Performance analysis of BPSK and DPSK systems in the presence

of Nakagami-m Fading and Noisy Phase Reference
C. M. Lo and W. H. Lam

Department of Electrical and Electronic Engineering
The University of Hong Kong, Pokfulam Road, Hong Kong

cmlo@ieee.org

Abstract : This paper presents closed form expressions to evaluate the average bit error rate (BER) for
coherent binary phase shift keying (BPSK) and differential PSK (DPSK) modulations over slow
Nakagami-m fading channel with noisy phase reference. The performance degradation due to the
noisy phase reference are then investigated with respect to the channel fading parameter m and the
phase error . The phase error is assumed to be uniformly distributed with maximum phase error .
At the average BER of 10-3, results show that the error performance of the BPSK with  > 440 is worse
than the DPSK with =0 for m=9, i.e., the advantage of employing the BPSK over the DPSK is
vanished.

Keywords: Nakagami-m fading, Noisy phase reference, Phase shift keying

Introduction fading channels with noisy phase reference has
been found.
For coherent communication systems, error
performance are usually evaluated by assuming This paper thus derives closed form expressions
that a perfect phase reference is available in the of the average BER for the BPSK and the DPSK
receiver for demodulation [1,2]. In practice, this systems in the presence of Nakagami-m fading
local phase reference is however reconstructed and noisy phase reference. Also, the analysis
from a noise-corrupted version of a received presented in this paper may be used to
signal, and thus a phase error, , is usually characterize the performance of an acquisition
resulted. The immediate effect of the phase error scheme for many wireless systems. In Section II,
is degradation of detection performance of the the derivation of the average BER for the BPSK
coherent systems. are presented and the corresponding derivation
for the DPSK are given in Section III. Numerical
Over the years, many researchers have results and discussions are given in Section IV.
investigated the error performance of binary
phase shift keying (BPSK) and differential PSK Derivation of the average BER for the BPSK
(DPSK) systems over an additive white gaussian
noise (AWGN) channel in the presence of noisy The average BER for the BPSK system in the
phase reference [2-8]. For BPSK systems, [9-13] presence of Nakagami-m fading and noisy phase
have evaluated the bit error rate (BER) in the reference is considered in this section. For fading
presence of Rayleigh fading and noisy phase channels, the conditional BER for the BPSK with
reference. The corresponding error performance phase error is given by [2]
in Rician and Lognormal fading channels have 1
Pe   ,    erfc (  cos  ) (1)
also been investigated in [9] and [12]. Amongst 2
the various fading channels, Nakagami-m fading where erfc(.) is the complementary error function
distribution is considered to be the most versatile, [14] and  is the instantaneous signal to noise
since it is better in modelling mobile channels ratio (SNR) per bit of the received signal. The
than the Rayleigh, Rician and the Lognormal phase error  is assumed to be uniformly
fading distributions. To the best of the authors’ distributed in a range of   ,   as in [13] and the
knowledge, no error performance of the BPSK in probability density function (pdf) of it is given by
the presence of Nakagami-m fading and noisy p    1 / 2 . (2)
phase reference has been reported. In addition,
no performance analysis for the DPSK over
In addition, the pdf of  for the Nakagami-m

32

c. (5) m fading and phase noise can be written as 1  1m m  m 1   m    0 0 Pe    exp  2  ( m)    0 By substituting (1) and (3) into (5) and using a     series representation for the erfc(. The average BER can then be Derivation of the average BER for the DPSK obtained using (1) to (3) as [13]   The case of the DPSK is now considered.  1 1 ( 1) sin(( 2n  1) ) ( m  n  0.5)( n 20 . n  1 / 2. (4) conditional BER for the DPSK is also an even function of  and given by [2] 1 Pe   .    exp( cos 2 ( )) . (11) Since the conditional BER in (1) is an even 2 function of .  (1  k ) 2 2 41 / 2  m /   2  where 2 F1 (a. d can be evaluated  be written as as in [16] and the (14) can then (8) 1m 1 m m m  m m 1 1   n Pe       F1  . b. the average BER in (4) can also be written as By substituting (3) and (11) into (5). 1) values from 0. the average   BER for the DPSK in the presence of Nakagami- 1 Pe     P   . m  n  1 / 2.   2 ( 1) n (n  1 / 2)  0 n 1 / 2 erfc (  cos  )  1    (2n  1)! (12) n 0 By using the following identity [5]  1  1 F1  n  .2n  2.2n  2.1. 2 F1 ( n  1 / 2.2n  2. The two integrals   1    0 exp m /   1 F1  n   2 . n  . / m) 2 where m is the fading severity parameter with  .  p  p  d d 0 e . m.2n  2.) as [4]  exp  2 (1  cos 2  ( )) d d .  p  dd e . Finally.5 to  and  is the average SNR per bit. x) is the confluent hypergeometric    1 m      m 1 function [14]. . The notation (m) denotes the gamma (10) function [14].5)  2   2 2 41 / 2  m /   2    2   n 0 ( 2n  1)! ( m)(2n  1)  / m   n 1 / 2 m   2 k mk  1 m 1 sin(2k )   1   1 m       ( m)  k 1 2k      4  2    1 1   2 F1  m  n  .  (9) ( m  k )  m  k m  k 1 1   2 2 m   2 F1  .2n  2. .  cos( 2n  1) d p    d   2 0 m     1 m 1 sin(2k ) (7)  I 0    d     2      ( m)  k 1 2k where 1 F1 (a. the average BER in (9) (15) can be written in closed form as After further manipulation.  cos( 2n  1)          2  exp  cos(2 )   I 0     2 I k    cos(2k )  2   2  2  (6) k 1 the average BER in (5) becomes (13) 1 1   ( 1) n ( n  1 / 2) the average BER in (12) becomes  n 1 / 2 Pe    m   2  ( 2n  1)! 1m 1  1 m   m 1 0 n 0 Pe    exp        2 ( m)   2   0  1  1 F1  n  . The Pe    P  . The two integrals in (7) can be   exp       I k   d evaluated as 0   2    2  1 1  (14) (1) n ( n  1 / 2) sin( 2n  1)  m  m Pe   2   n 0 ( 2n  1)! (m) 2n  1   where  I v (a) is the modified Bessel function of  m  n 1 / 2 the first kind of order v [14].fading channel is given by [15] 1 1  ( 1) n sin( 2 n  1) m  m m 1   m  Pe  2    ( 2 n  1) 2 p      exp  (3) n 0     ( m)    F1 (m  n  1 / 2. k  1. x) is the Gauss hypergeometric   function [14]. b. the average BER can be simplified as 33 .

Therefore. J. 1976. When  2366-2369. "BPSK system with The advantage of using the DPSK over the BPSK sounder to combat Rayleigh fading in mobile is that the DPSK has a lower probability of radio communication. more degradation [8] P. 42. "Approximate results for the bit error For the BPSK system over channel having m probability of binary phase shift keying with increasing from 1 to 9. Viterbi. 1966. The performance degradation for the BPSK and Englewood Cliffs. 364-365.1  m  k m  k 1  References 1  2 F1  . equal to about 0. the average SNR at the average BER of 10 -3 Aerosp. pp. In addition. k . Australia. Y. are 2." IEEE Trans. Bateman. 1973. V. 19. [6] P. k  1. Zhang. for all values of m. Y. T. the Technol. no." IEEE Trans. 2. Mar. Kam. "Performance of phase- locked loops in the presence of fading varying. . Comm. "Phase-shift-keyed signal channels with noisy phase reference are detection with noisy reference signals. BPSK with the channel fading parameter m 393-401. respectively. Fig. the DPSK systems over Nakagami-m fading [3] W. Jain and N. 1. vol. vol. the degradation with =600 communication channels. May 1976. Tjhung and P. 1973. 00. Seek." IEEE presented. respectively. 1020-1022. vol.. T. pp..  2 2 41 / 2  m /   2    [1] A. From the last [13] Y. vol. 34 .5. BPSK has superior performance over the DPSK [12] A. 34. respectively.69dB and 3.38dB for m=1 and 9. S. "Data degradation are approximately 1dB and 3dB for  transmission over UHF fading mobile radio = 400 and 600. pp." IEEE Trans. vol. respectively. 3. no. pp. K. BPSK and the DPSK with m and  ranging from [5] V. m  k . P.73dB and 3." IEEE Trans.. Fig. Veh. the [10] A.1.. 4. Feb. it shows that the performance of "Error probabilities for coherent BPSK and the BPSK is superior than the DPSK by 1. no. pp. 1 depicts the average BER for the Trans.08dB as  is error on DPSK error probability. C. McGeehan. "Detection of a PSK signal transmitted through a 2. C. K. On the other hand. Blachman. Bateman and J. Tjhung. for example. 2 F1  m  1. K. no. vol. M. row of Table 1. Furthermore. Principles of Coherent (16) Communication. Comm. Electron. Hence. ranging from 1 to 9 and the maximum phase error [4] P. 12. C. "A general analysis of bit when there is no phase error. July 1966. 4. 4. Telecommunication Systems Engineering. Kam.. 1   m  m m 1 1 therefore  equal when   520 and 440 for m=1 and P e  1    2 F1  . The with random phase noise. is large. Dubey and J. row of Table 1. . Mar. 1981. K. 7. no. is therefore resulted for higher values of m. 374. 7. Y. From Table 1. M. no. K. 5.87dB QPSK in a slowly flat fading Rayleigh channel and 1. Table 1 shows the performance hard-limited channel. J. Lindsey. F. "Error probability of 2DPSK with phase low values of . Prabhu. J." IEEE Trans. pp. 623-630. vol. 1985. 29. we are error probability for reference-based BPSK interested in the value of  that makes the error mobile data transmission. performance of the BPSK with phase error equal vol. it shows that the [7] N. pp. Syst. T. Comm. no. vol. the degradation are different as m is [9] W. the advantage of using the 2 2m  2 2 41 / 2  m /   2   k m  mk BPSK is vanished with  > 44 for the fading 0 sin( 2k )   1   m   1 m    k 1 2k        4   2   channel with m=9.  9. degradation due to =300 is increased as shown in 41." IEE Proc. 398-402. 1989." IEEE Trans. "The effect of phase degradation are 2. For  = imperfect carrier phase recovery. 2 and Table 1. 275- for different values of m are presented in the last 285. pp. the performance noisy phase reference. Weber. no. The case of the DPSK is shown in Fig.6dB for all values of m. 131.. Yokoyama.14dB for m=3 and 9. . III. As  is increased. no. 1. Sept. 35-40." IEEE Trans. pp. Simon. Comm. to the DPSK with no phase error. S. Inform. respectively. 24. K. degradation at the average BER of 10-3 for the Theory. having a phase error. increased from 100 to 600 for m=1 and 9. the degradation are similar and noise. vol. Fu. July 1993. vol. July 1994. no. [11] M. NJ: Prentice-Hall. Lindsey and M. Aerosp. Blachman. Electron. Y. 37. Numerical results and discussions [2] W. Some and T. 487-499." IEEE Trans.. For Sinha. New York: McGraw-Hill." Journal of Electrical error performance of both modulations are and Electronics Engineering. 19. July 1984. Syst. 2.. Apr. Similar conclusions can be drawn channels. pp. Teo.  = 300. "PSK performance with 3 to 9 and from 100 to 600. 364- for the DPSK as shown in Fig. Comm.

Brychkov and O. Ed.06 0.14 2. Prudnikov. Hoffman.92 0.20 0. 1972.20 0. Abramowitz and I.48 0.06 0.46 0.90 0.95 0. A. 10 0.19 0.20 3.00 9.98 50 1.93 0.96 0.19 0. W.44 0.47 0.44 0. I.20 30 0.04 0. 1960." Gordon and Breach Science.05 0.60 1.08 3. Y. Integrals and Series.68 1. Statistical Methods of Radio Wave [14] M.77 1.18 10. 1999.78 3. pp. [16] A. respectively.79 1.31 9. 1986.59 1. 3.96 3.90 10.82 60 2.95 0.12 13. New York: England: Pergamon Press.07 3.04 0. Table 1.no. The performance degradation of the BPSK and the DPSK at the average BER of 10-3 for m and  ranging from 3 to 9 and 100 to 600. Oxford. P. C.53 8.89 0. 123-128. "The m-distribution-A general Marichev. The average SNR with =00 are also given for m increasing from 3 to 9.18 0.05 0.50 35 . Dover. Sept. Stegun.50 40 0.72 1.75 1.19 0. Handbook of Mathematical Functions.19 0. Nakagami.06 0.05 20 0.31 Average SNR at 10 -3 (dB)  = 0 11.32 8.46 0. Propagation. Amsterdam: formula of intensity distribution of rapid fading. [15] M. Average BER at Degradation due to  (dB) -3 10 BPSK DPSK maximum phase m =3 m =5 m =7 m =9 m =3 m =5 m =7 m =9 error  deg. A. in.73 2.48 0.

36 . 9.1 The average BER performance of the BPSK with =00.5.9. Fig. 300 and m=1. 2.3.3. 100 Nakagami-m fading 10-1 No phase error m=1 Average Bit Error Rate Phase error  = 300 10-2 10-3 m=3 m=5 10-4 m=7 AWGN with no phase error m=9 10-5 10-6 0 3 6 9 12 15 18 21 24 27 30 Average SNR per bit  (dB) 100 Nakagami-m fading 10-1 m=1 No phase error Average Bit Error Rate Phase error  = 300 10-2 10-3 m=3 AWGN with no m=5 10-4 phase error m=7 m=9 10-5 10-6 0 3 6 9 12 15 18 21 24 27 30 Average SNR per bit  (dB) Fig.5. 300 and m=1.7.7. The average BER performance of the DPSK with =00.