Impact of jittered sampling on conventional spectral estimates

Abstract

We examine the effect of jittered sampling (perturbations about a regular sampling scheme) on spectral analysis from a theoretical point of view, developing a general model for the effect in one dimension and then specializing to examples of Gaussian jitter. We find that the spectral integral or process variance is unaffected by jitter, but weight is cut off the spectrum according to a roll‐off function which is the squared modulus of the jitter characteristic function (Fourier transform of the distribution function). This weight is redistributed evenly across the spectrum for uncorrelated jitter and nearly so for broadband correlated jitter. However, the redistributed weight can be concentrated by strong correlation or periodicity in the jitter. The paper concentrates on the effects of jitter correlation, but we note that the distribution function, which decides the roll‐off, is also of key importance.