Signal-noise neural network for use in optimisation of transistor performance

Abstract

A different approach is utilised in the optimisation of a microwave transistor performance, which can be described as the signal-noise neural network representing the performance characterisation for the transistor. The signal-noise neural network gives the signal S and noise N parameters as functions of the operating conditions which are frequency f, bias voltage V/sub DS/, bias current I/sub DS/ and configuration type CT, and the performance characterisation provides all the compatible noise, the input VSWR, gain (F, V/sub i/, G/sub T/) triplets and their associated termination couples which are the source and reflection coefficients. Using variations of all these compatible measured functions F, V/sub i/, G/sub T/ against operation conditions, various types of optimisation processes are defined and emphasised in the design of an active microwave circuit, especially in MMIC design.