Power Allocation for Capacity Maximization in Eigen-mimo with Output Snr Constraint

Water-filling is the power allocation that maximizes the parallel-channel eigen-MIMO informationtheoretic capacity. However, for practical designs, designing for error performance is also important. The overall uncoded error rate, in a system with water-filled eigen-channels using digital modulation, degrades through the errors in the data that uses the weaker eigen-channels. This is because in water-filling, a lower transmit power is allocated (and a lower proportion of the overall capacity) to the weaker eigenchannels. One general solution is to discard the weaker eigen-channels, but this does still not allow simple control over the values of, or trade-off between, capacity and the error rate. In this paper a design approach is presented for the control of, and improvement in, the overall error performance, while maintaining the ergodic capacity as high as possible. The approach is direct; the eigen-channel power allocation is optimized for capacity maximization constrained by a desired error performance on the eigenchannels via constraints on their SNRs. A useful feature is that the capacity and the error performance can be managed through the eigen-channel SNR constraint. Statistical simulations with Rayleigh channels quantify the significant SER improvement compared to full eigen-MIMO (maximum number of eigen-channels) with water-filling, when using a fixed modulation set. This set comprises pre-chosen constellations for a constant number of eigen-channels, and this remains fixed for the all the channel realizations, i.e., with time for the time-varying channel. The improvement in error performance is at the expense of a small loss in capacity.