My problem is the following:
am writing a procedure to calibrate a datron 4700 (calibrator) using a fluke8508A (dmm), wanted to know what and how many measure points to consider and how to validate the procedure. two points zero and max range are sufficientes? i want to draw a calibration curve and only two point are too little in my opinion. have you some document or links to suggest me?
thanks

P.S. sorry for my bad english , i'm italian :o

I would say a minimum of three for each range.....zero, max, and middle.....five points are better, but three minimum.....and that is for each range.

Hope this helps.

Hershal

From back in my PMEL days I thought it was the following:

1) Full Scale
2) 90% of Scale
3) 10% of Scale
+ one additional point.

Wayne

If I understand correctly, you wish to calibrate the multi-function calibrator with a high accuracy multimeter. I haven't calibrated those models, but I am familiar with them, and have calibrated many similar models.

I would suggest that to calibrate a multi-function calibrator, there will be many parameters, not all of which may be calibratable using only the Fluke 8508A.

The first method to create such a procedure should at the least, be by reviewing the manufacturer's documented procedure for the 4700 (if it is available).

In the case of the calibrator, if you need to write a procedure with no manufacturer's manual, determine each function and range (AC, DC, current, voltage). For the DC parameters, as stated previously, check zero offsets in all ranges, then at least 2 parameters (measurands) in each range. For the AC parameters, approximately the same would apply, with the addition of testing frequency flatness in all functions/ranges.

I just reviewed some documentation on the 4700. Also, assure each resistance decade is measured adequately (4-wire).

Also, since I do quite a lot of high accuracy calibrations (I cal 3458A DMM's regularly), I would recommend assuring you use adequate cabling (special cables designed to calibrate high accuracy meters and multi-function calibrators). I believe both Agilent and Fluke make some.

Wmarhel Is there a rule or a document which says this?

The probem is that near zero the tur is < 3 and if i draw calibration curve by two points i can't see if is really linear .to perform bias, linearity and stability studies, how many and which points do you use for every range?
Thanks

If your TAR is <3:1, the first question to ask if whether the 4:1 is formally recognized in your country.....if not, then you have to calculate uncertainty, and don't be concerned about TAR.....

Otherwise, if 4:1 is recognized, then still calculate the uncertainty, at least below 4:1.

Hope this helps.

Hershal

As you posted the question again, I will echo exactly what the individuals posted about the # of test points (and obviously the uncertainty ratio portion also).

Depending on what it is, generally 10, 50, 90% of full scale, plus one to two additional is the norm. If your characteristic of interest is linear, as we all know, 2 points define a straight line. However, how many straight lines exist in real world?

It is the classic risk and return. You can do fifteen points at each parameter if you want. You'll have a very expensive, 40 page calibration. Or, just do two point, and hope that your device does not have error representing a curve or parabola.

As a rough suggestion, if you are doing this as a service for an outside customer, I would suggest locating more industry-standard procedures for the meter you are calibrating. If this is being performed in-house, you may choose to focus on the particular use of the device, and verify the accuracy for that use. There are several pros/cons to this suggestion as you may well know, but it may give a place to start from.

In Italy 4:1 is not formally recognized,so i use iso 17025 and the GUM to calculate the uncertainty. on the other hand a low tur is a problem because the uncertainty near zero explode. an uncertanty so high increases the risk by forgeries positive, and furthermore in the measures near zero expands the uncertainty strip too much, I read from some part than neighbour zero the entry size must be at least 3s, this would solve the problem, it but I miss the normative reference which refer to.

Wmarhel Is there a rule or a document which says this?

The probem is that near zero the tur is < 3 and if i draw calibration curve by two points i can't see if is really linear .to perform bias, linearity and stability studies, how many and which points do you use for every range?
Thanks

I'm reciting from memory from when I was in the calibration field. I had completed the PMEL (Precision Measuring Equipment Laboratory Specialist) course at Lowry Air Force Base in Denver, Colorado. This was the training school for all the U.S. military branches and even a large number of foreign military personnel as well as some civilians.

It has been over a decade since I actually sat on a bench and did calibration or repair work. You might want to try this site for some additional information: PMEL (http://www.pmel.org/page3.htm)

Wayne

One approach to uncertainty since Italy does not recognize 4:1, is to break into somewhat larger ranges near the bottom end.....some of the specific uncertainty near zero will be averaged out that way.....but don't make the ranges too much larger.....

Hershal