Issue link: https://iconnect007.uberflip.com/i/1210212
14 DESIGN007 MAGAZINE I FEBRUARY 2020 lem for 20 years. Once they use the informa- tion they get from my class, it isn't a hard problem to solve; it's that the advice given in the application notes is all wrong. Big surprise there, huh? Shaughnessy: It's a matter of getting the infor- mation out. You were saying that you're going to teach a class next June on EMI in Germany. Ritchey: I have a three-day seminar, and EMI is one of the topics. I also want to add that there are two or three people who have the title of EMI guru, and they have a bunch of rules of thumb, but many of those are wrong. If I know that their ad- vice had been followed, I also know what to fix. There are only a couple of consultants who are well-trained. Happy Holden: There are a lot more seminars now too. May- be there's better education and awareness. Ritchey: I would like to say that, but based on the stu- dents in my classes, I don't think that's happening. Matties: You would have a firsthand point of view to rec- ognize the trend there. When you talk about the students in your class, what do you see? Ritchey: Universally, the core reason that we have EMI problems is that en- gineers don't know how to design their power delivery system correctly, and that was true for almost everyone in my class. Matties: Does it always come down to power delivery? Ritchey: I had only one failure that was not power delivery out of the 200–300 designs I've fixed. That's the source of the noise. one goes away, and that's the villain in almost all cases. Most of what I work on has nothing but dif- ferential pairs in it. Big networking products that you would think would be a problem are flying because of inherently quiet differential signaling. To give you an example, everyone has differential twisted pairs in your house. The energy on it is an EMI band, but we don't have any issues, even with unshielded wires. The reason is that the fields cancel each other on the two wires. That does not happen when you have a single-ended circuit. People having problems have something in their product, such as a par- allel bus, with the power de- livery system designed incor- rectly. That used to be how we made easy money. I'd get a call and say, "Did you follow the app notes?" If the answer was, "Yes," then I knew what their problem was. Howev- er, being a good consultant, I won't tell you over the phone because then you think it's free advice. You fly me out at great expense, we fix it, and you're happy. It's hard to get this concept through, but if you follow the app notes, you will likely have an EMI problem. If your problems are typical applications, then you will always have an EMI problem. Almost all of them say to use 0.1-uf and 0.01-uF capacitors in all sorts of dif- ferent ways. Neither of those capacitors is able to deal with EMI. That's my giveaway. I know what to fix. You don't have the right capacitors in your power system. Shaughnessy: You're saying if you design the board correctly from the start with an eye toward this, you can design out the problems beforehand. Ritchey: My clients have not had an EMI prob- Lee Ritchey