I agree that this will be the way forward. My rudimentary experiment was simply to test the water. I will need to acquire the necessary monitor calibration kit, and was already considering a new monitor, so shall most likely kill two birds with one stone.
It seems possible that all Sony cameras may be creating an identical issue, and that their respective owner’s calibration ventures could be producing very similar RGB corrections, which would be interesting to compare.
In my opinion do yourself a favour and get a very large 4K monitor. I work with a 43 inch monitor and I just love it for photo editing.
Experimenting to the reference value… I added only the basic modules… exposure, CC, CB one instance of tone eq and filmic…with lens correction and color denoise…not that it likely needed that one…I used the white of the eye for wb from the CC module but no correction for the colorchecker…
Relative and then Perceptual Renders
DSD01171_01.ARW.xmp (10.6 KB)
Bit different and simpler and with Sigmoid
DSD01171.ARW.xmp (9.9 KB)
Thank you for these contributions - the colours in your upper Sigmoid edit come pretty close to reality.
I have applied the ColorChecker profile to a few recent images along with a crude CC red channel correction circa 0.95, and this has unanimously improved upon the previous white balance, which had been based upon picking from the ColorChecker’s white card. Skin within the profiled images appears far more natural.
I struggle to comprehend/explain the difference, but it is undoubtedly much easier to to perceive with my grandchildren’s unblemished juvenile skin . In essence, I suspect that a touch of yellow gets added to the white and pink tones within a profiled image, which is absent from a white card image.
I fancy a larger monitor and shall heed your advice regarding 4K, which isn’t something that had previously occurred to me.
For what it’s worth , I stopped using CC for white balance on my Sony a7m2.
The camera profile (matrix) used by DT was the ine pushing things wrong. For me I could get good results I’d the original shot was close to daylight WB (actually closer to the reference WB).
The further away the WB was, the more wrong ‘reference’ looked. But that’s ok, reference doesn’t have to look good :). But then when the camera profile was applied later in the pipe , it sent the colours so wrong that CC can’t correct it anymore (or just sees the scene differently).
No setting or tweaking of the reference values could improve that, unless you set it to basically ‘as shot’ settings.
Setting the input profile to ‘linear 709’ as a test (so no camera profile really, just assume the camera data is close enough to RGB) made the red cast on skin disappear.
The red cast was there with or without filmic, and with or without sigmoid.
One commenter fixed it by using a camera profile matrix taken from the tungsten set , but never explained how he did that and I couldn’t manage it.
My experiments with taking a matrix from a DCP file and looking for a difference between the daylight set and tungsten set of values inside, didn’t improve things.
In the end I always got good results by doing the WB the legacy way.
Or course , the a7 m2 is an old girl by now, and there have been changes and improvements to Sony colour in the newer models. Maybe the newer sensors are more linear in response and dont have this behaviour.
Thank you for your input! I have recently re-read your thread regarding the red cast on your baby, who must now be 2+ and keeping you on your toes. Edit: I have spent some time with your image and suspect that we are potentially experiencing different phenomena (my ‘fix’ doesn’t work on your image - but nothing else seems to either).
Back in the day of film I was a reasonably competent portrait photographer but I struggle to make digital work as I want it to.
Digital cameras seem to be quickly baffled by studio lighting, which drives my need for accurate white balance. Having worked my way through numerous WB references over the years, I have always felt that something better must be possible, which is what led me to experiment with CC.
My current fudge of removing the red cast with an extra copy of CC does appear to produce better skin tones than I can achieve when picking a white card in legacy WB, but I will probably have a different opinion tomorrow.
I suspect that I may end up happier changing my camera, rather than trying to fix the Sony issue, because such a fix will ultimately prove to be yet another fudge.
It is my understanding that Aurélien Pierre’s interest in programming was originally driven out of his frustrations with digital colours. I note that he continues to produce lots of monochrome images 
There seems to be a major update in the new development version of darktable (4.7x) regarding white balance.
I also encountered color calibration problems during last weeks with images made with Sony A7series (A7RM4 in my case) with darktable 4.6 and 4.6.1. The white balance module defaults to “camera reference” and in many cases I struggled to compensate color casts with color calibration module, sometimes without success (especially with certain non-Sony lenses or old lenses like Super Takumar).
The good news is that in actual development versions of darktable the white balance module uses a new option “as shot to reference” as default and that makes my Sony images look a lot better and made it easy to refine things with color calibration module.
Yes there is. I don’t like it and have had a discussion here - Introduce as-shot to later D65 workflow (improved highlights) by jenshannoschwalm · Pull Request #15602 · darktable-org/darktable · GitHub 3 weeks ago, starting about a third of the way down. I think people should take a look and comment there or here.
In summary, I don’t like it as follows:
It has come to light that Camera Reference is not the best setting for certain modules, which work better with an accurate white balance in the WB module. Call these group A. Recall also that D65 is the best setting for various other modules, call these group B.
So a change has been devised which uses the camera As Shot setting for group A, and D65 for group B. But if you accurately white balance the image yourself, both group A and B will use the setting you have created (the WB module will show “user modified”). That is, the modules which should be using D65 will not be.
This change is achieved via a new item in the WB dropdown list - “as shot to reference”
Issues:
It seems to me it’s going to be assumed you’ve set the correct WB in your camera. But have you? It’s not been part of the scene-referred “Camera Reference” process previously.
The benefits of the change - optimal image quality by using the best WB coefficients for the modules in question - are only realised for the new option, and it’s plain wrong IMO that you don’t get the benefits if you white balance the image yourself.
My suggestion:
If you’ve chosen Scene Referred for your workflow in Preferences, then D65 should always be used for group B modules, and group A should use whatever comes out of the WB module. There’s no need for a new item in the WB dropdown.
