[fixed Wing] INAV 9.01 crash on first flight (pitch lock?)

[Bingo2023]

Don't know if this helps you guys to tackle down the pitch/servo issue with current INAV version.
Attached is a log + video with 1.5 min flight time, all ended in a crash & plane loss (RIP).

I'm a beginner with INAV, just some experience with ArduPilot some years back.
Also my plane flying experience is some years back, just drones meanwhile. So I can just provide the data.

Any idea why it came down so fast?

1. Auto launch
2. Angle-Mode: I lost sight contact (white plane, white snow, white clouds) → switched to Loiter-Mode [0:58] and grabbed goggles
3. Loiter-Mode: after goggles, I did some stick movements still in Loiter-Mode.
4. Acro-Mode: [1:45] after switching to Acro-Mode it was very twitchy and lost fast attitude. I panicked and tried to pull up and switched to Angle-Mode shortly before the tree.

I see some data glitches 7 sec before crash & also in the beginning.

crash6_LOG00005.TXT
crash6_klein.mp4

Inav Blackbox Explorer: Sync Log with "-29.8" (I just cut&downsized video)

[error414]

What FC do you use?

Look at artificial horizon on you OSD, when you switched to acro (video 1:15), artificial horizon jumped to weird position. Video 1:19, inav message "alt sanity" it means that something is bad in position estimator in inav. It seems like very noisy gyro

BTW: if plane does not fly well in acro mode, never switch to any autonomous mode, it will be always worse :(

[Jetrell]

@Bingo2023 It would appear from your log and DVR. That the issue was caused by flight controllers IMU experiencing excessive vibrations.
Which in turn lead to attitude drift on the pitch and roll. i.e. the flight controller lost track of level flight.
You happened to be in a level stabilized mode, ANGLE, which didn't help in this case. The only way you could have got out of this would have been if you had switched to Manual mode.

The thing that made it noticeable in the DVR, as mentioned by error414, was the severe AHI drift on the OSD.
The log showed the accelerometer vibration was intermittently high at the log break.
Like you mentioned. It is interesting to see how things like the OSD Altitude Sanity check appeared at a break in the log. And at that point acc[Z] momentarily showed 9.19g. Which would certainly trigger a sanity check. But no such movement of the plane equivalent to that occurred in flight.
However the filtered gyro data was very noisy below 70hz, on all axis's.

Was the motor loose, even a bit on its mount. Or the FC board.

[Bingo2023]

FC: MicoAir743v2-AIO-45A

Plane is very little, 320g.
Motor was tight, I used a anti vibrating motor mount made of TPU. Biggest vibrating source were the 3 silicon motor wires.
See below:

My comments:

• Artificial horizon also drift away right after takeoff, seems it is very sensitive to G forces -> but plane flies good
• Loiter is autonomous mode? It flies good in loiter mode (longest duration in this video)

[breadoven]

Is the OSD correctly synced to the video, looks out by a second or so ?

Altitude sanity just means the error between the estimated and GPS altitudes exceeds 20m although it doesn't last long because navEPV increases in these circumstances to the point the estimation resets back to current GPS values.

[Bingo2023]

With this:

Inav Blackbox Explorer: Sync Log with "-29.8" (I just cut&downsized video)

It is in sync (See current).

[breadoven]

With this:

Inav Blackbox Explorer: Sync Log with "-29.8" (I just cut&downsized video)

It is in sync (See current).

Ignore the comment above, I missed it's DJI which I assume doesn't have the OSD/video sync issues you can get with Walksnail if you postprocess the separate OSD and video files recorded by the googles to get the final combined video with OSD overlay.

[Bingo2023]

Can 330Hz for Servos cause Servo-Oscillating in combination with mode change?
I use thin 20cm long Servo cables...

How can I check this in logfile?

[sensei-hacker]

@loghatVector that sounds like an AI-generated comment?
Please don't do that.

[Bingo2023]

#11008 that seems very related.

There are some more the pitch/nose down issues for INAV. Meanwhile I'm afraid to install INAV on my other planes...

[breadoven]

Looking at this again it appears the problem is likely the attitude (AHI) is drifting due to motor resonance upsetting the gyro. If you look at the gyro at around 1700 throttle you get a burst of noise which is then followed by AHI drift (see 30s , 49s and 1:16 flight time). The AHI recovers when the throttle moves away from the motor resonance point and the gyro calms down. It went haywire after Poshold was cancelled at the end (1:16) because the motor throttle increased slightly to the resonance point when the auto throttle switched off causing a load of gyro noise which the AHI didn't like at all.

I found when trying vibration damping motor mounts that sometimes they made things worse and I could imagine resonance would be more likely.

[b14ckyy]

Motor was tight, I used a anti vibrating motor mount made of TPU.

I am pretty sure this is your problem. Never soft mount a motor and especially not with TPU.
Reason:

• TPU is (for small amounts of flexing) not good at dampening and acts like a spring
• depending on TPU thickness, length of the mount and motor mass, you will get a very specific resonance frequency
• With a mount of that length, that resonance can reach a high amplitude as the motor will start shaking like a dogs tail at a frequency far below the filter cutoffs, causing a lot of sensor aliasing and clipping.

Hard mount that motor. The foam body of your plane will still transfer high frequency vibrations but its dampening properties will keep the amplitude low. So high frequency noise with low amplitude can easily be filtered out. The rest could be done by silicon rubber mounting the FC if it is sensitive by itself (optional).

Then the problem should go away.

[sensei-hacker]

Agree with b14ckyy . A 1-2mm washer would be one thing. That's a whole other thing.

Also as he mentioned, to reduce vibration generally you want to avoid bouncy materials. Something like memory foam is good (for a flight controller). If the material bounces, you're increasing the movement because you get the original movement plus the bounce back.

[b14ckyy]

I have done Motor offsets too in the past on tiny 300g builds like my Dart 250g with a 40.000rpm motor setup. But I used 3mm thread nylon standoffs (like the ones for FC stacks) for that. Barely any flex. 3D Printed hard plastic spacers or cone mounts are common but never flexible. Even on quads, soft mounted motors just use rubber pads at most and movement is barely noticeable. Bit this here is screaming for trouble.

[Bingo2023]

@b14ckyy
thank you so much for that information. I believe that could be the root cause.
I want to analyse that more in detail.
What would be a normal range for accVib ?

[b14ckyy]

I never used the accVib value myself and usually focus on the spectograph. Unfortunately in your log you don't have raw accellerometer or raw gyro enabled. The filters are doing a great job to clean out the output smoothly but there is a weird 2 stage shadow in the graph during full throttle flight (and you are full beans most of the time).

The problem is, that if the vibration is so bad that the gyro or accellerometer is actually clipping (hitting its max measurement value on every oscillation cycle), the filters will remove these clipped off values but leave faulty measurements as a result. Since there are permanent unmeasurable spikes, giving out movement or rotational data, that are in reality not there. They go into the filter, are smoothed as valid data and then send to the AHRS.

best thing is: scrap that mount, install a rigid one. Then enable raw gyro and raw acc logging. Disable a few other logs or reduce the log rate to 25% (Or reduce loop rate to 1khz instead of current 2khz).
If you still have issues, that log will help us. If not: issue fixed.