Issue link: https://iconnect007.uberflip.com/i/1509873
14 PCB007 MAGAZINE I OCTOBER 2023 Shaughnessy: Some companies say that material variations make it tough to build high-rel boards. What do you think of that argument? No, materials are not the primary problem in 2023, and I have found that material suppliers have been really helpful with the data that they provide, for the most part. Sure, you can always be wildly off in material selection due to lack of due diligence, but I have found that the question with the materials suppliers is more like, "Is their data con- sistent with your baseline due to measurement methods?" e bigger question is, gener- ally, "Is your fab process consis- tent?" Can you model with their data or not due to MSA (measurement system anal- ysis) delta? Do you validate material properties when you qualify internally as an outcome of your process recipe? If your processes are con- sistent, then you may see some material varia- tion. But I have to say, I just built some pretty advanced factories, and material variation was quite low, even though I was also biased that it may not be. e problem is more with the fabs being under-controlled, as there are major vari- ables that most shops do not manage that are absolutely critical. I think that these are oen masked because the folks doing the failure anal- yses are not privy to the detailed bill of process used (and neither are many of the fabs due to under-control), and if the signal-to-noise ratio on the fab side is so poor due to low due dili- gence on recipe development or poor controls, the materials tend to stand out as "different." Johnson: What areas should have more focus for better reliability? One highly neglected topic is copper grain structure. is is something that no one hardly ever talks about or addresses, and it touches many areas. Reliability and signal integrity are highly influenced by intrinsic copper struc- all of their reliability equipment in-house and be testing it in-house as well. In the grand scheme of things, it doesn't cost that much when you're doing this. Outside organiza- tions (i.e., OEMs, consortia, agencies) might develop a test vehicle or program, but it will take a year—or at least months to execute. So, while they do help, the cycles are too long to be of any benefit to everybody, and if you fail the test, you can get DQed. Also, by the time they get some data, we're already looking at or using different materials. is is something that they could run in a week if they had everything in-house, and stay ahead of the tech curve. Andy Shaughnessy: I'm surprised that weak microvias are still a problem after four or five years of committees and groups trying to find the root cause. They've come up with workarounds such as avoiding stacking vias, but sometimes you have to stack vias. at's right. One of the "big finds" I have heard recently in microvia reliability is that some designers were only putting in a few microvias in a large area. Of course, then it blows apart because all the stress is on one or two microvias. It's just kind of silly that this is a research topic now. We have seen this relationship just by changing the coupon pitch for years, and if you run multiple cou- pons at variable pitch, you can get precise val- ues to build your own empirical model that is not just some theoretical interpolation of datasheets. is should be common sense by now for any designer or fab shop. Every- one knows, or should know, that you can't do that. e obstruction to technology advance- ment caused by compounding coopetition with scarce engineering resources is proba- bly largely to blame. Is your fab process consistent?