) and thus
k
= ,
for k J. The stationarity of the interference configuration
allows for an accurate estimate of the impairment covariance
Q which can be efficiently exploited for interference
mitigation. The MIMOOFDMA system provides two
dimensions that can be employed for interference reduction:
The frequency domain of the OFDM signaling and the spatial
domain offered by the array processing. To enhance the
interference rejection capability, we adopt a uniform linear
antenna array (ULA) with closely spaced apart antennas and a
beamforming processing over each subcarrier. The SINR
variate y
k
= y(E
k
,I
1
E
k
} (2)
to be exploited when establishing the branch metric of the
Viterbi decoder. The channel E
k
and the interference
covariance Q of this scenario are modeled according to the
beamforming model.
B. The Randomized interference
This scenario is modeled to exploit the maximum diversity
provided by the fluctuations of both channel and impairments.
Diversity is artificially introduced through the randomized
multiuser access approach which provides interference
fluctuations over the codeword. In this case, any interference
mitigation techniques is unfeasible due to the unpredictability
of the impairments configuration (i.e., the highly varying
covariance
k
cannot be reliably estimated). To exploit the
diversity provided by the MIMO channel we adopt an OSTBC
with antennas sufficiently spaced apart. The receiver is based
on coherent maximum likelihood (ML) OSTBC detector as
described in [8], where the Viterbi decoder exploits the
knowledge of the average noise power o
2
over the whole data
region (conventional decoder). In this case, the SINR to be
used for decoding is
y
k
=
P
0
N
T
tH
k
H
H
k

c
2
(3)
As lowerbound performance reference, we also consider the
optimal decoding based on the knowledge of the instantaneous
interference power o
k
2
on each subcarrier. For this genie
decoder the instantaneous SINR at the decision variable is
y
k
=
P
0
N
T
t H
k
H
H
k

c
k
2
(4)
For analytical purposes, it is convenient to rewrite the SINR as
a function of the N
R
N
1
1 normalized spacefrequency
channel vector.
k
= _
P
0
N
T
.:cc(
k
H/ 2
E
k
)
(5)
where the interference covariance is
k
= defined as in (7)
for the coordinated scenario,
k
=o
2
I
N
R
for the randomized
one with conventional decoder and
k
= o
k
2
I
N
R
for the genie
decoder. The SINR reduces to:
y
k
= [
k
[
2
(6)
Properties of this equivalent spacefrequency channel
k
for
performance analysis depend on the spatialtemporal
dispersion of the multipath propagation for the MIMO
channel E
k
, the spatial configuration of the intercell
interference from the covariance
k
and the fluctuations of
the interference power induced by the multiple access policy
(i.e., the variations of I
k
).
Fig 2. Example of multiuser access policies applied to the
cellular layout of Fig. 1. For the coordination approach (a)
each data region is mapped onto a set of contiguous
subcarriers, so that the outofcell interference can be assumed
stationary along the data regions. In case of randomization (b)
the subcarriers of each data region are spread all over the
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3224
bandwidth according to a different pseudorandom
permutation rule in each cell leading to nonstationary
interference.
V. ANALYTICAL PERFORMANCE ASSESSMENT
We derive the performance of the MIMOOFDMA system in
terms of the average bit error probability P
b
= P
(b
`
k
b
k
)
at the output of the CC. For the sake of clarity, here we focus
on BPSK modulation with transmitted symbols x
k
{1}.
Sec.VI will extend the analytical methodology to HQAM
modulations. The average bit error probability is evaluated
through the union bound as [10]
P
b
1
k
cc
[(c)P(c)
ccs(d) d>d
]rcc
=
1
k
cc
[(c)E
y
[P(cy)]
ccs(d) d>d
]rcc
(7)
Where Jrcc is the free Hamming distance of the code, (J)
is the set of all error events c having Hamming distance J
from the allzero codeword, [(c) is the input weight and P(c)
is the PEP for the error event c. The error probability P(c) is
obtained by averaging the conditioned PEP P(c y) with
respect to y. Notice that the J erroneous bits in the error path c
are mapped onto J different subcarriers = {1,...,J}of
the data region by the sequence of interleaving, physical
logical subcarrier mapping and scheduling. For every value of
J, we need to consider all possible patterns with J erroneous
bits in the set (J), as different arrangements of J subcarriers
may lead to very different channel correlation values and
consequently different error probability values. For a given
frame, the conditioned PEP P(c y) depends on the SINR
variates y = {y1,...,yJ} that are experienced along the
subcarriers , i.e. with instantaneous correlated fading gains
= {E1,...,EJ} and interference configurations
= {I
]1
,...,I
]d
}. This probability can be expressed as
P(cy) =P(cE,I) = (
2y
c]]
(E,I) )
(8)
Where the effective SINR y
c]]
(E,I) is defined as
y
c]]
(E,I) = y
k k
. To ease the analysis, it is convenient
to write the effective SINR as
y
c]]
(E,I) = [
k
[
2
= [
[
2
kcP
(9)
as a function of the JN
1
N
R
1 channel vector
= 
]
1
1
,,
]
d
1
]
1
(10)
This gathers the channel responses (6) for all subcarriers
associated with the error event C. The average of the
conditioned PEP, Ey[P(c y)], with respect to y = y(E,I) is
derived according to (1), first with respect to the channel E for
a given pattern I and then over the possible interference
configurations. In particular, the first average (E
[]) is
evaluated in Sec. IVA, taking into account that the SINR
variates y are generally correlated due to the frequency
selectivity of multipath channel. Generally speaking, the more
error bits are spread in the frequency domain, the lower is the
correlation among SINR variates and the higher is the
diversity offered by the channel. The second average over the
interference patterns (E
I
[]) is discussed in Sec. VC.
A. Average over the fading
The average of the conditioned PEP with respect to the fading
is
P(cI) = E
H
[P(cE,I)]
= (
2y
c]]
)p(y
c]]
I)Jy
c]]
(11)
Where we used the compact notation y
c]]
= y
c]]
(E
k
,I
k
) for
the effective SINR (9) and p(y
c]]
I) for its probability density
function (pdf). This pdf can be obtained from (9) by using the
Rayleigh fading assumption and the correlation properties of
as described below. Based on the Rayleigh multipath fading
model, the space time channel is characterized by a spatial
temporal dispersion which directly reflects on the correlation
of the space frequency channel response. According to the
definition (10), the spacefrequency channel (evaluated in the
positions related to the error event) can be expressed as
~ cJ(0,R
d
) where the covariance matrix is
R
d
=E
H
] = _
R
]
1
,]
1
R
]
1,]
d
R
]
d
,
1
R
]
d
,]
d
_
(12)
Here R
k,h
=E[
k
,E
h
H
] is the spatial N
R
N
1
N
R
N
1
cross
correlation for the MIMO channel between the kt and t
subcarriers, weighted by the interference contributions
{k,}according to (5). Notice that the crosscorrelation of
the equivalent channel is evaluated in (12) only among the
subcarriers selected by the specific error event c.
According to the channel model the correlation matrix of the
multipath channel response is:
R
k,h
= p
w
=1
R
S,
0
k
0
h
exp[]2n
]
k
]
h
N
:
r
1
(13)
Where the spatial correlation matrix R
S,
is given by:
R
S,
=EjIcc
k
H/ 2
A
] .Icc
h
H/ 2
A
]
1
[
= R
1X,
_
k

H
2
R
RX,
h
1/ 2
_ (14)
from the Kronecker model. For the evaluation of the pdf of
y
c]]
in (9), we introduce the eigenvalue decomposition (EVD)
of the covariance matrix (12), R
d
=uu
H
, where A =
Jiog{z
1
,...,z
dN
T
N
R
} gathers the eigen values and U the
corresponding eigenvectors. Using the EVD, we can
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3225
equivalently write the channel vector as
= ub, with
b cJ(0,A) representing the projection of the channel onto
the orthonormal basis U. It follows that the effective SINR,
y
c]]
=b
2
is the sum of JN
R
N
1
independent exponentially
distributed random variables with mean values {z
}
=1
dN
R
N
T
. The
moment generating function (MGF) of such a sum is,
H
y
c]]
(s) =
1
1x
i
s
dN
R
N
T
=1
(15)
We can evaluate (11) using as p(y
c]]
) the inverse Fourier
transform of (15) which yields the general expression
p(y
c]]
I) =
x
x
n
dN
R
N
T
n=1
m
y
c]]
m+1
c
y
c]]
/ Z
x
m+
(m+)
m=0
(16)
Where p =JN
R
N
1
, z
= min
n
{z
n
} and the coefficients
m
are obtained recursively, starting from
0
= 1, according
to
m
=
1
m
_ _1
x
x
]
]
dN
R
N
T
]=1
_
m
1
m (17)
In case all the JN
R
N
1
eigenvalues are different (the
eigenvalues are not assuming identical values, i.e. z
z
]
for
i, ] with i ]), the pdf can be simplified as
p(y
c]]
I) =
A
i
x
i
dN
R
N
T
=1
exp (y
c]]
/ z
)
(18)
Where the coefficients A
are:
A
=  (1 z
k
/ z
)
dN
R
N
T
k=1,k=
]
1
(19)
Employing (18), we can further approximate (11) adopting the
Chernoff bound of the function (i.e., (x)
1
2
cxp(x
2
/
2)) and integrating over the effective SINR. Hence, after
straightforward algebraic manipulations, the average PEP can
be written as
P(cI)
1
2
A
i
1+x
i
dN
RN
T
=1
(20)
B. Codeword bound
To evaluate the bit error probability (7) we have to account for
all the possible codewords of the set (J) for a given J. The
cardinality of (J) varies with respect to the specific CC and
it usually grows for increasing Hamming distance. Notice also
that interleaver and scheduler scramble the bits of each
codeword in different ways giving rise to a large number of
patterns of subcarriers carrying the unmatched bits. Due to the
selectivity of the channel, each configuration experiences a
different degree of diversity. For example, let us consider the
CC with generator polynomials [171, 133]: at the free distance
Jrcc = 10 the set (Jrcc) is composed by 11 possible
error paths. Hence the computation of (7) requires the
evaluation of at least these 11 configurations.
To avoid the exhaustive computation of all the error events for
a given set (J), in the high SINR region we simplify the
bound (14) by upper bounding each PEP with the PEP of the
Code word c
w,d
associated to the configuration of J bits that
achieves the minimum degree of diversity (or equivalently, the
maximum correlation of the SINR values over ) at the output
of the scrambling process, thus leading to P(c) P(c
w,d
).In
this way the system performance (7) is further bounded as
P
b
1
k
cc
[
d
P(c
w,d
)
d >d
]rcc
(21)
With [
d
= [(c)
cc(d)
. Hence, the worst error event c
w,d
is
the one that yields the highest PEP:
c
w,d
=argmax
cs(d)
{P(cI)} (22)
For the maximization we use the expression
P(cI) =
1
n
_ _ _1+
z
sin
2
0
]
1
J0
dN
RN
T
=1
n
2
0
for the conditioned PEP, upperbounded with sin2 (0) 1
(see also [37]), i.e.:
P(cI)
1
2
1
1+x
i
dN
RN
T
=1
(23)
A. Average interference
Here we derive the average of the error probability over the
possible interference scenarios I according to the two
interference management strategies. In case of interference
coordinated policy, the interference pattern is constant over
the whole data region so that I
K
= I
and
K
= ,k J;
thereby, the average over I is not needed. On the other hand, in
the randomized approach the permutation of the subcarriers
makes the interference scenario I
k
randomly fluctuate along k
and the average PEP has to account for all the possible levels
of interference. According to the assumptions in Sec. III, the
overall noise power depends on the number of interferers V
k
=
I
k
 superimposed on the kth subcarrier. The interference
power pattern over the error event c can be fully described by
the vector of cardinalities : = [:1 :J ]. The average
of the conditioned PEP can be calculated as
P(c) = E
I
[P(cI)] = P(c:) P
(:) (24)
Where P (c :) is the PEP conditioned to the interference
pattern v and Pr(v) is the probability of the interfering scenario
v, depending on the traffic load p of the neighboring cells. The
number of interferers can be considered as a random variable
independent from subcarrier to subcarrier, so that
P
(:) = P
d
k=1
(:
]k
) (25)
Where each term P
(:
]k
) follows the binomial distribution:
P
(:) =(
N
I
) p
(1 p)
N
I
(26)
The conditioned PEP P (c :) is obtained through the
equations (11) and (16), or directly (21), by averaging over the
effective SINR (16) for a given interference configuration. We
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3226
recall that each component y
k
of y
c]]
is defined according to
(3) or (4) depending on the interference information available
at the receiver (either conventional or ideal genie decoder).
The conventional decoder has knowledge only of the average
noiseplusinterference power, thereby, for a given
interference pattern, the power o
2
in (3) is the average over
the error event:
o
2
=
1
d
o
k
2
kc
:
P
I
+ o
bn
2
(27)
Where :
=
1
d
:
k kc
is the average number of active users
sharing the same J frequencies. The expression (24) simplifies
to the average over the (N
I
+1) values of the scalar variable
:
(:
) for a scenario v.
On the other hand, the ideal case of a genie decoder is based
on the instantaneous interference power knowledge (4). In this
case the computation of the conditioned PEP has to be
performed over each of the (NI +1) possible configuration of
v. This number is unfeasible for any practical performance
assessment. For this reason, in Sec. VI the expectation (24)
will be evaluated semianalytically by MonteCarlo averaging
over a number of random realizations of v generated according
to the distribution (26).
VI. PERFORMANCE ANALYSIS FOR BROADBAND
STANDARDIZED OFDMA SYSTEMS
The performance analysis of Sections VVI is here configured
for the assessment of the following standardbased wireless
technologies: the WiMAX standard IEEE 802.16d for fixed
or nomadic SSs, the WiMAX standard IEEE 802.16e for
mobility scenarios and the 3GPP LTE recommendations
(downlink only). The analytical performance is validated by
MonteCarlo simulation of the OFDM/OFDMA PHY layer of
each standard, whose main parameters are summarized in
Table I. Details on the adopted CC are specified in Table II.
Both the interference randomization and coordination policies
are considered using for simulation the two scenarios
described in Sec. III. The performance is evaluated with
respect to the average SINR defined as
SINR=
P
0
o
n
2
+ P
I
N
I
=1
= [
1
SNR
+
1
SIR
1
(28)
Where SNR = P0/
bn
2
denotes the background signalto
noise ratio, while SIR = P
0
/ P
I
N
I
I=1
i is the signalto
interference ratio. All SSs are assumed to transmit at the same
power level. According to the pathloss law, the SIR depends
only on the ratio between the distances SSiBS0 (i) and SS0
BS0 (o), for = 1,...,N
I
: SIR =
P
0
P
N
I
=1
=
1
(I
0
/ I
)
q
N
I
=1
for
pathloss exponent p =3. For simulation of the coordinated
scenario, at the BS we consider an antenna array with inter
element spacing 1.4z (where z is the carrier wavelength)
according to the optimized antenna array deployment, which
provides a spatial resolution of 0 =3.4deg. Channels are
simulated using the beamforming model, with a separable
spacetime structure defined as follow. The multipath pattern
is composed of w =wcwm paths, grouped into wc clusters
of wm paths each. All paths within a cluster have the same
DOA,
which is 0
p
= (p (wc + 1)/ 2)A0 + 0
0
for the pt
cluster, p = 1,...,wc. The TOAs in each cluster are defined
as multiple integers of the sampling interval I , i.e.
(30)
and Laplacian (double exponential) profile over the angles
a
(,
0
) =
a
(0,
0
) exp[

0

(31)
Parameters o
:
and o
0
represent, respectively, the delay and
angle spreads, 0
0
is the main DOA connecting SS0 and BS0,
while p
d
(0) andp
u
(0,0
0
) are factors employed to normalize
the overall profile so that p
w
=1
= 1. Signals from the three
interferers are simulated with main DOAs (with respect to the
BS array): 0
1
=46.1deg, 0
2
=0deg and 0
3
=46.1deg. The
interference generated by the it SS is temporally
uncorrelated but spatially correlated with spatial covariance,
for i =1, 2, 3. Similarly to the SS0 channel, the multipath
structure of each interferer shows w
I
=w
c
paths with DOAs
{0p}
p=1
w
C
as defined above and Laplacian powerangle profile
p
,
= p
u
(0
p
,0
).
The frequencyflat (FF) channel where the channel gains are
constant over the whole bandwidth, i.e., E
k
= E,k (as for a
frequencyflat channel with delay spread o
:
= 0). The
maximum frequency diversity (MaxFD) channel where the
channel gains are i.i.d. over the subcarriers (as for the ideal
case of a maximum delay spread o ).
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3227
A. Modeling the WiMAX IEEE 802.16 multicell
scenario
We first consider the uplink of a WiMAX system conforming
to the IEEE 802.16d standard. This is an OFDM system
where the active SS in each cell transmits using all the
available subcarriers: K =K
T
=192 (full traffic load p =1).
We assume a coordinated scenario with each SS employing a
single antenna (N
T
=1) and the BS using an ULA of N
R
=4
elements for interference cancellation. The singleinput
multipleoutput (SIMO) channels are simulated according to
the beamforming model described above with wc = 35
clusters of wm =64 multipath components each.
Fig. 3. Analytic and simulated BER vs SINR in coordinated
interference scenario. Performance of a SIMO IEEE 802.16d
system with (A)16QAM, convolutional code WiMAXcc1,
interference cancellation with MVDR filtering and full traffic
load (p = 1). (B) QPSK, convolutional code WiMAXcc2,
interference cancellation with MVDR filtering and full traffic
load p =1. (C) randomized interference scenario with ideal
genie decoder. Performance of a SISO IEEE 802.16e system
with QPSK, convolutional code WiMAXcc1, PUSC
randomization and traffic load p = 1/3. (D) Analytic and
simulated BER vs traffic load p in randomized interference
scenario with ideal genie decoder. Performance of a SISO
IEEE 802.16e system with QPSK, convolutional code
WiMAXcc1, Interference randomization with PUSC strategy
and fixed delay spread o
:
= 1ps.
Fig. 3(A) compares the simulated and the analytical BER vs
the SINR, for 16QAM modulation and code WiMAXcc1 in
Table II. The user in cell 0 transmits from the main DOA 0
0
=
5deg with SNR =20dB. The analytical BER is evaluated using
the union bound (21) with PEP (31) and SINR distribution
(16). The union bound is truncated to either the free distance
term J =10 or the two first minimum distances J ={10, 12}.
The worst error event configurations, c
w,d
, are identified using
the specific code and interleaver scheme prescribed by the
standard. Multipath channels are simulated with delay spread
o
:
{0.1; 0.4; 1}& angular spread o
0
{1.1; 1.7; 4}Jcg.
The two extreme cases of frequency selectivity MaxFD and
FF are evaluated as well to get lower and upper performance
bounds. The numerical results in Fig. 3(A) show that the
system performance improves when increasing the channel
spread o
0
and/or o
:
. As a matter of fact the more uncorrelated
are the channel gains in space frequency, the higher is the
diversity gain provided to the BICM. As expected, the best
performance is achieved with the MaxFD case (channel gains
are uncorrelated over the subcarriers and thus the diversity
degree is J =10) , while the worst performance is obtained
with the FF channel which does not provide any diversity gain
(the channel is constant over the whole codeword). In all
cases, the analytical BER accurately fits the simulated results
over the whole range of SINR values and spacefrequency
selectivity degrees.
Fig.3(B) draws the simulated and analytical performance for
the same system of the previous simulation but with QPSK
modulation and punctured convolutional decoder WiMAXcc2
(see Table II). Multipath channels are simulated with delay
spread o
:
{0.05; 0.2; 0.5; 1; 3}ps and angular spread o
0
{0.57; 1.14; 2.3; 4.6; 5.7} deg. Simulated results show that the
proposed analytical tool offers a tight performance bound even
in this scenario. We point out that, compared to the previous
case, here more error events (with J =6, 7, 8) have been
included in the computation of the union bound since the path
at the free distance is not enough to represent the statistics of
the error events. We now investigate the performance
provided by interference randomization simulating the IEEE
802.16e standard with the PUSC permutation rule. For this
scenario we consider a singleinputsingleoutput (SISO)
communication link (N
T
=N
R
=1) and an ideal genie decoder.
The total data subcarriers for each OFDM symbol is KI =
841.
The PUSC approach prescribes a division of the spectrum into
35 subchannels of 6 tiles each. A tile is a group of 4
consecutive subcarriers repeated for 3 OFDM symbols. The
data region is formed by a number of subchannels of 48 data
subcarriers each. The user SS0 requests 4 subchannels and
adopts QPSK modulation with WiMAXcc1 code (see Table
II). The interleaving and the PUSC permutation rules are
applied according to the standard specifications. At the
receiver, we assume SNR =30dB.
Fig. 3(C) draws the simulated and analytical BER vs the
SINR for the considered system with traffic load p =1/3 and
o {0.05; 0.1; 1} ps. Notice that, in this SISO scenario, for
decreasing values of the delay spread the analytical curves
obtained with only the free distance (J =10) do not closely fit
the performance. The same result has been observed for SISO
systems with constant interference pattern. The basic principle
behind this behavior is that in a strongly correlated channel
(without spatial diversity provided by multiple antenna), an
unfavorable channel realization affects the whole codeword,
the interleaver is less effective and the error event can be
easily longer than the free distance. The analytical
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3228
performance computation should include error events with
longer distances. An investigation of the analytical
performance for varying system load p is in Fig. 3(D) for a
selective channel with delay spread o
:
= 1ps, average SINR
{12, 16dB} and SNR =30dB. As expected, performance
degradation is experienced as the cell load increases since the
interference power grows providing less interference diversity.
We can notice that the analytic bound fits all the considered
traffic load situations, ranging from light load (p =0.4) to full
load (p =1, i.e., for constant interference level).
B. Modeling the 3GPP LTE multicell scenario
The downlink of a 3GPP LTE system is here simulated in a
randomized interference scenario. We consider a MIMO link
of N
R
=N
T
=2 antennas, with Alamouti code and maximum
ratio combining (MRC) at the receiver. The diversity channel
model is adopted with o
:
{0.5; 1.5}s and channel length
wm =128 taps. The main 3GPP LTE system parameters are
reported in Table I while the subcarrier permutation policy is
modeled as described in [31]. The data region assigned to the
user of interest is a timefrequency grid composed of 25
subcarriers and 8 consecutive OFDM time symbols, yielding
an overall number of K =200 subcarriers. These subcarriers
are employed to transmit 400 coded bits, using the LTEcc
code in Table II and QPSK modulation. The traffic load is set
to p =0.4, while the signaltonoise ratio is SNR =20dB.
Fig. 4. Analytic and simulated BER vs SINR in randomized
interference scenario with (A)conventional decoder.
Performance of a MIMO LTE system with QPSK, CC LTEcc,
traffic load p =0.4. (B) ideal genie decoder. Performance of a
MIMO LTE system with QPSK, CC LTEcc, interference
randomization with traffic load p =0.4.
Figures 4(A) and 4(B) show the BER performance vs the
average SINR for the conventional and the genie decoder,
respectively. The analytical BER is evaluated according to the
union bound (21) truncated to the first two error events with
Hamming distances J ={12, 14}, using (16) for the average
with respect to the random interference and (16) for the
average with respect to the fading channel. Simulation results
show that the performance bound is tight over a large SNR
range and for different degrees of channel correlation. With
this system set up, however, the analytical bound turns out to
be less accurate for small delay spread (e.g., for o
:
= 0.5ps):
this is the most crucial range of operation as error events with
large Hamming distance (not taken into account here) are
more likely to occur in highly correlated channels. The
performance gain of the genie receiver with respect to the
conventional one is approximately 1dB.
Fig. 5. BER vs SINR in coordinated interference scenario with
3GPP SCM Model III for the proposed analytical bound and
the EESM method. SIMO IEEE 802.16d system with QPSK,
convolutional code WiMAXcc1, interference cancellation
with MVDR filtering and full traffic load
C. Performance comparison with the EESM
methodology
The performance provided by the proposed method is
compared here to the EESM approach. Fig. 5 shows the
comparison between the two methods for the IEEE 802.16d
beamforming scenario with QPSK modulation and code
WiMAXcc1 in Table II. Both the EESM and the proposed
analytical methods provide a good matching with the
simulated BER performances. However, the two methods
require substantially different approaches for implementations
discussed below. The EESM is an empirical method to assess
the performance of the OFDM communication over a given
realization of the frequencyselective fading channel. In order
to obtain the average performance, we need to average the
outcome of the EESM over a number of channel and
interference realizations as in a common link level simulator.
Therefore, the lower is the BER that we want to assess; the
higher is the number of channel and interference realizations
that we have to simulate to obtain reliable performance
estimation. On the other hand, the proposed method provides
directly the performance averaged over the fading, using the
(11) with (18) which is based on the eigenvalue decomposition
of the spacefrequency channel covariance R
d
. Monte Carlo
averaging is used in (24) only for handling the nonstationarity
of the interference (by including all the possible interference
structures), and it can also be adopted in case the channel is
modelled with a stochastic process. Usually, the number of
iterations required for such an average is meaningfully lower
than for link level simulators, or EESM method. It is
International Journal of Computer Trends and Technology (IJCTT) volume 4 Issue 9September 2013
ISSN: 22312803 http://www.ijcttjournal.org Page 3229
important to notice that numerical averaging is independent on
the BER target. Therefore the proposed framework provides
accurate performance at any BER values with low
computational complexity, and it can be cast easily for a wide
range of propagation environments avoiding any kind of
empirically tuned parameters.
VII. CONCLUSIONS
This paper proposed an analytical framework for the
performance assessment of bitinterleaved coded multi
antenna OFDMA systems over spacefrequency selective
fading channels with application to practical 4G broadband
wireless standards. The proposed analytical method provides a
direct average fading and interference.
REFERENCES
[1] F. Wang, A. Ghosh, C. Sankaran, P. Fleming, F. Hsieh,
and S. Benes,Mobile WiMAX systems: performance and
evolution," IEEE Commun.Mag., vol. 46, pp. 4149, Oct.
2008.
[2] D. Astely, E. Dahlman, A. Furuskar, Y. Jading, M.
Lindstrom, and S.Parkvall, LTE: the evolution of mobile
broadband," IEEE Commun.Mag., vol. 47, pp. 4451, Apr.
2008.
[3] H. Sampath, S. Talwar, J. Tellado, V. Erceg, and A.
Paulraj, A fourthgeneration MIMOOFDM broadband
wireless system design, performance, and field trial results,"
IEEE Commun. Mag., vol. 40, no.9, pp. 143149, Sep. 2002.
[4] W. Y. Zou and W. Yiyan, COFDM: an overview," IEEE
Trans.Broadcasting, vol. 41, no. 1, pp. 18, Mar. 1995.
[5] M. Russell and G. Stuber, Interchannel interference
analysis of OFDM in a mobile environment," in Proc. IEEE
Veh. Technol. Conf., vol. 2,July 1995, pp. 820824.
[6] S. Mumtaz, A. Gamerio, and J. Rodriguez, EESM for
IEEE 802.16e:WiMAX," in Proc. IEEE Int. Conf. Comput.
Inf. Science, 2008, pp.361366.
[7].Daniele Molteni and Monica Nicoli Interference mitig
ation in multicell LTE systems: performance over correlated
fading channelsDip. di Elettronica e Informazione,
Politecnico di Milano, Piazza L. da Vinci 32, I20133 Milano,
Italy.May,2009.
[8]. Hamid Jafarkhani, and A. R. Calderbank, SpaceTime
Block Codes from Orthogonal Designs Vahid Tarokh, IEEE
transactions on information theory, vol. 45, no. 5, july 1999.
[9]. Giuseppe Caire, Giorgio Taricco, Ezio Biglieri, Bit
Interleaved Coded Modulation IEEE transactions on
information theory, vol. 44, no. 3, may 1998.
[10]. Gerhard Gritsch, Hans Weinrichter, Markus Rupp. A
Union Bound of the Bit error ratio for data transmission over
correlated wireless MIMO channels IEEE Published in the
proceedings of ICASSP, Montral, Canada, May 1721, 2004.
BIBLIOGRAPHY:
Mr. MANOHAR NAIK RAMAVATH, received the B.Tech
degree in Electronics and Communication Engineering from
G.Pulla Reddy Engineering College, Sri Krishna Devaraya
University, Anantapur, A.P in 2008. Currently he is pursuing
M.Tech in G.PULLAIAH COLLEGE OF ENGG & TECH
KURNOOL, JNTU Anantapur, A.P. His current research
interests include Antennas and Wireless communication.
Sri. B.A.SARATH MANOHAR BABU, received the B.Tech
degree in Electronics and Communication Engineering from
Madina Engineering college, JNTU Anantapur A.P. and
received his M. Tech Post Graduate in Digital Electronics and
communication system from Madanapalli Institute Of
Technology, JNTU Anantapur A.P. He has Life member ship
in M.I.E.T.E. and M.I.E.I. He published one technical paper
international journal of Electronics and Communication
technology. Currently he is working as Associate Professor
of ECE Department in G.PULLAIAH COLLEGE OF ENGG
& TECH KURNOOL, JNTU Anantapur, A.P. His current
research interests include VLSI System Design