Page 1 of 1

good propeller test values

Posted: Mon Feb 15, 2021 2:12 pm
by berkc
Good evening guys,
I am trying to tune a PID controller while using a script that lets the drone hover but every now and then the cf just flips at the start of the script.
I was going through the debugging process and to strike out rotor problems I was wondering what values for the propeller test in the cfclient are okay.
The values I get always seem to be different from rotor to rotor such as:
m1= 0.9 (z=0.9), m2= 0.7 (z=1) , m3=0.8 (z=0.6), m4=0.6 (z= 0.7)

I had one with 3.5 which I switch as the low-high range is 0-2.5 , but my question is in what range the values can differ without causing instability(as I had other cfs with higher differences).


While I am at it a second question:
While propeller testing I get the warning that the kalman prediction rate is low, I also seem to be getting drifting estimations so I guess these problems correlate. Am I right to assume that recalibrating the Lighthouse would fix this? ( already tried different cfs, decks and pcs)

Thanks!

Re: good propeller test values

Posted: Tue Feb 16, 2021 4:21 pm
by tobias
As for the the rotor values it is very unscientific and often values over 10 is needed to cause any problems, but lower is better.

Regarding the flipping problem, is this with the lighthouse? Perhaps it is related to this lighthouse reset problem?

Re: good propeller test values

Posted: Wed Feb 17, 2021 10:43 am
by wydmynd
unrelated to lighthouse issues, I also found that in some crazyflies the motor PWM values differ consistently while hovering. In many cases two motors will have higher values and the other two will have lower values. you can use motors.m1 .. motors.m4 log variables to test this while hovering.
Replacing motors helped me reach more consistent PWM values during hover. This greatly improved stability. I assume this is due to manufacturing differences in the motors.

Re: good propeller test values

Posted: Thu Feb 18, 2021 8:01 am
by tobias
Yes a more balanced system will fly better but for the Crazyflie to flip I believe the lighthouse system is receiving some outlier that makes the controller unstable. Improvements needed here, to make the lighthouse implementation more robust. The outlier could be caused by e.g. reflections of the sweeping laser beam.

The difference in hover thrust values are most often related to a yaw force, you will then see two motors higher and two lower? This can be caused by the motor mounts not pointing "straight up" but are angled a bit.

Re: good propeller test values

Posted: Tue Feb 23, 2021 10:52 am
by berkc
wydmynd wrote: Wed Feb 17, 2021 10:43 am unrelated to lighthouse issues, I also found that in some crazyflies the motor PWM values differ consistently while hovering. In many cases two motors will have higher values and the other two will have lower values. you can use motors.m1 .. motors.m4 log variables to test this while hovering.
Replacing motors helped me reach more consistent PWM values during hover. This greatly improved stability. I assume this is due to manufacturing differences in the motors.
Sorry guys I just now saw your replies.

I also saw that Happening when manually (per vf client ) Controlling the Cf to hover, but it was Alternating between a higher value for m1&m4 and m3&m2, But still in a more or lesS stable hover

Re: good propeller test values

Posted: Tue Feb 23, 2021 11:15 am
by berkc
tobias wrote: Tue Feb 16, 2021 4:21 pm As for the the rotor values it is very unscientific and often values over 10 is needed to cause any problems, but lower is better.

Regarding the flipping problem, is this with the lighthouse? Perhaps it is related to this lighthouse reset problem?
Yes with the lighthouse + deck
The reflectivenes should Not be the issue as a Week prior the flight worked Fine.


I also tested a few different Batteries that I Took from the charging station(at 100%) so I don’t think thats the reason but I will Test it again.


A Little Update: recalibrating did Not help we also tried using the newest and one of the older Firmwares Bit that did not help either .

Are there some tests you can suggest for Controlling the correct function of the radio? Maybe there is a problem there

Re: good propeller test values

Posted: Tue Feb 23, 2021 3:32 pm
by tobias
Do you have two crazyflies so you could compare? We have seen some rare cases with bad IMU (BMI088), but it often has problems leaving the bias calibration state in the startup then.

Have you tried logging the lighthouse angles or the position during the startup to see if there is a outlier or similar?