A PREDICTOR-CORRECTOR APPROACH FOR FAST UNCERTAINTY QUANTIFICATION OF SOI MOSFET AT HIGH FREQUENCY

Abstract

Due to the declining return on performance vs cost of conventional MOSFETs, silicon on insulator (SOI) MOSFETs have the potential to aid economic progress. With the advancement of technology, it is becoming increasingly vital for researchers to develop high-fidelity models so that the advantages of this technology may be transferred into practical circuit implementations. Instead of having desirable characteristics, SOI FETs is very susceptible to fabrication process fluctuations and manufacturing tolerances due to transistor scaling. As a result, the circuit's performance becomes unpredictable. Thus, it is mandatory to append the uncertainty at the initial stage of design process. Monte Carlo is one of the uncertainty quantification approaches used for this purpose, although Monte Carlo is inefficient for the uncertainty quantification of SOI MOSFET due to its sluggish rate convergence. In this thesis polynomial chaos-based approach is presented for a fast uncertainty quantification of SOI MOSFET at millimetre wave frequency. In this technique, the predictor PC metamodel is utilised to capture the device's broad statistical features, while the highly accurate corrector PC metamodel captures the finer statistical features ignored by the predictor PC metamodel. This method provides a great numerical efficiency as compared to conventional MC technique. Finally, a probabilistic model of device Y-parameters is developed in this thesis, which provides statistical information such as mean, standard deviation, and probability distribution function for each of the device's Y-parameters. To validate this probabilistic model two example partially depleted SOI MOSFET and full depleted SOI MOSFET is represent in this thesis.