BLOCK TRANSMISSION OVER DISPERSIVE CHANNELS - TRANSMIT FILTER OPTIMIZATION AND REALIZATION, AND MMSE-DFE RECEIVER PERFORMANCE

Optimal transmit filters for packet-based data transmission on dispers ive Gaussian-noise linear time-invariant channels are derived by maxim izing the mutual information, subject to a fixed input power budget. A quasi-stationary approximation to the optimal nonstationary input cov ariance process is derived and shown to exhibit negligible mutual info rmation loss from the optimal case, for situations of most practical i nterest, Moreover, this quasi-stationary approximation results in effi ciently computed lattice or pole-zero implementations of the transmit filter. By considering the popular finite-impulse-response minimum-mea n-square-error decision-feedback equalizer (FIR MMSE-DFE) as a receive r structure, we show that transmitter optimization results in an appre ciable improvement in the decision-point signal to-noise ratio. Finall y, we show that, as the output blocklength becomes infinite, the optim um finite-dimensional nonstationary input covariance process converges to a stationary process whose power spectrum obeys the well-known wat er-pouring distribution.