Design and optimization of nonlinear mapping in decision feedback equalization

We derive an optimal nonlinear quantizer to be used as a decision function in a decision feedback structure. The criterion used to determine the quantizer characteristic is the degree to which the propagation of decision errors is suppressed. We show that a soft saturation nonlinearity can outperform the standard nearest neighbour quantizer in a decision feedback equalizer structure in terms of error propagation suppression. This implies that on channels where noise is impulsive a soft saturation quantizer decision feedback structure is superior to a conventional decision feedback structure an a broad class of channels