Mirek6
First Officer
Flight distance : 609724 ft
Canada
Offline
|
JJB,
Good questions and analysis.
You say: “Distance to HP does not change much in the period 4m8s and 4m16.s , within 1 meter.
I don`t see high variations in the LatLon positions and drone pitch and roll angles are normal for a GPS stabilized hover.“
Yes to first sentence. However variations of pitch and roll are there. Consistent with unstable air (wind) and Spark hunting for position, or both.
Pitch variations in 12 seconds switch semi-randomly from -7.9 to plus 11 degrees (quite a wobble), and roll is between -3.8 to 7.7 degrees.
At 4m 1.8s the craft has finished its decent and it should hover in place with strong GPS signal. Even if there is some breeze.
Instead, its starts hunting in a circle. I suggested both wind (its circular path is somewhat consistent with wind direction in which Spark flew away a bit later) and magnetics interference from rock containing hematite or magnetite.
If it was gust of wind only, Spark would overshoot its commanded position and than start returning along ellipse while being pushed by wind and constantly correcting its position.
Here, Spark circled. Either wind was shifting (possible) or we had a considerable mismatch between IMU and compass causing toilet bowl effect.
I concluded that it was speed error since, as you mentioned, GPS position was hardly changing yet calculated GSP speed was very different than horizontal speed. I do not know how both speeds are derived. But they did not match.
We have also seen before that speed/yaw/magnetic errors are linked. Complex circuitry on board of Spark may misinterpret some mismatching signals erroneously. For example, if Spark is confused because of magnetic interference, its controller may throw in yaw errors, weak GPS signal etc. Once problem begins and deepens, controller and firmware may get confused. This is why I usually do not take any particular metrics at face value but compare to other metrics and try to see patterns.
Thank you for explaining how speed in logs is derived – makes sense. If we knew these details, the analysis would be easier. In this case we have issues with Spark combined with possible inaccuracies of how logs are calculated.
Yes at 4m2 to 4m 14s Spark was definitely flying low. I confirmed it with video. Magnetic interference is distinct possibility.
DJI should also look at stability of this Spark’s IMU. The final root cause may be a perfect alignment of several problems – some wind, some magnetic interference from rocks, some issue with IMU and firmware which got all confused.
One last thing – did you notice that Spark switched to ATTI around 4m 15 seconds (starts its drift with no stick input – outside of pilot truing to raise it up) yet logs show P-GPS until 3 seconds before crash?
Cheers!
Mirek
|
|