15kc to 20kc isn't even one bit of headroom. I was suggesting leaving four bits of headroom over and above the 12 bits of signal. That's more than enough; an extra six bits will make no difference.
(That's why some audio systems go to 40kc - maybe 2 bits of headroom - but not to 10Mhz, which is what you would need for 10 bits of headroom) I think we basically agree - a few bits of extra headroom is good. I was just suggesting that you don't need to go all the way to 22 bits. On Fri, Oct 13, 2006 at 10:09:09PM -0400, graywolf wrote: > I beg to differ, there is always a bit of a loss in any image > transformation, starting with a few extra bits can make up for some of > that loss. It is like hi-fi, most of us can not hear anything above > 15kc or so, but you can tell the difference between a system that goes > to 20kc and one that only goes to 15kc. It is called headroom. > > -- > graywolf > http://www.graywolfphoto.com > http://webpages.charter.net/graywolf > "Idiot Proof" <==> "Expert Proof" > ----------------------------------- > > > John Francis wrote: > > On Fri, Oct 13, 2006 at 04:35:46PM -0400, Perry Pellechia wrote: > >> Yes, I understand all about sampling, dynamic range and resolution. > >> In my field (NMR spectroscopy) the instrumentation typically uses 12 > >> or 16 bit ADCs. The four extra bits offer tremendous improvement in > >> dynamic range and fidelity. Adding another 6 bits would offer even > >> more improvements assuming what feeds the ADCs have enough dynamic > >> range and high signal to noise. > > > > Precisely. And that isn't true for the K10D. According to some > > figures posted here (and elsewhere) there are at the most 2^17 > > electrons in a full well on the sensor (and that may be an over- > > estimate by a factor of 10 or more), and the signal-to-noise ratio > > probably isn't any better than 2^12. > > > > Using a 16-bit processor is reasonable, to avoid quantization > > noise when applying non-linear transforms such as gamma correction. > > But there's no added benefit in going to a 22-bit processor today. > > > > > > -- > PDML Pentax-Discuss Mail List > [email protected] > http://pdml.net/mailman/listinfo/pdml_pdml.net -- PDML Pentax-Discuss Mail List [email protected] http://pdml.net/mailman/listinfo/pdml_pdml.net
