|M.Sc Thesis||Department of Electrical Engineering|
|Supervisor:||Assoc. Prof. Steinberg Yossef|
Multi-antenna channels have recently gained attention because of the possibility of increasing capacity in fading environments. This work is an investigation of the fading multi-antenna channel in the presence of imperfect channel state information (CSI) at the receiver. A modified nearest neighbor decoder is presented and is assumed to operate in concert with a random Gaussian coding scheme. An expression for the Generalized Mutual Information (GMI), the achievable rate, is obtained. It is shown that under certain conditions, the achievable rate is equivalent to that of a fading multi-antenna Gaussian channel. The fading is known to the receiver and is equal to the channel estimation, and the noise is due to both the channel noise and the channel estimation error. The nearest neighbor decoder is augmented with a scaling factor which proves to be important in attenuating "noisy" time and space dimensions in the metric calculation. The MMSE channel estimator is shown to be optimal in the sense that it achieves the highest GMI rate. The outage probability versus GMI rate is investigated, and it is shown that when the channel estimation is not perfect there is a positive probability of outage irrespective of the total transmit power. The GMI is used as a criterion for optimizing the training and information sequences in a block fading channel. New results for the training sequences and optimal antenna usage are presented. It is shown that the number of antennas which should be used for training and information sequences depends on the SNR and the energy allotted to training. A suggested optimal training scheme sends a pilot through each of the useful antennas, one antenna at a time, and then transmits information only via the trained antennas.