A regressive algorithm for multiple-symbol differential-detection of DPSK

Research in early 1990's showed that multiple-symbol demodulation of DPSK [1] could provide a gain of 1 to 2 dB over conventional DPSK techniques. However, the demand on additional hardware is very high. We propose a regressive algorithm that takes advantages of the redundancy in demodulated data to improve upon the ''optimal'' algorithm by Divsalar [1]. This algorithm also has two modes of operation, which cuts down the error propagation. By selecting only the most essential information needed in ML-DD[1], it can also reduce the receiver hardware significantly. Computer simulation results are also presented. With reduced hardware complexity, our algorithm still outperforms the ''optimal'' ML-DD algorithm.