Better CMUcam2 Mount
 

I was just fixing a bent PWM for the millionth time when I actually looked at the CMUcam itself and realized something. Why the heck is the whole board mounted on the moveable camera mount?! This is big, clumsy, and places the PWMs right infront of the lens. I noticed that the little camera sensor itself can detach from the board, and be plugged in with a 32 pin ribbon cable.

So my question, why don't we mount the board (with all the serial and power connectors on it) to a stable part of the robot, and just the camera (with a ribbon cable link) to the swivel stand itself.. Wouldn't this make much more sense?

Re: Better CMUcam2 Mount
 

My guess is that the chief reason the IFI mount is done that way is for cost reasons--32-pin ribbon cables have a nonzero cost, making it cheaper for them to mount the sensor directly to the board.

Of course, there's nothing stopping you from going that route should you desire. The main concerns I could see involve interference on the cable (as the CMUcam already faces a bit of motor interference as it stands) and the general concerns of another point of failure for the camera (the cable can break or get unplugged). Welcome to the world of pros and cons.

Re: Better CMUcam2 Mount
 

I think the main issue it that, while a ribbon cable would bend pretty easily in the up-down direction. But getting it to twist so you can do the panning motion.. I think that would be a good bit more difficult.

Re: Better CMUcam2 Mount
 

I suppose that problem depends on the way the cable (and, indeed, the mount) is done. You could just give the cable enough slack to twist easily when the camera pans, then use a bit of force to keep it from flapping in the breese (a la 233 this year with the wires to their grabber--load the high-res and look at the front side of the arm). Alternatively, if the main goal is just to give the main board some breathing room, why not mount the board so it pans as well? I don't think there's anything stopping you from mounting the board to the back of the IFI mount, or there's always doing what WildStang did in 2006 (according to the Behind the Design book) and using a turret to handle that part of the problem. Command the turret to move, and board and lens move happily together.

(Now, watch--someone's going to come along with a 32-pin cable that can be twisted all day. I suppose that four Cat5 cables together would do the trick, even if it'd be particularly heavy and time-consuming for the task.)

Re: Better CMUcam2 Mount
 

Did you see this?

http://www.acroname.com/robotics/par...M2-TURRET.html

I didn't see it until AFTER we shipped this year...we might check it out next time.

Re: Better CMUcam2 Mount
 

didn't see this until just now...but it should be easy to make ribbon cable bend both directions, if you just fold the cable over at a 45 degree angle, like they used to do on the original IBM PC floppy drive cables. (showing my age again)