Vous êtes sur la page 1sur 8

Capacity of orthogonal space-time block codes in MISO fading channels with

co-channel interference and noise


YUE DianWu and WANG Qian
Citation: Science in China Series F-Information Sciences 52, 1697 (2009 ); doi: 10.1007/s11432-009-0145-
2
View online: http://engine.scichina.com/doi/10.1007/s11432-009-0145-2
View Table of Contents:http://engine.scichina.com/publisher/scp/journal/Sci China Info Sci F/52/9
Published by the Science China Press

Articles you may be interested in

Differential space-time block-diagonal codes


Science in China Series F-Information Sciences 50, 747 (2007);

Performance of multiuser CDMA system withspace-time block coding in MIMO Rayleigh fading
channels
Science in China Series F-Information Sciences 52, 2296 (2009);

Robust linear receivers for STBC systems with unknown co-channel interference
Science in China Series F-Information Sciences 52, 1067 (2009);

Capacity bounds of transmit beamforming over MISO time-varying channels with imperfect feedback
SCIENCE CHINA Information Sciences 53, 1417 (2010);

Performance analysis of a noise-normalized FFH/MFSK receiver over Rayleigh fading channels with
partial-band noise jamming
SCIENCE CHINA Information Sciences 54, 129 (2011);
www.scichina.com
info.scichina.com
www.springerlink.com

Capacity of orthogonal space-time block codes in


MISO fading channels with co-channel
interference and noise
YUE DianWu1,2 & WANG Qian1
1 College of Information Engineering, Dalian Maritime University, Dalian 116026, China;
2 National Mobile Communications Research Laboratory, Southeast University, Nanjing 210096, China
Orthogonal space-time block codes (OSTBCs) are an efficient mean in order to exploit the diversity
offered by the wireless multiple-input multiple-output (MIMO) channel. This paper considers capacity
problems of OSTBCs over spatially correlated multiple-input single-out (MISO) Rayleigh fading channels
in the presence of spatially correlated Rayleigh co-channel interference and additive Gaussian noise,
and derives exact expressions of the ergodic capacity and outage probability (capacity distribution) for
such OSTBCs. Some numerical examples are given to illustrate the effect of co-channel interference on
the ergodic and outage capacity of OSTBCs.

multiple-input single-output, orthogonal space-time block code, capacity, co-channel interference, noise

1 Introduction
users share the same radio spectrum, which typi-
Orthogonal space-time block coding (OSTBC) is cally causes co-channel interference. However, the
an open-loop transmit diversity scheme that decou- capacity analysis of OSTBCs in the presence of
ples the multiple-input multiple-output (MIMO) both co-channel interference and additive Gaussian
channel, thereby reducing the space-time decod- noise is not available in the literature[9,10] , although
ing into a scalar detection process. This char- its counterpart with Gaussian noise alone has been
acteristic of OSTBC makes it a powerful tool, thoroughly investigated.
achieving full diversity over MIMO fading chan- In this paper, we investigate the capacity of OS-
nels, and requiring little computational cost for TBCs through spatially correlated multiple-input
both the encoding and decoding processes[13] . single-out (MISO) Rayleigh channels with spatially
Therefore, orthogonal space-time block codes (OS- correlated co-channel Rayleigh interference and ad-
TBCs) have captured much interest in recent years. ditive noise, and derive exact expressions of the
The capacity issues for OSTBCs have been studied ergodic capacity and outage probability for such
widely, yet with main focus on the scenario with- OSTBC-MISO systems. We also provide some nu-
out interference[48] . In cellular systems, multiple merical examples in order to observe the impact
Received October 23, 2008; accepted January 15, 2009
doi: 10.1007/s11432-009-0145-2

Corresponding author (email: dwyue@newmail.dlmu.edu.cn)


Supported by the National Natural Science Foundation of China (Grant No. 60672030), and the Open Research Fund of National Mobile
Communications Research Laboratory of Southeast University (Grant No. W200810)

Citation: Yue D W, Wang Q. Capacity of orthogonal space-time block codes in MISO fading channels with co-channel interference and noise.
Sci China Ser F-Inf Sci, 2009, 52(9): 16971703, doi: 10.1007/s11432-009-0145-2

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


of the co-channel interference on the ergodic and for the interferers, either. Our above-mentioned as-
outage capacity of OSTBCs. sumptions are consistent with that of the capacity
analysis of MIMO channels in the presence of co-
2 Channel model and formulation channel interference in refs. [1416]. Under these
assumptions, the optimal transmission strategy is
Consider a MISO system with t transmit anten- the equal-power allocation approach[13] . As such,
nas and one receive antenna operating in a co- we can assume that Rs = 1t I t .
channel interference environment, such that the Although there are some assumptions, our re-
intended t 1 signal vector s is corrupted by search results will be applied suitably in mobile cel-
interfering users. We assume that the ith inter- lular systems or cooperative relay networks[1419] ,
ferer is equipped with ti antennas thereby forming where the desired user inevitably will suffer co-
a ti 1 interfering vector si where i = 1, . . . , . channel interference from other users. Classically,
Since a scalar receiver is considered, we can repre- the transmitter with an antenna array, or the
sent all interfering vectors, equivalently, by a sin- source and several cooperative relays (whose an-
gle tI vector sI , such that sI = [s1 , , s ]T and tennas can form together a virtual antenna ar-
tI = t1 + + t [11] . Here and hereafter, we will ray although each node is only equipped with one
use superscripts T and to denote transposition antenna[18] ) transmit OSTBC signals to the des-
and conjugate transposition, respectively. The re- tination. The destination may be a base station
ceived signal, in complex baseband form, can be or a access point, can know precisely the CSI for
thus written as the desired user via a train sequence, and can also
know precisely the CSI for the interferes, which is
v = s h s + I hI sI + n, (1)
provided by other base stations or access points.
where n is zero-mean additive white Gaussian noise The transmitter may not have any CSI.
with unit variance; symbolically, we can denote Space-time block codes have been introduced to
n CN (0, 1). We assume that Rs = E(ss ), provide transmit diversity in wireless fading chan-
and sI CN (0, RI ) to have trace(Rs ) = 1 and nels using multiple transmit antennas and provide
trace(RI ) = 1, so that s represents the transmit the same diversity orders as maximal-ratio receiver
signal to noise ratio (SNR) and I the interference combing[57] . By decoupling the vector detection
to noise ratio (INR). We consider Rayleigh fading problem into simpler scalar detection problems, the
in this paper. We can write the distribution of effective STBC channel is a scalar additive white
the channel gain vector for the intended signal as Gaussian noise (AWGN) channel with a gain pro-
h CN (0, ) and its counterpart for the interfer- portional to the Frobenius norm of the matrix
ence gain vector as hI CN (0, I ). The covari- channel[5] . Let rs denote the rate of the OSTBC,
ance of the interference-plus-noise conditioned on and k k2 denote the squared Frobenius norm.
hI is Then we have from ref. [5] that the SINR at de-
R = 1 + I hI RI hI . (2) tection can be expressed as
Therefore, from refs. [12, 13] we have the following 1
= k s R1 h k2 . (4)
mathematically equivalent model to eq. (1): rs
Note that
v = s R1 h s + n.
p
(3)
k s R1 h k2 = s h hR1 . (5)
We assume that the receiver has perfect knowl-
Thus the SINR at detection can be rewritten as
edge about the channel state information (CSI) for
s x
the desired user, and also has perfect knowledge = , (6)
(1 + I y)
about CSI for the interferers, but the transmit-
ter has no knowledge about the CSI for the de- where s = s
rs t
, I = I
tI
, x = h h, and y =
sired user, and has no knowledge about the CSI hI (tI RI )hI .

1698 YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


" 1 2 #
3 Capacity analysis of OSTBC-MISO sys-

I 1 I 1
tems + + I + . (9)
pi s qk pi s qk

3.1 Probability density functions of SINR B. All {pi } are identical and all {qi } are iden-
The capacity analysis based on the SINR in eq. tical. When some elements of the eigenvalue set
(6) involves two random variables x and y. To de- {p1 , p2 , , pt } are equal and/or some ones of the
termine their probability density functions (PDF) eigenvalue set {q1 , q2 , , qtI } are equal, we ob-
fx (x) and fy (y), let {pi , i = 1, , t} and {qi , i = tain the results as limiting case on the right of
1, , tI } denote the eigenvalues of and tI I RI , eq. (8) via LHospitals rule. In particular, when
respectively. The PDF of x and y can be repre- p1 = p2 = = pt and q1 = q2 = = qtI , eq. (8)
sented in terms of these eigenvalues. The different can be simplified to
treatment will be used for different cases of {pi } ex xt1
fx (x) = ,
and {qi }. (t)
Since (x, y) is related to by the the transform (10)
ey y tI 1
defined in eq. (6), we can use the Jacobian method fy (y) = .
(tI )
to determine the PDF of yielding
Z After inserting eq. (10) and simplifying by using
f () = f,z (, z)dz eq. (35) of ref. [21], eq. (7) produces
z>0
t  
t1 e/s X t I j (j + tI )
 
(1 + I z)
Z
= fx f () = t .
z>0 s s (t)(tI ) j=0 j ( I + 1)(j+tI )
s
fy (z)(1 + I z) (11)
dz. (7)
s It is easy to obtain the following results in a sim-
A. All {pi } are distinctive and all {qi } are dis- ilar way.
tinctive. For the general case in which all {pi } are C. All {pi } are distinctive and all {qi } are iden-
distinctive and all {qi } are distinctive, invoking the tical.
results in ref. [20] allows us to obtain (1)t1 X pt2
t
e pi s
i
t f () =
pt2 e(x/pi )
Q
X s j6=i (pj pi )
fx (x) = (1)t1 Qi , i=1
tI (tI +1) 
j6=i (pj pi )
 
i=1 I I
(8) +1 + tI I +1 .
tI pi s pi s
tI 1
X q tI 2 e(y/qk ) (12)
fy (y) = (1) Qk .
k=1 b6=k (qb qk ) D. All {pi } are identical and all {qi } are distinc-
By inserting eq. (8) and simplifying, eq. (7) tive.

produces (1)tI 1 X
tI
qktI 2 t1 e s
t (I z+1) f () =
s t (t) k=1 b6=k (qb qk )
Q
I z + 1 pt2 e pi s
Z X
t1 i
f () = (1) Q
z>0 s i=1 j6=i (pj pi ) X t  
t
tI 1 I j (j + 1)
X q tI 2 ezqk j
(1)tI 1 Qk dz j=0

b6=k (qb qk )
 (j+1)
k=1 I 1
t X
tI + . (13)
X (1)t+tI 2 pt2 q tI 2 s qk
= Q Qi k
j6=i (pj pi ) b6=k (qb qk )
i=1 k=1
Z 3.2 Ergodic capacity of OSTBC-MISO
I 1
s 1 e pi s (I z + 1)e( pi s + qk )z dz systems
z>0 The ergodic capacity of OSTBC-MISO systems is
t tI
1 XX (1)t+tI pt2
i qktI 2 e pi s given by
= Q Q
s i=1 k=1 j6=i (pj pi ) b6=k (qb qk ) C = rs E[log2 (1 + )]. (14)

YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703 1699

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


Now define the following integrate functions the outage capacity[22] . The channel capacity de-
Z
fined above is a random variable. The outage ca-
I(c, a, n) = ex/c (1 + ax)n
0
pacity is the value which the capacity random vari-
log2 (1 + x)dx, (15) able falls below with a preset probability (outage
Z m1
x log2 (1 + x) probability), say w. Let c = 2w 1. Mathemati-
J(c, a, n, m) = dx. (16) cally, we can write
0 e (1 + ax)n
x/c
Z
Having obtained the PDF of , then we can di- Pout (w) = f ()d. (21)
rectly calculate the ergodic capacity for the four <c

different cases of {pi } and {qi }, as shown by the Now we give the expressions of the outage prob-
following equations. ability according to the four different cases.
A. All {pi } are distinctive and all {qi } are dis- A. All {pi } are distinctive and all {qi } are dis-
tinctive. tinctive.
t
rs X X
tI
(1)tI +t2 pt2 q tI 2 Pout (w)
C= Q Qi k Z
s i=1 k=1 j6=i (pj pi ) b6=k (qb qk ) = f ()d
   <c
qk
qk I s pi , I , 1
Z Z
s pi = fy (y) fx (x)dxdy
  y>0 x< c (I y+1)
2 I qk s

+ qk I I s pi , ,2 . (17) Z Z t
(1)t1 ex/pi fy (y)
s pi =
X
dxdy
p2t j6=i (pj pi )
Q
B. All {pi } are identical and all {qi } are identical. y>0 x<
c(I y+1)
s i=1 i
t   i I t c 1
t rs I (i + tI )J(s , s , i + tI , t) pt1 {e[ s (I y+1)]pi 1}
Z X
i
X
C= . = Q fy (y)dy
i=0
i s t (t)(tI ) y>0 i=1 (1)t j6=i (pj pi )
(18) t X
X tI
(1)tI +t1 pt1 q tI 2
C. All {pi } are distinctive and all {qi } are iden- = Q Qi k
j6=i (pj pi ) b6=k (qb qk )
tical. i=1 k=1
Z
t 1
pt2
c 1
rs (1)t1 X i {e[ s (I y+1)]pi 1}eyqk dy
C= Q 0
s i=1 j6=i j pi )
(p
tI
t X
   X (1)tI +t pt1
i qktI 1
I =
I s pi , , tI
Q Q
pi i=1 k=1 j6=i (pj pi ) b6=k (qb qk )
 s ( c
)
pi e pi s

I
+ tI I I s pi , , tI + 1 . (19) 1 s . (22)
s pi I cqk + s pi
D. All {pi } are identical and all {qi } are distinc- B. All {pi } are identical and all {qi } are identical.
tive. Pout (w)
tI
rs (1)tI 1 X qktI 2 tI
ec/s

C= s
s t (t) k=1 b6=k (qb qk ) =1
Q
(tI ) s + cI
t   t i1 
t X (c/ )i1 X 

X
I i (i + 1)qki+1 s i1
i (tI + k 1)!
i=0 i=1
(i 1)! k=0 k
 
I qk 
s I
k
J s , , i + 1, t . (20) . (23)
s s + cI
3.3 Outage probability of OSTBC-MISO C. All {pi } are distinctive and all {qi } are iden-
systems tical.
c

We have studied the ergodic capacity for MISO


t
pt1
i {e pi s ( c +
s pi
)tI 1}
s pi
X
Pout (w) = Q I
. (24)
channels. Another useful quantity of relevance is i=1
(1)t j6=i (pj pi )

1700 YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


D. All {pi } are identical and all {qi } are distinc- of correlation for interferers on the outage proba-
tive. bility is not serious.
tI 1 t
(1)tI 1 qktI 1 e qk I

X s
Pout (w) = Q 1+ .
k=1 b6 = k (q b qk ) cI q k
(25)

4 Numerical examples and remarks


To examine the theory with numerical results, we
consider the case where a single antenna is used to
receive the desired signal transmitted from t = 4
antennas, corrupted by several interferers each em-
ploying 2 transmit antennas. We make use of
an OSTBC with code rate rs = 1/2 for the de-
sired user. For simplicity, we adopt the correla-
tion model of exponential type, for which the cor-
relation between antennas p and q is defined by
C(p, q) = r |pq| exp(j|p q|/12). Physically, r |pq|
denotes the correlation magnitude. Moreover, we Figure 1 Ergodic capacity versus SIR for various interfering
let RI = t1I I tI . antenna configurations.
Here we first consider the general case A. The
correlation coefficient for the desired user is fixed
to be r = 0.5 for Figures 14. The ergodic capacity
versus the SIR (10 log 10 ( ((Is/t/t)I ) )) is depicted in Fig-
ure 1, where the three curves are shown for num-
bers of total interfering transmit antennas equal
to tI = 2, 4, 6, respectively. As expected, the er-
godic capacity decreases with increasing tI . Figure
2 depicts the ergodic capacity versus the SIR for
three different correlations with one interferer. The
three curves in Figure 2 are shown for interfering
correlation coefficients equal to r=0.05, 0.70, 0.95,
respectively. It can be seen from Figure 2 that the
impact of correlation for interferers on the ergodic
capacity is quite small.
Figure 3 shows the outage probability as a func-
tion of the preset capacity protection value w. In
Figure 2 Ergodic capacity versus SIR for different interfering
Figure 3, the three curves correspond to the num-
transmit correlations.
ber of total interfering transmit antennas tI =
2, 4, 6, respectively. Again, we observe that the Figure 5 is a figure that enlarges the part of Fig-
outage performance drops as tI increases. Figure ure 4 for 0 6 w 6 4. There is a performance
4 depicts the outage probability versus w for three threshold, denoted w0 , whose value is about 2.7.
different correlations with one interferer. The three It should be noticed that when w > w0 , the sys-
curves in Figure 4 are shown for the interfering tem performance becomes better with increasing r,
correlation coefficient r = 0.05, 0.70, 0.95, respec- but when w < w0 , the system performance is worse
tively. It can be seen from Figure 4 that the effect with increasing r.

YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703 1701

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


Figure 3 Outage probability versus the preset capacity thresh- Figure 5 A threshold of outage probability for different inter-
old for various interfering antenna configurations. fering transmit correlations.

Figure 4 Outage probability versus the preset capacity thresh-


Figure 6 Outage probability versus the preset capacity thresh-
old for different interfering transmit correlations.
old for different cases.

Simulation results are included in Figures 14 for for these cases. Just as shown in Figure 6, the
comparison. Each point in the simulation curves theoretical and simulation results are also nearly
are obtained by averaging over 100000 independent identical for cases B, C and D.
computer runs. The theoretical and simulation re- Theoretical and simulation results with case A
sults are nearly identical verifying the validity of are included in Figure 6 for comparison. In the fig-
the theory for case A. ure, the correlation coefficient for the desired user
In order to verify the validity of the theory for is r = 0.9 or r = 0. The number of interferers is
cases B, C and D, we have done some simulations 3, each employing 2 transmit antennas. For cases

1702 YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2


A and D, their correlation coefficients are 0.2, 0.4, and on the outage probability are not serious.
and 0.6, respectively. For cases B and C, however,
their correlation coefficients are all 0. By observ-
5 Conclusions
ing Figure 6 and a great deal of simulation data,
we can have the following conclusions: In this paper, we investigate the capacity of OS-
a) When the correlation coefficient for the de- TBCs over MISO correlated Rayleigh channels in
sired user increases, the system performance be- the presence of both correlated Rayleigh co-channel
comes worse; interference and additive Gaussian noise. Exact
b)when the correlation coefficient for the inter- solutions to the ergodic capacity and outage prob-
ferers increases, however, the system performance ability for OSTBCs are given, and the effects of co-
becomes better; channel interference on the capacity are observed
c) the effects of correlation on ergodic capacity by simulation.

1 Alamouti S M. A simple transmit diversity technique for wire- ternational Conference on Acoustics, Speech, and Signal Pro-
less communications. IEEE J Sel Area Commun, 1998, 16: cessing. Pacific Grove, California, USA, 2005. 11531156
14511458 12 Goldsmith A, Jafar S A, Jindal N, et al. Capacity limits of
2 Tarokh V, Jafarkhani H, Calderbank A R. Space-time block MIMO channels. IEEE J Sel Area Commum, 2003, 21(5):
codes from orthogonal designs. IEEE Trans Inf Theory, 1999, 684702
45: 14561467 13 Farrokhi F R, Foschini G J, Lozano A, et al. Link-optimal
3 Ganesan G, Stoica P. Space-time block codes: a maximum space-time processing with multiple transmit and receive an-
SNR approach. IEEE Trans Inf Theory, 2001, 47: 16501656 tennas. IEEE Commun Lett, 2001, 5: 8587
4 Zhang H, Gulliver T A. Capacity and error probability analysis 14 Blum R S, Winters J H, Sollenberger N R. On the capacity of
for orthogonal space-time block codes fading channels. IEEE cellular systems with MIMO. IEEE Commun Lett, 2002, 6(6):
Trans Wirel Commun, 2005, 4: 808819 242244
5 Sandhu S, Paulraj A. Space-time block codes: a capacity per- 15 Blum R S. MIMO capacity with interference. IEEE J Sel Area
spective. IEEE Commun Lett, 2003, 4: 384386 Commun, 2003, 21(5): 793801
6 Musavian L, Dohler M, Nakhai M R, et al. Closed-form capac- 16 Kang M, Yang L, Alouini M -S. Capacity of MIMO channels in
ity expressions of orthogonalized correlated MIMO channels. the presence of co-channel interference. Wirel Commun Mobile
IEEE Commun Lett, 2004, 8: 365367 Comput, 2007, 7(1): 113125
7 Dohler M, Aghvami H. Information outage probability of dis- 17 Pabst R. Relay-based deployment concepts for wireless and
tributed STBCs over nakagami fading channels. IEEE Com- mobile broadband radio. IEEE Commun Mag, 2004, 42: 80
mun Lett, 2004, 8: 437439 89
8 Maaref A, Aissa S. Performance analysis of orthogonal space- 18 Jing Y, Hassibi B. Distributed space-time coding in wireless
time block codes in spatially correlated MIMO Nakagami relay networks. IEEE Trans Wirel Commun, 2006, 5: 3524
fading channels. IEEE Trans Wirel Commun, 2006, 5: 807 3536
817 19 Scaglione A, Goeckel D L, Laneman J N. Cooperative com-
9 Li L, Li H, Yao Y -D. Joint channel estimation and co-channel munications in mobile ad hoc networks. IEEE Signal Process
interference suppression for space-time block coded systems. Mag, 2006, 23: 1829
In: The Thirty-Fifth Asilomar Conference on Signals, Systems 20 Mathai A M, Provost S B. Quadratic Forms in Random Vari-
and Computers, Philadelphia, PA, USA, 2001, 2: 989993 ables: Theory and Applications. New York: Dekker, 1992
10 Larsson E G, Stoica P, Li J. Orthogonal space-time block 21 Shin H, Win M Z, Lee J H, et al. On the capacity of dou-
codes: maximum-likelihood detection for unknown channels bly correlated MIMO channels. IEEE Trans Wirel Commun,
and unstructured interferences. IEEE Trans Signal Process, 2006, 5(8): 22532265
2003, 51: 362372 22 Simon M K, Alouini M -S. Digital Communication over Fading
11 Ye S, Blum R S. Some properties of the capacity of MIMO Channels: A Unified Approach to Performance Analysis. New
systems with co-channel interference. In: Proc 2005 IEEE In- York: Wiley, 2000

YUE D W et al. Sci China Ser F-Inf Sci | Sep. 2009 | vol. 52 | no. 9 | 1697-1703 1703

Downloaded to IP: 134.206.250.166 On: 2017-04-07 22:54:19 http://engine.scichina.com/doi/10.1007/s11432-009-0145-2

Vous aimerez peut-être aussi