[volt-nuts] Fluke 720A Kelvin Varley Divider Questions

Dr. Frank Stellmach drfrank.stellmach at freenet.de
Mon Sep 20 18:21:08 UTC 2010

  > Thanks to Greg, Bill, Frank and all the others for your valuable 
comments and insight on the Fluke 720A reference divider.

>  Frank, I will attempt to answer your questions about calibrating the 5440B.
>  I have the 5440B manual which states, "the recommended calibration procedure
>  uses the Fluke 732A Direct Voltage Reference Standard and the Fluke 752A Reference Divider
>  to establish the necessary voltage standards which are then compared to the calibrator with the Fluke 845A Null Detector."

>  I would really prefer to get my hands on a Fluke 752A because it would provide me the required highly precise 10:1 and 100:1 divider
>  ratios that would, in combination with the Fluke 732A's 10 V output, provide me highly accurate standard voltages
>  for all the required cardinal points of 0.1V, 1V, 10V, 100V, and 1,000V.
>  Unfortunately, the 752A does not seem to be readily available on the used equipment market (e.g., EBay, Craigslist, etc.)
>  and the Fluke 720A is available.  While the Fluke 720A may be overkill for providing only two precise divider ratios of 10:1 and 100:1,
>  it appears to be the only option that is readily available.  So, I understand your statement that I "don't need the 720A for calibrating a 5440B",
>  but I do not understand your statement that I "can't use a 720A".
>  Maybe I am not understanding a subtle point that you are making and I sincerely would like to understand it if that is the case.
>  I think I can use the 720A to provide the required precision 10:1 and 100:1 divider ratios specified in the 5440B calibration procedure
>  since I can't get a 752A.  Am I missing something?
>  Charlie


the 752A or a 3458A are able to calibrate the 5440B correctly, but not the 720A.
What you are missing is the correct interpretation of the 720A specification.

The 720A is specified "0.1ppm linearity error of input" for a setting S of 1.1 ..0.1.
That means, if you apply 10V input voltage ("Input" is the crucial word!), it gives an error of 1µV.
Related to the output (linearity!!!), that means the 0.1x divided output of 1V is unsecure 1µV, i.e. 1ppm of output, or 1ppm of ratio.

That means, that the 1V reference point for calibration of the 5440B is too unprecise, or "on the edge", as the reference should have 1/10
of the precision level of the UUT.

For the 100:1 transfer, 100mV range, the situation is even worse. 4ppm is the precision of the 5440B, plus EMF, but the 720A gives 4,6ppm on the 100:1 ratio only
For the high voltage range  transfers, you also have to add the nonlinearity of the thermal heating in the same manner (power coefficient of linearity .. of input..").

The 1000V =>  10V transfer is precise to 4+2,6ppm = 6,6ppm for the 720A (10W power dissipation!!), for 100V=>  10V it is 1,2ppm (on the edge again).

For high voltage, the 720A resistance is much too low.

The 752A is better suited for that, its divider resistors have 4 MOhm in the 1000V range, giving 250mW only.
This divider is specified for 0.5 and 0.2 ppm ratio (output) accuracy, thermal effects included.
Low voltage transfers can be done even more precise, ca. 0.2 ppm for a 10V =>  100mV, and 0.15ppm for 10V =>  1V transfers.
Those precision levels are at least 5 times better than the specified accuracy of the 5440B ranges - good metrological practice.

For the beginning I recommend two articles where this aspect is described:

Here you can check my statement about the correct calculation of the ratio error of the 720A:

page 11-19.

And this famous one describes the problem of checking the ultra low nonlinearity of the HP3458A, for which the 720A is not precise enough, due to the same root cause.

See pages 14 and 23 for discussion.

Here it is stated also, that the definitions of nonlinearities are not precisely given, what I can underline from painful experience.

regards Frank.

More information about the volt-nuts mailing list