Suspected IMU problem

[ccpost]

Hi all,

I have 2 Crazyflie 2.0 units. One of them flies beautifully (trim values at 0 even). The other seems pretty borked.

The attitude indicator appears level in the Python client when sitting on the ground. On taking off, the trim goes way off to the right slowly over the course of a few seconds until it seems to level off. I can fly it just barely with a ton of left roll. If you look at the attitude indicator in the Python client while I'm manually holding it level, it thinks the copter is rolled to the right what seems to be the amount I'm compensating. The copter has some high-frequency oscillations, but the high-frequency component isn't too much to be unmanageable on its own. The right roll bias floats a little bit in magnitude over a period of a few seconds while flying. On landing, the attitude indicator is usually slightly off to the right, but closer to level than in flight. Over the course of about 5 seconds it drifts a little closer to level.

I flashed the test firmware in viewtopic.php?t=1233#p6762 to see if it yielded any IMU problems. I can't tell if this means it's checking out OK or not since the other samples had many more status lines. Mine doesn't spit out anything more than this after around 20 minutes of letting it sit jus tot be sure.

Code: Select all

SYS: Crazyflie is up and running!
SYS: Build 6+:29624e7f6e3b (2014.12.0) MODIFIED
SYS: I am 0x37303638343551173D0049 and I have 1024KB of flash!
IMU: MPU9250 I2C connection [OK].
IMU: AK8963 I2C connection [OK].
IMU: LPS25H I2C connection [OK].
30 78 42 43 0 50 0 0 0 0 0 0 0 0 0 7D 
Status 0

Any ideas on further things to debug? I suspect that the IMU is just bad since it's idea of level is drastically different in-flight and on the ground, but I can't really confirm.

[arnaud]

Hi,
What we have seen so far with the IMU is that either it works, or it does not. But since it is MEMS chips, if they pass self test they are good (ie. we never saw big variation between IMUs that passes self test).

Your problem might be due to vibration, did you try to balance the propellers? (we have a documentation for that on the website). What version of the firmware are you running?

[ccpost]

I'm on 2016.09 firmware.

I haven't yet done a propeller balance. That was next on my list. I'm skeptical, though, since when I've had prop issues on other non-Crazyflie copters they just vibrate a ton, but have never exhibited any consistent bias where the attitude totally diverges from ground truth in the same direction and magnitude every time.

[ccpost]

Alright, so after a bunch of stability work here's where I am:

After balancing the props really carefully, all the visible vibrations dampened out, and there was a tiny bit of audible oscillation. It was still biasing towards right roll a few seconds after taking off, but wasn't nearly as bad. On my other one (the stable one) I have one of these prop guards installed: https://www.amazon.com/gp/product/B00EC ... UTF8&psc=1

In the interest of ironing out the only remaining difference, I moved the prop guard over to the problem copter. That seemed to add just enough mass (or something) to calm it down, and its flyable now.

I'm still a bit confused, since I didn't do any prop balancing with the stable one, and it's still more stable than the one I spent a bunch of time fiddling with.

The consistent roll bias caused by vibration is also odd to me since I haven't seen that failure mode on other quads. Is that a normal symptom of vibration?

Thanks for your help and for an awesome hackable product!

[tobias]

Like Arnaud wrote above we have manly seen sensors that work or doesn't work so I'm also thinking vibration. What if you move the propellers from the balanced CF2 to the problematic one? Also check if there is any bent motor axis.

The prop guard will help dampening the vibrations so that might be why it works better with that on.