Abstract
The search for optimality in the design of channel precoders and training symbols in block processing communication systems is one of paramount importance. Finding the best tradeoff in terms of power distribution between information and pilot symbols for frequency selective channels, when channel estimation via feedback is available, however, has not been fully addressed. In this paper, we solve the problem of finding the optimal power distribution between pilots and data symbols in the mean-square-error (MSE) sense when a delayless error-free channel feedback path is available to the transmitter. The novel approach adaptively designs the optimal precoders and training vectors based on the frequency domain estimates of the channel.