HDR to SDR 3dlut

Home Forums Help and Support HDR to SDR 3dlut

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #9200

    iSeries
    Participant
    • Offline

    Hi, I’m playing around with creating an HDR to SDR 3dlut from a previous calibration. I followed the instructions posted on the MadVR/ArgyllCMS thread, and selected source colour space BT2020, tone curve SMPTE 2084 (roll off) and entered the display peak luminance as measured. I changed black output offset to 100% as I don’t want a BT1886 type curve (I assume that is correct?), I want the same curve as my SDR 3dlut (REC 709 2.4). Using this profile everything seems too bright and washed out, like it is a BT1886 type curve which I don’t want. If I switch to MadVR’s own HDR to SDR conversion with my display peak luminance set there, things look much more like I’d expect (which I assume is doing tone mapping etc and then running it through my SDR 3dlut). Am I doing something wrong? If I open up advanced options in Displaycal, you can select ‘source colourspace’, but there is also a ‘content colour space’ which was defaulted to DCI P3. What’s the difference between content colour space and source colour space?

    • This topic was modified 6 years, 6 months ago by iSeries.
    • This topic was modified 6 years, 6 months ago by iSeries.
    #9203

    Florian Höch
    Administrator
    • Offline

    Did you actually set the HDR 3D LUT in madVR? Note that differences between madVR HDR to SDR shader or madVR with HDR to SDR 3D LUT should be quite small.

    #9205

    iSeries
    Participant
    • Offline

    Hi Florian,

    Yes I set MadVR to use the HDR to SDR 3dlut. When I switch to MadVR shader processing the image darkens and looks more like I’d imagine it should. With the 3dlut, like I said, it’s much brighter in the dark to mid tones. I had figured the differences should be quite small so I don’t know if it’s something I did wrong or this is just how it’s supposed to be. But definitely MadVR shader math gives a much more ‘contrasty’ look, and much closer to how SDR material looks through my SDR 3dlut.

    One more question I had was when I calibrated my TV, I had it in ‘wide’ gamut mode, does the HDR to SDR 3dlut take advantage of those extra colours?

    #9207

    iSeries
    Participant
    • Offline

    If I disable my SDR 3dlut from MadVR, switching between the HDR to SDR 3dlut and MadVR shader, the visual result is basically identical. When I enable my SDR 3dlut, then the difference is there as described. It’s like the gamma curve that is in the SDR 3dlut just isn’t there in the HDR to SDR 3dlut.

    EDIT – I just re-ran the HDR to SDR 3dlut generation in Displaycal, this time using 0% black offset, and the result is identical to the HDR to SDR 3dlut with 100% black offset. Is that right? I thought 0% offset would be a BT1886 type curve, and 100% offset would be the equivalent to my SDR 3dlut which was also created with 100% black offset.

    • This reply was modified 6 years, 6 months ago by iSeries.
    #9209

    iSeries
    Participant
    • Offline

    Disregard that about 0% and 100% looking identical, I had the 3dlut loaded into the wrong spot in MadVR. The difference between the HDR to SDR 3dlut and MadVR shader is still there though.

    #9213

    Florian Höch
    Administrator
    • Offline

    I don’t think there is a problem. If you use the pixel shader, madVR applies the normal Rec 709 3D LUT “on top” of the pixel shader output. In the case of a Rec 709 gamma 2.4 3D LUT, this means there is an additional gamma adjustment of +0.2 (I think madVR by default assumes 2.2 for Rec 709 video), leading to darker midtones.

    #9221

    iSeries
    Participant
    • Offline

    This is the look I’d like to achieve with the HDR to SDR 3dlut – as far as I can understand it, the HDR to SDR 3dlut employs ‘bt.2084’ which seems to be an eotf designed for HDR and HDR TV’s. As I’m watching in SDR on a plasma, is there any way I can make a HDR to SDR 3dlut that would mimic the tone curve of my SDR 3dlut (2.4 100% output offset)?

    • This reply was modified 6 years, 6 months ago by iSeries.
    • This reply was modified 6 years, 6 months ago by iSeries.
    #9226

    Florian Höch
    Administrator
    • Offline

    The important thing to understand is that BT.2084, unlike any ‘gamma’ (including BT.1886) before it, is an absolute (in terms of luminance) EOTF. Which means each code word maps to an explicitly defined luminance on screen. Needless to say, this approach is not really compatible with any gamma adjustment ‘after the fact’.

    #9229

    iSeries
    Participant
    • Offline

    I think I understand that, and it all makes sense if watching HDR in HDR. I guess what I’m not understanding is once HDR is converted to SDR, why is BT.2084 needed? At that point once everything is in standard dynamic range, why can’t any EOTF be then applied to it?

    #9231

    Florian Höch
    Administrator
    • Offline

    The HDR to SDR is simply a madVR distinction to make its configuration easier. All that any 3D LUT for HDR material does is the same thing as any 3D LUT does: It makes the displayed content conform to specific parameters. In case of HDR, it makes the display follow the BT.2084 EOTF. Any adjustment afterward is nonstandard.

Viewing 10 posts - 1 through 10 (of 10 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS