Impact of FFT algorithm selection on switching activity and coefficient memory size

Abstract

The binary tree decomposition allows for obtaining a large number of algorithms that can be used to calculate the fast Fourier transform. This paper analyzes the differences among these algorithms in terms of switching activity, which is related to the power consumption of the circuit, and size of the coefficient memories, which is related to the area of the circuit.Experimental results show the most efficient algorithms in term of area and power consumption. Furthermore, the paper shows the importance of a proper algorithm selection, since efficient algorithms can lead to savings of upto 45% in terms of the coefficient memory and even greater than 50% in terms of switching activity with respect to other less efficient ones.