2
$\begingroup$

I'm building an autonomous sail boat (ripped out the guts of an RC sail boat and replaced with my own mainboard etc.)

The controller board I have can accommodate both an MPU9150 and an HMC5883. Is there any advantage is using both magnetometers for a tilt-compensated heading? I'm thinking that I could compute the unit vector with soft/hard iron offsets removed for both, and then average the two vectors to get one slightly better one?

Not sure if it would yield a better result though.

$\endgroup$
  • $\begingroup$ One advantage of using only the sensors from the MPU9150 is that you wouldn't have to do sensor alignment. $\endgroup$ – marcv81 Oct 14 '14 at 10:08
  • $\begingroup$ @marcv81 very true. The funny thing is, though, that the accuracy of the built-in mag is substantially worse than the HMC. As a result, you may not get any better results, despite eliminating the cross-axis alignment issues. $\endgroup$ – kolosy Oct 15 '14 at 17:45
  • $\begingroup$ I'm not familiar with the MPU9150, but I've setup a MPU6150 + HMC5883L on a quadcopter. I only used the magnetometer to compensate for the gyro drift. The HMC5883L is good, but I can't see how worse it would have been with a poor compass: after all it only needs to provide vaguely consistent results on average and at a slow rate. $\endgroup$ – marcv81 Oct 16 '14 at 9:26
1
$\begingroup$

A brief google search turned up this paper on using multiple magnetometers. Depending on how far out at sea you expect to deploy this sailboat, you may want to consider a fallback mode where if the 2 compasses diverge by a certain amount then you consider one of them to be malfunctioning (since the average would be very inaccurate).

$\endgroup$
  • $\begingroup$ The link doesn't seem to show. Do you mind editing it back in? $\endgroup$ – kolosy Oct 7 '14 at 17:40
1
$\begingroup$

Not really.

With careful planning, experimentation and testing to minimize effect of servo motor magnetic field affecting the sensor (by placing sensor as far away as possible and mathematically iron compensation via angle table), one should be able to achieve accuracy of a few degrees, which should be sufficient for the said application. (This excludes the case of sensor error detection via multiple sensors)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.