PCB manufacture

I had some PCBs made with 22 boards per panel (small boards). I wanted a single panel for prototypes, but it turned out I could get 6 panels for only $500 more. This would get me through the first round of production if I didn't need any changes. The day the maker was supposed to ship them, I got an email saying they had a poor yield and only got 116 good boards out of 7 panels vs the 132 I ordered.

So now I have a concern about the quality of the boards. I was told the failed boards had problems with the plating in the 10 mil holes I used for vias. They claim that the "good" boards have no quality issues. My concern is that the vias may be marginal and open once they are temperature cycled a bit.

I am not sure how to handle this with the vendor. I guess I could just flat out tell them that I am concerned with investing some $18,000 building up over 100 boards only to have my customer see failures in the field. I don't feel like they have done 100% on this since they did not provide the quantity I requested.

They have offered to rebuild the entire set of panels if I am not happy with the result, but that would be a week delay. I also don't know if they are willing to let me use one of the current panels and remake the rest.

Anyone have experience with 10 mil via holes and reliability issues? If you have a problem with an order like this, is it reasonable to use one panel and reject the rest?

As you have figured-out, pushing the process limit could be expensive. We would not go below 16 mil holes, even if we need bigger boards. Are you sure you can't use bigger holes?

Sure, I can use bigger holes. But I can't use bigger pads. It was hard enough to make the layout work with 10 mil holes and 24 mil pads. This board is very small and defined by the customer's requirements. It is a daughter card on an existing product so I have no flexibility in size except for extending one end which I have already done as much as possible. Even that does not do a lot for the density issue since the board is over 5 times longer than it is wide. Extending the length more just means I have longer lines to try to squeeze through the same bottlenecks around the chips and connectors.

I guess I am confused about "process limits". I have seen vendors that claim to have capability for hole sizes down to 6 mil. They just say that they can't guarantee that the hole will not be plated closed which is fine with me. So I don't see where 10 mil holes should be a real issue. Did I pick a poor PCB maker? Am I just not grasping the realities of PCB manufacturing?

Did the vendor you chose give this claim? If not then there's your problem, choose a different vendor.

--
DaveN

Sorry, I meant did your vendor make the claim on 10mil?

Additionally did you order 132 boards or the yield acheived from the panels?

The realities with any process is that as you approach its limits, you will get higher fallout; edges of the bell curve and all that.

If your supplier is telling you that parts have failed then it tends to suggest that they have a reasonable quality system in place. They know the process and its limits and are therefore testing for those failures. It's an old but true saying that you can't test in quality, you can only design for it.

--
DaveN

On Thu, 10 Apr 2008 08:42:33 -0700 (PDT), in sci.electronics.design, rickman bloviated:

I am looking at a fab right now that is a mix of 10 and 7 mil drilled vias. Board has 8 BGA and is 2x7" so your process is not in any way pushing the envelope. If you are getting bad plating in 10mil vias then the most likely cause is insufficient agitation during the plating and air bubbles were trapped in the vias. You will need to section (slice & dice) the failing fabs to confirm whether the plating is actually thick enough across the whole of the fab. If it is air bubbles you are probably okay as that usually is batch related. If it is a function of the vendor skimping on the plating, you are screwed.

That being said, I have no empirical data but there does seem to have been an increase in the failure rate of commercial fabs in the last year or two. Usually in the plating of the vias. This is across several vendors US & Asian.

At least they have a good enough quality control that they can spot this problem. You can have them do a cross section of a good board to see if the vias look healthy.

Plating systems have to be top-notch to handle small via sizes. If the board house hasn't kept up on updating their equipment, they will have problems with small features.

I routinely use 8mil drill with 15mil pads. The pad size is marginal, but the four board houses I have used never had problems with that via size for the past 5 to 8 years.

There is one case back in the early 90s where the plated through holes would disconnect when moderate current (apx 1 amp) passed thru the plated hole after soldering. Before soldering, the PTH handled 10A, after soldering it would blow around 1A. Most of the bypass caps disconnected at power up.

--
Mark

I had problems with a vendor once, it was "weak" vias. They were 20mils though, done on early ROHS process. And yes, baking boards twice for smd and through-hole did create new broken vias, some immediately and some later. If I knew about this, I'd toss entire batch right away. Caused quite a big of grief. I ask my vendor now to leave them a bit longer in the metalization "soup" and he had to compensate for part of the damage (tarnished name can't be restored so easily).

You can try baking boards without components and than passing them through tester again, make sure testing current is set to high"ish" level.

Mark

Hi Mark, Have you done any PTH vias (0.012"D) that are in SMT solder pads that are filled either with a non conductive material before the solder is added? I have done this on a couple of jobs recently and it simplifies layout and the vias used for thermal flow, let's say six under a DD Pac work great. What say you? Harry

messagenews: snipped-for-privacy@p25g2000hsf.googlegroups.com...

Thanks for your suggestion. I thought of that, but I don't have a way to test the bare boards. Turns out I have four more weeks to order the panels, so I have some time to evaluate the boards in the prototypes.

I'm waiting to hear back from the board vendor. But I expect I will have to bite the bullet on this one and reorder the production panels from a better supplier. The whole reason I bought so many was that they were not much more than just buying prototype quantities. So I knew it was a bit of a risk. Not a big deal now that I have more time to deal with it. I was mainly worried about the schedule.

PCB vendor surely has a machine to check for continuity, even the cheap vendors do. Ask them.

Mark

We have used filled vias for a different reason, to get more space in a tight layout. Really strange to see pads with no visible drill hole! I normally avoid filled vias due to extra cost. When I put thermal vias under a Dpak, I make sure there's enough solder paste to fill the via. The pad, or Cu pour, on the opposite side of the board is mostly covered with soldermask, but the via drill hole isn't tented so the via can burp. I usually use around 50 to 70% pad coverage for solderpaste, depending on the part and/or if I use thermal vias in the body pad.

Harry, I've been having fun making switching supplies recently. Made a

50W forward converter and, something we chatted about briefly a few years ago, a SEPIC switcher to charge up a capacitor bank with constant current for a pulse amplifier. Both run at 600kHz. The new controllers sure make life easy. Both these supplies used the technique above for the power components thermal vias.

--
Mark

Thanks Mark, not many of us left. Cheers, Harry

messagenews: snipped-for-privacy@d1g2000hsg.googlegroups.com...

You are suggesting that I temperature cycle the boards a few times and send the blank boards back to the vendor for retest? I guess that is an idea. So far they have been a bit difficult to communicate with. I shouldn't say "difficult", just very slow getting answers from them. I ask a question and it seems to require a lot of research to get an answer. Then I hear back after a day or so. It is like they don't seem to feel like they screwed up and are not making any real effort to help me deal with the resulting repercussions.

I'll give them until tomorrow (Friday) to get back to me with something substantial. If we don't have some sort of dialog going that is likely to result in an improvement in my satisfaction, I will just cut my losses and move on. I need to keep my customer happy and I can't do that with an unresponsive vendor.

messagenews: snipped-for-privacy@d1g2000hsg.googlegroups.com...

Just be sure that you tell them that. There are a lot of native English speaking "Front ends" for Chinese fabs that are several time zones away. It could explain the lag and the poor information transmission quality.

messagenews: snipped-for-privacy@d1g2000hsg.googlegroups.com...

These guys are in the US. I did finally get someone who could at least try to get a little info, but I didn't get a lot. I was told they checked the chemistry and didn't find anything out of whack. He offered to do a destructive test, but never gave me a RMA. Finally someone called as a general follow up and I told her what had happened. It was a bit surprising that she didn't know ahead of time that there was a problem. She also didn't have access to any information on the problem or the contacts I had already had about this. She asked me to send her the last email about what they had found and promised to get back to me on Monday.

At this point I am ready to release the test fixture to be made and I am looking for someplace else to do it. It is 8" x 10", 4 layers and I only need a small number, say 4. Any suggestions?

[Snipped]

I have had very goods results from

They are based in china, but I have had no problems with their product. The quality is excellent, the price is very good and you can get a direct quote on their web site. I normally pay about US$60 to get the boards couriered to South Africa, which takes about 36 hours. They have Fedex and DHL as an option. The only problem I have had is that the status on their web site are sometimes not updated. To get the latest status it is best to email them.

Regards Anton Erasmus