On 06/07/06, Toralf Lund <[EMAIL PROTECTED]> wrote: > Toralf Lund wrote: > > Eric Featherstone wrote: > >> On 06/07/06, Toralf Lund <[EMAIL PROTECTED]> wrote: > >>> I was talking about the number of different voltage levels that may be > >>> output from the sensor itself. That number is not necessarily 4096; it > >>> is the analogue-to-digital converter that has 4096 different values. > >>> > >> Toralf, the sensor is an _analogue_ device, of course it produces more > >> than 4096 voltage levels. It produces an _infinite_ number of levels > >> > > I'm afraid not. A CCD can only output a certain set of distinct voltage > > levels. I think Godfrey explained this earlier. > I should perhaps add that there are at least two different factors that > come into play here. One is the actual charge at the photo site - which > can only take one of a certain number of discrete levels (i.e. it is not > a continuous range), the other is the electric noise in the system. > Adding noise to the system essentially means that you will no longer be > able to tell nearby charge-levels a part, thus reducing the number of > usable output "values". > > I'm not sure *exactly* what the numbers are in today's DSLR sensors, but > I don't think the final figure can be that much higher than 4096. I was > speculating that it might be, say twice as large or 4 times as large. > But maybe they don't dimension the systems that way; perhaps the fact > that a 12-bit A/D was chosen actually indicates that the number close to > 4096, or at least significantly smaller than 8192. > > And again, I certainly think that the actual relations here has impact > on the exposure considerations we are discussing. > Ahh, my mistake. I had rather foolishly assumed that the "resolution" of the CCD voltage, as determined by the maximum number of photons captured, would exceed by many orders of magnitude the digital resolution. i.e. sufficiently higher that it was for all intents and purposes infinite. This appears not to be the case: http://spiff.rit.edu/classes/phys559/lectures/gain/gain.html So please ignore what I said.
Eric. -- PDML Pentax-Discuss Mail List PDML@pdml.net http://pdml.net/mailman/listinfo/pdml_pdml.net
