Servers by jDrones

Dual-motor tailsitters

(Peter Hall) #1242

I am working on a more permanent fix, but in the mean time increasing Q_TAILSIT_VHGAIN and decreasing your Q_A_rat… gains proportionally may work. This has the effect of reducing the control surface gains while keeping the vector gains the same.

Have made progress on the code, got fed up with constantly re-flashing to test it out. I have purchased real-flight and will do testing using SITL, have been messing about abit the physics seem quite good.

(Claus) #1243

Hi, I am not able to Upload them …what do I do wrong ?

Gesendet mit der WEB.DE iPad App

Tried to zip it …seemto be too large :frowning: got an error …

(Claus) #1244

Today went out with the laptop and tried to adjust the q_a_rate parameter set…
Had my first hover and transitions ! worked pretty good despite the windy conditions…I had still some wobbling effects during landing … Back and forth transition worked good…
How are the axes defined during the tailsitter mode ?

the front is pointing upwards - so I guess pitch is leaning against the wind angle ?
then roll would be the horizontal angle meaning motor acceleration and thrust.
yaw would then be the rotation around the vertical axis …are these assumtions correct ?

@losawing :thanx for your info - checked the tilting constructionsand found 2 issues :slightly_smiling_face:
@lorbass: used your file to check my parameter set

here is mine in case someone is interested:
vtol_field3.param (16.2 KB)

(letpi) #1245

hi Claus,
Most people here use dropbox to share large files.
That’s a good news you were able to fly your wing and make transitions. Some oscillations (wobbling?)are not a problem if they are not too strong and only temporary. When the wing get some speed in VTOL modes oscillations can be very bad. This is because there is more airflow on control surfaces and because all gains are increased when throttle decrease. @iampete pushed a fix for that and the corresponding firmware is post 1034.
Your definitions of control axis are correct but only when the wing is pointing upward. The wiki definition is: the yaw stick controlling earth-frame yaw, and roll stick controls earth-frame roll.
You may find the aircraft has a weird behavior when you want to roll it since the lean angle is higher than 45°. I guess this is a consequence of the earth frame.
No problem with your parameters. Maybe q_a_rat_yaw_P and I are low. It worked for me to increase them a lot in order to withstand wind gust. Nevertheless, as @lorbass said, you will find It takes more than 50 flights to get all q_ parameters tuned.

( research air) #1246

check this out …

(lorbass) #1247

I just flashed V3.8.5, but no such Params to finde.

3.8.5 Vor Gestell.param (16.1 KB)


(Peter Hall) #1248

Not been merged yet, Tridge is busy with Chibios and OSD stuff I think

(Claus) #1249

Hi folks ,

I finally managed to spend some time on the field with the laptopto tune my Caipirinha2.
It seems the I tuned the params in the wrong direktion. The swining I saw on the axes was due to too low
parameter !! Now it feels pretty save when I do the transitions.
Q-stab and to transition to FBWA works fine in both directions.
Now I have probs with the q-hover …when I switch to this mode my transimitter input is delayed for about 2-3 sec . And it is hard to hover at a decent hight.
I shortly tried to q-loiter - but the flyer was acceleraton and drifting away from the wind - although I had a pretty good gps lock.
Perhaps someone has experience with a similar behaviour.
For now here is my actual Parameter file …

Greez ,

Clausvtol_field7lowwindok.param (15.3 KB)

(lorbass) #1250

My caipirinha can move in hover mode max. 10 kmh. After that it starts pumping.
So if the wind is stronger, it can’t hold the position even in Q_Loiter.
Therefore I never use Q_Loiter.
In Q_Hover I turn the wing to get it back with Roll when the wind is stronger.

Here an Idea for a planned improvement in poste 1219:

(Claus) #1251

Ah okay… :slight_smile:
by the way - how did you realize the 180° tilting of the rotors ?

(lorbass) #1252

I use this servos:

They can be programmed with the programmer DPC-11 for 178 ° on the PC.
Or the dealer can deliver it with 178 °.

(Mark Whitehorn) #1253

Interesting non-vectored VTOL design:

(letpi) #1254

So clever !! I don’t know if this is a vision of future but I like it

(Marek) #1255

New tailsitter design:


canard style with dual motor, but coaxial with counter-prop

2 x BE4114-400kV with 16x5.4 props
2 x ESC 40A
LiPo 6S
Pixhawk with GPS, Current/Voltage sensor, telemetry 433 MHz
Arduplane V.3.8.5
Weight - 2.7 kg

Now - tuning PID’s in QSTABILIZE.
with pitch - ok (P=0.3, I=0.1, D=0.001),
but with roll -> small P values (0.08 - 0.3) not work = great oscillations. Starting to be stable with P=0.8 - 1.2.
Any ideas with this high P values?
Parameters file -> EM-1.param (17.0 KB)

(Mark Whitehorn) #1256

How do you control body-frame yaw? Are the elevons the only control surfaces?

(Marek) #1257

Yes. Only elevons for yaw

(letpi) #1258

With a traditional ardupilot dual motor tailsitter, in all VTOL modes, roll control is given by differential motor thrust. I dont see a rudder too. I guess it means roll control is obtained with vertical canard surfaces ? That would be very new.
your parameters are OK for a first test except q_a_rat_yaw_p=0 and q_a_rat_pit_d=0

(Marek) #1259

Two canards -> one is elevator (pitch), second rudder (roll) plus elevons or flaperons (yaw).
There are also possibility to use difference in RPM of both motors as additional yaw ( similar to coaxial heli).
From first tests - elevons are very eficient in yaw.
For me - the best idea is to use this configuration in „vertical” modes than during transition switch off elevator and rudder and use flaperons as in normal delta plane. But have no idea how to change settings of both canards to fixed position in neutral during transition.

(Peter Hall) #1260

This has now been merged, I removed the second parameter because it turned out not to be useful, should be in todays daily build.

I have been messing about tiring to use real flight for code development, not having a huge amount of luck. First task is to make a real flight copy of the bronco, I have had it flying OK but it seems to do random stuff occasionally. Not sure if real flight is really good enough to evaluate the performance of hovering code.

Anyone got any tips?, what is the best thrust vectored model to to use? So far I started using the CUAV model as it was closest to my bronco and resized the components, its does sort of fly most of the time on the real parameters.


(Marek) #1261