Issue link: https://iconnect007.uberflip.com/i/1285883
44 DESIGN007 MAGAZINE I SEPTEMBER 2020 Brooks: There is a paragraph in IPC-2152 that says the way you size a via is to give it the same cross-sectional area as the equivalent trace. That's all it says. That's about the only thing in IPC-2152 that addresses vias. When I started working with Johannes with the soft- ware, the first thing I did was apply the soft- ware to the IPC-2152 graphs to see if we could simulate them and get the same answer and, therefore, determine whether the software had some validity from that standpoint. We deter- mined that it did, and we could fit the curves extremely well. Then, I asked Johannes, "Can we simulate a via?" I don't want to speak for Johannes, but I don't think Johannes had ever thought about looking at something as small as a via, and it took us a few trials with the software to get it modeled. But we did get it modeled, and it was a result of that modeling that showed us that the temperature of the via was not at all related to the current going through it. The reason is pretty obvious when you think about it. The trace is very much larger than the via. The trace is a heat sink. Whatever cur- rent you put into the via, and no matter how hot that via tries to get, the heat just pours right out of that via into the trace—top and bottom—and the via cannot heat up because of the heat sink. Holden: Does the via ever get as hot as the re- flow temperature? Brooks: The via only gets a little bit hotter than the trace. If you have a trace and a via that are the same cross-sectional area, the via is cool- er than the trace because the via is an internal trace, and the trace is an external trace. IPC- 2152 showed us that internal traces are cool- er than external traces because they are fully in contact with the dielectric, and the traces on top and bottom are only partially in con- tact with the dielectric; they're partially in contact with the air. The dielectric cools bet- ter than the air does. The via might be 10 de- grees hotter than the trace, but that's about it. I've run some simulations out to unreasonable situations—250-mil or 500-mil traces and a 10- mil diameter via—and you still don't get more than 10–15 degrees hotter than the trace. Shaughnessy: Where is most of the interest coming from? Adam: The most important branches of indus- try in Germany are the automotive industry and power electronics. This is obvious, but my users come from all areas of electronics development. These areas include consumer electronics, net- work technology, LEDs, or even technologists who try to find a good solution without having Gerber files or any designs. They create patches with copper and patches without copper, invent a stackup, move holes or vias to where they want them, and apply some current or power. Is heat spreading capable of dissipating the heat from the local heat source? Brooks: That touches on a parallel topic. When I put on seminars, and I show the simulations and stuff and show what happens when you change the trace width or trace thickness and how it impacts temperature, the question, "What should the temperature be?" always comes up. Remember that it's the board de- signer's job to design a set of specifications, but what the temperature should be is answered by a systems-level engineer, not a board design engineer. That may be the same person, but it's a system engineering function to determine what the temperature should be and what the allowable temperature is. It's not the board de- signer's function. Shaughnessy: Is there any advice you'd like to give the designers or EEs just now getting into thermal? Brooks: Find a book. Remember that current density is not related to temperature. All the automated tools out there talk in terms of cur- rent density, but it's not current density, and I build the case in my article from four different directions why it isn't. But the short answer is current density is a derived variable. It's cur- rent divided by area, but the area is one of the variables in the heating equation in the first