- This topic has 89 replies, 5 voices, and was last updated 3 years, 1 month ago by
.
-
Posts
-
I did not say such thing. Read again. Measurements are real. More patches, more accurate description of display, including all non ideal behavior (WHICH IS REAL).
I just found this thread here: https://hub.displaycal.net/forums/topic/just-curious-but-why-does-auto-optimized-test-chart-produce-uneven-edges/
Seems that the problem with jagged gamut edges is related to auto-optimized testchart, Florian said it’s normal and a better choice of profiling… so am I worrying for nothing? Though my 3D LUT gamut is more jagged than the one in that thread…
Vincent I didn’t get the point. Can you explain it better?
Unless device is broken (not ypurs), display varies with temp (OLED or such, not yours), or use wrong measurement mode (not you) or some auto dimming (not your display), what you measure is real… INCLUDING NON IDEAL BEHAVIOR OF DISPLAY. Your display behaves that way.
OK, Vincent, but engineering math promts me that smooth behavior is better for the cost of some tolerance loss. I also note a strange behaviour of my Epson 9900 printer: it shows small gap in black channel linearization chart, but when I crop this chart to black strings, the gap desappears. It is either rendering issue (I use RIP instaead of Epson driver) or print system specifics. No precise result can be achieved here and better to use smoothing. But low patch count may result in rude gradients on screen or low correction of some colors. Here we should find a compromise. Though, better way will be using high patch count and increased smoothing. DisplayCAL doesn’t have smoothing setting for LUTs, but I use this approach when build printer profiles with colprof Argyll command.
OK, Vincent, but engineering math promts me that smooth behavior is better for the cost of some tolerance loss.
Yes, that is the reason to recommend single curve matrix with BPC fro general color managed apps (like PS/LR) IF that kind of profile matches display to certain degree (dE)
I also note a strange behaviour of my Epson 9900 printer: it shows small gap in black channel linearization chart, but when I crop this chart to black strings, the gap desappears. It is either rendering issue (I use RIP instaead of Epson driver) or print system specifics. No precise result can be achieved here and better to use smoothing. But low patch count may result in rude gradients on screen or low correction of some colors. Here we should find a compromise. Though, better way will be using high patch count and increased smoothing. DisplayCAL doesn’t have smoothing setting for LUTs, but I use this approach when build printer profiles with colprof Argyll command.
I’m not saying that he must do bigger charts, I said more patches have higher chances to capture non ideal behavior, and that behavior he see is real regardless of what he wants his display to be.
I agree with “Here we should find a compromise” (regarding patches), just said that his display has that kind of irregular behavior. Hence instead of wasting time seeing gamut plot, just validate device link ( = LUT3D equivalent) and see if at X patch size it matches the colorspace you want to simulate (+visual gradient test).
PASS = OKHi!
I’m continuing to experiment with various combinations of settings, getting better results.
But now in DisplayCal settings dropdown menu I’ve got so many calibrations, I don’t need them, I’ve tried to delete them with the appropriate button next to it in the GUI, now they seem no more selectable, so I think they are deleted, but they are still in the dropdown menu, why? How can i clean it up?
In Windows system: close all color managed applications (better to reboot), open Control Panel (old one, not “gear icon”) – small icons sorting, Color Management (I realy don’t know menu name in English), All Profiles tab. Here you can delete them. Never delete profile files directly in Windows system folder (C:\Windows\System32\spool\drivers\color), – they’re also signed to specific registry, not general registry file as I see.
In Mac: simply remove from ColorSync system (/Library/ColorSync) or user (~/Library/ColorSync) folder or subfolder, for system folder administrative or even root rights are needed. User folder may not exist if no profiles are installed for current user only.
Linux: I realy don’t know, my experience here is essential, but modern GUIs have install/uninstall dialogue in mainstream packet set.
I was able to delete the previous icm profiles, displaycal’s dropdown menu hadn’t refreshed, don’t know why, now it’s all ok.
But I don’t understand LUT3D creation through resolve and a decklink. I’ve made some tests but when I verify the LUT3D I’ve just created (leaving the verification tab settings as they are, choosing the displaycal resolve preset), dE is always high, around 1.4 average and 2.6 peak. Why? Also I can’t reach 100% srgb, and again I don’t understand, the monitor is in native gamut mode, and has 100% srgb in its specs.
Can you guide me in understanding this situation? Thank you!
Huge errors in grey. You’ll need more patches fro LUT3D profile or better starting base calibration so appended VCGT is more accurate.
I run 2600 patches, and in the tone curve under calibration tab I set as measured, so the speed is unsettable.
Also I abandoned LG calibration studio because it’s too buggy, really. I thought that going with the monitor’s user mode (which has native gamut and customizable RGB values) could be good, too.
The best grey I’ve gotten (see attach) is with settings that I don’t think are suitable for LUT3D creation (in fact they aren’t chosen by the preset), like BPC enabled, “enhance effective resolution of colorimetric PCS-to-device table” enabled, relative colorimetric under profiling tab, tone curve 2.4 relative under calibration tab… But as you can see, the grey has improved, while the other colors have gotten worse…
If you are using LG in Resolve as GUI display, try to do what you’ll do for common desktop general purpuse profile:
Settings default
D65, calibration Slow, gamma closest to what you can choose in OSD (2.2 if no controls)
XYZLUT no BPC 400-1000 patches.
Then make LUT3D (with/without VCGT depending on profile active as default display profile)What is your current display profile? Is it based on single curve or XYZLUT-based? What calibration tolerance (fast/medium/low level) was used and how many patches were used to build profile?
The LUT3D I’ve made were all XYZLUT+matrix profile type, as the Resolve preset choose.
2600 patches auto-optimized.
Calibration speed was not manually set because if you leave tone curve “as measured”, displaycal doesn’t let you set it.
For Vincent:
With resolve I use the LG through the decklink, not as GUI display (and in LG OSD it’s possible to manually set gamma 2.2/2.4/2.6, if it’s in User mode).
