Digital generation of stochastic signals of arbitrary spectral shape

For computer-simulation experiments in the development of noise monitoring systems three methods of generating ergodic Gaussian random noise with specified spectral properties have been investigated: digital filtering of white noise optimum symmetric FIR filters, a modified Rice formulation and an approximation of the Kac representation of noise. The proposed modified Rice formulation is a new noise generation method which is most efficient with regard to the computation time. By windowing subsequent Rice sequences a smooth noise record of any desired length can be produced. (author)