Canon 5D Mark II for shooting stereoscopic movies

[BrianF]

Shooting stereoscopic (3D) footage is foreign to most indie filmmakers. But familiarity with 3D shooting is more common among still photographers. Even amateur still photographers have been playing with DSLR stereoscopic images for years. It's a genre of still photography with its own gear, clubs and websites. In many ways, full-frame 35mm DSLR stereoscopy seems more refined as a methodology than motion-picture 3D in any presently used form. It's certainly more accessible to the non-billionaire.

And now DSLR cameras are becoming video cameras. How hard will it be to adapt the still 3D methods to a moving image? They already have twin camera bars. They know the effects of combinations of different interaxial distances, subject distances and focal lengths. They use a remote to trigger both camera shutters at the same time.

I'm certainly not claiming that all of 3d photography's gear and techniques will translate to motion-picture videography easily. But the still-3d enthusiasts have blazed a trail that would seem to make things easier than, say, inventing the 3DVX, which delivers standard-definition images and which you can't own for any price.

Price of a DSLR-based 1080p 3D package? Two of the new Canon 5Ds with brand-new, top-notch 50mm primes would be a grand total of...less than $5600. Add a pair of 85mm lenses for another $800. This package would give you the two most-used focal lengths in 3D photography. Add a twin-bar tripod mount for $100 or so and you're still under $7K.

My HVX kit cost more than that. And it's missing a dimension.

No way is the path to 3D going to be as easy and cheap as the above fantasy makes it sound. There's no apparent system of synchronization past the initial trigger -- and will that even work for triggering the video-record function? Even if you can trigger both cameras at once, they will drift out of sync eventually, putting a time limit on any single shot (perhaps doubled with a tail slate). Minimum interaxial length may be a problem, forcing the use of longer lenses. Post production is a mystery to me. And then there's audience: Who is going to watch your movie and how?

But that's 3D in 2008. It's either a mess or a fortune no matter how you approach it. But if a masochist were to choose to navigate this mess, could double-5Ds be the best possible low-budget route?

 

[Matthew Bennett]

My HVX kit cost more than that. And it's missing a dimension.

Ha! Great, to add to the HVX's woes it is now also 'missing a dimension'....

HVX: "Why are looking at me like that.."
MAN: "I hate to tell you, but you are missing a dimension."

 

[Lee Wilson]

No way is the path to 3D going to be as easy and cheap as the above fantasy makes it sound. There's no apparent system of synchronization past the initial trigger -- and will that even work for triggering the video-record function?

I can trigger my HV10 and HV30 perfectly in sync with a single infa-red remote.

Even if you can trigger both cameras at once, they will drift out of sync eventually

Why do you think this ?

Do you expect one 5D mkII (for instance) to record at 30.194 fps and another to record at 29.896 fps ?

I would expect them to record at exactly the frame rate they claim they record at. (?)

 

[BrianF]

Lee,

Encouraging info on the simultaneous trigger.

As far as slipping out of sync, I believe that any two independent, unconnected devices playing or recording temporal media will eventually go out of sync, even if they are digital. For example, two identical music CDs started perfectly at the same time will have drifted some small amount out of sync by the end. Why? I don't know. But I do know that engineers concerned about absolutely perfect sync assume this phenomenon will occur. A perfect start does not mean sync will remain tight forever.

Imagining a two-minute take with the Double-5D, I doubt that the file recorded by Left Eye camera would be substantially out of sync with Right Eye camera by the end. You could probably play the L and R soundtracks together and hear little or no telltale reverb effect. There would probably be no lip-sync issues. But I believe with 3D video the tolerances are much lower than we're used to caring about. The audience apparently gets a headache even with a small amount of difference between the two eyes. I don't know how much drift it takes to cause a problem, though, or how quickly a take would reach that point.

 

[Lee Wilson]

For example, two identical music CDs started perfectly at the same time will have drifted some small amount out of sync by the end. Why? I don't know. But I do know that engineers concerned about absolutely perfect sync assume this phenomenon will occur. A perfect start does not mean sync will remain tight forever.

I have to disagree here - a CD playing at 44.1khz will play exactly at that sample rate - it will not shift, slow or speed up.

Imagining a two-minute take with the Double-5D, I doubt that the file recorded by Left Eye camera would be substantially out of sync with Right Eye camera by the end.

Unless you put forward a reason you think this might happen I can't see any reason to think it would.

You could probably play the L and R soundtracks together and hear little or no telltale reverb effect.

?

The L & R channel of either camera will be perfectly in sync with itself.

I am not sure what you mean by 'no telltale reverb effect' ?

Perhaps you mean that the L&R channels from one camera when played against the L&R right from a second camera - that is slightly out of sync - will exhibit 'phasing'. (and if really out of sync - by a whole waveform for instance - will exhibit attenuation, especially of the low frequencies/bass.)

There would probably be no lip-sync issues. But I believe with 3D video the tolerances are much lower than we're used to caring about. The audience apparently gets a headache even with a small amount of difference between the two eyes. I don't know how much drift it takes to cause a problem, though, or how quickly a take would reach that point.

You have still not put forward the reasons you think they might slip out of sync.

 

[Nicky]

Wouldnt it be more appropriate to use a camera that provided wide DOF so that everything is in focus, so that the viewer could choose their own z-space focal point?

I would imagine a narrow DOF 3d shot to be kinda pointless... no?

Im still learning

 

[BrianF]

Lee -- okay.

Nicky -- I believe you are right about shallow DOF presenting a potential problem for 3D. When a background goes significantly out of focus the viewer's eye apparently reads it as "flat," which means it can't go back any farther than the screen. So of the three zones you have to work with -- behind the screen, the screen, in front of the screen -- you rob yourself of one. Apparently. I'm still learning as well, so I may not have explained this correctly. But I do believe that 3D filmmakers in general do not value shallow DOF, as they have other ways to indicate depth, to put it mildly. So full-frame 35mm would not necessarily have an advantage, especially under conditions requiring a large aperture.

 

[Steve Castle]

Its an interesting idea that may be increasingly more important if 3D LCDs ever catch on.

For what its worth, Fuji has announced a stereoscopic system for "stills and movies" called Real3D, obviously it won't have anywhere near the video capabilities of a 5D, however its an integrated solution that takes care of things like such as focus, zoom range, exposure, etc, for both sensors.

http://www.dpreview.com/news/0809/08092209fujifilm3D.asp

 

[BrianF]

I just thought of another way in which the Double-5D would not be the ideal stereo camera in a simple side-by-side configuration. Ideal interaxial distance -- space between the center of the two lenses -- is considered to be somewhere around 2.5 inches, simulating the spacing of human eyes. A greater distance can work, but the measures taken to keep the image from inducing eyestrain (with a wide interaxial distance close objects would have a huge 3D effect but would cause the viewer's eyes to strain to move farther than their natural range, I think is the problem) are using longer lenses and/or staging the action farther away.

Longer focal lengths will bring up the 35mm full-frame DOF issue even more -- we actually want to lean toward the other direction to avoid shallow DOF. And staging action far away from a 50mm lens does not offer much opportunity for variety.

There are alternatives to the side-by-side approach, although they are more complicated. Haven't looked into them myself, but it might be necessary for the 5D.

 

[Lee Wilson]

I just thought of another way in which the Double-5D would not be the ideal stereo camera in a simple side-by-side configuration. Ideal interaxial distance -- space between the center of the two lenses -- is considered to be somewhere around 2.5 inches, simulating the spacing of human eyes.

Very good point.

Although I have heard (a long while back so can't remember the details) that it helps on wider spacings to go for multiples of the average eye spacing - so x1 is good, but x2, x4 and so on - can also work - but you would need to check this out.

Maybe a mirror construction on the front of the two 5Ds would help bring the 'eyes' together. (??)

 

[Psynema]

Lee -- okay.

Nicky -- I believe you are right about shallow DOF presenting a potential problem for 3D. When a background goes significantly out of focus the viewer's eye apparently reads it as "flat," which means it can't go back any farther than the screen. So of the three zones you have to work with -- behind the screen, the screen, in front of the screen -- you rob yourself of one. Apparently. I'm still learning as well, so I may not have explained this correctly. But I do believe that 3D filmmakers in general do not value shallow DOF, as they have other ways to indicate depth, to put it mildly. So full-frame 35mm would not necessarily have an advantage, especially under conditions requiring a large aperture.

Not necessarily true, as argued by a few 3d enthusiasts. The farther back you go, the more out of focus it is, the more in the foreground the less, so there's still a sense of depth and the subject of course will still be seen in 3d.

 

[Psynema]

Very good point.

Although I have heard (a long while back so can't remember the details) that it helps on wider spacings to go for multiples of the average eye spacing - so x1 is good, but x2, x4 and so on - can also work - but you would need to check this out.

Maybe a mirror construction on the front of the two 5Ds would help bring the 'eyes' together. (??)

I read somewhere as long as the subject is 30 x the distance between the two lenses away, you're good - no clue how accurate that is though.

 

[Lammy]

I'd also like to add that the d90 would be awful to do stereoscopic movies in. It's too soft.

The 5D would also be bad, because the full frame sensor and extreme shallow depth of field. Ew! Bokeh in 3D looks AWFUL, AWFUL! It's an eye rape.

k, that is all.

 

[beckspace]

Interesting, our eyes have bokeh like any optical mechanism, but in 3D movies it doesn't work because our eye's lenses dont need to focus the subject, the focus is on the screen. That's the missing part of the 3D whole thing, it gives the binocular ilusion, but the sensation of depth is not complete.

People with just one eye have perception of depth, they don't need binocular vision to see depth. and, of course, watching a 3D movie they only see a flat image.

For 3D to work well, nothing can be too close of the lens and everything needs to be in deep focus, which kills part of the 3D fun