On the error statistics of turbo decoding for hybrid concatenated codes design

In this paper we propose a model for the generation of error patterns at the output of a turbo decoder using a Context Tree based modelling technique. This model can be used not only to generate the decoder error pattern behaviour with little effort, avoiding simulations, but also to investigate – with no need of performing neither a turbo code distance spectrum analysis, nor the probabilistic characterization of log-likelihood ratios or of the extrinsic information at a turbo decoder output – the performance of hybrid concatenated coding (HCC) schemes having a turbo code as component code. These coding schemes combine the features of parallel and serially concatenated codes and thus offer more freedom in code design. It has been demonstrated, in fact, that HCCs can perform closer to capacity than serially concatenated codes while still maintaining a minimum distance that grows linearly with block length.