RED primes for s-3D

[istvanttt]

OK, I need some advise, I have a good understanding of space, actors in the space, how the textures of a surfaces influences a s-3D experience, know the rules of the geometry, but have a limited knowledge about optics, specifically how lenses are built.

I know (because I experienced it) that zoom-lenses can change their optical axes when zoomed. I know that ie 3Reality developed a database of offsets the camera has to do for different zoom-positions for different lenses. I heard also that others “matched” the zoom-lenses specifically for s-3D.

So here is my question: how different can prime lenses be from each other? In other words, if I order 2 sets of RED primes, should I ask RED Inc on my knees to match them? Is this question an exaggerated one, one you would never do to a producer, but let it do, if necessary, a specific lens-doctor?

I’m a strong believer that everything that can be done well on set should be done there and not in post.

Thanks for any help
Istvan

 

[Eric Lange]

I think it really depends on how precise you want to go for S-3D?

For the past few years various productions have managed to get nominally viewable stereo footage then has to be wrangled substantially in post. More recently there appears to be more of an appreciation for the benefits of more precise optical alignment and focusing for producing high quality stereo that is safe to view and does not require near impossible manipulations in post. If true 4k output is required then a higher precision is required. The aim is really to have pixel perfect epipolar lines on screen; but never the less, the human eye balls are able to make very fine motor adjustments to compensate for minor Y parallax providing the Y parallax shift is global. However things get more complicated when the Y parallaxes are not uniform and are as a result of mis-matches in magnification and radial distortion coupled with adverse relative camera orientations. The eyes cannot compensate for such conditions and headaches from eye strain start to creep in regardless of how the depths in the scene are managed.

I think it also depends on how you want to use the lenses. On the one hand Zooms could be good for quick and “dirty work”, providing you have real time stereo visual feedback, the advantage being that it is theoretically possible to compensate for miss-matches in magnification. However if the scene requires a smooth zoom-in or out and follow focus then this is very difficult to coordinate to a high precision. Depending on the zoom lens used, and the effect required it may indeed be necessary to “map” the lens especially if the zoom and focus movements cause a vertical displacement of the optical axis. The global Y offset can be fixed in post, but having pre determined curves for that gets the job done more quickly and precisely. In stereo photogrammetric applications we try to stay well away from zooms as they are nearly impossible to fully calibrate.

Lower end primes should be easier to use than zooms for precise work, providing you can adjust the back focal distance in a precise way (an autocollimator is handy for that purpose). With the focus rings locked off and things adjusted for good visual alignment trough real time feed back everything should be O.K. However as soon as you need to track and pull focus etc. things become more difficult as there is no guarantee that the left and right lenses are perfectly matched for magnification after a focus pull. Using standard servo controlled follow focus, the mechanical linkage via gear wheel, plus the mechanical backlash of the lens helicoids and gearing throughout this system, it is not very likely that you will get perfectly matched imagery after you do successive focus pulls. In old school stereo photogrammetry we like to know effective focal length to within 10 micron or better. In a standard cinematic set up I would be very surprised if you could get better than 50 micron. The Cinematic primes should be mechanically more stable and the optical axies should not drift vertically more than 25 micron (so 4 pixels), which will have to be fixed in post or left as is. Even If you were to couple high precision encoders to measure the rotation of the focus ring there is still an indirect linkage to the components that are actually being translated, and backlash could still be a problem for high precision work; the lens could be taken apart and adjusted to remove backlash, but this will shorten the life of the mechanical components and could cause problems at extremes of temperature. The Ultra Primes might be great for S-3d as breathing is compensated (constant effective focal length) and ZEISS claim a “backlash” free system that travels very straight, but they are hellishly expensive to purchase outright.

If I was building a lens focusing system for precision stereo work on the REDOne, I would be tempted to use a Schneider 2.8/28mm Digitar lens (this is a wonderful piece of glass) . This lens would be fitted to a precision translation stage and drive unit, with a hand held (closed loop) servo control system. The back focus for left and right lenses would be set using an autocollimator. Both lenses could be translated to 15 micron accuracy or better, and the crossed roller bearing stages are straight to within five micron or better. More advanced alignment techniques can be used to make the optical axis truly square to the sensor. In such as system assuming rigid mounting you can stay pretty much on pixel. What would be nice, is to deliberately introduce position and focus errors to determine acceptable tolerances for focus and alignment in S-3d work, and to determine how such a system could be downgraded in a controlled and informed way.

If you really need Zooms to zoom then getting a ‘matched” pair would be helpful, as far as primes go I think that there is so much other fiddling about in stereo set ups (especially if you are using “mirror boxes”) that I don’t think it would be necessary to get matched primes for current S-3d work. If you need greater precision, then lower end cinematic primes are perhaps not very suitable, and I would recommend that a purpose built focus and lens system be used instead.

It would be nice if there was an option on cinematic primes for a high precision internal encoder that could measure the actual translation of various mechanical elements inside the prime, this would make it much easier to calibrate cinematic primes for precision S-3d work.

That’s my two cents,

Cheers,:beer:

Eric

 

[Stefan Christou]

:couch:

 

[Eric Lange]

:couch:

How is Brixton these days?

I went to some pretty wild parties in the mid 80's, has it become all developed and gentrified like Notting Hill, or does it still have an air of danger about it? ( I live in the states now)

 

[istvanttt]

Eric, like always your contribution is excellent and very helpful!

As you know I believe that one of the most exciting part for a storyteller is the empatical feeling which the s-3D experience of the texture of materials and skins can give to a movie. True, todays theaters are still to dark to show that but I'm optimistic that in 1-2 years it will improve. So I prefer to plan already now for that precise shooting, even if...
Thanks
Istvan

 

[Eric Lange]

So I prefer to plan already now for that precise shooting, even if...

Istvan

even if...? ( I think I have an idea what the "even if" might involve)

(thanks for the much too kind words)

Cheers,

Eric

(I guess folks that are working on current stereoscopic productions are sworn to secrecy?)