Modeling and decision feedback equalization of uncertain channels

In this paper a methodology for uncertain linear communication channel modeling is evaluated. The resulting extended model is then used for the design of decision feedback equalizers (DFE). Considering an initial nonparametric description of the channel by experimental data, a model for nominal channel and uncertainty is developed by means of orthonormal function series expansion. The procedure shows good model to data fitting. Using this model a partially adaptive decision feedback equalizer is then designed. The fixed front end or precursor filter is designed as an optimal robust linear deconvolution filter. The postcursor filter is designed with a fixed denominator to optimally cancel residual past symbol interference and the numerator is treated as an adaptive FIR filter with a small number of taps. Simulation results show extended equalizer flexibility at low additional computational cost for the proposed DFE design