Having three heaters on single PCB

I posted this question on electronics SE, but was suggested that I post it here.

I basically having three heaters (serpentine pattern) on a PCB (fiberglass board), as the following image shows:

The ones at the end would be heated to 95°C, the ones next to it at about 50-55°C, while the middle one to about 72°C. They are spaced 0.3mm apart.

The problem is because the 55°C heater is between heaters with higher temperatures, the heat gets transferred onto that area and goes well beyond 60°C, which I'm not able to control.

I tried removing the area between the heaters to see if it would help somehow, but the problem persists.

What can I do to "confine" the temperature to the heater area?

• What is the PCB substrate actually made of? If you don't know, you're gonna have to measure its thermal conductivity.
– Fizz
Nov 26 '15 at 6:38
• The board is a glass epoxy FR4 board. So it's fiber glass I suppose. Looking it up on Wikipedia, I got 1.059 W/m·K in-plane thermal conductivity. I'm not sure how to interpret it though. Nov 26 '15 at 6:51
• Can you put a heatsink on the sections that are supposed to be cooler and blow air past it? Probably routing slots would help a lot, but without a lot more information on the heat loss mechanisms it's hard to tell. If you even just increase the average heat loss by some method the total power consumption will increase but the relative importance of the horizontal conduction will decrease. Nov 26 '15 at 7:20
• Generally double posting between SE sites is frowned upon, but if you don't get anything much better than the guesses/suggestions already said here, you could also ask on engineering.SE because it's mostly a heat transfer problem, and the experts in that are likely to be found mostly there (if anywhere on SE). Also there are some free textbooks on this web.mit.edu/lienhard/www/ahttv131.pdf but going from theory to practice may be quite a challenge on problems like this.
– Fizz
Nov 26 '15 at 8:02
• StackExchange discourages cross-posting questions like this. First off, there's more detail in the EE version of this question. Second, the other posting already has an answer and some constructive comments. By cross-posting, you have fragmented the question and made it more difficult for future visitors to read all of the information and answers. Please either flag this question or your EE version, and request a moderator to migrate the question to the other site. Once that's done a mod can merge the two questions.
– user16
Nov 27 '15 at 13:32

Increasing the insulation between the heat sources there is quite difficult.

Instead, you can sink the heat away:

1. Add a full copper fill on the backside of the PCB and connect that into a heat sink that is kept cool externally.
2. Add copper rectangles between the separate heating elements.
3. Connect the rectangles to back side with plenty of vias.

This should keep the areas between heaters from heating too much. If you need accurate temperature in the in-between areas also, it may be necessary to adjust the heat sink cooling based on feedback also.

Well, air has worse thermal conductivity than practically any solid, so chances are that a hole will decrease the heat transmission. There are however solids with thermal conductivity very close to that of air, so if that board was made (on purpose) from one of those... you won't see much improvement for your hassle.

And since you mentioned fiberglass while I typing this (doesn't quite look a common FR4 I have to say), then you might hope for halving of the conductivity given that Wikipedia give 0.024 Wm-1K-1 for air and 0.045 for fiberglass. I suspect with fiberglass though there can be a lot of variation between between various flavors because it's porous.

According to this presentation thermal conductivity of FR4 is a lot better at 0.25 W/m·K (5x) than that of pure fibeglass given by Wikipedia. (And actually that presentation agrees with https://en.wikipedia.org/wiki/FR-4#Properties but on the through-plane value). In-plane value is even greater there at 0.81 W/m·K to 1.059 W/m·K. I'm guessing the epoxy binder in FR4 contributes a lot to that. So on conducted heat alone you could hope for 10x (or even 40x) improvement in thermal isolation a with hole then. But then there's also radiated heat and depending how the board is oriented convection may contribute too. There are some examples/caculations on the web for radiative heat transfer between some simple geometries, e.g. here but I couldn't find something for yours.

There is software that can do this sort of complex thermal simulation, alas I don't know any free&good such, and it depends of course on having good material data. Actually, of the freebie stuff http://poweresim.com/newhelp/EN/thermal.jsp might help to some extent, but it seems it can only do some chunky blocks.

Actually, one thing you could do to estimate this is to position a termocouple in air at the same distance form the hot traces as the cold[er] traces are. And then measure what this registers. I suspect this would be a resonable approximation for what a hole could hope to achieve (ignoring he self-heating of the cold[er] traces).

• Thanks for the comprehensive reply! I've tried removing the area between -- but wasn't able to see much improvement. I didn't know about hyperlynx before - thanks for that! I think I'll try posting it to the other SE site as you suggested -- I'll probably also try filling the space with some thermal insulator to give it a try Nov 26 '15 at 11:24