An efficient linear statistical FET model

The commonly used FET model using model parameter statistics is examined and found to be at best a difficult structure for modeling an FET's performance statistics. A simpler linear statistical model based on principal-component analysis is proposed, which results in uncorrelated model parameters. An example using actual measured GaAs FET data uses just 13 uncorrelated random variables to model the FET's performance statistics from 1 to 11 GHz.<<ETX>>