How to calibrate multiple magnetometers for mass production

[samf48]

How would one go about calibrating several magnetometers in a production environment? The magnetometer is being used for the purposes of a compass.

Currently I'm calibrating them one-by-one in an open field free from magnetic interference (i.e. no hard or soft iron around) by rotating my magnetometer, and associated electronics, in all three dimensions and then compensating, through software, for the inherent magnetic interference of the on-board electronics.

I'm curious though how large companies scale this process without introducing magnetic interference. I am also assuming they don't go outside into an open field and do this so I wonder how you can control magnetic interference inside a building presumably with ferrous metal in the walls.

 

[MacGyverS2000]

I would imagine a fixture and a strong magnet. The fixture ensures the device is pointed at the faux magnetic North (the magnet) and calibration is performed. A nearby strong magnet would swamp out any Earthly magnetic field lines.

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

The typical fixture uses a Helmholtz coil, which provides the artificial magnetic environment, and can be used to perform all sorts of testing on magnetometers. I'm surprised you haven't scoured the websites of your competitors for such information.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[samf48]

Thanks for the replies, I guess the key is creating enough of a magnetic field so much so that I can rely on it drowning out the other sources of magnetic interference in the room. Believe it or not competitors don't post too much about their calibration techniques, I did find some whitepapers on using the 'least of squares' method of calibration which is what I am currently doing.

 

[IRstuff]

"Calibration" actually involves a number of calibrations:
> accuracy and matching of field intensity measurements
> orthogonality of the 3 axis magnetometers
> alignment to fiducials of the case

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[VEBill]

Using a strong local magnet will mask the local magnetic environment, but a very strong local field will also mask "the inherent magnetic interference of the on-board electronics" which was in the first post.

There's a clue in the universal approach used in mobile phones, where the user is directed to move the phone around in a Figure-8 pattern of movement to calibrate the internal magnetic compass. It seems obvious that such an approach would allow the software to adjust-out the device's own field (as it would be a constant offset during the movement).

I wonder how they deal with variability of the device's own field? More software? A bespoke magnetometer system might be designed with a constant current (in the circuits) to avoid this question.

A strong local field will allow calibration of the relative alignment of the sensor onto the device's circuit board. That would seem to also be a requirement as part of the Factory ATP.

 

[samf48]

Currently using a similar method to compensate for the hard-iron and soft-iron biases from my onboard electronics: [URL unfurl="true"]https://github.com/kriswiner/MPU6050/wiki/Simple-and-Effective-Magnetometer-Calibration[/url]

This method though, also looks at the field around the device as well as the magnetic field generated by the internal electronics. Since the end application is a digital compass, the readings are taken in a outdoor environment so as not to skew the final compass reading when it is used outdoors.

 

[IRstuff]

The Helmholtz coil is used to swamp out all external influences.

What accuracy are you trying to achieve?

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[samf48]

Pretty loose accuracy of +/- 15° for the end application of the compass...

 

[IRstuff]

Oh, OK, then most of what I listed are barely needed. Generally, then, a "tumble test," will get you pretty much everything you need. You could do that either with a Helmholtz (see below) coil setup or just reasonably far away from other things. A tumble test basically involves rotating the compass throughout 3 DOF in angle space; no translation is required, per se. The collected data can then be used to calibrate for both field strength and orthogonality. There's a Matlab file available for download that does calculations. You can use a 3-axis gyro to get the angle measurements.

This is similar to what we used for compass calibration, but the actual volume of controlled magnetic environment is pretty tiny.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[samf48]

Looks like there are some pre-packaged set ups out there: [URL unfurl="true"]https://matesy.de/en/products/magnet-measurement-calibration/helmholtz-coil[/url]

Since the goal is to do this indoors, could I simply perform the tumble test using the Helmholtz coil with my current method of calibration, with respect to the Least squares method? Basically I'm wondering if I can perform the calibration using the Helmholtz coil without having to change my software. I'm guessing that I probably can't do that because the strong magnetic field will skew the current calibration curve.

 

[IRstuff]

What "current calibration curve?" The tumble test will result in a complete set of calibration coefficients.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[VEBill]

If the UUT's own internal currents (and thus its own self-generated fields) are varying, then the software may need to include a table or algorithm to adjust the resultant internal corrections.