2
\$\begingroup\$

I'm building one of those DIY reflow ovens from a toaster oven, a K-type thermocouple, a MAX6675 breakout board, a SSR and an Arduino.

I'm having a lot of trouble getting a reliable temperature measurement from the MAX6675. With the Arduino powered (through a USB connection) but the mains part of the circuit disconnected, I get a good temperature measurement. But as soon as I plug the mains in, it goes haywire. It's not even particularly predictable; most often, it just reads zero, but sometimes it reads about 750 degrees, other times it wanders up and down between zero and a hundred degrees or so.

It's a bit hard to tell, but it seems that just plugging the earth pin of the mains circuit in is enough to cause this. I guess I have a noise problem, and since it seems to be the earth connection that does it, I guess it's a ground loop.

The thermocouple is a screw-in probe type, which I've insulated from the body of the oven using kapton tape. I've connected the braided shield to the Arduino's earth; on the scope, this improves the noise situation considerably, though it doesn't seem to make much difference to the measurement. Initially I had the MAX6675 mounted fairly close to the SSR, but have since moved this, again without it making much difference.

What else should I try? Would connecting the arduino's earth to mains earth be a good move? (The Arduino is being powered from a laptop that probably doesn't have a very good earth).

Thanks for any suggestions!

Update

FWIW, here's what the 'scope shows between T+ and T-:

Time- and frequency-domain plots of thermocouple input

The peaks every 40ms look fairly nasty, but it's not immediately obvious to me what would cause them. There is a solid state relay nearby, but it's just off at this stage, so wouldn't expect switching noise from it.

\$\endgroup\$
1

3 Answers 3

3
\$\begingroup\$

If your laptop is powered via a two-pin AC outlet plug then there will be no earth. In addition to there being no earth, the "0 volts" wire of the lap-top will be rising up and down to the tune of several volts p-p at the frequency of the laptop's charger. It, almost certainly will be a switching power supply running at about 100 kHz.

I would recommend that you do as you suggest but as a quick interim, connect a wire from the Arduino 0 volts to the body of the oven.

\$\endgroup\$
5
  • \$\begingroup\$ The laptop is powered from a 3-pin plug, but whether that translates into the USB ground being connected to earth is more than I can say for the moment. Looks like I need to break out the mains earth wire into the control box and establish a good earth as a first step. \$\endgroup\$
    – Tom
    Feb 14, 2017 at 9:30
  • \$\begingroup\$ Just try a wire between the arduino 0V and the metalwork of the oven. \$\endgroup\$
    – Andy aka
    Feb 14, 2017 at 9:43
  • \$\begingroup\$ It was easier to break out the earth lead next to the arduino, which I've done. So now the mains earth wire is connected to both the arduino ground and the shielding of the thermocouple. I've also added a clip-on ferrite bead, though I'll freely admit to magical thinking on that point. The result is an improvement, but it's still not great - at room temperature, plugging in the mains causes the measured temperature to drop from 19 degrees to 12 degrees. It's usable, but not wonderful. \$\endgroup\$
    – Tom
    Feb 14, 2017 at 15:56
  • \$\begingroup\$ Ok, try 100 nF across the input to the ADC you are using. \$\endgroup\$
    – Andy aka
    Feb 14, 2017 at 18:29
  • \$\begingroup\$ I've run out but have some on order. I'll report back. \$\endgroup\$
    – Tom
    Feb 15, 2017 at 17:57
1
\$\begingroup\$

It seems your CM noise coupling is too much. Standard reduction methods include high mu CM chokes and shunt caps to reference low impedance local 0V analog gnd.

If that gnd is floating then power can be carrying CM noise and another CM choke with cap to AC gnd to further increase impedance ratio to reduce noise levels.

Additional active shielding is necessary when above is insufficient grounded with sheath at ADC input only.

Then when this is not enough , good CMRR Op Amp with 50Hz notch filter can remove >60dB if CM is within acceptance range.

The trick is not to guess but try to make it worse then identify the source coupling, conducted and/or radiated then design the method of attenuation by impedance ratios and shielding or balancing the source of noise with same methods, twisted pair, shielding etc.

Often AC gnd is sufficient to make a big improvement but imperfect, then software averaging, or signal conditioning can also be useful.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks for your thoughts. Software conditioning is not much use at this point, as the noise is bad enough that the MAX6675 gives either zero reading or off the scale. I'll try adding a CM choke and see if it improves things. \$\endgroup\$
    – Tom
    Feb 14, 2017 at 9:54
  • \$\begingroup\$ I'm not sure what constitutes high mu - care to suggest a device? Bear in mind that the choke will have to be as close as possible to the MAX6675 breakout (strictly not supposed to introduce anything between the bimetallic wires coming from the thermocouple and the terminals of the conversion device). \$\endgroup\$
    – Tom
    Feb 14, 2017 at 15:40
1
\$\begingroup\$

Install an RC filter in both thermocouple leads. You don't need exact temperature, so a few degrees error with inserted-100-Ohm resistors and their solder joints is a worthy experiment. The filter also needs capacitors, so let's use THREE: one differential, after the resistors, right by the MAXIM; the other TWO go from V+thermcoup to GND right by the MAXIM, and from V-thermcoup to the same GND solder point. I'd go with 1uF ceramics. Result: 1.6KHz F3dB.

You could also bias the thermocouple at VDD/2 (2.5 volts?) instead of letting nature and its electric fields push up and pull down on that tiny differential voltage; make a voltage-divider from VDD to GND, and tie ONE of the thermocoup leads to the VDD/2 node.

\$\endgroup\$
1
  • \$\begingroup\$ The thermocouple is a grounded type (ie the shield is connected to T-) so at the very least that would require some care! \$\endgroup\$
    – Tom
    Feb 14, 2017 at 9:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.