>
> I think it was AG, earlier in this thread who suggested it could be a
> demosaicing artifact. I could see how that could be the case, although I
> don't see any meaningful difference using the three algorithms available
> in RawTherapee.
>
I just tested my theory with RT on an E-1 file with no joy. Ain't no pattern
noise even if I push the conversion +5 stops. My Minolta A1's pattern noise
is real--caused by internal RFI issues. Suppose I need one of those nifty
Canon files to try it on. :)
BTW, it's amazing, but +5 stops on an E-1 file is STILL usable. Noise is
there, sure, but color fidelity and tonal separation in the shadows remain
decent. I'll post samples tomorrow.
Whatever the analog ability of the sensor to react to tiny differences
> in light, the analog voltages have to be encoded digitally. Perhaps this
> is some sort of quantization effect where tiny diffrences in voltage are
> forced into "buckets" with average values that create too much contrast
> between pixels. This could magnify any subtle pattern of sensor site
> sensitivity differences resulting from physical sensor design.
>
Moose, I think you are onto something here. The premise that a sensor is a
totally linear device is not necessarily true. No two sensor designs see
light exactly the same way and the in-camera processing is remapping the
bits a little bit maybe during the A-D process or afterwards, but prior to
the RAW file write. To deal with this quantization effect you are
describing, there is induced noise-dithering. One thing I've noted about the
E-1 files is that there is a base noise to the image that reminds me a whole
lot of the dithering that we apply to digital audio. I believe that Olympus
uses an aggressive dithering algorithm to mask quantization artifacts. This
is why the Olympus files are "noisy" as compared to other cameras, but is
also attractive. The noise that is there is actually generated to be that
way. This is my theory which I cannot confirm has been confirmed.
(Now that I think of it, at what point do the sensor sites start showing
> the effects of the number of individual light quanta that they sense?)
>
Another great question. Unfortunately, these are closely held secrets of the
camera companies. Just try to find out what the spectral response of a given
sensor is! In the film days, spectral response curves were published for
nearly every film ever made. Those geeks among us can actually look at those
curves and get an idea of why one film responds differently than another to
certain colors. Kodachrome doesn't look like either Ektachrome or Fujichrome
Velvia, Provia or Astia. We picked and chose films for their inherent
ability to see colors certain ways--even though the white-balance remained
identical from one film to another. Those of us from the northwoods knew
that Kodachrome turned bark gray, Ektachrome turned it blue and the
Fujichromes would turn it brownish. Nothing else changed.
But in the digital era, we are denied the information regarding the most
basic knowledge of our imaging systems. All we, the dumb sheep customers,
know is pixel count, pixel pitch and some rudimentary information about the
Bayer pattern--but not too much information. Heaven forbid that we find out
what the RGB sensors have their peak wavelength sensitivities at or the
signal efficiency of each of the colors. Would you care to tell me what the
cut frequencies are for the UV and IR filters in the 5Dmk2? As somebody who
has an understanding of spectral response curves and knows a little bit on
how to exploit them, this is actually important information to me. Kodak
publishes some of this information, but nobody else does.
Doesn't matter, I guess. As long as it passes DPReview's comparative tests,
we're good.
AG
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|