Compensating Vcc drift in 200A hall sensor output

[elektrinis]

I have an interesting project at hand. I need to measure DC currents up to +/-200A peak and 100A long-term.
My requirements are:
1. Board-mount, preferably SMD. Current is carried by PCB.
2. Low power consumption (below 100uA) or quick switch-on/enable with output settling in 100us.
3. Measurement error below 0.5%
4. Zero offset below 0.1A
5. Less than 1W of heat at 100A

http://www.allegromicro.com/en/Products/Current-Sensor-ICs/Fifty-To-Two-Hundred-Amp-Integrated-Conductor-Sensor-ICs/ACS758.aspx
So basically it lead me to try ACS758. The setup is as follows:
* Vcc is 5V, from a high voltage buck. It is required for several other devices and can't be changed to anything else.
* The sensor draws around 20mA, which is quite bad, so mosfet switch has been implemented to turn it off when not needed. Vcc filtering cap is placed before the switch to make it transition quickly.
* Signal is being read by ARM mcu, running at 3.3V (from extra LDO and external Vref). A resistor divider has been added.

However there is one big issue: output voltage at 0A is fixed at 1/2 Vcc, so I have a really bad drift of zero over time due to tiny fluctuations in PSU voltage, by tiny I mean 20mV difference throws off my reading by 1A.


So I am looking for new ways of solving this issue. I have a couple ideas:
1. add a precision LDO 5.0->4.5V just for the sensor. Then I would have more or less stable zero drift, but increased error and possible range issues.
2. add an external Vcc/2 divider and use it as zero reference for differential ADC. In this case I would have to look in to filtering and compensating ADC input impedance, possibly by adding extra opamp. If this has been done, I would like to see a reference design.
3. Use different hall sensor. Are there any compatible products from other companies, that have a fixed internal reference voltage? I know there are many options in SO8 and similar packages, that you just solder on top of conductor, I have used them before. However they are also quite sensitive to mechanical stress and need calibrating.
4. Use a shunt with precision opamp? I remind you my maximum allowed power dissipation at 100A is 1W.

Any ideas will be helpful.

[Alex Nikitin]

4. Use a shunt with precision opamp? I remind you my maximum allowed power dissipation at 100A is 1W.

Any ideas will be helpful.

I would go with this. 0.1mOhm shunt , 10mV output at 100A, 10uV offset, 50uV max error . Not that difficult and no problems with the power consumption.

Cheers

Alex

P.S. 0.1 mOhm SMD shunts are available from stock at Mouser, for example

[elektrinis]

4. Use a shunt with precision opamp? I remind you my maximum allowed power dissipation at 100A is 1W.

Any ideas will be helpful.

I would go with this. 0.1mOhm shunt , 10mV output at 100A, 10uV offset, 50uV max error . Not that difficult and no problems with the power consumption.

Cheers

Alex

Sounds reasonable.
Did anyone do it for differential current? I mean I will have to measure negative drop somehow, as opamps negative supply rail would be connected with one end of shunt.
Will look in to what's recommended by TI and others.
Thank you.

[Alex Nikitin]

Did anyone do it for differential current? I mean I will have to measure negative drop somehow, as opamps negative supply rail would be connected with one end of shunt.
Will look in to what's recommended by TI and others.
Thank you.

Again, nothing out of ordinary, a low power chopper opamp with a common mode range down to the negative supply (i.e. LTC2066 etc) in an instrumentation amp configuration with the gain of 100 and a bias from the Vref of your ADC, say 2.5V.

Cheers

Alex

[elektrinis]

Thanks a lot, I will look in to it. Looks like a good solution.

Our product has the ACS758 on board already. If talking about this particular sensor, Is an LDO the only solution to stabilize it a little?

[woody]

Maybe I miss something obvious, but why not run the ACS758 from the same 3.3V supply as the ARM? The ACS is a 3.0 - 5.5V device, and running the sensor and the (reference for the) AD converter from the same power supply makes sure that 0.5Vcc is the midpoint of the AD converter, no matter how much drift there is.

[elektrinis]

Maybe I miss something obvious, but why not run the ACS758 from the same 3.3V supply as the ARM? The ACS is a 3.0 - 5.5V device, and running the sensor and the (reference for the) AD converter from the same power supply makes sure that 0.5Vcc is the midpoint of the AD converter, no matter how much drift there is.

There were some other reasons why same supply supply rail could not be used, but after all these issues it now looks like a possibility. Will look in to it.


Also there is another hall device with 5V +/-10% supply that can't changed (750A LEM DHAB series). It does not have a reference output and is referenced to Vcc/2 just like the ACS758.
Now it is supplied from the same 5V rail via C filter and mosfet switch. Extra LDO is no longer an option, as we will end up too low. Is there a "correct" way for creating a precise Vcc/2 reference?
1/2 divider and opamp? Currently 5V rail is being measured and current reading is compensated accordingly, but accuracy is not spectacular.

[woody]

Oops, upon further reading I see that the ACS758 is not for new designs. The recommended substitution is the ACS770x, which has 4.5-5V single supply operation. So there goes my solution 

[elektrinis]

I wonder why there is a replacement. Can't see any difference on a first glance.

[David Hess]

Maybe I miss something obvious, but why not run the ACS758 from the same 3.3V supply as the ARM? The ACS is a 3.0 - 5.5V device, and running the sensor and the (reference for the) AD converter from the same power supply makes sure that 0.5Vcc is the midpoint of the AD converter, no matter how much drift there is.

Hall effect sensors are ratiometric so using the same voltage for the hall effect sensor's excitation and the ADC reference will remove gain errors.  To make an offset measurement, a pair of precision resistors can make the hall effect sensor look like a bridge transducer with a differential output.

[Alex Nikitin]

Oops!  I failed to factor in the +/- current spec-- so, you will need to use the version of the INA240 that has a gain of 100, and also use the built-in resistor divider in the INA240 to offset the output to 1/2Vcc.  Then you calibrate as above-- with "zero" being some very stable ADC value (near 1/2 Vcc), and then in +/-10A steps all the way to +/-200A.  Done!

The INA240 also needs over 2mA of supply current. As I've said, a micropower chopper RRIO opamp with the gain of 100 or maybe 150, we don't need a lot of CMRR, so a plain four 0.1% resistors arrangement should work just fine (1K+100K/150K), with a bias from a 2.5V reference (ideally the same as for the ADC).

Cheers

Alex