A high throughput K-best detector on FPGA

The multiple-input-multiple-output (MIMO) technique is widely used in modern wireless communication systems because it greatly increases system capacity and improves communication reliability. However, MIMO detection is a challenging task. Sphere detection algorithms are preferred in practice as these can achieve bit error rate performances close to that of the maximum likelihood algorithm. Sphere detection algorithms also offer affordable computational complexity. In this paper, we propose a layer-wise detector framework on the basis of the K-best detection algorithm for 4×4, 16-QAM MIMO systems. The kernel module of the framework is based on the layer-processing module. In this case, we fully pipelined the module with a parallel sorting unit. By using the layer-processing module, we can assemble different detectors with different performances and cost trade-offs. We deployed four detectors with different amounts of layer processing modules. All detectors receive a throughput of over 300 Mbps. The detector with eight layer modules was fully pipelined. This detector generates one result per clock cycle and obtains a throughput of up to 2.68 Gbps at 168 MHz.