Calibration of transient FE simulation: Improvement of post-processing and simulation automation

Abstract

Reliability and lifetime of LED modules depend critical from drive current and junction temperature. Transient thermal analysis (TTA) is widely used to measure the transient thermal impedance Zth and the thermal resistance Rth of LEDs to access the junction temperature. To predict the junction temperature of an LED in an application during product design and development calibratedfinite element (FE) models are required. In this paper, the correlation between simulated transient temperature data and the experimentally by TTA measured forward voltage (Vf(t)) is analysed in detail. Using a test chip, it is demonstrated that post-processing the average temperature of the junction of the FE model is the appropriate approach to correlate the simulated transient temperature data to the V/t) measurements. A FE-model for a family of white high-power LED, i.e. different number of LED dies on ceramic sub-mounts of different sizes, is developed using ANSYS and calibrated to the Zth(t) curve measured by TTA. The calibration is done in the time-domain, i.e. using the Zth(t) curve and its logarithmic time derivation b(z). The FE model is fitted to the experimental data. Due to the phosphor conversion, the heat load must be divided on the epitaxial layer (EPI) and the phosphor. The heat load distribution influences the Zth(t) curve significantly from the μs-to the ms-time range and must be considered as important parameter when fitting the simulation model. The calibration of the FE model is done using the average and the maximum temperature value of the EPI from the FE simulation. A significant difference of 20% between the thermal resistances of the thermal interface layers is obtained. The thermal performance of the thermal interfaces is significantly overestimated when using the maximum temperature for post-processing.