Edit : from the posts below I deduce the issue seems specific to my config. Threfore I remove temporarily my request to add a line a caution in the wiki as it might not be a general problem for everyone, until further investigation and identification of the root cause.
This post is a follow up of this one : http://www.diydrones.com/group/arducopterusergroup/forum/topics/str...
I purchased this capacitor to serve as an APM power protection: http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=...
I thought this would protect my APM from very short voltage drops and ultimately avoid erventual APM resets in mid air. WRONG.
I connected this capacitor on the A1 to serve thus as a short power supply on APM.
What happens then when connected ?:
multicopter is on the ground (after landing for example), motors are armed, throttle is at zero. Then if you turn your radio off -> RTL is triggered ! -> motors start full throttle to reach RTL altitude. (of course at the same moment your radio is off so you can only observe a flyaway)
This behaviour was tested and repeated multiple times with same results (on my config).
Another scare we can avoid...very very bad idea this stupid capacitor.
Surely the problem is not the capacitor. The mistake is that you have an armed copter on the ground and you turned your radio off??
The capacitor can only supply power for a few mSec. Has nothing to do with the behaviour.
It's a little like the driver who changed the furry dice on his mirror, and then says 'my car started steering oddly'...
As a comment, why not just turn the radio on?. Make sure the throttle is at a safe level, and switch it on. Within a second, you have control back.
1. rule of R/C and Robotics: Turning your remote control off is the last thing you do, after everything else.
It does seem that at least 2 very important safety precautions were ignored.
A. Always disarm the APM with your transmitter when you have landed and are finished flying.
B. Always disconnect your battery before turning off your radio.
We have all violated one or both of those safety rules many times.
But it also seems to me that if your battery was connected and you didn't disarm, and you had Failsafe set to RTL it would do that whether you had a capacitor hooked up to the APM or not.
There is no way you could do this if power weren't still hooked up, even a large super capacitor wouldn't begin to have enough power to actually get your copter off the ground.
Really hard to see how a simple 3300mfd (filter) capacitor could be responsible for anything.
I have observed this behaviour as well when doing bench testing with capacitors. I believe the issue is this:
1) The 32u2 browns out before the AVR chip. We know this, it was revealed during the investigation of my last Octo crash.
2) When the 32u2 browns out, it stops sending a signal, so the AVR responds with a failsafe event.
3) Normally, this is not a problem, as the AVR browns out a few milliseconds later.
4) Large capacitance on the board will hold up the AVR just long enough that it can throttle up the ESC's before it also shuts down.
However, I don't think it's necessarily a reason to not use a capacitor, as they can serve a helpful purpose. This is mostly a problem when bench testing as you have observed. I almost got bit a few times myself!
If you are operating safely, it should not be a problem.