Crash with large quad in Autotune. Lots of testing

  • Large (32" prop) quadcopter
  • 12S LiIon battery, lots of capacity
  • AC 3.5.5 on a Cube 2.1
  • Dual Here
  • Dual Vcc supplies
  • Dual Hall effect current sensors
  • Independent V supply to optoisolators on ESC (logged as Vservo in tests, wasnt logged during the autotune)
  • Filters tweaked to reduce vibe noise and cube is mounted on an isolated tray
  • RPM1 is monitoring an internal cooling fan
  • Nothing to unusual with the setup otherwise

Aircraft was tested on the ground for 2+ hours before maiden flight - Power electronics loaded/heatsoaked/etc. Temperatures of critical items are closely monitored and logs reviewed. No indications of impending failure were noticed.

Initial few flights went without much drama, PID tuning was significantly out, however totally flyable.

Roughly 5 minutes into the autotune, while being repositioned (ie it wasnt tuning at the time), the aircraft wobbled in pitch and then dropped.

  • No indications of any structural failure
  • No indications of any power electronics failure (other than the drop)
  • Damaged aircraft was powered back up on bench supply, no abnormal current draw, all esc/motors functioned normally

We have been over the logs with a fine tooth comb, and cannot see an issue, so suspected some sort of desync condition with the ESC.

We have replicated as much as feasibly possible, the condition/setting/loads that were on the aircraft with it strapped to the ground, and have completed some extreme testing of the powertrain, trying to cause a desync with large/instant throttle requests at 400/490Hz, without “success” (see logs attached - we did this 3 times without a hint of desync or other failure). First ground test is on 3.6/NutX with RC passthrough on Ch1-4 and SERVO_SPEED set to 490Hz. 2nd test is on 3.5.5 with Ch1 hooked to 4 motors, arm check/crash detect disabled and in Acro. Both tests have a mix on radio providing harsh motor PWM changes.

There are a number of other, obscure, things it could be, but -

Would appreciate others looking over the logs (particularly those with indepth knowledge of how Autotune updates params on the fly?) to give us some ideas on the next steps in troubleshooting. Thanks!

1.BIN - Autotune log with crash
2.BIN - Power train test 3.6/NutX
3.BIN - Power train test 3.5.5

Bummer Check the GPS connector for a loose connection some have been loose mine dropped off this will also disarm the craft " i think" in flight. I also recommend not using the small carrier board for reasons above but use the Spekreworks custom carrier board it can handle 50 v. and has a separate usb port to communicate. For this size of a craft I think manually tuning PID would be best. I fly it land the craft make the adjustment and fly. As a side note I been testing 3.6-rc Nuttx and like it. My test with chibiOS rc-1 on the cube shows slow feedback on the hud and unstable flight so i would avoid for now.

I loaded the crash bin file and see the your vibrations are high in z "red "over the 30% my homemade hex is under 10%. for an example with camera.

Also with Auto analysis sates this error possibly the craft was moved around during arming.

Test: IMU Mismatch = WARN - Check vibration or accelerometer calibration. (Mismatch: 1.05, WARN: 0.75, FAIL: 1.50)