Thread: [dispcalGUI-users] How do I calibrate two displays to the same color?
Display Calibration and Characterization powered by Argyll CMS
Brought to you by:
fhoech
From: Jan N. F. <Jan.Niklas@Fingerle.org> - 2015-04-27 22:17:13
|
Hi, you /may/ have seen my problem on StackExchange[1], but it wasn't answered there and I'm using dispcalgui, I give it a try here. I've tried to calibrate two LCD displays, one brand-new LED one and an old-ish CCFL to "the same" colours (6500K, 140cdm²) on LinuxMint using my Colormunki Display (and two other colorimeters before that). With different settings for profiles etc. the outcome is essentially always the same: On a first glance the colours seem to match, on a second, closer look the CCFL display has a slight red-ish tint compared to the LED display. So, skin tones are very much off. Looking at the TRC curves in the LinuxMint Colour Settings, the curves for RGB lie almost on top of each other for the LED screen, while for the CCFL screen the red curve is a bit steeper than the other two. You can have a look at examples at StackExchange[1]. For the little I know, these two findings (red-ish tint and steeper red curve) seem to go hand in hand. I don't understand what's going wrong and why the TRC curves are that different, given that I calibrated the screens to the target colours and brightness before starting the calibration process. Is there anything I can do? Would you expect me to get closer matches with a spectrometer instead of a colorimeter? Cheers, Jan Niklas [1] http://photo.stackexchange.com/questions/61721/how-do-i-calibrate-two-displays-to-the-same-color-lcd-led-backlight-and-ccfl |
From: Florian H. <dis...@ho...> - 2015-04-27 23:31:47
|
Hi, Am 27.04.2015 um 23:50 schrieb Jan Niklas Fingerle: > you /may/ have seen my problem on StackExchange[1] the only forums that I frequent regularly are the official dispcalgui help forums on sourceforge, this mailing list, and the "display calibration" section on AVSforum.com. In short, posting here was the right idea :) > I've tried to calibrate two LCD displays, one brand-new LED one and an > old-ish CCFL to "the same" colours (6500K, 140cdm²) on LinuxMint using > my Colormunki Display (and two other colorimeters before that). > > With different settings for profiles etc. the outcome is essentially > always the same: On a first glance the colours seem to match, on a > second, closer look the CCFL display has a slight red-ish tint compared > to the LED display. So, skin tones are very much off. There's many variables here. - What colorimeter correction(s) have been used (if any)? The ColorMunki display has corrections for both white LED and CCFL which can be imported from the vendor software. - Does fullscreen white match visually between the two displays? If not, this is the first thing to do. If colorimeter corrections alone do not produce a satisfactory match, choose one of the screens as your reference, adjust the other to match the white. You can do this purely visually. Then, re-profile the adjusted display. - What's the gamut of each display? For a good screen-to-screen match, the gamuts of both displays should ideally encompass the most common image colorspace you're using (e.g. sRGB). - Are the applications you are using to view images color managed, i.e. do they actually use the installed display profiles? > Looking at the TRC curves in the LinuxMint Colour Settings, the curves > for RGB lie almost on top of each other for the LED screen, while for > the CCFL screen the red curve is a bit steeper than the other two. You > can have a look at examples at StackExchange[1]. Not sure how to interpret the curves due to missing axis labels, but I think the Gnome Colour settings TRC curve plot traditionally shows the amount of correction needed to hit D50. This is really only useful information if you want D50 on a non-D50 display. > Is there anything I can do? Possibly, see my comments/questions above. > Would you expect me to get closer matches > with a spectrometer instead of a colorimeter? A good visual match can usually also be obtained with a colorimeter. A spectro can remove a good deal uncertainty though. Florian. -- Florian Höch http://dispcalGUI.hoech.net GPG: 770C 1FC4 9A8C 33E7 5794 A9DF 8EA3 1827 1BFD 1EBE |
From: Jan N. F. <Jan.Niklas@Fingerle.org> - 2015-04-28 21:44:38
|
Hi, thanks for the fast reply. I'll try to answer as good as possible for now. Any testing and trying will have to be delayed until the weekend. Am 28.04.2015 um 01:18 schrieb Florian Höch: > In short, posting here was the > right idea :) great. :-) > - What colorimeter correction(s) have been used (if any)? The ColorMunki > display has corrections for both white LED and CCFL which can be > imported from the vendor software. I've tried that, but wasn't really able to make sense of the correction matrix names. I've tried a downloaded (from your site) correction for my specific LED screen with the ColorMunki Display, which didn't make much of a difference. For the CCFL display I've tried the two corrections downloaded from the vendor software that seemed to make some kind of sense (the name suggested CCFL, but not my specific display). > - Does fullscreen white match visually between the two displays? In not-color-managed software (Tunderbird, LibreOffice), it does. At least the CCFL may have a very, very slight yellow/yellow-green tint, but in no way the red-ish tint that I mentioned in my original mail and only if I really want to see a difference. My goal is to have a working environment for darktable, which is color managed and my primary test object: If I take the darktable window from the LED to the CCFL screen, the colours look essentially the same for the better part of one second (while darktable tells me that it is "working"), then the colours switch to the tinted version. > - What's the gamut of each display? For a good screen-to-screen match, > the gamuts of both displays should ideally encompass the most common > image colorspace you're using (e.g. sRGB). It's a (LED:) Dell U2412M and a (CCFL:) Fujitsu P24W-5. Both aren't wide gamut, but should be reasonably well equipped for sRGB. > - Are the applications you are using to view images color managed, i.e. > do they actually use the installed display profiles? Yes. At least (a) I'd expect it from darktable (b) the documentation states this explicitly (as well as that they handle two displays the right way[tm] and (c) the delay mentioned above seems to indicate that they're doing something differently for the other display. > Not sure how to interpret the curves due to missing axis labels, Well, yes, sorry, I clipped those, since I didn't want to put german labels on an english forum. The x axis is called "Eingangs-Antwort", the y axis "Ausgangsantwort" and it's explained that "eine Farbtonreproduktionskurve ist die Zuordnung der Leuchtdichte der Szene zur Leuchtdichte der Anzeige". (Sorry for the German to the non-german readers of this mailing list.) > but I > think the Gnome Colour settings TRC curve plot traditionally shows the > amount of correction needed to hit D50. This is really only useful > information if you want D50 on a non-D50 display. Maybe. I'm not at all an expert. Actually, I don't want anything fancy. Just the usual "the prints should fit the screen" for different kind of prints. I'm well aware, that this is easier said than done. But, for a start, it would be great to have the screens matching - but if I get the skin tones right on one, people look like freshly boiled (if I go from LED to CCFL) or like ghosts (the other way round) on the other. >> Would you expect me to get closer matches >> with a spectrometer instead of a colorimeter? > > A good visual match can usually also be obtained with a colorimeter. A > spectro can remove a good deal uncertainty though. OK, since a spectrometer isn't the cheapest thing to buy, it'd be great if it worked with the colorimeter. :-) Thanks, so far, for your help, cheers, Jan Niklas |
From: Florian H. <dis...@ho...> - 2015-04-28 23:16:29
|
Am 28.04.2015 um 23:44 schrieb Jan Niklas Fingerle: > For the CCFL display I've tried the two corrections > downloaded from the vendor software that seemed to make some kind of > sense (the name suggested CCFL, but not my specific display). Yes, they are generic corrections, but based on high res spectral data obtained from a laboratory-grade spectrometer (Minolta CS-2000). > In not-color-managed software (Tunderbird, LibreOffice), it does. At > least the CCFL may have a very, very slight yellow/yellow-green tint, > but in no way the red-ish tint that I mentioned in my original mail and > only if I really want to see a difference. Ok, if it's a reasonable match that's a good starting point. > My goal is to have a working environment for darktable, which is color > managed and my primary test object: If I take the darktable window from > the LED to the CCFL screen, the colours look essentially the same for > the better part of one second (while darktable tells me that it is > "working"), then the colours switch to the tinted version. Some sanity checking is probably in order. Pick one (or several) of the images you're having problems with. Open a terminal and use Argyll's cctiff or littleCMS' tifficc (if you have created the display profiles without black point compensation, I'd recommend the latter) to convert the images using the display profile: cctiff -p -ir <image>.jpg -ir <display>.icc <image>.jpg <out>.jpg or jpegicc -i <image>.jpg -o <display>.icc -t1 -b <image>.jpeg <out>.jpg <image>.jpg should be the path to the source image <display>.icc should be the path for each display profile <out>.jpg should be the output filename Then, view the images on each respective display in an image viewer that is *not color managed or has it turned off* (important!). Any viewer that explicitly allows to turn color management off should work (e.g. the Gimp). If you used calibration, make sure to load it for each display (dispwin <display>.icc). If you're seeing the same differences, then the next step would be a look at the profiles and the settings used to create them. If the differences go away, this would indicate a problem with darktable or the multi-monitor setup. > It's a (LED:) Dell U2412M and a (CCFL:) Fujitsu P24W-5. Both aren't wide > gamut, but should be reasonably well equipped for sRGB. The Fujitsu should be wide-gamut, and has 100% sRGB coverage according to prad.de, while the Dell has a standard gamut with slightly less sRGB coverage (93%, still good). Florian. -- http://dispcalGUI.hoech.net GPG: 770C 1FC4 9A8C 33E7 5794 A9DF 8EA3 1827 1BFD 1EBE |
From: Florian H. <dis...@ho...> - 2015-04-28 23:23:25
|
> If you used calibration, make sure to load it for each > display (dispwin <display>.icc). Correction: The command should read dispwin -d<number of display> <display>.icc Florian. -- http://dispcalGUI.hoech.net GPG: 770C 1FC4 9A8C 33E7 5794 A9DF 8EA3 1827 1BFD 1EBE |
From: Jan N. F. <Jan.Niklas@Fingerle.org> - 2015-05-01 16:24:37
|
Hi, > Some sanity checking is probably in order. thanks, I wasn't aware of these commands. Using these checks and viewing those converted images with the Gimp (color management deactivated) everything looked fine. Checking back with darktable, it was using Colord *and* XAtom to find the correct profile. The weird behaviour disappeared when switching to XAtom only. So, there's something weird going on, especially as darktable's command "darktable-cmstest" reports [...] HDMI1 the X atom and colord returned the same profile [...] HDMI2 the X atom and colord returned the same profile [...] Your system seems to be correctly configured Anyway, your help with the sanity checks helped me to look into the right direction. :-) > If the differences go away, this would indicate a problem with > darktable or the multi-monitor setup. Looks like it. >> It's a (LED:) Dell U2412M and a (CCFL:) Fujitsu P24W-5. Both aren't >> wide gamut, but should be reasonably well equipped for sRGB. > > The Fujitsu should be wide-gamut, One never stops to learn. I was sure, it wasn't. Anyhow, thanks for the help. Cheers, Jan Niklas |