AMS2750E and a Complete Lack of Measurement Uncertainty

Section 2.2.75 of AMS2750E defines “Traceable” (or Traceability) as “The ability to relate measurement results through an unbroken chain of calibration to NIST or equivalent agencies in countries outside of the United States.”

In the expansive and varied world of calibration, we have a dictionary. We call it the VIM (JCGM 200:201 International Vocabulary of Metrology.) Section 2.41 (6.10) defines “Metrological Traceability” property of a measurement result whereby the result can be related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty. This definition is followed by eight different notes to clarify the definition further and allow for special circumstances. None of those notes remove the requirement for measurement uncertainty.

To simplify all this, in the language of metrology and therefore calibration, you cannot have traceability without uncertainty. This part is non-negotiable to metrologists, and if you are dealing with an ISO/IEC 17025 accredited laboratory, it is not just an NC. It is considered a complete breakdown in traceability, and that’s a big deal to us.

Why should I care?

AMS2750E tends to treat thermocouple measurements as a chain of voltage calibrations. This would be all well and good if thermocouples worked that way, but they don’t. Voltage production due to the joining of dissimilar metals is only part of how a thermocouple works. The voltage generated is a function of the Seebeck effect first discovered in 1821. The voltage of the hot junction, however, cannot predict the temperature in and of itself. There needs to be a “cold junction” in the circuit. In the past this was performed by transitioning the thermocouple wire to copper and placing the junction in an ice bath, hence the name “cold junction.” Modern electronic devices measure the temperature of the transition from thermocouple wire to copper and compensate electronically. This is a very useful trick. However, it is also the single largest contributor of uncertainty to the measurement.

Consider the following example:

·       Published specification of a Fluke 5500 calibrator @ 50 mVDC is ±0.005mV

·       With a “K” type thermocouple 50mV is about 2250°F and 1°F is represented by a change of 0.02 mV

·       Based on the two statements above you would expect the accuracy of the Fluke 5500 to be ±0.25°F @ 2250°F (0.005mV being ¼ of 0.02mV)

·       The published specification of the Fluke 5500 for a measurement @ 2250°F is ±0.72°F. That is nearly three times the amount predicted by the difference in voltage. Where does this additional uncertainty come from?

The cold junction compensation is responsible for the majority of the inaccuracy of the measurement. The problem is that it is not very easy to measure the temperature of a set of connections and their contribution to measurement inaccuracy can be several degrees in extreme cases.

AMS2750E allows for the compensation of readings using “correction factors.” Corrections factors should never be applied to a system without regard for measurement uncertainty. In many cases when measuring thermocouples, the measurement uncertainty may be as large as the correction factor that is applied. In this case, we are adding arbitrary bias to the measurement that will likely cause the measurement to be more errant then it was before the “correction factor” was applied.

I hope to see a future version of AMS2750…F? that regards measurement uncertainty as a real contribution to measurement, instead of a nonexistent property that can be excused from a system using correction factors.