Thanks for pointing me. Anyway - my problem seems to be a bit different - the desired altitude is going down, but the input throttle is kept in the mid deadband in PosHold -> so the altitude should not change.
And the calculated climb rate seems to by on positive max. value in the same time (CRt parameter of CTUN).
You’re right. It looks that it isn’t related to the issue that I have pointed. It seems that the arducopter decides to not raise the motor throttle to keep the altitude despite the fact that DCRt wants that.
I’ll parse your parameters to see if I found something that can cause this behaviour.
I’ve noticed that too - but if you take a look at vibrations in IMU logs part - they are quite low (less than ±1 ~ ±0.1G for X a Y) - not sure what could be the reason.
There is something that causes, that EKF is not working for me as it should I think, but maybe I’m wrong.
Thanks for your time and effort.
Yeah, the imu graph looks nice. I think someone more used to efk could explain why the vibe levels are so high in contrast with imu. I don’t have any clue about that , but I trust this could be the cause.
Well, I’m using small kyosho gel pads under PH. I find them work well from my previous apm heli project. Propellers are already balanced, so the next thing I can try do is to use some damping under motor mounts.
I’ve took a look at IMU graphs once more and I’ve found big difference between IMU and IMU2 z vibration values - see attachment.
I’ve also noticed, that Randy did some changes in Copter-3.3.2-rc in z-axis filtering - so there are some tries to do yet.
I think Tabascoz has correctly identified the issues and in particular I think it’s caused by high vibration. The way to measure it is mentioned here but I’ve also attached some graphs. copter.ardupilot.com/wiki/common … vibration/
The levels are regularly climbing to the 40m/s/s range for many seconds (i.e. not short-term spikes) and we can see the conradiction that appears in the altitude vs climb rate. I.e. the climb rate is positive but he altitude is falling. This is the classic sign of high vibration levels messing up the EKF’s altitude estimate.