Thermal Cycle Reliability of Vapor Phase BGA Joints
In this paper the authors performed an evaluation to assess reliability of vapor phase processed BGA solder joints, with and without vacuum applied.
DOWNLOAD
Authored By:
Ward Gatza
Agilent Technologies, Inc.
Colorado Springs, CO USA
Tom Evans
Thomas C. Evans Consulting
Colorado Springs, CO USA
Summary
Prior to committing production boards to vapor phase soldering, we performed an evaluation to assess reliability and evaluate the vacuum soldering option. The reliability of vapor phase processed BGA solder joints, with and without vacuum applied, was evaluated by means of a test vehicle circuit board assembly.
The test vehicle was designed with daisy chain nets through multiple solder joints. These were designed with all of the solder balls in a chain having a similar distance from neutral point, so this factor could be part of the reliability assessment. The boards were temperature cycled for 8250 cycles of -5 to 95°C, by which time all of the outermost daisy chains had failed. The number of cycles-to-failure was analyzed using Weibull plots and characteristic life was calculated.
Conclusions
Referring to Figure 4, the sample sizes were not large enough to show that the characteristic life is statistically different between the different reflow methods. The reason is that the 90% confidence bounds surrounding Weibull plot lines for the vacuum process and no-vacuum process results overlap, and therefore don't show statistically significant differences. What can be stated is that the differences are consistent for each reflow method across four outermost daisy chain nets as shown in Figure 5, and have the appearance of indicating realistic values expected for larger sample sizes.
The reason for the vacuum soldered board showing shorter apparent characteristic life than the other samples isn't fully understood yet, but it is notsuspected to be caused by the slightly longer time-above-liquidus in the profile. This difference appears negligible and the values are well within recommended limits. Further destructive analyses of the solder joints may shed some light on this.
There is an instance of a statistically significant difference in two populations within this dataset. Vapor phase no-vacuum daisy chain net O3 has a characteristic life of 5303 cycles and a DNP of 25.5mm. For the same boards, net O4 has a characteristic life of 4690 cycles and a DNP of 26.9mm. At a 90% confidence level O3 sample data represents a population with characteristic life that is not part of the population represented by O4 sample data. The longer characteristic life of O3 is therefore attributed to a shorter DNP.
Initially Published in the IPC Proceedings
|
