Efficient SOR-based detection and architecture for large-scale MIMO uplink

Affected by relaxation factor ω, for large-scale MIMO uplink, successive over relaxation (SOR) detection involves low-complexity matrix inversion but unstable performance as ω changes. In this paper, a more stable and efficient SOR-based detection, which is nearly unaffected by ω, is proposed. First, the convergence of the proposed method is proved. Both analytic and numerical results have shown that, with lower computational complexity, its performance 1) stays almost the same when ω varies, and 2) is very close to that of linear minimum mean square error (MMSE) algorithm. Finally, the corresponding VLSI architecture is proposed in detail.