- This topic has 6 replies, 2 voices, and was last updated 5 years, 11 months ago by
.
-
Posts
-
Hi
First, I’m a newbie to color management, so forgive the questions…
I recently bought an iMac 27 and installed Windows 10 on it. iMac are factory calibrated, and indeed, the colors look very good under macOS without any adjustment (I know this can be validated only by a mesurement, but the subjective perception is very good).
Unfortunately, the dedicated ICC profile under macOS obviously doesn’t work under Windows (although I thought ICC format was OS independent) : colors are pretty awful, blue becomes dark blue, etc. The graphic card drivers come with “AMD settings” where you can adjust color temperature, but I left it to “Automatic” as setting it to 6500K generates oversaturated red. Then, I decided to buy an x-rite i1 Display and calibrate my iMac.
I used i1 Profiler and calibrated using the “White LED” correction and the following targerts : 6500 K, gamma 2.2, 120 cd/m². The result is not too bad (subjectively, though).
Then, I used DisplayCAL (correction : Spectral LCD white LED, targets : 6500 K, gamma 2.2, 120 cd/m², very large testchart for LUT profile). The result looks ok, too.
I used DisplayCAL “Verification” for both profiles. What suprises me is that if the profile from DisplayCAL obviously get better results (average deltaE 0.21, max 1.1 vs average 0.49 and max 1.75 for i1 Profiler), my perception is different : in very light shades of grey (from 247 to 255), a red shift is visible when using DisplayCAL generated profile, but not with i1 Profiler profile. And as many Windows apps use theses shades, it’s pretty disturbing.
When displaying gradients (black/white), I also can observe smoother transitions with the i1 Profiler profile. And in the very dark shades, a strange red shift too (whatever profile).
Should I upload both measurement reports?
In one of the topics (about greyscale issue), it was suggested to create a 3D LUT, as it’s more precise. Then, my (newbie) questions (I couldn’t get clear explanations after googling) :
Does Windows can natively handle 3D LUT, or only 1D LUT?
I’d like the correction to globally apply to Windows, is it possible with a 3D LUT?
If the answers to both questions are “no”, is there a solution to give priority to the greyscale correction over RGB correction when calibrating? This red shift is very disturbing to my eye!
Thanks for your answers.
Hi again
Forget about the gradients line, as after rebooting, the same gradients look ok whatever the profile I use. But while trying to tweak DisplayCAL yesterday, they looked awful. I saw there is a “DisplayCAL Visual Whitepoint Editor Temprorary Profile” under Windows color management. Could this be the cause of the temporary degradation of the gradients?
AMD settings:
Color temp “ON” and 6500K = no change
Color temp “ON” and other value = assumed native white D65 and sRGB/2.2 gamma, apply some transformation in GPU for white point correction.
Color temp “OFF” = assume native white D65 and sRGB/2.2 gamma, use display EDID data to emulate an sRGB display.Newer iMacs have widegamut backlight so windows desktop will look more saturated. Color managed apps should render images in the same way as in OSX.
Since the are widegamut, you should not use WLED correction, this is intended for white led sRGB displays.
Unfortunately Xrite does not bundle a specific correction for newer widegamut LED bakclights. You can use Xrite GBLED (RGphosphor) or some of the ones in DisplayCAL colorimeter correction database (but they may need some tweak at user side to “fix” some values).Your current display only supports 3x 1DLUT at hardware level, no matter what OS you use. iMac displays are like laptop displays with the same limitations when calibration them: you have just a brightness control, white point is corrected at GPU.
Some applications allow you to use LUT3D like madVR (windows video render) or Resolve (multiplatform). This is not a global calibration but a image window/canvas/fullscreen one and limited to that.
If you want global custom gamut emulation on the fly, an iMac will not provide it. You’ll need a LUT3D monitor like NEC PAs or some Eizo CG or an external lut box with whatever monitor you want. For a more limited version of custom gamut emulation, a widegamut monitor with hardware calibration may work (lut-matrix-lut hardware calibration).Since you hardware only supports 3x 1DLUT, grayscale-whitepoint is the only kind of calibration that you will get. BTW DisplayCAL only does that kind of calibration for its profiles and same goes to i1Profiler.
If you want a better grayscale calibration make DisplayCAL’s calibration slower (slider), use a proper spectral correction to your display (explained before) and limit profile type to “1 curve + matrix” in order to minimize some rounding errors in color managed apps (or OSX Desktop). You can apply this to Windows or OSX.If Windows non color managed desktop bothers you, try AMD’s sRGB emulation based on EDID data explained before. Maybe is what you miss from OSX, just keep in maind that it’s applied globally (=sRGB display).
AFAIK OSX takes EDID data to build “default” profile for a display. Since you said that OSX seems accurate, we could assume that EDID data is “visually accurate” and AMD sRGB emulation on the fly will work.Hello
Thanks for the answers.
Color temp “ON” and 6500K = no change
No, the saturation change is obvious, reds seem to pop out of the screen (color temperature : 6500K).
Since the are widegamut, you should not use WLED correction, this is intended for white led sRGB displays.
Unfortunately Xrite does not bundle a specific correction for newer widegamut LED bakclights. You can use Xrite GBLED (RGphosphor) or some of the ones in DisplayCAL colorimeter correction database (but they may need some tweak at user side to “fix” some values).I used “White LED” as I read somewhere on this forum it was the correction to use, but I’ll give a try to “GB-LED”, and I’m searching the colorimeter correction database.
Since you hardware only supports 3x 1DLUT, grayscale-whitepoint is the only kind of calibration that you will get. BTW DisplayCAL only does that kind of calibration for its profiles and same goes to i1Profiler.
I’m confused by this latter statement (I’m a newbie, I told you). What’s the point, then, of displaying thousands of color patches if the resulting profile deals with greyscale only? To be able to assess the global accuracy of the display once the greyscale is optimized?
I’ll try to optimize the grayscale calibration as you advised.
Thanks again for the answers.
Hello
Thanks for the answers.
Color temp “ON” and 6500K = no change
No, the saturation change is obvious, reds seem to pop out of the screen (color temperature : 6500K).
No, you’re wrong. When you set Color temp ON and 6500K, AMD application does nothing to windows desktop. No change.
You must read what I wrote below that in my previous reply: windows desktop (icons and file explorer) are not color managed. You are seeing each RGB values for each icon and each wallpaper sent as RGB values in native gamut , not as if they were encoded in sRGB. Same applies to other non color managed apps.
P3 “red” (255,0,0) is much more “saturated” than sRGB “red” (255,0,0) . This is what you see. AMD settings are not involved at all (unless yu mess up the configuration like explained before).
Since you hardware only supports 3x 1DLUT, grayscale-whitepoint is the only kind of calibration that you will get. BTW DisplayCAL only does that kind of calibration for its profiles and same goes to i1Profiler.
I’m confused by this latter statement (I’m a newbie, I told you). What’s the point, then, of displaying thousands of color patches if the resulting profile deals with greyscale only? To be able to assess the global accuracy of the display once the greyscale is optimized?
Your hardware only has LUTs provided by graphics card (GPU) to modify display behavior, like in a low end monitor. Calibration will use that 3 x 1DLUT (GPU) to neutralize grey to the same color as white, correct gamma to your desired target and set your desired white. Nothing more.
1 vector for each channel. Position in a vector is color vlaue “IN” and value stored in that position is color value “OUT”. Since you have 3 vectors, values in the same position aim for values than once sent to display produce a neutral grey (no color) with a desired bightness relative to “white” (gamma).Since your display is an iMac, it’s actually like a laptop display with laptop limitations. For example an office IPS sRGB monitor usually has RGB gain controls to correct whitepoint in monitor… a laptop does not have it, an iMac does not have it… so white must be corrected in those vectors limiting one or two channels maximum value . It does not matter if an iMac has a perfect D65 white… you maty want another white (printing) and that correction will be done at GPU. iMacs are not a good choice for photo editing for this thing (and other ones that will be out of topic).
Higher end monitors could be able to correct white, gamma and grey neutrality inside monitor (monitors with hardware calibration) and even higher end monitors could be able to emulate smaller gaumts than its native (and some of them use an internal LUT3D for better accuracy). So first of all you must neeed what can be done at HW level with your device an its limitations.
Now let’s get back to calibration done in GPU LUTs: A huge number of grey and near grey patches for calibration will bring more information to compute a closer calibration to your desired target (white, gamma and neutral grey) .
Once DisplayCAL or i1Profiler have a calibration computed (that 3 vectors), they load it into graphics card LUT hardware and measure how that display behaves: white, gama, grey neutrality and gamut (after calibration behaviour). It could be done with lots of patches for a “Table” profile (Xrite naming) or less patches for a matrix profile (a more idealized behaviour of your display). This is what you configure in i1Profile as number of patches… or in DisplayCAL sliders for profilling patches.
That info is stored in an ICC/ICM profile file. Calibration, I mean data loaded into GPU 3x1DLUT are stored as a field in that file (VCGT field)… think of it as an appendix in a book. The core content of a profile is how that display behaves when it was measured (after calibrated in your situation), calibration is an appendix aaded to that book in order to load it at OS startup/user logon. This is just a didactical explanation (core content of a book + calibration appendix), a visual idea, not a technical one.I think that you were not aware about this. One set of patches is for calibration (usually fixed, but DisplayCAL is configurable to some extent) and another set of patches for profiling (capturing how calibrated monitor behaves).
Color managed programs only are aware about profile information (“how calibrated monitor behaves”). They take some RGB numbers in a colorspace and transform them to another RGB numbers in your display colorspace (which is decribed by profile).
Windows desktop (icons and wallpapers) do not make that transformation by default (which is good for some things and bad for others).
Color managed apps like PS do it, same goes to OSX buttons and wallpaper.Hello
Thanks for the answers.
Color temp “ON” and 6500K = no change
No, the saturation change is obvious, reds seem to pop out of the screen (color temperature : 6500K).
No, you’re wrong. When you set Color temp ON and 6500K, AMD application does nothing to windows desktop. No change.
You must read what I wrote below that in my previous reply: windows desktop (icons and file explorer) are not color managed. You are seeing each RGB values for each icon and each wallpaper sent as RGB values in native gamut , not as if they were encoded in sRGB. Same applies to other non color managed apps.
P3 “red” (255,0,0) is much more “saturated” than sRGB “red” (255,0,0) . This is what you see. AMD settings are not involved at all (unless yu mess up the configuration like explained before).
I understand that, as Windows desktop is not color managed (unlike macOS), an “absolute” red (255,0,0) displays differently within different gamuts. So, the saturation change I can observe (I swear) when I switch “color temperature” to “ON” means the iMac switches from sRGB to P3 gamut? And as Windows desktop colors were probably designed for a generic sRGB display, they look flashy when displayed “roughly” in a wider gamut.
Since you hardware only supports 3x 1DLUT, grayscale-whitepoint is the only kind of calibration that you will get. BTW DisplayCAL only does that kind of calibration for its profiles and same goes to i1Profiler.
I’m confused by this latter statement (I’m a newbie, I told you). What’s the point, then, of displaying thousands of color patches if the resulting profile deals with greyscale only? To be able to assess the global accuracy of the display once the greyscale is optimized?
Your hardware only has LUTs provided by graphics card (GPU) to modify display behavior, like in a low end monitor. Calibration will use that 3 x 1DLUT (GPU) to neutralize grey to the same color as white, correct gamma to your desired target and set your desired white. Nothing more.
1 vector for each channel. Position in a vector is color vlaue “IN” and value stored in that position is color value “OUT”. Since you have 3 vectors, values in the same position aim for values than once sent to display produce a neutral grey (no color) with a desired bightness relative to “white” (gamma).Since your display is an iMac, it’s actually like a laptop display with laptop limitations. For example an office IPS sRGB monitor usually has RGB gain controls to correct whitepoint in monitor… a laptop does not have it, an iMac does not have it… so white must be corrected in those vectors limiting one or two channels maximum value . It does not matter if an iMac has a perfect D65 white… you maty want another white (printing) and that correction will be done at GPU. iMacs are not a good choice for photo editing for this thing (and other ones that will be out of topic).
I know that point and I didn’t buy this machine for that specific purpose (event if I occasionally do some photo editing), but that’s another topic.
Higher end monitors could be able to correct white, gamma and grey neutrality inside monitor (monitors with hardware calibration) and even higher end monitors could be able to emulate smaller gaumts than its native (and some of them use an internal LUT3D for better accuracy). So first of all you must neeed what can be done at HW level with your device an its limitations.
Unfortunately, the price tag of Eizo’s displays (4K, at least) was out of reach for me!
Now let’s get back to calibration done in GPU LUTs: A huge number of grey and near grey patches for calibration will bring more information to compute a closer calibration to your desired target (white, gamma and neutral grey) .
Once DisplayCAL or i1Profiler have a calibration computed (that 3 vectors), they load it into graphics card LUT hardware and measure how that display behaves: white, gama, grey neutrality and gamut (after calibration behaviour). It could be done with lots of patches for a “Table” profile (Xrite naming) or less patches for a matrix profile (a more idealized behaviour of your display). This is what you configure in i1Profile as number of patches… or in DisplayCAL sliders for profilling patches.Ok, I get it. That’s the reason why when DisplayCAL initiates the calibration process, Windows reverts back to no profile (default Windows display), and when the calibration is done, it is applied (visible color change in Windows desktop) and then the profiling process begins.
That info is stored in an ICC/ICM profile file. Calibration, I mean data loaded into GPU 3x1DLUT are stored as a field in that file (VCGT field)… think of it as an appendix in a book. The core content of a profile is how that display behaves when it was measured (after calibrated in your situation), calibration is an appendix aaded to that book in order to load it at OS startup/user logon. This is just a didactical explanation (core content of a book + calibration appendix), a visual idea, not a technical one.
Ahah thanks for the warning! As a computer engineer, I regularly try to explain to people how things work behind a software, “visual ideas”. In color management, I’m in the position of the one who needs to be explained to. Thanks again for the time you spend explaining things.
It means the majority of data in an ICM profile (core content of the book) is useless if you don’t use color managed apps ?
I think that you were not aware about this. One set of patches is for calibration (usually fixed, but DisplayCAL is configurable to some extent) and another set of patches for profiling (capturing how calibrated monitor behaves).
Indeed, and it’s much clearer now, thanks.
Color managed programs only are aware about profile information (“how calibrated monitor behaves”).They take some RGB numbers in a colorspace and transform them to another RGB numbers in your display colorspace (which is decribed by profile).
Windows desktop (icons and wallpapers) do not make that transformation by default (which is good for some things and bad for others).
Color managed apps like PS do it, same goes to OSX buttons and wallpaper.Hello
Thanks for the answers.
Color temp “ON” and 6500K = no change
No, the saturation change is obvious, reds seem to pop out of the screen (color temperature : 6500K).
No, you’re wrong. When you set Color temp ON and 6500K, AMD application does nothing to windows desktop. No change.
You must read what I wrote below that in my previous reply: windows desktop (icons and file explorer) are not color managed. You are seeing each RGB values for each icon and each wallpaper sent as RGB values in native gamut , not as if they were encoded in sRGB. Same applies to other non color managed apps.
P3 “red” (255,0,0) is much more “saturated” than sRGB “red” (255,0,0) . This is what you see. AMD settings are not involved at all (unless yu mess up the configuration like explained before).
I understand that, as Windows desktop is not color managed (unlike macOS), an “absolute” red (255,0,0) displays differently within different gamuts. So, the saturation change I can observe (I swear) when I switch “color temperature” to “ON” means the iMac switches from sRGB to P3 gamut? And as Windows desktop colors were probably designed for a generic sRGB display, they look flashy when displayed “roughly” in a wider gamut.
No. Windows renders desktop in whatever gamut configuration(s) your display has. Your display has only a configuration with near P3 gamut. Other displays have sRGB emulated gamut and you can switch between them but yours has not that features.
What you see is a special feature of AMD/ATI cards that has been available for a decade even if there was no visible in settings. This was included as part of the AVIVO engine. Pretty awesome for consumer cards back in 2004-2005, maybe even earlier. There was an ATI’s PDF about AVIVO engine with the details. It looks like it is some kind of de-gamma lut, matrix emulation, gamma reencode and calibration lut and temporal dithering unit to avoid banding like in some monitors with HW calibration… google that PDF, I do not remember the details.
When you switch off Color temp in AMD settings you are actually switching ON an sRGB emulation based on EDID data on your AMD graphics card. I mean it’s not a “part” of iMac… or a PC with an AMD GPU, it is just a special feature of that GPUs.
AMD settings are confusing, they lack of explanation to newbie about what is this and how it behaves.IDNK if nvidia offers that feature at HW level. AFAIK nvidia control panel does not offer sRGB emulation based on EDID or custom gamut emulation.
(to Florian)
IDNK if there is a public API for these features but it would be an awesome feature to add to DisplayCAL.Unfortunately, the price tag of Eizo’s displays (4K, at least) was out of reach for me!
27″ QHD widegamuts about $1000 from NEC or Eizo win hands down any 5k screen from Apple or LG… but this will be out of topic.
It means the majority of data in an ICM profile (core content of the book) is useless if you don’t use color managed apps ?
A profile is meant to be used in color managed apps! or at least as a source for building a transformation between colorspaces like a LUT3D.
VCGT is just a very very useful extra for computer display profiles (VCGT=Video Card Gamma Table). There are other devices that could be profilled and do not have VCGT field: printer, scanners…
