Low power implementation of digital filters using DBNS representation and sub-expression sharing

This paper presents a solution for low power consumption in digital systems designs. The Double Base Number System (DBNS) representation recognized by its high sparseness is used for digital filters coefficients representation to reduce the switching activity, a key factor in power consumption. In this paper, we show that the use of the Canonic DBNS representation for filter coefficients representation leads to a dynamic power reduction average of 30% compared to the 2psilas complement representation. The second contribution of our work consists in the application of sub-expression sharing on the filter coefficients, already expressed in DBNS representation. Strengths and limitations of this optimization will be discussed in details throughout this paper.