Sigma SD1 Merrill from ISO 100 to 6400

DMillier wrote:

xpatUSA wrote:

D Cox wrote:

The SNR depends on the number of photons collected during the exposure. If you adjust the exposure time and/or aperture for different lighting conditions, that number stays the same.

Don is absolutely right. I retract any statements to the contrary.

Don't confuse the numbers arriving at the lens with the numbers collected.

Pretty much what I did . .

Thank you.

OK, this helps clarify a bit. Let's reboot.

Does this mean we agree now that straitoutofhell's idea is not correct? Shooting high ISO in bright light is not less noisy than shooting in low light because of something about the nature of light itself (asssuming we have removed complications like point sources of light from out scene skewing exposure)? [This is still a question, not an assertion]

With my painted wall scenario, evenly lit in turn by soft bright ambient bright and dim light there would be no issue with getting exposure correct or extra high contrast subjects and all that will differ is the capture time. Any additional noise has to be caused by factors connected to the longer exposure as this is the only variable that has changed.

This feels like it ought to be correct but I don't actually know - are there other sources of noise that have been overlooked?

edit:

I guess the way to test this is to shoot the low light scene by changing the aperture rather than the shutter speed, maybe by using ND filters on the bright capture.

I'm now thinking that 'additional noise' was the red herring for me.

For a CMOS sensor we have 'reset noise'. That noise is say 40 electrons (SD9 white paper) introduced immediately by the reset switch opening just prior to exposure. Now say we collect 40,000 electrons-worth of light no matter what the scene brightness is. The shot noise for that is 200 electrons. Those two noises added together are 240 electrons (e-) and do not change for a given exposure.

SNR for that exposure: 167

But we also have the dreaded 'dark current ' (reverse-biased photodiode leakage current). This I think is the noise that "fills up" the sensor using long exposures, even with no light coming through the lens at all. (SD9 white paper value for dark current is 1.0nA/sq.cm, for example).

For one 5um pixel, that current density means about 1560 electrons/sec into the pixel well, so it would take about 32 sec to fill up a 50,000e- capacity well (if I got my sums right).

SNR for that exposure depends on shutter time. Like, for 5 sec, 50000/(1560x5+240) = about 6.4. Ergo, the exposure time has changed the SNR, not the shot noise.

Numbers are approximate for illustrative purposes.