How stable is a thermocouple?

[ftransform]

I have a old thermocouple reader which, when the filter caps were replaced, reads the temperature with a fluctuation of +- 0.2 degrees Celsius, meaning it will jump between 2.8 and 3.2 for instance. It still jumps when I put it inside of Styrofoam box. I have tested the atmosphere in my room prior with a RTD and found the temperature is much more stable then that.

It has some tantalum capacitors for filtering the input (which I checked the ESR of, they seemed fine), and some filter capacitors for the rectifier (which I replaced with excess).

I used this method for checking the capacitors:
http://fullnet.com/~tomg/esrscope.htm

Replacing the supply filter cap reduced the fluctuations significantly. They tested bad. The tantalums on the thermocouple input tested out fine. Is this type of fluctuation normal for a type T thermocouple? I also tried wrapping the thermocouple in ferrite rings but that did not really do anything. I had like 4 rings with 30 windings installed near the screw terminals on the unit, out of frustration, but that did not really do anything.

Maybe since this is an old unit it was just not built for the amount of noise that switchmode supplies throw out now?
I was thinking about maybe making some sort of matched low pass filter but that seems rather difficult for such low voltage measurements.
The wire is also twisted.

DUe to a lack of probe at this time I have just twisted together the extension wire to form my thermometer but I don't think this has any effect.

[amspire]

I have a old thermocouple reader which, when the filter caps were replaced, reads the temperature with a fluctuation of +- 0.2 degrees Celsius, meaning it will jump between 2.8 and 3.2 for instance. It still jumps when I put it inside of Styrofoam box. I have tested the atmosphere in my room prior with a RTD and found the temperature is much more stable then that

Because a lot of the Thermocouple readers often have a scale going hundreds of degrees or even over 1000, degC,  1 deg C can equal about 0.1% or less of full scale. Many old thermcouple meters had limited A/D resolution so a jump of +/-0.2 deg was pretty common. I think that is what you have - just digitizing error. Not much you can do about it.

[ftransform]

I have a old thermocouple reader which, when the filter caps were replaced, reads the temperature with a fluctuation of +- 0.2 degrees Celsius, meaning it will jump between 2.8 and 3.2 for instance. It still jumps when I put it inside of Styrofoam box. I have tested the atmosphere in my room prior with a RTD and found the temperature is much more stable then that

Because a lot of the Thermocouple readers often have a scale going hundreds of degrees or even over 1000, degC, room 1 deg C can equal about 0.1% or less of full scale. Many old thermcouple meters had limited A/D resolution so a jump of +/-0.2 deg was pretty common. I think that is what you have - just digitizing error. Not much you can do about it.

AH, I figured it was kind of silly to add the decimal point if the digitization error is there. Not totally useless but a bit irritating.
What is the span adjustment for?

Right now I adjusted the zero adjust so it matches up with my 80 degree F span mercury thermometer. Do you think its better to short out the unit and zero it or match it up with a good mercury thermometer? Intuition leads me to the thermometer. I used a rubber band to attach the thermocouple to the mercury bulb of the thermometer as a "calibrator"
I believe my thermometer is advertized as accurate to 1F, which is better then the 1C accuracy that seems to be said on various websites.

This reader was made for a type T thermocouple, which has a upper range of 300c.
This thing is old enough to have curved traces. and a 7 seg display.

[amspire]

Right now I adjusted the zero adjust so it matches up with my 80 degree F span mercury thermometer. Do you think its better to short out the unit and zero it or match it up with a good mercury thermometer? Intuition leads me to the thermometer. I used a rubber band to attach the thermocouple to the mercury bulb of the thermometer as a "calibrator"

This reader was made for a type T thermocouple, which has a upper range of 300c.

If you have a lot of ice in a glass of water, and you can keep the sensor from contacting the ice, then as long as the water is fairly pure, it is extremely close to 0 degC. You can do something like wrap the thermocouple in Glad Wrap and then a few layers of paper towel to separate it from the ice. A insulated mug or a polystyrene cup wold be good. Accuracy is better then 0.01 deg C without too much difficulty. I would use that as the reference for zeroing. Trying to use boiling water for 100 deg C reference is not much use - it takes a very good setup, and the boiling point changes with air pressure.

[ftransform]

Right now I adjusted the zero adjust so it matches up with my 80 degree F span mercury thermometer. Do you think its better to short out the unit and zero it or match it up with a good mercury thermometer? Intuition leads me to the thermometer. I used a rubber band to attach the thermocouple to the mercury bulb of the thermometer as a "calibrator"

This reader was made for a type T thermocouple, which has a upper range of 300c.

If you have a lot of ice in a glass of water, and you can keep the sensor from contacting the ice, then as long as the water is fairly pure, it is extremely close to 0 degC. You can do something like wrap the thermocouple in Glad Wrap and then a few layers of paper towel to separate it from the ice. A insulated mug or a polystyrene cup wold be good. Accuracy is better then 0.01 deg C without too much difficulty. I would use that as the reference for zeroing. Trying to use boiling water for 100 deg C reference is not much use - it takes a very good setup, and the boiling point changes with air pressure.

Ah I was trying to do that but for some reason this thermometer blanks out and displays ---- if cooled to bellow 0 degrees C 

[amspire]

My next choice would be the RTD if it is a Platinum RTD - if you can measure the resistance of the sensor accurately. The Platinum RTDs are very accurate.

Did you let the thermocouple actually touch the ice? The ice is below 0 degC. It is only the water that is at 0 deg C.

[ftransform]

My next choice would be the RTD if it is a Platinum RTD - if you can measure the resistance of the sensor accurately. The Platinum RTDs are very accurate.

Did you let the thermocouple actually touch the ice? The ice is below 0 degC. It is only the water that is at 0 deg C.

I dunked it in a cup of ice water. This was prior to the thermometer calibration. I might try some more stuff tomorrow but ive just had about enough of this thing.

Does the span adjustment have anything to do with the high/low temperature maximum? I'm scared to touch it
It would be nice to be able to measure cold temperatures.

But anyway, considering that this will not be more accurate then +-0.2, I think a rubber band to thermometer is more then accurate enough.
This thing already almost ended up being punted out a window. A issue popped up which I thought was related to bad solder connections due to problems with the display, board flexing/tapping just cemented my "theory".

I spent a bunch of time reheating solder joints and adding extra solder on some skimpy looking joints, took it apart several times now. It turned out there was a socketed IC behind a ribbon cable that I just kept on missing!! 

[IanB]

Did you let the thermocouple actually touch the ice? The ice is below 0 degC. It is only the water that is at 0 deg C.

Ice right out of the freezer will be colder than 0°C, but once it has been sitting in a glass of water for a while and had time to reach equilibrium then the ice will also be at 0°C.

It is always the case that when an isolated system reaches equilibrium then all parts of the system are at the same temperature.

[SeanB]

Zero and span are somewhat interactive, but just adjust zero with the TC in ice water until it reads 0.0 then transfer to the boiling water and adjust span till it reads 100.0 ( provided you are within 1000ft of sea level this will be about right) then place back in the ice water again and check it still reads 0.0 for a check. Adjust again if changed and repeat until it does not change. A higher test point than boiling water is if you have a solder pot with 60/40 solder that is cooling down from molten, when it is going crusty it is at 183C +-5C, as a check point. Probe has to be able to handle solder though, so a stainless steel probe tip is needed.

[ftransform]

So this is kinda weird, the display stabilized. It was oscillating like I mentioned for a while, so I stopped paying attention to it. It was unplugged for several days... now I plugged it in and its stable, deviates at 0.1 at low temperatures... Previously it was stable > 25 degrees, and it started to hunt at <25 degrees, now its stable at 19 degrees +-0.1...

I wonder what happened.. reformed capacitors maybe? It's not oscillating fast at 0.1 degrees either, it might be the air currents in the room, I have not tried stabilization yet.
I tested all the capacitors using the oscope and FG.


Could the filter tantalums have reformed?