Who Has The Best Home-Made 10Vdc Reference?

[CalibrationGuy]

Hello everyone! As some of you know, I have a calibration lab and I enjoy testing voltage references and have dabbled with creating my own. However, after perusing the threads here, I see that my efforts have been eclipsed by some epic efforts by the likes of Dr. Frank, Andreas, Branadic, TiN, et al. I have several Fluke 732 references. One of my older units, a 732a has an extraordinary drift history. I measured it once with a Keysight 3458a that was right back from a metrology lab calibration from Keysight (compared with their Josephson array) 20 months ago, and once again recently with the same meter back from another metrology grade calibration. The result: 0.3ppm difference! Yes, the temp/humidity was the same for both measurements, same cable connecting the units, RF field measured, ACAL ALL, 200NPLC, etc,, etc,, etc. An incredible result. Of course, if we apply all the uncertainties, a cynic can say that this is hardly definitive proof. Well, I concede that point, but now I have 2 more 732a units, a 732b, and 732c which will give me really good data points in the coming years.

So, sorry for the long winded intro, but my question is this: Who has the best reference here? I mean one that they've built. Also, you have to back up your claims by comparisons to a known voltage reference such as the 732. I have 4 3458a meters sitting on my bench right now and can see them drift over the course of 24 hours by 0.3 to 1.6ppm max difference between them.

Looking forward to your responses!

TomG.

My apologies for forgetting Kleinstein!

[Grandchuck]

Here is a shot of one of my reference units.  3 copies of Dr. Frank's design.  My 3458A readings wander about +/- 1.5 ppm.  I have checked calibration using the standard designed and built by TiN (USA Cal Club).  I don't think I can achieve anything more as a voltnut without going into more expense and time than I am willing to invest.

[CalibrationGuy]

That's a very nice looking project. If you have time, it would be nice to see the front panel with one of the outputs hooked up to your reference meter.

TomG.

[Grandchuck]

Sure TomG.  Thanks for the kind words!

I added a 7 hour graph.

[CalibrationGuy]

Thank you for that picture. I assume the panel meter is monitoring your power supply voltage. We are sending one of our Fluke voltage references to Keysight soon. It's part of our effort to get under the half part per million uncertainty certification @10Vdc. Once that happens we can characterize your home project if you are curious as to how close it is to 10.00000Vdc. You pay the shipping. The test is free, but I can't give you an official calibration report. Otherwise, my corporate clients would be really not happy.

TomG.

[DimitriP]

but I can't give you an official calibration report. Otherwise, my corporate clients would be really not happy.

No one will ever speak of it. Ever. To anyone. At all.

[CalibrationGuy]

but I can't give you an official calibration report. Otherwise, my corporate clients would be really not happy.

No one will ever speak of it. Ever. To anyone. At all.

I think you may be right. Unfortunately, some of my clients are on here, and there's nothing to stop anyone from posting images on this board. But yes, the chances are only slightly better than winning the lottery. And yes, that was pretty funny.

TomG.

[Grandchuck]

Yes, the panel meter is indicating the voltage before the 12 volt regulator.  It was once connected to one of the reference outputs but always displayed 10.000 ... how boring

[branadic]

From what I observed during my own experiments, these 5-digit panel meters create a lot of noise and interferrence, hency why I ditched them also in my low noise power supply and replaced them by analog panel meters. Just my 2 cent.

-branadic-

[Grandchuck]

Yes, the switch is to turn the panel meter off and it is almost always off.

[IanJ]

Yes, the panel meter is indicating the voltage before the 12 volt regulator.  It was once connected to one of the reference outputs but always displayed 10.000 ... how boring

I was wondering about that. How about displaying the internal temperature of the enclosure?

Ian.

[Grandchuck]

Hi Ian.  That was the setup a while back.  It worked but did not actually help my volt nutting.  My lab is almost always at a rather constant temperature (unless we have a power outage).

It turns out that occasionally one of the chopper amps latches up and this can be detected by monitoring the unregulated dc supply.  It happens rarely and only when connecting/disconnecting the output jacks.

BTW, my PDVS2mini is sill serving me very well.  I never adjusted/recalibrated it ... no need to!

[MiDi]

D105-10 (REF102) from Awesome14 (Kaysert - Tom Kayser) gets my vote


from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg600690/#msg600690






from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg606722/#msg606722


from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg602915/#msg602915

[CalibrationGuy]

That must be a revision b board. Mine had aluminum foil, so a few more ppm off.

TomG.

[Alex Nikitin]

You've made me to remeasure my home-made reference. I am not claiming it is best though it has kept its 10V output inside probably 10ppm for last 8 years or so (if measured at the same temperature as the tempco is not small, at 25C it is about -1ppm/C ). It is certainly different from the majority, I might yet to come back and do some more work on it now that I have the HIOKI DM7275 voltmeter which would make it easier to measure.

Cheers

Alex

[CalibrationGuy]

Now you're making me read yet another voltage reference thread! This forum is quite addictive, I must say.

@Alex: The performance of your reference is impressive!

TomG.

[Alex Nikitin]

@Alex: The performance of your reference is impressive!

TomG.

Thank you Tom, I suppose for a prototype assembly it is good. I've used a metal can JFET (available from stock at Mouser for about $3 each), and well aged Ultrohm WW resistors that seems to be very stable (as the 10V output depends both on the JFET reference output (around 7V) and the step up amplifier. I had some plans to improve this reference, perhaps now I will build another one with some updates, first of all the temperature control would be interesting, as the "zero" tempco point can be adjusted easily for a particular temperature.

Cheers

Alex

[CalibrationGuy]

Ok, looks like I've been inspired to throw my hat in the ring. I have an advantage in that I have a calibration lab at my disposal for testing a few designs. In particular, I have a 732a that has drifted only a few microvolts in over 20 months. Perfect for long term drift testing.

Some thoughts:

I was interested in using the ADR1001 because 10V and 5V outputs are useful to me, but they're not available readily yet.

The ADR1399 looks promising, but I was looking for 3458a level performance. Maybe 4xADR1399 is a possibility.

The LTZ1000 is another option. Reference circuit with a 7V to 10V amp/buffer circuit. The buffer circuit may be heated and/or compensated with resistors of positive and negative tempcos.

Of course the ADR1000 as an alternative to the venerable and widely tested LTZ is also an option, but I hear that it takes many months to stop drifting, and there is no long term drift data yet.

All of these options have been discussed in great detail here.

Suggestions as to which option I should try first are welcome.

PS - Low drift is priority one with noise a close second.

TomG.

[jewelie]

D105-10 (REF102) from Awesome14 (Kaysert - Tom Kayser) gets my vote

...

Christ on a bicycle!  What a rabbit hole that was!

[Overspeed]

D105-10 (REF102) from Awesome14 (Kaysert - Tom Kayser) gets my vote


from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg600690/#msg600690






from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg606722/#msg606722


from https://www.eevblog.com/forum/metrology/calibratory-d-105-dc-precision-voltage-reference-standard/msg602915/#msg602915

Hello

This '' thing'' is a nice demonstration on how to built something bad , metrology level can reached with DIY equipment but that request a bit of cost analysis and respect in design , except if you lost in the middle of no where or stuck during a weekend by bad weather ....

A descent quality PCB cost 2 Euro ...so why use a prototype board with ugly soldering


Regards
OS

[HighVoltage]

Hello

This '' thing'' is a nice demonstration on how to built something bad , metrology level can reached with DIY equipment but that request a bit of cost analysis and respect in design , except if you lost in the middle of no where or stuck during a weekend by bad weather ....

A descent quality PCB cost 2 Euro ...so why use a prototype board with ugly soldering


Regards
OS

You need to read the full thread to understand. 
It is the funniest thread on eevblog, but it is long, we had a good time.

[CalibrationGuy]

Ok, looks like I've been inspired to throw my hat in the ring. I have an advantage in that I have a calibration lab at my disposal for testing a few designs. In particular, I have a 732a that has drifted only a few microvolts in over 20 months. Perfect for long term drift testing.

Some thoughts:

I was interested in using the ADR1001 because 10V and 5V outputs are useful to me, but they're not available readily yet.

The ADR1399 looks promising, but I was looking for 3458a level performance. Maybe 4xADR1399 is a possibility.

The LTZ1000 is another option. Reference circuit with a 7V to 10V amp/buffer circuit. The buffer circuit may be heated and/or compensated with resistors of positive and negative tempcos.

Of course the ADR1000 as an alternative to the venerable and widely tested LTZ is also an option, but I hear that it takes many months to stop drifting, and there is no long term drift data yet.

All of these options have been discussed in great detail here.

Suggestions as to which option I should try first are welcome.

PS - Low drift is priority one with noise a close second.

TomG.

Let's not derail this thread. Any thoughts or comments on my questions?

[Doctorandus_P]

Marco Reps has build some amazing things. He has built:

https://github.com/marcoreps/ADRmu

And now he is working on a stepper motor driven switch box to do automated long term drift measurements between an array of those things, also featured on Hackaday:

https://hackaday.com/2024/05/27/reinventing-rotary-switches-with-stepper-motors/

[CalibrationGuy]

Marco Reps has build some amazing things. He has built:

https://github.com/marcoreps/ADRmu

And now he is working on a stepper motor driven switch box to do automated long term drift measurements between an array of those things, also featured on Hackaday:

https://hackaday.com/2024/05/27/reinventing-rotary-switches-with-stepper-motors/

Is that the "precious ppms" guy?

TomG.

[Roehrenonkel]

Is that the "precious ppms" guy?

Yes, and he has the "Safety-socks"!  :-)))

[CalibrationGuy]

He has a unique ability to be funny while discussing highly technical issues which don't really lend themselves to humor.

TomG.

[Whales]

It's not PPM material, but my AD2712 is useful 10VDC ref for checking my multimeters.  Not sure if I fit in here 

I thought my Extech was under-reading by much more than 1%.  Hmm.  I wonder if each set of range resistors can go bad independently. 

EDIT: I previously thought the Extech MN16 was horribly under-reading when testing AA batteries but perhaps I was imagining it, now it seems to be only 1.2% different from the Parameters 7080B when testing a 1.38V AA.
EDIT2: I don't know how to calibrate it, there are 6 pots inside and no documentation 

[Mickle T.]

Marco Reps has build some amazing things. He has built:

https://github.com/marcoreps/ADRmu

And now he is working on a stepper motor driven switch box to do automated long term drift measurements between an array of those things, also featured on Hackaday:

https://hackaday.com/2024/05/27/reinventing-rotary-switches-with-stepper-motors/

Is that the "precious ppms" guy?

TomG.

In my opinion, this guy deserves an honorable second place after Lymex Zhang (lymex/BG2VO) - the greatest popularizer, who has published hundreds of articles about sub-ppm/volt/ampere/ohm-nuts 

[Smokey]

If I were to assembly one of the existing reference designs, is there a consensus on which one is the "best"?

xdevs KX/FX, reps ADRMu, etc?

[CalibrationGuy]

If I were to assembly one of the existing reference designs, is there a consensus on which one is the "best"?

xdevs KX/FX, reps ADRMu, etc?

Best design depends on your level of expertise, your testing capabilities, your requirements, and yes, your budget. You can spend $200 on a single resistor. Start with requirements and budget.

TomG.

[Smokey]

   No hard requirements or budget.  I'm not trying to be difficult, it's just a thing to play with at this point.  Yes I am a professional, but I don't do anything that needs metrology level stuff.  My best meter at the moment is a Keithley 2700, so no I can't even really verify it.  Doesn't mean I won't get something nicer eventually if the opportunity arises.  Quit making me look bad in public already for not having a rack full of 3458A's!  I know I'm not officially in the club yet  .. 
And yes I know this is not an inexpensive adventure.  Which is sort of why it would be good to know if one is favored so I don't regret spending a bunch to built it.

My question was more about design maturity, reliability, and stability in this group of boards people have put together.  If it really is a design trade between the boards and there is not a clear "winner", what are the strengths and weaknesses of each?

[CalibrationGuy]

Alas, I cannot advise you as to which is the best design as I have not built any of them. xDevs, aka TiN is a friend of mine and we've tested his FX against our Fluke 732a/b/c array with great results. I also hear good things about Dr. Frank's designs. You can also prototype an ADR1399 board in bootstrapped configuration. In the ADR1399 thread, closer to the end someone posted a design that looked promising. Hope this helps.

TomG.

[aronake]

   No hard requirements or budget.  I'm not trying to be difficult, it's just a thing to play with at this point.  Yes I am a professional, but I don't do anything that needs metrology level stuff.  My best meter at the moment is a Keithley 2700, so no I can't even really verify it.  Doesn't mean I won't get something nicer eventually if the opportunity arises.  Quit making me look bad in public already for not having a rack full of 3458A's!  I know I'm not officially in the club yet  .. 
And yes I know this is not an inexpensive adventure.  Which is sort of why it would be good to know if one is favored so I don't regret spending a bunch to built it.

My question was more about design maturity, reliability, and stability in this group of boards people have put together.  If it really is a design trade between the boards and there is not a clear "winner", what are the strengths and weaknesses of each?

From above, and if around 200 USD is OK, sounds like a Dr Frank LTZ1000 should be a good start for you. It works very well, easy to solder with all through hole components and largely follow the reference design and it has a 7V to 10V stage. The LTZ1000, or if you can find an ADR1000 cost a bit of money. For the 5 "critical" resistors they are not too critical, so no need for anything too fancy or advanced to start with. The 7V to 10V resistors are very critical though, but you could upgrade these later as the 7V part has stabilized. This upgrade would be outside the 200 USD budget...

[ch_scr]

This is my cheap performance, based on Solartron 7081 circuit with tempco compensation, 2dw232 zener and "equal element divider" for 10V transfer as described here.
The "equal element" transfer network resistors are fairly benign Susumu RR1220P 25ppm/K
The logging is continuous with 34970A, it logs this among others like raw LM399.