An Improved K-Best Sphere Decoding Architecture for MIMO Systems

The sphere decoding algorithm (SDA) has been used for achieving maximum likelihood (ML) detection for today's multiple-input multiple-output (MIMO) systems. However, normal SDA suffers from variable computation complexity and non-fixed throughput. The K-best SDA is proposed for MIMO detections for its less complexity and fixed throughput. Most of the K-best SDA complexity comes from the path cost computation and sorting operations at each layer. In this paper, we first introduce a reduced complexity K-best SDA, based on Schnorr-Euchner (SE) strategy. Simulations show that for 4 x 4 64QAM system, we can save 25% path cost computation and 27% sorting operations with almost no performance loss. In addition, by exploiting the natural partial sorted results coming from the SE method, we develop an architecture applying a modified rank order filter to the sorting operation for the K-Best SDA. Compared with bubble sort algorithms, our architecture can reduce almost 50% sorting complexity, which is a significant contribution to the K-best SDA implementation for MIMO systems.