Skip to Main Content
A method of combining ZF-SIC (MMSE-SIC) and a reduced ML search is proposed (named ZF-SIC-SEARCH and MMSE-SIC-SEARCH) which gives V-BLAST decoding performance close to that of ML (maximum likelihood) while retaining the complexity comparable to that of ZF-SIC (MMSE-SIC). We show that ZF-SIC-SEARCH is same as ML algorithm for Mt = 2 (Mt denotes number of transmitting antennas). Our algorithm first determines a reduced set of candidate solution vectors and then chooses the one nearest (in the sense of Euclidean distance) to the received vector. For Mt = 2 in ZF-SIC-SEARCH, the diversity order analysis of one of these solution vectors named as true mapped vector shows that the true mapped vector enjoys additional diversity due to selection combining. We also present a method to further reduce complexity for higher order constellations by formulating the problem in an equivalent real domain. This alternative, though inferior to ML, still has significant performance improvement over conventional ZF-SIC/MMSE-SIC. Simulation results are presented for 5times5 MIMO system.