Electrical noise in temperature measurement?

I am working on and extruder for use at first with an existing CNC mill. I will be using the mills 4th axis to drive the extruder stepper, so I am using the mills 4 axis drive electronics. Since I'm not using the standard electronics, I've taken pieces from the existing electrical boards and put them together to make a temperature controller.
I have an AD595C with a K Thermocouple. Output of the AD595 is sent to a Duemilanove analog in pin, and the result converted and displayed, first in the serial monitor window, then on a separate serial display. Works great, controls the temp to within a couple degrees C, definitely enough for some rough testing. Until....
Flip on the CNC motor power supply, and the temp read goes haywire. Crazy noise. At first I thought there was an electrical connection through the thermocouple back to the board, wound up taking the thermocouple off the head, and having nothing but it and the supply on. The only electrical connection was that both were being run off power supplies connected to the same house circuit. Grounded the thermocouple shield wire with no effect either.
I've not read anywhere (particularly the datasheet) that these are particularly noise prone, but does anyone have experience with something like this? The DC supply that the mill runs off of isn't the best in the world, perhaps it's causing the issue. I need to scope a couple things, but I'm starting to run out of ideas. Anybody have any insight?

It would seem that the issue definitely stems from the mill power supply. With the scope probe grounded, I saw 50mV+ of noise when turning on the supply. Might be some of the problem

... driving stepper motors is extreme 'noisy', as you're switching really high currents with high frequencies.

Here you need really heavy filtering or better separating the measuring circuitry from the stepper drivers ...

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

Would shielding the electronics help?

That depends if it is conducted or radiated emissions causing the problem. For conducted you need filters.

I fixed a similar problem with a simple RC filter: [hydraraptor.blogspot.com]

You also need to make sure there is no motor current flowing in the ground connection from your thermocouple to the micro.

---

[www.hydraraptor.blogspot.com]

I did some more testing last night, I've come to the conclusion that it's noise radiated into the house voltage by the switching power supply. Stripped everything back until it was the only thing on and still got the noise. When I turned it off the noise was gone. I ordered an appropriate EMI filter for the supply today, I should have it early next week. It's a cheap supply, and I'm going to get a better one, but I want to see what the filter does first, if for nothing else than to have some numbers to compare the two.

NopHead, I checked every path from motor/supply to thermocouple, even went so far as to physically remove the thermocouple, board etc, to the other side of the room >10ft separation. Didn't have a stepper anywhere near the TC. Still noise with the PS on.

The most annoying part is that this has obviously been ongoing for a long time, and probably the source of a number of other "mysterious transient gremlins" that I put a lot of measures in place to fix. At least I seem to have found it now

Koko, when you said you grounded the thermocouple shield, exactly what did you do? The standard thermocouple (or at least the ones I have made from thermocouple extension wire and Harbor Freight freebies) does NOT have an outershield of braided or foil conductor wrapped around it like a coax cable. They do have a plus wire (yellow IIRC) and a negative wire, but neither one is a shield and even grounding the negative wire can cause problems with the very low voltage signal. If your thermocouple wire is not shielded, you should get some braided cable shielding, only $3 for 10' at McMaster-Carr and slide that around the thermocouple wire. Then ground only one end, the end next to the AD595C (good choice there!). If the problem is radiated electrical noise, this will go a long way to fixing it.

On the other hand, if it is conducted noise, try adding a filter to the AD595C input. The temperature changes very slowly, so putting a 10uF ceramic cap across the input pins, and 10K resistors in series should not change the stable temperature result, but will damp out a LOT of noise. Also make sure that you have by-pass caps very close to the AD595C. These should also be ceramic, and at least 0.1uF, but in a noisy situation, 1.0 or even 10.0uF would be better. If the power supply is VERY noisy, add a 10,000uF electrolytic cap to the by-pass cap to smooth the larger, longer noise pulses.


Hope this Helps.

Mike

rocket_scientist Wrote:
-------------------------------------------------------
> Koko, when you said you grounded the thermocouple
> shield, exactly what did you do? The standard
> thermocouple (or at least the ones I have made
> from thermocouple extension wire and Harbor
> Freight freebies) does NOT have an outershield of
> braided or foil conductor wrapped around it like a
> coax cable. They do have a plus wire (yellow IIRC)
> and a negative wire, but neither one is a shield
> and even grounding the negative wire can cause
> problems with the very low voltage signal. If your
> thermocouple wire is not shielded, you should get
> some braided cable shielding, only $3 for 10' at
> McMaster-Carr and slide that around the
> thermocouple wire. Then ground only one end, the
> end next to the AD595C (good choice there!). If
> the problem is radiated electrical noise, this
> will go a long way to fixing it.
>
> On the other hand, if it is conducted noise, try
> adding a filter to the AD595C input. The
> temperature changes very slowly, so putting a 10uF
> ceramic cap across the input pins, and 10K
> resistors in series should not change the stable
> temperature result, but will damp out a LOT of
> noise. Also make sure that you have by-pass caps
> very close to the AD595C. These should also be
> ceramic, and at least 0.1uF, but in a noisy
> situation, 1.0 or even 10.0uF would be better. If
> the power supply is VERY noisy, add a 10,000uF
> electrolytic cap to the by-pass cap to smooth the
> larger, longer noise pulses.
>
>
> Hope this Helps.
>
> Mike

I'm pretty sure I can tell if a wire has a shield on it or not. McMaster PN 3648K24 has a braided stainless shield.
You can almost make out the TC wire in the third pic down here. Sadly, grounding this shield doesn't do anything to kill the noise.

Additionally the circuit recommended on the wiki, and coincidentally on the datasheet shows the negative TC wire grounded for the AD595.

I will look into filtering the AD595 input as a last resort if the EMI filter I ordered for the supply doesn't cut it. As I said earlier with no connection whatsoever electrically other than they were plugged into the same house circuit, I saw 50mv+ of noise on a scope, with the ground of the probe connected to the tip when turning on the supply (of the mill/steppers, not the supply for the temperature board). Pretty sure that this needs curing regardless of any filtering to the TC.

Glad to hear that you have the higher quality thermocouple wire that includes the shielding. So that knocks out 2/3 fo my suggestions. Hopefully the power supply filtering will take care of the problem.

edit: removed


Mike

Edited 1 time(s). Last edit at 10/15/2010 03:02PM by rocket_scientist.