Authored By:
Jean-Paul Clech
EPSI Inc.
Montclair, NJ, USA
Summary
Cycles-to-failure versus Distance to Neutral Point (DNP) data for SnPb and lead-free assemblies under Accelerated Thermal Cycling (ATC) conditions are observed to follow three main trends: 1) no DNP dependence; 2) a power-law dependence with an exponent near -1; 3) a power-law dependence with an exponent close to -2. The first two trends are at a significant departure from standard Coffin-Manson types of models for SnPb and lead-free assemblies (e.g., IPC 9701). Deviations of DNP test data from the standard models can be significant and have serious implications for board designers: a) solder joint life based on Coffin-Manson type of models may be under-estimated by a large factor - and packages rejected - for new designs using a larger die or package; b) moving to smaller die or packages, the existing models may over-estimate the reliability gains associated with smaller size components.
This leaves board designers with significant uncertainties and reliability risks. This paper resolves the above differences by means of a simple strength-of-materials model that provides physical and quantitative insight into the combined effect of assembly pitch and DNP on thermal cycling life. The pitch, which affects the stiffness of the assembly, is a significant factor that is not accounted for in standard models. The proposed "pitch and DNP" life model accounts for the pitch stiffness effect and is validated against numerous ATC datasets.
Conclusions
A first order model has been developed that highlights the effect of the assembly pitch on solder joint reliability under thermal cycling conditions.
Besides determining the maximum DNP at critical corner joint locations, the assembly pitch has a direct effect on the stiffness of a slice of an assembly of width the pitch. Everything else being equal, a smaller pitch makes for more compliant assembly slices that stretch and bend more easily than large pitch slices during thermal cycling.
As a result of the pitch stiffness effect, solder joint life goes as the inverse of the product of the pitch and the maximum DNP. This is a simple design rule or "rule of thumb" that is of use when considering design changes from one chip size to a different chip size, smaller or larger. Concurrent changes in pad sizes and solder ball diameters, and their effect on solder joint life also have to be considered. The latter parameters are not factored in the present model. Their impact on solder joint life will be addressed in a future publication.
A consequence of the pitch stiffness effect is that, for assemblies with a fixed pitch, solder joint cyclic life goes as the inverse of the DNP when component size increases. This is at a significant departure from standard models and IPC-9701. The fundamental reason behind this discrepancy is that standard models are based on solder joint fatigue life being strain-range dependent (a la "Coffin-Manson"), whereas the present model is strainenergy based with the assembly stiffness (or compliance) relieving some of the shear strains associated with the thermal expansion differential between board and components.
The model, including the pitch stiffness effect, is supported by over a dozen experiments as well as SnPb and lead-free test data from independent sources.
While this paper highlights the pitch stiffness effect in a simplified algebraic model - equation (14) - and with supporting test data, the pitch stiffness effect is not new and is already factored in life prediction models developed by the author for SnPb assemblies (Clech, 1996) and SAC lead-free assemblies (Clech, 2005). The latter models include an assembly stiffness parameter K, and thus the pitch, in the slope of isothermal stress reduction lines that are used to compute stress/strain hysteresis loops during thermal cycling.
Standard models or modified versions of the standard models that do not include an assembly pitch parameter, and thus do not account for the pitch stiffness effect, should be examined closely and validated against relevant ATC test data prior to use.
Initially Published in the SMTA Proceedings
|
