DIY TEC controller

[mimmus78]

I was thinking to build a DIY TEC controller to help me with my metrology volt nut experiments but then I found this project that it seems quite up to the my needs and almost a turnkey solution.

https://hackaday.io/project/11218-i2c-tec-controller

What do you think? Worth a try don't you think?

[SvanGool]

There is at least one disadvantage, if you ever want to run it on batteries: a PWM driven TEC is not so efficient as an analog driven TEC.
Another one could be, that you don't want any heavy switching currents in the neighborhood of very sensitive circuitry.

[mimmus78]

So we need an analogue solution ...

There is at least one disadvantage, if you ever want to run it on batteries: a PWM driven TEC is not so efficient as an analog driven TEC.
Another one could be, that you don't want any heavy switching currents in the neighborhood of very sensitive circuitry.

[ap]

There are TEC controller hybrids being sold e.g. on the bay US. Also, an option may be used controllers from e.g. Thorlabs or ILX.

[kony]

So we need an analogue solution ...

Or simply filter the chopped PWM drive output to reasonable level of ripple (<10%).

[2N3055]

Datasheet for ADN8831 mentioned in that document mentions "no more than 0,5% ripple" after filter...
So more than good enough for TEC use...

[mimmus78]

Is it me ... or datasheet of ADN8831 is in chinglish?

http://www.analog.com/media/en/technical-documentation/data-sheets/ADN8831.pdf

"For best stability, the thermal sensor is to be closed to the object."
"Adjusting the PID network optimizes the step response of the TEC control loop. A compromised settling time and the maximum current ringing become available when this is done."

[tszaboo]

http://www.linear.com/product/LTC1923

Typical temperature setpoint accuracy of 0.1°C is achievable with the LTC1923. Adding an instrumentation amplifier front end allows setpoint stability of 0.01°C.

Should be easy-peasy.

[mimmus78]

So it seems worth a try, I really like this because I can connect it to my raspberry that already run GPIB stuff for the various instrument.
I will get this project pcb when I order pcbs for my next project. Let see how it works!

[bertik]

Oh here I can throw in my design, recently finished:

http://hololaser.kwaoo.me/electronics/Arduino_TEC/Arduino-TEC.html

BTW an analog controller is never a efficient as a PWM controller (assuming that the
PWM signal is filtered like for a DC supply). Actually to avoid noise this is mandatory for
critcial applications anyway.

[kony]

Not *that* mandatory. You can usually synchronise clock of such offending noisesource to sampling of data aquisition ADC, which makes the spurs disappear in the measured signal, as it aliases exactly on Nyquist zones boundaries (worst case adding some DC offset to the signal - but if you are into precision measurements, you are very likely to have the measured value riding on AC carrier signal anyway).

[Kleinstein]

The ADN8831 combines switched mode regulation for the supply voltage and a linear more MOSFET for the final adjustments and ripple reduction. So in principle this could give a good efficiency and very low ripple. However the final possible EMI problems depend an the construction.
The regulator itself is analog and thus adjustments of the PID loop might not be very easy and limited in range - it could be a problem with a very slow system. Also operation of the TEC near it's limit (e.g. more than 70% of nominal current) is not good, as linear operation is assumed.

For a large, relatively slow system a µC based regulator is likely better. This could include three more improvements over the simple analog PID regulator:
1) proper anti windup, to limit overshoot on startup or operation near the edge.
2) correction for the nonlinear current to power relationship of the TEC
3) use of a second sensor on the other side of the TEC. This can improve response, as the path from the hot side of the TEC is a major path for external disturbance. It is also highly recommended for protection against overheating.

[Andreas]

What do you think? Worth a try don't you think?

But only if the area to cool/heat is very small. (Like a laser diode).
1-3Amps is way too little for typical cigarette to cigar box large devices.
In other cases the peltier with its heat sinks is way too slow.

It is also highly recommended for protection against overheating.

Even cheap Peltier car cooler boxes use a mechanical thermostat switch.
I would never trust a processor alone.
How will you explain your fire assurance that you have trusted a microcontroller?

I have a different strategy:
use a peltier car cooler box. (50-60W)
Let the cooler run most of the time from 32 deg C down to 10 deg C and up to 15 deg C.
For the rest use simple heater foils 2*12W (with reduced voltage) on a aluminium plate to heat for fast temperature control.
I usually also use 2 temperature sensors: one at the aluminump plate (fast reaction P-controller)
 the other at the place of interest to do fine adjustment. (I-controller).

with best regards

Andreas

[Conrad Hoffman]

That control method sounds crazy at first, but it's exactly the way environmental chambers work- run the refrigeration full tilt, then adjust the outlet air termperture with very fast heaters. Works well, but it's inefficient. TECs are absurdly inefficient too, but they get the job done. I've never ordered from these people, but the deals seem to be good-
http://www.banggood.com/search/thermoelectric.html

[mimmus78]

So basically you stick a TEC to cool the outside box of the chamber and than you control inside temperature by circulating air over a heating resistor inside?

[Andreas]

some pictures are here:

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

1)
the TEC cooler box (under the table)

2)
the heat spreader (aluminium sheet) with the heater foils. (car mirror heaters).
I usually use a shoe box to put the heat spreader together with the D.U.T. in
to avoid direct air draft / dripping water by the cooler fans.

with best regards

Andreas

[Kleinstein]

TEC efficiency gets considerably better if you only use the current needed to get the required temperature difference. So it can really help to reduce the power of the TEC.

They can also be rather fast in reaction - it is mainly the heat transfer inside the chamber that sets the speed. The TEC itself can heat / cool in seconds.

The rather high thermal conductivity across the TEC is one weakness, as it gives an easy way for external changes to the regulated side. Using also a sensor for the outer side of the TEC allows for accurate control, as one already measures the main path of external disturbance.

[mimmus78]

So I tested my 120W TEC with 12V and 6A.

It went down to -10°C and up to more than 40°C from ambient temperature in pretty reasonable time.

The TEC was attached to one of this Hammond alluminum case 14x10x8 cm very poorly insulated.

So this 60W~ of power are more than enough for my small thermal chamber.

Anyway as for safety I will use thermo mecanical cut off switch (the same I used in my resistor heater) and maybe also a fusible switch as last resort and third level of redundancy.

I was also planning to use fiber glass as insulant and all not inflammable materials for the external case. Hope this will be more safe than the car fridge ...

Inviato dal mio Nexus 6P utilizzando Tapatalk

[TiN]

Even 20-40W TEC will be just fine, as for metrology needs you just need power to maintain thermal differential, and our DUT resistors or references does not generate much power to dissipate/cool.
Based on readings from 2510, TEC module barely draws 10W to maintain +50c with +23c ambient. Also be careful going below ambient temperature, as humidity and condensate can easily ruin your DUT.

[Echo88]

What would be the most cost efficient method to regulate a TEC in a winecooler-chamber to get <0,050°C stability? I thought about using a industrial PID-controller, which has a 4-20mA-Output (used, from ebay for little money). I would convert that current to 0-10V and use it to control a suitable power supply (dimmable Meanwell-supply for example) for the TEC.

Question is, if the controller is fast enough to stabilise the winecooler when i attach the temperature sensor on the cold side of the TEC.

[Andreas]

a TEC in a winecooler-chamber to get <0,050°C stability?

This stability without a oil bath with stirring oil?
you are kidding. 

With best regards

Andreas

[tszaboo]

a TEC in a winecooler-chamber to get <0,050°C stability?

This stability without a oil bath with stirring oil?
you are kidding. 

With best regards

Andreas

I linked a controller what supposed to do that...
I guess all you need is a big bag of thermal mass in it, and a PT100x to measure it correctly.

[Kleinstein]

How good the temperature stability is, depends on the temperature changes on outer side of the TEC. If this temperature is not changing fast (e.g. a massive heat sink with some thermal mass) a standard PIC controller could be good enough. It is more the thermal setup than the regulator that sets the limits. I don't know how good the typical wine coolers are in this respect. My guess is that a really well tuned PID regulator might be just good enough if the external temperature is not changing fast.  Things can get tricky at low temperatures, when more than about 50% of the maximum cooling is needed, as the system starts to get nonlinear and efficiency gets lower and thus more heating to the cold side. It get much more difficult if there are active heat sources inside the chamber - this is something that might require a stirred bath.

Having a second sensor on the outer side of the TEC too can react considerably faster and better. Actually measuring the temperature difference over the TEC could also reduce variations in the heat flow by some thing like a factor of 10 (possibly more if the adjustment is really good). In this case something like a small µC board and specially programmed µC would be the way to go. A PID controller that support feed forward from a second sensor would also work. As an approximation (linear part of TEC and thus small currents) one might use analog circuitry for the 2 nd sensor (e.g. add some of the sensor signal to the 0-10 V control).

For the power stage, a dimable power supply could be a problem, if one needs really low voltage and changing the direction of current flow. So one might need something like a relay to switch polarity and something to include this in the regulator.

[Andreas]

Hello,

who wants a "setpoint stability" of 0.1K (as offered in the linked data sheet)
when a actual stability of 0.05K is required?

When I do a temperature controller for a chamber and have 25.0 deg C at the sensor.
The DUT may be 2-3 deg C above due to self-heating
(45 mW for a voltage reference with 3mA*15V and a package of 80-150 K/W.)
And at different points within the chamber you also have some deg C temperature difference.

with best regards

Andreas

[Echo88]

If youre referring to the LTC1923: it states 0,01°C-setpoint stability, not 0,1°C. Depending on cooling/heating-mode even better 0,002°C stability is achievable: http://cds.linear.com/docs/en/application-note/an89.pdf

Of course the absolute temperatures of each module (DVM, reference...) in the thermal chamber doesnt really matter, as long as sufficient temperature stability ensures reliable measurements.