What determines the ISO on a DSLR? Canon vs Nikon

From what I understand it is only an equivalent of the film and is totally dependent on the maker correct? Is it possible that we are being tricked into thinking that some of todays DSLR cameras that shoot well very high ISO are in fact a little bit of an exageration? I mean the Nikon claim of ISO 25600 and be able to use it is just a little beyond belief and there does not seem to be really any jump in picture quality loss as the ISO goes up? I know that high of an ISO on film would be useless. So it makes me think that nikon is using its previous lack of good low light performing cameras as a marketing ploy to try sway some customers their way by doctoring or should I say exaggerate the numbers in their new line to put them in the good high ISO performing market. A test of the canon ISO 3200 and nikon ISO 6400 under identical conditions is in order ,me says.

What are anyone elses thoughts?

DSLR ISO is determined from the "gain" applied to the signal after it has been captured. A higher ISO means the internal processing applies more artificial boosting of the captured signal. This essentially gives you more light for free, but at a cost of noise. Noise is introduced through quirks in the circuitry that are amplified so much by the amplification process that they overtake the original data gathered, thus giving you a "noisy" shot.

Also, usability is different for different people. An uncropped image taken in bright sunlight with no harsh shadows, printed at 5x6, would probably be fine at ISO 25600 (or whatever). However, a 100% crop of a shadowed area with the same settings would probably look terrible.

As for doctoring the numbers, I don't think that is the case. I'm sure their light amplification software/hardware combination does in fact increase the "gain" by as many stops as they say it does. However, this does not mean the pictures will be good, or even usable. It just means that they can technically go that high. Also, underexposing the shot by a stop and boosting the exposure post-production is how Canon's get their expanded ISOs, so it's very likely Nikon does the same thing.

AirBrontosaurus wrote in post #4067626
DSLR ISO is determined from the "gain" applied to the signal after it has been captured. A higher ISO means the internal processing applies more artificial boosting of the captured signal. This essentially gives you more light for free, but at a cost of noise. Noise is introduced through quirks in the circuitry that are amplified so much by the amplification process that they overtake the original data gathered, thus giving you a "noisy" shot.

Also, usability is different for different people. An uncropped image taken in bright sunlight with no harsh shadows, printed at 5x6, would probably be fine at ISO 25600 (or whatever). However, a 100% crop of a shadowed area with the same settings would probably look terrible.

As for doctoring the numbers, I don't think that is the case. I'm sure their light amplification software/hardware combination does in fact increase the "gain" by as many stops as they say it does. However, this does not mean the pictures will be good, or even usable. It just means that they can technically go that high. Also, underexposing the shot by a stop and boosting the exposure post-production is how Canon's get their expanded ISOs, so it's very likely Nikon does the same thing.

Superb, clear answer! Thank you.

AirBrontosaurus wrote in post #4067626
Also, usability is different for different people. An uncropped image taken in bright sunlight with no harsh shadows, printed at 5x6, would probably be fine at ISO 25600 (or whatever). However, a 100% crop of a shadowed area with the same settings would probably look terrible.

As for doctoring the numbers, I don't think that is the case. I'm sure their light amplification software/hardware combination does in fact increase the "gain" by as many stops as they say it does. However, this does not mean the pictures will be good, or even usable. It just means that they can technically go that high. Also, underexposing the shot by a stop and boosting the exposure post-production is how Canon's get their expanded ISOs, so it's very likely Nikon does the same thing.

take a look at the d3 iso 12800 and 25600 shots...4x6 the noise will kill anything you print. at iso25600 i think my cell phone could take a better photo...

and the ISO really is dependent on the type of sensor. in the case of canon, there are two different "gain" circuits, one at the photodiode and one at the end of the sensor. depending on how it's used either both are on or just the latter is on (the case of long exposures with mkiii). now that all major companies use cmos for their top line cameras, you'll be seeing less physical differences and more software ones (iso12800 and 25600 are not real values,rather software bumps on iso)