T.C. measurements on precision resistors

[Vgkid]

I completely missed the Fluke 8508 in the stack, and immediately looked at the rack. With the 2 displays, they always remend me of a power supply.

[HighVoltage]

Sure ... basically the PTB are using a Fluke Transconductance Amplifier controlled by a Wavetek calibrator to generate the power. Then they have a range extender from MI (that blue box you can spot) to measure the current via a reference resistor (in the silver box). That is compared to the current through the bigger resistor.
The smaller 1Ohm resistor is used as a short to keep the current flow constant, when the bigger resistor is not loaded. That just helps stabilize the transconductance amplifier.

Thanks for the high resolutions pictures, it is really a pleasure, looking at them.
Interesting things to notice:
Fluke 732A, marked "BUND" (German military) but label partially covered.
Nice copper bar for ground connections, sitting on Styrofoam.
May be the 34401A is used to measure the temperature through resistance?
Some of the 4 mm banana plugs / cables do not look like metrology grade.

Great opportunity for you to go to their lab.




 

[Edwin G. Pettis]

To bring up an earlier discussion on this thread about ratio measurements using the 3458A (or any other DVM for that matter), this paper from Agilent/Keysight;

http://literature.cdn.keysight.com/litweb/pdf/5992-1058EN.pdf?id=2643219

lends some information about the accuracy and uncertainties of ratio measurements.  Two very important pieces of information about this type of measurement, the reference must be a known, accurate quantity including its uncertainty of accuracy and secondly ratio measurements consist of two sequential measurements, thus the effects of time and environment enter into the measurements, plus ratio measurements are very sensitive to noise.  While the filter function can reduce this some, it cannot eliminate it, particularly when the noise is varying.

[Edwin G. Pettis]

If you have a known accurate reference (and its uncertainty), within the parameters that apply to your DVM, you can still get a ratio that will be a bit more accurate than your DVM's actual accuracy as set out in the Keysight paper.  Just be very careful of noise pickup, ratio modes are quite sensitive to it and even with a filter on, you can expect an additional drop in actual ratio accuracy because of it.  A ratio is made up of two independent readings made at two different times (sequentially), even if the linearity of the ratio mode were perfect, the readings would still have error attached to them.  According to the engineers at Keysight, that could double the error of the readings.

[3roomlab]

If you have a known accurate reference (and its uncertainty), within the parameters that apply to your DVM, you can still get a ratio that will be a bit more accurate than your DVM's actual accuracy as set out in the Keysight paper.  Just be very careful of noise pickup, ratio modes are quite sensitive to it and even with a filter on, you can expect an additional drop in actual ratio accuracy because of it.  A ratio is made up of two independent readings made at two different times (sequentially), even if the linearity of the ratio mode were perfect, the readings would still have error attached to them.  According to the engineers at Keysight, that could double the error of the readings.

sorry i deleted the question
the question was : it seems using ratio method, an uncalibrated DMM can also make useful measurements?
then i thought, no way, impossible (deleted the previous post  )

now this is getting more interesting for me as i dont intend to calibrate my used DMM (as in send it to the fluke shop)

[Edwin G. Pettis]

In a manner of speaking, it does improve on the uncalibrated accuracy of your DVM, within limits, but it really isn't a substitute for a calibrated instrument.

[3roomlab]

well i have nothing back in working condition to do any measurement
i still have a few more things to try/mod to reduce noise
just curious what kind of noise level are we looking at when we say a good external ref and a DMM to use for this ratio thingy?

[acbern]

You need to have min 100NPLCs and average the result of a few measurements (or do more NPLCs, because the 3458A avarages lower count NPLC mesurements when you choose high NPLC numbers, details see manual). Typically I see the last digit jump on ma 3458As. No EMI surces arround, thermally stabilized cabling... Of course, any other meter has its own noise specs, so noise realy depends on meter used.
And yes, an uncalibrated 3458A can be used as a precise meter if you have a precise reference as the basis. I am aways doing this even with my calibrated 3458As because it leads to lower inaccuracies (a 732A/B is way more stable than a 3458A).
Now the problem is that few meters have transfer accuracies specified. So you need to use the 24h value then, which immediatelly makes uncertainties worse. Even the 34420 does not have it specified.

[Andreas]

Hello,

again some overview sheets of measurements.

For the 12K5 and 12K resistors I recognized that the AC excitation of the AC-Multiplexer
which I introduced on the 1K resistors to avoid thermal EMF voltages gave strange results.
(see also the discussion of Frank on his HP3458A).
https://www.eevblog.com/forum/testgear/precise-offset-compensation-ohm-measurements-and-validation-of-dmms/msg834537/#msg834537

So I did the measurements again with the DC-Multiplexer.
Due to the mounting of the resistors onto a metal sheet the thermal EMFs play no role during measurement.

From the 12K5 resistors all UP805 resistors specced <=3ppm/K were below 1ppm/K.
The Z201 resistors (typical spec 0.2ppm/K) play in the same ball park.

From the 12K0 resistors (2 different batches) the second batch are all below 1ppm/K.
#6 shows a "open hysteresis curve" (does not return to the old value during measurement).
It needs over night to return to the old value.

The 2 resistors from the first batch have higher values of T.C:

With best regards

Andreas

[Andreas]

Hello,

further measurements: this time on 70K resistors.

Measurements get more difficult.
With my standard setup (1.5m long twin lines) the noise increased from 10-15 uVpp to more than 40uVpp.
(10uVpp is the noise floor of my ADC for a single measurement, with averaging over 1 minute I go below 1uVpp).

So I decided to use shielded cables to go back to around 15 uVpp.

The Z201 70K resistors all measured below 0.5ppm/K
The UP805 70K resistors with shipped 1510 were between 1-2 ppm/K
and the 1532 shipped below 0.8 ppm/K

with best regards

Andreas

[Andreas]

Hello,

measured the 2nd batch of UP805 1K resistors with datecode 1532.
T.C. is around -1.2 ppm/K.

Overview attached.
next will be this undressed guy.
Lets see if the missing epoxy has a influence on hysteresis.

With best regards

Andreas

[Vgkid]

The Vishay VAR, this will be interesting.

[Andreas]

Hello,

and here the Vishay VAR results of my 2 samples.
Datecode B1532

Even from the same batch the stray of VAR is rather large compared to PWW-resistors UP805.

When regarding the drift over temperature there are 2 special things.
The hysteresis curve behaves like a "eight" and is not smaller than on other resistors.
(I hoped that the hysteresis on this resistors is near zero.)

The drift over one day (on the cold side) is around 5 ppm even at the same temperature.
Over night the drift restores almost fully so that the game at the next day is the same.
At sample 2 there is a remaining drift of nearly 2 ppm.

As comparison the hysteresis of a UP805 1K (#15) resistor.
and one of a Z201 1K (#5).

With best regards

Andreas

[Andreas]

Hello Ken,

I am shure that it makes a difference between continous ramping and only measuring the 3 interesting temperature points.
The reason is: it takes some time until the epoxy creeps (due to stress) and absorbs / is freeing the humiditiy.

For me the question is: what is more realistic for the application.
For that I would have to even ramp more slowly. (minimum a factor of 5-10).
But then the testing time gets longer than one day. (Which I do avoid).

With best regards

Andreas

[acbern]

Re. Z-foil humidity drifts, I had initially used standard (non hermetic) as references, and have seen drifts of about 5-10ppm over the year. That was the end of these references... The hermetic ones were up to Vishays spec (2 ppm in 6 years, low load, iirc).
So if you want to use standard z-foils in the voltage divider (multiplier) of an LTZ design to step up to 10V, they are probably not worth it. I have not done any tests with dissicators or putting them into oil, which my be worth a try, so cannot comment.
The other option /besides hermetic resistors/dividers) that may be more stable is dividers in one package. This way the humidity impact affects both resistors on the substrate (made from the same material) the same way and one would expect a more symmetrical behaviour.

[zlymex]

I have a lot of foil resistors to test their TC, I'd like the connections to be reliable, repeatable, and no soldering involved.
I have a 16 by 4 scanner that is very good to test 15 resistors at the one time, speeding up the process quickly.
However, how to connect resistors at the chamber has been the trouble for me.

While surfing this thread about 'ugly' wires, 'ugly' heat, I thought there may exist a 2-wire socket for one lead. I then went thru my stuff and found out a kind of socket that allow the leads go through and insert to another socket at the bottom as well. This will make 4 wire connection required.
The bottom socket will soldered onto a PCB and I'll find some way to firmly attach the top one before making fly wires to the side of the PCB.

[zlymex]

The other option /besides hermetic resistors/dividers) that may be more stable is dividers in one package. This way the humidity impact affects both resistors on the substrate (made from the same material) the same way and one would expect a more symmetrical behaviour.

That's the idea. I have ordered some VHD200 from Vishay that both hermetic and in the same package.

[Andreas]

The bottom socket will soldered onto a PCB and I'll find some way to firmly attach the top one before making fly wires to the side of the PCB.

Hello,

Perhaps you can reverse orientation (upside down) for the upper connector and use a 2nd PCB.

Measurement may work if you have a instrument that compensates for thermocouple voltages.
The 2nd problem is to sense the resistor temperature exactly.

With best regards

Andreas

[zlymex]

Hello,

Perhaps you can reverse orientation (upside down) for the upper connector and use a 2nd PCB.

The reversal seems to be a good idea but it doesn't work in reality, they allow one way insertion only.

[plesa]

While surfing this thread about 'ugly' wires, 'ugly' heat, I thought there may exist a 2-wire socket for one lead. I then went thru my stuff and found out a kind of socket that allow the leads go through and insert to another socket at the bottom as well. This will make 4 wire connection required.
The bottom socket will soldered onto a PCB and I'll find some way to firmly attach the top one before making fly wires to the side of the PCB.

For similar connection I'm using the thru-hole MicroJack connectors made by Keystone
http://keyelco.com/userAssets/file/M65p102.pdf

The plastic in the connectors which you would like to use can be source of error due to the poor insulation resistance and humidity sensitivity.

[zlymex]

The other option /besides hermetic resistors/dividers) that may be more stable is dividers in one package. This way the humidity impact affects both resistors on the substrate (made from the same material) the same way and one would expect a more symmetrical behaviour.

That's the idea. I have ordered some VHD200 from Vishay that both hermetic and in the same package.

Hate to ruin your day, but these don't seem to be consistent in production in regards to TCR tracking.  If you order a big batch, you may find that some of them TCR track very well, while the others maybe not so much.  In addition, there is no guarantee that they will age together.

For a much better ratio resistor set that actually will track, contact Edwin Pettis-- he can make you a set that are made from the same resistance wire that comes off of the supply reel one after another.  This will ensure that the two resistors track very well, and they should age together.  I suggest placing several wraps of copper tape around each resistor individually, then wrap those together with several more turns of copper tape.  This will help keep the resistors at the same local temperature.  If one of the resistors ends up being much smaller than the other, Edwin may suggest that the smaller one be placed in a smaller package-- this will help with tracking.  With this method, it is possible to get near zero TCR tracking coefficient.

Although Vishay is notorious deceiving people in the datasheets for tempco, they do have guarantees for some of their hermetic packages for shell-life of 2ppm/6yr, which is THE most important factor for voltage standards that worth trying out, and seems no other WW manufactures say something about the aging even close.

As for the tempco, foil resistors show much better straight-line curve with typical beta 0.01ppm/C2 or better in room temperature, once compensated(which is easy), the curve will be more flat than even the best standard resistors ever made.

I have two Fluke 731A transfer standards where they suppose to use ratio resistor set with the same technique as you described(same supply reel on one mica stripe), but the performance is not up to expectation. Fluke changed to hermetic WW of individuals in all their later versions(731B, 732A, 732B),.

Furthermore, I have a plan to put more than a dozen voltage references in a small traveling case and Vishay VHD resistors are perfect for that because of their small physical dimension.

[Andreas]

It would be interesting to put your same experiment inside of another box that has a few large bags of desiccant to see if you get the same overnight shift in value.

Hello Ken,

same experiment (12.03.) with a 1 kg (2 lbs) dry pack on ZVAR #2
As comparison the measurement of 25.02. without dry pack.

In the warm phase the hysteresis has changed to smaller values.
But no significant change during cold phase.

Unfortunately due to the large thermal "mass" I do not reach the same minimum temperature.
(only around 15 deg C instead of 13 deg C).

with best regards

Andreas

[Andreas]

Interesting.  I wonder if the greater thermal mass of the desiccant packs was responsible for the difference, and not humidity.  How do we test this?

Hello Ken,

I do not think so. The only difference is on the warm side where the heater is well in regulation so there is no difference in ramp speed. On the cold side the ramp speed is slowed down dramatically. This should give also a dramatically smaller Hysteresis on the cold side which is not the case.

On the other side: for me the pin pointing is not essential. The ZVAR has from my side of view no advantage over a Z201 resistor in precision application (which was my hope). The hysteresis behaviour is even worse. And they are very fragile. The ageing is undefined. -> I will not use them for my purposes.

With best regards

Andreas

[branadic]

The ZVAR has from my side of view no advantage over a Z201 resistor in precision application (which was my hope). The hysteresis behaviour is even worse. And they are very fragile. The ageing is undefined.

Well, putting the Vishay VAR into a bath of Galden such as HT110, ZT180 or similar to prevent humidity effects could be a benefit over the possible mechanical stress Z201 could be influenced by.
Some effects that still haven't been investigated yet is barometric pressure. As was found Fluke 732 showed an influence due to barometric pressure. As far as I remember the paper did not identify where this influence came from, resistor or voltage reference.

However, I would expect a bigger influence with molded package (Z201) compared to raw and unpackaged resistor (VAR). So putting the single resistor in a brass case with feedtrough capacitors and filling the case with Galden could make a good improvement here.

The better solution would be to have a vacuum inside the case. This way there would be no pressure conducting material between case and resistor.

EDIT: Don't we have guys over here that are able to put a VAR inside a glas tube "filled" with vaccum? That would be awesome.

It's not that vacuum resistors have not been existing, but none of the current manufactors had the idea to revive this technic.

[zlymex]

The precision film resistors also have their place-- they have relatively low current noise [than carbon or carbon-film] and are good at the higher frequencies-- even better at high frequencies than VPG foils because the resistors are so thin.  But they are at the bottom of the list if you are looking for the best long-term stability [Exception: Ta2N [Tantalum Nitride] film resistors, which can have excellent long term stability, but only "good" TCR unless highly selected]. 

That's very informative. One thing I'd like to ask is base on this:
J. Pickering must be the master mind behind the Wavetek 7000 voltage reference and he hold several US patents for that as well.
There was an article "A solid state DC reference system"(in NCSL Conf. Dig., 1995, p. 369, he was the co-author), which I cannot find but cited by this article:
"Design of a 10mA DC Current Reference Standard"
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1202069
(I attached an excerpt)

Obviously this is refereed to the Wavetek 7000, and he meant that they use TaN film resistor for the step up from 7V to 10V.

This also supported by his article "Setting new standards in DC voltage maintenance systems"
http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=771967
A practical approach to maintaining DC reference standards
www.elcal.ch/files/11749-eng-01-a.pdf

My question is:
Did 7000 systems really use TaN film resistors?
Are Vishay TDP resistor(that 7000 use)  TaN film resistors?
www.vishay.com/docs/60045/tdp.pdf
(They say it is Passivated nichrome)