We describe a novel method for simulation of correlated homogeneous random
processes with given second-order statistics of a general kind. These proce
sses are specified by their correlation and cross-correlation functions. Th
e simulated processes are constructed as sums of plane waves with random wa
ve-vectors. For non-periodical processes with wide spectra, such a method r
equires much less memory, than the independent generation of the Fourier-co
mponents.