@jamiek : … want to make a closed-loop air heat exchanger with thin printed air channels to allow cooling with no chance for dust ingress
@MakerJim : … can get TECs that allow completely sealing an enclosure- but they are not cheap. $200 range IIRC for around 12W of heat transport at 12V.
If sealed, then why not water, like gaming/mining rigs? Conformal/epoxy coat electronics, or whatever nvidia/nasa do for such situations.
Yes, I noted this (complete unit, no further assembly required).
Jamie then noted
Yes, you can get bare TECs dirt cheap, but then have to do design, assembly, integration, test- all the stuff. E.g. the following is one example
Edit: If it were just air-to-air, you might be able to make a heat exchanger from something ubiquitous to a V1 machine design (EMT tubing, anyone?). Fans on both sides moving air through the isolated-but-thermally-connected tubes. I admit it’s an interesting thought experiment if nothing else.
Part of me hates the chance to put debris on the controller, part of me says if hundreds of these machines get used all over the world with no issue then maybe it isn’t worth worrying about.
Always interesting thought experiments in the community!
In space, it’s completely different. No convection. You conduct the heat to a radiator and dump it to space. (or you heat the hell out of it, because one side of your stuff is hot as hell and the other side is cold as space- and your thing happens to live on the cold side)
The initial idea is something like this:
The interior of the box is sealed, not necessarily a pressure vessel but no meaningful exchange between inside and outside. The circulating fan (green) blows the interior air through some narrow channels that circulate back to the main enclosure.
Plastic is not the best thermal conductor, but a printed single-extrusion wall can be pretty thin, and perhaps it can be made to have pretty high surface area.
Then the exterior cooling fan (purple) helps by blowing exterior air past the outside of the thin channels.
There are also special geometries like counter-flow heat exchangers that can be highly efficient, but the core idea is lots of surface area and circulating fans.
As for the active cooling, with TECs, I might go with a ‘dumb’ solution that just integrates some of those plates into the printed air channels. One of the concerns is that there could be condensation on the cool side, but I think if it is just separated from the electronics and arranged so it can’t drip or have droplets blown onto the electronics, it should be okay. Since the side being chilled is nominally closed, it shouldn’t accumulate water continuously like it would if it were open to the room.
I believe that TECs will tend to self-regulate, as the temperature difference increases between the hot and cold side, it produces a reverse voltage proportional to the temperature difference. Maybe it would be enough to wire several TECs in series and let it run open loop from the power supply. That would be nice, not needing any extra thermal controller. It might not be the most efficient, and the cold-side temperature might not be super low, but the simplicity might make it worth the loss in performance.
Very interested to see what comes out of this. I just wrote up a bit in the other thread that I can move here if you want.
Personally, I wouldn’t go down the path of anything too exotic. We’re already pretty decent at getting heat into air cheaply with heatsinks. Things like tube structures tend to be VERY bad at this for both air and water due to surface effects meaning the flow is relatively isolated in the center until you get to high enough flow rates to induce turbulence.
For a TEC based cooler you can just use COTS heatsinks with fans and go heatsink-tec-heatsink with the TEC being in the middle of your enclosure.
For a TEC, bear in mind that you need to put in as much energy as you’re transferring to get 0 degrees of temperature delta, so getting 10W out via a TEC will need at least 10W into the TEC, meaning 20W dissipated on the far side. At that point the TEC is also just behaving in the same way as a block of aluminium, so to make good use of the TEC you need to develop a temperature gradient across them which means you need a higher ratio of input power to transferred power, so maybe 20-40Win for 10W of transfer which gives you 30-50W dissipated on the hot side but should give a 20-30degC delta. Condensation issues will need to be handled appropriately and if you’re trying to get temps relatively low they will definitely ice up and then stop working well because the ice insulated.
As I said in the other thread, though, I personally think just filtering/screening the input and making sure it’s in a location that it’s not dealing with the worst of the dust/chips is probably the better approach, but it’s definitely not ‘interesting’ so I’m keen to see where this leads if only just for fun.
They should do a little bit, but it’s not huge. I’ve used them a couple of times in the dim, distant past and I think we were using current limited supplies to make them a little more predictable from that perspective.
There are a bunch of effects that end up defining where the TEC actually ends up operating. One is hot side temp. Hot side temp depends on your heat-sink there, the amount of power into the TEC and the amount of power being transferred through the TEC. The amount of power into the TEC and the amount of power being transferred through are what set the thermal delta (or vice versa, depending on how you’re looking at it).
For me, that’s the key here. If I’d killed a controller or two then maybe I’d be a little more concerned. As it stands, mine had been open to the air for months, just sitting in the box it came in, then had a bit of cardboard in the way when it was getting aluminium chips thrown around, then I finally got around to sticking a lid over it and have barely used it since…
I’ve also seen enough marginal thermal designs (we have a truly insane one sitting out in our warehouse at the moment that’s similar to the TEC approach) that are horrifically over-engineered and needlessly complex to realise that a fan with a filter and some form of thermal warning goes a LONG way!
Here’s a different idea.
Take a 5015 fan, intake near the ESP32. Blow the output into a “U” shaped EMT tube that goes out, down the beam, then back into the enclosure. Have the EMT then dump air through a shroud that ducts air by the TMC2209 heat sinks.
This is sealed, the airflow is circular, but the EMT body itself is metalic and will dump heat to the room. Only need to have a TEC anywhere if you happen to be in death valley or antartica. Just flip the polarity in either case.
Only the fan moves, and the airflow pulls air by the ESP32, and blows air over the TMCs.