Input Amplitude Compression in Digital Signal-Detection Systems

An implementation for digital coherent signal-detection systems is considered in which analog-to-digital conversion is achieved by a uniform quantizer preceded by a signal amplitude compressor. This configuration is a practical approximation to the optimum coherent-detection quantization scheme which has nonuniformly spaced output levels. An optimum input compression curve is derived and is seen to be an output-scaled version of the locally optimum detection nonlinearity. Performance analysis and design of these compressors as considered for several noise models, with the conclusion that this implementation, when properly scaled, yields nearoptimum performance for only a moderate number of quantization bits.