A statistical analysis of the estimation error of sound power measured by sound intensity techniques

Abstract

The difference between the exact power passing through the measurement surface and the sound power estimated from a finite number of sound intensity measurements is defined as the estimation error. Based on the expression of the error derived in a previous paper [M. Q. Wu and M. J. Crocker, Proc. INTER-NOISE 86, 1129 (1986)], the statistical properties of the error were studied and frequency histograms of it were drawn for the cases where the sound source was assumed to be a point monopole with random locations over the source surface and the measurement surface was a 1-square-meter square parallel to the source surface. It is found that if the measurement distance is 0.3 m and if four sound intensities are measured over the measurement surface, then for 89% confidence the estimation error will be less than 0.9 dB. If the measurement distance is increased to 0.5 m, then for 99% confidence the estimation error will be less than 0.5 dB. Some estimation of the necessary number of sound intensity measurements for a given error limit has also been conducted.