Abstract
In this paper, an approach to determine the thermal impedance of a multi-chip silicon carbide (SiC) power module is proposed, by fusing optical measurement and multi-physics simulations. The tested power module consists of four parallel SiC metal-oxide semiconductor field-effect transistors (MOSFETs) and four parallel SiC Schottky barrier diodes. This study mainly relies on junction temperature measurements performed using fiber optic temperature sensors instead of temperature-sensitive electrical parameters (TESPs). However, the fiber optics provide a relatively slow response compared to other available TSEP measurement methods and cannot detect fast responses. Therefore, the region corresponding to undetected signals is estimated via multi-physics simulations of the power module. This method provides a compensated cooling curve. We analyze the thermal resistance using network identification by deconvolution (NID). The estimated thermal resistance is compared to that obtained via a conventional method, and the difference is 3.8%. The proposed fusion method is accurate and reliable and does not require additional circuits or calibrations.
Highlights
Power modules with high power densities are increasingly in demand for various applications, including electrified vehicles, transportations, and renewable energy systems [1,2,3]
Wide-band-gap devices such as silicon carbide (SiC) and gallium nitride (GaN) devices have been actively studied because of their potential to overcome the limitations of Si devices [4,5,6]
It is expected that such parallel SiC power modules would encounter various thermal reliability issues originating from their high power density
Summary
Power modules with high power densities are increasingly in demand for various applications, including electrified vehicles, transportations, and renewable energy systems [1,2,3]. After the transient curve is derived using the two methods (fiber optic measurement and simulation-based estimation), the thermal impedance is extracted using network identification by deconvolution (NID), which is a typical thermal impedance extraction method [26,27] This method yields the thermal resistance, which is compared. CTTohhneevettnhhteeirormnmaaalllMrreeetsshiisostdtaa(nnRcceeefeccraeannncebb)ee ccaallccuullaatteedd uussiinngg EEqquuaattiioonn ((11)) bbaasseedd oonn tthhee mmeeaassuurreedd ppoowweerr ddiissssiippTaahttieioonnthaaennrmdd atteelmmreppseeirsraatattuunrcree ddciiafffnfeerrbeeennccceeaolocffujjuluannteccdttiioounnstitonogccaaEsseqeuation (1) based on the measured power dissipation and temperature difference of RRRth. The resTuhletsreosfulrtespoef aretepdeaetexdpeexrpimeriemnetnstsaraeresshhoowwnn iinnFFigiguruer8e.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call