Issue link: https://iconnect007.uberflip.com/i/1210212
24 DESIGN007 MAGAZINE I FEBRUARY 2020 the connector attaches to the chassis and not to the ground plane on the card. Also, with unshielded twisted pairs, you want to have something called a magnetic—but it's really a ferrite—choke in the differential pair to re- duce the common-mode currents on unshield- ed twisted pair, every RJ45 connector that's shipped on a router or a switch has a built- in ferrite choke to reduce the common cur- rents. Then, if you want to reduce the common currents, that's where you use shielded twisted pairs, and you connect the shield to the chassis. All of these are well documented in Mentor's list of design rule checkers. I think Zuken also has a design rule checker for EMI. Todd Hub- ing's website is also a great source that has good information about design rule checking and kind of a checklist of dos and don'ts for ways to reduce radiated emissions. Holden: And you teach at the University of Colorado Boulder. Bogatin: I teach a practical PCB design class. It's a mixed class with seniors and graduate students. And everybody in the class will do three circuit boards in the classroom. I've de- veloped a five-week design cycle, and it's not so much about circuit design as it is implemen- tation. I've talked about it at my AltiumLive keynote the last few years. I've talked about some of the things that I do in the class and some things I've learned about teaching these concepts to students, but I think it applies to engineers in general as well. Holden: You wrote my SI chapter for The HDI Handbook back in 2007. In it, you have your basic diagram that says that all noise comes from one of four or five sources. Do you still adhere to the principle that SI and EMI prob- lems are from one or all these five fundamen- tal cases? Bogatin: Yes. One of the things I say in my class is that the first step in designing a board is get- ting connectivity correct, and once you have connectivity correct, then it's all about noise; everything is noise. I'm happy that you have a good memory. At the time, I said there were five families of noise, but now I'd broaden it to six families because I've included high-speed serial links (Figure 1). Figure 1: The six families of SI problems. (Source: Eric Bogatin)