5790A Uncertainty Budget

[sanuriel]

Hi all, I know this may be taboo but I'm trying to sort this out here since contacting Fluke yielded nothing. So the company recently acquired a 792A to calibrate 5790A/B, and I've been tasked with the uncertainty calculation. Working on the 1000V range with 1000V in I've come up with the following contributors:

732A Uncertainty (contains cal uncertainty and trended uncertainty)
752A Uncertainty
5730A Linearity (because we are adjusting from a non-nominal value to a nominal value) x 2 (positive and negative)
5730A Stability
8508A Linearity (100 mV RNG because we are verifying roughly 10V to 10V)
Resolution of UUT (I think only once, but I'm not sure)
Repeatability of UUT
792A Calibration Accy
8508A Linearity x 2 (because +DC after connected to 792A, -DC after connection, and the AC Measurement) - Figure I don't care what the value is just the difference between the measurements.

What am I missing as my uncertainty looks much better than Fluke's?

[chinman]

There is a 1000 V range resistor that goes with the 792A.

[briansalomon]

If you use the search tool on this website you will find a great deal of information on this subject in previous discussions.

I know Fluke does use a coverage factor of K=3 for some of their tolerances.

[N79]

I would guess your 792A measurement uncertainty would swamp most other sources you've listed, and besides the Type A I think you have a pretty good list. As a check, make sure all the contributors you combine are of the same units and use the same distribution parameter (usually 1 sigma). The unit check will not necessarily tell if your analysis is right, but it will definitely let you know if it is wrong.

[sanuriel]

There is a 1000 V range resistor that goes with the 792A.

That is given, otherwise I would have destroyed the 792A

[sanuriel]

If you use the search tool on this website you will find a great deal of information on this subject in previous discussions.

I know Fluke does use a coverage factor of K=3 for some of their tolerances.

I looked for 5790A uncertainty and 792A uncertainty and really the search does next to nothing.

[HoggyBank]

At 1000V at 1k Hz, the 5790B has an absolute accuracy of 38 ppm (1 year), the 792A (according to the manual) has a +/- 5C uncertainty of 27ppm.  Without any other contributions, you are already at 1.4 to 1. 

For your 792A, you should be using the AC/DC Measurement Uncertainty.  This is found on the certificate from Fluke or whoever calibrated your 792A.  Fluke gives 2 pages, 1 with corrections; the other with the uncertainties.  May be better or worse than what is in the manual.  Depends on what you pay for.

[measure]

I know Fluke does use a coverage factor of K=3 for some of their tolerances.

Fluke does not use k=3 for some of their tolerances; Fluke Calibration products are typically specified at 99%, where k=2.576.

[sanuriel]

I know Fluke does use a coverage factor of K=3 for some of their tolerances.

Fluke does not use k=3 for some of their tolerances; Fluke Calibration products are typically specified at 99%, where k=2.576.

You're incorrect here as Fluke themselves told me that the 792A was at a K=3.

[briansalomon]

I went back and reviewed a response by "USMC Kalibrater" from a previous discussion on this subject. If you use the search function and look for "coverage factor" it leads to his response:

In reply #54 He points out that a 95% coverage factor (K=2) is actually 1.96  (rounded to 2 places)

So K=2 and K=3 are a type of shorthand. The actual factors would be used for the calculations but are represented like this on certificates for simplicity.

[Hawaii596]

Yes, it is sometimes a little confusing, as some Fluke products have k=2, and some have k=3 specifications. It changes how you apply them in uncertainty budgets.  I had to go through and do some budget corrections due to this issue.