- This topic has 31 replies, 3 voices, and was last updated 3 years, 3 months ago by
.
-
Posts
-
To Vincent: Is my PV270 correction usable here (both ColorMunki and non-NIST i1p2 should be avaible)? The PV270 seems to have the same IPS panel.
You can share them of course. Thanks for sharing
Are they in colorimeter database?
Which one?
To Irfan: BenQ SW2700PT comes with Palette Master Elemets. This tool is limited and produces poor results cause of few pathces used, I’ve tried it with BenQ SW240 a month ago. DisplayCAL makes excelent calibration for SW240, but you can’t use native display superb gamut (SW240 does not have Rec.2020 setting to get it) and you can’t use your display for AdobeRGB or sRGB emulation (i.e. setting AdobeRGB as display ICC profile in system). This shouldn’t be a problem for modern software, but some hoary tools like 3ds Max are better with standard ICC emulation. To make colors limited to some standard gamut you should set it in display menu, then start calibration and profiling process. …Hm, try to download Palette Master (non-Elements) out of PV270 support pages. Does it recognize your display? Note: Argyll/DisplayCAL supports ICC v.2 format only, select it in other software to make 3D LUT later.
AFAIK PM is only for PV line (another i1Profiler clone like Dells or viewsonics). SW line is limited to PME. I would be glad to be wrong here.
Note on BenQ: I’ve met SW240 and PV270 twice. Old ones (including my display) have non-homogenous panels, but new ones (2019-2020 mfg years) are pretty good. The most homogeous panel I’ve met is 4K LG 27UL850-W that is sRGB+ display (good for Rec.709) with internal LUT. Tomorrow I will calibrate the second unit and I will test its proprietary software. I’ve never met Eizo here (it’s a dream).
Newer P3 (and only P3) multimedia displays seems to have good color uniformity. Nightmare comes when you move to AdobeRGB green.
I downloaded one from a Munki, the one with shortest name:
Seems fine, a native gamut sample of a QLED. Unusual choice of white point… but EDRs from Xrite store even more strange ones… Is it yours Alexei? Seems OK.
@Irfan, IDNK which one you selected before, but this one seems fine.I downloaded one from a Munki, the one with shortest name:
Seems fine, a native gamut sample of a QLED. Unusual choice of white point… but EDRs from Xrite store even more strange ones… Is it yours Alexei? Seems OK.
@Irfan, IDNK which one you selected before, but this one seems fine.Hi Vincent, sorry was on a shoot, here in this thread Florian had shared a cleaned version
https://hub.displaycal.net/forums/topic/benq-sw2700pt-2/
Which looks like the attachment. Should i use that or the one you sent in this link?
Other Questions
since i dont have native in osd controls, do i select Adobe RGB or rec 709 on osd? i have gamma controls, do i select 2.4 or 2.2? my sony reference is on 2.4.
the checkbox “apply calibration (vcgt)” should be checked or unchecked?
Also the tone curve in displaycal should be 1886 2.4 absolute with offset 0
Or custom 2.4 relative offset 100?
I downloaded one from a Munki, the one with shortest name:
Seems fine, a native gamut sample of a QLED. Unusual choice of white point… but EDRs from Xrite store even more strange ones… Is it yours Alexei? Seems OK.
@Irfan, IDNK which one you selected before, but this one seems fine.Hi Vincent, sorry was on a shoot, here in this thread Florian had shared a cleaned version
https://hub.displaycal.net/forums/topic/benq-sw2700pt-2/
Which looks like the attachment. Should i use that or the one you sent in this link?
They are the same, but the one you plotted seem to be not as warm as the one I linked, although for a CCSS it should not matter.
Use florian’s.Other Questions
since i dont have native in osd controls, do i select Adobe RGB or rec 709 on osd?
OSD preset is a gamut simulation where (simulated) primaries are produced as a mix of native primaries. That mix can be wrong. Measure. Choose the one that has more % of Rec709 inside. A priori we can think as AdobeRGB OSD preset as 1st candidate… but maybe its red is a little undersaturated , so if due to monitor limitations you are forced to use some factory calibration with some smaller than native colorspace simulation, measure first.
i have gamma controls, do i select 2.4 or 2.2? my sony reference is on 2.4.
Measure actual value for each setting (actual value may not match OSD label), not “mean/median value” but overall “gamma vs input 2D plot”. Choose the one better suits you.
Keep in mind that “usually” people wnat a multipurpose monitor. So as I said previously they set 2.2 gamma, calibrate & profile to get a VCGT calibration applied system wide and an ICC profile for color managed apps like Photoshop, Acrobat, some browsers (maybe browsers need simple matrix display profiles).. etc.
Then for a LUT3D meant for Rec709 2.4 content they make a LUT3D and that LUT3D has rec709 color transformation and 2.4 gamma too.IDNK what you want. If you want 2.4 system wide, try to get OSD controls close to that. If you want 2.2 on desktop apps choose osd preset close to that, then make LUT3D to 2.4. Is up to the color engine in that app that supports LUT3D the severity of rounding errors applying LUT3D, having dither and such. IDNK these details regarding Resolve. MadVR has no issues AFAIK.
the checkbox “apply calibration (vcgt)” should be checked or unchecked?
If you are going to apply VCGT calibration for all desktop apps, use that display profile as default display profile in OSD and use that ICC profile to compute a LUT3D for resolve and use that monitor as GUI, then uncheched.
ICC profile excluding VCGT stores display behavior AFTER calibration (after VCGT was applied). Hence color transformations from Rec709 to that display description do not need to know about which GPU calibration you applied (VCGT) AS LONG AS THAT CALIBRATION IS APPLIED PREVIOUSLY SYSTEM WIDE (like with DisplayCAL lut loader app in the taskbar tray).
With MadVR LUT3D VCGT data is used to make LUT3D because “usually” people configure it to clean GPU calibration when madVR is running, so there is no GPU induced banding (because LUT truncation to 8bit and all that stuff in nvidias and intel IGPUs). madVR engine does colro transformation at high bitdepth then truncate to whateever you set as output using its own dithering, hence no banding. But that’s madVR.
AFAIK Resolve has no button/configuration to clear GPU LUTs so it uses whatever is loaded by OS/3rd party apps like DisplayCAL (with the quality these apps can provide)
If you want that Resolve applies full calibration LUT3D, inlcuding grey (VCGT check) you cannot have VCGT appiled system wide (yiou cannot have as default display profile a profile with not linear input=output VCGT ).More versatile setup (if there is no GPU induced banding) is apply display profile and its VCGT calibration system wide on OS, then make GUI LUT3D for resolve with VCGT unchecked.
Sorry for long explanation, I was trying to explain why you need to use checked or uncheked when making LUT3D.
Firstly. Thank you for such detailed explanation. I read it several times to understand what is going on. And it makes a lot of sense. Thank you for being patient and extremely helpful! Much much more than all my googling and searching.
According to my understanding, there are 3 things.
Vcgt this stores the gamma tone information and on its face has nothing to do with colour transformations. I.e. It just tells what grey should be how according to the tone curve selected in tone curve settings.
ICC stores all color transformations and takes into account vcgt (irrespective of apply vcgt checked or unchecked) i.e tells every colour or the patches how they should be at every tone step of vcgt from source color space to my monitor.
3DLUT is ICC like but with greater information.
Some apps read icc like photoshop and some read LUT like resolve.
The checking and unchecking of apply vcgt is whether to apply this gamma tone curve information to the system OS wide or do not apply this information.
Unchecked situations
If I keep this apply vcgt unchecked, install ICC and keep ICC as default for my monitor in windows color management (displaycal tray profile association is a mirror of) and view a photo in photoshop I’ll see a proper gamma curve as well as color since icc took into account the vcgt information.
If I keep this apply vcgt unchecked, make a 3DLUT out of icc, install ICC and keep this icc default in windows color management, apply lut in resolve color viewer, then I’ll see a proper image and gamma curve cuz now the 3DLUT also has the vcgt information that it took from icc which icc too had from VCGT while making itself
Apply VCGT checked situations
If I keep this checked, it will apply this gamma tone information OS wide. If I then install ICC and set this as default in windows color management, view a photo in photoshop, I’ll see a wrong image cuz now I will be doubling VCGT. So essentially ICC+VCGT(the one taken into account while making icc)+VCGT (that is checked and applied os wide)
Similarly if I keep this checked install ICC and set this as default in windows color management, make a 3DLUT and apply in resolve, I’ll see a wrong image cuz now I am viewing VCGT(applied os wide)+3DLUT+VCGT(the one lut got from icc)
Is this a correct understanding? If yes I have some more questions.
I just calibrated with tone curve as custom 2.4 relative and offset at 100%. Attached is the report. I can see some bad drifts in colour in dark areas. is this just the limitation of this monitor?
from the tutorial
The “Resolve” preset is set up to not use iterative gray balance calibration, but as we are going to create a profile that will also be installed to the operating system, you may want to enable it. Set calibration tone curve on the “Calibration” tab from “As measured” to “Rec. 1886” (or another desired curve). Note that this setting does not influence the 3D LUT tone curve, which you can adjust in the next step.
This means, In 3DLUT tab, i can change the percentage of offset without needing to calibrate again right?
About the Tone Curve
from my understanding based on various Florian’s words in other threads, absolute is a technical curve. i.e. from the source definition itself. It doesn’t care about how the display being calibrated is or behaves.
Relative is a curve that takes into consideration the response of the display being calibrated. the offset from 0-100 manages how much it takes into consideration this display’s response. 0 being nothing (close to absolute) 100 being as much as possible. is this correct? please correct me if i am wrong
Also how does the black output offset work in Absolute?
The way we can change tone curve in 3DLUT without needing to calibrate again, is that not possible in creating an ICC?
Thank you
Vcgt this stores the gamma tone information and on its face has nothing to do with colour transformations. I.e. It just tells what grey should be how according to the tone curve selected in tone curve settings.
Not exactly like that. VCGT corrects in GPU grey so it behaves as you want. => “calibration curves” in displaycal profile info
Think of it like “training in a gym”Display gamma / tone response curve is actual display behavior in greys, usually measured after calibration and stored in ICC “TRC” tags. => Tone response curve in displaycal profile info.
Think of it as “taylor made suit” for grey. It can be measured without calibration or with VCGT appiled.ICC stores all color transformations and takes into account vcgt (irrespective of apply vcgt checked or unchecked) i.e tells every colour or the patches how they should be at every tone step of vcgt from source color space to my monitor.
Not exactly like that.
ICC file does nothing. It is just a taylor made suit from (“usually” after calibration) display measurements. It stores how display behaves. Nothing more nothing less.
App color engines deal with such transformations form one ICC profile to another. For example Adobe Color engine (ACE), or LittleCMS, or apple Colro Sync.
A LUT3D is just a crystalized instance of a particular transformation from one colorspace (Rec709) to another (your display). LUT3D (those particular transofrmation instances) have a ICC equivalent called “device link” ICC profile.
3DLUT is ICC like but with greater information.
No, explained above.
The checking and unchecking of apply vcgt is whether to apply this gamma tone curve information to the system OS wide or do not apply this information.
No, it need to be read in opposite direction “if OS applies this correction (VCGT) then do not apply it in LUT3D creator”, which is different than “by checking or unchecking VCGT i will make OS load VCGT”.
By checking or unchecking VCGT in LUT3D creator you do not force OS to behave that way. You check or uncheck that checkbox reacting to whatever OS (or LUT loadrs like DIsplaycal) are going to do, so you do not apply them “twice” or “never”, just “once”.
Unchecked situations
If I keep this apply vcgt unchecked, install ICC and keep ICC as default for my monitor in windows color management (displaycal tray profile association is a mirror of) and view a photo in photoshop I’ll see a proper gamma curve as well as color since icc took into account the vcgt information.
It took account of how display behaves after VCGT is applied. Photoshop or other toosl do not need to know that there was a VCGT applied. They just look how ICC display profile says that display behaves. If while running Photoshop you clean GPU calibration, display & ICC no longer match and you see innacurate grey colors.
If I keep this apply vcgt unchecked, make a 3DLUT out of icc, install ICC and keep this icc default in windows color management, apply lut in resolve color viewer, then I’ll see a proper image and gamma curve cuz now the 3DLUT also has the vcgt information that it took from icc which icc too had from VCGT while making itself
No, LUT3D expects dusplay to behave like ICC says. You may write it as “it expects that grey is corrected by some 3rd party agent”, so it does not care about it.
Apply VCGT checked situations
If I keep this checked, it will apply this gamma tone information OS wide.
no. It will apply VCGT calibration to LUT3D, nothing nore, nothing less.
If I then install ICC and set this as default in windows color management, view a photo in photoshop, I’ll see a wrong image cuz now I will be doubling VCGT.
No. These app usually do not use LUT3D. They use default ICC for display as configured in OS. As long as this display profile matches display behavior.. all this are ok
If you use such LUT3D in a LUT3D compatible app and VCGT correction/calibration is loaded into GPU you may see things wrong since it is applies twice. This “error” only apply for LUT3D compatible apps (or icc equivalente “device links”) and only for them.
I just calibrated with tone curve as custom 2.4 relative and offset at 100%. Attached is the report. I can see some bad drifts in colour in dark areas. is this just the limitation of this monitor?
Report looks fine, maybe grey range a*b* can be improved a little by choosing slower calibration for display ICC but IDNK your partiocular situation.
Ready to work with Rec709 content.
I cannot explain how helpful you have been Vincent. It all makes sense now. And I have a much much better understanding of how this works.
Hats off to Florian for making such a brilliant program. I will surely donate once I wrap my head around all this and can calibrate all the displays around me.
Last few questions if you can answer please!
In the documentation it says by default resolve preset is set to tone curve of ‘as measured” since usually a grey iteration (I understand this is the vcgt and 1D) is not required.
Question 1. If it’s set to as measured, then it never calculates and calibrates the grey. Is this right?
Question 2. If it never calculates then apply vcgt checked will have no effect on the 3DLUT cuz Grey’s was never calculated or calibrated. Is this right?
Question 3. What does usually not required mean? Should displays be used with a curve they naturally produce and is it a good practice not to apply a 1D correction?
Question 4. If you could explain or if you could send me to a link where I can understand absolute and relative in custom curves and what black output offset does in both absolute and relative. That would be wonderful! Cuz from what I understand absolute pertains to absolute values. 0 means 0. 50 means 50. But the display might not be able to do 0 and 50 precisely and this might cause inconsistency between displays. Relative modifies the curve to better match how visually separate displays look. Please correct me if I am wrong. But I don’t understand how the black output offset behaves in absolute and relative.
Hi, Vincent! This Munki’s ccss could be mine by its date (I tried PM and DisplayCAL both, and I made i1p2 correction too).
Bad new are: the second LG 27UL850-W was returned to shop cause of significant tint and brithness flaws. Also note that profiles by its True Color Pro for internal LUT produce dirty and posterized gradients.
After some investigation I’ve found that most sRGB (and wide gamut) IPS non-TV panels are produced by AU Optronics (AUO), LG, Samsung and Sharp and Innolux for notebooks. I don’t know what panels Eizo uses. So:
- Samsung makes mostly curved panels of 27″ size (they’re look pretty good). Probably flat ones are used in HP displays, but HP shows bad results.
- LG makes specific 21:9 panels (some with sRGB coverage), but its 4K 27″ panels production is not stable.
- I’ve only met Sharp panel in expensive Dell XPS 15 AdobeRGB+ and some Asus model notebook. Having superb gamut both had terrible flaws.
- AU Optronics is a main manufacturer for middle-class displays and third party assemblers like BenQ, ViewSonic, AOC etc. Not bright at one side, but sometimes good enough at the other. By example, BenQ PD2705Q has got CalMAN and Pantone certificates (Benq doesn’t declare technical values to customer, though). This means better quality control (factory panels selection, probably) and its test at prad.de shows pretty good results.
- Eizo CS and, better, CG models (LED has used since yr. 2016) have good enough uniformity, they have workable internal chroma correction, BenQ’s correction is weak. Eizo is very expensive, of course.
- I’ve never met full P3 coverage for PC platform and external non-Apple dispalys. Some PC notebooks have P3 coverage, it seems that they use Apple panels.
- The only inexpensive and good enough sRGB+ PC notebook panel is 16.1 inch Innolux N161HCA-EA3. I’ve met it in Honor MagickBook Pro, that unfortunately has crystal effect on screen.
I suspect truckers in all sins: here in Russia they may carry displays from China through severe frost.
Hi, Vincent! This Munki’s ccss could be mine by its date (I tried PM and DisplayCAL both, and I made i1p2 correction too).
Bad new are: the second LG 27UL850-W was returned to shop cause of significant tint and brithness flaws. Also note that profiles by its True Color Pro for internal LUT produce dirty and posterized gradients.
AFAIK LG has not and never had proper spectral correction for their HW calibration software for widegamut displays. At best it is using the wrong RGB LED correction for all: older GB-LEDs, newer P3 WLED PFS etc.
After some investigation I’ve found that most sRGB (and wide gamut) IPS non-TV panels are produced by AU Optronics (AUO), LG, Samsung and Sharp and Innolux for notebooks. I don’t know what panels Eizo uses. So:
Panel manufacturer is not that inportant, it’s QC. You can pay LG or AUO to get some certainty regarding panel quality… but that price is not going to be the same as general QC batches.
I’m pretty sure that veteran AH-IPS GBLED panels for nec or eizo were the same panel model used Dell or LG or HP counterparts for their widegamuts, but they requested (and paid) certain QC that others cannot pay if they want to offer a product in the price range they want.If low cost widgeamuts hace a target price for including panel parts, assembly, predicted warranty issues, HW cal software.. etc. you as display assembler (dell &etc) cannot pay more than X for extra QC because you go off your target price.
has got CalMAN and Pantone certificates
Calman & pantone certificates means little. Or technicolor. Try to find tolerances and number of points checked. At best is just factory calibration at center meeting some unknown targets. Guessing I would say at least at native white less than X in grey and RGBMY primaries for some colorspace simulation.
It does not mean PD2705Q is bad, quite the opposite, seems a good & affordable display. It just mean that all that stuff is mostly marketing.Eizo CS don’t have better uniformity because “internal chroma correction”, they are better because uncorrected panel is better (they paid more). You can see this in dC uniformity reports with max contrast/compensation turned off. Additionaly they can do better in brightness & contrast because uniformity compensation (UC).
If in a typical IPS 1000:1 you get 800-900:1 contrast with good uniformity after tweaking white to D65, thats because panel (+backlight) was uniform. If without manually set UC=ON you get 500-600:1… there is some bad stufff under the hood.“This means better quality control (factory panels selection, probably) and its test at prad.de shows pretty good results.”
Not really, they (widegamuts) are extremely poor in their newer SW-C line. Panels are trash so to make them behave uniform contrast is destroyed. 600:1 at D65 implies HUGE uniformity compensation. You can see them at prad.de. A 200 euro IPS can do better than them for video or works published to web.
In the documentation it says by default resolve preset is set to tone curve of ‘as measured” since usually a grey iteration (I understand this is the vcgt and 1D) is not required.
Question 1. If it’s set to as measured, then it never calculates and calibrates the grey. Is this right?
I do not remember. If all calibration tab is set to as measured it does not calculate VCGT grey color & gamma correction (it ask you to use none=linear or current).
If you set a target white but set TRC/Gamma as measured I do not remember behavior. I would exect to correct grey color but leave gamma (grey’ brightness) unmodified… but better check this. I do not remember. It is not a setting used by me. So if that guess is right, there is VCGT calibration data to make all greys look grey without color… but check it, I do not remember.Q2 is related.
Question 3. What does usually not required mean? Should displays be used with a curve they naturally produce and is it a good practice not to apply a 1D correction?
1D VCGT correction loaded into GPU LUTs causes banding on some GPUs and some setups.
A common sRGB display is expected to behave like a 2.2g sRGB-like display for a lot of non color managed apps, from video to web or office work. Hence aiming for a 2.2g for general work seems a good choice, it also is usually close to “most” of the displays out of the box TRC/gamma.
Question 4. If you could explain or if you could send me to a link where I can understand absolute and relative in custom curves and what black output offset does in both absolute and relative. That would be wonderful! Cuz from what I understand absolute pertains to absolute values. 0 means 0. 50 means 50. But the display might not be able to do 0 and 50 precisely and this might cause inconsistency between displays. Relative modifies the curve to better match how visually separate displays look. Please correct me if I am wrong. But I don’t understand how the black output offset behaves in absolute and relative.
There was an extensive explanation in AVS Forums. Myabe Dominc Chan or other in HCFR or Argyll/MadVR thread, but I do not remember date.
It’s all about your display finite contrast and what should be reference values to compare against. Validating a aginst true Rec709 2.4 power law will result in errors where youd isplay cannnot go (like under black level cd/m2 and closest blacks unles you clip and you really do not want to clip but preserve some near black brightness separation). So your validation setup from previous report seems OK to me.In the documentation it says by default resolve preset is set to tone curve of ‘as measured” since usually a grey iteration (I understand this is the vcgt and 1D) is not required.
Question 1. If it’s set to as measured, then it never calculates and calibrates the grey. Is this right?
I do not remember. If all calibration tab is set to as measured it does not calculate VCGT grey color & gamma correction (it ask you to use none=linear or current).
If you set a target white but set TRC/Gamma as measured I do not remember behavior. I would exect to correct grey color but leave gamma (grey’ brightness) unmodified… but better check this. I do not remember. It is not a setting used by me. So if that guess is right, there is VCGT calibration data to make all greys look grey without color… but check it, I do not remember.Understood. I checked the documentation again its like this.
Tone curve / gammaThe target response curve is normally an exponential curve (output = inputgamma), and defaults to 2.2 (which is close to a typical CRT displays real response). Four pre-defined curves can be used as well: the sRGB colorspace response curve, which is an exponent curve with a straight segment at the dark end and an overall response of approximately gamma 2.2, the L* curve, which is the response of the CIE L*a*b* perceptual colorspace, the Rec. 709 video standard response curve and the SMPTE 240M video standard response curve.
Another possible choice is “As measured”, which will skip video card gamma table (1D LUT) calibration.does this mean it is as i understood? keeping tone curve on as measured skips the VCGT and take the natural response of the monitor?
If yes, is there a way to get a detailed report like that of an icc of uncalibrated display and not the small window that “report on uncalibrated” gives? so that i can check the deviations and gamma and everything else like in a detailed report?
Q2 is related.
Question 3. What does usually not required mean? Should displays be used with a curve they naturally produce and is it a good practice not to apply a 1D correction?
1D VCGT correction loaded into GPU LUTs causes banding on some GPUs and some setups.
I noticed this, and i see less banding when i right click the tray and reset VCGT. Although the darks are slightly lifted and again there might be inconsistencies in grey holding its greyness accurately.
This makes me question, does then resetting VCGT like this (from tray) is equivalent to having a 3DLUT generated from icc without VCGT (tone curve set to as measured)? Or will this result in inconsistency since the ICC was originally created with VCGT set to a curve? (like custom 2.4 relative offset 100%)
A common sRGB display is expected to behave like a 2.2g sRGB-like display for a lot of non color managed apps, from video to web or office work. Hence aiming for a 2.2g for general work seems a good choice, it also is usually close to “most” of the displays out of the box TRC/gamma.
Understood this exactly. Since my monitor has a gamma setting in its OSD, which when i set to 2.4 and get a report on uncalibrated gives a mean/medium of 2.4. But i dont know how to generate a detailed report of uncalibrated so i can check how the gamma behaves throughout the curve.
Question 4. If you could explain or if you could send me to a link where I can understand absolute and relative in custom curves and what black output offset does in both absolute and relative. That would be wonderful! Cuz from what I understand absolute pertains to absolute values. 0 means 0. 50 means 50. But the display might not be able to do 0 and 50 precisely and this might cause inconsistency between displays. Relative modifies the curve to better match how visually separate displays look. Please correct me if I am wrong. But I don’t understand how the black output offset behaves in absolute and relative.
There was an extensive explanation in AVS Forums. Myabe Dominc Chan or other in HCFR or Argyll/MadVR thread, but I do not remember date.
It’s all about your display finite contrast and what should be reference values to compare against. Validating a aginst true Rec709 2.4 power law will result in errors where youd isplay cannnot go (like under black level cd/m2 and closest blacks unles you clip and you really do not want to clip but preserve some near black brightness separation). So your validation setup from previous report seems OK to me.Will try to find this.
Thank you SO SO SO MUCH Vincent. you have no idea how helpful you have been throughout. Really appreciate it!
any idea how i can insert a photo from attachment to here in middle of text?
