The total number of model parameters in compact MOSFET models is usually between 50 and 300, necessitating an elaborate extraction process. Gradient-based optimization techniques were found to be useful but somewhat limited in their scope. The main difficulty is that the error profile has numerous local minima which can trap the convergence process. Similarly, the error surface is not always conducive to successful application of classical methods. In this work, the feasibility of using Genetic Algorithm (GA) to alleviate these problems is investigated. To achieve unambiguous results we concentrate primarily on the effective mobility model development and parameter extraction. Consequently this investigation is restricted to wide long-channel MOSFETS described by combining a new form of a charge-sheet model with an elaborate effective mobility model.
Journal: TechConnect Briefs
Volume: 2, Technical Proceedings of the 2003 Nanotechnology Conference and Trade Show, Volume 2
Published: February 23, 2003
Pages: 314 - 317
Industry sector: Sensors, MEMS, Electronics
Topic: Compact Modeling