(Semi-)DIY thermal chamber, -40-85C

[Myrddin]

Brainstorming about a thermal chamber build, -40-85C (industrial) temperature range and “large” volume:

For several upcoming projects, I will need a decent thermal chamber. As the budget does not allow me to just buy a commercial “table-top” chamber (which are in the 15,000 Euro/16,000 Dollar range), I need to build something (hopefully) equivalent. The overall budget should stay under 2k.

EDIT: To clarify the use case and requirements:

The chamber will be used to check the temperature dependence of single components (photodiodes mostly), small PCB assemblies with and without casing. Sometimes to test whether a system survives a specific temperature (pre-HASS screening), also for battery-powered devices. The DUTs are (mostly) under 200mmx150mmx40mm and weight under 300g, enough power and room for an additional small vibration motor would be great. Good to be prepared for the unknown future and larger devices, but that's not an A priority.

Target test volume would be around 40-60l, the cooling rate is not critical (for stress testing a high rate would be great but again, no high priority). At least on the paper, 500W heating/cooling power is sufficient to reach the 50C/min (HASS “standard”) at a load of 500g aluminum (arbitrary, AL has high specific capacity so worst case).  But the main use case is precise temperature dependence measurements. A 0.5C stability (temporally and spatially) and ~10 minutes to reach this stability from a 1C step would be great.


Currently, two main concepts are under investigation:

• A TEC based system, DIY chamber and compressor assisted liquid cooling
• A compressor cooled system, based on a modified shock freezing chamber

1) The TEC based concept:

The first idea was to build a chamber similar to TiN's design (https://www.eevblog.com/forum/metrology/diy-large-thermal-chamber-for-metrology-tm-testing/), but using a stronger TEC module (~500W, 8x TEC1-12706, water cooled), a ready-made gastronomy insulated cabinet (robust and already has a door with hinges) and for the liquid cooling part a laser/process chiller (these common CW5000/CW5200 chiller e.g. https://www.teyuchiller.com/industrial-portable-water-chiller-cw-5202-dual-water-inlet-and-outlet_p247.html).

The chiller would provide water with a temperature of 5C (controlled) or even lower with anti-freeze. The TEC module needs to cool the remaining dT of 40-45C, which is doable, even with a slight thermal load. With an insulation of 60 mm, the thermal load is at least 20W from (30cmx30cmx42cm inner volume). The TEC1-12706 has over 10W capacity at dT=50C (at 5A). Seems doable. I tried the setup with a small transport box for dry ice and reached -25C quickly (with no load, 10C-15C cold cooling water from a bucket and crappy seals and insulation at the TEC module).

Now, as the complete build with CW5200 cooler, power supplies and control electronics, insulation etc. will cost at least 1k, another route might be better:

1) The compressor-based concept:

Shock freezer (or blast chiller) are used to cool down food to -18C rapidly. They commonly use (nowadays) R290 (propane)  as refrigerant and (easily) reach temperatures down to -40C. The great thing would be that small units are available for less than 1100 Euro/1300 Dollar, and they are fully integrated units. They are usually built as a stainless-steel chamber with circulating air cooling, perfect!
One would need to add a heater (well, that's easy…) and a control system (well, … not so easy).

E.g.: https://www.chefook.com/catalogo/files/abbattitore-CHABB23T-chefline-open-1_3.jpg

The two shock freezers in closer selection have the following (inner) dimensions (and compressor power):
380 mm x 360 mm x 320 mm (500W)
640 mm x 400 mm x 360 mm (750W)

I would currently very much prefer this solution over the DIY TEC solution, but I did not find any information how compressor cooled thermal chamber to the temperature regulation:
The compressor needs to stay turned off for at least 2 minutes after each turn-off and there is no (sane) way to regulate the cooling power, besides regulating the cooling fan. The heater control is easy, both the heater and the heater fan could be regulated from (almost) 0% to 100% smoothly. Does someone have any information on how a good temperature stability is achieved with compressor based thermal chambers? Cooling all the time and heating against it seems excessive. And I cannot see that turning it on/off like in normal refrigerators leads to an acceptable temperature stability, even with a fighting heating PID...

So in the end, it boils down to the following points:

	DIY TEC chamber	Mod. shock freezer
Price	Lower	Higher
Construction	Very DIY, no metal housing	Professional stainless enclosure
Thermal capacity	Limited	Very high
Thermal stability/Control complexity	High/Low	Low/High ?

For me, that is currently difficult to choose, if I see a way to precisely control the temperature with the shock freezer, I will follow this route. If not, I will most probably stick with the “safe” TEC topology.
Happy to hear your thought, maybe you have some experience controlling these kinds of systems.
Regardless which route I go, I will document the build here. I did not find a similar build in the forum, besides TiNs great TEC chambers, so perhaps it is interesting.

Thanks!
 

[tszaboo]

We had an engineer that had the idea to use a freezing unit, with a heater inside as a makeshift temperature chamber. It apparently damaged the rubber seal of the chamber, with expensive repairs. I'm not saying don't do it, but check first if it would survive it, or if it's easy to replace.

[Myrddin]

That was a problem I was considering with the usual freezer units, especially stressing the refrigeration circuits with high temperatures. Good to know that also the rubber seals seems to be sensitive..
Even the small shock freezers are built to handle 12 kg of 90C hot food, so with good thermal protection and control, heating it to 85C should be fine I guess (hope).

[EC8010]

We had a proper environmental chamber, big enough that someone could be squeezed inside. It went from -30C to over 100C and used a compressor and heater. It was the thick end of £20k ten years ago. Measuring its temperature with a PT100 and appropriate DMM logging to computer, you could see ramps 1C peak to peak. Before that, we had a chamber that cooled by allowing bottled carbon dioxide to expand and chill. That was outlawed the moment someone measured the carbon dioxide levels in the lab. And before that, I was told that someone modified a chest freezer to add a heater.

I made a temperature cabinet out of a very tinny Aldi table top oven costing £20 (new). I spent £12 on a PID (Proportional Integral Differential) controller plus Type K thermocouple to replace the horrible thermostat. Rewired all heating elements in series to reduce the power and make the heating more even. I added a computer fan inside the cabinet to stir the air (£1 from ham radio rally), and added a toroidal transformer etc to power the fan (£15). £50 all up, but no active cooling, so all runs have to increase temperature. But it works and gives useful results.

Something to bear in mind about temperature cabinets is that they are never very accurate. You need a fan to ensure an even temperature, so that means your equipment is effectively fan-cooled, which it wouldn't otherwise be. So testing to (say) 40C in a temperature cabinet is not the same as testing to 40C genuine ambient because of the air flow from that fan. Second, hysteresis in the control system is often visible (although my PID controller is quite good on that). It's hard to imagine a compressor-based solution that wouldn't show hysteresis. And when you put something in the cabinet, the fan doesn't work as well at keeping temperature throughout the chamber even. 1-2C error from demand seems to be about the limit for something based on heated/cooled air that can be varied. Obviously, you can hold tighter limits if you don't mind taking a very long time to get there.

The reason temperature cabinets (and some building air conditioning systems) alternate between powerful heating and cooling is to force the temperature to have equal ramp slopes heating and cooling because that allows tightest control. Wasteful of energy, but if you need tight temperature control, then you need equal heating and cooling time constants.

[miro123]

Quite ambitions project. The requirements are not stil not clear to me.
e.g max mass of tested object, Max temp ramp up /down rate. volume etc.
Then you can start with selection of materials and components, and not other way around.
E.g. Polystyrene has glass transition temperature of 100 C . The vicat softening point is just 85C - Can this material satisfy all your requirements?

[mzzj]

I’d probably build ”box in a box” and manage the cooling power by flaps and fans.

Ie freezer running non-stop with insulated box inside. Crude temperature control with air control flaps and fans, fine tuned with heater and pid controller.

[Myrddin]

Thanks, already, for all the comments and information.

I have updated the first post with some more specific information about the requirements and the use case. The setup shown was just a quick proof of principle with the cheap TEC module. Polystyrene will not be used in the final setup, and the shock freezer has a different insulation (rated mostly to 100C-130C continuous).

So a possible solution could be to shield the test volume partially from the cold side (evaporator), decreasing the load on the compressor during heating. This would be the compromise to a box in a box solution. The cold side fan can be part of the PID loop, acting on positive dT. This would reduce the cooling power, but also reduce load on both the heater (does not heat against the -50C evaporator) and the compressor. The cold side barrier could be removable for high dT/dt tests:

EDIT: The black block in the middle is the DUT, seems that all the transparent fills are solid…
EDIT2: Maybe it is beneficial to direct the hot flow from top to bottom (reversed to sketch), needs to be tested…

What would you use as a heater? A common convection oven heating element is quite slow (and very high power…). I guess I will look for a 230V heater which is quite flat (zero cross switching relay controlled). These 24V PTC heaters would add some overhead in terms of power supply and switching the high currents.

[Echo88]

Regardless of expected heat load the TEC path with single stage TECs like cheap 12706 to -40°C isnt doable.
The Tcold/Tambient difference of ca 70K (-40°C, 30°C coolertemp assumed) will make the TECs useless as they only will about zero heat pump power Qc: https://peltiermodules.com/peltier.datasheet/TEC1-12706.pdf
Youd need multiple stage TECs which would provide a bigger allowed difference temp at still tiny cooling power.
Those are way more costly compared to finding a suitable (maybe used) freezer or similar compressor solution.

I dont have a suitable concept to convert a freezer to a nice and at least somewhat efficient thermal chamber for your temp range, but here are my thoughts:
Id try to decouple the compressor from the varying temp (-40°C...85°C) by using a liquid chilling compressor and the cooled liquid would be circulated via controllable pumps into your chamber.
This would increase efficiency and should be safer than risking to increase the refrigerant temp to 85°C, which would most likely violate pressure limits of the system.
I did this stuff about 15 years ago using a DIY chiller, cooling ethanol to about -30°C (AFAIR) and pumping it with Laing pumps.

Heres an example for an industrial chiller: https://www.kaeltetechnikshop.com/verfluessigungssatz-tecumseh-cajn4492yhr-hbp-r134a-220-240v
Maybe there are also standard PC chillers that enable your -40C temp limit, but i doubt it, given their normal CPU heatload requirements.

[Myrddin]

The TEC solution would involve a two-stage cooling: First a CW5200 chiller cools the cooling liquid to ~5C. The rest down to -40C is done by the TEC module. A dT of 45C is feasible even with the cheap TECs. 70C is not, that is correct.

A second cooling circuit with a low temperature cooling liquid, controlled by pumps, is a nice idea, but quite challenging to build on a short timescale. That is a bigger project. I try to use as many off the shelf components as possible.

If I built the refrigeration system by myself (not completely out of scope, we did it before for a project) I would just add an electronic expansion valve (an EEV ). This would allow controlling the cooling capacity inside the PID loop without turning off the compressor (If all components are designed for this unusual state, basically a blocked evaporator). Here again, I think I need to buy as much as possible within the budget to complete the build in a reasonable timescale.

I am totally with you concerning the pressure limits, 85C is ~35bar for R290 (propane) refrigerant. That is most probably over the rating of the compressor or other parts, I am curious how it is managed in the original use case (cooling 12kg of 90C water). I would assume, if you are not turning on the cooling the evaporator would get close to 70-80C. Maybe it is rated to survive this temperature load. Or you have to turn it on. In the two chamber system this would be a solution, only heat if the compressor is running as well. This would prevent the system to overheat. Also one could wire the overpressure switch (which is hopefully installed) to the heater.
 

[Echo88]

Im sorry, seems i completely overlooked the suggested chiller + TEC solution at reduced DeltaT.
Seems quite good and doable with standard components.

[johansen]

500 watts thermal load at -40C with thermoelectric coolers dumping into 5C is only possible with on the order of 1000 watts worth of tec's which make for a 1500 watt thermal load, realistically you need 1 ton glycol chiller.

I would just modify a 1 ton r410 minisplit and try and get down to -40C. The compressor will be running hot.. but because the stator core is in direct contact with the shell, you can put it in a bucket of water.

[dietert1]

Can you separate oven and refrigerator into two units? That will simplify everything and the two units can be used at the same time.
I happened to put a DIY build into an oven at 50 °C nominal in order to cure two component epoxy and afterwards there were some PUR parts with deformation.

Regards, Dieter

[Psi]

Have you thought about humidity control, if you need it, and if you don't what side effects not having it will cause in the system?

[Myrddin]

@johansen: The thermal load will be ~20W from insulation losses + load of the DUT. Most devices will have <<10W thermal losses, definitely less than 50W. With a load of 500W, the TEC solution would be out of the game, no question. But still difficult to decide.

@dietert1: You mean testing the DUT in a refrigerator from room temperature to -40C and in an oven for tests from room temperature to 85C? I like the idea, but I am not sure whether it simplifies the design of the refrigerator part to keep the stability requirements. It would solve the problem with 85C in the freezer …
I tried to implement something similar with the two-chamber design. As long the cooling fan is not active, the heated chamber is (mostly) isolated from the cooling part. This is the best way I currently see for the compressor-based solution.

@Psi: Not really, to be honest. Humidity control in a sense that I can regulate the humidity during operation is definitely not needed for my application. Reducing the humidity to prevent (excessive) condensation on the DUT is more of a concern. The current plan is to flood the chamber with dried air before each run (from an air compressor unit), not sure whether this is sufficient.  “Fortunately” this problem occurs in both implementation, TEC and compressor-based. In the TEC solution, I would need to implement defrosting and condensation collection in a smart way, the shock freezer has it build in.

Thanks for all the input!