Issue link: https://iconnect007.uberflip.com/i/271756
28 The PCB Magazine • March 2014 combination of strain rate and board thickness will generate a separate distribution of failures. The experimental design is actually to produce an evaluation of the reduced factor list outlined in Table 1 at each of these six combinations. The strain rate dependence of fracture in ep- oxy resin systems is well documented in basic materials research, and has been widely incor- porated into existing industry specifications and guidelines, such as IPC/JEDEC-9702—Mono- tonic Bend Characterization of Board-Level In- terconnects [2] and IPC/JEDEC-9704—Printed Wiring Board Strain Gage Test Publication [3] . A proper treatment of this topic is beyond the scope of the current paper, but testing for this experiment wa s targeted at three specific princi- pal strain rates intended to cover the acceptable ranges of all major assembly processes. Th ose three targeted principal strain rates are 1000, 3500, 7000 micro strain per second (µe/s). Eight sub-lot variations of a mechanical test vehicle were procured. Solderable surfaces were plated in both OSP and ENIG to generate inter- faces based on both Cu6Sn5 IMC and Ni4Sn3 IMC systems. The PWBs were obtained in two laminates provided by Isola. A standard filled phenolic cured FR-4 and a non-commercial variant of the first which had been modified to reduce room temperature Young's modulus by approximately 40% in an attempt to toughen the resin system. Two versions of physical design were gen- erated, each with identical footprints and out- lines but with two distinct laminate stacks to represent incremental levels of assembly com- plexity. The first version was made up of 20 copper layers and had a nominal thickness of 0.100 inches (2.54 mm), the second contained 26 copper layers and had a nominal thickness of 0.130 inches (3.3 mm). The 185 x 185 mm test vehicles have a single BGA footprint for a 40 x 40 mm – 1.0 mm pitch device. A sectional view of the 0.100 inch version of the test vehicle is presented n Figure 4. The 40 mm 1517 I/O built up flip chip pack- ages were daisy chain devices provided by LSI. The package substrates were all plated with SAC305 over copper before spheres of the various alloys were attached. The BGA spheres were provided in Sn37Pb, SAC105, SAC305 and SAC405. Assemblies were preconditioned by one pass through the SMT oven prior to BGA attachment to account for the fact that these large devices usually exist on the top side of double sided assemblies. Where forced re- work was required assemblies were processed through two further hot gas cycles to simu- late removal and replacement of the BGA de- vice. Solder joints from these processes were properly formed with acceptable voiding and typical microstructure. Two thermal profiles were prepared for the testing program. The SnPb devices were at- tached with SnPb paste while all of the Pb-free devices were processed with SAC387 paste. The characteristic Pb-free profile is presented in Fig- ure 5. This induces minor modifications to all of the SAC alloys in the final solder joints but it is typical of results generate throughout the industry. Time zero cross sections for all pro- cess lots were inspected by optical microscopy. table 1: doe1 design Variables. Figure 4: typical structure—40 mm pkg on 20-layer board. SPHERICAL BEND TESTING continues