Who calibrates the calibrators?

[CaptainNomihodai]

I have a question that may be stupid, impractical, and/or bordering on the philosophical: who calibrates the calibrators?
My understanding is that, for example, if I have a DMM and need it calibrated I send it out to someone who uses what's essentially another DMM with more digits. But surely that needs calibration too, using a machine with yet more digits? But then who calibrates that? Where does it end and why does it end there? I know there are expensive voltage and resistance references you can buy, but how do we know those are accurate? The whole thing seems necessarily circular (i.e. you measure the accuracy of a meter against a reference whose accuracy was measured with a meter whose accuracy was measured against a reference and/or more accurate meter whose accuracy was measured against a reference, and so on...).
What am I missing?

Apologies if this has been asked and answered before (I'm sure it has), so feel free to bash me for not sifting through the search results more thoroughly.

[Avacee]

Dave's video of a visit the Keysight/Agilent Standards Laboratory in Melbourne might be of interest



with an explanation of what calibration is:



and how about a portable Cal Lab!

[ap]

The national standards laboratories (NIST, PTB...) are the ones who have the most precise representation of the various standards and are the ones responsible to maintain and improve them (i.e. decrease their uncertainties). They do this by applying new methods to define the standards (e.g. from other well defined standards; time/frequency is a preferred one due to the uncertainties achieved here) or by decreasing uncertaintes in existing setups.
These standards are then used to calibrate other standards used in cal labs. You can go to national standards labs yourself and have your gear calibrated. Not cheap though.
In a nutshell. There is a lot of information available from them.

[daybyter]

Some more info:

https://www.eevblog.com/forum/testgear/10-digit-voltmeter!/

[Moon Winx]

The national standards laboratories (NIST, PTB...) are the ones who have the most precise representation of the various standards and are the ones responsible to maintain and improve them (i.e. decrease their uncertainties). They do this by applying new methods to define the standards (e.g. from other well defined standards; time/frequency is a preferred one due to the uncertainties achieved here) or by decreasing uncertaintes in existing setups.
These standards are then used to calibrate other standards used in cal labs. You can go to national standards labs yourself and have your gear calibrated. Not cheap though.
In a nutshell. There is a lot of information available from them.

Yes, and to add an additional layer, the national standards labs periodically perform "key comparisons" with each other, coordinated by BIPM in France. It determines how well each nation's standards compare with each other. Just about every parameters can be created in a lab now aside from the kg which is still based on an artifact (although this will probably change in a year or so due to several national labs developing watt balances).

[CatalinaWOW]

It may be easier to understand what is going on by returning (in thought and memory) to the way these things used to be done.  There were physical items that by agreement were the standard.  The meter and the kilogram were physical pieces of hardware in a laboratory in France.  Those were the reference values, and people compared their secondary standards to those.  The comparisons grew more sophisticated over time, and there was always a great deal of effort spent understanding the sources of error in the comparison (and reasons why those standards might be changing over time).  The volt was defined by a specific type of battery operated in specific labs.

The process is fundamentally the same today, even though almost all of the primary standards have been defined by repeatable physical phenomena.  Your voltmeter is actually traced to one of those repeatable physical phenomena, a Josephson junction.  By agreement the voltage across that junction under a given set of conditions is the defined standard for a volt.  People measure that voltage with incredibly precise voltmeters.  In principal, they could measure your voltmeter with that incredibly precise voltmeter to provide a calibration, but you might not want to pay the bill.  So someone measures their less precise voltmeter with the really good one, and establishes bounds on its accuracy.  Since  it is a far simpler instrument and far easier to use while achieving its stated accuracy it can cost less to evaluate your instrument with this one.  That chain may continue until a price point is reached that meets your needs.

The key to all of this is agreement that something will be the standard.  Then everything else is just a daisy chain of comparisons.

[CaptainNomihodai]

The process is fundamentally the same today, even though almost all of the primary standards have been defined by repeatable physical phenomena.  Your voltmeter is actually traced to one of those repeatable physical phenomena, a Josephson junction.  By agreement the voltage across that junction under a given set of conditions is the defined standard for a volt. 

I think that's the answer I was looking for. So then it's up to science wizardry to try to determine more and more precisely the conditions under which a Josephson junction has a volt across it?

[CatalinaWOW]

The process is fundamentally the same today, even though almost all of the primary standards have been defined by repeatable physical phenomena.  Your voltmeter is actually traced to one of those repeatable physical phenomena, a Josephson junction.  By agreement the voltage across that junction under a given set of conditions is the defined standard for a volt. 

I think that's the answer I was looking for. So then it's up to science wizardry to try to determine more and more precisely the conditions under which a Josephson junction has a volt across it?

More accurately there is no such thing as a physical volt.  There is an agreement that a potential difference defined by certain physical conditions is called a volt.  That has no more relation to fundamental physics than the lump of platinum that used to define a kilogram.  Some alien species might choose a different lump, or a different set of physical conditions.

The definitions in terms of repeatable physical parameters are important in that they are believed to be stable over time, and repeatable at multiple locations.  These two conditions enable comparisons with measurement systems for calibration purposes.

[CrazyTiger]

Most of the time calibration lab calibrates multimeters using calibrators like Fluke 5720A, 5700A, 5730A (Fluke's latest model with touchscreen  )  those are in turn calibrated with a characterized voltage, resistance standard like the Fluke 732B DC Reference Standard and precision multimeters like 3458A.

These characterized voltage standard also called transfer standard it will be calibrated or compared with Josephson Junction Voltage System, usually these can be found in National Metrology Centers like NIST, OEM's like Agilent and Rohde & Schwarz have them too. These JJV's are bilaterally compared with other JJV's across the world, to constantly verify their accuracy. The most important thing in calibration is the measurement uncertainty, the lower the better. There is never an accurate lab, its just which lab have the least measurement uncertainty.


Another important part is, during calibration the lab take takes advantage of short term stability, usually when calibrating highly precise multimeter like 3458A, the reference multimeter will be required to be calibrated at 90 days interval, in certain case, the reference need to be calibrated at 24H Interval, this ensures the reference multimeter to be extremely accurate. Take a look at Agilent's Primary Standard Lab:

[Moon Winx]

...and the Josephson Voltage Standards accuracy are dependent on the accuracy of the frequency fed to the junctions along with defined constants. Time & frequency are parameters that can be measured to parts in 10^13 or better, so considering the JVS's 732B measurement accuracy is around 40 nV (@ 95% CL), the frequency is several orders of magnitude more accurate. In fact, I believe defined constants in the (planck's & electron charge) in the voltage equation have a much higher uncertainty than the frequency used to produce the volt.

[HighVoltage]

One related video from Keysight Loveland - USA:

Interesting video that can only be fully appreciated by stopping every few seconds.
What liquid do we see at 1:23 where all the resistors are stored?

[Echo88]

White Mineral Oil, look around 1:20

https://www.sonneborn.com/product/blandol-0

[SeanB]

Mineral oil,in an open system. Probably a SAE10 weight high purity napthetic mineral oil, selected and dried using a molecular sieve to get low water, and then kept stirred in the bath along with having a pump taking it through a micropore filter and some molecular sieves to keep it dry. Probably has a reverse current heat exchanger in the pump feeds to keep incoming and outgoing streams as close as possible in temperature so there are no thermal gradients in the bath.

Echo88 got the manufacturer, thank you.

[HighVoltage]

White Mineral Oil, look around 1:20

https://www.sonneborn.com/product/blandol-0

I did not see that, thanks

The video I have watched a few times now, just amazing.

[CalMachine]

One related video from Keysight Loveland - USA:

Interesting video that can only be fully appreciated by stopping every few seconds.
What liquid do we see at 1:23 where all the resistors are stored?

I spent over 30 minutes watching it    I wonder what they didn't show?  I would have thought their Volt would be more robust than what was portrayed. 

I did not expect to see the 3-4 3458As at each station.  I was happy to see the old brown Datrons still being used!

[TiN]

3-4 3458As at each station.

What did you expect, 8508A's or 2002's instead?
3458A has power of ACAL, and linear ADC which may be quite important for metrology use

[CalMachine]

3-4 3458As at each station.

What did you expect, 8508A's or 2002's instead?
3458A has power of ACAL, and linear ADC which may be quite important for metrology use

To be honest, I had expected 2 3458s and an 8508!  Then either a 5790 or 4920.  Not surprised most stations are using 5790s as the 4920 just went out of support from Fluke in April   I might reach out to keysight and see if they will support the 4920 (hopefully the will after putting out a video like this!  )

Looks like they have show some AH cap bridges for a split second, too!  Wish I could get one of those for the lab.

The 3458A ADC is definitely something to write home about.  It's phenomenal how linear it is...  I feel extremely ignorant for thinking I could improve the ADC just by increasing timebase accuracy by a few magnitudes. 

[Conrad Hoffman]

It's probably mentioned somewhere, but something important to understand is that even though establishing the standards can be difficult and expensive, almost anybody, with relatively simple equipment, can do comparisons to far better than 1 ppm. Thus, the daisy chain of comparisons that finally reaches your personal meter, loses very little of the original accuracy, if things were done properly.

[fcb]

I had a tour of the National Standards lab of Malta in 2007 (I had an office in the same building and needed to weigh my suitcase before getting on a plane...)

Saw the standard kilogram of Malta, etc.. but the best bit was the Temperature lab, not the mineral oil baths and amazing looking platinum resistor transducer things, but the things the measured these transducers with:

F18 and F900 thermometry AC bridges made by the English company ASL (Automatic Systems Laboratories) - huge pieces of kit, really very beautiful.  Looks like they have been aquired and the F900 has been superseded by the F9000 - read it and weep:

http://www.wika.co.uk/ctr9000_en_co.WIKA?ProductGroup=72441

Another thing of wonder were these massive isolated granite table/pillars that went through the floor down many feet and were bonded into to the bedrock.

[branadic]

Hi,

I found this whitepaper by NI today and thought I should share it with you.

How to Achieve a 0.01 µV/V Deviation on Your 10 Vdc Proficiency Test Without Using a Josephson Array

-branadic-

[Cerebus]

More accurately there is no such thing as a physical volt.  There is an agreement that a potential difference defined by certain physical conditions is called a volt.  That has no more relation to fundamental physics than the lump of platinum that used to define a kilogram.  Some alien species might choose a different lump, or a different set of physical conditions.

Not quite. Although the volt is arbitrary, it is not arbitrary in the way the kilogram is. The volt is not a measure of a fundamental physics phenomenon, but it is defined using fundamental physical phenomena. The kilogram is defined as "the mass of that lump of platinum iridium alloy over there (pointing towards the BIPM in France)". You couldn't communicate that to aliens because they don't have access to the prototype kilogram. Whereas the modern volt is defined in terms of fundamental physical constants that you could communicate to aliens and which they could then use to reproduce a volt themselves.

[AlfBaz]

Our version of NIST is NATA, they list testing facilities in Oz.
Note the difference in uncertainties between Keysight and our national measurement institute
Keysight
2.5uV at 1V nominal
25uV at 10V nominal
NMI
0.04uV/V at 1V and 1.018V
0.02uV/V at 10V

Not sure why keysight state "nominal" voltage and the NMI seem to be using a different standard for 1/1.018V and 10V

[Cerebus]

Our version of NIST is NATA, they list testing facilities in Oz.
Note the difference in uncertainties between Keysight and our national measurement institute
Keysight
2.5uV at 1V nominal
25uV at 10V nominal
NMI
0.04uV/V at 1V and 1.018V
0.02uV/V at 10V

Not sure why keysight state "nominal" voltage and the NMI seem to be using a different standard for 1/1.018V and 10V

I suspect that you're comparing apples and oranges there, e.g. one is for Keysight's run of the mill calibration service* (such as what they'd do for a handheld DVM) and the other is for a primary standard.

*Where the calibration chain might be National standard->HP 'house' standard->'gold' HP 3458A->calibrator->your instrument.