[filmscanners] RE: Newish Digital Tech

["Austin Franklin" <austin@darkroom.com>]

Hi Paul,

> At high light levels, as at high current levels,
> the number of photons or electrons is so large that the random
> fluctuations
> are a small percentage of the total.

What do you mean by “high levels”?  As far as “noise floor” goes, an
electrical signal that has a swing of -3V to +3V and has a noise of .01V,
has that as it’s noise everywhere along the full swing of the signal.  The
noise isn’t greater at 3V than it is at 0V.

> If you double the number of photons
> that you capture, the random variations only increase by the
> square root of
> two, improving the signal-to-noise by 3db.

Hum.  Do you have a source for that?  I don’t believe that is true, and will
have to think about your assertion.  What is the source of the “random
variations”?  I know there is some randomness in reception of photons,
simply because of atmospheric dispersion, and other causes...but I believe
it is linear and tracks directly with area.  Now, the ability to detect the
photons is a different story...and as I say below, the CCD has a noise floor
of minimum detectable photons, and that holds for the entire range of the
sensor...though there are other sources of noise due to process, but they
don’t change the photons that hit it, it’s just how the sensor responds.

Perhaps you are not separating out sensor behavior with actual photon
behavior?

> (Of course, other fixed noise
> sources, like electrical noise in the A/D amplifier, and
> quantizing noise in
> the A/D converter, provide a noise floor.) Conversely, if you halve the
> number of photons that you capture, the S/N degrades by 3db (as long as
> that's the dominant source of noise). This is true even if you
> count photons
> with 100% accuracy.

Ah, OK.  So you agree with what I said above...but then were is this
variable noise coming from?  There is a noise floor WRT the number of
detectible photons, and that is also the “noise floor” of the CCD.

> This explains, for instance, why my 2MP Digital Elph has lower
> noise in low
> light than my 5MP DiMage 7. The pixels are larger, and capture more light.

AND they are less susceptible to EMI.  There is a minimum number of photons
that imaging sensors need to “register”, and the larger the sensor, the less
photons per unit area are required.  That explains the difference, and I don
’t see what that has to do with noise.

> > But, let’s use what you say...the Bayer pattern CCDs has only ONE
> precision
> > color filter over each sensor...but the Foveon has up to THREE.
>  One would
> > reason, that by your claim, that the Bayer pattern sensors have,
> > therefore, less noise.  Right?
>
> No.

Well, YES ;-)

> In a Bayer pattern chip, each site makes use of one third of the
> spectrum, and discards the other two thirds.

YOU MISSED THE POINT.  Please go back and read what I said.  It’s clear that
the lower sensors in the Foveon get diminished light due to the fact the
light travels through other filters, and this is not the case for the Bayer
sensors.  Each site of the Bayer sensors get %100 of the light for THAT
color, in the Foveon that is not the case.

> In the Foveon chip, the other
> two thirds of the light is absorbed and measured in the other two layers.

There is physically no way the Foveon has %100 transmission to the lower
layers.  Also, the sensing area of the Foveon MAY be smaller than the
sensing area of a Bayer pattern CCD for the same information of the same
color.  You can not draw the conclusions you are based on the limited
information you have about the sensor’s architecture and performance.

There is a REASON the Foveon sensors have “issues” with low light
performance.

Austin

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body