manufacture an oven for measurements in a temperature controlled environment

[IRFP]

Hello,
I am looking for tips for designing an oven for measurements in a controlled environment.
The idea is to divert an insulated beverage box, see photo.
Maximum temperature 50 ° C, controlled by PID regulator and TC K or J.
For the heating resistor, I think to use a 3D printer aluminum tray.
Thank you for your advice

[Conrad Hoffman]

The big question is how good do you want it? That should work great for 1 C or so. You might want a very small fan to circulate the air for better uniformity. If you want to go better, two enclosures/ovens might be necessary. There's also the question of where to put the sensor. Best response time will be had if it's on the heater plate. Most accurate control will be had if it's on the DUT. I've potted sensors in small aluminum blocks with neo magnets for attachment. The problem with putting the sensor on the DUT is you need very long time constants for the PID to prevent low frequency oscillations. I've done 0.01C that way in a single (large) box, but with a thermally conductive DUT (aluminum or steel) and very long stabilization times.

[Kleinstein]

For accurate measurements thermocouples are not a good choice, as they need a 2 nd sensor for the "cold" side. The more practical sensors are PT1000, NTCs, temperature reading chips or just diode junctions. I have done temperature control to 0.01 C stability with thermocouples, but unless really needed for high temperature and small size, I would avoid them.

I don't think one would need a second enclosure to get good temperature control. If at all this would be more like having a 2 nd smaller metal box inside to get a homogeneous temperature and faster response. There is a general trend that a large oven will be slow, just as it takes longer for the temperature to equalize inside. Even of one gets a metal box to respond relatively fast - it can take quite some time for the DUT to reach the temperature. Depending on the DUT a fan can really help.

For heating I would also consider using transistors as heaters - so no extra heat loss in external control hardware.
For a larger system several separate heaters and control loops can help to speed things up.
As an oven is usually relatively slow the control is usually easier to do digital (e.g. µC or even DMM+PC) than analog.

[Kean]

I'm working on something similar right now - both heating and cooling, but also monitoring how certain batteries change behaviour over temperature during charge.

For a simple and reasonably accurate temperature sensor I'd suggest one (or more) LM34CAZ.

If you're set on using a TC, then type T will give better resolution in your limited range, but you may need to implement NIST corrections if you don't get a TC interface that does it as the type T means the usual TC approximations are not great.  And you may need a cold junction sensor like the LM34 anyway.  A MAX31856 or similar solves those problems, but the LM34CAZ is so much easier.

See here for a good comparison of pros & cons of different temperature sensors
https://labjack.com/support/app-notes/temperature-sensors

[Andreas]

Hello,

My setup (with a TEC controlled car cooling box so that I can cool + heat) is described here:

https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg462298/#msg462298

The idea using a 3D printer aluminum tray is not bad (especially since the heater foils that I use have gone up in price by a factor of 5).
They are usually designed for higher temperatures (>100 deg C) so you will need to use a rather low-ish power supply voltage.

And a over-temperature switch (mechanical thermostat) would also be a good idea.

For my car cooling box I use ~12W to heat up to 50 deg C.

As temperature sensor for a ~30 deg C range I usually use NTCs (can be linearized by a pull up resistor).
One sensor is mounted directly at the heat spreader to control the temperature for temperature ramps.
If I need a stable temperature at the DUT I use a 2nd NTC and trim the setpoint of the heater-NTC so that the final temperature at the DUT is met.

with best regards

Andreas

[Echo88]

Dont know if youve already settled on the mentioned box, but may i suggest the following:

https://pasteboard.co/ImkkFok.jpg

Winecooler Acopino BC48 ~100€ or cheaper (standard Keithley-case doesnt fit inside but 34465A does)
Depending on heating only via resistor or using a peltier module the following gear would be usable, depending on budget:
TEC-controller: LDT5910B or Arroyo 5235 from ebay.
Only-Heater-Controller: Conductus LTC-10/20. Very nice considering stability, autotune and used price.

Maybe its of interest for other users who dont want a DIY-controller.

[IRFP]

Hello everyone and thank you for all your answers.
I'm considering using a fan indoors, however, I'm afraid this will generate disturbing noise.
I just get the PID regulator, see picture of the model.
For the probe, it will be TC J or PT100, to be defined.
A safety regulator will be integrated.
To supply a switched or linear power supply?

[IanJ]

I made a small chamber using kingspan, a couple peltier coolers which are controlled to either cool or heat on a switch. Been meaning to add control at some point.
Can cool 12degC below ambient, and will heat quite a bit albeit never tried its limits.
The peltiers have small fans on the inside which also helps move air.
Kingspan soooooo easy to cut and build. I used metalized tape to hold it together.

Ian.

[Echo88]

Such a PID-controller produces an ON/OFF-ouput, no nice smooth analog output for low EMI.
Switching power supply dont belong in a voltnut-lab/you would spend your time hunting for mysterious voltage offsets in your measurements due to those SMPS.
If youre really aiming for high resolution/low noise-measurements then such simple PID-controllers will give you headaches.
Thats at least my experience after having dealt with them. 
Ive gone from such PID-controllers + SMPS, to PID-controller + linear supply, to the mentioned Conductus LTC-10 to real TEC-controller as LDT5910B and Arroyo 5235.
I remember that i could see the errors due the on/off-control-loop from the PID-controller very well and that it was very annoying to use it with a PC to get temp-curves running via controller-Analog-In/Modbus.
If you only want to use heaters id suggest getting a LTC-10, costs ~150$ used and youll be lightyears ahead of the cheapest PID-controller + SMPS-solution and youll love it for its many functions, GPIB-connectivity and stability.
Thinking back i wasted much time, instead of just investing a bit more money for good gear and be done with it.

Sorry if thats not want you want to hear, but i think its better if you can avoid the trouble which ive gone trough.
A fan doesnt add much noise (afaik), but you should remember to shield youre measured device against air drafts.
Maybe wrap it a bit in paper for example, that way the measurement box gets thermally equalized well and you avoid air drafts at the critical device.

What do you want to measure exactly?

[IRFP]

Hello,
To answer ECHO 88, I will make precision resistance measurements and other reference sources of tesion. The power supply will be of linear type, control in static PID of the heating plate, I made according to the regulator ...
Another question relates to the connection for measurements, between the inside and the outside, which type of connectors to use?
Thank you

[notfaded1]

Such a PID-controller produces an ON/OFF-ouput, no nice smooth analog output for low EMI.
Switching power supply dont belong in a voltnut-lab/you would spend your time hunting for mysterious voltage offsets in your measurements due to those SMPS.
If youre really aiming for high resolution/low noise-measurements then such simple PID-controllers will give you headaches.
Thats at least my experience after having dealt with them. 
Ive gone from such PID-controllers + SMPS, to PID-controller + linear supply, to the mentioned Conductus LTC-10 to real TEC-controller as LDT5910B and Arroyo 5235.
I remember that i could see the errors due the on/off-control-loop from the PID-controller very well and that it was very annoying to use it with a PC to get temp-curves running via controller-Analog-In/Modbus.
If you only want to use heaters id suggest getting a LTC-10, costs ~150$ used and youll be lightyears ahead of the cheapest PID-controller + SMPS-solution and youll love it for its many functions, GPIB-connectivity and stability.
Thinking back i wasted much time, instead of just investing a bit more money for good gear and be done with it.

Sorry if thats not want you want to hear, but i think its better if you can avoid the trouble which ive gone trough.
A fan doesnt add much noise (afaik), but you should remember to shield youre measured device against air drafts.
Maybe wrap it a bit in paper for example, that way the measurement box gets thermally equalized well and you avoid air drafts at the critical device.

What do you want to measure exactly?

I decided to take your advice and get a LTC-21 I found... it wasn't easy to find (it seems the company changed hands as I found it under Neocera company) but I found the manual for it in digital format.  Thanks for the tip.  Here are the stats:

Display
Sensor data may be displayed in units of Kelvin, Fahrenheit, Celsius, Volts or Ohms.
• All displays are six digits plus sign, a floating decimal point and a units indicator.
• Time averaging may be performed by operator selection of intervals between 0.5 and
16 Seconds.
Accuracy of Sensor Measurements
The accuracy of linear resistance sensor measurements, including Platinum, Carbon-Glass
etc. is 0.1% within the resistance range of 100Ω to 1KΩ , 0.25% from 1KΩ to 10KΩ and
1% between 10KΩ and 30KΩ.
The accuracy of diode sensor measurements is 0.05%.
The accuracy of sensors using the variable ac constant voltage biasing is given in the
following table:
Excitation Accuracy Resistance Range
1mV 0.1% 1 Ω to 1 MΩ
320uV 0.1% 1 Ω to 500 kΩ
 1% 500 kΩ to 1 MΩ
100uV 0.1% 1 Ω to 150 kΩ
 1% 150 kΩ to 1 MΩ
32uV 0.5% 1 Ω to 50 kΩ
 2% 50 kΩ to 500 kΩ
 10% 500 kΩ to 1 MΩ
10uV 0.5% 1 Ω to 15 kΩ
 2% 15 kΩ to 150 kΩ
 10% 150 kΩ to 1 MΩ
Table 1 Sensor Accuracy vs Sensor Excitation
All accuracy specifications are valid over the rated environmental temperature range.
Range of Temperature Measurements.
The instrument will measure and display temperatures in the range of 0K to 800K. (That's -459.67 F to 980.33 F)
The actual temperature range for measurement is, of course, dependent on the type of sensor
used.
Temperature Resolution
The temperature resolution of the LTC-21 is 1 part in 512000.

Bill

[notfaded1]

Another question... does anyone have a source for the cables for these Conductus / Neocera temperature controllers?  The main ones are basically:

• Analog Output Connector Assembly (AMP, 4-pin)
• Sensor Cable with mating connector, 5 meters (Lemo Inc. FGG.1B.304.CNAD52) - this is a 4-pin Lemo
• Heater Connector Assembly (Circular, AMP, 7-pin)
• Relay Connector Assembly (Unsure it says Circular, 8 pin)

I wish there was a kit of these you could buy but I'm thinking I may have to make them...  I wonder how many people use these anymore... seems like could be a lot since they made a lot but who knows?  No one seems to sell the temp controllers with the cables.

The serial db9 and GPIB are obvious.  Also sorry if this is off topic.

Bill

[notfaded1]

If anyone else is ever interested you basically have to make your own cables because Conductus which became Neocera stopped support of the temp controllers and passed it over to a company called 1st Quality Electronics see:  http://neocera.com/ltc/  They exhausted their supply of parts a few years ago they told me today.  These are great temp controllers but be prepared to make some cables... not like we haven't all done it before but that's the story.

Bill

[IRFP]

Progress of the converssion project of a beverage cooler in a thermostatic chamber.
Improvement of the insulation and installation of a heating block.
The heating block is made with a ceramic resistance of 72 W mounted on a heatsink of old graphic card.
On the top I installed a modified TESTO datalogger.
The control unit is finalized
The stability of the system must be tested.

[TiN]

Second life of 7900GTX cooler, nice 

[IRFP]

Hello TIN,
Exactly this graphics card 
The dissipation is very effective and uniform