Characterization of the Wireless Channel Impairments in the propagation channel have the effect of disturbing the information carried by the transmitted signal. Channel disturbance can be a combination of additive noise, multiplicative fading, and distortion due to time dispersion. The focus of this chapter is to characteriz2 the wire- less channel by identifying the parameters of the corruptive elements that distort the information- carrying signal as it penetrates the propagation medium, The corruptive elements are in the form of multipath delay spread, Doppler spread due t motion, and signal fading of the frequency- selective and non-frequency-selective variety. The propagation channel is normally time-variant. In this chapter, we first model the channel as linear time-variant, and then characterize Doppler spread, multipath delay spread, and fading in terms of second order statistics. 2.1 MULTIPATH PROPAGATION ENVIRONMENT ‘The wireless propagation channel contains objects (particles) which randomly scatter the energy of the transmitted signal. The scattered signals arrive at the destination receiver out of step ‘These objects (particles) are referred to as scatterers, Seatterers introduce a variety of channel impairments including fading, multipath delay spread, Doppler spread, attenuation, etc., and the inherent background noise. Background noise can be approximated as thermal noise and treated a8 additive white Gaussian noise (AWGN). Digital transmission over practical wireless channels is mainly limited by interference and distortion other than AWGN. 24Section 2.1 MULTIPATH PROPAGATION ENVIRONMENT 25 Scattering by randomly located scatterers gives rise to different paths with different path lengths. resulting in multipath delay spread. Consider a point source (a single tone sinusoid) as a test signal, If the propagation channel does not exhibit multipath delay spread, the point source would appear at the front end of the receiver as another point source. A multipath situation arises when a transmitted point source is received as a multipoint source, with each of the individually received points experiencing a different transmission delay. A pictorial view of the scattering phenomenon is depicted in Figure 2.1. The effect of multipath propagation on digital transmission can be characterized by time dispersion and fading. Time Dispersion. Because multiple propagation paths have different propagation delays, the transmitted point source will be received as a smeared waveform, Nonoverlapping scatterers give rise to distinct multiple paths. which are characterized by their locations in the scattering medium. As depicted in Figure 2.2. all scatterers are located on ellipses with the transmitter (Tx) and receiver (Rx) as the foci, One ellipse is associated with one path length. Therefore, signals reflected by scatterers located on the same ellipse will experience the same propagation delay. The signal components from these multiple paths are indistinguishable at the receiver. Signals that are reflected by scatterers located on different ellipses will arrive at the receiver with differential Scatterers ° =n Delay spread b 1 Transmitter Receiver Figure 2.1. Multipath spread due to channel scattering. Scanveesr Figure 2.2. Ellipsoidal portrayal of scatterer location.26 Chapter 2_ Characterization of the Wireless Channel delays. If the maximum differential delay spread is small compared with the symbol duration of the iransmitted signal, the chiannel is said to exhibit flat fading. If the differential delay spread is large compared with the symbol interval. the channel exhibits frequency-selective fading. In the time domain, the received signals corresponding to successive transmitted symbols will overlap, giving “ rise to a phenomenon known as intersymbol interference (ISL). IST is a signal-dependent distortion, The severity of ISI increases with the width of the delay spread, The IS} distortion in the time domain can also be examined in the frequency domain. ISI degrades transmission performance. Channel equalization techniques can be used to combat ISI, as discussed in Chapter 4. Fading. When the delay differences among various distinct propagation paths are very small compared with the symbol interval in digital transmission, the multipath components are almost indistinguishable at the receiver. These multipath components can add constructively or destruc- tively, depending on the carrier frequency and delay differences. In addition, as the mobile station moves, the position of each scatterer with respect to the transmitter and receiver may change. The overall effect is that the received signal level fluctuates with time, a phenomenon called fading. As an example, consider the transmission of a single-tone sinusoidal signal with frequency f. over a channel with two distinct paths, as shown in Figure 2.3. For simplicity. the delay of the line-of-sight (LOS) or direct path is assumed to be zero, and the delay of the non-line-of-sight (NLOS) or reflected path is x. The received signal, in the absence of noise, can be represented as (0) = ery COs(2at fot) + 2 Cos(2m felt — 7) Qi) where ) and cz are the amplitudes of the signal components from the two paths respectively. The received signal can also be represented as H(t) =acostrfit +). (2.1.2) where oP + a3 + 2ayer> cos(2x ft) and @ sinQr fet) ] ey + 3 cos(n fer) o= tan” [ aycost2 aft) Transmitter a,cost2af,@— 7) Figure 2.3. A channel with two propagation paths.Section 2.1. MULTIPATH PROPAGATION ENVIRONMENT 27 The received signal amy Figure 2.4 The amplitude fluctuation of the two-path channel with wy =2 anda) =1 are tne amplitude and phase of the received signal, Both a and @ are functions of a r. Figure 2.4 plots the amplitude a as a function of fr, where w; = 2 and a = | observed that the two received signal components add constructively when ft = 0.1.2... and destructively when for = 0.5, 1.5.2.5, .... In general, as the mobile station moves, @. @3, and t change with time. The received signal amplitude « and phase also change with time. When the signal components from the two paths add destructively. the transmitted signal experiences deep fading with a small value of the amplitude a, During each deep fade, the instantaneously received signal power is very low. resulting in poor transmission quality (i.e., high transmission error rate) Diversity and error-correction coding (with interleaving) are vilective w. combat channel fading tor better transmission accuracy, as discussed in Chapter 4 ‘The channel fading can be classified as long-term fading or short-term fading, The former is large-scale path loss, characterizing the local average of the path loss, as discussed in Section 24 ‘The latter describes the instantaneously received signal level variations with respect to the local average, as discussed in Section 2.5. In summary. multipath propagation in the wireless mobile environment results in a fading dispersive channel. The signal propagation environment changes as the mobile station moves and/or as any surrounding scatterers move, Therefore, the channel is time varying and can be modeled as a linear time-variant (LTV) system. In the following sections, we first study how to - describe a wireless channel using the LTV model in Section 2,2, and then focus on the correlation28 Chapter 2 Characterization of the Wireless Channel functions of the LTV channel in Section 2.3. The long-term and short-term fading models are then studied in Sections 2.4 and 2.5 respectively. 2.2 LINEAR TIME-VARIANT CHANNEL MODEL Consider a multipath propagation environment with A’ distinct scatterers. The path associated with the nth distinct scatterer is characterized by the 2-tuple (cm (1), ta(t)), where y(t) represents the amplitude fluctuation introduced to the transmitted signal by the scatterer at tiie f, t,(f) is the associated propagation delay, and = 1, 2,..., MW. Consider a narrowband signal #(7) transmitted over the wireless channel at a carrier frequency f., such that E(t) = xe"), Q20 where x(t) is the complex envelope of the signal and 3 denotes the real-valued component. In the absence of background noise, the received signal at the channel output is in(t)X(E = THA) = Rr(ei er}, where r(1) is the complex envelope of the received signal and can be represented as w H(t) = Yara (tye PHM c(t — tld) Note that the complex envelopes, x(#) and r(t), are respectively the equivalent representations of the transmitted and received narrowband signals at baseband. The channel can be characterized equivalently by its impulse response at baseband. Since the channel is time varying, the impulse response depends on the instant that the impulse is applied to the channel. 2.2.1 Channel Impulse Response Let us first review the impulse response of a linear time-invariant (LTD) channel, as shown in Figure 2.5(a), Let h(t) denote the impulse response (ie., the channel output when the channel input is an impulse applied at 1 = 0, 8(¢)). Here 8(¢) is the Dirac delta function and is defined by 6a) = {° st (2.2.3) and if 6) =1 (2.2.4) 00 Because the channel is time invariant, if the input is delayed by 11, the output is also delayed by 1) correspondingly. That is, the channel output in response to an input applied at time #1, 5(t—N),Section 2.2 LINEAR TIME-VARIANT CHANNEL MODEL 29 . att) Hy=hi) alee atte * \ : x(t) LTIchannet |_1(!) 8 ay ate 4) - r(t)=A(e- 4) 0% 0 (a) LTI channel oe) re) = y(t) t Cy mie ° x) LYV channel |_1(0) £ ht) S eet) Fe) = hy) # Aye 1) - i (b) ETV channel Figure 2.5 The linear channel model. is h(t — 11), AS a result, we can use the in,pulse response (+) to completely characterize the channel. The time variable 1 in the impulse response actually represents the propagation delay of the channel. The channel output r(2) is given by n ripamonarn= | x(t—tyh(r)dt, = 2, where * denotes the convolution operation Next consider an LTV channel, as shown in Figure 2.5(b), where /y(t) and z(t) denote the channel responses to the inputs 5(1) and 8(t — 1) respectively. As the channel propagation environment changes over the time duration (0, 11]. the channel output /ig(t) is not simply fy (1) delayed by 1 (ic. ho) # hile — 41). As a result, in order to characterize an LTV channel. we should define a channel impulse response as a function of two time variables, one describing the instant when the impulse is applied to the channel (initial time), and the other describing the instant of observing the channel output (final time). Definition 2.1 ‘The impulse response of an LTV channel, A(z, r), is the channel output at ¢ jin response to an impulse applied to the channel at 1 ~ t. In Definition 2,1, the variable + represents the propagation delay. From the definition and Eq. (2.2.5), the channel output can be represented in terms of the impulse response and the30 Chapter 2_ Characterization of the Wireless Channel channel input by rn) =| ACO = Edt, 2.2.6) The channel impulse response for the channel with V distinct scatterers is then y ACt.t) = Yo tg (eM — elt), (2.2.7) where 4,(0) = 2nfotq(t) represents the carrier phase distortion introduced by the nth scatterer. Thus, @, (1) changes by 2a radians whenever t)(z) changes by 1/fe. which is usually very small. This means that changes in 0, (t) have a far greater effect on the transmitted signal than changes in a(t), as a small change (such as motion) in the scatterer can cause a significant change in the phase 1), (1), but may not cause significant changes in the amplitude a7, (¢). Substituting Eq. (2.2.7) Eq. (2.2.6), we obtain Eq. (2.2.2). into 2.2.2 Time-Variant Transfer Function With the multipath channel characterized as a linear system, the channel behavior can also be examined in the frequency domain via a Fourier transformation, Time and frequency have an inverse relationship. Definition 2.2 ‘The time-variant transfer function of an LTV channel is the Fourier transform of the impulse response, A(z, £), with respect to the delay variable r Let H(f,r) denote the channel transfer function, as shown in Figure 2.6. We have the Fourier transform pair H1) = Frlh(r.)] f h(t, he Pl de A(c,t) = Fp [ACS 1)] aif Hf. et S df cd where the time variable can be viewed as a parameter. The received signal can be represented in terms of the transmitted signal and the transfer function as ro =i RG deaf, (2.2.8) where " RUA) = AL OX(F) and Xf) = Fla(o)). At any instant, say f = fy, the transfer function H(/. to) characterizes the channel in the frequency domain. As the channel changes with ¢, the frequency domain representation alsoSection 2.2 LINEAR TIME-VARIANT CHANNEL MODEL, 31 0) ae ri) ee A(r.t) — Hf.) | meet x0) al RUD = HON Figure 2.6 Frequency-time channel representation. changes with t. Therefore, we have the channel time-varying transfer function. If the channel is time invariant, then the impulse response is a function of the delay variable r and is independent of the time variable f; thus the transfer function varies only with the frequency variable f and is independent of t. This is consistent with the impulse response and transfer function of an LTL channel. 2.2.3 Doppler Spread Function and Delay-Doppler Spread Function Doppler Shifts. In general, the output signal of an LTI system does not have frequency compo- nents different from those of the input signal, That is, an LTT system does not introduce frequency shifts to its input signal. On the other hand, both nonlinear and time-varying systems introduce new frequency components other than those existing in the input signal. For a wireless prop- agation environment, due to the mobility of mobile users and/or the surrounding scatterers, the channel is linear but time Variant, As a result, a wireless channel introduces frequency shifts to the transmitted signal, a phenomenon called the Doppler effect and the introduced frequency shifts called the Doppler shifts Consider a scenario shown in Figure 2.7. where the base station (BS) transmits a single- tone pilot signal at frequency f., and the mobile station (MS) is moving along the x-axis with a constant velocity V, Let &(r) denote the angle of the incoming pilot signal viewed from the Figure 2.7 The Doppler effect.32 Chapter 2 Characterization of the Wireless Channel mobile receiver at time f, with respect to the x-axis. The received signal at the mobile terminal at time ¢ has a frequency of f+ 1(1). v(t) is the Doppler shift and is given by vy : vu) = — cos b(t), (2.2.9) € Where © is the velocity uf light, From Eq. (2.2.9), it is observed that the Doppler shift increases with the signal frequency. f. and uset velocity V. As a wireless channel can be characterized equival a channel being time varying in the time domain means a channel introducing Doppler shifts in the frequency domain. In fact, ay wireless channel usually introduces continuous Doppler shifts in a certain range including the zeny Doppler shift, the effect of the channel on the transmitted signal in the frequency domain is more spectral broadening than a simple spectral shift. This phenomenon can be illustrated by the following simple example Consider the wireless channel with N distinct scatterers as described by Eq. (2.2.7). Assume that the delay spread is negligible as compared with the symbol interval of the transmitted signal AS a result, the propagation delay can be approximated by its mean value £. For simplicity, we further assume that this mean delay does not change with time. The time-variant impulse response of the channel ean be approximately described in the form ently in both time and frequency domains, Att. t) = Z)8(t = (2.2.10) where Z(t) = Yoh yan (t)expl—j2xfetm(t)]. Given that the transmitted signal is x(e), the received signal in the absence of background noise is r(t) [ A(z, t)x(t =r) de Jose = [izose = #)jx@—r)de Z(t)x(t ~ 2). The overall effect of the channel is to provide a complex time-varying gain, Z(t), and a trans- mission delay, @, to the transmitted signal. In the frequency domain, the received signal is RA) =Flr)] = FIZ(t)x(t — 2) = FIZ@)] * Flatt — 7) = FIZ) IX Ne], where X(f) = F[x(¢)]. Given that the channel gain Z(2) changes with time, its Fourier transform F[Z(s)] has @ finite but nonzero pulse width in the frequency domain, As a result, the pulse width of R(f) is larger than the pulse width of X(f) due to the convolution operation. This meansSection 2.2 LINEAR TIME-VARIANT CHANNEL MODEL 33 that the channel indeed broadens the transmitted signal spectrum by introducing new frequency components, a phenomenon referred to as frequency dispersion Doppler Spread Function H(f, v). An LTV wireless channel can be characterized by the Doppler spread function H(f, v) which is defined by the relation between the channel input signal and output signal in the frequency domain. Let X(f) and R(/) denote the Fourier transforms of the transmitted signal x(t) and, received signal r(z), respectively. Then H(f. v) is defined by the following equation 20 ane f X(f —v)H(f — vy, v)dv, (2.2.11) Where v is a variable describing the Doppler shift introduced by the channel. From Eq. (2.2.11), H(f,v) is the channel gain associated with Doppler shift v to the input signal component at frequency f. Since both the time-variant transfer function H (f,r) and the Doppler spread function H(J,v) can be used to describe the same channel, there exists a relation between the two channel functions. It can be shown that Hf) =F LALO) f Hf. ne Pde HD = FMA Sf Hf wet? ay where the frequency variable f can be viewed as a parameter. The preceding Fourier transform relation verifies that being time-variant in the time domain can be equivalently described by having Doppler shifis in the frequency domain. Delay-Doppler Spread Function. A wireless channel can be characterized by the delay- Doppler spread function H(r, v) defined as the Fourier transform of the channel impulse response A(z, ¢) with respect 10 1, as follows: H(t.v) = Fhe of Atree Pde, (2.212) Given the channel input signal (7), it can be shown that the channel output signal is — 5 n= f Hp x(t — A(t, vel ™ dvdr. (2.2.13) ‘As both the transmitted signal and received signal can be represented either in the time ‘domain or in the frequency domain, we have four channel functions, A(t.t), H(f.0, H(t. v) and H(J, v) to characterize the relation between the transmitted signal and the received signal ~ These four functions are equivalent in describing the LTV channel. Preference in selecting any one of the functions depends on whether the transmitted and received signals are represented34 Chapter 2 Characterization of the Wireless Channel Impulse AG ssponse Ar de Pade FHL = if hr, he Pdr # op a Ln Delay-Doppler ansfer function spread H(r.v) ACh. af Fy Ha) = Fu thy. "Cg ne Pr FW(fv)y= Wwe Pdr Doppler spread Alf) Figure 2.8 Relationships among the channel functions. in the time or frequency domain. Figure 2.8 summarizes the relationships among the channel functions. 2.2.4 Example on the Channel Functions Example 2.1. LTV Channel Model Consider an LTV channel with impulse response given by A(z, t) = 4exp(—t/T)cos(2r), + 2 0, (2.2.14) where T = 0.1 ms and @ = 10x. a. Find the channel time-variant transfer function H(f,1) and the Doppler spread function Hf). b. Given that the transmitted signal is afb IlsT n= |f eet? (2.2.15) where Tp = 0.025 ms, find the received signal in the absence of background noise. c. Repeat part (b) if the transmitted signal is 4x2(t) = x\t—T)), (2.2.16) where 7; = 0.05 ms. d. What do you observe from the results of parts (b) and (c)?Section 2.2 LINEAR TIME-VARIANT CHANNEL MODEL, 35 Solution a. The time-variant transfer function is H(t) = Felhe, 1] = Fyldexp(-t/T)cos(@ry], 20 Acos(QnFrlexp(—t/T)], +20 _ AT cos(2t) “D+ jlafT” ‘The Doppler spread function is A(f.») = FAKE) AT cos(2t) = [rr a TE janft ar. See Qnv —Q)). Te jaapt eer Ore atv — 2) Filcos(2)] b. The received signal is calculated as follows: no=f Ae, txt = t)de u f 4exp(—1/T) cos(Qr nit — t) dr lb 0, t<-To Tr sr costeny[1 — ex ( 14h)). -IhTo The transmitted and received signals are plotted in Figure 2.9(a). c. Similar to part (b), the received signal is calculated as follows: L no= | A(t. xe — de =f 4exp(—t/T) cos(2t)x1(t — T, — t) dt ly36 Figure 2.9 The transmitted signals x;(t) and normalized received signals r(t)/T (f= 1 and 2 Example 2.1 is: Chapter 2_ Characterization of the Wireless Channel T lr a — x0) == (0-7 LOS Zz 0 = / s t x ji | \ \/ ails “| AIR lt ll 01-005 0 005 O01 O15 02 025 03 03 04 . t(ms) (a) x(t) and £(1)/T =15 L 1 L 02 03 (ms) (b)Section 2.3 CHANNEL CORRELATION FUNCTIONS. 37 0. mash 7 paddy 7 4 costa [1 ~ex0( =5+8)), T= er ST Th -t e AT cos(@t) [ew (--4=%) —exp (-=45)). 1>T +7 The transmitted and received signals are plotted in Figure 2.%b). 4d. From Figure 2.9, it is obsérved that: (1) the received signals have a larger pulse width than the corresponding transmitted signals because the channel is time dispersive: and (2) even though the transmitted signal x2(1) is x\(1) delayed by 7, the received signal r2(t) is not n(t) delayed by 7; because the channel is time varying. 2.3 CHANNEL CORRELATION FUNCTIONS & When the channel changes with time randomly, the channel impulse response A(¢, 2), time-variant transfer function H(f,1), Doppler spread function H(f, v), and delay-Doppler spread function H(x,v) are random processes and are difficult to characterize. Under the assumption that the random processes have zero mean, we are interested in the corvelation functions of the random processes. For simplicity of analysis, we assume that a. the channel impulse response h(t, ¢) is a wide-sense stationary (WSS) process; b. the channel impulse responses at 11 and 19, A(t), ©) and A(t», 1), are uncorrelated if 1) ¥ t2 for any 1 ‘A channel under assumptions (a) and (b) is said to be a wide-sense stationary uncorrelated scattering (WSSUS) channel. 2.3.1 Delay Power Spectral Density Under assumption (a), the autocorrelation function of h(x, ¢), defined as 1 SEU (ri h(a t+ ACL, is a function of 1, x, and Ar, and does not depend on 1. The superscript (*) denotes complex conjugation. The correlation function can be represented as ai Oa Cer 12, M0) = SETH Cer h(a. + A0)) (2.3.1) Furthermore, under assumption (6), the autocorrelation function can be represented in the form [12] Gn (TL, 12, At) = Ga(t1, ANS(er — 72), (2.3.2) or equivalently, Gn(T,T + Ar, At) = OR(t, At)5(Ar), (2.3.3)38 Chapter 2 Characterization of the Wireless Channel where a(t, At) foe t+ Ar, Andar. At At = 0, we define P(t) = by (t, ). (2.3.4) From Bqs. (2.3.1)-(2.3.4), we have nlt) = Farldn(t. t + At, ANMaro = Fa: [EU okte + acl} (2.3.5) From Eq. (2.3.5), we observe that ¢,(r) is the Fourier transform of the correlation function. According, to the Wiener-Khintchine relations!, the function @s(t) represents power spectral density (psd). It measures the average psd at the channel output as a function of the propagation Gelay, r, and is therefore called the delay psd of the channel, also known as the multipath intensity profile. Figure 2.10 illustrates an example of the psd, where the minimum delay is assumed to be zero. The nominal width of the delay psd pulse is called the multipath delay spread, denoted by Tn From the delay psd, we can calculate statistics describing the time dispersive characteristics of the channel. The nth moment of the delay, denoted by 1”, is given by [tenae . (2.3.6) foe dt tHe () o[-—7,— Figure 2.10 Delay power spectral density. |For a WSS random process, its power spectral density in the frequency domain is the Fourier transform of its autocorrelation function in the time domain, The relations between the autocorrelation function and the power spectral ensity are called the Wienes-Khintchine relationsSection 2.3 CHANNEL CORRELATION FUNCTIONS 39 The mean propagation delay, or first moment, denoted by @, is f toa(r) dt t= 23D \ / gn(t) dt and the rms (root-mean-square) delay spread, denoted by o:, is ia fo — iP on(rydr o, = | + (2.3.8) festoae In calculating a value for the multipath delay spread, it is usually assumed that Tm © 02. In the preceding calculations, the contribution of each delay r to the statistics is weighted by the psd level associated with the delay, and is normalized by the total power SL bu(rydr. 2.3.2 Frequency and Time Correlation Functions From the relationship between f(r, f) and H(f,1), we have the following: a, If A(x, 1) is WSS, then H(f, t) is also WSS with respect to rt. As a result, we can define the autocorrelation function of the time-variant transfer function H(f,t) as 1 ont hy fd? SEH NAC f t+ Ad), (239) b. The correlation function dz(fi. fo, 4t) can be represented in terms of the correlation function ¢)(r, At). bul fir fr 0) =f u(t, Ate PUPA ae =| Galt, Ate PROM de Zoyias.an where Af = fr—fi - The function $1(Af, At) is a time-frequency correlation function, Letting At = 0, we can write the frequency domain representation as a Fourier transform of the delay psd, i.c., ona) = Sef OHCF +4f0) f gare PR de (2.3.10)40 Chapter 2 Characterization of the Wireless Channel $4 (AF) is catled the frequency correlation function of the wireless channel, The delay psd, n(x). portrays the time domain behavior of the fading channel, whereas the frequency correlation function, $1(A/), portrays the frequency domain behavior. The nominal width of $4(Af), denoted by (Af), is called the channel coherence bandwidth, as shown in Figure 2.11. Since time and frequency have an inverse relationship, we have xe 4 (Ne = 3.11) (ie., a large multipath delay spread also means a small channel coherence bandwidth). Over the channel coherence bandwidth, all signal frequency components experience the corre- lated perturbation. When a signal propagates through the channel, the frequency components of the signal separated by a frequency width greater than the channel coherence bandwidth are distorted by the channel in an uncorrelated manner. The degree of fading experienced by the transmitted signal depends on the relationship between the channel coherence bandwidth and the transmitted signal bandwidth. Let W, denote the bandwidth of the transmitted signal. The fading channel can be grossly categorized as follows: If (Af)- < W,, the channel is said to exhibit frequency selective fading which introduces severe ISI to the received signal; If (Af) > Ws, the channel is said to exhibit frequency nonselective fading or flat fading which introduces negligible ISI. In the time-frequency correlation function (Af, An), letting Af = 0, we have L on(0 5 on O, An = ZEULDUCS + AOL. (2.3.12) x (At) is called the time correlation function of the channel, It characterizes, on average, how fast the channel transfer function changes with time at each frequency. The nominal width of ¢ 4 (A2) is called the coherence time of the fading channel, denoted by (At)-, as shown in Figure 2.12, lef af 0 (4). — Figure 2.11 Frequency-correlation function and channel coherence bandwidth.Section 2.3 CHANNEL CORRELATION FUNCTIONS a Neu (40)] 0 (4), Figure 2.12 Time-correlation function and channel coherence time. ‘The channel fading at time 1 will be very different from that at time ¢ —(At)e or earlier. If the channel coherence time is much larger than the symbol interval of the transmitted signal, the channel exhibits slow fading to the signal because the channel changing rate is much smaller than the transmission symbol rate. The time correlation function is independent of the frequency variable f due to the assumption that the channel has uncorrelated scattering. In other words, a channel exhibiting uncorrelated scattering in the time domain can be described equivalently as the condition that the channel transfer function is WSS in the frequency domain (with respect to f). 2.3.3 Doppler Power Spectral Density ‘The correlation function of the Doppler spread function H(f, v) is defined as 1 gElH hw H (fa. ¥2))- For a WSSUS channel, the correlation function can be represented in the form [12] DH (SF,m1)5(V1 = V2), where Af = fy— fy and (Af, v) can be obtained by Fourier transformation of the frequency- time correlation function 67 (Af, At) with respect to At. Therefore, Oy (Af.v) / oulAf, Ane Paar, (2.3.13) eo At Af =0, we have y(v) = On 0,v) = ip on(Atde "dar. 3.14) Equation (2.3.14) shows that the correlation function ®;(v) is the Fourier transform of the channel correlation function ¢4 (Ar). Based on the Wiener-Khintchine relations, 1,(v) is a psd in termsa Chapter 2_ Characterization of the Wireless Channel Oy (0) 0 a, Figure 2.13 Doppler power spectral density and Doppler spread. of the Doppler shift v. As a result, the function j(v) is called the Doppler psd. The nomi width of the Doppler psd, denoted by By, is called the Doppler spread, as shown in Figure 2.13 From the relation that $4 (At) and ©y(v) are a Fourier transform pair, we see that the coherence time is inversely proportional to the Doppler spread, ! (A), ® Ee 23.15) With the Doppler psd, the nth moment of Doppler shift, denoted by v", can be calculated by f rencar fener The mean Doppler shift, or the first moment, denoted by i, is va (2.3.16) feuooav fouoras and the rms Doppler spread, denoted by ay, is b= (2.3.17) 1R fo — bP Oy (dv ov (23.18) | y(v)dv As an approximation, it is usually assumed that Bg ¥ oy.Section 2.3 CHANNEL CORRELATION FUNCTIONS Frequency-time correlation by (Af Ad) af=0 idth, [Frequency correlation] [~ Time correlation , byl) (40 } tr iy! art Tu i Sheelapeneeen >t 1 Delay psd d4(7) Doppler psd yd) | — aes Figure 2.14 Relationships between the channel correlation functions and between the channel parameters. Figure 2.14 summarizes the relationships between the correlation functions and power spectral densities and between the channel parameters. 2.3.4 Examples on the Channel Correlation Functions Example 2.2. Calculation of the Channel Parameters © ——______———. Consider a fading channel which exhibits a Doppler frequency shift uniformly distributed between —10 Hz and 10 Hz. Determine a, the mean Doppler shift, b. the rms Doppler spread, and c. the coherence time. Solution The Doppler psd normalized with respect to the total power is -0> hh, we have di ~ d and dy * d. However, since the carrier wavelength 4 is very small as compared with c, a slight change in the propagation distance can cause a significant change in the received signal carrier phase. Depending on the difference between the carrier phases of the signals from the two paths, the received signal components from Figure 2.16 Two-path propagation over a flat plane.Section 2.4 LARGE-SCALE PATH LOSS AND SHADOWING 51 the two paths may enharice each other or cancel each other. The carrier phase difference is related to the difference between the two propagation distances dy and dy by an: R-A= Fh —d)). (24.7) ‘Taking into account the phase difference, the received signal power is P,(d) = P.GG, (a es)! [L + aye expli(e - #1”, (2.48) where ay and By are the amplitude attenuation and cartier phase shift, respectively, introduced by the reflection. If ay ~ I (ie., the reflection loss is negligible) and Ay = for @ « I, then < feaelA 2nd)... (2nAd\/? ra) = 2.6.6, (;47) |1 cos ( ) = ssin( 7 I ai? 2nd = 76.6, (45 4) [2=2eos( az | 2 int (EAA 2.49 = 406, (25 ) sin ( rs (249) where Ad = d> — dy. Given that d >> hy and d > hy, from Eqs. (2.4.5)-(2.4.6), it can be derived that Ng ees 24.10) d As a result, the received signal power is 2 A ‘2mhyh, P.(d) © 4P,G,G, (—— ) sin? (——* ee Ge) = ( Ad ‘The corresponding path loss is 1 all afd Datkihe wat =[4(4) vt (ue )} = -10e | (& 4) we (Mt) | (dB). (2.4.12) Figure 2.17 illustrates the path loss Lp(d) in dB as a function of the distance d, where fe 900 MHz, hy = 35 m, and fh, = 3 m, It is observed that (1) the path loss has alternate minima and maxima as the distance between the transmitter and receiver increases, and (2) in general, the path loss increases with the distance d. (24.11)52 Chapter 2 Characterization of the Wireless Channel Path loss 1, (d) (dB) 60 L 1 1 L L n 100200 300—S0—S=«SSC«ODSS~SSTO ~~ d(m) Figure 2.17. The path loss Lp(a) in dB versus distance d in the two-path model. 2.4.3 Log-Distance Path Loss with Shadowing As a mobile user moves away from its base station, the received signal becomes weaker because of the growing propagation attenuation with the distance. Let [.p(d) denote the log-distance path loss, which is a function of the distance d separating the transmitter and the receiver. Then, pd) x ($) .d2dy & or equivalently, Lyd) ad p(do) + 10« ween (2) dB, d > dy (2.4.13) lo where « is the path loss exponent and do is the close-in reference distance. Typically, do is 1 km for macrocells, 100 m for outdoor microcell, and 1 m for indoor picocells. Given dg, the value Lp(do) depends on the cartier frequency, antenna heights and gains, and other factors. Table 2.1 gives typical values of the path loss exponent in different propagation environments, Furthermore, as the mobile moves in uneven terrain, it often travels into a propagation shadow behind a building or a hill or other obstacle much larger than the wavelength of the transmitted signal, and the associated received signal level is attenuated significantly. This phenomenon is called shadowing. A log-normal distribution is a popular model for characterizing the shadowingSection 2.4 LARGE-SCALE PATH LOSS AND SHADOWING 53 Table 2.1 Path Loss Exponents for Different Environments Enyironment Path Loss Exponent, « free - z urban cellular radio 27035 shadowed urban cellular radio 3t05 in building with line of sight 16018 obstructed in building 4106 process. As a result, long-term fading is a combination of log-distance path loss and log-normal shadowing. Let (an) be a zero-mean Gaussian distributed random variable (in dB) with standard deviation o¢ (in dB). The pdf of eas) is given by Feias)() (2.4.14) Let Lp(d) denote the overall path loss with shadowing (long-term fading) in dB. Then, Lp(d) = Lp (d) + «ca (2.4.15) . da = Lp(do) + 10k logio ¢) +p) (AB), d > do. Since €4p) follows the Gaussian (normal) distribution with pdf given by Eq. (2.4.14), € in linear scale is said to follow a log-normal distribution with pdf given by (2.4.16) _ 20/1010 [-See*] $0) = Ta, OP a The first-order statistics of log-normal shadowing are characterized by the standard deviation o¢, which can be obtained from measurements. For example, 8 dB is a typical value for o in an ‘outdoor cellular system and 5 dB is a value for an indoor environment, 2.4.4 Okumura-Hara Path Loss Model ‘The Okumura-Hara path loss model (106, 60] was developed by curve fitting the measurement data collected-in Tokyo, Japan. It is suitable for outdoor macrocells. The path loss is represented as a function of (a) the carrier frequency f.-€ (150, 1000] MHz, (b) antenna heights of base station, hy € [30, 200] m, and mobile station, hm € [1, 10] m, and (c) the distance between the base station and mobile station d € [1, 20] km. The path loss in dB is given by A+ Blogio(d) for urban area - Lp(d) = {A+ Blogio(d)—C for suburban area , (24.17) A+ Blogio(d)— D for open area54 Chapter 2 Characterization of the Wireless Channel where A = 69.95 + 26.16 logio( fe) — 13.82 10g,9(hp) — alli), B = 44.9 - 6.55 logyo(hy), C HSA + 2logio( fe/28)F D = 40.94 + 4.78[logig(f-)]? — 18.33 logio( fe). and a(4m) is the correction factor for mobile antenna height, and is given by am) = (1.1 log fe) — 0.7}Am — [1.56 logyo( fe) — 0.8) for a small to medium city and 8.29[logig(1.54hm)]> — 1.1 for fe < 200 MHz Ahm) = > 3.2flogig(11.75h,) = 4.97 for fe > 400 MHz : for a large city. Figure 2.18 shows the Okumura-Hara path loss in large city, suburban and open areas respec- tively as a function of d in km, with parameters f- = 900 MHz. hy = 50 m, and hy = 3m. Large city ‘Open area Path loss Ld) (4B) (km) Figure 2.18 Propagation path loss using Okumura-Hara model.Section 2.4 LARGE-SCALE PATH LOSS AND SHADOWING 55 2.4.5 Lee's Path Loss Model Lee's model [81] can be used to predict area-to-area path loss. The model consists of two parts: (a) path loss prediction for a specified set of conditions. and (b) adjustment factors for a set of conditions different from the specified one. The moxlel ‘requires two parameters: (a) the power at a 1.6 km (I mile) of interception Py in dBm, and (4) the path-loss exponent «. The specified set of conditions is as follows: Carrier frequency fe = 900 MHz Base station antenna height = 30.48 m (100 ft) Base station power at the antenna = 10 W Base station antenna gain = 6 dB above dipole gain Mobile station antenna height = 3.m (10 ft) Mobile station antenna gain = 0 dB above dipole gain ‘The received signal power in dBm is represented by : P.(d) = Poapm) — 10K logio(4/do) — 10n logig( f/fe) + aoan), (2.4.18) Where dy = 1.6 km, d (= do) is the distance between the mobile station and the base station in km, and # is a constant between 2 and 3 dependent on the geographical locations and the operating, frequency ranges. n = 2 is recommended for a suburban or open area with f < 450 MHz and n= 3 for an urban area with f > 450 MHz. The parameter cyan) is an adjustment factor for a different set of conditions Aq 48) = 10 logy (a,a2a3a4as), (2.4.19) where ay — | NEW base station antenna height (m) = = 30.48 (m) = mobile station antenna height (m) | 3m) ay — BO transmitter power (W) 7 10 (W) dag = NEW base station antenna gain with respect to 2/2 dipole ign EW Dass atetlon tenn gain ‘with reepedt: to. 4/2 apo 4 a= as = different antenna gain correction factor at the mobile station The value v in ap is obtained from empirical data and is given by 2 for new mobile station antenna height > 10 m 1 for new mobile station antenna height < 3 m56 Chapter 2_ Characterization of the Wireless Channel Table 2.2 The Parameters in Lee's Path Loss Model in ‘Various Propagation Environments terrain Poin free space —45 ‘open area 49 | ‘suburban areas 617 3.84 urban area (Philadelphia) | —70 3.68 urban area (Newark) —6+ 431 urban area (Tokyo) —84 3.08_| ‘A 2.4B signal gain is provided by an actual 4 dB gain antenna at the mobile unit in a suburban area, and less than 1 dB gain received from the same antenna in an urban area for adjusting as. The two required parameters Poasm and x have been determined for various propagation environments based on empirical data and are given in Table 2.2. The propagation path loss using Lee's model for some environments given in Table 2.2 is plotted in Figure 2.19, where the base station antenna height is 50 m with a gain of 6 dB ‘with respect 0 a 4/2 dipole, the transmitter power is 10 W, the mobile station antenna height is 3 m, and the carrier frequency jf. is 900 MHz, 150 140 130) 120 0 90. Path loss L,,(d) (dB) 70, 2 4 6 8 Ww 2 14 16 B20 d (km) Figure 2.19 Propagation path loss using Lee’s mode!Section 2.4 LARGE-SCALE PATH LOSS AND SHADOWING 57 2.4.6 Radio Cell Coverage Radio cell coverage is the service area supported by each base station. The coverage depends on (a) service quality requirements, such as the required ratio of the signal power to interference-plus- noise power (see Chapter 5 for more details), or the required minimum received signal power level given the transmitted signal power, and (h) the propagation environment, For example, for free space transmission with an omnidirectional antenna, the cell coverage is a circle centered at the base station (transmitter) with a radius depending on the propagation loss. Given the transmitted signal power level, the minimum required received power level can be mapped to the maximum allowed path loss, Then from Eq. (2.4.3), the cell radius (d) can be determined. In practice, the path loss depends on the propagation environment, including the transmitter and receiver anteana heights, and may differ from angle to angle as seen from the base station transmitter antenna. As a result, the cell coverage will generally have an irregular shape (not a circle). Furthermore, because of the random nature in the path loss due to factors such as shadowing, the cell coverage is not deterministic but should be specified based on statistic parameters. The following example illustrates how to determine the cell coverage for a given propagation model, where the service quality criterion is specified in terms of the propagation path loss 3 Example 2.5 Radio Cell Coverage for the Log-Distance Path Loss| —_____— Consider the log-distance path loss model. Determine the cell coverage for the following wo situations: a, Without shadowing, the path loss is given by Eq. (2.4.13). It is required that, at the cell border, the path loss cannot be y dB larger than that at the reference distance do. b. With shadowing, the path loss is given by Bq. (2.4.15). Due to the random nature of the propagation path loss, the cell coverage is defined as the service area of the base station ‘over which the path loss (over that at the reference distance do) is limited to y dB with a pre-defined probability. Assume the reference distance do is very small compared with the cell radius. Solution a. The path loss in dB at a distance d (> do) from the base station is = ie d Lp(d) = Lpldo) + 10¢ logy (5) : a As the path loss monotonically increases with the distance d and is independent of the pfopagation direction as observed at the base station, the cell coverage is a circle centered at the base station with radius R specified by a sy. AL (d)la= g lank _ = d Lp(@ila-a ~ Eg(4) = We oe, (+) Solving the above equation for the radius R, we obtain RK dy x 1010/0,58 Chapter 2. Characterization of © _vireless Channel b. With shadowing, the path loss in dB at a distance d (> dp) from the base station is . : - fd L 5 (d) = L (do) + 10 logig (¢) + eB), a where the random variable (an) characterizes the effect of shadowing and is modeled by a Gaussian random variable with zero mean and standard deviation o, (in dB). The relative path loss at d (> dp) with respect to the loss at do is given by Aip(d) 2 L,(d) = Lp dp) = 10k logio (5) + eu). b The problem of estimating the cell coverage can be approached in two steps: Step 1: Determine the probability a; that the path loss at location r (> do) is below the threshold y, where the probability is averaged over the circumference: Step 2: Calculate the probability a2 of the circular area (defined by R) over which the path loss is below the threshold based on a), where the probability is averaged over the circular area (cell area). At the distance d = r (> do), the probability that the relati ay(r) = P(ALp(r) < y) P(€(aB) < y — 10k logyg(r/do)) 10k logy9(r/de =o( wile) cm where Q(x) & Px Ie exp(—1?/2)dr. Next, we want to find the probability a1 of the event “ALp(r) > do and dA = rdOdr as shown in Figure 2.20. Substituting Eq. (2.4.20) into Eq. (2.4.21), we have Looper R a ~ =n wf a\(r)rdr f ° (2 = 10K st)Section 2.4 LARGE-SCALE PATH LOSS AND SHADOWING _ 59 Figure 2.20 The cell coverage area where with é 10k logio(R/do) % an 10K I ‘ loBi0 % As a result, 1 en, ] dt | xdx = flae Lom atbiniR/ds) 4 1 thes =2 : = [ xdx eT Padt 5 Var | Jexwi%= * Love 7 owt] -8-(9) ao+0n(8) -oo(22)[o(e+3) ef beoe(S)) If do & R, we have wc on(25¢4)[o(er2)-o(-+ eo(2)))60 Chapter 2 _ Characterization of the Wireless Channel Given the required a2 value, the cell radius R can be determined numerically from the above equations. 2.5 SMALL-SCALE MULTIPATH FADING 2.5.1 First-Order Statistics ‘Suppose the multipath channel is characterized by N distinct scatterers in which the nth scatterer is associated with a gain aq(t) and a delay r,(t), where n = 1,2,...,.V. Consider digital transmission over the channel at the carrier frequency fc with a symbol interval much larger than the channel multipath delay spread. From Eq. (2.2.2), the received signal at baseband, in the absence of background noise, is N 1) = Doan tye PF (0 — m9) nal m Ny ® Yaane ee] -i). = ‘The approximation is reasonable as long as the delay spread is much smaller than the symbol interval. The complex gain of the channel is Z(t) = SV _jaq(t)e~/2*/e%(, Let Z(t) and ~Z,(t) denote the real and imaginary components of the complex channel gain, respectively, so that Z(t) = Ze(1) — jZ,(1). Then i Z(t) = Y an(t) cos Ont) @n(t) sin By (0), where 6,(t) = 2m fetn(t)- Furthermore. let a(t) = (22) + 220), 8) tan Z, (0)/Z-0)) be the amplitude fading and carrier distortion introduced by the channel. The fading characteristics can be studied by examining the pdfs of the envelope a(r) and phase 6(¢) at any time 1. The fading characteristics depend on whether the transmitter and receiver are in line-of-sight or not in line-of-sight. The former case is called LOS scattering while the latter case is referred to as NLOS scattering. LOS seattering has a specular component (from the direct path), and can be modeled as a Rician distribution. NLOS scattering does not have a specular component, and can be modeled as a Rayleigh distribution. A pictorial view of LOS and NLOS scattering is depicted in Figure 2.21.Section 2.5 SMALL-SCALE MULTIPATH FADING 61 NLOS path. ‘Transmitter Receiver Figure 2.21 NLOS versus LOS scattering. Rayleigh Fading (NLOS propagation). In this case, E(Z-(1)] = E{Zs()] = 0. (2.5.1) Assume that, at any time 1, for n = 1,2,...4., a. the values of 6,(t) are statistically independent, each being uniformly distributed over {0, 2m}; b. the values of a(t) are identically distributed random variables, independent of each other and of the 8,(¢)"S. : According to the central limit theorem, Z(t) and Z,(t) are approximately Gaussian random variables at any time ¢ if V is sufficiently large. For simplicity of notation, let Z- and Z, denote Ze(t) and Z;(t) at any time f. It can be shown that Z, and Z, are independent Gaussian random variables with zero mean and equal variance o: | Ela2}, where of denotes aq(f) at any time ¢. As a result, the joint pdf of Z- and Z, is 1 x? py? f2.2,&. 9) = sz 80 | —* , OO 0) and at the level R, which can be expressed by cy yi Maa, Yieardy. (2.5.9) For the Rayleigh fading environment studied in Subsection 2.5.1, it can be shown that [130] 2 Sai(X, Y) = ro) 420,00 ) 4/5, From the definitions of LCR and AFD, we have Ne oka = lim ae (2.5.12) Equation (2.5.12) provides a relationship among the three statistics (i.¢., LCR, AFD, and the cumulative distribution function (cdf) of the amplitude fading «). Thus, if any two statistics are known, the third one can also be determined. For the Rayleigh fading environment, the cdf of ais " 2 Pasn= f faldy =1~exp(-35). (2.5.13) isSection 2.5 SMALL-SCALE MULTIPATH FADING 67 By Eqs, (2.5.11)-(2.5.13), the corresponding AFD is _ expe?) = (2.5.14) ‘The AFD is a product of two components. The first component, 1/(v27Um). indicates that the AFD is inversely proportional to the mobile speed and the carrier frequency. The second term, [exp(p") — 1]/p, depends only on the normalized threshold p. Figure 2.26 shows how the component changes with the normalized threshold in dB. The value of the AFD increases dramatically as the threshold p increases much above the rms value. This can be explained from Figure 2.25. With a large threshold value, it is very unlikely for the amplitude level @ to cross the threshold, Therefore, the length of time that a stays below the threshold can be very long. Knowledge of the AFD value helps to determine the most likely number of bits that may be lost during a fade, This is useful for relating the received signal-to-noise ratio (SNR) during & fade to the instantancous bit error rate (BER) 108 fexn(y Ww 10! i z pldB) Figure 2.26 The normalized average fade duration of the flat Rayleigh fading channel68 Chapter 2_ Characterization of the Wireless Channel Example 2.6 The LCR Ng —— Consider a flat Rayleigh fading channel. Determine the positive-going level cros 6 =I, when the maximum Doppler frequency v,, = 20 Hz. Compute the maximum velocity of the mobile if the carrier frequency is 900 MHz. ing rate for Solution ! Na = Vitvmpe™?; vy, =20Hz, p= = Vin x 0x1 xe! = 18.44 crossings per second. The maximum velocity is Vn = vm «= 20 x 1S m/s = 6.66 mis = 24 knvhi Example 2.7. The AFD xx Suppose a flat Rayleigh fading channel exhibits a maximum Doppler frequency vp 200 Hz. a. Determine the average fade duration for a normalized threshold level p = 0.1. b. How does the result in part (a) changes when p is increased to 1.0? c. How does the result in part (a) changes when p is reduced to 0.017 Solution a. Atp=0.1, eon? oils x 200 x 0.1 b. Increasing the threshold level will increase the AFD value. At p = 1, Example 2.8 The LCR Ng and AFD xy? Consider a mobile cellular system in which the carrier frequency is fz = 900 MHz and the mobile travels at_a speed of 24 km/h. Calculate the AFD and LCR at the normalizéd level p = 0.294. Solution At f. = 900 MHz, the wavelength is. = = at = 4m. The velocity of the : = $82 _ mobile is V = 24 km/h = 6.67 mV/s. The maximum Doppler frequency is vm = V/2 = $5 = 20 Hz, The average duration of fades below the normalized level p = 0.294 is er (0-294)? _ 8 Pininp Vin x20 % 0.094 = 0.0061 s.ENDNOTES 69 The level crossing rate at p = 0.294 is, Ng = Vin vmpe™ = Viet x20 x 0.2946~10294" = 16 uperossingv/second. SUMMARY Unlike a guided wire, the wireless propagation channel is prone to atmospheric conditions. A consequence of this is that electromagnetic propagation through the wireless channel will suffer different degrees of impairment. To understand the channel disturbance on the transmitted signal, so as to facilitate transmitter and receiver design as°a means of combating channel impair- ments, we have examined commonly used analytical methods for modeling the channel char- acteristics. The next chapter will study modulation methods commonly used in & mobile radio environment. Z ENDNOTES L. For a general discussion on channel fading and its characterization, see the paper by Sklar [140] and the book by Kennedy [71]. 2. For detailed discussion of the LTV channel model, the channel functions, and correlation functions, see the paper by Bello {12} 3. For background on random variables, probability density distributions, and random processes. see the book by Papoulis [108] 4, For detailed derivation of the free-space propagation path loss, see the book by Pratt and Bostian [122] : 5. For path loss models in an outdoor environment, see the papers by Okumura et al. [106] and Hata [60] for the Okumura-Hata model, and Chapter 2 of the book by Lee [81] for Lee's model. Other models are given in {129}, [89], [43], [36], and [156] For cell coverage area, see the papers by Leonardo and Yacoub [83, 84]. Characterization of an indoor propagation channel is teviewed in the paper by Hashemin [59]. A statistical model for indoor propagation is given in the paper by Saleh and Valenzula [133]. . 8. For detailed analysis of the short-term flat fading characteristics (such as Rayleigh and Rician fading), see the papers by Clarke (30] and Lee [78], Chapter | of the book edited by Jakes [42], and Chapters 3 and 4 of the book by Lee [82]. For computer simulation of correlated Rayleigh and Rician fading channels, see the book edited by Jakes [42] and the paper by Young and Beaulieu [162]. ®. Other references on wireless propagation channels include the textbooks by Griffiths [56}, Parsons [109], and Rappaport [128]. . ae70 Chapter 2_ Characterization of the Wireless Channel PROBLEMS P2-1 For a linear time-iavariant channel with impulse response (7), given the channet input x(t), the channel output is 72) = x (0)eh(e). Verify that this channel input-output relation is a special cave of Bg. 2.2.6) P2-2 A wireless cha where T = 0.05 ms. Q = 10x, and 6 € [—7, +r] is a constant a. Determine the channel time-variant transfer function. b. Given that the channel input signal is l, Osr0. Assuming that T = 10 ns, determine a. the mean propagation delay, b. the rms delay spread, and c. the coherence bandwidth of the channel. > P2-7 The time-variant transfer function of a WSSUS channel is given by Hf, 1) = a(tel# a. Does the channel exhibit frequéncy-selective fading or flat fading? Explain, b. Derive the channel impulse response f(r, 1). c. Given that the transmitted signal is x(t), derive the received signal r(t) in the absence of background noise. From the relation between x(f) and r(0), find the mean delay and delay spread of the channel P2-8 Let W, be the signal bandwidth and Ty be the multipath delay spread of the channel. a. Discuss the conditions under which the channel exhibits (1) frequency selective fading, (2) flat fading.PROBLEMS n P2-9 P2-10 P.-L 2-12 P2-13 “P24 b. Specify the channel conditions urider which the time-variant channel transfer func- tion is Hf.) sae, where a(t) and 6(r) are real-value time functions, Does the channel introduce Doppler shifts? Explain, ‘The delay psd of a WSSUS channel is given by ay _ [0.5 +e0sQrz/T)), Ost hy. Consider the cell coverage under the assumption of the free-space propagation model. If it is required that the path loss at the cell border should not be larger than o dB, find an expression for the cell radius. In practice, the cell coverage depends on the propagation environment, Consider a log- normal propagation environment. The received signal power P, at a distance d (> dp) is Prcapwy = Pocsaw; — 10x logig(d/do) + €any, where Po is the received power at a properly chosen reference distance do, « is the path Joss exponent, é(gg) is a random variable uniformly distributed over [—a, -+a]. The cell coverage is defined as the service area of a base station over which the signal power (from the base station) received at a mobile station is larger than & with a probability of a. Find an expression for evaluating the cell coverage area. For the Rayleigh fading environment studied in Section 2.5, a, show that Z, and Z, are independent, and b. derive the joint pdf of Ze and Z, as given in Eq, (2.5.2). ‘The channel gain of a flat Rayleigh fading channel, Z. — jZ;, can also be represented as aexp(—j6), where the amplitude fading « = /Z2-+ Z2 and the phase distortion 6 =tan|[Z,/Ze]. a. Derive the joint pdf of a and 0, b. Show that @ and @ are independent ¢. Find Ela] and Elo]. Derive the pdf of the carrier phase distortion in the Rician fading, as given by Fq. (2.5.7),PROBLEMS 7B 2-22 Consider a cellular system with a carrier frequency of 2,000 MHz. Suppose the user is in- a vehicle moving at a speed of 60 knv/hr, Assuming that the channel exhibits flat Rayleigh fading, find a, the LCR at the normalized level of —3 dB, b. the AFD at the normalized level of ~3 dB 2.23 The LCR at the normalized threshold p for a flat Rayleigh fading channel can be repre- sented as Na =V2HVm MR: where ng = pexp(—p?) is the normalized LCR with respect to 271 Vy, and Ym is the maximum Doppler shift. a. Find the normalized threshold value po at which the LCR mg achieves the maximum value. b. Explain why the LCR at p decreases as p deviates from pp, 2-24 Figure 2.27 gives a block diagram to generate N’ uniform samples of the complex gain. Zelt) — jZq(t), of a flat Rayleigh fading channel [162]. Let 7, denote the sampling period ins, vm the maximum Doppler shift in Hz, and a the parameter in the Rayleigh distribution, The two Gaussian random variable generators generate two independent and identically distributed (jid) random sequences {XJ} and (Y4l|{p. each sample having zero mean and variance a? as follows: ELXi] = El%e] =0 ELXG) = EL¥e] ELGX) = EY ELM) =0, KLE iy, kal ‘The sequence {Yq} is converted {0 the imaginary component by multiplication with ~j The summed sequence, [Xi — j¥i). is a complex Gaussian sequence. The complex iid zero-mean Gaussian random \wariable generator iid zero-mean Gaussian random variable generator|{¥4) af Figure 2.27 Block diagram of a Rayleigh fading channel simulator.74 Chapter 2_ Characterization of the Wireless Channel sequence is weighted by a sequence of filter coefficients {Fi} given by 0, k=0 ha Rhee k= kn ke km + heveee N km —1 k=N—ky is k=N—km + 1,....N-2N=1 where fin = VmTs. and km = (fmN| is the largest integer less than or equal tof. An inverse discrete Fourier transform (IDFT) is then taken of this filtered complex sequence to form the complex time sequence of length N, {Zen — JZsn}|N ign WhETE Zon = Zc(nTs) and Zs, = 25(nT). Both zcq and zs, have zero mean and the same variance o? given by Sh eaof (#) ‘ As a result, both Zen = Zen/(V202) and Zyn = Zon/( 202) have zero mean and variance 1/2. The second moment of the amplitude fading is normalized to unity so that E[a2] = 1. a. Write a program to simulate the complex channel gain of the Rayleigh fading channel. Note that the IDFT block can be implemented by inverse fast Fourier transform (IFFT). The input parameters are the user velocity V, carrier frequency fe. sampling interval T,, and sequence length N. b. Given V = $0 km/hr, f- = 900 MHz, T; = 1 ms, and N = 1000, plot the empirical pdfs for the amplitude fading sequence {a} and phase distortion sequence (On), where an = /Z2, + Z2, and 0, = arctan(Zsn/Zen). Compare the empirical pdfs with the theoretical ones, respectively. Using the same input parameters as in (b), find the empirical LCR and AFD values for the amplitude fading sequence {a,) and compare them with the theoretical values. Zz