Archived posting to the Leica Users Group, 1999/10/13

[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]

Subject: Re: [Leica] Re: do your homework, please!
From: Alexey Merz <alexey@webcom.com>
Date: Wed, 13 Oct 1999 20:47:06 -0700

> Jim Brick:
>> People talk about "digital photography", "digital sensors",
>> etc... but in reality, a "digital" sensor is really an
>> "analog" sensor. Each pixel position has an associated
>> capacitor "bucket" to hold the number of electrons that
>> were allowed to flow in, based upon the number of photons
>> that the receptor received during the "exposure."

Anthony Atkielski:
> Note that both photons and electrons are discrete entities, 
> so you actually end up with a count, not a continuous signal 
> level.  The signal is thus digital rather than analog.

Nope. The electon signal accumulated in the CCD potential 
well has to be amplified prior to "counting" in the A/D
converter. Digital implies the lack of a physical model as 
you've said. Pixel intensity is NOT transferred as a straight
count of electrons but as an analog signal (roughly)
proportional to the electron count, to the A/D converter.

The representation is made symbolic only once it's passed 
though the converter as a digital signal. Thus a CCD
camera is NOT a purely digital device. As with silver halide
transparencies that are subsequently scanned, a digital image
is subject to significant sources of analog noise and 
distortion prior to digitization.

To be precise the CCD output is a wierd hybrid: pixel
LOCATION is digital, while pixel INTENSITY is analog. 

[...]
>The best charge ratios are around 100,000:1, I believe.

Again you use the term "best" in a nonsensical fashion. 
The "best" charge ratio is the one most suited to the 
specific application at hand.

>This is equivalent to about 17 stops in terms of dynamic range.

No it is not. This is for two reasons. First, In a real, 
room-temperature ccd unit a typical video rate CCD has an 
intrascene dynamic range of 1:100-300, a low noise CCD of 
about 1:1000. This is due to the presence of noise sources 
that I described in my previous post.

The relationship between photons entering a well and electrons
stored in that well is relatively linear (one of the best 
characteristics of CCDs) but that relationship does NOT 
correspond to "F-stops" because the SLOPE of the line 
defining this relationship varies depending on the CCD and 
how it is operated. 

This is true for films, too: their contrast (or characteristic 
curve) varies depending on emulsion composition, exposure, 
and developemnt.

In each case there are parameters that can be adjusted,
but not without trade-offs. Right now I get a better balance
of parameters with silver halide emulsions than with CCDs
for still photography. But CCDs allow me to do microscopic
work that would be insanely difficult with film. Such as 
marking the lysosomes (degradative compartments) of human 
cells with a fluorescent label, then watching what happens 
when those cells are invaded by Salmonella bacteria
(requires QuickTime plug in):

http://www.webcom.com/alexey/images/dynamo.mov

Different problems require different tools. As my father
sternly warned me: "pliers are NOT to be used as a wrench!"

- -Alexey