Sie sind auf Seite 1von 6

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO.

5, MAY 2008 1839

An Alamouti Coded OFDM Transmission for Cooperative Systems


Robust to Both Timing Errors and Frequency Offsets
Zheng Li and Xiang-Gen Xia

Abstract— In this letter, we propose an Alamouti coded or- and frequency offsets. For combating timing errors, we use
thogonal frequency-division multiplexing (OFDM) scheme for a OFDM as the modulation technique. With a cyclic prefix (CP)
cooperative communication system robust to both timing errors insertion, OFDM is robust to timing errors. However, OFDM
and frequency offsets. OFDM with cyclic prefix (CP) is used
to combat timing errors. In order to mitigate the intercarrier is sensitive to frequency offsets. Frequency offsets may destroy
interference (ICI) caused by multiple frequency offsets in the co- the orthogonality of the subcarriers and result in intercarrier
operative system, an ICI-self cancellation scheme is constructed, interference (ICI), which may severely degrades the system
which can suppress ICI effectively. Moreover, in the proposed performance. Many methods have been proposed to mitigate
scheme, if the channels are real-valued fading channels, the ICI for OFDM systems in the literature, among which the
received signals at the destination node have the Alamouti code
structure on each subcarrier and thus it has the fast symbol-wise ICI self-cancellation [14] or polynomial cancellation coding
ML decoding and when frequency offsets are not large, the new [15] method is attractive for its simplicity and particularly
scheme can achieve diversity order 2. interesting when there are multiple frequency offsets as in
Index Terms— Alamouti code, asynchronous cooperative di- cooperative systems as we shall see later. The rationale of
versity, multiple frequency offsets, OFDM. this method is to map one transmitted data onto a group
of subcarriers with proper coefficients. By doing so, the
I. I NTRODUCTION ICI signals generated within the same group can be ”self-
cancelled.” Later, in [16] a variation of ICI self-cancellation
PACE-time coding [1], [2] is an effective technique to
S exploit spatial diversity through multiple antennas in a
MIMO system and multiple transmissions from multiple nodes
scheme by using symmetric data conjugation is proposed. In
[17], it is proposed to combine ICI self-cancellation scheme
with space frequency codes to achieve full diversity as well
in a cooperative system [3], [4]. Different from a MIMO
as mitigate ICI in MIMO-OFDM systems. Based on these
system, in a cooperative system multiple transmitters/antennas
ICI cancellation methods, in this letter, we combine the ICI
are in different locations and have their own oscillators.
self-cancellation scheme proposed in [16] with an OFDM and
This means that the transmissions from different nodes in
Alamouti code in the asynchronous cooperative system to
a cooperative system may have different timing errors and
also combat the frequency offsets from the different nodes.
also different frequency offsets. A natural question is how a
Thus, our proposed new scheme is robust to timing errors
space-time code is designed in such a cooperative system with
and ICI can be reduced effectively. Moreover, when channels
both timing errors and frequency offsets to achieve the spatial
are real-valued fading channels, the received signals have the
diversity. While most of the existing studies on space-time
Alamouti code structure on each subcarrier and thus it has the
coded cooperative systems have assumed the synchronizations
fast symbol-wise ML decoding. Note that in [11], only timing
in both time and frequency, i.e., no timing errors or frequency
errors are concerned and the scheme is amplify-and-forward
offsets, there are a few space-time coding schemes that can
(AF) based and OFDM is implemented at the source node,
achieve spatial diversity when there are timing errors [5],
while in this paper, both timing errors and frequency offsets
[6], [7], [8], [9], [10], [11] and equalization techniques are
are considered and the scheme is decode-and-forward (DF)
used to combat multiple frequency offsets [12], [13]. To our
based and OFDM is implemented at the relay nodes as we
best knowledge, there is no space-time coding schemes in the
shall see later.
literature for a cooperative system that achieves full spatial
The remaining of this letter is organized as follows. In
diversity and possesses fast ML decoding and moreover is
Section II, the system model with both timing errors and
robust to both multiple timing errors and multiple frequency
multiple frequency offsets and Alamouti code is described.
offsets.
In this letter, we propose an Alamouti coded orthogonal The ICI self-cancellation scheme is proposed in Section III.
frequency-division multiplexing (OFDM) scheme for a coop- Simulation results are presented in Section IV. Finally the
erative communication system robust to both timing errors conclusions are given in Section V.

Manuscript received April 27, 2007; revised June 7, 2007; accepted August
30, 2007. The associate editor coordinating the review of this letter and
II. S YSTEM M ODEL WITH T IMING E RRORS AND
approving it for publication was C. Tellambura. This work was supported F REQUENCY O FFSETS
in part by the Air Force Office of Scientific Research (AFOSR) under Grant
No. FA9550-05-1-0161 and the National Science Foundation under Grant Consider a cooperative system with one source node, one
CCR-0325180. destination node and two relay nodes, as shown in Fig. 1.
The authors are with the Department of Electrical and Computer En- Every node in the system has only one antenna. To transmit
gineering, University of Delaware, Newark, DE 19716 (e-mail: {zhli,
xxia}@ee.udel.edu). the information from the source node S to the destination node
Digital Object Identifier 10.1109/TWC.2008.070443. D, there undergo two phases. In the first phase, the source node
1536-1276/08$25.00 
c 2008 IEEE

Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.
1840 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 5, MAY 2008

where Wk,i , i=1, 2, are the corresponding AWGN with zero-


R1 h1 mean and unit-variance.
When there exist timing errors in the system, without loss
S D of generality, we assume the signals from R2 arrive at the
h2 destination node τ samples later than the signals from R1 ,
i.e., τ1 = 0 and τ2 = τ in (1). If the length cp of CP is not
R2 less than τ , we can still maintain the orthogonality between
Phase 1 Phase 2 the subcarriers [9]. The delay in the time domain corresponds
to a phase change in the frequency domain:
Fig. 1. Cooperative system architecture. f τ = [1, e−j2πτ /N , . . . , e−j2πτ (N −1)/N ]T

with f = [1, e−j2π/N , . . . , e−j2π(N −1)/N ]T .


S broadcasts the information to the two relay nodes Ri , i=1, Now the received signals with timing errors for two suc-
2. Meanwhile, the two relay nodes receive the information. cessive symbol durations on subcarrier k are
During the second phase, the two relay nodes Ri , i=1, 2, send
the received signals to the destination node D. In this letter, Rk,1 = H1 Xk,1 + fkτ H2 Xk,2 + Wk,1 , (4)

we adopt the decode-and-forward (DF) protocol, i.e., the two Rk,2 = −H1 Xk,2 + fkτ H2 Xk,1

+ Wk,2 (5)
relay nodes have correctly detected the information symbols
in the first phase. Thus, we only consider the transmission in where fkτ = exp(−j2πkτ /N ). From equations (4) and (5), we
the second phase. can see that timing errors may cause a frequency-flat channel
Assume the channels between relay nodes Ri and destina- to be frequency selective when Xk,1 = Xk,2 , however, OFDM
tion node D are quasi-static flat fading and their coefficients is robust to timing errors and Alamouti code structure on each
are denoted by hi , i=1, 2. When there exist both timing subcarrier is still maintained [9].
errors and multiple frequency offsets, the received signal at We now consider that there exist timing errors and fre-
the destination node in the time domain is quency offsets simultaneously. We denote ε1 , ε2 as the normal-
ized frequency offsets, i.e., the ratios of the actual frequency
r(t) = h1 ejθ1 (t) x1 (t − τ1 ) + h2 ejθ2 (t) x2 (t − τ2 ) + w(t) (1) offsets to the subcarrier spacing, of the two relays R1 and R2
where x1 (t) and x2 (t) are two signals transmitted from the to the destination, respectively. In the frequency domain, after
two relay nodes, θ1 (t) and θ2 (t) are the phase rotations caused the DFT processing, the received signals for two successive
by different frequency offsets, τ1 and τ2 are the different symbol durations on subcarrier k are
delays from the two relay nodes to the destination node, w(t) N
 −1
is the AWGN. We next design the two signals xi (t) for the Rk,1 = Q10 H1 Xk,1 + Q1k−l H1 Xl,1
transmissions from the two relay nodes. l=0,l=k
At the relay nodes, the detected symbols are distributedly N
 −1
encoded into Alamouti code first. Then the coded symbols are +Q20 fkτ H2 Xk,2 + Q2k−l flτ H2 Xl,2 + Wk,1 , (6)
OFDM modulated to N subcarriers. Then each OFDM symbol l=0,l=k
is preceded by a cyclic prefix (CP) of length cp . Finally the
OFDM symbols are transmitted from the relay nodes to the N
 −1
destination node. At the destination node, the CP is removed Rk,2 = a(−Q10 H1 Xk,2

+ −Q1k−l H1 Xl,2

)
first for each OFDM symbol. Then the received signals are l=0,l=k
demodulated by N-point DFT. N
 −1
Consider two consecutive symbols on subcarrier k, denoted +b(Q20 fkτ H2 Xk,1

+ Q2k−l flτ H2 Xl,1

) + Wk,2 (7)
by Xk,1 and Xk,2 (in the frequency domain). According to l=0,l=k
the Alamouti code structure, during the first symbol period,
Xk,1 is transmitted from R1 and Xk,2 is transmitted from R2 where
∗ sin(π(εi + k)) 1
on subcarrier k, and during the second symbol period, −Xk,2
∗ Qik = π exp[jπ(1 − )(εi + k)], i = 1, 2,
is transmitted from R1 and Xk,1 is transmitted from R2 on N sin( N (εi + k)) N
subcarrier k, where (·)∗ represents the complex conjugation. (8)
We assume that the channel coefficients H1 and H2 in the a = exp(j2πε1 T /N ) and b = exp(j2πε2 T /N ), (9)
frequency domain (Hi = hi since we assume channels are
flat fading) are constant during two OFDM symbol intervals. T is one OFDM symbol duration, i.e., T=N+cp .
Thus, at the destination node, after the CP removal and From equations (6) and (7), we can see that the Alamouti
DFT processing, under the assumption of perfect timing and code structure on subcarrier k does not hold when there exist

frequency synchronizations, i.e., θ1 (t) = θ2 (t) = 0 and frequency offsets, and the parts in the summation signs
τ1 = τ2 = 0 in (1), the received signals for two successive in (6) and (7) denote the ICI components on subcarrier k
symbol durations 1 and 2 on subcarrier k, 0 ≤ k ≤ N − 1, are caused by all other subcarriers of the two relay nodes. In the
next section, we propose an ICI self-cancellation method to
Rk,1 = H1 Xk,1 + H2 Xk,2 + Wk,1 , (2) mitigate ICI as well as keep the Alamouti code structure on
∗ ∗
Rk,2 = −H1 Xk,2 + H2 Xk,1 + Wk,2 (3) each subcarrier.
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 5, MAY 2008 1841

III. ICI S ELF -C ANCELLATION S CHEME C OMBINED WITH where we make use of the symmetric conjugate-negative 
A LAMOUTI C ODE property of Qik , and the parts in the summation signs are
We adopt the symmetric data-conjugate method in [16] to the self-cancelled ICI components on subcarrier k.
construct an ICI self-cancellation scheme. The main idea of Similarly, in the second OFDM symbol, when channels are
the symmetric data-conjugate method is to utilize the symmet- real-valued, the received signal on subcarrier k is
ric conjugate-negative property of ICI weighting function Qik N/2−1
in (8) [18]. The symmetric conjugate-negative property of Qik 

Rk,2 = a (−Xl,2 H1 Q1k−l − Xl,2 H1 Q1−(N −k−l) )
in (8) are as follows:
l=1
Qi0 + (Qi0 )∗ ≈ 2 and Qik + (Qi−k )∗ ≈ 0, for i = 1, 2. N/2−1


Notice that in the frequency domain phase change vector +b (Xl,1 flτ H2 Q2k−l + Xl,1 fN
τ 2
−l H2 Q−(N −k−l) )
l=1
f τ = [1, e−j2πτ /N , . . . , e−j2πτ (N −1)/N ]T , there exists a re-
N +Wk,2 (13)
lationship of fkτ = (fN τ ∗
−k ) , for k = 1, ..., 2 − 1, while
N
k = 0 and k = 2 , the relationship of conjugation does not and on subcarrier N − k it is
always hold. We will also make use of this property of f in
N/2−1
the following design. 
∗ ∗ ∗
We use the symmetric data-conjugate mapping as follows: RN −k,2 = a (−Xl,2 H1 (Q1l−k )∗ − Xl,2 H1 (Q1N −k−l )∗ )
during the first OFDM symbol, at the first relay node, we l=1
∗ N/2−1
map the information symbols as XN −k,1 = Xk,1 while at 
the second relay node, we map the information symbols as +b∗ ∗
(Xl,1 τ∗
fN 2 ∗ τ∗ 2 ∗
−l H2 (Ql−k ) + Xl,1 fl H2 (QN −k−l ) )

XN −k,2 = Xk,2 ; during the second OFDM symbol, at the first l=1

relay node, we map the symbols as XN −k,2 = −(Xk,2 ∗ ∗


) = +WN∗ −k,2 (14)
−Xk,2 while at the second relay node, we map the symbols
when ε1 = −ε2 , we have a = b∗ and Q10 = (Q20 )∗ . With this
as XN −k,1 = (Xk,1 ) = Xk,1 , k = 1, ..., N2 − 1; when k = 0
∗ ∗
N property in mind, without loss of generality we next centralize
and k = 2 , we do not send any information symbol. Thus,
the two different frequency offsets to satisfy ε1 = −ε2 . We can
the symbol rate is 12 − N1 .
set the receiver’s frequency to the mean value of the relays’
After the symmetric data-conjugate mapping, in the first
frequencies to satisfy this assumption if the receiver knows
OFDM symbol, the received signal on subcarrier k is
the two frequency offsets of the two relays, the centralization
N/2−1
 is as follows. Denote the frequency of R1 as fc + f1 and
Rk,1 = (Xl,1 H1 Q1k−l + Xl,1

H1 Q1−(N −k−l) ) the frequency of R2 as fc + f2 , where fc is the carrier
l=1
frequency, f1 and f2 are the frequency offsets of the two
N/2−1
 nodes, respectively. We assume the receiver knows the two
+ (Xl,2 flτ H2 Q2k−l + Xl,2
∗ τ
fN 2
−l H2 Q−(N −k−l) ) frequency offsets. We can then adjust the receiver’s frequency
l=1
to be fc + f1 +f
2
2
and at the destination node, the two frequency
+Wk,1 (10)
offsets consequently become f1 −f 2
2
and f2 −f 1
2 , respectively.
and on subcarrier N − k it is Thus, after the normalization, the assumption ε1 = −ε2 can
N/2−1 be satisfied.


RN −k,1 = (Xl,1 H1∗ (Q1l−k )∗ + Xl,1

H1∗ (Q1N −k−l )∗ ) After the centralization, we can combine (13) and (14) as
l=1 follows:
N/2−1
 ∗ ∗ τ ∗
+ τ∗
(Xl,2 fN ∗ 2 ∗ ∗ τ∗ ∗ 2 ∗
−l H2 (Ql−k ) + Xl,2 fl H2 (QN −k−l ) )
(Rk,2 + RN −k,2 )/2 = r(a)H1 (−Xk,2 ) + r(a)fk H2 Xk,1
l=1 N/2−1
1 
+WN∗ −k,1 . (11) + [(aQ1k−l + a∗ (Q1l−k )∗ )(−Xl,2

)H1
2
l=1,l=k

When channels are real-valued, we have Hi = Hi∗ , i = 1, 2, +(a∗ Q2k−l + a(Q2l−k )∗ )Xl,1

flτ H2 ]
and thus we can combine (10) and (11) as follows: N/2−1
1 

(Rk,1 + RN τ + [(aQ1−(N −k−l) + a∗ (Q1N −k−l )∗ )(−Xl,2 )H1
−k,1 )/2 = H1 Xk,1 + fk H2 Xk,2 2
l=1
N/2−1
1  +(a∗ Q2−(N −k−l) + a(Q2N −k−l )∗ )Xl,1 fN
τ
−l H2 ]
+ [(Q1k−l + (Q1l−k )∗ )Xl,1 H1
2 1
l=1,l=k + (Wk,2 + WN∗ −k,2 ) (15)
+(Q2k−l + (Q2l−k )∗ )Xl,2 flτ H2 ] 2
N/2−1 Δ
1  where r(a) = (aQ10 + a∗ (Q10 )∗ )/2.
+ [(Q1−(N −k−l) + (Q1N −k−l )∗ )Xl,1

H1 We rewrite (12) and (15) into
2
l=1
+(Q2−(N −k−l) + (Q2N −k−l )∗ )Xl,2
∗ τ
fN −l H2 ]

(Rk,1 + RN τ
−k,1 )/2 = H1 Xk,1 + fk H2 Xk,2
1 1
+ (Wk,1 + WN∗ −k,1 ) (12) +Nk,1 + (Wk,1 + WN∗ −k,1 ), (16)
2 2
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.
1842 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 5, MAY 2008

1 0
10

0.9

0.8
−1
10
0.7

0.6
|r(a)|

BER
−2
0.5 10

0.4

0.3 Sys.1, ε =0.25, ε =−0.25, BPSK


1 2
−3 Sys.2, ε1=0.25, ε2=−0.25, QPSK
10
Sys.1 ε1=0.05, ε2=−0.05, BPSK
0.2
Sys.2, ε1=0.05, ε2=−0.05, QPSK
Sys.1, no frequency offsets, BPSK
0.1

−4
0 10
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 2 4 6 8 10 12 14 16 18 20
ε Eb/N0(dB)

Fig. 2. |r(a)| vs. ε Fig. 3. BER comparison between Sys.1 with BPSK and Sys.2 with QPSK
when the frequency offset estimations are perfect

∗ ∗ τ ∗
(Rk,2 + RN −k,2 )/2 = r(a)H1 (−Xk,2 ) + r(a)fk H2 Xk,1
to mitigate ICI and achieve the cooperative diversity but the
1
+Nk,2 + (Wk,2 + WN∗ −k,2 ) (17) Alamouti code structure may not be maintained, i.e., the
2 symbol-wise fast ML decoding may not be held. If in this case,
where Nk,1 and Nk,2 are the remaining parts in the summation the mapping methods proposed in [14], [17] may be alternative
signs in (12) and (15), respectively, which denote the sup- choices. To construct the same scheme as in equations (16)
pressed ICI components and can be approximately considered and (17) without pre-compensation looks challenging and it is
as additional Gaussian noises. From (16) and (17), we see that left for the future research.
the received signals have the Alamouti code structure as well
as the ICI self-cancellation on each subcarrier. IV. S IMULATION R ESULTS
In practice, wireless channels are complex-valued. In order We next show some simulation results for our proposed
to obtain real-valued channels at the receiver, we may make scheme. In this simulation, we assume that N=64, and the
the following assumption. Each relay estimates its channel length of CP is 16. The delay τ is chosen randomly from 0
to the destination node first, then in each re-transmission, to 15 with the uniform distribution. The simulation assumes
the complex channel is pre-compensated by multiplying the perfect channel state information knowledge the destination
normalized complex conjugate of the channel estimation. node. We also assume that the receiver estimates the two
We use the Alamouti’s combining scheme to get the fol- frequency offsets and adjusts its frequency to satisfy ε1 =
lowing combined signals for detection: −ε2 . We use the fast ML Alamouti decoding at the receiver.
Sk,1 = r(a)(|H1 |2 + |H2 |2 )Xk,1 The following two systems are considered for comparison:
+r(a)H1 Zk,1 + fkτ H2 Zk,2 , (17) 1) System 1 (denoted by Sys.1 in short): An Alamouti-
OFDM system in DF mode without ICI self-cancellation.
2) System 2 (denoted by Sys.2 in short): Our proposed ICI
Sk,2 = r(a)(|H1 |2 + |H2 |2 )Xk,2 self-cancellation scheme.
−H1 Zk,2 + (fkτ )∗ r(a)H2 Zk,1 (18) We first give the simulation results when perfect carrier
∗ Δ τ
frequency offsets knowledge is known at the destination node.
where Sk,1 = r(a)H1 (Rk,1 + RN −k,1 )/2 + fk H2 (Rk,2 + In Fig. 3, we show the performance comparisons of Sys.1
∗ ∗ τ ∗ Δ ∗
RN −k,2 ) /2, Sk,2 = (fk ) r(a)H2 (Rk,1 + RN −k,1 )/2 − with BPSK modulation and Sys.2 with QPSK modulation
H1 (Rk,2 + RN −k,2 ) /2, and Zk,i = Nk,i + 12 (Wk,i +
∗ ∗
since in Sys.2 the self-ICI cancellation mapping reduces the
WN∗ −k,i ), i = 1, 2. symbol rate by half. When frequency offsets ε1 = 0.05, ε2 =
From (17) and (18), we can see the diversity gain of the −0.05, we can see that the performance of Sys.1 is almost the
Alamouti code structure is determined by |r(a)| and the values same with the performance of Sys.2, which is because BPSK
of |r(a)| in terms of different frequency offsets ε are plotted modulation is robust to small frequency offsets. However,
in Fig. 2. From Fig. 2, we can see that |r(a)| is large when ε when frequency offsets are large, i.e., ε1 = 0.25, ε2 = −0.25,
is in the ranges from 0 to 0.05 and from 0.25 to 0.3. When the BER of Sys.1 is always around 0.5, but the BER of Sys.2
ε = 0, |r(a)| achieves the largest value 1, and the diversity is much better, which is because when ε1 = 0.25, ε2 = −0.25,
gain is the largest. the diversity gain of Sys.2 is large according to Fig. 2 and the
Note that, for complex channels, we need to do pre- ICI in Sys.2 is mitigated. Moreover, we can see that when
compensation to achieve the proposed Alamouti code struc- ε1 = 0.05, ε2 = −0.05, the slope of Sys.2 is the same as the
ture. When the pre-compensation is not used at the relay slope of Sys.1 when Sys.1 has no frequency offsets. It implies
nodes, we can also construct an ICI self-cancellation scheme that our proposed scheme can mitigate ICI effectively when
Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 5, MAY 2008 1843

0 0
10 10

−1 −1
10 10
BER

BER
−2 −2
10 10

Sys.1, ε =0.075, ε =−0.075, QPSK


−3 1 2 −3
10 Sys.2, ε1=0.075, ε2=−0.075, 16QAM 10 Sys.2, ε =0.075, ε =−0.075, 16QAM, Non−perfect
1 2
Sys.1, ε1=0.05, ε2=−0.05, QPSK Sys.2, ε1=0.075, ε2=−0.075, 16QAM, Perfect
Sys.2, ε1=0.05, ε2=−0.05, 16QAM Sys.2, ε1=0.05, ε2=−0.05, 16QAM, Non−perfect
Sys.1, no frequency offsets, QPSK Sys.2, ε1=0.05, ε2=−0.05, 16QAM, Perfect

−4 −4
10 10
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Eb/N0(dB) Eb/N0(dB)

Fig. 4. BER comparison between Sys.1 with QPSK and Sys.2 with 16QAM Fig. 6. BER comparison when the frequency offset estimations are perfect
when the frequency offset estimations are perfect and non-perfect for 16QAM

0
10
see that when the frequency offsets are small, the loss of
the performance is negligible. When the frequency offsets
are larger, the performance is slightly worse if there exist
−1
10
estimation errors.

V. C ONCLUSION
In this letter, we proposed an Alamouti coded OFDM
BER

−2
10

system for a cooperative system that is robust to both timing


errors and frequency offsets. OFDM was used to combat the
−3
timing errors and an ICI self-cancellation scheme was used to
10 Sys.2, ε =0.25, ε =−0.25, QPSK, Non−perfect
1 2
Sys.2, ε1=0.25, ε2=−0.25, QPSK, Perfect mitigate the ICI caused by multiple frequency offsets in the
Sys.2, ε1=0.05, ε2=−0.05, QPSK, Non−perfect
Sys.2, ε =0.05, ε =−0.05, QPSK, Perfect
cooperative system. With this new scheme, when the fading
1 2
channels are real-valued, the received signals at the destination
−4
10
0 2 4 6 8 10 12 14 16 18 20
node have the Alamouti code structure on each subcarrier and
Eb/N0(dB)
therefore the proposed scheme has the fast symbol-wise ML
decoding and in the meantime, can achieve 2-order diversity.
Fig. 5. BER comparison when the frequency offset estimations are perfect
and non-perfect for QPSK
ACKNOWLEDGMENT
The authors would like to thank the reviewers for their
frequency offsets are small and achieve diversity order 2 (note constructive comments, in particular suggesting the complex
that the Alamouti coded OFDM method has diversity order 2 channel pre-compensation.
[9] as we mentioned earlier).
In Fig. 4, we show the performance comparisons of Sys.1 R EFERENCES
with QPSK modulation and Sys.2 with 16QAM modulation. [1] V. Tarokh, N. Seshadri, and A. R. Calderbank, “Space-time codes for
From Fig. 4, we can see that even when frequency offsets are high data rate wireless communication performance criterion and code
construction,” IEEE Trans. Inform. Theory, vol. 44, pp. 744-765, Mar.
small, i.e., ε1 = 0.05, ε2 = −0.05, the performance of Sys. 1998.
2 is about 4 dB better than the performance of Sys.1 when [2] S. M. Alamouti, “A simple transmit diversity technique for wireless
Eb /N0 is around 16 dB. Moreover, the same observation can communications,” IEEE J. Select. Areas Commun., vol. 16, pp. 1451-
1458, Oct. 1998.
be obtained to imply that our scheme can achieve diversity [3] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation diversity,
order 2 when frequency offsets are small as in Fig. 3. part I, II,” IEEE Trans. Commun., vol. 51, pp. 1927-1948, Nov. 2003.
In practice, a frequency offset estimate may have error [4] J. N. Laneman and G. W. Wornell, “Distributed space-time-coded pro-
tocols for exploiting cooperative diversity in wireless networks,” IEEE
due to the additive noise and/or interference [19]. We next Trans. Inform. Theory, vol. 49, pp. 2415-2425, Oct. 2003.
show some simulation results when frequency offsets have [5] S. Wei, D. L. Goeckel, and M. C. Valenti, “Asynchronous cooperative
estimation errors. We assume the distribution of the estimation diversity,” IEEE Trans. Wireless Comm., vol. 5, pp. 1547-1557, June 2006.
[6] X. Li, “Space-time coded multi-transmission among distributed transmit-
values are uniform over a range [εi − εi , εi + εi ], where ters without perfect synchronization,” IEEE Signal Processing Lett., vol.
εi is the real frequency offset, i=1,2. In the simulation, we 11, no. 12, pp. 948-951, Dec. 2004.
let εi /εi =20% and compare the BER performance between [7] Y. Li and X.-G. Xia, “Full diversity distributed space-time codes for
asynchronous cooperative communications,” IEEE International Sympo-
perfect estimation and non-perfect estimation. The BER curves sium on Information Theory (ISIT), Adelaide, Australia, Sept. 2005, pp.
are shown in Fig. 5 and Fig. 6. From these results, we can 911-915. Also, IEEE Trans. Commun., Apr. 2007.

Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.
1844 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 7, NO. 5, MAY 2008

[8] Y. Shang and X.-G. Xia, “Shift-full-rank matrices and applications in scheme for OFDM mobile communication systems,” IEEE Trans. Com-
space-time trellis codes for relay networks with asynchronous cooperative mun., vol. 49, pp. 1185-1191, July 2001.
diversity,” IEEE Trans. Inform. Theory, vol. 52, pp. 3153-3167, July 2006. [15] K. A. Seaton and J. Armstrong, “Polynomial cancellation coding and
[9] Y. Mei, Y. Hua, A. Swami, and B. Daneshrad, “Combating synchroniza- finite differences,” IEEE Trans. Inform. Theory, vol. 46, pp. 311-313, Jan.
tion errors in cooperative relays,” in Proc. IEEE ICASSP 2005, vol. 3, 2000.
pp. 369-372, Mar. 2005. [16] K. Sathananthan, C. R. N. Athaudage, and B. Qiu, “A novel ICI
[10] Y. Li, W. Zhang, and X.-G. Xia, “Distributive high-rate full-diversity cancellation scheme to reduce both frequency offset and IQ imbalance
space-frequency codes for asynchronous cooperative communications,” effects in OFDM,” in Proc. Ninth International Symposium on Computers
in Proc. IEEE ISIT 2006, Seattle, Washington, USA, July 2006. and Communications, vol. 2, pp. 708-713, July 2004.
[11] Z. Li and X.-G. Xia, “A simple Alamouti space-time transmission [17] D. N. Dao and C. Tellambura, “Intercarrier interference self-cancellation
scheme for asynchronous cooperative systems,” IEEE Signal Processing space-frequency codes for MIMO-OFDM,” IEEE Trans. Veh. Technol.,
Lett., to appear. vol. 54, pp. 1729-1738, Sept. 2005.
[18] S. Tang, K. Gong, C. Pan, C. Pan, and Z. Yang, “Self-cancellation of
[12] Z. Li, D. Qu and G. Zhu, “An equalization technique for distributed
intercarrier interference in OFDM systems with phase noise,” in Proc.
stbc-ofdm system with multiple carrier frequency offsets,” in Proc. IEEE
IEEE PIMRC 2006, Sept. 2006.
WCNC 2006, vol. 2, pp. 839-843, Apr. 2006.
[19] K. Shi and E. Serpedin, “Coarse frame and carrier synchronization of
[13] D. Veronesi and D. L. Goeckel, “Multiple frequency offset compensation OFDM systems: a new metric and comparison,” IEEE Trans. Wireless
in cooperative wireless systems,” in Proc. IEEE GLOBECOM 2006, Nov. Commun., vol. 3, pp. 1271-1284, July 2004.
2006.
[14] Y. Zhao and S. G. Haggman, “Intercarrier interference self-cancellation

Authorized licensed use limited to: VELLORE INSTITUTE OF TECHNOLOGY. Downloaded on August 4, 2009 at 10:12 from IEEE Xplore. Restrictions apply.

Das könnte Ihnen auch gefallen