- This topic has 89 replies, 5 voices, and was last updated 3 years, 1 month ago by
.
-
Posts
-
Thank you guys for the detailed answers!
The fact is that my 31MU97 can’t really stores 2 different calibrations, I don’t know why, it’s always been like that since the beginning (with any true color pro / LG calibration studio versions, that doesn’t matter, but I’ll give the latest update a chance). So unfortunately I can’t rely on this feature…
And I’m on an nvidia 1070 GPU, so there’s no sRGB emulation options.
I also found that with hardware calibration alone I have a red cast, so after it I will always have to do a more detailed displaycal calibration (at least for professional grading purpose).
To Алексей Коробов:
I do color correction / grading works in a controlled environment, with only a 6500K 15-16lux led light behind the monitor and nothing more. So I always set a gamma of 2.4 instead of 2.2.
And I always leave all the monitor’s “smart features” off, including uniformity correction.
Other than that, I don’t undestand some of the settings you suggests,
I use a i1 display pro colorimeter, so what white point xy should I set? I thought that the Rec.709 standard 0.3127 0.3290 was fine…
why choose a black point correction other than 0%? I have experienced that this raises the black point, which I want to avoid at all costs as my monitor seems to reach a minimum of around 0.1235 cd/m2, quite poor…
Why select black point compensation? Displaycal warns that this will reduce color conversion accuracy.
And why choose relative as the rendering intent?
To Vincent:
Regarding the gamut corners, in the profile information, set A2B or B2A doesn’t smooth them, although their paths are slightly different… The only way that seems to smooth them is uncheck “LUT”, immediately to the left of the A2B/B2A selection, but I don’t know what this means.
So to understand what the problem is I have to redo the calibration but with a single curve+matrix profile and black point compensation unchecked, right? And then?
Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.I also found that with hardware calibration alone I have a red cast, so after it I will always have to do a more detailed displaycal calibration (at least for professional grading purpose).
This has been discussed before. LG has no i1d3 corrections for your LG. It corrects it with a RGBLED correction instead of GB-LED.
Also even measuring properly it may oversimplify calculations, like correction white first, then change brightness which may drift white point.So to understand what the problem is I have to redo the calibration but with a single curve+matrix profile and black point compensation unchecked, right? And then?
No. I said “in order to check if those irregular boundaries are real (or big) calibrate & valiate with a simpler profile”. That does not mean to force the use of simplified profile as display profile for critical work, just to check how good it is. Check if it describes actual display or not.
If it is not so good validating that display against simplified profile (no simulation profiles), then irregularities are real, your display is “as bad” as profile shows. Instead of worring about native gamut boundaries just check if LUT3D simulated colorspace (with that detailed irregular XYZ display profile) is accurate.If HW calibration is unusable, set display to sRGB/Rec.709/AdobeRGB (whatever you need) and start DisplayCAL process.
Black point correction – you’re right, this may decrease display contrast. Don’t use it if better contrast is preferable.
Black point compensation (BPC) – moves black from absolute zero to your actual zero signal level (0.1235 cdm). This have to be used in most cases, though BPC actually decreases saturation. You’ll get dark (very dark in your case) colors clipping instead.
Relative colorimetric intent – is the way of image conversion from its embeded ICC to display ICC. This correctly converts all colors that are inside display (target) gamut, but clips others (they’re stay at some edge of gamut). So, this is what you need for color managed editing. Perceptual conversion is compression of original ICC to target ICC, this changes colors (yes, this is also color management). Perceptual mode is needed to avoid clipping and to save gradients in some cases. It is usual for printing RGB images. Note: this is default setting, applications may ignore it. Read essential reference to color rendering intents and soft proofing.
Also: Gamma value – better to set value from display mode, i.e. 2.4 for Rec.709 and sRGB for sRGB (find others in Wiki), but your display probably sets separate gamma in corresponding menu. Profile test contains gamma graph that helps you to determine display actual gamma emulation (I think, LCDs always emulate gamma, a natural function of CRTs).
White point: I use 03123 0.3280 for my taste and my instruments (they’re not reference quality tools). Use 6500K if you don’t find better white tint with xy selection tool.
Follow to Vincent’s colorimeter correction recommendations, I’ve never met 31″ wide gamut LG.
Unlock RGB range setting in Option -> Show advanced options.
DisplayCAL screenshots with settings are attached. If you’re on Mac install X-Rite linear profile and set it for your display before calibration.
I’m sorry, but if you use limited gamut mode, colorimeter correction may be wrong. Probably, my BenQ SW240 for AdobeRGB correction is good in this case.
I’m sorry, but if you use limited gamut mode, colorimeter correction may be wrong. Probably, my BenQ SW240 for AdobeRGB correction is good in this case.
That LG is a vanilla GB-LED, not a WLED PFS like SW240. Using limited gamut does not change CCSS correction, it must be native always (all colors in limited gamut are a combination of native primaries with scaled brightness, hence native CCSS covers all).
Ok, big news! After I’ve updated LG calibration studio, now it seems I can store 2 different hardware calibration, as it should normally be!! Thanks a lot Алексей Коробов for the reporting, last time I checked for LG updates was few months ago.
So this is what I did next: I remade the native gamut hardware calibration as the starting point of it all (100 cd/m2, gamma 2.4, ICC v2, table based), and I had a 0.7 average dE.
Then in displaycal I went with almost the same settings Алексей Коробов suggests. For the BPC and the rendering intent, I then saw what was your point, after some deeper research, and set them as you wrote. Beyond that, I chose GB-r-LED IPS (Dell U2413) correction, 0.3127 0.3290 as white point coordinates, 100 cd/m2 white level, 0.0001 cd/m2 black level, gamma 2.4 relative, and 2052 patches. Also I unchecked interactive display adjustment as the monitor grays out custom adjustments in hardware calibrations mode.
And at the end of the process I finally had 100% sRGB coverage, 0.3124 0.3280 as white point, and a 0.41 average dE. Not bad, right? I attached the measurement report, let me know if it’s all ok or I missed something.
Anyway, in the profile information I still have those kind of gamut corners, don’t know how to get rid of them, but since I will work in a smaller colorspace (sRGB/Rec.709), I think they doesn’t matter, or am I wrong?
Now, if you can confirm that all is fine, tomorrow I will go ahead from here with the 3D LUT for resolve, with this settings (starting with the displaycal resolve preset):
In the display/instrument tab: full range output levels, GB-r_LED IPS (Dell U2413) correction.
In the calibration tab: all set on “as measured”, leaving interactive display adjustment unchecked.
In the profiling tab: all as the preset has already set, 2060 patches.
In the 3D LUT tab: “create 3D LUT after profiling” checked, gamma 2.4 relative tone curve, 100% black level output compensation, “apply calibration (vcgt)” checked (cause I’ll use a decklink), relative colorimetric rendering intent, .cube, full range input/output encoding, 65x65x65.
I think these settings are right, but do you have any notes?
Lastly, to have an sRGB “normal” desktop mode available, I was thinking to do the second hardware calibration (in sRGB gamut, 160 cd/m2, gamma 2.2), and then switch back and forth between the two from the monitor itself, leaving only displaycal ICC profile (the one whose report I have attached) in windows. In this way displaycal ICC should correct the LG calibration shifts, while the latter should limit the overall gamut to sRGB, right? I don’t need a super-precise desktop mode calibration, at most a common one.
Generally all is OK. However, black level of 0.0001cdm is mistake (your display can’t achieve it), but this may be tuned to real black automatically. Draw dark grey gradient (better to do in Photoshop for a new, non color managed image) and check for a gap near black. I think that 2052 patches is exessive amount for display. But not a mistake. I also think that 65x65x65 LUT size may also be exessive and will slowing video output, but I’m green in video tech, follow Vincent’s recommendation.
Seems OK, grey is not superb but Resolve LUT3D may improve it. Regarding LUT3D size IDNK Resolve optimal performance or where are located perfiormance penalties, try in LGG forum.
Yeah, I’m wondering the same thing on 3D LUT resolution, but I see people go with 65x65x65 for resolve, so I thought it was good. I’m gonna try to do a deeper research, but there isn’t too much about it.
For the black level, I set the minimum 0.0001 cause I thought it would pushes the monitor’s blacks a little further, I didn’t know it was a real mistake. If it is, then I have to redo the calibration… I will check in photoshop as well.
If I have to redo it, do you have any suggestions to further improve the dE and overall performances?
Thank you!
For the black level, I set the minimum 0.0001 cause I thought it would pushes the monitor’s blacks a little further, I didn’t know it was a real mistake. If it is, then I have to redo the calibration… I will check in photoshop as well.
On LCDs contrast is static, like a sliding window on a rule. Max window width (max contrast) is at native white. Choose a white brightness and panel static contrast will set you a miniumum black value, it won’t go below that value with that white brightess. Hence that value on DisplayCAL GUI will be ignored. No need to redo.
If I have to redo it, do you have any suggestions to further improve the dE and overall performances?
Slow calibration speed may improve grey range (color diferences between a grey and another grey). This means that grey calibraion stage for general purpose profiles (like the ones you may use as display profile for Photoshop use) is about 20-30min on a i1d3 pro (slow speed, up to 96 patches) + the time needed for profilling stage (which depends on profile patch size).
Ok, I’ve taken note, thanks Vincent!
Today I’ve run two 3D LUT calibration through resolve/decklink (one with all on “as measured” but white point in the calibration tab, and an absolute with white point scaling rendering intent in the 3D LUT tab, and one with exactly the same settings of previous windows ICC profile and a relative intent) and always end up like what you see in the attached image, and can’t reach 100% sRGB coverage (only 98%, apparently blues are smaller than sRGB standard), but I don’t know why, the settings are practically the same as in the ICC profile that I’ve made for windows, with 100% sRGB, so where’s the difference?
And in the 3D LUT profile information if I select B2A instead of A2B (that is the one I’ve attached) the plotted gamut is even more shifted/unstable.
Any suggestions? Honestly, I don’t know what to do.
Explained before, the more detail in measurements, the higher chance to capture non ideal behavior.
Run measurement report with Rec709 2,4 and use device link equivalent to LUT3D. If it’s OK you are done.So I have used too much patches? Honestly people have opposite opinions about it, some say 500 patches are sufficient, others say they should be 5000. I set it to 1500, I thought it was the best compromise.
Also I’ve made a test calibration with 500 patches and for the rest same settings that had given me 100% sRGB coverage (but with 1500 patches), and it ended up with 99% coverage. That lead me to think 500 were a little low, and 1500 a better choice…
