[attachment=0]Erroneous response.gif[/attachment]This behavior started in March on an APM 2.5 while still trying to tune it, all of a sudden it threw itself forward into the wall, propellers first. After getting new propellers and patient testing, changing the PIDs to values that have worked in the past, I now have a log that shows that negative elevator input results in negative Desired Pitch. However, positive elevator input also results in NEGATIVE desired pitch, of an exaggerated magnitude and with the wrong sign!
This has me at a loss, because I can’t tune if it is behaving erratically. What could cause this?
Update: Since a motor got bad I replaced them all. The acceleration is now about ±1 m/s2 on X, Y, Z. Made no difference.
Logs show that when not giving any stick input, the APM requests both positive and negative pitch commands. But, when using the elevator stick (which is reversed in the radio), stick forward gets translated just like it should, and with the correct magnitude. But stick towards you gets translated into a magnified stick forward command. Very magnified! Like in the power of two. As in an algorithm that calculates the absolute number by first taking the square and then the root, but forgot to take the root and also forgot to fix the sign.
This reminded me that the problem started happening at about the time when I upgraded from copter version 2.9.1 to version 3.2.1.
Does it ring a bell, anyone? Maybe I should just downgrade again.
After reloading the firmware (the latest, 3.2.1) the issue is gone. Obviously a bug caused by a flashing error.