Publication

Abstract : Approximate computing is a promising approach in digital circuit and digital signal processing applications which are implicit error tolerant circuits. In this project, four approximate multipliers are proposed with three approximate 4–2 compressors designs. To extend the applicability of the proposed approximate multiplier design highly configurable signed/unsigned multiplication is introduced using a high-level configuration parameter. Inner stage pipelining technique is used to maximize the operating speed with least path delay overhead which bring further improvements in performance efficiency, and optimal clock gating technique is used to reduce the power consumption rate, which bring improvement in energy efficiency. The proposed method was implemented in Xilinx ISE 14.7 version tool and synthesize in cadence genus tool with 45 nm technology. The four proposed multipliers are used in the application of image multiplication by discretizing image in MATLAB and Modelsim. As a result, the area, power, and delay were reduced from the exact multiplier, and the energy efficiency and performance efficiency is more when compared with the existing method.