cooling tests

[Tesa]

Hello everyone

I spent my weekend doing a lot of tests with different cooling.

My starting point was
one 60w china TEC powered by 14,8 V.
23.4 degrees room temperature
aluminum heat sink 150x100x50 mm on the warm side
50x50x04 mm aluminum plate on the cold, mounted with 1 M140 diode drive to 2W .

I tried many different configurations of 1 and 2 TEC .

my best attempts without insulation gave a diode temperature of 0.8 degrees.
with insulation I could reach -1.6

I tried to change the heatsink out with a cpu cooler , and then took it off . my best setup gave a temperature of -12 degrees.

I tried to add a 4mm aluminum plate between the TEC and the CPU cooler, but I could only reach -4 degrees .

the conclusion is (which I already knew ) that copper is better to conduct heat than aluminum .

BUT does anyone have experience with the copper as "optical table " ?

or by making a hole in the optical table so you can get a cooler directly under the module ?

Does anyone have any other ideas on what can be done to cool more , without using dry ice or nitrogen ?

I ordered a 300W TEC Europe made ​​, which I will test tomorrow . If anyone has any interest I will post the results here .

[The_Doctor]

The most important thing is to get the hot side temperature down. Your heatsink isn't big, but if it has fine vanes or vanes with ridges to increase surface area, you can get a lot more heat out by forced air to get it as close to ambient as possible, which will drag the cold side temperature down. To couple heat in the diode mount to the baseplate, a copper heat spreader might help, but it doesn't need to be big. Make the baseplate and hot-side heatsink big, of aluminium. All the copper has to do is match heat conduction from small heat source to large heat sink, so the copper area on the base plate needs no more than 5 times the area of the diode mount base. You won't do better there unless the mount itself is also copper. A better place to look is parasitic heat loading, so use stainless bolts, crinkle (crenelated) washers, fibre washers together with them as well. That gets the heat conduction low while remaining physically strong.

Unless you're using a TEC controller, and enough hot-side heat removal to give you lots of headroom, the result will not be a steady temperature but one highly dependent on hot-side temperature. Also, using a bigger TEC might not work well, they're less efficient in context because it's harder to remove enough heat unless you want so much aluminium in the projector that you'll have trouble lifting it easily! It might be far better to use two smaller TEC's, a solid aluminium block that fits one of them on a vertical side while it sits on the baseplate, so the cold sides meet like a wall and a floor, then get the diode mount in full contact with both. Use plenty of closed cell polyethylene as insulator. I tested that once, it's better than neoprene, polyurethane, etc.

If you can get hold of design guides from Melcor or Marlow, do it. It's not easy to work it all out but it will give you a clearer idea if you study it a while anyway. The main thing is don't use more TEC power than you need, because any extra is just waste heat you have to remove! Defining 'enough' is the tough bit. The only way to make it easier is to use insulation to reduce parasitic heat loads so the calculations are close enough to what you're actually removing from the heat source. And a 2W diode's heat will be roughly forward voltage times current minus 2W. Actual heat pumped will be more because of leaks. Then add the power used by the TEC(s) capable of pumping it to get the low temperature you want, taking into account hot side temperature to get difference between sides of TEC.

Bottom line: Unless your hot side heatsink is a water-cooled monster, that 300W TEC will annoy you, and may make enough heat to break something.

[planters]

I agree with the Doctor. As long as we assume you want to stick with TEC cooling then the heat load and the temperature differential need to be estimated and then you can pick the best combination of TEC modules and heat sinking. Because these devices are so inefficient you need to try to get them to operate as near as practical to their optimal range to get the best results. Adequate heat sinking the hot side is a lot more important than any heat distribution plates or the choice of interface material (such as thermal grease, thermal epoxy, indium film etc) because there is so much more heat to remove with the added load of the TEC's themselves. As a very general rule expect to get about 25 degrees C of cooling per stage if using multiple stage TEC's and because each stage adds its heat to the load it cools, each TEC down the stack needs to increase its cooling capacity. You can design this using the tables and graphs from the manufacturers or do this experimentally if you have some TEC's and can power each independently.

[Tesa]

My Heatsink don't get hotter than 28 degrees when I force cooling them, så i think that the ted is my limit right nom. am i wrong? I will go with TEC controller, but this is just to find the best way to get as cold as possible. In my final build i would like to have 6 or 8 ps P73 diodes driven to about 1,5w each. I read somewhere that it would require a steady temperature a 0 degrees. am i wrong here? therefor i though that i needed som headroom in cooling capacity, because of different room temperature.

[planters]

You will need to design for around 3V x 1.7A/diode or 36watts at a delta of 25 degrees. I could locate a TEC that would do this and link you there. Don't forget to seal the diodes from room air to prevent condensation. Take a look at this video and see if this is what you are trying to do.

http://www.youtube.com/watch?v=kAgmscFjrvM

[The_Doctor]

The very cheap TEC1-12709 modules on eBay look like being close to that spec. Qmax of 72W, and with a hot side at 35°C while pumping 36W they should get 0°C cold side temperature, and easily lower with hot side at 25°C. I'm basing that on Melcor's specs for CP-1.4-127-045L which is a very close match to this TEC. It may be better to go for TEC1-12712, also very cheap on eBay, because TEC1-12709 would need close to Imax to do this, and TEC1-12712 will do it more efficiently, closer to optimum current for best efficiency, and do it on a standard 12VDC supply with no controller, too. (They're both 40mm squared.) Don't be tempted to go stronger, it just wastes more energy that has to be got rid of...

EDIT:
I think TEC1-12709 would be more than good enough for a quad array, or a single mode RGB module, in case anyone;s interested.

[planters]

I would size these a little more conservatively, to allow for some ambient heating. These will be in atmosphere and the will have multiple wire leads. I think and this is just a guess, based on having run theses things on a few projects, that you should allow a 1.5 correction factor or 54 W. Also the performance curves tend (I don't know the precise ones for the module you identify) to run more efficiently closer to 1/4 of Q max. So, I would use three of these modules in parallel but, this is a rough estimate without seeing the curves.

[The_Doctor]

I already uprated a bit (assumed a need to get just below 0°C with hot side at 35°C while pumping 36W). With at least 3mm thick closed cell polyethylene foam on case ceiling and walls and any unused baseplate area inside the vapour-tight enclosure, parasitic loss should be very low with a compact build in comparison with the main load. The diode mounts and thermal coupling will be the weak points. If the baseplate is kept any lower than 30°C (thread poster suggested 28) the TEC1-12709 will get it there for sure, just not with ideal efficiency.

It's likely not good to go for ideal efficiency because of cost, and even if it were done it might better be done with one 50mm TEC to get big area instead of more than one, simpler wiring (fragile connections, routing more than one pair of high current wires in a tight space is asking for breakage somewhere), and a certainty that the thermal coupling won't be compromised with the risk of small height differences in the TEC's. (A thou or two difference can scupper the effort).

Melcor state that best performance is had at between 40% and 80% input power. I extracted the following ratios from their graph:
65% of Qmax at 40% of Pmax
85% of Qmax at 60% of Pmax
95% of Qmax at 80% of Pmax

As Vmax (at Qmax and Imax) for a 127 couple TEC is 15.4V, these standard TEC's are aimed at running at a nominal 12V. Which is nice because it takes a lot of the hassle out of this, that point was chosen as a good compromise for convenience and efficiency (and to make our lives easier).

While the uprating can be extended for efficiency the increase in Pmax (couple count and thus Vmax being same) is proportional only to Imax, which also makes it easy, just pick a bigger Imax. TEC1-12712 is working at 75% Pmax in this context, and TEC1-12715 is at 60%. All doable very cheaply with one 40 mm squared TEC. I think the 12V nominal value for a TEC with 15.4Vmax (same for all 127 couple TEC's) is chosen for this midpoint of 60% so the TEC1-12715 is probably the best choice, and barely more than my previous best shot at selection. Uprating for parasitic loss will not need a stronger TEC if the parasitic loss adds 50% to the load because you have 66% headroom before you hit Qmax. Better to go with one cheap TEC running on a simple 12VDC supply and just do all you can do get those parasitic leaks down. That way the diodes will get as cold as they get, usually below 0°C, and close enough even in a hot room if good air cooling is used on the hot side.

Some ideas to reduce parasitics:
Stainless steel wire for diode's main wires, branch off with copper on the cold side. Stainless conducts electricity a lot better than heat... Could be trial and error picking the gauge though.
Wire diodes in series to reduce wire gauge and its thermal conduction. (Presumably done already to minimise driver count).
Keep surfaces white or shiny, difference in temp may be enough for radiative losses otherwise, given closeness of hot and cold side materials.
Special attention to washers, fibre and crinkle types used together, gives some safe spring tension (avoids destructive errors in torque when tightening) as well as big reduction of heat loss.
If the enclosure can be 6mm wider to get another 3mm of closed cell polyethylene foam to make 6 in all, or more, do it.
Leave a cold trap area on the TEC to get vapour. Once it's all solid there it won't move again and it will not cause loss once it settles there.
Sealed TEC's perform a bit less well, but the long term performance is better if no water vapour can get in to the couples, so it's worth it to get a sealed one.

Reserve further efforts to keeping the hot side temperature down! Better that than add more waste heat with a yet stronger (or bigger) TEC.

EDIT: (Left for the record, is replaced by following post... The 12V nominal TEC voltage is likely chosen for cheap devices and safety margin, not efficiency, when using basic 12V supplies like batteries on chargers or in cars, etc.)
At the 60%-of-Pmax point, the TEC1-12715 will be using nearly 140W to pump nearly 40W plus parasitic losses, so maybe over 200W in all. Going for best-case 40% will still need about 150W+ to remove from the hot side, so there's not much gain there. While using a bigger TEC might help a bit, going to over 50 mm square won't be needed, and if you risk using more than one TEC there may well be more performance lost in thermal coupling problems than is gained in aiming for the 40%-of-Pmax best possible case. No matter what you do, if you have three colours all putting out similar waste heat, the projector will heat a room in winter. Just not as much as a gas laser...

[The_Doctor]

Bit more...
I went through the Melcor graphs carefully (first time in years...). It looks like that TEC1-12715 is a good choice for this, at close to Iopt too, even with a bit of parasitic loss if hot side is 30°C or less. (Conditions were assuming Th=35°C, Tc=-1°C, Qc=36W plus some slight losses).

That TEC has varying specs on the web for Qmax, but I saw between 138W and 150W. The Melcor guide indicated a need for a bit more than 136W Qmax to get Iopt of 0.4 of Imax.

The one thing that didn't check out was the voltage, to drive this TEC at Iopt wants as low as 7.6V, not 12V! (And still only 10V at my original assumption of 0.6 of Imax). Not that it matters other than for efficiency, but the good news is that the heat waste is a lot lower than I thought, 0.4 of Imax (15A) being 6A, times 8V, just 48W instead of 140W! (Plus whatever gets pumped). A lot better.

Going for a 50mm square TEC might help if the extra area and better thermal coupling is worth the price, and space is no object. Alternatively there may even be a 40mm squared TEC1-12718 to run at 6V or even 5V, but I don't know if there is one. Could be great if there is though, 5V at high current is cheap to get.

[shrad]

you could also use 4 units of a 20mm 5V TEC in parallel