Maximum likelihood detection of spatially multiplexed signals via loosely ordered depth-first sphere decoding

The complexity of depth-first sphere decoders (SDs) is determined by the employed tree search and pruning strategies. Proposed is a new SD approach for maximum-likelihood (ML) detection of spatially multiplexed, high-order, QAM symbols. In contrast to typical ML approaches, the proposed tree traversal skips the computationally intensive requirement of visiting the nodes in ascending order of their partial distances (PDs). Then, a new pruning method efficiently narrows the search space and preserves the ML performance despite the non-ordered tree traversal. This proposed approach results in substantially reduced PD calculations when compared to typical ML SDs and, for high SNRs, the necessary calculations can be reduced down to the number of transmit antennas.