New secondary voltage standard

[MiDi]

Had a very interesting talk with PTB about a project for a new secondary 10V standard they are currently working on.
As there is the permission to talk about it, I can give some details - even I do not really understand this sophisticated stuff in detail.

This project is aimed to fill the gap between todays secondary voltage standards and JJA.
The key specs are <0.14ppm/year drift, TC <0.014 ppm/K and LF noise <0.041 ppmpp (0.014-14Hz).
It is based on a standard voltage reference in conjunction with a sophisticated PLL.
This allows to calibrate/adjust the voltage with a frequency standard e.g. GPS disciplined oscillator - so this can be made on site without need for cal-lab.

The PLL consists of a OCXO as frequency reference and a voltage controlled oscillator.
The voltage is controlled by a very precision combined DAC consisting of a very precision DAC for the MSBs and an overlapping precision DAC for LSBs which is adjusted to 10V (+-0.14ppm) and maintains this value due to the PLL and the OCXO.
The standard voltage reference is boosted with the DACs to 10V (DAC output is connected to voltage reference) - they act as the voltage divider made of precision resistors in classic 10V standards.

Edit 2.4.: corrected all specs

[branadic]

I wonder how they made frequency to voltage conversion, as this is the tricky part, right? The best f-V-converter I'm aware of is AN14 "Ultra-Linear V→F Converter" that achieves "This V→F has a resolution of 1ppm, with linearity inside 7ppm (0.0007%)."

If this is simply a 1st of april fool's joke, I call that infantile and a comment that is not in the proper section!

-branadic-

[David Hess]

It seems dubious to me.  Voltage-to-frequency and frequency-to-voltage converters rely on charge metering in one form or another which ultimately depends on a voltage or current reference.  This might be useful if capacitor matching, including charge injection, is better than resistor matching but I am not convinced that is the case.  The exception might be precision dividers and multipliers where the capacitors can be identical and the frequency is fixed.

The DAC output should be divided as much as possible consistent with the reference accuracy for the minimum trim range to prevent DAC drift from contributing to reference drift.

I wonder how they made frequency to voltage conversion, as this is the tricky part, right? The best f-V-converter I'm aware of is AN14 "Ultra-Linear V→F Converter" that achieves "This V→F has a resolution of 1ppm, with linearity inside 7ppm (0.0007%)."

The best frequency-to-voltage converters I am aware of enclose a voltage-to-frequency converter inside of a feedback loop.

What I would like to know is, where are the precision delta-sigma DACs?

[guenthert]

[..]
If this is simply a 1st of april fool's joke, I call that infantile and a comment that is not in the proper section!
[..]

I had a laugh nevertheless
Well played.

[Dr. Frank]

That's no Aprils fool, or maybe one which went in the wrong direction.
The PTB obviously has a Reasearch Project with WEKOMM, for an improved voltage reference.
This is official, and branadic had linked this recently.
Can be found on the PTB site.
Frank

[David Hess]

I took it seriously because I have seen this proposal before.

[guenthert]

Works for me either way. 

[TiN]

Traceability of this setup would be worst metrologist nightmare, like we didn't had enough of that with SSRs and ageing prediction/etc.
Beauty of JVS is that it is "simple" and rely on fundamental physics. It's the details and implementation that are expensive to do 

[Andreas]

Edit 2.4.: corrected all specs

and why not a 1.4. V Standard?

[branadic]

Information in first post is somewhat limited.
However, I came across this recent short paper, with PTB obviously being involved: https://www.ama-science.org/proceedings/details/3798

So the mentioned approach of the PTB with GPS has nothing to do with the work PTB is doing together with Wekomm.

-branadic-

[ramon]

After reading, does that means that we can experiment with those old transistor arrays in DIP package and some good resistors (or resistor arrays) and get better performance than a LM399A? 

[Kleinstein]

There are already some low noise bandgap references available (e.g. LTC6655). Noise wise they can compete with buried zener types at a comparable power level - though I don't see a much better performance. The lower voltage may however be more suitable for single chip SD ADCs.

The weak point is however the long time drift, that is highly dependent on the case.
Separate transistor array would have a similar case problem.

[niner_007]

Is it LTZ1000 based? Is all the voodoo to get 10V? Why not just buffer the reference (7V out), you can't really beat that if you want ultimate stability. Wondering what the 1 year stability of a well aged LTZ1000 looks like, someone on the forum said <0.5ppm stability. Fluke can't get it under 1ppm, and they got all the fancy ovens.

[SilverSolder]

I'm patiently waiting for a Josephson junction array in a TO-18 case!

Maybe April 1, 2021? 

[razvan784]

The paper is a little short on details - construction, thermal matching, environment control measures...
Note how, eventhough the DYI bandgap achieves much lower uncertainty than the Zener on a ~200 s time scale, it begins to increase thereafter, indicating poor stability.
Maybe one can use power transistors (having huge junctions) to make a low-noise bandgap. Maybe those "perforated emitter" types that Noopy has been tearing down around here. They could be mounted on a heatsink, maybe surrounded with thermal insulation and then placed in an aluminium box - that should help with temperature equalization.

[niner_007]

One other thing, mentioned by other, the key spec of 0.14ppm/year, if it is stability or uncertainty, it doesn't add up, given the DAC and clocks, the error budget seems optimistic.

[bdunham7]

I Fluke can't get it under 1ppm, and they got all the fancy ovens.

I'm not sure that's actually true.  You're sort of making an apples-to-oranges comparison.

[niner_007]

You’re right, they are not using LTZ1000, but rationally, surely they’d deliver better than 1ppm if they could, they are not in a business shy of charging customers $

[bdunham7]

What I (also) meant was that comparing the results of a hobbyist who, perhaps accurately, determines the uncertainty or drift of his LTZ1000 reference to be 0.5ppm to Fluke conservatively specifying an uncertainty with guardbands and a 99% confidence interval is not an even comparison.  If you assume 20% guardbands and 3-sigma confidence (OK, that's 99.7%, but close enough) and a random-ish distribution, then half of the 732Cs will actually be within 0.2ppm.  Perhaps my math or something else is off, but I'm not seeing a 0.14ppm standard as such a huge advance unless it is economical or has some other advantage.

[Magnificent Bastard]

My Fluke 732B/AF is already at -0.133ppm/a...

[notfaded1]

My Fluke 732B/AF is already at -0.133ppm/a...

I wish I had some B's... only A right now and homemade so far.  Some new competition for the LTZ would be neat but time has shown it's not so easy minus Fluke's custom buried zener.  It seemed to me at first that there must be new technology that can rival it but people here and Dr. Frank have convinced me it's a lot harder than it seems like it should be.  With nano technology, quantum, and you'd think right?  Well at least there's some new tech on the timenut horizon vs. voltnut horizon.  Optical with microwave and ytterbium and strontium instead of rubidium and cesium:  https://spectrum.ieee.org/tech-talk/semiconductors/optoelectronics/optical-atomic-clock-advantage-expands-electronics

So we may not have a new voltage standard but we're defining the second a lot better these days.

[branadic]

Be patient, people are working on it, but it will still take some time to come up with the next generation of voltage references.

-branadic-

[The Soulman]

It kinda reminds me of:

https://www.eevblog.com/forum/metrology/building-your-own-voltage-reference-the-jvr/

[dietert1]

My Fluke 732B/AF is already at -0.133ppm/a...

Could you write a sentence or two to explain what you compare to and how you do the comparison? Can you show some kind of log. maybe of several years?

I am currently running a comparison of two well built references and although i see a standard deviation of only 0,029 ppm for the daily comparisons, even after 40 days the statistical uncertainty at 99 % c.l. of the drift is about +/- 0.4 ppm/a, let alone systematic uncertainties.

Regards, Dieter

[guenthert]

It kinda reminds me of:

https://www.eevblog.com/forum/metrology/building-your-own-voltage-reference-the-jvr/

How so?  The JFET based voltage reference goes back to an observation originally published by L E MacHattie in 1972 (May 26th, not April 1st), has been cited in patents (US4068134, US4158804, US3975648A), exhaustively discussed and can readily be reproduced.