in the last few test flights I noticed strange quadcopter behavior in gps-dependant modes (loiter, position hold, rtl). after brief investigation the problem was found - turns out gps output became very noisy compared to the previously logged flights. However I still cant figure out why it happened at the first place?
The only changes made to the quad was frame structure upgrade, but none whatsoever modifications in electronics or even relative components position. Also upped arducopter to 2.7.1 version.
Here is the graphs from log browser (GPS lat channel)
#1 is what its used to be
#2 what i have now
PS. the jitter is in range +/- 5-8 meters. Lock is good: 7-9 sats.
PPS. the second log with a same problem shows even more percular features
Update: 2.6 Arducopter under same conditions has no problem whatsoever. It looks to me like some sort of least significant bits of the gps coords are xored / inverted OR floating point precision loss.
Update 2: ArduCopter both 2.7 - 2.71 exhibit the same erratic coordinates behavior. Having a first peek in to the whole codebase and a numerous changes since 2.6 my first suspect is (and I could be very wrong though) AP_LeadFilter.
Update 3: Ok, turns out to be exactly that 'filter' thingy. Commenting out get_position to return a position solved a case.
Jason Short, your turn)
Thanks in advance.