Low ohm precision resistor standard and testing

[alm]

If that drift is just from ambient temperature, then it would translate to about 20 ppm/K. Much higher than the 10 ppm/K specified by Burster. But if I read the specs right, the temperature coefficient of the SMU alone could easily be responsible for ten times that value. The specs state that the temperature coefficient is 0.1 * accuracy / K. Accuracy is about 2400 ppm (the 200 µV offset for the 200 mV voltage measurement range is responsible for most of that). Better put the SMU in a thermal chamber .

One thing you may be able to do is assume the Burster resistor meets its specs, and measure the Russian resistors relative to it. Either that or measure for long time periods and use regression to try to compensate for temperature fluctuations.

[HighVoltage]

This 1 Ohm Burster resistor was turned ON with the Keithley 2460, since I got it.

When I started out with this Burster 1 Ohm resistor, it had shown 1.000 080 Ohm on the SMU
A day later, I ended up with 1.000 040 Ohm
Again, two days later, it shows 1.000 013 Ohm
It has fallen about 55 ppm

So, there is clearly a drift that is no longer associated to only temperature variations.
Somehow I had expected it to be stable after a few hours with a straight line in the graph.
But may be equilibrium has not been reached inside the resistor from the current flow?

I would assume this drift is related to the resistor and not the Keithley SMU?

Any comments?

[HighVoltage]

I was gone over the weekend and left the instrument running with this 1 Ohm Buster resistor.

Now, after 5 days, it has settled to a perfect value of 1.000 001 Ohm on the SMU

Without a real temperature chamber, it will probably not be of any real value to keep it running, So, I will start looking for some good temperature regulation equipment.
 

[HighVoltage]

For a new project, I need to measure 80A DC very precise and accurately.
Current clamp probes would not be accurately enough.

So I got one of these "Made in Germany" WEIGEL 100A shuts with 60mV /100 A of voltage drop.
This is a 0.5% class shunt.

Today I did the first test for up to 5A and look at the result.
Almost too good to be true.
Interestingly, I see no marks of calibration on this resistor. It is like they built it and it is spot on.

In the next few days I will test it for up to 100A, to see how linear this resistor is.

Anyone here having experience with this Weigel type shunts for metrology type purpose?

[alm]

No experience with this type. A current shunt should be calibrated at full current to account for self-heating. But since the temperature coefficient may have a parabolic shape, it could be that 5 A (close to room temperature) and 100 A (max temperature) end up at almost the same resistance value, which could account for why the 5 A reading looks so good. See figure 6 in this paper for a curve from an older (manganin-type) shunt. Obviously the curve for Evanohm would be flatter.

[ap]

If it has to be very precisely and accurately (whatever that is  ), then Isabellenhütte are nice, 1ppm/K. Or Ultrastab. Not cheap.

[HighVoltage]

See figure 6 in this paper for a curve from an older (manganin-type) shunt. Obviously the curve for Evanohm would be flatter.

That was a nice paper to read, thank you for the link.

Today I hooked this shunt resistor up to 100A DC but of course, I can only read my PSU to one decimal place and using a current clamp also does not help to get better resolution.

It seems I need an additional calibrated shunt, to get any kind of current accuracy at 100A.

At 100A and running for about 30 min, the resistor elements heat up to about 70°C and the 35mm² cable heats up to about 40 °C

Removing the resistor at this temperature and measuring the resistance again at 5A, gives exactly 0.000600 Ohm again. That was a little surprise.

[alm]

I was going to propose heating the resistor up and measuring the resistance at a lower current if you do not have the facility to measure 100A accurately. Is obviously not a metrology-grade measurement because the temperature distribution could be different, but it should be pretty close.

Not sure what the proper way would be (apart from sending it off to a cal lab that can generate 100A). Is it possible to do a current transfer with a stable current source with the current equivalent of a KVD?

[HighVoltage]

Not sure what the proper way would be (apart from sending it off to a cal lab that can generate 100A). Is it possible to do a current transfer with a stable current source with the current equivalent of a KVD?

Good question, I do not know, may be someone else with precision high current measurement experience can explain this.

All I have read so far is to use a calibrated precision shunt resistor.
But they are very expensive.

Also, in the paper you posted yesterday, there was a nice explanation that the DUT has to settle over time to reach equilibrium.

Here is my test:
There is a fast immediate drop and then it slows down.
In this test, equilibrium has not been reached after about 9 min.

This means, we must wait many minutes before we take current and voltage reading, when measuring such resistor.

[alm]

But how do they calibrate the calibrated precision shunt resistor?

Would heating the resistor (ovenizing it) to just under 70°C reduce settling time? Obviously it would have to be calibrated like that.

[e61_phil]

But how do they calibrate the calibrated precision shunt resistor?

I think they use a direct current comparator. With such a unit you can compare currents with the ratio of turn over a core. The "secondary" winding is controlled in a way to null the flux in the core. Therefor, you can measure the secondary (lower) current with another shunt and transfer the reading.

By the way I did some measurements on such a shunt a while ago. But with 1A only: https://lowcurrent.wordpress.com/2017/02/10/kleine-widerstaende-und-offset-compensation/ (sorry, it is in german)

In the mikrocontroller.net forum was a discussion about a 100A source and a guy (from a german cal lab) said, that this current shunts which looks like yours are in no way good enough for measurements below .5%.

[HighVoltage]

But how do they calibrate the calibrated precision shunt resistor?

Good question, I think I will ask a few cal labs here in Germany.

Would heating the resistor (ovenizing it) to just under 70°C reduce settling time? Obviously it would have to be calibrated like that.

Probably yes.
I had expected this shunt to be calibrated at 100A and was surprised to see spot-on numbers at only 5A
The problem is that this resistor will have a different temperature for each current level.

I would really like to measure the TC curve of this resistor from 1A to 100A to get an idea of its temperature drift. But that requires a second better quality shunt. May be I should get a 1000A shunt for a second one that would not heat up too much and will stay in the more linear area.
Hmm, more and more questions.

Plus... if I have two shunts ... I need a third for comparison.
The good thing about shunts, they can all be put in a series connection without loosing accuracy.
   

[HighVoltage]

By the way I did some measurements on such a shunt a while ago. But with 1A only: https://lowcurrent.wordpress.com/2017/02/10/kleine-widerstaende-und-offset-compensation/ (sorry, it is in german)

In the mikrocontroller.net forum was a discussion about a 100A source and a guy (from a german cal lab) said, that this current shunts which looks like yours are in no way good enough for measurements below .5%.

Nice write up, thanks for the link.
Your picture is funny to see the little alligator clips on the large screws. But I guess as long as the current flows, it should not make much difference.

So far, I am under the impression that this resistor is much better than the .5% it is posted at. Otherwise I would not read 0.000,601 Ohm at 5.000,000A (This one was with a heated resistor).
Or may be I was lucky to just get a "good" one?

 

[HighVoltage]

It seems this setup requires about 1 full hour to settle and find equilibrium at 100A.
Here is the 1 hour graph...

It would be interesting to hear, how official cal labs do this.
Do they really wait 1 hour for each measurement point to settle?

[alm]

I think they use a direct current comparator. With such a unit you can compare currents with the ratio of turn over a core. The "secondary" winding is controlled in a way to null the flux in the core. Therefor, you can measure the secondary (lower) current with another shunt and transfer the reading.

Ah, so there is something like a voltage divider that lets you transfer between current ranges. Thanks!

The problem is that this resistor will have a different temperature for each current level.

True. What you could do is thermally couple a second resistor to it, and dissipate approximately (100A - expected current)R in it to keep the total dissipation approximately constant. You are not aiming for single degree C temperature stability here, the temperature coefficient is supposed to be good enough for fluctuations from 25°C to 70°C. Not sure if there is any precedence on this, or if this is a completely ridiculous idea. It will probably slightly degrade accuracy compared to waiting an hour, but should produce a much better reading after waiting a minute.

I would really like to measure the TC curve of this resistor from 1A to 100A to get an idea of its temperature drift. But that requires a second better quality shunt.

How about just heating it using some other method (oven, other resistor) and measure the resistance at 5 A, like you have been doing? Once you know the tempco as a function of temperature, you should be able to estimate the relation to current. Obviously this is not the cal lab method (temperature gradients will be different), but it sounds like you are not equipped like a cal lab either .

So far, I am under the impression that this resistor is much better than the .5% it is posted at. Otherwise I would not read 0.000,601 Ohm at 5.000,000A (This one was with a heated resistor).

What is the uncertainty on that resistance measurement based on the detailed specs for the SMU? It could just be that the SMU error happens to cancel the shunt error.

[Kleinstein]

It seems this setup requires about 1 full hour to settle and find equilibrium at 100A.
Here is the 1 hour graph...

It would be interesting to hear, how official cal labs do this.
Do they really wait 1 hour for each measurement point to settle?

There can also be quite some settling in the SMU. I would expect the shunt inside the SMU to be of about similar level of precision. In the SMU there is not only the dissipation of the shunt but also the power source part.
So to make sure it is the shunt and not the SMU, one would first measure a dummy or a short and only after something like an hour switch to the shunt.

[HighVoltage]

Thank you for your feedback, Kleinstein and alm.

I just received another shunt resistor and I want to use this one as a reference for the 100A one.

This new one has following markings
- 240A
- 5 Ohm
- 60 mV

It seems to have a manufactures logo and a part number on the side but I have no idea who made this one.
May be someone here recognizes that logo / stamp?

Also interesting, this one is marked 5 Ohm but in reality has 250 micro Ohm
Is that just a stamping mistake, hmmm?

And another curiosity came up during my research in to shunts.
Why is it that most of them have 60 mV at nominal current ?
May be in the past there was a volt meter specifically made for shuts with a maximum of 60 mV
But why just 60 mV?

[lukier]

I got this 150A shunt recently: http://www.ebay.co.uk/itm/322631256627

It is also for 60mV full scale (thus 0.4mOhm). I still need to do the testing under high load, but I was surprised how well my Keithley 2001 can measure it (well at least the resolution was there).

I guess the 60mV comes from old D'Arsonval movement panel meters.

[HighVoltage]

I got this 150A shunt recently: http://www.ebay.co.uk/itm/322631256627

It is also for 60mV full scale (thus 0.4mOhm). I still need to do the testing under high load, but I was surprised how well my Keithley 2001 can measure it (well at least the resolution was there).

I guess the 60mV comes from old D'Arsonval movement panel meters.

Yes, I totally agree.
There is something about Keithley and accuracy when it comes to low Ohm measurements.

Today I gave my 100A Weigel resistor 2 new Hirschmann binding posts for the sense wires.
This resistor also measures very accurately with the Keithley SMU.

These measurements were taken at 5A and the resistor is 1 uOhm off on the Keithley and the min max don't even change at 70nOhm standard deviation over 683 readings.

[tszaboo]

You can use the Kirchoff's law to calibrate these shunts. That is how we made accurate (0.03%) 100A at the lab. Basically we had a bunch of current sources, each with 0.03% accuracy, and we just summed the currents. We could produce up to 4000A with 0.03%. Each current source had it's own water cooled shunt (the entire thing was water cooled), kept at 23-ish degrees, to avoid the dew point. Each shunt individually was calibrated in house. Calibrating a 10mOhm shunt is "easy".

We were selling the current sources/sinks (aka, battery testing equipment). The proper calibration waited for the shunts to heat up. But that heating up happened fast, since they were directly on a temperature controlled cold plate. Seconds.

Now, that is how we did it. Then we sent our reference shunts (good up to 50A) to the cal lab, and magic happened and they came back calibrated. To be honest, I dont know what the magic was.

[Edwin G. Pettis]

I have some Tokyo Seiden 30A, class 0.5 current shunts, they are calibrated for 180mV full scale.

[HighVoltage]

I have some Tokyo Seiden 30A, class 0.5 current shunts, they are calibrated for 180mV full scale.

Interesting!
So, they are then 60mV / 10A
May be this was an old standard, to build everything around 60mV full scale?
But why?

Also, why not today build it always around 100mV / 100A to make easy to use a DVM for readout?

[HighVoltage]

Calibrating a 10mOhm shunt is "easy".

What is your procedure to calibrate a 10 mOhm resistor in your lab?
At how many points between a low current and full scale current do you check for linearity?


Very interesting with your 4000A of multiple current sources.
With my two Agilent 6V / 100A PSUs I just can reach 200A.

[TiN]

Perhaps if you really interested in high accuracy high current stuff, you'd be better off getting LEM Ultrastab DCCT.
They pop up on ebay from time to time to relatively cheap prices. I've got two with best linearity, Ultrastab IT-600S.
It is 1:1500 ratio CT, and with linearity spec <1ppm, offset error <20ppm and tempco <0.2ppm/K, so high chance that your measurements accuracy with these DCCTs will be limited by your burden resistor stability. There are also lower (and higher) current Ultrastab's with similar specs. Buying older Danfysik DCCT (model 867 is same as IT-600S) might save few $$.

I did a brief test before, with 5700A and 5205A as current source.

CERN using these beasts to calibrate and maintain LHC superconductor magnet currents to ppm levels stability. You can find related research paper, covering their transfer and calibration apparatus. 

[HighVoltage]

Hello TiN

You just opened a complete new door for me to look at currents in ppm level.
Thanks, I was not aware of these current sensors and ordered one.