It was pretty depressing and frustrating to get to such a conclusions…
- Aerial Photo-Video vehicle is complex integrated solution where software works with hardware extremely tight so it’s impossible to separate them out.
- ArduPilot is software project and worth nothing without hardware, hence useless for video production since it covers Flight Controller only.
- Open Source HW as an entity is myth:
- Software is ‘air’, you need nothing extra to what you already have to start creating new piece of code, click a button – done. Any other person can copy and use it right away in exact same manner.
- Hardware – you need materials to start with, build something physically. Document every single step, provide bill of materials.
- Other person need to get these materials (pay for them) and repeat all your steps in order to replicate, it will not be same and will not work same way.
- It will be different item, experience will be different because the only way to replicate it in correct way – by creator, means job need to be paid. Don’t forget that DIY and Open Source documentation is bad, incomplete and inaccurate in most cases.
My advice: need working tool – do not waste even minute on DIY and Open Source.
DISCLAIMER: Stop here and think. Below are my personal thoughts and conclusions. They may hurt your delicate soul. If you disagree with above – go read something else.
It was started “long” time back when GoPro 2 appeared on shelves. DJI had frame wheel F450 and NAZA, ArduCopter firmware was run on APM 2.6. Not vary many frames, motors, props, ESCs are to choose from. Most cheap and DIY drone community was growing fast.
By that time I was searching for solution for auto or remote “cameraman” to get family videos. It’s pretty inconvenient when you cannot record whole family action, when at least one member has to control camera. There was one more requirement – it should be portable enough, not only vehicle but whole system
At that time ArduCopter project stated, they have tracking capabilities. Great! That was exactly what needed! Well, I need to build drone for it and learn how to fly first.
It didn’t take too much time to build first flying quad. It’s pretty strait forward even with simple Flight Controller (FC), many flights and crushes and within year you gain enough experience to start doing vehicle to achieve goal: flying “cameraman” with follow capabilities. At that time DJI had introduced first Phantom, compact enough for medium size backpack.
Here when small grain of salt started to sneak in. One by one.
First grain. 3DR and all variances did NOT have any vibration isolation, while DJI had it almost from the beginning. This “small” detail was simply overlooked as not important for DIY community. Well, do it yourself… Half of forums are about it. Almost every issue with copter not performing well boils down to person did not isolate FC properly. People who was able to compare APM2.6 with NAZA stated that after balancing props and motors APM still not performing as its rival. And if just step back a little it’s obvious that buying cheap Chinese part will not perform as well as more expensive Chinese part since almost all come from there except may be KDE which I’d like to try but because of very reason of this post it may not happen.
Second grain. New FC came out – Pixhawk – 32 bit ARM architecture, some redundancy, more advanced algorithms not fit in APM any more. And what? Same story – no vibration isolation. There were engineers worked full time on this new FC, a lot of SW and FW improvements, but no so highly anticipated feature. I wish be more farsighted there to see bigger problem and switch to DJI then.
Third grain. Finally I was able to build quad with “follow me” feature. Why it’s considered as grain? Well, it worked in one particular build only, and I was not able to reproduce it on other vehicle with same FC and gimbal at same level of accuracy. Other issue with ArduCopter solution: you need phone with connected telemetry module on you and better not blocking antenna, otherwise it’ll lose connection. And it may cause fly away.
Another thing: way they use for tracking is simple – phone send its coordinates and FC calculates drone position and camera pitch. This not very accurate, but DJI had same thing – remote was used instead of phone and results are same. So method itself has limitations, it’s fine.
Telemetry module and Tower App draws phone battery extremely fast, fully charged went down to 20% in 30 minutes. And this become an unacceptable: you have to have remote AND phone AND telemetry module on you in order this feature to work and you have 30 mins at max.
Fourth grain – portability. Try to pack everything you need for 3-4 flights. If just fly drone itself: vehicle, remote and 3-4 batteries, but remember. goal was to have video rig. It adds screen connected to remote of on tripod. But it all turned to point that we are limited to camera. It’s not difficult to make small drone, but it’s difficult to make small drone capable to carry camera you want to shoot video with. And this is a big problem. Why? Because there is only camera that is small enough – GoPro. Camera and any good 3-axis gimbal add 300-400 grams easily. Accounting video transmission and all ugly wiring you need to solve – everything adds up. Smallest setup I was able to build took two 24 inch DeWalt tool boxes, drone, 4 batteries, remote+2 batteries, screen, goggles, spare parts. It’s not any portable at all. No one go with two tool boxes on weekend hike with family. And GoPro camera is horrible choice for aerial video.
I have smaller version for 4S battery setup, but it doesn’t make much difference…
Fifth grain. Concept of ArduPilot project was wrong. Idea was to make FC firmware flexible and expandable. It’s good at some extent. Just in mid 2016 lead developers started to work on obstacle avoidance using lidars – interesting and extremely expensive devices, definitely not for hobbyists. Initially reading sensors and making decisions about distance supposed to be made by FC itself as everything else. This means every sensor has to have standard interface and protocol – in DIY area it’s really hard. Even MavLink is not standard yet. Luckily there are smart people in community and now we have a message “DISTANCE_SENSOR” and all free to use any hardware to generate it. Hooray! But it’s too late for such relatively simple solution
Sixth grain – no integration. Means there are vast of low quality electronics around which is meant to be as final product. In order to build something nice looking and good performing and not ugly duckling, one need to invest time and money into development and integration, which is nothing close to hobby.
Pre-Summary. Till this point I spent 5 years building, tuning, trying and rebuilding. I was not able to solve jello on camera no matter what I did. Interestingly the most clear video I’ve got was from drone with hard mounted GoPro on it, just a little of foam.
There were 5 summers went by where I had no vehicle ready because something was not working or was not tuned or system was so bulky to even think about taking it with me.
Drones are expensive hobby if one have relatively ambitious goals. After I accounted all spending on it – it’s about 3-4 DJI Phantoms, and I still far from reaching goal.
Finally I switched to DJI Mavic – optical tracking, DH video downlink, really portable and video quality is better than GoPro. And it’s actually cheaper than any of my builds – camera+gimbal+remote cost more that Mavic, simple math. Despite some restrictions it’s still years ahead.
Summary.
When go DIY? It’s simple – no goals, just love of tinkering.
When Proprietary?
– need tool to do the job;
– need flying camera with quality better than GoPro;
– integrated portable solution with HD down link and optical tracking.
Even if you have any DSLR or mirrorless camera and want to use it in air, go with DJI platform, FreeFly or others who already working fulltime on making these things usable. Do not waste you time joining together fully independent pieces never designed to be as one system.
3DR partnered with DJI – it should mean something for those thinking.
FC is just tiny bit, it’s like making decision which car to buy by looking at ECU – seems stupid.