Power Oriented Design of Linear Phase FIR Filters

Abstract

In the design of low computational complexity and low power FIR filters, researchers have made every effort to reduce the number of adders when coefficients multipliers are considered as the multiple constant multiplication problem. In this paper, for the first time, we propose a power oriented optimization of linear phase FIR filters, where a power cost is used as the criteria in the discrete coefficient searches and synthesis. The power cost is computed based on a newly proposed power model, which takes both the static power and dynamic power into consideration. With the new power model, a new coefficient synthesis scheme is proposed such that the synthesized coefficient consumes the minimum power. Compared to the adder-cost oriented algorithm, the proposed power-oriented algorithm has two advantages: First, the algorithm can optimize filters with lower power consumption, and second, the optimal design in the sense of power consumption is frequency aware. Unlike the adder-cost oriented algorithms that generate the same final coefficient set and the same synthesis of the coefficient set regardless of the frequency for a given filter specification, the proposed algorithm search and synthesizes the coefficients with the awareness of the working frequency; different designs may be resulted for the same filter specification but different working frequency, and each designed filter has lower power consumption in its specified frequency. Transistor level simulations of benchmark filters verified the above claims.