Irregular attitude instability leads to rollovers

[Gabriel]

Hello,

We have had a few flight tests where we will be able to takeoff and move the quad in Stabilized or Position modes and then when giving a Roll or Pitch input it almost seems like one side, or one motor, loses all power and the quad rolls over. I am including a log and video of a representative example of this behavior.

When analyzing the log I see:

1. At 1:21.4 log time there is a manual roll command of 0.6
2. The roll angle setpoint peaks at 20 deg @ 121.485 and starts decreasing @ 1:21.515
3. The roll angular rate setpoint peaks at 106 deg / s @ 1:21.455 and starts decreasing @ 1:21.515
4. The estimated roll goes from ~2.5 degrees @ 1:21.4, crosses the roll angle setpoint of 11.6 degrees @ 1:21.544, hits 90 degrees @ 1:21.954 and then the rollover takes over
5. The estimated roll rate setpoint goes from -6 deg / s @ 1:21.4 to 116 deg / s @ 1:21:528, dips down to 104 deg / s @ 1:21.583, and then shoots off again
6. At 1:21:4 motors 1 & 4 drop power and motors 2 & 3 rise in power, which makes sense for a roll to the right. Motor 1 seems to rebound to a much higher output than Motor 4, and after Motor 4’s initial rebound it drops power to almost 0 at 1:21.585 before rising somewhat
7. Vibration metrics seem fine, battery voltage / current seem consistent, there were no other warnings displayed by the flight controller other than “Attitude failure (roll)” @ 01:22

Flight Review Log:

Flight Review

Flight Video (crash around 0:51):

Hardware Setup:

• Flight Controller: ARK FPV Flight Controller
• ESC: ARK 4in1 ESC
  • Bidirectional Dshot 300 enabled
• Motors: BrotherHobby SE V4 950KV
  • All four spin on Arm Command
• Battery: 6s LiPo
• Props: HQ 10x5x3
• Frame: GEPRC Mark 4 10”
• GPS / External Mag: ARK SAM GPS mini (u-blox SAM-M10Q)

Tune:

• Roll Rate (K - 1 , P - 0.14, D - 0.002)
• Pitch Rate (K - 1, P - 0.14, D - 0.002)

Questions:

• Is this just a bad tune?
• Is it just an ESC / Motor failure?
• Is the GPS spoofing indicator something to be worried about?

We have examined the ESC connections and solder joints, and everything appears to be fine. Have not noticed any issues with the motors outside of the rollovers. Would really appreciate any help, thank you.

[Gabriel]

My coworker figured out the issue. We had to adjust the Timing Advance in the AM32 firmware for our ESCs. This was not something that could be done in PX4, but required flashing Ardupilot onto the board so we could passthrough into the ESC. Everything is flying fine now. Something not mentioned in the original post that led to this was that the motors would sometimes “buzz” at certain RPMs.