A Simple Direct Drive Multicopter Camera Gimbal

Here's a short hi-def video of a tilt/roll camera gimbal that's quick and easy to build, using parts from a nifty line of unique CNC'd aluminum hardware developed and marketed by Servo City (servocity.com) of Kansas and used mostly, it seems, for terrestrial robotics. (I hasten to add that I don't work for them etc., just want to pass on what I think is some really cool stuff.)

Before continuing I'll mention that I am aware of the drawbacks of driving camera gimbals directly with servos (more on that later), and I am also aware of the negative effects of the more or less random geometry employed here. But I'm not looking for professional results, but rather just a solid, simple mount for basic video that won't make the viewer seasick.

I don't like any of the cheap box-frame style gimbals I've been playing with for several years (on trad helis and now on multicopters) They tend to be rickety, sloppy and fragile and all look like they were cobbled together by a ten year old with an incomplete erector set. Yes, I know there's good stuff out there and I also know what it costs.

While experimenting with some alternative ideas, I recently ran across Servo City's modular servo mounting blocks. Two of these simply bolted directly together, with the camera attached directly to one of them, provides direct tilt and roll (a third one could provide pan, if desired). The structure is compact and rock solid with zero play or flex and depending on the servos used could handle a much heavier camera than my cased GoPro 3. The hardware connecting the assembly to the airframe is also from the same source, off-the-shelf, and no machining or other complications were necessary. All the pieces have (sometimes multiple) identical hole patterns (some tapped 6-32, some sized to pass a 6-32) so there are many ways things can be put together with no need for drilling etc.

From the top down I used an aluminum disk (again with all sorts of predrilled/tapped holes, which makes for all sorts of options for mounting it below the airframe), a tube-clamp fitting bolted to that disk, a short aluminum tube, then another tube-clamp with an integral extension having, again, the same bolt pattern, to which the first servo block is mounted.

These servo mounting blocks are the heart of the matter. They are made in versions for HiTec and for Futaba standard (full) size servos. There's an even larger version, probably not of much interest to us, but unfortunately they don't offer smaller ones, which might be useful here. They look expensive (at least to anyone who has bought trad heli or r/c car CNC upgrade stuff) but they're not. Rather than trying to describe them further here, I suggest having a look at them at servocity.com and in the photos in my video.

So. The second servo block is bolted to the first at a right angle, and the camera is bolted to the second servo's output shaft/flange. With this arrangement the first servo provides tilt by rotating the second servo and the thereto attached camera, while the second servo provides roll.

The camera is a GoPro H3 Black in its hardshell case (I refuse to fly my GoPro out of the case, tempting though it is). The case is attached to the output flange of the servo block by four 6-32 screws passing through holes drilled in a spare “deep” GoPro case back, the kind that comes with rear-mounted accessories. (In an earlier version I had mounted a thin aluminum plate to the servo output shaft and then used an industrial-grade Velcro-like material to attach the case. I didn't like that so much, next thought about gluing but since I had several deep backs on hand I ended up drilling one and bolting it on.)

I'll mention the servos briefly (as this is a topic that can go on forever). The servos de jour are HiTec analogs, model HS485HB. This is the third model I've tried, and so far gives the best results. I started with HiTec digitals, model HS562MG. The results were not good, with a lot of jitter in the video. Next I tried a pair of very inexpensive HiTec analogs, HS322HD. These were smoother but still pretty shaky. On the advice of a Servo City tech I installed the analog 485s, (cheap enough at under $20) and they are noticeably smoother than the others. I would like to try a pair of the new brushless Futabas, but they are over $100 a pop and I'd also need to buy Futaba-compatible servo blocks ($26 each) so I'm hoping someone else will try these and let us know how they are. From all of this it might be concluded that analog servos are categorically smoother than digitals, but I don't claim to know that, I just know what's happening here at the moment. And as indicated in the beginning, this is not a path that will likely lead to perfection, but it looks like with a little more tweaking it might deliver quite good results.

Which brings up another issue, one that's not limited to this setup, regarding the gimbal outputs on APM 2/2.5. As has been noted elsewhere, the supposed “fast” (490hz) outputs (RC8 -7) don't seem to work at all. Everything I've done, digital and analog, has been off of RC 11 and 10, which are supposedly running at 50hz. I don't know if the faster outputs would make a visible difference. It would be nice to be able to find out and I'm hoping this will be possible after the next firmware upgrade.

Folks to whom light weight is a primary issue probably won't like this setup. This hardware is robust, it's not designed for aircraft. But it's not all that heavy. Personally, I don't have a problem trading some flight time for the extra stability and smoothness that comes with weight, especially when flying manually.

Random Notes:

The 6-32 washerless Phillips-head (yuk!) screws supplied with the above referenced hardware are subject to backing out if not dead tight. Use Loctite.

If you want to replicate my setup and can't figure out exactly what to order PM me on DIYD and I'll give you a list.

Comments, suggestions and questions (other than “Why the f. would anyone do that?” ) are welcome.

For the record, the gimbal is currently mounted on a DJI Flamewheel 550 Hex (“The Witch”) with upgraded T-Motor model MT2216-11 900 Kv motors spinning Graupner 10 X 5 inch carbon “E-Props” and driven by 30 amp “Opto” ESCs which in turn are powered by a parallel pair of 3300 mAh 4S Turnigy “Nano” Lipos. Control is through a JR 12X Tx to a JR 921 Rx with two satellite antennas, feeding an APM 2.5 w. Ublox GPS. Electronics are powered off the mains via a 20 amp Castle BEC. Camera gimbal servos are powered via a separate 1200 mAh 2S Lipo (overkill) and a 10 amp Castle BEC. Two strips of ultra-bright LEDs, red and green, are powered directly off the mains and provide orientation. Weight is 6 pounds 12 ounces. Flight time as configured is 10 minutes with a 3 minute reserve.

I hope all of this is of some use to someone. I'll update as appropriate.

Have Fun & Fly Safe!

Views: 5762

Tags: Camera, Gimbal, Servo


Developer
Comment by Rob_Lefebvre on March 19, 2013 at 10:13am

Wow, congrats, that is a solid result!  I looked at using these things, and drew up a bunch of permutations of ways to assemble them, but was put off because of the typical statements that direct driver servos simply cannot work. Your result is pretty good!

I'm also a fan of using more weight for a stiffer structure, and thus better performance instead of going for ultimate light weight.  All the flight time in the world is useless if the resultant video is crap.  I have looked at every gimbal on the market, and am not happy with any of them, they are all flimsy junk IMO.  Well, except for the Zenmuse which is a really nicely made CNC construction.

You don't seem to have much vibration either.  Nice work.

Comment by Gary McCray on March 19, 2013 at 11:32am

Excellent video Oliver,

Amazingly good video from an extremely simple and inexpensive but very robust mount.

I have seen Oliver's setup and these little "Servo Blocks" from Servo city are truly a wonderful building block for us.

http://www.servocity.com/html/hitec_servoblocks__plain_shaft.html

They are machined out of aluminum and are as light as they could be while preserving maximum strength and flexibility.

The key to their success is a captive ball bearing race that maintains complete rigidity while permitting friction-less and fluid rotation.

I am planning on using these for scanning for SONAR, IR and LASER ranging platforms.

Oliver now has the simplest mount you can make with them and it works very well.

It is also possible to easily add a few aluminum or fiberglass plates or angles to produce a focal plain centered camera motion. 

Also, as Oliver says, at this point the main limitation for smoothness appears to be the servos themselves, or at least the way they are handled in our existing firmware. 

On his Hexacopter, the 2 high speed servo channels do not function at all so he is forced to use the low speed channels.

And in Digital servos that actually seems to result in more "jerkiness" than analog ones, because the digitals respond fast enough to actually stop and start between PWM updates.

As I understand it, even if the "fast" channels did work, the actual APM updates speed is still only 50hz which would still permit the digital servos to start and stop between PWM updates.

Even though that is really fast", the accumulated inertial forces show as a "jerkiness" in the actual motion.

That is why right now analog servos (with tight gear trains) are better.

In any case check out Servo City they have some really great stuff.

Comment by Oliver on March 19, 2013 at 10:22pm

I've gotten a couple of requests for a parts list for this gimbal. Here's a sketchy graphic with the various bits and their numbers and current prices. Note the the servos I'm using are now out of stock.  The vendor/manufacturer is ServoCity.

Comment by Hugues on March 20, 2013 at 12:19am

Thx Olivier for sharing. Can you explain how you connect/bolt the two servio blocks to each other ? It doesn't seem to have enough room to bolt anything ?

Comment by Paul Braun on March 20, 2013 at 4:19am

Oliver - thank you so much for posting this.  Very good stuff. I haven't any experience on that front yet (just learning to fly) but I'm building a hexa soon and would like to take it to this level. 

I'm curious, how are you operating the 2 servos to keep the camera stable while still making the hexa roll and pitch back and forth?  I can't visualize how you handle all those things from the radio.  Do the servos with the camera work automatically in some way?

Comment by Hugues on March 20, 2013 at 5:50am

@Paul, I can answer about the servo control : APM has a automated roll/pitch stabilisation built in on ports A10 & A11 (or you can even use other servo outputs). Very easy to setup, take a look there at the wiki:

http://code.google.com/p/arducopter/wiki/AC2_Camera?wl=en

 

Comment by Oliver on March 20, 2013 at 7:37am

Hugues, here's how the servo blocks are connected:

Comment by Hugues on March 20, 2013 at 7:56am

thx. And how do you isolate the Alu mount plate from the quadcopter against vibrations ?

Comment by Oliver on March 20, 2013 at 2:30pm

Here's what I used for mounting the gimbal assembly to the Flamewheel 550 Hex and for controlling vibration.
This is a rail-mount structure intended to be used with a regular little boxy gimbal mount, usually placed sticking out forward on the aircraft. The four crosspieces are mounted to the rails by rubber grommets encased in U-shaped plastic housings that have threaded metal inserts for the included mounting screws. The crosspieces can be positioned wherever desired on the rails  The two most narrow crosspieces are for attaching the whole thing
to the bottom plate of the airframe. Rather than bolting these to the airframe I use 3M Dual Lock pads, each
backed up with a zip-tie running around the rail and through existing slots in the airframe bottom plate. This is
very secure but can be  removed or installed very quickly.

The aluminum disk at the top of the gimbal assembly is mounted under the slotted crosspiece by means of four
"sandwich" vibration mounts. So, there are three points at which vibration is dampened: the Dual Lock, the grommets, and the sandwich mounts. I see no jello or other signs of vibration in the videos.

The landing gear is inadequate for a Hex with a load, the long curved arms flex a lot. I'm still using it but am working on something better that will be mounted on the airframe. Meanwhile a crossbrace between the skids would help, especially if you haven't yet developed a light touch in landing.

Here's a link to the relevant page at Goodluckbuy.

Have Fun & Fly Safe

Comment by Hugues on March 20, 2013 at 2:56pm
Thx, very nice triple vibration filtering/ absorption

Comment

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

Join DIY Drones

© 2014   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service