Perhaps I am wrong in my understanding by what it means to be traceable to the SI through NIST, and directly traceable to NIST
Our A2LA/17025 accredited 732B test report has the boiler plate text that includes "...this certification is traceable to the internationally accepted representation of the SI Volt through the Josephson constant (KJ-90 = 483597.9 GHz/V) and the use of a Josephson voltage standard operated by [my lab]". We definitely don't claim NIST traceability for the Volt. If you didn't have a JVS you would need to, but if you can create the volt in your lab that's where the traceability ends (well, actually it ends wherever your frequency source traceability ends, but that's a whole 'nother domain). For intrinsic standards like this and the QHR, the auditors from the accrediting agency look at your inter-lab comparisons and uncertainty calculations to decide whether you can claim a certain uncertainty on your scope of accreditation. NIST or other NMI traceability only applies to non-intrinsic standards.
Metrological traceability requires an unbroken chain of calibrations to stated references, all having
stated uncertainties – refer VIM [1]. The persistent misconception that metrological traceability may be
linked to a particular organization (e.g., “traceable to a specific National Metrology Institute”) fosters
continued confusion with regard to its nature. Metrological traceability pertains to reference quantity
values of measurement standards and results, not the organization providing the results.
2. NIST uncertainty (95% confidence) is spec'd to 5nV/V on their JVS. Fluke's uncertainty (95% confidence) their JVS is currently is 5nV/V PLUS extra 50nV if you want a traceable measure (That's where Fluke's measure spec uncertainty of 100nV or 0.01ppm on 10V scale comes from). NIST has the luxury of being the golden standard, so they don't have the extra added transfer uncertainty. They ARE the definition of the volt for US. That being said they still intercompare will all the bigger labs - everyone is trying to keep the standards in close agreement.
3. For our needs and doing business with our customers, we MUST have have the words "traceable to NIST" on both absolute value and uncertainty. "Traceable to Fluke" is worthless a lot of times. You will see in several places in Fluke documentation that yes you can get a measure at Fluke to better than 0.1uV/V U95 BUT that is relative to Fluke's Volt. The NIST traceability for uncertainty will be larger (that part appears in the fine print).
4. YOU CAN cut out Fluke and send your Fluke 732a / b to NIST for Measure and Cal, and they will measure it against the NIST JVS for 30 days, and give you a characterized report. The fee is $2311 and you pay for all shipping.
https://www.nist.gov/calibrations/voltage-measurements-calibrations
The official reports will use 1uV/V uncertainty. For the rest of this post I will assume we are on a 10V scale so that 1uV/V is 1ppm, and 95% confidence.
NOW - As explained by NIST, for any known Zener Diode Reference, about the best anyone is going to do is about 0.2ppm uncertainty. They have seen it go as low as 0.1ppm uncertainty but that is rare. So the Calibration is quoted as the best you'll get is 0.19ppm, and the normal range of uncertainty on any Zener Diode Vref they've seen is around 0.1 to 0.5ppm
The limiting factor here is the noise of the Vref. This includes Zener Noise, Temperature, Humidity, Pressure effects on the system. So in the words of NIST: They watch your Vref for 30 days at a measuring resolution of down to .005ppm and then look at the overall drift and noise - and assign an absolute voltage value and 95% confident uncertainty, which will be no better than 0.19ppm.
The other problem NIST pointed out: You can measure a 732a/b to some ridiculous fine resolution on a JVS and come up with some value. The problem is making use if that measured value: Unless you're going to use that measured value within a very short time frame, trying to get a transfer measure out of your Vref much below 0.1ppm is like trying to measure bowl of jello. And forget transporting the 732 and expecting that measure to still be valid at a very small uncertainty.
5. Zener noise of 732a's is spec'd at 1uV RMS or less (typically way less than that when they age, ours are down around 35 or 50uV). 732b is spec'd at 60uV rms (.1 to 10Hz), and usually measure around about that. You are never going to get a Cal certificate better than 0.2ppm uncertainty because anything below that uncertainty is going to be a measure of mostly the noise of the Zener Diode Vref - and not a very stable or usable measure.
So the bottom line is YES you could get your 732 measured at NIST - but all that happens is you might get 0.19ppm uncertainty. Send it to Fluke and measure against their JVS and you've dropped all the way down to 0.2ppm uncertainty and saved a few bucks. Plus Fluke will perform an adjustment if you need it.
NIST is only going to work with the "Traceable to NIST" uncertainties with values they can verify. And that lower uncertainty limit - for now at least - is about 0.19ppm for a zener diode reference.
Have Fun!
I am looking at a live data sample from what is considered a very good 732b testing at NIST right now on their JVS and I'm seeing more like +- over 200nV spikes over just a couple days, not weeks. It gets worse than that over 30 days.
So you might want to check with your buds and see what the real story is - otherwise you have some extraordinarily good 732's!
Yes, those are good math gymnastic stunts, and then on a simpler level: consider the Zener's inherent noise...say .5uV to a 1uV RMS, .1 to 10Hz - at any given time. These devices do have a noise floor at the moment you pull a measure from them. One measure is never enough.
Of course as dacman pointed out: the very best predictor is to develop an un-adjusted long term drift history which is what I pointed out a long ways back in the first post. A 732 with a long (many years) recorded calibration report history is much more valuable than a new unit, I say.
Back on topic: How does this all relate to purchasing a 732 -Traceable Calibration Certificate- at Fluke with better than 0.2ppm uncertainty in the first place? I'm still kind of wondering how that happens - or if that's something they just don't do anymore? I've talked to 4 people at Fluke, from sales to lab people - all say that just isn't available for the 732.
MisterDiodes did not say anything about taking one measurement and calling it a day, read it carefully, "One measure is never enough."
He is not talking about curve fitting either, he is talking about taking all of the actual readings over a period of years (even more than one reading a year if you're adventurous) and doing a simple average of it, that will give a rather accurate figure for the next drift measurement. These references tend to drift in a general direction at a fairly constant amount every year once they are old enough. No curve fitting or any other mathematical high jinks are needed.
It is also a fact that you simply cannot average out very low frequency noise over relatively short periods of time (we're talking days at least for an average), these low frequency components are indeed random and cannot be predicted or averaged by any math for short periods. It is this noise that limits the absolute accuracy and for 732s, that is about 0.2 PPM, sometimes a little lower but that is the best you're going to get and no amount of math fiddling is going to change that. If that was true, NIST and everybody else would be using such techniques to reduce the uncertainties further beyond the noise floor. Ask the experts, they'll give you the same answer, 0.2 PPM for a 732 , maybe 0.1 PPM for an exceptional one. That is all that needs be said, this was the intended subject of this thread in part, nothing to do with JVS as such although the information is indeed interesting.
I'm asking if there was a way to purchase a 732 traceable fluke cal certificate better than 0.2ppm uncertainty. It sounds like what dacman possibly got was a "measure" lab service product from Fluke, not a traceable cert. maybe? Again: Fluke currently says "No" to any 732 traceable cal cert any better than 0.2ppm (Service center or Park Cal, at least to the four people I've talked to), which is exactly how I reported the process on the first page of this thread.
I've also looked back at a bunch of 732 docs for older cals, and it's never been anything better than 0.3ppm to 0.5ppm for the "standard" NIST traceable cal certs from Fluke, even from over 20 years ago.
A Congressional act in 1904 established the U.S. Legal Volt to be a quantity defined by the National Bureau of Standards, now the National Institute of Standards and Technology (NIST). With the 1990 international agreement on the Josephson representation of the volt, NIST defined the U.S. Legal Volt to be the same as the international volt representation. Since the success of the first Josephson array voltage standards in 1984, their use has proliferated to more than 70 national measurement institutes (NMI's), military, and commercial laboratories around the world. This has resulted in some confusion about the traceability of non-NMI's that are in possession of a JVS that is, in principle, as good as the national standard. Some guidance on this question is provided in International Standards Organization (ISO) documents that state the general principle that intrinsic standards like the JVS, that have participated in a comparison with an NMI, can claim traceability.
While the voltage appearing across the terminals of a Josephson device is, in principle, given exactly by V = n f / K J, in any real measurement there are a variety of potential sources of error and uncertainty as listed in Table 2. In the case of a known error, such as a reference frequency offset or a known leakage resistance, a correction can be made. It is then the metrologist’s task to assign realistic numbers to all uncertainties including the uncertainty in the corrections. One method of doing this notes that only items 1 and 2 in Table 2 depend on the voltage across the Josephson array. All of the other components are about the same regardless of the array voltage. Therefore, the combined effect of items 3-8 can be quantitatively evaluated by making a set of measurements of a short circuit using exactly the same algorithm that is used for any other measurement. The standard error resulting from items 3-8 is just the root mean square (RMS) value of the set of short circuit measurements.[64] Additional experiments must be performed to estimate frequency and leakage uncertainty. Internationally accepted procedures for combining uncertainty and establishing confidence intervals is the subject of the BIPM’s Guide to the Evaluation of Uncertainty in Measurement.[65] Typically, the total uncertainty contribution of a Josephson system in a measurement averaging time of a few minutes is a few nanovolts. Since the most common use of these systems is the calibration of Zener standards with a noise level of 50-100 nV, the contribution of the Josephson system is negligible.
Finally the veil of bullshit is pulled away. Verified info from Fluke (lab manager #5 I've talked to) and NIST (PhD metrologist).
First, in response to CalMachine:
Yes Fluke will do a raw measure (to ISO17025) of a 732 and get you to .06ppm uncertainty against Fluke's JVS. As I have explained several times. That is a certified "voltage measure", not a calibration certificate. A Calibration Certificate (to ISO 17025 AND Z540-1-1994) from Fluke also includes a Test Uncertainty Ratio of 10:1 and the entire instrument is tested to be in spec. A Calibration Certificate is what I have to supply to my customers, they would never, ever go for just a certified measure. Yes an ISO17025 qualification includes traceable to SI through NIST, but you have to make the distinction between a calibration certificate vs. just a voltage measure.
The little tidbit you're leaving out from a Fluke .06ppm uncertainty measure is the warning from Fluke that uncertainty is valid for about 2 hours on a typical 732. That's why they can do this as on on-site measure down to .06ppm with some multiple freshly cal'd 732's, do your measure on-site, and then get back to the JVS and recal those 732's right away.
To recap: A Fluke Calibration Certificate on a Fluke 732 is performed to ISO 17025 and ANSI Z540-1, and is traceable to the SI through NIST, and if it is purchased thru Fluke that includes a -complete- instrument verification test and 10:1 Test Uncertainty Ratio (TUR). This is the practical usable limit of a single 732 if it's being shipped anywhere where traceability back to SI / NIST volt is maintained.
NOTE: You can also have a Calibration Certificate and more Certified Measures done at more frequent intervals to help characterize your 732 to lower uncertainty - but that will vary from unit to unit and has a drift history is developed.
Now for use in your own lab you can of course do whatever statistical math you like on your own 732 . For actual commerce beyond your lab it's the Calibration Certificate measurement uncertainty that counts as the basis for any product measure we send to our customers. At least that's only what my customers demand to see - If I tried to claim some measurement uncertainty based on some "certified measure" they'd just laugh at me. I have to be able to show the complete third party traceable calibration certificate.