1200 V 4H-SiC金氧半場效電晶體SPICE模型設計與優化

Abstract

近年來，隨著再生能源以及電動車產業的興起，世界對於能量的轉換效率有所重視，而碳化矽材料與功率金氧半場效電晶體常被運用在許多電力電子系統上。本論文則將探討1200 V 4H-SiC金氧半場效電晶體特性與建立高精準度之SPICE模型。 此論文將利用LTspice模擬軟體進行SiC DMOSFET SPICE模型建模，透過本研究的建模流程與方法，分別把SPICE模型內的EKV模型之順向與逆向模型、寄生二極體、寄生電容數學模型等元件之重要參數萃取出來，最後套用至Wolfspeed的C2M0040120D之Datasheet數據建立出新穎1200 V SiC DMOSFET之SPICE模型，並且與Wolfspeed所提供的原生SPICE模型做元件的靜態與動態分析模擬以及模型精準度比較。 在靜態特性分析比較上，元件之新穎SPICE模型在不同操作條件下都可以低於10 %以下，相對地，原生SPICE模型的誤差則都會大於20 %；在動態特性分析比較中，新穎SPCIE模型之寄生電容特性誤差約3.12 %～27.01 %，而原生SPICE模型為23.15 %～110.48 %。而在雙脈衝測試方面，新穎SPICE模型在E_on、E_off、E_total都達到9 %以下的誤差；原生SPICE模型則都大於9 %的結果。代表本論文建立之新穎SPICE模型在靜態與動態模擬上，都有較優的模擬結果，以利於元件提升後續電子電路系統之精準度之的開發。

In recent years, the world has paid more attention to power conversion because the industry of renewable energy and electrical vehicle were blooming. The SiC MOSFET is one of the switching devices applied in power systems. This work will discuss the characteristics of 1200 V 4H-SiC MOSFET and build a high-precision SPICE model. LTspice software is used to build the circuits and simulate the performances of SiC DMOSFET SPICE model. The SPICE model consists of the EKV model, body diode, and parasitic capacitance mathematical models. The parameters are extracted by the modeling process and the method proposed in this work. We build novel SPICE model of 1200 V SiC DMOSFET by using C2M0040120D datasheet, Cree. Inc. Then, the novel SPICE model is compared with the original one to understand the fitting accuracy. In the static characteristics analysis, the error of novel SPICE model can be less than 10 % under different operating conditions, while that of the original one is greater than 20 %. In the dynamic characteristics analysis, the fitting error of parasitic capacitance is 3.12 % ~ 27.01 %, while the original SPICE model is 23.15 % ~ 110.48 %. Moreover, the error of novel SPICE model is less than 9 % in E_on, E_off and E_total in the double pulse test, while the original SPICE model is greater than 9 %. It means that the novel SPICE model established in this work has better fitting results in both static and dynamic characteristics.