Coupled LDPC codes: Complexity aspects of threshold saturation

Abstract

We analyze the convergence behavior of iteratively decoded coupled LDPC codes from a complexity point of view. It can be observed that the thresholds of coupled regular LDPC codes approach capacity as the node degrees and the number L of coupled blocks tend to infinity. The absence of degree two variable nodes in these capacity achieving ensembles implies for any fixed L a doubly exponential decrease of the error probability with the number of decoding iterations I, which guarantees a vanishing block error probability as the overall length n of the coupled codes tends to infinity at a complexity of O(n log n). On the other hand, an initial number of iterations I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">br</sub> is required until this doubly exponential decrease can be guaranteed, which for the standard flooding schedule increases linearly with L. This dependence of the decoding complexity on L can be avoided by means of efficient message passing schedules that account for the special structure of the coupled ensembles.