56 PATH LOSS AND SHADOWING received signal power falls off more quickly. Table 2.4: Coverage area But recall that we have set Phin = B,(R), 50 for different y and a5 the faster power falloff is already taken into ac- ore count (i.e., we need to transmit at much higher —""— ower with y = 6 than with y = 2 for this a 8 rp 7 y y equality to hold). The reason coverage area in- 2 7 oF 8 creases with path loss exponent under this as- 4 8 sumption is that, as y increases, the transmit 6 9 8 7 power must increase to satisfy Pain = P,(R). This results in higher average power through- ‘out the cell, yielding a higher coverage area. PROBLEMS 2-1. Under the free-space path-loss model, find the transmit power required to obtain a re- ceived power of 1 dBm for a wireless system with isotropic antennas (G; = 1) and a carrier frequency f = 5 GHz, assuming a distance d = 10 m. Repeat for d = 100 m. 2-2. For the two-ray model with transmitter-receiver separation d = 100 m, fiy = 10m, and h, = 2m, find the delay spread between the two signals, 2-3. For the two-ray model, show how a Taylor series approximation applied to (2.13) re~ sults in the approximation Qate tx =D _ Amhih, * OTN 2-4, For the two-ray model, derive an approximate expression for the distance values below the critical distance d, at which signal nulls occur. Ag= 2-5, Find the critical distance d, under the two-ray model for a large macrocell in a subur- bban area with the base station mounted on a tower or building (ht, = 20 m), the receivers at height h, = 3 m, and f. = 2 GHz. Is this a good size for cell radius in a suburban macro- cell? Why or why not? 2-6, Suppose that, instead of a ground reflection, a two-ray model consists of a LOS com- ponent and a signal reflected off a building to the left (or right) of the LOS path. Where must the building be located relative to the transmitter and receiver for this model to be the same as the two-ray model with a LOS component and ground reflection? 2-7. Consider a two-ray channel with impulse response h(t) = 018(1) +.25(t —.022 ps). Find the distance separating the transmitter and receiver, as well as or and 2, assuming free-space path loss on each path with a reflection coefficient of —1. Assume the transmitter and receiver are located 8 m above the ground and that the carrier frequency is 900 MHz. 2-8. Directional antennas are a powerful tool to reduce the effects of multipath as well as interference. In particular, directional antennas along the LOS path for the two-ray model can reduce the attenuation effect of ground wave cancellation, as will be illustrated in this problem, Plot the dB power (10 logio P,) versus log distance (logyo d) for the two-ray model with parameters f = 900 MHz, R = —1, h, = 50m, h, = 2 m, G; = 1, and the followingPROBLEMS 57 d .d) 11: System with scattering for Problem 2-11 values for G,: Gy =1, 316, .1, and .01 (ie., G, = 0, —5, -10, and —20 dB, respectively). Each of the four plots should range in distance from d = 1 m to d = 100 km, Also calcu- late and mark the critical distance d. = 4/;h-/2 on each plot, and normalize the plots to start at approximately 0 dB. Finally, show the piecewise linear model with flat power falloff. up to distance hy, falloff 10 logy(d~*) for h, < d < d_, and falloff 10 logy(d~*) for d > d.. (On the power loss versus log distance plot, the piecewise linear curve becomes a set of three straight lines of slope 0, 2, and 4, respectively.) Note that at large distances it becomes increasingly difficult to have G, < G; because this requires extremely precise angular di- rectivity in the antennas. 2-9, What average power falloff with distance do you expect for the ten-ray model” Why? 2-10. For the ten-ray model, assume that the transmitter and receiver are at the same height in the middle of a street of width 20 m, The transmitter-receiver separation is 500 m. Find the delay spread for this model. 2-11, Consider a system with a transmitter, receiver, and scatterer as shown in Figure 2.11. Assume the transmitter and receiver are both at heights h, = h, = 4 m and are separated by distance d, with the scatterer at distance .5d along both dimensions in a two-dimensional grid of the ground — that is, on such a grid the transmitter is located at (0,0), the receiver at (0,d), and the scatterer at (.5d, 5d). Assume a radar cross-section of 20 dBm, G, = 1, and fe = 900 MHz. Find the path loss of the scattered signal for d = 1, 10, 100, and 1000 meters. ‘Compare with the path loss at these distances if the signal is only reflected, with reflectis coefficient R = —1 2-12. Under what conditions is the simplified path-loss model (2.39) the same as the free- space path-loss model (2.7)? 2-43. Consider a seceiver wit se power —160 UBin within the signal bandwidth of in terest. Assume a simplified path-loss model with dy = I m, K obtained from the free-space path-loss formula with omnidirectional antennas and f, = 1 GHz, and y = 4. For a trans- mit power of P; = 10 mW, find the maximum distance between the transmitter and receiver such that the received signal-to-noise power ratio is 20 dB. 2-14, This problem shows how different propagation models can lead to very different SNRs (and therefore different link performance) for a given system design. Consider a linear cel- lular system using frequency division, as might operate along a highway or rural road (see58 PATH LOSS AND SHADOWING 2km = Base Station/Cell Center Figure 2.12: Linear cellular system for Problem 2-14 Figure 2.12). Each cell is allocated a certain band of frequencies, and these frequencies are reused in cells spaced a distance d away. Assume the system has square cells, 2 km per side, and that all mobiles transmit at the same power P. For the following propagation model: determine the minimum distance that the cells operating in the same frequency band must be spaced so that uplink SNR (the ratio of the minimum received signal-to-interference or S/I power from mobiles to the base station) is greater than 20 dB. You can ignore all interferers, except those from the two nearest cells operating at the same frequency. (a) Propagation for both signal and interference follow a free-space model. (b) Propagation for both signal and interference follow the simplified path-loss model (2.39) with do = 100 m, K =1, and y = 3. (©) Propagation for the signal follows the simplified path-loss model with do = 100 m, K = |, and y = 2, while propagation of the interfererence follows the same model but with y = 4, 2-15. Find the median path loss under the Hata model assuming f, = 900 MHz, h, = 20m, ie = 5m, and d = 100 m for a large urban city, a small urban city, a suburb, and a rural area, Explain qualitatively the path-loss differences for these four environments. 2-16. Find parameters for a piecewise linear model with three segments to approximate the two-ray model path loss (2.12) over distances between 10 and 1000 meters, assuming h, = 10m, hy = 2m, and G; = 1. Plot the path loss and the piecewise linear approximation using these parameters over this distance range. 2-17. Using the indoor attentuation model, determine the required transmit power for a de- sired received power of —110 dBm for a signal transmitted over 100 m that goes through three floors with attenuation 15 dB, 10 dB, and 6 dB (respectively) as well as two double plaster- board walls, Assume a reference distance do = 1, exponent y = 4, and constant K = 0 dB. 2-18, Table 2.5 lists a set of empirical path loss measurements. (a) Find the parameters of a simplified path-loss model plus log-normal shadowing that best fit this data. (b) Find the path loss at 2 km based on this model. (©) Find the outage probability at a distance d assuming the received power at d due to path loss alone is 10 dB above the required power for non-outage. 2-19. Consider cellular system operating at 900 MHz where propagation follows free-space path loss with variations about this path loss due to log-normal shadowing with « = 6 dB.PROBLEMS 59 Table 2.5: Path-loss measurements for Problem 2-18 Distance from transmitter P,/Pr 5m 60.48 25m 80 48 65m 105 a8 110m 115 dB. 400 m 135 48 1000 m 150 48 ‘Suppose that for acceptable voice quality a signal-to-noise power ratio of 15 dB is required at the mobile, Assume the base station transmits at 1 W and that its antenna has a 3-dB gain. There is no antenna gain at the mobile, and the receiver noise in the bandwidth of interest is —40 dBm. Find the maximum cell size such that a mobile on the cell boundary will have acceptable voice quality 90% of the time. 2-20. In this problem we will simulate the log-normal fading process over distance based on the autocovariance model (2.50). As described in the text, the simulation first generates a white noise process and then passes it through a first-order filter with a pole at e~*”*+, Assume X, = 20 m and plot the resulting log-normal fading process over a distance d ranging from (Om to 200 m, sampling the process every meter. You should normalize your plot about 0 4B, since the mean of the log-normal shadowing is captured by path loss. 2-21. In this problem we will explore the impact of different log-normal shadowing pa- rameters on outage probability. Consider a cellular system where the received signal power is distributed according to a log-normal distribution with mean jz dBm and standard devi- ation oy dBm. Assume the received signal power must be above 10 dBm for acceptable performance, (a) What is the outage probability when the log-normal distribution has ry = 15 dBm. 8 dBm? 4 dBm, find the value of zy required for the outage probability to be less than 1% —a typical value for cellular systems (©) Repeat part (b) for oy = 12 dBm. (a) One proposed technique for reducing outage probability is to use macrodiversity, where a mobile unit's signal is received by multiple base stations and then combined. This can only be done if multiple base stations are able to receive a given mobile’s signal, which is typically the case for CDMA systems. Explain why this might reduce outage probability. 2-22. Derive the formula for coverage area (2.61) by applying integration by parts to (2.59). 2-23. Find the coverage area for a microcellular system where path loss follows the simpli- fied model (with y = 3, do = 1, and K = 0 dB) and there is also log-normal shadowing with o = 4 4B. Assume a cell radius of 100 m, a transmit power of 80 mW, and a minimum. received power requirement of Pain = —100 dBm.60 PATH LOSS AND SHADOWING 2-24, Consider a cellular system where (a) path loss follows the simplified model with y = 6 and (b) there is also log-normal shadowing with « = 8 dB. If the received power at the cell boundary due to path loss is 20 dB higher than the minimum required received power for non-outage, find the cell coverage area 2-25. In microcells, path-loss exponents usually range from 2 to 6 and shadowing standard deviation typically ranges from 4 to 12. Given a cellular system in which the received power due to path loss at the cell boundary equals the desired level for non-outage, find the path loss and shadowing parameters within these ranges that yield the best and worst coverage area ‘What is the coverage area when these parameters are in the middle of their typical ranges? REFERENCES LI] D. Parsons, The Mobile Radio Propagation Channel, Wiley, New York, 1994 [2] M. Patzold, Mobile Fading Channels, Wiley, New York, 2002. (3] IEEE J. Sel. Areas Commun., Special Issue on Channel and Propagation Modeling for Wireless Systems Design, April 2002 and August 2002, [4] IEEE J. Sel. Areas Commun., Special Issue on Ultra-Wideband Radio in Multiaccess Wireless Communications, December 2002. [5] _T. S. Rappaport, Wireless Communications ~ Principles and Practice, 2nd ed,, Prentice-Hall, Englewood Cliffs, NJ, 2001 [6] | W.C. Jakes, Jr, Microwave Mobile Communications, Wiley, New York, 1974 [reprinted by I Press] [7] A. S.¥, Poon and R. W, Brodersen, “The role of multiple-antenna systems in emerging open access environments,” EE Times Commun, Design Conf., October 2003, [8] T. Kumer, D. J. Cichon, and W. Wiesbeck, “Concepts and results for 3D digital terrain-based ‘wave propagation models: An overview,” IEEE J. Sel. Areas Commun., pp. 1002-12, September 1993 19] _N. Amitay, “Modeling and computer simulation of wave propagation in lineal line-of-sight mi- crocells;” IEEE Trans. Veh. Tech., pp. 337~42, November 1992, [10] J.W. McKown and R. L. Hamilton, Jr, “Ray tracing as a design tool for radio networks,” IEEE Network, pp. 27-30, November 1991. U1] K.A. Remley, H.R. Anderson, and A. Weisshar, “Improving the accuracy of ray-tracing tech- niques for indoor propagation modeling,” IEEE Trans. Veh. Tech., pp. 2350-8, November 2000 (12) H-J.Li, C.-C. Chen, T-Y. Liu, and H.-C. Lin, “Applicability of ray-tracing techniques for pre- diction of outdoor channel characteristics,” IEEE Trans, Veh. Tech., pp. 2336-49, November 2000, [13] A. Domazetovic, L. J. Greenstein, N. Mandayan, and I, Seskar, “A new modeling approach for wireless channels with predictable path geometries,” Proc. IEEE Veh. Tech. Conf., September 2002, [14] J. H. Tamg, W--S. Liu, ¥-F Huang, and JM. Huang, “A novel and efficient hybrid model of radio multipath-fading channels in indoor environments,” JEEE Trans. Ant. Prop., pp. 585-94, ‘March 2003. [15] A.J. Rustako, Jr,, N. Amitay, G. J. Owens, and R. S. Roman, “Radio propagation at microwave frequencies for line-of-sight microcellular mobile and personal communications,” IEEE Trans. Veh. Tech., pp. 203-10, February 1991 [16] W.C.¥. Lee, Mobile Communications Engineering, McGraw-Hill, New York, 1982. LI7] J-B Wagen, “Signal strength measurements at 881 MHz for urban microcells in downtown Tampa,” Proc. IEEE Globecom Conf., pp. 1313-17, December 1991.94 STATISTICAL MULTIPATH CHANNEL MODELS where Yp,; = [x1 cos O,(t) + y; sin On(1) 1227/2 for (x;, y,) the antenna location relative to the origin and @,(r) the AOA of the multipath relative to the origin of the antenna array. Assume the AOA is stationary and identically distributed for all multipath components and denote this random AOA by 8. Let A(@) denote the average received signal power as a function of 6. Then we define the mean and rms angular spread in terms of this power profile as J7, 0A@) 48 He = = —__ (3.67) ', A@) 40 and | 2.@ — H0)?A@) a0 _= |e 3.68) \ f,4@ade G8) respectively. We say that two signals received at AOAs separated by 1/0 are roughly uncor- related. More details on the power distribution relative to the AOA for different propagation environments (along with the corresponding correlations across antenna elements) can be found in [23] Extending the two-dimensional models to three dimensions requires characterizing the elevation AOAS for multipath as well as the azimuth angles. Different models for such 3-D channels have been proposed in [24; 25; 26]. These ideas were used in [22] to incorporate spatiotemporal characteristics into Jakes's uniform scattering model. Several other papers on spatiotemporal modeling can be found in [27] PROBLEMS 344. Consider a two-ray channel consisting of a direct ray plus a ground-reflected ray, where the transmitter is a fixed base station at height h and the receiver is mounted on a truck (also at height h). The truck starts next to the base station and moves away at velocity v. Assume that signal attenuation on each path follows a free-space path-loss model. Find the time-varying channel impulse at the receiver for transmitter—receiver separation d = vf sufficiently large for the length of the reflected ray to be approximated by r +r! = d + 2h/d. 3-2. Find a formula for the multipath delay spread 7,, for a two-ray channel model. Find a simplified formula when the transmitter—receiver separation is relatively large. Compute Ty, for h, = 10m, h, = 4m, and d = 100m 3-3. Consider a time-invariant indoor wireless channel with LOS component at delay 23 ns, a multipath component at delay 48 ns, and another multipath component at delay 67 ns. Find the delay spread assuming that the demodulator synchronizes to the LOS component. Re- peat assuming that the demodulator synchronizes to the first multipath component. 3-4. Show that the minimum value of f.r, fora system at f. = 1 GHz with a fixed transmit- ter and a receiver separated by more than 10 m from the transmitter is much greater than 1. 3-5. Prove, for X and Y independent zero-mean Gaussian random variables with variance o°, that the distribution of Z = y/X? + Y? is Rayleigh distributed and that the distribution of Z? is exponentially distributed.PROBLEMS 95 3-6. Assume a Rayleigh fading channel with average signal power 2o* = —80 dBm. Whatis the power outage probability of this channel relative to the threshold Po = —95 dBm? How about Py = —90 dBm? 3-7. Suppose we have an application that requires a power outage probability of .01 for the threshold Po = 80 dBm. For Rayleigh fading, what value of the average signal power is required? 3-8. Assume a Rician fading channel with 207 = —80 dBm and a target power of Py = —80 dBm. Find the outage probability assuming that the LOS component has average power s? = 80 dBm. 3-9. This problem illustrates that the (ails of the Rician distribution can be quite differ- ent than its Nakagami approximation, Plot the cumulative distribution function (cdf) of the Rician distribution for K = 1,5, 10 and the corresponding Nakagami distribution with m (K +1)?/(2K +1), In general, does the Rician distribution or its Nakagami approximation have a larger outage probability p(y < x) for x large? 3-40. In order to improve the performance of cellular systems, multiple base stations can re- ceive the signal transmitted froma given mobile unit and combine these multiple signals either by selecting the strongest one or summing the signals together, perhaps with some optimized weights. This typically increases SNR and reduces the effects of shadowing. Combining of signals received from multiple base stations is called macrodiversity, and here we explore the benefits of this technique. Diversity will be covered in more detail in Chapter 7, Consider a mobile at the midpoint between two base stations in a cellular network. The received signals (in dBW) from the base stations are given by Pa =W+Z, Pa=W+Za, where Z;,> are N(0,07) random variables. We define outage with macrodiversity to be the event that both Py and P, > fall below a threshold 7. (a) Interpret the terms W, Zi, Z2 in P,,. and P,.» (b)If Z; and Z» are independent, show that the outage probability is given by Pox = ([Q(A/OF, where A = W — 7 is the fade margin at the mobile’s location. (©) Now suppose that Z; and Zp are correlated in the following way: Z =a¥, +Y, Zy = a¥, + bY, where Y, ¥), ¥) are independent N(0,?) random variables and where a,b are such that a? + 6? = 1. Show that tem [aloe ) (4) Compare the outage probabilities of (b) and (c) for the special case of a o =8, and A =5 (this will require a numerical integration). yi,96 STATISTICAL MULTIPATH CHANNEL MODELS 3-11. The goal of this problem is to develop a Rayleigh fading simulator for a mobile com- munications channel using the method of filtering Gaussian processes that is based on the in-phase and quadrature PSDs described in Section 3.2.1. In this problem you must do the following. (a) Develop simulation code to generate a signal with Rayleigh fading amplitude over time. Your sample rate should be at least 1000 samples per second, the average re- ceived envelope should be 1, and your simulation should be parameterized by the Doppler frequency fp. Matlab is the easiest way to generate this simulation, but any code is fine (b) Write a description of your simulation that clearly explains how your code generates the fading envelope: use a block diagram and any necessary equations. (©) Turn in your well-commented code. (@) Provide plots of received amplitude (4B) versus time for fy = 1, 10, and 100 hertz over 2 seconds 3-12, For a Rayleigh fading channel with average power P, = 30 dB and Doppler fp = 10 Hz, compute the average fade duration for target fade values of Py = 0.dB, Po = 15 dB, and Py = 304B. 343. Derive a formula for the average length of time that a Rayleigh fading process with average power P, stays above a given target fade value Po. Evaluate this average length of time for P, = 20 dB, Po = 25 dB, and fp = 50 Hz. 3-14. Assume a Rayleigh fading channel with average power P, = 10 dB and Doppler fi 80 Hz. We would like to approximate the channel using a finite-state Markov model with. ight states and time interval T = 10 ms, The regions Rj correspond to Ry = {y : —00 = y <—10 4B}, R= (y : -10dB < y <04B), Ry = (y : 0dB (1) = sin 2nf(t + At), find the minimum value of At for which the channel response to 1)(1) is approxi- mately independent of the channel response to u2(t) (4) Will this channel exhibit flat fading or frequency-selective fading for a typical voice channel with a 3-kHz bandwidth? For a cellular channel with a 30-kHz bandwidth? REFERENCES a (21 Bi) lay [5] (6) 'T. S. Rappaport, Wireless Communications ~ Principles and Practice, 2nd ed,, Prentice-Hall, Englewood Cliffs, NJ, 2001 D. Parsons, The Mobile Radio Propagation Channel, Wiley, New York, 1994 R. S. Kennedy, Fading Dispersive Communication Channels, Wiley, New York, 1969, R.H. Clarke, “A statistical theory of mobile radio reception,” Belt System Tech. J, pp. 957-1000, July/August 1968. W.C. Jakes, Jr, Microwave Mobile Communications, Wiley, New York, 1974. M. K. Simon and M.S, Alouini, Digital Communication over Fading Channels: A Unified Ap- roach to Performance Analysis, Wiley, New York, 2000.196 PERFORMANCE OF DIGITAL MODULATION OVER WIRELESS CHANNELS a og suk | gE gE j if 2 | 4 i t | t 1 : } 4 voll 4 Figur 6 redler vers norman edly ren for Gain power daly rfl {Reprinted by permission from [25, Fig. 9], © 1987 IEEE.) more sensitive to the rms delay spread of the channel than to the shape of its power delay profile. Moreover, pulse shaping can significantly impact the error floor: for the raised co- sine pulses discussed in Section 5.5, increasing f from 0 to I can reduce the error floor by over an order of magnitude. An example of Chuang’s simulation results is shown in Figure 6.5. This figure plots the irreducible bit error rate as a function of normalized rms delay spread d = or, /T, for BPSK, QPSK, OQPSK, and MSK modulation assuming a static channel with a Gaussian power de- lay profile, We see from the figure that for all modulations we can approximately bound the irreducible error floor as Poor < d? for .02 < d < .1. Other simulation results [24] support this bound as well. This bound imposes severe constraints on data rate even when symbol e11ur probabilities on the order of 10~? are acceptable. For example, the uns delay spread in a typical urban environment is approximately 67, = 2.5 1s. To keep 07, < .1T, requires that the data rate not exceed 40 kbaud, which generally isn’t enough for high-speed data ap- plications. In rural environments, where multipath is not attenuated to the same degree as in cities, o7, © 25 us, which reduces the maximum data rate to 4 kbaud. EXAMPLE 6.8: Using the approximation Fhay (01, /72)% find the maximum data rate that can be transmited through a channel with delay spread a, = 3 4s, usingPROBLEMS 197 cither BPSK or QPSK modulation, such that the probability of bit error Py is less than 10°. Solution: For BPSK, we set Phoor = (07, /Th)* and so require Ti, > o7,, /V Phoor = 94.87 1s, which leads to a data rate of R = 1/T, = 10.54 kbps. For QPSK, the same calculation yields T; > 07, /V/Phoge = 94.87 14s. Since there are two bits per symbol, this leads to a data rate of R = 2/T, = 21.01 kbps. This indicates that, for a given data rate, QPSK is more robust to ISI than BPSK because its symbol time is slower. ‘The result holds also when using the more accurate error floors associated with Fig- ure 6.5 rather than the bound in this example. PROBLEMS 641. Consider a system in which data is transferred at a rate of 100 bits per second over the channel. (a) Find the symbol duration if we use a sinc pulse for signaling and the channel band- width is 10 kHz, (b) Suppose the received SNR is 10 dB. Find the SNR per symbol and the SNR per bit if 4-QAM is used, (©) Find the SNR per symbol and the SNR per bit for 16-QAM, and compare with these metrics for 4-QAM 6-2. Consider BPSK modulation where the a priori probability of 0 and 1 is not the same. Specifically, p(s, = 0) = 0.3 and p(s, = 1) = 0.7. (a) Find the probability of bit error P, in AWGN assuming we encode a 1 as 5)(1) = Acos(2f-t) and a0 as s2(t) = —Acos(2zf.t) for A > 0, assuming the receiver structure is as shown in Figure 5.17. (b) Suppose you can change the threshold value in the receiver of Figure 5.17. Find the threshold value that yields equal error probability regardless of which bit is transmit- ted — that is, the threshold value that yields p(t = 0 | m = p(n = 1) = pO =1| m = O)p(m = 0). (©) Now suppose we change the modulation so that s(t) = Acos(2zf-f) and s2(1) = —Bcos(2nf.t). Find A > 0 and B > 0 so that the receiver of Figure 5.17 with threshold at zero has p(i = 0 | m = 1)p(n = 1) = pt =1 | m = O)p(m = 0). (d) Compute and compare the expression for P, in parts (a), (b), and (c) assuming E»/No = 10 4B and No = .1. For which system is P, minimized? 6-3. Consider a BPSK receiver whose demodulator has a phase offset of ¢ relative to the transmitted signal, so for a transmitted signal s(1) = -Eg(1) cos(2/-1) the carrier in the de- modulator of Figure 5.17 is cos(2zrf-t + @). Determine the threshold level in the threshold. device of Figure 5.17 that minimizes probability of bit error, and find this minimum error probability 6-4. Assume a BPSK demodulator in which the receiver noise is added after the integra- tor, as shown in Figure 6.6. The decision device outputs a 1 if its input x has Re(x} = 0, and a 0 otherwise. Suppose the tone jammer n(t) = L.le/®, where p(9 = nz/3) = 1/6 for198 PERFORMANCE OF DIGITAL MODULATION OVER WIRELESS CHANNELS no so Decision | 1070 x frou 1x] Pevisin | Por? cos(2rfet) Figure 6.6: BPSK demodulator for Problem 6-4 v29 16-QAM 30 ee 5-QAM 9-QAM Figure 6.7: Signal constellations for Problem 65. n= 0,1,2,3,4,5. What is the probability of making a decision error in the decision device (ie., outputting the wrong demodulated bit), assuming Ae = 2/7) = | and that informa- tion bits corresponding to a 1 (s(t) = Accos(2xf-t)) or a0 (s(t) = —A, cos(2zfet)) are equally likely. 6-5. Find an approximation to P, for the signal constellations shown in Figure 6.7. 6-6. Plot the exact symbol error probability and the approximation from Table 6.1 of 16- QAM with 0 < y, < 30 dB. Does the error in the approximation increase or decrease with %2 Why? 6-7. Plot the symbol error probability P, for QPSK using the approximation in Table 6.1 and Craig's exact result for 0 < y, < 30 dB. Does the error in the approximation increase or decrease with y,? Why?