Issue link: https://iconnect007.uberflip.com/i/1285883
SEPTEMBER 2020 I DESIGN007 MAGAZINE 43 ber and distribution of heat sources? It hap- pens that under given conditions no reason- able temperature can be achieved, although it would be possible according to the datasheet. However, the datasheet does not know these special conditions. The software is also a virtu- al experimenting table. By the way, I am also a learner, and I like to work with committed EEs from whom I can learn. Brooks: For a board designer using board de- sign software, there's a learning curve that he has to come down on in any particular brand of software. Johannes' software is extremely powerful, but there's a learning curve coming down that, too. The work that I've done with that for PCB trace analysis probably touches on 10–15% of the capability of what the soft- ware can do. If you are a system engineer and want to look beyond traces at what happens on the rest of the board from the contribution of the components and stuff like that, that is a system engineering function, and this software is very well suited for doing that. Dan Feinberg: I'm really interested in the fact that Johannes is saying how quickly the soft- ware can respond to a thermal change. Can you give me an idea of how quickly that would be? Brooks: The software isn't real-time. You set up the parameters, push the button, and it will give you a measure of what's going on at the mo- ment you've pushed the button. You enter the data, but it's not a monitoring thing. If I'm do- ing some complex trace work on a moderately powered desktop computer, it will reach a solu- tion in one to three minutes at most. If I'm do- ing a via analysis, which is maybe two orders of magnitude more complex because of the size of the via walls and much larger matrices you're working with, a good, solid via analysis is going to be anywhere from 20 minutes to two hours. Adam: The software can even be faster than re- al-time; for example, if you wait 10 minutes in the lab for the steady-state, the calculation is done in two minutes. It is the size of the board and the number of layers that determine the CPU time and the effectiveness of heat remov- al. We calculate a network of up to millions of data points, which takes time. Coupling the electrical properties to temperature or tran- sient processes takes longer. However, one must not look at the comput- ing time, but at speeding up the time-to-mar- ket. For a laboratory test, the components and circuit must be ready for operation, but a sim- ulation can start much earlier. Happy Holden: This uses finite element analysis. Adam: It's not precisely finite element, but it's an equivalent. Holden: When I was at Mentor, I worked with FloTHERM and a computational fluid dynamic tool, which was three-dimensional. Adam: Yes. I was a field engineer for FloTHERM in my previous job. Holden: But when they gave me the computa- tional fluid dynamics, I went back and said, "Are you trying to fool me? Is this a video? Be- cause this is a 2006 Dell computer desktop, and it's not capable of doing real-time three-di- mensional 3D modeling finite element analysis with a single Dell." And they said, "No. That's not fake or a simulation. That's real-time, but it's all Russian software, and they do finite el- ement convergence in a totally different way than we do in the West." I got to play around a lot with three-dimensional computational fluid dynamics. Shaughnessy: What are some of the biggest misconceptions that you see out there for the designers and EEs who are working with ther- mal? What are some of the common mistakes? Brooks: I love that question. The temperature of a via is determined by the amount of current that goes through it. That's absolutely false. There's no relationship between current and temperature in a via. Adam: Or current density in general.