After several successful flying missions, the last one… wasnt. And we cant figure out why. =( Can we rely on Pixhawk 2.1 / firmware with the other aircrafts is the question until we find the problem.
Controller: HexAero ProfiCNC Pixhawk 2.1 with CarrierBoard
PM: 3DR 90A 10S
GPS: HexAero Here+ RTK
Telemetry: 3DR 433MHz SIK-radio
Receiver: Futaba R7003SB 2.4GHz FASSTest with 3300uF electrolyte capacitor for 5V
Camera: Sony A6000 with Drotag
LiPo’s: Dual (parallell) GensAce 3S 6000mAh 30C
(Transmitter: Futaba T14SG v5.0 2.4GHz FASSTest
Software:
Mission Planner
APM:Copter 3.4.6 (the latest at the time it was built and firmware’ed)
The multicopter have been flying great and all have worked great! But this last flight:
Stabilize mode, started flying up and away a bit as planned, switched over to Loiter (still flying up and away) and shortly after that switching to Auto. The multicopter started turning to WP1 (still increasing height) and then it started yawing fast in the other direction, faster and faster (1 revolution per second in the end). It still increases height and travelling to WP1, but rotating. Engaging RTL and it continues to WP1 a few seconds (same smooth curve) but then starts descending, faster and faster until it crashes feets down, not at Home position. This is seen in the KMZ-file as well. All this happens the first 20 seconds.
The LOG-file gives a lot of information, but we cant find out why it starts rotating and what makes it descend in the end.
I have desmantled the hole multicopter to find the fault.
Nothing have loosened (all things are there they should be)
No arms have turned
No arms have jumped out of the holders
All screws are tight
All cables are okay as I can see (the only connectors on the multicopter are the ones on the Pixhawk and the batteries)
your cube power seem to be dangerously low, it should be around 5.3V not 4.8-4.3. However it can be measurement error from a faulty power module or by a ground offset.
If you look at the VibeZ and Clip2 values, just before the rotation started, VibeZ jumped to from ~20 to ~70 and then clipping started to raising as well. This shows that something got loose onboard. Everything after this caused by the excessive vibrations.
Where do you find the Cube Power? Is it Power - Vcc you mean? Shouldnt it be 4.0-5.whatever it was (read it yesterday). It is powered through a 3DR 90A 10S PM. Should be 5.0V I guess. The module was ordered from 3DR and was used only in this multicopter, may have been a bad one? Shouldnt power be more stabile?
VibeZ… thats a lot of vibrations. What is clipping (Clip2)?
After a few hours of talking to the pilot, looking at the log, looking at the wreck… We found out that there could have been some parts that got loose or something.
If the Here+ got loose, wouldnt the compass-values from the Here+ and the onboard compass-values differ?
When the multicopter started “falling” the hood on the Tarot flew off. It is a good wind-catcher when going down. It has three points where it is fastened to the frame with rubber grommets. If one got loose, the hood/lid could have flapped a lot?
The battery power is 12,6V in the beginning (4,2V/cell), and after only 5 seconds, it was down to 3,6V/cell. Thats low. CURR - Volt.
One theory is that one motor-mount turned on one boom a little, but thats quite impossible, it isnt that loose at all. Have to use “man power” to rotate.
Could a propeller broke in the air? Shouldnt that increase X an Y vibrations?
Vcc should be 5.3-5.4 volts. A working power module should output 5.4v. However if the module is faulty, it’s ground reference could move with the power draw, and that makes the voltage meassurement in the flight controller invalid. But generally if power-vcc is below 5.0 volt then there is a problem with your power system.
X and Y vibes also increased significantly at the same time when Z. This is consistent with the loose item theory. Clip variables incremented when accelerometer is saturated, Clip0-X, Clip1-Y and Clip2-Z… you also see increasing values for all three axes.
Motor mount turning not cause fly away, just a momentary yawing… (Don’t ask where I know it )
VOLTAGE
I just measured the 3DR 90A 10S PM output voltage (connector to Pixhawk 2.1) with no load, and it measures 5,29 Volts. With Pixhawk connected. it measures 4.00-5,03 Volts.
And with the Power Brick (it cant handle higher amps) that came with the Pixhawk, it measures 5,36 Volts with no load and 5,32-5,33 when connected to the Pixhawk.
3DR module is not good for the Pixhawk according to what you said. What PM should we use? Better to use an UBEC for power for the Pixhawk (such as HobbyWing 10A UBEC that we use which is configurable)?
Lesson learned, use a power module that can keep the voltage, measure before use.
VIBES
Yes I see increasing Clip, Z increases a lot more than X an Y.
Motor mount couldnt have turned. Anyway, a lesson: Dont use round booms! =)
Well, I guess we never know what caused the start of the rotation. Something got loose, dont know what. The only loose things was the GPS, the lid/canopy and if a propeller-tip broke in mid-air. Or do you think there could be anything more to look at? I learn reading logs better and better. =)
I personally use Mauch sensors on all my serious copters/planes. I found inaccuracy with standard power bricks causing me to land early, when there was actually acceptable voltage to maintain flight for many minutes more. Switching to the Hall effect Mauch sensors has given those extra minutes back to me, so I consider well worth the expense. They come with calibration data supplied so every module calibrated in factory - just bung the supplied current/voltage dividers into parameters and good to go with trusted voltage/current readings. I use a separate bec to supply pixhawk power connector - also a Mauch, plus a bec to supply the Mauch itself as these need a supply of their own to function.
We bought a bunch of AttoPilot PMs for the newer builds. How are they compared to Mauch? AttoPilot is also recommended (ArduPilot), thats why we got them. Mauch was sold out everywhere at the time, if I remembered correctly why we choosed AttoPilot.
We never got the 3DR 90A 10S to show the proper voltage and current readings (on the multicoopter in this thread). It was easy to calibrate and it showed the right values in the beginning, but the voltage lowers faster than reality and failsafe came minutes before it was time to land.
It was the attopilot 180a sensor I replaced with Mauch in my 680 hexacopter. When I said I had issues previously with power brick, it was actually attopilot. I replaced a standard apm power brick in my fx61 LRS wing with Mauch also recently and have seen far more accurate readings. I was getting random low voltage readings before, but feel a lot more confident with the Mauch fitted. I think for commercial applications Mauch is a no brainer.
)