Abstract

Multiple correlations (cumulants) and their respective frequency transformations (polyspectra) are useful in analyzing skewed and/or nonlinear random processes. In particular, the triple correlation contains more signal information than does the autocorrelation (retention of phase relations) and is finding widespread use¹ in various 1-D signal processing applications such as ultrashort laser pulse reconstruction, detection of phase coupling, and nonlinear spectroscopy. One of the main drawbacks, however, is the computational complexity that is associated with the cumulant generation. In this paper we present an optoelectronic parallel processor for the real time computation of a 1-D signal's bispectrum. Due to the parallel nature of the architecture, the new bispectral processor eliminates the serial scan necessary to retrieve the third-order cumulants. In addition, since the cumulant generating architecture has no moving parts, the usual instrument aberrations inherent in the free-space triple-intensity interferometer are reduced and the SNR efficiency is increased. We demonstrate how this optical bispectral processor can be integrated with digital electronics for the real time digitization of a 1-D signal's bispectrum.

© 1989 Optical Society of America