the FW and BW works fine. The FW is very smooth, the BW a little ugly but reliable even in wind. In auto the FW transition does not work, I think there is a bug. the BK transition is fine. Without vector thrust, the stability is at the limit and in wind this is almost impossible to keep stability. Probably a clever design of the wing would better that but this is an other subject. You need to tune the PID, Autotune in fixed wing is fine. For the copter I would advise to held the wing nose up in qstbilize mode. Begin to set mixing_gain to 1.0 or 1.2. To set PID (Q_A_RAT_X) increase one D term until you get fast oscillation then reduce D. Increase the corresponding P until you get oscillation (slower ones), then reduce. Make I=P. You can play also with Q_AZ_P and Q_AZ_D to improve Qhover and transitions. Same method, increase D until oscillation then P. Did you set EKF to 3 ? Q_M_THST should be correctly tuned. My wing is flyable in manual mode, I have set a panic switch with my Tx that replace whatever mode by manual mode, so when something goes wrong I always be able to take control and land manually. I hope it helps
Regards
Pierre
Yep, I plan to tune the airframe. Thanks for the info. This flight was just to see how code can handle the auto transition as manual transition was already proof. My idea is that the FW transition should start with fw motion in copter mode to gain a certain speed before switching the airplane mode. As I also plan to scale up the airframe and I suspect that for larger airframe the transition might not work correctly. Except the airframe has enough thrust to weight ratio (which I don’t know the exact value).
No, I use EKF 2 (Default).
for EKF, read post 65 from tridge.
I saw looking at my log that the logic of the FW transition in auto is not the same that the one from qhover to FBWA. In Auto, there is no pitch order and the transition to fixed wing is ordered while the wing has no speed and is vertical. But if you read the post 825 by tridge you will see that he corrected a bug after the release of the 3.8.3 code, so I loaded the code again and will make a test tomorrow.
Vectoring the thrust is a huge improvement,
Just to clarify: tune PID holding the wing is not optimal, you can begin with but to complete, start again the same process flying in Qhover.
Ok. I will try to change EKF too. I don’t use thrust vectoring because I want it as simplest as possible. Maybe tuning for performance will make a transition smoother. Waiting to see your flight test result.
Does someone know if SITL already work for Tailsitter or not ?. I tried before and It can’t takeoff from ground due to (I think) wrong direction of nose
Cheers,
RW
Hi All,
Sorry if this question was answered before. Please point me to right direction…
Do you think regular plane shape (ex. Talon with modified dual motor https://www.youtube.com/watch?v=LOtqKuu9bew) can be used for Tailsitter airframe?
Thanks in advance
Here all about:
Regards, Otto
I don’t think so.
But if you want to test you need at least to modify the ailerons to get them full span and wider and also to make special motor mount to allow vector thrust. In copter mode, the problem is the stability with the pitch axis. My experience with a flying wing is that you can fly safely in copter mode in 20km/h of wind if you have the vector thrust and almost no wind without.
To make the talon a VTOL I would prefer the tricopter option with tilt motors.
Regards,
Pierre
a video of a normal flight: Q_hover to FBWA to Q_hover
Everything is fine
Takeoff in auto mode. Once the altitude of the transition to FW is reach, edge flight on the right side. Almost crash, manual mode to take control. I tried several time, always the same. I have played with parameters… the same. Tried with a fresh 3.8.3 code… the same. If I fly auto without the takeoff it is fine.
ESC are carefully calibrated, motor thrust is the same on the left and right side. I don’t understand what happen.
Two strange things:
During the climb I had control only over the roll (earth frame) but the roll was controlled by the yaw stick.
After aborting auto, I flew manual then FBWA and Q_hover. When switching to Q_hover, it just shut the motors.
pierre
I have realflight 8 now. Must say that the physics are totally off from reality. 
Also bought X-Plane 11 and this is awesome. Now I need to learn how to build models for this. SITL is very easy with X-Plane and you can even use the free Demo version for this.
@tridge Can you please release a beta version with the knife edge fix?
My name is Karl-Emil, I’m 19 years old and come from Denmark. I am currently building a tailsitter drone using the Pixhawk, but I’m constantly running into trouble… 
I have read most of this thread but I am still confused. I will now try to explain every variable about my aircraft and I hope that you will help me to set up the controller - you seem like the experts to ask.
My aircraft
Picture of my tailsitter
It is supposed to be a dualrotor tailsitter with elevons and thrust vectoring. From what I understand this kind of aircraft is possible with arduplane.
It is using Pixhawk controller from eBay with M8N GPS/Compass
It is using old 2-4s esc’s but I cant remember what kind exactly or if they run Oneshot etc. I have better ESC’s that I can mount if it is necessary
It uses two analog 9g servos for the thrust vectoring.
I think I have the physical build pretty well figured out. I have build lots of drones over the years and recently I got a 3d printer.
The hard part for me is the software - how do I set up the controller. I have used the APM 2.6 with autonomous drones, but the pixhawk is giving me a few problems. It seems that dualrotor tailsitters with thrustvectoring are quite rare which also makes the documentation pretty small.
I have read that you have to do the following:
- Q_ENABLE = 1
- Q_FRAME_CLASS = 10
- Q_MAV_TYPE = 21
Questions:
- What does Q_ENABLE = 2 (Enable VTOL auto)
- What should I set for the Q_FRAME_TYPE? (I would think 2:V beacuse im using a flying wing design)
- What shoild I set for the Q_TILT_TYPE? (I would think 2:Vectored yaw)
I am not very experinced with ardupilot so if you can outline all the steps pretty clearly then that would be a huge help to me!
Is this aircraft possible to make or should I try a non-vectored design instead?
Note: The goal is to make an autonomous delivery drone. I have made a YouTube video about this plane that I am building.
Simiilar to my vectored Tailsitter. I also had a lot of troubles.
Here my Params:
Vor wechsel auf 3 8 3 beta3.param (16.1 KB)
And here the last Video
Test 47 https://youtu.be/nIXRRIeIZe0 Neue Transition, Akku 12mm nach hinten.
For me, the servos are to small. Are they for 180 degrees?
And here the setup:
ARMING_RUDDER = 2
Q_A_ANGEL_BOOST =0
BATT_MONITOR=4
BATT_AMP_PERVOLT = 41.9
BATT_VOLT_MULT= 13.7
SERVO1_FUNCTION=77 Left Elevon Output 1 neu, nach kd0aij
SERVO2_FUNCTION=78 Right Elevon Output 2 neu, nach kd0aij
SERVO3_FUNCTION=75 Left Tilt Servo,Output 3 , nach Tridge
SERVO3_MAX = 2070 90° up
SERBO3_MIN = 900 90 °down
SERVO4_FUNCTION=76 Right Tilt Servo,Output 4 , nach Tridge
SERVO4_MAX = 2100 90° up
SERVO4_MIN = 900 90 °down
SERVO5_FUNCTION=73 Left Motor Output 5
SERVO6_FUNCTION=74 Right Motor Output 6
MIXING_GAIN = 0.5 0= ohne Elevons, >0 gemischt.
MIXING_OFFSET =0 (-1000, 0 , +1000 %)
KFF_RDDRMIX = 0 default 0.1 else crash (Video)
AHRS_EKF_TYPE=3
EK2_ENABLE=0
EK3_ENABLE=1
Q_ENABLE=1
Q_FRAME_CLASS=10
Q_ASSIST_SPEED=0
Good luck
Thank you for the fast response. I still have a few questions though.
What do I need to set the following parameters to:
Q_FRAME_TYPE
Q_TILT_TYPE
And also, what is the difference between Q_ENABLE=1 and Q_ENABLE=2?
Karl-Emil
Hi Karl-Emil, when your in Mission Planner there is a compare function to compare your set of parameters with the set Otto send you.
Im not very sure, but I think Q_FRAME_TYPE isnt used for tailsitters
Also here you can find an explanation for every parameter: http://ardupilot.org/plane/docs/parameters.html
I think your motors are too close to each other. The Aileron does not cover the whole propwash. Also consider buying Digital MG servos. I use this ones: https://hobbyking.com/en_us/hextronik-mg-14-14g-2-6kg-0-11sec-digital-aircraft-servo.html
Yes, David is right.
[quote=“palm369, post:851, topic:15302”]
I think your motors are too close to each other.
[/quote] looks ok to me, TVTS right?
I will try Otto’s settings. I have just bought the servos you linked.
Can anybody tell me what Q_FRAME_TYPE = 2:V looks like - wing design and/or motor layout? I have searched all over the internet, but I can’t find the answer.
Q_FRAME_TYPE is the motor Layout you use for the VTOL motors. Like theses here: http://ardupilot.org/copter/docs/connect-escs-and-motors.html#motor-order-diagrams
AFAIK this parameter is irrelevant for Tailsitters.
It seems like this software has been built in a way that you need some “inside knowledge” to know all the details. I would think that every parameter that is showing would be relevant for the aircraft. Where can I read what parameters that matter for a dualrotor tailsitter with vectored yaw?
If you read all the links, posted by Tridge in post 1, you are also an insider.
Yes you are right, the infos are spread all over the thread.
And parameter which are not or no longer used, parameters which are not explained, and parameters for developers only. It is a Research project not a kit for ready to fly. But you will learn a lot by doing as I did. 
See here: https://www.youtube.com/channel/UCeY_PBnw4JTeXcyFIekcLtw/videos?view_as=subscriber