Issue link: https://iconnect007.uberflip.com/i/884055
30 The PCB Magazine • October 2017 eating into that fringing capac- itance. That's something peo- ple don't generally think about and I wonder how many peo- ple have been able to see a char- acterization. It's enough to get your attention." Yogen said, "Companies have to start using the landless via, like what we produce. We invented our process in 1995 and have been using it for the last 20+ years. With the positive- acting electrophoretic photore - sist, whatever pad size you put that's what comes out, so if you have 6 mil holes then make the pads 7 mils; you get extremely reliable landless vias, and now with direct im - aging technology you can get even closer." At this point, Happy Holden referenced an article on landless vias that he wrote last year. "There are four different ways of doing it, one of which is the positive electrophoretic re- sist. But there are three other ways: the direct imaging machines, the Hewlett-Packard meth- od that they got from the Japanese (which doesn't require any change in imaging regis- tration), and then there's the Russian method. Anybody can use the Hewlett-Packard and Rus- sian methods; they are free and no change of equipment or process is required—just a change in artwork. I've made more landless vias than everyone in the world put together, but the ar- ticle mentions all four methods." When Steinberger asked why more compa- nies don't make landless vias, Holden's response referred to IPC specs, which require a minimum pad size to surround any via or hole. Holden: "Landless vias were never adopt- ed by the IPC, though we (HP) have data that shows landless vias are 10-times more reliable than vias with lands. The landless via and its reliability blows a hole in Class 1, 2, and 3 [of the IPC specs]. HP did extensive testing: on dif- ferent hole diameters, different thickness of boards and different sizes of vias, and we found out that our Japanese partners apparently knew a lot more than we did, and they were less sub- ject to influence from the IPC. You needed a pad EXPERTS DISCUSSION: SIGNAL INTEGRITY AND IMPEDANCE CONTROL 50 years ago because we used to crimp actual leaded com- ponents, not to the whole bar- rel but to the pad, and because we've been doing it for 60 years, the industry thinks we have to do it forever, even though it's detrimental to both reliabili- ty, density, and signal integri- ty. Hewlett-Packard used them exclusively once we discovered the secret, and nobody bothered to look at our boards closely enough to realize there weren't any lands on those vias. They're covered with solder mask, so you don't really see it." Yogen added, "We have data on 6-layer landless vias that pass 1,000 IST cycles; the IST person says it's much better, because there's no pad to anchor it and then crack. Nor- mally the pad will only anchor to the laminate and then crack at the knee of the hole. But if you've got no pads, then you've got no corners to crack." Steinberger mentioned that he intends to contact a few colleagues, who he said, "are ac- tively designing a system at 28 gigabits and starting to plan for 56 gigabits" and pick their brains about landless vias. Holden concluded this portion of the con- versation by reminding us of days gone by: "Hewlett-Packard, in my day, was operating at 150 GHz for microwave test equipment, and the test equipment had to be nearly 8 to 10 times better than what we were measuring; even back in the '90s and the 2000s, we were way over 150 GHz. So, making very sensitive test equipment, you do a lot of testing, and you find out a lot of myths aren't correct, and in fact they're det- rimental and not being controlled by military standards or things like that." This conversation next ventured into design territory, which we will excerpt in the October issue of The PCB Design Magazine. PCB Mike Steinberger