T.C. measurements on precision resistors

[babysitter]

Hi Edwin,

yes they are quite massive. Diameter ~20 mm (almost 0.8") , length even more, either ~31 mm or  ~26 mm.
The center of the resistor is a piece of brass tube passing right thru the resin body, I think that was more for fixing than inserting a thermometer.
There are residues from a yellow print with a ligature "TD"-like  logo, one showing a number "1069" - certainly the date code finally found!

[Alex Nikitin]

Most of the old Ultronix resistors I have are somewhat tighter in tolerance and much smaller in size.

I have quite a few of Ultrohm resistors like one on the photo, some 0.1%, some 0.05% and some 0.02%, from 50 Ohm to 1.8K. Even 0.02% ones are still in tolerance. I've used some to build the feedback divider in the 7V to 10V stage in the JVR reference. The whole reference did drift less than 5ppm over one year, so the resistors look OK for that use.

Cheers

Alex

[Edwin G. Pettis]

I'm rather curious about the internal construction, we didn't have anything with a brass tube through it, for whatever reason, Ultronix had very few drawings of earlier resistor designs in our library (huge set of drawers with lots of drawings on velum, long before computer records).  The large wire diameters were easier to work with and despite very high molding pressure could be reasonably stable, a lot depended on the type of termination that was being used, most likely these rather large resistors used a different type of termination than the regular axial (or radial) resistors.  Those lugs are far larger than the usual terminations on the smaller resistors.

[TiN]

Lookie what showed up in my mailbag. Precision wirewounds made by Edwin G. Pettis. It's a perfect 70K*2/1K/12.5K/120 set for LTZ-reference.
Here's photo with PTF56 75K and my standard VPG VHP203Z for size comparison.



I took some physical measurements, as there is chance I might use this type resistors in upcoming project.



My K7168+3458 setup is tied to reference measurements till next few weeks, so I'll have to use pair of Keithley 2002's + 2510 to run thru TCR curves. Conditions are same, 20C slow ramp up and down. Most of wirewound resistors I've tested have strong non-linear TCR curve, unlike foil resistors. So I'm personally very curious what these little candies will show. 

[Andreas]

Hello,

so they are smaller than the standard 805 style that I have. (13 * 6.4 mm)
Waiting how they compare ...

With best regards

Andreas

[mimmus78]

Ohh that's will be interesting to see how they compare to what I found. Have I published some of this charts from this resistors?

[TiN]

Here are first two, 12.5K and 1K. I have them thermally coupled with copper foil wrap.



Similar to other PWW's, there is significant nonlinear effect.
I decided to run second ramp-up/down as well to confirm absolute deltas, as end +20C values did not quite reach start points.

[mimmus78]

TiN may you post CSV? I'd like to compare with my charting program. There seems to be a parabola too, am I right?

I think this is the same thing as my overshoots. Can be the case that you get it before that rampup is finished because your rampup last 12 hours and mine that is only 30 minutes?

Need to find some time to programm a slower rampup to check if I can get it too "with anticip".

Inviato dal mio Nexus 6P utilizzando Tapatalk

[TiN]

Just click on graph to see live data, there is link to DSV in there.

[Galaxyrise]

During the 20 to 40 ramp, the resistance data is non-linear.  But during the 40-20 ramp, the resistance data looks linear to me.  Why is that?

[fluxamp]

Just looking at the plot, the curve for the 12.5k resistor strikes me as being odd. The curve for the 1k resistor changes slope every time the temperature does, but the one for the 12.5k only does so at the end. Given that both resistors behave similarly at the end, I'd suspect that the first half of the measurement is swamped by an effect different from the TCR of the resistors (though my knowledge about that stuff is too limited to be sure). Might also explain the non-linear vs linear issue that Galaxyrise noticed.

[Andreas]

Hello Illya,

I think you will have to do at least a 3rd run.

At minimum the first hot phase seems to be influenced by humidity which is/was within the epoxy of the resistor.

With best regards

Andreas

[MisterDiodes]

TiN:

I would like to offer a suggestion:  When running resistors tests, it's always better to run them under actual bias conditions - just doing a DMM ohm's measurement doesn't really tell you what the real stability performance is "in-circuit".  You want to take into account the effect of local heat flows and how stresses equalize across the resistors, and a resistor dissipating any non-zero power will provide some small buffer to humidity effects - the air around the resistor will always be very slightly above ambient.

In other words, any sort of direct DMM ohms testing only tells you how the resistor acts with your DMM Ohm's measuring current.  Real application situation might be a different story when you chase PPMs.

Remember also that the 3458a isn't really the best way to measure precision resistances - it is a good tool for voltage, but a bridge and / or real resistor current flow is your best friend for digging into precision resistance measures.

I would run the test again, but measure the performance at how these are actually used in your LTZ:  Measure the ratio of output voltage 12.5k/1k under your circuit typical bias across the pair - 7VDC or however you have your heater ratio biased.

Now look at how the Ratio output changes over temperature.  Now look at LTZ datasheet and see about how much temperature swing it's going to take to change your Vref 1ppm (which you can't really measure with confidence anyway, realistically).... Or whatever spec you need for your application.  These will probably be fine depending on how tightly controlled the spec is you're aiming for.

Remember, it's going to take at least 100ppm change in heater resistor RATIO to change the Vref output of LTZ by 1ppm.  Unless you have a JJ-Array handy, that's still beyond any real capability to measure absolute voltage values that accurately.  Nulling against multiple calibrated 732b's can get you in the 2ppm uncertainty area, but that's about the realistic confidence limit for any lab without a JJ-Array on-site.  Even relative voltage values are hard to accurately measure in the low PPMs over a few day's time, even with the best 3458a.  As you well know - be careful - those low PPM numbers are hard to measure carefully and accurately and with repeat-ability.

I would also suggest putting your resistors under bias for at least 24~48hrs before testing, especially after shipping.  You don't know how cold or hot that shipping container got during transit - and any precision resistor will need some thermal stress recovery time under bias before you can measure true performance.

[TiN]

Here updated graph, completed two runs and start of 3rd (with bit slower ramp).
TCR Box, resistors, wiring or meters are not disturbed between runs, and no work was performed nearby.

AC is always off, so there is only natural convection around in the room. You can see that from BME280 sensor readings,
with ambient temperature in range from +26 to +27C, RH in range 53-58%.



1K resistor behave normally and almost linear, other than about -10ppm shifts after each cycle.
However 12.5K one have positive tempco till around 32-35c, reversing to negative right after.
Resistance shift after first run was -27 ppm, and -13 after second run.

Make me think to bake these resistors for some good amount of hours and coat them afterwards to get more predictable behaviour.

Remember also that the 3458a isn't really the best way...

Indeed, even though actually pair of K2002's used for measurements above.
Just to note, most of foil resistor ramp up and ramp down linearly, and do not show resistance change after the shift (both hermetical fancy VHP's and regular S102's).
This can be also explained by much much smaller resistor element thermal mass in foil type resistors as well.

I would run the test again, but measure the performance at how these are actually used in your LTZ:  Measure the ratio of output voltage 12.5k/1k

I think this is great summary of the next test, using my fresh acquired K6221. 

Remember, it's going to take at least 100ppm change in heater resistor RATIO to change the Vref output of LTZ by 1ppm...

Great reminder to anyone playing with LTZ-REFs, I just mark this bold here for future lurkers. Indeed good point. 

I would also suggest putting your resistors under bias for at least 24~48hrs before testing, especially after shipping.  You don't know how cold or hot that shipping container got during transit - and any precision resistor will need some thermal stress recovery time under bias before you can measure true performance.

100% true.

Let's agree, that even with all going on these are great resistors, with very low tempco, suitable for usual LTZ1000 application, as there are no 20C variations in the circuit to even see any of these extreme resistance shifts. If one need guaranteed better than 2ppm/K TCR resistor, get ready to pay something in XXX $USD range per piece and be ready to invest days of testing to confirm that number with proper gear.



Here's a glimpse of one of my best resistors, measured by exactly same setup, 95 KOhm VHP101 with PMO. It was 90$ a pop and 30 weeks lead-time.  .

[Edwin G. Pettis]

A note to TiN, none of my resistance wire have positive TCRs and while stress can cause an apparent increase in TCR, the TCR of the wire cannot be changed, all of my wire spools are between 0 PPM and -3 PPM/°C, no positive TCRs at this time.  Unlike other manufacturers of wire wounds, mine do not produce different TCR polarities with the same spool of wire, this effect is due to the way they manufacture.  I cannot say what might be causing this apparent positive TCR but it is not the resistor.  I have never used a spool that has a dual ±TCR characteristic, which many spools of wire do have, mine are specifically manufactured to my specification, while I have a couple of spools with a dual TCR, it is still within my specs.  The 12.5K resistor you have has the dual TCR of -2.5 / 0 PPM/°C, the cross over point being the reference temperature of 25°C and this is not an abrupt change over.  Since I don't know where/who that resistor came from I cannot pinpoint the exact time of manufacture but that was the wire used on it, the 1K resistor would have been made with the same wire as usual.

[TiN]

Thanks Edwin, I'll be surely doing more tests, I agree that it's not clear what is the cause of such changes.

[Andreas]

1K resistor behave normally and almost linear, other than about -10ppm shifts after each cycle.
However 12.5K one have positive tempco till around 32-35c, reversing to negative right after.
Resistance shift after first run was -27 ppm, and -13 after second run.

Make me think to bake these resistors for some good amount of hours and coat them afterwards to get more predictable behaviour.

I would also suggest putting your resistors under bias for at least 24~48hrs before testing, especially after shipping.  You don't know how cold or hot that shipping container got during transit - and any precision resistor will need some thermal stress recovery time under bias before you can measure true performance.

100% true.

Hello Illya,

so besides the drying of the first run
your resistors behave very similar to my 805 samples.
the 12 or 12K5  resistors with around +1ppm/K
the 1K resistor with around -1ppm/K
(please correct the -1.5 and -1.35 ppm/K to positive in your graph for the 12K resistor as the slope of temperature and resistance are in the same direction).

Unfortunately for the divider ratio this gives around 2ppm/K in total.

Baking is a good idea to get better results (lesser hysteresis) if you measure immediately after the baking.
But I also suggest to let the resistors rest for 2 weeks on the shelf and then you should get something similar as in your current measurement.
So for practical use baking will not help.

Also running "under bias" has no influence for a LTZ cirquit.
The power dissipation of the LTZ resistors is between 0.25 and 3.5 mW in worst case.
so for a thermal resistance of 20-50 K/W of the resistor this will give no more than 0.2 deg C temperature increase under bias.

With best regards

Andreas

Edit: corrected 2 deg C -> 0.2 deg C

[Dr. Frank]

Hi Illya,

I thought, that Edwin Pettis would be able to produce the 1k and the 12.5k from the same wire, with virtually identical T.C.?

As tests of several contributors have shown, the mitigation factor for the 12k/1k divider is about 76 (not 100) in worst case, so these total ~2ppm/K contributes numerically 0.026ppm/K for the LTZ circuit.

Anyhow, this can be reduced further by using the compensation resistor.

Frank

[mimmus78]

TiN may you make a test with fast rump up and down?

Something like 30 °K in 40 minutes and maintaining the high/low temperature for 1h.

If my theory is right you will get not the parabolic behaviour  bust just a little bit of overshoot like I observed in my tests.

Inviato dal mio Nexus 6P utilizzando Tapatalk

[MisterDiodes]

TiN:
I suggest you try a test where you hold temp for a dwell time to find out the thermal time constant of your test setup.  In other words, change temp 5 degrees, then hold temp steady until you see resistor stabilize, then change again and wait until you see the resistors really stabilize, etc.  Until you do that test you won't know how fast your test setup can respond to temp changes.  You might find out your temp. ramp rate has to be much slower.  We don't use a constant temp ramp, we've found that a dwell time at each temp level is a little safer.  That will show you if you're outpacing the resistor's ability to stabilize to the new temperature.  This is where you're going to find out that the size of the resistor bobbin makes a difference.

You'll also find out differences where it all depends on how you've got the resistors setup, and ideally they are mounted on a real board in that would be like in your final application.

Andreas:  You never know what the thermal flow coefficients of any precision resistor is for any application until it is run under real world conditions - and clipping the resistor to a set of DMM leads is not at all the same as running in-circuit.  In other words for a real test the resistors are mounted up on the board and under real current flow. Your heat rise change and resulting wire stresses in the resistor at real current can be completely different from mine, depending on how you test.  How these are soldered to a board, the enclosure, weight of copper on the board, mounting style, do you bend the leads 90° at the end of the resistor, mount vertical, enclosure insulation, etc.,  etc.  All of those will have an effect on resistor stability performance.  It's always best to use real world conditions for your real application test.  Just running these on any DMM will -never- expose the real story on performance in your application.

Dr. Frank:  I won't argue with you on your results heater resistor ratio influence - but from a customer test we did 5~6 years ago on 10ea LTZ1000ACH - in a lab with access to JJ-array (measurement uncertainty ~0.3ppm, which is about the best possible we can expect anywhere) - we found the datasheet is closer to being correct.  We found the heater resistor -ratio- had to change at least 103~105ppm to deliver a solid >1ppm change on output voltage, at 25° ambient, PWW 3ppm TCR resistors, mounted on our 1.5oz copper board design inside our application enclosure with semi-insulated LTZ can.  Again this is one of those measurements where all environmental elements comes into play.  Your mileage may vary.

Even if we take your more conservative estimate of an LTZ Vref delta attenuation factor of 76 lower than the heater resistor ratio change - these resistors will be fine for most applications - especially given the fact the basic LTZ stability as ppm drift is down in the mud when it comes to most real lab's measurement capability (Usually better labs are at 2ppm uncertainty for voltage).  It all depends on the expected operating temperature range of your final application.

The take-away here is that for a typical LTZ 7V datasheet circuit - expensive Vishay's won't buy you much of anything in terms of overall profitable cost / performance benefit, and that foil will always get you a bit more noise over PWW's (although that noise effect is difficult to measure on LTZ output, but is there).

[Edwin G. Pettis]

Once again, (read reply 713 above) I do use the same wire to make both of the heater resistors, therefor they have by default the same TCR.  The TCR of the wire cannot be changed by any operating temperature, not even +150°C.  The apparent TCR can be made to look higher by external stress if applied directly to the wire, this is not a change in wire TCR nor can it change the polarity of the TCR.  I cannot stress enough the difficulties of measuring PPM (or sub-PPM) quantities, there are many artifacts that can affect the readings.  Again, the only reliable method of reading PPM quantities of resistance is with a high quality resistance bridge of sufficient accuracy and uncertainty.  Any other method brings additional uncertainties to the measurement process whether known or unknown.  I know the capabilities of my resistance systems, their accuracy and uncertainties.  The accuracy and uncertainty varies to some degree with the actual value in question.  I cannot measure a 10 ohm resistor to the same degree of accuracy as I can a 10K resistor because of system limitations and I know what those are.

Calculated values made from measurements with unknown uncertainty errors are not acceptable, calculations does not enhance the original accuracy and uncertainty of the measurement, ask a high grade calibration lab about that.  Whether using a 3458A (freshly calibrated) or a Keithley K2002 (or similar) does not equate to a high performance resistance bridge.  All these posted readings I have seen have never quoted an uncertainty factor (likely greater than 2 PPM), the readings are apparently taken at face value, errors and all.  Every reading has an uncertainty, often from multiple sources (you don't know accurately, the temperature to 5 decimal places or even 3 decimal places), this cannot be ignored.

While these measurements can be used as a rough basis for comparison, this cannot imply actual accuracy of the measurements without the actual accuracy and uncertainty declared.  None of the equipment mentioned here are capable of the accuracy implied, whether calibrated or not, the uncertainty of said measurements easily exceeds the implied accuracy.  Yes, you can compare 'relative' measurements but that still leaves wide open the actual accuracy involved and uncertainty (a real question is, relative to what?  At best, that means itself which is virtually meaningless).  The world of PPM is filled with unknowns and uncertainties unless you're the NIST (or equivalent), a fact you cannot get around.

I can read a 10K resistor on my 242D bridge to at least 10 digits, it may even repeat a reading to 9 or ten digits, that doesn't change the fact that the accuracy is within 1 PPM and the uncertainty is 0.5 PPM, that absolutely means the accuracy stops at the 6th decimal place and anything beyond that is essentially useless, it cannot be used to enhance the reading by math.

I understand the premise being pursued here, it can be fun and educational but it is a requirement of science to maintain accuracy of statements and measurements, state the known accuracy of your equipment if known or if it is unknown, and uncertainty involved and view the data in that light, it is of little use to compare unknown quantities, particularly when these quantities are being claimed in PPM.  As Bob Pease would say, that is just sloppy work.

[Dr. Frank]

Dear Mr. Pettis,

I don't want to lessen the features of your fine resistors, neither the capabilities of your measurement equipment.

Anyhow, all what you write about the uncertainty on resistor measurements, i.e. the capability of such a high precision bridge versus these 8.5 digit DMMs is correct, I assume, but for absolute measurements only, in this context.

Even if the absolute measurement value would off by many ppm, by using highly resolving DMMs, the mean T.C. can be determined quite well, because the change of the resistors value is quite big over this big temperature change, and for the case of a linear R(T) curve, Linear Regression calculus will give very good statistical uncertainty for the T.C.

The T.C. itself also has not to be so precise, a precision of several 10% would be good enough.

Especially in this case, and disregarding any precision figures, it is obvious that the 1k and 12k resistors have opposite apparent T.C., although they both are manufactured from the same wire, as you stated.


And that's very interesting, why this is the case , despite these both wires should have the same T.C. by design.

You mentioned stress as a possible root cause in the beginning, could you please comment on that a bit more?

I have used the General Resistor 'econistors' and also found too high T.C.s on 120 Ohm components, obviously due to too tightly packaged or molded windings, and during quite fast T.C. measurements.
G.R. determined their T.C.s at 2-3 fixed temperatures only, letting them relax for 24h or so before taking a measurement.
They also got T.C.s of inverse sign, than me, on the same samples. The T.C. was quite constant over temperature, so it's not been due to the box averaging method.

If slow relaxation effects falsify the apparent T.C., such resistors have a limited application, especially for our LTZ circuit.

Frank

Once again, (read reply 713 above) I do use the same wire to make both of the heater resistors, therefor they have by default the same TCR.  The TCR of the wire cannot be changed by any operating temperature, not even +150°C.  The apparent TCR can be made to look higher by external stress if applied directly to the wire, this is not a change in wire TCR nor can it change the polarity of the TCR.  I cannot stress enough the difficulties of measuring PPM (or sub-PPM) quantities, there are many artifacts that can affect the readings.  Again, the only reliable method of reading PPM quantities of resistance is with a high quality resistance bridge of sufficient accuracy and uncertainty.  Any other method brings additional uncertainties to the measurement process whether known or unknown.  I know the capabilities of my resistance systems, their accuracy and uncertainties.  The accuracy and uncertainty varies to some degree with the actual value in question.  I cannot measure a 10 ohm resistor to the same degree of accuracy as I can a 10K resistor because of system limitations and I know what those are.

Calculated values made from measurements with unknown uncertainty errors are not acceptable, calculations does not enhance the original accuracy and uncertainty of the measurement, ask a high grade calibration lab about that.  Whether using a 3458A (freshly calibrated) or a Keithley K2002 (or similar) does not equate to a high performance resistance bridge.  All these posted readings I have seen have never quoted an uncertainty factor (likely greater than 2 PPM), the readings are apparently taken at face value, errors and all.  Every reading has an uncertainty, often from multiple sources (you don't know accurately, the temperature to 5 decimal places or even 3 decimal places), this cannot be ignored.

While these measurements can be used as a rough basis for comparison, this cannot imply actual accuracy of the measurements without the actual accuracy and uncertainty declared.  None of the equipment mentioned here are capable of the accuracy implied, whether calibrated or not, the uncertainty of said measurements easily exceeds the implied accuracy.  Yes, you can compare 'relative' measurements but that still leaves wide open the actual accuracy involved and uncertainty (a real question is, relative to what?  At best, that means itself which is virtually meaningless).  The world of PPM is filled with unknowns and uncertainties unless you're the NIST (or equivalent), a fact you cannot get around.

I can read a 10K resistor on my 242D bridge to at least 10 digits, it may even repeat a reading to 9 or ten digits, that doesn't change the fact that the accuracy is within 1 PPM and the uncertainty is 0.5 PPM, that absolutely means the accuracy stops at the 6th decimal place and anything beyond that is essentially useless, it cannot be used to enhance the reading by math.

I understand the premise being pursued here, it can be fun and educational but it is a requirement of science to maintain accuracy of statements and measurements, state the known accuracy of your equipment if known or if it is unknown, and uncertainty involved and view the data in that light, it is of little use to compare unknown quantities, particularly when these quantities are being claimed in PPM.  As Bob Pease would say, that is just sloppy work.

[CalMachine]

a lab with access to JJ-array (measurement uncertainty ~0.3ppm, which is about the best possible we can expect anywhere)

I'm not trying to jump in here because a lot of this is over my head at the moment, but I would like to point something out.  I had our 732B cal'd directly against Fluke Everett's JJA in January and they reported a measurement uncertainty of 0.06 µV/V.

[Dr. Frank]

Dr. Frank:  I won't argue with you on your results heater resistor ratio influence - but from a customer test we did 5~6 years ago on 10ea LTZ1000ACH - in a lab with access to JJ-array (measurement uncertainty ~0.3ppm, which is about the best possible we can expect anywhere) - we found the datasheet is closer to being correct.  We found the heater resistor -ratio- had to change at least 103~105ppm to deliver a solid >1ppm change on output voltage, at 25° ambient, PWW 3ppm TCR resistors, mounted on our 1.5oz copper board design inside our application enclosure with semi-insulated LTZ can.  Again this is one of those measurements where all environmental elements comes into play.  Your mileage may vary.

Even if we take your more conservative estimate of an LTZ Vref delta attenuation factor of 76 lower than the heater resistor ratio change - these resistors will be fine for most applications - especially given the fact the basic LTZ stability as ppm drift is down in the mud when it comes to most real lab's measurement capability (Usually better labs are at 2ppm uncertainty for voltage).  It all depends on the expected operating temperature range of your final application.

The take-away here is that for a typical LTZ 7V datasheet circuit - expensive Vishay's won't buy you much of anything in terms of overall profitable cost / performance benefit, and that foil will always get you a bit more noise over PWW's (although that noise effect is difficult to measure on LTZ output, but is there).

Hi MisterDiodes,

there were four people in the LTZ thread, who virtually did the same measurement, like you, I assume, i.e. modifying these both resistors by some percentage, and measuring the LTZ1000 output deviation.. this varied from 74 (Andreas) to 105 (somebody on bbs38hot).
So 74 was worst case.

I just mentioned that, because even using these worst case T.C. numbers, the effect and the optimizing of the resistors T.C. are of 2nd order importance only ..
The residual T.C. will always be below 0.05ppm/k, and may be disappointingly too high, if you set your money on expensive ultra - low - T.C. Vishays, or T.C. matched sets.

The LTZ itself, i.e. up to now undisclosed / not yet explained effects, will play an important role, so a T.C. trimming to << 0.05ppm/K might be accomplished otherwise, e.g. by this T.C. compensating resistor

What interests me, in this context, how did you characterize the T.C. of your LTZ1000 devices, as most of the gear accessible to us amateurs has much higher T.C.?

Do you know  a possibility to accomplish that, say trimming to 0.01ppm/K , w/o the aid of a JJA standard?

Frank 

[TiN]

I've started step test. Here's python code to take 3000 samples on each step, with steps 28C->40C->20C->30C->40C.
Also updated graph data to bit slower 3rd ramp up and even more slower ramp down 40C-28C.

Code: [Select]

while cnt <= 400000:
    if (cnt < 3000):
        meter2510.deduct_tmp(28)
    if (cnt >= 3000) and (cnt <= 6000):
        meter2510.deduct_tmp(40)
    if (cnt >= 6000) and (cnt <= 9000):
        meter2510.deduct_tmp(20)
    if (cnt >= 9000) and (cnt <= 12000):
        meter2510.deduct_tmp(30)
    if (cnt >= 12000) and (cnt <= 16000):
        meter2510.deduct_tmp(40)

        get_THP() # Read Temp, RH, Press from BME280 sensor
        meter2510.get_data()
        k10v = k4.get_data()
        k16v = k6.get_data()


732B cal'd directly against Fluke Everett's JJA in January and they reported a measurement uncertainty of 0.06 µV/V.

That is measurement uncertainty, but your final 732B uncertainty at your site is higher, because it's sum of all uncertainties and 732B stability. Per 732B spec your end uncertainty at the box connector is +/-0.36 ppm/30 days, or +/-0.86 ppm/30 days or 2.06ppm/year assuming that there was no output change due to shipping, and conditions of your lab and Fluke's calibration facility are exactly the same.