Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

Accurate characterisation and modelling of SiC MOSFETs for transient simulation

Yang, Peng 2022. Accurate characterisation and modelling of SiC MOSFETs for transient simulation. PhD Thesis, Cardiff University.
Item availability restricted.

[thumbnail of PhD thesis]
PDF (PhD thesis) - Accepted Post-Print Version
Download (14MB) | Preview
[thumbnail of Electronic Theses and  Dissertations Publication Form] PDF (Electronic Theses and Dissertations Publication Form)
Restricted to Repository staff only

Download (117kB)


Silicon carbide (SiC) metal-oxide-semiconductor field-effect transistors (MOSFETs) possess properties that are superior compared to their silicon counterparts, such as low conduction and switching losses, high thermal conductivity and operating temperatures, etc. SiC MOSFETs need to be evaluated either experimentally or through simulation to fully exploit or understand their benefits. Compared to experimentation, simulation is more time and cost-efficient so is preferred at the initial converter design stage. However, accurate and fast models of SiC MOS FETs need to be established for such simulation, which is challenging due to fast switching speed and frequency of SiC MOSFETs. This thesis focuses on improving the accuracy and speed of models of SiC MOSFETs and simplifying the modelling process at the same time. Firstly, the characteristics of SiC MOSFETs required for the modelling were analysed. It was found that the I-V and C-V characteristics in their dynamic state have a significant impact on the accuracy of the models. However, the existing methods to measure and extract these characteristics are complex due to the multiple measurement equipment configurations are required. This thesis proposes a simplified dynamic-state characterisation method. This method analyses the relationship between characteristics and the switching waveforms of SiC MOSFETs, and extracts these characteristics directly from the switching waveforms measured by a double pulse tester to simplify the measurement process. These measured dynamic-state characteristics, combined with the conventional static-state characteristics, were used to built the SiC MOSFET model.The relative root-mean-square (RMS) errors of the model can be reduced by at least a factor of 3, compared to the model that only considers the conventional static-state characteristics. Based on the extracted characteristics, a measurement-based hybrid data-driven modelling method is proposed. Conventional equation-based models have drawbacks such as a complex modelling process, poor adaptability, low accuracy and slow simulation speed. The proposed modelling method utilised a hybrid data-driven model based on artificial neural networks to simplify the modelling process and improve the adaptability. The switching waveforms simulated by the proposed model are 1.5 ∼ 3 times closer to the experimental waveforms, compared to the commercial equation-based Angelov model. At the same time, the proposed model is 30% faster than the Angelov model in terms of simulation speed. However, equipment required for the measurement-based modelling may not be available for some converter designers, therefore, a step-by-step datasheet-based modelling method is proposed, which is completely based on the datasheet without the use of any further data or equipment. Compared to the measurement-based modelling method, the datasheet-based modelling method results in 24% increase in RMS errors and cannot accurately match the gate driver resistor used in practical experiment. However, the datasheet-based modelling method features a simpler modelling process and 15% faster simulation speed so provides a more cost and time-efficient process for converter designers to quickly validate their converter design.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Engineering
Uncontrolled Keywords: SiC MOSFET, Electrical characterisation , Electrical modelling , Transient simulation , Double pulse tester , Artificial neural network
Date of First Compliant Deposit: 14 April 2022
Last Modified: 14 Apr 2022 09:25

Actions (repository staff only)

Edit Item Edit Item


Downloads per month over past year

View more statistics