Aeriel Cinematography - Build or Buy

3689697042?profile=original3689697072?profile=original3689697113?profile=original

I've been traveling around the world the past couple years and on/off of it's occurred to me that it'd be great to have a drone to get some of the crazy shots at especially scenic locales.

The latest generation of consumer drones are almost there - the Phantom 4 and (to a lesser degree) the Yuneec Typhoon H PRO have decent image quality in a small package (1.4kg, 2kg respectively). As a photo/video geek however, tiny sensors with middling dynamic range, rolling shutter, and substandard bit-rates rankles. Having the flexibility of mounting a higher quality payload like the BMMCC, or a 360 solution like the 360fly 4K would be much better.

The tradeoff for DIY drones is that while adding payload flexibility (and a potentially better travel form-factor), they seem to have fallen behind AIO devices in terms of flight/camera automation.

$1399 - Phantom 4 Positives:

  • A very nice phone/tablet UI that includes smart drone and camera control as well as video.
    • Tap to fly
    • Optical object tracking (ActiveTrack)
    • Orbit mode
    • While it doesn't have explicit "cable cam" modes, you can course lock and set waypoints and manually control the gimbal
  • Lightbridge is a 2.4GHz digital Wimax-like protocol that advertises 5km range and does integrated transmit, telemetry, and video (220ms latency)
  • Has stereo-optical frontal sense and avoid and active obstacle avoidance
  • It also has a downward facing camera and ultrasonic sensors that give it best-in-class low-altitude hover
  • 1.4kg flight ready
  • Quick release rotors

Phantom 4 Gotchas/Mods:

  • Replacing landing gear - the form factor isn't ideal for strapping on the back while hiking, this, and landing gears in the shot can both be fixed by replacing the default skids w/ retractable landing gear (see also: 1, 2, 3, 4, 5, 6)
  • The gimbal/camera is quite complex looking, hasn't been reverse engineered, you're likely stuck w/ it
  • You can probably take the Phantom apart and weatherproof it (CorrosionX everything); it may be possible to replace the lens

$1899 Yuneec Typhoon H Pro/Realsense:

  • Sadly, it's almost not worth talking about this because while there are a many good things, the camera is terrible - it shoots a mystifyingly low bit rate (29Mbps vs the Phantom 4's 60Mbps) and doesn't have a true flat profile. The default lens is super-crappy and focused improperly, so if you buy one, consider a lens upgrade mandatory.
  • The Typhoon H is a hex, and can fly if a motor/prop goes out. It also has quick release rotors.
  • It comes with a retractable landing gear and allows 360 motion on the gimbal out of the box, but that doesn't help portability much - the arms fold down, making a big cylinder that's probably not the most convenient thing to carry.
  • The ST16 controller is quite large, but integrates an Intel Atom touch screen system into the controller w/ 1080P transmission w/ <200ms latency all via a custom 5.8GHz (802.11ac-like?) protocol (1.6km range)
  • It uses front-facing ultrasound and Intel Real Sense camera for obstacle avoidance
  • Camera modes:
    • Orbit Me
    • POI
    • Journey (selfie)
    • Curve Cable Cam (preset curves for flight and camera position via waypoints)
    • Watch Me/Follow Me (GPS)

And of course, there's the DIY Drone option. I will probably end up here, since I've started going down the rabbit hole here, but some thoughts.

First, portability is the pretty key here. Although flying weight will probably be a bit heavier than the Phantom 4 (just over 2kg), we can build a drone that can lift over 1kg of payload that packs up to about the size of an A3 piece of paper using an Alien 680 frame (Skylark, H4/H4).

Above is what it looks like folded up

The foldable props are particularly sexy.

For travel, separate servoless retracts is probably the way to go, but life may be easier w/ some known standard retracts.

We're of course responsible for our own camera/payload here. I have a Point Grey Grasshopper 3 lying around that'd be worth trying to get to work. Notably, it has a Sony IMX252 sensor which is FHD+, 120fps+, has a global shutter, and about 12 stops (72dB) of dynamic range, far better than the DJI/Yuneeq cameras assuming I can get something reasonable to record with (ideally an XU4MinnowBoard MAX, or less ideally, a TX1/J120).

In addition to that we'll need a gimbal. Variations of the Alexmos/SimpleBGC controllers seem to be the most common/well known. This undoubtedly requires some more research, but there's no end of control/programmability there.

The same lightweight gimbals should be usable for mounting something like the Fly360 4K.

Gimbals that can handle heavier setups (BMMCC, other m43 cameras) are more expensive and heavier, but should be doable. (if my eCalc numbers are accurate)

You can of course spend a lot of time working on the power train. I'm almost tempted to buy some motors and a testbench simply to confirm some of these, but there's no point over-optimizing before doing a build, I suppose.

Lastly, and most importantly perhaps are the navigation bits.

There's plenty of 915MHz telemetry modules that are quite cheap, although it might be worth considering a DragonLink for ridiculous range, handling the RC transmission, and bt connecting to handhelds.

The UBLOX M8N GPS modules seem pretty standard and can be bought for dirt cheap (cheap one w/ EMI protection), you might benefit from a really large pro version (or an expensive smaller one) - there's a tiny comparison here. For those serious about their GNSS it looks like Emlid's Reach is pretty cool (DrotekandPiksi are two others)

For the Pixhawk flight controllers, two good options look like the PixRacer which has slightly newer sensors and built-in wifi, and the PixFalcon which has a failsafe mc and some extra power filtering. Both are pretty tiny.  (I'm still researching whether it's worth going with the newer RPi-based controllers like the Erle-Brain 2 and Navio2 - right now it may be better/worth it to simply connect a separate board via MAVLink for more complex stuff, but have a more reliable mc for basic flight controls).

My initial plan for video transmission is to use an analog 5.8GHz transmitter w/ HDMI support, but the Amimon Connex Mini looks pretty darn cool (1080p60 @ 1ms latency, 500m range).

Ultimately, I'd like to primarily use autopilot/telemetry for most of the flying, and at the very least reduce radios (the DragonLink would combine the telemetry and RC control), but RC control, at least for a backup/emergency is of course required.  This is probably it's own project, but it seems like it'd be possible to build a tinyduino/PPM+receiver+minjoystick 2/4-channel doohickey.  Until then, a janky $15 controller might be the most compact option.

Obviously, just about everything for this setup requires some work, but the big projects for me would be:

  • optical-flow/object tracking/gimble movement (optical follow me)
  • gimble/camera direction waypointing (curved cable cam)
  • laser/sonar obstacle avoidance/stability
  • fly-by-telemetry (radio/control reduction)

For those interested in a parts list/notes:

https://docs.google.com/spreadsheets/d/1TkPLaImvb76bMIvJKKhjp9y2UGUzCGRrSXMRtdTIkas/edit#gid=0

My plan is to start putting some of this together when I get back in the US in Sept/Oct

E-mail me when people leave their comments –

You need to be a member of diydrones to add comments!

Join diydrones

Comments

  • if you re looking for a foldable high quality frame, i've probably the world largest choice !

    https://multir-concept.fr/

  • Hi I to am interested in Aerial Photography/Cinematography and believe the biggest hurdle at present is the Redundancy aspect, as generally these types of Rigs are going to be large and be carrying expensive Cameras etc, in Germany for example if your Rig has sufficient Redundancy built on-board you can obtain authority to be able to fly over crowds/gatherings for example. DJI have obviously cornered this market with the recent introduction of their Matrix 600 and A3 pro flight controller as this gives you the redundancy capabilities. But there are other companies working on giving you this built in redundancy for example I have recently discovered http://www.topxgun.com/en/ who appear to be one a a select few that have the capabilities to offer their own solutions the T1-S and T1-Pro have dual IMU/GPS redundancy at affordable prices that can compete with DJI and their newly released D1 goes one step further with built in Dual redundancy and RTK accuracy no one else besides DJI is able to offer this at present.

    上海拓攻机器人有限公司-让天下没有难用的无人机
    拓攻机器人是专业无人机飞控系统及应用解决方案提供商,提供无人控制领域的核心部件和整机方案服务。 我们坚持“开放、合作、共赢”的理念,积极寻求合作伙伴共同创新,形成健康良性的无人机产业生态系统,推动无人机行业持续发展。 以飞控为核心,我们面向各个行业推出针对性的无人机应用解决方案,目前已涵盖安防警用…
  • @Gary. Just to be clear that 250 is not mine but it inspired me to get a zmr 250 and Feiyu Mini3D Gimbal and start work on something like it. I will be using a Storm32 controller where you can set the gimbal yaw follow to not see the legs. A Hero 4 with a good after market lens is still in the quality range of a Phantom. I have most of the bits and a ZMR frame is cheap. Ditto a used gimbal. Still, not really a newbie project for sure.

    I think Pixracer on 3.4 is just about there now.
  • A couple thoughts:

    • Ambarella upcoming 14nm H2 SoC will do 4Kp60 HEVC and 4Kp120 AVC as well as support 10-bit HDR. They provide cameras for GoPro (expecting the GoPro 5 and Karma to have a big jump in quality) as well as DJI. I could just wait for the next gen and probably be satisfied image quality wise.
    • I may end up getting a consumer drone for OOTB usability but also build another one to hack on. I program in Python, and have been working w/ embedded ARM boards, doing image processing and automation for the past few years. Between DroneKit, Flyt, and other high level layers, it looks pretty straightforward to add a lot of the missing camera functionality, and it might be fun to play around w/ the radio stuff like Wifibroadcast and openLRSng and all the whizbang sensors (ultrasonic, px4flow, laser altimeters, rplidar, etc)
  • Hi Marc,

    I know the Pixracer runs AC 3.4, but that there also remained the last time I looked a few unresolved issues.

    I think it is a fine controller and is now being fully supported by the ArduCopter development team.

    I was actually referring to your 250 quad as an example of what could be done.

    My only concern is whether a GoPro will satisfy his needs, if not it rather ups the size requirements..

    He may also be concerned about the 3 axis gimbal and the legs providing interference with the camera.

    Maybe it's worth mentioning that there are 2 flavors of 3 axis gimbal.

    By far the most common one only provides a small amount of yaw to allow the camera to compensate for undesirable airframe movements.

    The less common one like on the Inspire and top end Yuneec with retracts provides for full directed motion yaw.

    As long as you are willing to let the air frame provide the yaw, there is no reason to use retracts, you can design the landing gear so it is simply outside the field of view.

    I think your 250 is dynamite and while I love the Solo for all it's advanced video dollying capabilities, I will be trying to put together something like yours for carrying around more compactly, I think we will be seeing a lot more pro video quads like that in the future.

    Best,

    Gary

  • @Gary "If you only want basic control, the Pixracer or even NAZE controllers do an excellent job."

    There is no reason to choose a Pixhawk over a Pixracer. Pixracer runs AC 3.4 and is ideal for any autonomous quad. As for stuffing a gimbal and camera on a 250 quad -- look at the video I posted. It can be done right now with the only compromise being a flight time of 8 minutes. I am going for easily removable carbon spar legs and a Feiyu Mini 3D gimbal that has quick release. But as you say a Solo or Phantom is the way to go.

  • Hi lhl,

    You do present an interesting problem.

    For traveling around you really want the smallest quadcopter package that will work for what you want to do.

    (You really won't benefit in any substantial way from a hex or larger, quad motor / ESC reliability has improved to a point where a single motor failure is a very uncommon mode of failure.)

    Basically you will want a quadcopter.

    The Phantom 4 is a very versatile and well proven package at this point from every aspect if the camera is satisfactory for you.

    As soon as you go to something more substantial than the current state of the art GoPro (and these manufacturer specific clones) you really need to think about building your own.

    And the Skylark package you show does fold to a very portable package.

    If you only want basic control, the Pixracer or even NAZE controllers do an excellent job.

    If you want actual semi autonomous repeat dolly capabilities, get a Pixhawk.

    For that matter if you really want a copter well designed to assist with video shooting, get a 3DR Solo, it has really well thought out built in semi autonomous video dollying capabilities.

    For now you would be stuck with a GoPro Black, but it does a really nice job of providing a high res FPV output to an IPad Mini albeit with about a half second delay (used for video photography works fine, FPV racing not so much).

    I have a Solo and it is very easy to use and the video dollying modes make it really easy to set up and get the shots you want.

    Marc above is  recommending a 250 racing quad, I think the entire future of multicopter video/photo use is going to be in smaller and smaller frames (and for your use this is a good thing), the main limiting factor right now is stuffing in a decent camera and gimbal.

    Maybe the best place to start is to figure out what the smallest camera / gimbal pair you can be happy with and then get the smallest high performance quadcopter frame you can stuff it on to.

    The rest is easy.

    Also, regarding retracts, while it is true you need a 3 axis gimbal for the smoothest possible video, you don't necessarily need to yaw the camera itself for adjusting your shots.

    Multicopters can either yaw the camera or yaw the entire copter with equal control.

    In fact, yawing the entire copter can be easier to use especially with decent semi-autonomous assistance (like the Solo does).

    Then you don't need retracts which add weight and complexity and are kind of Mickey Mouse.

    Sure the Inspire and equivalent Yuneec do a great job with them but those are expensive, big bulky multicopters and really not the best for packing around with you.

    With my Solo I got their backpack which works really great for carrying the Solo and a few batteries, but your kind of stuck for lunch unless you want a power bar.

    Just a few of my thoughts and experiences.

    Best Regards,

    Gary

  • @Justin. Think much smaller and look at my link and watch the video. That is a ZMR 250 racing quad with a Feiyu 3 Axis gimbal and Hero 4 Black. <900g AUW. You could use a Pixracer on that frame as well. That is smaller and lighter than a Phantom. It would be a little work though....

  • 3702276377?profile=originalThis has worked out great with Pixhawk, 15" props and a 10,000 mAh 6S, with at least 1.5 kg of payload, and could be made into an X8 that still folds down. This is for mapping though, and with a 660mm frame (or the one shown at 800mm), it doesn't have the portability you're looking for.

    If you want to go the DIY route and carry on foot, I really like Marc's idea of thinking much smaller, like a TBS Discovery, but it probably can't carry the sensor you want? Otherwise the P3 Pro is really a great value, you'll be flying instead of building and troubleshooting, and there is nothing like Lightbridge. Experienced DJI users tell me the P4 is bloated.

  • @James: that sounds like a pretty compelling argument for the Pixracer. Any thoughts on how a PXFmini/RPi Zero compares?

    @Marc: yeah, it may make sense to just buy a P4, there are a number of things (especially looking at the GCS UI/integration) and the current state of gimbals where it would definitely be a very intense "hobby" experience to get things up and running. It might be nice to have something that just works reasonably well out of the box.

This reply was deleted.