Clustering Algorithm for the Screening of SiC MOSFETs Connected in Parallel

• Main Authors: Lv, Fang, Mao, Saijun, Lei, Guangyin, Tan, Qiuyan, Yang, Shuhao, Ding, Yujie
• Format: Conference Proceeding
• Language: English
• Subjects: classification; Clustering algorithms; current sharing; DBSCAN; Manufacturing processes; MOSFET; parallel-connection; Resistance; Semiconductor device modeling; SiC MOS-FET; Silicon carbide; static characteristics; Statistical analysis
• ISSN: 2329-3748
• DOI: 10.1109/ECCE53617.2023.10362020

Parallel connection of SiC MOSFETs is widely used for high-current and high-power applications. Due to manufacturing process immaturity, the parameter dispersion of SiC MOSFETs is severe, the current imbalance for the paralleled SiC MOSFETs is a critical challenge. Firstly, the model of the relationship between the current disparity of paralleled devices and static characteristics, such as threshold voltage and onstate resistance of SiC MOSFETs is proposed in this paper. Then, the concept of multivariate statistical analysis is introduced to quantify the similarity of static characteristics into distance parameters, and DBSCAN (Density-Based Spatial Clustering of Application with Noise) algorithm is developed for the screening of SiC MOSFETs to mitigate the current imbalance. To compare the proposed method with traditional screening methods, a double pulse test platform with four SiC MOSFETs parallel connection circuit is built. Less than 4% imbalance between the drain-source current for the screened SiC MOSFETs with parallel-connection has been achieved.