Hi Felix,
I have been following this thread but haven’t had time to spend looking at the data. The issue does not seem clear cut.
@Goobisofty looking at the picture of the servos, the servo arms look quite different from what I have typically seen. Are these part of the heli kit or did you make these or buy them separately?
One looks like it has a bolt on top and the other does not and it looks like there is a bolt that squeezes the two halves together.
You may want to check and see if any of the servo arms have slipped on the servo. I would also look at the flexibility of the arms, meaning check that they are not too flexible. Flybarless designs require a lot of torque to hold and move the linkages. The blades constantly want to go to flat pitch.
I will try to look at this in the next day or two.
Hello Goobisofty,
In the above picture of the helicopter, I noticed one of your servo arms is missing a critical screw. Is there a reason why this screw is removed? It makes me wonder if there are other critical screws missing in your helicopter.
Yes at time time of building the screw was no there cause I was centring the servo arm and adjusting the pushrod length and also the picture which u saw was the other heli not this crashed one and the arms were not included with the kits so we used metal horns as suggested by the manual and the heli which was crashed were using even thicker servo arms than those u were seeing in the picture by
This was the servo horn we were using in this heli
But still I am confused of that 180 degree yaw turn by itself could I’d be servo bec power loss
Thanks for sharing the video. That’s really helpful. There is one thing that instantly caught my attention. Right after takeoff, the heli is in a hover a few centimeters above the ground and then suddenly skyrockets. That is not in line with the stick inputs of the pilot.
The pilot slowly raises the collective (yellow line). The altitude plot (pink line) shows that the pressure under the heli changes, but in the moment the heli starts to rise very quickly, there wasn’t a change in collective that could explain that behavior. Before I saw the video, I interpreted that as the heli somehow sticking to the ground (for example in high grass) and then suddenly coming free. But that is definitely not the case here. The heli is already in a hover and then suddenly climbs. This climb wasn’t commanded by the pilot and as I already mentioned, in stabilize mode, collective pitch is under full manual control, so the climb wasn’t commanded by the FC either. I also can’t see large changes in the RCOU signals right before the climb. That again leads us to a mechanical failure. Something must have happened that caused the swash plate to go up.
That is probably what the pilot reported as sudden climb. I always looked at the second climb right after the heli started to pick up speed to its left side. But the problem definitely occurred earlier.
No, it was a result of the mode change from stabilize to acro. As you can see in this plot, it happened right after the mode was changed and it was indeed commanded by the FC. The pilot did not actively command it by stick input, but he showed a fast and adequate reaction and accepted the movement (he made a stick input in the same direction). We could dig deeper into the logic of the FC and why this movement happened after the mode change, but this was definitely not the cause of the crash. It happened long after the pilot had completely lost control over pitch and roll movements.
Initially the take off was completely in control but suddenly after take off it climbed unintentionally and then it was doing its own thing and we never intended for aggresive collective or cyclic maneuver as this heli was a scale heli so we were going nice and slow but sometimes going nice and slow doesn’t help and yes the pilot later on switched into acro mode after failing to take over the control but that didn’t work at all
Please post the log file for the maiden flight where you said the aircraft behaved well. As for the aircraft yawing 180° during this last flight, I have a theory regarding acro mode and how it handles yaw when aircraft target attitude goes inverted. I will need more time to review the code and understand how it handled the situation. What bothers me is that the target angle wasn’t limited as it should’ve been because acro trainer mode was being used.
@Goobisofty I have confirmed using the simulator that the aircraft will perform uncommanded yaw when large pitch (elevator) stick inputs are made in the trainer mode that you had set (the default settings). Even with a well tuned aircraft this will happen. So I will have to check with @Leonardthall to see if this is expected behavior for the trainer mode.
I think the poor tuning combined with the over control in acro contributed to your crash. Since you were using the trainer feature in acro mode, you should have been able to just center both pitch and roll sticks and the aircraft would seek level attitude. Personally, I think it is better to stay in stabilize mode for initial testing.
That is a heart breaking crash. Sorry to see all that work go into the ground like that!!
The first thing we need to understand when discussing “uncommanded yaw” in acro is that the ATT message is the Euler angles while acro is being controlled in the body frame. This means that you don’t need to command any yaw input to change the euler yaw angle or heading of the aircraft. For example if you roll 90 degrees then pull the pitch stick you will not see any change in Euler pitch angle, the roll angle will stay constant at 90 degrees and the yaw angle will change quickly.
The other thing to understand is the target euler angles are where you have commanded the aircraft to go and when an aircraft is not able to control it’s attitude then it will be a combination of both the attitude you have commanded and the effect of the actual aircraft attitude dragging the targets around due to limits in maximum angular error. This can make looking at a log like this tricky.
One final point before I get to the log. Both acro and Stabilise are both flying the aircraft in the same way. If you can’t fly an aircraft in Stabilise you can’t fly it in acro. The reason we recommend Stabilise is if you centre the sticks the aircraft controller will do it’s best to level the aircraft. It will do this even if the angular errors are getting out to 30 or 40 degrees. However the same angular errors in acro make it impossible for the pilot to fly because they can’t see where they are telling the aircraft to go. They are commanding a ghost aircraft and the real aircraft is trying to get to where the ghost is. But the pilot can only see the real aircraft. This isn’t a problem when the difference between the ghost and real aircraft is only 1 or 2 degrees. But when it is 30+, nothing good happens.
Ok. Lets look at the flight.
As soon as you took off the attitude error started to grow. By the time you switched into Acro, the aircraft had achieved a roll error of 29 degrees and a pitch error of -45 degrees. At the moment of the switch the aircraft had a roll error of 6.74 degrees and a pitch error of -40 degrees.
As soon as the pilot switched to Acro they commanded the aircraft to do a 360 degree flip as they attempted to lift the nose. What I mean by this is the ghost aircraft did a full 360 degrees rotation in pitch but due to the poor control the real aircraft couldn’t follow that at all and only managed to pitch up by … well it hardly moved.
As you expect the yaw target flipped as the ghost aircraft went inverted then level again. As the pitch command is reduced there is a lot of roll that gets commanded and the ghost aircraft rolls inverted for a second. Again the roll and pitch control is so poor the aircraft doesn’t do this.
This is the roll and pitch errors through the flight:
Thanx very much for the explanation, in short what I understood that if we had heli in stabilize mode it wouldn’t have done a yaw flip but there is still one unsolved mystery remaining and that it why in the first place heli gained so much altitude on its own the pilot was well aware of acro and stabilize mode but wasn’t expecting a sudden climb in altitude that panicked us and top of that it was not even responding to our stick movements in stabilize mode but prior to this flight everything was going very normal heli was flying fine but we did not verify by looking at logs I can even share the video if u want of that flight but unfortunately I don’t have logs for that flight and I cannot go back to pickup that sd card cause it was some freelance project work and my job was only to setup ardupilot on these helis but flying part was not my part due the pilot have much more experience flying these helis in acro mode .
Sorry but Unfortunately I don’t have the log file of that maiden flight I have to go there to recover it and that is 2000km away from me at the moment I might go back at some point in future and make a new thread mentioning this one.
Previously the throttle curve was like this during maiden 0-25-50-75-100 percentage but after maiden it was something like 65-66-67-68-69 percentage don’t remember the actual numbers but was similar to this
Thank you very much, @Leonardthall for this detailed analysis and impressive insights! I learned a lot!
So what we do know now is that the bad attitude response of the aircraft lead to all these complicated problems which finally resulted in the crash. But there is still the open question what caused the attitude error to grow so much. Was it just poor tuning? Actually, I don’t think so, as the values don’t look completely wrong to me. I would expect at least some kind of reaction. But the heli showed almost nothing. I might have overlooked something, some filter setting, servo limit… that would significantly reduce or slow down the control response. But actually, the servo output channels are moving quite a bit and again, I would expect at least some kind of reaction from that.
That might be helpful as we could see, if the heli showed a fast control response in the first flight. In that case something must have changed, probably failed. But I don’t know, if a failed linkage rod or servo arm is enough to explain the behavior of the second flight. There was almost no response in pitch AND roll. And right after takeoff we saw that sudden climb that was also not commanded by the pilot or FC.
Yeah the first flight was mostly hovering and stuff but it was completely under control and the first battery pack gave us 12 minutes of hovering time will upload video and share link thank you
So did you know what rotor speed you were targeting? Did you measure it when you decided to use around 60% for all the points. If so what was the rotor speed?
