>> It's been mentioned about two point six zillion times by now that you
>> probably don't, though. The sensor itself doesn't have a lot more than
>> 12-bits worth of latitude, so producing more bits in the A/D doesn't
>> help a lot.
>>     
>
> I disagree with that a little. It's still the same sensor reveiving the 
> photons no matter if you set the ISO to 100 or 1600 in a normal
> DSLR. When setting the camera to ISO 1600 you just capture a lot
> less photons in one pixel well than you would if you set ISO 100
> (having the same amount of light available on both exposures). This
> is because you will have less light entering the sensor to get the
> correct exposure. Every single photon generates one electon (not
> exactly but to about 70% success rate) and there is a lot of
> space for those electrons. A good modern CCD sensor does have 14-16 bits
> worth of depth (or latitude). Don't know about these DSLR camera
> sensors though, maybe they are just 12 bits :(. 
I think the actual sensors do have the 14-16 bits you mention - which 
literally means that the max charge is up to some 65000 electrons. 
However, they also have a readout noise of at least 10 electrons, which 
is equivalent to 3 or 4 bits. This essentially means that you are left 
with a usable dynamic range or latitude equivalent to 12 bits; any 
additional bits would just be "measuring the noise", if they contained 
any information at all. That's why a 12-bit A/D is traditionally used.

Furthermore, when you amplify the signal for a higher ISO setting, you 
also amplify the noise, thus shifting up the 3 or 4 bits so that contain 
noise, so it enters even the 12 bits you keep.

But like I said, this has been discussed a lot in the past few weeks. In 
in several rounds before that, too. Enough for now.

I'm not sure I've read the data sheets of the *exact* sensors used in 
the Pentax cameras, either, by the way (I've seen technical data for 
various ones of similar type...)

- Toralf


-- 
PDML Pentax-Discuss Mail List
PDML@pdml.net
http://pdml.net/mailman/listinfo/pdml_pdml.net

Reply via email to