LG C8 Lut

Home Forums General Discussion LG C8 Lut

Viewing 15 posts - 121 through 135 (of 213 total)
  • Author
    Posts
  • #21579

    chros
    Participant
    • Offline

    Note that Home is accurate in HDR10 but …

    Let’s just talk about HDR10 for now (I did/do), even dealing with this mode only is way more complicated than it seems.

    Ted misunderstood my intentions…

    He didn’t even talk to you, have you read the link I posted? 🙂

    “3) Baked Tone/Gamut mapping can’t be disabled… If you want to perform a 3D LUT for HDR10, the display need to have gamma based tracking (not PQ), to be able to provide you the max output possible (700-800 nits) with gamma 2.2-2.4, and don’t have enabled any kind of tone/gamut mapping. … BTW the Panasonic EZ1000 supports 66-Point Cube 3D LUT for HDR10, and there is a trick to disable the tone mapping …”
    And read on.

    Due to this, so far I thought (and still think, note that I haven’t tested it) that PQ is active even in calibration mode:
    – you said that it’s recommended to put the curve of the TV to 700 nits (e.g. with madvr) before calibration
    – that’s why I wanted to create a PQ 3dlut and not gamma (when I first tried)

    The math based HDR10 tone mapping IS disabled in calibration mode. At least it should be.

    Can you also test this as well? (I haven’t yet):
    – start calibration, then upload unity 1dlut and unity 3dlut
    – fire up HCFR and run 20 points grayscale

    Your 3DLUT targets D65 because the Rec2020 profile has a profile white point of D65.

    Oh, thanks, it makes sense. I wasn’t sure about this, because DCI-P3 has a separate option for D65.

    I am not sure why it seemed to work fine for me. I will try to calibrate the C9 this way tomorrow and show you screenshots with a validation report.

    Thanks. Can you also upload all the screenshot of DisplayCal settings (as I did) along with the ti1 testchart, to be able to fully reproduce the test case?

    Lastly I’m not sure whether it’s useful to run a validation at all (especially if the TV is still in PQ mode), instead:
    – compare the result against an unprofiled preset with different contents / scenes
    – check for black crush with Mechanik’s Black clipping patterns
    – etc…

    #21581

    János Tóth F.
    Participant
    • Offline

    He didn’t even talk to you, have you read the link I posted? 🙂

    Haha. My bad. I though you referred to these posts: ConnecTEDDD and my response where he explains what I already knew: math based HDR10 tone mapping can’t be disabled for normal mode but calibration mode only, thus HDR10 calibration has to target gamma 2.2. Or you have to use an SDR mode with it’s reduced peak white (and different color and ABL handling, etc).

    By the way, the C9 allows the user to turn on the “W boost” for SDR but I am not sure how much brightness is to be gained (it’s probably still limited in SDR compared to HDR) and I am still not excited to calibrate an SDR mode for HDR10 with a 3DLUT targeting PQ TRC. It would also require source-side workarounds (possible for movies, not so much for video games).

    “3) Baked Tone/Gamut mapping can’t be disabled… If you want to perform a 3D LUT for HDR10, the display need to have gamma based tracking (not PQ), to be able to provide you the max output possible (700-800 nits) with gamma 2.2-2.4, and don’t have enabled any kind of tone/gamut mapping. … BTW the Panasonic EZ1000 supports 66-Point Cube 3D LUT for HDR10, and there is a trick to disable the tone mapping …”

    Due to this, so far I thought (and still think, note that I haven’t tested it) that PQ is active even in calibration mode:
    – you said that it’s recommended to put the curve of the TV to 700 nits (e.g. with madvr) before calibration
    – that’s why I wanted to create a PQ 3dlut and not gamma (when I first tried)

    Already answered above: “math based HDR10 tone mapping can’t be disabled for normal mode but calibration mode only”.

    Can you also test this as well? (I haven’t yet):
    – start calibration, then upload unity 1dlut and unity 3dlut
    – fire up HCFR and run 20 points grayscale

    Again, just look at the damn picture with your naked eyes in calibration mode! What do you Ccccc? 🙂 I tell you what: ~P3 gamut, ~10k WP, gamma ~2.2. No HCFR or sensor required, only your bioassy to tell it’s not PQ.

    Your 3DLUT targets D65 because the Rec2020 profile has a profile white point of D65.

    Oh, thanks, it makes sense. I wasn’t sure about this, because DCI-P3 has a separate option for D65.

    The original DCI-P3 has a different WP than “Display P3”, hence the options to pick a WP for P3. But your target profile is Rec2020, not P3.  😛

    Thanks. Can you also upload all the screenshot of DisplayCal settings (as I did) along with the ti1 testchart, to be able to fully reproduce the test case?

    Lastly I’m not sure whether it’s useful to run a validation at all (especially if the TV is still in PQ mode), instead:
    – compare the result against an unprofiled preset with different contents / scenes
    – check for black crush with Mechanik’s Black clipping patterns
    – etc…

    The last time I did this on the C8, I ran validations in both calibration and normal modes: against gamma 2.2 and PQ respectively. (Yes, I repeat, against gamma 2.2 in calibration mode.)

    #21584

    j82k
    Participant
    • Offline

    I actually tested this at my previous calibration session.

    First pic: calibration mode enabled, factory 1dlut, unity 3dlut
    Second pic: calibration mode enabled, unity 1dlut, unity 3dlut

    So yes it is tracking gamma 2.2 when calibration mode is enabled. The one with the factory 1dlut uses my 2 point low adjustments, that’s why it has a lower gamma than the one with unity 1dlut.

    • This reply was modified 8 months, 1 week ago by j82k.
    Attachments:
    You must be logged in to view attached files.
    #21589

    chros
    Participant
    • Offline

    So yes it is tracking gamma 2.2 when calibration mode is enabled.

    Thanks!

    Again, just look at the damn picture with your naked eyes in calibration mode! What do you Ccccc? 🙂 I tell you what: ~P3 gamut, ~10k WP, gamma ~2.2. No HCFR or sensor required, only your bioassy to tell it’s not PQ.

    😀 I wasn’t sure anymore 🙂 Now you convinced me! Thanks!

    #21591

    János Tóth F.
    Participant
    • Offline

    @chros – You can find everything here (onedrive folder). But before somebody shrieks at the sight of the dE values of R,G,B primaries, do note that these validations were done against the DCI-P3 and Rec2020 R,G,B primaries, NOT against the measured native gamut coverage (read the file names).

    Oh, and yes, forget my laziness: I re-used the C8 sensor correction (I will create and upload a C9.ccss at a later date) and didn’t adjust the OLED Light to hit 700nit (or created new synthetic reference profiles for the measured ~750 nit with PQ hard-clip at 750).

    #21593

    chros
    Participant
    • Offline

    @chros – You can find everything here (onedrive folder).

    Thank you! I’ll take a look later …
    So, what do you see when you compare it to a unprofiled preset? Is the image darker for you as well? 🙂

    Oh,  and looking at you screenshots, I noticed that there’s a new DisplayCal version, and the alpha7 chipset (17 points cube) is also supported. 🙂
    Thanks @Florian!

    • This reply was modified 8 months, 1 week ago by chros.
    #21595

    János Tóth F.
    Participant
    • Offline

    @chros – Yes, the profiled HDR10 Technicolor mode is significantly darker than the factory Cinema mode. But dark grays look subjectively way too bright on this C9 in all factory modes, SDR included. The tone response difference wasn’t nearly this prominent when I used this workflow (or CalMAN) on the C8. So, that doesn’t really mean anything.

    According to the validation report of calibration mode (LG’s tone-mapping OFF), dark patches are too dark (they would fit on a gamma 2.4 curve) and bright patches are too bright (would fit gamma 2.0). So, yeah, may be a carefully done 1DLUT calibration –or– BFI during profiling (and possibly a bigger patch set) could significantly improve the results. Although CalMAN’s long iterative 1DLUT calibration probably suffers more panel drift in HDR mode than this (the 100% white can drift up ~10% if the patch area is stressed enough).

    #21596

    chros
    Participant
    • Offline

    @chros – Yes, the profiled HDR10 Technicolor mode is significantly darker than the factory Cinema mode. But dark grays look subjectively way too bright on this C9 in all factory modes, SDR included.

    I’m “glad” to hear that, so it’s not just me 🙂

    The tone response difference wasn’t nearly this prominent when I used this workflow (or CalMAN) on the C8. So, that doesn’t really mean anything.

    That’d be interesting: maybe you used some different settings, mode, etc …

    According to the validation report of calibration mode (LG’s tone-mapping OFF), dark patches are too dark (they would fit on a gamma 2.4 curve) and bright patches are too bright (would fit gamma 2.0).

    Exactly what I got here (very similar to my last result, I also had to lower gamma to 2.13 this time as last time). I reprofiled HDR10 with your settings, only difference I used:
    – I used madvr as display (with 5 seconds BFI) instead of LG
    – embed calibration was deselected and Linear was selected (there was no screenshot about that)

    I also measured gamma in calibration mode, I got the same that you mentioned:
    – 1st image: unity 1dlut + unity 3dlut
    – 2nd image: unity 1dlut + generated 3dlut
    So it’s way off. Interesting thing is that @jk82 got completely different results from us with his replaced panel 😀

    Otherwise resulting coverage / gamut is like yours ~97%.

    1 more thing I noticed on your validation screenshots, that I noticed while watching some test scenes:
    – the green color is completely off 🙂

    So, yeah, may be a carefully done 1DLUT calibration –or– BFI during profiling (and possibly a bigger patch set) could significantly improve the results.

    BFI is not an issue with DisplayCal. So we have to figure out what the next step should be.
    How can we create 1dlut with DisplayCal? @Josh’s util already supports uploading it.

    But as I mentioned earlier: I tried out factory 1dlut + WBRGB maxtrix based 3dlut and this suffered from the same color problems (except for it didn’t crush blacks), e.g. the green is completely off there as well (in my Technicolor preset).

    Attachments:
    You must be logged in to view attached files.
    #21599

    Josh Bendavid
    Participant
    • Offline

    The 1D LUT can be created by running the calibration procedure (ie by having Tone Curve set to anything other than “as measured” in the Calibration tab in displaycal).  Then the .cal file created with the profile can be uploaded.  Or was the question something different?

    #21600

    János Tóth F.
    Participant
    • Offline

    I did a quick test on the C9: set the Cool WP preset to 192-192-192 in EZ-ADJUST, measured the gray ramp with SDR Technicolor set to Cool and Wide, then the same preset after a Full DDC Reset. The colverify tool prints very high dE numbers. I am not sure I am using it correctly, so I made a semi-log chart in excel instead. This confirms my subjective assessment that dark shades look brighter in Technicolor after a reset than in ISF Dark Room set to Cool and Wide (no reset). So, the factory LUTs are probably not neutral (or the Full Reset permanently alters some other processing parameters for the picture mode which alter the near-black response).

    Attachments:
    You must be logged in to view attached files.
    #21602

    Josh Bendavid
    Participant
    • Offline

    Are you sure the dithering behaviour is not changing when resetting the 1d LUT ?

    #21603

    chros
    Participant
    • Offline

    Then the .cal file created with the profile can be uploaded.

    Thanks, that was the question.

    @Janos, I quickly measured the lowered gamma (2.13) unity 1dlut + generated 3dlut in calibration mode, compare it to the 2.2 one:
    – the curve is basicly the same, just shifted

    So maybe we have to do calibration as well in DisplayCal and include it in the 3dlut to compensate the native gamma.

    Attachments:
    You must be logged in to view attached files.
    #21605

    János Tóth F.
    Participant
    • Offline

    @chros – Did you notice the validations were done against DCI-P3 and Rec2020 primaries (respectively in calibration and normal mode). Since the panel’s native gamut doesn’t fully cover P3, let alone Rec2020, those numbers will be be fairly high (huge in case of Rec2020). If you want pretty numbers on the reports, you have to create a synthetic profile with the measured primaries and validate against that or create a test chart which only validates R,G,B triplets which should produce in-gamut colors (alternatively: use a different software for validation reports which automatically does so).

    I wonder why the DisplayCAL 3DLUT doesn’t correct the tone response. I think it should. May be the profile is flawed due to the WRGB magic and “W boost”.  I was lazy to create a high quality profile and validate it with neutral LUTs. I guess I shall do that next.

    #21607

    János Tóth F.
    Participant
    • Offline

    I should have set dICtCp for the validation report and/or looked at the gamma plot. 🙁

    Attachments:
    You must be logged in to view attached files.
    #21609

    Josh Bendavid
    Participant
    • Offline

    One thing to be aware of.  One shoudl NOT run “profile only” unless the white point is correctly calibrated to start with.

    Ie, even if one wants to use only the 3d lut for the TV, then one should still run the calibration step in DisplayCal (but in this case check enable “Apply calibration” in the 3D lut settings.)  Doing things this way with the factory 1D LUT makes the most sense with e.g. colour temperature set to cool and the service menu white point settings for cool set to 192/192/192/64/64/64 as has been discussed.

    In order for “Profile only” to behave correctly, one should instead set the color temperature to “Warm 2” and then calibrate the “warm” color temperature settings in the service menu interactively to D65, leaving cut values at 64/64/64, and red gain at 192 as has been discussed in some of the lightspace profiling threads.

Viewing 15 posts - 121 through 135 (of 213 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS