The Space Plane Project is an attempt to lift a fix wing UAV to ~30,000 metres and have it fly back home. It is sponsored by Equal Experts and ground support is supplied by Addingham Scout Group.
This article is for those very geeky people who would like to keep up with the state of the build. There’s far too much information in here but if any of it is helpful to others, it will be worth it.
15th January 2016
Prototype Maiden flight
After all those crashes when launching by hand, the bungee launch worked really well. The autotune flight mode did not have enough input to really smooth out the oscillation and the flight was quite short. Only the RunCam on the underside of the nose.
Second flight was with the 360 camera bolted to the bottom. This did seem to slow the ascent rate compared to the first flight but it’s difficult to tell without looking at the logs. I suspect that a 4S battery would make a difference when the 360 camera is on there.
Autotune for longer
- Buy a 4S battery which is the same weight as the two 2200mAh batteries currently in there.
Editing 360 video
Tried editing the video on Alex’s windows computer with the supplied Action Manager software but the quality was very low. Instead we used the Samsung S7 to stitch the raw video, copied it to the Mac and used Final Cut Pro with 360 Tools plugin to reorient the view so it started looking forward. Uploaded to an unlisted YouTube video.
Tuesday 17th January 2017
Received the Bufallo FX-79 air frame and ordered motor, ESC, propellers.
Fitted GoPro into nose of X5 and moved the RunCam to the tail. CoG looks good but only with both cameras on.
Thursday 19th January 2017
Early morning flight of the X5 to perform the Autotune exercises. The RunCam must have been powered on by accident in the bag because it was out of juice. Only the GoPro working. Did not fit the 360 because it was gloomy and footage not worth it.
Flight went well and it flew at max 40m and averaged around 44mph. Climb rate very slow on Autotune and Cruise modes. Maybe because it’s very heavy? Must try on Training mode it see if it is better or max ascent rate is too low. Forgot to try the new Flight Display on the controller.
The new pin on the airspeed sensor fixed the problem, the airspeed now looks accurate.
Fitted secondary GPS to X5 but not yet programmed module for Flight Mode.
Looks like Pixhawk will pick the GPS module as the most accurate based on the number of satellites it can see. This may not work for us because over 12,000m both GPS modules may have the same number of satellites connected but the modules set up for low level will simply not report a position. Code review will be needed.
After some research, it looks like the Pixhawk can be calibrated mid flight using the PREFLIGHT_CALIBRATE command. If it would not break anything else, this could be used to calibrate at different heights and before release from the balloon.
Soldered the sonar range detector but it did not work when plugged in. It may have worked but the APM status screen did not show sonar distance. Unplugged and left for another night.
Turned off the airspeed auto-calibration. It looks like the speed is accurate now.
Investigate in-flight calibration
- Why did the ranger finder not work? Try with battery power instead of USB.
- Investigate GPS module failover, will it work with GPS accuracy instead of number of satellites?
- Try Flight screen on Taranis.
Test airspeed now that the auto-calibration is off.
- Try on training mode to see if ascent rate is better
- Check max ascent rate for auto modes
- Estimate stall speed and adjust min speed on Auto modes accordingly
- Set Flight Mode on 2nd GPS module
Saturday 21st January 2017
Flew from Otley Road again. This is a good site because it provides softish grass and heather which protects against heavy landings and there’s less chance of damaging the 360 camera hanging from the bottom.
The Flight Deck software did not work, it just sits on the “Start your autopilot” screen – still needs investigation on bench.
I’ve not changed the Flight Modes yet so took off on Autotune but switched to Cruise shortly after. Cruise mode seems to use much less throttle than Autotune.
Switched to RTL but it just started climbing so I switched back to Cruise after a few seconds. The RTL height is probably set to 100m which may have been in the low clouds today and I didn’t want to risk not being able to see the UAV. It will have to change to ~40m for this testing.
Just the one flight today. It lasted about 5m 27s and took 516mAh. With 4400mAh available, the X5 should fly for about 40 minutes before the voltage drops low enough to force a landing.
I’ve received the motor and servos for the new FX-79 so I can start to build it now. Must balance that with testing on the X5 though.
Configured the RTL altitude to be 40m.
Setup the correct auto tracks to speak the flight mode when the controller is switched. Setup the Flight Modes to Training, Manual, FBWA, Cruise and FTL.
Trying to set up the APM Power to report the correct time left. On the last flight, the batteries used 516mAh but the CurrTot in the log says 4613.6 (461.36mAh ?). The current BATT_AMP_PERVOLT is 18 but should maybe be 20 to get the right figure. I’ll try that another day.
Found that Flight Deck requires the Pixhawk firmware ArduPlane 3.8.0beta and I’m currently using the stable 3.7.1. Asked the question on the forums as to how difficult it is to upgrade without going through all the calibration and tuning again. May just use the beta version on the FX-79.
Sunday 22nd January 2017
Alex and I met Nic, Michael and Theo on the moor. Nick launched the plane three times. We filmed the launches from multiple angles.
The external LED was not flashing but I could see blue fix light on the GPS and a green light inside the body. Later found that the i2c wire had been unplugged but the plane flew fine without external compass.
Tested RTL mode which worked really well. FBWA was the same as Autotune.
One heavy landing where the plane body was penetrated by a sharp stick.
Climbing is still very slow. Must see if an increase in max climb rate parameter will have an effect or if it’s just the heavy weight that’s the problem. I forgot to test in Training mode.
Talking with Monroe in the US about the NASA launch, we think that a V tail should be fitted to the FX-79. The V should probably have control tabs fitted to enhance the stability but this will need lots of PID tuning which is both a risk to the plane and quite difficult.
There could also be holes in the wings to “fool it into thinking it’s got thinner wings”. The hole would need to be closed at lower altitudes and open on release. Not should how to do this and it sounds like a very big risk, maybe too big.
Started to build the FX-79 just using the spare Pixhawk. I need to know the total weight so that I can do the balloon CAA forms. Put in the Pixhawk, power BEC and vibration damper, mounted the motor and glued in the receiver with the wires going through the bottom of the plane.
Much more room than the X5 and much better build.
- Test in Training mode
Test with greater max climb Maybe try with 4S battery to increase climb rate
Monday 23rd January 2017
More work on fitting the electronics. Had to remove the Pixhawk again to solder on the ESC power cables. Left of the connectors on this plane to save weight.
Ordered the carbon fibre rods and sheets for the V tail. Started to design the tail section. Monroe is right, the best place for the rods to be attached is under the uptight foam tail fins. This leaves all ailerons free to move but the rods will be quite far apart meaning that the tail V will be quite large.
Tail design questions:
- length of sheet to use
- angle at middle of V
- where to fit the servos
- how to attach the centre of the V (bent fibreglass? square rod? round rod?)
- size of tabs
- angle of attack (probably in line with main body at level flight)
Had an idea to test the tail angle by putting cameras on the wings, facing backwards and a smoke micro-generator in the middle to check turbulence of the tail sheets.
Wednesday 25th January 2017
Try the desktop Dropbox and sync the Video directory to Hannah’s shared folder.
- Upload the
steady camand mobile video of last Sunday’s flights. Rename file instead of tape log
Thursday 26th January 2017
More build of the FX-79 since it is priority to find the weight so I can apply to the CAA for the first flight.
With all the innards and cameras, with two 5500mAh batteries, the FX-79 weights 4.5kg. Far too much. Each battery is 900g, it looks like one will have to be enough since the estimated max weight to get to 31km is 3.5kg. I’ve ordered a 4S 5800mAh battery which is 562g – more power for less weight. Tests will be required to see how long this battery will last.
Also had to order a new ESC since the instructions that come with the motor state that min 70 amp ESC is needed, not the 40 amp ESC recommend by HobbyKing.
Also also had to order a new pixhawk 1, the UnmannedTech version because the cheap copy that I ordered from ebay is just too slow and dodgy. The IMU seems to have a delay which will cause wobble when flying. I was hoping that the Pixhawk 2.1 would arrive next week but the manufacturers have warned of delays because of supply chain problems. The pixhawk 2.1 requires a special GPS2 cable to add a secondary GPS but these are not available until the beginning of March.
Saturday 28th January 2017
Fitted the new Pixhawk to the FX-79 and it found the radio straight away and there was no delay in IMU. Only the motor will not arm so need more investigation.
Also had another go at designing the V tail. It looks like the carbon rods can be fastened to the underside of the wing but may need at least a couple of through wing clamps to take the strain.
Why motor not arming? (because the pixhawk main power cable was loose)
- Talk to model shop and see if there’s anything he can supply to fasten the carbon rods to the foam wings.
Sunday 29th January 2017
- Auto flight mode – mission with auto-land
- Get some more 360 video
I was hoping to test auto land but when running the mission the plane seemed to oscillate, maybe due to low cruise speed, so I changed back FBWA after the second waypoint. On analysis, the airspeed sensor reports a steady drop in speed from 17m/s to 14m/s. Maybe cruise speed should be 15m/s or more. Maybe 45% throttle on FBW modes is not enough with this weight of plane.
If the auto mode worked, I could have chased the plane with the Inspire to get some close up film of it flying.
Three flights – first was a little bit heavy landing and the runcam came off the tail. Second flight was with the 360 camera. The camera mount is loose but I risked launching anyway because there’s no one to be hurt if it came off and not much chance of good weather for the next week. Alex pointed out that the camera defaults to 2K recording quality but will do 4K so I need to change to the higher quality when starting to record. Both the last two flights were without the runcam so it was more nose heavy but it flew fine, even with the 360 camera attached.
I forgot to try training mode. Doh.
Noticed that the X5 prop was spinning without the motor turning so I tried to remove the prop shaft adaptor to see if there was anything wrong. As I pulled it off, one of the shaft adaptor leafs fell off. I don’t have a spare so the X5 is out of action until the replacement arrives.
Fix 360 video mounting Up the cruise speed and minimum FBW speed – 15m/s. Change cruise throttle from 45% to 55% Change 360 video quality to 4K – will it stay on that setting between recordings?
Thursday 2nd February 2017
Received the Pixhawk 2.1 and would like to fit it straight into the FX-79 but the GPS cable is not compatible with the UBlox M8N. The new cable includes the safety switch so I’m loathed to modify it but it may be necessary. The original Pixhawk 1 is installed at the moment so I’ll do a few test flights before installing the 2.1. Flight Deck cable is also not compatible with the Pixhawk 2.1 and the custom firmware needed for Flight Deck probably has never been tested on this hardware.
- Maiden flight with the Pixhawk 1 – test airframe and build
- Install V tail and tune PIDs
- Install Pixhawk 2 and do it all again
Still tempted to butcher the cables and install the 2.1 so I don’t have to tune the PIDs twice.
The FX-79 frame is almost done and with an evening available, I could maiden it this weekend, given good weather.
Not sure about having a servo cable splitter or to have the 4 wing servos plugged into separate outputs on the Pixhawk. There should be enough power to run two servos from one output but it would be nice to control the levels more and have the inner tabs move less than the others.
Look up multiple servo control on the Pixhawk
Fixed the 360 mount by adding more glue and a washer/flat nut on the base. It seems really stable and secure now. Famous last words…
Replaced the slipping propeller shaft adaptor and rebalanced the props. There should be much more power now.
Fitted the large 4S battery, which fits well and needs no ESC configuration. Must check the power detection on the Pixhawk settings.
Test the new 4S battery on the X5 Test the new prop fitting Test the new 360 mounting
Friday 3rd February 2017
Beautiful day so I thought I’d grab the chance for a flight before going to the office. I chose a new launch area, one with a few copses of trees, thinking this was a more interesting landscape.
When I pressed the safety switch and armed the motor, the prop started spinning, even on min throttle. I tried to reset a couple of times but it kept spinning on FBWA so I switched to manual mode to launch, so I could safely hold the back of the plane and stretch the bungee. It took off OK but was very unstable on Manual so I switched to FBWA after about 5 seconds. It was heading towards the trees so I pulled up but it would not climb, I stopped the motors and pushed the nose down but not in time. The plane was stuck in the top of the tree. I climbed to the top of the tree and managed to get it down. Both cameras had been ejected and the 360 camera mount was broken off. I found the GoPro further into the woods. The 360 camera was on the floor below the tree.
Look at min throttle setting Take the spacers out of the motor mount and see if that helps ascent Fix 360 camera mount Fix airspeed sensor mounting Check new battery is OK after full discharge
X5: Fixed the 360 mount, removed the spacers on the engine mount, added screw fitting to airspeed sensor. Had to force recharge of the 4S battery because it was under-voltage. Changed the FBW min throttle to 0% instead of 25%.
4th February 2017
Morning: Successful mission run – Alex filming – aborted auto-landing due to walkers near the landing area
Afternoon: Successful autolanding after mission. Tried to film from the air using the Inspire but the plane flies too fast to get many good shots. Took 16 minutes to get a couple of OK fly-bys.
5th February 2017
Quick flight over Addingham Moorside. Range test with FBWA – (look up distance from take-off)
Did not test Training mode because it’s very rocky there and a risk. Did not test emergency radio off for the same reason.
Need to get practice on Training/Manual so I can take it high and stall to recover using Auto, Loiter or RTL.
360 camera broke off on landing – no-where to land in long, soft grass or ferns. Need to add landing protection for the 360 camera.
Flew higher than I wanted to, the controller said 150 metres before I realised. Need to add a geofence/alarm on transmitter to stop it flying over 120 metres and stay within the law.
I’d like to try a mission with a bit of distance at this location, maybe following the ridge and going over the trees. Even as far a the Flesher’s farm.
Used 1310mAh on this flight with lots of climbs. The telemetry on the radio stated ~15.5v most of the time, even at the end. I’d like to check and graph the voltage during a longer flight to the limits of the battery. Also need to check what happens to the plane on low voltage. It should RTL and lowish voltage and then force land on very low voltage. This needs testing on a very open area without public near.
Trying to finish the FX-79 but it will not find the radio signal to calibrate. Asked on the audupilot forum. If this was not a problem, I could have finished the rest of the build.
Ordered a camera hot shoe mount to repair the 360 camera mount
Monday 6th February 2016
On the way to the office, it was a beautiful day with light winds and the sunrise was reflecting off the clouds in rainbows, perfect for a quick flight. I walked up to the usual test launch site and plugged everything in. Unfortunately the GoPro was either too cold or had discharged in the carry bag because it was dead. I’ve not had time to mend the 360 mount so I packed everything up and walked back to the car. Need to make sure the cameras are warm and charged from now on.
Editing 360 video
For the knowing.digital conference, we are getting Google Cardboard units made up with the EE logo and a QR link to the video playlist. So far the playlist only has a single 2D video of the maiden flight so making a good 360 video is priority.
I tried to import the 360 video from the weekend but Final Cut Pro X says it’s incompatible. After a little research, I found that the mp4 files need to be encoded using H.264 or FCPX will reject them. Boo. My i7 laptop takes about 7 hours to convert a 20 minute 360 video so I left it overnight and will get back to editing it tomorrow.
Replaced the Pixhawk1 with the Pixhawk2 in the FX-79 because I found that the telemetry cable fits the 2nd GPS port so I adapted it to connect to the UBlox M8N. I also replaced the servo line splitters with extra extension cables. The long outside tabs are connected to MAIN 1&2 and now the smaller extra inside tabs are connected to MAIN 5&6. There is a setting on ArduPlane to make the 5&6 connections extra ailerons or rudders but this does not allow the evelon mix output to go to these tabs. Put a call for help out on the forum again.
Maybe the inner tabs can be used as differential spoilers or try the MAIN 5 function to be aileron and MAIN 6 to be elevator to match the 1 & 2 channels.
Tuesday 7th February 2017
Trying to get the differential spoilers to work on channels 5&6 but it does not seem to be working. …symptoms….
Bought FCPX because my copy came with the Mac and would not upgrade. The 360 plugin upgraded but would not work with FCPX 10.2 and could not downgrade.
Converted all 360 videos to h.264 and the new files import into FCPX.
Thursday 9th February 2017
Fixed 360 mount on the bottom of the X5 and added a strip of carbon rod to protect it on landing. Not sure the protection will work well in bushes but if landing on grass or flat ground, it should deflect the pressure.
Installed APSync on the Pixhawk2 and connected via wireless UDP from the APM Planner. This will make it much easier to calibrate since it’s not tangled in the USB lead.
Tried to get the MAIN 7 servo working on the Pixhawk2 (Service Bulletin advise) but it’s not using elevon mixing. Must try duplicating the channel 1 output on channel 7 from the remote.
Created a 360 video and uploaded to YouTube. There is copyrighted music on there so adverts may show if there’s enough views.
The poor MacMini was struggling to edit the 4K 360 video and the Dashwood stitching is only as good as the stitching on the phone except it shows a black hole directly above and below, the phone generates a blurred area which looks nicer.
Going to change some selections of the video to show the motoX bikes and see if there’s any phone stitched video I can use instead of stitching on the Mac.
Saturday 11th February 2017
Did a session with the scouts to film them learning more about the Space Plane project. I was hoping that they could see the prototype fly but the weather was wet and windy.
Found that the elevon output mixing code does not yet support differential spoilers so I’m going back to the original idea of a servo cable splitter. Oh well, I tried.
Tried an auto-take-off and longer mission. The X5 took-off OK with what sounded like a full throttle (need to check the log) but when it reached take-off height, I could hear the prop speed slow and the plan slowly lost height before hitting the ground. From the log, it looks like the airspeed sensor is not calibrated and was reading a much higher speed than the GPS, nearly double in fact. There was very little wind so it must have been an inaccurate reading rather than a tail wind. The plane thought it was flying at 30m/s but wanted to fly at 22m/s. GPS says it went from 15m/s to 10m/s which is below the stall speed.
It managed to find the dry stone wall at the end of a very big field. The GoPro glass was smashed again.
On the X5, set the airspeed calibration on again, the arming check was changed to check everything (mask = 254) to check everything (mask = 1). When I tested the new settings, it showed that the GPS2 was still configured and the compass calibration was off. Disabled GPS2 and recalibrated the compass. Replaced the glass on the GoPro.
Thursday 23rd February 2017
Great show at the Equal Experts conference in Leeds. Loads of people very interested in the Space Plane. Preparing for the show has used time that would normally have been building and flying but it was worth it.
Monday 24th February 2017
Moving house and the weather is bad with storm Doris making flight impossible. I’d like to test the changes to the X5, maiden the FX-79 without the tail and start to build the first balloon box but packing to move and high winds mean not much happened this week.
I did replace the 40 amp ESC with a 70 amp one. I had to solder the motor cables to the ESC because the HobbyKing ESCs do not come with the bullet connectors. The Turnigy ESC was easier to configure for throttle throw, the HobbyKing ESC beeps 4 times for the 4S battery but does not spin up when armed. Need to spend time finding out what the issue is.
Tuesday 22nd March 2017
Started the build for the first balloon test launch. The payload is fitted with a GoPro HERO 4, two GoPro HERO 5 Sessions and a GoPro Karma Grip with HERO 5. Lots of GoPro but since the HERO 4 proved to be the most reliable of all the cameras sent up last time, I decided to stick with what works best. The OSMO has been replaced by the Karma and I’m hoping the this will last the whole journey. It’s much heavier than the OSMO and sticks out the back from the payload box but the payload has to be ~4kg to test the final weight of the FX-79 anyway.
I’ve tried to get the FX-79 working with a 70 amp ESC but both the hobbyking and turnigy escs don’t work. The hobbyking esc does not come with bullet plugs so I soldered them to the motor but this means I needed to order a replacement motor when I decided to go back to the 40 amp ESC. It’s now looking good for a maiden flight on Saturday.
Friday 24th march 2017
First flight of the X5 since it crashed on a long distance auto-mission. It was a problem with low default speed so that has been increased. Last time I tried to fly, the APM would not arm because the compass needed recalibration. This time there were no problems flying. The airspeed sensor is on calibration mode but in use and I mostly flew on Autotune. The RTL worked well even with the blustery wind, which was > 10 MPG AGL.
Tried Training mode but it was very gusty and the plane flew past the sun and next view I had it was heading away from me and towards the road (about 750m away) so I changed to RTL to get it closer.
The lower C battery worked well and the CoG is about the same despite the slightly lighter battery. It climbed well on both the 20C and 40C batteries and I could not see any flying difference.
I wanted to fly the Inspire at the same time but forgot to pack the USB cable for the tablet. Single point of failure.
Saturday 25th March 2017
Crashed the X5 into a wall. Again.
This time I was trying a long duration auto mission. Take-off was good but 25 metres AGL was not enough to clear a the wall in the next field. The port wing broke off but it was a clean break and it glued back on with hardly a visible line.
Monday 27th March 2017
Quick test of the repaired X5 and it flew fine.
Tuesday 28th March 2017
Box Payload Build
I was supposed to release the balloon today but it was so foggy that it the video would have been very boring. The KPI was predicted to be 6 or 7 so I cancelled the flight due to possible damage to electronics and rubbish GPS signal, which would have negated the test anyway.
Trying to configure the UBlox M8N GPS receiver for high altitude flight but the ucenter software can’t connect to the receiver. On the macbook I’ve installed Parallels with Windows 10 and the latest Mission Planner, followed the instructions here to connect via MAV but ucenter says nothing is listening on port 500. The FTDI cable I ordered a while ago is still packed/missing since the house move. I have asked for help in the forums. Sunday looking good for a balloon launch so I hope to get it configured by then.
Thursday 30th March 2017
Found the problem with the GPS configuration…. the plug on the new M8N GPS module was for the older APM and not the Pixhawk. Luckily I had a spare plug and just changed it over to get it working. The latest PX4 firmware has configuration for the type of GPS so I don’t need to connect the u-centre software apart to confirm that it’s working.
I bundled all the cameras and pixhawk together in the payload box and hung it outside overnight. The cameras ran for 4.5 hours, recording 4K@30fps which will be enough for the whole balloon flight. It may not be enough for the plane flight since that may be 2 hours to release and 4 hours of flight, maybe. Anyway, all is set-up for the first balloon flight on Sunday morning.
The CAA have kindly sent an exemption for Sunday and setup a NOTAM but misunderstood that I wanted the second launch to be during darkness and the exemption is from 03:00 to 09:00. Plan is to get up early and start to fill the balloon at 05:00. Hopefully the payload will land before 09:00.
Saturday 1st April 2017
Called Leeds ATC and then the NOTAM office and they moved the NOTAM from early morning to afternoon on Sunday when the weather was forecast to be much better.
Sunday 2nd April 2017
Balloon release was at 5pm and it went really well. Alex was in control of the DJI Mavic to film the take-off and was right in the flight path and managed to dodge out of the way just as the balloon passed by.
We were held for a few minutes until ATC called back to say go, they had a whole queue of planes taking off and they didn’t want the balloon going by when there was lots of traffic. The cameras and electronics were all on and running for about 10 minutes before we got the call.
Alex and I chased the balloon down the predicted path, which it was following exactly. We waited by the side of a road new Doncaster for the signal to return and then raced along country roads to get there as close to landing as possible. We were beaten to it when we received a call from a young lad who had been playing in the park and the payload had landed while he and friends were playing. I had some strange calls from parents on the drive home.
The 360 and the GoPro Session had run through-out the trip but the Karma had stopped recording after an hour or so.
Even with ~3kg payload, the balloon reached 34,600m according to the blackbox, so one of the mission goals was a success.
The pixhawk worked all the way through but the GPS lost signal from about 20.5km. It could have been the configuration – airborne4G – or it could have been the cold which hobbled the GPS receiver. Not sure how to find out other than do another balloon launch.
The next mission, test release mechanism, can’t go ahead without accurate GPS at altitude above 30km.
Easter Weekend April 2017
Built the Tarrot Hugin body but it does not come with engine mount, servos or servo extension leads. Without the servo leads, the tail build cannot be started. Apart from the missing bits and no help from the vendor, it’s a well build airframe. There’s loads of room inside for electronics but there is no thermal insulation so foam will have to be added if this fuselage is going to be used at high altitude. It is about the same weight as the FX-79, when taking the V tail into account. The inverted V tail from the Hugin could be used on the FX-79 but I’ll try it on the Hugin first.
Also built the Skywalker Naja because this has a V tail and so the configuration will be very similar to the Hugin and the FX-79 with tail. The Naja is also less brittle than the other two so it should survive the initial testing.
Another attempt to maiden the FX-79 (without the V tail) but I’ve lost the launch bungee and could not cobble something together with the short sailing bungees I had in the garage. Managed to configure the Pixhawk 2 for elevon mixing but reversing both channels on the transmitted and both channels on the aurdupilot radio setup. Don’t know why it needs this, tried every other configuration, this works so I’ll just leave it.
Found out why there are black holes at the top and bottom of my 360 videos when edited though Final Cut Pro…. When I import the stitched videos, they are imported at a standard 4K resolution but the Gear 360 produces slightly shorter resolution than the standard. FCPX puts black bands on the top and bottom of the video which appears as black holes on the finished clip. Solution: import at a custom resolution which is the same as the original Gear 360.
Saturday 22nd April 2017
Successful maiden flight of the FX-79. First try was not so successful, the elevons did not have much throw so it just plowed into the heather and the wings came off at the link. Some of the carbon rod was damaged and a big chip out of the foam. After sticking it all back together and increasing the throws by attaching the push rods to lower holes on the control horns, the launch was good. It’s a very slow flier and seems to drift around turns, coming out almost going sideways. It’s probably not worth configuring the differential spoilers because it’s going to have a V tail attached soon.
Built the Naja electronics. It’s all ready to maiden apart from configuration, the air speed sensor and the prop, which has not arrived from HobbyKing yet. The Naja is V tail to I can hopefully iron out most of the configuration kinks before trying the Hugin airframe and then the FX-79 with the Hugin tail.
Could not find servos to really it the Hugin V tail so I jammed some low profile rover steering servos in there. Not ideal but the are solid.
Built most of the next balloon payload. There’s a couple of servos on the outside and instead of a plastic dome, the GoPro karma is free on the outside of the box. That’s a risk because the gimbal may freeze and there’s a much higher chance of damage on landing. It’s worth the risk for much better video though.
Sunday 1st May 2017
After lots of ground tuning and tweaking, the Naja V Tail was launched for the first time. I’ve not flown a V tail before so was nervous and the wind was Easterly so we had to move to a spot away from our normal testing ground. The Naja flew without drama and glided for a long time without power. Not camera mounted on the plane because this is just a test setup for the big V tail but Alex took lots of still photos. I flew to low and slow, trying to make it easier for Alex to snap pictures, but it stalled and came down hard. The wings had detached and the propeller broken but, other than the prop, no damage.
There should be enough capacity for the Naja to have the S7 phone in the payload and have the Mavic follow the phone and film the plane. We would have tried that at the Naja maiden but I only brought one flight battery and the phone was low on power. Must try it ASAP for the Naja and the FX-79.
Bad weather. CAA Application
Another CAA Application. More bad weather. Shadow 200 build.
Tarot X8 octocopter
We bought an old Tarot T1000 octocopter from ebay and installed a Pixhawk. After finding that I’d put the motors on in the wrong order and correcting it, the copter flew really well. Landing seems to be a problem and it almost always tips over no matter how gently or carefully try. Some of the self-tightening props are now stuck on but because this is only going to be used for release tests, there should be no issue.
Starting with a paper plane then with a toy glider, then the skywalker X5 and eventually the big space plane, we plan to do release tests. For the smaller aircraft, up to the X5, the balloon payload foam box will have a Pixhawk 2.1 iwith Edison running apsync OS and DroneKit script. Mission script is now open source. The script looks at the current altitude and triggers a release mechanism as a preset height.
So far the paper plane and the toy glider have been released from 10, 25, 50 and 100m heights.
Test servos on side of payload. Script to repeatedly move test servos.
The long line to the payload caused a pendulum effect which made the octocopter too difficult to handle and therefore not safe enough. I’ve moved the payload to between the drone legs which means the copter flies well again but may cause problems with the larger wing span planes and the prop downwash.
We are working on simulating the mission script on a laptop without having to copy to and start real hardware.
We tried multiple times to launch this plane but only managed to get it to fly once and that was taking-off into a gusty 15 mph headwind. When it did fly, it flew very well, even with very little tuning. We tried several more times in gentler winds with longer bungie and 6s instead of 4s batteries but no luck, it just skimmed the grass and then stopped. Conclusion: the FX-79 is much easier to launch and test.
I’ve ordered replacement props and will try again with the Shadow but I’m now skeptical about it’s ability to survive the ground tests, let alone the record attempt. I can now be a bit more energetic about launches, worrying less about damage unless I find a safe and repeatable method to launch.
After lots of successful glider drops from the octo, it looked like the mission script was working well. We released a balloon early on Sunday morning which dropped the glider at about 30,000m. The video captured the glider falling and tumbling – it would have done better with a tail drop instead of a flat drop.
Overall a successful test but three things went wrong:
- The balloon shattered but some of it was caught in the parachute. I think the line between the parachute and the balloon was not long enough.
- The dataflash logs were not on the sd card. The flight was over 3 hours, including searching and retrieval, so the logs files would have been quite big and maybe over the 4 gig filesystem limit.
- The mission script logs were overwritten when I powered-up the pixhawk at home. Need to change the script so it rotates or backs-up the logs on start.
Prototype version 4
We’ve built another Skywalker X5 and squeed the Pixhawk 2.1, a 5500mAh battery and radio telemetry inside. It flies really well but is quite inefficient and will only fly for 30 minutes to 40% battery. I discovered the CoG page on ardupilot which describes how to find the correct CoG for a particular airframe. I’ve balanced the X5 exactly to the manufacturer’s instructions but the logs show it needs lots of up elevator just to fly level (average PIDP.I = 13 and should be 0), which is caused by either the motor having the wrong angle or CoG being too far forward. Since the CoG is spot on, I’m going to try remounting the motor so it pushes the nose up a little, hopefully making the flight much more efficient. I’m hoping for 90 minutes flight time on 5500mAh.
Next task is to update the script to start a flight when the release mechanism is triggered.
I have found that I can’t warm up the Pixhawk 2.1 in the car and carry it to the launch site because when I try to arm, the IMUs are inconsistent. Either the plane needs to be armed in the car (bad because motor may spin and the home point is initially wrong) or rebooted (bad because the Edison is running linux and forced reboots are not nice) or I find a way to recalibrate the IMUs from the script.
Successful launch of the Shadow 200 clone
After 19 unsuccessful trips up onto the moor to try different methods of launch, we finally have a repeatable successful method. The battery is now a 6s instead of 4s, the propeller is spinning and throttle at 50% before take-off. It’s too dangerous to hold the wings while the prop is spinning so the tail has been reinforced with fiberglass where it is mounted to the wings, the carbon rods can now be handheld while the bungie is stretched. A head wind helps too.
The max airspeed reached on 80% throttle is 18m/s (40mph) which is much lower than expected. 20mph winds on the ground usually mean 40mph or more at 100m AGL so groundspeed is almost zero.
On the first couple of flights, I had reversed the rudder direction to make it work but, even though manual control and FBW worked fine, auto, RTL and guided modes made the plane fly sideways as the tail and the yaw fought to turn in different directions.
Despite the tail configuration the plane flew really well and the unpowered glide is very efficient, to the extent that it was difficult to land in a strong wind because it did not want to descend. I’m looking forward to stall recovery tests because I think it will recover easily.
A guessed CoG looks to be almost correct, the log shows a PIDP.I average of +1 which is about as perfect as I could hope for. The CoG will need more tuning as more batteries and cameras are added.
X5 Recovery Tests
Taking the Skywalker X5 up to 100m, switching to manual, throttle to zero and allow to stall. After a few seconds, I switch back to FBW (RTL mode) and allow the flight computer to recover from the stall. Every test has been successful so it’s now time to try dropping from the octocopter.
X5 Centre of Gravity
In an attempt to make the X5 prototype more efficient, I’ve tilted the motor down so that the tail is pushed down and it should need less input from the control surfaces just to keep it level. I’ve changed the battery from a 50C to a 25C since the max measured current is 15amps. The smaller battery size allows the battery to be moved back and therefore the CoG to be further back. The result is still an average PIDP.I of 8 which is not very good. Anyway, this is just an expendable prototype so no more time tuning allowed!
Shows and talks
The planes and multicopters have been dragged around the country as we give talks on TV and at conferences. We’ve been up to Edinburgh and down to Brighton. We’ve even had two days at SwanseaCon.
Telephone and email conversations with the CAA are going well, they seem open to a launch very soon but are currently overwhelmed with work so our project is seen as low priority.