Antenna selection for MIMO-OFDM systems with channel estimation error

We study space-time-frequency coded multiple-input-multiple-output (MIMO) orthogonal frequency-division multiplexed (OFDM) systems employing receiver antenna selection (AS), where training has been utilized to perform linear minimum mean square error (LMMSE)-based channel estimation. To minimize the mean square error (MSE) of this estimator, equispaced and equipowered training symbols are used, with the number of training symbols per OFDM symbol equal to the channel length. The maximum received signal power AS rule is proposed, which decouples the AS process from the channel estimation process. By upper bounding the pairwise error probability (PEP) expression, we show that the system with a channel-estimation error (CEE) still achieves maximum spatial and multipath diversity. The performance of the system for the special case of square unitary and orthogonal codes has been analyzed. For the case of orthogonal codes using constant modulus symbols, we also derive the exact bit-error-rate (BER) expression. The degradation in performance, due to the presence of the CEE, is captured by the loss-in-coding-gain (LCG) and loss-in-performance (LP) expressions. To minimize the loss due to the CEE, an optimal power-allocation scheme that distributes the total available power between training symbols and data symbols is defined. Compared with the perfect channel knowledge case, equal power training for the CEE case performs at least 3 dB worse, whereas optimal power training for the CEE case suffers a loss less than 3 dB always. Analytical and simulation results are presented to validate our analysis and results.