Statistical estimation of small-signal FET model parameters and their covariance

A statistical approach to the problem of parameter extraction of small-signal FET models is presented. This approach makes it possible to accurately assess parameter estimates and their variance and covariance, due to measurement uncertainties, without utilizing time consuming Monte-Carlo simulations. The method presented uses a maximum likelihood estimation with the widely used cold-FET technique to determine the parasitic elements and their covariance from two different gate bias conditions. These are thereafter used to perform a corresponding maximum likelihood estimation of the intrinsic elements from an active bias condition. Thereby, maximum information available from the measurements are brought into determining the model parameters as accurate as possible. The accuracy of the intrinsic and parasitic covariance are validated using Monte-Carlo simulations.