Can we believe a calibration certificate?

[Badwolf]

Hi everyone
I've been looking at this forum for a long time now and I finally decided to post a message (yes my first one !!!  )

Can we believe a calibration certificate?

I recently bought a 8-1/2 DMM (Adventest R6581T) and I received a nice calibration certificate (only the paper is nice, not its content  , ...)

I suggest a good exercise for beginners in metrology, in the form of a small game: find the errors in a certificate.
This will exercise you on reading and interpreting the specifications of a DMM, on the method of calculating the different limit values ​​and to be critical on tools used for a good calibration / verification.

To analyze this certificat, you will find the user manual on the xDevs.com: https://xdevs.com/doc/Advantest/R6581T/R6581-manual-OCR.pdf
Ok, it's in Japanese, but it's also a good exercise to quickly find information in a user manual.
You will also find on xDevs a good article on teardown / repair of this DMM: https://xdevs.com/fix/r6581t/

(Please, for the experts, they will recognize each other, give enough time to beginners before publishing the solutions  ).

Good luck ...


Sorry for my spelling / grammatical mistakes, I speak English like a Spanish cow and Google is my best friend / teacher for translation

[saturnin]

The calibration "certificate" (errors/mistakes are obvious) you posted is notably similar (cal. equipment used) to one I found in one eBay auction (see attachment). Does it come from the same company? 

[Badwolf]

The calibration "certificate" (errors/mistakes are obvious) you posted is notably similar (cal. equipment used) to one I found in one eBay auction (see attachment). Does it come from the same company? 

Maybe, maybe not  , my goal is not to accuse someone but to provide a good training exercise to understand the specifications of a DMM and to read a calibration certificate.
'Cause it's easy to get tricked when you don't understand all the information.

[Dr. Frank]

Well, the 1000 DCV is outside the MAX limit, but PASSED.

The whole specification is much worse than any reasonable 8 1/2 digit DMM, like the 8508, or the 3458A.
That's the real pain in the ...

Frank   

[Badwolf]

yep, the DC error is the first one (too easy), but evil is in details  .

For the specification, i think is due to lack of zero calibration, the last is from 2001, even if 10v and 10k are from this year.
So may be lot of drift / offset in the input amplifier. I'm running out of time right now to do a good check.

[bdunham7]

Well, where to start?

First, the resistance min/max acceptable are calculated from the nominal, not actual, values.   

Second, the actual readings are 6 and 7 digits, not 8 or 8.5.  I'm not sure it actually matters to the results in all cases, but still....

Third, the calibrator is a Fluke 5101B which is several orders of magnitude less accurate than the meter?? 

EDIT:  I didn't read the manual, but is it appropriate to only check the calibration at one spot in each range, right at the nominal top?  Is linearity guaranteed separately somehow?

Is there more?

[Badwolf]

Yep one more... details,... details ....specification ...

[Dr. Frank]

Well, where to start?


Third, the calibrator is a Fluke 5101B which is several orders of magnitude less accurate than the meter?? 

...

That's not really correct.
The 5101B serves as a reasonably stable stimulus only, but the reference is another 3458A.
This directly indicates, that this R6581T DMM is not worth its 8 1/2 digits. The 3458A itself might be a very good DMM, but is usually no accepted as a reference. (Like a 732A and a SR104)

I would calibrate 6 1/2 digits with a 3458A, at most

Frank

[Badwolf]

Indeed, the use of 5101 is not a certificate error.

But I think the calibration was not done in the best way.
If the 3458A had been used to check the output of the 5101, we would have approximate values, as for the resistances,
not exact values as it is. But hey I was not present during the calibration, so in doubt I do not consider this as an error, but a point to check.

[bdunham7]

That's not really correct.
The 5101B serves as a reasonably stable stimulus only, but the reference is another 3458A.
This directly indicates, that this R6581T DMM is not worth its 8 1/2 digits. The 3458A itself might be a very good DMM, but is usually no accepted as a reference. (Like a 732A and a SR104)

I would calibrate 6 1/2 digits with a 3458A, at most

Frank

You know more than I do about this subject and I have no idea about the actual requirements for issuing a cal certificate, but I have a few questions about how this would be done.

I saw the 3458A in their list of equipment (4ppm no less!), but it appears to be obvious that they did not actually use it as a reference.  I would expect, as the OP mentioned, that the actual values would be listed from reference to 8.5 digits and the min/max accepted would be calculated from that.  Now the cert is probably entirely fraudulent, but if they did anything, I can't see how they used the 3458A for the voltage measurements.

So, if you were going to do this quasi-properly with the equipment at hand, would you measure the output of the calibrator with the reference and then switch to the DUT, relying on the short term transfer stability of the calibrator?  Or would you connect all three in parallel? 

If the first, it seems the short-term specs on the 5101B insufficient here.  If the second, isn't there an issue with the DMMs interfering with one another, as I've read about on this forum?  And what about the noise specs on the 5101B which are way higher than the measurement tolerances--there's no guarantee that both meters will respond the same way to a particular kind of noise, is there?  The maximum error implied by the cal cert is 7ppm for the 1V range and 5.2ppm for 10V.  The ripple and noise spec for the 5101B at 10V is 0.02%+50uV, so 2.05mV noise.  That's 205ppm unless I screwed up.  Isn't that a problem?

[Zoli]

One question: how the DC 1000V range is passed? For me, 1000.0072 V is 18mV lower then 1000.009 V   

[Kleinstein]

The ripple specs for the 5101 are really poor. However the DMM will not react that much to ripple.

With the relatively poor source one would have the three instruments in parallel. The interaction should be weak - though the R6581 does cause quite some current spikes at the input - so a difficult one in this respect. One way around this would be running the 3458 with 1 plc and skip the points effected by switching.

Still the 3458 is not a proper reference to calibrate a 8 digit meter.
With a calibration to a lesser standard the limits from the references should be noted. With a 8 digit meter it would not be unusual to not guarantee the full specs with a budget cal., but those limits should be noted.

The limits for the 1000 V range look relatively tight - there often is additional tolerance for >500 v or so. So the 1000 V are likely borderline: not sure to be in spec, but also not sure to be out.
I would expect test with negative voltages too. With the 3458 as reference there the measured value is missing.

[dietert1]

As far as i understand the 1000 V error spec in the user manual p. 15-1 is 2.5 ppm in 24 h, 5 ppm in 90 days and 7 ppm in 1 y. The instrument is in calibration, but the certificate has the limits wrong. And i agree with Dr. Frank, that the instrument doesn't do 8.5 digit measurements. Even if you calibrate every day, you could not even do 6.5 digit measurements without careful error calculations.
But that is more or less true for all 8.5 DVMs.The built in references of those instruments are very simple. I would say those LTZ1000 references are very good, yet optimized for cost and convenience and not good enough for eight digits.

Regards Dieter

[deadlylover]

Here is a cal certificate I got when I asked a local lab to do an artefact calibration on my R6581 (back when I was a bit of a volt nut...oh god it's been 5 years?!).

In Australia, NATA accreditation means the lab has to perform to the ISO 17025 standards (or other applicable standards). And of course naturally, they will always give out the uncertainties of the standards used. You can actually look up a lab's accreditation on the NATA website and see their least uncertainties.

I think it's safe to say that the "calibration certificate" that you've been provided is more like a basic operational check by the seller? 

This is why you'll sometimes see two prices when you get stuff sent to calibration. A normal calibration and a more expensive accredited calibration.

[Kean]

Here is a cal certificate I got when I asked a local lab to do an artefact calibration on my R6581 (back when I was a bit of a volt nut...oh god it's been 5 years?!).

I'm surprised that document doesn't provide any details of the device being calibrated.  I assume there are other pages to the report despite the attachment being page 1 of 1.

[deadlylover]

I'm surprised that document doesn't provide any details of the device being calibrated.  I assume there are other pages to the report despite the attachment being page 1 of 1.

Nope that was it, I only wanted the 10v and 10k ohms done for you know....volt nut tendencies. The full calibration+verification would have cost something like 2.5 grand I think? 

Oh there was the main page with customer details, instrument details like model/serial number and stuff if that's what you meant. (I only uploaded the "attachment" to that report)