A customer called the Help Line with an issue regarding an area array device that is similar in scale to a chip scale package...
A customer called the Help Line with an issue regarding an area array device that is similar in scale to a chip scale package. They were experiencing low-profile, fine pitch, ball grid arrays (LFBGA) failures and wanted to determine the root cause. The customer manufactured a PCB (printed circuit board) that employed the use of two LFBGAs. One of the LFBGAs exhibited a high solder ball void rate, while the other was relatively free of voids. The PCB was processed with lead-free solder. Their investigation into the contract manufacturer’s processes did not reveal any issues that caused the anomaly. The customer could only determine that the solder voids appeared at the solder ball to PCB pad interface and not the solder ball to IC pad interface.
The customer supplied the suspect PCBs to ACI for analysis as well as bare boards and pre-reflowed components. They specifically requested assistance in determining the presence or absence of BGA ball excessive voiding, black pad, ball drop, gold embrittlement, cold solder joint or other soldering defects. Additionally, the customer requested investigation into any noted issues that would prevent or significantly degrade either the solderability of the BGA or the PCB pads, or the performance of the solder paste.
The customer stated that the boards were subjected to overnight burn-in at ambient temperature, and were stacked up, powered on, and monitored for electrical failure. One LFBGA began experiencing intermittent failures at temperatures ranging from 30˚C to 70˚C after 2 hours of burn-in.
The materials supplied by the customer were:
• A populated board having “suspect” 144 pin,
15 mil pad LFBGA
• A new LFBGA balled, never soldered
• Two (2) bare boards
• Solder reflow process diagram
The SMT board components were soldered in a reflow oven, followed by wave soldering to join the through-hole parts. When the reflow process was changed to accommodate lead-free solder, the board failure rate increased.
The following test methods were utilized:
• XRF - Elemental analysis was conducted on the new solder balls and on the pads
• Micro-sectioning - The failed LFBGA was micro-sectioned through multiple rows
• Scanning Electron Microscopy/Energy Dispersive Spectroscopy - The solder ball joints were examined first with an optical microscope, and then with SEM and EDS.
XRF analysis results indicated that the solder contained 93% Sn, 3.5% Ag and 3.5% Cu. Although the copper content was high, the results did confirm that this was a Sn-Ag-Cu alloy and that no lead contamination was present. Gold and tin thickness measurements for the PCB pads showed that the gold averaged 0.096 μm and the underlying nickel measured 4.7 μm. A high phosphorous level was not detected during XRF analysis of the bare board, ruling out the possibility of black pad. There was no evidence of gold embrittlement, cold solder joints, or excessive voiding.
X-ray inspection of the suspect LFBGA package showed that two die were present inside the package. No shorts, obvious opens, broken wire bonds, or significant voiding of the solder balls was observed. The solder balls were round, centered on the pad, and exhibited less than 10% voiding. Discussions with the customer and observations of images eliminated the possibility of excessive ball collapse.
Endoscopic inspection of the mounted BGA showed that the solder balls were round and shiny and the solder fillets were well formed. The solder appeared to have been reflowed properly and there was no evidence of cold solder joints. Examination of the reflow profile and comparison to the lead-free water soluble solder paste R520A showed that the profile was proper.
Micro-sectioning revealed solder balls that displayed a lack of wetting and incomplete solder joints. A corner ball displayed solder de-wetting and the separation was apparent and prominent through more than 50% of the connection area. Figure 3-1 shows that the solder has de-wetted from the pad and a nickel-tin intermetallic remains in some areas. Lead (Pb) was not detected or visible in the solder, and a typical tin nickel copper intermetallic that has a diffuse structure (non-layer) is present as expected. The adjacent ball also showed a separation through more than one third of the connection area. The opposite corner ball also showed the same non-wetting behavior and was present all the way through the solder joint. In another instance, the separation extended through the entire interface on that cross-sectional plane. This can cause intermittent electrical behavior during temperature excursion.
Discussion and Conclusion
The analysis resulted in the following observations:
• External examination of the boards showed that the boards did not have workmanship issues or display evidence of poor assembly practices.
• Internally, the BGA solder joints displayed incomplete solder joints, large separations, and poor wetting.
• The solder joints also displayed a normal lead-free solder joint structure. There was no evidence of Pb-contamination. Some small shrinkage cracks were observed. These cracks are acceptable but may be improved by controlling the cooling rate of the boards after soldering.
• Most importantly, intermetallics of proper thickness were formed during the reflow process. In most of the samples, the intermetallic was not difficult to resolve. Those detected were of proper structure, diffuse, and of expected thickness.
• There were several observations of a poor wetting angle and lack of solder joint connection. Except for the separation, the rest of the solder ball structure had good homogeneity, and lacked voiding.
Because the intermetallic was formed properly, it can be concluded that the solder was in contact for some amount of time, the intermetallic formed, and the solder then separated at the pad. Consequently, it is safe to conclude that the failure was caused by severe de-wetting, and that the de-wetting occurred either at some time during the reflow cycle after the intermetallic form prior to solidification, or during the subsequent wave soldering cycle.
There are also several other factors that may have aided the de-wetting. The separated solder joints occurred primarily at the corner BGA balls and some edge balls, which implied that some mechanical contribution from either expansion differences were present. Also, excessive heat input and longer dwell times may also cause severe de-wetting.1
Several influential factors should be considered to improve wetting behavior. Decreasing the reflow soak time with a slower cooling rate might improve the overall solder reflow behavior. The wave solder profile should also be examined, as solder separations can occur if the topside preheat is too high.
The customer was very satisfied at the level of analysis that went into drawing the conclusions and felt confident in making adjustments to their profiles.
From follow up conversations, the customer has stated that they cut the failure rate by 65%. They gave the process window a slightly hotter spike, shortened the soak time, but raised the soak temperature a bit and allowed for slower cooling. They were also in the process of testing out smaller PCB pads, and hope to see the remaining failure rate drop further.
1Reflow Soldering Processes and Troubleshooting SMT, BGA, CSP and Flip Chip Technologies. N. C. Lee, Elsevier, 2001 pg 6-110.
|
