-
Now i am wondering whether it may be possible to mod the controller into a linear one by setting hysteresis to 0 °C. This should result in a 0.5 Hz PWM signal that could be fed into an analog integrator with a linear output stage - essentially a supply with remote voltage control.
Will try that when the drok controllers arrive. Already made a chamber with a 240 W used Dr. Neumann cabinet cooler from ebay.
Regards, Dieter -
Already made a chamber with a 240 W used Dr. Neumann cabinet cooler from ebay.
I'd be interested if you can share details on that build! -
Apologies if I have misled people, but I did say in the beginning that it was a simple ON/OFF type controller. I was not referring to pulse width modulation or full proportional–integral–derivative control. The DROK is a super cheap kit with tuning parameters. When I purchased the device, I wanted to spend as little as possible so that I could learn about the characteristics of my thermal control box with the intention of upgrading to possibly a TE Tech controller.
-
No complaints here, the drok is dirt cheap and with its remote control it fits into an automated test setup. It will be useful. The proposed supply voltage control is something they also do in the Arroyo temperature controllers to avoid extra heat in the controller. If that can make the drok cycle on-off-on... every 2 seconds, wonderful. Of course one would replace the relay by a Mosfet and use the relay to switch the TEC between cooler and heater mode. We'll see.
Regards, Dieter -
After receiving feedback, I decided to put the Sensirion inside a closed metal container to compensate for the oscillations and overshoot. Nothing to write home about now since this would be the scenario for most DUT anyways.
I had to make a phone call while the test was underway. The Bluetooth dropped out a few times due to the poor signal broadcasting from the Sensirion in a metal enclosure and it continued unnoticed. Regardless, there are no immediate indications of poor performance at two setpoint temperatures (27.5 and 29.5 °C). A thermal mass is all that it took to dampen the oscillations and overshooting. The noise is about < 0.1 °C. For 16 USD, it is a bargain.
-
Of course one would replace the relay by a Mosfet and use the relay to switch the TEC between cooler and heater mode. We'll see.
I would love to see what you come up with.
-
Although the TEC chamber is still on the bench and in the state of furniture, here some images. Size is 53 cm x 60 cm x 70 cm, so even a HP 3456A should fit inside. There is a suitable linear power supply with 6x 2N3055 and a 400 W transformer. The cooler takes about 10 A at 24 V and its outer fans take another 1.5 A at 24 V. For thermal insulation i bought 3 cm PUR boards.
Regards, Dieter
-
Today i had to cancel the DROK order i made and paid for on April 21st. They did not ship until now and won't tell a date for shipment. And China to Germany would take another 5 or 6 weeks.
Hope to find another solution. In fact last year i made a MSP430 digital controller similar to an Arroyo with a full bridge DAC controlled current output stage and a Pololu step down module modified to deliver the required TEC voltage + 2 V. For the chamber i can make something similar, just with bigger power stage.
Regards, Dieter -
Now i have a PID controller for the Cotek AE-800 i got two days ago. The PID controller reads the Sensirion SHT35 sensor over I2C and operates the COTEK over its 4800 Bd Uart port (labeled RS232 by COTEK). Host connection for logging is via the CDC Uart included in the STLINK programmer on the nucleo board. All fast and easy - no analog except Uart optical isolation. PID cycle time is 500 msec.
The log is from heater operation at 50 °C. Calmed down after protecting sensor from air drafts inside the ceramic tube of the resistor. I think noise is below 0.02 °C standard deviation. During this test the fans of the COTEK remained off.
Next step is to implement the polarity switch for TEC operation and resolve operation near zero output. Then it's ready for the TEC chamber.
Regards, Dieter
-
For regulating a TEC it makes sense to go a bit further than just simple basic PID. There are at least 2 areas one can improve:
1) the TEC is nonlinear in the control current. It is more like a parabola with the max cooling at some point that is usually a little lower (as no heat sink is perfet) than the nominal current. Si this is just a simple function on the output of the PID control.
2) The second is also measure the other side of the TEC. One has 2 ways to treat this and they are effectively equal. One is taking the the other side temperature as a known disturbance treated by feed forward control. The alternative is to inclose the other side temperature in the function to calculate the actual TEC ouput power / heat flow.
A 3 rd point is protection, in case the cooling fails.
-
The PID i implemented as exercise includes the nonlinear heater characteristics. Assume it will be similar for the TEC cooler. And the I2C port is ready to drive a second Sensirion sensor. Those have one address pin.
The Dr. Neumann cooler has a built in over temperature switch. It should connect to the COTEK supply to implement hardware protection. Anyway i think inside the chamber there should be redundant sensors and maybe more fans to avoid temperature gradients.
Regards, Dieter -
Yes, i followed your recommendation and ordered after translating and reading your Russian blog. Mine is from 2018. Those supplies are 3x stronger than needed, so i hope it will last. Other parts were found in our drawers. Did you also make a digital controller?
One step still missing with my controller is an interlock to prevent 30 V output as long as the controller is without USB connection or not in remote mode. The poor SHT35 inside the heater went up to 120 °C once, but it seems to be OK. And this is what the heater exercise was meant for.
Regards, Dieter -
Hope it works well for you. I also thought about it, yet i found it a bit expensive and with this programmable Cotek power supply so little is missing.
The polarity switch is just 2x PVI1050 dual optical isolators plus four high current Mosfets, parts that we have. Maximum loss will be about 2 * 4 milliOhm * 10 A * 10 A = 800 mW. And yesterday i found those nonlinear TEC equations Kleinstein always mentions.
Regards, Dieter -
I didn't calculate like that. For me it's interesting work - a bit like sports. After looking at those Arroyo controllers in some detail, i would like to implement a controller with two sensors on hot and cool side. Also i could look at the Seebeck voltage of the TEC: turn off current for some milliseconds once per second to measure the TEC generated voltage.
Regards, Dieter -
Meanwhile i implemented the Cotek AE-800 startup. After pulling the two default bridges of the inhibit signal on CN2 (blue and white wires) the supply remains inhibited after power-on. Yet the auxiliary output does turn on, so the controller can send the "REMS 1" and "POWER 1" commands whenever it wants to and the main output will remain off until then.
I wired in a third optocoupler driven by the auxiliary output, so the controller knows whether the AE-800 is on/came on and it should initialize the supply before running the PID loop. This procedure seems to work and prevents unwanted initial output from the supply. If the controller stops by pulling its USB cable during operation or turning off the USB host, the supply will continue with the same voltage. Yet the polarity switch will turn off, as it's PVI1050 opto-drivers are powered from USB. So the TEC cooler turns off automatically.
Regards, Dieter -
Tin do you have results for 8508A and 3458A?
-
Meanwhile i implemented the polarity switch for the Cotek AE-800. It supports three state mode:
+, off, -. This simply happens using one of the STM32 GPIO pins in three
state mode. Below the recommended 10 % of 30 V = 3 V limit of the
AE-800 the controller firmware starts using the off state of the polarity switch
to implement a PWM mode for the heater.
Here i have a test log, still using the resistor heater for test. Target temperature
is 24 °C, little above ambient temperature. Now the heater runs at low power.
The diagram shows the heater voltage as blue dots. Due to PWM mode the dots make two
lines at 0 V and 3 V. In order to show what happens the green curve is a low pass filter
of the heater voltage. First the heater runs at an average voltage of about 2.1 V.
After covering up the heater, average voltage drops to about 1.6 V (about half
of previous heating power).
The PWM work-around keeps temperature variations at 0.015 °C (StDev).
The controller also got the second temperature sensor and a PWM output for fan control. As
the fans in the Dr. Neumann cooler don't support PWM control, i ordered another Cotek 24V 4.5 A supply
and hope to mod it for running the fans at 12 .. 24 V. PWM signals are good with optocouplers.
Regards, Dieter
-
Here i have a result for our TEC chamber.
The Dr. Neumann TEC cooler is specified at 240 W. It takes about 10 A at 24 V.
Inner and outer fans currently run on 12 V (half speed).
An Advantest R6581T DVM is operating inside the TEC chamber, so the TEC
cooler is always in cooling mode in order to take out the power dissipation
of the DVM. During the data taking period of several days the inner
temperature sensor of the DVM was 38.5 +/- 0.1 °C while ambient temperature varied.
The TEC cooler runs at voltages of up to 7 V. As far as i understand i could
put another meter into the chamber, maybe even three of them when
using a mains voltage stabilizer with reduced output voltage.
The extrapolated measurement shows that without TEC drive, the inner
cooler temperature will be above the outer cooler, but only about 1 °C
or so. This is good for operational safety: It means that the DVM will
not overheat as long as the fans are running and as long as the TEC is
protected from heating (inhibit voltage reversal).
One disadvantage of the COTEK AE-800 power supply used for the TEC cooler
are its two small fans - they sing like dentist tools. Need to find
a quiet solution. Today i got chamber temperature control depending on
R6581A inner temperature sensor, will show results later.
Regards, Dieter
-
Here i have a log showing successful chamber operation over several weeks. My aim was operating an instrument or a metrology setup at constant temperature. So there is an Advantest R6581T running inside the chamber. The diagram shows instrument internal temperature (LM35 sensor inside R6581T). The orange curve is from an ambient temperature/humidity sensor (SHT35) and exhibits the daytime modulation. Chamber control is based on two more SHT35 sensors, one mounted close to the outer heatsink of the Dr. Neumann cooler, the other one inside the chamber. In period A the inner sensor was close to the inner heatsink of the cooler to get the fastest possible control loop. Then the chamber was off for some days. I had to rewire the sensors with shielded I2C bus cables as they crashed once a week. Also i replaced the noisy fans of the Cotek AE-800 power supply by something quiet: A Sanyo Deiki fan with 3D printed adapter.
In period B the inner sensor moved to the center of the chamber. This improved temperature stability although the control loop became slower. Then in period C i added some fine tuning of chamber set temperature based on the R6581T internal temperature readings. Since then the R6581T has been running for a month at internal temperature = 38.85 +/- 0.014 °C (standard deviation).
Right now the TEC cooler operates at about 2 V with a design limit of 24 V. Inner and outer fans of the cooler run at 12 V, that is half speed. The control loop for the TEC voltage executes twice per second, the DUT fine tuning every five seconds.
Regards, Dieter
-
It is this time of the year again, when we build some more active TEC thermal chambers. I've managed to bring my resistance measurement capabilities to a magnitude better resolution and uncertainty, and this required additional chamber for cross-comparison measurements. Also needed to measure standards at different temperatures, so just one main chamber is not going to cut it anymore. Large chamber that I've shown last year is still mostly happy and working daily, but does require few minor improvements and fixes for discovered shortcomings. That will be story for another day, but today I'm showing lower power mini-TEC chamber built from scrap from megaTEC one.
Base material is same R13 50.8mm thick metallized foam board. Construction is also same, just boring box out of sheets despite some wierd aspect ratio this time.
You can tell that geometry was not my favourite subject in school. I don't care if it's look ugly, all that matters here is functionality and low cost/effort.
It's about a meter long, so deep devices such as Fluke 732A or Fluke 752A would have no problem fitting in. But most of time this chamber will just have passive resistors such as Fluke 742A, ESI SR104, L&N 42**, etc. in it.
Same foam board glue Loctite PL 300 was used to seal/glue walls together.
I have generously applied it on all corners and junctions to ensure airtight connection.
End cap walls for front and rear would be less thick with Foamular NGX panels + styrofoam.
Inner volume of result chamber happens to be 406 x 305 x 924 mm. Just enough to fit two ESI SR104, perfect result.
First victim size check:
Hopefully this weekend I can put 40W watercooled TEC unit, hook up Arroyo TekPAC with Raspberry Pi xDevs.com TECkit python datalogger/controller and get some logging to check stability.
-
Thanks for sharing.
As usual - clear step by step explanation and pictures.
I have one question - how are you going to achieve constant temperature across such big area?
At the moment I'm struggling with this problem.
1. Forced air is kind of double-edged sword.
2. I've tried to increase the thermal mass inside of chamber, using thick metal sheets
3. I consider to use MIMO control - I have no idea what it will bring -
Same as in big chamber and most of commercial chambers, it is forced airflow with internal fans.
Large thermal mass inside of chamber is a problem if you want fast response to control. -
Talking about thermal mass, I doubt there is much lack of such. Just took a photo of contents in megaTEC chamber, all kept at happy +23.0°C during intercomparisons.
-
Here is some data on using two instruments inside a large thermal chamber.
The Advantest R6581T was first and the chamber temperature controller included slow temperature fine tuning of the R6581T internal temperature, with a target temperature of 38.85 °C.
On 2021-12-21, two days after i inserted the K2182A, i turned off the fine tuning. I thought with two instruments it doesn't make sense to fine tune one of them. Since then the internal chamber sensor SHT35 gets used for control, with a target temperature of 23.25 °C. Hourly averages of that sensor show 23.25 with a standard deviation of 0.003 °C, so the chamber controller works well.
The diagram shows temperature logs over several weeks. K2182A internal temperature remains much more stable, as it sits closer to the SHT35 sensor. The chamber shows internal temperature differences, although one of the internal fans has a tube to improve air exchange.
Regards, Dieter
-
What is the meaning of the plot three axis? No axis labels. Did you test without tube? Wonder if that makes any measurable difference
