R. Nordin, S. Armour, J.P. McGeehan, “Overcoming Self-Interference in SM-OFDMA with ESINR and Dynamic Subcarrier Allocation”, Proceedings of IEEE 69th Vehicular Technology Conference, 2009. VTC Spring 2009, pp. 1-5, Apr. 2009

1. Overcoming Self- Interference in SM-OFDMA
Vehicular Technology Society
with ESINR and Dynamic Subcarrier Allocation
Rosdiadee Nordin, Dr. Simon Armour and Prof. Joe McGeehan
Centre for Communications Research Laboratory, University of Bristol
Rosdiadee.Nordin@bristol.ac.uk
1. INTRODUCTION 5. SIMULATION PARAMETERS
 Combined benefits of MIMO architecture and OFDM(A) technology make it an  Using a 2 × 2 SM-OFDMA system, with 64 QAM, ¾ rate modulation
attractive technology for future wireless communications, such as 3GPP LTE, scheme in an uncorrelated Rayleigh fading channel environment.
WiMAX (802.16) and MBWA (802.20).
 Adopted ETSI-BRAN Channel Model ‘E’, which simulates typical large open
 However, MIMO suffers from severe impairment effects due to Co-channel space outdoor environments for NLOS conditions, with excess delay spread
interference (CCI), also known as self-interference, inherent in the MIMO of 1760ns, sampling period of 10ns and RMS delay spread of 250ns.
architecture.
 This research aims to minimized the CCI effect in the MIMO-SM scheme by  Simulated for 16 MS users and Operating frequency 5 GHz
exploiting the knowledge of the channel gain and combine it with Dynamic each MS is allocated a single sub- Bandwidth 100 MHz
Subcarrier Allocation (DSA). channel consisting 48 useable FFT Size 1024
Useful Subcarriers 768
subcarriers. Subcarrier spacing 97.656 KHz
10.24 s
 2000 independent identically Useful symbol duration
2. OBJECTIVES distributed (i.i.d.) quasi-static
Total symbol duration 12.00 s
Punctured ½ rate convolution code,
Channel Coding
 Utilise the multi-user channel to mitigate channel fading and exploit it as a source random channels samples per constraint length 7, {133, 171}octal
of diversity. user.
 Consider the interference that exists between the spatial sub-channels.
6. RESULTS& ANALYSIS
 Provide fair benefit across all users.
Probability(Normalised perceived ESINR metric > abscissa
1
Channel Gain
0.9 ESINR
0.8
Random  From the CCDF comparison, DSA-ESINR
3. SYSTEM MODEL metric outperforms the random allocation
0.7
0.6
strategies by up to 7 dB.
H1
Transmitter at
X1
X1
With index 1
Tx1 channel 1
Rx1 0.5  DSA-channel gain has slightly lower gain;
Base Station
X1 OFDM
H3
channel 3
0.4 approximately 2 dB compared to DSA using
User k Input
With index 2
H2
0.3 the ESINR metric.
Data Scrambling/ FEC/ Symbol Serial to Parallel& DSA channel 2 0.2
Puncturing/ Interleaving Mapping Spatial Multiplexing mapping X2 Tx2
X1X2
With index 3 0.1
H4 Rx2
X2 X2 OFDM channel 4
0
-4 -2 0 2 4 6 8 10 15
With index 4
Normalised perceived ESINR metric (dB)
Index 1 2 3 4 10
ESINR from DSA
Index 4
 DSA-ESINR have consistent high ESINR 5
Normalised ESINR metric (dB)
Index 3
other users Scheme
Index 2 metric compare to DSA-channel gain. 0
Index 1
-5
 This shows that DSA-channel gain does
ESINR1
ESINR2
-10
not consider the effect of self-
G2 interference (and its impact upon system -15
ESINR
G1 capacity and BER) for the selected -20
Receiver at calculation
G2 DSA
Mobile Station k
MMSE info. G1DSA subcarriers. -25
Channel Gain
-30 DSA
S1 Y1 DSA
OFDM DSA-ESINR
Deinterleaving/ S1 S2 Parallel to MMSE Demapping -35
User k Symbol 0 100 200 300 400 500 600 700
Depuncturing/ Viterbi Serial& De- Linear 0
Output Data Demapping Y2 10 Subcarrier index
Decoding/ Descrambling Multiplexing S2 Detection DSA Random
Demapping OFDM
10
-1
Channel Gain
ESINR
 DSA-ESINR has better BER performance
than DSA-channel gain by 4 dB (at 10-5 BER).
4. METHODOLOGY 10
-2
 This implied that DSA-ESINR has the
benefit of minimizing the effect of self-
BER
-3
10
Allocation strategy interference, resulting in improvement of
10
-4
BER.
ESINR metric has the knowledge of self- DSA allows user with the lowest ESINR to 10
-5
interference from the other transmitted have the ‘best’ ESINR that is available for the -6
4 Channel Gain
ESINR
10
data stream within the same channel. next transmission. Random
Average (dB)
0 5 10 15 20 25 2
SNR(dB)
0
-2
(1) Initialization  DSA using the ESINR metric has the -4
2 4 6 8 10 12 14 16
smallest variance and the largest average User number
Set ESINRk,m’=0 for all users, k=1…K; Set allocated subcarrier index, channel response, compared to other
15
Channel Gain
Ck,s,m’=0 for all users, whereby s is the allocated subcarriers and types of allocation algorithm.
ESINR
Random
10
spatial sub-channels, m’={1,2,M’}; Set s=1
Variance
 It shows that the proposed algorithm 5
(2) Main Process offers fairness whilst maximizing the
0
While Nm’≠0Nsub= average channel gain of any given user 2 4 6 8
User number
10 12 14 16
{ (a) Sort users, start with users with less ESINR metric. Find user without minimizing it in the other users. Algorithm ESINR Channel gain Random
k satisfying: Mean (dB) 1.815 1.658 -1.336
ESINRk,m’ ≤ ESINRi,m’, for all i, 1≤i≤ K Variance 0.6061 1.882 4.972
(b) For the user k in (a), find subcarrier n satisfying: 9
ESINR-DSA
Gk,m’ ≥ Gj,m’, for all j N 8
ESINR-Random
ESNR-DSA
(c) Update ESINRk,m’, N and Ck,s,m’ with the n from (b) according to 7
ESNR-Random
Capacity bound  Average data rate performance for both the
ESINRk,m’ = ESINRk,m’+ Gk,m’ ESINR and ESNR systems increases at
Throughput (bps/Hz)
6
approximately twice the capacity of random
Nm’=Nm’-n 5
subcarrier allocation.
Ck,s,m’=n 4
 The proposed algorithm is able to achieve
s=s+1 3
the desired balance between fairness and
(d) Go the next user in the short list define in (a), until all 2 system capacity.
users are allocated another best subcarrier, Nm’=0 1
-2 0 2 4 6 8 10 12 14
} Eb/N0 (dB)
7. CONCLUSIONS
 Initial investigation has revealed that the next generation of wireless system would be capable of improving the capacity performance while
providing fair gain and improved BER performance.
 The proposed algorithm considers co-antenna interference within a SM-OFDMA system.
 Future work will focus on investigating BER and capacity performance of the system in correlated channels where the effect of self-
interference is more dominant. Based on the initial results, the proposed algorithm can be expected to provide even greater benefits.
ACKNOWLEDGEMENT: Travel grant for this conference is funded by University of Bristol’s Alumni Foundation, whose financial support is gratefully acknowledged.