Autotune Crash

ArduCopter Copter 3.3
1

Hi all,
I have been upgrading some of my rigs from the crappy Dji A2 to the PixHack and Pixhawk FC.
YES UPGRADING.
I have has an unfortunate crash during an Auto tune. This is a large H8 T-Motor U7 400kv 15x5 blades, 2x 16,000 6 cell. BeSteady gimbal.
I have test flew the rig a few times with the default PIDs and it flew relatively well on the defaults.
It was time to run an Auto tune which I have done several times on other rigs.
First part I had the Autotune bitmask set to 1 for the roll. This went very well and gave me very sound results.
Set Autotune bitmask to 2 for the Pitch. All was looking good until about 29 minutes in when all of a sudden the rear motors sounded like they revved up high. The rig flipped over the pitch quite spectacularly about 7 times until it smashed into the ground. Roll held in solid.
All took place too fast for me to react and jump out of autotune.
I have looked over the flight log several times and can see the point, just after 29 mins when this happened but I can’t see anything standing out to which could have caused this.
I have several rigs with the same setup ready for tuning but now I’m too afraid to fly them until I find a cause of this crash.
This is the first time any of my Arducopter rigs have failed on me and I have many of them.
Would some kind person be willing to look over the log and see it they can find anything odd in there. I’m still in the process of learning the system.

Thanks
Anthony

2

https://www.dropbox.com/s/2qa4m2nsu2mn81m/24.BIN?dl=0

Tried uploading log but keeps failing so hares a link.

3

ACPUK

The logfile seems to be truncated. It ends at 9.31 meters from the ground. Mission planner auto-analysis points an unusual voltage fluctuation in px4 power, which I don’t have in any of my models:

I couldn’t find any indication of a crash in the logfile. It could have been a brownout and the log has ended before that.

Also you should take care of the vibrations. They seem to be very hight and also some clipping is occurring , which is not good.

Take a look at copter.ardupilot.com/wiki/common … n-damping/

4

thank you tabascoz.
I have never managed to get the Auto Analysis to work, I always get a “time error” followed by Bad input file.

I did not spot the vibrations in the log. I did a live Vibration test from MP and the bars stayed very low.
Clearly this needs addressed.

There is a cut off from in the log. the batteries were thrown out at the 6th flip.

May I ask how you derived "Test: VCC = WARN - VCC min/max diff 0.395v, should be <0.3v"
Is this part of the auto Analysis?

I suspected the power to be the issue all along. But when I compared graph from another log powered from the USB it looked as if it had the same fluctuations.
My initial thought was the power module failed causing the crash. Hence the re-build has triple power supply.

Thank you again for looking at this for me. Most appreciated.

Regards
Anthony

5

Antonhy,
You’re welcome!
Yes, it was from auto analysis

I think Missionplanner hasn’t been updated to handle the new time format on the logfile from arducopter-3.3 on the stable version. To make it work, go to config/tunning and mark the checkbox "beta updates " . After that, you can force an update going to help - check for BETA updates . After it should work.
I never did the vibration test in MP, but I would be best to guarantee that the vibe levels are within normal range through graph analysis.

Analysing again your power graph, it seems that most of the time the voltage keeps oscillating between 5.3 and 5.2v which is normal, however there is a few peaks going to 4.9v. The board should run fine in 4.9v, but it it is oscillating it could have went to a lower level , thus triggering the odd behaviour.

Regarding the log, looking to the pitch att graph looks like it hasn’t recorded the loops. My guess is that autopilot has crashed before the batteries were ejected.

Keep in mind that autotune push the machine to its limits. That could trigger many failures that could happen in a normal flight.

Do you have a power module? I could not see any data regarding the current draw.

Regards, Fernando

6

Thanks again Fernando for your help. I will update MP to beta.
Yes I do use a power module but I don’t supply the ESC from it as it is a large rig and can easily pull 250A if motors go full power so could not risk blowing the unit. I do however have a hall effect current sensor installed. Just not set up a voltage divider to limit the voltage swing for the pixhawk ADC.

I will be installing a new power module as it is the most likely cause. The low spikes could cause a brownout. I am also installing a triple redundant power setup. Power in, USB in and RC out rail.
Just a little confused on weather I need the Zener diode for triple power, I suspect I do.
The power will be 3x ubec units.

Just not sure as yet which way to set them up as I want power backup for the companion computer also. (which was not connected during the Autotune, likewise the Lidar)

I also discovered a chunk of carbon fiber jammed under the FC which may have caused the vibrations.

The rig is re-built now so have to suss out the power this evening ready for test flight and autotune tomorrow.

Thanks
Anthony

7

Looks like tabascoz has done some good analysis here. The logs suddenly ending is the classic sign of a brown-out and this fits with the voltage change warning and also hearing that the copter can draw 240amps. I’m going to close this issue if that’s ok.
Thanks tabascoz!

8

I agree, its the most likely cause. i have installed 3 power supplies now. Power in, USB and RCout rail.
Not tested yet as been held up with another rig.

thanks again for the VERY helpful replies.
p.s. how to you close / mark a topic solved?

9

OK finally had a chance to fly again today.

3x power sources for the Pix hawk.

Autotune went perfect.

Tested again with heli strapped down and done some fast power-ups on the motors with just the Pixhawk power unit and has 3 power fails. Pixhawk browned out but only on aggressive motor changes. Not had a chance to test with a different power unit.

I need to do some more tests as this could be the LiPos dropping voltage rapidly or spikes on the power line.

Same test with triple power and no problems. Note though the UAB power is independent.

I have a heavy week ahead but I will do some more tests and post the results.

10

I have a Tarot T960 Hex Copter and have recently go to fly it. Using 3.3.3 APM and everything had been working well until today. I have used default PIDs and Copter worked relatively well and I have successfully flown in stabilize mode, loiter mode and Alt Hold Mode. I had also tested RTL several times. All good so I tried AutoTune today and things appeared well for about 5 minutes. The autotune for roll was progressing well then just before the crash, the autotune seemed to go divergent with big moves left and right and before I could get out of Autotune, it flipped and crashed. The log attached.

Can someone look at the log and see if there is something wrong. I looked at it and can’t find anything. The compass had an error but this was after it hit the ground and broke off the 3dR GPS/Compass module.

How do I share the log file (bin file, size is 3.5MB)? Board’s size limitation is 2MB. I can email to anyone that might help.

Thanks for any guidance or comments. I have ordered parts to fix the T960 and the damage wasn’t too bad. I want to understand what went wrong before I try this again!

11

@mthomson51,
Use dropbox or Google Drive.
Regards,
TCIII GM

12

[quote=“mthomson51”]I have a Tarot T960 Hex Copter and have recently go to fly it. Using 3.3.3 APM and everything had been working well until today. I have used default PIDs and Copter worked relatively well and I have successfully flown in stabilize mode, loiter mode and Alt Hold Mode. I had also tested RTL several times. All good so I tried AutoTune today and things appeared well for about 5 minutes. The autotune for roll was progressing well then just before the crash, the autotune seemed to go divergent with big moves left and right and before I could get out of Autotune, it flipped and crashed. The log attached.

Can someone look at the log and see if there is something wrong. I looked at it and can’t find anything. The compass had an error but this was after it hit the ground and broke off the 3dR GPS/Compass module.

dropbox.com/s/kmfwnmmdv5xtd … n=69095336

Thanks for any guidance or comments. I have ordered parts to fix the T960 and the damage wasn’t too bad. I want to understand what went wrong before I try this again![/quote]

13

It was gradually reducing roll rate P from initial 0.113 down to 0.034…
Looks like relatively huge change, but I can’t tell for sure if that’s a problem, because:
if you plot the final wild oscillations , then you see it still uses a lot authority to control roll, with no success.
(plot AHRS.Roll and RCOU.CH1 and RCOU.CH2)
So I am suspecting motor sync problems…

14

Thank Andre for looking at the data. Given the hex seemed to fly reasonably well, it seems odd that the P value would be reduced that much. Is that typical? Is there any way for me to tell for sure if it really was a motor sync issue? As I watched (sorry no video), it seemed to go divergent making bigger and bigger pitch angles until it couldn’t recover. Unfortunately, I wasn’t expecting it and next time, I’ll have my finger on the auto tune switch in an attempt to recover the situation.

15

it’s first now I realize how late I answered this thread, did not see the other replays before now.
Anyway - the thing is that the log, does not show increasingly big angle toward the end, could it be that the AP got loose ?
I think it’s on the low side, but given the PWM recorded, I don’t think that was the cause. - it seems it still was in control.

16

I had similar issue yesterday. Started in alt-hold mode, switched autotune, roll went ok and during pitch tuning evrything switched off.
Looks like brownout but there is another power source attached to servo rail.
I don’t zener diode but only small tilt digital servo is attached (apart from motors).
Interesting fact is that X4R receiver powered by Pixhawk reported 4.5 volt.
On my previous flight I saw two very short gliches but nothing happened.

Logs are here (last log is from crash):

17

Hi, I have a large experimental quadcopter that is 3m long and 1.3m wide, using 4 of 22 x 8 props and 4 of 6S 8000mAh batteries, all connected in parallel. It was running a Pixhawk clone, but now a Radiolink Pixhawk and GPS.

The original Pixhawk started loosing its horizon, and the drone would lean left and land heavily.

I changed over to the new controller 2 days back, and after all the calibrations and setup, the drone was flying ok on Stabilise and Altitude Hold, but on Position Hold it would start to oscillate, lean left, pitch forward, lean right and pitch backward, around and around getting worse. I decided an Autotune would probably fix the problem.

When I invoked Autotune, (all axes), The drone started pitching left and Right like it should. About 1 min into the Autotune, the drone suddenly went into a rapid roll, and rolled about 3 times before hitting the ground with lots of damage.

Logs end part way into the Autotune Roll Left at about 12° of lean. Some previous movements were up to 20°.

This was not a brown out. Board voltage is pretty stable at between 5,2V and 5.35V, right to the end.

The only thing I can see in the logs is that the motor outputs were getting pretty close to maximum, a couple of the motors are running at 1764 average and going to 1954(max) (RCOUT values) during the rolling process. Its possible (but unlikely) that 2 of the motors lost sync together (perfect rolling motion, no pitching). there is actually very little variation in RCOUT for the motors in the final logs, so I don’t expect it was fast motor speed changes, as I’ve seen in other posts.

Perhaps it just leaned too far and couldn’t recover. The roll speed was very quick when it was falling. I think it did around 3 complete rolls from about 15 meters in about 1 second.

Anyway, I’ve attached the logs if anyone wants to have a look.00000008.BIN (876 KB)

Maybe I just need 22 x 10 props?