Two-level-search sphere decoding algorithm for MIMO detection

Breadth-first search sphere decoding (SD) is considered as one of the most promising methods for multiple-input multiple-output (MIMO) detection due to its low computational complexity, capability of providing fixed throughput and pipelining features. However, the performance of the conventional breadth-first search SD is significantly degraded if the number of candidate symbols kept at each level of the tree search is not sufficiently large. To solve this problem, we present a novel two-level-search sphere decoding algorithm for MIMO systems. In our proposed algorithm, symbol detection is concurrently performed on two levels of the tree search, and the search space is extended to find good candidate symbols. The use of simultaneous two-level search is to greatly reduce the possibility of discarding good candidates at early stages. Compared to the traditional breadth-first search SD algorithm, our approach provides a performance gain of about 2.5 dB at a bit-error-rate (BER) of 10^-4. Our work also exhibits a low computational complexity in term of the number of multiplications. Furthermore, we also discuss some VLSI implementation aspects of the proposed algorithm for practical applications.