The first test chip encapsulated into a ceramic DIL package was used in various case studies. In one of our experiments we investigated the effect of an IC socket.
We had two setups, in which the only difference was, that the test chip was soldered directly into the test board or was plugged into an IC socket soldered into a similar PCB having the same geometrical and thermal properties. The final results obtained thermal transient measurement are shown by the structure functions plotted in Figure 7-4. The increased thermal resistance towards the ambience introduced by the socket is clearly visible on the plots. This value is about DRth = 25 K/W (see the difference of the singularities of the curves).
Figure 7-4: Effect of a chip socket, demonstrated by the differential structure
function.
In the second experiment the test chip was plugged into a socket and in one of the measurements a cooling mount was attached to the top of the IC package. The obtained structure functions are plotted in the same diagram in Figure 7-5. This clearly shows that the cooling mount introduces an excess thermal capacitance (the corresponding "bump" in the structure function is marked by an arrow) as well as it reduces the thermal resistance by a great extent, compensating for the excess thermal resistance introduced by the IC socket.
A third experiment was performed in order to demonstrate the sensitivity of our tester. The same setup was used: the test chip plugged into an IC socket that was soldered into a printed circuit board. The difference between the two measurements shown in Figure 7-6 was, that in one case the test IC was fully plugged into the socket, while during the other measurement it was only slightly plugged in: to realize the electrical connections only. In this second case the heat removing path was extended by the extra pin-length that remained outside the socket. This excess heat resistance can be identified as the difference between the singularities of the two curves, which is about DRth= 5 K/W, corresponding to 1.5 mm excess pin length of the IC.
Figure 7-5: Effect of a cooling mount, demonstrated by the differential structure
function.
Figure 7-6: Effect of the pin length
Die attach failures are very dangerous packaging problems, strongly influencing the reliability and difficult to detect.
We have elaborated a method with which die attach problems of packages may be detected fast. It is based on the evaluation of the thermal transient curves that can be obtained by thermal transient testers. Evaluation of the thermal transient curves leads to the structure function of the heat flow path, presented on the screen of the transient tester equipment. The measurement was carried out using T3Ster, with a resolution of 1m s and 0,012 ° C, in the arrangement of Figure 7-7.
Figure 7-7: The measurement arrangement
Figure 7-8: Measured thermal transient curves
Figure 7-8 presents measured thermal transient curves of good and bad devices. It is visible that there are slight differences in the curves, but it is very hard to detect the source of the distortion of these curves. To find the origin of the distortion in the heat flow path we have to examine the so called Structure Function, which can be generated by direct transformation from the measured thermal transient curves. The differential structure function (Figure 7-9) gives the cross sectional area of the heat flow path measured from the chip with respect to the cumulative Rth thermal resistance, measured also from the chip.
Figure 7-9: Differential structure functions of good and bad devices
Figure 7-10: The differential structure function of the reference device. The
arrows point to characteristic locations of the structure.
It can be noticed that there are characteristic differences between the presented functions. To understand these differences we discuss first the differential structure function of C08, the known good reference device (Figure 7-10). The left-hand side of the curve refers to the chip, the right hand end to the cold plate. Arrow 4 shows this point. The value read on the horizontal axis gives the steady state thermal resistance between the chip and the cold plate, it is 3.2 K/W. The zigzagged beginning of the curve shows the presence of some noise, but an average K=0.1 value can be considered. In case of silicon material this is equivalent to a 19.7 mm2 cross sectional area. The next peak 1 refers to the heat capacitance of the transistor case, determined by the dominant heat capacitance of the copper base plate of the case. The next peak 2 is the heat capacitance of the copper island of the mounting plate, peak 3 is the heat capacitance of the mounting plate itself. After locating these characteristic points, the partial thermal resistance values can be read from the figure. The thermal resistance between the 1-2 points is about 0.6 K/W, this is the thermal resistance component of the transistor soldering. Between points 2-3 the thermal resistance of the plastic coating can be read, in our case this is about 1.3 K/W. The thermal resistance between the mounting plate and the cold plate determines the distance between the points 3 and 4.
Figure 7-11: Comparison of the differential structure functions of C02 and C08.The
shift in peak 3 suggests soldering error
Comparing the structure function of C02 to the reference function (Figure 7-11), we notice the following. Although peak 2 may be recognized on both curves, at C02 a characteristic minimum is visible at the right hand side of it and the thermal resistance to the next plateau is much (2.5 times) higher. This suggests the presence of a soldering problem.
The differential structure function of C17 is presented in Figure 7-12. In case of C17 peak 1 is shifted to the right with a 0.4 K/W value and the entire rest of the curve shows the same right shift. This means the presence of an extra thermal resistance between the chip and the copper platform of the case, which indicates that the chip is not attached to the platform appropriately. Note, that the die attach problem does not manifest in a characteristic deviation in the steady state thermal resistance, but in the differential structure function the deformation is characteristic.
Figure 7-12: The differential structure function of C17 referred to the structure
function of C08. The shift of peak 1 suggests die attach failure
It is interesting that these problems can be noticed already on the measured transient curves. Examining the measured transient curves we can notice that the measured curves of both devices are running above the nominal curves with about 20-25% in the 0.1-0.2 sec range of the transient measurements. This is a very important experience, suggesting that die attach failures can be noticed by short transient measurements, offering the possibility of using the method even for in-line testing.
Soldering errors of the module can be recognized from steady state thermal resistance measurements as well, but such measurements are more time consuming. We found however that steady state thermal resistance measurements can be approximated by short transient measurements. In our example the steady state was reached in about 300 sec but all the problematic devices could be detected with a 10-30 sec transient measurement.
Figure 7-13: Comparison of the differential structure function of C05 and C08.
The shift of peaks 1 and 3 shows the presence of both die attach failure and
soldering problem