Hi all,
I have a little problem with my quad, and as I tried many usual solutions, I'm posting here.
I notice a slight (but visible) roll/pitch instability when hovering, in Acro/Stabilize/AltHold modes (and other). It doesn't looks like an oscillation, but more like a kind of random roll/pitch instability. To test this behavior I flied today, a good-weather and absolute-no-wind day. I hovered in AltHold mode for about 20 seconds, but the same issue could be observed in Acro or Stabilize modes.
The image log result of Roll/Desroll and Pitch/Despitch datas is attached ("log1.png"). As you can see, I have an amplitude in the Roll and Pitch axes of about 2 degrees. I am confused as I flown with a Dji Naza (an other configuration) and this behavior is not observed.
The second log image file attached ("log2.png") comes from the same graph with an other scale : I did first one right then left maximum roll angles (45, -45), then the same forward and backward. As you can see, the roll angle is less responsive than the pitch axis as I use the same gains but my battery is heavy and along the right-left direction.
The third file attached is the complete log file, my 20 seconds of AltHold hover test are between log lines 5600 and 9300.
Where do you think it can come from ?
Many thanks.
Francois
Here is my config :
- dji F450 frame
- Pixhawk with AC V3.1.5, fixed on the frame with 4 little slices of earplugs (vibration levels are slow, AccX in [0.2, 0.78], AccY in [-0.34, 0.31], AccZ in [-10.6, -9.35])
- 3DR uBlox Lea-6H GPS + Compass
- 3DR Power module (the one that comes with the Pixhawk card)
- T-Motor MS2216 kv1100
- 4S LiPo 3.7 AH
- DJI 1038 plastic propellers
- HobbyKing F-30A ESC flashed SimonK
- as my quad is overpowered, autotune does not work so I did all PID settings manually. As you can see in the second image attached, the behavior is good.
Replies
Your PID values look odd to me...generally these days P and I are the same numerically, and your D value looks very low. Typically these days D is 1/10 of P and I. You said that these were done manually. I think you need to revisit these.
Looking at your rcouts, your copter does look nicely powerful, but not rediculously so...so I would think auto tune should work, but I have had to manually tone my latest very small 200mm quad...so it doesn't always get good results, but a 450 size one should be ok.
also if you used 3.2, you could try EKF to see if it improved things.
Thanks for analyzing my log.
I tried to change my PID values yesterday, and the result is :
- the P-gain of about 0.1 is good, in acro mode I have a very sharp response without oscillations. I have a slight overshoot, but the phenomenon is probably explained by this page in the "advanced tuning parameters" section.
- I increased the D-gain from 0.006 to 0.008, which is not to far from 1/10 of the P-gain. Starting from 0.009 I get some high-frequency oscillations.
- the I-gain looks more subjective to me. I can see the difference between no I-gain and "some" I-gain, but once activated with non-zero values the response is not so clear even in fast translations. To my mind, I need a very windy day to see the difference.
I also updated the firmware to 3.2, enabled AHRS_EKF without noticeable difference. I didn't performed the "20s untouched hover" as before, but I will do it ASAP to compare the logs.
I'll keep you in touch.
OK cool. I think that overshoot in that link is caused by the I term building up, but you may be seeing a slight over shoot due to high P because higher P is on its way to oscillating if increased.
Yeah too much d will give high speed oscillations. The 1/10 is just a guideline....my mini quad isn't using a 1/10th ratio.