MOSFET Simulation Using Neural Networks

Electrical simulation is mainly based on accurate mathematical models of the electrical circuit’s elements. Each MOS transistor contains a set of basic parameters forming its mathematical models. Among the main parameters there are length and width of its conductive channel, which affect the basic characteristics of the transistor. There is an inverse task in the electrical circuit’s modeling, that is namely to find the width and the length of the conductive channel to obtain a mathematical model of an unknown transistor. These measurements are selected by enumeration taking place onto the convergence between its measured characteristics and these ones which are calculated by any modeling program. One of the solutions to obtain the width and length of the transistor’s conducting channel is to train a computer model based on a neural network. The purpose of this study is to valid the approximation of a function setting the width and length of the transistor’s conducting channel in accordance with its measurements by means of training of neural network. As a result of the training and the final statistical analysis, it was proved that the algorithmic task approach, based on the neural network training, provides the necessary funсtion approximation.