TEC cooling

[sdouble]

Hi Guys,
I currently need to cool down my preamps to reduce serie noise of the input j-FET.
At best, I would like get -30 Celcius. The TEC cooler must be on top of my Aluminium Enclosure.  So the full enclosure will be cooled down.
Total dissipated power in the board is 2 W. I plan to use 2 stage TEC glued on top of the enclosure. Heat sink above the TEC and good quality fan above the Heatsink.
Pretty bulky and heavy but I can cope with that.
The design I have in mind is shown there :
https://www.aliexpress.com/item/Free-Shipping-DIY-Peltier-Air-condition-refrigeration-plate-TEC12706AJ-12V-Cooling-fan/783238242.html?spm=a2g0s.9042311.0.0.KDWeEX
but I found strange to have 2 metallic screws attaching the cold plate and the heatsink.
I plan to use a different TEC than that one.
All design suggestion are very much welcome since I have no experience with TEC cooling.
Moreover, would you know a good producer for TEC units. I do need something really reliable

[jbb]

I'm not a Peltier expert, but getting down below 0 C seems to be difficult.  We tried it at work, thinking "how hard can it be?" and the answer is "pretty hard."

The killer seems to be that Peltier elements leak heat internally. So you apply more current to pump it out. Which generates more heat.  It's absolutely critical to a) get low thermal resistance on the hot side and b) have the smallest possible cooled object on the cold side.

Here's the informal advice we got:

• Apparently the data sheet 'maximum current' is a fantasy horror in which you pump heaps of current in, only to get high I2R losses and not get any cooler. So if you're after high temperature changes, try not to go over 50% of rated current.
• The semiconductor elements inside are a bit delicate and don't like PWM and really don't like on-off (thermostat) control.  Apparently they should be supplied with smooth, steady DC. For testing, a bench supply should do nicely.

Also look out for condensation and frost.

Good luck.

[fourtytwo42]

At best, I would like get -30 Celcius. The TEC cooler must be on top of my Aluminium Enclosure.  So the full enclosure will be cooled down.

Have you worked out the surface area of your box, the temperature differance to ambient and hence its heat gain, unless your box is really tiny impracticable.
You cannot hope to get to -30C without severe condensation problems unless you enclose the cooled device in a sealed enclosure full of dry air or gas.

[sdouble]

Condendation is really an issue.
I planed to integrate 2 dessicant bags :
https://www.aliexpress.com/item/10pcs-lot-Silica-Gel-Desiccant-Moisture-Absorber-10g-bag-Desiccant-Gel-Packs-Silica-Gel-Desiccant-Reusable/32441582001.html?spm=a2g0s.9042311.0.0.kQfsdE
into the enclosure. Not sure however how will this works below 0 celcius.

humidity is a major concern for my circuit : high impedance, high voltage both don't like humidity.

[Pack34]

TEC's have a maximum temperature difference between the hot and cold sides. This means that if you want to get that cold side down to where you want it, then you need to be able to remove heat from the hot side as efficiently as possible.

I would definitely keep the two points below in mind.

1. Thermal feedback loops.
Be sure that the hot side of the TEC is appropriately thermally isolated from the rest of the system. Otherwise, you'll be heating up the box that your widget is inside of that you're trying to cool. This will make the TEC have to work harder to achieve that temperature. This will hit a practical limit and you'll lose your temperature setpoint. You can achieve this by either keeping the heat sink assembly floating or by using a metal with a lesser thermal conductivity as a bracket to interface the two.

2. Airflow
Airflow is key. Keep in mind the dead spaces caused by fans. I've found that the ideal fan placement is on the side and to have a defined air channel around the heat sink so that air can't escape half-way through. All of the air has to go across the entire heatsink.

[ejeffrey]

Make sure the hot side is adequately cooled.  Water cooled if possible.

Insulate the cold side as well as possible.  You have to cool not only the heat generated by the TEC and the load, but also conduction, convection, and radiation heating the cold stage.

Watch out for condensation.  The heat of vaporization is *huge*, so condensing water releases an enormous amount of heat.  Also, if liquid water gets near the TEC element itself it will conduct heat between the hot and cold stages very well.

For a 2-stage solution, you need to make sure that the wires going to the upper stage are thin.  Fat copper wires will form a thermal short-circuit and reduce the effect of the first stage.  You want to balance resistive heating vs. heat conduction.  Likewise, make sure the wires powering the load are as thin as practical, and thermally anchored to the intermediate cold stage.  You probably want to use something like thin magnet wire with a thin varnish coating to heat sink, it is really hard to cool wires through thick insulation.

Don't use PWM or on-off control.  Filtered PWM is fine.

Feedback loops are hard to get right.  Especially if you are trying to get the coldest possible temperature.  The problem is that "cold side temeprature vs current" is stupidly non-linear, and of course if you exceed the current for coldest temperature, it starts heating again as I^2R heating overwhelms the ability of the TEC to move heat.  This makes it hard to get a feedback loop that is stable in all these conditions, and can easiliy be pushed into positive feedback and overload your TEC.  Start off with manual constant-current control.

[Kleinstein]

With a two stage cooler the lower temperature TEC should be considerably smaller (e.g. 1/4-1/10 the size). Depending on the power of the circuit there is an optimum size of the TEC elements: much larger ones will need much more power, as  much of the power is needed to overcome internal heat flow and I2R power loss.

The nominal maximum current is for the case of ideal heat sink and contact - real maximum currents with a finite heat sink will be lower. If much lower (e.g. more than 25%)  this is indicating a poor TEC or poor thermal contact/heat sink.

A 2 W power loss in the circuit sounds like a lot to me - if possible reduce this. It can make cooling a lot easier. A high impedance source does not sound like is needs a lot of current in the FETs.

[sdouble]

I do parallelize 4 fets to reduce the noise. My input is rather capacitive (about 500 pF). I need about 150 mS of gm to get out of the noise. about 12 mA per fet + 20 mA for the second stage amplification.

[Kleinstein]

With the usual low noise JFETs one can get something like 10 mS per mA of current. So one might only need 15 mA to reach a total of 150 mS. However this might need a few more FETs, as they get less efficient at higher current.

Also 20 mA for the second stage sounds like an awful lot of current - I don't think one will such a low noise for the second stage.
Even then 4*12 mA + 20 mA = 68 mA  at something like 5 V would be only some 350 mW and not 2 W.

Cooling takes quite some effort so it might be worth to reduce the power first. The 2 nd stage may not need to be cold at all.

[sdouble]

10 ms per mA , are you sure ?
might be ok up to 2 mA, but not above.
that would require many units and much space that i don't have