T.C. measurements on precision resistors

[ltz2000]

FYI I'm getting a pretty awesome 10K metrology grade transfer standard soon from Wekomm
2ppm tolerance
Long term stability better then 1ppm / year
Temperature stability better than 0.3 ppm / °C
And they are conservative specs, and they are working on producing 0.001ppm tolerance parts.

Teardown anytime soon...?


Now available from Datatec for 3750,00 EUR each

http://www.datatec.de/Labor-Messgeraete/Adapter-Kabel-Stecker/Wekomm-Widerstandsnormale/index.htm

[Edwin G. Pettis]

@Itz2000,

"Teardown anytime soon", you are aware that opening the standard will not only invalidate the warranty but it will also invalidate the calibration certificate as well, not to mention any accidental damage possibly caused while opening it.  Otherwise, congratulations on the purchase of a primary standard, now what are you going to do with it besides a teardown?

Frankly, I would have purchased an SR-104 standard used, for around that price you would have gotten a resistance standard with very long term proven performance which the Wekomm has yet to prove.  While its stated specifications are very similar to the SR-104, it still must prove it to be considered a standard of the same calibre as the SR-104.

As to the claim of 0.001 PPM, pure nonsense, they can't measure to that accuracy, no national primary lab in the world can measure to that accuracy, the uncertainty will likely be reduced to 0.1PPM in the proposed 2018 standards but that is still a factor of 100 over 0.001 PPM.  Any one who makes such a silly claim is suspect and is not worthy of consideration.  Any serious metrologist would laugh at such a claim.

[ltz2000]

"Teardown anytime soon", you are aware that opening the standard will not only invalidate the warranty but it will also invalidate the calibration certificate as well, not to mention any accidental damage possibly caused while opening it.

The same happens to a $10k spectrum analyzer or whatever the victim of the teardown is.

The manufacturers send their products to the blogger to get publicity. Which can be, and usually is, worth much much more than the cost of the instrument they give for free. Advertising in the traditional way is very expensive.

But there are always risks. In this case the manufacturer probably underestimated the possibility that someone points out that their product is basically a $50 component in a box with a $4000 price tag.

[Andreas]

But there are always risks. In this case the manufacturer probably underestimated the possibility that someone points out that their product is basically a $50 component in a box with a $4000 price tag.

Hello,

the $4000 is not for the component inside the box. It is only for the box itself.
Of course including the sticker on the box (and the service from calibration).

I am shure that Dave is already doing a "extended shelf life test"   
until he gets a instrument which is better than his 34470A so that he can verify the result.

If I had such a stable device it would have to work hard. 
(against the 12K or 12K5 resistors).

With best regards

Andreas

[Edwin G. Pettis]

From what I've seen and read on this blog site, I very likely have the most accurate resistance measurement system of anybody here.  DVMs are NOT considered standards under any definition, just check with your National Primary Calibration Laboratory and they will tell you the same thing.  Even the 3458A is not considered a standard under any measurement conditions, they are not used as standards by any metrology lab I know of.

At 10K, my 242D system is accurate to 0.2 PPM as calibrated against an SR104, I also have the same 0.2 PPM uncertainty as everybody else.  That gives me a 5:1 accuracy ratio when comparing a resistor (standard or otherwise) over the range of 120R0 to 1Meg, accuracy begins to drop off above and below that range.  Therefore I am barely accurate enough to test the Wekomm standard under the required specifications.  While I have a resolution of  up to 0.0001PPM, it has no bearing on the accuracy.  A National Lab must be able to to maintain a 10:1 accuracy ratio.  Under these requirements I do not see anyone on this blog (unless unknown to me) that has the required measurement system to verify the Wikomm 10K standard or even make an accurate measurement of it.

I, for one, wouldn't mind seeing the inside of the Wikomm standard out of curiosity. If someone wants to open one up for all of to view, I certainly would not object.  I could say the same thing for an SR-104, however I am somewhat less curious about it as I know a fair amount about what is in there, just some minor details lacking in my knowledge of its innards.  I certainly wouldn't want to pop one of those open if it was still in working order, it would definitely destroy its value as a standard.

[bingo600]

At 10K, my 242D system is accurate to 0.2 PPM as calibrated against an SR104, I also have the same 0.2 PPM uncertainty as everybody else. 

Edwin what's a 242D system (link please)

I think i read that Quarks have a SR104 (lucky him) , i could use it for cal'ing my 3458

/Bingo

[Vgkid]

Here is the manual, have fun...
http://www.ietlabs.com/pdf/Manuals/esi242D_im.pdf

[Edwin G. Pettis]

@bingo600,

The ESI 242D system is the best measurement system short of a primary lab, most secondary labs have them and any serious resistor manufacturer has one or an equivalent of it.  While the general specifications give up to 1 PPM measurement accuracy (at best), the instrument is capable of being calibrated to 0.2 PPM accuracy over the main ranges and only begins to lose accuracy above and below the range I specified earlier.  Since it is time consuming to calibrate to that level of accuracy, many metrology labs will settle for the 1 PPM level which is quite good to begin with.  A 242D new goes for somewhere around $35,000-$40,000 USD these days.  It also requires an SR104 and a large pile of transfer standards to achieve maximum accuracy.  They are still in high demand at secondary and lower level metrology labs around the world.  Even with an older 242D and 'used' standards, you are still looking at a good $20,000 USD plus traceable calibration requirements.

Below about 10K, the DCC bridge can achieve a higher accuracy but even used ones are quite expensive and few labs need such instrumentation (remember, the 0.2 PPM uncertainty still applies to this bridge as well).

Thank you Vgkid for posting the link to the current manual.

[bingo600]

Thank you vgkid & Edwin

Manual study time (even though its 100% out of my budget)

/Bingo

[Dr. Frank]

...

As to the claim of 0.001 PPM, pure nonsense, they can't measure to that accuracy, no national primary lab in the world can measure to that accuracy, the uncertainty will likely be reduced to 0.1PPM in the proposed 2018 standards but that is still a factor of 100 over 0.001 PPM.  Any one who makes such a silly claim is suspect and is not worthy of consideration.  Any serious metrologist would laugh at such a claim.

That Wekomm standard is not serious, I agree.
Without a thermometer and a calibrated R(T) curve, it's nonsense to claim for a reasonable secondary standard.
Also, current SI Ohm is not better defined than 0.02ppm, 0.2ppm,  therefore no standard on the world can be less uncertain.

SR104 may be stable and uncertain to 0.1ppm  , and that's the limit of todays artefact standards.

If the SI will be redefined in 2018, Ohm will be uncertain to zero ppm, in definition.

The underlying quantum standard, the von Klitzing Hall effect resistor, is proven to be uncertain to 1e-19. (Definition equals mise en pratique)

Transferring this to any secondary standard, made of any metal alloy, is limited to about 1e-9 (0.001ppm), due to thermoelectric voltages of that order.

Frank

[ltz2000]

From what I've seen and read on this blog site, I very likely have the most accurate resistance measurement system of anybody here.  DVMs are NOT considered standards under any definition, just check with your National Primary Calibration Laboratory and they will tell you the same thing.  Even the 3458A is not considered a standard under any measurement conditions, they are not used as standards by any metrology lab I know of.

At 10K, my 242D system is accurate to 0.2 PPM as calibrated against an SR104, I also have the same 0.2 PPM uncertainty as everybody else.  That gives me a 5:1 accuracy ratio when comparing a resistor (standard or otherwise) over the range of 120R0 to 1Meg, accuracy begins to drop off above and below that range.  Therefore I am barely accurate enough to test the Wekomm standard under the required specifications.  While I have a resolution of  up to 0.0001PPM, it has no bearing on the accuracy.  A National Lab must be able to to maintain a 10:1 accuracy ratio. Under these requirements I do not see anyone on this blog (unless unknown to me) that has the required measurement system to verify the Wikomm 10K standard or even make an accurate measurement of it.

The organisation I used to works for carried out a comparison of their resistance calibration systems at 10 kohm. The three competitors were an ESI 242D, a Guildline current comparator resistance bridge and an in-house built automated system made of 3458As measuring voltage, a nanovolt scanner and a stable current source.

The standards used were four stable SR104s with a long history and stored in an air bath. A set of ten cross measurements between the standards were made with each system during the day and another set a week later.

I couldn't find the paper so I am missing the exact data and the analysis. But from memory the 3458A system was the best. The results of the cross measurements were consistent within a few hundreths of ppm which was quite amazing. The current comparator was working at the high side of the measurement range but still performed much better than the specifications. The 242D was a more than a decade behind the current comparator and would have been even worse if used outside temperature stabilised lab.

Even with an older 242D and 'used' standards, you are still looking at a good $20,000 USD plus traceable calibration requirements.

Manual study time (even though its 100% out of my budget)

www.ebay.com/itm/221765322225

I got one for free and bought another from Ebay for $300 a few years ago. Both are the high accurary version 242D / SP3632 manufactured in the late 1980s.

[acbern]

If you look at the manual of the 242D, it is very apparent that this instrument is very high in resolution but relatively 'bad' in accuracy. An accuracy of 0.2ppm (in transfer mode) with the 242D in practical use is hard to guarantee. And in absolute measurement mode, it is anyway 'specified' much worse. E.g. the best drift rate of the 8 digit 925 resistance decade, which needs regular calibration, is 3ppm/K. In other words, at only 0.1K delta between the calibration temperature and the working temperature, some undefined number of resistor(s) within the instrument may (or may not) have drifted by 0.3ppm, other resistors even more (worst spec is 15ppm/K). And you cannot tell how that translates into the precision of a specific measurement and how it affects the above stated 0.2ppm accuracy, as you dont know what drifted how. A GUM error propagation calculation would essentially be impossible. And maintaing 0.1K stability over the whole decade is anyway merely impossible, so realistically we talk about several ppm of possible drift for the decade. Now you may say that this is a short term comparison measurement only, little temperature changes during that periode, but then the fact that between the reference measurement and the DUT measurement totally different decade resistors may be used (depending on the Standard to DUT ratio), is not taken into account. And these different sets of resistors may have drifted appart isnce last cal. Using the 240C deviation bridge, above 6ppm of difference (not  a lot), is only recise to 1ppm (para. 2.4.2). And it goes on with missing information about drifts for the resistors, stability of the voltage sources, trimming adjustment of resistors to 1,5ppm or worse and so on and so on.
So the 0.2ppm stated here, sounds just impossible to guarantee under standard conditions (even though of course, typically, it may be fine, but you cannot use this in a calibration).
242s sell at a relatively low prices used, see the link above, and I have always refrained from buying one, as using a 3458A in voltage comparison mode, has a well defined and much better accuracy, and has much less unknowns and allows to do a error propagation calculation. It is more work to do a resistance measurement though.
The highest accuracy devices today are still DCCs, see current data sheets of Measurements International for example or Guildline. They are pretty expensive though, and only are justified for primary labs or manufacturers of high precision resistors.

[Edwin G. Pettis]

@Frank,

The current standard SI uncertainty is quoted at 0.2PPM for a 'regular' calibration, but if the uncertainty is reduced to 0.1PPM in 2018 this only implies an improvement in the absolute measurement accuracy of the SI standard, it does not imply any improvement in the accuracy of the Quantum standard as it is already considerably higher than 10:1 accuracy.  Every standard of measure has some value of uncertainty in the measurement, that cannot be removed by any foreseeable technology.  I would anticipate that after 2018 my SR-104 may have an uncertainty of 0.1 PPM during that calibration in which case the uncertainty of my 242D will also be reduced by the same figure.

@Itz2000,

Technically, I can calibrate my 242D to have a zero PPM error as calibrated against my SR-104, however the limiting factor is still the 0.2PPM uncertainty, there is no way to know for sure what the exact value of the SR-104 is, that is why there is an uncertainty factor applied to the measurement.  My 242D can resolve 10K to well below 0.1PPM and it may even repeat but that doesn't eliminate the uncertainty of the measurement.  Now you could send the SR-104 to a National Metrology lab, such as NIST and request that they provide a measurement which exceeds the 0.2 PPM uncertainty.  There are two problems with that, one is that they are going to charge you several thousand dollars to do the measurements IF they agree to do it at all and secondly, they may question whether or not the standard could survive shipping without any change in the calibrated value.  Unless your SR-104s were calibrated to beyond the 0.2PPM standard uncertainty, your measurements cannot be any better than that.   You may have been able to measure stability to better than 0.2PPM but that does not remove the uncertainty.  You can indeed quote better measurements, however the uncertainty of the standard still takes precedence over those measurements, therefore you certainly can say that the repeatability of your system was within ~0.05PPM but the accuracy of the measurement is not the same.  So the automated system, in the end, has no better accuracy than the 242D (apparently that system was not as well calibrated as mine).

Considering that you say they achieved better than 0.1PPM, I'm betting that system cost a whole lot more than a 242D and a SR-104 plus with so many more instruments and connections involved, keeping thermals and errors well below 0.1PPM would be nothing short of a primary lab capability.  I do not know why you are complaining about a 242D in a temperature stabilized room, that is what a metrology lab is by definition and I rather doubt the rest of your equipment would have performed as well outside of a controlled lab environment either so your comment is rather mute on the subject.

The Guildline has about the same accuracy as the 242D at 10K, where it really shines is with low ohm resistors where the 242D has more limited resolution (this was a design decision not so much a limitation of technology), the Guild line can exceed <0.05PPM, I have heard of 0.01 PPM with careful calibration and depending on the value.  Again, the Guildline is still limited by the 0.2PPM uncertainty even though it can measure to ten times better than that.

Finally, my original comment was that I do not believe that anyone on this blog has the required measurement system, such as I do to verify a Wikomm standard.  The fact that a large company may have the where with all to spend tens of thousands of dollars on a near useless exercise does does meet the stated parameters I put forth.  I am certainly not in competition with NIST over resistance accuracy, for that matter, very few companies have the same resources that NIST has nor are silly enough to claim so.  0.2PPM uncertainty is certainly good enough for any of my customers unless they are in the market for an SR-104 (same uncertainty) and I am not ready to take on that as yet.

[Edwin G. Pettis]

@acbern,

The manual (any manual) is not going to tell you what the capability of an instrument is in actual practice.  I have used and calibrated 242Ds since 1973 and they are quite capable of exceeding manual values.  Since it appears you have not had any experience with an 242D, you are merely speculating on what may or may not be its capabilities.  Direct measurement mode wholly depends on how good the calibration of the machine is and if you are depending on another cal lab to do it, I guarantee they will absolutely not take the time to calibrate it to capability only to 'spec' and yes that includes your 3458As as well.  I've been in calibration for over 40 years so unless you manual readers have any hands on experience, you do not qualify as a accurate source.  The uncertainty which I specify is the uncertainty of my SR-104, I can calibrate my 242D to practically the exact same value and in-turn calibrate the rest of the 242D to nearly the exact same accuracy on the main decades, I am currently within 3 PPM at 10Megohms despite the fact the manual says 10 PPM.  Since I cannot adjust the lowest decades, the accuracy does indeed fall off, I do have other methods of greater accuracy including direct comparison against low ohm standards improving accuracy considerably.  Over the main adjustable decades, I am within the uncertainty limit.

Your drift speculation is only valid if the room temperature varies significantly (yes, I can indeed measure this) that is why thermometers were invented, I can check any calibration point on the 242D quite quickly if need be and if needed, make the appropriate adjustment if needed, it is exceedingly rare than any customer would need resistors made to even PPM levels let alone sub PPM levels.  Unless you are a Metrology lab or a manufacturer making very precision measuring equipment, there is absolutely no reason for PPM resistors, even I rarely need resistors like that and I have business reasons for them.

The 242D can indeed hold tight accuracy over more than just a few hours since I have thousands of hours on 242D, I can be considered an expert on the subject.

As to the 3458A accuracy, just where does it say in the manual that it has PPM or sub-PPM accuracy except on possibly the 10V DC range.  The linearity of the ADC is not the same as accuracy, in using the 3458A in ratio mode to compare two components, you still must have a reference component of sufficient accuracy and stability for the measurement to compare to.  If you are using a SR104 as the reference, that also means that the voltage across the unknown is limited as is the ratio mode voltage range of the 3458A.  Even if you are using the internal 3458A reference you still must provide a reference component to compare to the unknown.  In the example of an SR-104 compared to an unknown 10K, you've already lost half of the adc range on the resistor divider.  Linearity is not the only specification in play here, check your manual and you will find several other specifications in play as well, that possible 0.02PPM linearity is likely swamped out by the other factors and lets not forget, DVMs also have uncertainties just like everything else.  It is also temperature sensitive just like most other instruments.  All in all, I think I'll trust my 242D system over your 3458A.

You mentioned stability of the voltage source for a bridge?  A resistance bridge does not rely on voltage source stability, it only needs to be reasonably clean (not too much ripple or large amounts of noise), the measurement is differential, the PSRR is very high, much higher than a DVM.  The main concern is keeping the wattage down to reasonably low values in the resistors.  Your 3458A may have more problems with dirty or unstable voltage sources.  The use of math to reduce errors is questionable, particularly if it is not used with care, just because a DVM has math doesn't mean it makes your measurements any better.  There are limitations to reducing errors with math, statistics can easily lie, I don't need math with the 242D and an accurate measurement with the 242D does not take that long, I don't know why you might think that except that you haven't used one.  Just how long does it take to set up an accurate measurement with a 3458A in ratio mode?  Now be careful here, I know people and metrology labs that do use 3458As so no fudging. <grinning>

For those who have a 3458A, it is convenient to use it for various measurements, nothing wrong with that, just so long as you know the limitations.  Primary cal labs (such as NIST) do not and will not consider a DVM as a standard, not even the 3458A, it can be used as a monitor for instance, NIST has a number of Fluke and HP DVMs in use but not in any capacity as standards, they do not qualify as such.  I know of 3458As which are used to monitor LTZ1000A voltage references in laboratories and when they go in for calibration, they get the 'deluxe' calibration which costs a lot, they are handled with kid glove care, HP actually picks them up, transports them to the lab, calibrates them and transports them back and you folks (me included) aren't in that league of calibration but they still aren't standards even though they are cycled every 6 months.

 Again, DCCs can out perform the 242Ds on low ohms, by the time they are up to 10K, they are about the same accuracy as the 242D (never mind the stupid manual), a DCC can be justified if you need the low ohm accuracy, otherwise you are better off with a 242D.  DCCs do not measure resistance by the same method as the 242D, they have a very intricately wound set of ratio transformers that do the real work.

Let's say one of you gets a Wikomm 10K and you just happen to have a calibrated 3458A (but you don't know what it is actually performing to on the various ranges, it is just within specs), what are you going to use to compare the Wikomm to using the 3458A's ratio mode?  You already have some unknowns in the 3458A so you're going to need another standard such as an SR104 (that is calibrated of course) for the ratio comparison otherwise you're dead in the water before you even got started.

[Edwin G. Pettis]

Curious, I did not underline all of that!

In the modify window, the wayward underlining is not showing up at all.  Just a minor annoyance, sorry folks.

[acbern]

Edwin,

obviously, neither a 3458A nor a 242D are standards, they are meters. Standards, by definition, are e.g. a SR104 standard resistor or 732 voltage reference or a rubidium clock. Meters are used to compare standards with DUTs. This works the better (precision-wise), in a nutshell, the lower the transfer inaccuracies of the meter are (such as temperature drift, applicable aging, non-linearity...; and not to forget other gear involved). Under GUM, you cannot use the 3458A for direct resistance comparision, unless you have done a lengthy, and pretty limited (to essentially only the condition you validated) statistical analysis for your specific instrument. The 8508A or even the 7081 are better spec'ed here. The resistor transfer accuracy unfortunatelly is not specified by Keysight, although of course, when you try, and many here have, you are somewhere in the 0.1ppm region. Still, you cannot use this accuracy for an error budget statement under GUM.
To do exact resistance measurements, you would do that in voltage divider mode. Here, the 3458A has very precise transfer specifications, which actually are independent of wether you do a gold standard cal or a normal cal. You do a comparison, comparing (voltage drops across) e.g. a SR104 and a DUT. This is essentially a linear relationship (the SR104 being the standard), under mathematical consideration of the various error contributions and weighing factors. You can do repeatable measurements to say 0.1ppm ball park, but when you do the math, depending on the other gear you use, you rather are in the 1ppm ball park (even with very stable voltage references). It may not sound very exciting (or be considered conservative) to have a factor of say appr. ten between how precise you think you measure and what you can prove mathematically with k=2, especially as sub ppm accuracies seems to be kind of en vogue here. More details are in the 3458A spec and related historical HP/Agilent design background data, frequently discussed here.
And yes, unfortunatelly, we manual readers do rely on what data we get from gear vendors, thats how the procedures work, and how the cal labs work, and unless we want to do a research exercise specifying a gear ourselves, as per GUM rules, and with potentially unclear results, we use what we get specified to establish a error calculation. If no data is available, it is in most cases more efficient to use something that has appropriate data. The 242 has not.

It may well be that your 242 typically is more precise than spec in some (or many) ranges (you do not know for sure, unless you have a more precise instrument of which we dont know yet), but you cannot prove this in compliance with GUM error budget guidelines. Maybe you have done that, let us know, we would certainly be interested to see it, maybe you also have specs other than what is in the manual. Or, maybe, your statement that you 'dont need math when using the 242' says it all.
After having had a look at the 242 manual long ago, I decided for me it is not an exercise that is worth it, given all the missing data and (specified and unspecified) drifts. This is a pitty, as the 242 is well priced in used condition.
With respect to accuracy data of DCCs, I would recommend to consult MI's data sheets, they are factors more precise than the 242, yet at a different price tag too (50k region), allowing use in national labs. The only option worth it, for me, would be an old 9975.

[engiadina]

@Edwin G. Pettis

So you are telling us, that this document

http://www.kalibriercentrum.de/pdf/DAkkS_Urkunde_Bayern.pdf

is basically rubbish ?

Wow.

[mikerj]

Curious, I did not underline all of that!

In the modify window, the wayward underlining is not showing up at all.  Just a minor annoyance, sorry folks.

You have three opening underline tags around the word "accurate", and the closing underline tag at the end of your post.

e.g.

Code: [Select]

[u]accurate[u] [u]

...
...
...
[/u]


[TimFox]

@Edwin G. Pettis

So you are telling us, that this document

http://www.kalibriercentrum.de/pdf/DAkkS_Urkunde_Bayern.pdf

is basically rubbish ?

Wow.

My German isn't as good as it used to be, but which items on the calibration facility list disagree with E G Pettis' posting?

[Edwin G. Pettis]

@engiadina,

While I do not read German, I do understand the listings for the various measurement types, I do not understand why you think what I said negates this document.  While a few of the resistor limits (if I understand it correctly) are a bit different than some of the ones I stated, that does not in anyway negate the document.  My calibration is in compliance with our NIST standards and uncertainties which may be better than in some other labs.  I see nothing wrong with the document you presented.  It is nothing more than a statement of capabilities of this particular calibration lab which most labs compile for their customers.

@acbern,

You are half correct, DVMs are not considered standards period, ask a metrology lab and they will tell you so, the 242D is indeed a standard in its own right and by definition is a standard.  It is fully calibrated and traceable to which ever national metrology lab your country maintains, here it is the NIST.  I see no contrary explanation for that.  It is certainly not a 'meter', while it has a null detector, that in no way makes the instrument a meter, that is just plain silly.

As to the unnecessary underlining in my earlier post, it is of no consequence so it does not warrant any further effort.

[Dr. Frank]

Dear Mr. Pettis,

I would like make some interesting clarifications about the Ohm uncertainty:
The BIPM lists on its KCDB page all National Metrology Institutes (NMI), i.e. their capabilities and uncertainties: http://kcdb.bipm.org/
When you search for the German PTB, or the US NIST, you will find the hint, that the Ohm is relative to R(k-90), that is the von Klitzing / Quantum Hall Effect (QHE) standard, not relative to S.I. Ohm!!

These NMIs either use a DCC, Direct Current Comparator, which gives 0.2ppm or less uncertainty, or the CCC, Cryogenic Current Calibrator, which allows comparisons as low as 0.004ppm uncertainty, directly to the QHE! See listing for the NIST in the attached document!

Explanation of the DCC and CCC bridges used, can be found here: http://www.nist.gov/calibrations/upload/tn1458.pdf

According to that, it's very well possible to make measurements /comparisons with much lower uncertainty than 0.1ppm, even outside an NMI, and without direct access to a QHE.

Therefore, this Wekomm resistor, which Dave received, and which is a prototype only, really had been calibrated to 0.1ppm uncertainty, by this Bavarian calibration lab! Dave published the Wekomm calibration document, where you can find further details:

The capability of the lab relies on their (decades old) GenRad 1444A-10kOhm standard, which is the predecessor of the IET / ESI SR-104.
This is an ultra stable artefact, much more stable than the usual SR-104, and which has been calibrated many times directly at the PTB, so its stability of about 0.01ppm/yr. has been characterized since a long time.

They also use an ultra precise bridge, an MI 6010B, which obviously allows direct 1:1 comparisons below 0.1ppm level.
(Neither do I know that bridge, nor how it compares to the 242D bridge of Mr. Pettis.)

There is obviously a demand by the NMIs, especially our PTB, for more stable working-standard resistors, than these Thomas type 1 Ohm and the SR-104, which are mostly too unstable to reach the 1e-9.. 1e-10 comparison region of the QHE and CCC bridge.

The Wekomm resistor, when a precise thermometer and T.C. characterisation will be implemented in the future, is a candidate to improve these standard resistors by orders of magnitude regarding stability. Dave already mentioned, that our German PTB is involved somehow.

I had an interesting discussion with 'engiadina', who is the design engineer of this Wekomm resistor, who confirmed that in detail.

It was a big surprise for me, what are the weak points of the current Ohm metrology, and what seems to be possible with these upcoming Wekomm resistors, implementing  Vishay BMF technology.

Anyhow, in 2018 the difference between the QHE and the S.I. will probably vanish, so that the S.I. Ohm really will be precise to exactly zero ppm. requiring also better working standards.

Frank

[ltz2000]

The capability of the lab relies on their (decades old) GenRad 1444A-10kOhm standard, which is the predecessor of the IET / ESI SR-104. This is an ultra stable artefact, much more stable than the usual SR-104, and which has been calibrated many times directly at the PTB, so its stability of about 0.01ppm/yr. has been characterized since a long time.

The GR 1444-A was introduced 5 years later as a competitor to the ESI SR104. It generally turned out to be less stable and never became popular. But there still can be (and in this case obviously is) very very good individual units.

There is obviously a demand by the NMIs, especially our PTB, for more stable working-standard resistors, than these Thomas type 1 Ohm and the SR-104, which are mostly too unstable to reach the 1e-9.. 1e-10 comparison region of the QHE and CCC bridge.
The Wekomm resistor, when a precise thermometer and T.C. characterisation will be implemented in the future, is a candidate to improve these standard resistors by orders of magnitude regarding stability.

A typical SR104 drift rate is approximately +0.07 ppm per year. Magnitudes better would really be something!

In a document I posted earlier the VSL measured their Vishay VHA518-11 resistors against the Quantum Hall. The Wekomm uses the VHA518-7 which is the same type except there are only 7 bulk metal foil chips instead of 11 inside the hermetic can.

The two resistors measured by the VSL showed drift rates of 0.4 and 1.9 ppm/year. And most importantly the drift rate it is not (yet) predictable.

Even if Vishay had some magic to improve the stability that much, why would they share that secret only with a small German toy train manufacturer?

[Dr. Frank]

A typical SR104 drift rate is approximately +0.07 ppm per year. Magnitudes better would really be something!

In a document I posted earlier the VSL measured their Vishay VHA518-11 resistors against the Quantum Hall. The Wekomm uses the VHA518-7 which is the same type except there are only 7 bulk metal foil chips instead of 11 inside the hermetic can.

The two resistors measured by the VSL showed drift rates of 0.4 and 1.9 ppm/year. And most importantly the drift rate it is not (yet) predictable.

Even if Vishay had some magic to improve the stability that much, why would they share that secret only with a small German toy train manufacturer?

Obviously, it's not the standard VHA518-11, Mr. Everybody would get.

It was a very interesting story, and Dave already made some hints about cooperation between Wekomm and PTB.

Maybe, engiadina may tell some background info on his own, I dare to do so.


Frank

[engiadina]

Let me give some answers:

why would they share that secret only with a small German toy train manufacturer

Well, we are not exactly a toy train manufacturer. We develop and produce electronics, and yes, some of our customers are toy train manufacturers. On the other hand, developing electronics for those applications does not qualify or disqualify a company in any way. We have car manufacturers, big stock industries and high technology companies as our customers. Our WEB-Site could be improved, there I have to agree totally. Will do that !

Regarding the relationship towards Vishay. First of all, you should know that VPGSensors is now seperate from Vishay. VPGSensors make their own business and focus on precision resistors and strain gauges. So the company itself is much smaller.
Second, it is a question of demands. We did for some years research with standard VPG Resistors and identified some areas where the performance could be optimized, let us call it this way. Then we addressed VPG with our results and with a bunch of proposals how to implement those optimizations. We were extremely lucky to find some open ears at VPG. They realized that our proposals could improve their products as well so a very open and friendly cooperation developed.

Some ideas are much too expensive to put it in series production but their facility in Israel is able and willing to implement them in the resistors we get. They are surely extremely selected, not by nude virgins of course but rather by men with the grey beard. During our research we found a lot of issues which can happen if someone does not treat the resistors in a specific way. Of course, if you take a VHA518-11 and have it specified at 0.01% (10ppm) then nothing special has to be regarded. But do not forget, we take the drift up to 1ppm and the uncertainty to 0.1ppm.

One key lies in the ominous PMO, mentioned in the datasheet. This "Post Manufacturing Operations" help stabilize the resistors quite a lot if done in the right way. As we found out, some effects of the PMO can be reverted, if the resistor is mounted afterwards. So we developed together with VPG a distributed PMO process, which is partly done in the factory of VPG and partly done at our place. By this way we can guarantee the properties of the resistor much better.

It is not only the resistor itself which matters. It is every piece of the construction which has to be engineered properly. We had our own cable type manufactured for example, which provides lowest thermal voltages and best isolation. The mount of the resistor itself keeps it from every strain of outside impact, etc.

Every step we took was checked and verified with our own 242D Bridge. But this is only an indication if the direction is correct. Often we have put the resistor to a standards lab which is using a MI bridge and some Thomas resistors in an oil bath. They could verify the success or in many cases showed us some weak points.

So developing a resistor with that precision is not a fast task. Certainly not going to Vishay, asking them for a precision resistor and putting a box around. There is much more behind it, using years of experience and is hidden sometimes in very small details. The reason why we are now on the market is that we verified, that we can produce reliably those resistors according to our specifications, which are mostly much better that a competitor product from F...e.

Of course we showed the resistors to the german national standards lab, the "Physikalisch Technische Bundesanstalt" or PTB where they have a Cryo Comparator to measure and define the "technical" Ohm down 1E-10. What I can say for now is that they checked our resistors against that Cryo Comparator and are extremely happy with the results.

So I think we are on a good path for now. We have a second type of resistor in development which has included temperature sensors and features some much more rigid specifications in term of stability. Even if this takes time, we even now are sure that this goal can be achieved. There is still some potential in those Vishay Foils which has to be activated by the right techniques.

But for now our actual resistor series are already better (and cheaper) than most of the so called resistance standards. Only ESI 104 and GenRad 1444A have better specs, but they are way more expensive.

Oh, before I forget. The mentioned VHA518-7 resistor mentioned in the marketing paper from VPG refers to a previous, much earlier development status. Now we use a wekomm resistor, manufactured by Vishay Precision Group. This is probably the best way to describe it.

[Vgkid]

Thanks for the response.