This thesis proposes a method for determining the optimal antenna element radiation characteristics which maximize diversity gain given a specific power angular spectrum of the propagation environment. The method numerically constructs the eigenfunctions of the covariance operator for the scenario subject to constraints on the power radiated by each antenna as well as the level of supergain allowed in the solution. The optimal antenna characteristics are produced in terms of radiating current distributions along with their resulting radiation patterns. The results reveal that the optimal antennas can provide significantly more diversity gain than that provided by a simple practical design. Computational examples illustrate the effectiveness of adding additional elements to the optimal array and the relationship between aperture size or the description of the impinging field and the array performance. A synthesis procedure is proposed which uses genetic algorithm optimization to optimally place a reduced number of dipoles. The results from this procedure demonstrate that using the framework in conjunction with optimization strategies can lead to practical designs which perform well relative to the upper performance bound. Finally a novel array architecture is proposed where subsets of antennas are combined together into super-elements which are then combined in the same manner as the optimal array. The simplifications that result from the genetically optimized small array or the super-element array provide a design options which are feasible in many communication applications.

While the capacity of multiple-input multiple-output (MIMO) systems has been explored in considerable detail, virtually all literature on this topic ignores electromagnetic considerations. This dissertation explores electromagnetic effects on the capacity performance of these multi-antenna architectures. Specifically, it examines the impact of superdirectivity for compact antenna arrays, the effect of antenna mutual coupling, and MIMO performance of multi-mode optical fiber with non-linear detection. Superdirectivity can lead to abnormally large capacity bounds in a MIMO communication system, especially when the antennas are placed close together. Because superdirective behavior is difficult to achieve in practice, this work formulates an approach for limiting the impact of superdirectivity by introducing finite ohmic loss into the capacity expressions. Results show that even a small amount of ohmic loss significantly affects the achievable system capacity and suppresses superdirective solutions. This formulation allows a more detailed examination of the capacity of MIMO systems for compact arrays. For channels which do not vary in time, placing antennas closer together generally reduces the system capacity. However, recent work has demonstrated that for a MIMO system operating in a fast fading environment where the transmitter and receiver know the channel covariance information, the capacity increases as antennas are placed near each other due to an increase in spatial correlation. Analysis of this behavior illustrates that when these capacity gains (due to closely spaced antennas) are observed the radiated power is also increased. Constraining the radiated power leads to superdirective solutions in which the ohmic loss constraint developed must be used to properly determine the capacity behavior of this system. Application of this constraint then leads to an optimum antenna spacing in contrast to the findings of previous research which indicate that antennas should be as close together as possible. Additionally, this section provides an analysis regarding the number of spatial modes that can be used for various system configurations. Recent research has shown that it is possible for MIMO communication techniques to be used with multimode optical fibers to increase the available distance-bandwidth. However, implementation of traditional MIMO schemes requires the use of coherent optical detection which can lead to high system complexity and cost. This dissertation proposes a multimode fiber MIMO system architecture which allows simultaneous transmission of unique streams to different users on the same fiber while using incoherent detection with amplitude and phase modulation at the transmitter. The resulting capacity scales nearly linearly with the number of transmitters and receivers. Because the architecture requires channel state information at the transmitter, a training scheme appropriate for use with optical intensity detection is also discussed.