Issue link: https://iconnect007.uberflip.com/i/960594
42 SMT007 MAGAZINE I APRIL 2018 size, area and thickness. Similarly, the same material sets were used: mold compound, die attach, and assembly factory. Assembly The PCB and assembly details are in Table 3. Assembly of daisy-chain parts to boards followed industry norms. Solder paste printed to boards used the alloy Sn3.8%Ag0.7%Cu (SAC387) and a no-clean flux system. Place- ment of parts to boards used a dual eyepiece placement machine for aligning parts to solder- paste print. Finally, boards were run through a 10-zone reflow furnace with a peak tempera- ture between 235°C and 245°C. Experimental Matrix The six experimental cell combinations shown in Table 4 were built and tested. Sixteen from each were subjected to monitored TCoB -40°C/+125°C cycling, with additional unmonitored units included for crack propaga- tion measurements. Non-destructive T0 char- acterization was performed on all cells. Cycling, Electrical Testing and Data Analysis Assemblies were tested in an Air-to-Air Thermal Shock (AATS) dual chamber system whereby one chamber remained hot (+125°C) and the other remained cold (-40°C). An eleva- tor system moved test boards between these chambers within about 10 seconds. Both chamber dwell times were set at 30 minutes totaling 1 cycle/hour. Typically, ~5 minutes was required to reach equilibrium, yielding ~25 min dwells. Figure 3 displays a typical temperature profile obtained by placing thermocouples in the assembly solder joints. Assemblies were monitored in-situ during cycling using a 1.2mA current through each net. An event detector logged a failure when a net resistance exceeded 300 ohms. Failures were defined per IPC-9701 [4] . Generally, daisy- chain resistances were few ohms at the begin- ning of an experiment. Net resistance did not immediately change measurably during early stages of solder joint crack growth, but climbed quickly as the crack approached 100%. Therefore, a net failure was logged when any one of the solder joints in that net had a crack near 100%. For each test cell, cycling continued until at least 75% of the samples failed, after which the data were fit to a 2-paramter Weibull distribution using maximum likelihood esti- mate (MLE). Three metrics of solder joint life- time were extracted from each distribution: (1) characteristic life (Eta), (2) extrapolated Table 3: PCB and SMT assembly details. Figure 3: AATS temperature profile. Table 4: Experimental matrix.