gabebalazs wrote in post #15749935
I don't know if anybody's following my logic.
Shutter speed certainly plays a large part in the success of EC. However, given the choice of over-exposing ISO 1600 by +.6EV and hitting my minimum shutter speed (lets say 1/40th), and over-exposing ISO 3200 but +1.6EV with the same shutter speed, I personally would take the ISO 3200 option all day.
Being able to over expose by +1.6EV, means that the scene has 1.6EV of headroom over and above what the camera determines to be correct exposure. By choosing ISO 1600 @ +0.6EV, means we still wasted 1EV of data (dynamic range)
The shadows are the key here. The further down the exposure (the further left of the histogram) we go, the lower the signal to noise ratio (the higher the noise). If the camera didn't nail the exposure (under-exposed), and/or, you decide in post process to raise the shadow areas, you will have more noise to contend with (@ ISO 1600 +0.6EV). If the ratio of noise at lower light levels (EV) was 1:1 with the ratio of noise from a 1 stop ISO bump, it would be a tie. Unfortunately for us, the ratio of noise in EV is a logarithmic scale (halving of the EV = more then double the noise)
Take a look at the logarithmic Signal to Noise Ratio at DXO.
@ ISO 100, from 100% on the brightness scale to 10%, we went from 45dB to 38dB (approx), a 7dB reduction in Signal to Noise for a 90% drop in brightness.
Now look at 10% brightness (38dB) to 1% brightness (24dB). Only 9% brightness difference for a 14dB reduction in Signal to Noise. And that sucker just keeps getting worse the darker it gets! A further measly 0.9% reduction in brightness results in a whooping 18dB loss of Signal to Noise.
Conversely, looking at the SNR 18% results for ISO, we see that we only lose 3dB for an ISO bump of 1600 to 3200.
It's very late here. So apologies in advance if I messed something up above, or it's incomprehensible.