Fluke 742A-10K and HP3458A

[Alex Nikitin]

Not to push the Calibration Club thread further off-topic, here I would share some data on my new toy - the Fluke 742A-10K Resistance Standard. I also will copy the photos here.

It appears that the HP3458A resistance measurements have considerably larger temperature coefficient and/or drift without ACAL that the Fluke. I've run overnight test (OHMF, NPLC100, OCOMP ON, DELAY 1) and in the morning the HP3458A did show about +2ppm from the value yesterday afternoon. The temperature variation in the lab was not much and the internal temperature of the meter did change by about 1.2 degree C. Yesterday at 2pm, after ACAL ALL with the internal temperature 38.4C and the resistor temperature 24.9C the measured value was 10.000058K . This morning at 8:40am the measured value was 10.000078K (+2ppm), internal temperature was 37.2C and resistor temperature was 23.0C. After ACAL ALL the HP3458A measured 10.000060 (internal temperature 37.3C, resistor temperature 23.0C).

At least it looks like just after ACAL the measurement values are consistent. I will repeat the measurements at some intervals (possibly weekly or bi-weekly) to see the measured value changes in time .

I also measured the 742A with the Keysight 34465A meter, the measured value just after ACAL was 10.00007K (NPLC100, OCOMP ON)

Cheers

Alex

[HighVoltage]

... and the resistor temperature 24.9C

It looks like you got a stable Fluke resistor.
How did you measure the temperature of the resistor?

Will you open it up and show us the inside?

[Alex Nikitin]

... and the resistor temperature 24.9C

It looks like you got a stable Fluke resistor.
How did you measure the temperature of the resistor?

Will you open it up and show us the inside?

I've measured the temperature with a thermocouple attached to the resistor's side.

And I won't open it. There are photos of the internal construction of this Fluke already available here on the forum and as I plan to use it and have no particular reason to open it, I'll pass.

Cheers

Alex

[Theboel]

I believe this the picture alex talk about.
Just dumb question why paralel conection ?
More watt for safety ? More stable because more area ?

[RobK_NL]

why paralel conection ?

By combining resistors with opposing temperature coefficients, a set can be obtained with a much lower temperature coefficient than a single resistor.

Also, four devices in parallel means less self heating, thus higher stability.

[zlymex]

......
Just dumb question why paralel conection ?
More watt for safety ? More stable because more area ?

@RobK_NL said it all.

As for the might be extended question, why 4 instead of 2? 4 can compensate two parameters while 2 can compensate only one.

Why parallel connection instead of series connection? That will depend on what range of values the designer think are the best. Fluke use two 4.05k in series in 5450A indicating 4k is better than 16k. I really don't know fluke think 40k is better than 2.5k. Another reason for this is mechnically easy to construct.

[Alex Nikitin]

Thank you zlymex, your photos did help me to part with not so small amount of money and purchase the 742A-10K.

Cheers

Alex

[Edwin G. Pettis]

To RobK_NL, el al,

Your answer is not quite correct, these resistors are 'selected' out of a larger batch of resistors possibly made from the same spool of alloy (in most cases), the TCR normally should not vary that much from resistor to resistor although that can be affected by manufacturing processes.  Some manufacturer's end results may have variances of several PPM in TCR.  With sloppy TCR results, TCR selection becomes much more important.  Depending on the alloy's heat treatment, the TCR can have either +, - or ± TCR relative to above or below the 23°C reference.  Most of the TCRs I've seen that are used for these standard resistors fall into the ± TCR category which allows them to do a bit of 'selection' of TCR, as their manufacturing processes tends to produce a small range of ±TCR values.  One drawback of this melding of resistors to produce a low TCR around the reference temperature is the likely hood of nonlinear TCR with temperature, this generally doesn't present too much of a problem within the 'normal' ±3°C to ±5°C range.  Vishay makes use of this technique in part to minimize TCR over a restricted range of temperature but it is rarely a linear TCR curve.  If you have seen the TCR curve of an SR-104, it tends to have a slightly parabolic curve to it around the 23°C reference temperature, this is because it is made up of multiple resistors and each one of them has a nonlinear TCR curve so the end result is a nonlinear TCR curve with temperature.

A single resistor can be made with a very low TCR as well, it doesn't have to be multiple resistors.  Multiple resistors in modern resistor standards are used as more of a convenience than necessity because of the manufacturing procedures.

Self-heating is a result of power dissipation in the resistor, stability is defined as the drift of a resistor's value with time, this can be anything from sitting on a shelf to power cycled, temperature cycled or just on all the time.  Self-heating will slide the resistor up the TCR curve but when removed the resistor should return to the same or nearly the same value depending on the stability with power figure.  Resistor standards, particularly primary types require kid glove care, while power dissipation of 100mW is generally of no concern, power is kept to a minimum to prevent possible significant changes in value, Manganin is considerably more sensitive in this respect than Evanohm standards.  Primary standards like the SR-104's stability is just about the same as shelf stability given the normal care it is supposed to have.  The small amount of self-heating should not be a significant factor in these larger standards.  Transfer grade standards are almost always made of single resistors with low TCRs although very high value standards can be made up of multiple resistors such as the SR-1050 10Meg unit which has three main resistors per step.

There are multiple conditions which can affect the absolute value of a standard when being measured, all of them must be taken into consideration when attempting very high accuracy measurements, it is not just any single factor.

Major factors in the resistor's TCR performance are the wire alloy's TCR spec followed by the resistor manufacturing processes that produces the end result.

[zlymex]

1. Fluke 742A may be much inferior than SR104 for long term stability, but as far as TCR is concerned, they are the same grade, if not better.

2. Matching 4 resistors does make the extreme low alpha TC possible. The level of alpha for Fluke 742A may be very low, even below 0.03ppm/K, this cannot be achieved or not feasible to achieve by just a single resistor.


3. Fluke does make PWW resistors with TCR below 1ppm/K, even below 0.3ppm/K, but that is not sufficient for 742A as a single resistor.

4. As can be seen from photo of a 742A below, there are four resistors with TCR of +0.1ppm/K, -0.1ppm/K, +0.2ppm/K and -0.2ppm/K respectively, therefore, they are indeed selected for match.

[Edwin G. Pettis]

Zlymex, you appear to be basing your opinion on a very small sampling.  I did indeed say these resistors were matched or 'selected' from larger batches of resistors.  I have been involved in the manufacture of precision wire wound resistors for over 40 years, I know precisely how these resistors are made, treated and selected for such applications.  If you think that Fluke actually produces batches of resistors with TCRs under say ±0.5 PPM/°C, you are quite mistaken.  Yes, I can make resistors with TCRs near or practically zero TCR but it is neither simple, nor easy to do, if it was everybody would be doing it and I have one of the most repeatable manufacturing procedures available anywhere.

The individual TCRs do not have to be sub- 1 PPM/°C to be matched to a near zero composite TCR, it can be done with almost any TCRs also long as the sum total of the TCRs add up to near zero.  There is nothing special in that.

Multiple resistors can decrease individual self-heating, but so can a larger resistor or even a smaller resistor depending on its thermal capabilities.

Again, Fluke does not use unsuitable resistors from a given batch, what you see inside of a 742 series resistor is selected resistors, that does not reflect what the characteristics of the total batch of resistors had.

I have made resistor standards myself for major corporations, including General Electric that operated under much more demanding conditions than a Fluke 742 would ever dream of being run under.  How about running at 10 watts and holding tight specification, don't try that or they're toast with a 742 or SR-104.

[saturnin]

@Edwin G. Pettis

Based on your knowledge, does TCR of precision resistors change with time? (i.e. is it likely that TCR of a given Fluke 742 determined during manufacturing is still the same after years of service?)

[Edwin G. Pettis]

The TCR of a resistance alloy is produced by high heat treatment, generally a complex process to create a given TCR in a given alloy.  Part of the problem for the alloy/wire manufacturer is that each ingot of alloy has slightly different characteristics so that the required heat treatment procedure is never exactly the same to get the same TCR, a really good metallurgist can minimize the unknown to some degree but it still something of a trial and error, if the 'finished' TCR is not achieved, it can be put back through the process again for another shot at the correct TCR.  Tight control of the proportions of metal elements in the alloy makes for more consistency in the ingots but it is difficult to get much better than about 0.1% variation in metal contents.

For Evanohm, the heat treatment is far above the normal operating temperatures of even military resistors rated to operate to +125°C, in order to 'mess up' the TCR of an Evanohm alloy, you would have to raise the temperature well above 300°C, the top recommended operating temperature for Evanohm is in the range of 200°C and at this point the TCR is not the same as it is at +125°C.  So as long as you don't get Evanohm really hot, the TCR will not change over time to any significant degree.

Mangain however is much more sensitive to temperature and any significant time of temperatures over 60°C can cause permanent changes to it.  Therefore it is never a good idea to heat your Manganin resistor to anything over about 40°C-45°C for any length of time.  Manganin is also sensitive to low temperature as well although it may be more of a temporary change than permanent if the temperature is not too cold or prolonged.

Another factor affecting what TCR is possible is the diameter of the drawn wire in Evanohm, the larger the wire diameter, the harder it is to achieve very low TCR, the smaller the diameter, the more comparatively easy it is to achieve low TCR and to some degree this depends on the ability of the wire manufacturer.  Tighter low TCR specifications on the wire tends to produce higher wire costs, ask me how I know!

For the vast majority of precision wire wound resistor manufacturers, the end result TCR is not a tightly controlled end result, the TCR of wire can be affected by the manufacturing process of the resistor manufacturer because of various sources of stress, some of which are easier to mostly remove than others.  Getting finished resistors with TCRs under ±1 PPM is not that easy and there is normally a variation in the end TCRs, how much depends on the manufacturer's processes.  I should note that technically stress acting on a resistor's value does not actually change the wire's inherent TCR, it is still the same, the stress merely causes the apparent TCR to be larger than the real TCR, if it was 'relaxed' and freed of stress, the true TCR would appear again.

[Alex Nikitin]

A quick update. After almost one month I've brought the 742A-10K back to the HP3458A, gave it full five hours of warm-up, as the resistor did travel in my backpack for about an hour, half of that time in cold (~5-6C) outdoor conditions (though I did pack it well in some thick bubble-wrap). After a fresh ACAL ALL the readings over 30min were pretty stable at 10.000060K average at 23C in the lab, so there is no detectable shift over that period of time between the Fluke and HP meter. Vertical scale 0.5ppm/div

Cheers

Alex

P.S. at the same time the Keysight 34465A did drift by about 2-3ppm on 10K resistance range, as it was measuring +1ppm from HP3458A and now it is +3..+4ppm (after ACAL). On the voltage range there is no detectable drift between two meters though.

[Alex Nikitin]

Another three months passed and I've brought the 742A back to the HP3458A. Unfortunately, due to a very warm day here the temperature in the lab was around +27C. After ACAL ALL the measured value of the 742A is 10.00005K, 1ppm less than what I've measured in December. If I use the Alpha and Beta values printed on the unit, the calculated resistance change would be about -0.3ppm. However the HP3458A most likely is responsible for the rest of the difference. I am planning to send the 742A to Fluke UK for Traceable Calibration with Data service as that should give me a good uncertainty in resistance reference for the next year or so.

Cheers

Alex

[Edwin G. Pettis]

Actually, for a 'young' standard, the stability is quite good, it will take some time, several years probably, to 'settle' down to a lower rate of change.  For now, something around 1 PPM/year or so is not bad at all.  In the mean time, expect little variations up or down with use and temperature variation.  Treat it with kid glove care and it will settle into a nice stable resistor.

[Alex Nikitin]

Actually, for a 'young' standard, the stability is quite good, it will take some time, several years probably, to 'settle' down to a lower rate of change.  For now, something around 1 PPM/year or so is not bad at all.  In the mean time, expect little variations up or down with use and temperature variation.  Treat it with kid glove care and it will settle into a nice stable resistor.

Hi Edwin,

Thank you, however this 742A-10K is not that "young", most likely about 15-20 years old. I might ask Fluke if they can establish the date of manufacturing from the serial number. I should send it for calibration soon and after that would know exactly how much it has drifted since it was made. Fluke UK provides 2ppm calibration uncertainty for 10K resistor, AFAIK. I've asked the lab if they would re-measure the Alpha and Beta coefficients, however they only provide the value at 23C+/-0.5C. On the other hand the lab confirmed that the temperature coefficients are unlikely to change with time, only the resistance drifts, so the printed Alpha and Beta can be used with the new calibrated value.

Cheers

Alex

[ManateeMafia]

Alex,

I purchased a used 742A that was missing the alpha/beta coefficients. I emailed Fluke and their response was that they only keep records for 5 years. Whether or not that was true, I never got the data. I hope you have better luck.

[Alex Nikitin]

Alex,

I purchased a used 742A that was missing the alpha/beta coefficients. I emailed Fluke and their response was that they only keep records for 5 years. Whether or not that was true, I never got the data. I hope you have better luck.

At least it is possible to measure these coefficients if you have a bridge or a meter with a sub-ppm resolution and short term stability, and a temperature chamber. According to the manual for 742 the measurement is done at 18C, 23C and 28C and the coefficients are calculated from the measured values. I may try to verify the coefficients for my 742A-10K after I'll get it back from calibration, however it might be not that easy  .

I've run HP3458A + 742A-10K overnight and it appears that the meter has a much larger tempco than the resistor. The change was about +5ppm for about -5C change in the ambient temperature and a fresh ACAL ALL in the morning only changed the measured value by -1.5ppm. The internal temperature of the meter dropped from 38.8C to 34C overnight. If we believe the Alpha and Beta printed on the resistor the value change from 27C to 22C should be only +0.15ppm . So this HP3458A has about -0.7ppm/C tempco on the reference resistor after ACAL. I need a temperature stabilized chamber for the meter (or for a bridge)  .

Cheers

Alex

[HighVoltage]

It seems Fluke is very hesitant in releasing old cal documents.
I bought a very expensive SPRT (Temperature Probe) for a very low price because it came without the detailed coefficients and no cal paper. And for sure I expected Fluke to have these based on the serial number and I was so wrong. But they offered a new calibration for lots of $$.

I need a temperature stabilized chamber for the meter (or for a bridge)  .

Yes, it would be nice to have a temperature controlled lab for these kinds of measurements.

[Alex Nikitin]

As the temperature in the lab is reasonably stable at the moment around 24C, I did ACAL ALL on the HP3458A and moved the Fluke from the bench (where it's temperature, measured by an attached thermocouple, was 25C) to a (rather warm) top of the Keithley electrometer. Over an hour the temperature of the resistor has climbed up from 25C to 29C - with less than -1ppm change in the measured resistance value. The change predicted by the Alpha and Beta coefficients for this resistor and temperature change from 25C to 29C is -1.1ppm . Looks good to me though it was a very crude way of checking the tempco. With the maximum predicted value change from 18C to 30C of less than 2ppm it is not an easy to check though.

Cheers

Alex

[Alex Nikitin]

Here is the graph for the 25C-29C measurement of the 742A-10K.

Cheers

Alex

P.S. I've added a graph with calculated from Alpha and Beta coefficients Resistance values v Temperature in 18C-30C range.

[chinapp]

the 2ppm drift was totally contribute by the 3458A,most of the 3458a has some low drift ,the reason is the A/D board of 3458a,a new A/D board will solve this problem