Thanks everyone for your input,
I move the compass/GPS another 30mm away from magnetic sources as far as I could get it, performed another calibration and it works. Testing later today! I will let you know how I go.
Ok Tried some testing this arvo with no luck. Takeoff pitch forward breaks prop, takeoff pitch rearwards breaks prop and so on.I did give it plenty of throttle for takeoff.
What I have. Tailsitter 2 motors (T motors 2814-770kv running 13x5 props) 2 elevons with pixhawk, looking from the rear the right motor spins clockwise left anticlockwise.
CG seems ok.
If I tilt right side down in STABILIZE MODE right motor spins up quicker to counteract. Tilt left side down left motor spins up quicker.
Is this correct? Q_FRAME_TYPE 0
left throttle on channel 3 and right throttle on channel 4
SERVO3_FUNCTION=73
SERVO4_FUNCTION=74
I Did have problems getting my DX8 setup the only way I could get surfaces to move in the correct direction was to use -WING:DAUL AIL- I could not get -WING:ELEVON- to work
Tailsitter.param (16.1 KB)
Hi Andrew, can you do the same check as you did for motor, but for elevons ? For instance, tilt forward, what is the eleven reaction ? ( should be both up).
What is the total weight ? What is the max thrust of one motor with your prop ? (use a scale)
To me the props are looking big for those small motors, what battery are you using ?
Ps:any pictures, video will help 
I donāt understand what you tried to do with your DX8. But do not use your Tx to reverse a channel or to make elevon mixing.
You first have to understand if you have a configuration problem (reversed elevon, bad parameterā¦) or a PID tuning issue.
Some general advise:
Breaking propeller is very normal. if you succeed flying your tail sitter with less than a dozen of broken propeller you are better than me.
When switched to FBWA or Q_hover, elevons must react the right direction when you tilt the wing.
you must have a GPS 3D fix or the EKF does not work. Read post 65 from tridge.
You have to tune you PID and increase P terms as much as possible. Mixing_gain is also very important, increase it up to 1,2. In the post 832 I describe the method I used with success.
Without vector thrust, the max pitch angle the wing can recover from is very limited so I suggest to reduce Q_angle _Max to 20.
Q_Frame_type is not used for tail sitter
Read post 281 from Kd0aij to understand servo group and pwm output.
Hi everyone,
I just wanted to apologise for not answering posts the last few weeks. Iāve been deep in the ChibiOS port of ArduPilot, plus preparing for a couple of conferences.
I hope to start doing some catch up on this forum soon.
Cheers, Tridge
2 Likes
Hi @tridge where can we read more about this port or what it might mean for Ardupilot? Thanks.
Thanks @tilt, I shouldāve phrased myself better.
Wanted to know what it means when Ardupilot is ported to this OS as Iām not aware of boards we can use that are using this OS.
Hope this clears my question
looks like f3cc spracer might do all that pix can? but just speculating
Hi to all!
I am trying to configure a tailsitter with a Pixhawk px4.
I am using, as SERVO_FUNCTION for motors, values 73 and 74
And for elevons, values 77 and 78.
I am having troubles, when I try to use the mode QHOVER o QSTABILICE, when I move the sticker of my rc, like a 3 - 5 %, in the elevons, it moves all the way through (100% of the movement allowed for the elevon)
But when i put any wing mode (like FBWA, Stabilice o Manual) it moves correctly, I mean, for a difference of 40% in the RC, the elevon moves a 40%.
I have read that some of you are using SERVO_FUNCTION values of 4 and 19. Could this solve the problem?
Or it could be a problem of not configuring correctly the value of Q_A_RAT_PIT_P? Now it has the default value.
We tried to fly it one time, and in Q mode it is very unstable (we wouldnāt be able to change to Wing mode and later we saw the problem with the difference between the RC and the movement of the elevons.
(sorry about my English, itās not my native language)
I am designing a dual-motor tailsitter configuration in which each rotor has full collective and cyclic pitch. Currently I am planning to use a separate mixer for the CCPM but it would be great to eliminate these and feed the cyclic servos directly from the Pixhawk. Is there any interest in adding this to ArduPilot? If so I may have time and/or budget to look at it but would like some advice on how much work it would require.
Collective pitch is a nice add-on and helps with having efficient propulsion setup both in hover and FF. At least if you can compensate for the probably less efficient props you have to use in this size.
But for what you need cyclic pitch? (Aint this a Chinook then? 
)
Also you need 3 more servos for each prop and add super much complexity. Though eliminating the main advantage of the Tailsitter configuration.
All this is totally normal. Please read the full thread and/or use search.
You need to tune PID, I described the way to go already a while ago.
You have more infos/pics about your design? How you did the fuselage?
The cyclic is an alternative to vectoring the motors, which we know helps a lot with pitch control. I donāt think vectoring scales up very well to larger models though, and cyclic will probably give more authority. I agree it is more complicated.
Hi everybody,
Just want to share the flight test result of Arduplane 3.8.4 with conventional dual motor tailsitter (no vectoring). Full auto mode works properly now especially resolving forward transition bug from the previous firmware. However, there is still an issue during landing. itās quite tricky to land without stumbling in auto mode because copter code wonāt disarmed immediately after touching the ground causing ground effect. So, sometime, we switch to Qhover/QLoiter instead during final descending. As we tried to land with QHover/QLoiter we have learned that before the plane touch the ground, if we cut throttle completely, the plane would land perfectly. Sill not sure, will try to test more about auto land.
1 Like
Your idea is very interesting.
You need 10 servos output (2 for ESC, 2 for elevons and 6 (or 4) for collective and cyclic pitch) which is fine because the pixhawk has 12 output.
But to control cppm servo directly from the pixhawk you need to mix at least input from pitch and throttle which is to my knowledge not possible with the pixhawk loaded with arduplane. The solution could be to use a flybarless head and connect pixhawk output to a flybarless unit like Kbar. At banggood the tarot 450 size flybarless head is only 15 ā¬ and the kbar is 25 ā¬. Kbar is also a 3 axis gyro and I donāt know how it will behave specially during transition. I am not sure it will work, you have to test. Also to manage the roll stability, do you plan to use throttle or blade pitch ? Not easy but interestingā¦
Be aware that you are going to break a lot. When a standard tail sitter fall on its belly, the worst case is a broken propeller but with a heli rotorā¦
regards,
Pierre
Looking good from ground.
But the log shows all the axis are oscilating. How did you do the tuning?
What airfoil do you use? Symmetrical?
Why you think so?
Agree, makes sense. Also the authority wouldnāt be thrust dependent which is known to cause stability problems while descending.
Maybe you can put the collective on the flap channel and adjust the three possible settings to three flight stages like hover (low pitch), acceleration (medium pitch) and fast level flight (high pitch).
The cyclic can be hooked up as an pitch channel when you use only one servo for it. You might not need more then two servos, when you donāt use cyclic roll. (which I can see no reason to use cyclic roll in any case. You wonāt be able to copter yaw with differential cyclic roll, only to roll and for this you better use collective differential pitch)
So you could use it with stock code on the sum of 8 CH. (Except for differential pitch when you hook it up via flap function)
getting interesting 
The moment of inertia of a rotor goes up rapidly with scale. Even on RC helicopters the rotor disk has a lot of inertia, and if you try to vector it mechanically the rotor generally stays still and the rest of the heli moves! So vectoring will be less effective.
Good point on the stability at low throttle, I had not considered that. I like the flap channel idea.
I agree the roll cyclic is not that useful. However it would require an electric or mechanical mixer to avoid controlling all 3 servos from Pixhawk, so you almost get it for āfreeā.
Today I have been retuning my tailsitter running 3.8.4) in the rig via Ziegler/Nichols method, because the parameters I had didnāt work so nicely anymore after the added PID scaling with throttle in 3.8.3.
Again I found that it is very difficult to get the pitch axis tuned because of the pitch induction when the rudder moves. There is a high frequency oscillation first due to rudder, then only with higher P-term you get to the actual oscillation your looking for with this method. Roll and Yaw are fine and very easy to tune.
To fix this I thought that I might change the Q_A_RAT_PIT_FILT of the stock 10 Hz to something else. I tried 1 Hz and alo up to 100 Hz and this makes absolutely no difference at all in any case. (Is this parameter actually working?)
I also tried changing the P, I and D-term. Found that the only thing that helps is when I reduce P-term to unusable low values like 0.10 and less instead of my calculated 0.45
Everyone should run into this problem with tailsitters, exept if your rudders have no mass 
So what to do???
Also: Anyone here fully understand how the FF term works? Can I reduce P and compensate this with FF or somehow else?