Tau Labs Camera Stabilization

[Gary Mortimer]

Not merged yet but very soon. Some more great work from James with thanks also going to Reddog for helping him with some hardware.

So this will take two outputs and use them for servos to compensate for roll and pitch.

Probably a safe bet to say it will be in the next peoples GCS build.

 

[Bartman]

your laboratory looks like it was styled by the same guy that did my bathroom.

that's a very nice application of the CC board. i've been thinking about doing that with my new Okto to see if I can get better roll stabilization than the MK board is capable of. it always shows a tilt when you're rolling left or right.

edit: i thought the camera application was already available. looks like it's in development. carry on!

Bart

 

[Kilby]

Very nice. I was actually just wondering about this feature this morning. You guys read my mind sometimes. ;-)

 

[Gary Mortimer]

Focus your thoughts hard in the next two months....

 

[Macsgrafs]

1 problem I see with this setup...it has poor resolution! This will show on the final video as jerks when the aircraft rolls/tilts etc.

Ross

 

[Crash]

1 problem I see with this setup...it has poor resolution! This will show on the final video as jerks when the aircraft rolls/tilts etc.

Ross

Here is James' initial post that may explain the jerkiness. I think he was working on the firmware rather than dialing in a cheap mount.
http://forums.openpilot.org/topic/5426-camera-stabilization/page__view__findpost__p__17271

The roll is a little sloppy looking - it's not the code but the piano wire on the camera mount has a bit of wobble. We'll see what it looks like in flight and I'll update the video once I've mounted it to my quad.

 

[peabody124]

Yeah as crash said the roll jerkiness is the camera mount, not the code. Pitch is quite smooth. I also spent all of 5 minutes finding the right scale. I just wanted to get that out there. I think I can modify it to use a heli link and ball joint to have less slop.

 

[Bartman]

In theory the CC board should hold the camera level in roll and pitch right? it seems that the MK camera controls don't do so well in roll.
anyone?
thx
bart

 

[Droider]

If this can keep my camera level while I am tracking sidways ill have one! (another one).

Be good to see how you have set it up. This is mine and many others issue.

http://www.youtube.com/watch?v=hsEVz_ehXoM&feature=mfu_in_order&list=UL

DAve

 

[peabody124]

well that's definitely not how it should look IMO. Hopefully we won't have that but if so we'll try and address it. Does roll work properly when facing forward?

 

[Droider]

Yep.. It seems to be when I fly away from take off and then yaw round to film.. well like the bikes! Filming mountain biking is what I got in to this whole thing for and although you may not always be able to tell the mount is not correct it would make the hill look remarkably steep or very tame!

If I can mount a CC board to take care of the just the roll Ill be a happy man. The nick is fine.

DAve

 

[peabody124]

Well I took it out and tested it today and I'm QUITE pleased with the result. There's vibes due to my crappy mount and the fact the quad was dragging almost a pound of lumber as a makeshift landing gear.

 

[peabody124]

Ok video update to include actual flight. I don't think there's a roll problem. There is a vibration problem due to my mount You don't want to see the flights where the screw came out of one side!

 

[DennyR]

I'm getting to be very interested in Open Pilot as I think they will emerge as the leader in FC technology with their modular INS technology. These guys go about it the right way.
The CC is actually doing as fantastic job of stabilisation in that video. The problem is that the camera mount is not mounted in an idea way. For every action there is an equal and opposite reaction so every time the servo inputs a strong signal there is a slight deflection. The key to great stabilisation is keeping everything very slow. you need about 5:1 gear reduction and hang your mount system with the soft mounts as far away from each other as possible. It's looking good thus far!!

 

[DennyR]

Purely for information that the developers may be interested in I use an IMU that has .1 us resolution and a 20 deg/sec gyro that drives a G2 hitec servo board running in continious rotation feedback mode. 12/100 reduction. 7.2 volts.

 

[CopterCam]

Excellent work Guys !............... wish I could contribute.

If you need a crash tester.......... I'm yer man.

Sid

 

[peabody124]

DennyR,

Definitely my setup is very suboptimal. I feel bad underselling CC that way cause I'm sure it could lots better. It's just not my thing enough to spend a lot of money on a good camera rig. Now the code is out there hopefully others can produce much better demonstrations of what it can do Or at least the code will be in the next release, just select an Camera Stabilization from a list of optional modules.

 

[Bartman]

Purely for information that the developers may be interested in I use an IMU that has .1 us resolution and a 20 deg/sec gyro that drives a G2 hitec servo board running in continious rotation feedback mode. 12/100 reduction. 7.2 volts.

Denny
I see a few words I recognize but could you please translate your post for us mortals?

At what point does the CC board become ready for prime time in terms of us using it for camera stabilization?
thanks
bart

 

[peabody124]

Denny
At what point does the CC board become ready for prime time in terms of us using it for camera stabilization?
thanks

On a quad I would say it already is, although we won't package up a new version for another week or two. The mechanical structure is independent of our software.

For hexs and octo's, Sambas has been doing some work do use the PWM inputs as outputs but that's probably a bit further out. That gives up to 12 outputs.

 

[DennyR]

Well at the moment Bart. I see CC as a great little board that offers really good stabilisation out of the box. The camera outputs can only be as good as the camera mount that accepts those outputs. That is the limitation of the product at this time. We cannot expect anymore from the developers until we define exactly what type of camera mount we are using. i.e. is it direct drive or reduction gear, are the servos running in proportional mode or are they continuous rotation feedback mode, servo resolution is also like chalk and cheese. etc. From what I have seen I believe it is a good deal better than MK's 8 bit offering.

I am soon going to release a part two of my video series which will explain in quite a lot of depth how one can take stabilisation to the next level. But we must consider that it is not everybodys priority to have an expensive and quite heavy pro system. Peabody124 should not feel bad, because he is very typical of the vast majority who are happy with a simple set-up that can be tweaked and fixed to improve as it goes along. That is all part of the fun. 99% of all multirotor videos will never be shown on anything beyond vimeo etc. With improvements in post stab. software, the number one priority should be focussed on dealing with the vibration issue so that the new cameras out there can do what they are getting very good at. (In camera Stab.) The nuts and volts of my pro system will soon be revealed in part two of the video series and the jargon fully explained with some interesting testing footage. As camera technology develops it will reach a point where this level of stabilisation can also be done via another route. providing you deal with the number one problem first which as always is vibration.
I would have had part two posted a while back but I have discovered what I think is something almost ground breaking which relates to aerodynamic stabilisation of the model itself.
Consider this, If you want to move say to the left, then you initiate that move with a bank in that direction. With my new system, as power is increased to one motor then a condition, known as the Coander Effect, induces flow onto another aerodynamic device known as a Gurney (all race cars have a gurney) and that device pushes the model sideways with minimal bank angle. Wait til u see FMC-4