeg ety ac are unknown a prt ionospheric proms igs Gr(f) and the receiving filter, with frequency response Gr(f), to be matched to G7(,), Therefore, IGr(PlGR(A)| = Xrelf). (9.4.17) Then, due to channel distortion, the output of the receiving filter is © yO) = YO agx(t — nT) + wt), (9.4.18) n5=00 where x(t) = gr(t) * c(t) * gr(t). The filt sodi produce the sequence ter output may be sampled periodically to es Ym = > anXmn + Wm 3 = 200m + 5° anna + Um: oa ay am where x, = 2x(nT),n = ' ; ; Eavation ne te .. The middle term on the right-hand side niniber Of e a system, it is reasonable to assume that the IST affects # fini mymber of symbols. Henes, we may assume that x, = 0 for n < —Ly and 2 Ls the ontpat of thé cost finite, positive integers Consequently, the ISI observed a data sequence receiving filter may be viewed as being generated by passing ! {am} through an FIR filter with coefficients {xq,—L1 <1" < Lah*9.4 System Design j Section ‘stem Design in the Presence of Channel Distortion 509 (Gn) iby th L_, ouput = 3 21am Figure 9.18 Equivalent discrete-time channel filter. shown in Figure 9.18. This filter is called the equivalent discrete-time channel filter. Since its input is the discrete information sequence (binary or M-ary), the output of the discrete-time channel filter may be characterized as the output of a finite-state machine with L = L, + Lz states, corrupted by additive Gaussian noise. Hence, the noise-free output of the filter is described by a trellis having M¢ states. Maximum-Likelihood Sequence Detection. The optimum detector for the information sequence {am}, which is based on the observation of the received sequence {Ym} and given by Equation (9.4.19), is an ML sequence detector. The detector is akin to the ML sequence detector described in the context of detecting partial response signals that have controlled ISI. The Viterbi algorithm described in Section 13.3.2 provides a method for searching through the trellis for the ML signal path. To accomplish this search, the equivalent channel filter coefficients (xn) must be known or measured by some method. At each stage of the trellis search, there are M surviving sequences with M" corresponding Euclidean distance path metrics. Due to the exponential increase in the computational complexity of the Viterbi algorithm with the span (length L) of the ISI, this type of detection is practical only when M and L are small. For example, in mobile cellular telephone systems that employ digital transmission of speech signals, M is usually selected to be small, ic, M = 2 or 4, and 2 < L < 5. In this case, the ML sequence detector may be implemented with reasonable complexity. However, when M and L are large, the ML sequence detector becomes impractical. In such a case, other more practical but suboptimum methods are used to detect the information sequence {am} in the Presence of ISI. Nevertheless, the performance of the ML sequence detector for a channel with ISI serves as a benchmark for comparing its performance with that of suboptimum methods. Two suboptimum methods are described next.519 Digital Transmission through Bancimited AWGN Channels Chay, Linear Equalizers. For channels whose frequency response characteriy, are unknown and time variant, we may employ a linear filter with adjustable pars eters, which are updated on a periodie basis to compensate for the channel isons Such a filter, having parameters that are adjusted periodically, is called an adgpn re First, we consider the design characteristics for a linear equalizer from a fre. quency domain viewpoint. Figure 9.19 shows a block diagram of @ system iq employs a linear filter as a channel equalizer. The demodulator consists of a receiving filter with the frequency Fesponge Ga(f) in cascade with a channel equalizing filter that has a frequency respons. Ge(f). Since Gr(f) is matched to Gr(f) and they are designed so that they product satisfies Equation (9.4.17), |G_(f)| must compensate for the channe} dis. tortion. Hence, the equalizer frequency response must equal the inverse of the channe} response, i.e., 1 1 cH em where |Ge(f)| = 1/|C(f)| and the equalizer phase characteristic 0¢(f) = -6,(,). In this case, the equalizer is said to be the inverse channel filter to the channel response. ‘We note that the inverse channel filter completely eliminates ISI caused by the channel. Since it forces the ISI to be zero at the sampling times t = nT, the equalizer is called a zero-forcing equalizer. Hence, the input to the detector is of the form Ge(f)= eHN, IF < Ww, (9.4.29) Ym = Am + Wm, where wm, is the noise component, which is zero-mean Gaussian with a variance Nae a= | snicanrice near ES =f" SDN Lw ICP a if, (9.4.21) Input Fi data _ | Tansmitting ivit " tector filter} | Caan Rearene| Equalizer [Tod Grif) ” ater, Fy Gah) Noise n(t) Figure 9.19 Block diagram of a system with equalizer.