Advice/Experience for Keysight DMM Calibration?

[fiddleback]

Just to finish up here, regarding 34465A calibration/adjustment by Keysight:

• The meter was returned within about a week, and Keysight kept me very well posted regarding its progress. In fact I was able to retrieve the complete Measurement Report from their website several days before I actually received the unit back.
• The unit was indeed "adjusted" as promised verbally (but contrary to the wording of the service agreement)
• As promised, very comprehensive 'as received' and 'as adjusted' reports were provided
• Somewhat surprisingly to me, 1 year upper and lower limits were used in the reports - even for the 'as adjusted' comparison. All of the 'as adjusted' readings on my unit appeared to be within the 24 hour spec. But this does indicate that I could have received a unit back that just barely met 1 year limits and it would have "passed" calibration.
• Somewhat surprisingly to me, the unit was adjusted at a temperature of 25.6C, and ACAL'd at 25.3C. (Working mostly with boat anchors, I was expecting 23.0C.) I'm guessing that since this unit has an ACAL capability (that I'm wary of using), Keysight judges that any adjustment temperature is OK. Personally, I don't exactly buy into this, because ACAL only corrects for DC voltage and resistance measurements. Others (current, AC voltage) are spec'd based on the "CAL" temperature.
• As to performance, +10VDC was found to be 5PPM low (it was 1PPM low at time of manufacture). After adjustment it was <1PMM low.
• 10K Ohms was found to be 3PPM high (it was <1PPM low at time of manufacture). After adjustment, it was dead on.

As an aside, I've found the DCV tempco of this unit to be approximately 0.20 PPM/C, which I find to quite respectable.

All-in-all, I'm fairly pleased with the Keysight calibration and would do it again, although I have to say that if a different model were being calibrated, I'd go through the entire up-front communications effort again - to find out what they ACTUALLY do, not what the website SAYS they do.

[Andreas]

Hello,

the 25.6 deg C are they stated as room temperature or is this internal device temperature?

On the "as received" data is there any reading outside of 70% of the 1 year spec?
This is the limit that I found on a formerly Keithley web site (no longer available) for doing adjustments.

with best regards

Andreas

[e61_phil]

• As to performance, +10VDC was found to be 5PPM low (it was 1PPM low at time of manufacture). After adjustment it was <1PMM low.
• 10K Ohms was found to be 3PPM high (it was <1PPM low at time of manufacture). After adjustment, it was dead on.

Such numbers only make sense together with uncertainties. Could you also say how big the uncertainties of the calibration are?

I also read in the 34401A manual that HP suggested to always adjust these "lower precision" meters every time to ensure it stays in spec over one year.

Calibration against the values of the used calibration intervall (normally one year) is normal. Otherwise you would need lower uncertainties which is more expensive and not needed. The 24h are over before the unit has been returned.

[fiddleback]

the 25.6 deg C are they stated as room temperature or is this internal device temperature?

Internal temperature.
The calibration page on the display shows the current internal temperature in real time. It also shows the difference between the current temperature and the calibration temperature to a tenth of degree Celsius, and it updates a couple of times per second. So adding the difference to the current temperature give you what was presumably the internal temperature at time of  last calibration.
Note that this device has a fan that runs all the time, and doesn't produce much heat, so I find that internal temperature is generally not all that much hotter than room temperature.

On the "as received" data is there any reading outside of 70% of the 1 year spec?

No, everything was very close to nominal - way inside even the 24 hour spec. If I get a chance, I'll calculate PPMs for each measurement - there are very many! - and report on my findings.

• As to performance, +10VDC was found to be 5PPM low (it was 1PPM low at time of manufacture). After adjustment it was <1PMM low.
• 10K Ohms was found to be 3PPM high (it was <1PPM low at time of manufacture). After adjustment, it was dead on.

Such numbers only make sense together with uncertainties. Could you also say how big the uncertainties of the calibration are?

Keysight DOES offer a calibration with an uncertainty report, but, of course, this is more costly. They do, however, indicate the calibrator used (5730A), and the certificate number of its current calibration. The "cal due date" on the calibrator was 30Aug2020, so I'm guessing that the calibrator may have been calibrated only a few weeks prior to calibrating my unit.

The 24h are over before the unit has been returned.

True, of course. But I suspect that 24 Hour specs are intended more for use in transfer measurements. The HP 33401A Service Guide says:

• "The 24-hour accuracy specification indicates the multimeter's relative accuracy over its full measurement range for short time intervals and within a stable environment Short-term accuracy is usually specified for a 24-hour period and for a +1 C temperature range."

I suspect that the key word here is "relative". Even if the absolute accuracy of the unit is way off, the relative accuracy is typically more meaningful in transfer measurements.

Thanks for the great questions and observations!

[Grandchuck]

• As to performance, +10VDC was found to be 5PPM low (it was 1PPM low at time of manufacture). After adjustment it was <1PMM low.

[/quote]

My 34465A has drifted low on the +10VDC by an amount similar to yours.  It has not been sent back for calibration.  I also have a 3458A and some trusted references and that is how the drift has been determined.
Has any one else seen this type of drift?  Not complaining BTW ... just curious 

[e61_phil]

Keysight DOES offer a calibration with an uncertainty report, but, of course, this is more costly. They do, however, indicate the calibrator used (5730A), and the certificate number of its current calibration. The "cal due date" on the calibrator was 30Aug2020, so I'm guessing that the calibrator may have been calibrated only a few weeks prior to calibrating my unit.

of course?!?!?! Without uncertainties I wouldn't call it a calibration certificate. Guessing isn't the right approach in metrology
Even the most 3458A certificates from Keysight I saw stated uncertainties of a couple of ppm even in the 10V range.

It is nice that it shows readings close to the nominal values, but as already said it means nothing. I often saw DMMs which agreed perfect to each other and after comparing them to a well known standard one could show both are wrong in the same direction and value. And even with a freshly calibrated calibrator it doesn't mean that the calibrator was adjusted, if it was still within spec (3.75ppm/a for 10V, everything else is worse). It depends how they use the 5730A (and if they do it like Fluke suggests). Normally you run an artifact calibration from time to time. That expands the calibration cycle of the 5730A to two years. Otherwise Fluke recommends to calibrate the 5730A yearly. In the end it is the decision of their department.

[Gandalf_Sr]

I just had my 34461A calibrated by Keysight in Roseville and, like fiddleback, I found their web site to be as clear as mud. I too called a couple of times and got passed around until I eventually spoke to someone in their calibration department called Neil who confirmed what fiddleback said above; their $240 calibration service does a pre adjustment test, then the meter is adjusted, and then tested again; the results of both the pre and post adjustment tests are included in the calibration report and they send you a paper calibration certificate.

On the subject of logistics/cost, it's cheaper to take the $228 'Keysight Calibration Agreement' option ($228) because, apart from the$12 saving, you only have to pay for the shipping TO Keysight and they ship it back at their cost; the 'Keysight Calibration - Per Incident' option ($240) performs the same calibration but you have to pay the shipping both ways.

When you take the Keysight Calibration Agreement option, you have to fill in a load of information and then you get a receipt that says you've paid for a service agreement which was very confusing and, at first, I thought they'd conned me into paying twice but, in reality, the agreement you purchase includes the ability to send the meter back for calibration at any time during the year of coverage (you pay to ship it to them with an RMA number they issue after you request the calibration service) and then they ship it back to you at their cost.

[mzzj]

of course?!?!?! Without uncertainties I wouldn't call it a calibration certificate. Guessing isn't the right approach in metrology

That's how it goes with big names like Keysight and Fluke.  Factory calibration certificates are often good as toilet paper  even from the big names. 

For example National Instruments sells you traceable calibration without uncertainties 
https://www.ni.com/services/calibration-compare/

Consensus among metrology folks is pretty strong that you can't have traceability without uncertainties...

[e61_phil]

of course?!?!?! Without uncertainties I wouldn't call it a calibration certificate. Guessing isn't the right approach in metrology

That's how it goes with big names like Keysight and Fluke.  Factory calibration certificates are often good as toilet paper  even from the big names. 

The certificates that came with the Keysight 34470A were such factory calibration certificates without any use (we need them for accreditation). But the Fluke 8508A and also the Fluke 5730A came with proper calibration certificates. But that is calibration gear and therefore probably not comparable.


I don't know how the test rig for the adjustment and end test of such meters look, but if there is a rig with a calibrator which is used to adjust the meter the first time after producation and after that test ("calibrate") the meter it is no surprise, that all these meters are "spot on". That would be even the case if the used calibrator is way off.

[dl1640]

iirc, the 5730A is calibrated on 2-year basis, in between maybe to execute a cal check to get a bird view of drift, if they calibrate 8.5 digit meter, the reference is a reference meter, the 57xx is used as a stable source.

[tszaboo]

of course?!?!?! Without uncertainties I wouldn't call it a calibration certificate.

The uncertainty is written in the datasheet of the multimeter. There are like 3 pages of it.

[dl1640]

no, uncertainty on calibration certificate is not your meter’s data sheet. the uncertainty is evaluation of how uncertain the reported measurement data is, usually it includes short term repeatability of your meter, finite display resolution of your meter, measurement standards uncertainty, measurement procedure uncertainty, tempco if relevant etc., in general it is limited by the class of your meter because finite resolution, intrinsic noise etc., you can’t get better reported uncertainty of 34401 than 3458, different class, even the cal source is the same 5730.

[tszaboo]

no, uncertainty on calibration certificate is not your meter’s data sheet. the uncertainty is evaluation of how uncertain the reported measurement data is, usually it includes short term repeatability of your meter, finite display resolution of your meter, measurement standards uncertainty, measurement procedure uncertainty, tempco if relevant etc., in general it is limited by the class of your meter because finite resolution, intrinsic noise etc., you can’t get better reported uncertainty of 34401 than 3458, different class, even the cal source is the same 5730.

Specification (spec)
The warranted performance of a calibrated instrument that
has been stored for a minimum of 2 hours within the operating
temperature range of 0 to 55 °C and after a 60-minute warm
up period. All specifications include measurement uncertainty
and were created in compliance with ISO-17025 methods. Data
published in this document are specifications (spec) only where
specifically indicated.

[dl1640]

well, your meter’s spec normally includes calibration uncertainty of manufacturer, this is true in many of Fluke calibrators, they call this absolute uncertainty and is reported as 3 sigma to give you more confidence but sometimes conservative, however if you have better measurement standards than Fluke have, you are able to have better meter spec...but i don’t think that is needed for daily applications unless you are a true nuts.