The relative simplicity of the Soft Output Viterbi Algorithm (SOYA) in Turbo code decoding makes it more efficient in hardware implementation, compared with the more complex Maximum a Posterior (MAP) Algorithm. However, the performance of SOYA-based decoding is generally inferior to that of MAP-based decoding in terms of Bit Error Rate (BER) and Frame Error Rate (FER). Therefore, for the emerging applications of turbo code, it is very desirable to improve both the performance and the hardware efficiency of the SOVA-based Turbo code decoding.
We propose a Bi-directional SOVA (BI-SOYA) and a serially mixed SOYA (SM-SOVA) consists of the original SOVA (a forward SOYA) and the backward directional SOYA (a backward SOVA) to improve the performance of SOVA based Turbo decoding.
At first, the backward SOVA is compared with the forward SOVA in turbo code decoding. We find noticeable performance improvement for the backward SOYA when it is not terminated, which turns out to be due to a smaller reliability value, indicating that the termination conditions of the turbo encoder strongly affect the performance of the backward SOYA decoder.
Bi-directional SOYA (BI-SOYA) consists of the original SOYA and the backward SOVA in parallel. The advantage of our BI-SOVA is that we do not need either parameter extraction or threshold value evaluation. The disadvantage is that the required area is about two times larger. The simulation results show that out BI-SOVA produce performance similar to that achieved by Max-Log MAP algorithm.
We also propose a hardware efficient serially mixed SOYA decoder composed of the forward SOYA decoder and the backward SOVA decoder. Although the SM-SOVA is similar to BI-SOYA, in combining forward and backward SOVA``s together to improve BER performance, its hardware algorithm is quite different. First, while both forward and backward decoders are required for the Bi-SOVA, in the SM-SOVA, the backward SOYA decoding can be done in the forward SOYA decode...