Issue link: https://iconnect007.uberflip.com/i/1221561
MARCH 2020 I PCB007 MAGAZINE 97 Dickson: Yes. Typically, when people talk about elongation with copper, they discuss stress cracks built in, and I'm old enough to remem- ber when there were additives that you had in your plating solutions that the elongation would get so high that it would crack from nor- mal thermal cycling. But with the maturity of the chemistries we've had for the last almost 15 years, especially once pulse got in, the high elongation deposits were pretty easy to capture. Johnson: Are you far enough into the pro- cess to give a general sense of reliability? Your qualitative statement was very encouraging, of course. I know you're not done yet, but are there some numbers that you can share? Dickson: We have done our own testing, and we've been able to do IST thousand cycles that had very good results. WUS has done some industry-level testing. We're currently doing HDP User Group's, but that one is very ag- gressive. For example, it has a 2.7-millimeter deep blind via. A 0.1-millimeter HDI blind via is deep; this is 2.7 millimeters deep, and it's on a 0.5-millimeter pitch, and you couldn't even run HDI at that pitch. Johnson: You're in a completely different ball- park with that kind of testing. Dickson: We went to furthest possible limits of where it was—not necessarily manufactur- ing, but what we even understood about doing the blind innerconnect. We can do a through- hole VeCS, with back route to remove the stub, much thicker than that, but those are some of the deepest blind applications we have done. We're working through those tests now. We have had some limited success with it. The main issues that we have seen are similar to through-hole. We still have the issues where if you have metal features and you're routing a slot, the slot to the metal feature or drill to metal features is still critical. Those are there because the materials used included glass re- inforcement, but the trace to the trace and the trace to the reference across from it has had pretty positive results so far. However, we don't have enough detail yet. We're producing multiple design test vehicles now, building in a reliability structure, so we'll be able to publish more details by July with some sophisticated information. It's enough to warrant people to look at doing their own reli- ability testing. Johnson: VeCS allows for some routing den- sities that are in line with what the industry is starting to demand. It allows for packaging that includes some of the larger, more complex, problematic packages. It also seems to address some of the via failure issues seen in HDI. Given that the OEMs are pushing on us, as an industry, to deliver two orders of magnitude more reliable than we can currently achieve, it starts to sound like VeCS is a technique that has found its time. Dickson: That's why WUS dove in and pushed this. The difficulty with doing any type of new technology is there's so little history on reli- ability. Alternative mature technologies, even if the performance is bad or not perfect, if it's known, that has a large margin of security. Peo- ple have said, "I need 10,000 samples at 1,000 hours of my special design before I'd even be interested." Okay, that's good. Two years from now, you'll have your point. The difference with this technology than other technologies is that we have many customers telling us, "We don't have an option. We need this because we must have it." And I tell them, "Look, it takes nine months to learn how to design." That's your paradigm shift. You can't say to a designer, "Design me a VeCS board." Joan Tourné is brilliant. He's a designer, an engi- neer, and has two brains. Even with him, you need to be able to do this, and the customers that we have now are learning the things that we knew. We're building out a DFM rule base that is reasonable; if you follow that DFM rule base and not try to do something outrageous, we think the reliability numbers are going to be superior to what you would expect from performing through-hole and HDI. VeCS is not a replacement for those two; it's in addition to those two. There are very few