Pseudorandom signals for the dynamic analysis of multivariable systems

The extension of the input/output-crosscorrelation method of system dynamic identification, using a pseudorandom input signal, to multi-input/multioutput systems is considered. The system impulse responses may be obtained directly from the measured crosscorrelation functions (in a similar manner to the single-input/single-output case), provided that the system-input signals have autocorrelation functions of delta-function form and are uncorrelated with each other, it is found that it is not possible to meet both these requirements without resorting to a very prolonged integration time. Methods of overcoming this difficulty by using signals of practicable lengths, the properties of which approximate to those required, are therefore proposed. Both 2- and 3-level signals are considered.