Bitcraze Forums

The Crazyflie 2.0 seems to be extremely unstable out of the box. We have yet to find person who can fly it more than 10 seconds. So first let me ask this: Is this normal? We are using XBox 360 controller and PC client app running from source under Windows 10.

Here are my observations on this issue:

1. Just hold Crazyflie on table and only do thrust. Observe that not all 4 motors are NOT driving equally. This will create random drifts.
2. Hold Crazyflie in hand, apply thrust and orient crazyflie in different direction. Observe that motors don't attempt to compensate and they even randomly stop sometime without getting recovered.
3. Hold crazyflie and apply thrust+pitch or thrust+roll and observe the 4 motors are essentially getting powered without much regard to the intent.

So do some tests like above and it's very clear that stabilizer in Crazyflie is not only completely ineffective but is even detrimental to stability. I see huge gaps beyween desired pitch/roll and actual pitch/roll. I updated with latest firmware and still see these issues.

I first want to confirm if these are known issues or if there is some magic config or settings that needs to be done here. I've 4 crazyflie and all are showing same symptoms.

Hi,

This is definitely NOT normal. Out of the box the Crazyflie is quite stable and should be easy to fly.
There is always some drifting involved with ground effects, but once you're about knee high, it should be fine.

In general this sounds like a hardware problem. This may sound too easy, but please triple-check that you have mounted the right propellers in the right way and not upside-down!

Observe that not all 4 motors are NOT driving equally.

This sounds problematic. Does that happen with all four Crazyflies?

When you connect a Crazyflie to the Crazyflie Python client and tilt it with your hand, does the artificial horizon react to the movements of the copter accordingly?
Please also check the log output in the Crazyflie Python client if it shows any errors.

If possible you might also want to try a different client (e.g. Android or iPhone) just to isolate the problem.

Regards,

Fred

I'm having this same issue. After tinkering with the settings for a while, I've made it better but haven't managed to fix it entirely.

Things that helped:

-Adjusting pitch and roll to compensate for the random movements
-Nudging the battery back and forth by millimeters
-Increasing the thrust slew so the copter doesn't do that odd delay before I can shut it off.

I've also observed what OP has described with the copter not self correcting. If I hold it with the small side tabs, and tilt it, it doesn't fight me to try and remain upright like my little $16 Chinese clones do. I'm not new to quads, I run Devo TX on a handful of models from a fullsize flamewheel, to tiny chinese micros, and this one just doesn't seem very flyable. As well, it has randomly disconnected from the crazyradio a couple of times, and instead of the killswitch dropping it out of the air, it flies by itself until it hits something, which it always does.

I'm running the latest version of everything, inside the VM from my win7 machine, the artificial horizon reacts, and the problem is WORSE on the android client.

Something is definitely wrong. Does the Crazyflie 2.0 vibrate a lot? Could be a bad propeller or motor. When you are saying it drops the connection is it from a mobile, the crazyradio or maybe the DevoTx? Have you ever gotten a selftest failure (red flashing 5 times quickly over and over during startup)? Could be a sensor problem.

Little to no vibration, but the connection drop occurs mainly with the crazyradio. Standing between the radio and the quad is often enough to knock it out.

I've never had a self-test failure.

The radio link can be interfered. I suggest you test by changing channel and datarate. Try e.g. 2Mbit and channel 100. As the radio currently doesn't do any channel jumping, finding a good frequency can make a big difference. Also higher data rate has a less chance of collision with e.g. WiFi but less range which is better inside.