Issue link: https://iconnect007.uberflip.com/i/1348195
14 DESIGN007 MAGAZINE I MARCH 2021 costs more, and works poorer than if you'd just waited and started two months later. Matties: And you'll probably be doing multiple re-spins on it as well. Kolar: Exactly. ere are a lot of different sources of poor data: ere's just lack of knowledge, schedule pres- sure, sloppiness, and there is misinformation. ere have been a number of times that we have engineers send us reference designs and say, "Just do what the reference design says." en we get to argue that the reference designs have nothing to do with manufacturability, and you repeat that cycle over and over. From our perspective, when you're getting input data, ideally, you're getting enough vari- ants of it that you have something to validate against. Maybe I get a board file that has some of the parts loaded. en, maybe I also have a STEP file or mechanical data that I can sanity- check against that, or the schematic, which is also going to have some callouts. Matties: You keyed in on some of the problems to get good data. One you mentioned was tim- ing. People want this start date. What are some of the other problems that would lead us to the situation where the data's not there? Kolar: Product intellectual property rights can be one. It may be that it's a proprietary datasheet that the engineer has access to, but you, as the designer, don't necessarily have, or they're not sure what you have access to. It can also be confidential projects and you're only getting a tiny subset of the information, so you're not understanding the big picture of how it might be used, and that's leading you to make false assumptions. We don't know what it's being used for, and we don't ask. Matties: What other obsta- cles prevent the data coming to you? Kolar: I think it's this constant mix: Some people are just sloppy, some just don't have time or the knowledge, and some people are just pulled in so many directions that they're just throwing some- thing over the wall to check it off their list. I see a lot of errors get introduced when inheriting a previ- ous design and leaving in the leover "cru." We talked a lot about this with fab notes and assembly notes in a previous issue: "I'm going to piece together these parts of the schematic and move forward, or I'm going to grab this PCB design and start with this." Well, I might set up my new layers, but I don't remove all the old ones that I don't care about. I just leave a bit of cru, and that cru piles up, and you go to the next person that borrows it, and the next person borrows it, and you just get all this cru over time. Matties: You covered a lot of problems. Mark, do you have any additional problems that you may see? Mark Thompson: Well, a couple of different things. Jen touched on stackups—things that change, controlled impedances that change from design to design, from Rev. A to Rev. B to Rev. C. Frequently, we'll get a stackup from customers dictating the dielectrics, cop- per weights, effective dielectric constants on subsections. It doesn't do us any good to get controlled impedance changes and a previ- ous stackup because all of that stuff now has to change. You'd have to redo a new stackup. It's important to query the fabricator ahead Jen Kolar