Precision measurement with current transformers, how difficult?

[woodchips]

I have only used current transformers for mains single and three phase applications, so 100A upwards. Also large and claimed accuracy, not resolution, is not much better than 1A over temperature, moisture etc range in the real world.

The parts I have used tend to be recycled from other uses so whilst good quality might be couple of decades old.

[uer166]

A 40,000:1 dynamic range for a current transformer is pretty readily achievable?

Are your sure? My limited experience suggests even a 4000:1 range is very optimistic? Mains power meters seem to be not a lot better than 1%.

A current transformer relies in mechanical arrangements and stability to work, like other mechanical sensors like pressure, load etc. Not like a a resistor or even a capacitor which are mechanically robust with much less sensitivity to external events, noise, influence etc.

Yah I'm sure since I designed similar 3-phase meters that would achieve that dynamic range in single digit % accuracy. OP said they wanted resolution more than absolute accuracy. You can (with some effort, but trivially in a sense that it's been done before) get 4000:1 with 0.1% accuracy, or say 40,000:1 with 1-2% accuracy. CTs don't rely on mechanical construction being accurate, rather the magnetic material they're made of, (Hc, Br, hysteretic loss, overall shape of BH curve, etc)..

[wilhe_jo]

I'm actually in the game as well... however, I tend to make this a partly commercial product.
It's gonna be a harmonic current analyzer.

I investigated hall and GMR based sensors etc.
Besides their non-linearity, my problem is that I want to get at least up to 9kHz... so ATM, I'm rather sure, I will take the shunt approach.

From my preliminary research, 24bit audio ADCs should be fine... but there are relatively affordable 32bit audio ADCs with 768kHz sampling rate around.
My intention is to have them isolated and floating at mains potential and doing U/I per phase with the 2 channels.

Not the cheapest option, but it should get me there.

However, I just need to measure 50mA on 16A... that's only 1:300

[coppice]

I investigated hall and GMR based sensors etc.
Besides their non-linearity, my problem is that I want to get at least up to 9kHz... so ATM, I'm rather sure, I will take the shunt approach.

Hall probes are interesting. Hall probe chips have serious stability problems. Its hard to make things which are better than 2% accurate over a modest dynamic range with Hall probe chips. However, at least two companies have their own in house ASIC which integrates a Hall probe with power measuring electronics on a single die, and achieve really good temperature stable accuracy for power measurement in utility energy meters. Assume they have some well tune compensation on those chips.

GMR is also interesting, but the devices I've played with ended up rather disappointing.

From my preliminary research, 24bit audio ADCs should be fine... but there are relatively affordable 32bit audio ADCs with 768kHz sampling rate around.
My intention is to have them isolated and floating at mains potential and doing U/I per phase with the 2 channels.

A lot of audio ADCs do not permit the use of an external voltage reference. However, if you look around some have a pretty temperature stable internal reference, and can work well for accurate measurement. I'm not clear why they made an effort to achieve those stabilities for the audio market, where its irrelevant.

[wilhe_jo]

I'm not clear why they made an effort to achieve those stabilities for the audio market, where its irrelevant.

Well, I'm pretty sure, they just use some available standard-cell-reference (that was cheaply available from the fab) instead of inventing a poorer new one

73

[coppice]

I'm not clear why they made an effort to achieve those stabilities for the audio market, where its irrelevant.

Well, I'm pretty sure, they just use some available standard-cell-reference (that was cheaply available from the fab) instead of inventing a poorer new one

If you want 20ppm/C or better you need to make a chip by chip effort to achieve that. Been there, done that. You are talking about <20 cent chips. You really don't want to add any cost to those.

[Kleinstein]

Besides the audio chips, there are also the ADCs made for power metering application. They are somewhat similar to the audio ones, but may offer better gain stability and often an external reference at least as an option.

For the current transformer it is about keeping the magnitizing current small and thus a small voltage per turn and a relatively low resistance in the secondary side. If one really wants to push it, one could used a separate winding for sense and secondary drive in a TIA like configuration.  It is extra effort, but no magic.
For higher harmonics one may want a thin layered core (e.g. nonoperm) to avoid eddy current problems.