Improved Decoding with the Bi-Directional SOVA for Turbo Codes

Abstract

The soft-output Viterbi algorithm (SOVA), which is a soft-input soft-output (SISO) Viterbi algorithm, is widely adopted for turbo decoding by taking advantage of its moderate complexity at the expense of some performance loss. Since the coding gain achieved by SOVA is generally about 0.7 dB less than that of maximum a posteriori (MAP) decoding, some research has been done to improve the performance of SOVA decoding for turbo codes. Bi-directional SOVA (Bi-SOVA) was proposed to be a good algorithm, approaching max-log-MAP, or even better. However, the performance improvement of Bi-SOVA degrades when the signal-to-noise ratio (SNR) gets higher and performs like a conventional SOVA. We present an improvement of the Bi-SOVA decoding scheme for turbo codes by using a scaling factors scheme. It is concluded that Bi-SOVA with scaling factor can achieve a stable performance that is better than max-log-MAP within a wide SNR range. In addition to BER performance, the computational complexity of Bi-SOVA with scaling factor is examined and discussed.