Design considerations in power spectra measurements by diffraction of coherent light.

The estimation of the spectral density of a spatial random process by diffraction requires that the parameters of the diffraction system be compatible with the statistical properties of the process. Otherwise, the estimate of the spectrum can be unstable or unduly biased. The relation between the second-order statistics of the process and the parameters of a basic diffraction system that are required for reasonable spectral estimates is investigated. Such parameters include aperture dimensions, wavelength, focal length of the lens, and scanning slit size. With the typical diffraction system, it is difficult to estimate the power spectrum because of the presence of the zero-order aperture-diffracted light. A technique is furnished in this paper for spectral estimation at dc. Experimental results are furnished for film grain and total film noise. It is shown that the residual phase noise of film at dc even under so-called matched conditions is large enough to prevent the accurate estimation of the dc value of grain noise.