Issue link: https://iconnect007.uberflip.com/i/960594
48 SMT007 MAGAZINE I APRIL 2018 ture range. One sample per cell was measured. Pad design was not expected, nor observed, to influence package warpage. Therefore, the three samples for each substrate material type were considered to be from the same popula- tions. Packages using the low CTE dielectric warped considerably less on the cold side (-40°C to +25°C), 25 µm on average. This was expected based on the following analysis. Package warpage is driven primarily by the CTE mismatch between the mold compound and substrate. Below Tg, the mold compound expansion is 9ppm/°C. The low CTE dielectric expansion is 11ppm/°C, but the standard dielectric is 16ppm/°C. Therefore, the package will warp as it is cooled from the neutral temperature (roughly 150°C to 175°C.) But since the delta between mold compound and substrate is greater for the standard dielectric, the warpage will also be greater. Solder Joint Characterization One unit from each of two hybrid cells 2 and 6 were selected for T0 solder joint characteriza- tion after mounting to the PCB. Standoff and solder joint diameter were measured for four joints of each pad type (SMD and NSMD) for each sample. Examples are shown in Figures 9 and 10. Surface evolver predictions based on single joints anticipated the NSMD solder joints would be about 7 µm shorter. However, the actuals were much closer since attaching both types in a single package forces them to a common standoff. The standoff differences observed here between the two types possibly resulted from PCB or package warpage. Overall the joints were shorter and wider than predicted. Temperature Cycle Electrical Test Results Sixteen components from each of the six cells described in Table 4 (p. 42) were mounted on PCBs and cycled -40°C to +125°C as described above. The test was terminated after all cells had at least 12 units fail (75%). The results are plotted in Figure 11. The legend symbol n/s represents the number of units tested (n) and number of suspensions (s). Also shown are the fitted two-parameter Weibull character- istic life (Eta) and shape factor (Beta). Good fits were obtained for all except cell 1, which had an outlier first failure. Inspection of the graphs suggests that the six populations can be divided into three groups. From least to most reliable: (A) cell 1; (B) cells 2,3&4; (C) cells 5&6. This suggests that the low CTE dielec- tric or either hybrid design offered an improve- ment over the baseline, and that combining them was best. Three metrics (Eta, extrapolated 1% fail- ure cycle and first failure) were extracted from each distribution and have been summarized in Table 5. Each of these metrics was linearly Figure 9: Cross-section of typical SMD solder joint after mounting to PCB. Figure 10: Cross-section of typical NSMD solder joint after mounting to PCB.