Accuracy of junction temperature measurement in silicon power transistor

This paper reports experimental measurements of junction temperature obtained by various electrical and thermal methods. Electrical measurements were based on temperature-sensitive junction parameters including pulse sampling of CCBFand VBE, steady state hFE, and pulsed ΔVCBF. Direct measurement of peak junction temperature was made using an infrared microradiometer. Junction temperature measurements were obtained as a function of device operating conditions, since the junction temperature distribution is a complex function of operating conditions. Comparison of these results clearly indicates why inconsistent results have been obtained in the past using the various electrical methods. The VCBF, VBE, and ΔVCBFmeasurement indicated temperature much lower than the true peak temperature obtained from the mieroradiometer, but they did indicate the correct variation with operating conditions. The hFEdata, on the other hand, indicated temperature values higher than the actual values and did not indicate the correct variation. Thermal impedance values calculated from the VBEor VCBFjunction temperature data are typically one-third to one-half of the actual value based on peak junction temperature. This error is shown to be a direct function of the nonuniformity in junction temperature. This error arises primarily because the calibration data is not applicable to actual operation with nonuniform junction temperature. Devices with voids or nonuniform current distributions are shown to have large errors. These discrepancies explain much of the confusion involving thermal limitation and results of reliability screening tests.