Absorbing heat?

[T i m]

Hi all,

I have some neat extruded ally cases for my std Raspberry Pi's and I was about to put my new 3B+ in one and wondered, because it seems to get the warmest of all the Pi's so far ... if the interior finish (shiny / matt) / colour (white to matt black) would make any difference to how efficiently the case would re-radiate the (Pi) heat?

eg, If we assume the case is sealed (it isn't but doesn't have much in the way of airflow) then the heat coming off the Pi to the inside of the case would be direct in the form of radiation ... and indirect by the way of conduction from the warm air, that distributed / transferred via convection to the inside of the case (and hence via conduction to the outside etc).

So, the case has a slightly shot blasted / gold anodised finish throughout and I was wondering if painting the inside matt black (acid etch spray primer and matt black spray) would have any cooling advantage at all or would it in fact actually make matters worse because the paint would act like a thermal barrier?

I'm pretty sure the Pi would survive irrespective as I believe it has the ability to clock-throttle when it overheats but equally if you can help it run 'cool' it will run at full speed for (and possibly last a bit) longer?

I did consider making up a heat-pipe (a 'U' shaped copper strip) allowing me to conduct heat from the main chip to the case but that requires a bit more 'engineering'. ;-)

I've ordered the one of the little 3 x heatsink kits that may add a little thermal capacity (to the chips) and the ability to conduct the heat away from the chips themselves more efficiently but that may only slow the raise in temperature of the chips but could increase the speed the inside of the case is heated, especially if it runs longer before throttling?

Anyone confirm the practical science behind it all please?

Cheers, T i m

[Martin Brown]

There is some difference notably shiny metallic surfaces are terrible radiators of thermal IR at ambient temperatures. Almost any colour of matt paint is black in the thermal IR band. How much difference it makes on a plastic box which is intrinsically a poor conductor I don't know

It will probably help improve things by painting it black. I once built a linear 12v PSU using the then new 78xx series linear regulators and it went into thermal foldback mode when put into a shiny aluminium box. Painting the shiny metal matt black solved the problem completely - it still ran warm but nowhere near going into self protection mode.

Arranging it so that there is a vertical path and warm air rises may help to encourage it to stay cool.

[T i m]

I hadn't really considered how will the case would re-radiate the heat.

Ok.

Sure, but this is ally so I thought it might be more predictable etc?

So from what you said above that means inside and out?

(A bit like I have here etc).

That is my normal solution with PC's and the like but in this case I was sorta interested to see if I could 'get away' with it by just changing the colour etc.

I was sorta working from the idea that many laptop 'brick' type PSU's are in sealed / plastic enclosures ... often get warm or even fairly hot without much in the way of negative consequences (what feels 'hot' to us isn't generally that hot for most electronics etc).

I was just trying to remind myself of the science behind the ability of a surface to reflect or absorb heat and if it was a function of it's colour or not, especially when the heat wasn't being radiated in the visible light spectrum?

Cheers, T i m

[Brian Gaff]

Its a shame these devices do not have a thermal mounting point to enable the metal case to cool it. I think from experience matt black paint does insulate but can also get hot in itself. To be of any use it needs to be part of the material and share its thermal qualities.

Brian

[T i m]

Agreed Brian but they probably don't (yet anyway) because the designers consider it unnecessary?

The thing is I've not read anything about how (thermally) restrictive an enclosure you can put such things in but am trying to use common sense / experience to not make it worse than it need be.

Yes, if you had two sheets of ally and pained one matt black and left one polished silver, I'm guessing the matt black one would get hotter if you shone an incandescent spotlight equally across both. I'm not sure there would be any difference if you blew a fan heater across both or touched both with a clothes iron, so it's only radiated (not convected or conducted heat).

I agree, that would be the best solution but conscious it may be overkill in this situation (even as part of just experimenting etc).

Cheers, T i m

[Martin Brown]

If you can connect the aluminium box to the heatsinks then that is the best way to get the heat to the outside.

I was surprised at the time that it made such a difference. When an object becomes around 30C above ambient temperature the radiation losses start to become significant (scale like T^4 in absolute temperature).

Forced air ventilation is the other way to keep things cool.

There are some very cute designer materials that are capable of cooling when placing in sunlight under a clear blue sky. They reflect most of the visible light incident from the sun whilst radiating incredibly well in the thermal band IR.

Priced at the unobtanium level at present but it is an interesting technology if a cheap plastic film version can be manufactured.

[Andy Burns]

would get in the way of the expansion connectors (not that everyone uses those)

[Theo]

ISTM that if you have a closed system, all that matters is the flux to the outside. The nature of construction of the closed system will affect distribution of heat within the system, but fundamentally it's the amount of heat you can get rid of that matters.

Here I'm modelling the system as $hot_interior_stuff surrounded by some kind of case shell. That shell is not convecting, because it's a solid. There is no air convection through it, because it's sealed.

So the overall thermal resistance of the shell is:

a) absorption of heat from $stuff to shell b) conduction of heat through shell c) radiation of heat from shell to environment

(I'll declare convection, conduction of heat from the shell to the environment as orthogonal for now)

I think I'd agree that a black interior would absorb more internally radiated heat, but you'd achieve a lower resistance by conduction directly from $stuff to the shell.

So I think you maximise by colouring it black for a) and c), but you don't want to make b) worse by painting it. For instance, a black anodising layer would be very thin. If that's not an option, the thinnest paint you can find followed by scraping off interior paint to achieve a better conductive interface might be appropriate.

I suspect the difference in absorption a) based on interior colour is going to be outweighed by adding additional shell layers hindering b). However you'd really need to run numbers to find out.

Theo

[T i m]

Given that we aren't talking about 'loads' of thermal energy here, my idea was a 'U' of copper strip with one outer face of the flat part of the U on the chip and the opposite side bolted to the underside of the case top. Being formed into a (as long as practical) 'U' should provide a modicum of flexibility and the thickness would be a compromise between thermal conductivity and allowing said 'give'.

Maybe even brass 'shim' material formed into said 'U' would be better at conducting the heat away and to the case than the air would with convection?

Cheers, T i m

[T i m]

Agreed.

Correct.

Check.

Check.

Check. I had considered the use of an external finned heatsink to increase the passive cooling *of* the case but wasn't sure if the case itself would typically be better at re-radiating the hat from the case than it might be accepting it in the first place (convection / radiation).

Ok.

Ok.

Agreed.

Agreed.

That's why I considered a spray acid-etch primer (spray likely to be thinner than brush on) and a light spray of matt black.

Unfortunately the RPi PCB is held off the bottom of the case by two brass standoffs on one side and the HDMI, uUSB and audio ports poking though the case on the other (so the metal shells of the HDMI and USB / Ethernet ports would slightly help conduct internal heat to the ally case).

And why I thought I had better ask the scientists here. ;-)

I have been reading up on some of this and apparently the internal temperature sensor of the Pi isn't that accurate so some sort of external (thermocouple / temp sensor) might be required.

That said, basically I was just looking to see if there was something easy I could do before I box this Pi up in it's case that might help the thermal situation, preferring things to run cold rather than hot.

Alternatively I could perforate the top and bottom of the case but there isn't much room around the PCB for any airflow. Perforating the sides could allow airflow across the top of the board at least, as long as I keep the case on it's side?

I do have a couple of small 5V fans and some have powered them from the GPIO pins but others have suggested the (electrical) noise may not help the stability of these higher clocked models. Nothing stopping me running the fan from the external PSU though (that might isolate the noise to some degree) but I would prefer it to be a KISS / silent solution.

I'll lave a look to see what sort of temperature monitoring you can get on the Pi / Raspbian and then run it open / enclosed and see what happens.

Cheers, T i m

[Thomas Prufer]

Or fill it with an insulating liquid. This means that all the heat transfer runs via conduction and convection, and the entire case would be at a similar temperature. (I'm guessing that inside radiation would lead to a hot spot, and that this would be less pronounced if the inside were liquid-filled.)

If you think this is a silly idea: Maybe, but it's done by PC and Raspi overclockers, -->"overclock oil".

If you think that a heated sealed container of oil with wires going through rubber grommets is a potential oil spill, you are probably (oil) spot on.

Thomas Prufer

[T i m]

Funnily enough I'd already thought of that one Thomas. ;-)

Quite. Rather than filling the existing case I could just drop the bare RPi into a container filled with said liquid, that way I wouldn't have to worry about seals etc.

No, I saw a CRT being run in a liquid years ago and aren't the Cray CPU's (directly) liquid cooled?

I'm still trying to find a slot to setup and run some thermal tests, firstly with the RPi just out and open, then with it in the metal case as-is and then with the little heatsink kit and lastly with the GPIO slot uncovered.

The thing is I know air has density and so there may be a limitation re how small a hole or slot that would be of any use re convected air cooling (from within the case).

A mate of mine found air doesn't always adapt to 'scale' when he built an RC submarine and felt the holes in the deck looked too big and were out of scale. So he filled them and re-drilled the deck with the correct scale holes, only to find that the air bubbles wouldn't pass though the smaller holes so the thing maintained too much buoyancy because of the trapped air ... and wouldn't submerge!

Cheers, T i m

[Rob Morley]

Me too, but I thought it would be too much hassle if you wanted to access the board, e.g. the GPIO pins. :-)

[Rob Morley]

I suspect heatsinks fixed to the chips with thermal adhesive stand a better chance of making good thermal contact than a flexible strip that you can only glue at one end, even though they rely mostly on the air inside the case to lose heat.

[Rob Morley]

Perhaps it's time to have a go at home anodising? Because that would probably put the "black" in very good thermal contact with the alloy surface.

[Tim Streater]

Absorbing heat was what was happening to me while cutting the grass, so I filled myself with an insulating liquid.

[T i m]

Hmm, whilst I like the idea of that I think it may be a step too far in this case. ;-)

Absolutely ... and I don't know why they weren't made in black in the first place.

That said, I felt my OMV RPi server yesterday and that's been on for weeks and the same case felt pretty cool. That said, being just a file server means the CPU is rarely heavily loaded or loaded for long periods and mine is a RPi2, not a 3B+ ... displaying a video for 2 hours?

*If* matt black (anodising) would be the best finish in and outside the case for thermal reasons, I may ask the manufacturers if they can supply them in that finish ... or a local anodiser if they could re-coat some for me?

Cheers, T i m

[T i m]

If I was to use a thermal epoxy to bond a 'heatpipe' to the main chip, I would then bolt the other end to the inside of the case, also using thermal compound.

Quite, and then we are back to how well the case itself might relay that heat out.

I was hoping the heatsinks would provide a little thermal inertia and also a better chip-to-air transfer. More heat transferred into the air within the case should make the air warmer (and the chip cooler) and therefore the case may be then able to re-radiate more of that heat to the outside world?

An external heatsink bolted to the outside of the case (and with thermal paste) and with the heatsink fins arranged vertically re the typical airflow should silently help to keep the case itself cool.

What about attaching a (matt black) heatsink to the *inside* of the case (again, with paste or thermal glue) to increase the cases heat absorption rate (and thermal mass as an aside)? ;-)

Cheers, T i m

[T i m]

As an aside ... I wonder what proportion of RPi users actually use them to interface anything (using the GPIO pins)?

I have several RPi's and I've *never* used the GPIO pins for anything, whereas I have used all the various i/o interfaces on all my Arduinos.

A mate bought me a BBC Micro:bit and after the Arduinos I found that really complicated and confusing to use. ;-(

Cheers, T i m