Max voltage for Hall effect current sensor board

[bitwelder]

Hello,

I have a project in mind about a uC device that monitor/datalogs currents (at mains voltage level).
As active component I found an Hall-effect sensor (Allegro ACS712 chip).
The datasheet of this part says that the chip itself has a 2.1kV insulation and all is fine and dandy, but what I'm interested at is the PCB where the chip has been mounted (see pictures).

Looking at the track creepage, what would you rate as suitable max voltage for the primary terminals?

Thanks!

P.S. here is the link with PDF of the Allegro demo board, a bit more spacious than what I have :
http://www.allegromicro.com/Products/Current-Sensor-ICs/Zero-To-Fifty-Amp-Integrated-Conductor-Sensor-ICs/ACS712/~/media/Files/Demo-Boards/ACS712-85-0322_Layer_Prints.ashx

[SeanB]

50V max. Lose the ground plane around the whole top section of the board around the input pins, and make a 2mm wide slot under the chip extending at least 1mm out each end of the chip to provide increased creepage distance.  Board will have to be a little wider to keep it as strong. To use it with mains you will have to clean after assembly and apply a conformal coating around the IC to keep it clean and dry.

[David_AVD]

I have a potential use for one of these ACS712 devices, but have a question about the isolation voltage.

It will be measuring the current drawn by a small (< 3A) 240Vac mains motor.  There will be external connections for controlling the motor via an RS485 port.

In the Allegro FAQ, they go on about basic and reinforced isolation.  What does this really mean when it comes to isolation between the mains and the ELV portion (RS485 port) of the design?

[M. András]

for higher voltages i would use the bigger 758 sensor there is enough clearance between the legs so a nice new board os solder the wires directly to the legs is a possibility btw its approx 3-4mm wide and 2-3mm thich each

[David_AVD]

Hmmm... just checked the 758 out, but it's only 40mV output per Amp sensitivity compared to 185mA per Amp of the 712.   

The motor current in some applications may be less than 1A, so the 712 would give me approx 100mA resolution with an 8-bit A2D.

[M. András]

Hmmm... just checked the 758 out, but it's only 40mV output per Amp sensitivity compared to 185mA per Amp of the 712.   

The motor current in some applications may be less than 1A, so the 712 would give me approx 100mA resolution with an 8-bit A2D.

and you have to filter its output pretty well the thing has nice amount of noise

[David_AVD]

There is a filter pin which the app notes say to use instead of trying to R/C filter the output directly.

One application I have in mind just needs to know if the motor is pulling current. The other would benefit from knowing if the motor was unloaded (broken mechanism) or overloaded (jammed mechanism).

[alxnik]

For AC monitoring you can go with inductive sensors which are completely non-invasive so you don't need to care about creepage. Even for DC you can find non-invasive hall effect sensors - I think honeywell sells a lot of these - and be done with it.

[David_AVD]

I had a quick play around with a 20A version of the hall effect chip (eBay PCB).  It works as advertised, but you sure don't want to get a magnet near those things - massive changes in output (as expected).

It looks like I'll go with a small PCB mount CT feeding into a LTC1966 RMS converter.  That seems more appropriate considering some of the actuators are actually low voltage DC motors with a SMPS integrated into the outer casing.

[M. András]

i played with mine with the hanger magnet for my fluke meter. it goes way beyond the fullscale 200A output voltage when the magnet is touching the epoxy, then it recovers to almost zero current output then if i flip the magnet to reset then it goes to the midscale range. i have the bidirectional version so its 2.5v. the appnotes and datasheet have information about when subjected to full scale current residual magnetism remains in the core. wondering about how could you simple reset that. didnt brothered with power off and on. but could be a small electromagnet on top of the device to reset the magnetisation

[Maturola]

I had designed the ACS712 and 711 into industrial load control units, they had been tested and UL approved to work for up to 240VAC load at the primary terminal. the 2K+ insolation is between the primary and the Sensor.

[M. András]

I had designed the ACS712 and 711 into industrial load control units, they had been tested and UL approved to work for up to 240VAC load at the primary terminal. the 2K+ insolation is between the primary and the Sensor.

what resolution did you archive with it?

[Maturola]

About a 180mV per A

[David_AVD]

I had designed the ACS712 and 711 into industrial load control units, they had been tested and UL approved to work for up to 240VAC load at the primary terminal. the 2K+ isolation is between the primary and the Sensor.

In that FAQ I linked to they mention 2100V isolation.  Then they seem to say it's good for 240Vac applications where there is a protective earth (basic isolation), but no external connections (reinforced isolation).  Maybe I'm misunderstanding what they say?   

[Maturola]

I think you are confusing the Isolation between the Load side and the Sensor side (which is 2.1 kVRMS minimum isolation voltage from pins 1-4 to pins 5-8) and the voltage that is supported by the clearance of the input pins.

This sensors are place in line with the current you want to measure, there is not voltage potential between the input pins when current is flowing.

[David_AVD]

They also mention the metal slugs ( connected to pins 5 and 8 ) protruding each end of the package.  I checked the actual chip and you can see bare metal on each end.

So this bare metal is 2mm from pins 1 and 4.  How does this impact the isolation on a real board?

The proposed device will have an exposed RS485 interface.  I was wanting to keep the signal 0V floating with respect to mains earth to avoid loops (via the PE) when multiple devices are interconnected.

My take on it was that there is only 2mm isolation between the current sense side and the output side due to those end pieces.  Cutting a slot below the sensor helps with the creepage figure between the SOIC package pins of course.