PX4 v1.14: High Vibrations and Accelerometer Clipping on Pixhawk 4 Mini and Matek H743 Slim

Flight Testing & Log Analysis v1.14
1

I’m using two different quadcopter setups: one based on the Pixhawk 4 Mini and the other on the Matek H743 Slim. Both are running PX4 v1.14, and with each I’m experiencing accelerometer clipping.

For the Pixhawk setup, I’ve already tested various PX4 firmware versions. With identical PID tuning and hardware, I saw significantly worse vibration signatures in hover (see Same hardware, same MC parameters, but significantly different vibrations in hover - Where does this come from? - #4 by knolle) in comparison to v1.10 which was perfectly fine. This demonstrates that neither the PID tuning nor the hardware is the root cause of the clipping—instead, the PX4 software itself has a significant impact on vibration levels.

A similar thread discusses high vibrations when switching from ArduPilot to PX4, though in that case the hardware appeared to be the cause (see High Vibration and Accelerometer Issues - #6 by dronesalot). Comparable issues have also been reported on Cube Orange flight controllers under PX4 (see High vibration in Cube Orange+ - #110 by sibujacob). Interestingly, the Cube Orange does not exhibit these vibration problems under ArduCopter (see Accel 1 Clipping with PX4, no problems with Arducopter - #4 by mwshafer).

On the Matek setup I tested both PX4 v1.14 and ArduPilot without any hardware changes. I can confirm that vibration levels are vastly lower under ArduPilot; PX4 still suffers from clipping. Thus, the hardware cannot be the issue here either, as ArduPilot flies it without problem.

In summary:

  1. I have ruled out hardware as the cause of the excessive vibrations and clipping.
  2. I have ruled out frame-specific effects.
  3. I have ruled out the choice of flight-controller hardware.
  4. I have ruled out PID tuning as a factor.

The only common denominator is that PX4 v1.14 exhibits significantly worse vibration behavior than other firmware (versions) - those other versions allow safe flight, whereas v1.14 does not. What changes were introduced in PX4 v1.14 that degrade vibration performance compared to other firmware releases? How can I mitigate or work around this issue?

2

v1.14 should not degrade the vibrations as such but it will show them because I believe they are filtered less by default, and so more visible.

If you share a logfile (via logs.px4.io) we can have a closer look, otherwise it’s hard to say.

3

I was not expecting v1.14 to degrade the vibrations, but I was expecting it to not increase them. As demonstrated in the comparison available at Same hardware, same MC parameters, but significantly different vibrations in hover - Where does this come from?, the vibrations are indeed different across PX4 firmware versions. I am still trying to understand what specific changes between v1.14 and earlier versions might explain the altered flight behavior, as adapting the notch filters did not solve the problem. The differences in vibration characteristics are even more pronounced when now comparing v1.14 with ArduPilot on the same hardware.
Here is the log file for the PX4 flight: https://logs.px4.io/plot_app?log=15bb12fb-5e52-4182-aed9-750d2fee19ef

4

It doesn’t “increase” them, it just makes them visible, and potentially requires some filter adjustment: MC Filter Tuning & Control Latency | PX4 Guide (main)