A Very Near Optimum Interference Cancellation Algorithm for CDMA Signals

Abstract

In this paper a new algorithm for detection of DS/CDMA signals is presented that by definition of an annealed probability for users' data estimated in each iteration a fraction of multiple access interference (MAI) proportional to this probability is cancelled from the output of the matched filters. This algorithm is simulated and compared with some of the well-known multi-user detectors such as minimum mean square error (MMSE), decorrelating, maximum likelihood (ML), and single user lower bound for randomly spread signals which is an efficient spreading model for long codes or code-hopping based DS/CDMA systems in under-load, full-load, and over-load conditions. It is shown that the proposed algorithm has optimum performance in under- load conditions and that it has only about 1 dB difference in full-load conditions. In all cases, the proposed algorithm outperforms the other well-known multi-user detectors. The proposed algorithm is simulated for large number of active users in full-load condition and its performance is compared with other suboptimum multi-user detectors, and single user lower bound. It is observed that the proposed algorithm greatly improves the performance and is very close to single user lower bound especially in the high SNR region. Finally, the proposed algorithm is also simulated for combating near-far problem. It is shown that the performance loss of the proposed algorithm as compared with the optimum detector is negligible