Issue link: https://iconnect007.uberflip.com/i/1307491
NOVEMBER 2020 I DESIGN007 MAGAZINE 15 pay attention to the details. A lot of times, a company will qualify with one or two spe- cific suppliers. Then, once a prototype moves into produc- tion, they want to start mini- mizing the expense, so they start looking to shop around to see where they can get it cheaper. And then that just blows everything out of the water at that point, because who knows what you're get- ting with regard to quality? Holden: But if you want to read really good explanations about vias, go back to some of the publica- tions by Intel and other aerospace companies around 2000, when they were starting to qual- ify microvia vendors and using the IPC bench- mark PCQR 2 , and you'll see there's a report on the lots. Every page is loaded with data because the qualification is that you have to be able to fabricate panels three separate times at least two weeks apart. I always joke because we used to disqualify Chinese companies that thought they could fool us by building 25 panels, picking the best 15, and then shipping five of them two weeks apart. They didn't understand that the testing machines were so sophisticated that they expected to see three peaks when they'd analyze each work order. If you have one lot, you're only going to see one peak. They were disqualified for an inter year because they cheated, and they'd ask, "How did they do that?" I had to explain to them how it's tested and built and that there's a reason why they want you to build three separate lots: to find out what your lot-to-lot variance is. Building one lot, you might think you can fool them, but not these testing machines. The PCQR 2 is just an outstanding benchmark- ing and qualification tool, and it doesn't cost that much because the test is all automated for it. But most people have never heard of the PCQR 2 except the aerospace and military guys. Shaughnessy: Has your com- pany had any issues with vias? Chavez: We are a large corpo- ration. We have technical fel- lows, subject-matter experts (SMEs), and our appropriate departments in place to vali- date this type of information. But are the lower-level busi- ness units following that? Are they taking those lessons learned and applying them in their design? And what you would find is, in some cases and companies, they aren't— even though they've already benchmarked and said, "These are our lessons learned." You'll find that at the low level, in the weeds, these one- off divisions of small groups of engineers are just pumping stuff out. That's where you start getting anomalies because they're not following the lessons learned or industry best practices. Holden: You need a rigorous qualification pro- cess, which is not going to be that expensive but is going to help you in the long run. You also need a miniaturized coupon on that quali- fication so that you can put a coupon on every one of your SMT runners so that you have sample vias or a net on every single panel that you assemble; then, you can hold those for three or four months, or if something happens, you can go back and test them. You might also require a report in every shipment that tests those coupons to the fun- damental qualification, so you have the confi- dence that, "You passed our qualification, but we want to know that your process is still in control now, six months later." The one thing I've learned in 50 years of making PCBs is that no matter how good you are, the PCB process can go south on you before you know it just because you looked at it cross-eyed. It's so convoluted and complex, and it has so many complicated steps. We're lucky that we can make multilayers at all, but we do. You have to be on top of this thing all the time, Happy Holden