Fast thermal profiling of power semiconductor devices using Fourier techniques

Abstract

Accurate prediction of temperature variation of power semiconductor devices in power electronic circuits is important to obtain optimum designs and estimate reliability levels. Temperature estimation of power electronic devices has generally been performed using transient thermal equivalent circuits. In the presence of varying load cycles, it has been typical to resort to a time-domain electrical simulation tool such as P-Spice or SABER to obtain a time series of the temperature profiles. However, for complex and periodic load cycles, time-series simulation is time consuming. In this paper, a fast Fourier analysis-based approach is presented for obtaining temperature profiles for power semiconductors. The model can be implemented readily into a spreadsheet or simple mathematical algebraic calculation software. The technique can be used for predicting lifetime and reliability level of power circuits easily. Details of the analytical approach and illustrative examples are presented in this paper.