Dual-motor tailsitters

@ag-sys Congratulations on the foamboard tailsitter. First video of a forward transition too.
Your airframe is definitely cheaper than my Stryker at $23 USD: https://tinyurl.com/kfu9y4w

Are your motor mounts also made of foam board?

@ag-sys Gratulation for your courage to translate in an early flight.
If the elevons are to small for hovering, they are really insufficient for the descent.
Here the solution of a high scool of Zurich. https://www.youtube.com/watch?v=QADvPDWtgFU
The elevons are larger than 1/3 of the wing.Seems to be nescessary.
I think, once the wing flips the FC reacts cracy. Tridge will find the reason.

Sound Bug in Plane 3.7… 3.8…Beta and Master with Pixracer Board. (V4)
After booting the well known sound “Start Up OK” is to hear: http://download.ardupilot.org/downloads/wiki/pixhawk_sound_files/PX4_StartUpOk.wav
and then a double tone similar to “Failed to Arm” : http://download.ardupilot.org/downloads/wiki/pixhawk_sound_files/failedToArm.wav but with longer tones.

But later Arming is possible.
With FW Rover or Copter this last doubletone is not to hear.
Doubtfull if this will have consequences during flight.

welcome, and congratulations on your first flights! Looks like a nice simple build.

I’ve had a bit of a look, and the main thing is the pitch PID tuning in hover. Here is what the PID is doing:


There are two key things in this:

  • when hovering, the CoG or trim is a bit off, so that the plane wants to naturally lean forward. The hover pitch controller quickly saturates its pitch integrator
  • it is oscillating in the pitch PID output, driven by the P gain primarily
    Can you move things around a bit so that the weight is a bit more even? You need more weight on the top side of the wing or less on the button. Otherwise try to angle the motors up a bit more. You should also increase the Q_A_RAT_PIT_IMAX to 0.8 (its currently 0.44) to give it more ability to compensate.
    If you can get the weight right then your pitch will improve a lot. I suspect you should also lower the P gain a bit, but try it first with the weight distribution fix.
    Let us know how it goes!
    Cheers, Tridge

My elevons are mechanically trimmed for QSTABILIZE mode, and I need about 30% up deflection in fixed wing flight. This should be OK since very little elevon travel is needed in FW modes.

Tridge, thank you very much for taking the time to look at my logs, I wish I was half as good at analyzing them as you are!

The weather has been too windy for testing all week but I made it to the field this morning. I implemented all the changes you suggested. I also halved the roll and pitch P and I gains for fixed wing flight. My battery is attached under the wing so I was able to alter the weighting by using a smaller battery as a quick fix. My first flight was just a hover test. This time the craft felt much more controllable, responding as expected as long as I kept my inputs small. Whenever it started to get squirrely, a quick burst of power brought it level again. I was able to land with no issues.

For the second flight, I went with the original larger battery. The craft was still very controllable as long as I took it easy on the inputs. I think changing Q_A_RAT_PIT_IMAX from 0.44 to 0.8 had the greatest effect. Here’s the second flight with transition and landing:
https://youtu.be/Maz__AOTGhI

And the log:
https://drive.google.com/open?id=0B860_VUHDPWnQnEtamFIT2dhdVU

I’m thrilled with how this has been going, it’s a cool piece of code! Next time I’m going to push it a bit harder and check how it handles in mild wind. Does anyone have any advise regarding wind? I tend to land quads and planes into the wind but I feel like I should try to land at 90 degrees to the wind so as to present as small a surface area as possible, correcting my position with roll. I don’t know, we’ll see.

1 Like

and I’m thrilled that it is flying your plane so well! I think you have the most successful dual-motor tailsitter with ArduPilot so far.
When you fly next time, could you try some QHOVER and QLOITER? The QHOVER will test how well it maintains altitude automatically, and the QLOITER will test how well it maintains position too.

I suspect leaning into the wind is the right approach, but I really don’t know. If you can experiment and offer advice for others then that would be great!

Hi, congrats! I find into the wind is most pridictable and controlable, and do transitions upwind
Tridge , I volenteered us for NASA Mars project, see last comment on mvi 0160
Rick

cool! That would be really fun. Thanks!

@tridge,
I’m sorry, I have checked the safety procedure with 3 Controllers, each with Copter stable, Plane
stable, beta and latest. All systems could be switched to a stable Light on with nothing ready, just the controller
powered via USB. (Of course arming is then not possible)
The Pixracer PX4 V4 also but only with Plane stable and Beta of April 4.
But NOT with PX4 V4 latest.
The Light looks like: https://www.youtube.com/watch?v=iF4ZGx5QM3A
And if everything ist OK (no error messages) arming via Throttle is possible and the mots are running.
Bug also confirmed by.

For I had a shocking experiance with ignoring a strange behavior of the safety switch with a copter,
I do not dare to continue with this issue.
Here the result after one week, nerves cutted should help to understand my wariness.

@ag-sys Nice flight, congratulations. The transitions at 1:22 and 2:40 look great to me, and you seemed to have good control on the final descent and landing. I’ll be interested to see how well it hovers in the wind. Thanks for posting the log, too.

What are the current spikes on the transitions?

I’m not sure what the current spikes are. There are also corresponding spikes on the CTUN.ThrOut graph but not on RCIN.C3. The motor / prop combination will pull about 25A each on 3s so 58A seems about right for full throttle spooling up.

The motor mounts are made of 2mm ply wood, same dimensions as the Flite Test Mini Arrow.

I’m also interested to see what happens in the wind. Today it’s 15 - 20 mph so a little too much for my liking. I want to work up to that level gradually. I’ll be trying out QLOITER on my next flight too.

you’re right, and I’ve reproduced the issue here. It seems to be specific to Pixracer. I’ll see if I can find the bug soon.

I’ve done a PR to fix this:
The Light looks like: https://www.youtube.com/watch?v=iF4ZGx5QM3A
it should go into master in the next day I expect

@Tridge,
Thank you verry much, I’m happy!!!

Still waiting for new props.
Otto

Hi Tridge,…is it yaw +tilt ? I am getting servo throw that sometimes point 180 deg in opposit direction to the point that the props will hit the mounts but not consistently? this is on illusion with 180 deg servos I can reprogram the servos for less throw, what paramiter to reduce yaw and tilt on the tilt servos?
thanks

Hi Tridge et al. Could the master be adapted to fly a quad motor tailsitter? Something like https://www.flitetest.com/articles/x-craft-review

yes, it could be, but it would require some significant code changes. I do have that one on my list of vehicles to support, but I doubt I’ll get to it for a while. The key thing it needs is to be able to use AP_Motors when in fixed wing flight for rate control

I discovered that APM Planner2 had most of the KMZ export functionality necessary to display aircraft orientation from DF logs. Current WIP is in https://github.com/ArduPilot/apm_planner/pull/1040

This is what my first attempt at a back transition looked like. (blue line is FBWA, green is QSTABILIZE)

very nice!
I can see this being very useful for analysing logs from users