Magnetometer Meeting Notes March 28, 2007

Resolution on Magnetometer Resolution

So, we arbitrarily decided that 1 degree resolution on the new magnetometer is OK and here's why:

the magnetometer's job is to de-bias the gyros, which over time, doesn't need to be that accurate since it just needs to pin gyro drift.
our pointing accuracy doesn't need to be sub-degree (this might even extend to accuracies necessary for orbital insertion trajectories).
we can keep it cheap and still our 1 degree resolution... shouldn't cause us to need a lot of crap like external 16b ADCs (in otherwords, stick to the ARM's internal ADC and use cheap op-amps and make Jason deal with it).

Note that the accuracy of the system is dependent on our magnetic models and sensor calibration.

We'd like to the magnetometer to do +/- 2 Gauss. We think this is enough to deal with local hard iron interference, and gives us enough slop for temperature and whatnot.
- This means that our range is 4 Gauss, really maybe even 5 Gauss with slop says Tim, which now means we only get about 2 degrees per bit at 10 bits.
- This forces us to oversample: from http://en.wikipedia.org/wiki/Sigma-delta_modulation,
- rms noise voltage = (quantization noise) / sqrt(oversampling factor)
- So, if we want to take our 10 bit ADC (quantization noise = 1 * 10^-3) and change it into a 12 bit ADC (quantization noise = 244 * 10^-6) , that means that we need factor of 16 oversampling ratio. That means for a bandwidth of 10 Hz, we need 2 * 10 * 16 = 320 Hz.

Jason is going to look for other magnetic sensors besides honeywell: Jason, go ahead and put together an email and mail to it to Tim who will put it on the SensorComparison page.