Getting very low temperatures with TECs

[texaspyro]

I was thinking of doing a beefy temperature chamber and came across this.  He got around -55C from a stack of three cheap TEC modules (with a water block cooled hot side into a large radiator).   Looks like he used two modules in series connected in parallel with the third module.

https://www.youtube.com/watch?v=aSGwaQWA05U&frags=pl%2Cwn

[David Hess]

Stacking TECs absolutely works and is necessary for larger temperature drops but quickly reaches diminishing returns.  Careful consideration of insulation and condensation needs to be made.

[wraper]

Stacking TECs absolutely works and is necessary for larger temperature drops but quickly reaches diminishing returns.  Careful consideration of insulation and condensation needs to be made.

I would suggest larger cells / more cells stacked used only at a fraction of their rated power because they are much more efficient at such conditions. Then returns won't diminish so fast.

[jeremy]

Isn’t the radiator sitting outside in below freezing temperatures? If you live in Texas as your name suggests, I can’t imagine this is something you would have regular access to?

[Kleinstein]

Stacked TECs need the TEC at the hot side to be much larger than the cold one, as it has to move additional heat. Using 1 element for the cold side and 2 of the same on the hot side is probably still to little. From what I have seen it's more like 1 : 9 , usually with a small unit and a large one of 9 times the area.

It is sometimes used for things like IR detectors - at very low power.

[Conrad Hoffman]

I think the makers offer triple stacked units. Look at those designs for clues. More than that probably isn't worth the effort. You end up with a very small area of low temperature and you'll want good insulation (like a vacuum) to really take advantage of it. If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

[extide]

You can definitely get pretty cold temps with TECs, but of course all of the power you dump into it turns into heat that you need to cool as well. I remember about 15 years ago or so I ran an overclocked Pentium 4 system under a TEC with a water block cooling the hot side. Was kinda fun at the time. They are kinda tricky to use if you actually need to cool something that is putting out a lot of power, though. As long as whatever is in your temp chamber isnt putting out too much heat it should be pretty doable.

[edpalmer42]

The other thing to remember about a temperature chamber is that it has to deal with the heat generated by the DUT.  This limits the usefulness of TEC coolers.  A bar fridge might work better.

[wraper]

If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

Quite stupid suggestion frankly. If you are so concerned about thermal conductivity of wires, use wires with small cross section. You certainly can use smaller cross section with copper compared to other materials.

[daqq]

Peltier stacks exist and totally work, but as was mentioned, there is a limit. What I've seen done is a very small peltier on top of the stack, on it is a sensor. Below is another peltier, bigger, and below that another one, bigger still.

See:

https://www.digikey.com/en/product-highlight/m/marlow/multi-stage-series-thermoelectric-coolers-tecs

[David Hess]

Stacking TECs absolutely works and is necessary for larger temperature drops but quickly reaches diminishing returns.  Careful consideration of insulation and condensation needs to be made.

I would suggest larger cells / more cells stacked used only at a fraction of their rated power because they are much more efficient at such conditions. Then returns won't diminish so fast.

A single TEC will only support a maximum temperature differential.  At some point it is better to stack them than to add more parallel units.

As Kleinstein pointed out, the TECs on the hot side have to pump more heat so should be made larger.

Just informally messing around, stacking a smaller TEC onto a larger one easily resulted in ice accumulating on the cold side of the smaller one even with minimal heat sinking on the hot side.  In a real application, the cold side will need to be protected against condensation.

[Conrad Hoffman]

If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

Quite stupid suggestion frankly. If you are so concerned about thermal conductivity of wires, use wires with small cross section. You certainly can use smaller cross section with copper compared to other materials.

Having built commercial TEC cooled devices with a lot of wires where thermal conductivity was an issue, my real world experience probably trumps your opinion. Try to keep an open mind.

[jeremy]

If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

Could you suggest something? Stainless steel perhaps? Although I imagine it would be difficult to find stranded and insulated stainless steel wire?

[Jeroen3]

I was thinking of doing a beefy temperature chamber and came across this.

Did you research how to operate commercial temperature chambers cool when the EUT is emitting some heat?
A normal industrial freezer already has problems with this. Let alone the all the condensation problems you will be having.

Typical chamber goes to -20/30 C, then it needs external cryogenics to get lower.

[RoGeorge]

Another thing to mention is the elements must be powered with constant DC (for best efficiency), not PWM.  That is because, for the same target cooling temperature, a pulsed DC will produced more heat than a lower and constant DC.

[Kleinstein]

If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

Could you suggest something? Stainless steel perhaps? Although I imagine it would be difficult to find stranded and insulated stainless steel wire?

The material of the wires is not that critical. The ratio of thermal to electrical conductivity does not change very much with metals. So a thin copper wire is as good as a thicker brass one. Getting thin copper wires is often easier than isolated brass of phosphorous-bronze.

[Conrad Hoffman]

The problem is when you have a lot of wires, or need to go a very short distance, say for an IC of some sort. The gauge of copper would be impractically hard to work with. You'd want something moderately better, but not too annoying to work with, so it should be solderable. I'd start with constantan or manganin, which have lower thermal conductivity. Choose the gauge so you can just barely live with the increased resistance.

[texaspyro]

Although I imagine it would be difficult to find stranded and insulated stainless steel wire?

Stainless steel fishing line leader...

[edpalmer42]

Rather than using exotic materials, you could just make the copper leads very long.  The dual-oven version of HP's 10811 oscillator has the leads wrapped a few times around the oscillator and fully thermal insulated.  Nobody's quite sure why that was done, but the best guess is to provide thermal isolation and low thermal gradiant for those leads.

https://www.realhamradio.com/GPS-oven-journey.htm

Ed

[pwlps]

The material of the wires is not that critical. The ratio of thermal to electrical conductivity does not change very much with metals. So a thin copper wire is as good as a thicker brass one. Getting thin copper wires is often easier than isolated brass of phosphorous-bronze.

The ratio of thermal to electrical conductivity also depends on temperature:
https://www.lakeshore.com/products/cryogenic-accessories/wire/pages/Specifications.aspx

In cryogenic applications I'm using phosphor bronze or manganin wires  for all signal/measurements terminals, actually I don't have active components in my apps (passive sensors only) but I imagine the thermal losses might be minimized using copper for power lines and some of these alloy wires for signal etc.

[mzzj]

Kryotherm Frost-74 sitting on top of 4 parallel connected Frost-74 modules would get the intermediate temp to -20C and cold side to -65c or so with 2 watts load.
Not for free tough, some 300 Watts dissipation at hothot side. So you get beefy power dissipation but tiny chamber(something like 40x40x40mm is doable)

[Bassman59]

If the thing you're cooling has leads, don't use copper. Use something much less thermally conductive, even if it has some resistance.

Quite stupid suggestion frankly. If you are so concerned about thermal conductivity of wires, use wires with small cross section. You certainly can use smaller cross section with copper compared to other materials.

When I worked for the local observatory doing infrared imaging, the sensor was cooled to 25K with liquid-helium refrigerators. The cold heads got down to 4K and the thermal straps from the heads to the sensor cold block was copper braid. Constantan wire was used to connect the sensor outputs and control signals to the hermetic feedthrough connector, for precisely the reason Conrad suggests. That temperature differential is significant!

Oh yeah, the dewar in which the sensor and filters and diffraction grating and stuff lived was the size of a clothes washer and took about a week to pump down to sufficient vacuum.

[Bassman59]

I was thinking of doing a beefy temperature chamber and came across this.  He got around -55C from a stack of three cheap TEC modules (with a water block cooled hot side into a large radiator).   Looks like he used two modules in series connected in parallel with the third module.

How big is your temperature chamber?

Remember you want to pull a vacuum on it so you don't get snow on whatever you put in the chamber.