Modular networks for direct processing of delta-modulated signals

Modular networks for arithmetic signal processing are described for operation with delta-modulated input and output signals. It is shown that designs based on the concept of the digital delta-modulation encoder provides a flexible approach to implementation of arithmetic units using full adders and flip-flops encompassing both serial and parallel forms. Schemes for the elementary operations of addition and multiplication by a constant coefficient are presented and maximum quantisation error is determined. The effect of input errors when the elementary unit is embedded in a larger network is also discussed including idling sequences introduced at the inputs as approximations to zero. Networks for signal mixing and digital filtering are developed by combining elementary units and the complexity and maximum quantisation error of each are determined. It is shown that in most cases greater economies in hardware and/or less quantisation error can be achieved in comparison with existing techniques.