Massive MIMO real-time channel measurements and theoretic TDD downlink throughput predictions

This paper is the first to evaluate the realistic performance of Massive MIMO using real-time channel measurements conducted at 3.51GHz. Our testbed consists of 64 National Instruments Universal Software Radio Peripheral Reconfigurable Input and Output (NI USRP RIO) software defined radios (SDRs) with 128 dipole antennas at the Base Station (BS) side, and 6 USRPs configured as 12 single-antenna User Equipment (UE) nodes. The system operates in a Time Division Duplex (TDD) fashion with a Long Term Evolution (LTE) like Physical Layer framework which, when fully deployed, will enable instantaneous throughput measurements on the Uplink (UL) and Downlink (DL) according to a user-defined frame schedule. In this paper we present real-time UL channel measurements at all BS antenna ports simultaneously, which allows the temporal characteristics of the Massive MIMO channel to be recorded. Eigen-beamforming precoding and a Received Bit-level Mutual Information Rate (RBIR) abstraction simulator are then used to predict the theoretic DL performance. We quantify the co-located Massive MIMO throughput for differing numbers of BS antennas (i.e. 32, 64 and 112 elements) and UE locations for an indoor environment.