Finite Random Matrix Theory Analysis of Multiple Antenna Communication Systems

Multiple-antenna systems are capable of providing substantial improvement to wireless communication networks, in terms of data rate and reliability. Without utilizing extra spectrum or power resources, multiple-antenna technology has already been supported in several wireless communication standards, such as LTE, WiFi and WiMax. The surging popularity and enormous prospect of multiple-antenna technology require a better understanding to its fundamental performance over practical environments. Motivated by this, this thesis provides analytical characterizations of several seminal performance measures in advanced multiple-antenna systems. The analytical derivations are mainly based on finite dimension random matrix theory and a collection of novel random matrix theory results are derived. The closed-form probability density function of the output of multiple-input multiple-output (MIMO) block-fading channels is studied. In contrast to the existing results, the proposed expressions are very general, applying for arbitrary number of antennas, arbitrary signal-to-noise ratio and multiple classical fading models. Results are presented assuming two input structures in the system: the independent identical distributed (i.i.d.) Gaussian input and a product form input. When the channel is fed by the i.i.d. Gaussian input, analysis is focused on the channel matrices whose Gramian is unitarily invariant. When the channel is fed by a product form input, analysis is conducted with respect to two capacity-achieving input structures that are dependent upon the relationship between the coherence length and the number of antennas. The mutual information of the systems can be computed numerically from the pdf expression of the output. The computation is relatively easy to handle, avoiding the need of the straight Monte-Carlo computation which is not feasible in large-dimensional networks. The analytical characterization of the output pdf of a single-user MIMO block-fading channels with imperfect channel state information at the receiver is provided. The analysis is carried out under the assumption of a product structure for the input. The model can be thought of as a perturbation of the case where the statistics of the channel are perfectly known. Specifically, the average singular values of the channel are given, while the channel singular vectors are assumed to be isotropically distributed on the unitary groups of dimensions given by the number of transmit and receive antennas. The channel estimate is affected by a Gaussian distributed error, which is modeled as a matrix with i.i.d. Gaussian entries of known covariance. The ergodic capacity of an amplify-and-forward (AF) MIMO relay network over asymmetric channels is investigated. In particular, the source-relay and relay-destination channels undergo Rayleigh and Rician fading, respectively. Considering arbitrary-rank means for the relay-destination channel, the marginal distribution of an unordered eigenvalue of the cascaded AF channel is presented, thus the analytical expression of the ergodic capacity of the system is obtained. The results indicate the impact of the signal-to-noise ratio and of the Line-of-Sight component on such asymmetric relay network.