Asymptotic analysis of MIMO wireless systems with spatial correlation at the receiver

With a unified approach, this paper investigates the asymptotic performance or, equivalently, the large-system properties, of various point-to-point systems with antenna-array-based multiple-input multiple-output (MIMO) channels having spatial correlations. Using the replica method originally developed for statistical physics, we provide analytical solutions to the input-output mutual information of MIMO systems: 1) at the transmit side, with arbitrary inputs, and 2) at the receive side, with either the optimum space-time joint decoding or various suboptimum spatial equalizers followed by a bank of temporal decoders. Important physical meanings revealed though the analytical solutions to those more practical combinations, such as how the input-output mutual information is affected by the channel spatial correlations, are highlighted along our derivation. Moreover, we provide a novel waterfilling algorithm to determine the data-rate-maximizing transmit signal covariance matrix when only the channel long-term spatial correlations are available at the transmitter.