[4.1.2] Unwanted Yaw mouvements

[Luis-Sanaria (Luis Sanabria)]

Hi everybody,

Until today I’ve being flying with the ArduCopter 4.0, but now I would like to change to 4.1 because of the improvements that have been added in general.

Sadly, after some tests I can see a behavoiur that I did not see in some versions before.

Before sending my drone in AUTO, I always make sure it responds fine to every external input that can occur during the flight (like strong winds). To do this, I test the drone with some roll/pitch/yaw inputs individually, and after that I try to make some circles moving pitch and roll like in this image :

Normally after I release the sticks the drone stays in the same position it was before. The problem now is that when I do this mouvement, with the 4.1, the drone begins to deviate in YAW. I can see the yaw measure vary so I want to imagine is not a mag problem. I also checked the orientation and the distance of my motors.

In the end, I though it was a problem with the fact of using the EK3 instead of the EK2. This is why I made some tests changing the version and the EKF used.

Results :

In 4.0.7
EK2 - no problem
https://drive.google.com/file/d/1eTLrSPf9gysdoeS2xnmewC6nXV38M39Z/view?usp=sharing
EK3 - no problem
https://drive.google.com/file/d/1uVuDhQ9iDCO1Rd2IxUcmZas8u9zOOFAe/view?usp=sharing

In 4.1.0
EK3 - Problem
https://drive.google.com/file/d/1pfwpVy9dGEIpYuIDfoGQM-dd-0adNGBI/view?usp=sharing

In 4.1.2
EK2 - Problem
https://drive.google.com/file/d/1LnRRnRboEekAt1ZROACN2aBSAYQnikfk/view?usp=sharing

EK3 - Problem
https://drive.google.com/file/d/1-SmN82wjbqLnsKr5orxhFZW9w2I6LaOH/view?usp=sharing

With the logs available, I add some screenshots where I plot the PITCH (blue), ROLL (green) and YAW input (yellow) along with the DesiredYaw (red).

With the evidence of these results, the first thing I did was to check the release notes and I noticed there are some changes that have some relation with yaw behaviour. I am not an expert, so I would like to know if maybe this behaviour is actually desired for some reason I ignore. If this is the case, I would like to understand just why, so I can adapt the mouvements of my drone.

Even if I like the idea of this being a new feature, I hardly think this is the case. I would like to know if I need to change some parameters in order to make the yaw work again as I did it in the previous version, or if this is just something that needs to be changed.

Even if you have a little idea, that would be really helpful.

[xfacta (Shawn)]

Exactly what yaw issue are you seeing? In the last “EK3 - problem” log Yaw is actually the best and pitch and roll need more tuning.

You are doing lots of large stick movements, it could just be that Yaw is not keeping up with your expectations considering the amount of pitch and roll.
Your compass performance looked good, that doesnt seem to be an issue.
Try increasing ATC_ACCEL_Y_MAX,11284.98 to about 20000

You can set these:
ATC_THR_MIX_MAN,0.5
INS_FAST_SAMPLE,1
LOIT_BRK_DELAY0.3
PSC_ACCZ_P,0.26
PSC_ACCZ_I,0.52

You dont have your battery failsafes or voltages set - I strongly suggest you do the “Initial Parameters”. Update MissionPlanner to latest beta, go into Setup, Mandatory Hardware, Initial Parameters.
You’ll probably want to accept everything you can find on that page, except the T-Motor Flame ESC settings, and then it would be best to do Autotune again.
Ideally you would set up Harmonic Notch filter too, and I think this would fly really well, since it’s responding well to your stick inputs (tuning considered).

For Harmonic Notch filter start with these settings and do a short hover and a few gentle movements, nothing radical
INS_LOG_BAT_MASK,7
INS_LOG_BAT_OPT,0
INS_HNTCH_ENABLE,1 ← set this then refresh params to see the rest
INS_HNTCH_MODE,1
INS_HNTCH_REF,0.2
INS_HNTCH_FREQ,80
INS_HNTCH_BW,40
INS_HNTCH_ATT,40
Then we can check the FFT in logs, fine-tune the settings and then run another Autotune.

I wouldn’t worry about testing with EK2 since EK3 has had plenty of testing and real-world use now.
I can see you are being thorough about your testing though, trying to narrow down the causes.