I was testing the Auto Commands for ArduCopter 2.9.1 using APM 2.5 with MTK GPS and something strange happened.
I had been doing successful auto flights (takeoff, waypoint, RTL) for about a week and I was starting to change some parameters for RTL. I changed two parameters and then did an auto mission but instead of landing the quad dove toward the ground at around 4 m (during the RTL landing phase). I switched back to my default parameter, but the same thing happened. I also noticed that it twitches in auto flight but not in stabilize flights. I tried to fix the twitch by changing the PPM encoder (as it is listed in the wiki) and this stopped the motor spin ups (which were happening while it was flying successfully) but the twitch remains.
So, I thought that RTL was the problem and I started using waypoint with land instead of RTL. I also reflashed the firmware on it to 2.9.1b. I got one good flight out of it. When I went to fly it a second time the quad took off, flew to the first waypoint (this is into the wind) fine and was on its way to the second waypoint (wind at its back) when it dove out of the sky. Both times it looked like a motor stopped spinning.
I have attached the .tlog for the last flight (I can find the others if you want to see them). I noticed that there is a throttle spike (went to 100) right before the quad started to lose altitude.
I thought that this might be a hardware problem, but shouldn't the motor cut out in stabilize flights as well?
I'm going to look at all of its solder joints and all the hardware connections on Monday. Any ideas for places to look for faults (in the hardware or the log)?
How are your esc's/motors wired? That sound like a classic symptom of loose bullet connectors - they lacken off after a while...
which param shows you the throttle spike?
I have DJI 920 Kv motors connected to the DJI ESCs. Been flying (and crashing) the same setup for a while and the bullet connectors always seemed to have a really tight connection.
How would I test the motor/ESC connection without flying? Do I just see if the connectors wiggle while on the bench? I was planning on running the motors (without props) and wiggling the connection to see if I can get it to turn off.
Under the VFR Hud there is a box labeled throttle. That is one that is showing me the spike.
As Dave_C wrote, it was a loose ESC connection. One of the female bullet connectors on the ESC wasn't connected to the ESC and was being held on by the shrink wrap.
Replaced the ESC and just had a bunch of scuessful Auto flights.
Bullet connector caused crashes have been a continuously recurring problem on this site.
I've got two FlameWheels with DJI OPTO ESC's and have never had a problem.
The bullet connectors fit straight into the sockets built in to the ESCs and seem to be a tight fit.
And I do use a little contact grease, but I still don't trust them in the long run.
When you are flying a quad, even a momentary disconnect can reset the ESC and you are in for a painful consequence.
I think the common practice of putting heat shrink over them only makes the problem harder to see and does nothing to solve it.
In the end, the only solution is really soldered connections or some other and better kind of high current connector.
Too bad, other than inescapable bullet connectors, the DJI Opto ESCs seem to be a really good and reliable ESC.
Our problem wasn't from a bad connection between bullet connectors but from the bullet socket on the ESC board breaking its solder pad. When we took the heat shrink (that DJI puts on) off the ESC this socket just fell off the board.
Otherwise, the bullet connections appeared to be solid.