Hover learn value reset 3.4rc5

After few test on a small f450 on 3.4 rc5 where everything worked flawless decided to try it on a bigger quad. Main problem is that i am unable to hover. Copter shoots in the sky and very difficult to land. Last attempt ended up with broken legs …
Hover learn is set to 2, calculated thrust hover is 0.5238545
Based on ecalc this setup should hover at 35% throttle

Before another attempt with 0.5238545 value that in alt hold will climb at mid stick i would like to reset that value, but where ?
What will be a good starting point ? can hover be eyeballed based on ecalc approximation ?

Here are few logs with the current setup


From what I was told or heard is that the devs recommend to not use hover learn until they can address some issues with it. Is that correct? For that reason since I have rc5 on mine I haven’t touched it.

That is bad news …
i have made another flight only in stabilize maintaining to quad in 1m alt difference. From my calculation throttle hover must be 0.19 instead of calculated 0.4867024.
Also huge difference between desalt and baralt

Attached the log just in case someone is interested in comparing data


I had set throtle hover at 0.2 and made a stabilized flight of about 2min. Throtle hover was calculated and set at 0.216…
Althold improved a lot. Desalt and baralt are well bounded.

Hover learn has worked well for me also. I was flying a F450 frame quad with 3 different batteries (4Ah, 5Ah & 6Ah) and each time the throttle hover parameter was quickly re-set and all was well.


Ok, so I think it’s working ok for you now?
I don’t use ECalc much but I’m not surprised that the hover throttle and the ecalc value don’t match. The ECalc value would likely be expressed as a percentage of the pwm output range while the thrust-hover value is before the esc->thrust calculation is done.

The AC3.4-rc5’s hover learning feature should be working ok. There’s still one small problem that will be fixed in -rc6 (which will go out this week) but it’s quite a minor issue related to using a filtered vs unfiltered value for the learning.

wiki page re thrust scaling: http://ardupilot.org/copter/docs/motor-thrust-scaling.html

Thanks again for your testing and reporting back, it really helps us get the release solid.

1 Like

from my understanding the calculated value is set only if it is bigger than previous. If the copter gets lighter learned value will not get lower.