Issue link: https://iconnect007.uberflip.com/i/1285883
16 DESIGN007 MAGAZINE I SEPTEMBER 2020 research to define current in vias and micro- vias. We just didn't get there. Happy Holden: I'd like you to relate the first time the two of us ever talked when I first met you in Colorado 35 years ago. I reviewed these mea- surements on a satellite or some kind of space project you were working on, as well as the temperature of these traces, and they didn't match the IPC standards. Eventually, that led to you becoming the chairman of IPC-2152, but I haven't heard the whole story. Jouppi: I don't think it was quite that long ago, but I was working at Lockheed Martin on a sat- ellite project called Stardust. It was the first time in my career as a thermal analyst that an electrical engineer provided power dissipations for the traces in the design in addition to the component powers. It was a significant amount of power, and it changed the design. I asked the designer how he calculated the power values, and that took us to trace sizing charts in IPC-D-275, a predecessor to IPC-2221. When I read IPC-D-275, it stated that vacuum environments should be taken into consider- ation, but it didn't say anything about where the data came from to create the charts or how to take into account the vacuum (space) envi- ronment, so I called IPC. I ended up contact- ing a fairly new IPC employee at the time, John Perry, to ask how the charts were derived. John told me that he talked to Dieter Berg- man. All he could find was a technical paper written by Dr. Charles Jennings from Sandia Labs, and it was published in an IPC technical paper. We were able to get that paper, and I just followed what they had done. I ran some ex- periments, got some data, and created both in- ternal and external traces on these new boards. The biggest revelation was to see the difference between the internal trace temperatures and what was published in IPC-D-275, and what they're giving you for the internal traces there. I wrote a paper and submitted that to IPC, and presented the paper at an IPC conference. Back then, Ralph Hersey was the chairman of the task group that dealt with those charts, and he'd also been doing a lot of work trying to fig- ure out where the numbers came from. At this meeting, he stated that the testing that I did validated the old charts and that we should go forward, using them as they were. I raised my hand and said, "No, I don't think that my paper really validates this. It does the opposite." That didn't go over that well with Ralph. He ended up letting go of that leadership role, and I end- ed up taking over the task group. I also ended up talking to a designer named Jim Yohe in Cal- ifornia who had come across an old Navy re- port that documented the charts in IPC-D-275. I ended up calling the Navy and asking them, and they said they didn't have it, but they said to contact the National Archives. I did, and I was able to find that old document that had the original data. It was work that was done by the National Bureau of Standards, and it was fund- ed by the Navy in the early 1950s. Printed cir- cuits were a new technology, and a study was needed to assess how much current could be applied to the traces. There were no multi-layered boards at the time. They collected data on a bunch of dif- ferent two-sided board configurations, with dif- ferent board thicknesses, materials, and cop- per thicknesses; some of the boards had cop- per planes on one side, and some did not. They took all of those configurations as one data set and drew a line through them to create the ex- ternal trace sizing chart that was last published in IPC-2221. The problem was that this infor- mation didn't get shared with all of the design community. What people don't realize is that those charts represented the external trace de- sign guidelines for the current and cross-sec- tional area and temperature rise. To this day, I've never been able to truly track down when the multilayer board internal trace data charts came out. There's no data that went with those. All they did was take half the cur- rent from the external ones and say, "That's what the limit will be for the internal ones." But there were two people at Hughes that did some curve fitting for the external charts and the internal charts, and the difference was a factor of two on the internal ones, so then it suddenly made sense that nobody really did any internal trace testing to come up with the