General optimization and extraction of IC device model parameters

Abstract

VLSI circuit simulation necessitates specification of model parameters. For the device equations to predict the measured <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I-V</tex> characteristics, the extraction of these parameters is required, and conventionally, measurements are performed sequentially to determine one or a few parameter values. These procedures are specialized to a particular model, and neglect the parameter interaction, as well as they are performed on a device-by-device basis. In this paper, a complete set of parameters for any arbitrary model is estimated by a modified Levenberg-Marquardt method operating in linear constrained space, and the result is a nonlinear least-squares fit of the data to a set of width/length-measured device characteristics. This model-independent extraction approach is enhanced by a sensitivity analysis to check for redundant parameters and is implemented in the SUXES computer program. It has proved to be an effective tool for device characterization and for the development and evaluation of new models.