Precoding for TDD and FDD in Measured Massive MIMO Channels

This paper analyzes the performance of well-known precoding schemes for massive multiple-input multiple-output (MMIMO) systems. The investigations are based on extensive measurements made with a sounding system capable of capturing the dynamic channels towards users moving in many different outdoor scenarios. Assuming ideal channel state information (CSI), results show that the mean sum-rate of the maximum ratio transmission (MRT) precoder varies considerably with the scenario, e.g., from 6.5 to 14.5 bit/s/Hz (10%- and 90%-percentiles) for a 64 element uniform linear array (ULA) at the base station (BS), while the zero-forcing (ZF) and signal to leakage and noise ratio (SLNR) precoders are more robust and higher performing with variation from 13.4 to 16.3 bit/s/Hz in the same conditions. However, when the CSI is non-ideal the performance drops. With the CSI delayed corresponding to movement of about 1/5 of a wavelength, the ZF and SLNR mean sum-rate is 60-92% of that achieved with ideal CSI (10%- and 90%-percentiles). More statistics for different massive array sizes with both delay and frequency offset CSI are given in the paper.