Anupam Choubey and Reza Ghaffarian, Ph.D.
Jet Propulsion Laboratory, California Institute of Technology
Pasadena, CA, USA
Even though commercial off the shelf (COTS) column grid arrays (CGAs) have been widely used in high-reliability electronics-system applications, the ball grid array (BGA) versions are also becoming attractive because of newer technology availability, much lower cost, and lower CTE mismatches onto polymeric printed circuit boards (PCBs). For these reasons, and because of availability of extensive reliability data, university and industry sectors have developed and now offer software for projecting solder joint reliability for BGA assemblies subjected to thermal cycling conditions.
A commercial software recently developed was evaluated to determine its accuracy in projecting solder joint reliability comparing test data for BGAs and land grid arrays (LGAs) with 1156 balls/terminations. In addition, two dimensional finite element analyses (FEAs) were carried out to determine stress distribution for the LGA/BGA as well as for a CGA with 1272 columns. Only the FEA analysis was presented for CGA since the commercial software does not yet offer projection for this packaging style. Finally, this paper summarizes these findings and presents an easy to use analytical model developed to predict the behavior of BGA and CGA assemblies under thermal cycling conditions.
Within the last two years, the IPC/JEDEC team collaborating on guidelines on "Reliability & Design Finite Element Analysis Standard" indicate the need for more effective use of FEA. Projection approaches, as well as standardization of input parameter in FEA, are key in achieving acceptable projection of solder joint reliability with acceptable errors. Both a simple 2D FEA and the commercial software with a background theory were utilized to determine the effects of key package/board/solder parameters on assembly reliability and whether projection results reasonably agree with test results generated under various conditions. A summary of results are as follows:
1. The commercial software showed projected reliability for the BGA1156 reasonably well considering most industry data are for plastic BGAs. It underestimatedreliability of its LGA counterpart.
2. The trends for the effect of die size and thickness are in the right direction for the PBGA as is established by industry. However, the level of changes may not agree with industry data.
3. FEA was used to project the difference between BGA/LGA and CGA assembly reliability since the commercial software package covers only the BGA and possibly the LGA style packages.
4. Another simple to use analytical method showed the effect of height increase for BGA vs CGA as well the effect of double-sided mirror assembly configurations for BGAs/CGAs. The current analysis projected a lower reduction in thermal ycle fatigue life for double-sided mirror-image CGA assemblies than that for BGAs. No industry data is yet available for CGA double-sided mirror image.
The FEA modeling and software analyses are considered to be a preliminary exercise. Further work on these areas should shed some light on answering questions, both by FEA and by testing, as the level of reduction in fatigue life for LGAs and methods for their reliability improvement using tin-lead solder assembly. For SAC assemblies, literature data on the LGA indicate higher than expected assembly reliability under thermal cycling, which postulated to be due to larger grain sizes.
Initially Published in the SMTA Proceedings