Issue link: https://iconnect007.uberflip.com/i/1511130
NOVEMBER 2023 I DESIGN007 MAGAZINE 13 ing with, and the netlist indicates that this will require HDI. Instead of using 127 vias, you should have used 5,000 microvias, because that would have reduced the layer count and every- thing else. But now, he's all the way through the design, and he's got via starvation. People mistakenly believe that HDI costs too much. Well, HDI costs too much because designers don't know how to use it, and they've never been properly trained that if you go with blind or buried vias, then you have to start out from the very beginning and use one of the models to indicate how many signal lay- ers you'll need. Shaughnessy: This sort of thing is in the IPC standards, so familiarity with the standards would definitely go a long way. Holden: Oh, yes, absolutely. at's why we wrote the first HDI Design Guide, which is not a standard. We included some of the different density modeling equations so that before you ever start, do some of these equations to see whether you will need blind and buried vias. If you think you will do HDI without it, then you will get trapped at the very end of the process, and it will be expensive. Moyer: I will give you a perfect example. I recently consulted with a company that builds products that go in one direction really, really fast and don't come back. ey had designed a coplanar transformer for their board, which was rigid-flex, using their magnetics so- ware. eir magnetics soware said that the arc here should be this wide, and you should have this much gap to the next arc, and so on. All was good. ey said that to carry this much current, it should be this thick. From a purely theoretical physics point of view, everything they did was correct. What they came up with was a coil transformer design, set to adjacent layers, so you'd have primary and secondary. e traces were 5 mils wide with 4-mil spaces, and they designed it out of three-ounce copper. is was physically impossible to build. With three-ounce copper, you need at least 10 to 15 mils of gap between traces. e narrowest cop- per feature that most fabricators will touch is about 10 mils, but they prefer about 12 mils with three-ounce copper. e soware that the engineers used gave them some numbers and they designed the whole thing, and they could not get a fab shop anywhere in the world to touch this. I told them what they needed to do to change it, but they didn't listen to me, so I stopped consulting with them. Shaughnessy: So, there are probably hundreds of ways that a board can be made overly complex. Moyer: e fun thing about a printed circuit board is that you can overconstrain it and overdesign it and it will still work. But my favorite question is, "What's the definition of optimal design?" at's really difficult to answer, because given a bill of material and a schematic, there's an almost infinite number of ways that the board can be physically designed. Kris Moyer