Author Topic: T.C. measurements on precision resistors  (Read 278738 times)

0 Members and 2 Guests are viewing this topic.

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #25 on: June 17, 2014, 10:58:32 pm »
Hi gazelle,

I forgot to add one more detail, particularly in the case of matching resistors.  It won't do you much good to order two (or more) 'matching' TCR resistors if they are dissipating different power levels, as is the case in the LTZ1000/A circuits.  If both resistors are dissipating the same power levels, then the self-heating will be the same and the TCR will track as expected, all else being equal.  If the power levels are different, then the self-heating is going to cause an apparent TCR mis-tracking which will be blamed on the resistors.  Surprisingly, I've seen this mistake made by engineers pretty often.

So in the case of the 12.5K / 1K divider, there is roughly a 15:1 power ratio, if this is not taken into consideration when selecting the resistors....it isn't going to work like the man (or lady) said it would.

I will get back to you shortly on specific sizing and other details, sorry for the delay.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #26 on: June 18, 2014, 12:16:05 am »
Hello gazelle,

To accommodate the materials in stock, these are the best sizes; for the 1K ohm resistor, my 704 style, axial lead, 24 AWG, body size 0.125" D x 0.250" L, for the 12.5K ohm resistor, my 807 style, axial lead, 20 AWG, body size 0.250" D x 0.750" L, the 70K ohm resistor would be a 807 style, the 120 ohm resistor would be an 805 style, axial lead, 20 AWG, body size 0.250" D x 0.500" L, the 30K ohm resistor would be a 805 style.  Any values near these would use the same style.

The resistor sizes for the 12.5K and 1K resistors are for best power balance and also best matching  tracking TCR.

One other note, the standard spacing for measurements of the resistor is 0.375" from the resistor body, this can be changed at the request of the customer if longer spacing is needed.

 I just received a new pricing list for some of the wire I use, the increase was a jaw dropper.  Thank you commodity materials and speculators, I hope that sink hole under your house is big enough when it opens.
« Last Edit: June 18, 2014, 12:18:51 am by Edwin G. Pettis »
 

Offline CaptnYellowShirt

  • Frequent Contributor
  • **
  • Posts: 447
  • Country: us
  • Scooty Puff Jr.
Re: T.C. measurements on precision resistors
« Reply #27 on: June 18, 2014, 02:22:30 am »
Subscribed...
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #28 on: June 18, 2014, 07:33:26 am »
Quote
I am also not shure if it is a good idea to work with (possibly hand picked) samples which are not easily avaliable from stock of common suppliers.
How can I be shure that this would not give a wrong picture? And how can other hobbyists get those resistors?

I appreciate that, not because to blame Edwin G. Pettis, but as expected there is somewhat emotion in the posts which make them sound less objective and objective results on resistors are what Andreas want to present.

I guess if we had some guy from Vishay in here there would be a flame war and fairy tale exchange. There is still some needless myth around the welding process of wirewound resistors and the CTE / TC compensation of metal foil resistors. Resistor manufactor act as if they had understood all the effects, but I bet they don't and that they are only on the right way of understanding it. But it would be a shame to agree such a statement.
Instead some marketing guy creates stories, foil resistor manufactor fight against wire wound resistor manufactor and vice versa, but both of you agree that thin film, thick film and carbon resistors are bad. But hey, I understand that, you guys want to sell resistors. You don't want to offer and share your knowledge, you don't want to tell us the disadvantages of your resistor technology and I'm sure you can tell some, because there is always some imperfection and parasitics is at least only one of them.

So Andreas, keep going on in your fully independant research, even if the test environment and the test conditions are not perfect. I don't know of any independant and objective research on the topic yet.
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #29 on: June 18, 2014, 05:21:07 pm »
Gentlemen,

If I sound like I'm attacking anyone here personally, it was not my intention and I apologize if I offended anyone, my intention here is to draw everyone's attention to the facts (not personal opinion) of both resistors and measurement techniques.  If you will carefully read my posts on page 2, I do talk about characteristics and parasitic comparisons between PWW and foil in both general and specific terms.  These 'facts' have been verified by independent engineers and laboratories, not just by me.  Have I personally examined every possible PWW manufacturer's components on the planet, of course not, however the resistors of nearly all major resistor manufacturers have been tested and examined in detail by my colleague (who worked for many of them over the years), myself or the laboratories of the top aerospace, military and precision instrumentation suppliers which has included some EU facilities as well.

When measurements involve PPM and sub-PPM accuracy, no matter who is doing it, it requires the utmost care to prevent the many possible sources of error from sneaking in, it has happened to the best of us, measuring nanovolts is no easy task, even for a standards laboratory let alone smaller, less well equipped facilities and even though I will encourage it quite heartily for the experimenter as a learning tool, the best of intentions and efforts may not produce accurate results, it may produce apparent consistent results but not accurate results.

Just because a DMM offers 6, 7, or 8 1/2 digits of resolution, that does not mean all of the digits are accurate, in fact, even in the short term, they are not.  The same facts applies to ADCs and DACs, just because there are 24 digits of resolution doesn't mean there are 24 digits of accuracy.  Sorry if I sound preachy here, I am just trying to draw attention to just some of the error sources present in the measurements being attempted here.  Careful examination of the specifications will show that there are several sources of internal errors which can and do add up to significantly higher figures than the resolution of the unit.  In the case of a DMM, usually the DCV is the most accurate function and is limited to 6 or 6 1/2 digits of accuracy, the other functions have even less accuracy.  The ratio function does tend to have terrific linearity, the top DMMs can possess very high linear specs of around 0.1 PPM but that does not guarantee accuracy of the measurement in itself.  Many of the sources of error are external to the DMM and are very 'sneaky', thermal EMFs to name only one of them.

The ADC Andreas is using is an excellent ADC, I've used it and some of its 'brothers' but while it has a resolution of 24 bits, it accuracy is closer to 20 bits (1 PPM) just like other 24 bit ADCs,  resolution cannot be confused with accuracy, they are two distinct, different entities.  Math cannot wholly compensate for error, statistics have limitations (proven every day) and frankly, I do not rely on math to iron out inconsistencies, at best you might be able to gain one more digit of accuracy, probably closer to a half of a digit, the rest of it is all noise and error sources being averaged out.  Unless the readings can be compared to a known accurate source WITH accuracy, then the readings cannot be considered valid, no matter how much they may repeat.

In a calibration lab, the rule is that a minimum of 5 times the accuracy of the unit being measured (under controlled conditions) and 10 times the accuracy is often required, particularly for very accurate measurements.  This includes the uncertainty of the measurement and all of the instruments and standards have uncertainties in their measurements which must be taken into consideration.  That is precisely why your SR-104 resistor has an uncertainty specified in its measurement and is part of the accuracy of the SR-104.  For an accuracy of say 1 PPM, the comparing standard and instrumentation must have a minimum of 0.2 PPM total accuracy and uncertainty verified by a primary lab which is usually something comparable to our NIST.

I am not attacking Andreas' test setup per se, considering cost it is a pretty decent setup, what I was trying to point out was that in my opinion the setup is not capable of producing the results Andreas was hoping for with the accuracy needed.  There was absolutely nothing personal in my comments.  I have many years of experience in calibration (my first position out of college) as well so I know what I am talking about.  One of my jobs, among others, was to design and build test equipment capable of measuring such quantities and it is not easy to eliminate or account for all of the error sources that can creep into such measurements.

I stand by all of my statements concerning the flaws in resistors, both PWWs and foils, there has been no evidence presented to the contrary by any of the other resistor manufacturers, nothing more than airy claims, "our resistors are welded", no they are not and the evidence proves they are not welded, both in physical examination and in actual performance data.  Until proven otherwise, my claim that my resistors are the only welded PWWs available (and the performance data backs that up) stands.  It is very easy to prove welding on my resistors.  There are manufacturing details about my resistors which I am not willing to publicize as those details are proprietary.  The PWW industry has a long history of 'borrowing' manufacturing ideas from one another, even manufacturers from the EU had done it and I suspect the Asians have also done it.  The evidence is right there in their resistor construction of the wide spread 'borrowing' of each others bobbin designs and that is further proof that they all share in the same flawed designs.

If you have any questions about resistors of either kind, I will be happy to answer them, I do try quite hard to keep my statements factual and if I sound like I'm beating my own drum, at least I have evidence to support it.

Andreas, I would be most willing to give suggestions about your test setup if you would like, you are perfectly free to use them or not.  I do encourage you to read my comments on the TCRs concerning the LTZ1000/A.  The Z hermetic foil resistors are quite unnecessary and will only result in excessive cost.  This has nothing to do with the Z's performance, claimed or otherwise, it is a question of where to apply the correct specifications.

I am still willing to send you some samples to try out if you will tell me a range of resistance that you can use, if I have any stock in that range I will send them and I guarantee that there will be no 'hand picking' involved, just randomly picked, standard line resistors with no special treatment that meets my specifications as given in the data sheet I posted earlier. I will even post the resistor values here as well.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #30 on: June 18, 2014, 11:01:21 pm »
Hello,

although I wanted to wait with responses until I have my first results ready I think its time to read some of the posts.

For the first: it is good to have someone with so long experience on this area. So we can all learn from your input.
And thanks for the links to your articles. You saved me the time for finding them.

There is no 'sweet spot' in resistors,

Is this valid for wire wound resistors only or also for metal foil?
I can imagine that the WW resistors if the base material always has a (more or less) positive TC will have no sweet spot except for divider ratios.

On metal foil vishay claims the expansion coefficient of the base material and the foil TC cancel out.
Of course the quality of matching could have something to do with the foil thicknes (resistance value).
The thicker the foil (low resistance values) the more positive the resulting TC could be.
The 10K metal foils of Dr. Frank are between -0.3 ... -1 ppm. My first measured 1K is around 0.9 ppm at 25 degrees.
So what is in between? Of cause not all metal foil alloys will have the same TC so the values might vary from batch to batch.
But these are only my speculations.

One more thing, the 'lumpy' TCR curve of foil resistors is inherent in the design, you should have seen the original TCR curves 30 or 40 years ago, it would make you seasick. 

would be really interesting do you have such a curve?
Up to now I found only one in the www (see attachment).

You didn't seem to have any problem with testing their sub-standard parts so why are you questioning mine?

I also have some reservations about your ADC test setup as well, from your statements, it sounds like you have to use a lot of math on those 'measurements' due to various influences such as noise, I am very leery of measurement results that use statistics to iron out garbage.  It calls into question the results accuracy.

I am not trying to throw cold water on these projects, on the contrary I think they are quite commendable and a valuable teacher but I still question a hobbyist's need for such component performance.

If you want to know more about my resistors you can ask here or I can tell you how to contact me more directly.

I hope you do not take it personally. In the mean time I have understood that your resistors could also be a serious alternative to other sources even for hobbyists.

And yes you are right. I have to live with the inaccuricies of my setup. ADC noise is one of them.
The measurements are meant only on a comparison basis. So in this case the repeatability and stability is of most concern.
I do not claim absolute accuracies. And the measurements are only on few samples and should not be generalized.
If someone other has better equipment we all would appreciate further comparison results.

By the way: the largest problem that I see at the moment is not the ADC but the temperature measurement.
With 3 ppm/K of the resistor and 1 K temperature difference between sensor and resistor the error is already 3ppm.
And I have at the moment large temperature differences between the 3 sensors of my UPW50 measurement.
So at the moment I have to work at the mechanical setup.
I know I should use a temperature controlled stirring oil bath. But I want to use the resistors afterwards for a reference.

To the hobbyist needs: You have to understand that its a little bit of "yes I can" and "mine is better than yours".
If you are building only 1 or 2 references they should be the "golden selection".

It would be a good idea if one of the resistor manufacturers would sell complete sets for e.g. LTZ1000(A) references.
If I would specify such a set:

For the 120 Ohms and 70K the absolute tolerance could be in the 1%(or even 5%) range.
Only long term stability and T.C. are of concern.
Standard T.C. would be 3ppm/K.  The golden selection set would have perhaps 1ppm/K.

For the temperature setpoint we would need a matched ratio set. 12K / 1K for LTZ1000 and 12,5K / 1K for LTZ1000A
If I could wish me the perfect resistor it would be made from the same wire (same lot) on the same bobbin within the same case if this is possible.
Ratio tolerance I would specify no more than 2 mV error (1 degree setpoint) since otherwise I would have to put the setpoint unnecesarily higher. 2mV / 600mV = 0.3% maximum ratio error. I hope that the T.C. mismatch is also below 0.3 ppm / K. (factor 10 improvement due to matching from same wire).
Do you agree with the specs?

The large question is:
So what would 2 sets with standard T.C.  3ppm/K or "golden selection" T.C. (1ppm/K) cost including shipping worldwide?

The other questions: do you have pictures from your resistors (sizes , forms). Is it more the 0816 axial style or radial style. I am a bit confused from your descriptions.

So long for today. (its already late here).

With best regards

Andreas

Edit: just to illustrate what I mean I have attached a curve with NTC temperature differences over temperature of the 2 NTCs attached to the UPW50. One near bottom and one near top. There is no concern that they differ absolute and over temperature. The problem is that there is a difference between cooling down and heating up of nearly up to 1K at the same temperature. So which one is the temperature of the resistor? NTC1 NTC2 or none? If I do not meet the exact temperature of the resistor I will see a hysteresis on T.C. curve which is not from the resistor but from the temperature displacement.
« Last Edit: June 19, 2014, 06:23:05 am by Andreas »
 

Offline quarks

  • Frequent Contributor
  • **
  • Posts: 830
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #31 on: June 19, 2014, 10:34:49 am »

The foregoing is standard calibration procedure, you may want to check a resistor's characteristics at something higher than minimum power levels. 
...
The 801 detector is about average in performance, you can substitute a better null detector such as a Fluke 845A...
...
The Fluke 8508A is an excellent DMM, in many instances I have been able to calibrate Flukes to much better accuracy than the specifications...
...
The ESI 240C and RS925A forms the most important parts of a Kelvin bridge and it is there that your accuracy in resistance readings resides.  The power supply is not critical but a good analog null detector is, you are 85% of the way to where you need to be with the two Kelvin bridge components you already have, in conjunction with your Fluke for the higher resistance readings, you are very well set for good accurate readings.

Hello Edwin,

thank you for your reply and your hints.

Because I do have all gear for a "242 like setup" (inkl. Fluke Null Detector 845), the Fluke 8508A was only meant to compare to the "ESI 242 system" results. 

Can you share / describe how you setup your own ESI 242 system to do your TCR measurements?

bye
quarks
« Last Edit: June 19, 2014, 10:36:40 am by quarks »
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1753
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #32 on: June 19, 2014, 10:54:23 am »
Hello Andreas,

first of all, I really appreciate your T.C. measurements.

Those ppm/K measurements are really very delicate.
Although you are using relatively crude measurement setup  (i.e. this nominal 24bit A/D), you can clearly demonstrate the T.C. over T curve of the Z201.

Before you continue your measurements, I'd like to propose some simple improvements in physical aspects..

In your R(T) curve, a hysteresis is visible.

As the temperature range is relatively small, i.e. below +/- 20°C, I assume that this effect originates mainly from the bad thermal coupling between DUT and temperature sensor.
You should improve your setup to distinguish such an effect from the known thermal hysteresis of the metal foil technology.

In such physical temperature experiments, an 'isothermal block' usually is used.

This is realized by a copper block with drill holes, inside which the DUT and the temperature sensor are mounted, both assembled with thermal conducting grease.
The copper serves two purposes, first it lags outer temperature changes by its heat capacity, and 2nd, it  tightly thermally couples DUT and sensor by its high heat conductivity.
Then you are left only with the heat transfer resistance inside the resistor, in my case, the oil inside the VHP package.

To summarize, your current setup suffers greatly from too high heat transfer resistivity, and you should simply add an isothermal aluminium block.
 
 
Then, instead of a forced gas flow, it's better to use temperature reservoirs, as for example cryogenic liquids  (He4, N2) at the bottom, below the isothermal block, i.e. no direct contact, but using the corresponding contact gas to slowly cool (or heat) the isothermal block. That mitigates hysteresis  effects from thermal imbalances greatly.

For the measurement on my VHP202Z, I have used an aluminium block (16x16x30 mm³), see photos.
The sensor (black leads) is a precision epcos NTC.
The outer aluminium box serves as an additional thermal shield, to further minimize air draught and thermal imbalances.

A cooling pad (-18°C, out of the freezer), or a container with hot water (90°C) inside a cool box served as the heat sink / source, and the natural air convection served as the thermal contact to the experiment.

The slower the experiment, the lower the possible temperature differences between resistive element and the temperature sensor, and the smaller possible hysteresis effects from temperature differences inside the experimental setup.

I'll show raw measurement data later.

Frank
« Last Edit: June 19, 2014, 01:41:34 pm by Dr. Frank »
 
The following users thanked this post: razvan784

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1753
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #33 on: June 19, 2014, 11:25:43 am »
Hello Edwin,

I also appreciate your interesting narrations about wirewound resistor technologies, and fully share your opinion about the Vishay metal foil technology.

I've encountered the same experience in my experiments on the LTZ1000 reference, and on my 5 VHP202Z reference resistors.

I disagree in one aspect: The Vishay hermetically sealed, oil filled resistors really are longterm stable, as Vishay claims, i.e << 1ppm/year.
My 4 year long monitoring of agroup of 5 VHP resistors show a drift of less than 1 ppm.

In contrast to that, I did not find (yet) wirewound resistors, which are promoted to have a similar good long term stability.

Only the ESI SR104 assembly is specified for a stability in a similar order of magnitude.

I also appreciate your  flaming appraisal  ;) of the wirewound resistor technology, as I agree, that this is mostly on par with the metal foil technology, but lacking hysteresis effects.

And I own a nice collection of old ww resistors and - sets, from Fluke, HP, Cohu, KINTEL, Alma, Burster,.. really nice handmade constructions.

Frank
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 815
  • Country: de
  • pushing silicon at work
Re: T.C. measurements on precision resistors
« Reply #34 on: June 19, 2014, 12:40:28 pm »

For the measurement on my VHP202Z, I have used an aluminium block (16x16x30 mm³), see photos.
The sensor (black leads) is a precision epcos NTC.
The outer aluminium box serves as an additional thermal shield, to further minimize air draught and thermal imbalances.


I know this little bugger!  :P

Frank, I got the cheapest available SR1010 from e-gay, cleaned it thouroughly and gave it the missing case screws and even cloned the optional  shorting bars (thanks SAH-Präzision for the great work), that are 12 pieces of very old 10 Ohm WW resistors... want to take a look ? :)
I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #35 on: June 19, 2014, 03:18:18 pm »

Those ppm/K measurements are really very delicate.

To summarize, your current setup suffers greatly from too high heat transfer resistivity, and you should simply add an isothermal aluminium block.
 
 For the measurement on my VHP202Z, I have used an aluminium block (16x16x30 mm³), see photos.
The outer aluminium box serves as an additional thermal shield, to further minimize air draught and thermal imbalances.


Hello Frank,
 
you clearly point out the largest error in my measurement setup.
And much thanks for your photos of your measurement setup.
This will help a lot, I will see how I can come close to that.
This is the best input on my problem up to now.

But I will not handle with cyrogenic liquids.
On the other side the thermal block will increase the effort for measuring different sizes of resistors.

I already asked myself how you did your 0.3 ppm/K measurement.
I hope you will deliver some details soon:
- temperature range
- temperature gradient

It could be possible that the bond of the metal foil to the ceramic substrate has a time constant regarding hysteresis.
So I think the ramp speed could also have a effect on hysteresis amount even if the temperature sensor is mounted correct.
I have seen such a effect on the LT1236AILS8-5 reference. But on the other side it could be also the temperature sensor mounting.

with best regards

Andreas
 

Offline ManateeMafia

  • Frequent Contributor
  • **
  • Posts: 719
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #36 on: June 19, 2014, 05:00:26 pm »
Hello Frank,

Would you consider a combination of technologies too much of an issue with measurement accuracy? For example, if Andreas used frozen blocks similar to this :

http://www.thermosafe.com/model/FPP56?

and used it with his thermoelectric cooler to slow down the process, perhaps disabling the TEC device before measurements. This would mitigate the need for any potentially hazardous substance in the home. The packs are flat and retain shape as they cool. I was given one of these the other day and thought it might come in handy for a similar test.

Thanks,

ManateeMafia
 

Online Conrad Hoffman

  • Super Contributor
  • ***
  • Posts: 1156
  • Country: us
    • The Messy Basement
Re: T.C. measurements on precision resistors
« Reply #37 on: June 19, 2014, 05:27:20 pm »
Edwin,

I'm curious what you think of the various Julie Research resistors, the bobbin wound units in plastic, the similar ones used in their Kelvin Varley dividers, and the ones sealed in oil? I've also got a collection of old L&N standard resistors, that I assume are manganin. They don't have a good reputation, but seem OK if the temperature can be controlled. Finally, I've got a roll of H.P. Reid (now Kanthal) Rediohm 800 wire. It seems quite good, but can't be soldered. Are there any welding tips you can give me?
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #38 on: June 19, 2014, 07:47:30 pm »
Wow, lots of questions, it may take me awhile to respond to all of them but I will get to them.

To Andreas (post #33), previous mis-understandings we will just chalk it up to mis-communication and forget about them.  Thank you for the compliments and I certainly do encourage your efforts, it is a very good learning experience.  Now for some of your questions......

As to the 'sweet spots', all resistors, no matter what type, yields both plus and minus TCRs, if a particular resistor specification only indicates a negative or positive TCR, the resistors are being selected, nothing wrong with that per se.  If such TCR specifications leads to customer misconceptions, then the manufacturer is guilty of a cover up as such.  All resistor specifications should have a TCR spec such as 0 +/- 1 PPM/°C for example, the distribution of TCRs always have both polarities.

A customer can specify a particular TCR of either polarity if needed, that does reduce the yield of that TCR from the batch and increases the price due to additional labor and time to 'pick out' a given TCR.  Also note that the customer has to accept a given tolerance for that particular TCR, asking for an absolute TCR is really going to cost.  Most designers would like a zero TCR, which in a given batch of resistors, there may be some resistors that come very close to that but yields are likely small and the price is inversely related to yield, i.e. smaller yield, bigger price.

In the various threads concerning the voltage divider required for the LTZ1000/A (actually two dividers would help....can anybody guess where the other divider goes?), the hunt for the 'perfect' resistor is very consistent.  Along with what I've said earlier about the voltage divider, another technique which is potentially less expensive is to try and match opposite polarity TCRs reasonably closely which will actually cancel each other out to possibly a high degree.  This approach could cost somewhat less than trying to find resistors (such as the Z foil) with near zero TCRs that match.  With this method, any two TCRs of opposite polarities will cancel each other out and the chances of finding those kinds of pairs in any given batch is substantially higher, any two resistors within a standard TCR specification could be used, even 3 PPM/°C TCRs as long as they were opposite in polarity.  Of course, matching two TCRs of the same polarity can work just as well as the opposite polarity match can if you have the right resistors.

I think I am talking myself out of some nice fat resistor charges...but the customer should be given the best possible solution to their needs, the supplier should not just think of themselves.

Vishay has been working on the physical problems of the film/foil resistors since they were invented   over 50 years ago, they actually started out as strain gauges, which they are still used as today, the 'precision' resistor part of it came later, Felix Zandman claimed the first patent on the precision film resistor (the patent(s) were later broken in court by the French company Sfernice) I believe around 1962.  The early TCRs were hyperbolic (like the one you posted, which by the way is a much more recent resistor) with the peak of the curve sitting at 25°C and the curve going in a negative direction on either side, some of their resistor still exhibit the same curve albeit with somewhat lower TCR/°C values.  As they worked on 'ironing out' this pesky curve they managed to find materials and binders which reduced the physically caused 'lumps' in the resulting curve to a 'wavy' sine wave like TCR, the early versions (both types of curves) had TCRs generally higher than 10 PPM/°C ranging up to as high as 50 PPM/°C.  For the lower TCRs that customers demanded, Vishay 'cherry picked' (as we call it over here) lower TCRs out of the batches, with this technique, they could supply a limited amount of TCRs under 5 PPM/°C.

As the years went Vishay continued to slowly improve the TCRs of their best resistors and the alloy suppliers were able to improve their TCR results as well, benefiting both wire and foil resistors.  The 'Z' foil line is the result of many years of work to improve the foil resistor, if you look carefully at the 'Z' foil TCR's box in their spec sheet, you will see the same 'wavy sine wave' TCR curve as the earlier resistors had, only they have managed to reduce it to a very small sub-PPM value within a limited temperature range, it is still not flat by definition.  The shape of the curves have remained pretty constant over the years only the TCR values have decreased.

I should interject here that while the earlier precision wire wound (PWW) resistors had flaws as I related in my EDN article, their TCR curves tended to be flatter than the Vishay resistors over most of their temperature operating range, approaching a much more linear curve, a fact which Vishay did their best in advertising to discredit along with presenting their resistors in the best possible (albeit somewhat distorted) light.  The flaws in the early PWW resistors did not help their case but for some curious reason, the PWW side of the industry did not fight back in the advertising arena
and allowed Vishay's effective attacks to further discredit the PWWs.  PWWs have been effectively written out of many projects due to that ad campaign even to this day.
 
I believe someone, maybe it was Andreas, asked about my TCR curve, well technically, it isn't a curve, it is a flat line across its temperature range with small changes in the TCR at the extreme temperature limits (-55°C to +125°C).  No curve balls, no spit balls, no screw balls, maybe just a fast ball.

Now, one of the major (among several) problems in measuring minute quantities is controlling temperature and (very importantly) controlling air flow around the D.U.T.  In a calibration lab, the air temperature is usually controlled as closely to 25°C as possible, but we aren't in a calibration lab so we improvise.  As Andreas mentioned, he is not trying to measure absolute quantities, which would be very difficult in his circumstances.  So we work with less absolute accuracy but stable (over the necessary period of the test) and repeatable results.  This can give a good indication of a measurement that the system is not directly capable of.  If the measurements are stable and repeatable and some other sources of error are observed, reasonably accurate results can be expected.  As we say, the devil is in the details.  TCR measurements have a lot of devil in the details.

A small chamber made out of insulation or Styrofoam big enough to hold your thermal cooler/heater assembly and the D.U.T. with the holes for wiring, temperature sensor, ect., plugged up will work nicely and the chamber is comparatively cheap to make, metal isn't as good since it doesn't insulate very well.  Unfortunately you don't want to put the ADC circuitry in there since you need its temperature to remain stable throughout a test run.  Put a Styrofoam cup over the ADC circuitry to keep air currents from messing with it.  While it would be nice to be able to run a TCR sweep over a 100°C range, it is not practical with a home brew very easily.  A 50°C range from hot to cold is sufficient to give a decent indication of average TCR  (remember Z foil curves are not linear), it will actually be a better TCR than is actual over the range because of the shape of the curve being nonlinear but it is impractical to have the needed equipment on an experimenter's budget, that takes a standards lab grade to get the job done.

A fan inside the D.U.T. container is okay as long as it is sealed, if it has any significant leaks, it will cause temperature errors.  I would mount the fan near the 'cold' plate to circulate the air around.  TCR measurements at resistor houses are usually done in large temperature chambers using CO2 tanks and heaters, there are table top chambers but they are expensive, even the used ones.

The next bug-a-boo is thermal EMFs in your connections to the measurement circuits and D.U.T.  This can be difficult to fix, as a microvolt of EMF is very significant, make sure the solder joints are good and the number of connections are kept to a minimum.  The easiest way to minimize thermals is to keep everything at the same temperature inside the containers.

Noise is another source of error and often is very difficult to get rid of it, it is bet to get rid of it at the source, but is often no possible.  In Andreas' case, it can be tricky filtering out noise at the input, partially due to the input impedance of the ADC which tends to vary with input.  Filtering with active circuits can be effective but can also introduce some errors of their own, offsets and their own additional noise.  To some degree, if you can quantify how much noise you have and it is relatively stable, it can be subtracted out of the measurement with fair accuracy.  However, the relative stability of many noise sources tends to drift with time and attempting to use long term averaging techniques will not result in better accuracy.

I think the majority of your measurement problems is controlling the air currents followed by noise and the thermal EMFs, try the setup I suggested, it may surprise you and the best part of all, it is relatively cheap!  Sorry, I haven't had the time to look closely at any of the photos of test setups as yet, I'll try to catch those as well.

I will answer your questions about the resistors for the LTZ1000/A circuits in another post, sorry the turn around time here is so long.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #39 on: June 20, 2014, 01:29:02 am »
Andreas,

I took a good look at the LTZ1000/A data sheet and curiously, many of their data sheets will specify resistor characteristics, even down to TCRs, this one is not the case.  Aside from the voltage divider resistors and the 120 ohm resistor whose TCRs are easier to determine, the other resistors, depending on just what circuit you are using (such as the positive (or negative) reference in the application section needs further examination.  Some of them have no significant effect on performance while others can.  I will post when I have completed my examination.  Quite a few variations are possible to enhance performance.....interesting.
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #40 on: June 20, 2014, 01:55:20 am »
To quarks,

(reply #34)  Since something along the lines of a full scale temperature chamber is not at my disposal at this time, I designed and built a small chamber out of high temperature insulation using a good glue that with standards higher temperatures than I intend on using.  I made a double thickness door out of two sheets of the insulation glued together, one that fits inside the opening and the other that fits across the entire side of the chamber.  At the top, I installed a fairly large TEC with a fan mounted to circulate air within the chamber, an accurate temperature sensor (platinum RTD) is connected to a custom designed temperature controller (by me) which should keep the temperature within less than a 0.25°C variance within the chamber and control the nominal temperature setting within 0.01°C.

The circuit board material is at least an FR4 grade or better, solder joints are kept to a minimum and clips are used to connect the resistor in four terminal mode through a bunch of teflon coated wiring through a small hole in the side to a switching console which, in-turn is connected to my ESI 242D.

I usually try for at least a 50°C temperature range or better between the hot and cold temperatures that the TEC is capable of.

That's it in a nut shell, if you are curious about any details, please ask.  Sorry, a bit short on time at the moment.

 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1753
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #41 on: June 20, 2014, 03:41:58 am »



But I will not handle with cyrogenic liquids.
On the other side the thermal block will increase the effort for measuring different sizes of resistors.

I already asked myself how you did your 0.3 ppm/K measurement.
I hope you will deliver some details soon:
- temperature range
- temperature gradient

It could be possible that the bond of the metal foil to the ceramic substrate has a time constant regarding hysteresis.
So I think the ramp speed could also have a effect on hysteresis amount even if the temperature sensor is mounted correct.
I have seen such a effect on the LT1236AILS8-5 reference. But on the other side it could be also the temperature sensor mounting.

with best regards

Andreas

Hello Andreas,

the story about liquid Helium or Nitrogen was an example only.
At the physics institute, I used to measure that way, and also calibrated our thermometers in the temperature range of  2K ... 300K.

 
For this purpose here, simply use a  cooling pad, or even your heater instead.

And I really meant that using an aluminium block is a simple thing..

Aluminium bars, 1m long and 16x16 or 25x25 are available in your next DIY store (Baumarkt).

Saw off a piece of appropriate length, and drill several holes completely through that block, so that you may attach different resistors in that block.

For tubular resistors and the NTC it's very easy, add 1/10..2/10 mm for the diameter.

For rectangular resistors, you have to drill and file a slot, well right, that's a little bit more effort.

Build one additional block for your reference resistor.


I use an HP3458A in 4W Ohm mode, with offset compensation and 0.1ppm resolution.
Consecutive readings on the same resistor are also stable to <= +/- 0.2ppm, at 100 NPLC (4sec measuring time).
 
Measuring my 5 VHP202Z one  after each other several times within 10..20 minutes,  I always get repeatable readings  which differ not more than 0.2ppm for the same resistor.

The ambient and internal temperature is stable to <= 0.1°C in that time period.

Therefore, the transfer stability is about 0.2ppm.

Even for much longer times, keeping the temperature stable, gives stabilities in that order of magnitude.


The T.C. measurement on the V334 took about 8 hours, see diagrams.
The vertical graticule is 0.2ppm wide.

Each measurement point is simply read once from the instrument, without further averaging than the NPLC 100 measuring time. I also apply a calibration factor for the HP3458A (derived from the group of these 5 VHP202Z), but that's secondary for the T.C.

There is a jitter from point to point about 0.2ppm wide.
Those Ohm measurements are relatively sensitive, and the DUT was connected with four 1m long, unshielded cables only.
I simply let the linear regression function average out all this jitter for determination of the T.C.

In the beginning, heating was done too quickly, so the curve deviates from the average line by a bigger amount. Therefore, about the first 100 points were skipped for the T.C. calculation.

For the rest of the temperature cycle, over 6h, the measurements follow nicely the linear regression line, without a noteworthy hysteresis.
If you compare the starting measurements around 25°C (#185) to those at the same temperature after 6h (#378), they agree within +/- 0.2ppm.
That's quite stable, I think.
The internal temperature changed not more than +/-0.2°C.

There are several glitches visible in the measurement, caused by myself, when I moved the heat reservoirs inside and outside of the cooler box.


This resistor itself obviously shows no hysteresis.

If you would repeat that measurement with a much faster temperature gradient, you might see a hysteresis loop, which would be caused by the temperature lag between DUT and T-sensor only.
By the shape of the hysteresis loop, you could judge, how good your temperature coupling is.

Frank
« Last Edit: June 20, 2014, 05:34:38 am by Dr. Frank »
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 815
  • Country: de
  • pushing silicon at work
Re: T.C. measurements on precision resistors
« Reply #42 on: June 20, 2014, 06:22:59 am »
@Andreas

Would you accept if I donate a metal slab (might be brass or whichever fitting slab I find) with drills (say 2 mm interval) and a tube of thermal grease? My own observations with medical thermistor catheters shows a big difference in thermal coupling between putting reference sensor, heater and DUT only in close proximity inside a small space surrounded thermal insulator to a drilled metal block surrounded by isolation.

BR

Hendrik


I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4095
  • Country: us
  • xDevs.com/live - 24/7 lab feed
    • xDevs.com
Re: T.C. measurements on precision resistors
« Reply #43 on: June 20, 2014, 11:41:08 am »
I can play with liquid nitrogen :)
Just got some foils today, and will build a test setup for TC and resistance measurements using TEC module.
Hope TEC will give me enough stability.

Ill post in my KX thread or create article on my site after all done.
don't want to hijack this nice thread.  :-DMM
YouTube | Chat room | Live-cam | Have documentation to share? Upload here! No size limit, firmware dumps, photos.
 

Offline Galaxyrise

  • Frequent Contributor
  • **
  • Posts: 526
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #44 on: June 20, 2014, 05:30:24 pm »
In the various threads concerning the voltage divider required for the LTZ1000/A (actually two dividers would help....can anybody guess where the other divider goes?)

First: Welcome to the forums, sir! I very much appreciate having you here.

Now to your "riddle": Are you referring to the technique (I first saw explained by Dr Frank  https://www.eevblog.com/forum/projects/ultra-precision-reference-ltz1000/msg240405/#msg240405) of using one divider to trim the zener output to exactly 7, then using 10 "identical" resistors to implement the gain divider to get to 10V?
I am but an egg
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #45 on: June 20, 2014, 09:42:40 pm »
Quote
Would you accept if I donate a metal slab (might be brass or whichever fitting slab I find) with drills (say 2 mm interval) and a tube of thermal grease?

I had the same idea and have send some brass to Andreas today. Hopefully the mail is fast enough to arrive tomorrow.
« Last Edit: June 21, 2014, 07:36:20 am by branadic »
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #46 on: June 21, 2014, 02:02:37 am »
To Dr. Frank (reply #36),

I was almost done with this post, this morning, had to stop and this afternoon, Chrome decided it needed to crash...argh!

First, thank you very much for the compliments.  now if I just remember what I'd written before.....

I should probably clarify some of my statements on stability, I work mostly with what would normally be called 'commodity' resistors, even if I don't consider mine in such a category.  So sometimes I kind of forget the boundary, as such, between resistor standards and regular resistors, even though, in my case, these is quite a bit of cross over between them.

Generally, stability is defined as the lack of change in a resistance with time and/or use, the less the change, the happier everyone is.  Stability is specified under two conditions, one for standards, one for other resistor types.  For a primary standard, such as the SR-104, stability is defined in terms of time, under the standard definition of use of a standard, i.e. very low power, constant temperature, no physical shock and kid-glove care.  This is very similar to the drift specification on a shelf non-powered of non-standard class resistors.  This drift can be anywhere from PPM to hundreds of PPM per year, it all depends on the type of resistor and its manufacturing processes.  The second stability test, which isn't applied to standards, is the powered long term drift specification.  Depending on the resistor type this can be anything from <10 PPM / year to hundreds of PPM / year.  This powered stability test can eliminate the boys from the men as such.

The real heavy duty stability test involves thermal shock MIL-STD-202 which has been around since the dinosaur age, at a minimum of 5 cycles between -55°C to +125°C and then checked for drift and failures.  To my knowledge, no Vishay Z foil resistors have come out of this test smelling like roses.  On the other hand, mine have gone through this wringer with no more than 12 PPM change and an average of only 8 PPM change plus the real show stopper, no failures which no other resistor has achieved.  One laboratory ran my samples through 50 cycles of thermal shock (why, I don't know), the stability shifts were still within the same specs and still no failures.  Frankly, I challenge any resistor manufacturer to put their resistors through that test and come out anywhere near as good.

I do know that the shelf and power stability specs on my resistors are nearly identical, we specify shelf as 0 +/- 5 PPM/year and full power @ 125°C 0 +/- 10 PPM/year but I've never seen one of the resistors come close to those limits.  To be honest, I do not have any specific data on the measured drift of these resistors, no past customer has ever come back and complained, it seems self-evident by the thermal shock tests, that these resistors have very good inherent stability and reliability designed in.

As to sub-ppm figures in TCR, I know that a significant percentage of any given batch TCR is within 0 +/- 1 PPM/°C with no special treatment.  Just how many of them are close to zero TCR I do not know and that would take some time to find out.  Tain't easy to measure folks!

Some of the old resistors did have some craftsmanship in them and even a bit of art to them.  The only problem with the old resistors are that they have probably drifted out of the initial spec some time ago which may not be an issue.  Unless they have opened they likely have some good like left in them.  The earliest resistors were pretty crude dating back to the 1910s and 1920s when mass production became more common, more 'automated' ways of making resistors were thought of, partly because even then, miniaturization was beginning to knock on the door.  The 1930s brought in ceramic bobbins of various designs for both regular (not really precision yet) and powers, these bobbins stayed with us, mostly unchanged until the era of plastics arrived in the 1950s and miniaturization was knocking even harder at the door, bringing troubles with it (ask Murphy).

I have a small collection of various bobbin resistors including some early ceramic precisions which were about half an inch in diameter and at least an inch long, some are encapsulated in bakelite, with a mounting hole in the middle.  Then there are some smaller ceramic precisions a little bit smaller than a half inch in diameter and only about 5/8th of an inch long but very similar construction.

The construction of the 'flat' mica bobbin resistor was originated by General Radio back in the 1940s, maybe a little earlier and that design has stuck with the standard resistor boys ever since.
« Last Edit: June 21, 2014, 04:36:27 am by Edwin G. Pettis »
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #47 on: June 21, 2014, 04:36:30 pm »
Hello Andreas,  (reply #38)

I don't know if any of you have seen the insides of a foil resistor so I've attached one, sorry it is a little fuzzy, the original lacked definition.  The main thermal resistivity is from the inside to the outside world, although the foil is bonded to the ceramic substrate with a fairly low thermal resistivity, the resistivity from the ceramic to the outside through the molded case is substantially higher being made from plastic.

Obviously, the rubber pad protects the 'strain gauge' from outside stress, without this rubber pad (which also has very low thermal conductivity), the foil's resistance value would vary quite a bit.  Putting this 'strain gauge' inside a hermetic can virtually isolates it from stress except from temperature effects, both from self-heating and ambient.

The construction of the Z series resistors are even more complex inside to further isolate the 'strain gauge' from stress.  They have been quite successful after working on the problem for about 55 years.  Granted, it took my colleague and myself a bit less than 25 years to solve the welding problem and other improvements to the stagnated PWW.
« Last Edit: June 21, 2014, 04:38:44 pm by Edwin G. Pettis »
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 417
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #48 on: June 21, 2014, 05:17:58 pm »
To Conrad,  (reply #41)

The Julie resistor design varied over time since his original patent (which really wasn't all that good of a resistor but good for its time), he did fiddle with the resistor design over time and tweaked some characteristics into better performance.  My colleague, Mike Chesselet, had more experience with Julie's resistors directly than I have, Mike was not too impressed with the designs, on the whole, partly because Loebe was fussy about details, they typically outperformed most of the wire wounds on the common market.  I would say they did not compare with the military grade resistors we were making at Ultronix during the 1970s but close.

One market Julie went into which the other commodity makers didn't go into was standards, mainly working/transfer types but Julie also tried to outdo the SR-104 and Thomas one ohm standards which he failed at.  At one time he even claimed that his laboratory was more accurate than NIST which made him the butt of many jokes.  In the case of his decade boxes and other transfer standards, they performed quite well but there were quality control issues which at times, bit him in the butt.  Most of them were mainly traced to poor solder joints inside the units which could be repaired.

On the whole, Julie's resistors tended to be better than average but not the best except in a few limited cases.  Loebe was a really terrific engineer and had many good designs to his name but he wasn't the best resistor design engineer.  As I told another colleague who asked about Julie and mentioned the patents, I replied that a patent does not guarantee a good device, it only guarantees that it may be unique.

The old L&N tubular secondary standards can be quite good and very stable over time but, like other standards, they must be carefully handled and unfortunately, many of these old standards were mistreated over the years decreasing their stability and permanently changing the value.  There are still good ones out there but there really isn't any way to tell if it is good without checking them.  They are made with Manganin (still are by IET) which is a somewhat fragile alloy, it can exhibit excellent stability if treated right but it is easy to abuse them and ruin them.  The 'good' ones still deserve an excellent reputation, finding one is the problem.  I have several of them myself and nearly half of them show signs of misuse.

I presume you have a spool of bare Evanohm wire, it must be welded to another 'hard' metal to get a good weld joint and at the right pressure and energy which varies with the wire size and whatever you are welding it to.  Unfortunately, it can be welded to a hard brass but you will also encounter soldering difficulties with the brass.  You can weld it to a nickel alloy but that doesn't get you much closer to a solderable joint.  Stainless Steel alloy 305 can also be used to terminate Evanohm but it also has some solder issues.   Most of the industry uses alloy 180 ribbon terminated by an arc (or butt) weld to make the joint, alloy 180 is solderable but has a relatively high TCR of its own.  Unfortunately arc welders aren't cheap, even used ones.  A really good crimp joint might work but depending on the metal, it could create nasty thermal EMFs and very small wire is hard to crimp.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #49 on: June 21, 2014, 08:03:51 pm »
I took a good look at the LTZ1000/A data sheet and curiously, many of their data sheets will specify resistor characteristics, even down to TCRs, this one is not the case.

Hello Edwin,

the "spec" for the resistors is not in the datasheet but in the famous AN86 (APPENDIX I).
"R1 THROUGH R5: VISHAY VHP-100 0.1%"
It seems they have cut out only a part from the full cirquit in the data sheet.

I also think I have to do a review of the setup when I install the isothermal blocks.
Perhaps I can put the solder junctions closer together (especially on the UPW50),
to further reduce possible thermocouples.
On one of my measurements I have seen a
"oscillation" on the NTC difference curve (only if the cooler is active)
from which I think it cannot be physically possible.
I think I have to install a EMI-Filter on the Multiplexer of the NTCs too (as I have already done on the resistor inputs from the beginning) to reduce the effect.

In the mean time I also think that the time constants are so large that I have to do other strategy for the hysteresis measurement. I will only have 5 setpoint temperatures 25, 10, 25, 40, 25 degrees an a appropriate settle time.

Another question is: why do I see always a non-linear (convex) T.C. curve?
I think I have to check NTC linearity influence. (and possibly correct the curves).

With best regards

Andreas


« Last Edit: June 21, 2014, 09:04:01 pm by Andreas »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf