b'Engineer Innovation | Electronics & Semiconductoraccuracy, (shown later), and solutionfrom the active area of the die to the time. package case is also measured. The overall package surface is included in Thermal resistance measurements the measurements. The results from Using the example of a QFN and athis method are also much more HFCBGA package, when we comparerepeatable. the results differences from the transient test method used byPerformance of the calibrated Simcenter T3STER with that ofSimcenter Flotherm modelstraditional static test methods, we seeThe improvements in the temperature one order of magnitude difference withimprovements are illustrated in Figure the high-performance flip chip BGA9. In the bar chart, the calibrated IC (HFCBGA) package (20.6 percent)models of the QFN and HFCBGA compared to the QFN (2.6 percent), aspackages are less than one percent shown in Figure 8. The powerdifference compared to the measured dissipation for the QFN package (2.5test results. In comparison, the non-watts) was much lower than thecalibrated models have larger errors of HFCBGA package (100 watts). ASEover two percent difference. surmise that with an increasing power dissipation, the differences in resultsAccuracy of compact thermal modelsbetween the static and dynamic testWe can also compare the accuracy of methods are magnified.the exported dynamic compact thermal models (DCTMs) from the Simcenter In the static test methodology, pointT3STER measurements. A transient temperatures are measured on the casesystem level model of an Access Point surface. Consequently, the location ofRouter was created (Figure 10), the thermocouple on the top surfaceoriginally with calibrated detailed will have an effect of the calculatedmodels of the QFN and HFCBGA resistance. On the other hand, thepackages on the PCB. 10dynamic test method uses an electrical Junction to Case Thermal Resistance (W/K)test (JEDEC Standard JESD51-1) to senseThe detailed IC models were then 7.663 7.467 -2.6%8the junction temperature. The heat flowreplaced with the DCTMs. Plotting the path taken by heat being dissipatedjunction temperature response curves 6QFNHFCBGAR 1 R 2 R 3 R n4Driving Point 2+20.6%C 1 C 2 C 3 C n0.162 0.2040Figure 7: Dynamic Compact Thermal model in the form of a R-C Ladder Network Static TransientMeasurement MethodBoard Level -2.6% +0.9%10 10092.1 89.7 92.990Junction to Case Thermal Resistance (W/K) Junction Temperature (C)7.663 7.467 -2.6%8 8070 -2.8% -0.2%6 60 57.9 59.5 58QFN QFNHFCBGA 50 HFCBGA4 40302 20+20.6%0.162 0.204 100 0Static Transient Exp. Sim._Ori Sim_ModMeasurement MethodFigure 8: Differences in measured Junction to Case ThermalFigure 9: Improvements in Junction Temperature predictions with Resistan Boarusi velStatic and Transient measurement met +0.9% calibrated modelsces d Le ng -2.6% hods100 92.1 89.7 92.9907080Junction Temperature (C)70 -2.8% -0.2%60 57.9 59.5 58QFN50 HFCBGA403020100Exp. Sim._Ori Sim_Mod'