The Volterra representation of an electronic device using the Neural Network parameters

STEGMAYER G., CHIOTTI O.

Arequipa, Perú,

Congreso; XXX Latin-American Conference on Informatics (CLEI); 2004

CLEI

Many electronic devices present nonlinear characteristics, which are often difficult to express analytically. Generally it is easier to perform measurements of the device parameters than to develop an analytical model of its behavior. As Neural Networks can be used to learn a system dynamics from input-output data only, we have developed a Neural Network model which reproduces the nonlinear behavior of an electronic device, in particular a Field-Effect Transistor (FET), using simulation data. However, electronic devices nonlinear analysis requires an analytical model (i.e. an equation representing the current-voltage relationship), described as a closed-form function, that allows to draw conclusions about the device, such as the Volterra series model. In this work, we want to show how the neural model and the analytical Volterra series model of the transistor are totally equivalent. Therefore, we show here how it is possible to build an analytical expression for a device nonlinearity, the Volterra series, with parameters of a standard Neural Network, trained with the device measurements or simulation data.