Efficient Sphere-Decoding for Soft-Output Multiple-Symbol Differential Detection of DPSK

The sphere decoder (SD) is a powerful algorithm to reduce the computational complexity for multiple-symbol differential detection (MSDD) of differential phase-shift keying. To further increase the quality of the transmission, soft-output sphere decoding produces reliability information in the form of LLRs which are fed into the channel decoder. In this paper, we compare the basic SD algorithm with two variants, centered metric computation and LLR accumulation, with respect to performance and computational complexity. We show that especially the latter is able to improve the performance significantly and additionally allows a flexible tradeoff between performance and complexity by adjusting the number of accumulated LLRs.