Application of Dither to Low Resolution Quantization Systems

Abstract

A significant problem in the processing chain of a low resolution quantization system is the Analog to Digital converter quantization error. The classical model of quantization treats the error generated as a random additive process that is independent of the input and uniformly distributed. This model is valid for complex or random input signals that are large relative to a least significant bit. But the model fails catastrophically for small, simple signals applied to high resolution quantization systems, and in addition, the model fails for simple signals applied to low resolution quantization systems, i.e. one to 6 bits resolution. This paper will discuss a means of correcting this problem by the application of dither. Two methods of dither will be discussed as well as a real-life implementation of the techniques.