Jump to content

As of July 17, 2015, the LabJack forums here at forums.labjack.com are shut down. New registrations, topics, and replies are disabled. All forums are in a read-only state for archive purposes.

Please visit our current forums at labjack.com/forums to view and make new posts. To post on the current forums, use your labjack.com login account. Your old LabJack forums login credentials have been retired. There are no longer separate logins for labjack.com and LabJack forums.


Attempting 0.1 oC accuracy with Thermocouples!

  • Please log in to reply
4 replies to this topic

#1 jhenise

  • Members
  • 26 posts

Posted 20 June 2013 - 12:00 PM

I work in a biotech company and am trying to set up a temperature data logger for quality control. The logger will use Type T thermocouples that I will attempt to read to 0.1 oC accuracy across a small range of 34 to 40 oC. I plan to use a LM34CAZ temperature sensor that will me mechanically attached to an well insulated cold junction to provide cold junction calibration. I have heat sources in the lab that I can trust to 0.1 oC, and a NIST certified thermometer that I can calibrate the sensors against and make my own scalar and linearization factors to get the absolute readings correct.

I posted a video on You Tube showing the reading of the E1034 (LM34CAZ based silicone temperature sensor) temperature sensor sold by LabJack, and of some Type T thermocouples using DAQ Factory software.

I am getting very good results reading the Type T thermocouples (purchased from Reotemp, San Diego, CA) directly as single ended inputs into the U6-Pro with the input range set to +/- 0.01 V (gain of 1000x).

The thermocouples are the grounded type in 3/16 diameter stainless steel housings. They have about 5 times the signal noise of the E1034, but this is still on the order of only 0.02 oC. I am very surprised that two of these identical thermocouples differ by only less 0.075 oC in their readings of the same environment. So this is looking possible!

I will post more results after calibrating the sensors in the lab!

#2 LabJack Support

LabJack Support
  • Admin
  • 8677 posts

Posted 28 June 2013 - 07:29 AM

Indeed interesting that the 2 thermocouples differ by only 0.075 degC. I wonder how that would play out across a larger temp range and more thermocouples?

Since you have "grounded" thermocouples, you need to watch out for complication #5 from our thermocouple app note:


In fact, I would slide a piece of paper between the 2 thermocouples in your video to make sure they are not having an electrical effect on each other.

I see in the video that you have a U6-Pro, so if you are using x1000 gain and resolution=12 you can expect an effective resolution of about 0.2uV and a noise-free resolution of about 1uV:


A T-type thermocouple gives about 40 uV/degC, so if the thermocouple signal has no noise at all the U6 noise-level would result in temperature resolutions of about 0.005 degC effective and 0.025 degC noise-free.

You said you are seeing a resolution of 0.02 degC with the thermocouples, which agrees well with the 0.025 degC ideal noise-free number. Likely the thermocouple signal has more noise, but the sigma-delta converter is getting rid of it.

It will be interesting to see what you find as you delve beyond resolution and get into absolute accuracy. If the U6 is at steady-state (stable temperature etc.), the limiting factor for the U6 would be the nonlinearity. Which from Appendix A of the U6 User's Guide is stated as 0.1 % full-span for the x1000 range, which is +/-20uV. That is the level we test each U6 to at that range, and limited by the ability of the calibration equipment to handle such small voltages, so the U6 could actually be more like 0.01%.

In your case you are only using the span of 1.36mV to 1.61mV, so I would expect you can achieve better linearity across that small span. For your desired +/-0.1 degC accuracy, you need linearity within +/-4uV.

#3 jhenise

  • Members
  • 26 posts

Posted 15 July 2013 - 11:02 PM

[font="calibri;"]Thank you for all the helpful information!  I am a bit of a hack with electronics so even the basic stuff is helpful…[/font]


[font="calibri;"]I have begun to do some testing with achieving absolute temperature readings.[/font]


[font="calibri;"]When hooked up as single ended inputs with a common ground , the grounded thermocouples did start having some noise issues (random spikes) when placed in certain electrically heated water baths etc. I hooked them up as differential inputs and this solved almost all of these issues. In the future I will go with ungrounded thermocouple probes since I can deal with a slower response rate.[/font]


[font="calibri;"]I am using 1000x gain and resolution set to 12 to read the thermocouples on the U6 PRO.[/font]

[font="calibri;"]There is about 0.02 to 0.03 oC of noise all the time in the signal, this correlates very well with your prediction, and is perfectly tolerable for reading to 0.1 oC accuracy.[/font]


[font="calibri;"]I am using the CB37 terminal board for the inputs. I attached an LM24CAZ silicone temperature sensor directly to the screw terminal housing to monitor the cold junction temperature. I then insulated the whole CB37 board in Styrofoam.  This works very well for cold junction compensation.[/font]


[font="calibri;"]I have been experimenting with calibrating the thermocouples in a 37.0 oC water bath I have in the lab.[/font]

[font="calibri;"]After adding some additive constants to offset the values of the thermocouples to 37.0 oC, I can take the thermocouples in and out of the water bath and they will return to 37.0 oC within +/- 0.03 oC, in agreement with my NIST standard thermometer. So reproducibility is very good![/font]


[font="calibri;"]I will soon begin to measure over a wider temperature range to test linearity…[/font]


[font="calibri;"]But so far I am very happy over all! [/font]

#4 nightowl5473

  • Members
  • 1 posts

Posted 06 November 2013 - 12:50 PM

Realize that regardless of whether or not the thermocouples are reading the same thing you need to use another method to check the accuracy in an absolute sense. The published accuracy of type T thermocouples is 0.5°C or 0.4% of the reading. This accuracy is a limitation of the TC and has nothing to do with noise. Two TCs from the same "batch" could agree very closely with one another and still be off from the "true" temperature.



#5 LabJack Support

LabJack Support
  • Admin
  • 8677 posts

Posted 06 November 2013 - 04:41 PM

I see that jhenise does mention having a trusted heat source and reference thermometer to evaluate absolute accuracy, but your point is a good one.  In fact, the 0.5degC figure is for class 1 Type T thermocouples.  A class 2 Type K at 1200 deg C is rated for just +/-0.0075*T => +/-9degC.



0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users