A double integrated neural network for identification of geometrical features dependency in lumped models

A novel identification technique for lumped models of general electronic circuits (i.e. MOSFET, BJT, monolithic integrated circuits and filters) is presented. The approach is based on a neural network having a supplementary layer and an adapted learning process, whose convergence allows the validation of the device model. The supplementary layer is another neural network trained offline on the model under exam. The inputs of the network are geometrical parameters and the neural network output represents the lumped circuit parameter estimation.