An Optimal Nonlinear Detector for Digital Data Transmission Through Non-Gaussian Channels

The problem of binary synchronous communication in additive non-Gaussian noise is considered. It is shown that, for a class of additive noises, an optimal detector consists of a two-input nonlinear device followed by an integrator and decision box. This system suppresses large excursions from the signal level in an optimal manner before integration of the received signal, so that the effects of large amplitude noise waveforms on the signal decision are minimized. Asymptotic performance characteristics are obtained indicating the relationships between the basic system parameters, signaling waveforms, signal energy, system bandwidth, decision time, noise dispersion, and error probability. Explicit expressions and curves for system performance in large variance (Cauchy distributed) noise are given for various basic signaling waveforms.