DRV421: The relationship between the size of Rshunt and the drift of zero point

Stuart,

Thanks for using the forum. From your information on the core I did a rough FEMM simulations and got a gain of about 500uT/A. From this we can do the same calculation like the other thread.

If your core gain is 500µT/A at our sensor location. The error on the primary current measurement will calculate as follows. If you give me more details on your internal shield we can get a better simulation.

8µT/(500µT/A) = 16mA (Primary current error)
This they calculates as 16mA/1000 = 16µA on compensation coil
With Rshunt = 1kΩ there is a 16mV at the input of the differential amplifier
Therefore after the differential amplifier you can get 64mV.

This is a little higher than what you are currently getting but it is not that far off. Please remove the large dummy shunt resistor and see if you get better results. Sometimes the imbalance of the Rshunt and load on the H-Bride does not work great. If you can place a resistor on other ICOMP pin that is not connected to RShunt to see if this helps. You would also need to bypass this resistor with diodes as you will limit the max current during degauss.

I do not understand the large offset. I have never seen anything like that. Please make sure the input voltage is zero. You could connect the Vref pin to the Rshunt when you disconnect the coil to make sure the voltage is close to normal operation.

Hello Javier,

Thank you for your suggestions I will try them out shortly.

One thing I had noticed is that there is up to 38mV p-p noise on the ICOMP pins. As I am a fan of trying to trap noise at source would you advise a filter incorperated into the resistance I place in series (bypassed by diodes) at the same time? If so what sort of roll off frequency is advised?

Best regards,

Stuart

Stuart,

The DRV421 samples the flux at about 250KHZ.  I would filter the signal only going into the Differential amplifier.  If you would like to filter that signal coming from the sampling rate of the fluxgate please place something below 250KHz.  I would place this filter so that it does not change the response of the transient signals from the coil but rather only going into the signal path. 

Hello Javier,

I have paced a 4k7 resistor with back to back diodes across it, in series with ICOMP1 as well as changing back the dummy shunt to 43R. I am still using a using the DRV421 evaluation module for my testing.

Doing this, with the reference voltage measured to be 1.6521V, the output voltage after a demag cycle was at 1.631V and over a period of 3 mins rose to 1.656V and maintained this for the next 5 mins +/-3mV.

Power cycling without DEMAG enabled the output was 1.692 falling slowly to 1.682 over the next 3 mins.

As a sanity check I removed and disconnected the compensation coil (after powering down), and the output  was found to be 1.5274. With Vref (1.6521V) connected to AINP the output changed to 1.5271V. With Vref connected to AINN the output was 1.1 mV.

For reliable operation does the shunt resistor need to be much lower and then utilise an external amplifier to bring the output up to a suitable level for our ADC?

I also looked at the TI document sboa199.pdf which intimates that this part was a good fit for the application of a class B RCD relay, being able to measure in the less than 30mA region. Do you have any further details on the best approach to achieve this in a robust manner.

Best regards,

Stuart

Just to note the Vref was connected via an 820R resistor to the AIN in each case for the sanity check.
As a further check I placed a short across the shunt resistor and the output was 1.6502V

Best regards,

Stuart

As a further note: I checked an unmodified DRV241-EVAL board (apart from the swapping of R3 and R4) and found the reference to be 1.6528V and the output was 1.6500V

Best regards,

Stuart

Stuart,

From all the number you shared I do not think that the part is much off from the values I would expect. I will address two of the items that I believe more information.

"As a sanity check I removed and disconnected the compensation coil (after powering down), and the output was found to be 1.5274. With Vref (1.6521V) connected to AINP the output changed to 1.5271V. With Vref connected to AINN the output was 1.1 mV."

Completely disconnecting the compensation coil will open the loop. The only thing make sure there is a forced voltage on AINN and AINP. if you connect the pins with a resistor or the Rshunt then connect one pin to Vref that would be ideal for verifying the offset of the Differential Amplifier.

"For reliable operation does the shunt resistor need to be much lower and then utilise an external amplifier to bring the output up to a suitable level for our ADC?"

You can use either method and it is a matter of trade-offs. I will mention if you are having signals that are higher in frequency normally our DRV421 differential amplifier has better CMRR which is crucial for high frequency changes as that will equate to high common mode changes at the inputs AINN and AINP. If you are going with low frequency signals you may not have to worry about it and either method should work.

Hello Javier,

Thank you for your patiance and support.

My concerns over the amplifier offset is that this design is for a 30mA Class B modular circuit breaker and would have to be calibrated at the factory and the only further demagnetisation oppertunities would be in special documented circumstances not available to the end user.

One major concern is repeatability of the up to +/-8uT offset of the fluxgate sensor front end. Is this figure a function of the chip and can be calibrated out by the host system or does it vary within those limits over time/conditions? I see to minimise the effect of this parameter we would need a much higher core gain. I am in disscussion with our core manufacturer for much higher permeability materials.

The behaviors I am experiencing that are concerning me the most with the core that we have been using is that if I do a demag cycle and use the resulting output as our reference level and then apply a 30mA DC current for a period of time and then remove the primary current the voltage output equates to a current of -10mA. When 30mA is then re applied the voltage output only equates to 20mA. This behavior is repeatable in various combinations in both directions.

I see from the SBOA199 tech note that this device is suitable for a 30mA RCD so I would expect an absolute measurement accuracy of at least +/-3mA at turn on without a demag cycle, as we have to react within 30mS of start up. Is there a reccomended combination of core gain, compensation turns and shunt resistor for this application.

Best regards,

Stuart

Hello Javier,

Do you have any thoughts on my questions and observations yet?

Best regards,

Stuart