| Over the last decade, it has been well established that multiple antenna systems provide a spatial degree of freedom (DoF), exploiting which, enhanced data rates or reliable performance or a combination of the two aspects can be achieved. These initial works are based on an i.i.d. channel model which is realistic only in wireless applications where large antenna spacings and a rich scattering environment are possible.;However, measurement campaigns show that, realistic channels are spatially correlated. It is unclear as to what the potential gains with multiple antenna technology are in these situations. Motivated by this problem, in this dissertation, we study the impact of spatial correlation, which is modeled with a recently proposed eigenbeam framework, on capacity and limited feedback performance.;In the first part of this dissertation, we show that the exact rate of capacity scaling with antenna dimensions is determined by the channel power, a measure of the channel's multipath sparsity level. Unlike rich scattering, sparsity provides a new dimension over which channel capacity can be optimized, namely, distribution of the channel's DoF in the available signal-space dimensions. Spatial distribution of the DoF can be realized by reconfigurable arrays where antenna spacings can be adaptively reconfigured in response to (a very limited) knowledge of the sparsity level.;We study the impact of reconfigurable arrays on capacity of sparse channels and provide insights on the spatial distribution of the DoF that lead to optimal capacity scaling laws, both as a function of antenna dimensions and SNR. Surprisingly, for all practical purposes, just three canonical antenna configurations are necessary to maximize capacity over the entire SNR range.;In the second part of the dissertation, we consider the case where limited feedback information is available at the transmitter. We first focus on transmit beamforming where the multiple antennas are used to enhance reliability and show how to optimally utilize the limited feedback resource. In contrast to the i.i.d. model, spatial correlation aids in constructing limited feedback codebooks that perform as good as perfect channel information beamforming. Codebook design guidelines for multi-mode signaling that trades-off reliability with multiplexing gain are provided. |
