After my crash near Harrismith, I threw a bit of a temper tantrum and took a break from everything drone. The crash really baffled me and killed the autopilot in the process. So I could not ever retrieve telemetry to see what went wrong. I did figure out the reason eventually (see details below).
Life went along and my job at ESRI South-Africa assigned me to do a talk on UAVs and their use in aerial photography in the upcoming summit. The topic is covered over a whole day and includes a demo and processing of the data.
So, I had to procrastinate later and get the drone up and flying again. Since the nose was a bit mashed after the last crash, I decided to implement and idea that I have toyed with for a long time – rebuild the nose completely. So I started removing what was left of the nose:
My experimentation with 75x75mm PVC gutter downpipe on other planes were quite successful and matched the width of the old body quite closely. I decided on a design that allows the nose to be detached. The “avionics package” (RX, autopilot, speed controller, BEC, etc.) will stay in the body of the plane, whilst the “payload” (cameras, batteries, etc.) will go into the detachable nose. This way I can make noses custom crafted to the mission.
The new nose has way, way more space than the old one. Gone are the 2-3cm thick foam walls, replaced by 2-3mm of PVC pipe wall. This is going to make packaging much easier in the future. The autopilot and camera equipment will also now be nestled in the middle of the plane body, far away from crashes. The canopy is made from old 5 litre purified water bottle heat formed over a wooden blank. So the canopy also has very thin wall thickness. I also repainted the plane white as I was tired of the ugly orange that did not really help much with visibility of the plane.
I also decided to enlarge the rudder fin. I always felt it is too small and have read that other people had good results enlarging this on this plane body. In the process of rebuilding it, I finally found out what caused the plane to crash. The rudder servo was dead. It explains the behaviour of the plane right before it crashed. The auto pilot was flying the let-down pattern and was on the downwind leg, and on trying to turn to the base leg, it would not turn. The plane banked, but would not turn and kept heading away. I took over from the Autopilot and tried to turn it manually, giving it more bank and pulling more on the elevator. It suddenly spinned and was on the ground in two rotations. My guess is that the servo died with the rudder stuck hard over (I could not see it after the crash as the rudder was ripped off) and it was banking in the opposite direction. The low speed, high angle of attack with lots of rudder in the steep bank was probably just too much and physics took over.
This is the second servo failure on this plane. I tried to use good quality metal gear digital servos. Bot failures so far was the expensive ones. The cheap replacements have never died in other planes nor this one. The first failure was an aileron failure and did not cause much trouble. I have now rebuilt the tail with a bigger fin, and a split elevator with a servo driving each half, giving me redundancy on the elevator (the possibility of a servo failure on the elevator has kept me awake for long enough now). Once I get my replacement autopilot, I will setup it back to it’s default of not using the rudder at all. It makes for less tidy flying, but since I don’t have redundancy on the rudder, I will leave it for manual flying and compensating for wind on landings only.
Here is a side view showing the new tail a bit better:
I hope to maiden the new setup this week-end and will post updates and more photos of the insides.
Since I started this blog, I have uploaded various bits on my progress and thought it prudent to do an update on where the project stands now.
In the startup post, I put up some goals for my project. Though I did not meet everything, I am fairly satisfied that I exceeded some expectations that took the project into a different direction.
Very Good Loiter time (1 – 2 hours)
I did not nearly make this target. I did approach a theoretical time of around 40 minutes with my initial designs. My current design is much heavier than anticipated with a much better payload than anticipated. The “mission profile” has also changed from flying around using FPV, and staying in the air for a very long time, to shorter missions to cover an area with aerial photography. With the more capable plane, I ended up reaching a practical limit of around 20 minutes.
Medium Range (3 – 5km)
This was an FPV target that I set. Since the plane’s focus move to aerial photography, I sort of abandoned FPV and only added it near the end again as a fun element. Even though I suspect that my FPV setup will reach at least half the target range, my 2.4GHz radio does not reach near that. The best I managed so far was around 800m. It seems as if I will have to switch to a long range radio setup if I want to go further. Since the plane has a full autopilot, can cruise at around 40km/h for around 20 minutes, it has a theoretical operating range of around 6km. I have however decided against letting the plane fly out of radio range and my fail-safe setting is to return to launch should I lose radio signal.
Pan / Tilt system for training the camera on a point of interest
This was achieved, but later removed. Since the plane’s focus changed, I simplified the FPV setup to a back-pack that can be easily removed.
Waypoint and orbit functionality, leaving the pilot’s concentration on operating the camera and not flying the plane at those times.
This was achieved and exceeded. The plane can take-off and fly fully autonomous. A telemetry link is also present that enables the operator on the ground to direct the plane on a map.
Since the original project targets moved a bit, I’m quite happy at the results so far. The plane performs very well as a cheap aerial photography platform. The autopilot acts as a very good safety-net for starting off FPV flight. I have given the controls to numerous people at the club to give them their first taste of FPV flying without real fear to the plane’s well-being. With the autopilot in Fly By Wire mode it’s practically uncrashable if I pay at least a bit of attention to the person flying it (barring of-course malfunctions to equipment).
The biggest limitation of the current plane is the inefficient airframe. It can carry a lot, but is very “draggy” and uses too much energy in cruise. I really want to move to a more aerodynamic airframe. Even though the current airframe can lift a lot, it’s internal space is really unusable to a big degree.
I am considering two airframes as a replacement:
This plane is built from the ground up as an FPV aircraft. It also has a down looking window for a camera. It gets good reviews from the community and friends using it. Internal space is huge.
- Around the same size as the airframe I’m currently using, so I can re-use the current airframe’s equipment.
- Efficient cruising
- Lots of internal space
- Easy to hand launch, or ground-takeoff
- FPV ready
- I won’t be able to fit a bigger still camera than I’m using.
- I won’t be able to fit a roll mount for the static camera for turn compensation.
This is a HUGE flying wing with a LOT if internal space. It gets very good reviews and is also personally recommended by a friend already using one for Aerial Photography.
- HUGE internal space that is easily accessible.
- Can fit a bigger camera
- Can fit a roll mount for the camera
- Needs beefier hardware than my current setup – need new motor and batteries.
- Can be tricky to launch alone
Due to cost, I’m leaning towards the Penguin as my replacement aircraft.
I’ve had the question put to me whether the telemetry radio is really needed for the Autopilot. Though strictly it’s not required, I would strongly recommend to get the telemetry link with the Autopilot.
Here are some of the deciding factors for me:
1. Don’t have to plug in the plane every now and again to update a mission.
2. Updating and changing of missions while in the air (handy if it heads back to Pretoria while in Limpopo!)
3. You can fine-tune the autopilot while in the air. Carefully of course!
4. You can follow your aircraft on a map while it’s doing it’s job. Also handy to see how well it handles your mission, so that you can learn to plan missions better in the future.
5. Live HUD showing all kinds of handy info. Access to around 90 parameters that can be displayed live.
6. You can control the plane by sending it to a point on the map. Should you have a gimballed camera, you can also control that by pointing it to a point on the map (guided mode).
Below is a video (speeded up 10x) of a flight and what the telemetry looks like whilst in flight.
The latest on-board videos that I take, I take using a keychain video camera.
This camera is available on e-bay at the following link:
South-Africans will have to contact the seller directly and ask for the item to be shipped and that YOU accept responsibility for our unreliable post office.
The camera records at 720p (Half HD) video with surprisingly good quality considering the size of the camera.
I installed a new radio module with two way telemetry and better features. My mom turned 70 this week and we went to visit her the weekend before in Tzaneen, Limpopo.
I took the plane with to take some aerial photos of the area and also to test out the new module.
In the second flight, I lost line of sight to the aircraft, then signal and then it never came back. A frantic search followed and using the directionality of the new module’s patch antenna, I could home in on the location of the plane.
Finally I could get close enough to get a telemetry link again to the laptop, where it revealed it’s position to me on the map.
I recovered the on-board video, and after a lot of head scratching and speculation, downloading of logs and analyzing of logs, I found that the plane had switched out of autopilot, and into manual mode.
It turns out I understood how the fail-safe worked wrong and that it was set up to set the autopilot into manual, rather than “return to launch” mode when it lost signal!
Luckily the damage was very little and some hot glue and 5 mins later, it is ready to fly again,
The video below tells the story from the plane’s perspective, crash included.
It’s been a hectic few weeks and it’s about time for an update.
My drone/UAV/flying foam terror can now fly nice grid patterns with the help of an Arduino APM 2.5. This enables me to get good coverage over an area that is needed for getting nicely overlapping photos for building an aerial photo.
My first aerial photo test went quite well, despite the windy conditions and the autopilot not being tuned in all that well for these conditions.
Below is a link to a map service serving the resulting 500 megapixel ortho photo and DEM (digital elevation model). The photo was built up from around 400 photos.
Open the link below and use the mouse to pan around, mouse wheel to zoom in and out (or SHIFT+CLICK+DRAG to box zoom). Sweep the vertical red line left and right by click-dragging it to reveal either the ortho photo or the DEM.
Be patient, though, it’s large and takes a while to serve out the high res images as you zoom in. It’s only a desktop serving the data, not a proper server!
Click here for the map.
It’s been a while since I did an update as I’ve been very busy getting a new plane up and running for a client that is interested in using it as an Aerial Photography platform.
I strapped a Canon SX 260H into it and did my first trial at the PRF club this afternoon. I was pleasantly surprised by the results.
The resulting orthophoto and DEM came out quite well taking into account that I am still waiting for the autopilot that will make taking the photos in good organized lanes MUCH easier.
Click the image below for a bigger version of some screenshots of the 3d model:
and here’s a video of the model: