New Sony Sensor

[Jack]

Interesting write up over at the Online Photographer for the new Sony chip --- should find its way into a Nikon D3X this year

http://theonlinephotographer.typepad.com/the_online_photographer/blog_index.html

 

[Guy Mancuso]

Nikon has really been on the move lately. Great for the industry at large that is for sure.

Now this one maybe worth waiting for D3x

 

[Jack]

Now this one maybe worth waiting for D3x

Yup. Especially now that they added that 24 TS lens

 

[LJL]

Do either of you guys think the 12-bit design on this new Sony sensor might slow its sale/use compared to 14-bit on the D3 and on the Canon sensors? I know there is a lot of theoretical discussion on this, but just curious if Sony is coming in a half-step behind on this or not.

Jack, I agree that if/when Nikon hits with a D3X(?) at FF and a bunch of pixels, their new T-S lenses are going to grab some more attention.

LJ

 

[Jack]

The 12/14 bit debate is an interesting one. Personally, I believe that previous generation "12-bit" sensors were really more like 10 anyway, with perhaps the green channel maybe almost making 12. I suspect if we compared real 36-bit (3x 2^12) color to real 42-bit (3x 2^14) color, I doubt many of us would notice much visible difference.

I mean we're talking 68 Billion colors versus 4 Trillion, but the human eye can only distinguish about 100 Billion distinct shades --- and that's inside the total visual spectrum. Factor in that the capture space of most cameras is smaller than the total visible spectrum by nearly half, then dividing the lesser colors present by 68 Billion gets us to the levels where 12-bit is getting pretty darn close to the absolute limit of human vision inside any likely capture space...

 

[LJL]

Thanks, Jack. At this point, I really do not have enough frame of reference or examples to know. I was just thinking about other very large MP sensors, like those on MFDBs, where they are 14 or 16-bit to get the most color depth from the capture. It does mean moving a lot more bits, and 24MP at 6fps speeds, as this Sony sensor is being advertised as having is an awful lot of bits already ;-) My personal preferences would be for getting as much info as possible at capture. If there really is not that much noticeable difference between 12-bit and 14 or 16-bit, it may be moot, but it is hard to know what you are missing when you do not have it to begin with ;-)

LJ

 

[Lars]

D3x rumors have been around for years, perhaps we'll finally see something this year.

What 12 bits translates to in a final image depends on several factors

- The inherent noise in the analog section of the chip
- The camera manufacturer's decision on where to place full exposure
- The interpolation algorithm of the Bayer interpolator in the raw converter
- Exposure and curve adjustments after conversion to an image raster.

Also keep in mind that this is linear bits - 14 stops dynamic range fits into an 8-bit gamma 2.2 space, it's not like the 8-bit limit clips half the information ( though some precision is lost).

I think the only thing that can safely be said is:

Everything else being equal, higher bit depth is less limiting on image quality.

However, everything else is never exactly the same, which is why we wait for the reviews to see the actual performance once the chip is in a camera. Until then, a theoretical discussion around chit bit depth is just theoretical.

There is a parallel to the same discussion for scanners - however that is a different situation: scanner bit depth is most important when scanning negs, as negs only use a limited center section of the dynamic range. This limited range is then inversed and expanded, and high bit depth will mean smoother tonality. Since most of us don't use our DSLR's to shoot negs on a light table, that consideration is hardly applicable to cameras.

 

[Jack]

Since most of us don't use our DSLR's to shoot negs on a light table, that consideration is hardly applicable to cameras.

Lars, next time you're in town, I need to introduce you to Jan Brittenson