Issue link: https://iconnect007.uberflip.com/i/1276973
AUGUST 2020 I DESIGN007 MAGAZINE 29 or three-ounce copper, and you're not calculat- ing exactly how much break you're going to lose off that dielectric value, your impedances could be way off. Shaughnessy: Happy, do you have anything to add? Happy Holden: I spent my first 30 years not allowed to use IPC specs. Our many packag- ing specs were far more detailed than IPC, and they were written around our fabrication. Everything was being built in Asia—not in China, but in Asia, where the engineers were not native English speakers. I cleaned up a lot of their stuff with pictures and diagrams because using words constantly got us into trouble in terms of different translations. Each engineer might translate it differently. Thompson: Exactly. And that brings up a really good point about the dielectric constant. I fre- quently get phone calls, where people will say, "What's the Dk of 370HR?" And I'll say, "What preg ply? What dielectric? What speed?" There are a number of different questions I'm going to have to ask to be able to get to where they want. Holden: Because all these things are built out of different cores with different weaves of glass, which all have a different glass-to-resin ratio, and that's not even counting what frequency you're operating in. Thompson: It makes a big difference between 500 megahertz and 120 GHz. With 120 GHz, that Dk goes way down. Holden: I always start with material because that can get you in trouble. And if an RF ampli- fier is not working on a cellular tower in Loui- siana in the middle of the summer, they blame the fabricator. Did you happen to understand that this material that you selected, when it's humid and hot, the loss tangent goes com- pletely away, and your signals go away too? It has nothing to do with the board. It's the fact that you didn't understand enough about Shaughnessy: Are the fab notes incorrect, or are they primarily just incomplete? What is typically the biggest problem? Thompson: Gosh, we see everything. We have over 5,000 customers. We see inaccurate notes and notes that are not relevant to the job and therefore require a phone call. "It looks like you're calling out blind vias on this job, and there are no second drill files associated with a layer 1 to 2." If they don't exist, don't call them out. Simple as that. Shaughnessy: It just amazes me because many of these designers have been doing this for 40 years. But a lot of times, fabricators just accept the data because you want the job, and you fix the problems. Thompson: Even the new guard, the new kids that are coming into PCB design, just don't know what they don't know. I try to assist them in any way I can. Shaughnessy: But it sounds like there are often no consequences to submitting bad fab notes. Thompson: Other than the hit on time. Because either a salesperson or I have to call the cus- tomer and clarify those notes. That takes time, and we sell time. Shaughnessy: You said you have your regular customers, and that you have them trained. What are some of the things that a trained cus- tomer will provide you? Thompson: For example, nesting, which is how much prepreg is going to get sucked up based on the copper weight that is an interface on an inner layer. If it's half-ounce, it's going to have less of an impact. If it's two-ounce, three-ounce, or four-ounce copper, it's going to be a huge impact. It also depends on the configuration of the layer, whether it's a full signal, a split plane, or a plane. If it's a plane, it's going to have the least impact. If it's going to be a split plane, it's going to have more impact. If it's a signal, if it's a pure signal, two