Combining experiment and FEA into one, in device characterization

Experimental thermal transient data taken to characterize a specific packaged semiconductor device is often not as useful as desired. For instance, the temperature sensitive parameter (TSP) structure accessible for temperature measurements may be thermally far from the actual source of heat input to the device, resulting both in time delay in the measurements and a high degree of uncertainty in the amount of temperature gradient not measured. A specific device, in this case a USB Filter, is presented as an example of the need to carefully combine experimental measurements with Finite Element Analysis (FEA), in order to construct an accurate thermal characterization of the device. In this example, the temperature measurements were made at TVS diodes almost adjacent to the series resistors, and the heat was input at the series resistors. Yet the FEA model showed that the temperature gradient between the hottest spot in the resistor and what would actually be measured at the TSP was a significant fraction of the overall thermal resistance of the device. This paper discusses several facets of the process, from FEA transient accuracy, to experimental limitations on speed and location of the measurements.