Originally posted by biz-engineer Hi folks.
I know how critical it is to expose correctly to avoid blown high lights. In single shot technique, I expose to the right with about +2ev above mid level, for the brightest part of the frame. In this way, the full range of the sensor is used.
That said, I now play with blending multiple exposures of a stack of images. While still having a reference exposure point for the highlight, I don't really know how to expose for the less bright part of this image since it could go down as low as total black at -14ev (considering raw) or -8ev (jpeg).
Is there such thing as the best level of exposure of shadows for exposure blending ? Thanks.
The concept of ETTR is not to use the "full range" of the sensor, but to capture the greatest amount of information within the image file. A digital image file is not linear (in terms of data capture), the highlight arrays can store more data than the shadows arrays. The concept of ETTR is to save as much color information about the scene as possible, and to do that, you want to use the bigger "buckets" set aside.
(A way to test this is to look at the file size of the same scene underexposed versus overexposed. The overexposed file with be larger than the under exposed file.)
I've always understood that if you want to do exposure stacking, you would expose your darks at as close to 128,128,128 as possible, which obviously obliterates close to 50% of your image, but you don't care about that anyway since you will be layering masking those parts out.
But by pushing the shadows that high, you get a really good signal to noise ratio and a lot more color information. When you do drop the exposure levels, you are getting cleaner darks.
From Wikipedia:
"ETTR was initially espoused in 2003 by Michael Reichmann on his website, after purportedly having a discussion with software engineer Thomas Knoll, the original author of Adobe Photoshop and developer of the Camera Raw plug-in.[1] Their rationale was based on the linearity of CCD and CMOS sensors, whereby the electric charge accumulated by each subpixel is proportional to the amount of light it is exposed to (plus electronic noise). Although a camera may have a dynamic range of 5 or more stops, when image data is recorded digitally the highest (brightest) stop uses fully half of the discrete tonal values.
This is because a difference of 1 stop represents a doubling or halving of exposure. The next highest stop uses half of the remaining values, the next uses half of what is left and so on, such that the lowest stop uses only a small fraction of the tonal values available.[20][21] This may result in a loss of tonal detail in the dark areas of a photograph and posterization during post-production. By deliberately exposing to the right and then stopping down afterwards (during processing) the maximum amount of information is retained."
Disclaimer (if needed): I am a huge supporter of Luminous Landscape, and owe a lot of my knowledge base to articles from Michael Reichmann et al. It's that article referenced in 2003 that really brought into clarity for the me the difference between film and digital. Coupled with my degree in software design, it made too much sense. I have been an ETTR believer ever since.