The theory of dithered quantization

Abstract

A detailed mathematical model is presented for the analysis of multibit quantizing systems. Dithering is examined as a means for eliminating signal-dependent quantization errors, and subtractive and non-subtractive dithered systems are thoroughly explored within the established theoretical framework. Of primary interest are the statistical interdependences of signals in dithered systems and the spectral properties of the total error produced by such systems.

Regarding dithered systems, many topics of practical interest are explored. These include the use of spectrally shaped dithers, dither in noise-shaping systems, the efficient generation of multi-channel dithers, and the uses of discrete-valued dither signals.