CSI Feedback Based on Spatial and Frequency Domains Compression for 5G Multi-User Massive MIMO Systems

Accurate channel state information (CSI) acquisition with less feedback overhead is a hot topic for multi-user massive multiple-input multiple-output. In 3GPP Release 15 New Radio, Type II codebook which consists of wideband beam group and beam combination coefficients (BCCs) of subband has been introduced to obtain high resolution CSI for performance improvement. But it requires large feedback overhead. In this paper, to reduce feedback overhead, CSI feedback based on spatial domain and frequency domain (FD) compression by using their correlations is studied. We first propose the BCCs are calculated based on channel covariance matrix projecting onto wideband beam group to enhance system performance. In order to make them be compressed effectively, a phase preprocessing algorithms is adopted before performing compression in FD. Then, the obtained compression coefficients after compressing BCCs are quantized to report to base station. To achieve better tradeoff between performance and overhead, a variable bitwidth which is exploited to quantize phases of compression coefficients is proposed. System-level simulation results show that the proposed calculation method of BCCs could improve system performance, and variable bitwidth quantization could achieve better performance with less feedback overhead compared with conventional fixed bitwidth quantization.