I've been working on a personal project for months now, and am getting to the point of designing my own enclosure - which I will be casting and machining from aluminium which will have a number of fins to dissipate heat.
Now, to fit my PCB into the space specification that I have set - I need to pack all my components in very tightly indeed, so there will be minimal area on the PCB to dissipate the heat.
I could of course design an aluminium/copper insert that slots into the enclosure with the PCB, that contacts 'heat pads' on the PCB, and conducts the heat away into the enclosure body - but that of course takes up valuable real estate.
What I propose to do instead, when the PCB slots nicely into it's enclosure - there will be ~1-2mm airgap from the board & components to the inside of the enclosure. I would like to then fill this gap with a dielectric fluid that would conduct heat from the components to the enclosure itself, which of course then dissipates heat to the environment.
We're not talking huge amounts of heat here by any means, but up to tens of watts maximum.
So, a couple of questions really;
Can anybody recommend, know or, or have experience of a specific dielectric fluid with a high thermal conductivity?
Does anybody know if perhaps such a fluid would interact with typical components, ICs, passives, the boards FR4 material etc?
Tens of watts is pretty consequential for an enclosed system.
I know a lot of overclockers have put their PC motherboards into pure mineral oil. It's non-conductive and carries heat better than air, to my knowledge. I doubt it has thermal conductive properties equal to proper heatsinking, though.
To go this approach, you need to guarantee a few things.
All components are sealed. ICs are good. Caps must be airtight. Switches and potentiometers would cause problems if the oil were allowed to create a barrier between wiper/contact and element.
The case itself is sealed. Obviously.
The mineral oil is really pure, and stays that way. With liquid flowing around, you can't have metal shavings or dust hanging out in there.
When I worked for the Admarlty Underwater Wepons establishment in the 70s we used freon to do this. However we were not dealing with such high powers as tens of watts. I think the problem you have is that with such high powers most low viscosity electrical inert fluids will boil.
The cards were packed right on top of each other, so the resulting stack was only about 3 inches high. With this sort of density there was no way any conventional air-cooled system would work; there was too little room for air to flow between the ICs. Instead the system would be immersed in a tank of a new inert liquid from 3M, Fluorinert.
That pretty much sums up how my PCBs will be stacked, albiet one two as of now.
However we were not dealing with such high powers as tens of watts. I think the problem you have is that with such high powers most low viscosity electrical inert fluids will boil.
My rough estimate of tens of watts is for a future design, that really - will likely be 20 watts absolutely max, but distributed from a much larger number of transistors all over the boards whose sum reaches 20 watts. I guess only experimentation will tell.
I know a lot of overclockers have put their PC motherboards into pure mineral oil. It's non-conductive and carries heat better than air, to my knowledge. I doubt it has thermal conductive properties equal to proper heatsinking, though.
To go this approach, you need to guarantee a few things.
All components are sealed. ICs are good. Caps must be airtight. Switches and potentiometers would cause problems if the oil were allowed to create a barrier between wiper/contact and element.
The case itself is sealed. Obviously.
The mineral oil is really pure, and stays that way. With liquid flowing around, you can't have metal shavings or dust hanging out in there.
Yes, a mineral oil will have superior thermal conductivity & capacity compared to air, but inferior to say copper.
Your point regarding switches and pots is a good one, my current design uses a few jumpers for hardware setup - but then again, the next revision is likely to use analog mux's to sort this problem. As for caps, are standard electrolytic caps airtight - I figured all caps would be?
Instead the system would be immersed in a tank of a new inert liquid from 3M, Fluorinert.
We used this stuff in the refinery I worked at for sealing (filling) sensing leads for pressure transmitters used in measuring acid service (pressure, flows, and vessel levels measurements) as it's inert and the stuff has a specific gravity of nearly 2.0. Stuff cost like $800 for a 5 gallon can as I recall.
Freons are covered by legislation these days so I'd give up on that - heavy inert oil seems reasonable, silicone
oils are the most inert and least flammable I believe. You'll need a way to alleviate pressure due to thermal expansion
that doesn't result in leaks. Choose an oil that's not viscous - you want convection.
Alternatively helium gas has a much higher thermal conductivity than air (due to low atomic mass). Painting the inside of
the enclosure with radiator black paint will increase its ability to absorb thermal radiation - a lot of people forget this
with metal enclosures, though it is a fairly small effect without large temperature differences.
And lastly any dielectric fluid will increase stray capacitances around the board, this may affect some things.
MarkT:
Alternatively helium gas has a much higher thermal conductivity than air (due to low atomic mass).
Bear in mind that it is extremely difficult to prevent helium gas from leaking - that's why it's used in leak detectors! I still think the first priority should be to see if the power consumption of the circuit can be reduced.
I still think the first priority should be to see if the power consumption of the circuit can be reduced.
This is of course my first priority, at the moment I'm not dissipating anywhere near that power - but a future version will include the possibility for multiple high side, high current power switches which will up the dissipation.
(...just because I've just done the calculation...) as an example, on the board are 8 transient suppressors dissipating up to ~400mW each, so we're at 3.2W.
Now include the 12V regulator, 5V regulator, 3.3V regulator, 8 MOSFETs driving not insignificant currents, 4 IGBTs that are self clamping, h-bridge IC, stepper motor IC all tightly packed into a 100 x 80mm double sided PCB.
When I get a moment at some point today, I'll do a full calculation of the worst case power dissipation and report back.
jtw11:
as an example, on the board are 8 transient suppressors dissipating up to ~400mW each, so we're at 3.2W.
Surely a transient suppressor should dissipate almost no power under normal conditions, and only dissipate significant power for very short periods when there is a transient?
jtw11:
Now include the 12V regulator, 5V regulator, 3.3V regulator...
If the load currents are significant, use switching regulators. What will the 12V regulator be feeding?
jtw11:
...8 MOSFETs driving not insignificant currents...
If you choose mosfets with a sufficiently low Rds(on) then you can get the power dissipation down to a low value, unless either the current is very high or you are using fast PWM.
jtw11:
...4 IGBTs that are self clamping...
They are more of a problem, if you are using them to drive inductive loads at a moderate frequency.
I'm using the suppressors to deal with flyback when switching inductive loads, as per my other thread you've helped in - hence the power dissipation.
The MOSFETs have an on resistance of only a few mOhms, and are only switching ~5A at the 10s of kHz.
The IGBTs are driving inductive loads, at currents of ~10A up to around 75Hz. The h-bridge is a VNH2SP30, so no - no bipolars here thankfully! Stepper chip uses MOSFETs in the bridges too.
The problem will arise in the future I foresee, with up to 20 high side switches potentially, so this thread was more to get a feel to see if the solution was viable re cooling.
It's worth measuring the resistance of the inductive load, then you can work out what proportion of the stored energy will be dissipated in the load and what will be dissipated in the TVS diode. Also, consider what fall time of the current you need in the inductive load. The higher the voltage rating of the TVS diode, the faster will be the fall time - but the greater the proportion of energy dissipated in the TVS diode.
