Digital BW, The Print

The noise is electrical noise. Every electrical circuit has noise - it 
is a results of random motion of electrons. The power of that noise is 
proportional to the temperature and something called "bandwidth." In 
fact there is a ever present electrical back ground noise on the earth 
of about 10^-15 watts. The only way to avoid electrical noise is to cool 
the circuit to absolute zero where it would be a little uncomfortable to 
take pictures and the shutter wouldn't work.

How this impact the image is when the sensors that count photons are 
characterized by the electronics (there is some photon noise but 
electrical noise is by far the limiting factor). The S/N (signal to 
noise ratio)  is the measure of the signal power -" number of photons" 
vs. electrical noise background in the circuit. The reason there is 
higher noise at higher ISO is the signals are amplified prior to the 
analogue to digital conversion to bring the voltage into the range of 
the A/D so another type of noise won't be a problem (quantization 
noise).  The amplification amplifies the noise along with the signal.

The 12 bit A/D will perform better than the 8 bit mostly because there 
is more margin in the dynamic range. It is unclear if 16 is needed but 
the more bits the better. I have seen RF sensor system where 24 bits 
were required.  One way to get better S/N is with "bigger" sensors (more 
are at each sensor location)  in the array that count more photons in a 
fixed shutter time vs. a small sensor. You don't eliminate the noise but 
provide a bigger signal to the A/D for a higher S/N. That's why the new 
Sony F585 is so noisy compared to the DSLR out there. However to get 
bigger sensor at each location you need a bigger array - the bigger the 
array the more expensive it is to produce ( cost will go up with area).

The noise in a Canon, Nikon, etc. is going to "look different" than in 
the Sigma because of the interpolation. The noise that exist in nature 
is what is called "white" that is there is no correlation between 
neighboring pixels. Because of the interpolation to get a full number of 
pixels in the mosaic sensors the noise is correlated in neighboring 
pixels. See  http://www.dpreview.com/reviews/sigmasd9/page15.asp where 
this is shown and is pretty clear what is going on.

Dpreview has pretty good reviews of digital cameras, http://www.dpreview.com

Truman


>
>
>Steve Kale <stevekale@...> wrote:
>Does anyone know where I can take a look at key performance stats for several digital 
>cameras?  I am keen to compare the Canon 1Ds and 1D MK II against Leaf and PhaseOne 
>digital backs.  I know these are completely different solutions but a I struggle with 
>understanding the price gap.  If one extrapolates the 1D MK II pixel count up to a full 
>frame 35mm hopefully-to-be-announced-in-the-not-so-distant-future 1Ds MK II then it 
>could be around 16-17 megapixels which really becomes very interesting.  I understand 
>the Canon is 12 bit vs 16 bit for the higher order digital backs (but then not so long ago 
>we did almost everything in 8 bit.....).  So I assume the signal to noise ratio is where the 
>battle is.  (I can process the RAW data of either a Canon or PhaseOne in CaptureOne and 
>so that is an even platform.)  I just took a look at the portrait shots on Canon's website 
>from the 1Ds and did not think these were particularly sharp vs shots from a 6mp 
>PhaseOne back I have used.  What other key factors should I be looking at?  (please ignore 
>portability, handling ease, fps etc  -  just image quality.)
>
>Thanks
>
>Steve
>
>  
>