Dual-motor tailsitters

Hi,

I would like to take a plane that can do torque rolls and glue a wooden stick or whatever to the end of the fuselage to protect the rudder and elevator. This would allow me to rest the plane on the wooden stick and balance it by holding it at the wingtip. I guess launching should be no big problem, but what about landing in hover ? Does the plane balance itself on the stick until I grab the wingtip and kill the motor - I guess yes with thrust vectoring, but otherwise no ?
Will it be stable enough in real world conditions to hover without ground contact until I grab the wingtip and reduce power ?

You will lose pitch and yaw control as soon as the stick contacts the ground, since that will pin the tail down.
But you obviously knew that… I think if the plane is small enough to hand launch into a hover, then you could probably do the reverse in calm conditions. But I’ve never considered that to be practical for launch or recovery myself. @tilt certainly has the best idea for T&Ls, and non-tilt tailsitters just can’t compete with that.

hi, Tridge,
I am not clear on the setting I to =p on all 3 axes? don’t want to get it wrong
thanks

I got four more flights in on my airframe today, after shifting the C.G. 2 cm back. The results were really positive! Here are the highlights:

  • I was unable to get the airplane into an unrecoverable pitch attitude while in Qstabilize or Qhover mode. No matter how far I pushed the nose over, in either direction, the flight controller was able to get it back to vertical. It did require a couple quick shots of power though, which leads me to believe that Leonard’s patch is going be necessary.

  • Pitching forward in Qstabilize caused me to gain altitude VERY quickly. When I would add full nose-forward input for a second or to in an attempt to get to an unrecoverable attitude, the airplane would gain 20-40 feet as soon as I developed any airspeed at all. Be careful if you try this under trees, as the throttle reduction required to keep from climbing might reduce your control authority enough that you won’t be able to pitch back up.

  • There is a lot of earth-frame roll to earth-frame yaw coupling. It’s still controllable, but every time I rolled the airframe I would end up with a pretty big yaw excursion that would take a while to correct. I think I need a lot more yaw P than the default value, since this airframe type inherently doesn’t have a ton of yaw authority when hovering, but what would be even better is a roll to yaw mix so the PID controller isn’t having to make the entire correction.

  • I was able to take off from the pitched-forward attitude that my airframe has while it’s on the kickstand landing gear. It just requires a pretty large throttle input so the wingtips and kickstand clear the ground before it has a chance to rotate forward.

  • There is definitely less control authority during the descent. I felt pretty comfortable up to about .5 m/s, but any more than that and I think it might cause trouble.

Considering that I hacked some motor mounts into a flying wing that was never designed to hover, and I haven’t done any tuning beyond bumping my MIXING_GAIN up to 1.0, it amazes me how well this thing hovers!

1 Like

great! (and obligatory request for dataflash log files …)

definitely. The defaults are setup for multi-rotors, which do yaw using differential motor torque. That is totally different from yawing with elevons. You will need quite high yaw P gain, and about the same amount of I gain as well. You may also find some FF gain is worthwhile.
If you post a log I can recommend some gains to try. The PIQY log message is the key to tuning yaw. It logs the contribution to the yaw controller from yaw P, I, D and FF.

yes, this is expected. You have less airflow over the control surfaces. That is fixed by the vectored airframe design.

1 Like

yes, the basic rule of thumb is that I scales with P. So set them equal for roll, pitch and yaw. You will find that the default for yaw is a small I, but that is for multi-rotors where yaw is totally different.

did you try any transitions btw? It should work fine I think. Just flick to FBWA and start flying. You will need to tune PTCH2SRV_* and RLL2SRV_* a bit. Maybe read up on the fixed wing autotune code.

I sent you a message with links to the logs; did you see it?[quote=“tridge, post:91, topic:15302”]
You will need quite high yaw P gain, and about the same amount of I gain as well.
[/quote]

That will put the yaw I gain out of limits (according to mission planner). Will the code handle that ok?

I haven’t yet; I want to get the hover controller tuned a little better before I introduce that new variable. I still haven’t tuned the forward flight controller either. If I change to AUTOTUNE from a Q mode, will it handle the transition the same way as FBWA?

@mrjadkowski Keep in mind that high yaw demand will compete with pitch control. You may have to give up some yaw authority in return for greater pitch authority.

yes, those limits are setup for multicopters. You can set any value

yes, that will work fine

no, I didn’t see it. What type of message? email? Gitter? discuss forums?

PM on discuss forums

Hi, Tridge still not clear on witch ajustments to make, the s/w I have been using has seperate rates and pids, could you spell out were to ajust? looks like I and P ranges are not the same? also did yoiu get my logs? 1 more question I installed a pixracer would It be better to install a pixhawk?
thanks

@tilt The pixracer has no “safety” coprocessor and fewer PWM output channels (6 vs 14), but is otherwise equivalent to the pixhawk1. The coprocessor implements “manual” fixed-wing mode by passing AERT RC inputs through directly to the output mixers. In the case of an FMU hang, this can save a fixed-wing aircraft (assuming the human pilot is capable), but is of no value for multicopters and other aircraft which aren’t flyable in manual mode.

1 Like

the resason I am asking about the controller is that it says bad AHRS, bad mag, intermittantly and 2 excursions uncommanded.
thanks

pixracer is a good choice I think. It has 6 outputs, which is the right number for your vectored tailsitter.
It also has better IMUs than the Pixhawk1.

there are separate controls for max angular accelerations, and PIDs, and many other find grained controls. The recommendations I’m making in this forum are just the rough first-cut settings to get it flying. A proper tune changes a lot more parameters.
Some key ones are:

  • Q_A_ACCEL_* for max angular accelerations
  • Q_ANGLE_MAX for max tilt angle
  • Q_PXY_* and Q_VXY_* for horizontal position control
  • Q_A_RAT_*_FILT for rate controller filtering

I’d need to see DF logs to look into that. If you let me know which log to look at and what point in the log I can take a look

no, sorry, where were they again?

I thing I sent them google drive