Practical aspects of iterative decoding
- « Previous Article
- Table of contents
- Next Article »
The convergence problem of iterative, block-mode, turbo decoders is discussed and the performance of a practical convergence criterion is presented. A fixed-point approach is used, whereby the saturation and stability characteristics of the extrinsic-probability vector for each MAP decoder are determined by simulation and used to terminate iteration. If these vectors are saturated and identical, or non-saturated and stable, the decoder has converged to a fixed point. The paper also examines the effect of interleaver design and machine precision effects on convergence. Sometimes, finite precision can lead to a limit-cycle effect, and practical solutions are discussed. Once convergence has been established, it can also be used to determine with high confidence the effective dfree of the decoder, even for large block lengths.
