EXIT-Optimized Index Assignments for Turbo Decoders with Unreliable LLR Transfer

Abstract

This letter considers a Turbo decoder for which the log likelihood ratios (LLRs) exchanged between constituent decoders are subject to independent bit errors. The resilience of the decoder against these errors can be improved - without increasing decoding complexity - by optimizing the index assignment (IA) of the LLRs with respect to the extrinsic information transfer (EXIT) characteristic of the decoder. To enable an effective optimization of the IA we propose an approximation of the mutual information of the extrinsic decoder output that can be computed efficiently without histogram measurements. Numerical results in terms of bit error rates are reported for different scenarios to validate the proposed approach.