EKF Variances and crazy PFD

Link to log: https://mega.nz/#!zhY2AYbA!8KmJ3HaI5dL4Be31NoORfJaTu4J4cDBKVNTVfor6hq4

Hi all,

I have been flying for about 4 months with Pixhawk and a custom airframe, weighing about 10kg fully loaded. Every so often I get severe EKF errors such as velocity and position variance. I attribute these errors to GPS glitches, because in the map you can see the plane jumping from place to place. These errors result in severe problems with the PFD: the pitch and the roll can be seen jumping all over the place and the plane has to be taken over in manual mode. The pitch can go from -70 to +70, something this airframe would never be able to do. Sometimes it happens when on the ground (when in FBWA, the control surfaces go crazy for a few seconds), sometimes it happens when flying. Most of the time I don’t get this error at all and the plane flies beautifully.

When I fly with my Skysurfer, I sometimes get EKF errors but the PFD is still normal and the plane can still navigate without problem.

The NK* values in the logs are all over the place, and I can’t pinpoint what could possibly cause this problem. The Pixhawk is located at the middle of the airframe and affixed using Velctro. The wingspan is about 3 meters and the power source is 2 8000mAh 6s batteries in parallel. The magnetometer and the airspeed sensor are not used. I have never had any problem with vibrations with this airframe. I believe it is well tuned as I already made many tuning flights. The GPS is mounted on top and has a clear view of the sky, but I believe it has trouble during turns.

The Pixhawk itself is a genuine 3DR product, as is the GPS (Neo7). It is about 6 months old. We use custom made connectors that are microscope inspected.

I hope I provided enough details and I’d really be happy to have some suggestions. thank you very much

PS The attached log is a bit long, sorry about that, but it takes us a while to takeoff for many reasons

Hit a decryption key required for download…

Posted new link. sorry!

This appears to be related to a bug in EKF2 when flying without a compass. Paul has a patch in and a hotfix version of Plane 3.5.3 will be coming out fairly soon to address this issue (as well as a bug with negative loiter radiuses). In the meantime flying with EKF1 rather then EKF2 for a no compass setup would avoid the bug.

Further review from @tridge or @MagicRuB would be welcome of course, but this looks to be like several other reported crashes with compass disabled.

I’m pretty happy to hear that! I’d like to know if activating it requires changes to any settings other than the filter selection, because we won’t have any test flights before the big flight! Thank you