LG C8 Lut

Home Forums General Discussion LG C8 Lut

Viewing 15 posts - 226 through 240 (of 265 total)
  • Author
    Posts
  • #29074

    Josh Bendavid
    Participant
    • Offline

    Do you have a reference where it is claimed that (only) a change in the dolby vision config file is sufficient to change/fix the near black behaviour?  I would have thought that this is more so depending on the 1D and/or 3D LUT.

    The 2019 config file can be easily converted to the 2018 format.  A config file can also be generated from scratch following the procedure used by Calman, but I don’t know what the default values are which should be used for the C8 wide mode colour primaries (I could measure them from my set, but given the way the white subpixel is used in HDR, there is an ambiguity whether to measure the primaries at peak brightness or at the onset of desaturation)

    I do still have the C8, but I’m not aware of any publicly available dolby vision test patterns.

    #29089

    chros
    Participant
    • Offline

    Do you have a reference where it is claimed that (only) a change in the dolby vision config file is sufficient to change/fix the near black behaviour?  I would have thought that this is more so depending on the 1D and/or 3D LUT.

    No, and it seems I was wrong again, sorry. I had to look it up and they manually calibrate 22 points WB on C9 (to get gamma 2.2)! (Although I don’t know how because on B8 there are only RGB controls for 20 points IREs).

    I do still have the C8, but I’m not aware of any publicly available dolby vision test patterns.

    They were linked in posts I linked:
    sspears’s one
    BlackPLUGE one

    Play them back in pitch black room, you will see that there’s elevated black (and read this post one more time).
    So, I’m only curious about whether we can get rid of the elevated black and slight black crush on our 2018 sets, no matter which way 🙂

    • This reply was modified 1 month ago by chros.
    #29096

    Josh Bendavid
    Participant
    • Offline

    Right, so it can be done in principle by applying 1D and/or 3D luts derived from the HDR10 calibration to the Dolby Vision mode together with a synthetic dolby vision config generated as described in the previous post (using the BT.2020 primaries).

    I’m not very motivated to try it without the ability to properly verify the results though. (I have the means to do proper test pattern generation for HDR10, but not Dolby Vision currently)

    #29177

    chros
    Participant
    • Offline

    I’m not very motivated to try it without the ability to properly verify the results though.

    I understand, but we can test easily whether we can get rid of the raised black without adjusting brightness / contrast controls.
    My point is that if we can’t get rid of the raised black then we don’t even have to play with this anymore.

    it can be done in principle by applying 1D and/or 3D luts derived from the HDR10 calibration to the Dolby Vision mode

    If it is the only way that we know of then can you provide me DisplayCal settings to try this one out?

    Last question to you and @János Tóth F.:
    do you have any luminance adjustment (apart from IRE 5-100) on your C8 / C9 or just the RGB controls under 20/22 points WB setting?
    Because if you don’t have it then I don’t know how the ClolorSapce guys modified the 22 point WB on their C9 (to modify gamma 2.2) … 🙂

    • This reply was modified 1 month ago by chros.
    #29186

    Josh Bendavid
    Participant
    • Offline

    The point is that if you change the LUTs then you could completely mess up the gamma and colour if things are not working as expected, even if the black point is fixed.  So it is not sufficient to just verify the black point.

    To calibrate HDR10 you just have to put the set in calibration mode for one of the HDR10 picture modes, and calibrate to gamma 2.2, BT.2020.  (but you have to be able to properly generate test patterns in HDR10, and it can be tricky because a LOT of devices will apply some kind of SDR-to-HDR tone mapping which can’t be switched off.)

    #29187

    Josh Bendavid
    Participant
    • Offline

    Ok, sorry I misread your message.  Yes I see your point, but if it doesn’t fix the raised blacks it’s also extremely difficult/impossible to understand what the problem is without proper test patterns.

    #29228

    chros
    Participant
    • Offline

    Ok, sorry I misread your message.  Yes I see your point, but if it doesn’t fix the raised blacks it’s also extremely difficult/impossible to understand what the problem is without proper test patterns.

    I agree, that’s why I said that I don’t want to play with it anymore then 🙂

    So, what about this: creating 1dlut only (for now) with HDR10 and try to upload it into one of the DV modes (e.g. Game preset, it behaves the same as Cinema).

    If you agree, can you write down the steps (e.g. settings in DisplayCal) to create a 1dlut with madvr and upload it? I can try this out this weekend, and we can carry on from there…

    Thanks

    • This reply was modified 4 weeks, 1 day ago by chros.
    #29230

    Josh Bendavid
    Participant
    • Offline

    I could not get this to work with madvr because current NVIDIA drivers apparently apply some SDR to HDR tone mapping when outputting windows gui in HDR and I couldn’t figure out how to turn it off.

    #29231

    chros
    Participant
    • Offline

    I could not get this to work with madvr because current NVIDIA drivers apparently apply some SDR to HDR tone mapping when outputting windows gui in HDR and I couldn’t figure out how to turn it off.

    No worries, I assume you use the new drivers, I went back to 398.18 so that’s fine (with 1060 card).
    I still need your input which steps need to be taken 🙂

    #29232

    j82k
    Participant
    • Offline

    So, I just figured out that you can actually use the 1DLUT (.cal file) to do a manual calibration with as many adjustment points as needed.

    Ever since I got the C8 replacement panel in 2018 it had a terrible non-linear near-black range and I’ve never been able to completely fix this. Not with a manual calibration as there are no adjustment points between black and 5% and also not with a displayCal generated 1D/3D LUT.

    Basically I just opened the .cal file with a text editor and deleted all the data points except 0 and 100% and then did grayscale and near-black sweeps in HCFR, adding adjustment points wherever needed. I didn’t think this would work but it does and it’s perfectly interpolating between the manually added data points, no matter how many and where you set them. The “NUMBER_OF_SETS” value has to be changed accordingly.

    I ended up with just 12 adjustment points wich resulted in a good grayscale response and great near-black tracking (see attached screenshot).
    After that I just did a quick 175P 3DLUT in displaycal and uploaded that too. I’ll probably do a larger test chart 3DLUT next time though.
    I watched a few movies so far and also looked at various near-black pattern and this messed up C8 has never looked better in dark scenes.

    Just for reference, this is the cal file I ended up with:

    CAL

    DESCRIPTOR “Argyll Device Calibration Curves”
    ORIGINATOR “Argyll synthcal”
    CREATED “Tue Mar 09 05:55:59 2021”
    DEVICE_CLASS “DISPLAY”
    COLOR_REP “RGB”

    NUMBER_OF_FIELDS 4
    BEGIN_DATA_FORMAT
    RGB_I RGB_R RGB_G RGB_B
    END_DATA_FORMAT

    NUMBER_OF_SETS 12
    BEGIN_DATA
    0.0000 0.0000 0.0000 0.0000
    0.0140 0.0298 0.0267 0.0259
    0.0300 0.0371 0.0334 0.0324
    0.0500 0.0481 0.0425 0.0399
    0.0950 0.0875 0.0845 0.0729
    0.1200 0.1053 0.1023 0.0887
    0.2000 0.1738 0.1680 0.1467
    0.3000 0.2631 0.2543 0.2218
    0.4000 0.3520 0.3450 0.2990
    0.6000 0.5576 0.5462 0.4746
    0.8000 0.7720 0.7580 0.6500
    1.0000 1.0000 0.9730 0.8400
    END_DATA

    My panel required major adjustments at 0.0140 (1.4%) and 0.0300 (3%) which isn’t surprisig as by default the luminance drops off a cliff below 5%.
    I know that generally it’s recommended to leave the 1DLUT alone but for my messed up non-linear panel this is a godsend.

    While I haven’t seen any near-black overshoot yet with this I wonder if there is still no way to keep the new stronger dithering when uploading a 1DLUT? How does calman handle this, does it also revert back to the original weak dithering when touching the 1DLUT?

    • This reply was modified 4 weeks, 1 day ago by j82k.
    Attachments:
    You must be logged in to view attached files.
    #29235

    Josh Bendavid
    Participant
    • Offline

    @chros interesting, do you know in which driver version the behaviour changed (for tonemapping of GUI in HDR output)?

    Once you have the test patterns working in HDR10 mode, the procedure is just to put the TV in calibration mode, then calibrate to gamma 2.2 BT.2020 (everything else should be the same as for SDR)

    You can do this with the 1D and 3D luts, or only 3D lut as you prefer.

    Let me know if you need more details.


    @j82k
    yes this is expected based on how the code was implemented.  The interpolation in aiopylgtv originally exists in order to deal with the fact that DisplayCal/Argyllcms are generating 256 points for the 1D LUT, but the TV needs 1024 points, but the interpolation code from numpy is completely generic.

    You can also do this programmatically in python if you prefer, generating any 1024 element LUT you want.  Curious why you weren’t able to get good results with the Displaycal-generated luts though.  Is it just a question of not enough points used near black for the calibration?  (The highest granularity setting accessible from the UI would give you 96 points for the greyscale calibration, so that should be even slightly more fine grained near black than you have here)

    #29236

    j82k
    Participant
    • Offline

    Curious why you weren’t able to get good results with the Displaycal-generated luts though.

    I think the problem with displaycal and my panel is that my i1Display Pro is unable to correctly read these dark patches especially colored ones and the non-linearity of my panel in that range makes this even worse. The displaycal generated calibration always ends up with either the near-black range still being way too dark or weird looking dark colors, color banding etc… Even if I manually adjust IRE5 to the correct luminance, at 3% it again becomes so dark that my meter can’t get a reading.

    i1Display Pro on Amazon  
    Disclosure: As an Amazon Associate I earn from qualifying purchases.

    #29237

    chros
    Participant
    • Offline

    The displaycal generated calibration always ends up with either the near-black range still being way too dark or weird looking dark colors

    I can confirm this, using DisplayCal for madvr SDR 3dlut, no matter which calibration settings you use (fast, slow, default) the result is completely random, sometimes it’s just slightly crushed other times more along with the weird colored  near blacks. You can get it sort of right let’s say 1 out of 10 🙂

    do you know in which driver version the behaviour changed (for tonemapping of GUI in HDR output)?

    I don’t know the exact version number for this, I wrote a summary about this. Which card and OS do you have? (here 1060 with v398.18 on Win10 1809) If you have a Pascal card then just revert back and hope it will still work with your Win10.

    I just opened the .cal file with a text editor and deleted all the data points except 0 and 100% and then did grayscale and near-black sweeps in HCFR, adding adjustment points wherever needed.

    Just for reference, this is the cal file I ended up with:

    That’s really interesting! Exactly what I try to achieve with DoVi!
    If we could get just 1 file like this that works then I bet it would be good for most of our sets (without replacement panel).
    Can you clarify how did you come up with the numbers in the cal file? Especially the 2nd (luminance) column? Di you just experiment with it?

    #29238

    chros
    Participant
    • Offline

     put the TV in calibration mode, then calibrate to gamma 2.2 BT.2020 (everything else should be the same as for SDR)

    You can do this with the 1D and 3D luts, or only 3D lut as you prefer.

    Let’s suppose I only want to upload 1D lut for now, is this right? (So I only selected 175 patches and Cal speed is High).
    Any comments about the attached settings?
    And what is the command line to upload the resulted cal file?
    Furthermore, what GPU settings shall I use? Full RGB 12 or 8 bit?
    Thanks

    Attachments:
    You must be logged in to view attached files.
    #29243

    j82k
    Participant
    • Offline

    That’s really interesting! Exactly what I try to achieve with DoVi!
    If we could get just 1 file like this that works then I bet it would be good for most of our sets (without replacement panel).
    Can you clarify how did you come up with the numbers in the cal file? Especially the 2nd (luminance) column? Di you just experiment with it?

    I just looked at a .cal file that displayCal created during the calibration which has 256 data sets and used that as a reference. The first value basically matches the % shown in HCFR when outputting 16-255 with madTPG. The 3 others are RGB adjustments that you can adjust after doing grayscale sweeps or just measuring single patches…

    For example from my reference file “0.00784314 0.00675731 0.00640922 0.00498094” would  be ~8% in HCFR so I add that and upload it to the TV then measure and adjust the RGB values until I’m happy.  I only used 4 decimal places though, not 8 like in the original file.

    The workflow is really similar to a manual TV calibration but with the advantage of being able to set as many adjustment points as necessary. For me this is mostly interesting for the low end range where I added most points. I might also use this for some minor adjustments after creating a 3DLUT when displaycal decides to brighten or darken my carefully adjusted near-black range.

Viewing 15 posts - 226 through 240 (of 265 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS