Thanks for the link
Of course the heart will be Arduplane firmware.
Cheers!
I have installed 4.1 DEV from Mission Planner (Ctrl + Q) for testing but I notice strange behavior loading parameters when connecting to Mission >>> it only takes 0.2 sec to load parameters
This is very strangeā¦,
I think thereās PX4 inside, i see QGC in the final part.
Does anyone know if you can fly a non-vectored tailsitter in Qstabilize and FBWA without a GPS and without a compass?
thanks,
Shef
@iampete @kd0aij
I begun to test the disk theory gain scaling on a non vectored dual motor tailsitter with max angle set to 80Ā° and airspeed enable. The calculation gave me a value of 10 for the disk load parameter. I had to lower this value to 2 to get rid of oscillations. Low speed stability, medium lean angles or transitions are not affected so the overall result is very good.
About the plane: wingspan is 86cm, 1Kg, 25cm propeller, larges control surfaces, CG set to 4% static margin. This may be not obvious on the video but the recovery from medium lean angle (lets say 45Ā°) require almost full elevon deflection which is very bad and I had expected a better result with this frame. Set the static margin to 0% improve things a lot but is not flyable in plane mode. I will try to implement a variable CG in the future.
This test was made with a freshly downloaded arduplane 4.1 firmware and I found a problem with tailsit_input, it seems to be broken, only copter style input is still working and the body frame roll is weird.
5 Likes
Congratulation. I have never seen a non vectored with such a lean performance in Hover Mode.
This is crazy! Looks better than wingtra! 
Thanks for pointing this out. Iāll try to find the cause of that problem.
I had some hopes this plane would hover like Wingtra but it is very far. In the range 30 to 60Ā° lean angle the pitch balance need almost full elevon deflection so no room left to stabilize body frame roll. Lets say the next one will be betterā¦
Thanks
Wow ā¦Pierre, you are the king of DMTS. 
@IamPete commented about Q_TAILSIT_DISKLD, better results may be found a little higher or lower than calculated. but change it from 10 to 2 is a big change for it to work OKā¦ 
Thanks, but this is a small kingdom.
I donāt know but the important point is oscillation reduction at high lean angle while keeping same stability.
Found the problem; PR here: https://github.com/ArduPilot/ardupilot/pull/16841
builds for sitl and fmuv3 here:
https://drive.google.com/drive/folders/1Xddn0ywvpFXy81XHB-H56HCfFDiS7GcN?usp=sharing
2 Likes
@kd0aij I tested the fix this evening, plane type input (set to 3) is OK and body frame roll allows to fly like a plane at high lean angle, thanks, the result is truly amazing.
@iampete, I made a short q_acro test and thanks to disk theory gain scaling I did not observe a single oscillation whatever speed or attitude, really impressive resultā¦considering very larges flaps with 45Ā° throw. The flight is much more pleasant in q_acro than it is in FBWA.
2 Likes
Just a short clip of my Albabird TVBS conversion using two fpv cameras with scripting auto-switch when changing flight modes from fixed wing to VTOLā¦the time delay is set at 2 seconds, but needs to be reduced a bit for this plane, it transitions quicklyā¦havent started tuning it yet so a bit wobbly on initial move to VTOL stanceā¦although I have 5 TVBS planes, every new one is a different tuning challengeā¦our defaults are pretty poor for tailsitters in many casesā¦unless its a copter tailsitterā¦we really need to put together a tailsitter tuning guideā¦one for vectored and one for nonā¦
I tear my hair out tuning, and on tailsitters its actually dangerous compared to any other horizontal stance quadplaneā¦I almost always damage a tailsitter model at least once trying to tune itā¦
4 Likes
@losawing Thanks for testing.
Do you have an airspeed sensor on this plane?
We should consider eliminating my gain/speed scaling option (Q_TAILSIT_GSCMSK bit 1) if Peteās new disk theory method is better in all respects.
Yes, airspeed sensor enable.
The att-throttle gain scaling gave very good results on the biwing and as far I remember there were some manageable oscillations only at low throttle and high speed (descending part of a loop for example). I had also some failed back transition on my non vectored batwing. I never tested it on a TVBS.
The disk theory is easy to tune, very effective to eliminate oscillations and seems to not reduce gain at low speed so yes it seems to be better.
1 Like
I would be glad to help but this is not easy : A Tailsitter (vectored or not) is difficult because it is a plane and as a plane it must be balanced and stable. Balanced means that the sum of moments must be 0 with the lowest control surfaces deflection (and from 0 to 90Ā° lean angle) and stable mean CG in front of neutral point. To my mind parameter tuning is only a second step. The most important one is to get a balanced frame. We should also think what āāflyingāā means when prop hanged.
prop hanging tilt vectored VTOLs are usually pendulum stable in pitch and rollā¦the motors are always well ahead of CG (assuming slow vertical velocity up or down)ā¦I almost never have an unstable hover in pitch or rollā¦yaw is always a monster for me to get stableā¦it has no inherent stability in that axisā¦
they should be a way to easily inspect the PIQx reports and desire/actual axis angle and determine at least the direction one or more of the PIDs should be movedā¦I just dont have the insight,I can look at tar/act PID and tell if the actual is over or under, but I really cant tell from the P/I/D what needs whatā¦
for exampleā¦that flight aboveā¦yaw was horribleā¦
the target pid and actual are way offā¦but I havenāt a clue as to what to change
btwā¦the horrible roll on takeoff was a glitch in my T16sā¦it gave full roll when I did notā¦a cable inside is loose I thinkā¦have a request for a warranty repair in to Grayson, but they never respond
I agree, once you manage to give the pitch axis enough authority, yaw is always the problem. But are you able to say if it is a PID tuning issue or a frame limitation ? Do not underestimate the role of control surfaces, they are more effective than thrust vectored.
Competition between pitch and yaw is also a problem. If pitch balance already need full control surface deflection yaw cannot be stable (and vis versa).
I can not help with log analysis, I am quite bad at this game.
I fly normal planes converted to tvbsā¦control surfaces have minimal impact, they are very small areasā¦definitely not the rudder or elevatorā¦way too far from prop wash and too tinyā¦ailerons perhaps a tiny bit for body frame roll, but the vectored thrust is much more dominant I believeā¦these could never fly as non-vectoredā¦control surfaces way too smallā¦