Electronics today are designed into dynamic and often unpredictable environments. The dynamic thermal behavior of the device influences its functional and lifetime reliability. Failure to properly capture the dynamic thermal behavior in thermal design models leads to overdesign, more field failures, and longer design cycles. The challenge for designing reliable electronic products consists of developing accurate predictive models that respond correctly to power dissipations that vary as a function of time in an acceptable amount of time.
Developing accurate predictive package thermal model has been constrained by the finite amount of engineering time available, and the quality or appropriateness of the available measurement data. Using thermal measurement data that captures the transient response of the device offers the best opportunity for designing products that experience changes in power dissipation or environmental conditions. Thermal models calibrated against fewer measurement results increases the chance of matching measurement results for the wrong reason and also limits the thermal model scope that it will provide accurate predictions. Calibrating a thermal model against multiple data points increases the probability that the model will be accurate for any operational environment.
This presentation discusses the risks associated with designing with un-calibrated models, the challenges associated with thermal model calibration and introduces a new approach for automatically calibrating a thermal model against multiple transient response curves.