Camera Converged or Parallel?

[Rich Schaefer]

I use Marshall OR-70-3D due to the size limitations, but it does the job.

Its interesting you say that. I was evaluating that very monitor last week, but I returned it because it can not do a basic convergence and therefore no parallel 3D measurements... but I really liked everything else it did! EDIT: Unfortunately, both tech support and the sale rep felt skeptical that Marshall would fix or add the ability for parallel shooting with some in monitor convergence. It might be added if customers speak out and request this option. My Panasonic dose it but needs Decimator boxes to work. I know transvideo dose it.

 

[Pawel Achtel]

hey, Rich

it has other ways to skin the cat:
1. Checkbox, with size of the box adjustable
2. H marker - a ruler for 3D depth control

 

[Rich Schaefer]

Pawel, true, I figured out those functions, but how would you measure positive and neg parallax if cameras are parallel and not converged, if the monitor is not doing a convergence?

 

[Maciej Orman]

Pawel, true, I figured out those functions, but how would you measure positive and neg parallax if cameras are parallel and not converged, if the monitor is not doing a convergence?

That is simple: All what is in front of the rig up to infinity will be negative parallax and there will be no scene geometry in positive parallax.
Also means that stereo window will be at infinity.

 

[Pawel Achtel]

Maciej beat me to it
Yes, you only have negative parallax when shooting parallel. Positive parallax is always zero, making it easier to deal with only the total parallax limit.

 

[Jason Goodman]

Maciej beat me to it
Yes, you only have negative parallax when shooting parallel. Positive parallax is always zero, making it easier to deal with only the total parallax limit.

This is a potentially confusing discussion. Shooting parallel does not limit you to negative parallax only. Shooting parallel without HIT does, but no one would do that. If you want to shoot parallel properly get a monitor that does HIT. There are many.

Also, I must add, lens shift is not analogous to the human eye movement and not a flawless solution to this problem. The lenses in our eyes are designed to rotate around their nodal points and our eyes never get closer together or further apart. Shifting lenses moves the entrance pupil and changes your interaxial. Shifting the sensors or the active area of the sensor behind stationary lenses would be the preferred method.

A good rule to keep in mind is that most things 3D related that are optical or mechanical have already been tried, in many cases 100 or more years ago. There is a reason why people use beam splitters for live action 3D photography. As cumbersome as they are, they offer the most benefits and the least shortcomings. It's that simple.

 

[Rich Schaefer]

I agree with Jason on this one. I totally think we can post converge and bring objects into positive when shooting parallel. We have to do it well and deliberately and we need to measure how much we are dialing in in both pos and neg to make good choices. That's why the Marshall OR-70-3d with no HIT was a bummer.

 

[Pawel Achtel]

This is a potentially confusing discussion. Shooting parallel does not limit you to negative parallax only. Shooting parallel without HIT does, but no one would do that. .....

Yes, I agree, but dealing with only total parallax and ability to quantify it easily on set allows me to gauge the total depth budget and was sufficient for my purpose because it is the total that I am mostly interested in (which is the sum of positive and negative parallax that would be ultimately adjusted in post as desired).

I could not find any 3D field monitor that would be as small as the OR-70-3D. Also, Marshall has very good remote interface allowing me to control all the important modes and functions from the controls on my underwater housing.

 

[Dave Blackham]

BTW, with 5k or 6k resolution to spare, HIT may be practically implemented incamera firmware by windowing the sensor. How about this as a feature request to RED?

I agree, if theres one company that could implement this I think its Red. Possibly it could be a metadata tag recorded in camera allowing this to be adjusted in post if needed. So camera could be set up 5k or 6k and allow the HDSDI monitoring output to crop and actively scanned via a HIT function to allow monitoring. The HIT metadata could be added to the R3D data.

 

[Maciej Orman]

This is a potentially confusing discussion. Shooting parallel does not limit you to negative parallax only. Shooting parallel without HIT does, but no one would do that. If you want to shoot parallel properly get a monitor that does HIT. There are many.

Also, I must add, lens shift is not analogous to the human eye movement and not a flawless solution to this problem. The lenses in our eyes are designed to rotate around their nodal points and our eyes never get closer together or further apart. Shifting lenses moves the entrance pupil and changes your interaxial. Shifting the sensors or the active area of the sensor behind stationary lenses would be the preferred method.

A good rule to keep in mind is that most things 3D related that are optical or mechanical have already been tried, in many cases 100 or more years ago. There is a reason why people use beam splitters for live action 3D photography. As cumbersome as they are, they offer the most benefits and the least shortcomings. It's that simple.

The image in an eye is projected on the retina and when eyes converge they do not rotate around retina but around the center of eye ball thus their entry pupils get close together equivalent to shrinking stereo base.
If eyes ware rotated around center of entry pupil only then stereo base would remain constant.

 

[Jason Goodman]

The image in an eye is projected on the retina and when eyes converge they do not rotate around retina but around the center of eye ball thus their entry pupils get close together equivalent to shrinking stereo base.
If eyes ware rotated around center of entry pupil only then stereo base would remain constant.

The eye-brain to stereo camera analogy does not hold water for a lot of reasons. Mainly because your eyes have a brain connected to them that does a lot of behind the scenes work to make sense of the visual world. Moving the sensors would be a more desirable approach than shifting the lenses. Both are impractical. HIT on set and in post works well and is easily achieved.

 

[Maciej Orman]

To make it short, HIT is beneficial if one has camera system with excess resolution.
One must remember that such method always create inconsistent view angle as the stereo window is cropped arbitrary depending on the stenographer's parallax optimization techniques.
Shift lens is available now and shift sensor is not. Thus for system with HD cameras, shift lens is the only solution which preserves HD resolution.

 

[Jason Goodman]

How exactly would a FIZ system work with a tilt shift lens in a cinema application? If FIZ cannot work, how do you propose focus and iris remain synchronized throughout the day of a complex 3D film shoot? How would one even pull focus for that matter?

If you are shooting 5K or even 4K for a 2K delivery, extra pixels are no problem. By the time 4K is prolific, we will be shooting 6K, 8K or more. Furthermore, even when shooting 1920x1080, the small amount of blow up required is arguably a reasonable compromise when compared to the sacrifices required by other methods.

With regard to "One must remember that such method always create inconsistent view angle as the stereo window is cropped arbitrary depending on the stenographer's parallax optimization techniques." let's eschew obfuscation shall we. This is an overcomplicated way to describe something that in reality is not a problem. You can also lose the word arbitrary from that sentence if we assume the cinematographer / stereographer take this cropping into account as they are composing and recording the shot. They would hopefully be involved in the post process that executes the cropping as well, so it's not arbitrary it's part of the plan.

I am unaware of any practical applications of tilt shift lenses for stereo cinema that are currently being used. Discussing theoretically "perfect" ways to do things is fine, but for people who read this forum seeking practical advice, what you suggest does not seem appropriate.

There is hard science behind stereoscopy and stereo viewing and there is an art to 3D filmmaking. Surgical, industrial, military and other applications may benefit from more "accurate" imaging techniques, but they do not require artistic input and aren't necessarily encumbered by the same practical realities as motion picture production. The bottom line is, if the image looks good, it is a good 3D image. How you got there does not matter, parallel, converged, converted or otherwise.

 

[Maciej Orman]

Here are the shift lenses which can be used with RED Epic camera: http://www.northlight-images.co.uk/article_pages/tilt_and_shift_ts-e.html

 

[Dave Blackham]

What about moving the iris ?

 

[Maciej Orman]

Yes, that works too but unfortunately all modern lenses have iris inside where all chief rays cross.
It is designed that way to reduce vignetting.

 

[Jason Goodman]

There is no doubt that this could be made to work and might yield interesting results in an experimental setting. The limited choices of focal lengths, coupled with the difficulties of controlling iris and focus in a dual stereo setup make the tilt shift lens approach impractical for most applications. Compared this to the relative simplicity of using any lens and just HIT'ing your 5K image a few pixels in post.

 

[Dave Blackham]

Yes, that works too but unfortunately all modern lenses have iris inside where all chief rays cross.
It is designed that way to reduce vignetting.

I mention it as it has specific uses for 3D macro, possibly long lens work with custom lenses, though long lenses have all manner of issues for 3D in itself. If you use a large format lens for instance and the image circle is large then moving iris can work as a means of controlling IA with out vigneting. I'm not suggesting this is a common approach but it still can work. If may also be possible so use twin Iris on something like a 6k or 8K sensor to capture 3D side by side using a common objective lens, thats if you can get a lens with a big front element to make the IA large enough to be useful.

 

[Maciej Orman]

Yes, that should aslo work with the limitation you've pointed out.
I am working on the ultimate solution and that is an adapter for single RED Epic with two prime or telephoto lenses two relay lenses and two routing mirrors.
Such design should deliver distortion free side by side stereo with HD+ resolution (becasue of off-axis stereo window control) with base adjustable from 50mm to 120mm.

 

[Dave Blackham]

There is hard science behind stereoscopy and stereo viewing and there is an art to 3D filmmaking. Surgical, industrial, military and other applications may benefit from more "accurate" imaging techniques, but they do not require artistic input and aren't necessarily encumbered by the same practical realities as motion picture production. The bottom line is, if the image looks good, it is a good 3D image. How you got there does not matter, parallel, converged, converted or otherwise.

I agree.

The other means of filming is 'French converged' or 'Method Derobe'. Id think Alan Derobe who probably knew more about 3D than most of us put together here, certainly me. If Alan used his system of converged 3D cameras for filming 3D, it will be good, and I know from using the method it does result in very good 3D images and is very easy to control.