Could you make a binary for a pixhawk 1
Sorry, I forgot about that; PixHawk1 - Google Drive
this flight was done with bodyframe-roll update
The control is very good, I havenât seen any bug. Something to mention when q_yaw_rate_max is too high, lets say 300°/s, there are glitches.
Sometime, attitude went bad but this is a frame limitation.
To my mind the updated bodyframe-roll is very good and I will keep flying with it but the difference with the actual is not easy to see because it it very difficult to hardly roll the plane. I need to come back to plane 4.2 to compare as I hadnât flown this plane for around 2 months.
I havenât tried to fly inverted aggressively, thatâs truly very difficult , but yes I saw several time a bad control when flying inverted (with 4.2).
Other things to mention, there are less control surfaces oscillations and transition to Q_loiter is now so Smoooooth. The second transition you can see was FBWA>QLOITER.
Congratulations to developers.
Thatâs some pretty exciting low-level flight. Was that all QSTABILIZE up to 4:21?
And that transition from FBWA to QLOITER really does look great.
And yes, if the yaw error gets beyond 90 degrees, it prioritizes roll/pitch to avoid loss of stability.
But it will eventually reach the target yaw. We might want to let the yaw target âslipâ instead, since it probably looks like uncommanded yaw when it finally catches up.
it was qhover up to 4:21. The autopilot make a good job at managing the throttle and this is easier for me.
An other video to show what I call yaw rebound that I get with the last binary you made for me. q_yaw_rate_max=180, q_a_rate_y_max=360
When the yaw stick is released, the yaw reverse for about 60°. This phenomena appears with Qhover (3 first attempt) but not with qloiter (2 last). I believe the plane is able to roll (bodyframe) at a rate much higher than 180°/s and I am wondering whether it makes the control more difficult when flying around.
Thatâs also apparent in the python simulation, and it happens because the roll stick is directly affecting bodyframe roll and also commanding an euler yaw rate at all pitch angles.
When the nose is vertical, bodyframe roll and euler yaw are on the same axis, so when you return the ailieron stick to neutral you get a reduction in yaw rate (down to zero) while bodyframe roll also goes to zero. A slow return to neutral will mask this effect. I could also try to modify the code to ramp the bodyframe roll component of the aileron stick down to zero as pitch goes to nose-up. That would be more like the current behavior in stable and master.
@losawing I generated a new binary with reduced yaw-error limit (down to 30 degrees from 90) and the change in roll behavior with nose vertical. Itâs branch bodyframe-roll-update2:
https://drive.google.com/drive/folders/1UY_vLZx04QyfLcrxSB7ztUWLwkM2X9kE?usp=sharing
Be careful with this one since the yaw change could cause a crash if it was done wrong. Youâd have to bail to QACRO or an FW mode if things go badly. Seems OK in SITL though.
Thanks for the explanation, not something straightforward.
I also modified the python program to show the change in roll response. The âreboundâ when hovering is gone with this change.
https://www.glowscript.org/#/user/kd0aij/folder/QuadPlane/program/bodyframe-roll-update
@kd0aij, this binary is OK, no rebound and good feeling in qhover mode. 2 flights this morning.
hi lorbass
Could you please share the drawing of the skywalker x6 vector motor mount, the format is stp, if you can, I will be very grateful
I donât know .STP Format.
Here you will find the Files in the Format for the App Blender and
the 3D Print Files as .STL
https://drive.google.com/drive/folders/107l5s97osKd71IBtoJb4g8KYK-DOmwyg?usp=sharing
@losawing, @iampete , Close to the development front I stuck in a Tuning Issue.
As a final test bevor upgrading to the latest Version I tested QLOITER with a shocking acrobatic figure
to see at 0:35 of the following Video.
The Log File ist here:
https://drive.google.com/file/d/1161npXHGBCPLN4WVKdwsJe6iubvAh29t/view?usp=sharing
May be Q_A_RAT_PIT_P is a little âhardâ with 0.2 but with 0.15 I had a similar issue with QHOVER.
With P=0.2 all other operations, even landing in strong wind or pitching 60°, are perfect.
Why in QLOITER it starts shaking? And then, back to QHOVER, the system nearly collapsed with
a lot of error warnings until calming itselves.
I think, QLOITER behavior is important in a emergency case with QRTL.
Coulâd you analize the issue and find the fault in my tuning?
First step should be to update to 4.2. Lots of improvements for tailsitters. If there is still a issue it should just be a matter to turning down the position controller gains.
@lorbass
instability you get look like the problem I reported there
I had this stability issue at least 20 time with plane 4.1 but never never never never with plane 4.0 and previously. The instability always happen right after a lane switch message or a EKF yaw reset message. It happen in any copter mode and affect only the belly sitter. The non vectored is OK with 4.1 and 4.2. I think I had the problem with 4.2beta but I am not sure. Sometime instability does not stop and the only way to escape is to switch to FBWA but when it come right after take off you just crash the plane.
As I was the only one to report this issue I believed it was a hardware problem so I tried to disable internal compass or IMU2 (to avoid the lane switch) without success. At the end I gave up and I fly only the non vectored.
hi,lorbass
Thank you for your reply, maybe it is the question I expressed, I may need the airfoil template of skywalker x6, or the servo mount you designed, in STEP format, because I need to make a second modification to match my servoďźthankyou
@losawing, @iampete, Thanks for your help.
After analizing the Log on the base of your advise, I found both are right.
As to see in the pic, there are two sections. Left the oscillating QLOITER and right the part with the strange behaviour after changing back to QHOVER.
If a master user like losawing give up with the vectored Tailsitter, the issue must be verry serious.
Therefore I also will give up (with 4.1.7) and escape to 4.2 as recommended by Peter.
With controller position do you mean: Q_P_POSXY_P ?
I would half:
Q_P_POSXY_P
Q_P_VELXY_D
Q_P_VELXY_I
Q_P_VELXY_P
They should be keep more or less in proportion.
A Miracle. I loaded FW 4.2.1 and compared all Params with the old ones.
And exacly the mentioned Params are reduced of about half.
The old ones where also default.
Will report the result after a bad wether perdiod.
Thank you for helping.
No shaking QLOITER, because no flying possible with 4.2.1 as vectored dual Motor Tailsitter.
With the same Params as with FW 4.1.7 (execpt Q_P_âŚ) there is no Output (Zero) on Servo5/6.
And when set Trim to 1500 after restart it has changed again to 1100.
Tested with all Arming Checks OK and with no Arming checks.
Arming with left stick possible as usual, Elevons and Tiltunits move but no action on motors.
I couldnât find a description of a new function in the Release Notes.
What did I miss?
Vor Test 17.param (20.3 KB)