Precision Resistor Standard

[echen1024]

I'm sorry. I know there's a distinction, but forget about it form time to time.

[Dave Turner]

The 'simple' definition that I was taught was to imagine shooting at a target.

Precision was grouping shots within a small circle whether on or off the center of the target.

Accuracy was when the centroid of the shots was on the bull.

Not mutually exclusive but statistically different.

It all depends on what one is attempting to prove.

[linux-works]

precision is about repeatability.  hitting the same thing over and over again.

but that thing would be, in absolute terms, 'off'.  and that's where accuracy comes in.  accuracy is how well your view of 'x' matches with some authority's view of 'x'.

[Dave Turner]

Exactly - my definitions hold. - no contest

[echen1024]

I was also taught a "pi" example in Elektor. It stated that:
3.14= accurate, but not precise
3.34573894579834798347= precise, but not accurate
3.141592653589793238462643383279502884197= accurate and precise.

[HighVoltage]

These days in my world, everything is all about Six Sigma compliance and all measurements need a Cp and Cpk value attached.
So, we have to use 6 1/2 digit meters, although for the real application, this kind of precision is not needed.
Well, to calibrate or check the calibration on a 6 1/2 digit meter, we need to have at least one more digit to check against.
It really is overkill in most respect, but required by some customers.

[echen1024]

Vishay HZs are coming tomorrow!

[Vgkid]

How much did those cost you?

[KJDS]

These days in my world, everything is all about Six Sigma compliance and all measurements need a Cp and Cpk value attached.
So, we have to use 6 1/2 digit meters, although for the real application, this kind of precision is not needed.
Well, to calibrate or check the calibration on a 6 1/2 digit meter, we need to have at least one more digit to check against.
It really is overkill in most respect, but required by some customers.

I wasted an hour of my life years ago in a debate with a customer. We had HP 34401 6.5 digit meters measuring the supply current on an amplifier. Their spec was less than 650mA and we were typically drawing 350mA. A quick look at the production stats showed the maximum for the last 10,000 unit's was 374mA, so we are always a long long way in spec. However no-one had documented the required measurement uncertainty from the set up so that was a QA fail. It took a long long discussion and sufficient maths to confuse him that it wasn't an issue.

[echen1024]

How much did those cost you?

Samples. I have to write a case study.  They are apparently around $150 each.

[HighVoltage]

I bought these from Mouser

Y145310K0000V9L
Vishay Precision Group Metal Foil Resistors - Through Hole
TARIC:8533210000 ECCN:EAR99 COO:IL
10 kOhms
Euro 19,14


Y00071K00000V0L
1000 Ohms
Euro 15,17

Y1453100R000V9L
100 Ohms
Euro 21,15

Then inserted them in to an aluminum housing with some good quality 4 mm gold plated banana plug sockets
They work extremely well, if I compare them on different high end meters

[acbern]

The key with resistance standards is stability, not so much abs. Tolerance. That means essentially temperature stability and aging. Temp stability can easily be achieved by vishay foil resistors, they are 0.05ppm/k stable arround room temp. Aging, mostly caused by humidity variations is more of an issue. According to vishay, a foil resistor can vary by 20 ppm caused by humidity. This also seems reversible, as they state. For precision calibration (eg calibrating a 3458a), this is nor acceptable. Also, btw, esi 1010 or similar are not much more stable. The only way to solve this is packing thm in oil. Vishay does this with their hermetic products. Therefore the question raised in this blog absolutely makes sense to put a vishay standard foil resistor in oil. Not for temp stability but for aging. Has anybody done that?

[Thor-Arne]

If I remember correctly, the issue of long time stability versus humidity was discussed in another thread about voltage references.

I don't think there was a conclusion on the topic.

But, one thing that was discussed was to use a silica gel humidity indicating desiccator (like rs#  838-732) in a sealed enclosure. That would be a lot less messy and no bad effects on the devices from the oil.

[Dr. Frank]

Temp stability can easily be achieved by vishay foil resistors, they are 0.05ppm/k stable arround room temp.

Nope. Mostly all Vishay metal foil resistors are specified 2ppm/K max. T.C., measured in a certain temperature box, 15..35°C at best.
Those 0.05ppm/K is a typical value only.Vishay does not guarantee at all, that this low T.C. is achieved on real samples, especially not at room temperature of 25°C.

But it is possible to determine the physical T.C. or the R(T) curve.
So the real value can be corrected from measurement of the ambient temperature to high precision.
My real 5 resistors VHP202Z have T.C.s from -0.3 to -1.0ppm/K,  linear between 15 and 35°C, which makes a correction easy.

Aging, mostly caused by humidity variations is more of an issue. According to vishay, a foil resistor can vary by 20 ppm caused by humidity. This also seems reversible, as they state. For precision calibration (eg calibrating a 3458a), this is nor acceptable. Also, btw, esi 1010 or similar are not much more stable. The only way to solve this is packing thm in oil. Vishay does this with their hermetic products. Therefore the question raised in this blog absolutely makes sense to put a vishay standard foil resistor in oil. Not for temp stability but for aging. Has anybody done that?

Molded resistors drift by influences from humidity and oxygen, as Vishay claims.
This should be an irreversible effect (corrosion).

It makes no sense to put those types in oil, as they already offer those hermetically sealed, oil filled versions.
The resistive element is directly embedded in oil, therefore removing any humidity and oxygen from the start.

If using molded resistors and an external oil bath, the oil will presumably not soak into the mold compound, and will not embed the resistive element.
It will only avoid that additional portions of humidity and oxygen will reach the resistive element.

So that's a cumbersome, 2nd best solution to the problem.

Frank

[acbern]

Of course the temp corfficient would have to be determined, not being specified. However this is pretty simple as has been said. Humidity effects seem to be reversible, which is strange, but has been stated by vishay in a paper. Using the right oil is key, should be silicone oil, there are spcial versions available for high voltage/low conductivity electronics applications. Of course it may be a mess if not properly sealed, but this is solvable.
For calibration of my 3458a i bought a 4 wire hermetic 10k resistor, put it in a metal box and attached a 0.1grade pt100 to it, however for my other values i did not want to spende that money. The 10k hermetic cost me about 100euro, compared to 20 for a standard. So for 1 ohm to 10megaohms, the delta price is beyond 500 euro, worth considering. Plus some of these are not available in 4-wire hermetic anyway.
If baked prior to embedding i am pretty shure an effect close to an hermetic is achievable. Soking is not needed, all humidity in the epoxy should have gone, before. Baking is an accepted standard procedure anyway in electronics. Using dessicant seems interesting, i am hesitant, even though there are some nice products available avoiding the usual bags.

[Dr. Frank]

Of course the temp corfficient would have to be determined, not being specified. However this is pretty simple as has benn stated.

It's not trivial, and not easy.
Such extremely low TCs require sophisticated and ultra stable equipment.

How did YOU do it?

Humidity seems to be reversible, which is strange, but has been stated by vishay in a paper.

Can't remember which one.. Do you have a reference or title of this paper for me, please?

For calibration of my 3458a i bought a 4 wire hermetic 10k resistor,

I did the same, 10ppm grade plus test protocol, hoping for even better uncertainty in the 1ppm range. Later it turned out, that the absolute value of the resistors were not measured correctly, probably.

Anyhow, 10ppm is not sufficient for calibration of a 3458A, isn't it?

If baked prior to embedding i am pretty shure an effect close to an hermetic is achievable. Soeking is not neede, as long as all humidity in the epoxy has gone.

Baking is not good either, will cause several ppm of hysteresis, which will give long term creep drift, or a steady offset.
What makes you so sure, besides really testing that?
And oxygen is also an important factor.

[acbern]

In wtxplc.com e.g. there is a vishay prec group glossary that under shelf life declares complete reversability by drying/baking, which would also indicate no hysteresis effect.There are also other docs. Of couse, not to be confused with distructive life tests such as pressure cookers.
Generally important is to keep the resistors under 10mw load to avoid drifts.

Measurements of voltage drop on two resistors, relative to each other, one at constant temperature, one in an oven/peltier heater/coooler, powered by a precision voltage such as 732a or 5440/5700/5720 or 4808, using the very high linearity of the 3458a adc allows precise measurements of small drifts. I also use this methode to calibrate other resistors from the 10k, which saves cost. the error calculation shows surprisingly small errors piling up as you move away from the 10 k. It is a pitty agilent has not specified transfer accuracy of res measurements.
It must be noted that the z-foil such as the z201 is specifed to 0,2 + 0,6ppm/k max over the full military temp. Range, so if you have measured 1ppm, it was out of spec., the 0.05 ppm arround room temp is usual, getting to above 0,5 is pretty unlikely.

I had the 10k calibrated to 2ppm, and hope the lab was correct, at least i paid well for it, unfortunatelly. I have no other way to verify at that level, other than redoing it in a year's schedule. I would certainly be interested in some cross comparisons. I have also decided to spend the money at least for the 10, 100, 1000 and 100 k hermetic resistors to get less drift. Got a good deal.

[Dr. Frank]

In wtxplc.com e.g. there is a vishay prec group glossary that under shelf life declares complete reversability by drying/baking, which would also indicate no hysteresis effect.There are also other docs. Of couse, not to be confused with distructive life tests such as pressure cookers.
Generally important is to keep the resistors under 10mw load to avoid drifts.

Sorry, on my  hermetically sealed, oil filled VHP202Z, I measured a pronounced hysteresis of 5ppm after heating them to 125°C.
Therefore, that has nothing to do with humidity inside, obviously.
This technology is similar to DMS elements (resistive strain gauge), which also suffer from hysteresis effects

I was able to remove that shift only by temperature cycling.
Clearly, I always have treated the resistors for 'shelf life' mode only, P < 10mW.

Measurements of voltage drop on two resistors, relative to each other, one at constant temperature, one in an oven/peltier heater/coooler, powered by a precision voltage such as 732a or 5440/5700/5720 or 4808, using the very high linearity of the 3458a adc allows precise measurements of small drifts. I also use this methode to calibrate other resistors from the 10k, which saves cost. the error calculation shows surprisingly small errors piling up as you move away from the 10 k. It is a pitty agilent has not specified transfer accuracy of res measurements.

Yeah, that is no easy measurement, and you have very sophisticated gear, haven't you?

I 'only' use the 3458A for such measurements.
I could demonstrate a transfer stability of 0.2ppm,  over 20min. and at constant temperature of +/- 0.1°C.

It must be noted that the z-foil such as the z201 is specifed to 0,2 + 0,6ppm/k max over the full military temp. Range, so if you have measured 1ppm, it was out of spec., the 0.05 ppm arround room temp is usual, getting to above 0,5 is pretty unlikely.

Well, Vishay in the past published several too optimistic specifications, with very low T.C.s in the describing text
Reading more carefully revealed that those were typical values only, and in the small table on the first page, at the bottom, they always specified +/-0.2 ppm/K + 2ppm/K max. spread for the T.C.
Therefore, my resistors of 5 years ago were well inside the actual spec, which was the best one available.
When I complained about those resistors, Vishay at that time could not deliver better guaranteed T.C.s, as they obviously could not control the process.

You are right, lately they again specified resistors ,VH102Z, I think, with 0.6ppm/K max. spread.
But if this is true now, and whether they cancel that spec again, I don't know.
The spec looks odd, and the parts are not oil filled. Strange.
 

I had the 10k calibrated to 2ppm, and hope the lab was correct, at least i paid well for it, unfortunatelly. I have no other way to verify at that level, other than redoing it in a year's schedule. I would certainly be interested in some cross comparisons. I have also decided to spend the money at least for the 10, 100, 1000 and 100 k hermetic resistors to get less drift. Got a good deal.

Well, that explains how you got such a low uncertainty.
Definitely, the VHP resistors are very stable over time and very predictable over temperature, even with high T.C.s , so they really make very good precision standards, and the calibration is well worth its money. Although, an ESI SR104 would be more convenient.

Perhaps you really have got better technology inside your resistors, meanwhile.
Which T.C.s did you measure on your resistors?

I'm also very keen to make a comparison with your gear!
Just tell me place and time.

Frank

[acbern]

could it be that if the baking was performed at say 60C such effect would not occure. given increased temperature and low rel. humidity in the oven, that may be sufficient to bake a standard z-foil. I will give this a try later this year when I have my other hermetic resistors and the normal ones are not needed.
my hermetic 10k measured 0,18ppm/k, over the entire rangefrom 10 to 33C. a change of 0.1ppm is not that easy to measure as you pointed out. the coeffizient seems to slightly increase towards the low temperatures, respectively at normal room temp seems to get close to 0,1ppm. so this one seems to be ok re. latest spec. it was manufactured about a year ago. maybe they really have improved the alloy. I have not measured any of my standard z-foils as I have no pt100 attached to them in the housing. I may do a comparison at a later stage though.

always open to do a comparison.

[Edwin G. Pettis]

Vishay has a long history of buttering up their data sheets with various methods and then sometimes hiding the 'true' specs somewhere in the fine print if they think about it.

Like all tight spec'd resistors, they are selected from a batch by running a TCR on all of them and picking the ones which appear to be within the 'real' specs, the remaining resistors which fail are sold with reduced TCR specs at still impressive prices.  No, you cannot 'adjust' a resistor's TCR like you can 'trim' a foil resistor into tolerance by cutting shorting paths on the foil pattern.  The TCR is created by heat treating the raw foil to adjust the TCR down to near zero but it is a hit or miss process which can require repeating the treatment until it is within range.  It is not possible to control the several metal quantities in the alloy to a fine enough mix to get the same results every time.  Hence, it is expensive.

The 'z' foil resistors are good resistors, even if you go by their 'true' specs, however these foils have limitations.  A restricted temperature range and power level to achieve these impressive specs, very often they require hermetic sealing to get the job done which invites really impressive pricing.  The drift specifications are again under very limited conditions, we refer to them in the industry as 'kid glove care' otherwise not so good things happens.

A really good engineered resistor will have great specifications even at full power for prolonged periods, they will have excellent TCR over a wide temperature range, they will have very low long term drift under both power and un-powered conditions.  They will show a very small drift in value after a minimum 5 cycles of thermal shock with zero failures. A really good resistor is not affected by humidity.   A really good resistor does not need to be hermetically sealed to meet all of these requirements.  Sadly, marketing hype is a very poor way of achieving a great resistor.

Sub-PPM TCRs are difficult to achieve consistently with current technology.  While we all would like a 'perfect' resistor, one has yet to be built.  In the mean time, buyer should be wary of sub-PPM claims, as pointed out earlier, they are very difficult to measure accurately and that is the key word here.  Just ask any primary calibration lab and they'll let you know just how difficult it is to do by the calibration fee they will quote you.  Everything is controlled, fleeting measurements under questionable conditions still has significant uncertainty error and it doesn't appear that any of the measurements quoted here is taking that into consideration.  Even ratio measurements have uncertainty gentlemen, remember that the next time you are quoting sub-PPM measurements.  You'll find that the uncertainty is much larger than the measurement you are quoting.

[acbern]

some comments to the above:

a calibration resistor would e.g. not be operated at full power over long times. the purpose of the resistor is to define a standard and should be used as carefully as possible, as is general metrology practice. you would never operate your esi 104 at its max power over lng time, its drift will increase, a simple result of the arhenius law. you cannot change physics. I would as far as cal resistors are concerned, not agree with the statements above about what a good resistor is.

you would not do a drift measurement at that precision level without knowing what the errors are. the 3458a e.g. has no specified resistor transfer tolerance, although people have demonstrated good results. you need to e.g. do a voltage comparison measurement (again, applying failure propagation analysis) with a 3458a and a very stable voltage reference. alternatively the solartron 8 digit meter has a resistance transfer spec of 0.1ppm. or use a guildline 8 digit bridge. all not cheap stuff, but neded if you need to work at a certain accuracy. I would think most people using this equipment know what they are doing. and as you do temperature drift measurements the absolute value initially is not that critical, as you can measure the drifts and then do a recalc to its known cailbrated standard value at the cal temp. and attach a failure propagation calc to it.

so although I am not a vishay emplyee  I still believe it is a very good resistor for what it can be used and considering its price. you need to know what its limits are and use it accordingly. have a look at e.g. the fluke resistors selling in the 1k range, and look for the true specs (it is a different doc, not in the manual, they revised it) re. aging and tempco. 

I measured my hermetic 4-wire foil resistors at about 0.2pm/k, which is ok for many users, and if thats not good enough you can operate them in a slightly heated (say 35C) copper block (only heat it up when needed). a national metrology lab has written a doc about this and they have made a lot of comparisons with the ESI 104 gold standard, and results were very good. i paid 30 euro for one, high resistives are more expensive, at 0,1% abs (I cal them anyway, so abs value not critical as long as i know what it is exactly).

[Edwin G. Pettis]

acbern,

You are comparing apples and oranges here, the SR-104 is a primary calibration standard and hence has to have certain limits applied to its operation for long term stability, that has been traditional for primary standards for decades, they are by nature a bit delicate.

The Vishay foil resistors being discussed here are not primary standards, not intended to be primary standards and are not under normal conditions standards.  If you are going to attempt to use them as a working standard, the rules are similar but the specifications are relaxed in general.  At best, under controlled conditions, you might be able to call them working or transfer standards.

The resistors I was referring to are not primary standard resistors (you seem to have jumped to the conclusion that I was comparing them to an SR-104 which is incorrect) but real resistors which can operate under the conditions I mentioned and exceeds the specifications of transfer standards and in some respects even a primary standard.  Nothing I said about those resistors defy the laws of physics.  They are by any 'working' resistor comparison, exceptional precision wire wound resistors.  They are designed to take severe operating conditions and be very stable and reliable.  I would suggest you check out the thread; T. C. measurements on precision resistors, it contains much more information about the real 'good resistor'which I will not repeat here.  One of the points I'm making here is that stability and reliability has to be designed into the resistor, if it can't exhibit exceptional characteristics under duress, then it is just another common resistor.  Hermetic sealing is just a band-aid to cover up a flawed design.

The SR-104 (and other similar resistors) undergo complex manufacturing processes in order to achieve the 104's impressive performance, I am familiar with the general details of these processes and it is quite impossible to use them for other classes of resistors, the processes are, in total, expensive and time consuming.  On the other hand, Vishay foils are not the only resistor to achieve sub-PPM TCR performance but they are, inherently more fragile than a wire wound resistor both physically and in specifications.  By comparison, they require more of the 'kid glove care' to hold on to their specifications.  There is not enough space (or time) to go into details about the processes used to manufacture exceptional resistors (the SR-104 is, of course, a wire wound resistor of the same alloy that I use), in fact the Vishay foil resistors are also made from the same alloy, albeit with different processing, they are made with a rolled sheet form unlike wire wounds.

I do not use DMMs for measurement of resistors characteristics except in the most general terms, I use an ESI 242D resistance bridge as my primary measuring instrument and an SR-104, plus other transfer standards, to calibrate it.  The only resistor measurement instrument 'commonly' available which can exceed the 242Ds capabilities is the Direct Current Comparison bridge which is only better than the 242D below 10K ohms.  Yes, I am not only a calibration tech (since 1973) but an electronic design engineer who has been working the precision resistor industry since 1975.  I designed and built the machinery to make WW resistors, test them to military standards and co-designed the best precision wire wound resistors available period.

I did not say that Vishay resistors were not good resistors but they have a long and documented history of buttering up their specifications, I know, I've been around for most of Vishay's history including all of the foil history.  I've even seen the manufacturing processes they use so I know more about their resistors than just about anybody else, out side of Vishay.  A lot of other people also agree about Vishay's spec games, it is not just me.

Almost any 'good' resistor will exhibit very good characteristics if treated in the manner you refer to, i.e., mounted in a controlled heated copper block and a few milliwatts of power applied, that amounts to a resistor sitting on a shelf un-powered, I'm not impressed, how long was this 'shelf' test for?  Vishay's own specification of drift, un-powered, well exceeds the SR-104 and mine for that matter. 

I also understand that you and the others have not had the opportunity to access the resistors I refer to, I hope to make them more available in the near future.  I will likely be posting further documents on the other thread I referred to as time allows.

To some degree, we are making comparisons of five different classes of resistors, the primary standard, working or transfer standards, my regular resistors, foils and hermetic.  There are differences between them, there is some over-lap of specifications to some degree and there are specifications/characteristics somewhat unique to each of them as well.  Each must be taken into consideration for the application and this must be done carefully. 

Please take the time to review the thread, consider and then ask questions, it may change your mind at least a little.

[Flump]

Interesting read Edwin 

[babysitter]

Hello Edwin,

when people settled on the unweldable, unsolderable evanohm, have there been no experiments about electroplating the ends of the wire with some solderable metalisation? I am aware of a supplier of resistive heating parts used to weld plastics foil that puts a (nickel?) gold coating on the end electrodes. They don't solder, it is made for screw connection, but might be favourable for soldering against the evanohm wire.

There are some other alloys - Isabellin, Konstantan, Manganin and their variants are the ones I am aware of - have there been trials of making resistors out of two different alloys with canceling TCR? I think I read even at this forum someone wound a layer of cooper around a existing resistor to add a small resistance and the cooper TCR countered the resistors TCR well.

[Edwin G. Pettis]

Hi babysitter,

Evanohm is definitely weldable, the big problem was that everybody (except the primary Evanohm standards and myself) was trying to weld an un-weldable joint, those bad welds are still being used in the industry which generally limits the performance and stability of those resistors.  As per my articles, the bad solution of baking the heck out of them weeded out the worst offenders and those with tight crimp joints made it through.  As I admitted elsewhere, I have not personally inspected every brand of PWW resistor on the planet but their performance speaks volumes.  The PWW industry has a long and jaded history of borrowing 'ideas' from each other.  All of the 'sandwich' terminations discussed in the articles were thought of almost entirely in the 1960s, I have not seen anything new except possibly some really odd combinations of those sandwiches since.

While solving the weld problem removed a major source of trouble, other, more subtle effects were revealed which also contributed to poor performance, once those were taken care of then state-of-the-art performance was possible, until further material improvements can be made, it is as good as it can get for the time being.

Yes, it is possible to use both the Evanohm alloy (which goes by several other names too) and other similar alloys to create different combined TCRs, it has generally not been done due to some difficulties encountered in the manufacturing process, it has been done in the past but in a clumsy manner which turned out to not be good enough for a production line.  The technique is called splicing.

Manganin and its similar alloys are mostly copper and can be soldered (carefully, Manganin is highly sensitive to high heat and can be permanently changed by too much soldering heat, it has a very low temperature of annealing), in fact most resistance standards made from Manganin are soldered.  It is not possible to weld Manganin to Evanohm easily, Manganin is too soft compared to Evanohm and the two will just mush together making a less than terrific mechanical joint.

The 'trick' of putting a bit of copper wire in series with a PWW resistor does work if you aren't expecting tight, real stable TCRs and you have to solder the copper to the resistor lead of course.  The compensation tends to be accurate over a limited temperature range.  If the copper resistance is a very small percentage of the total resistance, the 'stability' issue of the copper is accordingly minimized.

There is always the devil in the details.  Thank you for asking.  Cheers.