Can I trust this cal cert

[grumpydoc]

Found a 2001 for reasonable money with a cal cert and couldn't quite resist.

Not that I need 7.5 digits but it would be nice to get an idea if my motley collection of 2000's and 2015's and were within spec, as well as my KH 522 (about which I've previously posted - alarmingly in 2015).

Now I realise the limitations of a 2001 in this regard - the extra digit not necessarily bringing all that much in the way of extra accuracy, certainly not an order of magnitude as the best range on the 2001 is 25ppm of reading + 2ppm of range compared with the 2000's 30+5 (on the 2V and 10V scales respectively).

7.5 digits is actually new territory for me though - as is external calibration reports. The document provided looks impressive in that it suggests the 2001 is, in fact, an order of magnitude better than spec with an error on the 20V range of around 1PPM.

But it neglects to mention (other than in an opaque fashion) what standards were used, or what their uncertainties are, or when *they* were last calibrated - should I expect this information?

I guess it will be something like a Fluke 5730A - something in the 3.5-2.5PPM range so I can probably take 1PPM with a pinch of salt and assume that the meter is pretty close - certainly within 10PPM but I can't help thinking that, at this level, there should be more information.

Also - the meter thinks that it was last calibrated in 2009, so they didn't actually take it through the cal procedure, should I be worried about that (I suppose they just took it through a calibration check and as it was in spec left it at that).

I've just included the 1st page the rest is basically the same for the other ranges.

[J-R]

I agree, it does seem slightly odd.  There are 1-2 least significant digits missing as well...

Briefly I had a couple 2001 DMMs to play with and actually they have an 8.5 digit mode that with a good calibration/adjustment, attention to detail and filter settings can be of some use for DCV at least, IMHO.   So actually looking at the data provided I would personally get it professionally calibrated/adjusted.

[Martin72]

Hi,

But it neglects to mention (other than in an opaque fashion) what standards were used, or what their uncertainties are, or when *they* were last calibrated - should I expect this information?

We are certified under various standards and are regularly audited to ensure that we meet these certifications.
If we were to present such a protocol to the auditor, we would lose all our certificates, including the request to recall all our products that were measured with the device...Fatal.
A calibration record must detail the standards after calibration has been performed.
The calibration laboratory itself has to list its accreditations and finally, of course, the equipment with which the calibration was performed and to which standard it is calibrated and when this calibration expires.

[thm_w]

The cal is from 2011(?) any value it has is long gone.
Either send it off to get an up to date cal, or get a voltage/resistance/etc standard thats been cal'd to test it against.

[coromonadalix]

fishy

Here at my job   we have way more information than this paper, with tester / technician signature, and yes the reference(s) tools are specified  not an Sxx gimmick

We have the before CAL and after CAL  .....

[bdunham7]

I guess it will be something like a Fluke 5730A - something in the 3.5-2.5PPM range so I can probably take 1PPM with a pinch of salt and assume that the meter is pretty close - certainly within 10PPM but I can't help thinking that, at this level, there should be more information.

Also - the meter thinks that it was last calibrated in 2009, so they didn't actually take it through the cal procedure, should I be worried about that (I suppose they just took it through a calibration check and as it was in spec left it at that).

It appears fairly worthless to me.  And any cal lab using a 5730A would probably mention it!  And they'd also do a whole lot of other things differently.  What is the "1ppm" you mention?  There's no mention at all of the accuracy of the calibrator or the lab.  And what does 11/09/06 mean--is it a date?  November 9, 2006?  Or September 11, 2006?

[Psi]

There is definitely some fraud in the cert industry, with cal/emc certs generated without the work being done.
I'm sure many cal/emc companies do it 100% legit and take pride in being accurate, but there is definitely dodgy stuff going on too.

There's been a few threads on eevblog like this suggesting dodgy cal/emc certs.
The last thread I recall was one where the company had obviously copy/pasted the cert from a prior tested unit and changed the S/N. There was test results listed for optional features that the specific unit submitted for cal did not have.  Had that not been that case no one would have have known.

Many business asking for cert see it as a box-ticking exercise, and don't really caring enough to even read the piece of paper they get.  So the fact we do see some dodgy certs means there's likely a lot more of them that never get noticed.

[grumpydoc]

The cal is from 2011(?) any value it has is long gone.

This "cal" is from 11^th May 2023, I only included one sheet out of four for brevity - there's a front sheet which gives a few admin details but apart from mentioning IOS 9001 and ISO 10012-1 doesn't give any real info, then three pages of results. I've redacted the name of the company to protect the guilty party.

The 1PPM is my interpretation of what's presented - 19.00002 for an applied voltage of 19V is a 20uV error or roughly 1PPM - it's a bit higher on some of the other ranges e.e. 2.5PPM "out" on the 2V range.

[mendip_discovery]

It is a 9001 lab, not a 17025 lab, so they have more wiggle room. 9001 you state what you will do and as long as you do that everything is great. 17025, they send around some people who know their stuff and you have to prove to them you are competent etc. and they have a bunch of requirements that help you do things right.

Even for 17025, the customer doesn't need to know what the kit is or when it's due for calibration because under the continuous monitoring of the equipment that may change due to it potentially drifting off from specification. Plus a poorly looked-after calibrator used in a large temperature swing lab may perform differently than a well-looked-after, documented/characterised calibrator in a tightly controlled lab. Oddly enough it is not even required for the customer to know the idents of the kit used.

9001 allows them to get away with no stating uncertainties, that is partly because for a lot of labs this is the leap that takes them from 9001 to 17025. Some cal software will help them with this but if it's a small lab they may not have got their head around this yet. Technically unless its got an uncertainty you have no way to judge the measurements.

With regards to recalibrating stuff, unless you have good standards in the lab you risk making things worse if you adjust. Some test gear has some odd requirements (11.3 V at 33 GHz) which means that unless you have that very expensive calibrator then don't even bother adjusting. Also if you have a meter that is well within specification why would you adjust it.

I agree, it does seem slightly odd.  There are 1-2 least significant digits missing as well...

I have seen this before even in reputable labs. It's because lots of software likes to truncate numbers and unless you have set the certificate up correctly then you can lose the odd digit. I have a bit of a reputation for complaining with some labs, enough that they mentioned me to their 17025 assessors. This is mostly because I have to input the results from the certs into a spreadsheet so I can track the calibration over the years.

Just looking at the cert one of the things that stand out to me is the specification for the DC stuff, PPM then goes to show µV. I also don't like the 10 Ω statement as its very rare to have a 10 Ω exactly out of a calibrator, I would no doubt be ok for the 3.5 digit meters but at 7.5 you will be in the area where a few µΩ/Ω of error is visible.

I would say use this cert as it is better than nowt. How many bits of kit are sold on the bay each week with no history and statements such as "not tested" etc. At least you are buying with some confidence even if it is just a small lab doing a cheap calibration.

[Kleinstein]

A few of the zero readings look odd and would be out of specs: the 20 V offset and the 4 wire ohm short readings for 200 ohm and 200 K.
The 20 V offset looks odd in combination with the +-19 V readings.

[AVGresponding]

This can be a useful resource for someone looking for cal labs in the UK:  https://www.ukas.com/find-an-organisation/browse-by-category/?cat=708

[grumpydoc]

A few of the zero readings look odd and would be out of specs:

I assume these are readings are with the inputs shorted.

FWIW I get +00.00002 on the 20V range when set to 6.5d resolution - and +00.000027 in 7.5d mode.

Given that the spec seems to be ±counts in 6.5 digit mode I think that's OK and my meter is actually within specification.



But I'm not sure about the voltages that they specify - 2 counts on the 2V scale would be ±2.0µV, not 2.9µV and 20µV not 40µV for the 20V scale.

How do you get a 2001 to display 8.5 digits?

Thanks to mendip_discovery for an explanation of the differing standards levels.

Technically unless its got an uncertainty you have no way to judge the measurements.

This bugs me - as I said, I have assumed that they used a calibration source which itself was stable enough for a meter with a 12 month accuracy in the 20-30PPM range. If the rule of thumb is to have a source with 1/10 the uncertainty then you'd need to be in the vicinity of 3.5 to 2.5PPM - hence my example of the 5730A.

With regards to recalibrating stuff, unless you have good standards in the lab you risk making things worse if you adjust.

Which makes it reasonable that they would not run the meter's calibration and just run a check and only adjust if the meter was out of spec.

Just looking at the cert one of the things that stand out to me is the specification for the DC stuff, PPM then goes to show µV.

I guess working out the acceptable deviation in volts is easier for people to understand. 52µV is actually correct for the maximum deviation on the 2V range for a 1.9V input - but only for the 2001 itself. I really hoped they would have included their calibrator/meter's uncertainty as well - for approx 3.5PPM you'd have to add another ±7.0µV making the actual allowable deviation more like ±59.0µV

I also don't like the 10 Ω statement as its very rare to have a 10 Ω exactly out of a calibrator, I would no doubt be ok for the 3.5 digit meters but at 7.5 you will be in the area where a few µΩ/Ω of error is visible.

I don't know calibrators handle resistance - just switch in a precision component?

I guess the big boys will offer better precision but the best that Mouser or Farnell stock is in the order of 0.01%

I would say use this cert as it is better than nowt. How many bits of kit are sold on the bay each week with no history and statements such as "not tested" etc. At least you are buying with some confidence even if it is just a small lab doing a cheap calibration.

When I frequented rallies (I don't have time these days but might try to get to Newbury on the 25^th of June as that always used to be good) we always used to observe that there would be two bins - "Tested, working" and "Not tested".

The suspicion was always that "Not tested" meant "Not working" and I've tended to carry that logic forward to eBay.

I've no idea what the 11/09/06 date represents. It's not the date of this calibration check.


I've messaged the lab to ask what they used - if they tell me, great. If not, well, I'm no worse off than I am now and it will be a lab I know not to use

[Kleinstein]

Having an uncertainty 10 x better than the DUT specs is an ideal situation, but rare for the higher end instruments.  They can get around with less, but than should note it, when there are some results in the grey zone (reading + calibrator uncertainty gets out of spec for the DUT).
So it is not adding the calibrator uncertainy to the meters specs, but subtracting it to get the good zone. With a not so good calibrator chances get larger to get a values in the gray zone where they can tell if the meter is still in spec or not.

Getting a zero reading of 20 µV or 27 µV in the 20 V range is still not a good result. The specs seem to be for zero stability only, not the absolute value. Stability would need a time frame for that and with the waiting time not a spec one would like to test.
The data shown in the table are 240 µV and that would be bad - makes me wonder if it could be manual reading and a typo ?

I don't know the details on the K2001, but ideally there should be a point to adjust the zero. This may be even in the more user domain, outside the normal adjustment procedure. The short (or open reading for the current) to not need a very special cal instrument. It is a very real option to adjust the zero readings, but not the gain.  If the data shown are correct that shpould have been done with the last calibration.

The resistors from the calibrator may be actually precision resistors with trim, at least for the more standard values. Some calibrators have simulated resistance. It depends on the system used if they actually get accurate 10 ohm, or have some known deviation that needs to be taken into account for high accuracy.

I would not worry so much obout getting 8.5 displayed digits. The K2001 is more like noise and already the 7th digits is more like borderline. Thus they already give the stability with only 6.5 digits.
Chances are the GPIB provided data have more digits and one rarely wants to write down that many numbers anyway by hand.

[grumpydoc]

Having an uncertainty 10 x better than the DUT specs is an ideal situation, but rare for the higher end instruments.

Granted, once you get into the territory of 8.5 digit meters with single digit PPM uncertainties it might be a little difficult.

So it is not adding the calibrator uncertainy to the meters specs, but subtracting it to get the good zone

They didn't do that either - but I see your point.

Getting a zero reading of 20 µV or 27 µV in the 20 V range is still not a good result.

I don't disagree, it's my first 2001 though so I don't know what to expect.

It's 1-2µV on the 200mV and 2V ranges.

The specs seem to be for zero stability only, not the absolute value. Stability would need a time frame for that and with the waiting time not a spec one would like to test.

The spec mentions "variation in zero, 1 hour" so do seem to have a time-frame in mind.

I would not worry so much about getting 8.5 displayed digits. The K2001 is more like noise and already the 7th digits is more like borderline. Thus they already give the stability with only 6.5 digits.

Strictly for amusement value only - actually I found out how to get it, GPIB, as you correctly guessed, allows 8.5 digits to be enabled directly or you can use the "secret menu" (hold Math + Freq down at power-on - see https://xdevs.com/guide/keithley_secret/) and then 8.5 digits appears in the resolution menu options.

I agree - 7.5 digits on the 2001 is something of a triumph of precision over accuracy, especially given (as you say) it is widely recognised as a noisy meter.

My Krhon Hite 522 is almost certainly a bit high though - if we assume the 2001 is as close as the calibration report suggests it's putting out 10.000470V when set to 10V - the max error is supposed to be 250µV for 10V output.

[tautech]

I would say use this cert as it is better than nowt. How many bits of kit are sold on the bay each week with no history and statements such as "not tested" etc. At least you are buying with some confidence even if it is just a small lab doing a cheap calibration.

Sig:
Motorcyclist, Nerd, and I work in a Calibration Lab :-)
--
So everyone is clear, Calibration = Taking Measurement against a known source, Verification = Checking Calibration against Specification, Adjustment = Adjusting the unit to be within specifications.

This, all this.

Never noticed your Sig before .....too much speed reading. 

One of our customers has from what I have seen, the best Cal lab in NZ of which I've had a tour of a couple times.....after visiting they need mop the floor of all the drooling everywhere tautech has been. 

[J-R]

For sure don't expect the 8.5 digit mode on the K2001 to be comparable to a true 8.5 digit unit, but I tested it out with various DCV references that I had calibration data for and found it better than expected.  You want maximum NPLC as well as advanced filtering if I recall correctly.  As with most DMMs with such a feature, one of the primary areas I found it works well is at the lower end of the range.  So don't expect too much at 1.9V for the 2V range for example, but it can be great at 3V in the 20V range where you'd otherwise lose a whole digit.

When performing such operations you really need the calibration data and you need to reproduce the calibration conditions.  Then these older units can do more than what is expected, IMHO.