Вы находитесь на странице: 1из 16

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO.

1, JANUARY 2006

271

[10] A. Martinez, A. Guilln i Fbregas, and G. Caire, New simple evaluation of the error probability of bit-interleaved coded modulation using the saddlepoint approximation, in Proc. 2004 Int. Symp. Information Theory and its Applications, Parma, Italy, Oct. 2004, pp. 14911496. [11] G. Poltyrev, Bounds on the decoding error probability of linear codes via their spectra, IEEE Trans. Inf. Theory, vol. 40, no. 4, pp. 12841292, Jul. 1994. [12] D. Divsalar, H. Jin, and R. J. McEliece, Coding theorems for turbolike codes, in Proc. 36th Allerton Conf. Communication, Control, and Computing, Allerton House, Monticello, IL, Sep. 1998, pp. 201210. [13] F. W. J. Olver, Asymptotics and Special Functions. New York: Academic, 1974.

On the Distribution of SINR for the MMSE MIMO Receiver and Performance Analysis
Ping Li, Debashis Paul, Ravi Narasimhan, Member, IEEE, and John Ciof, Fellow, IEEE
AbstractThis correspondence studies the statistical distribution of the signal-to-interference-plus-noise ratio (SINR) for the minimum mean-square error (MMSE) receiver in multiple-input multiple-output (MIMO) wireless communications. The channel model is assumed to be (transmit) correlated Rayleigh at-fading with unequal powers. The SINR can be decomposed into two independent random variables: SINR SINR , where SINR corresponds to the SINR for a zero-forcing (ZF) receiver and has an exact Gamma distribution. This correspondence focuses on characterizing the statistical properties of using the results from random matrix theory. First three asymptotic are derived for uncorrelated channels and channels with moments of equicorrelations. For general correlated channels, some limiting upper bounds for the rst three moments are also provided. For uncorrelated channels and correlated channels satisfying certain conditions, it is proved that converges to a Normal random variable. A Gamma distribution and a generalized Gamma distribution are proposed as approximations to the nite sample distribution of . Simulations suggest that these approximate distributions can be used to estimate accurately the probability of errors even for very small dimensions (e.g., two transmit antennas).

where xt 2 p is the (normalized) transmitted signal vector and yr 2 m is the received signal vector. Here p is the number of transmit antennas and m is the number of receive antennas. H W 2 m2p consists of independent and odentically distributed (i.i.d.) standard complex Normal entries. Rt 2 p2p is the transmitter correlation map2p ; c ~2 ; . . . ; c ~p ] 2 ~k = ck m trix. P = diag[~ c1 ; c p , where ck is the signal-to-noise ratio (SNR) for the k th spatial stream. This denition of SNR is consistent with [1, Sec. 7.4]. H = H W Rt P 2 m2p is treated as the channel matrix. nc 2 m is the complex noise vector and is assumed to have zero mean and identity covariance. Note that the power matrix P has terms involving the variance of the noise. The correlation matrix R t and power matrix P are assumed to be nonrandom. Also, we restrict our attention to p  m. We consider the popular linear minimum mean-square error (MMSE) receiver. Conditional on the channel matrix H , the signal-tointerference-plus-noise ratio (SINR) on the k th spatial stream can be expressed as (e.g., [1][6]) SINRk =
1

MMSEk

01=

1
Ip

where I p is a p2p identity matrix, and H y is the Hermitian transpose of


H

+ mH

yH 01

01
kk

(2)

Index TermsAsymptotic distributions, channel correlation, error probability, Gamma approximation, minimum mean square error (MMSE) receiver, multiple-input multiple-output (MIMO) system, random matrix, signal-to-interference-plus-noise ratio (SINR).

I. INTRODUCTION This study considers the following signal and channel model in a multiple-input multiple-output (MIMO) system:
yr

. Note that (2), in the same form as equation (7.49) of [1], is derived based on the second-order statistics of the input signals, not restricted to binary signals. For binary inputs, Verd [4, eq. (6.47)] provides the exact formula for computing the bit-error rate (BER) (also see [7]). Conditional on H , this BER formula requires computing 2p01 Q-functions. To compute BER unconditionally, we need to sample H enough times (e.g., 105 ) to get a reliable estimate. When p  32 (or p  64), the computations become intractable [4], [8]. Recently, study of the asymptotic properties of multiuser receivers (e.g., [2][4], [6], [8][11]) has received a lot of attention. Works that relate directly to the content of this correspondence include Tse and Hanly [11] and Verd and Shamai [6], who independently derived the asymptotic rst moment of SINR for uncorrelated channels. Tse and Zeitouni [3] proved the asymptotic Normality of SINR for the equal power case, and commented on the possibility of extending the result to the unequal powers scenario. Zhang et al. [12] proved the asymptotic Normality of the multiple-access interference (MAI), which is closely related to SINR. Guo et al. [8] proved the asymptotic Normality of the decision statistics for a variety of linear multiuser receivers. [8] considered a general power distribution and corresponding unconditional asymptotic behavior. Based on the asymptotic Normality results, Poor and Verd [2] (also in [4], [8]) proposed using the limiting BER (denoted by BER1 ) for binary modulations, which is a single Q-function BER1 = Q( E(SINRk )1 ) =

p1

H W Rt P

xt

+ nc =

p1

H xt

+ nc

(1)

0t

=2

dt

(3)

E(SINR )

Manuscript received January 20, 2005; revised September 9, 2005. P. Li is with the Department of Statistics, Stanford University, Stanford, CA 94305 USA (e-mail: pingli@stat.stanford.edu). D. Paul is with the Department of Statistics, University of California, Davis, Davis, CA 95616 USA (e-mail: debashis@wald.ucdavis.edu). R. Narasimhan is with the Department of Electrical Engineering, University of California, Santa Cruz, Santa Cruz, CA 95064 USA (e-mail: ravi@soe.ucsc.edu). J. Ciof is with the Department of Electrical Engineering, Stanford University, Stanford, CA 94305 USA (e-mail: ciof@stanford.edu). Communicated by R. R. Mller, Associate Editor for Communications. Digital Object Identier 10.1109/TIT.2005.860466

where E(SINRk )1 denotes the asymptotic rst moment of SINRk . Equation (3) is convenient and accurate for large dimensions. However, its accuracy for small dimensions is of some concern. For instance, [8] compared the asymptotic BER with simulation results, which showed that even with p = 64 there existed signicant discrepancies. In general, (3) will underestimate the true BER. For example, in our simulations, when m = 16; p = 8; SNR = 15 dB, the 1 of the exact BER. In asymptotic BER given by (3) is roughly 10000 current practice, code-division multiple-access (CDMA) channels with m; p between 32 and 64 are typical and in multiple-antenna systems arrays of 4 antennas are typical but arrays with 8 to 16 antennas would be feasible in the near future [9]. Therefore, it would be useful if

0018-9448/$20.00 2006 IEEE

272

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

one could compute error probabilities both efciently and accurately. Another motivation to have accurate and simple BER expressions comes from system optimization designs [13], [14]. This study addresses three aspects related to improving the accuracy in computing probability of errors. First, the known formulas for the asymptotic moments can be improved at small dimensions. Second, while the asymptotic BER (3) converges very slowly [4, p. 305], the accuracy can be improved by considering other asymptotically equivalent distributions using higher moments. Third, the presence of channel correlations may (seriously) affect the moments and BER. In fact, when the dimension increases, the effect of correlation tends to invalidate the independence assumption [15, p. 222]. To the best of our knowledge, the asymptotic SINR results in the existing literature do not take into account correlations. Mller [16] has commented that the presence of correlations makes the analysis very difcult. The channel model in (1) takes into account the (transmit) channel correlations. This model does not consider the receiver correlation and assumes a Gaussian channel (Rayleigh at-fading), while [3] and [8] did not assume Gaussianity. Removing the Normality assumption will still result in the same rst moment of SINR but different second moment (see [3]). Ignoring the channel correlation, however, may produce quite different results even for the rst moment. For example, it will be shown later that, with equal power and constant correlation  (equicorrelation model), the rst moment is only about (1 0 ) times the rst moment without correlation. Our work starts with a crucial observation: assuming a correlated channel model in (1), the SINR can be decomposed into two independent components
+ Tk = Sk + Tk SINRk = SINRZF k

Gamma random variable, combining the exact distribution of Sk , we are able to produce very accurate BER curves even for m = 4. We will use binary signals as a test case for evaluating our methods. We assume binary frequency-shift keying (BFSK) for obtaining the same limiting BER (3) as in [2], [4], [8]. It should be clear that our methodology applies to other constellations such as M -QAM. The correspondence is organized as follows. Section II describes ZF how to decompose SINRk = SINRk + Tk . Section III has the derivaZF tion of the exact distribution of SINRk and its independence from Tk . Section IV contains the derivation of the rst three asymptotic moments (or upper bounds on moments) of Tk . Asymptotic Normality of Tk is proved in Section V. In Section VI, the Gamma and generalized Gamma distribution approximations are proposed. Section VII has a demonstration about how to apply our results on SINR in computing the probability of errors. Section VIII compares our theoretical results with Monte Carlo simulations. II. DECOMPOSITION OF SINRk Let A = I p +
1
m

H yH . Apply the following shifting operation:


(7)

where ak 2 p21 stands for the k th column of A; ak(0k) 2 (p01)21 is ak with the k th entry removed. A (0k;0k) 2 (p01)2(p01) is A with the k th column and k th row removed. Then
[A

A = (a1 a2 . . . ak . . . ap ) Akk ay shift k(0k) ~ 0! =A ak(0k) A(0k;0k)

01 ]kk = [A ~ 01 ]11
=

(4)

01 Akk 0 ay (A(0k;0k) ) ak(0k) k(0k) 01 A(0k;0k) 01 ak(0k)


I p01 +
1 1

01

(8)

where SINRZF k , denoted by Sk , is the corresponding SINR for the zeroforcing (ZF) receiver, which, conditional on H , can be expressed as (e.g., [1], [3], [17])
ZF = SINRk

From (8), it follows that (2) can be expressed as SINRk =


=

01 1 H yH m

:
kk

Akk 0 ay m
1

k(

(5)

hy hk
k

0k)

01
01 y hk H( 0k)
(9)

It will be shown that Sk has a Gamma distribution. Because Sk and Tk are independent, we can focus only on Tk . As Sk is often the dominating component, separating out Sk is expected to improve the accuracy of the approximation. For uncorrelated channels and channels with equicorrelation, we derive the rst three asymptotic moments for Tk (thus, for SINRk also), which match the simulations remarkably well for nite dimensions. For general correlated channels, some limiting upper bounds for the rst three moments are proposed. We prove the asymptotic Normality of Tk for correlated channels under certain conditions. The proof of asymptotic Normality provides a rigorous basis for proposing approximate distributions. Since Tk is proved to converge to a nonrandom distribution (i.e., a Normal with nonrandom mean and variance), the same asymptotic results hold for the unconditional SINRk , by dominated convergence. The subtle difference between conditional and unconditional SINRk becomes important if one would like to extend to random power distributions or random correlations (see [8]). As an alternative to (3), a well-known approximation to the true BER for binary modulations would be BERa =
0

1 y 0m h H0
2
k

k)

y H (0k) H( 0k)

where hk
m

UDV U DV y ; U
(p

2 m2(p01) ; D 2 (p01)2(p01) ; V 2 01)2(p01) ; U y U = I p01 ; V y V = V V y = I p01 . Let U c be the orthogonal complement of U , implying that U y U c = 0(p01)2(m0p+1) , y y
~ = I m . Plug these into (9) to get ~ = [U U c ] satises U ~ U ~ =U ~U and U

Consider the singular value decomposition (SVD): H (0k) =

2 m21 stands for the kth column 2(p01) is H with the k th column removed.

of H ; H (0k)

SINRk =

yU ~ h 0 ~U hk k

hy UD 2 I p01 + 2 k

D2

01

U y hk :
(10)

Now, expand
~ h = U k

U y hk Uy c hk

tk ;t 2 sk k
1

(p

01)21 ; sk 2
1

(m

0p+1)21 :

Then (10) becomes SINRk =


=

1 1 0 px p2 e dt dFSINR (x)

(6)

m
1

y sk + sk
m

0p+1

y I p01 + tk
2

where FSINR is the cumulative distribution function (CDF) of SINRk . (6) converges to (3) as long as the asymptotic Normality holds. We propose using a Gamma and a generalized Gamma distribution to approximate FSINR . In particular, if Tk is approximated as a generalized

m i=1 = Sk + Tk

ks k
k;i

01 kt k2 k;i
1+
1
m

D2

01

tk

i=1

2 di

(11)

where di is the ith diagonal entry of D .

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

273

The same shifting and SVD operations on SINRZF in (5) yields k SINRZF k

ZF + Tk . So far, no This proves the decomposition SINRk = SINRk statistical properties of H are used, hence, the decomposition is not restricted to Gaussian channels. In the case p > m, SINRk in (2) is still valid. The same SVD yields
SINRk

1 sy s = S : =m
k k k

(12)

1 =m

=1

kt k2 = T 1 + 1 d2
k;i m i

(13)

because H (0k) has only m nonzero singular values when p > m. The results from random matrix theory that we will use later also apply to p > m. However, in this study, we restrict our attention to p  m.

tk The distribution of Sk and its relationship with Tk can be obtained


AND

III. DISTRIBUTIONS OF Sk

p T Fig. 1. The ratios , obtained from 5 simulations, with respect to m , S for two (equal power) SNR levels (c 0 dB, 30 dB), and three levels of ; : ; : . equicorrelations 

E( ) 10 = ( = 0 0 5 0 8)

from the properties of the multivariate Normal distribution (see [18, Ch. 3]). Here, R = P R tP , which is assumed to be positive denite, can be considered as the generalized covariance matrix. We denote a Normal distribution as N(mean; Var), and a complex Normal distribution as CN(mean; Var). Thus, each row of H W  0; I p ), and each row of H  CN(0 0; R). CN(0 Observe that

The rst three moments of Sk are

p+1 E(S ) = m 0m 6 Var(S ) = m 0 p + 1 62


k k

4 6k =

where rkk is the (k; k)th element of R , rk(0k) is the k th column of R with the k th element removed, R (0k;0k) is R with the k th row and k th column removed, [R 01 ]kk is the (k; k)th entry of R 01 . Conditional on H (0k) ; hk ; tk ; and sk are all Normally distributed

~k = c [R 01]kk [Rt 01 ]kk y (R(0k;0k) )01 rk(0k) = rkk 0 rk (0k)

Sk(S ) = E
k

k m2 (Sk 0 E(Sk ))3

(14)

Lemma 2: Conditional on H (0k) ; ktk;i k2 is a noncentral Chi-squared random variable with a moment generating function (MGF)

The following lemma concerns the distribution of tk .

p+1 3 = 2m 0 6: m3
k

(18)

0; 6kI m ) hk j H (0k)  0H (0k) (R(0k;0k) )01 rk(0k) + CN (0


tk = U y hk j H (0k)  0DV y (R(0k;0k) )01 rk(0k)
k p k c k k k m p

(15) (16) (17)

0; 6 I 01 ) + CN (0 y U H 0 I s = h j (0 )  CN (0; 6 0 +1 ) : (Recall that H (0 ) = UDV y ; U y H (0 ) = DV y ; U yH (0 UDV DV y = 0 .) U yU


k k c c

2 ktk;i k2 is a standard noncentral Chi-squared conditional on H (0k) , 6 random variable with two degrees of freedom and noncentrality param2 kzi k2 , which gives the MGF in (19). eter 6

k z k2 = 1 01y6 exp 1y0 (19) y6 01 where z = 0d (V yR( 0 0 ) r (0 ) ) . For uncorrelated channels, 2 kt k  G(1; c ~ ), independent of H (0 ) . Proof: From (16), t j H (0 )  CN(z ; 6 ). For uncorrelated ~ hence kt k2  G(1; c ~ ). In general, channels, z = 0 and 6 = c
E exp(kt k2y) j H (0
k;i i i k k

k;

k;i

k;i

k;i

We state our rst lemma.

ZF = Sk is a Gamma random variable, Sk Lemma 1: SINRk 1 G(m 0 p + 1; m 6k ). Sk is independent of Tk . Proof: Recall that

For future use, we give expressions for three moments of ktk;i k2 , conditional on H (0k)

1 0p+1 ks k2 : Sk = k;i
m

E(kt k2 j H (0 ) ) = 6 + kz k2 Var(kt k2 j H (0 ) ) = 62 + 2kz k2 6 Sk(kt k2 j H (0 ) ) = 263 + 6kz k2 62 :


k;i k k k k i k;i k i k k;i k i k

(20)

=1

The independence of Sk and Tk follows from (16) and (17). Conditional distribution of sk given H (0k) is independent of H (0k) . Therefore, (17) is the unconditional distribution of sk . The elements of sk are 0; 6k ), and thereindependent and identically distributed (i.i.d.) CN(0 fore

The signicance of the decomposition of SINRk can be seen from ) as a function of p; c (SNR), and correlation , Fig. 1, which plots E( T S for m = 16. Here we consider equal power ck = c case with R t being an equicorrelation matrix (i.e., Rt consists of 1s in the main diagonal and s in all off diagonal entries). This gure suggests that in the major p ; Sk might be the dominating component. range of SNR and m IV. ASYMPTOTIC MOMENTS

1 S  G(m 0 p + 1; 6 ): m
k k

ZF is a Gamma random variable Gore et al. [19] proved that SINRk for the equal power case. Both Mller et al. [17] and Tse and Zeitouni ZF . [3] used a Beta distribution to approximate SINRk

1 01 kt k2 = 1 01 kt k2 = 1 ty 3t (21) m =1 1 + 1 d2 m =1 m where  = 1+ 1 ; 3 = (I 01 + 1 D )01 = diag[1 ; . . . ;  01 ].


Tk =
i p p k;i k;i i k k i m i i d p m p

This section derives the asymptotic moments of Tk , written as

274

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

We use some known results for the empirical eigenvalue distribution (ESD) of the product of two random matrices (e.g., Silverstein [20], Bai [21], Silverstein and Bai [22]), to study the asymptotic properties of Tk . 01 ! 2 ; . In We work under the regime: p ! 1; m ! 1; pm the rest of the correspondence, whenever we mention in the limit, we refer to this condition. p01 Suppose that fi gi =1 are the eigenvalues of R (0k;0k) . Suppose furP Rt P converges as p ! 1 to a nonther that the ESD of R random measure F R . By Theorem 1.1 of [20], under some weak regularity conditions on F R (including that the support of F R is compact 1 Hy and does not contain ), in the limit, the ESD of m (0k) H (0k) , denoted by J , converges to a measure J , whose Stieltjes transform, denoted by MJ , satises

(0 1)

MJ z and its derivatives are not sufcient for computing the moments of Tk , which involve ktk;i k2 . We will combine the conditional moments of ktk;i k2 in (20) with MJ z to get the asymptotic moments of Tk . Three cases are considered in the following three subsections. First, we consider uncorrelated channels with unequal powers and obtain the asymptotic moments of Tk explicitly, although the results cannot be expressed in closed form. Next, we show that, assuming equal powers, the asymptotic moments of Tk for uncorrelated channels have closedform expressions. Finally, for general correlated channels with unequal powers, we derive some limiting upper bounds for the moments of Tk and give some sufcient conditions under which these upper bounds are the exact limits.

()

()

MJ z

4 ( )=

( )= x0z
J dx p0

(1

dF R  0 0 zMJ z

()

( )) 0 z

(22)

A. Asymptotic Moments of Tk for Uncorrelated Channels In this case, R t ck P; k I p; R 1 2 2 ; kz i k 2 di k V yR(0 rk(0k) 0 k;0k) rk(0k) i k 2 ktk;i k  ; ck , as shown in Lemma 2. The following three lemmas are proved in Appendix I.

(see [8]). For nite p, the last integral in (22) can be approximated by

1 (1 0 0 zMJ (z)) 0 z : Note that for uncorrelated channels, i = c ~i . In general, (22) requires = fz 2 0 : Im(z) > 0g, and MJ (z) is the unique soluthat z 2 g. However, since MJ (01)  0 tion in fM 2 : 0 z + M 2 and since R is positive denite (implying that all i s are positive), and we only consider  1, it follows that  0 0 zM z 0z and its derivatives are bounded in a complex neighborhood of z = 01, hence MJ (z ) and its derivatives are well dened at z = 01 by the bounded 1i
1 =1

p01

= 0 G(1 ~ ) E

= =

6 = ~ = ck m p; ) = 0, and

i

Lemma 3:

p Tk p0

4 c M (01): ! ck  ;c = k J

(26)

Lemma 4:

(1

( ))

4 c M 0 (01): Var ppp0 1 Tk ! ck  ;c = k J


2 2 2

(27)

convergence theorem. Therefore,

Lemma 5:
1

3) = 1 p0 1 = 1 J^(dx) tr(3 p01 p 0 1 i 1 + m di 1+x p 4 ^ = x 01 (01) J (dx) 0! MJ (01) =  ;c ; 3 ) = 1 p0 tr(3 1 p01 p01 i 1 + m di p 4 ; 0 (01) = = (x 0 (101)) J^(dx) 0! MJ ;c 3 ) = 1 p0 tr(3 1 p01 p01 i 1 + m di 1 p 1 00 4 = (x 0 (01)) J^(dx) 0! 2 MJ (01) =  ;c
1 =1 2 2 1 =1 1 2 2 2 2 3 1 =1 1 2 3 3

p0
(23) (24)

1 Sk(pTk ) ! 2ck  ;c = 4 c M 00 (01): k J 1


3 3

(28)

Therefore, asymptotically, the rst three moments of Tk can be approximated as

1  ;c E(Tk )  ck p 0 p 0 1 Var(Tk )  ck p p ;c 0 1  ;c : Sk(Tk )  2ck p p


2 2 2 3 3

(29) (30) (31)

(25)

Combining with the moments of Sk in (18), and using independence of Sk and Tk , we have

and where 0! denotes converge in probability, and 00 z are the rst and second derivatives of MJ z , respectively. In MJ 0 z , and MJ 00 z have to be solved numerically general, MJ z ; MJ except in some simple cases (e.g., an uncorrelated channel with equal powers). 0 z , and MJ 00 can be i 0 0 zMJ z 0 z . MJ Let i z approximated by solving

()

()

()

()

()

0 (z ) MJ

1 ( ) = (1

( ))

()

1  ;c E(SINRk )  ck m 0pp + 1 + ck p 0 p 0 1 Var(SINRk )  ck m 0pp + 1 + ck p p ;c 0 1  ;c : Sk(SINRk )  2ck m 0pp + 1 + 2ck p p


2 2 2 2 2 3 3 3 3

(32) (33) (34)

1 p0 i z = 1 p0 i MJ (z) + 1 ( ) 10 p0 1 i 1 i (z ) p01 i 1 i (z ) p0 p0 0 (z ) 2 i z i MJ 00 (z ) 1 0 1 MJ = p 0 1 i 1 i (z ) p01 i 1 i (z ) p0 2 (i MJ (z) + i zMJ0 (z) + 1) + p0 1i 1 i (z )


0 z MJ
1 1 =1 2 =1 2 1 1 =1 2 =1 2 1 =1 3

y H (0k) converges 1 cH W . The ESD of m H( In this case, H 0k) to the well-known MarcenkoPastur Law ([21, Theorem 2.5]), from which one can derive closed-form expressions for the moments by (tedious) integration. Alternatively, we directly solve for MJ 0 from the simplied version (i.e., taking i c for all i) of (22) to get
:

B. Closed-Form Asymptotic Moments of Tk for Uncorrelated Channels With Equal Powers

= p~

=~  0 (~ c(1 0 ) + 1)  ;c = MJ (01) = 2~ c

( 1)

(35)

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

275

Similar results can be obtained if we compare our approximate variance formula (33) with the asymptotic variance given in [3]. See the technical report [23] for details. C. Asymptotic Upper Bounds for Correlated Channels Recall that

E(ktk;i k2 kH (0k) ) = 6k + kzi k2


where

kzi k = di k(V y (R 0k;0k )0 rk 0k )i k :


2 2 ( ) 1 ( ) 2

There is an important inequality


 ~ Fig. 2. The ratios  ~ 20 dB.

 ~ and  ~

, with respect to m and p=m, for SNR

p01 i=1

where 

~(1 + ) From 1 + c

= c ~ (1 0 ) + 2~ c(1 + ) + 1. Similarly ~(1 + ) 0  2  ;c = MJ (01) =  ;c 0 1 + c 2~ c  1 c ~ 2  ;c = MJ (01) =  ;c 0 3 : 2 


2 2
0 00


(36) (37)

p01 i=1

1 kzi k2 1 kz k2 = p01 m i i m 1 + 1 d2

i=1

m i

V y (R(0k;0k) )01 rk(0k)

= kV yR(0k;0k) )01 rk(0k) k2 4e : = kR(0k;0k) )01 rk(0k) k2 = k

(41)

 , it follows that  ;c   ;c   ;c :
2

Some limiting upper bounds for the rst three moments of Tk are provided in the next lemma. (38)

01 in our computations. In the literature We always replace by pm p is often used. It can be shown that the (e.g., [3], [4], [8]), = m choice of can have a signicant impact for small dimensions.1 Our ~) is approximate formula (32) for E (SINRk ) (denoted as 
 ~=c m0p+1 1 + cp 0 ;c p p p01 1 =c ~0 1 ~ (1 + p )2 + 1 0 4 m c c ~ (1 0 p ) + 1
2

0 1 6k MJ (01) + ek E(Tk )U = p m Var (T )U = p 0 1 62 M 0 (01) + 1 6


k

Lemma 7:

(42)
k

m2

1 +2 2 2 

2 ek + ek

Sk(Tk ) =
U

(39)

3e 6 1 +2 k k m

(43) p 0 1 3 00 8 =9 2 6 21 3 6 M (01) + m2 6k ek + m2 6k  ek + ek m3 k J

ek + 2

16

2


ek + 2

16

01 , the corresponding  ~ is denoted by When is replaced with pm p  ~p01 . If instead is replaced with m , it is denoted by  ~p . We can ~ with the well-known asymptotic rst moment (e.g., [3], [4, compare  ~R ; eq. (6.59)]), denoted by 
 ~R

0 (44)

where  and  0 are the same constants in the proofs of Lemmas 4 and 5. Proof: See Appendix III. We are interested in the special case where ek ! 0. If ek ! 0 at any rate (faster than O(p0 ) or faster than O(p0 )), ignoring all the terms involving ek ; E(Tk )U (Var(Tk )U or Sk(Tk )U ) is still the true limit. Under these conditions, we propose the approximate moments for SINRk

01 or = p is Similarly,  ~R;p01 and  ~R;p indicate whether = pm m ~R . Obviously,  ~ ~ used for  R;p01 =  p01 .  ~  ~ p Fig. 2 plots  and  as functions of m and m . It is clear ~ ~ ~p01 and  ~R;p at small dimensions can be that the difference between   ~ substantial. For example, when m = 4; p = 4;  is almost 3. On ~ ~p01 and  ~p is not signicant. the other hand, the difference between  This experiment implies that our proposed approximate formulas are not only accurate (see simulation results in Section VIII) but also not as sensitive to the choice of . ~ with  ~R algebraically. The following lemma compares 
Lemma 6:  ~p   ~p01 =  ~R;p01 Proof: See Appendix II.

=c ~0 1 ~(1 + p )2 + 1 0 c ~(1 0 p )2 + 1)2 : 4( c

(40)

p+1 0 1 6k  ;c + ek E(SINRk )  m 0m 6k + p m Var(SINR )  m 0 p + 1 62 + p 0 1 62 2 + e2


k

(45)

(46) k k m2 m2 k ;c p+1 3 0 1 63  ;c + e3 : (47) Sk(SINRk )  2 m 0 6k + 2 pm k k 3 m3

 ~R;p .

that in our notations, c P in [3]. and equivalent to 


1Note

A ~ = cm p is equivalent to 

in [4, eq. (6.59)]

2 3 Note that ek ; ek ; ek are retained in these expressions because simulation studies show that keeping these terms improves accuracy at very small dimensions. It turns out that in the equicorrelation situation, ek ! 0 at a good rate under mild regularity conditions. The equicorrelation is the simplest correlation model (e.g., [15, eq. (4.41)]) and is often used to model closely spaced antennas or for the worst case analysis [24]. For more realistic correlation models, see Narasimhan [25].

276

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

Lemma 8: If the correlation matrix R t consists of 1s in the main diagonal and s in all off-diagonal entries,2 then ek ! 0 in the limit, f as long as i6=k c c = O (p ) for f < 2. Sufcient conditions for respectively. Proof: R t can be written as R t = (1 0 )I p + 1 p1T p , where 1 p denotes a column vector of p 1s. Recall that R = P R tP .
T P (0k;0k)1p011p R (0k;0k) = (1 0 )P (0k;0k) + P 01P (0k;0k)

ek ! 0 faster than O(p0

4 ) and O(p0 ) would be f < 3 2 and f < 3 ,

Although (51) is bounded by ek = k(R(0k;0k) )01 rk(0k) k2 , it is challenging to show that (51) converges (not necessarily to zero). For the remainder of this section, we assume that ek ! 0 faster than O(p0 ), which makes the contribution of the term (51) asymptotically negligible. Lemma 9:

W
D

with an inverse

0k;0k) (R(0k;0k) )01 = ( 10

P 01

where =) means converge in distribution. Proof: See Appendix IV. Corollary 1:

3) = 0 6k tr(3 D 2 2 = mTkp ) N 0; 6k ;c p01

(52)

0 P k;0k)  : (48) 0 (100 1 1T P 0  (p 0 1) + 1 0  p01 p01 (0k;0k)


P kk P (0k;0k)1p01 . Therefore, = P

mTk 0 (p 0 1)6k  ;c pp 0 1

D 2 2 = ) N 0; 6k ;c

(53)

Note that rk(0k)

Proof:

R(0k;0k) 01 rk(0k)  0 = 10 1 P P  kk (0k;0k) p01   0 0 10 P P 1 (p 0 1)  kk (0k;0k) (p 0 1) + 1 0  p01  0 = (p 0 1) (49) + 1 0  P kk P (0k;0k)1p01
and

mTk 0 (p 0 1)6k  ;c pp 0 1 3) + 0 6k tr(3 = mTkp p01


We have shown that

3) 0 6k tr(3 (p 0 1)6k  ;c : pp 01

mTk 0 6k tr(3 pp 0 1 3c )

D 2  2 ): = ) N(0; 6k ;c

ek =

= =

2  T 01 (p 0 1) + 1 0  P kk1p01P (0k;0k)1p01 2  ck (p 0 1) + 1 0  i6=k ci

2 R(0k;0k) 01 rk(0k)

The fact that

3) 0 (p 0 1) ;c ! P pp 6k tr(3 0 01
(50) follows from Theorem 1.1 in Bai and Silverstein [22], which implies 3) that (p 0 1)( tr(3 p01 0  ;c ) is stochastically bounded. Equation (53) now follows from the Converging Together Lemma (or Slutskys Theorem) [26]. Corollary 2:
p+1 p01 SINRk 0 m0 D m 6k 0 m 6k  ;c = ) m0p+1 m

i= 6 k c = O(p ), for some f < 2. If we want ek ! 0 faster than O(p0 ) and O(p0 ), we should

from which it follows that ek

! 0 if

In the equal power case, i6=k c c, c = p 0 1 and 6k  (1 0 )~ which implies that the rst (second, third) moment of SINRk for an equicorrelation channel is roughly only 1 0  ((1 0 )2 ; (1 0 )3 ) of the moment without correlations. To conclude this section, we mention that our asymptotic results also apply to the real channels (e.g., in [3], [8]) with a modication that involves multiplying the variance by a factor of 2 and the third moment by a factor of 4. V. ASYMPTOTIC NORMALITY OF Tk This section proves that, if appropriately normalized, Tk converges in distribution to a Normal random variable, both in the case of uncorrelated channels as well as the correlated channels if ek ! 0 faster than O(p0 ). The latter includes the equicorrelation channel with regularity conditions as in Lemma 8. The difculty in dealing with channels with arbitrary correlations is due to the term
p01

4 have f < 3 2 and f < 3 , respectively. This completes the proof.

2 + p01 62  ;c 2 6k k m

N(0; 1):

(54)

Proof: This is a direct consequence of the independence of Sk and Tk . VI. DISTRIBUTION APPROXIMATIONS Once the asymptotic Normality of Tk (and SINRk ) is proved, it is possible to rigorously dene the approximating distributions. Since Sk and Tk are independent, we do not need to approximate the distribution of Sk . However, we keep the option of approximating the distribution of Sk + Tk itself. Our approach is to match the rst two asymptotic moments of Tk with the corresponding moments of a target distribution (e.g., Gamma). The asymptotic Normality of Tk (Lemma 9) ensures that the approximate distribution converges to the true limit. The Normal approximation is not accurate for small dimensions, and to an extent this is because Tk is positive and a Normal random variable has zero third central moment. We expect a Gamma distribution, which has nonnegative support and nonzero third central moment, to be a better approximation. It turns out that only considering the rst two moments may not be sufcient for accurately computing error probabilities. Therefore, we consider a generalized Gamma distribution, which matches the rst three moments and hence is likely produce better results, as another approximation.

m i=1

1 kz k2
i

p01 i=1

1 2 1 + 1 d2
m di m i

V y (R(0k;0k) )01 rk(0k)

2
i

: (51)

2In order for

1 <  < 1. Rt to be positive denite, we have to restrict 0 p0 1

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

277

In this section, the same conditions are assumed as in proving the asymptotic Normality of Tk . That is, ek ! 0 faster than O(p0 ). When this condition is not satised, we do not have a rigorous asymptotic result. In this section, the symbol  is also used to denote the approximate distributions. A. Normal Approximation Asymptotic Normality of Tk implies that
2 2 Tk  N p 0 1 6k  ;c ; p 021 6k  ;c : m m

(55)
Fig. 3. RST and RSSINR for selected SNR levels, over the whole range of , for equal power uncorrelated channels.

m 0 p + 16 + p 0 16  ; SINRk  N k m m k ;c m 0 p + 1 62 + p 0 1 62  2 : k m2 m2 k ;c
B. Gamma Approximation whose parameters are determined by solving

Also, the asymptotic Normality of SINRk means that

(56)

like RS = 1. It will be shown later that RS is also critical for the generalized Gamma approximation. The following inequalities hold for uncorrelated channels with equal powers. Lemma 10: Assuming equal powers and no correlations

Tk can be approximated by a Gamma random variable G( T ; T )

The Gamma approximation of Tk is therefore,

0 1 6k  ;c E(Tk ) = T T = p m 0 1 62  2 : 2 Var(Tk ) = T T = pm k ;c 2
2 2 (p 0 1) ;c 1 6  ;c ; : k 2  ;c m  ;c

RST RSSINR RST

1  1 for any p  m  RSSINR in the limit:


RST
and

(63) (64) (65)

Proof: See Appendix V. (57) Fig. 3 plots some examples of power uncorrelated cases.

Tk  G Tk should be

RSSINR

, for the equal

According to the Gamma approximation, the third central moment of


3 2 T T 4 0 1 63  ;c = 2 pm : k 3  ;c

C. Generalized Gamma Approximation A generalized Gamma distribution can be described by a stable law or an innite divisible distribution [27], [28], [26, Chs. 2.7, 2.8], which involves the sum of i.i.d. sequence of random variables. In our case, conditionally, Tk is a weighted sum of non-i.i.d. random variables, hence Tk is not exactly a stable law nor innite divisible. The Gamma approximation of Tk can be generalized by introducing an additional parameter to the original Gamma distribution [27], [28], i.e., assuming Tk  G( T ; T ; T ). The regular Gamma distribution is a special case with T = 1. Assuming a generalized Gamma distribution, the rst three moments of Tk would be

(58)

We can also approximate SINRk by a Gamma distribution again by matching the rst two moments SINRk

G( ; ),

0 p + 1 + (p 0 1) ;c )2 ;  G (m 2 m 0 p + 1 + (p 0 1) ;c 2 1 6 m 0 p + 1 + (p 0 1) ;c : k m m 0 p + 1 + (p 0 1) ;c


2 2 2 63 m 0 p + 1 + (p 0 1) ;c : k m3 m 0 p + 1 + (p 0 1) ;c

(59)

The third central moment of SINRk then would be (60)

E(Tk ) = T T 2 Var(Tk ) = T T 3 Sk(Tk ) = (T + 1) T T :

(66)

Dene RS to be the ratio of the third central moment of the approximated Gamma distribution to the asymptotic third central moment. For Tk this becomes

Equating these moments with the asymptotic moments of Tk will lead to the same T and T as for the Gamma approximation. The third parameter T will be

RST =
For SINRk ;

4  ;c :  ;c  ;c

(61)

T =

RST

2 0 1:

(67)

RSSINR RST

01 + p01  ;c 2 2 1 0 pm m = 01 + p01  ;c 1 0 p01 + p01  ;c : 1 0 pm m m m

(62)

Similarly, we can also generalize the Gamma approximation of SINRk by assuming SINRk  G( ; ;  ). The third parameter will be

and RSSINR are indicators of how well the Gamma approximation captures the skewness of the distribution of Tk . Ideally, we would

=

RSSINR

0 1:

(68)

278

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

When  > 1, the generalized Gamma distribution with these parameter does not have an explicit density in general. However, it can be described in terms of the stable law and has a closed-form MGF [27], [28] MGF (s; G( ; ;  )) = exp

1 where the gamma function 0(y ) = 0 ty01 e0t dt, and the incomplete y 0 1 e0t dt, and 2 F1 ( 1 ) is the hypergamma function 0( ; y ) = 0 t geometric function
2 F1

When  < 1, the generalized Gamma distribution is a compound Poisson distribution with an MGF MGF (s; G( ; ;  )) = exp

01

1 0 (1 0 s)

1; 1=2 + ; + 1;

(69)

1= 1=2 + 1= 1 0(1=2+ +n) 0(1=2+ ) = 0( +1+n)


n=0
0( +1)

n 1= 1=2+1=

n!

(75)

10

1 1 0 s

01

B. BER by Generalized Gamma Approximation on SINRk According to the results of [30], [31, Ch. 9.2.3], under the generalized Gamma approximation on SINRk , the BFSK BER (denoted as BERgg ) can be expressed in terms of the MGF as BERgg

(70) From Lemma 10, it follows that T > 1 and  > 1 for uncorrelated equal power channels. VII. ANALYSIS OF THE PROBABILITY OF ERROR Computation of the probability of errors using the distribution of SINRk is a way of measuring how successfully the proposed distribution approximates the truth. For this purpose, we will compute the BER using (6), which is equivalent to the BFSK BER. It should be straightforward to apply our methods to other types of (nonbinary) constellations. To simplify the notations, the subscript k in BERk will be dropped for the rest of the correspondence. In this section, we will provide BER formulas under the Gamma and the generalized Gamma approximations, and compare these results with the exact formula for BER, denoted by BERe , given in [4, eq. (6.47)] or [7]. If the asymptotic Normality of SINRk holds, as in Corollary 2, then BERa

=1  =1 

0 0

MGF

1 ; G( ; ;  ) d 0 2 sin 2

exp

1 0 1 +  01 2 sin2 

d
(76)

which is for  > 1 and can be evaluated numerically. We can similarly write down BERgg for  < 1. C. BER by Generalized Gamma Approximation on Tk Under this approximation, when T MGF for SINRk = Sk + Tk as MGF

> 1, we can write down the

1 1 0 p e dt dFSINR (x) p 0 x 2 4 Q E(SINR ) : ! BER1 = k 1

4 1 =

(71)

s; G m 0 p + 1; 6k + G ( T ; T ; T ) m 1 = m0p+1 10 6 ms

For uncorrelated channels with equal powers c2 (1 0 )2 + 2c(1 + ) + 2 + c(1 0 ) 0 E(SINRk )1 = (72) 01 = p = . However, for correlated channels, since by treating pm m it is not convenient to compute E(SINRk )1 , we will use the our nite-dimensional moment formula to compute E(SINRk )1 , which is different for different p and m. Next, we will derive a variety of BER formulas corresponding to various approximation schemes, based on BFSK, which can be easily generalized to other types of constellations. A. BER by Gamma Approximation on SINRk Denote the BER computed using the Gamma approximation by BERg . Integration of (6) by parts yields BERg

2 exp T T0 1 1 0 (1 0 T T s)
The corresponding BER (denoted as BERg+gg ) would be BERg+gg

(77)

=1 

1+

m0p+1 6 2m sin 

T T 2 exp T T0 1 1 0 1 + 2 sin2 
which has to be evaluated numerically. VIII. SIMULATIONS

d

(78)

Replacing FSINR bution G( ; ) as dened in (59), and using the results from integral tables [29], we have BERg

1 1 0 1 x0 dx: p e FSINR (x) 2 (73) 2 0 (x) with F ; (x), the CDF of the Gamma distri-

1 1 0 1 x0 dx p e F ; (x) 2 2 0 1 0 0 1p = x e 0 ; x dx 20( ) 2 0 0 (1 = 2 + ) (1 = ) 1 p = 20( ) 2 (1=2 + 1= )1=2+ = 22 F1 1; 1=2 + ; + 1; 1=21+ 1=

Our simulations consider m = 4 (p = 2; 4), and m = 16 (p = 8; 16), for the equal power case. Both uncorrelated and correlated (with equicorrelation  = 0:5) channels are tested. The range of SNRs is c = 0 dB30 dB. Without loss of generality, the rst stream (i.e., k = 1) is always assumed. H W is sampled 106 times for every combination of (m; p; c, and ) for computing the empirical moments, distributions, and BER, except when computing the exact BERe for m = 16; H W is only sampled 105 times. A. Moments The theoretical moments are computed using (45), (46), and (47). Fig. 4 plots the rst three moments of Tk computed both theoretically and empirically from simulations. For m = 16, the theoretical moments match the simulations very well, especially the rst moment. When m = 4, the curves for the rst moment are still quite accurate except for the correlated cases at very small SNRs. Note that due to the log scale, the errors at small SNRs are largely exaggerated in the

(74)

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

279

Fig. 4. Theoretical and empirical moments of Tk. The vertical axes are in the log10 scale. Note that the variances and third moments are in terms of their square roots and cubic roots, respectively.

Fig. 5.

Theoretical and empirical moments of SINR.

gure. When m = 4, for the second and third moments, there seem to be some signicant discrepancies between theoretical results and simulations at large SNRs. However, these errors contribute negligibly to the second and third moments of SINRk (see Fig. 5). For example, when m = 4; p = 4;  = 0, exact Sk(Sk ) = 314:983 . Although the empirical Sk(Tk )(= 16:403 ) differs quite signicantly from the theoretical Sk(Tk )(= 5:333 ), the theoretical Sk(SINR)(= 314:983 ) is almost identical to the empirical Sk(SINR)(= 314:993 ). The theoretical and empirical moments of SINRk are compared in Fig. 5 for m = 4. As expected, the curves match almost perfectly except for the observable (due to the log scale) errors at very small SNRs. B. Distributions Fig. 6 presents the quantile-quantile (qq) plots for distributions of Tk based on Gamma and Normal approximations against the empirical distribution, at a selected SNR = 10 dB. The gure shows that the Normal approximation works poorly for small m or p. The Gamma

approximation ts much better in all cases. Fig. 7 gives the same type of plots for SINRk . It shows the Gamma approximation works well, p = 1 . When p = 1, the Gamma distribution approxespecially for m 2 m imates the portion between 1%99% quantiles pretty well. We have shown that RST  RSSINR for uncorrelated channels with equal powers in Lemma 10, which implies that a Gamma could approximate the distribution of SINRk better than that of Tk . Also, p = 1 than at p = 1 , Fig. 3 shows RSSINR is much smaller at m m p = 2 1. which helps explain why Gamma approximation works well at m 2 C. Error Performance Fig. 8 plots BERs versus SNRs for uncorrelated channels. The gure shows that the BER curves produced by the Gamma approximation p = 1. are almost indistinguishable from the simulated curves for m 2 p When m = 1, the Gamma approximation still works well for moderate SNRs (e.g., SNR < 15 dB). At larger SNRs, the Gamma approximation slightly overestimates BER. The gure also shows that the generalized Gamma distribution produce almost perfect ts to the simulated BER

280

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

Fig. 6. Quantile-quantile plots for Tk . The triangles on the curves indicate the 1% and 99% quantiles. The range of quantiles is 0.1%99.9%.

Fig. 7.

Quantile-quantile plots for SINR . The triangles indicate the 1% and 99% quantiles. The range of quantiles is 0.1%99.9%.

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

281

Fig. 10.

= 16.

Compare the uncorrelated and correlated BER curves. m

= 16,

1 Fig. 8. BER curves for uncorrelated channels. and are used for 2 computing BER in both (a) and (b). BER stands for the exact BER. BER and BER are computed by assuming SINR is a Gamma and a generalized uses the exact Gamma distribution for Sk and Gamma, respectively. BER approximates Tk by a generalized Gamma. BER is the asymptotic BER.

=1

curves. All gures indicate using the asymptotic BER (BER1 ) formula will seriously underestimate the error probabilities at large SNRs.3 Fig. 9 presents the BER results for the correlated channels with equicorrelation  = 0:5. We can see the similar trends as for the p =1 uncorrelated cases, i.e., Gamma approximation works well for m 2 and the generalized Gamma approximation performs remarkably well. Finally, to see the difference between correlated and uncorrelated cases more closely, Fig. 10 plots the interesting portions of the BER curves for both cases, which illustrates that the differences could be signicant. IX. CONCLUSION This study characterized the distribution of SINR for the MMSE receiver in the MIMO systems, for channels with nonrandom transmit correlations and unequal powers. The work started with a key observation that SINR can be decomposed into two independent component: SINR = SINRZF + T , where SINRZF has an exact Gamma distribution. For uncorrelated channels as well as the correlated channels under certain conditions, T is proved to converge in distribution to a Normal and can be well approximated by a Gamma or a generalized Gamma. Our BER analysis suggested that these approximate distributions can be used to accurately estimate the error probabilities even for very small dimensions. APPENDIX I PROOF OF LEMMAS 3, 4, 5 Restate Lemma 3 as follows:

E
Proof:

p Tk p01

4 c M (01): ! ck  ;c = k J

(79)

p Tk p01

=E =E

p01

1 E(pT j H ) k (0k) 1
p01

p E(ktk;i k2 )i p 0 1 m i=1 3) p = 6k m E tr(3 p01 p ! 6k m E(MJ (01)) = ck MJ (01)


Fig. 9. BER curves for equicorrelation  : . Note that the BER s are different because we did not simulate the true limit of the mean.

= 05

3Note that, in order to produce comparable BER results with the classical references (e.g., [2], [4], [8]), we actually used real channels (only for the BER curves).

282

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

by the bounded convergence theorem [26, Sec. 1.3b], because

3)  1: tr(3 p01
Restate Lemma 4 as follows:

We can show that the rst term in the right-hand side of (84) con3 M 00 (01) as follows: verges to ck J
p

1 0 1E E =
p3 m3 p

pTk

0E

pTk H (0k)
p01 i=1

3 j H (0k) 3
H (0k)

4 c2 M 0 (01): 2 2 = Var ppp0 1 Tk ! ck ;c k J


Proof:

(80)

1 E p01 E kt k2 0 6 p3 =m k;i k 3 p01


p3 1 =m 3 p 0 1E p 3 = 2m 3 6k E
i=1 p01 i=1

1 0 1E E

ktk;i k2 0 6k i
3 3 
i

Var ppp0 1 Tk

2 1 p01 kt k2 = pp Var i k;i 01 m


i=1 i=1 p01 i=1

3 3 26k i

2 1 p01 kt k2  = pp Var E i k;i 01 m 2 1 + pp 0 1 Var E m

H (0k) H (0k)

p 31 ! 2m 3 6k 2 E

tr 33 p01
MJ

00 (01) = c3 M 00 (01): k J

(85)

ktk;i k2 i

32 ) + p2 62 (p 0 1)Var tr(3 3) p2 2 tr(3 =m 6 E k k 2 p01 m2 p01 2 M 0 (01) ! ck J


3) 0 because E( tr(3 p01 ) ! MJ (01) by the bounded convergence theorem, 3 tr(3 ) and (p 0 1)Var( p01 )!0, which can be proved using the results from the concentration of spectral measures for random matrices [32]. The result we need is Corollary 1.8b in [32], which can be stated as follows:

We can show the last two terms in the right-hand side of (84) tend to 0. Expand
p

1 0 1E E

pTk H (0k)

0 E(pTk ) 3
3) 0 E(tr(3 3)) tr(3 p01 p01
3
:

p 3 2 =m 3 6k (p 0 1) E

Apply the concentration theorem one more time

3) 0 E tr(3 3) tr(3 p01 p01

E
>

 2e0(p01) 

(81)

for any  > 0.  depends on the spectral radius of R (0k;0k) , the logarithmic Sobolev inequality constant, and the Lipschitz constant 1 . We can easily check that the function of g (x) = f (x2 ) = 1+ x 1 f (x) = 1+x is convex Lipschitz and g (x) has a nite Lipschitz norm. Therefore,

3) 0 E(tr(3 3)) 3 tr(3 p01 p01 1 2 3) 0 E tr(3 3) > x  3x P tr(3 p 0 1 p 0 1 0 1 2 0(p01) x 3 1  3 x 2e dx = 2 (p 0 1)3
0

dx  3

(86)

3) = 1 2xP tr(3 3) 0 E tr(3 3) > x Var tr(3 p01 p01 p01 0 1 0(p01) x 2 dx =  4 xe  (p 0 1)2
0 3
Restate Lemma 5 as follows:
p

3 1 E((E(pTk jH which implies p0 (0k) ) 0 E(pTk )) ) ! 0. 1 The last term in (84) also tends to zero, again using the concentration theorem and the following inequality:
p

dx:

(82)

3) which implies (p 0 1)Var( tr(3 p01 )!0. This completes the proof.

0 1 Cov(E(pTk j H (0k) ); Var(pTk j H (0k) )) 3) ; tr(3 32 ) p3 3 tr(3 = 3m 6 ( p 0 1)Cov k 3 p01 p01


3) p3 tr(3 3  3m 3 (p 0 1)6k Var p 0 1

0 1 Sk(pTk ) = p 0 1 E((pTk 0 E(pTk )) ) 4 c3 M 00 (01): 3 = ! 2ck ;c k J

32 ) Var tr(3 p01

(87)

We have shown (83)

3)  2 : Var tr(3 p01  (p 0 1)2


Similar arguments will show that

Proof:

1 E((pT 0 E(pT ))3 ) k k p01 1 E E pT 0 E pT j H 3 j H (0k) = p0 k k (0k) 1 1 E E(pT j H ) 0 E(pT ) 3 + p0 k k (0k) 1 3 Cov E pT j H + p0 k (0k) ; Var pTk j H (0k) 1

32 )  2 Var tr(3 p01  0 (p 0 1)2


for a different constant  0 . Therefore, the last term in (84) also tends to zero. Combining the results for the three terms of (84) together completes the proof.

: (84)

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

283

APPENDIX II PROOF OF LEMMA 6 Restate Lemma 6 as follows:


p

APPENDIX III PROOF OF LEMMA 7 We will frequently use the following three inequalities:

~  ~ 01 =  ~ 01   ~
p R;p

R;p

(88)

~p Proof: Expand 
p m0p+1 1 cm 0 1 p 0 1 (1 0  ) ~ =1 +2 2c p p2 2 p
p

Zk

4 (1) =

(89)

 kR(0k;0k) )01rk(0k) p01 1 2 kz i k 2 4 (2) = m Z  1e


k

=1

01 1 kz k2
m
i

=1 1 + 4e k2 =
i k k

01 1 kz k2 i
m m

1 d2
i

(91) (92)

where
p
p

(p )2 = (cm + p 0 cp)2 + 4cp2 = (cm + p)2 + c2 p2 + 2cp2 0 2c2 mp:


Expand  ~p01
p01

m2 c2 2 p

p 1 0 m + 2c m 1+ m +1 p

p 2

Zk

4 (3) =

=1 p01
i i

=1

1 + 1 d2 2 2 1 6 kz k 2 8 3  9e 1 2 1+ d
m i m i m i

(93)

x 2x  1 ; 6x  8 , for any x  0.)  1; (1+ (Note that 1+ x 2 (1+x) 9 x) We begin with the rst moment. Conditional on H (0k)

m0p+1 1 (1 0  ) =1 02 01 2c p
p

E
(90)

Tk H (0k)

where
p01
p

1 01 6 + kz k2 =m 1 2 =1 1 + d 1 6 tr(3 3) + Z (1) =m 3) + e :  1 6 tr(3


p k i i m i k k

(94)
k

2 p
p

(p 01 )2 = (p )2 + c2 + 2c2 m 0 2c2 p 0 2cp:


Therefore,
p

c c 2c 2m +p 2 + 2c p2 0 2 p 0 p ;

3) Recall that tr(3 p01 then given by

! M (01), the limiting upper bound of E(T ) is


P J k U k

0 16 E(T ) = p m
:

MJ

(01) + e

(95)

~ 0 ~ 01 = 1 2
p

cm p2

1 + p 0 1 0  +p 01 p
p p

We can derive the limiting upper bound for the variance of Tk by following the proof of Lemma 4. We know that

It sufces to show

()(cm + p)(p 0 1)p  (p 0 1)(cm + p)2 0 cp(p 0 1)(cm 0 p) ()(cm + p)p  (cm + p)2 0 cp(cm 0 p) ()(cm + p)2  (cm 0 p)2
p p

2 cm + p + p(p 0 1)p 0 p p01  0 2 + 2(cm + p)(p 0 1)p ()(cm + p)2 + (p 0 1)p2 p p 4 2  p p01

Var ppp0 1 T

= pp 0 1E

1 01 62 + 2kz k2 6 1 2 2 m2 =1 1 + d 2 1 01 6 + kz k2 + pp Var 1 2 01 m =1 1 + d
2
p k i i m i p k i i m i

(96)

The rst term in the right-hand side of (96) can be written as


p2 m2

which is true. Therefore, we prove  ~p is equivalent to. ~p01   ~R;p , note that To show 
p01

 ~ 01 . Here we use() for


p

32 ) + p2 1 6 E(Z (2) ) 62 E tr(3 p01 m p01 p2 1 p2 2 00 1 m 2 6 M (01) + m p 0 1 6 2 e


k k k k J k

(97)

in the limiting sense. Using the fact that

0 ~

R;p

=1 2

c p

+  01 0 
p

Var(x + y)  Var(x) + E(y2 ) + 2 Var(x) E(y2 )


we can bound the second term in (96) by
p2 p m2

Sufces to show
p 01  p 0 c ()p2 201  p2 2 + c2 0 2cp ()c2 + 2c2 m 0 2c2 p 0 2cp  c2 0 2cp ()p  0cm + cp + p
p p p p p p

3) + p2 E Z (1) 2 ( 0 1)62 Var tr(3 p01 p01 2 3) E Z (1) 2 + 2p 6 Var tr(3 m p01 p2 262 + p2 e2 + 2 p2 6 m 2  (p 0 1) p 0 1 m(p 0 1)
k k k k k k

2e


(98)

which is true. Hence, we complete the proof.

where  is the same constant in Lemma 4.

284

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

Combining (97), (98), and ignoring the higher order term in (98), we can get a comprehensive upper bound

Combing all the bounds and limits, we can obtain a limiting upper bound for the third moment

Var ppp0 1 Tk

p2 2 p2 1 1 =m 2 6k MJ (01) + m 6k p 0 1 2 ek
0

0 1 63 M 00 (01) + 8=9 62 ek + 6 62 1 ek + e3 Sk(Tk )U = pm k k k 3 k J m2 m2 


(99)

2 p2 2 2 +pp 0 1 ek + 2 m(p 0 1) 6k  ek

3e 6 1 +2 k k m

ek

16 2 + 2m k 

ek

16 2: + 2m k 0

(104)

which, for convenience, is written as

0 1 62 M (01) + 1 6k 1 + 2 2 Var(Tk )U = pm 2 k J m 2 
0

ek

2: + ek
(100)

APPENDIX IV PROOF OF LEMMA 9 Restate Lemma 9 as follows:


W

We can get the limiting upper bound for the third central moment by following the proof of Lemma 5. Recall
p3 p

3) = 0 6k tr(3 D 2 2 = mTkp ) N 0; 6k ;c p01

(105)

1 3 0 1 Sk(Tk )= p 0 1 E E pTk 0 E(pTk j H (0k) ) j H (0k) 1 E E(pT j H ) 0 E(pT ) 3 + p0 k k (0k) 1 3 Cov E pT j H + p0 k (0k) ; Var pTk j H (0k) 1
Expand the rst term of the right-hand side of (101) to get

2 2 ), where E(ejyW ) ! exp(0 y2 6k ;c

Proof: It sufces to show that the characteristic function of W ,

E(ejyW ) = E(E(ejyW j H (0k) ));


Conditional on H (0k)
:

= 01 :

(106)

jyW

j H (0k)
jy

(101)

= E exp

mTk

p3 m3

33 ) + p3 1 62 Z (3) 3 E tr(3 6k k k p01 m2 p 0 1 3 p p3 1 28 3 00 m 3 6k MJ (01) + m2 p 0 1 6k 9 ek :

(102)

Expanding the second term of (101) and selectively ignoring some negative terms, we can get a bound

0 1 E E(pTk j H (0k) ) 0 E(pTk ) p3 1 3) 0 E(tr(3 3)))6k + mek )3 m 3 p 0 1 E((tr(3 


p3 m3

3) p01 E exp jykt k2 k tr(3 p6 = exp 0jy i k;i p01 i=1 3) k tr(3 p6 = exp 0jy p01 jy p01 01 p p01 2 1 0 pp10 1 jy6k i exp kzi k2 jy  p6 1 0 i=1 p01 3) L (p)L (p): k tr(3 p6 = exp 0jy (107) 1 2 p01
Note that i can assume

3) 0 6k tr(3 p p01

H (0k)

1+

 1. For any xed y 2

, since p

! 1, we

3(3 1 6k 2p01

 3

3 +3 ) + p 0 1 ek

p3

p3 m2

2 6k

 p

( 0 1)

ek :

y j6 k  i jyj < 1; jp pp <1 01 p01

which allows us to conduct complex series expansions. We can write


L2 p

We can also bound the last term of (101) by

3 0 1 Cov E pTk j H (0k) ; Var pTk j H (0k) 3 Var E pT j H  p0 Var Var k (0k) 1 Var E

( )=

pTk H (0k)

jy p p01 jy  p6 1 0 i=1 p01 jy p01 p p01 = exp kzi k2 jy  p6 1 0 i=1 p01

p01

exp kzi k2

(108)

We know
pTk H (0k) pTk H (0k)

Then

2 2 + 2p 6 2e  p2 ek k k m 

k log(L2(p))k  
:

p01 i=1 p01 i=1

kzi k2

and

Var Var
4

p (2) 2 32 ) + mZk =m 4 6k Var 6k tr(3 2 p4 2 2 22 2 1 m 4 6k 6k  0 + m 2 ek + 6k mek  0

(103)

(1)  2jyj ppm 0 1 Zk  2jyj ppm 0 1 ek ! 0

ji 2 kzi k2 pjy p01

jy p p01  p6 1 0 jy p01

(109)

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

285

because of the assumption that ek ! 0 faster than O(p0 P L2 (p) ! 1, hence can be ignored.
p01

). Therefore

APPENDIX V PROOF OF LEMMA 10 First we can show that RST =   2 To simplify the expressions, let  ;c 2  ;c 0 Q =  ;c 0 O 0 Q, where


01 L1 (p) = 1 0 pp10 1 jy6k i i=1 p01 6k i pp = exp 0 log 1 0 jy 01 i=1 p01 1 n (jy6k i ) = exp n (p 0 1)n=2 i=1 n=1 1 p01 (jy6  )n k i = exp n ( p 0 1)n=2 n=1 i=1 p01 p01 2 jy 6k i 0y 2 62 k i = exp exp 1 = 2 (p 0 1) 2(p 0 1) i=1 i=1 1 p01 (jy6  )n k i 2 exp n(p 0 1)n=2 n=3 i=1 3) exp 0y2 62 32 ) L (p): (110) 6k tr(3 k tr(3 = exp jyp 3 2(p 0 1) p01 We will show that L3 (p) converges to 1 in probability. Note that in the
above derivations, we can expand the complex logarithm and switch yj the order of summations because of the condition pjp 01 < 1

4   ;c  ;c .  1, i.e.,  ;c =  ;c 0 O and  ;c =
c ~ Q = 3: 

O
Then

~(1 + ) + 1 0  ; =c 2~ c 

4  ;c  ;c 0  ;c

2 0 Q 0 4 =  ;c  ;c ;c 2 O0 Q =  ;c (112) ;c c ~ (1 + ) + 1 0  2 = 0  ;c ;c 2~ c  (~ c(1 0 )2 + (1 + ) 0 (1 0 )) = (113) 2  2 c(1 0 ) + (1 + ) 0 (1 0 )) 2 O = (~  ;c 2  (~ c(1 + ) + 1) 0  2 2~ c  1 2 = 2~ ( 0 (~ c(1 0 ) + 1) 0 2~ c ) (114) c 2 2 0c ~(1 0 ) 0 1 c ~(1 0 ) 0 1 : ~ = 0c  ;c Q = 2~ c 3 2 3
(115)

1 p01 (jy6  )n k i klog(L3 (p))k = n=2

(jyj6k )n  n(p 0 1)n=2 n=3 i=1 1 p 0 1 (jyj6 )n k  n=2 3 ( p 0 1) n=3


jy j 6 p 0 1 (p01) = 3 y j6 1 0 (pj0 1)
After cancellations
P ! 0;

n=3 i=1 1 p01

n(p 0 1)

Plug (114) and (115) into (112),

4  ;c  ;c 0  ;c

1 = 2~ c 2 3

3 0 2 (~ c(1 0 ) + 1)

0 3~ c  + c ~2 (1 0 ) + c ~

(116)

Therefore, i.e., L(p; c) ! 1:


P

ejyW j H (0k)

32 ) 0y 2 62 P k tr(3 ! lim exp p!1 2(p 0 1) y2 2 2 = exp 0 2 6k  ;c

(111)

which is true. It is easy to check to make sure 2 completes the proof for

4 0  ;c  ;c 0  ;c 3 2 c(1 0 ) + 1) 0 3~ c  + c ~2 (1 0 ) + c ~ () 0  (~ 2 2 ()( 0 3~ c )  (~ c(1 0 ) + 1)( 0 c ~ ) ~ )2 c 2 )(2 0 c c )2  (2 0 4~ ()2 (2 0 3~ ()2 (2~ c )(22 0 4~ c )  4~ c (2 0 c ~ )2 2 2 ()2~ c 0

0

0 3~ c  0. This

which is the characteristic function of a Normal random variable with 2 2 . mean 0 and variance 6k ;c To show

RST

E(e

jyW

) ! exp +
y

2 2 2 0 y 6k ;c

with equality held when c = 0 or = 0. It is easy to show RSSINR  1 by noting that

=   ;c 1 
;c ;c

it sufces to show jE(exp(jyW

E exp

jyW

+ 2

y2 2 2 6k  ;c y2

62  2

k ;c

)) 0 1j ! 0.

2  ;c
which implies

 ;c  p ;c  ;c   ;c + 2

01

RSSINR

2 2 + 2 6k ;c H (0k) 0 1 2 2 2 E(exp(jyW )jH = E exp y2 6k (0k) ) 0 1 ;c

= E E exp

jyW

2 2 exp 0 y 62 2 0 1 = 0: ! E exp y2 6k ;c 2 k ;c
by the bounded convergence theorem. This completes the proof.

01 + p01 p ;c  ;c 1 0 mp m  1 0 p01 + p0 1  ;c 1 0 p01 + p01  ;c m m m m 01 )2 + 2 p01 (1 0 p01 )p ;c  ;c + p01 2  ;c  ;c (1 0 pm m m m = 01 )2 + (1 0 p01 ) p01 ( ;c +  ;c ) + p01 2  ;c  ;c (1 0 pm m m m 1 with equality held only when RST = 1 holds.
2

286

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 52, NO. 1, JANUARY 2006

The remaining task is to show that RST 01 by , thus In the limit, we can replace pm

 RSSINR

in the limit.

 RSSINR 2 2 4 +  ;c 0 )2 + 2 (1 0 ) ;c () (1 0 )(1 2 + (1 0 )( ;c +  ;c ) + 2 ( ;c  ;c ) 4  ;c   ;c  ;c 4 ()(1 0 )  ;c ;c 0  ;c 2 2   ;c  ;c ( ;c +  ;c ) 0 2 ;c  ;c


RST
2 2 which is equivalent to c ~  0, after pages of tedious algebra. This completes the proof for RST  RSSINR , in the limit. See the technical report [23] for details.

ACKNOWLEDGMENT The authors wish to thank Persi Diaconis for reading the original draft of this manuscript and Amir Dembo for the helpful advice in the revision. Ping Li thanks Tze Leung Lai and Youngjae Kim for the enjoyable conversations. He also acknowledges Antonia Tulino, Sergio Verd, and Dongning Guo for the e-mail communications during the preparation of the manuscript.

REFERENCES
[1] A. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless Communications, 1st ed. New York: Cambridge Univ. Press, 2003. [2] H. V. Poor and S. Verd, Probability of error in MMSE multiuser detection, IEEE Trans. Inf. Theory, vol. 43, no. 3, pp. 858871, May 1997. [3] D. N. C. Tse and O. Zeitouni, Linear multiuser receivers in random environments, IEEE Trans. Inf. Theory, vol. 46, no. 1, pp. 171188, Jan. 2000. [4] S. Verd, Multiuser Detection. Cambridge, U.K.: Cambridge Univ. Press, 1998. [5] U. Madhow and M. L. Honig, MMSE interference suppression for direct-sequence spread-spectrum CDMA, IEEE Trans. Commun., vol. 42, no. 12, pp. 31783188, Dec. 1994. [6] S. Verd and S. Shamai (Shitz), Spectral efciency of CDMA with random spreading, IEEE Trans. Inf. Theory, vol. 45, no. 2, pp. 622640, Mar. 1999. [7] A. Hjrungnes and P. S. R. Diniz, Minimum BER prelter transform for communications systems with binary signaling and known for MIMO channel, IEEE Signal Process. Lett., vol. 12, no. 3, pp. 234237, Mar. 2005. [8] D. Guo, S. Verd, and L. K. Rasmussen, Asymptotic normality of linear multiuser receiver outputs, IEEE Trans. Inf. Theory, vol. 48, no. 12, pp. 30803095, Dec. 2002. [9] A. M. Tulino and S. Verd, Random matrix theory and wireless communications, Foundations and Trends in Communications and Information Theory, vol. 1, Jun. 2004. [10] L. Li, A. M. Tulino, and S. Verd, Asymptotic eigenvalue moments for linear multiuser detection, Commun. Inf. and Syst., vol. 1, no. 3, pp. 273304, 2001. [11] D. N. C. Tse and S. V. Hanly, Linear multiuser receivers: Effective interference, effective bandwidth and user capacity, IEEE Trans. Inf. Theory, vol. 45, no. 2, pp. 641657, Mar. 1999. [12] J. Zhang, E. K. P. Chong, and D. N. C. Tse, Output MAI distribution of linear MMSE multiuser receivers in DS-CDMA systems, IEEE Trans. Inf. Theory, vol. 47, no. 3, pp. 11281144, Mar. 2001. [13] D. Gesbert, Robust linear MIMO receivers: A minimum error-rate approach, IEEE Trans. Signal Process., vol. 51, no. 11, pp. 28632871, Nov. 2003. [14] A. Hjrungnes and D. Gesbert, Exact SER-precoding of orthogonal space-time block coded correlated MIMO channels: An iterative approach, in Proc. 6th Nordic Signal Processing Symp., NORSIG 2004, Espoo, Finland, Jun. 2004, pp. 336339.

[15] E. Biglieri, Transmission and reception with multiple antennas: Theoretical foundations, Foundations and Trends in Communications and Information Theory, vol. 1, p. 222, 2004. [16] R. R. Mler, On the asymptotic eigenvalue distribution of concatenated vector-valued fading channels, IEEE Trans. Inf. Theory, vol. 48, no. 7, pp. 20862091, Jul. 2002. [17] R. R. Mller, P. Schramm, and J. B. Huber, Spectral efciency of CDMA systems with linear interference suppression, in IEEE Workshop on Communication Engineering, Ulm, Germany, Jan. 1997, pp. 9397. [18] K. Mardia, J. T. Kent, and J. Bibby, Multivariate Analysis. San Diego, CA: Academic, 1979. [19] D. Gore, R. W. Heath, and A. Paulraj, On performance of the zero forcing receiver in presence of transmit correlation, in Proc. IEEE Int. Symp. Information Theory Lausanne, Switzerland, Jun./Jul. 2002, p. 159. [20] J. W. Silverstein, Strong convergence of the empirical distribution of eigenvalues of large dimensional random matrices, J. Multivariate Anal., vol. 55, no. 2, pp. 331339, 1995. [21] Z. D. Bai, Methodologies in spectral analysis of large dimensional random matrices, a review, Statist. Sinica, vol. 9, no. 3, pp. 611677, 1999. [22] Z. D. Bai and W. Silverstein, CLT for linear spectral statistics of largedimensional sample covariance matrices, Ann. Probab., vol. 32, no. 1A, pp. 553605, 2004. [23] P. Li, D. Paul, R. Narasimhan, and J. Ciof, On the Asymptotic and Approximate Distribution of SINR for the Linear MMSE MIMO Receiver, Dept. Statist., Stanford Univ., Stanford, CA, Tech. Rep. 2005-15, Jun. 2005. [24] H. Shin and J. H. Lee, Capacity of multiple-antenna fading channels: Spatial fading correlation, double scattering and keyhole, IEEE Trans. Inf. Theory, vol. 49, no. 10, pp. 26362647, Oct. 2003. [25] R. Narasimhan, Spatial multiplexing with transmit antenna and constellation selection for correlated MIMO fading channels, IEEE Trans. Signal Process., vol. 51, no. 11, pp. 28292838, Nov. 2003. [26] R. Durrett, Probability: Theory and Examples, 2nd ed. Belmont, CA: Duxbury, 1995. [27] P. Hougaard, Survival models for heterogeneous populations, Biometrika, vol. 73, no. 2, pp. 387396, 1986. [28] H. U. Gerber, From the generalized gamma to the generalized negative binomial distribution, Insurance: Mathematics and Economics, vol. 10, pp. 303309, 1991. [29] I. S. Gradshteyn and I. M. Ryzhik, Tables of Integrals, Series, and Products, 2nd ed. New York: Academic, 1980. [30] M. S. Alouini and A. J. Goldsmith, A unied approach for calculating error rates of linearly modulated signals over generalized fading channels, IEEE Trans. Commun., vol. 47, no. 9, pp. 13241334, Sep. 1999. [31] M. K. Simon and M.-S. Alouini, Digital Communication Over Fading Channels: A Unied Approach to Performance Analysis, 1st ed. New York: Wiley, 2000. [32] A. Guionnet and O. Zeitouni, Concentration of the spectral measure for large matrices, Electron. Commun. Probab., pp. 119136, 2000.

Оценить