New secondary voltage standard

[e61_phil]

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

On my last visit at the PTB I had the chance to compare my LTZ1000 against a 732A which was used there. The 732A was periodically compared to a JJA (I saw the calibration certificate) and they said that the 732A isn't moving by more than ~0.1ppm/year

[retroware]

There is a patent from someone at Advantest documenting this technique.

https://patents.google.com/patent/US6091281A/en

[guenthert]

  The patent is interesting, I would like to see an implementation though.

  In the patent it is written that "Therefore, in the present invention, because of the fixed relationship between the control voltage supplied to the TCXO and the oscillation frequency of the TCXO, the control voltage shows a high precision DC voltage."  This hinges on the relationship between the control voltage and the oscillation frequency of the TCXO being fixed, doesn't it? 

  And if the TCXO is controlled via a variable capacitance diode, then how 'fixed' is the relationship between control voltage and capacitance really?  I'd think the geometry of the diode changes with temperature and with that the capacitance.  In a GPDSO that doesn't matter, as the error is calculated and corrected periodically using the GPS time, but how is this supposed to be handled here?

[guenthert]

  Just noticed, that his has been discussed here before: https://www.eevblog.com/forum/metrology/another-method-to-get-a-precision-voltage-reference-does-it-make-sense/

[David Hess]

Temperature compensated varactor diodes can be very stable.  Do a search for Bob Pease's comments about varactor diode based operational amplifiers.

[Jay_Diddy_B]

Hi,

I have looked at the patent. The circuit described in the patent is essential a GPS disciplined (VC)TCXO. The reference voltage output is the tuning voltage of the VCTXO.

OCXOs are generally considered to be an upgrade from TCXOs.

So if were to monitor the tuning voltage in my Z3801A (for example) I should have a stable reference voltage?

If this correct, then the converse should be true, if I feed the OCXO tuning voltage with a fixed voltage I should have a very stable frequency, no need for the GPS disciplining.

But if I put the Z3801A in the holdover mode, by disconnecting the antenna the frequency will drift.

As a voltage reference it seems very complicated to me. There are lots of opportunities for drift and aging in this circuit:



Image is from the patent https://patents.google.com/patent/US6091281A/en

Does anybody know if Advantest ever used this reference in a product?

Regards,
Jay_Diddy_B

[bdunham7]

Hi,

I have looked at the patent. The circuit described in the patent is essential a GPS disciplined (VC)TCXO. The reference voltage output is the tuning voltage of the VCTXO.

OCXOs are generally considered to be an upgrade from TCXOs.

So if were to monitor the tuning voltage in my Z3801A (for example) I should have a stable reference voltage?

If this correct, then the converse should be true, if I feed the OCXO tuning voltage with a fixed voltage I should have a very stable frequency, no need for the GPS disciplining.

But if I put the Z3801A in the holdover mode, by disconnecting the antenna the frequency will drift.

As a voltage reference it seems very complicated to me. There are lots of opportunities for drift and aging in this circuit:

(Attachment Link)

Image is from the patent https://patents.google.com/patent/US6091281A/en

Does anybody know if Advantest ever used this reference in a product?

Regards,
Jay_Diddy_B

I had the same feeling, although I don't want to dismiss out-of-hand an idea that I may not fully understand.  Just because something is patented doesn't mean it is a good idea.  Perhaps someones pointy-haired boss wanted to market a frequency-related voltage standard thinking people would associate it with the Josephson effect.  The problem is that there isn't any fundamental principle of physics being exploited here and I don't see anything that would give this system any more stability than current methods.  I guess we'll see.

[Kleinstein]

I see no real advantage use a crystal and GPS clock as a reference for the varactor capacity. One would probably better just use a brigde with varactor and capacitor (possibly twice to make it a full bridge). The idea has the positive effect that the reference part itself (cap + varactor) would not produce significant heat and could thus be easily temperature stabilized. However there is parasitic capacitance and this tends to be effected by humidity. So if at all it would need to be in a small sealed container.  My first estimates do no look that optimistic, as there is quite some TC to the varactors (some 2 mV/K).

[guenthert]

[..]
So if were to monitor the tuning voltage in my Z3801A (for example) I should have a stable reference voltage?
[..]

  I believe in principle that to be true, however (as discussed in the older thread) the relationship between the tuning voltage and the frequency is not in favor for a voltage reference: the tuning voltage swings a few volts for a few ppm of frequency.

[guenthert]

Temperature compensated varactor diodes can be very stable.  Do a search for Bob Pease's comments about varactor diode based operational amplifiers.

Those are interesting little buggers with input bias current of less than 10pA and CM voltage of up to 200V.

In https://books.google.com/books?id=Kcv8BAAAQBAJ&pg=PA72&lpg=PA67&dq=Philbrick+P-2&source=bl&ots=spTAKv58_J&sig=UkIDXapzfihaGuwCDrMTuKLhUhk&hl=en&sa=X&ved=0ahUKEwi1-8LQ-YXOAhXCFR4KHXY1CEwQ6AEIKTAC#v=onepage&q=Philbrick%20P-2&f=false
however I found an estimation of the capacity to voltage dependency of varactor diodes (delta of 0.03pF for a delta of 1mV) and some anecdotes of stability of the P2A over time which doesn't look too promising (better than 1uVpp over an hour or two).

[Magnificent Bastard]

Mine was a special "cherry picked" unit that was sold to the US Air Force (for one of their PMEL labs).  It has lower than normal noise and lower than normal annual drift compared to the average 732B.  My unit gets calibrated by Fluke-- and only a JJA has enough stability to get the uncertainty down to these low levels.  It (unfortunately) was made with one of the early LTFLU's (the non-A version), and so is more sensitive to barometric pressure changes.  This is calculated into the calibration and compensated for at my local altitude.  I was only able to get the one unit, and it is unlikely that I will ever find another one with these ultra-low drift specs.  My interest is in building a number of LTZ based references to build out a sustainable array of references that can be statistically monitored.  It's been a long road, but I'm getting closer.

[dietert1]

Mine was a special "cherry picked" unit that was sold to the US Air Force (for one of their PMEL labs).  It has lower than normal noise and lower than normal annual drift compared to the average 732B.  My unit gets calibrated by Fluke-- and only a JJA has enough stability to get the uncertainty down to these low levels.  It (unfortunately) was made with one of the early LTFLU's (the non-A version), and so is more sensitive to barometric pressure changes.  This is calculated into the calibration and compensated for at my local altitude.  I was only able to get the one unit, and it is unlikely that I will ever find another one with these ultra-low drift specs.  My interest is in building a number of LTZ based references to build out a sustainable array of references that can be statistically monitored.  It's been a long road, but I'm getting closer.

The references i am observing are LTFLU 'A's and their voltage difference exhibits a barometric pressure correlation of about -0.003 ppm/hPa. Do you have some numbers for LTFLU non-A and LTFLU A?
So barometric pressure variations may contribute +/- 0.03 ppm typical - a significant fraction of the observed residual (stdev = 0.029 ppm). I am just installing some BMP388 sensors for continuous pressure monitoring, trying to convince myself that a standard deviation of the daily average below 0.01 ppm is realistic. As far as i understand that would be required to predict yearly drift to within 0.1 ppm.

Regards, Dieter

[SilverSolder]

[...]
So barometric pressure variations may contribute +/- 0.03 ppm typical
[...]

That's impressive resolution - is that something that can be reliably achieved outside an underground shielded bunker? 

[dietert1]

Until now that isn't precision but resolution. It's a small setup in our lab with the two references and a HP 3456A to measure the voltage difference, with low thermal EMF wiring. Then +/- 0.03 ppm of 10 V are +/- 300 nV. The difference voltage is in fact about 190 uV and in it's lowest range the HP 3456A has a resolution of 100 nV  and that is also its observed offset voltage stability. If i just leave it running, apparently it is that stable. That HP 3456A came to us through ebay about 15 years ago. The in-guard power supply was recapped and got some mods.
Later i will play with the wiring and observe changes caused by EMI and tune things to make those as small as possible, but that's another story: to make the voltage reference something useful. Build the references into a case with binding posts, with battery operation etc. Currently it is just a little research by a physicist.

Regards, Dieter

[Johnny B Good]

I see no real advantage use a crystal and GPS clock as a reference for the varactor capacity. One would probably better just use a brigde with varactor and capacitor (possibly twice to make it a full bridge). The idea has the positive effect that the reference part itself (cap + varactor) would not produce significant heat and could thus be easily temperature stabilized. However there is parasitic capacitance and this tends to be effected by humidity. So if at all it would need to be in a small sealed container.  My first estimates do no look that optimistic, as there is quite some TC to the varactors (some 2 mV/K).

 Would it not be better to derive the voltage reference from the OCXO's Vref pin rather than the variable VFC pin voltage? 

 I wanted to improve on the 1mV resolution of my cheap Mestek "9999 counts" DMM when monitoring the VFC (currently 2,281 mV +/-500uV) in my mark 2 homebrewed basic GPSDO (hardware PLL with a 5000s TC) and hit upon the bright idea of using the OCXO's 5.127v Vref to drive a 1mA current through a 220 ohm 'ballast resistor' and a set of 5 hand matched 1K resistors to realise a series of DC offset voltages in 1 volt increments from zero to 4 volt in order to subtract in this case, the excess 2 volts off the meter reading allowing me to use the 999.9mV range, to show VFC changes in 100uV increments.

 I used a 50k trimpot with a 22k resistor in series wired across the 5 k resistor string to trim the volt drop across each 1k resistor as close as was possible with this cheap meter to exactly 1.000 volt each. Exactitude would be nice but in this case, temperature stability was the most vital parameter for my purpose.

 I'd already experimented with the mark 1 GPSDO using a 3.16v CR2032 cell (a surprisingly stable voltage reference for my much more modest voltage stability requirement) before using a TL431 from my wallwart salvaged parts collection to create a 3.000 volt reference (the OCXO in this case needed circa 3.311v to tune it dead on 13MHz - don't ask!) but would drift by just over a mV for modest 2 to 3 deg changes in room temperature.

 Short term, this was not really much of a problem in observing the minute to minute changes in the ionosphere's electron density on the phase of the 100KHz pulses out of my fake NEO M8N gps receiver modules that I was phase locking the OCXO's to but I wanted a better voltage reference for my mark 2 GPSDO attempt and making use of a temperature controlled voltage reference built into most OCXOs (including the seven 12v 10MHz CQE units I'd managed to get hold of dirt cheap), just seemed the most obvious solution to saving a few hundred quid on an overpriced 4 1/2 digit bench voltmeter just to be able observe the VFC change in 100uV increments.

 Previously only being able to observe in 1mV increments meant I could see the effect on the DSO display before I could see any change in VFC voltage whereas now, it's the other way round. 

 Why anyone would think the VFC voltage would be a better choice than the thermally stabilised Vref voltage of your typical OCXO, I just can't imagine, especially when you consider that there's no need to complicate the arrangement with the addition of a GPS timing receiver module - just use the OCXO, almost any OCXO (even a broken one provided the Vref was still intact), on its own in a suitable enclosure.

 Anyhow, that's my two dollar's worth on a subject I'm only just starting to get to grips with. 

JBG

[Kleinstein]

The Vref. pin on an VCXO is usually just an internal classical voltage reference (e.g. bandgap or zener). The main purpose is likely to have a stable for the control and it may be needed internally for the temperature regulation anyway.
I don't think it would make much sense to used the VCF pin as a reference, but the general idea behind it may work: The VCF pin usually control a varactor diode, and the voltage versus capacitance curve of a varactor would be used as a reference. Just don't measure / regulate the capacitance with an oscillator, but directly with a bridge. The advantage over the classical zener would be that the varactor bridge has little intrinsic noise and also low heat production. The problem are parasitic capacitors and changes of it. So it would kind of requite a compact, possibly monolytic design to really become very stable.

[Johnny B Good]

@Kleinstein

 Thank you for the clarification on the thinking that had prompted the suggestion of using the VFC as a voltage reference. Obviously you can't use the VFC for this simply on account of the ageing effect of the crystal itself.

 My thinking, as far as solving my own little problem of using a reasonably temperature stable DC offset to utilise a cheap "9999 Counts" DMM to monitor the VFC was that the Vref, even if derived from a cheap commodity  1117-5 ldo used to power the XO and temperature controller would be a far better option than the TL431 I'd tried in my mark 1 gpsdo simply because it was being temperature stabilised and providing a constant (in this case 14mA) current to the XO and thermal regulator circuitry plus one additional fixed 1mA external load (my series string of five 1K matched resistors).

 I actually ran a voltage comparison test between two of my CQE 12v 10MHz OCXOs just to verify how well these temperature stabilised sources were likely to be. I did see some slight variation of around 100uV between them (5.128 versus 5.172 volts - a 44mV difference) over a two or three day test run. Since this was a significant improvement over the TL431 I'd used in the mark 1, I picked the one out of the two I'd been using for this voltage stability test which seemed to be making the least contribution to the 100uV variation for use in my mark 2 gpsdo.

 Although it's unlikely I'll be needing anything other than a 2v offset using my chosen OCXO, I designed it as a multi tapped potentiometer of five 1K resistors partly to cater for different OCXO voltage offsets but mainly because I'd still be able to use my cheap "9999 counts" Mestek DMM to trim the volt drops across each individual 1K resistor to within 100uV.

 I know this level of measurement accuracy falls far short of that normally dealt with by members in this topic thread but it meets my immediate requirements (for now ) and might help other cash strapped hobbyists work out a cheaper alternative method to get that extra magnitude of accuracy they crave before being forced to dig ever deeper into their pockets for better test gear.

JBG

[branadic]

Thinking about it, the only approach that I know of and that could fit the description by MiDi in the first post is based on NV centers.

Robust and accurate electric field sensing with solid state spin ensembles

A solid-state quantum voltage reference

I know that there is a group at PTB working on NV centres as many people do.

-branadic-