MMSE-based MIMO precoder using partial channel information

This paper presents a precoder design based on the minimum mean squared error (MMSE) criterion and using the knowledge of only the transmit and receive correlation matrices of the underlying MIMO channel. The optimal transformations are shown to be eigen-beamformers, which transmit the signal along eigenvectors of the transmit correlation matrix. Power loading across the eigenbeams are determined based on eigenvalues of both transmit and receive correlation matrices and can be viewed as a waterpouring policy. Performance of the proposed precoders using partial channel information in various MIMO channels is evaluated and compared to that of precoders based on full channel knowledge at transmitter. It was shown that as the number of the transmit and receive antennas increases, the MMSE-based MIMO precoders using partial channel knowledge can achieve the performance of precoders using full channel knowledge. Furthermore, the channel that gives the best performance is the one with only one strong eigen mode for cases with a large number of receive antenna or a small number of transmit antenna. For small number of receive antennas or large number of transmit antennas, the full rank channel with equal gain eigen modes will gives the best performance.