Correlation across transmit antennas in multiple-input multiple-output (MIMO) systems has been studied in various scenarios and has been shown to be detrimental or provide benefits depending on the particular system and underlying assumptions. In this dissertation, we investigate the effect of transmit correlation on the capacity of the Gaussian MIMO broadcast channel (BC), with a particular interest in the large-scale array (or massive MIMO) regime. To this end, we introduce a new type of diversity, referred to as transmit correlation diversity, which captures the fact that the channel vectors of different users may have different, and often nearly mutually orthogonal, large-scale channel eigen-directions. In particular, when taking the cost of downlink training properly into account, transmit correlation diversity can yield significant capacity gains in all regimes of interest. Our analysis shows that the system multiplexing gain can be increased by a factor up to [M/r], where M is the number of antennas and r ≤ M is the common rank of the users transmit correlation matrices, with respect to standard schemes that are agnostic of the transmit correlation and treat the channels as if they were isotropically distributed. Thus, this new form of diversity reveals itself as a valuable “new resource” in multiuser communications.
This work further investigates how user grouping, more specifically, group partitioning, affects the sum-rate performance of a two-stage multiuser MIMO precoding scheme referred to as joint spatial division and multiplexing (JSDM). Taking both independent and clustered user geometry models into account, we propose new criteria for user grouping to noticeably improve the sum-rate performance of JSDM. Numerical results confirm that user grouping based on the proposed criteria yields substantial performance enhancement.