Building, flying and not crashing a large OctaQuadPlane

After the unfortunate crash of our last QuadPlane build with a failed ESC CanberraUAV wanted to build a new version with more redundancy in the VTOL part of the build. The result is the above OctaQuadPlane which we successfully flew for the first time yesterday.

When we were first designing a large QuadPlane we did consider an octa design, but rejected it due to what we thought would be a high degree of complexity and unnecessary weight. Needing wiring for both 8 vertical lift motors plus 4 controls for fixed wing flight, and extra controls for ignition cut and auxiliary functions like remote engine start, choke and payload control we worked out we'd need 15 PWM outputs. The Pixhawk only has 14 outputs.

After the failed ESC on the previous plane lost us the aircraft we looked again at an octa design and found that it would not only be possible, but could potentially be simpler for wiring than our last aircraft and be lighter as well, while having more lift.

To start with we looked for motors with a better power to weight ratio than the NTM Prop Drive 50-60 motors we used on the last build. We found them in the t-motor 3520-11 400kV motors. These very well regarded motors have a considerably better power to weight ratio, and back-to-back mounting of them was extremely simple and light with the above clamping arrangement.

To manage the complexity of the wiring we added support in ArduPilot for high speed SBUS output, and used one of these SBUS to PWM adapters embedded in each wing:

http://www.hobbyking.com/hobbyking/store/__24482__SBD4_4_Channel_S_...

that allowed us to have a single wire from a Y-lead on the SBUS output port of the Pixhawk going to each wing. We modified the BRD_SBUS_OUT parameter in ArduPilot to allow setting of the SBUS frame rate, with up to 300Hz SBUS output. For a large QuadPlane 300Hz is plenty for multi-rotor control.

The arm mounting system we used was the same as for the previous plane, with two 20x20x800 CF square section tubes per wing, mounted on a 300x100x1 CF flat plate. The flat plate is glued to the wing with silicon sealant, and the CF tubes are glued to that plate with epoxy.

For ESCs we used the HobbyWing 40A, which has a burst rating of 60A. That is well above our expected hover current of 15A per motor. Combined with the redundancy of the OctaQuad and the better reputation of HobbyWing ESCs we were confident we wouldn't have a repeat of our previous crash.

For batteries we switched to 6S, using one 5Ah battery per wing. That gives us over 4 minutes of hover flying time while keeping the weight well below our last build (thanks largely to the lighter motors and ESCs).

For this initial test flight we had the fuel tank mounted horizontally unlike the previous vertical arrangement. This was OK as we only filled it a small amount for these test flights. We will be converting to a vertical arrangement again for future fights to reduce the impact of fuel slosh causing CoG oscillations. We may also fill it with fuel anti-slosh foam as we have done on some other aircraft (particularly the helicopters).

Overall the build came out about 1.5kg lighter than the previous build, coming in at 12.5kg dry weight. With a full load of fuel we'd expect to be about 13.5kg.

We did three test flights yesterday. The first was just a quick hover test to confirm everything was working as expected. After that we did the first transition test under manual control, which worked very nicely.

The copter part of the tuning could definitely do with some work, but we thought it was stable enough to do a full auto mission.

It was a short mission, with just two full circuits before landing, but it nicely demonstrated autonomous VTOL takeoff, transition to fixed wing flight, transition back to hover and auto landing. The landing came in within a meter of the desired landing point.

We had set the distance between the transition point and landing point a bit short, and we hadn't included a mission item for the plane to slow down before it transitioned which led to a more abrupt transition than is really good for the airframe. Going from 100km/hr to zero over a distance of 77 meters really puts a lot of stress on the wings. We'll fix that for future missions. It is nice to know it can handle it though.

The transition to hover also caused it to climb a fair bit which meant it spent more time in VTOL landing than we would have liked, chewing through the battery. That was caused by it having to pitch up to slow down enough to achieve the stopping point in the mission which caused it to climb as it still had a lot of lift from the wings, combined with a high angle of attack. It didn't help that we had a slow slew rate on the petrol motor, reducing the motor from full throttle to zero over a 3 second period. We chose a slow slew rate to reduce the chance of the engine cutting due to fast throttle changes. We can fix that with a bit of engine tuning and a longer transition distance - probably 130 meters would work better for this aircraft.

The climb on transition also took the aircraft into the sun from the pilots point of view, which isn't ideal but did result in a quite picturesque video of the plane silhouetted against the sun.

The full flight log is available here if anyone wants to see it. The new features (new SBUS output support and OctaQuad support for quadplanes) will be in the 3.5.1 plane release.

We're not done yet with QuadPlanes. Jack is building another one based on the Valiant, which you can see here next to the Porter:

Combined with the GX9 tradheli build that Greg has put together:

we are all set for lots of VTOL fun!

Many thanks to everyone who helped with the build and as always to CMAC for providing a great flying field!

Views: 7741

Comment by DBX Drones on March 14, 2016 at 1:40am

Good job Tridge!
I am so glad you made that beast to fly!


Developer
Comment by Andrew Tridgell on March 14, 2016 at 1:49am

@Hughes, I'm not good at drawings, but basically we have:

  SBUS output port on pixhawk -> servo-y-lead -> SBD SBUS to 4-way PWM output adapter for each wing

then you use the SBD programmer to assign channels to each of the outputs.

We chose to connect the motor ESCs on each wing as:

 1) top-front

 2) bottom-front

 3) top-rear

 4) bottom-rear

so we programmed each SBUS adapter to produce the right assignment for an OctaQuad.

The CF rectangular arms are from hobbyking

http://www.hobbyking.com/hobbyking/store/__80920__Carbon_Fibre_Squa...

Comment by JB on March 14, 2016 at 3:27am

Thx Tridge.

I didn't realise 35m/s is possible with that airframe and so you still had a significant margin. I'm glad you still get nervous on your flights...and I'm not the only one! ;-)

With the wind at the OBC; although I can agree to plan for the worst, hope for the best, do you really think we'll be flying in 25knots at that time of year in that location? Just because they allow the flights doesn't mean it's going to be that windy. The law of averages, based on weather history for Dalby, would predict a much lower windspeed which needs to be overcome? 25knots is a considerable wind...and that's coming from a Sandgroper who's by nature familiar with wind! ;-)


Developer
Comment by Andrew Tridgell on March 14, 2016 at 4:36am

@JB, our competition flights in both 2012 and 2014 were in over 20knot winds (although average may have been a bit below that). Remember that the 25knot limit for halting the competition is the wind measured on the ground. The wind that matters for the mission time is the wind at the flying altitude.

Even worse if the wind changes at an inconvenient point in the mission! It could effectively be even higher if it is into the wind both ways.

How is your D2 prep going? Any photos or videos to show off?

Cheers, Tridge

Comment by Steve Mitchell on March 14, 2016 at 4:45am

Awesome stuff Tridge!

I now want to dust off my 1/3 scale Cub that's hanging in the shed and try it out with a sling load:)

Comment by JB on March 14, 2016 at 6:04am

Yeah we were a bit luckier with the wind then you were in 2014.  I forgot that. You had to fly after a thunderstorm and it even looked like they were going to cancel the rest of the event. Glad they didn't! :-)

In saying that wind in flight will likely be less tricky then when in hover with a quadplane though. Hopefully the wind will be better this time.

Mad rush for the D2 as usual and we still need to make our videos. Flat out at work too which doesn't help. Would be nice to be able to do it full time! D2 needs to knock some 30 teams out as well this time. Is going to be tough, hope we make it through!

Comment by Rob_Lefebvre on March 14, 2016 at 10:58am

Good work Tridge. Glad the HobbyWing ESC's are working out well so far.

What is your hover throttle and current like?


MR60
Comment by Hugues on March 14, 2016 at 11:01am

TX for the info. Are these cf arms real twill or rather pulltruded ?

Comment by Rob_Lefebvre on March 14, 2016 at 11:48am

Look real to me.


Developer
Comment by Andrew Tridgell on March 14, 2016 at 1:17pm

@Rob, average about 50% throttle in hover, but significantly more power to CCW props than CW props in our auto mission. It seems to be the yaw controller, presumably fighting the wind on the tail. Yaw integrator grew to about 30% while landing.

Hover current was a bit below 15A on motor1 (sensor was only on motor1). That motor was one of the higher ones, as it is CCW. Total current would have been around 110A I think.

We will try and arrange current sensors to cover all the motors in the future.

Comment

You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2017   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service