ACTUAL Pixel resolution of "35mm" CCD sensors.

[John Sprung]

a system that uses a 12-bit analog to digital converter

<{POST_SNAPBACK}>

I checked Panavision's web site. They say "10 bit log output", but nothing about the A/D.

 

 

 

-- J.S.

[Mike Brennan]

I checked Panavision's web site.  They say "10 bit log output", but nothing about the A/D. 

-- J.S.

<{POST_SNAPBACK}>

 

Nolan Murdoch said that the production version of the Genisis will be 14bit.

 

In general Sony are going 14 bit. (even the FX and Z1)

Replacement of the HDC 950 and other new Sony cams will probably have 14bit AD.

 

 

Mike Brennan

[Guest ADRIAN MCTIERNAN]

Arriflex is the only competitor talking about the details of their chip.  It's CMOS, 3018 x 2200 pixels, with a Bayer mask.  In video mode, it uses a subset of 2880 x 1620 pixels.

 

A Bayer mask is a very specific pattern of color filters.  You can think of the chip as being divided up into two by two groups of pixels, and each group has the pattern:

 

G  R

B  G

 

You could treat the data from a Bayer mask very simply by treating each two by two group as one pixel, averaging the greens, to make, for instance, 1440 x 810 from the Arri chip in video mode.  But because you have data from separate areas within the two by two group, especially two different green samples, and green is the dominant component of luminance, and luminance is the dominant factor in human perception of resolution, it turns out that there are smarter ways to process the Bayer data and extract more resolution from it.  You'll never get up to the full 2880 x 1620, but you can land somewhere between that and the simplistic 1440 x 810.  Recording the raw Bayer output lets you do the processing slower than real time, and you can go back and do newer filtrations on old data as they get invented.

 

Panavision isn't saying much about their chip other than 12.4 Megapixels and RGB filters, but not a Bayer pattern.  This is a custom chip from Sony.  It could be that they have patents in the works, or things that they intend to keep as trade secrets rather than patents.  I can only speculate that they might be doing things like a non-rectangular array or subpixels of different shapes and sizes.  It could be that they looked at the sampling structure and the filtration together, improving each with knowledege gained from working on the other.

 

Laying a mask on a single big chip is a whole bunch easier to do than making a prism block with dichroics and sticking three little chips to it in precise alignment. A single big chip gets you away from the restrictions on f/stop and back lens element position caused by the optical block.  It lets you use the huge investment in existing film lenses, and gets you the familiar depth of field you have with 35mm.  The downside to big chips is, the bigger the chip, the lower the yield.  Tiny dust particles in the manufacturing process can make a chip NG, and the bigger the chip, the more likely it is to get hit.

 

As for the actual resolution coming out of any of these systems, the right approach is to put the deep theory aside, and shoot tests. 

-- J.S.

<{POST_SNAPBACK}>