| Multi-input multi-output (MIMO) systems, that employ antenna arrays for transmission and reception, promise dramatic improvements in system capacity and reliability for emerging high-speed wireless communication systems. However, most existing works assume an idealized channel model that is inadequate to describe physical channel characteristics encountered in practice. Thus, they fall short of addressing the critical issue of how statistical channel correlation affects system capacity and performance. In this work, we study communication over physical MIMO channels from two key perspectives: (i) assessment of fundamental system capacity (maximum achievable data rate), and (ii) space-time coding techniques aimed at achieving the capacity and diversity afforded by the channel. Our work builds on the virtual channel representation, a recent advance in MIMO channel modeling, that combines the accuracy of physical models with the simplicity of statistical models. A methodology is developed to characterize the inherent channel correlation structure and to analyze its impact on capacity, and code design and performance. Specifically, we propose a D-connected model that provides a simple and tractable abstraction of physical channel characteristics: the connectivity parameter D captures the level of correlation and diversity in the channel. Our results show that a quadratic growth in the number of physical propagation paths is necessary to sustain the linear capacity growth with the number of antennas. We then develop criteria for designing space-time codes for general correlated channels, and show that the analysis is greatly simplified by the virtual representation. Our results subsume designs based on the idealized model, reveal the impact of intrinsic correlation patterns of realistic MIMO channels on space-time code design, and are illustrated by novel code constructions that exploit the channel correlation structure to improve system performance. |
