Look-Ahead Sphere Decoding: Algorithm and performance evaluation

Multiple Input Multiple Output (MIMO) systems are recognized as a key enabling technology in high performance wireless communications. Among known MIMO detectors, the Sphere Decoding Algorithm (SDA) is capable of optimal performance with a reduced computational complexity, compared to a Maximum Likelihood (ML) detector. However the hardware implementation of the SDA still is a major design challenge, especially with high throughput constraints. This paper proposed a modified version of the SDA, which achieves a significant throughput increase and guarantees ML performance, at the price of limited additional complexity. This new solution, called LASDA (Look-Ahead SDA) is mainly based on formal algorithm transformations, namely look-ahead, pipelining and retiming, applied at signal processing level. In addition, a modified tree search strategy, called test&restart, allows a limited increase of the iteration number. The analysis of the proposed LASDA shows that it is almost twice as fast as SDA, while simulation results exhibit ML performance.