Hybrid Precoding Processor for Millimeter Wave MIMO Communications

Abstract

Millimeter wave (mmWave) multiple-input multiple-output system is one of the most important techniques to increase throughput for 5G communications. To overcome the severe pathloss in the mmWave channel, the hybrid analog/digital precoding technique jointly designs radio frequency (RF) beamforming vectors and a baseband precoding matrix to approach the optimal spectral efficiency with reduced RF chain power and baseband complexity. This brief presents a hybrid precoding algorithm based on multiple orthogonal beamforming codebooks and a local search scheme, which reduces computational complexity by 40% with negligible bit-error-rate performance loss compared with the state-of-the-art algorithms. The proposed hybrid precoding algorithm was also designed and implemented by using TSMC 90-nm CMOS process technology. The fabricated hybrid precoding processor chip achieves 11.1M/10.4M/6.9M/4.3M matrices per second for 1/2/3/4 data streams, respectively. Compared with a previous work, the designed processor increases throughput by 65%/55%/40%/7.5% for 1/2/3/4 data streams with only 28% of the core area.