Measuring accurate milliohms

[BradC]

Ok, I was mooching around E-bay looking at calibrated shunts and I stumbled across this listing :

https://www.ebay.com/itm/Calibration-Shunt-Current-Amps-OHMS-tested-good-calibrated/192453957647

I can't find any good reference on how you would go about measuring such a low resistance with that level of precision (let alone accuracy).

How would you go about it?

[amspire]

To get anywhere near that, you would probable have to use a DCC bridge - direct current comparitor.

This allows you to compare a current through the tested resistor up to 2000A with a reference standard resistor running at 10mW.

It is using the balancing of flux in a transformer as the means of balancing the currents. If one winding has 10 turns at 100 A, and a second winding has 10,000 turns at 10mA, the flux from the two windings will exactly cancel out.

Since the transformer windings also provide isolation, it is easy directly comparing the voltage across your 1 milliohm resistor with the voltage across a 1 ohm standard resistor, or a 10K ESI standard resistor box (if you have the correct turns ratios.).

The galvanometer is just balancing the DC voltages across the resistors. The variable winding Np in the diagram is a bit more complex then in the diagram - it is a whole series of tapped windings and perhaps a circuit involving two transformers to give a 7 digit resolution. Unfortunately, you cannot buy a transformer set like this off the shelf anywhere. It has to be custom made at a very high cost. I am not sure how many people there are left who know how to build such a transformer.

A cheaper solution will be something like a guarded Wheatstone bridge, but it will be pretty hard to get anywhere near 7 digits accuracy.

[ap]

DCCs are commercially available. They are manufactured by Guildline and Measurements International, both in Canada. The technology was originally developped by the Canadian metrology department IIRC. 
What can also be done is to apply the currrent needed to the shunt, measure it e.g. with a precision current sensor (LEM ultrastab, uncertainties at the ppm level), and measure the voltage accross the shunt by a nanovolt meter (34420A or the like).
A calibrated value given on the shunt only makes sense / is usable if you know the condition under which it was tested, because the self heating will change the shunt resistance and depends on the current applied.

[amspire]

DCCs are commercially available. They are manufactured by Guildline and Measurements International, both in Canada. The technology was originally developped by the Canadian metrology department IIRC. 

It is great to hear that there are still people who can make this stuff. Had a look at the Guidline site and the specifications are pretty incredible: Best accuracy of 0.015ppm, Linearity 0.01ppm. Resolution .0001ppm.

There are extended to be able to measure at test currents up to 10,000A.

[HighVoltage]

That label of the shunt with 0.000,998,905,8 Ohm is total nonsense with such resolution!
If you wiggle the sense wires a little on that shunt or change the position slightly, the result would be all over at least the last 5 digits shown.

So, someone is making a label that has clearly not much meaning, to say the least.
And if the resistor warms up a little, it will change the resistance dramatically too.

There is only value in a calibrated shunt, when the calibration certificate will show the resistance at different current and different temperatures in a table. Or at least say the condition of current and temperature for this one measurement.

Also, if you hook up a constant current and wait a little, the voltage drop will change.

Making precision low ohm measurements is not easy, as I have discovered myself:
 https://www.eevblog.com/forum/metrology/low-ohm-precision-resistor-standard-and-testing/

[amspire]

That label of the shunt with 0.000,998,905,8 Ohm is total nonsense with such resolution!
If you wiggle the sense wires a little on that shunt or change the position slightly, the result would be all over at least the last 5 digits shown.

So, someone is making a label that has clearly not much meaning, to say the least.
And if the resistor warms up a little, it will change the resistance dramatically too.

If it was measured on something like a Guildline DCC, it probably would show that many digits of resolution. The test current may have been high - say 50A. There is always some kind of meaning to a high resolution reading. They might want to make matched sets of shunts for three phase power. Resolution never has to mean accuracy.

I doubt anyone was thinking that this resistor is 0.1ppm accurate. A sticker is definitely not a calibration certificate. Some people hate to see extra digits of a reading when the accuracy is a lesser number of digits. I like the idea of recording a whole reading.

For one thing, if you came across several of these resistors, you could compare the relative drift over time.

[BradC]

I appreciate the info guys.

I used amspires original post for some keywords and wound up at the Guildline site. I was there a couple of days ago checking out their precision current measuring kit.

I'd like to get a couple of vaguely accurate shunts in the order of 1 & 10 Amps. I've been checking out the Riedon RSN Series. 0.1% and +/- 15PPM/C. They work out about $50 bucks a throw, which makes both the accuracy and price in my ballpark.

By the time they're mounted in an enclosure with some forced ventilation and decent leads/terminals I should get some repeatability of measurement if not great absolute accuracy.

That label of the shunt with 0.000,998,905,8 Ohm is total nonsense with such resolution!
If you wiggle the sense wires a little on that shunt or change the position slightly, the result would be all over at least the last 5 digits shown.

I just saw that shunt and the label and figured someone was taking the piss, but then there's a lot I don't know so I figured I'd ask here.

Making precision low ohm measurements is not easy, as I have discovered myself:
 https://www.eevblog.com/forum/metrology/low-ohm-precision-resistor-standard-and-testing/

Yeah, I followed that thread.

It'd be a sad day if I didn't learn something

[ap]

If you look at the specification, and always provided the shunt is temperature stable (RUG-Z...), the accuracy of these DCC bridges is extremely high. See also data sheets, sub ppm. So the digits shown on that shunt could make reasonable sense (data given with 0.1ppm res.), again provided this specific shunt is stable. The latter may not be the case with this one. These bridges use a SW, and typically a calibration consists of say 10 measurement cycles, with 10 measuremens each, generating an average and calculation the standard deviation, which is then transferred into the appropriate measurement uncertainty stated on the shunt cal doc. Some cal labs using these bridges here certify in the 2ppm range uncertainty at 1mOhm, one additional digit is given most of the time, brings you to 0.1ppm resolution. More of a problem is the aging drift. This may e.g. be a couple 10ppm pa with a RUG-Z not operated at high currents frequently, and much more under heavy load conditions. Also the load current applied strongly affects the measured value.
And of course with a simpler setup (DMM and current source) such resolution does not make any sense.

[BravoV]

That label of the shunt with 0.000,998,905,8 Ohm is total nonsense with such resolution!
If you wiggle the sense wires a little on that shunt or change the position slightly, the result would be all over at least the last 5 digits shown.

I was wondering about that too on the Kelvin connection points, also with that with full exposed resistance element, isn't that generates less confident about it's long stability ? CMIIW 

Say compared to the sealed big resistor like these ?

10 miliOhm 0.1% 50ppm/C TCR  with the big sensing terminals (smaller one). Datasheet (PDF)
(Click to enlarge)


Same as above with 0.02% tolerance, but with Teflon coated for the Kelvin sensing wire.