Author Topic: T.C. measurements on precision resistors  (Read 396377 times)

0 Members and 1 Guest are viewing this topic.

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #975 on: April 21, 2018, 07:34:45 pm »
Would you please post a CSV file of the PTF56 data? 
see above below the overview pictures


Andreas,

I was asking for the actual measurements, not the analysis.    I really don't feel like digitizing your graphs by hand.

What you have provided in the overview and the CSV file version  is an approximation based on an assumption that the T.C. is a straight line.  It is quite obviously not.  It's pretty close to a 2nd order polynomial.  Your three plots show the left wing, the center and the right wing.  The coefficients of that polynomial are random variables.

I want to find the form of the polynomial.  It appears to have both a shift and a rotation of axes.  However, it may be more complex.  Without data I cannot tell.  With functional fits to R1(t) and R2(t) calculating V(t) of a divider is simple and much more accurate than selecting resistors like the #8 10 K unit by testing lots of resistors.

Once I have the correct polynomial form I can generate a random set of resistors and determine how many measurements are needed  to calculate the change in the voltage of a divider over a specified temperature range.  This is an afternoon's work if I have your actual measurements. And many days if I have to collect the data myself.

I've received a lot of flack from various quarters for saying I could reduce the error of a voltage reference quite substantially.  You more than anyone have the data that would led me show what I can or cannot do.  Despite asking multiple times both in the forum and by PM I have not gotten any data I can actually use.

To state the matter succinctly, the plots show that that the resistance, R(t), is a simple function of temperature and the hysteresis, H(t), is a simple function of the history of the temperature.  If one accepts the premise that the actual resistance doesn't matter so long as one knows precisely what it is,   the residual error goes from 10-20 ppm to less than 1 ppm, most of which is noise.

Edit:  I digitized a few points from 10k#8.  With a very modest amount of effort I was able to reduce the residual error to the hysteresis.

Edit: I found my rolling parallel rule.  Attached is the fit for 25k#8.  Again, the residual is the hysteresis which I am not going to bother with unless I get the actual measurement data with all associated metadata.

Edit: Just for good measure, here's another plot.  The computer I was using for the first is not feeling goo and had to be shut down.  The point of all this is that the tangent of the temperature-resistance curve at a single point is not a good metric for choosing resistors in the 21st century.  At least not for metrology.

with best regards,

Reg
« Last Edit: April 22, 2018, 12:34:08 am by rhb »
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
A couple of questions Re: T.C. measurements on precision resistors
« Reply #976 on: April 22, 2018, 03:48:57 pm »
I received a lot of flack for asserting that a precision voltage reference could be made from low cost parts and after a few years of calibration achieve low ppm level performance.  I particularly got flack over suggesting  the use of metal film resistors rather than metal foil types.

What is shown here are trivial 2nd order polynomial fits done using the excellent Marquardt-Levenberg solver in gnuplot.  Most of the error you see is undoubtedly due to my being forced to digitize the curves by hand rather than receiving the measurements in machine readable form.

I am an experienced data analyst.  I'm not an experienced metrologist.  Aside from the fact that it would require a  good deal of time and effort to develop the skills to make good measurements, to account for aging effects one needs years of data,   

My questions:

Does anyone still assert that with measurements with a 3458A  I cannot predict the value of a random resistor to within a few ppm?

Will those of you who have large volumes of data help by providing the data and information about the experimental procedures used?

Edit: added sample gnuplot script
« Last Edit: April 23, 2018, 12:59:00 am by rhb »
 
The following users thanked this post: zhtoor

Offline zhtoor

  • Frequent Contributor
  • **
  • !
  • Posts: 337
  • Country: pk
Re: T.C. measurements on precision resistors
« Reply #977 on: April 22, 2018, 04:14:09 pm »
@rhb

it seems that any divider composed of any zero-hysteresis resistors could be modeled / corrected using measured coefficients
and thereby accurately determining the output voltage.

hysteresis looks like the real problem, because it is a function of temperature history, could you do any of that analysis?
are you proposing to maintain a temperature-historical model for each resistor to accurately account for the hysteresis?
if so, it might make sense to make a composite resistor with a temperature sensing element like a thermister and a main resistor OR
use some kind of an isothermal block to house all critical resistors for a common temperature sensor.

since there may be a number of other problems like hygrometric ingress, voltage-dependent changes etc, it seems to me that in-situ
monitoring (for a given set of operative conditions) would make more sense.

best regards.

-zia
« Last Edit: April 22, 2018, 04:16:54 pm by zhtoor »
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #978 on: April 22, 2018, 06:27:24 pm »
Hysteresis needs measurements over a range of temperature excursions.   I have no way of telling which of the hysteresis curves is for rising temperature and which is for falling.   It  appears that excursions below a certain threshold are not accompanied by hysteresis.  There is also self heating to be dealt with.  I cannot work on those without additional data.

I'm working on a wide range temperature chamber large enough to hold my shielded enclosure.  For that I need to do heat flux calculations for the sides to determine how much insulation I need and how many Peltier devices.  I'd like to cover the full industrial range, but that may prove beyond my budget. Peltier devices, heatsinks, fans and aluminum plate are cheap.  But they do have limits.

My purpose here was simply to demonstrate that there are better approaches than using the tangent to the TC curve at a single point as the figure of merit for a resistor.  And spending a lot of time and money finding ones which are near the maximum of the polynomial.

There is also the question of how effective the thermal cycling that Fluke introduced with the 7001 is at eliminating hysteresis.

For example,  for a traveling standard, log environmental conditions using an MSP430.  It is obviously desirable to hermeticly seal a reference to minimize humidity effects in an enclosure stiff enough to reduce the magnitude of internal pressure changes. An MSP430 and a 3 V coin cell would log temperature via the internal sensor for years at a BoM of $1-2.

The uncertainty produced by 1/f , thermal and current noise are the biggest limitation.  The only way to deal with them is many measurements over very long periods of time.  That implies the need to accurately characterize the aging process of a reference.
 

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #979 on: April 22, 2018, 10:02:54 pm »
Hello,

sorry your wording is difficult to understand at least for me as a non native speaker.
I never could imagine that you really want me to take my precious hobby time to grab out all the data.
Up to now you cry for data without telling what you really want to do.

I want to find the form of the polynomial.

Now with the pictures I get the image that you want to solve a problem that is long solved on my side
every time I draw a LMS curve it is actually a 3rd order polynom.
2nd order does not fit in all cases and 5th order usually does not improve the result.

What you have provided in the overview and the CSV file version  is an approximation based on an assumption that the T.C. is a straight line. 

No, the 25 deg C value is actually the linear (1 st order) coefficient of the 3rd order polynominal.
Interestingly the 2nd order coefficient is around 0.032 +/- 10..15% on this batch of the PTF56 resistors.
So the 25 deg C value is a good measure for the fitting.

Once I have the correct polynomial form I can generate a random set of resistors and determine how many measurements are needed  to calculate the change in the voltage of a divider over a specified temperature range.  This is an afternoon's work if I have your actual measurements.
you are loosing yourself in a dream world in generating artificial data.
Reality will differ: in the PPM range you have to treat every resistor as a individual.

Edit:  I digitized a few points from 10k#8.

attached the normalized result of 1 minute averages of deviation from 25 deg C value (in ppm) over temperature difference to 25 deg C in (deg C)
(ignore the first 3 lines they are only the instruction for my solver).

good luck

Andreas

 
The following users thanked this post: Dr. Frank

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #980 on: April 23, 2018, 12:54:06 am »
Hello,

sorry your wording is difficult to understand at least for me as a non native speaker.
I never could imagine that you really want me to take my precious hobby time to grab out all the data.
Up to now you cry for data without telling what you really want to do.

My apologies.  I had assumed that the data was as accessible as the plots.  I use a single directory for small projects.  For large ones I use more elaborate structures.  One really does not want to put 40,000 files in a single directory.  Access gets rather slow.

I want to develop  models of the variation of precision references caused by age, temperature, humidity, barometric pressure and any other identifiable sources of variation which are measurable and relevant.  The purpose of these models is to be able to make measurements of a device during an initial test period which will be able to accurately predict the  value  and uncertainty  of the reference for some period of time into the future.

The motivation is quite simple.  In metrology, the particular value is of little consequence so long as it is known.  It is the uncertainty of that value that matters.  Traditional reference designs have expended great effort and cost producing specific values when all that is really required is  any nearby value which is known with low uncertainty.

The length of time that such predictions are accurate is dependent upon how long a history one has for the device and whether any extraordinary events take place such as happened to one of @cellularmitosis' references.

Hysteresis associated with  environmental changes requires having the environmental history.  Such data is easily collected with an inexpensive MCU which can run off a coin cell for several years.  But to be useful for reducing the uncertainty, one must have sound predictive models of the relationship between the reference value and the environmental history.  That requires a lot of data to assure that the models are statistically valid.


Now with the pictures I get the image that you want to solve a problem that is long solved on my side
every time I draw a LMS curve it is actually a 3rd order polynom.
2nd order does not fit in all cases and 5th order usually does not improve the result.

I omitted the result for 10k#8 because I had questions about the proper equation and with a sample size of one did not wish to pursue the matter.

What you have provided in the overview and the CSV file version  is an approximation based on an assumption that the T.C. is a straight line. 

No, the 25 deg C value is actually the linear (1 st order) coefficient of the 3rd order polynominal.
Interestingly the 2nd order coefficient is around 0.032 +/- 10..15% on this batch of the PTF56 resistors.
So the 25 deg C value is a good measure for the fitting.


Then the information you presented is misleading. A proper presentation should have included the other coefficients and the form of the polynomial.  As presented, the reader would expect that the curve was the RMS average of the rising and falling passes oddly labeled by a non-native speaker.

Once I have the correct polynomial form I can generate a random set of resistors and determine how many measurements are needed  to calculate the change in the voltage of a divider over a specified temperature range.  This is an afternoon's work if I have your actual measurements.
you are loosing yourself in a dream world in generating artificial data.
Reality will differ: in the PPM range you have to treat every resistor as a individual.

The creation of synthetic data is often used and well respected in the scientific community.  There are a tremendous variety of ways of creating synthetic data.  Which should be used is dependent upon the problem.  In exploration seismology multicompany consortia expend millions of dollars to create synthetic datasets for which all the relevant parameters are known so that one can evaluate how well algorithms for recovering such information from field data perform.  Most of the cost is the many hours of supercomputer time it takes to create the datasets.

In this instance, given the mean and standard deviation of the coefficients of the polynomial, one creates a suite of resistors from which one chooses randomly.  Generally referred to as the "Monte Carlo" method. This method is used for combinatorial problems because they are inherently NP-hard and cannot be solved for even a rather small number of samples. So one makes enough trials to determine the mean and standard deviation of the combination being studied.

Edit:  I digitized a few points from 10k#8.

attached the normalized result of 1 minute averages of deviation from 25 deg C value (in ppm) over temperature difference to 25 deg C in (deg C)
(ignore the first 3 lines they are only the instruction for my solver).

good luck

Andreas

Two series are rather less than needed to arrive at any definitive result on the hysteresis. However, it is more data than I had,  so I shall be happy with that for now.  And press forward with the construction of my own temperature chamber.

As you have taken the time to determine the hysteresis and have much more data than you have made available to me, would you be willing to share your insights?  Is the behavior on the RH & LH portions of the curve the same as in the center?

This is rather important to the construction of traveling references.

Have Fun!
Reg
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #981 on: April 23, 2018, 01:54:11 pm »
I've given a good bit of thought to the physics behind the hysteresis.  In a previous life I was heavily involved in deformation of porous media, so except for being rather rusty, it's familiar territory.

The only explanation I can find for the hysteresis is plastic deformation as the result of differential expansion.  In the case of thin film resistors, the most likely candidate is the solder between the end cap and the ceramic substrate.  Ceramics have  a low coefficient of expansion, a very narrow elastic range and almost no plasticity.  The elastic range of copper is large enough that it seems unlikely that it could be forced into plastic deformation. But the solder between the two could easily be stretched or compressed.  Also solder will exhibit much greater creep than copper.

Most of my experience with equations of state have been liquids.  I can't recall ever dealing with the EoS of a solid, but this spanned about 5 years over 10 years ago.  I do know that getting EoS data is very difficult.  However, NIST has a large database with a web interface so you can get graphs of the properties of interest.  In this case bulk modulus is the first order term.  Creep data is particularly hard to get, however, creep is a likely mechanism for aging of thin film resistors.

In general, EoS are very difficult to fit.  Common practice is to compute very large numbers of candidates and then seek the best fit to the available data.  This is an area where sparse L1 pursuits are probably a big benefit.  However, most of the difficulties are in the regions around phase transitions.  So a simple empirical approximation would probably suffice for our purposes.

For a resistor constructed from a ceramic tube and metal end cap soldered to the tube, the hysteresis should exhibit a threshold effect. Small temperature changes will deform the tube, solder and cap within the elastic range.  So if the original temperature is restored, the resistance will return to the original value.  Once the stress reaches the elastic limit of the solder, plastic deformation of the solder will take place.  However, the ceramic tube and the cap will be subject to elastic strains after returning to the original temperature.  Over time, the solder will creep and the elastic stresses in the tube and cap will relax.  This is probably a major component of aging behavior.

The magnitude of the residual stress applied to the solder will depend upon the magnitude of the excursion from normal operating temperature.  Creep is in general non-linear, so the aging behavior will depend on the magnitude of the excursion.

Heating will place the solder in tension.  Cooling will place it in compression.  In general material properties are different for compressive and tensile forces.  So one would expect to see the different behaviors of hysteresis seen in the data that Andreas has presented. Rocks have very high compressive strength and very low tensile strength.  Metals exhibit less extreme differences between the two directions.  Liquids, of course, have no tensile strength.

The experimental protocol to investigate this would be to cycle a resistor from an initial temperature to a new temperature and back starting with a very small excursion and gradually increasing the excursions.  This will allow finding the point of plastic deformation.  As this will vary with temperature, a complete characterization would require normalizing resistors to a number of different temperatures.

Normalization of a resistor to a particular temperature requires thermal cycling between temperatures above and below the target normalization temperature with the excursion diminishing over time.

Does anyone know of literature on the hysteresis effect?  Comments?  Andreas, TiN, MisterDiodes?  You all have a great deal of experience and expertise in the matter.  What do you think is the cause?
 

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 457
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #982 on: April 23, 2018, 05:01:57 pm »
RHB:

Comments below are based on what I do for a living, not from the hobbyist point of view - and it is meant to be constructive only.

Be careful about making too many assumptions on how resistors are constructed.  You are at the very  tip of the iceberg of knowledge known since the 30's... For instance there probably isn't a conductive path -thru- "solder" on a resistor component, and you want to learn why solder is more of a joint stabilizer, but not the preferred  conductive route in a precision circuit component.  You'll be looking at older books or older references in the IEEE archives, etc. Some information on exact construction details won't ever be known to you or not shared here on a public forum because it is IP owned by the manufacturer.  Also remember a manufacturer might change a construction detail at any moment that might render your prediction models inert - or at least require you to re-characterize a resistor based on a different construction method.  Which could take a long time.

When you look at Andrea's data and discover the TC is a curve and not a straight line (never has been) - there is no discovery there.  Unless you're talking about small temp ranges.  Resistor TC is always some combination of parabolic curves depending on what temperature ranges you're talking about. Typically over some temperature range you'll have an Alpha & Beta coefficient for the parabolic curve at that temp range - but that will change somewhat with power dissipation aging, thermal cycles, bias point, etc.

Also remember Andrea's data is an out-of-circuit resistor at one bias point measured probably at a higher Pd than datasheet.  That doesn't make the data wrong but it is just a very small dataset.  The temperature rise per mW, Power Dissipation drift per 1000hr, stress effects, Voltage Coefficient, etc.  all play into the system.  All of those effects depend a great deal on the final assembled mechanical structure of the electrical and thermo-mechanical circuits - Andreas' data is showing just a very small part of that.  For instance:  Is the thru hole resistor mounted with simple bends either end, and the bends are at what radius?  What weight of copper PCB trace?  Or does the resistor have stress relief bends on the leads - those extra bends change the resistor's relative stress and thermal conductivity to the board also. Thermal flows to ambient all dictate how the final TC / VC / PC and long term drifts will work.

Then of course you mention hysteresis effects - that's going to be a hard one to model.  As I pointed out before:  If you have no clue about the past history of a resistor, with no knowledge of the time integral PD drift effects or where you are on the hysteresis cycle...It is not clear to me how you are going to generate a prediction model that reduces uncertainty of measurement or increases accuracy.

When you mentioned that you could measure a resistance on a 3458a suddenly 1ppm, head's up on that:  A) A 3458a offers only 7.5 digits on resistance but balance that with B) Your 24 hr traceable accuracy is at best 5ppm (read the side notes on the spec sheet).  Jumping to a "1ppm" accurate measure would be very hard to be convincing - IF you're talking about absolute uncertainty.  If you mention "1ppm" around a cal lab you'd better have something to back that up, and verified on appropriate instrumentation.

That's because I'm mostly interested in real, no bullshit traceable measurements to a CALIBRATED and TRACEABLE absolute value, and I'm interested in actually reducing the uncertainty of measure AND increasing confidence of measurement.  Something you can talk to an ISO auditor about when discussing calibration techniques and how current estimated uncertainties are calculated from last cal date.

So when you propose that you can use a 3458a to measure resistance  to 1ppm the very first thing that pops into my mind is "1ppm relative to what??"

Otherwise when you're measuring resistances -accurately- you're not using a DMM - you'll typically be using a bridge method working against a Standard Reference Resistor to get down below 5ppm uncertainty.   AND the resistors under test are mounted in an accurate thermal and mechanical simulation of the final application - because every application is different.

Otherwise you're going to be building prediction models and making assumptions based on just a few resistance measures - and that's really not going to show you the whole range of what happens in the real world.







« Last Edit: April 23, 2018, 05:12:15 pm by MisterDiodes »
 
The following users thanked this post: TiN, Andreas, Edwin G. Pettis, Alex Nikitin, 2N3055

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #983 on: April 23, 2018, 06:48:15 pm »
MisterDiodes,

Thank you.   I am generally aware of all this, though not at the level of detail at which you deal with it.  I'm certainly not aware of all the limitations of  a 3458A.  I'm unlikely to have the privilege of using one.

Much of what I dealt with professionally is unmeasurable.  You cannot take a shale core plug from 20,000' BSL without it being permanently altered.  Nonetheless, we take core plugs and we measure their properties in the lab as they move tiny fractions of a micron at different frequencies and pressures in various directions of applied pressure and modes and directions of elastic wave propagation.  In the case of measurements at actual seismic frequencies (1-200 Hz) there are very few machines capable of making such measurements.  I know of one, am certain there is another, but am not aware of a 3rd.  The one I am most familiar with took a member of the original construction team over a year to bring up after it sat idle for 10+ years. I am acutely aware that reality is very untidy.

I certainly never asserted I had discovered anything.  Merely that  the thermal behavior can be much better approximated than a first order polynomial can provide and that presenting only the first order coefficient of a 3rd order polynomial gives the reader a rather misleading impression.

In a scientific investigation one proceeds by defining an objective. One then makes some guesses as to what might be relevant to that objective and proceeds to investigate them.  One's initial hypotheses are often wrong.  When that happens you formulate a new hypothesis and try a new set of experiments to prove or disprove that.  One proceeds in this fashion until one either reaches the objective or runs out of funding.

There is an old expression, "You eat an elephant one mouthful at a time."

I just spent an hour or more searching for papers on the thermal hysteresis of resistors and found nothing other than a NIST paper on thin films deposited on silicon wafers as a means of calibrating optical thermal sensors in a semiconductor fabrication process from 100-600 C.  Somewhat outside the temperature range of my interest.  I am certain there is far more work on the topic, but that much of it is locked away in company archives.  If anyone is aware of published papers I'd very much appreciate being informed of them even if it's just an author's name.

I don't think that conduction in the solder contributes significantly to the thermal hysteresis observed in Andreas' data.  I think it more likely that elastic strain of the ceramic element is the principle mechanism.  Thermal hysteresis implies the existence of some plastic deformation.  For the reasons I cited, the solder attaching the end cap to the ceramic body is the most likely element to be undergoing plastic deformation.

I have suggested that an MCU be used to log the thermal history of a reference.  An MSP430 can measure the temperature once a minute for years running on a single coin cell.  Having the history is not a problem.  It appears that understanding how to use that thermal history to reduce the uncertainty of measurement is relatively unexplored territory.

The measurements to which you are accustomed to making, as difficult as they may be on a day to day basis, pale in comparison to trying to determine the properties of a substance at a few hundred MPa and 1000 C.  Pure water at those temperatures and pressures will dissolve almost  anything it touches.  And those temperatures and pressures are minor compared to those in the earth 100 miles below your feet.  Simply put, such data will never be abundant, so you learn to do as much as possible with what you have.

If you will allow my hypothesis that plastic deformation of solder is a likely contributor to the thermal hysteresis of a metal film resistor, then I have suggested the following experimental protocol:

Thermally equilibrate the device by temperature cycling over a range of ever narrower temperature excursions in both directions converging on the desired test temperature. Then measure the hysteresis for small steps in one direction, each time dropping back to the equilibrated temperature.  Requilibrate the device and measure in the same manner in the opposite direction.

This process needs to be repeated at a number of equilibration points spread over the temperature range of interest.  As there is unquestionably an aging component related  to relaxation caused by creep, one would need to hold the equilibration point temperature for some time to measure that.

Is this the best procedure?  I have no idea.  It seems reasonable.  I do not expect that the properties of the solder will be constant over temperature, so this is a way to allow for it changing.  Is it a complete model of  the continuum mechanics?  Absolutely not.  But it includes the major factors:

plastic deformation due to differential expansion and contraction
temperature dependence of elastic moduli
plastic creep

If this model is not acceptable, please state why and propose a better model and a procedure for testing it experimentally. But please restrict yourself to the case of the thermal hysteresis of  a thin film resistor under ideal conditions.  The many other effects and considerations you insist on raising are interesting, but not germane to establishing a model for thermal hysteresis of a metal film resistor of construction similar to the PTF56.

We would not have electronics if Faraday, Ohm, Maxwell et al allowed themselves to be dissuaded from their work by the fact that reality is more complicated than the simple models they devised for idealized cases.


 
The following users thanked this post: zhtoor

Offline MisterDiodes

  • Frequent Contributor
  • **
  • Posts: 457
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #984 on: April 23, 2018, 11:06:57 pm »
RHB:  I think you meant to tell me to restrict my comments to "Metal" film, not "Thin" film.  Those are completely different resistor types.  I will also comment however I see fit or as time allows, sorry.  I'm just trying to give you a head's up here on some real effects you might encounter that you might want to be prepared to model.

Are -all- the effects contributing to drift germane to a PTF56?  Of course!  Because if a PTF56 resistor was immune to power dissipation effects it would cost a lot more and be a Bulk Metal Film...

Remember: if a theory is a good one it will always hold up and withstand any attempts to break it, when looked at from any angle.  Sometimes if someone has a skeptical point of view that might spark an even better idea at your end.  Or not.  I wish you the best of luck on your experiment, no matter how it turns out!  I promise!

RE: Eating Elephants (your analogy) I think you've got even more than one elephant here.  As Edwin says you've got just a two terminal device with multiple effects going on at once.  Some of the resistor MIL test standards will show you how to separate out some of the effects.  Some effects like TC have multidimensional gotcha's:  For instance if you measure the TC, and then heat cycle the resistor to a much higher or lower temperature and then check TC you'll have something different.  Or try doing a slow heat with a fast cool shock and vice versa :  Now you've got another TC that's a little different.  Does the resistor recover over time?  Sometimes yes or no.  Now do some rapid thermal pulses...maybe something new, maybe not!  And so on.  VC is another gotcha that creeps in if your bias voltage changes rapidly.  All of that depends on the mechanical and thermal circuits too.  Many of these items will not be in text books - you're better off gathering your own data,

The more interesting thing about cheap metal films is what happens when you don't pay a little extra for stability:  The power dissipation load life keeps changing resistance value over time. And changing and changing and changing.  SOMETIMES you'll see it stabilize a bit over time but give it a thermal shock and you might be wobbly again.    That's why PTF56 load life is spec'd in percent, not ppm.   Usually a plastic deformation model grows for a while and slows down, maybe stops and reverses somewhat upon removal of the stress.  Try that with a PTF56 and let me know if that worked 2 or 5 or 10 years from now.  Creep?  Yes...Plastic deformation?  You've got multiple processes going on there and you want to look at the longer time frame.

Based on experience, I think using another CPU to monitor and log the temperature of a cheap resistor over time has some pitfalls:  Since the interesting driving factor on a PTF56 long term drift is more power-dissipation dependent. you're probably going to need to monitor not only Temp, Humidity and Baro Pressure but also the resistor bias point (power) over time.  More or less variables as required.  And start computing the time integral effects of all those factors into your calculations.  If your logging battery goes dead then you're just "running in the blind" again.

Also consider:  At the end of the day is the labor time of characterization for a specific application, extra expense of the logging CPU, battery and sensors really worth it to use a cheap resistor, or do you spend a few extra bucks and use a decent resistor in the first place?  Which way is safer, lower noise and more robust??   Or does your technique make good resistors even better??  There isn't a correct answer here, your test results will show how that all plays out.  Let's see how you do!

I really suggest you start making some tests of your own and get a "feel" for some lower ppm measures.  Maybe visit a cal lab and watch a resistance bridge in action, and compare that to how a "rustic" 8.5 digit DMM works in comparison.  You'll need some good data first, and across several bias points over different time frames and temperatures.  It's not easy, and you really want to consider a precision resistor as part of the SYSTEM in the application, not a component on it's own. Because a resistor is affected by everything going on around it - electrically, mechanically and thermally.  It is the sum -total effect- of the parts in the system that matters in the end... :)

I'll leave you to your experiments now!  Good luck!
 
The following users thanked this post: TiN, Edwin G. Pettis, zhtoor

Offline thermistor-guy

  • Frequent Contributor
  • **
  • Posts: 365
  • Country: au
Re: T.C. measurements on precision resistors
« Reply #985 on: April 24, 2018, 02:17:21 am »
...
For a resistor constructed from a ceramic tube and metal end cap soldered to the tube, the hysteresis should exhibit a threshold effect. Small temperature changes will deform the tube, solder and cap within the elastic range.  So if the original temperature is restored, the resistance will return to the original value.  Once the stress reaches the elastic limit of the solder, plastic deformation of the solder will take place.
... 

Is this correct? I thought the end caps were force-fitted to the ceramic body, not soldered:

https://www.electronics-notes.com/articles/electronic_components/resistors/metal-film-resistor.php
"...Once the film has been deposited, a metal end cap is pressed over the deposited metal. This makes contact with the resistive film and has the leads incorporated..."

http://www.radio-electronics.com/info/data/resistor/metal-film-resistor.php:
"...Once the basic rod is complete, caps are pushed onto the ends to connect with the conductor and these enable the wire connections to be made...."
 

Offline Edwin G. Pettis

  • Frequent Contributor
  • **
  • Posts: 451
  • Country: us
  • The plural of anecdote is not data.
Re: T.C. measurements on precision resistors
« Reply #986 on: April 24, 2018, 02:47:56 am »
Yes, the end caps are press fit, for the PTF series, they are copper, you cannot solder to the metal film, at best you'll have a cold solder joint.
 
The following users thanked this post: MisterDiodes, 2N3055, razvan784

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #987 on: April 24, 2018, 04:36:02 am »
Hello,

as already said by the experts: there are too many influences affecting hysteresis.
And if some plastic is in the housing: humidity will also have a influence (see my baking test some pages before).

My strategy would be to select components with low hysteresis (in the temperature range of interest).
Of course the ageing also plays a role for the total "precision score" in a cirquit.
But this would be another thread.

For the PTF56: I am not shure if he has really end caps as I am missing the "bubbles" on each side.

with best regards

Andreas
 

Offline thermistor-guy

  • Frequent Contributor
  • **
  • Posts: 365
  • Country: au
Re: T.C. measurements on precision resistors
« Reply #988 on: April 24, 2018, 06:04:01 am »
...
For the PTF56: I am not shure if he has really end caps as I am missing the "bubbles" on each side.
...
Andreas

The Vishay data sheet refers to DSCC drawing 89088, rev. F (2013) of which is here:
https://landandmaritimeapps.dla.mil/Downloads/MilSpec/DsccDwg/89088.pdf

Vishay is listed as the only approved vendor. I can't see any clues about end-cap construction. A careful tear-down or x-ray of a PTF56 sample would be interesting. My guess is the interface, between end-cap and the resistive metal film, is an important contributor in a resistor's hysteresis. The fact you are seeing almost no hysteresis has me intrigued.
« Last Edit: April 24, 2018, 06:06:18 am by thermistor-guy »
 

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #989 on: April 24, 2018, 04:37:25 pm »
The fact you are seeing almost no hysteresis has me intrigued.

Hello,

as you can see above, not all PTF56 resistors have that low hysteresis.
The 1K resistors are much worse.

Example here:
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1415602/#msg1415602

and overview sheet ... where is the overview sheet with all 25 resistors?
Dammed I fear I did not do this up to now due to the AD587LW project and Jasons VHP resistors.
So another point on the todo list.
But one has to set priorities ....

with best regards

Andreas

 

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #990 on: April 24, 2018, 09:52:29 pm »
Hello,

as the topic is "T.C. measurements" here a first result of the replacements of branadic on his 8G16 1K resistors.

The 1704 datecode had a very large T.C. between 6-12 ppm/K instead of <=5 ppm/K as on the data sheet and also large hysteresis of up to 39ppm.
https://www.eevblog.com/forum/metrology/t-c-measurements-on-precision-resistors/msg1412212/#msg1412212

The first 1K replacement resistor datecode 1807 (8G16_1K#04) shows at least a T.C. of around 3.7 ppm/K (box method).
Hysteresis is still at +/-8ppm.

LMS coefficients:
A 0 =  5.4822777936661778E+0000
A 1 =  3.9043113572905335E+0000
A 2 = -8.9087314593799532E-0003
A 3 = -5.5098001147050806E-0004

Drift was 5 ppm during 4 days cycling. Most of it on the first day then stabilizing.
But since I had changed (soldered) the reference resistor also to 1K it can be that also the reference resistor needed some time to stabilize again.


with best regards

Andreas
 

Offline rhb

  • Super Contributor
  • ***
  • Posts: 3476
  • Country: us
Re: T.C. measurements on precision resistors
« Reply #991 on: April 25, 2018, 01:23:06 am »
Andreas,

Would you be so kind as to point me to a description of the experimental protocol you used for the PTF56 resistors?

My interest is the impact of a temperature excursion on a reference.  It's not clear to me that your data is usable for addressing that.

Thank you for your time and attention.

with best regards,
Reg
 

Offline texaspyro

  • Super Contributor
  • ***
  • Posts: 1407
Re: T.C. measurements on precision resistors
« Reply #992 on: April 25, 2018, 01:32:11 am »
The Vaisala HMK15 manual has a good discussion of using salts to calibrate humidity.   There is a table of how the humidity changes with temperature.

https://www.vaisala.com/sites/default/files/documents/HMK15_User_Guide_in_English.pdf
 

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #993 on: May 06, 2018, 09:09:51 pm »
Hello,

and the remaining resistors 8G16 1K datecode 1807 and a overview over some 1K resistors (mainly PWW).

all are within spec (max 5 ppm/K) but have also a hysteresis around +/-10 ppm over my 30 deg C extended room temperature range.

There is also one outlier with large drift over the 3 days measurement.

with best regards

Andreas
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: T.C. measurements on precision resistors
« Reply #994 on: May 06, 2018, 09:15:30 pm »
Hi Andreas,

would you mind to bake the resistors for 8h @ 150°C and repeat the measurements for one batch, say the 1k resistors? I could imagine that the expoy needs a post mold process.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 
The following users thanked this post: Andreas

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #995 on: May 09, 2018, 09:09:00 pm »
Hello branadic,

my kitchen boss sent me a big dislike.   :--
But I am shure we will find a solution.
Snail mail sent.

In the mean time the first results of a 8G16_120R (#5) resistor with datecode 1807
This time I give the measurements of all 3 days because obviously the resistor is quite drifty.
So I actually have the hope that baking will give some advantage.

on 3rd day we have 1.19 ppm/K Box T.C. with +/-14.8 ppm Hysteresis and -1.9 ppm drift against first day
on 2nd day the drift was -6.6 ppm so quite jumping around.

with best regards

Andreas
 

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: T.C. measurements on precision resistors
« Reply #996 on: May 12, 2018, 05:55:40 am »
A TEC box is on my project list, for now I can really only do a one way TC plot.  Picked up a Vishay 10K VHP101 (Y407810K0000V9L) for a small reference box project.  Here is a quick and dirty test from 15 to 36C, the plot is zeroed at 23C.

« Last Edit: May 12, 2018, 02:58:59 pm by kj7e »
 
The following users thanked this post: Andreas, eurofox

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: T.C. measurements on precision resistors
« Reply #997 on: May 14, 2018, 10:16:17 am »
Quote
Hello branadic,

my kitchen boss sent me a big dislike.   :--
But I am shure we will find a solution.
Snail mail sent.

Hello Andreas,

the resistors arrived today and are already in the bakery :)
I think they are ready for transport tomorrow.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 
The following users thanked this post: Andreas

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: T.C. measurements on precision resistors
« Reply #998 on: May 15, 2018, 01:22:02 am »
Here is a comparison between a Vishay Z201 and VHP101, both 10K 0.005% tolerance.  The hardest part was keeping my room temp within 0.5C for consistent results on my instrument.  This was especially true on the VHP101, almost a flat line TC;



Z201:     18-28C = 0.8ppm/C.    10-40C = 0.7ppm/C.
VHP101: 18-28C = 0.05ppm/C.  10-40C = 0.05ppm/C.



I found placing the parts in a plastic bag to keep air drafts out, then in tin foil with a jumper to my earth bar greatly helped minimize noise;


First, the box was cooled using frozen cold packs to about 1 deg C, then I applied slight energy to the resistor bank to warm it up.  I managed to keep the slope 6min-8min/deg;



« Last Edit: May 16, 2018, 02:21:41 am by kj7e »
 
The following users thanked this post: Andreas, try, cellularmitosis, eurofox, hwj-d, thermistor-guy

Online AndreasTopic starter

  • Super Contributor
  • ***
  • Posts: 3222
  • Country: de
Re: T.C. measurements on precision resistors
« Reply #999 on: May 20, 2018, 04:52:02 am »
Hello,

while the 12K and 1K resistors were traveling to get a true 150 deg C treatment
 I made a  around 70 deg C treatment in my 2nd cooling box for the  120 R resistors #7 to #9
for several hours.

Measurement of 120R#7 on 13.05.2018 showed a reduced hysteresis
(especially on low temperatures).
But next day the resistor was drifted by about 14 ppm.
On 3rd day (15.05.2018) the drift was around 18 ppm with again increased hysteresis.
Day 6 (18.05.2018) shows that the 18 ppm have stabilized.
So the treatment of #7 at 70 deg C has brought only a improvement
on hysteresis for the first measurement after the treat.

with best regards

Andreas


 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf