Coded BER minimization for MIMO multicarrier systems with imperfect channel estimates

Linear transmitter designs that minimize the MSE or the uncoded BER do not guarantee good bit error performance when any channel correction code is introduced. As an alternative, the cutoff rate, that has been considered for a long time as the practical bound on the performance that could be achieved by finite length channel codes, can provide a good performance not for an specific channel code but for the average over the ensemble of codes with block length N. This work presents a robust power allocation strategy that maximizes the cut-off rate using partial channel state information (CSI) at the transmitter. A Bayesian formulation is used to design an optimal transmitter when CSI is noisy. A closed-form solution for the transmitter design is obtained. In addition, the design adapts to the CSI quality, providing an algorithm that tends to the open-loop design when no CSI is available at the transmitter (i.e., the same power is allocated across all antennas and subcarriers).