Adaptive filter theory (2nd ed.)
Adaptive filter theory (2nd ed.)
Information Theory and Reliable Communication
Information Theory and Reliable Communication
Hi-index | 0.00 |
Optimal transmit filters for packet-based data transmission on dispersive Gaussian-noise linear time-invariant (LTI) channels are derived by maximizing the channel throughput, subject to a fixed input energy budget. A quasi-stationary approximation to the optimal nonstationary input covariance process is derived and shown to exhibit negligible throughput loss from the optimal case, for situations of most practical interest. Moreover, this accurate approximation results in efficiently-computed lattice or pole-zero transmit filters. By considering the popular finite impulse response minimum-mean-square error decision feedback equalizer (FIR MMSE-DFE) as a receiver structure, we show that transmitter optimization results in an appreciable improvement in the decision point SNR, when the output block length (N) is comparable to the channel memory. Finally, we show that, as N becomes infinite, the optimum finite-dimensional non-stationary input covariance process converges to a stationary process whose power spectrum obeys the well known water-pour distribution.