Real-Time, In Situ Degradation Monitoring in Power Semiconductor Converters

This paper develops a methodology to identify a power semiconductor assembly’s spatiotemporal dynamic thermal response changes with degradation, and for subsequent extraction of these varying properties to estimate degradation in actively switching converters. The methodology treats both boundary (e.g. cooling) and internal (e.g. solder layer voiding) thermal-mechanical degradation modes. The paper first identifies transient thermal response sensitivities to degradation using electrothermal impedances viewed with the frequency response function (FRF) metric. A sensitivity function method is then developed to relate location of temperature measurements, harmonic content of semiconductor device losses, and degradation mode, in terms of normalized impedance spectra. Loss model parameter sensitivity analysis reveals an advantage in tracking thermal response delay variations extracted in situ with electrothermal impedance spectroscopy. To complement spectroscopy approaches using non-parametric FRF data for estimating degradation, a method to directly estimate degradation-sensitive physical parameters is developed. Experiments show automatic estimation of varying thermal resistance.