Comparing Colourspace CMS and Argyll/DisplayCAL Calibrations for LG C3 OLED

[Justin Stephenson]

TL;DR
I’m comparing calibrations done in Colourspace CMS (CS) and Argyll CMS/DisplayCAL (ADC) on an LG C3 OLED used as a secondary monitor in an SDR post-production suite. I have specific questions about EOTF differences in the calibrations and calibration workflow in ADC.

Background
– Filmmaker working in animation and post (not an engineer or calibrator)
– Main monitor: Flanders Scientific DM250
– Secondary monitor: LG C3 OLED
– Video pipeline: BMD Decklink SDI 4K Pro
– Seeking to optimize calibration workflow

Setup
– Windows: Colourspace CMS; Rocky Linux: Argyll py3
– Test Pattern Generator: RPi4 PGen (Resolve server for DisplayCAL)
– LG C3 settings:
– ABL accessible and disabled
– Service Menu whitebalance precal to REC709 white point, 124nits at 109% superwhite (255)
– Automatic settings off
– Using Expert Bright picture setting slot
– Workflow: Rec709 2.4 100nits at 235 white
– CIE1931 2-degree observer
– Equipment:
– i1Pro3 Spectro X-RITE (2024 certified)
– i1D3 OEM (2023 model)
– Profiles are probe matched

Calibration Workflows

Colourspace CMS (CS)
1. Probematch
2. Profiling: LG’s “Extended Range” (16-255)
3. Characterization: 17^3 cube, 0.3s full field black stability patchs
4. LUT creation: Two-step process with CS “profile augmentation”
5. Verification: Legal range (16-235)

Argyll CMS/DisplayCAL (ADC)
1. Probematch
2. Profiling: Legal Range (16-235)
3. Characterization: Slow speed, 3500 patches, 0.3s full field black stability patches
4. Calibration settings: As measured white point/levels, custom 2.4 relative tone curve
5. LUT: Absolute Colorimetric rendering, custom 2.4 tone curve
6. Verification: CS using Legal range (16-235)

Verification Results
– CS: dE Average 0.811, 139 points above dE1, 0 above dE2.3
– ADC: dE Average 0.334, 13 points above dE1, 1 above dE2.3
– CS: EOTF rises above 0 after 0.2 and stays there
– ADC: EOTF dips between 0.75 and 0.9

Subjectively, compared to the CS Cal LUT, the ADC Cal LUT appears slightly better, with more neutral shadows when referencing the Flander DM240.

Questions
1. Are the Diff EOTF values(*) within acceptable tolerances for post-production monitoring?
2. Could the ADC EOTF dip in the upper range be due to lack of a defined panel preroll stabilization sequence?
3. Is there a way to incorporate a preroll stabilization sequence in ADC profiling?

(* I can pull a more granular report based on the diff eotf if that would be more useful here.)

I’ve tried various settings in both CS and ADC but haven’t found a way to address the EOTF differences or implement a stabilization sequence in ADC. Any insights or suggestions would be greatly appreciated.

• This topic was modified 1 year, 8 months ago by Justin Stephenson.

Attachments:

You must be logged in to view attached files.

Calibrite Display Pro HL on Amazon  
Disclosure: As an Amazon Associate I earn from qualifying purchases.

[Vincent]

Do black to blue  / to green/ to red ramp readings match measured with ADC and CS?  I mean, choose  a sample solid color in resolve with one LUT3D active, whatever it is. Then measure that color (spotread in Argyllcms, remeber to add colorimeter correction, see doc; CS should have a free measure feature)
Are you using the same matrix or are your creating a new one for ADC and another for CS?

Maybe you see different de colors in calibrated primaries because each program saw different coordinates while profiling the uncalibrated display.

AFAIK (older versions) CS matrix correction were computed of the fly by selecting Reference WRGB readings (from i1pro) an current uncorrected probe WRGB  readigs (i1d3).
On ArgyllCMS an actual CCMX 3×3 matrix is computed and then saved. Maybe you can play with “ti3” files used by DisplayCAL to compute a matrix from the same readings as CS, expecting them to be closer.
Also remember that DisplayCAL will try to use high res mode for i1pros while CS will use them at 10nm

[Justin Stephenson]

I am unable to use the same matrix. I have written a python script to translate the t3 probe match readings I do in ADC (normalized XYZ) to CS (raw reading Yxy). CS does, infact, calculate the match on-the-fly based on the reading. I do 4 readings of each patch (WRGB) and average them for ccxxmake and colourspace.

Those two profiles and LUTs were done a couple days apart. I will do another test where I do a CS calibration and and ADC calibration back to back and see what I get. I will also figure out how to create and run a CS/ADC readings comparison patch sequence (primaries and greyscale ramps?) to see what each program is seeing.

I am in the middle of production right now, so it may take some time to get to this. It is on my list.

Thanks for your input.

• This reply was modified 1 year, 8 months ago by Justin Stephenson.

[Vincent]

It’s a pity that CS, ColorNavigator, SpectraviewII, etc… won’t let you to put manually a 3×3 XYZ to XYZ correction using default “__GENERIC_CMF__” CCSS/EDR correction.
Specially for Vendor HW calibration software like those from Eizo/NEC/Dell/Asus/LG/Benq… lots of issues could been avoided due to the lack of proper EDR correction in the SDK.

ArgyllCMS & Calman can.

• This reply was modified 1 year, 8 months ago by Vincent.

[Author]

Posts