Originally posted by jsherman999 Having an ISO 6400 shot in the near future look like an ISO 400 shot of today's top-tier cameras is something I eagerly await.
Can't see this happening due to shot noise. ISO400-> ISO6400 is 16x (4 stops gain). So for equal output image brightness, Assuming no other sources of noise except the light itself, in the ISO6400 shot the sensor has collected 1/16th the number of photons as in the ISO400 shot. The photonic shot noise is related to the sq root of the sample size. So if a sensel (photo-electric element) outputted 16000 photo-electrons in the ISO400 shot, it would have a shot noise SNR of 16000/126.5 = 126.5:1. In the ISO1600, in the same situation, it would only output about 1000 photo-electrons, for a photonic SNR of 1000/31.6 = 31.6:1. So 1/16th the signal has 1/4x the SNR. Expressed as dB, 1/4x = -12 dB. For example compare the 18% SNR for these 3 Sony low read-noise cameras:
DxOMark - Compare cameras side by side
Manufacturers ISO
ISO400: 35.4 dB
ISO1600: 23.3 dB
Difference: -12.1 dB
Since 18% grey relative to saturation is a high level, the vast majority of the noise is photonic shot noise. At low levels, the predominate noise source is read noise, mainly ADC noise.
The sensor output level gets smaller as ISO goes up (at base ISO the sensor should be able to reach FWC, but with pre-ADC gain, the ADC will clip instead earlier and earlier at lower sensor output levels at higher & higher ISOs). So, the max. amount of signal is seen as 1/16th lower if 16x pre-ADC gain is applied. Since this a direct relationship rather than a sq-root one, the DR between ISO1400 & ISO6400 should fall (assuming no high-ISO raw NR) to 1/16th or -4 stops.
Using the NEX-6 DR figures from the same link above:
ISO400: 11.75 stops
ISO6400: 8.03 stops
Difference: -3.72 stops.
The droop in the expected DR at low ISO is due to the presence of ADC read noise. The lower the read noise is, the more consistent the DR improvement at low ISO. Look at the DR curve for the K-5.
So, I can't see the 18% SNR figure (shot-noise determined) improving dramatically enough to make the SNR of a ISO1600 shot look like an ISO400 shot. It is possible for the read noise to drop further, so some DR improvement is possible, but this will probably only be for high-ISO DR, since lens flare/glare is probably more significant for the real-world DR and will place a limit on the max. DR possible.
You could, of course, take a burst of 16 ISO6400 shots and combine them, either in-camera or in PP, to achieve a similar effect to taking 1 ISO400 shot. For the same light-level and f-number, the ISO400 shot would have a 16x longer shutter speed. It's possible that non-consistent motion that occurred during the much faster 16 ISO64000 shots could be easier to align when combining the stack.
Dan.