Shop OBEX P1 Docs P2 Docs Learn Events
Thermocouple Formula? — Parallax Forums

Thermocouple Formula?

davidsaundersdavidsaunders Posts: 1,559
in Propeller 1
Now that I am aware of the formula for calculating a thermistors reading, I am wondering if there is an equilevent simplification for calculating the reading from a thermocouple.

I have a bunch of thermocouples that I either have in use with the aid of tables for there value, or that I intend to put into use. As such a way to simplify things would be great.

Comments

  • DigitalBobDigitalBob Posts: 1,513
    You have to use the millivolt table for that type. Such as J,K,T, etc.. I don't believe they are very linear so your best bet is to use a chip like the AD594/AD595, LTC1025 or Max31855.
  • davidsaundersdavidsaunders Posts: 1,559
    DigitalBob wrote: »
    You have to use the millivolt table for that type. Such as J,K,T, etc.. I don't believe they are very linear so your best bet is to use a chip like the AD594/AD595, LTC1025 or Max31855.

    I have been doing it manually, and yes it is difficult to get the ADC setup correctly using sigma delta. Though I began with a copy of the circuit from the poor mans digital oscilloscope, and tweaked it from there until I got usable readings.

    Though thank you.
  • evanhevanh Posts: 15,916
    edited 2017-03-19 07:17
    Linear conversion works okay for most applications. It's around 50µV/°C for type-J and around 40µV/°C for type-K.

    Your biggest problem will be keeping the noise out. If you get this sussed then improving the above into a table lookup for more accuracy isn't too hard to add.

    EDIT: Corrected my brain-fade on the scale of things.
  • evanhevanh Posts: 15,916
    edited 2017-03-18 06:43
    Oh, and measured 0mV is ambient temperature at the terminals of the thermocouple cable. So you also have to know what that temperature is.
  • Tracy AllenTracy Allen Posts: 6,664
    edited 2017-03-18 16:28
    Thermocouples are fairly linear over a narrow range (say 50°C), but they wiggle substantially over their full range (say 1000°C). If you look at the thermocouple tables from Omega Engineering, you will find that they list the coefficients for a "forwarrd" and a "backward" polynomial. In using either the polynomials, or the table, or a linear approximation, you generally have to go both forward and backward. The reference cold junction temperature needs to be converted into an equivalent voltage, which is added to the measured thermocouple voltage, and that is then converted back to temperature. If you are using the table, that means a lookup followed by a lookdown.

    The integrated circuit thermocouple conditioners will take care of the cold junction compensation and they will add the appropriate voltage to the thermocouple voltage, and amplify the sum. They assume a linear approximation for the cold junction (assuming it is near room temperature), and they use a linear amplifier based on the nominal sensitivity in µV/°C. They do not apply any curvature correction.



  • davidsaundersdavidsaunders Posts: 1,559
    Tracy Allen wrote: »
    Thermocouples are fairly linear over a narrow range (say 50°C), but they wiggle substantially over their full range (say 1000°C). If you look at the thermocouple tables from Omega Engineering, you will find that they list the coefficients for a "forwarrd" and a "backward" polynomial. In using either the polynomials, or the table, or a linear approximation, you generally have to go both forward and backward. The reference cold junction temperature needs to be converted into an equivalent voltage, which is added to the measured thermocouple voltage, and that is then converted back to temperature. If you are using the table, that means a lookup followed by a lookdown.
    Thank you for that information, very useful.
    The integrated circuit thermocouple conditioners will take care of the cold junction compensation and they will add the appropriate voltage to the thermocouple voltage, and amplify the sum. They assume a linear approximation for the cold junction (assuming it is near room temperature), and they use a linear amplifier based on the nominal sensitivity in µV/°C. They do not apply any curvature correction.


    No extra unneeded IC's in my projects. Yes it means that I have to do some more work in often cases, though it is worth it to me.

  • DigitalBobDigitalBob Posts: 1,513
    It all depends on your level of accuracies. If you want instrument quality of TC reading I'd go with an ic. If it's just a room temp. monitor for hobby use then an ADC would probably work.
  • davidsaundersdavidsaunders Posts: 1,559
    DigitalBob wrote: »
    It all depends on your level of accuracies. If you want instrument quality of TC reading I'd go with an ic. If it's just a room temp. monitor for hobby use then an ADC would probably work.

    Neither, though it can be off by a few percent. Just need to be within 3% for temperatures in the range of 180C to 300C (356F to 572F), so nothing to perfect.
  • evanhevanh Posts: 15,916
    Oops, I was out by a factor of 1000! It's around 50 µV/°C, not the millivolts I said earlier. Sorry about that slip up.
  • jmgjmg Posts: 15,173
    davidsaunders wrote: »
    Neither, though it can be off by a few percent. Just need to be within 3% for temperatures in the range of 180C to 300C (356F to 572F), so nothing to perfect.

    Years ago we did a thermo couple correction using x^2 feedback via analog switches.
    There, we focused on room temperature, and the zone of interest, and were less worried about errors in the passes-while-warming zone.
    That was because room temperature errors would likely cause customer questions, but they have no idea if 100' reads as 120' :)

    You should be able to do a similar 'two zone' approximate fit, as you only need match two slopes.

  • JDatJDat Posts: 103
    What about ready to use ICs like MAX3185, MAX6675 or something similar?
Sign In or Register to comment.