"Floyd L. Davidson" wrote in message
...
"Peter" wrote:
"John Sheehy" wrote in message
...
"DRS" wrote in
. au:

Why do Nikons seem to have a base ISO of 200? Is it a trade off of
some
kind?

The D3 uses a low silicon fill factor, and makes up for it with large
microlenses, which, of course, fill the wells faster.
In order to meter at
ISO 100, the camera would have very little RAW highlight headroom, so
ISOs
under 200 are only offered as special modes. Apparently, they use the
low
fill factor to get more noiseless electronics at the photosites. This
keeps read noise down a bit, but increases shot noise a little, too.



Would you please explain this concept in English? I know about SNR and why
I
get more noise at a higher ISO, but I don't understand the difference
between read noise and shot noise.

"Photon shot noise", or "photon noise", or "Poisson
noise" or "shot noise" are all references to the fact
that light doesn't hit the sensor at a constant steady
rate. The rate variation results in noise because the
actual exposure is for some specified interval. Two
adjacent pixel locations may be illuminated exactly the
same, but in the particular interval that light was
allowed to fall on them it is very likely that one pixel
will get more light (light for that pixel was arriving
at a faster rate than for the other). The difference
between the signal from those two pixels is shot noise.

Shot noise looks like a paper covered with a mix of salt
and pepper! It is also greater for more exposure (the
variation in rate increases as the amount of light
increases). Hence the place it is usually seen in
images is in the sky, or clouds, or other featureless
highlights. (So maybe we should say it looks like a
paper covered with a mixture of white salt and light
grey pepper.)

Above a certain amount of exposure, the SNR of an image
is very likely to be determined by photon noise, in
which case it is called "photon noise limited".

Read noise is all noise that is in the image as a result
of reading the data from the sensor (even if the lens
cap is on and absolutely no light has hit the sensor and
therefore the signal is 0). The base level of analog
voltage fluctuation on the sensor adds to it. Things
such as clock frequency variations in the ADC add to it.
In most discussions quantization distortion is included
in the "read noise".

Read noise has, like shot noise, a Gaussian distribution
so it is essentially random, and will mostly be seen in
the shadows. To continue the analogy used above, it
looks very much like a paper covered with mixture of
black pepper and grey pepper.

Below a certain amount of exposure, the SNR of an image
is very likely to be determined by read noise instead of
shot noise. (Generally that is not a good thing.)
Because read noise will be seen in the shadows it is
generally more annoying to the human eye.

Thanks for that explanation. My interest as a photo artist is simply to make
a nice picture. When Nikon says I can shoot at a higher ISO with less noise,
I am skeptical. I have a friend who claims he shoot a D300 at ISO 1,600
without little perceptible noise. Yet I have not found that to be the case
on my D300.



--
Peter