Archived:Unexpected Noise or Fluctuations from Thermocouple Measurement

Updated May 13, 2024

NI does not actively maintain this document.

This content provides support for older products and technology, so you may notice outdated links or obsolete information about operating systems or other relevant products.

Reported In


  • K-Type Thermocouple
  • E-Type Thermocouple
  • J-Type Thermocouple
  • T-Type Thermocouple
  • SCXI-1100
  • SCXI-1122
  • SCB-68
  • SCB-68A

Issue Details

I'm getting wrong, unexpected, fluctuations or oscillations of several degrees in my thermocouple readings. They might center around what I expect the output to be, especially if I'm reading instantaneous data. I may be using a BNC to BNC-banana jack adapter for my thermocouple.  How can I reduce the noise in these measurements?


  • Determine if your thermocouple is floating or grounded. If it is floating, connect the negative terminal directly to AIGND.
  • If your thermocouple is ground referenced, use a differential measurement. 
  • Average the instantaneous measurements from your thermocouple within your application.
  • Select the maximum gain available, as thermocouples operate in a very small voltage range.
  • Ensure the negative lead of your thermocouple is grounded if you're using an AMUX-64T or SCXI-1100. 
    • If you are using an SCXI-1100 at sampling rates faster than 2 Hz, disable the 4 Hz filter. 
    • If you are using an SCXI-1122, the cold junction sensor is a thermistor and not a linear IC temperature sensor, so convert the voltage to temperature appropriately. 

Additional Information

  • A differential measurement helps eliminate noise by rejecting noise that is common to both lines of the differential pair. if these lines are not close together, they can have independent or out-of-phase noise that would carry more noise along the line. 
  • Voltmeters and thermocouple meters typically average 100 points for each measurement, which increases the signal-to-noise (SNR) by approximately 10%, because noise is reduced proportionally to the square root of the number of averaged samples.