T.C. measurements on precision resistors

[babysitter]

Hey Branadic, go get your thermal camera to Andreas and figure out the time how long it takes for a given power to heat up the shell/ cool after a given (even tiny) power pulse/how hot it gets/etc. !

[TimFox]

A former coworker had worked for a company who always used a very green color for their panels.  My buddy dummied up a unit with a more "normal" looking combination of metal and black and showed it to the company president.  He merely replied, "Green is the color of money", and kept the panel color.

[janaf]

Andreas, do you normalize all measurements to 25C? I should use the same as you 

[Andreas]

Hello,

first results from Ultrohm Plus resistor (UP805) 1K #1  0.1% +/- 3ppm/K
shipment date 1510 (not printed on device).

First the pictures of the test fixture.
keeping the wires (solder junctions) as good as can to the same temperature.
Similar to the latest Z201 test fixture.
Remembering that most of the heat is transferred by the wires.

Measurement of 23.03.2015

T.C. Box :  -1.147786492 ppm/K
T.C. LMS @ 25 deg: -1.14174590659788 ppm/K
max deviation from LMS: 2.62806722886730 ppm (Hysteresis + noise)

Measurement of 24.03.2015 (reverse polarity of reference voltage)

T.C. Box: -1.118012147 ppm/K
T.C. LMS @ 25 deg: -1.15136479955225 ppm/K
max deviation from LMS: 2.22772720963622 ppm (hysteresis + noise)

Measurement of 25.03.2015 (normal polarity again)

T.C. Box: -1.153801472 ppm/K
T.C. LMS @ 25 deg: -1.14061584826123 ppm/K
max deviation from LMS: 2.35918437692836 ppm (hysteresis + noise)

On the measurement of 23.03. it was not clear wether there is some ageing drift or if it is only some creeping effect (hysteresis) at lower temperatures.
but since the measurement of 25.03. has only 0.5 ppm difference at 25 deg LMS offset value (which is within hystersis) I think that it is not ageing.

The hysteresis of this sample is comparable to my UPW50 resistors.
T.C.  (1.15ppm/K) is well below the spec of 3ppm/K.
And better than most of my UPW50 samples.
(and well below the 8G16 resistors of Frank).

@Edwin: There remain following questions:
Is the temperature treatment to 150 deg a standard procedure or only on request?
In the datasheet the wattage of the 805 style resistor is not given.
Which material is the connection wire? (copper?)
Is it possible that the epoxy still hardens? Is it hardened thermally or by a hardener?

with best regards

Andreas

[Andreas]

Andreas, do you normalize all measurements to 25C? I should use the same as you 

I do this only for the LMS calculation.
(and the difference to LMS).
For me the T.C. around 25 deg is the most important.

In metrology you would use 23 deg. (+/- 1 .. 5 deg).
Since the temperature in my "lab" ranges from 18-32 deg (winter/summer) I use the 25 deg.
All components (resistors, references) are usually also normalized to 25 deg.

And this is mainly due to the fact that the mathematical (rounding) errors
during LMS calculation are lower when I normalize the values near zero.
Although this is usually not relevant for 3rd order approximations with double/extended floats.
But if you do a 5.th or 6.th order approximation you will see differences.

You can use what is more convienient for you.

With best regards

Andreas

[janaf]

first results from Ultrohm Plus resistor (UP805) 1K #1  0.1% +/- 3ppm/K
shipment date 1510 (not printed on device).

So your has about -1.1ppm/C while mine has something like +0.7ppm/C at this temperature range 

[Andreas]

@Andreas:

You should probably move the sense wire closer to the body of the resistor-- Edwin's spec is 3/8-inch I think [but check the data sheet].  Copper has about 4000ppm/K TCR, and because this is a lower value resistor [1K?], this might affect your data a little bit [and certainly in the sub-ppm range].

I'm not believing the hysteresis, I think that there is a time lag between what temperature the resistor is at and the reading from the temperature sensor.  I could be wrong, but if you sweep the temperature slower and pulse the drive current on the resistor [and sensor if it takes current] only while taking a reading, then I think they will track better.

That said, I think Edwin's resistors are showing very well so far...  Time for me to order some...

Hello Ken,

as Frank already calculated for his 120 Ohms resistors the wire lenght does not really contribute to the T.C.
Edwins wires are more stiff than others. Actually the wire diameter is 0.8 mm (AWG20) so having 33 uOhms / mm.
so if I would shorten the wire lenght by 30 mm on each side I would get a lower resistance of 2 milli Ohms.
With respect to the 1 K resistor the +4000ppm/K contribute only 1/500000 to the total T.C.
So for the 1K resistor please subtract these 0.008ppm/K from the result (giving rounded -1.16 ppm/K instead of -1.15 ppm/K)

Since I cannot reduce further the ramp speed I will do a accellerated test today with 0.3 K/minute instead of 0.12K/minute.
(just to see if the temperature lag increases).

With best regards

Andreas

[ltz2000]

The cal sheet for my Wekomm standard
http://www.eevblog.com/files/WekommResistanceStandardCalSheet.pdf

Kalibriercentrum Bayern. Not a primary lab, but very high capability. Volt/ohm nut friendly???

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

[Edwin G. Pettis]

Andreas (reply 416),

The standard process thermal conditioning which exceeds 150°C is done on all resistors, this reduces aging and drift to below the given specs.  There are additional processes which I can apply to further reduce such effects, some customers do require such processing but most do not as the standard stability exceeds most other PWWs to begin with.

That power rating of the 805 bobbin is 0.33W @ 125°C ambient.

The leads are oxygen free copper, tinned.

The epoxy uses a hardener, the manufacturer recommends room temperature curing but also specifies a 60°C bake if time is short.  The epoxy sealant on the end uses an epoxy based color ink which does recommend a short bake @ 121°C for two hours or 150°C for 15 minutes, this colored epoxy is the same stuff used for printing on the resistors.  I'm curious, why do you ask?

[Andreas]

The leads are oxygen free copper, tinned.

The epoxy uses a hardener, the manufacturer recommends room temperature curing but also specifies a 60°C bake if time is short.  The epoxy sealant on the end uses an epoxy based color ink which does recommend a short bake @ 121°C for two hours or 150°C for 15 minutes, this colored epoxy is the same stuff used for printing on the resistors.  I'm curious, why do you ask?

Hello Edwin,

oxygen free copper: you could sell them to audiophiles for a much higher price. 

I only asked because on other resistors up to now I have seen potrusions from baking the epoxy.
On your resistors I see shrinking epoxy like on some capacitors.
I am just curious to find out how the components are built.

On the first day of measurement I thougth that there could be some ageing drift due to hardening of the epoxy.
But since I measure only 0.5 ppm during 3 days (cycles) which could also be on the reference resistor or
due to hysteresis I do not think that there is a significant ageing drift.
I think I will let the resistor in the thermal chamber and do one comparison measurement after all the other 1K resistors.

And the power rating should be on the invoice the next time in germany (just in case the customs asks).

With best regards

Andreas

[TimFox]

If nothing else, OFHC copper can handle many more mechanical cycles of bending than can ETP copper.

[janaf]

I have seen the comments on hysteresis or not. I do not believe they are time lags. For example the Z #7, the difference in temperature domain is at maximum around five degrees between up / down. It is imo very unlikely that you have such a big temperature difference with the type of setup and slow ramp that you have. Could you simply stop the cycle at mid point, both up and down, let it stabilize for an hour to see if the effect is still there or not? 

[Andreas]

I have seen the comments on hysteresis or not. I do not believe they are time lags. For example the Z #7, the difference in temperature domain is at maximum around five degrees between up / down. It is imo very unlikely that you have such a big temperature difference with the type of setup and slow ramp that you have. Could you simply stop the cycle at mid point, both up and down, let it stabilize for an hour to see if the effect is still there or not?

I already did this experiment with a UPW50 resistor: (see page 1)
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg462300/#msg462300
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/?action=dlattach;attach=98751

and also with Z201#3
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg606730/#msg606730
https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/?action=dlattach;attach=135525

even with 120-200 minutes (2-3 hours) settling time there is a remaining offset.
It needs one whole night to reach the previous value.
For the Z201 in the data sheet a thermal settling time of 1 second is specified.
So after 1 minute the resistance should be settled within 0.001 ppm or less.

With best regards

Andreas

[acbern]

The cal sheet for my Wekomm standard
http://www.eevblog.com/files/WekommResistanceStandardCalSheet.pdf


Kalibriercentrum Bayern. Not a primary lab, but very high capability. Volt/ohm nut friendly???

Well, my experience, pretty expensive, and thus not volt/ohmnut friendly in my view. I did a search some time ago in Germany for calibration of 10k and 10V standards. Got many quotes and found that ESZ is pretty cost efficient for both (ISO certificates), and when asking for an ISO certificate they can certify accuracy limits below 0.5ppm for both, so I use them for my standards. And: I can drive by and drop off my standards and pick them up, thats good too (and from time to time have a little chat with the lab guys). For thermal converters I use Testo, because they have an automated procedure, so very good accuracy and at the same time cost efficient because few labor hours involved. RF power I do at Rohde/Schwarz, acceptable prices and they can adjust the EEPROM in their power heads.
These are all the calibrations I need.

[Edwin G. Pettis]

Hi Andreas (reply 423)

All resistors have many interacting mechanical forces on them, they are complex in nature and like most things, decreasing the effects of one force may amplify another.  While the mechanical forces in PWW and film/foil resistors share many common effects, there are differences because of the different types of construction as well.

PWW have hoop stress from winding the wire onto a bobbin, film/foil do not but film/foil have a similar type of stress from being attached to a ceramic substrate.  Encapsulation stress is present in both types for different reasons, both have partial remedies.  Film/foil are sensitive to moisture, partly due to absorption through the encapsulation and around the leads, mine have no significant affects.  Film/foil are particularly sensitive because the resistive element is bare and the circuit paths are very small and close together allowing water molecules to bridge them easily, baking is a temporary fix.  Foil/film resistors have an applied voltage coefficient, PWWs do not, again this is a artifact of film/foil construction.  Both types of resistors use very similar alloys, the basic aging rates are normally nearly identical.  Drift rates are complicated, mainly the result of residual mechanical stresses which are very difficult to completely remove.  This involves additional 'enhanced' (as Vishay likes to call it) processing to further reduce those stresses, i.e. costs more.  Oil filled hermetic sealing tends to be more beneficial to film/foil as it removes more of the external sensitivities that film/foil has to their environment such as moisture and barometric pressure which PWW resistors have insignificant sensitivity to.  One other important characteristic, film/foil are noisier than PWW resistors, particularly mine.

"I only asked because on other resistors up to now I have seen protrusions from baking the epoxy.
On your resistors I see shrinking epoxy like on some capacitors." - Andreas

I am not sure what is causing the protrusions you are seeing on the other resistors, it may be caused by a filler they are using in the epoxy.  The epoxy end 'seal' on my resistors is purely mechanical in nature, to hold the resistor inside the shell and support the lead assembly, the epoxy may not form a complete seal around the metal lead, this is of no concern as it does not affect the resistor's electrical characteristics.

Aging tends to be of a nonlinear nature, as exhibited in the better resistors, aging (which is a form of drift) tends to decrease with increasing time reaching a fairly level, stable rate after some period of time.  Just how much the initial rate is and the eventual stable rate becomes depends on the resistor technology and quality of construction.  As I've pointed out before, the more 'perfect' a resistor you want, the more expensive it is going to be and that cost tends to rise exponentially.  You are also going to rapidly run into the always present law of diminishing returns, that last little bit of improvement becomes so expensive, it isn't worth the trouble.

Speaking as an design engineer, respectfully:

Fluke, HP (Agilent, Keysight), ect. all recognise this, that is why their voltage references perform to a certain level for that cost because it is a happy balance.  It is much cheaper to have those references checked against a primary standard periodically than to try and put a primary standard inside their instruments.  Would anyone be willing to pay say five or ten times more for a 3458A with a primary standard inside?  I really doubt it, the performance of the LTZ1000/A with good resistors around it provides the necessary performance at a reasonable price.  The Fluke voltage standard ups the anti by using multiple LTZs to improve the performance but those aren't cheap and no one is going to make a single LTZ perform as well as that Fluke standard no matter how good the resistors are.

Those LTZ resistor sensitivity numbers given in the data sheets/apps are only approximate as stated clearly there, every LTZ is a bit different.  If you want to know about this stuff, just ask Linear Tech, the people who designed the chip.  If there were any relatively 'easy' tweaks that could be done to improve the performance, don't you think Linear Tech or Fluke would have already discovered them after all this time?

[Andreas]

If there were any relatively 'easy' tweaks that could be done to improve the performance, don't you think Linear Tech or Fluke would have already discovered them after all this time?

Hello Edwin,

perhaps they have already discovered some tweaks. But they have to earn money with their products. Some tweaks may be so time consuming that they will not get paid for that.
For me its a challenge to find out how the components work in ppm ranges. And to see how far can I go with self made equipment to create stable references.

With best regards

Andreas

[Andreas]

Hello,

as promised some further measurements

on 26.03. I did a "fast" measurement with 0.3K/minute to see if its temperature lag of the NTC-sensors which creates the hysteresis.
If its temperature lag the hysteresis should increase significantly which is not the case.

on 27.03. a normal slow measurement 0.12K/minute again.
Interesting is that the hysteresis/creeping effect at low temperartures
 is decreasing significantly after several temperature cycles.
(compared to measurement of 22.03.)
So my guess is that there is some humidity effect.

Today I am running the first cycle of UP805#2 1K resistor.

with best regards

Andreas

[janaf]

If there were any relatively 'easy' tweaks that could be done to improve the performance, don't you think Linear Tech or Fluke would have already discovered them after all this time?

Hello Edwin,

perhaps they have already discovered some tweaks. But they have to earn money with their products. Some tweaks may be so time consuming that they will not get paid for that.
For me its a challenge to find out how the components work in ppm ranges. And to see how far can I go with self made equipment to create stable references.

With best regards

Andreas

If there were any relatively 'easy' tweaks that could be done to improve the performance, don't you think Linear Tech or Fluke would have already discovered them after all this time?

Hello Edwin,

perhaps they have already discovered some tweaks. But they have to earn money with their products. Some tweaks may be so time consuming that they will not get paid for that.
For me its a challenge to find out how the components work in ppm ranges. And to see how far can I go with self made equipment to create stable references.

With best regards

Andreas

Edwin, Andreas,

LT would have an interest if they did not already "own" this market. No competition in sight. Fluke have no interest spreading their knowledge but obviously have the knowledge to build standards approaching 0.1ppm/year stability.

As Andreas wrote in his first post in this thread, his ambition was to see if it's possible to select & match resistors so they cancel errors. I think that is an approach that the commercial actors might not be able to do, just about nobody would be prepared to pay the cost. I have never seen a survey like the one Andreas is doing. I certainly learned a lot.

The measurements I have done on the LTZ1000ACH, I have not seen elsewhere. They seem to show systematic and significant differences compared to the datasheet (repeated on several different devices by me and by others). Who knows where this ends.................

[Edwin G. Pettis]

Jan,

Just what systematic and significant differences are you talking about?  If you are referring to the data table in the data sheet/app that shows the effects resistors have on the output of Vref, those are approximate and nothing more, they are going to differ from chip to chip and there are no specific limits put on them.  There are differences in all of the parameters of the LTZ, some do have specified limits, others do not.  There are intrinsic differences, chip to chip, which cannot be compensated for by external components, long term.  You might be able to do a tweak which will reduce drift temporarily but it will not be permanent.

Basically what Fluke is doing is well known, they are aging and measuring their LTZ chips over a long period, then the chips characteristics, within possible limits, are selected and put together in multiple LTZ summing circuits, just like the technique that was used with the LM399/As years ago.  This time consuming method can and does produce a unit device which can produce stabilities in the neighborhood of the 0.1PPM you are quoting, that may actually be on the better side of average frankly.  The long term stability of an LTZ varies, chip to chip and cannot be predicted accurately, it will settle down to within a given smaller drift range over time which, with multiple averaged chips, could produce the 0.1PPM but with rare exceptions, no single LTZ chip can produce such stability, certainly it could be less than 1 PPM/year in time and may even drop below 0.5PPM / year with more time but otherwise the plain answer is no.  Even if you put perfect zero TCR resistors around an LTZ you are not going to get any better drift rate than the intrinsic rate of the chip.

You are right in that specific aging techniques have been found over the years to age the LTZ chips faster than just letting them sit powered up for many months, from what I hear it is relatively involved and also takes quite some time as well, months.  You cannot subject the chip to too much abuse without causing unwanted changes, for that matter, you cannot subject most components to abuse without causing unwanted changes.

It would seem to me that a Vref stable to around 1 PPM/year that cost less than $100 has more than enough stability to meet most demands.  I would be putting more time into figuring out how to get an stable  ~10V output from this chip that is close to comparable to the LTZ's stability.  There are ways of achieving this, besides, contrary to common view, a 10V standard does not have to have an absolute 10.000000V output, only a stable output, it is easy to calibrate the scale to that offset and it isn't any less accurate, maybe not as easy as an absolute 10V but no cal lab usually has an absolute 10V standard, your DVM is calibrated to that offset voltage in the lab, simple calculator math.

[janaf]

Edwin,

Lots of things to reply to. The short version would be; please see my previous posts, the last one from the day before yesterday. But here is and example:

The datasheet states an attenuation of 100 of "R1" the current setting resistor. I have measured, on three different devises from two batches, two different board designs, an attenuation of -700. This has also been done by one Chinese poster and verified by TiN on this forum. We all measure an attenuation of near (-)700 (+/-5%) on five different ICs , versus the datasheet +100. I call that a significant and systematic difference.

Another undocumented aspect is how the time drift over temperature looks. We pretty much know that lowering temperature lowers drift, but by how much, what's the spread? I honestly don't even know the sign or if there is one or not. The measurements have also pretty much shown that these values are "known" errors, not uncertainties in the sense that the attenuation is near -700, not +/-100 as you might get the impression in the datasheet. The same is true for the other resistors in the LTZ100ACH circuit, but the differences are less dramatic.

I do not rule out the possibility that the datasheet data is valid for the thermally un-insulated LTZ1000CH.

If, like Andreas, you are trying to match TCR of resistors to the circuit based on the info in the datasheet, you'd definitely be mislead.

Fluke: I'd compare to resistors, it's "well known" how to make them but the devil is in the details, and surely include in-house secrets and lots of details and experience that take you from "good enough" to "the best". To my knowledge, the Fluke standards use a single LTFLU reference each, while it's common for large institutions to have four 732B units in one rack. The now obsoleted 7001 used a single LTZ1000 per module but was also available in multi-unit racks.

10V; yes surely looks a bigger challenge than "only" the raw LTZ1000 circuit output voltage. For some the 7.xx volt stability may be enough, for others approximately 10V is enough while others may require 10.000 000 0
volt.

In the end, what we do depends on our objectives. On a tight budget? Want to learn something? Volt-nut? Need a specialized reference? Are you happy with the 4ppm/year of the 3458A? Then fine. You can buy a stand-alone board on Ebay. If not, simply lower the temperature, maybe up-grade resistors and decrease drift to 1ppm-ish/year as many have reported.
 
I'm in for the learning. I'd like to get down to the nitty gritty details of this. It would have been a lot cheaper for me to buy a used 732B if I "only" wanted a good voltage reference.

[Andreas]

Hello,

Results of 2nd 1K resistor UP805#2:  shipment date 1510
28.03.2015 first measurement unfortunately had problems with communication connection of one ADC so started too late for a complete cycle.
29.03.2015 reverse polarity measurement. Due to switching to daylight savings also one hour shorter than usual.
30.03.2015 normal polarity like on 28.03.

Evaluation of 30.03.2015

Box: 37.70573087 ppm / -32.51923093 K = -1.15949024 ppm/K

LMS evaluation:
A 0 = -3.89958540734108E-0001
A 1 = -1.20262068020544E+0000
A 2 = -2.06316059338346E-0002
A 3 =  5.13160024309751E-0004

so T.C. at 25 deg is -1.20 ppm/K
max. deviation from LMS: 2.87036703474386 ppm (Hysteresis + noise)

So values are very similar to UP805 #1 with slightly higher T.C. and hysteresis / creeping effect.

https://www.eevblog.com/forum/projects/t-c-measurements-on-precision-resistors/msg637341/#msg637341

With best regards

Andreas

[Edwin G. Pettis]

@DiligentMinds,

Looks can be deceiving, the devil is in the details of the resistor construction and manufacturing processes.  I've been busy processing a complicated set of resistor orders and they have to come first of course.

First hand data on what you are asking for is quite limited, secondary information is limited to the data which was provided by other testing labs at potential customers.  The data sheet I posted earlier on the Ultra Precision Reference thread gives considerable indication as to long term accelerated aging, I refer to the 50 cycles of thermal shock (1500 component hours) MIL-STD-202 treatment showing unprecedented small changes in resistance, as is the norm, changes decreased with time.  The amount of change did vary slightly depending on the bobbin style and the winding operator's (one of the significant mechanical variables) abilities.  Unfortunately, the lab only provided the cumulative data figures and no individual resistor data.  As pointed out in the data sheet, one resistor bobbin style had unusual shifts which were later determined to be an error in the selection of wire size, the bobbin was slightly overfilled with wire.

The first hand data comes from a set of matched resistors that Bob Pease had requested 5+ years ago, as is my usual practice, extra resistors were wound in case any were lost during production.  The resistors were 200R0, 807 bobbin (0.250" D x 0.750" L), two resistors turned out within less than 2 PPM of each other  (tolerance), the other two were within <10 PPM.  I sent the closer matched pair to Bob and kept the other set, over the years my two resistors have stayed within the measurement uncertainty / accuracy of my 242D which is always kept calibrated to capability, not specs.  This performance was the result of the enhanced processing carried out in addition to the standard processing of my resistors, I cannot guarantee this level as I do not have accumulated physical detailed data beyond what I've mentioned here.  The long term drift under enhancement is less than half the standard long term drift factors which I've mentioned here as well.  The first hand and second hand data tends to indicate that long term drift appears to be under 2 PPM/year or less. for most bobbin styles.  Without further hard data I would not be willing to absolutely say that enhanced resistors will have essentially no drift as my resistor set has demonstrated, it is too small a data set to base a specification on at this time.

All of the reference resistors I make are enhanced, roughly, a 10K, ±0.005%, 0±3 PPM/°C, 807 bobbin is approximately $12.  Other specifications are additional such as a tighter tolerance, lower TCR, certificate of calibration, etc.

[Andreas]

Hello,

Results of 3rd 1K resistor UP805#3:  shipment date 1510
31.03.2015 first measurement
01.04.2015 reverse polarity measurement.  (no april hoax)
02.04.2015 normal polarity like on 31.03.

Evaluation of 02.04.2015

Box: 28.92999192 ppm / 32.37332611 K = -0.893636688 ppm/K

LMS evaluation:
A 0 = -1.23983866206313E-0001
A 1 = -9.09276513282291E-0001
A 2 = -2.23106751095097E-0002
A 3 =  4.32564638954812E-0004

so T.C. at 25 deg is -0.91 ppm/K
max. deviation from LMS: 2.42265360986628 ppm (Hysteresis + noise)

so this resistor is below 1ppm/K

With best regards

Andreas

[Andreas]

Hello,

Results of 4th 1K resistor UP805#4:  shipment date 1510
03.04.2015 first measurement
04.04.2015 reverse polarity measurement.
05.04.2015 normal polarity like on 31.03.

Evaluation of 05.04.2015

Box: 35.77847918 ppm / 32.16232419 K = -1.112434504 ppm/K

LMS evaluation:
A 0 = -3.05581250800723E-0001
A 1 = -1.12251269574014E+0000
A 2 = -2.22106743141866E-0002
A 3 =  4.18486182138598E-0004

so T.C. at 25 deg is -1.12 ppm/K
max. deviation from LMS: 2.04536528569842 ppm (Hysteresis + noise)

so this resistor T.C. is again very near to #1 + #2

While the other resistors where below 100ppm against my Z201#6 in resistance value.
This one has around +150 ppm offset.

With best regards

Andreas

[Andreas]

Hello,

Results of 5th 1K resistor UP805#5:  shipment date 1510
06.04.2015 first measurement
07.04.2015 reverse polarity measurement.
08.04.2015 normal polarity like on 06.04.

Evaluation of 08.04.2015

Box: 35.28859792  ppm / 32.40994882 K = -1.088819921 ppm/K

LMS evaluation:
A 0 = -2.23342937070244E+0000
A 1 = -1.13611717010502E+0000
A 2 = -1.86452762076308E-0002
A 3 =  4.20604123685255E-0004

so T.C. at 25 deg is -1.14 ppm/K
max. deviation from LMS: 3.36389728597688 ppm (Hysteresis + noise)

so this resistor T.C. is again very near to #1 + #2 + #4
But hysteresis is rather high and the ageing drift during the 3 days was around -2.2 ppm
Offset against Z201#6 is around +370 ppm
Generally the "creeping effect" during cold cycle decreases over the 3 days.

And again a overview over the 1K resistors.

With best regards

Andreas