Optimizations in max-log-MAP LLRe VLSI architecture

One of the most popular channel iterative decoding systems is the BCJR (MAP) decoding algorithm. It is used in turbo code schemes, and has enabled them to reach 1dB from Shannon´s theoretical limit. This kind of algorithm requires a large area and, therefore, any area saving, without penalty in critical path or BER performance, would be worthwhile. A new scheme is presented for the computation of LLR^e values in MAP-based systems. Based on exploiting the convolutional code Trellis properties, this scheme saves from 27 to 52% of the required area in LLR^e calculations, with a small penalty in critical path and without any change in BER performance.