Small thermal chamber for leakage measurements

[jeremy]

Hi all,

I am looking to build a small thermal chamber for measuring leakage currents of small circuits at various temperatures, and possibly battery ESR or current at low temperatures. I have a few points that I am not totally sure on, so if someone could point me toward some references or similar it would be greatly appreciated. Ideal testing range would be -20C to 50C, but I’m not sure going that negative is realistic. I wouldn’t be against having two chambers, one for negative and one for positive temps. I have had a look at the xDevs articles but they are more related to large air volumes and near ambient temps.

I don’t need fast cycling, I’d just like to be able to hold it within 1C indefinitely. The devices I would be measuring are fairly small, maybe a D cell battery (or an equivalent PCB) would be the largest size. I have an 8V/4A arroyo TEC controller that I plan to use with this, along with a PT1000 sensor. I’d like to use a single TEC. The heat generated by the device in the chamber will be negligible.

My questions are:

1. Is more thermal mass in the chamber better or worse? I am thinking it is better because it will be able to basically low pass filter any temperature fluctuations, but I don’t know if it will cause enough delay to confuse the PID loop of the controller. I was thinking of basically putting a small copper block or heat sink on the inside of the chamber, and maybe a tiny fan for airflow.

2. Is there a standard feed through technique/connector used here in order to keep good thermal isolation? Or should I just be going for tiny wires and some sort of sealant?

3. I was thinking of trying to build it around a small thermos or other vacuum insulated food container, has anyone tried this before (ie, replace the lid with peltier and heat sink)? Main issue is that they all are made with a screw thread to engage the seal, which would make running wires through the lid quite tricky if I wanted to reuse it. Or is this not worth the hassle and I should just use polystyrene foam around a die cast box?

4. Does anyone have a rough idea whether -20C would be feasible with a peltier controller of this scale? Assuming room temp is 25C (ie delta T of 45C) and hot side of the peltier has a nice big fan on the heatsink or one of those AIO PC water coolers

Also, I am aware that humidity will be a problem at the subzero temps (especially where I live!). But I can maybe solve that with some air drying or dry nitrogen or similar.

Edit: one final question; PT100 or PT1000? Assuming the controller supports both. Is there a good reason to go one way or the other?

[TiN]

My first TEC boxes were small, and overall design is easy to scale down.

1. Depends, but smaller the mass, the faster control loop should be. Your 1C stability is very relaxed, so easy to meet that target. For -20C you'd need decent insulation, and some margin for power. One can go even lower, like -40...-50C with TECs but that require care and limits you to only passive or low power DUTs.

2. Small gauge wires and minimal amount of thermal conductive coupling between outer world and inner box is a key. No need fancy feed interfaces unless you dealing with very high sensitivity or want control pressures/humidity properly too.

3. PS foam around box is cheap and easy to experiment with. But feel free to try different ways.

4. I'd recommend AIO watercooler, gives you more freedom and you don't have as much worry about very thin but very good thermal insulator between hot and cold interfaces (since you can insulate whole waterblock/pump head together)

5. Humidity, using dry gases means you need gas tight seals and pressure management. Doable, but adds complexity.

Many arroyo boxes don't support PT1000. So RTFM carefully. I prefer PT1000, less self-heating issues.

[jeremy]

Thank you for your insights! Yes, I checked and I must have misread before; the controller only supports PT100 (but it is not problem for me to change).

[miro123]

I am looking to build a small thermal chamber for measuring leakage currents of small circuits at various temperatures, and possibly battery ESR or current at low temperatures.

They look two different application with different requirements which result in different mechanical and control implementations.
1. PCBs - has small thermal mass and low power - approach with  thin connection wires. Copper slid will serve with dual purpose - equalize the temperature/due to small thermal mass +-5C at different PCB area is normal/,   reduce temeprature fluctuatin due to Aooryo control limitations. -20C in this case is easy achievable. Estimated current consumtion 10mA
2. Battery max current and ESR - High power big heat dissipation high thermal mass, equal temperature distribution. Estimated current consumtion 1..10A

1. Is more thermal mass in the chamber better or worse? I am thinking it is better because it will be able to basically low pass filter any temperature fluctuations, but I don’t know if it will cause enough delay to confuse the PID loop of the controller. I was thinking of basically putting a small copper block or heat sink on the inside of the chamber, and maybe a tiny fan for airflow.

2. Is there a standard feed through technique/connector used here in order to keep good thermal isolation? Or should I just be going for tiny wires and some sort of sealant?

3. I was thinking of trying to build it around a small thermos or other vacuum insulated food container, has anyone tried this before (ie, replace the lid with peltier and heat sink)? Main issue is that they all are made with a screw thread to engage the seal, which would make running wires through the lid quite tricky if I wanted to reuse it. Or is this not worth the hassle and I should just use polystyrene foam around a die cast box?

4. Does anyone have a rough idea whether -20C would be feasible with a peltier controller of this scale? Assuming room temp is 25C (ie delta T of 45C) and hot side of the peltier has a nice big fan on the heatsink or one of those AIO PC water coolers

Also, I am aware that humidity will be a problem at the subzero temps (especially where I live!). But I can maybe solve that with some air drying or dry nitrogen or similar.

Edit: one final question; PT100 or PT1000? Assuming the controller supports both. Is there a good reason to go one way or the other?

1. You Arroyo is using ancient single loop PID control. Higher thermal mass has double edge sword effects - Low thermal mas has higher BW but also higher ripple
2.  Can you test battery ESR with thin wires? - see my comments above
3.  Some eevblog members are using dewar mor their volatge references  - I have no experience. It is in my ToDo list
4. AIO or hetpipes is the way to go - Peltier elements specify max temp difference between heat and cold side
- PT100/PT 1000 - In general RTDs - PT100/1000 has lower sensitivity but higher mid-longterm stability than NTCs. I have no idea does your arroyo supports NTC
===============
My approach
I've looked for existing solutions. There ware two categories
1. General purpose -  low cost /Arroyo price range/easy to use,  fast, you get some reasonable results but it has limited capability.
2. High end - metrology grade - Price-tag outside my budged - huge, heavy , cumbersome to use. For ulitimate performance oil bath is the way to go
https://kambicmetrology.com/claibration-climatic-chamber/
https://www.fluke.com/en-us/product/calibration-tools/temperature-calibrators/fluke-calibration-6109a-7109a
 
Project status - far from publishing for public. Only internally used in my metrlogy projects results are prommising. Tuning and ussing the software is not for bread public - It requires deep knowledge in control theory
some design notes are here, it is partly in Dutch  - https://github.com/mirtcho/LM399/wiki/TC-Oven
1. Multy loop control
- Fast loop - NTC near actuator/Peltier element delay  Peltie->NTC - 50..250msec 
- slow loop NTC + PT100/PT1000 at point of interest - one week stability within +-1mK range
- feed forward channel - sensing enviromental Temperature humidity and air pressure
2. Mechanical
  air tight is the most important parameter in case of single low-BW control loop. It is mostly underrated parameter
  insulation takes only care for the absolute performance - It is mostly overrated parameter
  combination of two makes good design.
  thermal gradient on DUT surface can be as high as +-20 degree. Optimal placing  of temperature sensor can be an challenge

[Fidel]

What stability do you need? An excellent & basic TEC controller we use in the lab is the Ascon TLK33. It accommodates 12-24V, +8A rated (But can handle up to 12A) and can do both heating/cooling.  In our 1cu ft chamber stability is within 0.1C.  The TLK33 outright is around $250USD, but you can find them cheap on ebay in medical surplus used in CTC MN03-01 in a nice case with 24V supply.  The ones used in the CTC MN03 are equipped with RS485. Buy a USB adapter and you can use their free program to both monitor and program the unit or use Python with the Modbus RS485 library for your own custom controlling.

We've gotten down to -60C with a 3x stacked Peltier. It's all about the cooling and insulation.

[jeremy]

Thanks for the tip! Unfortunately those units look pretty expensive for me, they are around $800 USD including shipping to Australia on eBay. But I will keep an eye out for them. Stability for me isn't a big issue, +/- 0.5C is fine.

In your stacking configuration, I assume you used vacuum insulation? I have seen cryocooled image detectors with stacked plates which get to that level, but they have the benefit of being in a vacuum chamber.

[Fidel]

We used a 4" layer of expanding spray foam and aerogel insulation. The hard part is keeping the hot side of the TEC cold. Forced air allowed us down to -30C. A very large copper heatsink partially submerged in an ice bath got us to -60C.

There is an CTC Analytics AG MN 03-01 on ebay right now for $149, though shipping may be the killer to AU. They are handy no-non sense units that work well, though not as efficient as others as it uses an H-bridge PWM for control.

[jeremy]

Yes, unfortunately for the last 6 months or so USPS will not ship to Australia, so most shipping prices are a bit silly.

[ch_scr]

As for controllers, there are these TCB-NE TEC thermostat in a few different variants. While they also implement "ancient single loop PID control" they are AFAIK the cheapest option to get new. Communication is over RS232 and all configuration can be changed on the fly. They will hand out an english pdf with all the commands. I bought one and they do work, build quality seems good.

[openloop]

The cheapest and quickest approach is to use tools you already have. I simply hooked up a multimeter I can read from PC to get the temperature and beefy power supply I can control from the said PC. Plus, like 5 lines of python PID loop to read temperature and yank around power supply's constant current setting. If I remember correctly, I've got to about +-0.5C on the first try .

[langwadt]

buy a freezer, it'll go to at least -18'C, add a heater inside and run it "class-A" for lowish temperatures ?

[alligatorblues]

I built a 36V 12A unit with 3 12V 12A tecs in series. I put 2 sandwiched between (2) 1/4" copper plates, and one on top of that into the chamber. I PUT a 8 pound Al heat sink with 120mm fan on the hot side, and a small 40mm heat sink fan on the cold side. I used a trial and error PWM controller, a DMM with temp, a 900W PSU for 12A @ 36V.

The hot side cooled so well I had to turn the fan off to let it get going, or the cold side would not cool much. The temperature differential is within certain parameters, not absolute, at least with the tec chips I used. But the massive heat sink would in short order heat up a lot, so I had the fan for that. That cost me about 50.00USD, Because I already had the DMM ANd psu .

Putting it together took about six hours. I get temps down to -12C with no insulation. The chamber is about 0.5 cu foot, 1.8" hard plastic electrical junction box for waterproof installations. It fits together like a clam shell. I used foam rubber to seal the holes for temp probe and leads for testing.

[Doctorandus_P]

I built a 36V 12A unit with 3 12V 12A tecs in series.

In itself a single TEC can probably get you to -12, so the temperature itself is not the problem.
If you put three in series, then the last one has to move the heat generated by the first two.
The difference between putting just two in series is probably very small, but is easy to test.

You probably get the best results by putting two of them in parallel, and then put the third on top. (with a heat spreader in between).

[grizewald]

What stability do you need? An excellent & basic TEC controller we use in the lab is the Ascon TLK33. It accommodates 12-24V, +8A rated (But can handle up to 12A) and can do both heating/cooling.  In our 1cu ft chamber stability is within 0.1C.  The TLK33 outright is around $250USD, but you can find them cheap on ebay in medical surplus used in CTC MN03-01 in a nice case with 24V supply.  The ones used in the CTC MN03 are equipped with RS485. Buy a USB adapter and you can use their free program to both monitor and program the unit or use Python with the Modbus RS485 library for your own custom controlling.

We've gotten down to -60C with a 3x stacked Peltier. It's all about the cooling and insulation.

Sorry to necro this thread, but I thought your suggestion sounded like an interesting idea and managed to pick up an MN03-02 from eBay at a reasonable price to try this out for the temperature controlled chamber that I am building.

Maybe it's the auto-tuned PID settings that my TLK33 came up with but I didn't find it particularly stable at all. Mine oscillated up and down over more like +/- 1°C. If you have any settings you could share, I'd appreciate it.
I was also quite disturbed when I measured the current applied to the Peltier element to see that the TLK33 rapidly alternates between applying full power heating and then full power cooling. That really can't be good for the Peltier element!

It looked to me like the TLK33 is just a reversible on/off thermostat which simply applies as much current as the element's resistance dictates at the voltage supplied to it and reverses that current if the set point is exceeded only to switch back to heating polarity as soon as the temperature drops below the set point. There did not seem to be any attempt to control the temperature by varying how much current was supplied to the Peltier element.

I was also quite shocked at how this piece of Swiss made laboratory equipment was put together. Given the new price of these units, I'd expect professional construction which justifies the price tag.
What I found was poor to the point of being potentially dangerous! In reality the MN03-02 is just a on/off thermostat stuck in a generic aluminium extrusion housing together with a Chinese switched mode power supply. There is no innovation or custom made parts, it's all off the shelf. Even worse, the Swiss manufacturer hasn't even followed basic electrical rules when integrating these off the shelf parts. For example, the power supply used is a "Skynet SNP-Z10 Series" which is a 24V supply capable of delivering 4.5A constantly and 5.4A peak. Here is how they have connected it to the TLK33:



The power supply has six Molex 5045 series output pins, three for +24V and three for the ground. CTC have used just one of each. The problem with this is that the Molex 5045 connectors are rated for a maximum current of 3A per pin! This breach of the permissible load per pin will cause the pins to overheat and the contact resistance will start to rise making them even hotter. Eventually, this can lead to a fire.

The other breach of safety rules is in how the connections have been made to the screw terminals on the TLK33. All the wires are stranded copper. They are all soldered at the ends which are inserted into the screw terminals.



In some cases, two wires have been soldered together and then inserted into one screw terminal. This is a dangerous way to use screw terminals. The wires should never be soldered! Instead, they should first be inserted into a correctly sized ferrule and then crimped with the correct tool before being clamped in the screw terminals. The reason it should be done this way is that solder is relatively soft and if you screw soldered wires into terminals that are repeatedly heated and cooled, the wires will eventually loosen and walk out of the terminals due to the repeated expansion and contraction caused by the temperature changes. This can also lead to a fire risk.

I really was not expecting a piece of lab equipment made by a seemingly reputable Swiss company to be so cheaply engineered and poorly assembled!

[cellularmitosis]

Here is a simple little rig I made for resistor tempco measurements

https://www.eevblog.com/forum/projects/mini-tempco-characterization-rig/

[grizewald]

Here is a simple little rig I made for resistor tempco measurements

https://www.eevblog.com/forum/projects/mini-tempco-characterization-rig/

I remember that one; very nice indeed.

My chamber is slightly larger at six litres. Since TiN and Marco Reps highlighted the excellent Arroyo controllers, used prices have gone way outside of my budget, hence my search for an alternative.
What I'd really like is one of the boards pictured in the oddly corrupted attachment.

It's serial programmable, functions as a variable DC current controller and is very configurable. The only problem being that they cost 300 USD before tax and shipping.
The final price in my hand would be closer to 400 USD which kind of stings. In the end, I might just have to stump up the cash.
First though, I thought I'd give the TCB-NE controller, mentioned by ch_scr upthread, a try. One of those is enroute to me for the much less painful cost of 122 USD all-in.

[grizewald]

Great CNC porn! I wish I had something like that at home. I don't think my neighbours would though!