On Sep 12, 2006, at 1:58 PM, Toralf Lund wrote: > It just occurred to me that there may be one minor improvement to the > noise situation, though: I'm not 100% sure if I'm right, but the way > I've understood things, the ISO setting has in the past been > implemented > via a variable gain stage in the analogue domain. With this > oh-so-wonderful image processor, it is probably applied digitally > to the > 22-bit data instead. This does not do anything about the original > noise > or signal to noise ratio, of course, but at least it means that > *additional* noise won't be introduced by the gain stage itself. > > Differently put, some of what this 22-bit thingy is about, is probably > that the signal is converted to digital a bit earlier in the sequence, > so as to speak. This does eliminate some sources of noise - > probably not > the most significant ones, but still... > > Also, it has been suggested that the image processor will properly > remove some sources of variation that aren't strictly speaking noise, > like the effects minor differences in the sensitivity of the pixels. > (This, type of variation may unlike actual noise, cancelled out, as it > is *not* random.)
Certainly an interesting conjecture ... Much as I enjoy understanding how the little beastie works, however, I'm likely going to refrain from stripping it apart to deconstruct and understand all the details... I'm more interested to see how it performs in picture taking than on a lab bench. Godfey -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
