White Noise Quality and the Results of Simulated Identification

Abstract

The results of simulated identification of stochastic time-series and stochastic systems depend to some extent upon the quality of random number generators used. A new type of generators of sc alar and multichannel orthogonal white series is presented, which increases the accuracy of simulated identification as compared with conventional random number generators. It is based on the frequency domain definition of a vector orthogonal multisine series, which is transformed into the time domain using the Fast Fourier Transform. The properties of such series are discussed. Special cases of vector white multisine series in the form of scalar and bivariate white multisine series are presented. The enhancement of the accuracy of simulated identification using those series are illustrated by a number of examples.