Issue link: https://iconnect007.uberflip.com/i/1359517
APRIL 2021 I DESIGN007 MAGAZINE 19 of a market for simulation tools either. In fact, in the past, anyone who did well ended up going to the U.S., so you lose a lot of talent when a company becomes successful; it just disappears offshore. Holden: Where do your customers come from? Olney: Lately, the U.S. Probably 70% of my customer base is in the U.S., 25% in Europe, and 5% from other countries. Shaughnessy: Do you have any recom- mendations for somebody dealing with simulation challenges right now? Any tips? Olney: Sure. e biggest thing is forgetting what you learned at university. Circuit theory is all DC, but when you get into the frequency domain, with high frequency, fast rise time signals, and a distributed system, you need to think in the way of the electromagnetic fields, coupling, and radi- ation. It's a different world altogether. You really need to educate yourself before you can become a good high-speed designer. e other point I'd like to make is that simulation tools don't com- plete the analysis toolset. My eyeball is just as good as any signal integrity tool. at's because I have (embedded in my skull) 30 or 40 years of IPC design rules, design for manufacturability, and SI and PI requirements, so I can see a lot of issues before I really need to simulate. Just look- ing at the topology of the routing, for instance, gives me an idea of whether a design will per- form to expectations. On the other hand, simu- lation gives you another set of eyes and allows you to visualize other issues. When I do an analysis, I think outside the square. I drill down through the board; I do a teardown of the substrate. First, I look at the impedance and different technologies that must run on each layer, the materials used, and at the minimum cost, of course. It must provide adequate performance, but it doesn't have to be the best material, and it also has to be stocked by your fab shop. So, you must choose the right material for the right job. ese days, one of the largest problems is the amount of power supplies you have on a board. I think the last design I did had 35 power supplies. Some believe, "Routing is simple. You just route between two planes and keep the criti- cal signals isolated, etc." But until you get in amongst it (as a PCB designer, which I am) you come across the real problems of space. You just don't have the space to put in everything, while working out where these 35 copper pours go, and not interfering with anything else is a major problem in board layout. If you have a trace that crosses from one power plane to another, you are creating return path discontinuities and it can cause untold problems with EMI. So, you really need to plan how you're going to place these power supplies to get the most effective route. If you can't route a board it may not be because you don't have enough layers. People oen say to me, "How do you know when you've got enough layers? Is eight layers enough, or do I need 12 layers?" Aer a while you get a feel for it. But a good way to test it is just to let your autorouter go loose, and if you don't get 85% completion, then you've got something wrong. If the board is difficult to route, then it may be because of placement. You need to open channels for the router. An In-Circuit Design field solver simulates a DDR3 stackup.