New plane, two crashes in three flights

This is a duplicate post, I’ve recently discovered this ardupilot.com forum and thought the forum title is a more accurate representation of the help I want so I’ve posted here. The original is http://diydrones.com/forum/topics/anatomy-of-a-200km-crash

[youtube][/youtube]

[youtube]http://youtu.be/TkMIakF3jD4[/youtube]

I’ve got a Cyclops EPO that I’m setting up with a view to attempting a 200km flight (lovely plane, I’m sure it’s capable of it), but I’ve crashed it twice out of three outings, and I’ve got no idea what’s going wrong.
The maiden flight in manual mode was a success (it came back in one piece), but the test pilot said it was pretty lively and had trouble trimming it.
The second flight was in manual mode too, it flew a circuit or two before it banked hard right and spiralled into the ground. The speed was about 90 km/hr so it wasn’t a stall. The pilot said it just stopped responding to his input, we switched to Stabilize after he realized he’d lost it, but it made no difference.

[attachment=1]crash1 1024.gif[/attachment]

The telemetry logs show a large right aileron input (green trace, see the white asterisk on the graph, high=right) to level it after a left turn, but it kept rolling right and didn’t respond to the subsequent left input. The trace also shows some big right inputs, I don’t know if that is panicked pilot input or a dodgy radio/connection. You can see the aileron trace go low when I switched to Stabilize, but it didn’t make any difference to the roll (red).

I repaired it and tested everything I could think of and found nothing unusual, so I flew again, this time in FBW A mode. It took off and flew very smoothly, but I didn’t have much authority. Up elevator input barely made a difference, and it only turned very gently when I gave it right aileron and rudder. Then it never really levelled out. The right roll kept increasing (red trace below). I gave it some left aileron (I’m not sure that I gave it full left as indicated on the trace), which didn’t do much, and then it banked right and hit the deck.

[attachment=0]crash 2 1024.gif[/attachment]

In both cases it rolled to the right. In the first case it was like it kept rolling after a large right input, and didn’t react to the left input corrections. In the second case it flew the whole flight with too much right roll and didn’t react to the correction from the APM or from me. It feels to me like an aileron servo problem (failure, loose connection, brownout, jam, insufficient power).
Any insight or expert log analysis would be very much appreciated.

After two weeks of late nights building then repairing and analysing logs, I’m a bit burnt out and it seems hopeless rebuilding it if I haven’t found the problem.

Some of the tests I did:
[ul][li]Removed the power module, replaced JP1 and powered the whole thing from the UBEC. I monitored VCC while I moved and loaded the servos to see if the was any problem with the UBEC voltage/connection. It stayed at 5.0 V +/- 0.2V.[/li]
[li]Moving servos through their full range, and jiggling all the power/servo connections to see if they would fail, loading the wing to simulate the load in flight.[/li][/ul]

Things I haven’t tested:
[ul][li]Anything at high power. Perhaps the ESC or motor caused interference. They both sit about 4 inches above the APM.[/li]
[li]I don’t know if there is any telemetry record of the servo rail voltage if JP1 is removed?[/li][/ul]

Materials and specs:
[ul]Most gear used successfully in my Maja up until now.
[li]DX8 radio, Orange DSM2 receiver with one satellite (range tested to 30m before flight)[/li]
[li]APM 2.5, firmware 2.78, 3DRpower module (JP1 removed)[/li]
[li]Turnigy 3A low noise UBEC powering servo rail[/li]
[li]Turnigy D-Lux 30A ESC with red BEC wire disconnected[/li]
[li]Hacker A30-52 motor[/li]
[li]Vtail – mixing done in APM, RC is setup as T tail (according to instructions here)[/li]
[li]Hobbyking telemetry module[/li]
[li]Hitec HS-65HB servos x 6 (ailerons, flaps, v-tail)[/li][/ul]

Considerations/Complications
[ul][li]I have combined the power supply wiring for the flaps and aileron servos in each wing, and in the v-tail. Both get their power supply from the same red and one black wire, each have their own signal wire. Is one 26g wire enough to supply two HS-65HB 11g servos?[/li]
[li]When I first connected everything I plugged the UBEC the wrong way around on the servo rail for 30 seconds or so. Could that have fried something? It “appeared” to work fine after I corrected that. Is there a way to check?[/li]
[li]There is a large amplitude v-tail servo twitch that I get when I have the flaps switch in position 2 (50% flaps, no Elevator) if I have the elevator trim switch at 2 or 3 clicks of up elevator. Bizzare, it doesn’t happen at any other time. I still have normal manual control via the sticks; it goes away as soon as I move them more than a few mm.[/li][/ul]

Three log files attached from the three flights.

This really has to be something mechanical/electrical. As you noted, the SERVO_OUTPUT_RAW.servo1_raw was tracking the RC_CHANNELS_RAW.chan1_raw perfectly, and then suddenly roll became totally uncorrelated with it. Twice.