Conceiving Extrinsic Information Transfer Charts for Stochastic Low-Density Parity-Check Decoders

Abstract

[EN] Stochastic low-density parity-check decoders (SLDPCs) have found favor recently both for correcting transmission errors as well as for improving the hardware efficiency. The main drawback of these decoders is that they require hundreds of time periods to decode each frame, but their chip area is smaller than that of their fixed-point counterparts, so they can achieve higher hardware efficiency and may consume less energy. In this paper, we propose a novel extrinsic information transfer chart technique for characterizing the iterative decoding convergence of all the sequences involved in the SLDPC. We have conceived a new model, which takes into consideration not only the sequences exchanged between the decoders but also the sequences generated inside the variable-node decoder (those which are stored in the edge memories). In this way, the model is able to predict the number of decoding iterations required for achieving iterative decoding convergence, as confirmed by own decoder simulations. The proposed technique offers new insights into the operation of SLDPCs, which will facilitate improved designs for the research community.