Citation

Abstract

Most Fourier transform algorithms, when seeking a single frequency at the sampletime midpoint, simply choose the largest component of the set. The error is at least +0.5 component, and often larger. This article describes an addition to such algorithms that reduces the variance error by three-to-one (nominal). The addition achieves this by averaging all components within a predetermined “‘window,’’ selected as a function of frequency rate, and it is quite easy to mechanize within the existing algorithm.

Details

Volume
42-53
Published
October 15, 1979
Pages
155–160
File Size
371.8 KB