That is very clever rig Frank. The Canon HF21S's are great choice I think.

For those of you not against the stereoscopic glasses, Zeiss is coming up early next year with what could be really cool 720P eyewear, that has cool interfacing and OLED display technologies for about $500. IFA 2010: Carl Zeiss Presents New 720p OLED 3D Glasses - DigitalVersus

Thanks Pavel.
I primarily use the rig to shoot 3D video of the local wildlife.
The biggest problem is keeping the cameras in sync within about 4 or 5 ms.
Unfortunately, the HF21s tend to start quite a bit out of sync when you power them off and then on and stay that way no matter how many power cycles you give them.
I am working on a replacement for the Lanc Shepherd that gives me a lot more control including starting them at slightly different times to compensate for this.
It will also have manual focus which the lanc shepherd doesn't support.

It's about time somebody made them!

That sounds very interesting. Do you have to do a lot of zooming to capture wild life from a tripod? Do you need to re-space the rig? Are you going to share the design or manufacture the controller? Are you planning to incorporate an "in-sync" indicator? We really need a genlock port. It is not so hard to implement. Every old analog TV had to genlock to syncs everytime one changed channel or a source, so it is not such an amazing or exotic "professional" feature. It would probably cost dollar or two to implement in volume and they could charge $50 for it. It would sure convince many of us buy new camcorders (two at a time - Sony, Canon, Panasonic marketing, please wake up).

BTW, you probably know this, if the power up is consistently off, the problem could be in the camcorders, their H/W or even F/W if it is different or batteries. If for some reason one has bigger filter caps or power-up reset sense threshold or is otherwise different, you could have consistent discrepancy. The Lanc Shepard and similar controllers depend on consistent components. I know that Werner Bloos was matching camcorders for the pairing purpose, and usually wanted to try at least six pairs. If you have a chance, you may want to see if you can try to power up with a different camcorder, store demo, or friend, etc. My camcorders definitely do not power up in sync every time, but most of the time I am within .5 ms, which is good enough for running water, flopping wings, etc. Of course if you design your own, you should be able to scatter the power-up commands. I believe that it is what the original LANC Shepard did as well.

"Do you have to do a lot of zooming to capture wild life from a tripod?"
I wouldn't say a lot but some for sure.
"Do you need to re-space the rig?" No, but I plan on building a new small beam slitter unit with a servo so I can change the interaxial distance remotely also.
"Are you going to share the design or manufacture the controller?"
Too early to say. At the moment, it's breadboarded with four processors with the main processor running an Apache server. I don't see it as a hand held unit for now.
This is what it is going to do:

1. turn cameras on
2. turn cameras off
3. start recording
4. stop recording
5. set zoom speed
6. zoom in
7. zoom out
8. set auto focus
9. set manual focus
10. focus near
11. focus far
12. set max sync error in milliseconds
13. enable auto resync when max exceeded
14. disable auto resync

I don't want the camcorders to start in sync, I want the slower camera to start first, in my case that would be around 2 milliseconds before the faster one.

Frank I have used servos to control some lipstick cameras with marginal success but the same servo system wouldn't support the weight of a conventional camcorder. Are your servos only controlling the toe in or do you also adjust pan and tilt? If you're controlling the pan/tilt can you tell us what servo you are using that can handle the weight of two cameras? We were a bit disappointed with the amount of adjustment we got from the servo, not nearly fine enough to really aim the cameras to the same point unless they were perfectly aligned to start. We were using separate servos for each camera thinking we could adjust them to the same target using the controller but that didn't work. The Huber mount that someone else has mentioned would be nice but they are extremely expensive.

Pavel if you were starting from scratch with the intention of building this type of stereo rig what cameras would you use? I am thinking that block cameras might be the best way to go but I am not familiar with their video quality and capability, but their design looks right for this type of project. Also it would seem that progressive image capture would be of great benefit for stereo video along with the external sync of course. The Sony FCB-H11 would seem good but it is interlaced and no ext sync from what I can tell. Once you get a block camera with ext sync the price gets out of hand.

"Pavel if you were starting from scratch with the intention of building this type of stereo rig what cameras would you use? I am thinking that block cameras might be the best way to go but I am not familiar with their video quality and capability, but their design looks right for this type of project. Also it would seem that progressive image capture would be of great benefit for stereo video along with the external sync of course. The Sony FCB-H11 would seem good but it is interlaced and no ext sync from what I can tell. Once you get a block camera with ext sync the price gets out of hand. "

The camera choice is highly dependent on your needs, e.g. size and weight, portability, targeted results (movie theaters vs TVs vs computer screens), close-ups, general videos, landscapes, etc. I like portability and fairly quick preparedness, so I would use the HDR-CX550s and Werner's latest controller ( digi-dat Dienstleistungen - Anlagen - Technik ). But I can see why others would use other brands. The choices of camcorders would be between the the Panny HDC-TM700 (sync?), Canon HF-S21 and Sony HDR-CX550. I would pick the Sony. With the right spacing (the Sony being fairly narrow) one should be able to look into both viewfinders at once and skip a lot of baggage. The Sony only has limited shutter control and only 60i at full HD, but has other advantages (you may wish to see this discussion: (Yahoo! Groups ))

I am not familiar with the Sony FCB-H11, but it seems to have only analog outputs, that would be a big hassle.

There are times I could use mirror rig, but fairly rarely. To me it takes too long to get ready, but I will probably do one sometimes in the future. It's got a lot of opportunities to do things wrong, specially if you have to shoot quickly, but the only way I know of to get really close and have flexibility.

The only servo I am using is the toe in out servo.
The pan/tilt is a standard inexpensive unit that I am controlling with a phidget 8 relay module.

Thanks for your thoughts Pavel. The FCB-H11 does some sort of LVDS digital but I will confess I don't know what that truly means. Also a conventional camera has internal storage although I have the opinion that you are capturing the video streams on your laptop. I am still concerned with the use of interlace capture for 3D work, I may not understand how the interlace is actually captured in the camera but I would think that the time lag between fields (probably wrong term) could lead to false depth clues in the image. I will look at the yahoo link you provided, maybe it will be more clear how the interlace impact is mitigated.

Thanks Frank that pan/tilt looks pretty nice. I am going to assume it has the same level of control as a standard, inexpensive servo, about 1/2 degree of movement with each increment so using it to adjust toe in on a per camera basis is not practical. The 6lb capacity is certainly a big plus compared to the tiny pan/tilt heads I have used with small servo motors.

LVDS stands for Low-Voltage-Differential-Signaling. That is a serial digital stream. SDI and HDMI are also pixel serial digital uncompressed streams, so you might be o.k. with that output, depending of what you like to interface the camera with. No doubt they would have some sort of recording interface.

In general, interlacing doesn't do any harm to stereoscopic cues. That is what my cameras are. Unless you capture in progressive fashion, there is always de-interlacing taking place somewhere in the chain, because there are not many HD interlace display devices left, and none for 3D, as far as I know. Even TV's de-interlace the i-stream first, convert everything into the native progressive scan, before sending the display controlling stream to the panels. The only interlaced displays I can think of are deflection driven = CRTs - mostly gone. Interlacing was invented mainly to deal with decaying phosphors in CRT displays, to avoid flicker. Later on it was useful to give smoother motion with limited bandwidth. These days it is used to get full 1080 lines with less bandwidth requirement, specially for MPEG2 compression and VSB modulation, in case of ATSC broadcast in the US, but not directly used by the TV display(panel).

For camcorders, specially those at consumer prices, storing the HD stream on Hard Drive is much easier to accomplish, as well as the compression ( and on playback to de-compression) at half the bandwidth provided by the i-stream.

Of course computers and movie projectors also need progressive stream, so everything will be converted at some point.

Depending to what you are doing, the interlace will actually give you the equivalent of 60p. For example if you are distributing above/below stream, you can just use the individual fields and end up with "de-interlaced" 60p a/b stereo stream. As far as quality goes, there are couple of adaptive de-interlacing technologies, that yield excellent results.

@ 1080 lines, if you will be interfacing the playback computer player to a TV directly, (in the US) 60i/p capture would be just fine.

HDMI 1.4 deals with the limits of 1920x1080p frame packed stream high bandwidth by limiting the scan rate( of the pair of frames) to 24p. Of course no TV works in 24/48p, so there is always scan rate conversion required. If that is not done correctly, the 3D cues are completely destroyed. That is one of the reasons why specific 3D TVs, dealing with these conversions properly, are required.

If you will be targeting 3DBD or movie theater, you are best off with NATIVE 24p capture.

My beam splitting mirror will be here tomorrow and I'm excited to start working on the new rig.
Should be interesting trying to design and build it in my garage with only hand tools.
After running a large machine shop for over 30 years I enjoy the simplicity of trying to do it this way for some reason.