Sensor Design for 3-D Video Cameras

Cesar Rubio · October 22, 2011

The minimal pixel size I recommend in sensors is 5.5 microns. If we go smaller than that we get lower sensitivity “lower base ISO/ASA”, less dynamic range, lens diffraction problems etc)

For 3D-HDTV productions we need 1080p/2K Bayer resolution for a final true 720p output. So a 2/3” sensor size is needed:

2048 pxs (horizontal resolution of 2K) x 0.055=11.2 mm

For regular 3-D Cinema we need at least 3K Bayer resolution for a final true 2K resolution.

In order to get 3K Bayer resolution without making the pixels smaller than 5.5 microns size, we obviously need to increase the sensor size:

3072 pxs (horizontal resolution of 3K) x 0.0055 ( pixel size)=16.89mm (horizontal sensor size needed)

A Four Thirds sensor size will be good for that kind of resolution:

http://en.wikipedia.org/wiki/Four_Thirds_system

For IMAX 3-D we need at least 9K Bayer resolution:

9216 pxs x 0.0055mc=50.68mm

A Medium Format (645) sensor is needed for that kind of resolution:

http://en.wikipedia.org/wiki/File:SensorSizes.svg

But with Stereo 3-D photography in order to get the deepest DOF, especially in interiors with artificial light and wide lens apertures, we strive for lower sensor sizes.

What should we do then?

A clever way to resolve this dilemma is to use a sensor design as the Sony F65 camera has!

pro.sony.com/bbsccms/assets/files/.../F65_Camera_CameraPDF.pdf

That means we can pack double the resolution in any given sensor size!

For a 2/3” sensor we could pack up to 4K resolution. That resolution would gives us regular 3-D Cinema resolution (3K) plus room for crop in post when shooting parallel (why in the world we should “toe-in” the cameras or use “convergence”?)

I would settle for 12.5 percent more of horizontal resolution for cropping in post, so for 3K we would need:

3072=12.5%=3456 pxs. Easily achievable in a 2/3” sensor size.

For IMAX 3-D resolution I would stick to a S35mm sensor size (24mm horizontal):

24mm/0.055=4364x2 (F65 sensor design )= 8728 pxs.

The Sony F65 camera has 8K resolution, even if is not 9K, it's pretty close and I would not jump to FF35mm size (36mm horizontal) just to achieve 1K more resolution.

Also the prime lenses sizes for S35mm are smaller than FF35mm, and will permit us to work with side-by side camera configurations easier. But in order to do that, first we need to implement small "camera heads" a' la MVC's (Machine Vision Cameras)...

How we should call the Sony F65 sensor design? Bayer+ or Bayer x2?

Thanks,

Cesar Rubio

Wisconsin & LA.

http://dna-rubio-3d.blogspot.com/

http://dnarubio3d.wordpress.com/

Edited October 22, 2011 by Cesar Rubio