Application of the Fast Fourier Transform to Third-Octave and Octave-Band Analysis of Noise Produced by Business Machines

Abstract

A computer subroutine prepared by J. W. Cooley has been used together with additional programs to analyze the acoustical noise produced by typical business machines. The subroutine utilizes the Cooley-Tukey algorithm for the rapid computation of the Fourier coefficients of a discrete time series. The digital recording system described in an earlier paper [J. Acoust. Soc. Am. 40, 1273(A) (1966)] has been used to acquire data in a format suitable for computer processing. Emphasis is placed on computation of third-octave band levels from the Fourier coefficients, the filter characteristic or “window” associated with the analysis, and the spectral fluctuations that occur in short (100–250 msec) samples of the noise. The digital spectral estimates are in good agreement with analog measurements. For one machine studied, the spectral fluctuations are larger than would be expected for samples of random noise. Analysis of successive bursts from a second machine has indicated that spectral fluctuations occur even when the envelope of the pressure-time pattern is observed to be repetitive.