Analyzing & Understanding – Variation Effects

In recent years, the use of the electrical test method (ETM) has become more common to measure the temperature of semiconductor junctions. This method makes use of correlating the relationship of the semiconductor forward voltage to the temperature of the device, which is often linear but is always a unique correlation. This method has been recognized by JEDEC, and in particular for LEDs in 2012 when standards were issued (JESD 51‒50 through 53). The device under test (DUT) must be calibrated, or the Vf-Tj relationship must be determined to use this method. This is a straightforward method, and after completion the DUT may be used in an assembly to fully characterize a system’s thermal performance.

However, one issue that may occur is that of an assembly where the DUT may not be easily removed for calibration or, where for various reasons, it is not desired to do so. Without having the DUT calibrated, the unanswered questions are how one may find the junction temperature of the device and to what uncertainty, if at all. This paper attempts to answer these questions for a sample LED population and make some generalizations through the use of population statistics and voltage bins of LEDs.

Two voltage bins of LEDs were obtained from a major LED manufacturer, each populated with 15 LEDs. The LEDs were mounted on small metal core printed circuit boards (MCPCBs) and first calibrated individually. Per the ETM, the calibrations were conducted with low input currents (5 mA); this was significantly lower than the normal current used for Vf bin determination and became significant in the statistical analysis. The results show that when an LED population has a Vf spread of under 0.06 VDC at the 5 mA forward current, the Tj of the LEDs is known to a 12 °C band with 95% confidence and 7 °C with 75% confidence. A tighter bin spread less than 0.035 VDC provides lower bands of 8 °C and 5 °C with 95% and 75% confidence, respectively. The Vf spread at 5 mA drive current is the key factor in determining how close the Tj may be known with reasonable confidence.

Analyzing & Understanding – Variation Effects