Letter FFT Based Generation of Bandlimited Gaussian Noise Variates

Rapid generation of time series samples of stationary, zero-mean, correlated Gaussian noise will accelerate digital communication system simulations. In this letter, we show that a fast Fourier transform (FFT) based method is well suited to rapidly generate such noise samples. The FFT method requires O(N) memory elements and O(Nlog₂ N) floating-point operations to generate each sequence of N variates. Sequences that are bandlimited incur an aliasing error in the correlation function of the sequence, but for practical simulations we show this error is negligible. We quantify this error for arbitrary sampling rates and correlation functions.