Errors in the measurement of non-Gaussian noise spectra using rf spectrum analyzers

Abstract

We discuss the nature of errors which may occur when the spectra of random signals not obeying Gaussian statistics are measured with typical rf spectrum analyzers. These errors depend on both the noise statistics and the process used to detect the random signal after it has been passed through a narrow bandpass filter within the spectrum analyzer. In general, for random signals not obeying Gaussian statistics, the output of the bandpass filter must be measured with a square law detector if the resulting measurement is to be strictly proportional to the power spectrum of the input signal. We compare measurements of the power spectra of non-Gaussian noise using a commerical spectrum analyzer with its resident envelope detector, with measurements by the same analyzer fitted with a square law detector. Differences of about 5% were observed.