Re: [darktable-user] calibration question off topic
Thank you, I managed to achieve very close calibration - sRGB 99.89 / Adobe RGB 80.1 I still have to try all the settings below. My brightness is likely too high. I calibrated at 190 cd (my room is bright) but will try to lower more - aiming 130 cd. Regards, B On 2017-04-25 08:52 PM, Guillermo Rozas wrote: Great writing, I didn't know about it. It follows the same path I used (only that I used DisplayCal's GUI), so it saves me from writing it ;) My config on DisplayCal (and some comments): --Display & instrument: In the particular case of the SpyderColor5 you should first use the menu "Tools/Import colorimeter corrections from...", and extract those corrections from the SypderColor software (pointing DisplayCal to the windows installer should be enough). These proprietary corrections match the color response of the colorimeter to the particular light source of the monitor. Then: - chose the instrument mode that correspond to your monitor technology (for your Benq GW2765 it should be White LED if I'm not mistaken). - optionally enable white and black level compensation (not strictly necessary for a LED IPS panel, it's a trade-off of longer times for a marginally better profile in this case) - leave correction at Auto, it knows what to do (if you used a proprietary mode for the instrument it will use None, as the correction is already done by the colorimeter itself). --Calibration: As mentioned in Pascal's post, you can use "Tools/Report on uncalibrated device" to check the base properties of the monitor (remember to reset all color/contrast options before). This will give you an idea of how far the monitor is from your objective: the farther it is, the more artifacts you'll get in the end. For example, trying to calibrate a monitor with a natural color temperature of 8500K (very bad laptops) down to 6500K will probably result in horrible color banding. - color temperature = 6500K (unless you have some specific needs) - white level = as measured - tone curve = gamma 2.2 (unless it's a Mac or other monitor with gamma close to 1.8) - calibration speed = high For a desktop monitor with color/gamma controls the best procedure is to first use those hardware controls to get as close as possible to the objective, and only latter do the calibration. In order to do that, check the option "Interactive display adjustment": before starting the calibration DisplayCal will show you a real time estimation of color temperature and brightness. Use the RGB controls of the on-screen monitor menu to get close to 6500K, and use the brightness controls to get close to 130cd (you'll need to iterate as one affects the other). Once you get something reasonably close, start the calibration. This ensures that the software corrections will be smaller, reducing the chances of artifacts. Very important: note down all the values of the on-screen monitor menu, the calibration is only valid for those values (change them and you'll need to recalibrate). --Profiling: -profile quality = high -test chart = auto optimized -amount of patches = 425 These settings are a good compromise of a moderately long profiling session (a couple of hours) for a high quality profile. I hope this helps! Regards, Guillermo On Tue, Apr 25, 2017 at 2:55 PM, johannes hanikawrote: you've seen pascal's old writeup about display colour profiling, right? https://encrypted.pcode.nl/blog/2013/11/24/display-color-profiling-on-linux/ -jo On Wed, Apr 26, 2017 at 5:35 AM, Guillermo Rozas wrote: As per https://wiki.archlinux.org/index.php/ICC_profiles "Note that the system on which the profile is generated must host the exact same video card and monitor for which the profile is to be used" And this contradicts to some extend with http://www.colorwiki.com/wiki/Cross_Platform_Color "They will function the same way on any operating system, and can be easily moved from machine to machine." I am actually using 2 different machines with 2 very different cards (one windows and one linux). I think one item that changes the behavior a lot is the color temperature. I set it manually to 6500K. Windows would set to this value by default while Linux would try to set to 7600K by default. On sRGB - windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% windows vs 79% linux. Yes, I think the colorwiki link was referring mainly to printers, is a bit misleading (my bad to include it here). Is the same discussion as with the driver: in principle the graphic card should not affect the colors that are sent to the monitor, but graphic cards can be so different between them that in practice it does matter. Same generation of the same vendor may be OK, two completely different GPU is probably not safe. And yes, color temperature is very important: the profile tries to match the color temperature you ask for, so a profile with a target of 7600K will look a lot bluer than one with a target of 6500K. You probably
Re: [darktable-user] calibration question off topic
Great writing, I didn't know about it. It follows the same path I used (only that I used DisplayCal's GUI), so it saves me from writing it ;) My config on DisplayCal (and some comments): --Display & instrument: In the particular case of the SpyderColor5 you should first use the menu "Tools/Import colorimeter corrections from...", and extract those corrections from the SypderColor software (pointing DisplayCal to the windows installer should be enough). These proprietary corrections match the color response of the colorimeter to the particular light source of the monitor. Then: - chose the instrument mode that correspond to your monitor technology (for your Benq GW2765 it should be White LED if I'm not mistaken). - optionally enable white and black level compensation (not strictly necessary for a LED IPS panel, it's a trade-off of longer times for a marginally better profile in this case) - leave correction at Auto, it knows what to do (if you used a proprietary mode for the instrument it will use None, as the correction is already done by the colorimeter itself). --Calibration: As mentioned in Pascal's post, you can use "Tools/Report on uncalibrated device" to check the base properties of the monitor (remember to reset all color/contrast options before). This will give you an idea of how far the monitor is from your objective: the farther it is, the more artifacts you'll get in the end. For example, trying to calibrate a monitor with a natural color temperature of 8500K (very bad laptops) down to 6500K will probably result in horrible color banding. - color temperature = 6500K (unless you have some specific needs) - white level = as measured - tone curve = gamma 2.2 (unless it's a Mac or other monitor with gamma close to 1.8) - calibration speed = high For a desktop monitor with color/gamma controls the best procedure is to first use those hardware controls to get as close as possible to the objective, and only latter do the calibration. In order to do that, check the option "Interactive display adjustment": before starting the calibration DisplayCal will show you a real time estimation of color temperature and brightness. Use the RGB controls of the on-screen monitor menu to get close to 6500K, and use the brightness controls to get close to 130cd (you'll need to iterate as one affects the other). Once you get something reasonably close, start the calibration. This ensures that the software corrections will be smaller, reducing the chances of artifacts. Very important: note down all the values of the on-screen monitor menu, the calibration is only valid for those values (change them and you'll need to recalibrate). --Profiling: -profile quality = high -test chart = auto optimized -amount of patches = 425 These settings are a good compromise of a moderately long profiling session (a couple of hours) for a high quality profile. I hope this helps! Regards, Guillermo On Tue, Apr 25, 2017 at 2:55 PM, johannes hanikawrote: > you've seen pascal's old writeup about display colour profiling, right? > > https://encrypted.pcode.nl/blog/2013/11/24/display-color-profiling-on-linux/ > > -jo > > On Wed, Apr 26, 2017 at 5:35 AM, Guillermo Rozas wrote: >>> As per https://wiki.archlinux.org/index.php/ICC_profiles >>> >>> "Note that the system on which the profile is generated must host the exact >>> same video card and monitor for which the profile is to be used" >>> >>> And this contradicts to some extend with >>> >>> http://www.colorwiki.com/wiki/Cross_Platform_Color >>> >>> "They will function the same way on any operating system, and can be easily >>> moved from machine to machine." >>> >>> I am actually using 2 different machines with 2 very different cards (one >>> windows and one linux). I think one item that changes the behavior a lot is >>> the color temperature. I set it manually to 6500K. Windows would set to this >>> value by default while Linux would try to set to 7600K by default. On sRGB - >>> windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% >>> windows vs 79% linux. >> >> Yes, I think the colorwiki link was referring mainly to printers, is a >> bit misleading (my bad to include it here). Is the same discussion as >> with the driver: in principle the graphic card should not affect the >> colors that are sent to the monitor, but graphic cards can be so >> different between them that in practice it does matter. Same >> generation of the same vendor may be OK, two completely different GPU >> is probably not safe. >> >> And yes, color temperature is very important: the profile tries to >> match the color temperature you ask for, so a profile with a target of >> 7600K will look a lot bluer than one with a target of 6500K. You >> probably want 6500K, as that is the standard for monitor viewing. >> Brightness is also important, to a lesser degree. >> >>> There are just a lot of options in Display Cal and even there is >>> documentation -
Re: [darktable-user] calibration question off topic
Thank you All! I was not aware of the article. It is very informative indeed. @Guillermo - if you create such sane options that you use - it would be very beneficial too. I can see there is quite a bit to learn in this area... admittedly I have been doing a few things the wrong way. Regards, B On 2017-04-25 10:55 AM, johannes hanika wrote: you've seen pascal's old writeup about display colour profiling, right? https://encrypted.pcode.nl/blog/2013/11/24/display-color-profiling-on-linux/ -jo On Wed, Apr 26, 2017 at 5:35 AM, Guillermo Rozaswrote: As per https://wiki.archlinux.org/index.php/ICC_profiles "Note that the system on which the profile is generated must host the exact same video card and monitor for which the profile is to be used" And this contradicts to some extend with http://www.colorwiki.com/wiki/Cross_Platform_Color "They will function the same way on any operating system, and can be easily moved from machine to machine." I am actually using 2 different machines with 2 very different cards (one windows and one linux). I think one item that changes the behavior a lot is the color temperature. I set it manually to 6500K. Windows would set to this value by default while Linux would try to set to 7600K by default. On sRGB - windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% windows vs 79% linux. Yes, I think the colorwiki link was referring mainly to printers, is a bit misleading (my bad to include it here). Is the same discussion as with the driver: in principle the graphic card should not affect the colors that are sent to the monitor, but graphic cards can be so different between them that in practice it does matter. Same generation of the same vendor may be OK, two completely different GPU is probably not safe. And yes, color temperature is very important: the profile tries to match the color temperature you ask for, so a profile with a target of 7600K will look a lot bluer than one with a target of 6500K. You probably want 6500K, as that is the standard for monitor viewing. Brightness is also important, to a lesser degree. There are just a lot of options in Display Cal and even there is documentation - some areas are lacking detailed explanation for a beginner. Yeah, it also took me a while to understand which options to use, and even after deciding I was second guessing my configuration all the time. I'll post tonight the options that I found the most "sane", maybe it's helpful for you as a start point. If I find the time I'll try to also write a small walk-through of what I did when I calibrated my monitor (including the color temperature options). Regards, Guillermo darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
you've seen pascal's old writeup about display colour profiling, right? https://encrypted.pcode.nl/blog/2013/11/24/display-color-profiling-on-linux/ -jo On Wed, Apr 26, 2017 at 5:35 AM, Guillermo Rozaswrote: >> As per https://wiki.archlinux.org/index.php/ICC_profiles >> >> "Note that the system on which the profile is generated must host the exact >> same video card and monitor for which the profile is to be used" >> >> And this contradicts to some extend with >> >> http://www.colorwiki.com/wiki/Cross_Platform_Color >> >> "They will function the same way on any operating system, and can be easily >> moved from machine to machine." >> >> I am actually using 2 different machines with 2 very different cards (one >> windows and one linux). I think one item that changes the behavior a lot is >> the color temperature. I set it manually to 6500K. Windows would set to this >> value by default while Linux would try to set to 7600K by default. On sRGB - >> windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% >> windows vs 79% linux. > > Yes, I think the colorwiki link was referring mainly to printers, is a > bit misleading (my bad to include it here). Is the same discussion as > with the driver: in principle the graphic card should not affect the > colors that are sent to the monitor, but graphic cards can be so > different between them that in practice it does matter. Same > generation of the same vendor may be OK, two completely different GPU > is probably not safe. > > And yes, color temperature is very important: the profile tries to > match the color temperature you ask for, so a profile with a target of > 7600K will look a lot bluer than one with a target of 6500K. You > probably want 6500K, as that is the standard for monitor viewing. > Brightness is also important, to a lesser degree. > >> There are just a lot of options in Display Cal and even there is >> documentation - some areas are lacking detailed explanation for a beginner. > > Yeah, it also took me a while to understand which options to use, and > even after deciding I was second guessing my configuration all the > time. I'll post tonight the options that I found the most "sane", > maybe it's helpful for you as a start point. If I find the time I'll > try to also write a small walk-through of what I did when I calibrated > my monitor (including the color temperature options). > > Regards, > Guillermo > > darktable user mailing list > to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org > darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
> As per https://wiki.archlinux.org/index.php/ICC_profiles > > "Note that the system on which the profile is generated must host the exact > same video card and monitor for which the profile is to be used" > > And this contradicts to some extend with > > http://www.colorwiki.com/wiki/Cross_Platform_Color > > "They will function the same way on any operating system, and can be easily > moved from machine to machine." > > I am actually using 2 different machines with 2 very different cards (one > windows and one linux). I think one item that changes the behavior a lot is > the color temperature. I set it manually to 6500K. Windows would set to this > value by default while Linux would try to set to 7600K by default. On sRGB - > windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% > windows vs 79% linux. Yes, I think the colorwiki link was referring mainly to printers, is a bit misleading (my bad to include it here). Is the same discussion as with the driver: in principle the graphic card should not affect the colors that are sent to the monitor, but graphic cards can be so different between them that in practice it does matter. Same generation of the same vendor may be OK, two completely different GPU is probably not safe. And yes, color temperature is very important: the profile tries to match the color temperature you ask for, so a profile with a target of 7600K will look a lot bluer than one with a target of 6500K. You probably want 6500K, as that is the standard for monitor viewing. Brightness is also important, to a lesser degree. > There are just a lot of options in Display Cal and even there is > documentation - some areas are lacking detailed explanation for a beginner. Yeah, it also took me a while to understand which options to use, and even after deciding I was second guessing my configuration all the time. I'll post tonight the options that I found the most "sane", maybe it's helpful for you as a start point. If I find the time I'll try to also write a small walk-through of what I did when I calibrated my monitor (including the color temperature options). Regards, Guillermo darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
Thank you.. I think there are few details that I have been missing. As per https://wiki.archlinux.org/index.php/ICC_profiles "Note that the system on which the profile is generated must host the exact same video card and monitor for which the profile is to be used" And this contradicts to some extend with http://www.colorwiki.com/wiki/Cross_Platform_Color "They will function the same way on any operating system, and can be easily moved from machine to machine." I am actually using 2 different machines with 2 very different cards (one windows and one linux). I think one item that changes the behavior a lot is the color temperature. I set it manually to 6500K. Windows would set to this value by default while Linux would try to set to 7600K by default. On sRGB - windows came at 100% and linux came at 99.9 while Adobe RGB came at 81% windows vs 79% linux. There are just a lot of options in Display Cal and even there is documentation - some areas are lacking detailed explanation for a beginner. For the record - I am using Monitor Benq GW2765 calibrator Spyder 5 Pro Both systems windows and linux are very different in hardware. I think the 2 are quite a bit closer now. Thank you all again! Regards, B On 2017-04-25 05:39 AM, Guillermo Rozas wrote: Another (more complete) link: https://wiki.archlinux.org/index.php/ICC_profiles There is actually a warning along the lines of "it should be OS-independent, but check that the OS and drivers are not doing funny things behind your back". Regards, Guillermo On Tue, Apr 25, 2017 at 9:34 AM, Guillermo Rozaswrote: Quick search: http://www.colorwiki.com/wiki/Cross_Platform_Color On Tue, Apr 25, 2017 at 9:30 AM, Guillermo Rozas wrote: 2017-04-25 13:58 GMT+02:00 Guillermo Rozas : In the end, the ICC profile should be independent of the operating system and only depend on the graphic card + monitor combination. Hum I would have thought that it is also dependent on the actual graphic/display driver, no? If I understand it correctly, the ICC profile is only a mapping of "RGB values" to "corrected RGB values that, when feed into the driver, produce the intended color when translated to monitor light intensities". In that sense, it's true that the driver may change something. However, I would expect that the part of the driver that does the RGB -> light intensity translation is so low level that it SHOULD be independent of the operating system, specially if you use the proprietary drivers on both systems. Sadly I currently don't have my color profiler at hand to test this. Regards, Guillermo darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] Basecurve for LGE Nexus 5X.. how to do it?
Hi Malte, you could try darktable-chart . Of all the methods that I have tried it seems to give the best comparison to the in camera jpeg. It not only matches the tone curve of the image but also the color. Have a look at, http://weeklyedit.com/camera-profiling-darktable-chart/ You will need a color chart though. I used a wolf-faust scanner calibration chart that I already had because it has 288 color patches and therefore should produce a more accurate color match you can get one from, http://www.targets.coloraid.de/ Hope this helps. Tim. On Monday, 24 April 2017 23:26:18 BST Malte Cornils wrote: > Hi Tim, > > thanks - it's certainly helpful for the process, and I redid a whole ISO > series now. Unfortunately, I still don't get anything resembling a proper > curve. > > Any other ideas? > > -Malte > > 2017-04-23 15:42 GMT+02:00 Tim Rolph: > > Hi Malte, have you seen this video on basecurves? > > > > http://weeklyedit.com/basecurves/ > > > > Tim. > > > > On Saturday, 22 April 2017 23:14:13 BST Malte Cornils wrote: > > > Hi, > > > > > > although there is now even denoise support for the RAW DNG's created by > > > LG's Nexus 5X phone main camera, the basecurve between JPG and RAW seems > > > very much off. I was trying to follow > > > http://www.darktable.org/2013/10/about-basecurves/ to create my own > > > basecurve, but the result was much worse: > > > > > > While dt-curve-tool-helper worked following a mixture of the article and > > > the tools/basecurve/README.md for both RAW DNGs/JPGs, and the curve tool > > > invocation generated a useable shell script (attached), this did not > > > look > > > like a curve at all and had horrible colours. > > > > > > I tried with both a black/white picture following the blog instructions, > > > > as > > > > > well as a normally exposed one. You can find them here (my own work and > > > CC0-licensed): > > > > > > https://srv.cornils.net/DSC_0132.DNG > > > https://srv.cornils.net/DSC_0132.JPG > > > https://srv.cornils.net/DSC_0135.DNG > > > https://srv.cornils.net/DSC_0135.JPG > > > > > > (will leave them up there for at least six months) > > > > > > How can I get a reasonable basecurve out of those images? They were > > > generated by FV-5 camera app, which has DNG RAW support. > > > > > > Thanks > > > -Malte > > > > > > > > > > > > > > > darktable user mailing list > > > to unsubscribe send a mail to darktable-user+unsubscribe@ > > > > lists.darktable.org > > > > > > > > > > darktable user mailing list > > to unsubscribe send a mail to darktable-user+unsubscribe@ > > lists.darktable.org > > > darktable user mailing list > to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
Another (more complete) link: https://wiki.archlinux.org/index.php/ICC_profiles There is actually a warning along the lines of "it should be OS-independent, but check that the OS and drivers are not doing funny things behind your back". Regards, Guillermo On Tue, Apr 25, 2017 at 9:34 AM, Guillermo Rozaswrote: > Quick search: http://www.colorwiki.com/wiki/Cross_Platform_Color > > On Tue, Apr 25, 2017 at 9:30 AM, Guillermo Rozas wrote: >>> 2017-04-25 13:58 GMT+02:00 Guillermo Rozas : In the end, the ICC profile should be independent of the operating system and only depend on the graphic card + monitor combination. >>> >>> >>> Hum I would have thought that it is also dependent on the actual >>> graphic/display driver, no? >> >> If I understand it correctly, the ICC profile is only a mapping of >> "RGB values" to "corrected RGB values that, when feed into the driver, >> produce the intended color when translated to monitor light >> intensities". In that sense, it's true that the driver may change >> something. >> >> However, I would expect that the part of the driver that does the RGB >> -> light intensity translation is so low level that it SHOULD be >> independent of the operating system, specially if you use the >> proprietary drivers on both systems. Sadly I currently don't have my >> color profiler at hand to test this. >> >> Regards, >> Guillermo darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
2017-04-25 13:58 GMT+02:00 Guillermo Rozas: > In the end, the ICC profile should be independent of the operating > system and only depend on the graphic card + monitor combination. Hum I would have thought that it is also dependent on the actual graphic/display driver, no? -- Pascal Obry / Magny Les Hameaux (78) The best way to travel is by means of imagination http://photos.obry.net http://www.obry.net gpg --keyserver keys.gnupg.net --recv-key F949BD3B darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] calibration question off topic
I'm nowhere near an expert on this (I've only used DisplayCal twice so far), but: - Have you checked that the configuration on both cases is exactly the same, including the (optional) light source corrections? (the latter come with some vendor specific drivers for some detectors). You can use the profile generated in Windows as the configuration source in Linux to be sure ("load configuration", I think), although I remember it silently discarded options if they were not available (for example, those light source corrections). - Another thing to check is how repetitive is the profile generation. How many points are you using? Too few points can result in inconsistent fittings, specially for the low light part. You can check it by repeating the profile generation in Windows and looking at the resulting curves: slight differences are reasonable because of the empirical fitting involved, big differences mean either a bad calibration procedure or a bad and inconsistent monitor. - I also remember that on Linux DisplayCal threw at me some warning about 8bit vs 16bit support for the graphic card's LUT. I had no time to check what was the origin of this (or if it made any difference), but it MAY be the reason you get two very different profiles. In the end, the ICC profile should be independent of the operating system and only depend on the graphic card + monitor combination. What I did at the time was to just use the profile generated under Windows (I think there was no warning then), pending some investigation of the Linux generation procedure (which I never did...). If you want I can send you the configuration I used latter today when I get home. Regards, Guillermo On Mon, Apr 24, 2017 at 10:08 PM, I. Ivanovwrote: > Hi All, > > I know it quite a bit off topic but does anybody have best practices for > calibration with display cal? I am using it on Ubuntu 16.04 and what I > noticed is - if I use it - it only achieved 99.7% sRGB and ~77% Adobe RGB. > If I calibrate the same monitor on windows - it will achieve 100%sRGB and > ~81% Adobe RGB. > > Sadly - the pictures look quite different in DT. I am comparing the 2 by > taking the windows produced ICC and copying it on the linux. > > I am testing LUT profiles and Matrix profiles but I am quite puzzled. > > Thank you... > > B > > > darktable user mailing list > to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org > darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org
Re: [darktable-user] Pentax KP RAW (not?) fully supported?
Am Sun, 23 Apr 2017 11:17:20 +0300 schrieb Roman Lebedev: Hi Roman, > > I downloaded some RAW-files from > > http://www.photographyblog.com/previews/pentax_kp_photos/ and it > > looks like these PEF-files from the Pentax KP are not fully > > supported, eg with white balance. > (we do have a dng+pef on rpu, so that is good.) sorry, I do not understand what rpu means. > > Is the Pentax KP on the to-do-list? > https://redmine.darktable.org/projects/darktable/issues/new I cannot access this link, because I have to login and I am not registered. How long do you think it takes, that the PEF from a KP is supported? Al darktable user mailing list to unsubscribe send a mail to darktable-user+unsubscr...@lists.darktable.org