Quad sudden descend (DF log attached). Help needed to analyze

Quad descended still with high battery level
For 2 days(27 & 28 Jan) in a row, my quad S500 running Pixhawk and Copter 4.1.2 descended during loitering. The last one (28 Jan) results in quad overturn with no damage sustained. This is after changing the 4 motors from original Topseed S2312 920Kv (9 x 4.5) to Emax MT2216 810Kv (10 x 4.5). Attached are the DF logs for each of the 2 flights and the param.

FYI , the MOT_THST_HOVER on Jan 27 was 0.2532225 (was 0.3283042 before the flight).
and the MOT_THST_HOVER on Jan 28 was 0.2296536. thru auto learning.

I am aware that the actual roll and pitch were far off from the desired roll and pitch from the DF logs. This is most probably due to my temporary use of very soft and flexible props(10x 4.5). Assume this may not be the cause of sudden descend.

I am also aware that the motor of higher pwm is above the motor of lowest pwm by over 200-250 pwm. But I checked the CG before every flight and found it is proper. Refer to 2) below.

Can anyone who can explain the cause of the descend?

Questions on LogAnalyzer

  1. Attached DF log clearly logs Vcc. However, LogAnalyzer Test shows Unknown Vcc – no CURR log data. The LOG_BITMASK is 983039 checks all the available param. However, the drop down list does not include, as in previous MP version, Board voltage which I think logs Vcc.
  2. Previous MP version tests Motor Balance by showing average PWM of all 4 motors, but the attached LogAnalyzer shows UNKNOWN.
  3. The line PM=FAIL in LogAnalyzer says 7 slow loop lines found. max 7.70% on line 136574. Can anyone explain ?
  4. Test for GPS=UNKNOWN. But the DF logs clearly logs No. of satellites and HDOP.
  5. Test for Event/Failsafe = FAIL - ERRs found : CRASH GPS_GLITCH. Again DF log deos not reveal any GPS glitch.

Why does LogAnalyzers stop showing the test results of most param ?

https://drive.google.com/file/d/12E4TuMBdbPXBVFL6K7-c6mfLvZan3e_7/view?usp=sharing
https://drive.google.com/file/d/19PWGAcRYg9czGjz4h87fry6-WAXIXpE0/view?usp=sharing
https://drive.google.com/file/d/1h6vR-w2DzqBBkEJQYB-ROjUTnPOhA9uI/view?usp=sharing
https://drive.google.com/file/d/1rfA26moP29IybCIAzjJ6B3UjdirzG2T-/view?usp=sharing

The spread in motor PWM is probably because you have a twist in the arms or the motors aren’t aligned. There is a yaw imbalance since the directional pairs 1&2 (CCW) are spinning faster than 3&4 (CW). It’s not C of G.

As for the descent, there’s a problem with motor 2. Either the motor or ESC are failing, the prop came loose, or there’s a wire issue. Either way the PWM signal went to max and the others tried to balance it. I’m leaning to something mechanical, like the prop came loose or the arm moved or twisted. Only because the quad didn’t flip right over or drop right away.

The log analyzer in mission planner is out of date. Don’t use it. It doesn’t accurately reflect the recent firmware.

Thank you for yr analysis.

The photo of the quad is attached. MT2216 810kv were replaced few days ago and motor mounting screws were checked twice for tightening after mounting.

After reading your reply, I tested the frame by hand twisting each of the 4 arms and found no apparent twisting. The special angle from which I took the photo of the quad shows the 2 arms far away from the camera were hidden completely behind the 2 arms nearer to the camera, meaning the 4 arms are aligned without twisting.

Motors are also found rock solid. All props are checked after crash and found no loosening. Also checked for cold solder joint on ESC wires to PDB for Motor 2 whose PWM went to max, but found nothing unusual. The only thing is, I did not remove the power (red) and ground (black) wires from 3 of the BEC cables. Proper way is to have only 1 of the 4 cables connected with all 3 wires.

I still need to configure out why motor 2 went to max PWM by swapping it with other motor/ESC.

By the way, what plotter or log viewer did you use ? Thanks

you seam to confuse twisted arms with bent arms.

Please check again.

2 Likes

Thank you for your reply.

My understanding of bent arm, in engineering mechanics, is the formation of a curve, or deformation of a slender element due to application of an external force at right angle to its longitudinal axis. Visual inspection does not reveal any bent arm. All the bolts securing the 4 arms to the upper and lower frame plates are secured with no loosening.

For twisting of an object, it results in torsion from a torque applied along the longitudinal axis of the element. That is, the longitudinal axis remains unchanged after torsion. In my quad, there is no known external torque that causes torsion of any of the 4 arms.

Please clarify further if necessary. Thank you.

The torsion or twisting force is the motor torque in this case, and those frame kits are famous for it. Unless you got a good one from HobbyKing with the CF rod through the arms.
I’m not sure how good the genuine Holybro kit is, maybe they are good.

MavExplorer, part of MavProxy.
https://ardupilot.org/mavproxy/docs/analysis/mavexp.html