Another method to get a precision voltage reference / does it make sense

[pelule]

Just read a thread in [volts-nuts] regarding the US patent 6091281 (attached)
https://www.mail-archive.com/volt-nuts@febo.com/msg03027.html
Thought a bit about it:
1 - use a rubidium based GPSDO (I just working one)
 +  the rubidium deliveres a high short term stability
 + the GPSDO deliveres long term stability/accuracy
 == quite good reference frequency
2 - yet unclear for me
  ? what kind of VCO to use
  ? does ait really need high lineary (my guess not, but good short term stability)
  ? need good (low) long term drift

Does that approach make sense?
It seems worth to give em a try.

[Vgkid]

I will say that that was a rather interesting bunch of posts to read through.

[pelule]

Yepp, I also followed that intersting thread. That 400k$ is far, far out of my scope, even I also own a 3458A and a Datron 1281.
My thought during reading that thread was:
If I don't need it for cal business (with return of investment), there would be a cheaper way fitting my needs.
F. e. buy a quite good voltage reference (f. e. Fluke 732B, say cost is 5k$) and let calibrate it regular, let say 10 times a year.
Assuming a good qual cal lab ask 1k$, it could be done >300 times for the same amount of money.
if cal is done 10 times a year, that means 30 years cal history data - a good statistical basis.

Regarding the "another method" - had same thougth, that's why I have noted questions to answer:
  ? does it really need high lineary (my guess not, but good short term stability)
  ? need good (low) long term drift
I fully agree, the drift (time, temp. humidity) are in any case the main quality criterias for any reference.
As anything is drifting, this "reference" would drift also. Just how much is the question.
If a drift below 2ppm/yr is realistic it would worth to to try.

[branadic]

I can say nothing about the presented approach, but the question is, what are you really looking for? Do you really need the VOLT inside your homelab or do just want to know what's the real output of your voltage reference?

I know about activities realizing the latter. That means setting up a measuring system allowing you to precisely measure voltages without the need of an ADC + voltage reference by precisely measuring E-fields. I can't go into details, as all this is very novel and at the very beginning and far away from being integrated into a small device yet.

[pelule]

@branadic
My interest is just "interest" and "shall I try to build one" for fun in working on precision analog.
I am just finishing my Rubidum based GPSDO - and have seen that thread.

I don't need "the VOLT" and have no urgent need to know the real output of my voltage references.
I regular test my meters against my references (Guildline:9152T/P4, HP735) just to detect potential issues, I also log the meaures for high drifts detection (fully accepting the increasing uncertainty).
It's the typical men with two clocks problem - just I don't need the exact time, I just like to know, if the clock are operating.
My 3458A got the last cal 6 month ago - that's my "reference".
The measure logs show less than 0.2ppm/yr difference of the 1281 against the in-cal 3458A.

[David Hess]

I checked some datasheets and it looks feasible except for one thing.  Short term drift should be better than an LTZ1000 but long term drift over a period of years is about equal.  The problem may be noise which my guess is will be an order of magnitude higher at low frequencies; varactor based DC amplifiers were not known for low DC noise.

Given the cost of complexity, a reference using several LTZ1000s in parallel could be built for the same price.

[Kleinstein]

I don't think an VCO makes much sense - usual VCOs use a varactor diode. So the sensible way would be a bridge with a varactor diode and an other capacitor that is forced into balance with the right voltage. However there are quite some difficulties with stray capacitance and nonlinearity of the varactors. Also the TC of the usually varators is high (e.g. 2 mV/K).
To control the stray capacitance problem one might to have the core of the reference on a single chip.
Still I don't  think this can compete with a good zener reference.

There is one attractive feature compared to a zener diode: a capacitive bridge does not dissipate much power, so it is easy make it larger and keep temperature controlled.


There might be a chance to get a good current reference though, based on a high temperature squids. Somewhat similar to the JJA, but it is known to work at 77 K and easier to design than an JJA.

[David Hess]

I don't think an VCO makes much sense - usual VCOs use a varactor diode. So the sensible way would be a bridge with a varactor diode and an other capacitor that is forced into balance with the right voltage. However there are quite some difficulties with stray capacitance and nonlinearity of the varactors. Also the TC of the usually varators is high (e.g. 2 mV/K).

To control the stray capacitance problem one might to have the core of the reference on a single chip.
Still I don't  think this can compete with a good zener reference.

For VCTCXOs and VCOCXOs, the control voltage has to be taken into account for the aging and stability.  It does not need to be linear but it sure has to be stable.  If the control voltage is held constant, then the output frequency needs to remain constant within the specifications of the oscillator.

The varactor temperature coefficient is 100s of ppm/C so they must be compensating it in some way.

[pelule]

So my conclusion so far is:
the method would make sense, if following conditions could be archieved:

• the reference oscillator keeps stable over temp and time, at least below the target precision
          seems to be feasable with a rubidium based GPSDO
• the voltage/frequency relation of the VCO keeps stable over temp and time at least below the target precision
          this seems to be the most critical part of the design
          temperature dift may the less critical effect, if the VCO is kept inside a temperature controled enclosure
• noise may a critical issue
          advantage may, the follower circuits, like low pass filter etc. would have no impact into the drift calculation

a low long term drift, say 0.1ppm per year would be acceptable, it just requires re-cal time to time as any reference.

[Kleinstein]

The VCXOs are just using the crystal oscillator to sense the capacitance change of the varactor at the input of the VCO. I see little use in testing the varactor by means of a crystal oscillator instead of a more normal bridge circuit. A crystal oscillator has rather low sensitivity to capacitance and thus needs the high stability clock.  With a adjustment range of 100 ppm for the VCXO the drift of the crystal itself gets a huge problem: 0.1 ppm drift of the crystal would be about a 0.1% change in the control voltage. So the stability of the voltage to frequency relation is a big problem one that is likely not solvable unless you find a crystal stable to the ppb per year range.

Linearity of the varactor can be a problem, as with a nonlinear response the amplitude of the AC testing-signal will have an influence.