Cooling for thermal stabilisation, instead of heating?

[RoGeorge]

- Ovenized voltage references or oscillators are standard practices against thermal drift, but why not cooling, so to get lower noise?
- Peltier cooler are affordable, but all I see is heating.  How come that everybody is heating the references instead of cooling?
- Don't have any experience with metrology grade equipment, but in theory would make sense to cool instead of heating, or am I missing something?

[NaxFM]

Well, never thought about cooling honestly...
It does makes sense, but I guess we don't do it just because it's harder than heating.
Electronics components produce heat, each  in a different amount, so probably it would be easy to have hot spots inside the chamber. Also it would require way more energy and air circulation, since to produce cool air you also need to make the air on the other side hotter. Hot air that must be evacuated fast and without disrupting the thermal equilibrium of the chamber.
If someone like fluke does not use cooling for its World class standards like the 732, I guess they know it isn't worth it.

[wraper]

Well, I can think several reasons not to do it. Water condensation, much larger power consumption, way more complicated.

[Microdoser]

To make one thing cold, you generally have to make another thing hot. You now have a hot thing and a cold thing in the same device, which adds other problems with temperature differentials, heat removal and, I'm sure, more.

If you just make something hot, then you just insulate it, and you're done.

Always choose the simplest solution that actually fixes the problem.

[MegaVolt]

For the thermal noise of a resistor, to reduce it by half, you need to reduce the temperature by 4 times. Those. 293/4= 73.25 K (-200 C). This is a difficult technical task.

Lowering by 20-30 degrees gives little gain.

For semiconductors, the situation looks better. But still, the cold causes condensation and it will also have to be fought.

[dietert1]

Some people are using peltier elements with good results. It's easy.
The idea isn't really cooling but using the peltier element for cooling and for heating. If a device is to be used in a lab environment of say 23 °C average temperature and it produces a temperature rise of 7 °C by self heating, you can adjust the temperature of the peltier "oven" to 30 °C.
Then at 23 °C room temperature the peltier element will do nothing. At 20 °C it will be heating a little, at 26 °C it will be cooling a little. Power consumption will be small, definitely less than permanent heating. Those are realistic numbers for a metrology lab.

Regards, Dieter

[oz2cpu]

EXACTLY Dieter
if the wanted regulated temperature is close to normal indoor temp,
you will use much less power, over a heated system,
and you get into regulation, instantly, assuming indoor temp is very close to the wanted..
but we people already regulate our indoor temp, so we humans feel good,
so this means indoor temp will not change a big deal.

[Conrad Hoffman]

IMO, even if the device is a few degrees below ambient, fighting that film of water will be an issue if the dew point is high. I think you'll need a sealed enclosure with dry nitrogen and desiccant.

[dietert1]

Yes, the peltier "ovens" i tested are with a sealed aluminium box and a small bag of desiccant inside. Instead of connectors i was using flat cables similar to the Fluke 732B, with two component epoxy as seal.
With voltage references the disadvantage of low temperatuer operation is slow aging, i mean a long initial drift, maybe two years. With LM399 running at 80 or 90 °C that happens much faster and i think already after half a year they assume small predictable drift.

Regards, Dieter

[Kleinstein]

The usual peltier elements have quite some thermal conductivity. So even if the rest of the system is well insulated, the usual peltier elements need quite some power to sustain a temperature difference.  It needs careful planing to find a suitable small peltier element that is not too large and still powerfull enough. The extra thermal link still provides a relative fast path for disturbance from the outsider.
A heat only oven can use good isolation and this way keep faster external temperature changes out and ease on regulation.

Condensation and even just getting a RH higher than some 70% that increases the chance to get surface leakage is another problem. For leakage a temperature some 10-20 K higher than ambient is good as it ensures RH of <50% and thus usually no surface water layer and less surface leakage.

Active cooling is mainly used for sensors (e.g. IR ) that really need the lower temperature.

[mawyatt]

Some people are using peltier elements with good results. It's easy.
The idea isn't really cooling but using the peltier element for cooling and for heating. If a device is to be used in a lab environment of say 23 °C average temperature and it produces a temperature rise of 7 °C by self heating, you can adjust the temperature of the peltier "oven" to 30 °C.
Then at 23 °C room temperature the peltier element will do nothing. At 20 °C it will be heating a little, at 26 °C it will be cooling a little. Power consumption will be small, definitely less than permanent heating. Those are realistic numbers for a metrology lab.

Regards, Dieter

In 1980 we used a Peltier device to cool and heat a special coated Alumina target surface for IR Spectral Radiometry instrument (XM21) calibration. We needed to "servo" the surface to ambient + delta and ambient - delta and a couple absolute temperatures around 300 Kelvin. Worked really well, although closed loop system time constant as  fairly long.

Best,

[KT88]

Although using a Peltier cooler around roomtemperature sounds intriguing, comparing it with paralleling multiple references the effort seems to be at least questionable.
For example 2 References in parallel result in an improvement of 3dB, 4x gives 6dB.... This is in particular viable with lower cost options.

[eplpwr]

I'm in the process to try cooling. However, not for Vrefs but for everything JFETs, like discretes and ICs. There are a lot of great metrology uses for JFETs, but the performance suffers badly from heating, self-heating as well as environmental.

I'm aware of the side effects of cooling when RH is high, so there will be a rather complicated solution to keep some JFETs cooled - and dry, but it's worth a try. 

[KT88]

Cooling a JFET is a different story as the Gate Leakage can be reduced significantly: https://www.researchgate.net/publication/245114771_Some_investigations_of_the_JFET_gate_leakage_current_as_a_function_of_temperature

[MK]

Gate leakage will be improved but if you choose the wrong temperature band then G-R (generation-Recombination noise may bite you badly. I saw a noise vs temp curve a few years ago, it was up and down like the Alps!

[fluxgate]

The 32 Bit ADC fromm CERN HPM7177 uses peltier element for Temperature stabilization.

[AVGresponding]

Josephson junction voltage standards have to cooled rather aggressively... not exactly a practical proposition for a bench meter!