Re: [Paparazzi-devel] Arduimu test flight #2

[Christophe De Wagter]

Hi Chris, nice to hear about the developments. Here are ours: 

a ) Since July we have done many many hours of IMU flights in all weather conditions (from 0 to 14m/s wind, from -8 to +30 degrees C, with and without turbulence, with circle radius's of 100m down to 20m: please mention your flightspeed + circle radius when doing IMU tests) in a 15m/s cruisespeed TwinStar with an XSens MTiG connected to a YAPA autopilot. This is considered as the expensive but most reliable paparazzi setup we know of. Big connectors, mounting holes for everything, IMU well calibrated including cross axis + non-linearity analog devices, Ti DSP in XSens and simple safe code in the paparazzi. When compared to for instance the latest MicroPilot you end up with similar(/better?) performance  but with the amazing open-source flexibility of paparazzi! (PS: the XSens is not as good for quadrotors because it is relatively slow: 100Hz and quadrotors benefit a lot from faster innerloops). 

b) This week we also flew with a different setup. Old MicroJet + Paparazzi Tiny 1.1 (4-layer) + new paparazzi YAI IMU (=Booz imu with cheap Invensense sensors). The fixedwing code was modified to allow for much faster rates (Tnx Felix for merging our branch into master now). We flew with 960Hz IMU measurements, 240Hz AHRS propagation and 60Hz AHRS correction all inside the Tiny. The IMU was calibrated including cross-axis components. The expensive magnetometer was not soldered yet but we did use the wind-compensated heading instead of gps_course to correct the heading which in all cases improves the estimates. In the agile and not always perfectly stable microjet (e.g. microjet has fast pitch oscillations when pulling too much at low speeds) turn rates of 400 degrees per second do occur but thanks to the fast sampling and integration these short peaks are integrated quite well. [http://www.youtube.com/watch?v=q56BqmaaYYg http://www.youtube.com/watch?v=xq0S69zwEBE]. This plane also flew in December at minus 8 degrees. The gyro's only had a visible bias but no visible scale error due to temperature and were no problem. The Accelerometers also have a temperature related bias which was now corrected manually using the imu_pitch and imu_roll variables. Only other problem so far was the aligner that did not work when inside the car with running motor. I think the vibrations are too high while the aligner waits for steadiness. Furthermore it really helps to let the plane get the right temperature for a (few) minute(s) before takeoff.

We hope to test-fly a nice temperature calibrated CHIMU soon, which should solve many issues. 

Keep posting you IMU related achievements!

-Christophe

On Fri, Feb 18, 2011 at 8:37 PM, Chris <address@hidden> wrote:

I forgot to say that todays flight included flying with more than 10m/s wind.
That is about 5 to 6 Beauforts (36km/h up to 43km/h)
The circling of course was a bit distorted, well actually an oval course but the plane managed to fly well
and follow it's flight plan consisting of following 3 waypoints in a line.
At end of the flight a parachute was used for recovery and you can imagine me searching for the plane in a field as the wind carried the plane 200m away from the runway.
I am writing a function that estimates the parachute release point and although the function was telling me that the parachute should be deployed 200m away, i didn't believed it and decided to go for a manual deployment.
It was a big mistake :-)
Chris


_______________________________________________
Paparazzi-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/paparazzi-devel