Parametric identification of two-port models in the frequency domain

The problem of parametric estimation of a two-port model from noisy input-output data is treated. The input and output voltages and currents are measured simultaneously in the frequency band of interest. To compensate for the hardware limitations, a calibration procedure is proposed. A Gaussian maximum likelihood estimator for multi-input, multi-output systems is described. It takes the input noise as well as the output noise into account. The covariance matrix of the estimated parameters is readily obtained. To verify the ability of the estimator to handle real measurement problems, an experiment was performed. It was possible to predict the behavior of a two port within an error of +or-0.01 dB on the magnitude and +or-0.1 degrees on the phase.<>