Capacity optimization for Rician correlated MIMO wireless channels

We develop an optimization program to calculate the ergodic capacity and the optimal input signal covariance of a Rician correlated MIMO wireless channel. The program employs the Newton method and the barrier interior-point method to solve convex stochastic formulations, using efficient gradient and Hessian calculations. We then use the program to study impacts of the channel mean, the transmit correlation, and the K factor on channel capacity and the optimal input signal, revealing curious optimal signal characteristics at high SNR, where mode-dropping can always occur for a channel with strong mean or strong correlation. We also compare the capacity with the mutual information obtained by the signaling solution optimal for the Jensen bound. Results reveal a close approximation for systems with equal or fewer transmit than receive antennas at all SNR, and for those with more transmit than receive antennas at low SNR. The approximation diverges for the latter system at high SNR, depending on the transmit correlation and the channel mean, or the Rician K factor