Gradiometric Wheatstone Bridges Calibration

[Guille]

Hi everyone,

We are building a gradiometric magnetic field detector. The detector is based on a certain magnetic Wheatstone bridge. Two bridges are placed at different distances from a magnetic source, both signals are amplified and substracted by INA-217s, thus the common noise is reduced, the problem I see with this gradiometric setup is that both signal chains before substraction have to have very similar Gains and Offsets for the common noise to be efficiently removed. I'm tasked with optimizing this problem.

The biggest offender here is the Wheatstone bridge, it's resistors are not properly matched and thus has an offset <  /-20mV/V (for every volt of supply you apply to it's leads you'll get at max 20mV offset signal), additionaly it's sensitivity is 7~15mV/V/Oe (For each volt you supply you get 7~15mV signal for each Oersted of magnetic field applied).

Each sensor has a different sensitivity which is problematic for the gradiometric reading (unmatched gain). One way to adjust them is by changing the excitation voltage (Ve) fed to them so that both sensors output the same voltage for a common magnetic field. Like so:



Each sensor has a different offset voltage, one way to adjust for this is to introduce an oposite offset voltage into the REF terminal of the INA like so:



Wheatstone signals come from Sens 2 and Ref connectors: They are amplified by both INAs and one of them can be offset through the trimmer in the boxed opamp stage to match the other offset, ignore that offscreen INA. Finaly both signals are substracted by the last INA stage.

The Calibration Problem

The problem with this design is how both the sensitivity and offset of the sensor depend on the excitation voltage, let's say you do the following calibration:

• Adjust Offset: Place both sensors in a mu-metal box, measure the offset and adjust with trimmer
• Adjust Sensitivity: Measure output of system and adjust voltage fed to one sensor until both outputs are the same

Now that you've changed Ve the offset of that sensor has changed, you need to repeat step one. Additionaly you cannot be sure that you matched both sensitivites with step 2, who cares if you got the same ouput for a given magnetic field? If the offset changed again teh output matching for a given magnetic field won't guarantee that their slopes (Sensitivities) are the same.



Solutions

I thought of a couple of solutions, I would like to have some advice on which do you think is better or if you have any different ideas:

• Performing Offset and Gain adjust after the sensor: Adjust gain on the INAs with a trimmpot in series with RG, thus offset will no longer change when adjusting gain, cons: tempco, and drift of the trimmpot will f up the calibration over time
• Iteratively adjust Offset and Gain: Go through the steps described above iteratively and pray I reach some sort of optimal value, cons: I don't even think it's guaranteed I'll reach it
• Measure slope of response curve (Sensitivity) with a controllable magentic field: Measure response for different magnetic fields and calculate both slopes, adjust Ve to match sensitivities, measure offset and adjust, cons: Complicates the calibration setup a whole bunch introducing coils and current sources, this needs to be done on a hospital
• Calibrate Offset and Gain only in digital domain: No tempco/drift problmes with adjusting Gain by trimming, cons double the DACs, plus this sensor will eventualy be an array which makes that doubling costly.
• Adjust sensor resistors with parallel trimpot: This kind of offset calibration is not affected by Ve, however again I fear tempco and drift problems
• Wheatstone AC measurement techniques: No offset calibration needed, only need to calibrate sensitivity, cons I need to get a specialized DAC that can do this, any recommendations?
• Your awesome suggestion: ?

Any help, input is appreciated as I'm kinda stuck. Thanks for the interest.