Precision Isolated current sense (16bit+)

[ali_asadzadeh]

Hi,
We have a product that uses CT's to measure the 50Hz currents, we have used it about 10 years (it's in protection relays), but we want to upgrade our product, so I have seen these hall effect sensors for some years, I want to know if there are better and newer devices for CT's replacement, since the main problem with them is that they are custom made, pricey and big! also they produce some phase shifts in the current waves, but you can calibrate these things in software!

So regarding CT's and hall effect sensors or other methods of Isolated multi channel current measurement, please share your experience or Ideas.

[jbb]

Yes, CTs can get quite large depending on accuracy and burden requirements.  What sort of environment are you working in?  Low (<1000V AC), Medium (<30 kV AC) or high?  How large are the primary currents?  Are you measuring primary current in the switchgear or secondary current in the relay (1A / 5A nominal)?

For primary current over 100A, I would stick with CTs (or Rogowskis) because they are so robust.  For secondary current, I guess Hall Effect sensors would offer better bandwidth and thus phase accuracy which is relevant for metering and precision protection.

There are a couple of gotchas with Hall Effect sensors, though:

• They can be sensitive to external fields (depends on make)
• They need temperature compensation (usually built in)
• For closed loop types: if input current is applied while the sensor is unpowered, the 'iron' core can be magnetised and pick up a DC offset.
• If high currents are applied (e.g. phase to phase fault), the 'iron' core can be magnetised and pick up a DC offset

For maximum accuracy, you could also look at placing a precision shunt and ADC (and digital isolators) on the reference potential.  This will entail a certain about of screwing around with isolated supplies etc.

If you need the best possible accuracy, closed circuit fluxgate sensors seem to be the thing.  This one is specified in ppm, but hold on to your wallet!

[ali_asadzadeh]

Yes, CTs can get quite large depending on accuracy and burden requirements.  What sort of environment are you working in?  Low (<1000V AC), Medium (<30 kV AC) or high?  How large are the primary currents?  Are you measuring primary current in the switchgear or secondary current in the relay (1A / 5A nominal)?

We are measuring the secondary current in the relay (1A/5A) but these inputs can overload to 40A, my main problem for now is size and cost of the CT's also they are somehow sensitive to shock and vibration.
Regarding the Hall effect sensors, can they be calibrated! I mean do they provide good accuracy over the temperature and supply voltage variations, do we have low temperature drift ones with reasonable prices?

[coppice]

Yes, CTs can get quite large depending on accuracy and burden requirements.  What sort of environment are you working in?  Low (<1000V AC), Medium (<30 kV AC) or high?  How large are the primary currents?  Are you measuring primary current in the switchgear or secondary current in the relay (1A / 5A nominal)?

We are measuring the secondary current in the relay (1A/5A) but these inputs can overload to 40A, my main problem for now is size and cost of the CT's also they are somehow sensitive to shock and vibration.
Regarding the Hall effect sensors, can they be calibrated! I mean do they provide good accuracy over the temperature and supply voltage variations, do we have low temperature drift ones with reasonable prices?

Precision 5A CTs are fairly small and cheap these days, even one with a high DC tolerance. They can have a remarkable dynamic range and phase linearity compared to the 1990s. If you are paying more than 40 cents in reasonable volumes you should definitely be looking for a cheaper supplier.

[ali_asadzadeh]

Precision 5A CTs are fairly small and cheap these days, even one with a high DC tolerance. They can have a remarkable dynamic range and phase linearity compared to the 1990s. If you are paying more than 40 cents in reasonable volumes you should definitely be looking for a cheaper supplier.

Do you know suggest some part numbers, also are they small?

Also I have found the isolation amps like AMC1100, do you have any idea about them too?

[coppice]

Precision 5A CTs are fairly small and cheap these days, even one with a high DC tolerance. They can have a remarkable dynamic range and phase linearity compared to the 1990s. If you are paying more than 40 cents in reasonable volumes you should definitely be looking for a cheaper supplier.

Do you know suggest some part numbers, also are they small?

Also I have found the isolation amps like AMC1100, do you have any idea about them too?

Isolatedn amps, like the AMC1100 series, have a number of good qualities, but they don't usually have a great SNR, and most of them need isolated power for the input side of the amp. The poor SNR may or may not be a problem, depending on the dynamic range you need. They cost more than a CT.

[Jeroen3]

See if Sensitec has some hall-effect units you like.

[ali_asadzadeh]

See if Sensitec has some hall-effect units you like.

Thanks for the tip

[capt bullshot]

From my experience, the performance (in terms of linearity / accuracy / noise) of such kind of CT (for 50Hz / 60Hz measurement / metering applications)
http://www.coilws.com/index.php?main_page=index&cPath=208_291_14_241_143
isn't matched by any of the alternative technologies.

Open loop (Hall) sensors have large errors and sensitivity to stray fields. Closed-Loop Hall (or GMR for the Sensitec ones) are better in these terms, but still have DC-offset issues. Speaking of stray field sensitivity and offset drift, the LEM ones are better than the Sensitec (because these don't have a closed loop magnetic core). There are closed loop Flux-Gate sensors made by VAC these are quite good in terms of offset and accuracy, but add quite a lot of noise to the signal (due to their working principle).

Isolation by the mentioned AMC1100 (or many similar) devices will add substantial noise and non-linearity.

If don't need isolation in the current sense part, the cheapest solution would be to use a shunt resistor and connect it directly to a suitable metering IC.

[ali_asadzadeh]

Isolation by the mentioned AMC1100 (or many similar) devices will add substantial noise and non-linearity

Thanks for the Tips, AMC1100 has an appnote claiming under 0.5% accuracy in all conditions, also TI has a newer and better Device AMC1301 which claims under 0.1%, is these devices better than Hall sensors? Shunts are the best so these device basically use shunts on the isolated side

[capt bullshot]

Using a shunt having 40A overcurrent capability leads to a rather small value shunt. Something in the ballpark of 2mOhm to keep the power dissipation in a acceptable range. With 2mOhm, one has 10mVrms across the shunt at 5Arms nominal current - the AMC1301 and similar devices have 250mVpk FS input range, there are a few of these devices offering 60mVpk input range at elevated noise levels.
So, with 14.4mVpk in a 250mVpk range, the specified noise of these isolator devices is 24dB worse at your rated current, in the ballpark of 50 ... 60dB SNR. Nearly any closed-loop Hall device will perform better.

[coppice]

Isolation by the mentioned AMC1100 (or many similar) devices will add substantial noise and non-linearity.

Actually, the AMC1100 is pretty good for linearity, as long as you follow TI's design guidelines. It can appear somewhat non-linear if you don't, but it has been some time since I touched these devices, and I forgot what the issue is. Its noise which is the greater issue with these devices. Its fine for most protection applications, but if you are looking for a big dynamic range for measurements it can be a limitation.

[ali_asadzadeh]

but if you are looking for a big dynamic range for measurements it can be a limitation.

How many bit's of dynamic range can I expect from these? can they do 16Bits?

[BravoV]

but if you are looking for a big dynamic range for measurements it can be a limitation.

How many bit's of dynamic range can I expect from these? can they do 16Bits?

Actually you can do the math your self, what are you doing is like trying to build a weighing scale can weight "precisely" from tons (like a truck/car) down to grams (like gold/diamond or few grains of rice) ... accurately at the same measurement equipment you're going to make.

You should do your homework 1st on the error budget, see if its realistic before randomly choosing the components.

My 2 cents.

[jbb]

You should do your ... error budget, see if its realistic before randomly choosing the components.

Yeah, if you need something special it pays to do a full error budget.

You may be able to do something with multiple ranges (eg a ‘nominal’ range and an ‘overcurrent’ range).