A Proposed Remedy for Ball-in-Socket Defects
December 31, 1969 |Estimated reading time: 8 minutes
Ball-in-socket defects often occur randomly on BGA components without any obvious root cause. Of concern to customers is that these defects are not always caught during inspection or functional test. This article covers several commonly proposed root causes of ball-in-socket defects, and the use of an experimental reflow profile to solve these defects.
By Brian Smith
Ball-in-socket defects often occur randomly on BGA components without any obvious root cause. Figure 1 shows a typical ball-in-socket defect in which the solder sphere appears to be connected to the bulk solder, but actually is resting in a depression without making a metallurgical connection. Of considerable concern to customers experiencing these defects is that they are not always caught during inspection or functional test. Due to the formation of the defect typically resulting in partial contact between the bulk solder and the solder sphere, the final circuit often passes functional, optical, and in-circuit testing (ICT). Because there is no real metallurgical connection, weak joints may fail shortly after passing through all post-soldering tests. Circuit boards with ball-in-socket defects often fail during post-solder assembly processes, shipment, upon thermal expansion and contractions, or in the field.
Commonly Proposed Root Causes
There are several proposed root causes for ball-in-socket defects.
Solderability of the paste or sphere. One belief is that some combination of a difficult-to-solder component (or component contamination) and a solder paste that is not sufficiently active enough to wet the component surface will cause ball-in-socket defects. However, solderability tests on raw components rarely support this claim. Because this defect commonly occurs in easier-to-solder applications (wetting of tin/lead solder to tin/lead solder spheres), it is unlikely that poor component solderability and/or weak flux activity is the culprit.
Variable height of paste deposits. It has been proposed that this defect can occur as the result of random paste deposits that lack sufficient height or volume, preventing the component from being placed into the paste properly. This could occur due to a clogged aperture, poor paste release, or stencil damage. However, this would create open solder joints that are more obvious than typical ball-in-socket defects.
Lack of co-planarity of a BGA. Another proposed root cause of this defect is the co-planarity of BGA spheres. Co-planarity issues may be a factor in ball-in-socket defect formation as they can result in a loss of contact between the solder and the component during soldering. However, solder spheres are manufactured to tight tolerances, making it unlikely that sphere co-planarity would be the root cause behind ball-in-socket defects.
Charting the locations of ball-in-socket defects across a BGA shows that the defect occurs more frequently on one side or corner of the component than the other. Customers have noted that more than 95% of ball-in-socket defects occur in this manner. Poor solderability, inadequate paste height/volume, or component co-planarity could not cause this type of repeatability in one geographical area of the component. There must be another culprit.
Figure 1. Side view of a ball-in-socket defect.
After careful consideration and examination of the set of circumstances within several customer investigations, it became clear that the defects often appeared on the side or corner of the component that was cooler than the opposing side or corner. These determinations were made by attaching thermocouples to all four corners of the component to establish the thermal gradient (∆T) across the device. Defects occurred repeatedly on the cooler corner of the component, meaning that the top-left corner of the component would have been cooler by at least a few degrees compared to the bottom-right corner. This led to the hypothesis that the ∆T across the component may be generating the defect.
Based on this finding, some customers increased the temperature across the entire board to also increase the temperature at the cooler portion of the component. However, such efforts did not reduce defect levels. This indicates that, while the reflow profile is somehow involved in the defect formation, the solder not fully melting or not having enough time to adequately wet to the component did not drive defect formation.
Figure 2. Customer’s original reflow profile.
One customer experiencing a high incidence of ball-in-socket defects agreed to try an experimental reflow profile as a potential solution. This customer reported an original ball-in-socket defect level of about 0.9%. They investigated commonly proposed root causes, but the defects continued to occur at the same rate. The customer provided a copy of their reflow profile and analyzed the defect incidence levels and locations. It was evident that the defects were not occurring randomly from a geographical standpoint across the component. Nearly all defects were clustered in one corner of the component. The customer was using a reflow profile similar to the one shown in Figure 2, which represents a ramp-to-spike (or tent) profile with no discernable soak zone. The ramp rate was a steady 0.9°-1.0°C/sec. for the pre-reflow portion of the profile, with a 210°-215°C peak temperature. After measuring the temperature on all four corners of the component, it became evident that ball-in-socket defects were occurring on the coolest corner of the component. Furthermore, the ∆T across the component was ~7°C as the hotter corner reached the reflow temperature (183°C), leading to a significant difference in melting times. This temperature difference was caused by components surrounding the large BGA and was exacerbated by the fact that the hotter corner was at the leading edge of the component, and would experience the higher temperature of subsequent reflow zones in the oven sooner.
The customer agreed to try a reflow-profile adjustment. The hypothesis is that ∆T issues are forcing uneven wetting across the component, leading to some degree of component tilt. Component tilt may force component leads in the coolest corner of the component to lose contact with the solder paste. This may only occur if the hottest corner reaches a molten state sooner than the coolest corner begins to melt and wet to component leads. If the solder can be forced to melt at a more consistent rate at all corners, this tilting may not occur and ball-in-socket defects may be eliminated. Figure 3 shows a proposed profile to achieve this.
Figure 3. Proposed reflow profile for defect reduction.
Figure 3 includes reflow-profile data for two locations on the board, representing the hottest and coolest corners of the component where defects are located. The first goal of the revised profile was to minimize the ∆T across the component as the solder reached its liquidus temperature. This can be achieved by introducing a long, hot soak zone into the process. The target for soak time is 160°-180°C for 75+ sec., with an ultimate goal of a soak temperature near 175°C with a minimized ∆T. The second goal of this profile was to quickly transition into the solder’s liquidus phase. With a high-soak temperature and minimized ∆T, rapid heating between 175°-190°C is critical to force the solder to melt nearly simultaneously across the component. The customer implemented the experimental reflow profile without any additional process or material changes.
The customer had reported a defect level of 0.9% over several weeks for several thousands of boards using the original profile. After converting to the experimental reflow profile, zero defects were created over the next 1,000 boards built.
Based on these findings, a logical mechanism for defect formation was proposed. When using the previous reflow profile, the ∆T across the component forced the solder to melt on one corner of the component - creating wetting action between the solder and the component in a localized region. This wetting action exerts a downward “pull” on the component in the hotter corner, which can lead to a slight tilt to the component, while the hotter corner is above liquidus and the cooler corner is below liquidus. This force is assumed to not be strong enough to uproot component leads from the paste in the cooler corner. However, when the defect is formed, we assume that solder spheres in the cooler corner lose contact completely with the solder paste. If the solder sphere exits the paste cleanly (without any paste remnants), metallization is void of any oxidation prevention typically derived from the paste, resulting in rapid oxidation of the solder sphere. As the solder sphere exits the paste deposit, the paste itself is free from a massive heatsink, and can become molten before the sphere is forced back into the paste. Paste melting will cause solder wetting on the pad, meaning the flux will drop to the board level to promote spread and likely will not prevent oxidation on the molten surface of the bulk solder. Once the solder sphere is driven back into the molten solder, both surfaces have become oxidized sufficiently such that the bulk solder may not wet the sphere. The result is a sphere that appears to be resting in the bulk solder, but is not metallurgically connected.
Conclusion
A reflow profile with a hotter and longer soak, coupled with a quick transition into the solder’s liquidus phase, proved beneficial in reducing ball-in-socket defects. This proves that many ball-in-socket defects that were once considered random have an assignable cause, probable mechanism, and corrective action. Only a profile with a ∆T minimized using phase transition of the solder can eliminate ball-in-socket defects. The profile shown in Figure 3 eliminated all defects caused by component tilting. This approach should eliminate such defects in situations where they occur repeatedly within a small geographic area on a component.
The ramp-to-spike profile will exacerbate these defects as it will drive higher ∆T across a component and induce component tilt. Any profile with significant ∆T across a large BGA through the phase transition of the solder may create ball-in-socket defects. These can be virtually eliminated by designing a profile with a long, hot soak and rapidly increasing the temperature above the solder’s liquidus temperature.
Brian Smith, market development manager, Kester, may be contacted at (847) 297-1600; e-mail: bsmith@kester.com.