Wanted: Someone that can measure 10 V DC within 30ppm

[The Soulman]

My two 4,5 digit meters disagree by about two lsd, and for normal people that isn't a problem but for me..   
So I'm planning to build a 2,5 volt reference (or perhaps two, or ...) and after some testing/aging send them to someone
who is able to measure them within 100 ppm uncertainty of a traceable standard.
Preferable in the Netherlands but Germany is OK also.
Is this even feasible?

Reactions here or PM.

[The Soulman]

Do you have a local KS/KEI/Tek sales? Maybe you can pay them a visit and "demo" their units...
Or just go to a university to get someone to do this for you.
You don't need a traceable equipment. It's highly likely even a calibration overdue 6.5 digit meter is way more accurate than your 4.5 digit.

Don't take this offensively and I appreciate your input but:

1) I don't like salespeople.
2) I haven't stepped inside an educational institute other than elementary schools for a long time, and not sure who or how to contact anyone.

Some 6.5 digit should do..

[Vgkid]

Try sending a pm to Pa4tim, he is in the Netherlands. There are most likely others as well.

[SvanGool]

Bummer, I can't make the 100 ppm on my 34401A, but if 123 ppm is enough (https://www.eevblog.com/forum/metrology/lt1021cmh-based-mini-reference/msg1172332/#msg1172332)? Or if you can bring them within 90-14 days=76 days from now, my 10.00000 V is even 15 ppm better = 108 ppm, very close !  .
Unless Alex invalidates the intermediate results.

[HighVoltage]

If you can not find anyone in NL, you can send it to me and I will take measurements on my 3458A, 34470A, DMM7510

[try]

Soulman,

how fast do you need it?

I can be reached near Venlo a couple of times over the year.
Could bring in a 34401A.

Regards
try

[blackdog]

Hi The Soulman,

I can help you if you want.
100PPM is not a problem for me.
DMM4050 DMM's en a HP3458A on the bench :-)

Location Amsterdam
Send me a e-mail.

Kind regards,
Blackdog

[PA0PBZ]

Maybe if the Soulman can reveal his location a bit it would be easier. I guess my 34461A is up to the task too.

[SvanGool]

Bummer, I can't make the 100 ppm on my 34401A, but if 123 ppm is enough (https://www.eevblog.com/forum/metrology/lt1021cmh-based-mini-reference/msg1172332/#msg1172332)? Or if you can bring them within 90-14 days=76 days from now, my 10.00000 V is even 15 ppm better = 108 ppm, very close !
Unless Alex invalidates the intermediate results.

Just for the record: I mixed my Vs and Rs, for the 10V range my 34401A's accuracy is now 4+3+25=32 ppm for the 90 days period and 4+3+40=47 ppm for the 1 year period, well within the requested 100 ppm   

[TiN]

Gotta give credit to TS, great way to lure those voltnuts out. Just ask to measure something with low accuracy, and suddenly whole crowd is out in the daylight. 

[Theboel]

TiN, try to lure out the Voltnut in my country I think You need a full fleet of Fish ship to get a few 

[The Soulman]

Wow thanks for the response guys!
If someone is near me (west Drenthe) maybe we could meet so I don't have to build a reference, but on the other hand it is not bad to have one , or two..

[PA0PBZ]

but on the other hand it is not bad to have one , or two..

And that's how it starts... 

[SvanGool]

If someone is near me (west Drenthe) ..

And now he is also luring the German Voltnuts out, because that is close to their border   

BTW, I am not too close, mid-south.

[e61_phil]

If someone is near me (west Drenthe) ..

And now he is also luring the German Voltnuts out, because that is close to their border   

BTW, I am not too close, mid-south.

I looked into google maps and you'r absolutely right. I'm near Bremen. It's about 190km from Drenthe. I'm also able to measure 5V much better than a 100ppm.

[The Soulman]

I guess now I can build a reference and thermal cycle and age that a bit.
That could take at least one month or probably two, will contact people that responded here when I'have something put together.
Tomorrow I'm receiving a bunch of used precision wire wound resistors so I can include some of those as well to my travelling standard.

Any ideas for a ref IC that can be used? Maybe LM399 (8ppm/1000hr) or perhaps something cheaper and more simple so that multiples can be fitted in a small box?
Also 2,5 volts would be preferred but long term stability is a must.

Thanks again guys! 

[TiN]

Go for LM399.  But keep in mind that in 2-3 years you may end up having 2 x 8.5-digit DMMs, 4 x 7.5-digits ones, few standards, few calibrators, and bags of 50+$ components in your lab. 

[try]

Soulman,

Any ideas for a ref IC that can be used? Maybe LM399 (8ppm/1000hr) or perhaps something cheaper and more simple so that multiples can be fitted in a small box?
Also 2,5 volts would be preferred but long term stability is a must.

Thanks again guys! 

Taking into account your current uncertainty of 100ppm building some transfer standard is a no-brainer.
Buy the KKMoon reference, it already has a nominal 2,5V setting.
Please doublecheck before how far if at all your 2.5V measurement range exceeds that 2,5V barrier.
It's not useful having a reference giving 2,50080V which forces your DMM to switch into the next bigger range. 

If you want to build something yourself, use the MAX6325, p.e.
There is no need to use expensive wire wound resistors at this stage of your precision discovery.

Building some LM399 based reference is only important when having a longer term time horizon in mind as you can't buy aging time.

There is no need for building some device with multiple references at your current stage.
That is as useful as gold plating the cooler in your car.

Always bear in mind the precision you are targeting at and the cost involved. Both factors are highly likely to be left out when people are discussing their unrealized plans in this forum.

Have fun!

[The Soulman]

My main meter is 30.000 count, hence the 2,5 Volt.
Long term stability is a must to be able to recheck stuff and when the two meters and the reference drift apart (unlikely but you never know..)
it is useful to have multiple reference standards to determine what is drifting where..

Regarding cost, it should be possible to keep this project under 100 euro's (aprox. 60 for the device and 40 to ship it a couple of times?)
In my opinion that is still a fair price to pay to strengthen the confidence in the accuracy of my daily measurements for years to come.

[try]

Soulman,

with all respects to your needs for stability, this is where ignoring the quantitative aspects is not a smart decision as I critized above.

Your DMM has 30.000 counts.

One count represents a change in value of 33ppm.
Your device probably allows for 1-2 counts of variation.

Obviously your still to build reference has to move a lot before that change is showing up on or DMM which only has 4 2/3 digits.
How much time do you think will that take?

You might want to look up some drift numbers for a buried zener.
Checkout the last column on page 2:

http://www.ti.com/analog/docs/litabsmultiplefilelist.tsp?literatureNumber=sbva001&docCategoryId=1&familyId=720

I guess your device will not cost more than €15,- even with a metal housing and a LM399 if you insist on it (power supply excluded).

[Andreas]

Hello,

the sad truth is: you will need a minimum of 3 (better 4) references to judge about the long term stability.
If you have only one: how can you tell wether the reference driftet on its way back from calibration if you do not have the possibility to check against your home references.
And how can you tell which one of the references drifts most if you have only 2?

The mentioned MAX63xx/MAX62xx are not bad. But all my samples of MAX6250A/MAX6350 (in PDIP8 package) are drifting by about +10 ppm/year.

Many LM399 are much lower (in the range of 1-2 ppm/year) but of cause you have to select out the "stinkers" to get a good result.
To get around the 2.5V a good (long term stable) possibility is the use of a capacitive 3:1 divider with help of a LTC1043 + a unity gain buffer.

With a 30000 count DMM and a stable reference one should also be able to determine the tempco of the instrument.
My 80000 count DT80000 multimeter has a T.C. of about 84 ppm/K.

with best regards

Andreas

[try]

Andreas,

your proposed solution is overkill - shooting with canons at sparrows as the German saying goes.

Regards
try

[Andreas]

perhaps in the "beginners" section.
But never in the metrology / volt-nut section.

There is no such thing having too much DMMs and voltage references. see:

https://www.eevblog.com/forum/chat/test-equipment-anonymous-(tea)-group-therapy-thread/msg1114480/#msg1114480

With best regards

Andreas

[The Soulman]

Here is the plan: 5V (usb) input-> isolated dc/dc converter -> linear voltage regulator with 1mA current limit -> LT1021DCN8-7 -> over voltage/reverse voltage protection diodes-> glass fuse -> binding post output.
Parallel to the protection diodes a 10uf tantalum in series with a 4,7 Ohm resistor and obviously some capacitance around the linear regulator.
The above copied 3 or 5 times in one enclosure with one shared input, with the isolated converters I can connect the reference outputs in series for
higher voltage output.

Sounds like a reasonable plan?

Also and perhaps in the same enclosure I would like to build some sort of a resistance standard, 0R1,0R9,9R,90R,900R, in series with dual binding post in between so the individual resistors can be measured with 4 way connections but also multiple in series, could also act as a current shunt.
and 10K,10k (matched to 1ppm) and 100K,100K (matched to 1ppm) also with dual binding post for 4 way connection, could also act as single sides of a wheatstone bridge.
These resistors I already own, but also want to include higher values (1M and 10M) but not sure what resistors and connections to use?

While we are at it also add some capacitors to be calibrated, 100nF, 1uF, 10,uf ? Wima MKP10 good enough?

Suggestions much appreciated, parts will be ordered this week.

So the question changes a bit, Who can measure 6,95V, a large range of resistor values and some capacitors, to within 100ppm?

Having this device measured by several people probably isn't a bad idea and that way others can get some use-full data from it as well.

[TiN]

Capacitors within 100ppm going to be real challenge here .

I'd say loose USB input (it's very noisy), loose dc/dc converter and power the linear preregulator with some decent quality linear brick mains power supply.
Glass fuse will not do anything there too, if you reach currents that will break even smallest fuse (plugging mains into output? ), then whatever is behind fuse would be already charcoal dead.

[The Soulman]

Capacitors within 100ppm going to be real challenge here .

I'd say loose USB input (it's very noisy), loose dc/dc converter and power the linear preregulator with some decent quality linear brick mains power supply.
Glass fuse will not do anything there too, if you reach currents that will break even smallest fuse (plugging mains into output? ), then whatever is behind fuse would be already charcoal dead.

Thanks for your input 

Forget about the capacitors, bad idea and not that useful.

Yes USB wall-wart supplies are noisy and so are dc/dc converters but I do like the idea of having all of the references powered thru one single power input and having something standard like USB means I don't have to ship a dedicated PSU with it.
Also having isolated dc/dc converters means I can connect outputs in series.
This power supply section can be housed in a separate metal compartment inside the metal reference box.
Personally I'm not to worried about induced HF noise (not really HF but I'm a audio guy so anything above 20KHz is HF to me..  ).
Alternative could be to use separate mains transformer for each reference but then leakage current may become troublesome when references are wired in series and 50hz induced mains frequency noise is harder to get rid off (its much closer to DC obviously).
Or, 9 volt battery's, huhm, have to work out the math for that. 

Yes the fuse in combination with the diodes won't help when the output is connect to mains but it does help when it get accidentally connected to
a bench supply or some other silly event, just few cents in costs and from my calculations no measurable negative effects, so why not.

[Vgkid]

Capacitance to 100 ppm is hard to do. The most accurate rlc meters are only 0.02% accurate, with a bridge going from 0.01(1620), 0.001(GR1621 / AH2700) to 5ppm(AH2500/2550) Which I would guess as being really pricy to calibrate.
 I notice one sold on e-pay lately. Tin did you buy it? 

[chris_11]

Supplying a reference with a DC/DC and lots of filtering can be done, if the ADC the Ref is for is in the same box. But since you plan to check external meters with it, the common mode noise of the DC/DC and the USB supply will kill any precision and stability. Go for battery power and conventional mains transformer PSU.
Characterising a low common mode noise DC/DC is very demanding for that application.
There is a reason that they went for this classic approach in the 3458A

[ap]

Capacitance to 100 ppm is hard to do. The most accurate rlc meters are only 0.02% accurate, with a bridge going from 0.01(1620), 0.001(GR1621 / AH2700) to 5ppm(AH2500/2550) Which I would guess as being really pricy to calibrate.
 I notice one sold on e-pay lately. Tin did you buy it? 

No, not Tin, was me. May do a teardown at some time later, currently still sealed with warranty tags (for what its worth). Testing it at the moment, looks good so far. Plan to use it in conjunction with the GR 1620 (and 1689) bridges for increased accurracy and faster calibration of capacitance standards. May take a little, need to get down a bit with internal lab standards uncertainty.

[TiN]

No, not Tin, was me. May do a teardown at some time later, currently still sealed with warranty tags (for what its worth). Testing it at the moment, looks good so far. Plan to use it in conjunction with the GR 1620 (and 1689) bridges for increased accurracy and faster calibration of capacitance standards. May take a little, need to get down a bit with internal lab standards uncertainty.

Whoa, cool. It was bit sad to see that unit go, not to see it again. Glad it's still in community. I can't have it all anyway  .
I'd love for the separate thread about it, once you ready. 

[Vgkid]

Capacitance to 100 ppm is hard to do. The most accurate rlc meters are only 0.02% accurate, with a bridge going from 0.01(1620), 0.001(GR1621 / AH2700) to 5ppm(AH2500/2550) Which I would guess as being really pricy to calibrate.
 I notice one sold on e-pay lately. Tin did you buy it? 

No, not Tin, was me. May do a teardown at some time later, currently still sealed with warranty tags (for what its worth). Testing it at the moment, looks good so far. Plan to use it in conjunction with the GR 1620 (and 1689) bridges for increased accurracy and faster calibration of capacitance standards. May take a little, need to get down a bit with internal lab standards uncertainty.

Very cool, I look forward to a teardown. The theory of operations is rather interesting. The msbs are achieved by a normal bridge balance, with the remainder calculated with an adc (calculating the difference). With the internal references beig ovenized. You got it for a good price, one sold a few years ago for about 2-3x that. I'm not sure what they go for new

[The Soulman]

Reference chips are on their way, ordered 10 of them just to be sure, buried zener in a hermetically sealed package so 2,5 volts is out the window... 

I'm not a voltnut, not a voltnut. 

[Andreas]

Hello,

believe me it is absolute normal to have at least 4 references!
Its like having clocks.
How many clocks do you have in your household?

with best regards

Andreas

[The Soulman]

How many clocks do you have in your household?

Two that I use, one on my phone and one on my computer they're pretty much in sync  , we also have a bunch of analog clocks around but I don't look at those.. 

[nctnico]

Reading this thread I'm wondering whether you can expect a 4.5 digit meter not to wander more than 100ppm over time. The references in those won't be as stable as in (for example) a 6.5 digit meter.

[The Soulman]

Reading this thread I'm wondering whether you can expect a 4.5 digit meter not to wander more than 100ppm over time. The references in those won't be as stable as in (for example) a 6.5 digit meter.

Correct, however these two meters I have are pretty old(metrawatt ma5d and keithley 178),
and therefore I assume that they don't drift much more over time, the fact that they are consistently  no more than a couple of counts
apart from each other supports that idea.
And if not the internal references will be replaced. 

[The Soulman]

The references have arrived this week, 10pcs LT1021-DMH10.
https://www.arrow.com/en/products/lt1021dmh-10/linear-technology
These are the worst spec D grade with poorer tempco (max 20ppm/C) and initial accuracy (10V +/-50mV).
Long term stability specification is the same for all varieties, so at 3 euro each these are pretty cost effective in my opinion.

I did just trow something together with a random picked one and the MA5D is steadily reading 10,001V for a couple of hours.
Pretty promising so far.
Will hook up a second reference and measure the voltage between the two with 10uV (1ppm) resolution and see how stable that is. 

[The Soulman]

The first one reading 10,001V must have been pure luck, the second one reads 9,982V.
Now measuring the voltage differential between the two, after 15 minutes its close to steady 18,58mV at 15Volt input
and 21 degrees Celsius.
Will let it run for 24 hour and see how things change. 

[The Soulman]

The first one reading 10,001V must have been pure luck, the second one reads 9,982V.
Now measuring the voltage differential between the two, after 15 minutes its close to steady 18,58mV at 15Volt input
and 21 degrees Celsius.
Will let it run for 24 hour and see how things change. 

It's been a bit more than 24 hours, three weeks actually, running the references 24/7 and now back at 21 degrees Celsius (we have had some nice weather) all readings now are exactly the same as before.
I've learned from individually warming these references up (between my fingers) that these two references have approximately the same temp-co but opposite from each other.
At highest ambient temperature 27 dgr. C. the change between these two was 20ppm.
So reference nr1 (10,001V) has a temp-co of approx. 1,67 ppm/dgr. C.
And nr2 (9,982V) has a temp-co of approx. -1,67 ppm/dgr. C.
Assuming they both contribute to the 20 ppm over 6 dgr. C. by the same amount.

Either way this data points out that this pair would be sufficient to use as a travelling reference for my purpose.
Next up building a enclosure..

[macboy]

If someone is near me (west Drenthe) ..

And now he is also luring the German Voltnuts out, because that is close to their border   

BTW, I am not too close, mid-south.

Not close! 
You live in a tiny country, where nothing is much more than 200 km from anything else, probably much closer.
I once drove from my hometown to the Altantic ocean coast, never leaving my own country. It took five days to travel those >6000 km... and that was one-way. Another five days to drive back home!

[The Soulman]

If someone is near me (west Drenthe) ..

And now he is also luring the German Voltnuts out, because that is close to their border   

BTW, I am not too close, mid-south.

Not close! 
You live in a tiny country, where nothing is much more than 200 km from anything else, probably much closer.
I once drove from my hometown to the Altantic ocean coast, never leaving my own country. It took five days to travel those >6000 km... and that was one-way. Another five days to drive back home!

Well when you drive a car perhaps, not when you're riding a bicycle with wooden shoes, 200km here also takes 5 days. 

[Conrad Hoffman]

If you can get a good measurement on it, some NP0 caps are very stable. Multi-layers are even good and will get you a higher value, say up to 0.01 uF or so. The most accurate mechanical traditional bridge, save for the ratio transformer types, was the later production GR1608 at 0.05%. You need to find somebody with a GR1615 (or the big 1621) and a calibration standard to confirm it.

[The Soulman]

Recently I was able to buy a General Radio 722 precision condenser.
But I decided not to  , as I have no real purpose for it as of yet, nor any capacitance standard.
It would looked nice together with my GR 602n decade box tho.

[The Soulman]

.............And two years later..........

My references are long finished and the voltage references (lt1021-dmh10) have many hours on them.

Out of the ten I've purchased four are still in the original packaging, one decided it was happier being a 12k Ohm resistor,
five undergone further tempco testing (passive dummy load in a cardboard box method) out of which three where chosen:

Two are in a portable self powered (dual 9v battery) enclosure. and may have around 1000 hours on it.

One is in a bench-top enclosure with a passive switchable voltage divider (10x ESI 1K Ohm)
with trimmer networks to match each resistor to 1ppm.
I've haven't rechecked the resistors lately but any check from the outside makes me assume they are still spot on.
It has been powered on 24-7 for close to two years, (15000 hours?).


Oddly the dual "travel reference" never reads more than 2 ppm apart from each other/it self, amazing after all that time being on and mostly off.

The 24/7 "mother reference" remains within a range of 7 ppm of the travel reference most of that as far as I can tell is due to
tempco.

The Metrawatt ma5d only reads one lsd low when cold, after a 30 min warm-up it is spot on.
So it did not drift visually in the last two years.

Some poor quality pics are attached.

[The Soulman]

In the resistance department I have build a box with five terminals and four ESI resistors:

black-100R-green-998R-bleu-9K98-yellow-99K8-red (absolute values not sure as of yet..)

The terminals are gold plated hirschmann pki10Au and can be used with kelvin connections, two spades, two bananas.

In the picture from the previous post a 10mA current source is connected to the 100R and the voltage drop across it is measured
with the metrahit 2+. 

And another (this time plastic) box with hirschmann safety sockets and three values:

black-1M-green-9M-bleu-90M-yellow

The 1M is a single 0,1% tolerance generic (cheap..) metal film resistor chosen from a bunch of ten for having a value closest to the average of the ten.
The 9M are the remaining 9 in series.
The 90M are made of 9 high voltage (500V) 1% resistors and one or two additional resistors to bring them as closely to 9x the sum of the 1M and 9M as feasible, so also within 0,1% of the supposedly absolute.

Should also come in handy measuring (very low energy!!) high voltage.

[Edwin G. Pettis]

Another high accuracy capacitance bridge is the ESI 707A/B, ±.02% or better depending on range.  Not that bad for calibration either.

[The Soulman]

Another high accuracy capacitance bridge is the ESI 707A/B, ±.02% or better depending on range.  Not that bad for calibration either.

Thanks Edwin, but at the rate I'm going it will take another two years before I've build a couple capacitance standards and probably
also diy a bridge, these old bridges are fantastic but out-dated and to big for my "lab".

[1audio]

I have 3 ESI videobridges that claim .02% midband. The references are resistors so easier to make and make stable (for ESI). I have no clear way to verify its accuracy. Ease of use is definately good. Old CRT display not great today and the fans are noisy. But a good value on eBay.