Help with thermal calculations

Den tirsdag den 19. januar 2016 kl. 20.46.38 UTC+1 skrev Tim Wescott:

you can probably find some rule of thumb for flat plates in various orientations and come of with some number

but why not get a cheap alu box of roughly the roight size, stick a 20W resistor or light bulb inside and see where it takes you for various orientations

-Lasse

Can you put heat sink type fins on the outside? How hot can it get? As one data point I've got a lamp that needs to be heated (to 120 C inside). The whole thing sits in a blue pomona box (1.5" x 2.5" x 3") Starting up it draws ~0.5 A @28 V, but once running it's about 200 mA (call it

Hey maybe mock up a box, stick a few 10W power resistors inside, turn it up to 20 Watts and see how hot it gets. (nothing like real data!)

George H.

We use a box amlost that same size.

Sitting flat on a wood workbench, lab environment, no fans, the surface of the box rises about 2K per watt dissipated. The internal PCB rises very roughly 6 K/w above ambient.

One product was running pretty hot, FPGA and ADC and some regulators, and the customer was concerned, so we (among other things) added a gap-pad between the PCB and the extrusion, and the PCB/component temps dropped about 12K.

That was my next step. I'm thinking it'll be wise to make one and send it to my customer.

Thank you John and George.

I think that for most uses they won't be demanding the full 20 watts -- but I want to make sure they understand the implications if they go there.

There are calculations for that such as

but in the end I'd do the same thing, test it with a real box.

Better yet, go there with the box and tell them "Let's do some test runs. It's only 12 volts so it's safe to touch. Here, hold this box for a few minutes ..."

Heat-pipe, fan that sounds about like a jet engine, or both... passive cooling probably won't cut it in that form-factor. But it should make a nice coffee-cup warmer...

Can you bolt the box to some metal surface to carry heat away? You'd also have to transport the heat from the hot parts to the case.

Fan?

Creative ambient conditions? Ablative cooling? Isolated hot spot while the rest is cool? Defective instruments or sensors? Heat conduction to table or mounting structure. Lots of ways to do it wrong.

You left out the part where you're not allowed to change anything and where the deadline is already in the distant past.

What material is the case made from?

Well, I can give you a very rough framework. You'll need the specific heat of the material. The calculation is done in three stage. The initial heating to operating temperature and any subsequent melting and vaporization. I think we can safely ignore the 2nd and 3rd stages.

I'll assume that the 1x3x6" block is 50% solid aluminum. At 0.098 lbs/in^3 = 18 in^3 * 0.098 = 1.8 lbs.

The specific heat of aluminum is 0.22 BTU/lb.

Temp_Rise_F = (Watts * 3.142 * time(hrs)) / (Weight(lbs) * specific_heat(BTU/lb))

Temp_Rise_F = (20watts * 3.142 * 1hr) / (1.8lbs * 0.22) = 158F = 70C temp rise

However, that's for an insulated system and offers no consideration for radiation and conduction losses to the environment. With no heat loss, double the heating time and the change in temperature will also double. Without knowing more about the packaging and environment, I can't even guess(tm) at the heat losses involved.

You might also mention to them that semiconductor life is cut in half for every 10C raise in temperature.

Methinks you have it backwards. Sitting on the table, some heat is conducted away from the device and to the table. In other words, you're using the table as a heat sink extension. Dangling in free air will create a cloud of stagnant hot air, which will act as an insulator, causing the device temperature to climb continuously until it reaches meltdown. Of course if the air is moving, that will help draw heat away from the device, and help keep it cool.

Oops. I forgot to throw in the 50%. So, the numbers I posted are for a solid block of aluminum.

I think I'm going to send them a box roughly the size they want with a resistor inside dissipating 20W, and instructions to try it out.

Or maybe just hit them with a temperature rise measurement.

The 20 watts is for a corner case in their performance envelope -- they may just decide not to go there.

Send him a 25W incandescent light bulb. Oughta be close enough for a first estimate. I just did the experiment. Ask him to review the requirements for ruggedness. Anybody who picks it up is likely to drop it when they let go of the "hot potato".

In free air, convection currents will set up immediately and take the heat away.

.... Phil

As John Larkin points out, you've got a well-defined surface area, and it h as to get to 40K above ambient to get rid of the 20W.

If you want it to run cooler, you are going to have to dedicate about half that volume to a finned heat sink, and perhaps a bit more to a little fan t o pump air through the heat-sink - even if you mount it vertically 6" isn't enough height to drive much natural convection, and your customer sounds s illy enough to want to bury it in an enclosed space.

Farnell sells that kind of little fan

Huh, I've got this design in my head where I'm hoping to dump ~10+ Watts out of a back panel... 4"x ~12-16".. With some sort of Al radiator/ heat sink. And (maybe) a 10 degree (C) temperature rise? (15-20 C at max would be OK) I guess I better do my own measurement sometime.

George H.

"my"* diode laser mounts right to 2' x 3' Al bread board. A very nice heat sink.

George H.

Jeff, I think air starts flowing with something as small as a few C temperature rise... I'm not really sure, but some measurements I did made me think so. But it wasn't worth investigating...

Re: on a table, I guess it all depends on what the table is made of Al - good Styrofoam - bad Wood - I don't know... it burns near 451 F. :^)

George H.

Tempera

Yeah that's what I would say.

George H.

When I worked at FLIR we had some mechanical engineers who were very sophisticated in their thermal modeling (we had some pretty tight thermal budgets).

Even so, before they'd commit on power budgets to the electrical team, you'd find them in the lab with housing mock-ups and light bulbs and whatnot.