Further Improvements in SOVA for High-Throughput Parallel Turbo Decoding.

Abstract

In this letter, we present two mechanisms based on soft-output Viterbi algorithm (SOVA) technique that achieve a performance close to maximum a-posteriori (MAP) decoding. First, we propose a new technique called alternating direction SOVA, which allows for improving the performance of SOVA in similar terms as bidirectional SOVA while barely affecting the area complexity. Then, we introduce the idea of parallel turbo decoding based on path metric/state border exchange. Unlike other techniques, overlapping between adjacent subblocks is not required, thus reducing the decoding latency dramatically. We evaluate their performance in the context of 3GPP-LTE, showing that the combination of both mechanisms achieves a BER performance close to parallel MAP decoding while having a lower complexity