Rec709 and sRGB are both early 90s standards - they used a linear section to crush noise in the shadows.
As sensor technology and denoising has come on leaps and bounds, quite a few cameras no longer have the noise crush - so even some very expensive video cameras don’t use Rec709 as standardised.
I was referring to the basic Rec.709 2.22 gamma. I should have made that clear.
You are referring to Rec.709-1886. The use of an unqualified general term in photography strikes again - in this case, “Rec.709” has both kinds of curve.
I’m not referring to ITU-R BT.1886, that’s an EOTF (based on CRT). I’m referring to the linear portion in ITU-R BT.709 (the camera OETF). Many high-end cameras now just use a gamma function (or hypergamma function) for the EOTF (and therefore aren’t technically BT.709 compliant)
The first part of the curve is a linear segment:
https://www.image-engineering.de/library/technotes/714-color-spaces-rec-709-vs-srgb
It’s designed to work in conjunction with BT.1886 gamma curve to reduce the noise in the image.
Thanks. I was under a false impression that the linear segment was a later addition but I found a copy of Version 1 here and it was in there from square one:
Frankly, I find the naming of these options unfortunate in Darktable when it comes to working color spaces. The Rec.whatever standards are for video transmission, so they specify resolutions, frame rates, transfer functions, black levels, and whatnot.
Darktable just uses the color representation part; for practical purposes these choices determine working space gamut; given 32bit internal precision per channel the rest is irrelevant.
