Current transformers at milliamps reasonable? [SOLVED, answer is "Yes."]

[Paul Moir]

Hello,
I've got a little one off project.  I need to monitor the current going into three 120VAC devices and trip an alarm if they drop below a value.   Each consume about 200mA.  I'd like to be able to trip with an under 10mA resolution; what I mean is I would like 180mA to never trip and 170mA to always trip.

Current transformers, the type with a molded in hole for the wire, seem like a nice solution as they allow simple isolation.  I'm not getting into a custom PCB for this, and these keeps the high voltages off the board.  But I'm concerned about using them at these lower levels since I seem to be getting off the graphs on the datasheets I've looked at.  Oh incidentally I don't have to be very fast.  Several seconds to a trip is OK so I can integrate out some of the noise.

What do you think? The environment isn't too bad electrically speaking, but there is nearby equipment.

[coppice]

Hello,
I've got a little one off project.  I need to monitor the current going into three 120VAC devices and trip an alarm if they drop below a value.   Each consume about 200mA.  I'd like to be able to trip with an under 10mA resolution; what I mean is I would like 180mA to never trip and 170mA to always trip.

Current transformers, the type with a molded in hole for the wire, seem like a nice solution as they allow simple isolation.  I'm not getting into a custom PCB for this, and these keeps the high voltages off the board.  But I'm concerned about using them at these lower levels since I seem to be getting off the graphs on the datasheets I've looked at.  Oh incidentally I don't have to be very fast.  Several seconds to a trip is OK so I can integrate out some of the noise.

What do you think? The environment isn't too bad electrically speaking, but there is nearby equipment.

200mA isn't low for a current transformer these days. It used to be, but the dynamic range of cheap CTs has increased enormously. The ones we use for utility meters measure from 10mA to 100A with 0.1% accuracy.

[diyaudio]

Hello,
I've got a little one off project.  I need to monitor the current going into three 120VAC devices and trip an alarm if they drop below a value.   Each consume about 200mA.  I'd like to be able to trip with an under 10mA resolution; what I mean is I would like 180mA to never trip and 170mA to always trip.

Current transformers, the type with a molded in hole for the wire, seem like a nice solution as they allow simple isolation.  I'm not getting into a custom PCB for this, and these keeps the high voltages off the board.  But I'm concerned about using them at these lower levels since I seem to be getting off the graphs on the datasheets I've looked at.  Oh incidentally I don't have to be very fast.  Several seconds to a trip is OK so I can integrate out some of the noise.

What do you think? The environment isn't too bad electrically speaking, but there is nearby equipment.

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

It seems your needs is low sense with HV isolation

What about a shunt and a opto-coupler  ?


 

[timb]

Have you looked into the current sensors from Allegro? Something like the ACS712. There's a nice little breakout from Sparkfun. The chip provides at least 1000V of galvanic isolation.

[coppice]

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

That isn't even in the ballpark of what you can do with inexpensive CTs these days. As I said before, a cheap modern CT can measure 10mA to 100A with an accuracy of 0.1% of reading across the entire range.

[diyaudio]

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

That isn't even in the ballpark of what you can do with inexpensive CTs these days. As I said before, a cheap modern CT can measure 10mA to 100A with an accuracy of 0.1% of reading across the entire range.

 10mA to 100A...inexpensive CTs?, I cannot seem to find them on the net.. care to share an example of one?

[Paul Moir]

Thank you for your replies!

Yes a shunt and some sort of isolator is an alternative but one I wish to avoid since I can keep my PCB entirely low voltage that way.  What I mean is this is going to be executed on perfboard, and it's hard to get good safe isolation on that.  I agree that this is the safe bet if looking for success.  It just isn't as easy. 

I actually had a *very* brief look at the ACS712, and I assumed it was for low current non-isolated DC applications and dismissed it.  Now that I see what it really is, and that Digikey can send me the SparkFun breakout board in a day, makes this a viable option. 

[coppice]

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

That isn't even in the ballpark of what you can do with inexpensive CTs these days. As I said before, a cheap modern CT can measure 10mA to 100A with an accuracy of 0.1% of reading across the entire range.

 10mA to 100A...inexpensive CTs?, I cannot seem to find them on the net.. care to share an example of one?

Try looking for CTs made for the utility meter market, from companies like ShengKe. Less than 50 cents in quantity, DC tolerant, and measuring up to 60, 80, or 100A, depending on the model. There are also CTs made for operation up to 6A, for use with a multiplier CT outside the utility meter. These are very standard parts in the utility meter market.

If you want to see the kind of results you can get with those CTs in a meter design, try looking at app notes for utility meter devices, like http://www.ti.com/lit/ug/tidu291/tidu201.pdf . The tests there only go down to 50mA, but the accuracy doesn't degrade much down to 10mA. That is 10mA to 100A measuring active power to an accuracy of about 0.1% of reading. If you want to measure RMS current, the noise in your analogue front end can spoil things a little. However, if you properly subtract room temperature AWGN during the measurement process you can measure RMS current down to 10mA with not much worse than 0.1% accuracy. If you measure RMS current in a more bull at the gate manner you may be down to a few percent accuracy at 10mA.

[Paul Moir]

Oh I meant to say, coppice, my ears are wide open! 

Accuracy is not necessary.  I only need to be repeatable.  The whole thing will need to be calibrated for the application anyway.

[Ian.M]

Put five turns of wire through each CT and you are up to 1A equivalent for your 200mA trip point. 

[AlxDroidDev]

Have you looked into the current sensors from Allegro? Something like the ACS712. There's a nice little breakout from Sparkfun. The chip provides at least 1000V of galvanic isolation.

I've had nothing but failure with those.  One of my projects included using a ACS712-5 (that measures up to 5A), but its output signal is quite noisy. Others mentioned I'd be having this trouble in a thread I opened here on EEVblog, and indeed they were absolutely right.

Even the ACS712-5 is only able to perceive changes above 300mA.

Now I'll try with a INA219, which is based on a shunt resistor. I bought a breakout board from DX and I'm waiting for it to arrive.

[Paul Moir]

IanM, thank you.  I didn't address that because I thought that the CT's noise would be independent of the number of turns.  I'm comfortable with 10, but not 100. 


AlxDroidDev - Many thanks for saving me time! 

[coppice]

IanM, thank you.  I didn't address that because I thought that the CT's noise would be independent of the number of turns.  I'm comfortable with 10, but not 100. 


AlxDroidDev - Many thanks for saving me time!

CTs don't really produce much noise. The noise comes from the front end of the electronics you connect them to. If you connect a CT to a really high SNR input it looks very very clean. What I described before about subtracting noise to get the best results for RMS current measurement relates to the use of utility metering devices for the signal processing. Use a higher SNR ADC and you can measure RMS current to 0.1% of reading from 10mA to 100A without any compensation.

[timb]

Have you looked into the current sensors from Allegro? Something like the ACS712. There's a nice little breakout from Sparkfun. The chip provides at least 1000V of galvanic isolation.

I've had nothing but failure with those.  One of my projects included using a ACS712-5 (that measures up to 5A), but its output signal is quite noisy. Others mentioned I'd be having this trouble in a thread I opened here on EEVblog, and indeed they were absolutely right.

Even the ACS712-5 is only able to perceive changes above 300mA.

Now I'll try with a INA219, which is based on a shunt resistor. I bought a breakout board from DX and I'm waiting for it to arrive.

I dunno, I've used them in a handful of projects and they've worked fantastic for me. You need to know how to set the filter pin to reduce noise. You can easily sense under 50mA by simply running the output through an op-amp. They also have other versions with lower and higher current ratings.

For what it's worth, if you want to sense low DC currents, an INA219 and shunt can't really be beat. TI also has those new INA chips with the integrated current shunt that are pretty awesome.

For AC or large DC currents, the ACS sensors are very nice.

[coppice]

Have you looked into the current sensors from Allegro? Something like the ACS712. There's a nice little breakout from Sparkfun. The chip provides at least 1000V of galvanic isolation.

I've had nothing but failure with those.  One of my projects included using a ACS712-5 (that measures up to 5A), but its output signal is quite noisy. Others mentioned I'd be having this trouble in a thread I opened here on EEVblog, and indeed they were absolutely right.

Even the ACS712-5 is only able to perceive changes above 300mA.

Now I'll try with a INA219, which is based on a shunt resistor. I bought a breakout board from DX and I'm waiting for it to arrive.

I haven't used those particular devices, but they are based on Hall probes, and Hall probes are generally rather noisy devices. Your description doesn't sound surprising at all.

[AlxDroidDev]

For AC or large DC currents, the ACS sensors are very nice.

True, but my case was very much like the OP's situation: I needed to sense currents as small as 10mA. I was using the most sensitive ACS712, the 5A model. It's output changes 185mV for every 1A measured. For 10mA, its output should change 0.00185mV, which is more than the Vpp measured on the output noise. Basically, I had a hard time telling noise from very small current changes. Like others mentioned in my thread a couple weeks ago: even if I amplify the signal with a opamp, I'll just be amplifying the error/noise.

Thanks for the tip on the new INA devices from TI. I'll definitely look into those.

[timb]

For AC or large DC currents, the ACS sensors are very nice.

True, but my case was very much like the OP's situation: I needed to sense currents as small as 10mA. I was using the most sensitive ACS712, the 5A model. It's output changes 185mV for every 1A measured. For 10mA, its output should change 0.00185mV, which is more than the Vpp measured on the output noise. Basically, I had a hard time telling noise from very small current changes. Like others mentioned in my thread a couple weeks ago: even if I amplify the signal with a opamp, I'll just be amplifying the error/noise.

Thanks for the tip on the new INA devices from TI. I'll definitely look into those.

How did you have the filter pin set?

[diyaudio]

You cannot use a ACS712  for low current measurements example look at the data for x05B 5A figure it has Peak Noise of 21mV, Ta=25*C @ 185mV/A 

Its important to read and understand the datasheet, nobody seems to be stating the obvious facts.

[diyaudio]

This will do the trick..
 
http://www.winson.com.tw/Data%20Sheet/WCS2702.pdf
http://www.ebay.com/itm/WCS2702-Current-Detection-Sensor-Module-2A-Power-Module-/400948963206?hash=item5d5a6ba786:g:~3QAAOSwgQ9Vk5q6

[Zero999]

You could use a mains transformer.

The ratio of current between the primary and secondary will be equal to the ratio of the voltages. If you have a 6V:120V mains transformer and connect the 6V winding in series with the load, then the current through the 120V winding will be 1/20 of the 6V winding. If a 400Ohm resistor is connected across the 120V winding, the voltage will be equal to 1V/100mA.


Note that just going by the transformer's rating won't give you the correct result. It's the voltage when no load is connected which will determine the turns ratio.  If your 6V transformer gives 9V with 120V on the primary and nothing connected to the secondary, then the current in the 120V winding will be 13 ¹/₃ of the current in the low voltage winding, so this will need to be taking into account.

[Helix70]

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

That isn't even in the ballpark of what you can do with inexpensive CTs these days. As I said before, a cheap modern CT can measure 10mA to 100A with an accuracy of 0.1% of reading across the entire range.

Being modern has little to do with it. The core material, burden and number of windings will determine its low current performance, as this sets the magnetising current. Using a high permeability core will be better for this application. A number of primary turns may also help.

[coppice]

A shunt will be your best option for light AC loads for up to 200mA, applications using a  CT is used for anything AC above 1A and more (generally).. Also the CT noise floor and gain used to sample currents as low 50mA AC with say a 10mA resolution seems unavailable.

That isn't even in the ballpark of what you can do with inexpensive CTs these days. As I said before, a cheap modern CT can measure 10mA to 100A with an accuracy of 0.1% of reading across the entire range.

Being modern has little to do with it. The core material, burden and number of windings will determine its low current performance, as this sets the magnetising current. Using a high permeability core will be better for this application. A number of primary turns may also help.

Try looking at what has happened to the core materials in the last 10 to 15 years. 10 years ago the expensive CTs from people like VAC were considered state of the art, but look really poor when compared to a 50 cents CT made in the last 5 years.

[Paul Moir]

This is just a follow up to say that the CT option worked out very well.  I ended up using the Amgis CT1005 (http://media.digikey.com/pdf/Data%20Sheets/AlfaMag%20Elect%20PDFs/CT1005.pdf) with a 500 ohm load and only one pass through the core.  Milliamp resolution was in fact quite easy and noise wasn't a problem either.  I did have a little high frequency noise from the load but a simple 100Hz RC filter cleaned that up completely.
Thank you again for all the comments!