Cyber Journals: Multidisciplinary Journals in Science and Technology, Journal of Selected Areas in Telecommunications (JSAT), January Edition, 2011
Precoded Differential OFDM for Relay Networks
Homa Eghbali and Sami Muhaidat
Abstract — We study the performance of differential space time codes with linear constellation precoding (LCP) for orthogonal frequency division multiplexing (OFDM) cooperative networks over frequency selective channels. Through exploiting the unitary structure of the orthogonal STBCs, we design a low complexity differential STBCLCPOFDM receiver for cooperative networks. We assume the amplifyandforward protocol and consider both single relay and multirelay scenarios. Under the assumption of perfect power control for the relay terminal and high signaltonoise ratio for the underlying links, our performance analysis demonstrates that the considered scheme is able to exploit fully the spatial diversity.
Index Terms — Linear constellation precoding, orthogonal frequency division multiplexing, differential detection, cooperative communications.
I. INTRODUCTION
There has been a growing demand for high data rate services for wireless multimedia and internet services. Spatial diversity offers significant improvement in link reliability and spectral efficiency through the use of multiple antennas at the transmitter and/or receiver side [1][4]. Recently, cooperative communications have gained much attention due to the ability to explore the inherent spatial diversity in relay channels [5] [9]. The idea behind cooperative diversity is that in a wireless environment, the signal transmitted by the source nodes is overheard by other nodes, which are also known as partners. The source and its partners jointly process and transmit their information, creating a "virtual antenna array" although each of them is equipped with only one antenna. Most the current works on cooperative diversity consider coherent detection and assume the availability of perfect channel state information (CSI) at the receiver. In fading channels where the coherence time is large enough, the channel estimation can be carried out through the use of pilot symbols [10]. For fast fading channels where the phase carrier recovery is more difficult, differential detection provides a more practical solution. In [1014], differential detection has been investigated for cooperative transmission scenarios. The works in [1014] assume an idealized transmission environment with an underlying frequencyflat fading channel. This assumption can be justified for narrowband cooperative scenarios with fixed infrastructure; however, it is impractical if wideband cooperative networks are considered. In [26], the applicability of differential STBC to broadband cooperative transmission over frequencyselective channels was
H. Eghbali and S. Muhaidat are with the School of Engineering Science, Simon Fraser University, Burnaby, B.C., Canada (email: hea7@sfu.ca, muhaidat@ieee.org).
59
investigated for OFDM systems. However, the proposed system could not fully exploit the underlying multipath diversity. Motivated by this practical concern, we extend the work in [26] to linearly coded and grouped precoded OFDM systems for cooperative communications. LCPOFDM was developed in [25] for multicarrier wireless transmissions over frequency selective fading channels. Not only LCP improves the uncoded OFDM performance, but also doesn't reduce the transmission rates of uncoded OFDM, and
guarantees symbol detectability [24]. LCPOFDM was extended to GLCPOFDM to exploit the correlation structure of subchannels, such that, the set of correlated sunchannels are split into subsets of less correlated subchannels. Within each subset of subcarriers, a linear constellation precoder (LCP) is designed to maximize diversity and coding gains. The LCPs are in general complex and could possibly be nonunitary. While greatly reducing the system complexity, subcarrier grouping maintains the maximum possible diversity and coding gains [25]. GLCPOFDM is a considerably flexible system that offers maximum multipath diversity, as well as large coding gains, and guaranteed symbol detectability with low decoding complexity. Related work and contributions: Although there have been considerable research efforts on differential STBC (conventional and distributed) for frequency flat fading channels (see for example [10][17]), only a few isolated results have been reported on conventional differential STBC for frequencyselective channels [18][20]. Distributed differential STBC multicarrier transmission for broadband cooperative networks was investigated in [26], yet was suboptimal. Our contributions in this work are summarized as follows:
1. We propose a distributed differential linear constellation precoded OFDM (DDLCPOFDM) STBC scheme for broadband cooperative systems with amplifyandforward (AF) relaying. The proposed scheme can be considered as an extension of the DDOFDMSTBC scheme proposed in [26]. Carefully exploiting the underlying orthogonality of distributed STBC, the proposed scheme is able to fully exploit the available underlying diversity.
2. To further reduce the decoding complexity while preserving the performance, we extend the DDLCP OFDM STBC scheme to grouped linear constellation precoded (GLCP)OFDM system. We observe that the optimal performance of the DDGLCPOFDM relies on the design of the GLCP precoder.
3. We present a comprehensive Monte Carlo simulation study to confirm the analytical observations and give insight into system performance. We extend our work
on singlerelay scenarios to multirelay scenarios and analyze system's performance via simulation results.
vectors _{h}
zeromean complex Gaussian with power delay profile vectors
,
and
that are normalized such that
and
denoted
_{R}_{D} , and _{h} _{S}_{D} , are assumed to be independent
j
SR ^{,}
^{h} j
[
2
(0),
= σ
σ
SR
2
SD
j
v
by
2
(
L
[
2
2
)]
SR
=
σ
SR
∑
L
RD
l
RD
= 0
2
σ
RD
(
K σ
,
,
SR
l
RD
)
= 1
(0),
(
L
SR
)]
The rest of the paper is organized as follows: In Section II, the transmission model is introduced. The differential scheme under consideration for distributed LCP0FDM STBC is described in Sections III, and is extended to DDGLCP OFDM STBC in section IV. Section V extends our analysis on single relay scenarios to multiple relay scenarios. Numerical results are presented in Section VI and the paper is concluded in Section VII.
* , (.) ^{T} , and (.) ^{H} denote conjugate, transpose,
and Hermitian transpose operations, respectively. ⊗ denotes
denotes
entry
Kronecker product,
the Euclidean norm of a vector,
of
tr _{(}_{·}_{)} denotes the matrix trace. I _{N} denotes the identity matrix
of size N, and 0 _{M} _{×}_{M} denotes allzero matrix of size _{M} _{×} _{M} .
For
a
FFT
matrix
A
circularly symmetric complex Gaussian random variable is a
random variable
2 ) , where _{X} and _{Y} are
by
v
SD
∑
∑
SR
SR
=
SD
SD
=
[
0
0
σ
RD
2
(
l
(
SD
K σ
,
RD
K σ
)
,
= 1
2
SD
(
)
= 1
L
,
RD
SD
)]
v
RD
L
l
L
l
= σ
SD
2
(0),
SR
l
,
. The CIRs are assumed to be constant
over four consecutive blocks and vary independently every four blocks. We consider LCPOFDM transformation [25] where the LCP is defined by the N × N matrix _{Φ} that has entries over
the complex field. _{Φ} should satisfy the transmitpower
constraint
into frames, where each frame consists of two information blocks. Let the N ×1vector y _{S}_{R}_{D} represent the data vector that is transmitted to the relay terminal during the first block of each frame, whose entries are complex MPSK symbols that are generated through differential spacetime (ST) encoding.
Notation:
(.)
denotes the absolute value,
the
[.] _{k}
k
th
, _{l} denotes the (
entry
of
a
a
=
[
a
0
K
a
) where
] ^{T}
Q represents
element
is
) th
vector,
(
tr ΦΦ
H
)
_{=} _{N} . Information symbols are parsed
a
matrix,
[.] _{k} denotes
and
a
vector
)mod
N −
1
, After linear constellation precoding, the N ×1 precoded
block y
matrix Q ^{H} to yield the discrete time signal _{Q}
further
inserted
block and is removed from the corresponding received block.
In the following, without loss of generality, we drop the index _{j} for brevity. The signal received at the relay terminal during the first signalling interval (broadcasting phase) of each frame after the CP removal is
OFDM
cyclic prefix of length
To
is IFFT processed by the inverse FFT
[P a]
N
_{s}
=
N × N
q
a((
N − s + q
N
matrix.
l k
( ,
) th
_{P} _{N}
the
is
N × N
given
%
= Φ
y
SRD
SRD
the
permutation
whose
H
Φ _{y}
SRD
.
remove
L
SR
,
L
RD
,
L
SD
IBI,
a
l ≥
max(
) is
Q(l, k) = 1/
N (0,
2
σ
2
N exp(− j2π lk / N ) where
Z = X + jY ~ CN (0,σ
_{0} _{≤} _{l}_{,} _{k} _{≤} _{N} _{−}_{1}
.
per
transmitted
i.i.d.
bold lowercase letters denote vectors.
) . Bold uppercase letters denote matrices and
r
R
=
H
SR
Q
H Φ
y
SRD
+
n
R
,
,
(1)
II.SYSTEM MODEL
A singlerelay assisted cooperative communication scenario is considered. All terminals are equipped with single transmit and receive antennas. We assume AF relaying and adopt the user cooperation protocol proposed by Nabar et. al. [9]. Specifically, the source terminal communicates with the relay terminal during the first signaling interval. There is no transmission from sourcetodestination within this period. In the second signaling interval, both the relay and source terminals communicate with the destination terminal.
The CIRs for S → R , S → D and R → D links for the
,
transmission block are by =
j
th
h j
SR
h
j
SR
[0],
K
,
h
j
SR
[
L
SR
]
T
h j
SD
=
h
j
SD
[0],
K
,
h
j
SD
[
L
SD
]
T
,
and
j
h
RD
All
,
L _{S}_{D} , L _{R}_{D} denote the corresponding channel memory lengths.
, experience frequency selective Rayleigh fading. The random
links are assumed to
L _{S}_{R} ,
=
h
j
S → R
RD
[0],
K
,
j
h
RD
[
L
RD
]
and
T
respectively,
where
S → D
,
R → D
Where E _{S}_{R} is the average signal energy over one symbol
period received at relay terminal, H _{S}_{R} is the N × N circulant
matrix with entries _{[}
is the additive white Gaussian noise vector with each entry
2 per dimension. To
ensure that the power budget is not violated, the relay terminals normalizes each entry of the respective received
signal
to ensure unit average energy and
retransmits the signal during the second signalling interval
(relaying phase) of each frame. After some mathematical
manipulations, the received signal at the destination terminal during the relaying phase is given by
H
SR
_{]}
m ,
_{n}
=
h
SR
((
m − n
by
)
mod
N
)
,
and _{n} _{R}
having zeromean and variance of
E
(
[r
R
]
n
_{[}_{r}
R

2
)
_{]} _{n}
,
= E
SR
n = 1,2,K,N
+ N
0
N
,
a
factor
of
r
D
Q
H
Φ
+ n
D
y
,
SRD
(2)
where E _{R}_{D} and E _{S}_{D} are the average signal energies over one
60
symbol
entries
and
is
conditionally (conditioned on _{h} _{R}_{D} ) complex Gaussian with
H _{R}_{D} and
period
are
the
h
RD
received
N × N
at
destination
terminal,
with
and
H _{S}_{D}
circulant
matrices
respectively,
[
[
H
H
RD
SD
]
]
m
m
,
,
_{n}
_{n}
=
=
h
SD
(( 
m − n 
) 
mod 
N 
) 
(( 
m − n 
) 
mod 
N 
) 
,
n _{D}
R
6447448
Qr
Qr
*
( )
t
t
Φ
(
+
1
)
=
Qn
Qn
*
(
t
t
)
1
+
y
SRD
y
SD (
( )
+
Φ
RD
Λ
Λ
* Φ
SR
SD
−
SD
Φ
Λ
*
SR
(
t
)
.
t )
(5)
1 +
Assuming without loss of generality the symbols to have
1 , applying the minimum mean square
error (MMSE) equalization we have
m )
2
zero mean and variance
by 
a 

where 
ρ 
= 
1 
+ 
α 
= 
1 
+ 
σ
2
n
N
=
D 0
h RD
(
. variance
2
2
σ = σ
y
SRD
y
SD
=
The destination terminal further normalizes the received signal
,
R ,
A
γ
2
Λ
SD
that
this
*
SD
)
(6) 

− 
1 

. In the 
α
does not affect the SNR, but simplifies the ensuing presentation [21]. After normalization, we obtain
H
ΦΦ Λ
the receiver is provided with the timedomain observation r in
(3).
Inspired by Alamouti code [4], we can further extend (3) to a distributed (D)STBCLCPOFDM scenario, by using the
and
SRD
(
t +
1
)
= −
y
*
SD
( )
t
(4)
Exploiting the circulant structure of the channel matrices
i ,
a diagonal matrix whose
DFT coefficient of h _{i} . Thus,
transforming the received signal r _{(}_{2}_{t}_{)} to the frequency
domain by multiplying it with Q matrix and further writing the result in matrix form we have
H
_{i}
H _{S}_{R} , denotes
_{R}_{D}
,
(
n n
,
) th
and
SR,
H _{S}_{D} , RD,
we
SD,
have
is
_{H}
i
=
_{Q}
H Λ
i
_{Q}
, where
Λ
element is equal to the
n
th
following, we will discuss the distributed differential (DD) LCPOFDM system and the ST encoding and decoding
procedures employed by it.
The
data
III.
vectors
DDLCPOFDM
K
d
i
( )
t
_{=}
d
0
i
( )
t
,
,
d
N
i
−
1
( )
t
T
i = 1,2,
represent the OFDMSTBC symbols, where t is the time
n = 1, K , N , are drawn
index and complex symbols
from an unitenergy MPSK constellation. We are encoding the
and
into their linear constellation precoded
LCPOFDMSTBC
d
n
i
( )
t
vectors
data
d
1
( )
t
= Φd
1
( )
t
d
2
( )
t
= Φd
2
( )
t
differentially encoded frequency domain counterparts
and
y
SRD
(
m
)
=
0
y
SRD
(
m
)
y
N
−
SRD
1
(
m
)
_{}
T
,
…
,
(
SD
n
( )
t
y
Y
)
where
D
n
( )
t
and
Y
n
( )
t
having
n
y
SRD
n
SD
y
(
(
2
2
+
+
1
1
0
SD
y
(
n
SD
y
n
y
SRD
(
(
2
2
t
N 
− 1 
( 
m ) 
_{} 
T 
m = 2t,2t+1 as following 

y SD 

(7) 

d n 2 ( d 1 n ( ) ( )) t t * 
, 
(8) 

n 


y 
SD 
( 
2 t ) 

, 
(9) 

y 
n SD ( 
2 
t + 
1 
) 

t 
)) * 
, 

* 
(10) 
))
.
=
m
(
m
)
,
…
,
t
t
)
)
= −
=
(
Note that (7) is relating 4 frames of information in a differential manner. Applying (10) at the sequence level to the OFDM blocks _{(}_{2}_{t}_{)} and _{(}_{2}_{t} +1) in (9) and substituting them
into (4), we obtain
61
r (
+
(11)
1
2
(
n
%
r
%
r
n
(
(
(
(
n
r %
2
2
r %
(
n
2
t
+
)
t
t
2
−
−
t
2
1
))
*
)
1
))
*
−
(
_{=}
r %
r %
n
n
(
(
n
1
U
(
U
t
t
n
2
2
2
−
−
(
(
1
2
)
t
t
)
))
))
*
*
+
^{} d d
n
^{}
1
n
2
t
)
.
(
( )
t
(15)
To perform MMSE equalization, we can further cascade every
which represents the two consecutive received OFDM frames _{(}_{2}_{t}_{)} and _{(}_{2}_{t} +1) at the destination terminal.
from
the differentially encoded received signals in (12), we exploit the circulant structure of the channel matrices H _{R}_{D} , H _{S}_{R} , and
To recover the OFDM data vectors
d
1
( )
t
and
d
2
( )
t
n
th
elements to use the vector forms in (15) as follows
1
r %
r %
*
(
(
2
t +
t
)
(
r %
(
*
r %
2
2
t
t
2
−
−
2
1
)
)
1
r %
2
U
U
t
t
)
−
=
r %
*
(
(
2
−
−
1
*
2
(
t
t
)
2
(
1
)
)
)
+
1
2
d
d
(
(
t
t
)
)
.
(16)
d
n
1
(
)
and
n
d
th
n
2
subchannel
n
th
( )
t
. Considering the
R
n
( )
t
n %
n
( )
t
n
th
Thus, performing blind MMSE equalization with no access to
CSI, we have
B
=
(
H
%
r
H
Φ
H
%
r
H
d 
1 


d 
2 

( 2 t 
− 
2 
) 
B 

( 2 
t 
− 
1 
) 
B 

% 

R 
%
r
(
2
t
−
2
)
ΦΦ
H
%
r
H
(
2
t
−
2
)
(
2
t
−
1
)
%
r
(
*
2
(
t
2
−
2
t +
+
γ
2
%
r
(
B
)
1
B
T
)
%
R ,
2
t
−
1
)
IV.
DDGLCPOFDM
H
%
r
H
(
(17)
2
and
t −
1
))
Φ
where
γ
1
(12)
*
ΦΦ
Note that in here, we are performing the differential decoding
upon the
t
(12) in the matrix form we have
subchannel, writing
subcarrier to recover
and
Due to high decoding complexity of LCPOFDM, an optimal subcarrier grouping technique was proposed in [25] in which the decoder's complexity is reduced by dividing the set of all subcarriers into nonintersecting subsets of subcarriers, called subcarrier groups. In this approach, every information
symbol is transmitted over subcarriers within only one of these
subsets. Note that if the subcarrier grouping is properly done,
not only the decoding complexity is reduced, but also systems
performance is preserved [25].
The GLCP matrix θ is designed such that the decoding
complexity is reduced, while preserving the maximum
diversity and coding gains. Assuming that N = KM , in DD GLCPOFDM STBC system, the information symbols
,
where ψ is a K × N permutation matrix built from the rows
64748 64748
differential detector, R
the previous input, R
n ( t )
n
(
t −
, for each subchannel is related to
1)
, according to
d
i
( ),
t
i = 1,2,
m
are divided into _{M} blocks, d
m
i
( )
t
=ψ
m
d
i
( )
t
=
D
We
R
n
( )
t
n
( )
t
R
=
n
(
set
Y
n
n
D
t
−
R
n
(
t
−
1
)
−
n %
n
(
t
−
1
)
64748
Y
n %
−
( )
)
1
t
n
(
t
1
)
Λ
n
+
n
n %
n %
( )
(
t
+
( )
( )
n
t
−
D
n
t
n
t
−
1
)
.
14444244443
U
n
( )
t
(14)
(0)
_{;}
n = 0,K, N −1 _{t}_{o}
I and linearly detect
_{2}
D
n
( )
t
from Y
n
( )
t
using the orthogonal structure in (13). This
can be done by rewriting (14) as follows
_{(}_{m} −1) K +1 → mK of I _{N} ; and then precoded by the GLCP
matrix θ . Thus, as an example, (12) can be rewritten as
%
r
(
+
(18)
62
−
1
.
where
i = SRD,SD , are differentially encoded from the GLCP
t
(7).
Following [2425], for any _{K} , QAM, PAM, BPSK, and QPSK constellation, the optimal can be constructed through LCPA, which can be generally written as a Vandermonde matrix as following
OFDM symbols g
following similar steps as in
( )
Λ
i
=ψ Λ ψ
m
m
i
m
i
m
( t )
_{,}
i = SR,RD,SD
= θ
d
m
i
_{,}
T
_{a}_{n}_{d}
m
g
i
(2 )
t
, normalized received signal to the destination terminal, and the
destination receives
_{D} t denotes the received signal at the destination terminal at
time slot _{t} . Assuming that the Information symbols are first
parsed as four streams of N ×1 blocks
encoded to c _{t}_{i} using the orthogonal spacetime block code design C , (7) holds true with
where
n
x
x
n
1
x
n
2
x
n
4
n
3
* (
* (
) *
(
−
,r
D
,r
D
n
2
n
1
x
n
4
n
3
*
x
)
)
n
4
*
)
*
n
3
)
x
n
3
−
*
,r
D
n
4
n
12
x
n
3
)
*
(
x
n
4
n
1
)
n
12
x
(
x
,r
D
) *
*
)
*
,r
D
,r
D
x
i
_{,} i
,
,r
D
=
4
r
D
8
12
16
20
24
28
32
,
r
1, 2, 3,4 _{a}_{n}_{d}
x
x
− x
−
−
θ
=
1
β
1 α
α
1
L
L
M
L
1 α
α
2
M
1
M
α K
M
K
α
K
−
−
−
where _{β}
parameters _{{}
is
a
constant
K
1
K
2
1
1
1
,
x
x
n
1
x
(
−
(
(19)
the
−
x
n
2
n
1
x
x
(
x
D
n ( t )
=
1
)
(
)
− (
)
−
−
( x
1
n
2
x
( x
n
4
n
3
*
(20)
such that
tr θθ
(
H
)
_{=} _{K}
,
and
α
k
_{}}
K
k
=
1 are selected depending on _{K} [25].
V.EXTENSION TO MULTIPLE RELAY SCENARIOS
We consider a multiplerelay assisted cooperative wireless communication system with a single source S , _{R} halfduplex
relay terminals
The source, destination, and all relays are equipped with single transmit and receive antennas. We adopt the transmission protocol in [27] and consider nonregenrative relays. Note that unlike [27], we assume that there is no direct transmission between the source and destination terminals due to the presence of shadowing.
R , i = 1,2,K, R , and a single destination _{D} .
and Y
changes to
n ( t )
changes accordingly. Following this, R
4
D
r
8
D
r
12
r
D
16
r
D
20
r
D
24
r
D
28
r
D
32
r
D
T
.
n ( t )
in (13)
(21)
i
R
n
( )
t
_{=}
VI. NUMERICAL RESULTS
In this section, we present MonteCarlo simulation results for the proposed receiver.
T Fig.1. depicts the SER performance of the DDLCPOFDM
The CIRs for
S → R
i
and
links for the
→ D
R
h
i
SR
i
_{=} h
SR
i
[0],
, h
SR
[
L
i
SR
i
]
, respectively , where L
SR
i
th
relay
^{,}
terminal are given by
h
R
i
D
_{=} h
R D
i
[0],
, h
R D
i
[
L
R D
i
] _{}
T
STBC scheme assuming for the following three different
scenarios:
1)
2)
3) L
For LCPOFDM to achieve maximum diversity order, maximum diversity encoders should be used. Two classes of maximum achievable diversity order (MADO) enabling LCP encoders are introduced in [24], namely: Vandermonde encoders and cosine encoders. In here, we are using Vandermonde encoders and we assume 4PSK modulation. To
further minimize the receiver complexity, we are applying the
lowcost minimum meansquare error (MMSE) equalizer. Our simulation results indicate that with the optimal design of LCP encoder matrix, the DDLCPOFDM STBC system is able to achieve full spatial and multipath diversity,
and
All
the
selective Rayleigh fading. The random vectors h
are assumed to be independent zeromean complex Gaussian with power delay profile vectors denoted by
and
that are normalized such that
L
L
SR
SR
=L
=L
RD
RD
=L
=L
SD
SD
= 0,
= 1,
L R _{D} denote the corresponding channel memory lengths.
i
S → R
i
and
v SR
i
= [
2
σ
SR
i
(0),
v
=
R D
L
l
i
∑
SR _{i}
SR _{i}
= 0
[
2
σ
R D
i
(0),
2
SR
σ
i
(
l
SR
i
)
R → D links are assumed to be frequency
i
SR
i
and h
R D
i
,
2
σ
SR
i
(
,
2
σ
R D
i
L
SR
i
)]
(
L
R D
i
)]
= 1
and
∑
L
l
D
R
i
R
i
D
= 0
2
R D
σ
i
(
l
R D
i
Molto più che documenti.
Scopri tutto ciò che Scribd ha da offrire, inclusi libri e audiolibri dei maggiori editori.
Annulla in qualsiasi momento.