Issue link: https://iconnect007.uberflip.com/i/1210212
16 DESIGN007 MAGAZINE I FEBRUARY 2020 Matties: That was a captive facility, though. You have resources at different levels in a cap- tive facility than an independent would. Holden: All mistakes come back to the compa- ny; it's not transferred across legal entities. We could help a new designer out of college by pairing them with somebody that's been de- signing RF for 35 years and make a great pro- duction engineer. Ritchey: Sure, but that won't work anymore. Why I say that is because we have four things that matter in a stackup, of which only one used to mat- ter. The nature of the glass weave now matters with these high-speed differential pairs, the plane capacitance, and the crosstalk impedance. You can't expect the fabricator to do anything except the imped- ance part of that; I would not expect that personally. That's not what a fabricator's skills are. About the history of imped- ance, I was around Silicon Val- ley when it started to matter. First of all, my career was in ECL from day one, and I always worried about im- pedance. When the engineers started to worry about that, they didn't know how to deal with that. They would tell the fabricator, "You have to figure out how to make this design 50 ohms," which is probably what happened to you, Happy. They scrambled around to find a way to get 50 ohms. I probably got a couple of hun- dred calls from around the world from fab shops, asking, "How do I do this?" They tried to do it, and hats off if they did. It was not their responsibility, but they tried to do it. The accuracy with which I see it done is all over the map. Probably half the boards that come in that are supposed to be 50 ohms aren't even close. Holden: That's one of the reasons we spent a lot of money characterizing all of our standard Holden: On a power supply design network, is a power supply impedance and the need for lower impedance part of the difficulty with the noise? Ritchey: Yes. The problem is that the places where the noise is coming from are above the frequency where capacitors are useful. Holden: That's when you get to distributed ca- pacitance with materials and printed circuit fab- rication. Of course, those materials have to be designed in. A fabricator can't switch materials because they realize that a 5-mil dielectric between power and ground is not going to be much- distributed capacitance. Ritchey: Exactly. You're sug- gesting that fabricators are per- mitted to change the stack- up. If a fabricator did that to me, they would never have my business again. Holden: They could always recommend or point it out to someone. Ritchey: No one does my stack- ups but me. Matties: How would they know? Ritchey: They don't know. They can't do it for you. If you look back a couple of decades, you're only worried about impedance and not much else. Holden: One of the things I did as an engineer- ing manager in fabrication was to hire the best RF engineers we could find to help us with problems. Ritchey: That's not where the engineering deci- sions should be made. Holden: But it helped us discover that they're going to have a problem with prototypes.