The White Noise Approximation
Stephen B. Weinstein · IRE Transactions on Communications Systems · 1969
It is proved that the random-variable coefficients of the orthonormal vector components of a Gaussian noise process converge to uncorrelated and identically distributed random variables as the power density spectrum of the process approaches a constant almost everywhere. An example is then given to suggest the rate of convergence and the ratio of noise bandwidth to signal bandwidth required for good approximation of the noise process by the white noise process.