Issue link: https://iconnect007.uberflip.com/i/1285883
SEPTEMBER 2020 I DESIGN007 MAGAZINE 45 place. You're taking a true variable and current and creating a derived variable and current density; that uses one of the other indepen- dent variables in the first place. You can't treat it as a constant and also treat it as a variable at the same time, and I show illustrations where there's absolutely no relationship between cur- rent density and temperature. Adam: My advice would be to not trust data- sheets, especially not the prediction of the component temperature (laughs). You'll find a value called thermal resistance junction to ambient, and for the temperature rise, multi- ply this numerical value on the calculator by the watts of the component. Hopefully, there's a footnote in the datasheet that reads, "We used a JEDEC-51 like a test board with four layers, and the component was mounted on exactly this test board." What the datasheet then tells you is the thermal resistance of the test board and not so much the thermal re- sistance of the component—certainly not the thermal resistance of your board. You have to be very careful. Brooks: If you think very simply to the trace level, the trace heats by I 2 R. It cools primar- ily by conductivity into the board material, and there's a parameter in the board materi- al called the thermal conductivity coefficient, which is the most important factor in the cool- ing equation. The thermal conductivity coef- ficient is the most important factor, and the thermal conductivity coefficient is measured in two different directions. It's different in those directions. One is perpendicular to the trace, which is the through-plane measure, and the other is parallel to the trace, which is the in- plane measure. Coming back to the datasheet, the situa- tion is a bit better than it was, but it's still not good. The dielectric datasheets typical- ly don't have the thermal conductivity coef- ficient. If it does, it has the single value of the thermal conductivity coefficient, and it doesn't tell you which one it is. One of my calls to the industry is for all dielectric man- ufacturers to fully characterize their offer- ings by the thermal conductivity coefficient in both directions. Adam: It's even worse with specific heat capac- ity and mass density. If you want to run heat- ing curves or on/off states, you have to know the heat capacity of this FR-4 mix of glass and resin, and it is a very sad situation that we don't know the values properly enough. Shaughnessy: It seems like the EDA and the PCB design communities are getting a little more interested in thermal. Brooks: There's more acceptance and recog- nition of the need, but I don't see much ef- fort in learning what it's all about. You see old-fashioned rules of thumbs that are tossed up in the EDA datasheets or application notes and stuff that reflect they don't fully under- stand it. It turns out if you go back to the original IPC data, which was the NBS report published back in 1956, Mike Jouppi did some superb work back in the late 1990s and the early 2000s in tracing down the origins of that stuff. He told me one day that one of the original two au- thors has passed away, but the other was still living. Mike went to visit him, and he still had the original data in his garage in some box- Douglas Brooks