>>> Check the 'ordered m K10 today'. Rob is dropping heavy hints (about >>> weather sealing, mainly), and someone called 'gerasimov' from Bulgaria >>> posted a credible list of features containing a 22 bit A/D coverter. That >>> could be Aaron's 67-ditching feature... >>> >>> >> Hmmm. This has been discussed to great length before, but unless they >> have done something radical in other stages, the sensor itself doesn't >> have a dynamic range anywhere near the one associated with a 22-bit A/D >> converter, so why would they do that? >> >> - Toralf >> >> > > There should be enough for at least 5 more bits for a total of 17. I.e. > the original 12 plus each ISO setting. Depends on how you see it. Yes, I think there are enough different charge or signal levels for a few more bits, but the problem is that thanks to the noise in the system, there will be a certain amount of random variation in the signal so that essentially you can't reliably tell nearby levels apart. As far as I understand, the A/D is usually chosen so that the number of "bits" that get thrown away is equivalent to the magnitude of the noise at the native ISO setting. That's why you get "no noise" at the lowest ISO, and starts noticing an increasing amount as the value gets higher. > They could theoretically have 5 > analog amps and 5 A/D convertors and a post conversion combiner for an > effective total of 17. 22 however, would require another 5 bits on top > of the 17, which is a whopping 32 times more detail. > Exactly. I certainly won't believe that until I see it.
- Toralf -- PDML Pentax-Discuss Mail List PDML@pdml.net http://pdml.net/mailman/listinfo/pdml_pdml.net
