Cryptography Reference
In-Depth Information
•
The first uses an implementation of the turbo decoder in the form of
a
pipeline
by cascading several elementary modules, each module imple-
menting one detection and one decoding iteration.
•
The second uses a single hardware module, implementing the successive
iterations sequentially, by looping back on itself.
The first architecture presents a smaller latency, so is better adapted to applica-
tions requiring high data rates. On the other hand, the second solution enables
an economy in the number of transistors and therefore in the silicon surface. In
order to further reduce the surface used, some authors have proposed sophisti-
cated architectures enabling part of the SISO algorithm to be shared between
the equalizer and the decoder, despite the different structure of the trellises con-
cerned [11.36]. This approach also enables a reduction in length of the critical
path, and therefore in the global latency of the system. This last factor can be a
major obstacle to the practical implementation of turbo equalization (and turbo
systems more generally) since not all applications may tolerate an increase in
the processing delay at reception. Resorting to analogue electronics will perhaps
soon enable this obstacle to be overcome. An analogue implementation of a sim-
plified MAP turbo equalizer has thus been reported in [11.24], with promising
results.
From the algorithmic point of view, the application of MAP turbo equaliza-
tion to the GSM system has been the subject of several studies [11.15, 11.43,
11.6, 11.18]. The traditional turbo equalization scheme must thus be revised in
order to take into account the specificities of the standard (inter-frame interleav-
ing, different levels of protection applied to the bits at the output of the speech
encoder, GMSK modulation, ...). Simulation results show generally moderate
gains in performance, in return for a large increase in the complexity of the
receiver. This can be partly explained by the fact that the conventional GSM
system faces only a limited level of ISI on the majority of the test channels
defined in the standard. On the other hand, the introduction of 8-PSK modu-
lation in the context of EDGE greatly increases the level of interference. This
scenario therefore seems more appropriate for the use of turbo equalization, and
has given rise to several contributions. In particular, the authors of [11.40]
3
have
studied the implementation of a complete turbo equalization system relying on
a SISO equalizer of the DDFSE type with pre-filtering, coupled to a channel
estimator. They have obtained gains of the order of several dB, depending on
the modulation and coding scheme considered, compared with the performance
of the classical receiver. Furthermore, they have also proved the fact that the
equalization and iterative decoding principle could be carefully exploited in the
context of the ARQ retransmission protocol defined in EDGE (the
Incremental
Redundancy
scheme) to improve the global quality of service at reception.
3
See also the references cited in this article.
