Weird phenomenon when measuring temperature with thermocouple

[dhm04042]

I'm using USB-9213 connecting with bunch of thermocouples, which are in the same thermal condition.

And the result is this.

X-axis is time(second), and Y-axis is temperature(degree C). As you can see, there's weird cycle with about 30 minutes.

The thermal condition has not changed at all. Are there any possibilities that could effect the result like this by hardware or software?

[paul_cardinale]

My best wild guess is that the data are correct; the temperature really is fluctuating.

[Henrik_Volkers · Accepted solution]

Well, somewhere is a temperature cycle ....

Since the result of a TC measurement is not only a function of the thermal condition of the measurement junction (tip 😉 ) of the TC, instead it's a function the integral temperature differences along the TC and both endpoints* , I suspect the cold junction is drifting....

If you use one (additional?) channel with only a shortcut (short copper wire) at the cold junction (CJ), it should also read the CJ temperatur, how does it look like?

*) and all the rest ... found in the complete measurement uncertainty budget ....   :^)

Another source of slow changes I have seen in the past is a not perfekt alignment to the power line frequency in the voltage measurement, but that is more erratic...

[wiebe@CARYA]

I'd do a test with ice water.

That should be 0 deg.C... If that fluctuates, it's a technical problem.

Boiling water will work too, but might be impractical.

[dhm04042]

Well, this USB-9213 has its own CJC detection, so I select this option of course. Will it be better if I use independent CJC channel?

Thx for reply anyway.

[dhm04042]

I believe thermal condition is relatively stable, I guess.

[Kevin_Price]

I agree with others.  It looks real to me.  Any measurements I've done in open air in well-controlled lab spaces show fairly similar temperature cycles.  That's just the nature of HVAC systems. 

-Kevin P 

[RTSLVU]

You didn't say what type of thermocouple you are using but K type is probably the most common and K type thermocouples have a general accuracy of "Greater of 2.2°Cor 0.75%" and your 1.5 degree fluctuations are within their basic accuracy.