Michael Tapes said:
I would also think that it depends on the lighting conditions. I have always believed in the shoot at he require ISO, as pushing it will deliver lessor results.
Yes, I agree. First decide on what you want to capture and decide on the trade-off between aperture and shutterspeed. Then worry about the ISO needed to get a technically correct exposure level (i.e. "expose to the right").
However, from a technical point of view, from most reliable tests I've seen the evidence seems to indicate that up to and including ISO 1600 the 'sensitivity' boost is performed with an 'analog gain' amplification of the original signal levels prior to quantization (the analog gain is an input parameter for the ADC). At ISO 3200 the signal is basically the quantized ISO 1600 result, multiplied by 2, before linear to adjusted gamma conversion.
The reason that 'analog gain' produces lower noise, is because at the lower ISO settings each DN (digital number) is built from multiple photons per DN (even the shadow ones) so it is less prone to (quantization) rounding errors. There comes a point in low exposure levels where very few photons are converted to electrons, maybe 1 per digital number (especially in the shadows). Then there is no averaging benefit possible, and any method of boosting the signal levels will produce the same result as postprocessing.
The above is somewhat simplified to explain the principle, and I am aware that a signal level of 1 electron will in most cases be swamped by Photon shot noise and read noise, to mention a few noise sources.
Another point of attention is that by boosting the ISO, the dynamic range is reduced . But then that's just a reminder to use as low an ISO as feasible to allow a good capture.
There is little use in getting a technically good shot that suffers from motion/camera shake artifacts. Getting the shot is the prime concern.
Bart
