Next Generation Monitoring of SiC mosfets Via Spectral Electroluminescence Sensing

This article presents a new methodology for monitoring next generation SiC
mosfet
-based converters. The methodology uses the spectral distribution of electroluminescence that is emitted by the body diodes of SiC
mosfet
s and includes information on device current as well as the junction temperature. Intelligent estimation algorithms based on artificial neural networks allow processing the extracted information to separately estimate device current and junction temperature with high bandwidth. The resulting unified monitoring of device current and junction temperature galvanically isolates the sensors via optical guides from the power module. Thus, it can be effectively integrated into future power electronic modules to replace external temperature and current sensors and enable monitoring for safe and reliable long-term operation. Previous publications successfully investigated how either temperature or current information can be extracted from the intensity of the emitted light if the other variable is known. However, they did not aim to extract both variables at the same time from one measurement. This work presents how multiple optical sensors with different wavelength sensitivities and artificial-intelligence techniques can separately extract both variables. The proposed technology is evaluated using an automotive-grade SiC power module. Its utilization for monitoring future SiC-based converters allows reducing converter size and cost while increasing reliability.