Generation of Gaussian noise with improved quasi-white properties.

Problems of hitherto used and suggested random signals for the identification of non-linear systems are reviewed. Two processes for the practical generation of random signals are proposed which avoid most of the shortcomings of the former test signals.