We knew that already (that is the definition of the illuminant white point).
I found a UK reference that listed it as 72% NTSC so I think that supposed to be around 99% srgbā¦but most references to the monitor just state a number of colors so Iām not sure if that is accurateā¦lots of the specs posted on random sites for hardware are often not accurateā¦
statement deleted - it was only posted āby the wayā as indicated by āin any caseā which is missing from your quote above.
Not posting my full quote makes me look a fool, as if I thought that nobody knows what the color temperatures of D50 and D65 are.
I simply did not understand why you added that part, and wanted to avoid repeating a longer text; it appeared to me it was some kind of āresultā. I did not mean to make you ālook a foolā; my comment included a link back to the original post:
Which now links back to a post that no longer has that part. :-/ I think we should stop here.
@kofa I asked this in another thread as it happens, it is still a nuance that I am not sure about. So when setting the white point, what we are really doing is setting an additional transform to all colours? Is that any different to changing white balance in post processing?
I imagined a 3d gamut, with a line cutting right through the centre, going from white at the top to black at the bottom. Everything along the line is a shade of neutral grey. I can set that white point during display calibration, which shifts the colour of white (1,1,1) and also slightly tilts the line one way or another, which has the effect of shifting colours at the top of the gamut more than those at the bottom.
Now think you are most likely right and it is just a simple D65 to 5600k transform.
@cedric Thatās an interesting link. I am interested in the profiles that can test the VCGT, that will prove useful for me while I am experimenting.
Yes, we are transforming colours ā for display. No, itās not the same as changing white balance in post-processing.
What I write below is my current understanding. It may well be all wrong. Read with caution.
White-balancing is done to make sure that if you photograph something, a neutral surface, such as a grey card, would appear as neutral (and all the colours are correct relative to that neutral surface). For RGB, this means that an ideal neutral surface must be represented by equal red, green and blue components in the pipeline and in the output (ādraggingā all colours in the right direction, too). It affects all viewers. A simple form of white balancing is simply multiplying each channel with a given value to achieve neutral patch ā equal RGB (see e.g. here: Developing a RAW photo file 'by hand' - Part 1). Darktableās color calibration module performs a more sophisticated process, chromatic adaptation(*), which takes into account how our eyes and brains perceive colour.
The way I understand it, your choice of display white point could depend on the ambient lighting, as well as the displayās capabilities. If you need to match prints illuminated by D50 standard lights in a professional print proofing environment (see here: Product range of colour matching devices | JUST-Normlicht, for example), you would use a D50 white point for your display, too. Thatās calibration (adjusting the display so itās close to some desired state).
Then, your display profile tells your colour management system about your displayās properties to make sure that all image RGB values are displayed correctly relative to the chosen white point of your display (the colour (hue/saturation), as well as the brightness, is correct).
See Monitor profile, calibrate, system install
Iāve tried D50 at home, too, found it a bit too yellow; went back to D65, found that a bit too blue, after being used to D50, but I adjusted in a few hours.
(*) Chromatic adaptation makes sure that all colours appear correct relative to the white point. A neutral pixel will have R = G = B in your JPG or TIFF or whatever, regardless of whether the output format uses D65 (e.g. sRGB or AdobeRGB) or D50 (ProPhoto RGB), but will have different āabsoluteā (XYZ) coordinates. The XYZ coordinates of non-neutral pixels will also be adjusted, of course; however, they will not be represented by the same RGB values under different illuminants: they will be remapped according to the white point of the colour space, so they look correct relative to the chosen white point. Think of this as looking at the same printed image under different lights (e.g. incandescent light, direct sunlight, light on a cloudy day) - your brain will know it has to interpret the reflected light according to the light falling onto the paper. Even though the āabsoluteā colour (which is the product of reflectivity for each wavelength and the intensity of light with that wavelength in the light falling on the paper) is different, your brain will ācorrect for itā if given enough references (e.g. the white edge of the print, memory colours etc).
Then a further step is to convert the remapped XYZ values to the output (file or device) spaceās RGB (or CMYK) values (thatās not chromatic adaptation, but rather about transforming the XYZ coordinates of colours in terms of the axes defined by the primaries of the output space).
1 Like
@kofa Thanks for that explanation. The link for developing a RAW photo by hand looks interesting, it is now in my reading queue.
Have I understood correctly then, that white balance is to make sure nuetral grey is mathematically neutral (perhaps best checked with a colour picker) whereas display calibration is to make sure neutral grey is visually correct (and visually correct will be different depending on calibrated white point dictated by my own taste/requirement/use case/artistic whimsy) ?
In fact, a general comment to all, really appreciate the engagement and long form responses, I find it all invaluable.
They both are there (and you need both) to ensure neutral gray displays as neutral gray (and that all other colours also are displayed as correctly as possible*).
They are also both purely technical settings, not primarily meant for artistic expression. For that, you are probably better of using modules like ācolor balanceā. Although white balance is often adjusted āto tasteā. And sometimes thatās the only thing you can do (no neutral patch in the image, and unknown illuminant).
Iād advice strongly against adjusting your display calibration to suit your personal taste or artistic whimsey: you calibrate and profile your display to make sure that others with a calibrated and profiled display see what you intend them to see when displaying your images. Modifying your display settings for artistic whimsy defeats that purposeā¦
(* : some colours cannot be displayed on a three-channel display: the primaries have to fall within the colour āhorseshoeā, and they delimit the triangle which holds the displayable colours. There will always be colours outside that triangle and within the horseshoe. Colour spaces like Prophoto RGB are not bound by that choice of primaries, but that means you can end up with āvirtualā colours that fall outside the horseshoe)
@rvietor Thanks for your input. Understood regarding display calibration being a technical choice, I suppose in an ideal world we would all have displays calibrated the same. I get why this is typically D65, gamma 2.2, brightness 120cdm.
I have probably obsessed over this enough for now, I definately have a better understanding. I think at this stage I just calibrate and move on.
I am going to make 1 profile with calibration to D65 2.2, and 1 profile with the monitor in itās native state without a calibration just to satisfy my curiosity, I think I have a good enough basic grasp to make a more informed test.
I made a fresh calibration with displaycal and argyllcms 3.3.0. First, I reported on the uncalibrated display.
Then I reset my monitor to factory defaults, which set these levels:
Brightness = 100%
Contrast = 50%
Red = 100%
Green = 100%
Blue = 100%
I defined the calibration settings ā¦
ā¦ and the profile settings:
I ran the calibration and set the brightness and white point ā¦
ā¦ adjusting the RGB channels and the brightness on the monitor.
The profile was generated and I imported it into the Gnome colour management.
I used darktables colour management test to check it was being used by ācolordā.
The X atom isnāt set, but I can also set that using ādispwinā from argyllcms. I havenāt bothered because I think colord is working correctly and loading the VCGT and setting the profile, and I have set darktable to specifically use the colord profile.
However, I have also specified the display profile directly in the darkroom to be sure ā¦
ā¦ along with the standard input colour profile and sRGB output colour profile.
All thatās left now is to get back to editing some photographs 
Any comments or criticisms always very welcome.
How do you find the results. Are you happy with how things lookā¦
I donāt know about anyone else, but Iāve always done this physically on the monitor itself, thru the (always crappy) physical buttons and menus on the monitor.
2 Likes
That is the way I do it but now since I have them balanced it seems pretty stable and they remain good over timeā¦ I saw on one forum someone mentioning to just move red and blue and try to match to the green as sometimes moving one up drops the other one and you can be playing the yo yo for a whileā¦
Not all profiles support the perceptual intent, even if you select it. I think daisies falls back to relative colorimetric, if perceptual is not supported.
@priort will be editing over the weekend, I have a few photos in mind that I want to have another go at. I will follow up with a reply.
@paperdigits I did adjust the RGB levels on the display, just did it using a DDC interface to make it easier.
@kofa interestingly, no matter what intent I select in the display profile (in darktable), I donāt see any change. I assume because I am not using an image that really pushes the colour space. Any suggestions how best to test intents ?
1 Like
Well, maybe your profile does not support perceptual. Here is what Ella has to say about this:
LUT profiles require a source color gamut to create the various LUT profile tables, and the perceptual intent table in the resulting profile depends on both the monitorās color gamut and the color gamut of the source color space. I think DisplayCAL uses the sRGB color gamut. I donāt have any clue what ArgyllCMS does ā the documentation clearly tells how to specify a source color gamut but I couldnāt figure out what happens if the user doesnāt specify a source color gamut. In any event, the resulting profiles require that the user be able to figure out when to use relative colorimetric intent (the correct choice for general editing purposes) and when to use perceptual intent, and then make the appropriate configuration changes in the browser and in all the editing and viewing software the user might be using.
(Making and evaluating monitor profiles)
So, it I understand that correctly, unless you made a profile that has tables for darktableās default Rec2020 ā display perceptual mapping, you wonāt be able to use it. Or maybe the display profile is applied after output? Iām not sure. It would sound somewhat silly, as itād mean if you output in sRGB, youād never see colours wider than sRGB on your screen. I know there was a description about this, but I cannot find it. But if memory serves me well, from the output space, one can go to the soft proofing space ā back to display, if soft proofing is on. The description of the histogram space seems to support this:
histogram profile
Set the color profile of the histogram. None of the available options are ideal, however, āsystem display profileā is probably the least bad setting, since all other profiles are derived from the display color space and at least the values will conform to what you see on screen.
(darktable user manual - soft proof)
I thought the output profile was only used at export? It was mentioned in another thread I am engaged in ā¦
I have no deep desire to use perceptual, I just left it at that. Doesnāt actually look like my profile supports any intents. If this is the case, might I just as well de-select littlecms2 in the preferences?
Does your display profile support intents? Do you use them? Is it a profile you generated yourself?
Just when I thought I was making headway 
I used DisplayCal to generate mine. I think itās used with the relative intent (that one does not need custom mapping-from-some-input LUTs).
The profile viewer does say the default intent is perceptual, though. Iām out of my depth, here.
Here it is, if anyone wants to inspect it:
U2720Q #1 2024-07-13 10-06 D6500 S XYZLUT+MTX.icc (946.9 KB)