Steven Saner <[email protected]> writes:

> It would, therefore, be cool if the altimeter could in some way detect
> the apogee failure and trigger the main ejection early. Obviously, this
> defeats the intention of the dual deployment, but I would much rather
> have to hike after the rocket than for it to be destroyed and/or come in
> hot and unsafe.

I've thought about this a bunch, and in fact all of the Altus Metrum
flight computers have some idea of their descent rate on the way down
already; computed as a part of the Kalman filter outputs.

The big issue is that with only a barometric sensor in a tumbling
airframe, you've got a *lot* of noise in the data, so computing anything
like an accurate speed turns out to be a bit tricky. You can see what
the output of the Kalman filter looks like by plotting a .telem file --
the telemetry data includes the output of the Kalman filter while the
.eeprom file does not. We typically see a range of 50-100 m/s in speed
reported this way; that's what AltosUI or AltosDroid report for descent
rates, and you may have seen the crazy numbers from that...

What we don't have today is a bunch of high-precision descent data. We
reduce the data recording rate during descent from 100 samples/sec down
to 10 samples/sec to save flash space. This means that most of the data
used for the Kalman filter on the rocket during descent is not stored
anywhere. This makes doing development pretty hard; without any real
data, it's impossible to test a filter on the ground.

So, in order to make this work, we'd first need to hack the firmware to
capture high-rate descent data from a variety of flights. Then tweak the
Kalman filter parameters to generate reasonably smooth speed
data. *then* we could adjust the flight software to include a 'hail
mary' mode.

-keith

Attachment: pgpMyZXnILNIa.pgp
Description: PGP signature

_______________________________________________
altusmetrum mailing list
[email protected]
http://lists.gag.com/mailman/listinfo/altusmetrum

Reply via email to