Performance of stochastic quantizers employing nonlinear processing

Dithered quantizers are often employed in numerical instrumentation to improve the resolution and accuracy of the analog-to-digital converters used. To this aim, linear post-processing is usually performed on the quantized samples. In the paper, a different processing scheme is described that exhibits a nonlinear behavior. The characteristics of both linear and nonlinear topologies are analyzed, and expressions are given for the quantization error average and variance that make possible the comparison of the two algorithms. It is shown that the nonlinear one provides better performance, so that it can be employed to further improve the estimation capability of a dithered quantizing system. The reported results have been validated by means of meaningful simulations