Digital Signal Processing Applications with Iterative Logarithmic Multipliers

Abstract

Many digital signal processing applications demand a huge number of multiplications, which are time, power and area consuming. But input data is often corrupted with noise, which means that a few least significant bits do not carry usable information and do not need to be processed. Therefore, approximate multiplication does not affect application efficiency when approximation error is less than noise introduced during data acquisition. This fact enables usage of faster and less power-consuming algorithms that is important in many cases where processing includes convolution, integral transformations, distance computations etc. This paper discusses logarithm-based approximate multipliers and squarers, their characteristics and digital signal processing applications based on approximate multiplications. Our iterative multipliers and squarers contain arbitrary series of basic blocks that involves only adders and shifters; therefore, it is not power and time consuming and enables achieving arbitrary accuracy. It was shown that proposed approximate multipliers and squarers can be used in several signal processing applications without decreasing of application efficiency.