Online proceedings - EDA Publishing Association

q (RthΣ, t)q (R th , t )10.80.60.40.200 1 2 3 4 5 6 7 8 9 10R th Σ [K/W]Figure 11: Heat flow distributions for the 1D solder dieattach device at the diverging point of the Zth-curves withand without TIM (i.e. at t = 0.003s). The area high-lightedin gray corresponds to the steady-state Rth-JC of this device.In this picture the steady-state Rth-JC equals the areaunder the graph of q = 1 (i.e. the steady-state heat flowdistribution) limited by the package boundary. This area hasbeen highlighted in gray in figure 11 (solder die attach) and12 (glue die attach).While for the solder die attach package (figure 11) thearea under the q-graphs (the Zth) is approximately equal tothe gray area (the steady-state Rth-JC), the same is not truefor the glue die attach package (figure 12). In the latter casethe high thermal resistance of the low conductive glue layerdelays the formation of the heat flow through the package.While there is already a non-negligible flow through theinterface layer, causing the two Zth-curves to diverge, theglue die attach package contributes still less than half of itssteady-state Rth-JC to the Zth. I.e. in this case the TDIMmeasurement of this package would produce an Rth-JCvalue which is only about half of the real value.q (R thΣ, t )10.80.60.40.20PackageTIMInterface resistance without TIMSolder die attachPackageGlue die attachwith thermal grease0 2 4 6 8 10 12 14R thΣ [K/W]Figure 12: Heat flow distributions for the 1D glue die attachdevice at the diverging point of the Zth-curves with andwithout TIM (i.e. at t = 0.0078s). The area high-lighted ingray corresponds to the steady-state Rth-JC of this device.TIMwithout thermal greaseInterface resistance without TIMwith thermal greasewithout thermal grease24-26 September 2008, Rome, ItalyFigure 11 shows that even in the solder die attach packagethe heat flow is not yet fully developed when the Zth-curvesstart to separate. The TIM / interface resistance-contributionsof the heat flow more or less compensate the missing partinside the package.Naturally the simulation results for a 1D heat flow pathcan only provide a qualitative analysis of the TDIMmeasurement. Actual Zth-measurements suggest that thedifference between the Rth-JC obtained by comparing thederivatives of the Zth-curves and the steady-state Rth-JCvalue is smaller than predicted by the 1D analysis.V. EXPERIMENTAL RESULTSContrary to simulations where the steady-state Rth-JC caneasily be computed, the Rth-JC of a real semiconductordevice is initially unknown and not as easy to determine.However the evaluation of the structure function asdescribed in [5] should yield results accurate enough to serveas reference for the steady-state Rth-JC of the samples.Table II shows the measurement results for three groups ofsemiconductor devices:TABLE IIComparison of the Rth-JC values obtained by transient measurementsevaluating the structure function (R th-JC Stf.) and the difference of thederivatives of the Zth-curves (R th-JC |Δ(da/dz)|).TO device with solder die attachSample R th-JC Stf. R th-JC |Δ(da/dz)| Deviation1 3.11 K/W 3.09 K/W -0.6 %2 3.21 K/W 3.21 K/W 0.0 %3 3.09 K/W 3.11 K/W 0.6 %4 3.33 K/W 3.15 K/W -5.4 %Exposed pad device with thin glue die attach (approx. 30μm)Sample R th-JC Stf. R th-JC |Δ(da/dz)| Deviation1 2.97 K/W 2.69 K/W -9.4 %2 2.93 K/W 2.80 K/W -4.4 %3 3.27 K/W 3.12 K/W -4.6 %4 4.15 K/W 3.70 K/W -10.8 %5 3.71 K/W 3.70 K/W -0.3 %6 3.67 K/W 3.48 K/W -5.2 %Exposed pad device with thick glue die attach (approx. 50μm)Sample R th-JC Stf. R th-JC |Δ(da/dz)| Deviation1 4.65 K/W 3.99 K/W -14.2 %2 4.50 K/W 3.98 K/W -11.6 %3 4.16 K/W 3.62 K/W -13.0 %4 3.89 K/W 3.53 K/W -9.3 %5 4.68 K/W 4.09 K/W -12.6 %6 4.58 K/W 4.12 K/W -10.0 %©EDA Publishing/THERMINIC 2008 18ISBN: 978-2-35500-008-9