High vibration in Cube Orange+

We have discussed this in today’s maintainers call, and few propositions were raised:

  1. Perhaps the IMU that isn’t vibration isolated (out of 3) was being used as primary IMU for EKF2. Could you all check on that?

  2. Without access to the two logs with exact same vehicle configuration, but different PX4 version (v1.12 and v1.13), it would be hard to conclude that it’s a software problem (vibration metric alone isn’t definite clue, we would need to analyze actual IMU data)

@sibujacob could we have access to the two logs from v1.12 and v1.13 you had above?

Separate from that, Cube orange plus target was not available for v1.12, so we were wondering how that target binary was built for the test!

1 Like

@junwoo0914 regarding point 1) that makes sense for thw raw accel values but why would the vibe metrics look so high for the other 2 IMUs in the logs?

Have any of the developers flown with an Orange?

1 Like

@ junwoo0914…

thank you for the feedback…mine is a Cube Orange since beginning… Px4v1.12.3 log review was placed earlier based on another flight demo back in Jan …
If you require a flight demo using Px4v1.12.3, i will downgrade the px4v1.13.3 currently installed but i am not sure if it can install anymore in the first place …will try to give it a shot today …

I used plot juggler instead to get the device ids of the imu by doing the following action
import data from two PX4 log files and checking the imu_device id graph. i hope this is correct. It looks the ids matched for both PX4 v1.12.3 and PX4 v1.13.3

Also attached a PX4 v1.12.3 log file taken back in Jan / Feb 2023 at the same spot
https://review.px4.io/plot_app?log=51df1330-7291-4b93-ada8-25f956907dcb

Hope this is acceptable …

1 Like

I myself haven’t recently but I work with several clients that fly Cube Orange and/or Orange+.
I’m planning on test flying it again, I just need to mount it in my test quad again.

I had a look at the IMU sensors used in Pixhawk 6C, 6X, Cube Orange, and Cube Orange+. From a brief analysis it looks like the sensor drivers are mostly configured to:

  • high FIFO rate
  • max range (2000 or 4000 deg/s, and 16 G or 30 G, depending on sensor)
  • no sensor-internal filtering

Here is a bit more data: IMU sensor comparison Pixhawk 6C, 6X and Cube Orange and Orange+ - Google Sheets

So given that I don’t expect a huge difference in logged vibrations between these.

Just for comparison, this is what the accel log looks like for Pixhawk 6C (with high rate and sensor comparison logging enabled). That’s mounted with the included 1mm double sided tape. It looks noisy but I believe that’s because we’re looking at unfiltered raw data. It’s flying smooth for me.


One more note: make sure the Cube is not loose (wiggling) in the base board. Sometimes the screws might need some tightening.

2 Likes

Do you see those peaks in the angular velocity FFT?

Those vibrations are probably generated by your drone and amplified by the rate control loop. You should:

  1. reduce the gyro filter cutoff frequency (to ~30Hz)
  2. reduce the d-gyro filter cutoff frequency (to ~15-20Hz)
  3. reduce the roll and pitch rate gains

The vibration profile changed drastically between those 2 flights. It could come from how you attached the battery.

1 Like

Hi…thank you for your feedback. In fact the only difference between those flights were that initially I carried the drone in my hand to the launching pad while it was switched ON . The second improved graph is while i placed the drone on the launching pad after charging and then switched ON the power !! I am also surprised by this findings but otherwise the batteries are in the same place locked and tight and did not remove for charging purposes…

Edited - Also should I decrease the ‘K’ gain of pitch and Roll ?

Edit 2 - After changing the gyro cut off Hz , i get this message
Accel 1 message clipping…not safe to fly

I just did a sanity check test flight with OrangePlus on a X500v2 airframe and it flew beautifully with default tuning, that’s using the release/1.14 branch.

Again, the plots look noisy because it’s pre-filtering but there is no clipping and attitude as well as position tracking was spot on.

2 Likes

How can we download the v1.14 firmware. Will it show in the QGC firmware menu ?

Hi All,

After downgrading my firmware to PX4 v1.12.3 , I am having a huge relief …
a) The drone works well even with chopped off propellers which actually happened on
PX4 v1.13.3 .
b) The best think i like is the noise is very very less with the same chopped propellers. On PX4 v1.13.3, the noise from the props were very high. ( Strangely i see less noisy prop from the vibration graph of v1.13.3 )

c) It was extremely easy to handle the drone in all modes especially position mode. The take off and landing was so smooth that i didn’t touch the RC controls for the entire flight
( although i did some jerky shakes to test the integral response )

The same quadcopter config with PX4 v1.13.3 had me do fine tuning for hours with severe consequences or very less improvement and also where the drone tumbled over at one time. All the components remained the same for the demo test .

Therefore I have to conclude that I am prohibiting myself from flying the Quadcopter drone with v1.13.3 unless I see the same performance as with Px4 v1.12.3 . ( Awaiting for the v1.14.x )

i completely validate @ [mwshafer] point of view

log review of Quadcopter PX4 v1.12.3
https://review.px4.io/plot_app?log=42eda46d-73df-45cb-8d3e-e6003c9b6cf7
https://review.px4.io/plot_app?log=19ed77d6-5bc1-4295-951e-18c104ac32a8

To be confirmed, but I think we had more filtering enabled on the IMU itself in 1.12. We removed it to be able to control the whole digital filtering settings using IMU_GYRO_CUTOFF. This would explain why the raw data looks less noisy (it’s already filtered). However, setting the gyro cutoff correctly should give you the same behavior.

yes

This can’t be the only thing. Some screw might be loose.

Having clipping is usually a sign that something is resonating or rattling on the drone.
However, did you see any difference while flying with those settings?

2 Likes

@bresch
thank you for your feedback . my comments as follows
Having clipping is resonating …on the drone. However, did you see any difference while flying with those settings?… "

**The drone doesn’t allow me to takeoff with accel clipping message . I do have a Herelink CUBE ORANGE QGC setup. " All the CBRK parameters have been disabled for bypassing safety .

This can’t be the only thing. Some screw might be loose.I can assure you sir that the drone was tight enough for the entire flight setup. Also i did shake rigorously and tumble upside down manually with my hand to ensure nothin is loose !!

Currently i am on v1.12.3 i am really not in a confidence mode to go with v1.13.3 unless the Dev team assures me that they will help all the way . I also did not report here that fire broke out on one of the quad motors :grinning: after fine tuning while it was in the air …( i am not joking !! ) That was with a new temporary motor that i used for flying awaiting OEM . It was working fine for 2 days @ 5 to 10 minutes running but on third day during flight, the incident happened. Can a 360KV motor( temp) along with 3 nos x 380KV motor create such a rush of current on the quad from the esc. I have replaced the motor with the 380KV and using the same escs. These are premium range Sunnysky escs and motors. ( my 500 flights with DJI M2PRO and Parrot Anafi Thermal never saw such an incident :grinning: )

Today we did some testing as there were similar issues with accels clipping which occured after updating from v1.10.1 to v1.13 and did not occur anymore after downgrading back to v1.10.1.
For this, we used the same UAS (Holybro QAV250) with the setup being unchanged except the PX4 firmware. We tested with the firmware versions v1.10.1, v1.12.3, v1.13.3 and v1.14.0-beta2. Apart from observable changes in flight behavior and handling, differences between the firmware versions in terms of vibrations shift can be seen in the acceleration power spectral density plots provided below. Also, the power consumption changes noticeably (which is probably due to the changed flight behavior).

v1.10.1

v1.12.3

v1.13.3

v1.14.0-beta2

1 Like

Right, you cannot arm if you have that message, but if you have it before arming, it means that the autopilot was hit by something.
Unfortunately I don’t have a cube orange to reproduce the issue.

@knolle Is that also with a cube orange+ ?

@bresch No, it’s a pixhawk 4 mini - sorry for not specifying.

Ok, please open a new topic for that, this one is for Cube Orange+ only. And also share logs if possible.

@ Bresch

When is the v1.14.x official release . I thought better i wait for that release so that i could also start working with ROS2

Done: High vibration in Pixhawk 4 Mini after updating firmware to v1.13

1 Like

I can’t give you an exact date for the official release (should be in the next weeks) but you can already test the release candidates (there will only be some minor changes between those and the official release) to see if it works better.

1 Like