High Vibration and Accelerometer Issues

[Qejyu]

Hello,

I am experiencing some issues with accelerometers and vibration. I am using a Pixhawk6C on a quad. When I was using ArduPilot, the flight performance was excellent. However, after switching to PX4, I encountered significant flight vibrations.

Here are the logs and vibration graphs:

https://logs.px4.io/plot_app?log=076fbf64-c1e0-46db-a746-ee0ef385ab01

I identified that the issue was resonance caused by the damping pad by examining the logs. So, I removed the pad. The flight performance improved, but the vibration levels in the flight logs remained bad. Additionally, when I attempted to fly using an onboard computer later, the vibration levels were incredibly high, leading to a crash. I also occasionally experienced clips during flight testing.

Here are the vibration graphs and logs for RC-controlled flights:
https://logs.px4.io/plot_app?log=dfcf7072-5a96-4c13-9fc2-f4fd22efb1ff

And here are the ones for onboard computer flights:
https://logs.px4.io/plot_app?log=4afed8e7-f766-48c9-afe4-5cf8266ea246

I later tried adjusting IMU_DGYRO_CUTOFF to 15 and IMU_GYRO_CUTOFF to 30, which significantly improved the flight conditions, but the vibration levels are still high. I am unsure how to further improve the situation.
https://logs.px4.io/plot_app?log=307c3123-c907-49cd-93b9-9a23d0565223

The drone’s structure is very simple, consisting only of motors, ESCs, a power distribution board, GPS, an RC receiver, a laser rangefinder, and the Pixhawk6C, all tightly secured. The Pixhawk6C is mounted on the frame using foam padding.

I believe it’s a hardware structure issue, but I am not sure how to address it. Any advice would be greatly appreciated.

Thanks.

[mwshafer]

I’ve reported similar issues using a Cube Orange+ with no resolution yet. Have you done a test flight with high rate logging turn on? That’s usually a good plave to start. That other thread is here:
https://discuss.px4.io/t/high-vibration-in-cube-orange/32194
I’ve seen this issue on multiple frames and actually switched over to Arducopter forb a while which didn’t have these issue. I’ve recently been looking at trying to use to notch filters parameters in PX4 to try to filter out what I think might be my prop pass frequency. If you turn on high rate logging you’ll be able to see peaks in the FFT plots for the angular rates and actuators that will give a start at least.

[Anil_m]

Try changing the foam pads with new ones!
Are you getting similar vibrations without props?
And unbalanced props could make up such noise as well!
Check motors base fixtures, possibly dismantle and re-tight motors to frame …possibly entire frame…
Some debugging insights might help!!

[Qejyu]

Thank you, I’ve resolved the issue by removing the internal vibration-damping foam from the Pixhawk6C and using external damping pads.
I believe the issue stemmed from three factors:

1. There were minor vibrations in the frame, which didn’t affect flight but were hard to eliminate.
2. The internal damping in the Pixhawk6C wasn’t effective enough to filter out these vibrations, and may have even amplified them, pushing the accelerometer beyond its limits.
3. There might be a problem with the PX4’s EKF2 filter, which wrongly estimated the resonance between the damping pads and internal damping, reflecting it in the flight behavior.
   At first, I tried addressing the issue structurally, such as reinforcing the frame and changing the propellers, but it didn’t work. So I removed the Pixhawk’s internal vibration damping and added external damping pads. It’s working well now, with minimal vibrations, and the accelerometer is within acceptable limits. However, I believe the final solution lies in resolving issues within the PX4 filter itself. I don’t know much about filters. perhaps when I have more time, I can study and contribute to improving this.
   Maybe these insights could also apply to the Cube Orange+, and I hope you find this helpful.

[mwshafer]

Thanks for the update. I too have seen better performance with Arducopter. Arducopter has a low pass filteron the accelerometer (INS_ACCEL_FILTER) which might explain the better performance I’ve see when using that firmware. I can’t seem to find anything in the PX4 documentation that suggests any filtering is done on the incoming acceleration data. Let me know if you know of anything different.

[dronesalot]

I just found this topic. I’m running the same frame and an Orange+.

The supplied anti-vibration pads

Removed with only thin silicone vibration bumpers.

I don’t yet have enough experience with PX4, but I believe there is a resonance being set up with the motors trying to compensate. You can see in the second plot that it disappeared for a split second, and the drone itself even sounded significantly better.

This may be a characteristic of the frame coupled with insufficient filtering/damping. I will see if I can find some damping silicone plugs to stick i in the motor arms and see if it has any impact.