Characterized by multiple antennas at both ends of a wireless link, multiple-input multiple-output (MIMO) communication systems potentially offer a tremendous increase in data rate compared to single-antenna systems, without the need for additional power or bandwidth. This is possible because several spatial subchannels are opened between the transmitter and the receiver, through which several data symbols can be spatially multiplexed. These spatial subchannels can be used to simultaneously serve either multiple users in a space division multiple access (SDMA) system or a single user in a time division multiple access system. Ideally, the antennas are placed so that the corresponding MIMO channel matrix has uncorrelated elements. Unfortunately, because of practical physical constraints on the positioning of antennas, correlations within the channel matrix are unavoidable. These correlations diminish the degree of independence between the subchannels, which in turn significantly reduce channel capacity as compared to the ideal, uncorrelated, case.; To help reduce the channel matrix correlations, in this thesis we recommend distributing the basestation antennas as small arrays of antennas at several different sites around the microcell, instead of placing them all at a single site. We show this to be a very effective strategy for recovering some of the throughput loss compared to the uncorrelated case.; To further reduce the degradation due to correlations, we propose novel correlation reduction algorithms (decorrelation algorithms) based on antenna and user selection. In addition to reducing system complexity (since only a limited number of antennas may be active at any given time), these algorithms aim at improving the MIMO channel ergodic capacity by reducing channel matrix correlations while enhancing the spatial separability of the mobile users in a multi-user environment.; Performance evaluation results, obtained through computer simulation, confirm the advantages of distributing the basestation antennas; and demonstrate the effectiveness of the decorrelation algorithms. We consider a frequency selective Ricean fading MIMO channel with lognormal shadowing and correlated channel matrix entries (according to the angles of departure/arrival, angular spectrum and angular spread at each array) and with different pathlosses to each array site. We assume perfect channel knowledge at both the transmitter and receiver, and orthogonal frequency division multiplexing (OFDM) is used for transmission.
展开▼
