I think what I'm trying to get across is that yes obviously if I try to bring back lots of detail out of shadows I'll increase noise and clearly if my darks are hitting the wall I can't do anything to retrieve detail but do people have a simple rule of thumb where they say enough is enough it's time to blend two exposures.
Hi,
There is not a simple/hard rule, because it depends on how much you want to lift shadow detail brightness.
However,
it is a good question because until someone tries it it remains an unanswered question, and it remains a guess if and how much bracketing one should add in a once-in-a-lifetime situation that offers the opportunity of adding a bracketed exposure.
Although the sensor of the Nikon 810 has a good read noise performance, it still has to obey the laws of physics. That means that a very low exposure level will have a much lower Signal to Noise ratio (due to photon shot-noise) than better exposed parts of the image. Of course dark part of the image should probably remain relatively dark in the resulting image, but heavy processing may boost the noise to disturbing/distracting levels, also because the shot-noise will affect demosaicing of color. which is not only affected by brightening of the shadows, but is also affected by changes in saturation and contrast. So it is not immediately obvious how much they will suffer, but suffer they will.
Should this lead to distracting artifacts in the image, then we may wish we had addressed it when we still could, at the time of shooting, by adding longer exposures.
How much should one add? Well, in theory (barring ghosting and registration issues due to subject motion or camera shake) you could prevent most of the issues if you added enough exposure to the shadow regions to get them, say, half way the histogram after gamma conversion, that is 3 stop below clipping at the highlight side of the histogram. So if you want to significantly/visibly improve the shadow quality, you'd probably want at least the exposure levels that are some 4 to 6 stops darker than mid-tones to get at mid tone level. A 4 stop higher exposure will reduce the shot-noise to 50%, 6 stops to 41%, and both are visually quite noticeably better. You also need to take steps that are not
too big, otherwise the noise differences between images will make HDR stacking fail, and produce visible noise jumps in smooth gradients.
That means that you probably do not want to exceed an exposure interval that is more than 2 stops at the most (I use 1.3 or 1.6 at most), if you use HDR techniques. You can get away with a 3, maybe 4, stop interval if you do manual tweaking and masking, and only fill in clipped highlights with a lower exposure bracket,
after first aligning the exposures very accurately (a technique used by Guillermo Luijk in his ZeroNoise program, and an LR based manual masking method as described by Hans Kruse recently here on LuLa).
Does it make enough of a difference to bother? You will only know until you've tried it (and took precautionary brackets to begin with...). I know it does make a difference on my Canon shots, but we know that Canons have a lower tolerance for sloppy exposures than the recent Nikons (which may lead to sloppier exposures). ETTR is still important.
Cheers,
Bart