Académique Documents
Professionnel Documents
Culture Documents
Mohamed-Slim Alouini
Dept. of Electrical & Computer Eng. University of Minnesota 4-174 EE/CSci Building 200 Union Street S. E. Minneapolis, MN 55455, USA. E-mail : <alouini@ece.umn.edu>
Marvin K. Simon
Jet Propulsion Laboratory California Institute of Technology Mail Stop: 238-343 400 Oak Grove Drive Pasadena, CA 91109-8099, USA. E-mail: <marvin.k.simon@jpl.nasa.gov>
ABSTRACT
We obtain generic expressions for the average error rate of multichannel reception of digital signals over correlated slowly varying Nakagami-m fading channels. The results are applicable to coherent detection with maximal-ratio combining as well as to di erentially coherent and noncoherent detection with post-detection equal-gain combining. Aside from extending previous analyses of diversity reception over correlated Nakagami-m fading channels to include the performance of M -ary modulations, these results provide equivalent forms for known expressions corresponding to the performance of binary modulations. Because of their simple forms, these results o er a useful analytical tool for the accurate performance evaluation of various systems of practical interest.
1. INTRODUCTION
In studying the performance of diversity systems, the usual assumption made is that the combined signals are independent of one another 1, 2]. As discussed in 3, 4, 5, 6, 7, 8] there are a number of real life scenarios in which this assumption is not valid because, for example, of insu cient antenna spacing in small-size mobile units equipped with space antenna diversity. As a result, the maximum theoretical diversity gain cannot be achieved and hence one must revamp their analysis to account for the e ect of correlation between the combined signals. Along these lines, several correlation models have been proposed 3, 6, 9], and using these models several authors 3, 4, 5, 6, 7, 8] have analyzed special cases of the performance of various systems corresponding to speci c detection, modulation, and diversity combining schemes. For instance, Pierce and Stein 3] considered the performance of binary coherent and noncoherent systems over correlated identically distributed Rayleigh fading channels. In particular they obtained the average bit error rate (BER) of coherent binary phase-shift-keying (BPSK) when used in conjunction with maximal-ratio combining (MRC) 10], 11, Section 5.5.3, p. 24], and of noncoherent frequency-shiftkeying (BFSK) when used with postdetection equal-gain combining (EGC) 11, Section 5.5.6, p. 253], 12, Section 4.4, p. 298]. Miyagaki al. 4] analyzed the outage probability and the average symbol error rate (SER) performance of M -ary phase-shift-keying (M -PSK) for various dual-branch diversity receivers over correlated identically distributed Nakagami-m fading channels 9]. Al-Hussaini and Al-Bassiouni 5]
obtained a closed form expression for the average BER of noncoherent BFSK with dual-branch MRC reception over correlated non identically distributed Nakagami-m fading channels. More recently Aalo 6] analyzed the outage probability and the average BER of various coherent, di erentially coherent, and noncoherent binary modulations with multichannel MRC reception over identically distributed Nakagamim fading channels with two correlation models, namely the constant (equal) correlation model and the exponential correlation model. Finally, Patenaude et al. 7, 8] extended the results of Al-Hussaini and Al-Bassiouni and of by providing closed form expressions for the average BER of orthogonal noncoherent BFSK with postdetection EGC reception over two correlated nonidentical and also D equicorrelated identically distributed Nakagami-m channels. In this paper we obtain general results for the exact average BER or SER of M -ary coherent, di erentiallycoherent, and noncoherent modulations over equicorrelated, exponentially, and arbitrary correlated slowly varying Nakagami-m fading channels. Aside from allowing for many detection, modulation, and diversity combining cases not previously treated, these general results provide often simpler forms for the average BER or SER expressions corresponding to the special cases treated by the cited authors. In particular, each average BER or SER expression is obtained in the form of a single integral with nite limits and an integrand composed of elementary functions. The remainder of this paper is organized as follows. In the next section, the various channel correlation models under consideration are described. Section 3 provides in a generic fashion the performance of coherent modulations with MRC reception. Section 4 extends these performance results to di erentially coherent and noncoherent modulations with postdetection EGC reception. Due to space limitations, numerical examples for the error rates of particular modulations of interest over various correlated fading channels are omitted here but will be presented in an extended version of this paper 13] which in addition will cover some applications of the framework presented in this paper to the performance evaluation of hybrid diversity systems (such as space-multipath or frequency-multipath systems) over correlated fading channels.
pi ( t ) es t d t :
(1)
In this case the PDF of the combined signal envelope, pa (rt ), is given by 9, Eq. (142), p.34]
pa(rt ) = (m)(
2rt
1 2 (1
))m
rt2 m
1 2
2 Im 1 ( rt2 ) e rt ; rt 0; 2
(2)
where I ( ) denotes the th-order modi ed Bessel function 14, Section10.2, p. 443],
1 ; r2 ) = p cov(r 2 )var(r2 ) ; 0 var(r1 2 2 2
< 1:
(3)
is the envelope correlation coe cient between the two signals 1 , and the parameters d (d=1, 2), , and are de ned as follows: d (4) d = m ; (d = 1; 2);
2) ; = 2( 1+ (5) ) 1 2 (1 2 )2 + 4 1 2 : 2=( 1 (6) 2 2 (1 4 1 )2 2 By using a standard transformation of random variables, it can be shown that the PDF of the combined SNR per symbol, pa ( t ), is given by
pa ( t ) = (m)
where the parameters 0 and 0 are normalized version of the parameters and , and are given by m( 1 + 2 ) ; 4 0= = (8) Es =N0 2 1 2 (1 )
2 (1
m2
2 0
1 2
Im 1 ( 0 t) e 2
t;
0;
(7)
) 1 =2
: mp t (1 ) ; t 0;
(9)
pa ( t ) = (m)p1
1 m m+ 2
t p
r
1 2
exp
(1
mt
Im
1 2
(10)
For the Rayleigh fading case (m = 1) using the identity 14, Eq. (10.2.13), p.443] sinh z 1 2z I 2 (z ) = z ; (11)
1 We use the envelope correlation coe cient throughout this paper as a measure of the degree of correlation between the fading signals since, as pointed it out in 3, 6], experimental data on the correlation between fading signals are typically given in terms of this gure because of its relative ease of measurement.
where sinh( ) denotes the sin hyperbolic function 14, Section 4.5, p.83], it can be shown that (7) reduces to ; t 0; (12) ( 1 + 2 )2 4 1 2 (1 ) which itself reduces to the well-known expression for the case of identical Rayleigh channels originally derived in 15] and which can also be found in 3, Eq. (40)] 1 exp t t pa ( t ) = 2p exp (1 p (13) (1 + p ) ) ; t 0:
pa ( t ) =
exp
p(
+ )2 4 2 (1 )
1 2 1 2
1 2
(1 )
exp
+ 2 + ( 1 + 2 )2 4 2 1 2 (1 )
1 2
(1 )
2.1.2. MGF
Substituting (7) in (1) then using the Laplace transform 16, p. 1182, Eq. (110)], it can be shown after some manipulations that the MGF of pa ( t ) is given by
4 M (s) = 1 ( 1 + 2 ) s + (1 ) 1 2 s2 Ma(s; 1 ; 2 ; m; ) = a m m2
m
; s 0:
(14)
2.2.1. PDF
Based on the work of Gurland 17], Aalo showed the the PDF of t is given in this case by 6, Eq. (18)]2 (16) t 0; (1 m where 1 F1 ( ; ; ) is the con uent hypergeometric function 14, Chapter 13, p. 503]. For D = 2 using 14, Eq. (13.6.3), p. 509], namely
1 F1 (m; 2m; 2z ) =
2
pb ( t ) =
m t Dm 1 exp
mt 1 F1 (1 p ) p )m(D 1) (1
m+ 1 2
1 z m+ 2 z 1 (z ) e ; Im 2 2
(17)
It should be noted at this point that in 6, Eq. (18)] the symbol is used to denote the correlation coe cient of the underlying Gaussian processes that produce the fading on the channels. This correlation coe cient is equal to the square root of the power correlation coe cient. Based on the work of 18, p. 62] it is shown in 3, Appendix V] that for all practical purposes the power correlation coe cient can be assumed to be equal to the envelope correlation coe cient which is denoted by throughout this paper so as to follow what seems to be the more conventional usage of this symbol.
as well as the identities 14, Eq. (6.1.12) and Eq. (6.1.8), p. 255] yielding 1 3 5 7 (2m 1) p m+ 1 (18) = 2 2m then it can be shown after some manipulations that (16) reduces to (10) of Model A, as expected.
2.2.2. MGF
together with the identity 14, Eq. (15.1.8), p. 556], 6, Eq. (A-5)] a 2 F1 (a; b; b; z ) = (1 z ) ; it can be shown that m (1 p + Dp ) 4 M (s) = 1 Mb (s; ; m; ; D) = s 1 b m
Substituting (16) in (1) then using the Laplace transform 16, Eq. (4), p. 864], Z 1 (b) F a; b; c; k ; jsj > jkj; b > 0; s > 0; s > k; xb 1 1 F1 (a; c; kx) e sx dx = s b 2 1 s
0
(19) (20)
(1 p )
m(D 1)
For D = 2, as a check, it can be easily shown that (21) agrees with (14) for
1 = 2.
; s 0: (21)
2.3.1. PDF
Based on the work of Kotz and Adams 19], Aalo showed the the PDF of t can be very well approximated by a gamma distribution given by 6, Eq. (19)]3
pc ( t ) =
where4
3
2 mD 1 r
exp
r mD
mD t r
mD2 r
mD2 r
0;
(23)
r = D + 12 p
D=2 D 11 p :
(24)
Based on the work of Kotz and Adams 19], Aalo 6] points out that the approximation (23) is valid for high values of but is still accurate for values of D as small as 5. 4 We remind the reader that contrary to its usage in 6] the coe cient denotes in this paper the envelope correlation coe cient.
D
2.3.2. MGF
Substituting (23) in (1) then using the Laplace transform 16, Eq. (3.381.4)], namely Z 1 x 1 e sx dx = ( ) ;
0
(25)
r mD s
2 mD r
; s 0:
(26)
Md(s; f
"
D X d=1
!#
= E 1; 2; ;
"
exp s
D X d=1
d m d
!#
(27)
Md(s) =
D Y d=1
1 smd
1
12
p
1
1 sm1
12
1 sm2 1
p p
1 smD 2 D 1 sm D
1D
1 3 m 1 7 7
7 7 7 7 7 7 7 7 7 5
;
D D
(28) where j M ]jD D denotes the determinant of the D D matrix M . For D = 2, as a check, it is straightforward to show that (28) reduces to (14) of model A. For the case of identical channels ( d = ), (28) reduces after some manipulations to
2
1D
1 sm1
2D
1 sm2
Md (s) =
s m
mD
6 6 6 6 6 6 6 6 6 4
m 1 s p
12
m 1 s
12
p p
1D 7 2D 7 7
7 7 7 7 7 7 5
:
D D
(29)
1D
2D
m 1 s
Pi (E ) =
We now present speci c results for di erent coherent modulations with MRC reception.
P (E j t ) pi ( t ) d t :
(30)
d ; (31) sin2 1 where g = 1 for BPSK, g = 1=2 for orthogonal BFSK, and g = 1 2 + 3 for BFSK with minimum correlation. The average BER, Pb (E ) is then obtained by averaging (31) over (7), (16), or (23) yielding after interchanging the order integration the desired generic result
0
gt
g d : (32) 2 sin 0 For model B, (32) with (21) should be contrasted with the previous available equivalent result 6, Eq. (32)] which required the evaluation of the Appell's hypergeometric function, F2 ( ; ; ; ; ; ; ), which is typically not available in standard mathematical software libraries such us Mathematica, Matlab, and Maple, and which is de ned either in term of an in nite range integral of a special function 6, Eq. (A-12)] or alternatively as a double in nite sum 6, Eq. (A-13)]. For model C, (32) along with (26) is equivalent to 6, Eq. (40)] which is expressed in terms of the hypergeometric function 2 F1 ( ; ; ; ) 14, Chapter 15, p. 555].
Z =2 Pbi (E ) = 1 Mi
gpsk d : (34) 2 sin 0 It should be noted that for model A (34) together with (14) is an exact expression equivalent to 4, Eq. (54)] with the advantage of being expressed in terms of a single nite-range integral (contrary to 4, Eq. (54)]) and whose integrand is much simpler than 4, Eq. (54)], and hence easier to compute for any arbitrary value of the fading parameter m.
Z (M 1 Psi (E ) =
1) =M
Mi
For noncoherent detection of equal energy, equiprobable, correlated binary signals, = 1 and
l l+2 l 1 l cos(l( + =2)) : l l+1 l 1 l+1 cos((l 1)( + =2)) L l l=1 In (37) the parameters a and b are modulation-dependent and are de ned in 12, Eq. (4B.22)], and = v2 =v1 , with the parameters v1 , v2 de ned in A number of special cases are of particular importance.
f1 (L; ; ; ) =
L X
2L 1
l=1
L l
a=
1 j j2 2
!1=2
j j2 b= 1+ 1 2
! 1 =2
(38)
where (0 j j 1) is the complex-valued cross-correlation coe cient between the two signals. The special case = 0 corresponds to orthogonal noncoherent BFSK for which a = 0 pand b = 1. Furthermore, in the case di p erential phase-shift-keying (BDPSK), a = 0, b = 2, and = 1. Finally p p pof binary a = 2 2, b = 2 + 2, and = 1 correspond to quaternary di erential phase-shift-keying (DQPSK) with Gray coding. It should be noted that as a ! 0 (37) assumes an indeterminate form but the limit converges smoothly to the exact conditional BER expression.
Pbi (E ) = 2 (1 + )2L
b2 L
Letting a ! 0 and b = 1 (b = 2) in (39) yields the average BER performance of orthogonal BFSK (BDPSK), and in this particular case (39) is equivalent to the closed-forms 7, Eq. (21)] and 7, Eq. (16)], for models A and B, respectively.
f L; a b; ; 2 2 (a + b + 2ab sin ) Mi
a2 + b2 + 2ab sin
2
d :
(39)
5.
REFERENCES
1] M. -S. Alouini and A. Goldsmith, \A uni ed approach for calculating the error rates of linearly modulated signals over generalized fading channels," in Proc. IEEE Int. Conf. on Commun. ICC'98, Atlanta, GA, pp. 459{464, June 1998. 2] M. K. Simon and M. -S. Alouini, \A uni ed approach for the probability of error for noncoherent and di erentially coherent modulations over generalized fading channels." To appear in IEEE Trans. Commun. 3] J. N. Pierce and S. Stein, \Multiple diversity with nonindependent fading," Proc. IRE, vol. 48, pp. 89{104, January 1960. 4] Y. Miyagaki, N. Morinaga, and T. Namekawa, \Error probability characteristics for CPSK signal through m-distributed fading channel," IEEE Trans. Commun., vol. COM-26, pp. 88{100, January 1978. 5] E. K. Al-Hussaini and A. M. Al-Bassiouni, \Performance of MRC diversity systems for the detection of signals with Nakagami fading," IEEE Trans. Commun., vol. COM-33, pp. 1315{1319, December 1985. 6] V. A. Aalo, \Performance of maximal-ratio diversity systems in a correlated Nakagami-fading environment," IEEE Trans. Commun., vol. COM-43, pp. 2360{2369, August 1995. 7] F. Patenaude, J. H. Lodge, and J. -Y. Chouinard, \Error probability expressions for non-coherent diversity in Nakagami fading channels," in Proc. IEEE Veh. Technol. Conf. VTC'97, Phoenix, AZ, pp. 1484{1487, May 1997.
8] F. Patenaude, J. H. Lodge, and J. -Y. Chouinard, \Noncoherent diversity reception over Nakagamifading channels," IEEE Trans. on Commun., vol. COM-46, pp. 985{991, August 1998. 9] M. Nakagami, \The m-distribution- A general formula of intensity distribution of rapid fading," in Statistical Methods in Radio Wave Propagation, pp. 3{36, Pergamon Press, Oxford, U. K., 1960. 10] D. Brennan, \Linear diversity combining techniques," Proc. IRE, vol. 47, pp. 1075{1102, June 1959. 11] G. L. Stuber, Principles of Mobile Communications. Norwell, MA: Kluwer Academic Publishers, 1996. 12] J. G. Proakis, Digital Communications. New York, NY: McGraw-Hill, second ed., 1989. 13] M. -S. Alouini and M. K. Simon, \Multiple diversity over correlated Nakagami-m fading channels." In preparation for submission to IEEE Transactions. 14] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables. New York, NY: Dover Publications, ninth ed., 1970. 15] K. S. Packard, \E ect of correlation on combiner diversity," Proc. IRE, vol. 46, pp. 362{363, January 1958. 16] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products. San Diego, CA: Academic Press, fth ed., 1994. 17] J. Gurland, \Distribution of the maximum of the arithmetic mean of correlated random variables," Annals of Math. Statist., vol. 26, pp. 294{300, 1955. 18] J. A. Lawson and G. E. Uhlenbeck, Threshold Signals. New York, NY: McGraw-Hill Book Co., 1952. 19] S. Kotz and J. Adams, \Distribution of sum of identically distributed exponentially correlated gamma variables," Annals of Math. Statist., vol. 35, pp. 227{283, June 1964. 20] A. S. Krishnamoorthy and M. Parthasarathy, \A multivariate gamma-type distribution," Annals Math Statist., vol. 22, pp. 549{557, 1951. 21] J. W. Craig, \A new, simple, and exact result for calculating the probability of error for twodimensional signal constellations," in Proc. IEEE Milit. Commun. Conf. MILCOM'91, McLean, VA, pp. 571{575, October 1991. 22] R. F. Pawula, S. O. Rice, and J. H. Roberts, \Distribution of the phase angle between two vectors perturbed by Gaussian noise," IEEE Trans. Commun., vol. COM-30, pp. 1828{1841, August 1982. 23] M. K. Simon and D. Divsalar, \Some new twists to problems involving the Gaussian probability integral," IEEE Trans. Commun., vol. COM-46, pp. 200{210, February 1998. 24] C. W. Helstrom, Elements of Signal Detection and Estimation. Englewood Cli s, NJ: PTR PrenticeHall, 1995. 25] M. K. Simon, \A new twist on the Marcum Q-function and its applications," IEEE Commun. Letters, vol. 2, pp. 39{41, February 1998.