Copter 4.0.3 crashes during pitch autone

For testing and learning I use a F450 type X-style quad with a Pixracer R14 (Chinese clone) FC which has been running well for a couple of years using PX4 and QGroundcontrol.

I wanted to test Arducopter and Missionplanner so recently I made the change and ended up in a learning curve where the main problems were some sudden crashes for which I sometimes could not pinpoint the reason.

After recalibrations, buying new batteries and exchange of some components everything worked. All modes like AltHold, PosHold, Loiter, Land, RTL worked well so apparently I had found some working settings.

First autotune round (roll, pitch and yaw) went well but I felt that parameters were on the tough side so I reran and roll and yaw completed fine but pitch ended up in crashes which I could not explain. Crashes were of same style as the ones I had earlier experienced, ie a sudden fall from the sky. The first time of pitch autotune was quite smooth but the second one did visually appear to be much rougher.

I enclose a link to uploaded logs labeled 1-…, 2-… and 3-… The #1 and #2 logs are recorded during two separate runs for autotuning of pitch and #3 is recorded after repair of legs and just shows a few normal tests (no autotune) which all went well.

From log #1 I found some peculiar curve points for Att-yaw just before the crash. Att-pitch and Att-roll I am not experienced enough to judge but my guess is that they are OK. For RCOE C4 went steeply high while C3 went low and my conclusion was that the C4 motor and/or esc were failing. So I replaced the C4 Esc.

Next autotune run for pitch is shown by log #2. The pattern for RCOE curves for C3/C4 are similar but now C3 is the one running high while C4 is going low. And that is why I am writing to ask for some assistance with the interpretation of these logs. C3-C4 axis seems to have a problem but I am not yet experienced enough to read from the logs what might be causing this (my interpretation might also be totally wrong).

Log #3 is just a test run directly after repairs caused by latest crash. AltHold, PosHold, Loiter tested OK a nothing extraordinary happened. A jpg of RCOE curves is also uploaded together with the logs and the link is here:!AuPg1RhV9PQ4zE4TR3Ww64k3ltW1?e=s5yTgv

I will be very grateful for any assistance which can help me find the root cause of these problems.

Reviewing log 3 I don’t see much to complain about. The tune is not bad, Vibes are reasonable given that frame (I have had a few PixRacers on those frames) and the motor outputs look OK. It looks like you may have reviewed and set the pre-tune parameters?

AutoTune cranked up stabilize Pitch/Roll quite a bit, maybe too high. Perhaps another Autotune with aggression set a bit lower.

What I have found works well for me is to set the autotune aggressiveness to .05 then run the autotune. This will probably take the longest. Then once that completes, rerun it at .1

Seems to work much better for the crafts that I have

Thanks for taking the time to answer. I have not touched the pretune parameters and because of this I found it a bit strange that the stabilized pitch and roll values were increased so much by autotune. What would be a “normal” range based on your experience?

The drone should not fall from the sky during tuning. Can you see any reason to the crashes and to what makes the C3/C4 output behave like that?

The autotuned parameters went much higher than I expected so I am planning to lower aggressiveness and try again. But first I am trying to find out what might be the cause for the crashes because I sincerely hope that it is not the autotuning process itself which produces the fall from the sky or do you think this could be the case?

Before the crash in Log 2 Motor 3 was commanded to the Max which dropped the command to Motor 4 (opposite arm) to Min to attempt stabilization. Then it oscillated wildly in roll and down she went. So, loss of thrust on Motor 3 initially. ESC, motor, prop. Autotune puts a large demand on these components.

Thanks for responding.

Also the .jpg graphs I did upload showed this behavior for both autotune runs but my assumption was that this was due to faulty components or vibrations etc. rather than some kind overload. I have not used autotune before so I am learning.

The esc’s of this drone are the cheap yellow ones, I might change them one day but as previously mentioned this is a drone used for learning and testing only.

As Dave suggests I will make a few more runs but with aggressiveness set lower and hopefully this will solve the problem.

Ah yes, the infamous Red (or Yellow) Wrapper ESC’s. Sold as generic SimonK 30A or some such for ~$4 each? I would replace those right away.

I agree, I knew these were of questionably quality but the esc’s came with the frame kit and have been OK for testing and learning to fly. But once things are seemingly working there is a barrier of laziness to climb over in order to fix things not actually broken :slight_smile: I will change them now.