22-09-2012, 12:35 PM
EXIT Charts Aided Hybrid Multiuser Detector for Multicarrier Interleave Division Multiple Access
1EXIT Charts Aided Hybrid.pdf (Size: 122.4 KB / Downloads: 19)
Abstract
A generically applicable hybrid MultiUser Detector
(MUD) concept is proposed by appropriately activating different
MUDs in consecutive turbo iterations based on the Mutual
Information (MI) gain. It is demonstrated that the proposed
hybrid MUD is capable of approaching the optimal Bayesian
MUD’s performance despite its reduced complexity, which is at
a modestly increased complexity in comparison to that of the
suboptimum Soft Interference Cancellation (SoIC) MUD.
INTRODUCTION
The Interleave Division Multiple Access (IDMA) [1], [2]
system exchanges the classic position of Direct Sequence
(DS) spreading and interleaving employed in traditional coded
CDMA systems, leading to chip-interleaving instead of bitinterleaving,
where users are differentiated by their unique
user-specific chip-interleavers. There are several advantages
associated with IDMA system. It is observed that the userspecific
chip-interleavers effectively enlarge the free-distance
of the channel code employed and the resultant codewords
of the different users constitute unique noise-like random
signatures [3]. The employed chip-interleaving is also capable
of increasing the achievable time diversity in time-selective
channels, where the burst errors are dispersed and hence the
achievable decoding capability is improved [4]. Furthermore,
IDMA is capable of striking the best tradeoff between the
channel-coding rate and DS-spreading factor, hence improving
the attainable error-resilience [5].
HYBRID DETECTOR DESIGN
EXIT Charts Analysis
EXIT charts were introduced by ten Brink [14], which are
widely used in the analysis of turbo-type iterative systems.
This technique relies on computing the MI of the constituent
soft components. It evaluates the nonlinear function Ie =
χ(Ia), which maps the input a priori MI Ia ∈ [0, 1] to the
output extrinsic MI Ie ∈ [0, 1], where the amount of output extrinsic
MI Ie gleaned from the input a priori MI Ie determines
the convergence behaviour of this soft component. Since the
extrinsic information generated by the first component acts as
the a priori information for the second component and vice
versa, in the EXIT charts we alternately swap the abscissa and
ordinate axes, depending on which of the two components acts
as the source of a priori information, corresponding to the
abscissa.
Mutual Information Gain
For the sake of determining the activation instant of the
different MUDs, we have to monitor the MI gain of the MUDs
between the consecutive iterations at the receiver. In practice,
the MI is generally unknown at the receiver and there is a
difference between the EXIT charts based MI prediction and
the practical decoding trajectory. Thus, we have to contrive
a realistic measure of the MI gain used for controlling the
activation of the constituent MUDs. This can be achieved by
estimating the extrinsic LLRs’ average magnitude E(|Le
mud|)
at the output of the MUD.
SIMULATION RESULTS
1) Performance: We now investigate the attainable performance
of the proposed hybrid MUD in comparison to the two
constituentMUDs.We employ a rate-1/2 convolutional code in
conjunction with a rate-1/2 NSC code, resulting in an overall
code-rate of r = 1/4. An information packet length was set
to 1024 and a 3 path SWATM channel was considered, where
the number of subcarriers and the length of CP were set to
128 and 32, respectively.. Fig. 5 portrays the attainable BER
performance of the SoIC MUD, of the Bayesian MUD and
of the proposed hybrid MUD as a function of both Eb/N0 as
well as of the number of users K. Observe in Fig. 5 that at a
high normalized user-load of β = Kr = 12/4 = 3, the SoIC
MUD fails to converge, while the proposed hybrid MUD and
the Bayesian MUD achieves an infinitesimally low BER for
Eb/N0 > 10dB.
CONCLUSION
In this paper we proposed a novel hybrid MUD scheme
by switching between different MUDs based on the MI gain.
Our scheme combines the benefits of both low complexity and
high performance and strikes an attractive trade-off between
the constituent MUDs employed as benchmarkers. Simulation
results quantified the performance advantage of employing the
hybrid MUD, at normalized user-load of β = 3, the SoIC
MUD failed to converge, whereas the proposed hybrid MUD
achieved an infinitesimally low BER, which was similar to that
of the optimal Bayesian MUD, while imposing only about half
the complexity.