Modeling finite-memory nonlinearity in unit DAC elements, binary storage channels, and BPSK data channels

Spectral shaping of quantization noise can allow a few (or one) oversampled D/A converters restricted to two output levels, unit DAC elements, to replace high-resolution converters. In high-resolution or high-speed settings however, even the performance of systems with only two output levels can be limited by conversion-circuit nonlinearities that introduce quantization-noise intermodulation products into the signal band through nonlinear intersymbol interference (ISI). Nonlinear ISI also affects the communication and magnetic storage of data. Here a simple but very general model structure for such binary-in, analog-out nonlinearities is proposed. Curiously, its structure is that of a convolutional coder whose output bits are separately LTI filtered before being finally summed to form the analog output. The largest-energy filter responses have natural interpretations if this convolutional code is chosen carefully. One such code is presented here.