The test consists in the measurement of the noise intensity at the output o
f a bandpass filter, and the estimation of the accuracy of the measurement,
The confidence interval for the estimate is calculated under the assumptio
n that the noise is stationary and Gaussian. This assumption is called "zer
o-hypothesis". If a considerable part of the experimental data is outside t
he confidence interval, then the zero-hypothesis does not hold. The effect
of noise being non-Gaussian is investigated analytically. Experimental data
of 1/f noise in GaAs epitaxial films are presented and discussed. (C) 2000
Elsevier Science B.V. All rights reserved.