On a local forum, we discussed the possible causes of an elevator or arduplane malfunction that unfortunately could not be reproduced in subsequent flights, but which occurred 4 times during one flight, causing the aircraft to each time raise up or lower down its nose by up to 90° for no apparent reason.
An FPV video of the flight exists: https://youtu.be/qa0ujWTrmYk
(see 0:19, 0:51, 0:58, 1:09)
Log file: MagentaCLOUD (00000031.bin)
For comparison, the log of the following, completely uneventful flight: MagentaCLOUD (00000032.bin)
The plane is an Albabird converted to a QUAD/X (both front motors tiltable), Mateksys F405-VTOL (firmware target F405-TE), ArduPlane V4.5.7 (0358a9c2).
In order to help you find your way around the log file more quickly, I have provided a graphical representation of the events:
The aforementioned sudden pitch ups and downs occurred a total of 4 times (marked A, B, C and D in the following diagrams) before the pilot decided to safely land the VTOL with Q-Loiter.
B and C occurred in pure FBWA fixed wing mode, A and D shortly after or during a transition. With the exception of B, all pitch events were so far off target pitch that Q-Assist cicked in. Airspeed (in km/h) and pitch are plotted on the left ordinate, while the PWM values of the tilt servo and one of the tail motors are plotted on the right ordinate to make it easier to see in which phase and at what time the events occur.
I have omitted the tilt servo in the following diagram, as the flight phase can also be seen from the PWM value of the tail motor. The PWM values of the elevator servo (elevator, purple) are shown:
In each case, the elevator fully counteracts against the disturbance. Only at B was the disturbance not strong enough to activate Q-Assist.
The events are enlarged here: You can see that the aircraft rises shortly after the tail rotor is switched off at the end of the transition and the elevator is fully extended.
I initially had the impression that the elevator deflection could have triggered the pitch-up, as the roll change seems to occur later. Apart from the fact that the effective direction of the elevator is nose-down at higher PWM values, the Attitude control already reacts to changes in the rotation rate, and these can be recognized much earlier as they are not fused first: In the following graphic, I have entered the rotation rate around the transverse axis (GyrY) (multiplied by 50 for better visualization). You can also clearly see the different frequency of the less frequent recordings of the servo outputs compared to the IMU recordings:
The second event (B) occurs in the middle of fixwedwing operation (fbwa):
The third event (C, nose up) also occurs in pure FBWA flight. However, Q-Assist is triggered.
While the Q-Assist triggered by event C is still being discharged, a new malfunction occurs at a downward angle of almost 90°:
Q-Assist kicked in, almost simultaneously the pilot has activated Q-Loiter and landed the VTOL safely in q-mode.
In the discussion, we were ultimately unable to find a fault in the control by Arduplane, so we assume sporadic servo failure. As all previous and further flights with unchanged hardware have so far gone without a hitch, I have posted the logs here in order to possibly draw attention to further considerations.
Rolf