Allegro ACS758 current sensor IC

T

Teh Stork

1 kW
Joined
May 25, 2011
Messages
463
Hi ES :)

I see quite a few diy inverters using Allegro ACS758 current sensor ICs for current sensing.

I've used them myself, the last batch I received went into a high prescision test bench for BLDC motors. Their use is for sensing DC and AC currents. I tested the hall ICs against a meter capable of 30A, but still with an accuracy of 1%. The results wasn't good, this was for the 200A version.

AC current, 50Hz:
~11% linear measurement error (measures too low).

DC current:
Nonlinear measurement error. 0% error at 15A and up. Nonlinearity between 0A and 15A, up to 8% error.

Quick check list:
- Minimum load resistance; Added 3k9 ohm to fullfill this point.
- Software to store zero point: yes - LabVIEW.
- Stable power supply: Upgraded to stable labsupply.
- Data Aquisition: CompactDAQ with NI9205 and NI9207
- RC filter: 10kHz, C=2.2uF and 45 ohm. C is far over reccomended 10nF max capacitance, sensor scoped for oscillations. Since it is in series with 45 ohm oscillations shouldn't spur - and they wasnt observed either.

Anyone had any similar experiences?
 
I've never used them, but have you tried measurements without the RC filter?

I think I have one of those, I may do some tests today or tomorrow.
 
Hmm, no - I'll try to squeeze that test in this week.
 
Internal resistance of the IC output is probably significant, so the R in your filter will be a lot more, which may be the cause of AC measurements being off.
 
Teh Stork said:
Quick check list:
- Minimum load resistance; Added 3k9 ohm to fullfill this point.
Click to expand...
In the datasheet it says:
Output Load Resistance RLOAD(MIN) VIOUT to GND minimum 4.7kΩ
Click to expand...
I think this means that the output impedance should be larger than 4.7kΩ. This is confirmed from the Allegro ACS758 FAQ found here: http://www.allegromicro.com/en/Products/Current-Sensor-ICs/Fifty-To-Two-Hundred-Amp-Integrated-Conductor-Sensor-ICs/ACS758/ACS758-Frequently-Asked-Questions.aspx

Q: What happens if I try to drive less than the specified 4.7 kΩ minimum resistance with the output of the sensor IC?
A: The sensor IC may not produce an output, as its output driver will not be able to supply sufficient current.
Click to expand...
This must mean that the resistance must be higher for the IC's output driver to be able to drive the load. So your 3.9kΩ is too much for the output driver to handle reliably, you should use a higher value than 4.7kΩ, or no load other than your voltage measure device.
 
You are correct futterama. Intoducing a load resistance increased the error furter. I thought a load was needed, but no - it is not.

Still, with this fixed - accuracy is not the same as a true analog current meter (calibrated to 1% in 2012, valid for 5 years). Error is still around 5% in the 0-25A range (200A range sensors). I think the sensors are affected by the switching noise inherent to the inverter.
 
Brushless motor inverter, it was used to measure phase current.
 
Their stated measurement range is 50-200A so it can be expected that the accuracy is less below 50A, but 5% is still a bit disappointing especially when this is likely to get worse when they are used in a real work situation not on a bench.
 
i have found that there can be as much as +-35mv of noise on the output ( this also varies between devices and 70mv is the worst I have seen)this will make a huge difference to accuracy on the higher reading ic's , you need to get the RC correct on Vout... the IC do not like driving Capacitive loads so there will have to be a compromise for a low pass filter on the output if thats what you want. also, as been already stated the accuracy will also depend on Vs, this can be be really bad if using a adc without taking a measurement of Vs for the adc and the IC before taking the current sample. With a little care on RC selection and thoughtful programming i have managed to get the error down to approx 2% @ 100A using a 200A sensor ( think it can be even better than this but im still playing), Vout is also not 100% linear and how bad this can get is dependent on temp
 
@Teh Stork - Did you redo the tests with suitable RC filter?
The 2.2uF and 43ohm would try to draw way to much current from the device Zc is 1/(2.pi.f.C)
And any spikie noise from supply Vcc will be feed through to sensor output and thE RC filter will look like a 43ohm load.
This overload can collapse the sensor internaly Vcc and the output will fall.
Try again and 1k5 and 10nf for 10kHz filter.

And be carefull/ check what the NI daq cards have as their in put impeadance and filters.

Which sensor did you test the 0 - 200 or +-200A one ?
Did you have a 0.1uF cap on Vcc to gnd at the sensor?

@ IanFiTheDwarf - The 50 - 200 A description is indicating seperate sensors models designed for selected ranges, not a sensors limited capability.

I hope this can be cleared up.
The spec suggests +/-1% error for mV/A gain of Nominal 10mV/A for 200B model.
Noise 3mV (or 0.3A)
The Vcc/2 0A offset +/-5mV (or +/-500mA )

If you scale your error for 25A instead of 200A that's a factor of 8. So 1% will become 8% !!!
Maybe you should use the 50 A model?

Good luck getting it sorted.
NB I've used the LEM devices but not the pcb pickup aleggro packs.

Is it worth editing th OP to indicate the device was way overloaded?
Has Lebowski got these on his demo board for his motor controller?
 
I'm using the 50A bidirectional one in my recumbent, and the 150A bidirectional ones in the big lebowski controller.

what amazes even me is that the 150A sensors in the big controller still allow precise low current control. If I set this
controller at 8A max phase current I still have a nice smooth variable throttle, the big sensors do not turn the
controller into an all-or-nothing at these low current levels.

I hardly filter the current sensor outputs before going into mu controller IC (1K and 10nF I believe), but there's lots
of digital filtering etc in the controller IC...
 
I am using the +/-200a version in one of my Lebowski controllers. I played with it on the bench a bit and was able to measure 1.0a when I ran 1.0a of current through the sensor. I then reversed the current direction and measured -1.0a. The highest current I verified through it was a ~150a pulse from a capacitor bank while I had a shunt resistor in series with it. I don't remember the details, but the sensor worked as advertised in my application.

I also verified it met the 5uS response time by sending 5uS on pulses through it at 25khz.

None of theses measurements were done with precision as I was only spot checking its performance.

I also found they have usable range of around -220a to +220a, but did not verify linear response beyond looking at the slope and the voltage it cut off at.
 
You must log in or register to reply here.
Back
Top