Errata Introduction to Wireless Systems P. Mohana Shankar • Page numbers are shown in blue • Corrections are shown in red October 2011 Page 11 below Figure 2.4 The power detected by a typical receiver is shown in Figure 2.5. Page 13 Page 14 Pr d Pr dref , dref d 2 d dref , (2.4) Example 2.1; Answer The wavelength at 900 MHz is 3.10 8/9.108 = (1/3)m. Expressing the frequency in MHz, eqn. (2.6) can now be expressed as Lfree 32.44 20log10 f 20log10 d , (2.7) where d must be larger than 1 km. The unit of f is in MHz and d is in kilometers. Page 17 top of the page to account for the terrain. Additional correction factors can be included to account for other factors such as street orientation. These correction factors They take into account the following: Page 15 Figure 2.7 Transmitted power = 100 mW instead of 100 dBm Page 19 Example 2. 3 Find approximate values of the loss parameter, n, using Hata model for the four geographical regions, namely large city, small-medium city, suburb, and rural area. Answer: Using Figure 2.12, the loss values at a distance of 3 km are 131.36 dB, 131.34 dB, 121.40 dB, and 102.8 dB, respectively, for large city, small-medium city, suburb, and rural area. Page 27 Page 28 Figure 2.18 X axis label…, Envelope a or Power p Sentence below eqn. (2.40) Eqn. (2.40) Page 34 ………where p0 is the average power p pout 1 exp thr p0 2.3.3 Frequency-dispersive Behavior of the Channel Page 39 2.3.5 Frequency dispersion vs Time dispersion We have seen that fading can be in the frequency domain or in the time domain. It is possible to treat the fading in wireless communications systems as constituted by independent effects. The channel shows time dispersive behavior when multipath phenomena are present. At the same time, independent of this effect, the channel will also exhibit frequency dispersion if the mobile unit is moving. Page 39 Signal bandwidth Figure 2.31 Time dispersive/ frequency selective Slow and Frequency Selective Fast and Frequency Selective Bc Time selective/ frequency dispersive Slow and Flat Fast and Flat Tc Bit duration Page 40 Top of the page Page 41 The Rician probability density function is shown in Figure 2.33 for different values of the parameter K. The transition from Rayleigh to Gaussian (or Rician to Gaussian) can be seen clearly in Figure 2.33 as the value of K continues to increase. Page 41 Figure 2.33 0.7 0.6 K= - dB 0.5 K=6 dB K=14 dB K=18 dB A f (a) 0.4 0.3 0.2 0.1 0 0 2 4 6 envelope a 8 10 12 Page 42 Eqn. 2. 66 f pdB p p exp dB2 2 av dB 2 1 2 2 dB Page 46 2.4.5 Summary of Fading The various fading mechanisms and the attenuation described can be summarized in a diagram shown in Fig. 2.38. Note that Rician and Rayleigh arise out of multipath effects and Nakagami can represent them both. This is not shown in the figure. For most of the cases, analyses based on Rayleigh or Rician fading are sufficient to understand the nature of the mobile channel. A number of recent publications (Alhu 1985, Anna 1998, 1999) have suggested the use of Nakagami fading models to provide a generalized view of fading in wireless systems. Page 53 Exercise 2.1 Using MATLAB, generate plots similar to the ones shown in Figure 2.7 to demonstrate the path loss as a function of the loss parameter for distances ranging from 2 Km to 40 Km. Calculate the excess loss (for values of n >2.0) in dB. Page 54 Exercise 2.18. Generate a plot similar to the lognormal fading shown in Figure 2.36. Page 55 Exercise 2.17 Compare the maximum data transmission capabilities of the two channels characterized by the impulse responses shown below. dBm 0 dBm 0 -10 -10 -20 -20 -30 -30 60 85 (a) 110 ns (delay) 0 4 8 (b) microsec (delay) FIGURE P2.17 Page 62 Line below equation (3-21) where W ( = 2fm) is the maximum frequency of the modulating signal and Df is the peak frequency deviation. Page 75 P(f) Figure 3.18 f0 frequency f0+R |M(f)I2 f 0- R 0 R 2R frequency Page 79 (below eqn. 3.45) The output noise, n(T), has a spectral density, Gout(f), given by (See Appendix A.6) Page 80 Eqn. (3.51) Use Min in place of Mout H f KMin f exp j 2 fT Page 84 (3.51) Equation (3-70) s1 t 2E T cos 2 f 0t E 2 T cos 2 f 0t E 1 t s0 t 2E T sin 2 f 0t E 2 T sin 2 f 0t E 0 t (3-70) Page 87 The probability of error can be expressed (Taub 1986) as (Exercise 3.21) p e 12 erfc Page 90 E 2 N0 pBPSK e 12 erfc (3.79) E N0 (3.83) The plot of error probability for different values of E/N0 (signal-to-noise ratio) is shown in Figure 3.28. Comparing eqn. (3.83) with the bit error rate for ASK systems (eqn. (3.79)), it is seen that the performance of a coherent BPSK is 3dB better than that of coherent ASK. Page 93 Table 3.2 DPSK encoding: dk bk dk 1 Page 92 pav e p BPSK e | Df f Df df 1 2 2 f2 erfc E N0 Page 94 In Figure 3.32, use xk and yk in place of Xk and Yk. Page 98 Data fn 11 /4 -1 1 3/4 -1 -1 3/4 or 5/4 /4 or 7/4 1 -1 Table 3.4 cos Df exp 2Df 2 d Df Phase encoding in QPSK 2 f (3-86) Page 100 Figure 3.35 In Figure 3.35a, the line at 2 should be broken as shown. QPSK 1 (a) 0.5 0 -0.5 -1 0 1 2 3 4 5 6 7 8 QPSK (shifted by /4) amplitude 1 (b) 0.5 0 -0.5 -1 0 1 2 3 4 5 time (units of T) 6 7 8 Page 103 Figure 3.41 In the Figure the line at 1 should be broken as shown. 1 0.8 0.6 amplitude 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 0 1 2 3 4 time(units of T) 5 6 7 Page 104 Figure 3.43 {-1 –1} Page 116 Figure 3.57 cos[2(f0+Df)t T dt 0 + S FSK _ z decision (threshold = 0) 1 if z > 0 0 if z < 0 T dt 0 cos[2(f0-Df)t Page 120 The relationship between and a is not given (equations 3-312 and 3-133). a 2 Page 124 Detection and Reception of MSK and GMSK Since MSK can be generated starting from OQPSK, the bit error performance of MSK will be identical to that of BPSK, QPSK or OQPSK. However, MSK and GMSK can also be detected using a 1-bit differential detector, 2-bit differential detector or a frequency discriminator. The block diagrams of these receiver structures are shown respectively in Figure 3.69 a, 3.69 b, and 3.69 c. Page 120 c t qg t 1 4T 1 4T erf erf nT t 2 n1T t 2 (3-133) n 1 T t T n 1 T t erf 2 nT t erf nT2t (3-134) n1T t nT t 2 2 2 2 2 e e 2 4T22 Page 124 Detection and Reception of MSK and GMSK Since MSK can be generated starting from OQPSK, the bit error performance of MSK will be identical to that of BPSK, QPSK or OQPSK. However, MSK and GMSK can also be detected using a 1-bit differential detector, 2-bit differential detector or a frequency discriminator. The block diagrams of these receiver structures are shown respectively in Figure 3.69 a, 3.69 b, and 3.69 c. Page 125 Ts log2 M T (3.139) As the number of levels of modulation increases, the minimum power required to maintain a fixed bit error rate decreases increases. This can be seen from the decision boundaries shown in Figure 3.71. Page 129 C B log 2 1 R B E N0 log 2 1 R B bits / s E N0 R B (3.145) (3.146) Page 141 Figure 4.5 Frequency reuse pattern of cells. The alphabets (except D) represent the different channels. D is the distance between cells having the same frequency. D DR 3R 1 4 2 3R 21R Page 143 Because of this, q =D/R= 3N c is also known as the frequency reuse factor Page 143 TABLE 4.2 i Page 143 j Nc q S/I(dB) Figure 4.6 Expanded view of the cell structure showing a seven cell reuse pattern. H is the channel reused. Page 144 Figure 4.7 A three-cell pattern (Nc=3) showing six interfering cells D q 3N c 3 R Page 155 Example 4.4 In order to keep an acceptable level of performance, a service provider is planning to maintain an outage probability of 2%. What should be the power margin needed to achieve this goal (assume the worst case scenario), when the standard deviation of fading has been found to be 5 dB. Answer: The outage probability is given by eqn. (4.23) which can be rewritten as and the power margin in dB is given by Pout R 0.5erfc loge ( m ) 2 and the power margin in dB is given by M 10 log10 m The relation ship between and dB is given in eqn. (2.68) as . We, therefore, have = 1.15. Outage probability 0.02 0.5erfc loge ( m ) 1.6282 1 ln 10 10 dB As we have done in Chapter 3, we can solve this using the MATLAB function erfinv. We get m = 10.6 and M to be 10.3 dB. Page 169 p exp , 0 0 0 1 (5.5) Page 175 5.2.2 Effects of Frequency Selective Fading, Co-Channel Interference As discussed in Chapter 2, frequency selective fading arises when the coherent bandwidth of the channel is less than the message bandwidth. The error rates vary with the form of modulation and demodulation used. The performances of the modems also depend on the ratio of (d/T), where d is the r.m.s delay spread (eqn. 2.42) and T is the symbol period. The exact equations governing the error probability, taking frequency selective fading into account, are once again very complex, and are beyond the scope of this book. Numerical results are available in a number of research papers (Fung 1986, Guo 1990, Liu 1991a,b). Page 176 Page 189 Figure 5.5 f The legends should read dB instead of db. M 1 2M 1 M M 2 M 1! 0M (5.34) The performances of the three signal processing schemes are shown in Figure 5.17, with av equal to the signal-to-noise ratio after combining. They are also given in tabular form in Table 5.1. Page 195 H c f A f exp j f Eqn. 5.47) Page 213 Figure 6.8 two bits a T chip sequence b K Tc Tc modulo 2 addition of chip and data sequences c Page 214 First Paragraph The PN sequence is unique to each user and is almost orthogonal to the sequences of other users. Thus, the interference from other users in the same band will be much less. The number of orthogonal codes, however, is limited. As the number of users increase, the codes become less and less orthogonal more and more correlated (orthogonality is compromised) and the interference will increase. Page 220 Example 6.1 In a DS-CDMA cell, there are 24 equal power channels that share a common frequency band. The signal is being transmitted on a BPSK format. The data rate is 9600 bps. A coherent receiver is used for recovering the data. Assuming the receiver noise to be negligible, calculate the chip rate to maintain a bit error rate of 1e-3. Answer: Assuming that there is no thermal noise, the bit error rate is given by eqn. (6.15) where K is the processing gain and k is the number of channels. BER 10 3 0.5erfc z where z kK1 Using the MATLAB function erfinv (.), we can solve for z erfinv1 2* BER 2 chip rate We get z = 4.77. We are given k=24; K=23*4.77=109.82. Since K data rate chip rate = 109.82*9600=1.05 Mchip/s. Page 222 Figure 6.18 Instead of BPSK, it should be BFSK Page 223 Figure 6.20 Instead of FH/BPSK signal, it should be FH/BFSK Page 225 Figure 6.21 Block diagram of the transmitter associated with an AMPS system Change Page 231 to Eqns. 6.21 and 6.22 The radio capacity of the system, m, is defined as m channel bandwidth Bc * numberoft cells Nc total allocated spectrum B radio channels/cell. (6.21) Making use of the relationship between q and Nc (Chapter 4), q 3Nc, (6.22) Page 233 3rd line from the top Let us start by considering a single cell with N (= k of eqn (6.11)) users who share the cell. Page 233 E N0 S R N 1 S Bc Bc R 1 N 1 (6.27) where R is the information bit rate and Bc is the RF bandwidth. We have divided the signal power (S) in the numerator by the bandwidth (data rate R) of the message data, while the signal power in the numerator has been divided by the bandwidth (B c) occupied by the interfering signal. (See Section 3.2.7) The quantity (Bc /R) is the processing gain K of the CDMA processing, defined in connection with Figure 6.7. 0 Bc N0 (6.29) Page 233 E N0 min Page 234 Nmax K N max 1 NE0 s NE0 min 1 K E E N0 N0 min s Bc N0 1 S (6.33) Page 241 Table 6.7 Comparison of Standards Under the column for IS 95, change 1.288 kchips/s to 1.288Mchips/s Page 250 g t a n n exp jn2 f 0t ( A.1.7) Page 263 Normal or Gaussian distribution f x 1 2 x2 x exp 2 2 x 2 , x Page 266 where G is the gamma function G z wz 1e wdz 0