We have rejects in 2-3 per 15,000 PCBs that look like Head in Pillow. Should we tinker with the reflow process, or is this reject rate acceptable? Board Talk
Board Talk is presented by Phil Zarrow and Jim Hall of ITM Consulting.
Process Troubleshooting, Failure Analysis, Process Audits, Process Set-up CEM Selection/Qualification, SMT Training/Seminars, Legal Disputes
Phil Zarrow
With over 35 years experience in PCB assembly, Phil is one of the leading experts in SMT process failure analysis.
He has vast experience in SMT equipment, materials and processes.
Jim Hall
A Lean Six-Sigma Master Blackbelt, Jim has a wealth of knowledge in soldering, thermal technology, equipment and process basics.
He is a pioneer in the science of reflow.
Transcript
Phil
And welcome to Board Talk with Jim Hall and Phil Zarrow, the Assembly Brothers, here to talk about assembly processes - equipment, materials, and other things.
And speaking of other things, today I see we have a solder defect question.
Question is, "We have rejects that look like Head in Pillow. We see rejects in 2-3 per 15,000 PCBs. Should we tinker with the reflow process, or would you consider this reject rate acceptable?"
Jim
Head and Pillow is a real issue. First up I want to say before you think about tinkering with anything, are you using an up-to-date solder paste, because solder paste manufacturers have done significant work to try to minimize this with newer formulations.
As to what is an acceptable rate, it's very hard to say with the numbers quoted, 2 or 3 rejects per 15,000 printed circuit boards.
So, that would be about 1 in 5,000 circuit boards or about how many parts per 5,000. That's 3,000 parts per million. That's not too bad.
That's a three sigma process.
But that's not really the way we would define defects, because if we're talking head and pillows occurring on BGA parts, all of which have multiple balls, any one of which could have the defect.
So, the proper way to do this would be to define it not per number of PCBs, but for parts per million opportunities, so you would have to count all of the BGA balls on those 15,000 PCBs.
Since that's probably greater than 100, that would drop that down to about 30 parts per million, and that is certainly a very low defect level to try to go optimize just in general optimization strategies.
Phil
So, congratulations, A.B. The answer is probably not to tinker with the reflow process.
Movement towards continuous improvement is good and trying to fix the defect rate would be good, but it may be things beyond your immediate control.
Jim
If you really want to address a defect such as head and pillow, which is going to become more common, because there's going to be more and more BGA parts, thinner BGA parts, which are going to warp more - seriously look at getting an up-to-date solder paste.
That's true for all of the newer types of defects, such as graping, and so forth. Solder paste technology is evolving daily or at least monthly, and you may be making your life a lot harder than you have to by not evaluating a modern, up-to-date paste that could significantly reduce your defects such as head and pillow right out of the box.
Phil
Well, I think we answered that question hopefully to AB's satisfaction.
Comments
Depending on the location of the HiP balls the reflow profile is effeminately suspect. If the HiP sites are central to the Ball Grid array then yes the reflow profile needs to be tweaked b/c sufficient heat is not reaching the center of the BGA device. If the HiP sites are on the perimeter , near other components then yes; the reflow profile needs to be tweaked.
Ike Sedberry, ISEDS
All of the above are important, but a frequently overlooked cause is the co-planarity of the balls for the BGA. For example, a BGA with a 0.2mm (7.87mil)co-planarity is going to have issues with a 5 mil stencil. You will need to look at stepping up your stencil if your co-planarity is greater than the thickness of your stencil. Those balls need to be in/on the paste!
Alan Woodford, NeoTech
Before messing with reflow profiles it would be worthwhile to examine solder volume using a 3D AOI. How old is the paste? How long has it been sitting on the stencil? Are you following the recommendations from the paste supplier? Have those times been validated? In a past experience if the line went down for any length of time the paste dried out and HIP and other solder defects went up. HIP can be a hard one to iron out. How much bow and twist does the PCB display? What does it do when heated to 250? Good luck.
Bradley Fern, Entrust Datacard
HiP defect is actually very hard to capture. Only those extreme ones that cause immediate open can be detected in the inhouse ICT or functional test. BGA scope could only check the quality of exterior few rows of joints. X-ray laminography is mostly capable of HiP detection. If you don't have 100% X-ray laminography inspection, then it's recommended that you randomly sample some of your products to run a thermal cycling test (refer to IPC9701 for conditions) and post ICT/functional test; if no more failure is captured, then you may conclude to say the batch is basically safe to use. HiP is always a sensitive and critical failure mode, so treat it cautiously all the time.
Colin Wang, Eolane Co., Ltd
In the question posed, the manufacturer is rejecting 2-3 per 15,000 PCBs. You state in the discussion that this equates to 1:5000 defects or a defect rate of approx 3000 ppm, but when taking into account an example of 100 BGAs per PCB, this drops the defect rate to 30 ppm. This is where I am a little confused. If they are rejecting the boards with head-in-pillow in their entirety, then it could be any number of the solder balls that have the defect, unless they are cross sectioning or X-Ray testing for every defect and recording them, correct? If there was only a single occurrence of head-in-pillow per board that was rejected, then the defect rate would be 30 ppm, right? My apologies if these are basic questions! Maybe I am just not looking at it from the right perspective.
Jordan Koologe, MacDermid Enthone Electronics Solutions
From my point of view, the warpage of BGA during soldering is the most important source of Head in Pillow! That means the component supplier is responsible for such a problem.
The question is this rate acceptable or not should also to be discussed under the view: Do we find all the Head in Pillow Defects? I think, not at all. A little bit more printed solder paste can be help to minimize this problem, but this is not the root cause.
Heinz Wohlrabe, TU Dresden, Germany
One consideration not addressed is the detection of the defect. The HiP defect is very difficult to detect since you need to have an open or significant change in joint geometry for x-ray to detect. The marginal conditions that escape inspection can lead to latent field failures which in some product markets can be catastrophic.
