Improved Parameterization Methodology for a Field-Stop Trench-Gate IGBT Physical Model by Switching Feature Partitioning

The accurate simulation of insulated gate bipolar transistor (IGBT) characteristics is an essential task to predict its electrical behaviors. Numerous physics-based IGBT models have been developed and proven to produce satisfactory simulation results. However, unavailability of device parameters is the key roadblock to the application of these physics-based models. In addition, the coupling between device parameters renders parameter extraction very challenging and heavy. In this article, a novel parameter extraction scheme by switching feature partitioning is proposed, which considerably relaxes the negative impact of parameter coupling during parameter extraction. The IGBT inductive turn-off process and static characteristics are divided into nine typical feature segments under an inductive load. Then, the physical principle of each feature segment is analyzed, and the corresponding influencing parameters of each feature segment are found. During the parameter extraction, the corresponding weights are assigned to the parameter search according to the error of each feature segment. The proposed method for parameter extraction is elaborated. Using the extracted parameters, the simulation results by the physics-based IGBT model are in excellent agreement with the experimental results of the tested IGBT, and the effectiveness of the proposed parameter extraction program is verified. Finally, the extracted parameters are evaluated and verified by global sensitivity analysis.