Performance evaluation of large-scale MIMO systems with compressed channel feedback

In the paper, a large-scale (massive) multiple-input multiple-output (L-MIMO) system is considered. To support multiple users with multiuser (MU) precoding, the channel state information (CSI) needs to be known at a transmitter. To do this, the CSI needs to be fed back from a receiver, yet the feedback requires a significant overhead of the networks. To reduce the feedback CSI, we modify an interpolation method of an existing, time-domain based CSI compression method and evaluate its bit-error-rate (BER) performance. Numerical results indicate that the modification can improve BER performance significantly. For example, when a compression ratio is fixed by 20 and a zero-forcing (ZF) MU-MIMO precoding is employed, the uncoded BER with 16-quadrature amplitude modulation can be improved from 3×10^-5 to 3×10^-7. We also provide performance comparison with various MU-MIMO linear precoding schemes including ZF, minimum mean-squared-error, and conjugate-beamforming based methods, which guides further extension of this work.