Strange Roll behavior in stabilize mode

Equipment: Standard quadcopter with APM1

Hello,
After the update to v3.1 I found following strange behavior:

I hovered and I gave some impulses to the transmitter roll axis in order to check the PID quiality.
Please see the attached sreenshot from the log (RollIn:red and Roll:green), after some RollIn impulses the copter suddenly leaned strongly to the opposite direction (40°).
[attachment=0]Crazy_Roll_.jpg[/attachment]
This behavior can be reproduced. I tried it with different Rate PID settings (No D, No I, LowP, highP, HighD, HighI).
I didn’t try this with the former used version 2.7. So I don’t know, if this bahavior is connected with the software version.

Can someone explain this?
Can someone give me a hint, which parameters can I change to avoid this behavior?
I have attached the complete Log (see at line 12500).

Thanks in advance.

[quote=“diyohl”]Equipment: Standard quadcopter with APM1

Hello,
After the update to v3.1 I found following strange behavior:

I hovered and I gave some impulses to the transmitter roll axis in order to check the PID quiality.
Please see the attached sreenshot from the log (RollIn:red and Roll:green), after some RollIn impulses the copter suddenly leaned strongly to the opposite direction (40°).
[attachment=1]Crazy_Roll_.jpg[/attachment]
This behavior can be reproduced. I tried it with different Rate PID settings (No D, No I, LowP, highP, HighD, HighI).
I didn’t try this with the former used version 2.7. So I don’t know, if this bahavior is connected with the software version.

Can someone explain this?
Can someone give me a hint, which parameters can I change to avoid this behavior?
I have attached the complete Log (see at line 12500).

Thanks in advance.[/quote]

This kind of thing is generally hardware. Look at this graph - it’s always pitching forward and rolling right. Check the front-right motor.

Hello jschall
thank you for the analysis and the hint concerning the hardware. It was a plausible issue, because the ball bearings of some motors have been very very bad.

So,

  • I ordered and mounted new motors (1100kv instead of 950kv).
  • I removed cameras and FPV transmitter.
  • And I tried it again. Hovering at 30 cm height and giving pulses to the transmitter roll axis.

[attachment=2]strange_roll_2nd.jpg[/attachment]

The effect is the same (see at line 6300). I think, this effect appears in both directions and also with pitch pulses.
I attached

  • a new log,
  • a picture with the strong roll reaction of the copter
  • and a analysis roll-rollIn and pitch-pitchIn similar to your analysis. Looks this ‘roll-rollIn and pitch-pitchIn’ graph better than the old one? (I can’t judge it.)

[attachment=1]strange_roll_2nd_diff.jpg[/attachment]

What can I do as next? I have no confidence to fly the copter higher.

  • Something with ESC or ESC calibration?
  • Is the reason possibly the gruond effect in 30cm height?
  • Should I use the APM2.5 board instead of APM1?

The solution is quite simple.
ESC calibration has not been complete.
Apperrently the motor output signal fell below zero baseline of ESC input during attitude control.
Consequence: Delay of power increase and stability problems.
So the advice to check the hardware was correct.
Thanks for your help,
Jens

[quote=“diyohl”]The solution is quite simple.
ESC calibration has not been complete.
Apperrently the motor output signal fell below zero baseline of ESC input during attitude control.
Consequence: Delay of power increase and stability problems.
So the advice to check the hardware was correct.
Thanks for your help,
Jens[/quote]

Ahh, yes that can happen. Will start mentioning it.