High current shunt - temperature drifts

[CosteC]

Hello everyody

I am in need to measure about 400 A AC (continious) with as good accuracy as possible and fits 6 1/2 digit multimeter (~0.06%). Current is somewhat distorted 50 Hz sine.
Easiest way seems shunt. I have found 0.2 class and 0.25 class shunts:
https://www.lumel.com.pl/en/catalogue/product/shunts-b2-60mv-3 CLASS 0.2 but temperature coefficient seems 100 ppm/K which is rather a lot, but also specified for ambient temperature not shunt temperature.
Other https://docs.rs-online.com/18e7/0900766b813bb936.pdf is CLASS 0.25 and has thermal coefficient of 15 ppm/K which is bit questionable, as manganine coefficient is 0ppm for 25'C but -42 ppm for +100'C Obviously there are slightly different alloys under the same name, varying a bit in performance.

What is your recomendation for such measurements?

[Kleinstein]

400 A is quite high. A good shunt would habe to be phyical large to get good power handling capability. One may want a special low noise amplifier to get away with a low burnden voltage. For AC only the DC drift would be less of an issue. 60 mV burden would still be 24 W and the current path outside the active area can add to this.

The shunt would still need to be low TC to handle the heating - no need to care about 100 C, as the shunt should not get that hot at all.
Manganin is still a good choice, because of the low resistivity, low thermal EMF (important for DC) and good contacts.

The 2 shunts shown are way to tiny - for good accuray consider 10 x or more physical size. This could be shunt for a higher current (e.g. 1000 A) and using less burden with a better amplifier. Here AC is much easier than DC.

An alternative to consider would be a good current transformer instead of directly a shunt.

[Hydron]

This is definitely a job for a CT rather than a shunt - far safer and easier. You should be able to get one that matches the 0.2% shunt accuracy fairly easily (ratio error, so you have to add the meter AC current error too), or I suspect there are other better options if you want to spend more.

[TimFox]

You can use a CT terminated with an appropriate low-valued resistor to develop maybe 100 mV rms across that resistor and drive an AC voltmeter to avoid any error on an internal current shunt in the DMM.

[CosteC]

This is definitely a job for a CT rather than a shunt - far safer and easier. You should be able to get one that matches the 0.2% shunt accuracy fairly easily (ratio error, so you have to add the meter AC current error too), or I suspect there are other better options if you want to spend more.

Any proposals for 0.2% class or better CT?

You can use a CT terminated with an appropriate low-valued resistor to develop maybe 100 mV rms across that resistor and drive an AC voltmeter to avoid any error on an internal current shunt in the DMM.

Well. Not adding low value resistor in most of cases will ruin CT accuracy... And I studied DMM current measurement accuracy. It is not-so-great for currents > 100 mA.

[Hawaka]

DaniSense or a LEM Ultrastab should have easily the accuracy you need. But be ready to spend some $$$.

[CosteC]

DaniSense or a LEM Ultrastab should have easily the accuracy you need. But be ready to spend some $$$.

IN-xxx series from LEM is indeed almost magically accurate and equally expensive

I made extensive digging in specs and spoke with LUMEL laboratory department.

Conclusions:

• LUMEL does not require de-rating up to +55'C ambient. Murata require de-rating to 2/3 rated current for continuous operation at +25'C
• LUMEL keeps 0.2% up to +55'C . Over it it is 100ppm/K extra error.
• MURATA require vertical mounting due to blade shape resistive elements, LUMEL uses round bars which are not as demanding.
• If Murata accuracy is specified as 0.2% at +25'C then thermal drift makes it 0.295% class at +55'C... unless it is shunt temperature, not ambient. Then it is far worse.

So, LUMEL it will be. Maybe with lazy fan to be sure it is away from +55'C limit of ambient temperature.
In meantime lets check how much will cost their factory calibrated one with papers.

[tszaboo]

You can measure this with shunts, easiest way to improve them is to dump them into a bucket of mineral oil.
Or direct fans at them if you are not comfortable with oil.
LEM sensors are also very good.

[CosteC]

Mineral oil require cooling too, especially if measurements take hours
LEM is great, but there is lot of issues except of price. I am really not measuring current, I am measuring ratio between two currents (calibration/accuracy control) so I would need to synchronise LEM with other things. Not willing to do it as far as I can

Thanks!

[Phil1977]

Can you feed the currents you want to compare through the same CT?

[CosteC]

Can you feed the currents you want to compare through the same CT?

No. I am testing CTs. Or Rogowski Coils.
Generally, I need to measure large current and its relation to "small current" or "small voltage". In practice those are two "small voltages" - voltage from shunt(s) or voltage output from integrator (rogowski coil)

But I would like to hear more about your idea. Can you share?

[Phil1977]

It´s just a very easy idea: A current clamp or CT physically sums up all the currents flowing through it´s port area.

This can be used to enhance the measurement accuracy. If you e.g. want to compare two lines carrying around 100A then you need 1% of measurement accuracy to resolve a difference of 1A. If you direct both lines antiparallel through your current sensor, then you just read the current difference, so that a current difference of 1A with a 1% CT gives you a resolution of 10mA.

If the ratio is not one to one but any reasonable rational number, you could also feed several windings through the CT - though of course 400A lines may be too thick for that.

[Fungus]

What is your recomendation for such measurements?

Bigger heatsinks on the shunt.

Edit: I should probably say "better thermal management of..." - too many literal pedants around here.