LG C8 Lut

[chros]

Out of curiosity I have uploaded your cal file into Dolby Game preset, and although there is still no true black but the glowing with test patterns is gone 🙂

The uploaded cal is also does something:
– colors getting a bit pale vs the untouched Cinema preset
– “shadow” details are getting brighter (although I couldn’t see much difference on the test patterns, blinking squares are the same in both modes)

So maybe it’s worth playing with it, I used only these command with dolby game preset:

aiopylgtvcommand 192.168.1.78 start_calibration dolby_game
aiopylgtvcommand 192.168.1.78 set_1d_2_2_en dolby_game
aiopylgtvcommand 192.168.1.78 set_1d_0_45_en dolby_game
aiopylgtvcommand 192.168.1.78 upload_1d_lut dolby_game
aiopylgtvcommand 192.168.1.78 upload_1d_lut_from_file dolby_game "j82k.cal"
aiopylgtvcommand 192.168.1.78 end_calibration dolby_game

• This reply was modified 3 years, 1 month ago by chros.

[chros]

– “shadow” details are getting brighter (although I couldn’t see much difference on the test patterns, blinking squares are the same in both modes)

This one was wrong: after switching between unity and j82k.cal there’s a clear difference in the flashing boxes as well.
If I add this into the dataset, more black clipping happens:
0.0080 0.0000 0.0000 0.0000
So it seems the factory 1dlut causes the weird glowing in near blacks for whatever reason?

• This reply was modified 3 years, 1 month ago by chros.

[chros]

And checking with dark scene with fog in the background, guess what happened: the cyan/magenta issue kicked in in DoVi game mode as well! 😀

[j82k]

That .cal file was for SDR 100 nits, gamma 2.4.

I just tried a manual calibrated 1DLUT for HDR. Setting the TV to calibration mode and then calibrating to gamma 2.2.  I guess that’s working fine but stupidly I didn’t set the 100% point to 700 nits so my PQ EOTF is now about 13% too bright over the complete range.

Only required 10 adjustment points. If you’re gonna upload this again to your TV I’d probably delete the 0.5% and 1% points as they’re there to fix my too dark near-black range..

.cal file looks like this:

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 10
BEGIN_DATA
0.0000 0.0004 0.0004 0.0004
0.0050 0.0128 0.0118 0.0118
0.0100 0.0160 0.0148 0.0142
0.0200 0.0233 0.0210 0.0193
0.0400 0.0442 0.0432 0.0372
0.1000 0.1008 0.0983 0.0861
0.2000 0.1995 0.1960 0.1690
0.3000 0.3093 0.3038 0.2620
0.5000 0.5110 0.4965 0.4310
1.0000 1.0000 0.9700 0.8420
END_DATA

Attachments:

You must be logged in to view attached files.

[j82k]

And checking with dark scene with fog in the background, guess what happened: the cyan/magenta issue kicked in in DoVi game mode as well!

Regarding the HDR cyan/magenta issue … I think I found a fix by creaing a synthetic 3LUT, so no messy HDR calibration needed. 🙂

LG C8 HDR 3DLUT cyan/magenta issue

In the synthetic profile creator I created a Rec2020 and a DCI-P3 D65 profile. Black level 0, white level 10000, gamma 2.2 (not sure these settings even matter).
Then in the 3D Lut maker I used the Rec2020 profile as source and th DCI-P3 one as Destination, rendering intent ‘Perceptual’.
I tried all 10 rendering intent settings and some fix the issue and some don’t.

With perceptual from what I can tell most scenes will look identical to the factory 3DLUT, only these darker colored scenes look different. And it doesn’t mess up hue/saturation in brighter scenes like the unity 3Dlut does.

I did a factory reset on the C8 before doing these comparisons and used technicolor (set to warm 2) and cinema HDR presets for comparison, one with the original 3DLUT and one with the synthetic one.

With this fixed + the 1DLut which I use to fix my messed up near-black range I think I might be able to use the C8 in HDR mode again. I’ve basically only used SDR (madVR tonemapping) for a very long time.

Now I’m wondering if only some LG C8 have this messed up factory HDR 3DLUT or is it all of them and nobody notices?

[chros]

That .cal file was for SDR 100 nits, gamma 2.4.

Cheers, I know I just wanted to try out if it does anything at all.

So, I tried your new cal file (I haven’t modified anything, so I know it’s brighter in the shadows), and I do NOT see the cyan/magenta issue in DoVi game preset anymore!!! So maybe this can work?! 🙂
Obviously it brightens up the “shadows” but I have to wait until dark to say more about this 🙂

Now, I have to understand how you did this and try it for myself:
– what GPU output settings did you use? RGB FULL 12bit?
– in HCFR: SDR gamma 2.2 power BT.2020 in automatic mode? In the madvr settings: I guess HDR is off?

[chros]

Regarding the HDR cyan/magenta issue … I think I found a fix by creaing a synthetic 3LUT, so no messy HDR calibration needed.

Wha’???!!! Wooow!!! I want to try it out right now!!! 😀
The images look good from here.

In the synthetic profile creator I created a Rec2020 and a DCI-P3 D65 profile. Black level 0, white level 10000, gamma 2.2 (not sure these settings even matter).
Then in the 3D Lut maker I used the Rec2020 profile as source and th DCI-P3 one as Destination, rendering intent ‘Perceptual’.

I guess you talk about DisplayCAL-synthprofile.exe and DisplayCAL-3DLUT-maker.exe right?
Can you please upload screenshots of these (because of the settings) and maybe all the resulting files (in 7zip format) somewhere?

I tried all 10 rendering intent settings and some fix the issue and some don’t.
With perceptual from what I can tell most scenes will look identical to the factory 3DLUT, only these darker colored scenes look different. And it doesn’t mess up hue/saturation in brighter scenes like the unity 3Dlut does.

I played with them 2-3 years ago but only for SDR: there were (at that time for me) 2 groups: absolute vs the rest.
What about absolute, did it fix the issue?

used technicolor (set to warm 2) and cinema HDR presets for comparison, one with the original 3DLUT and one with the synthetic one.

Perfect! Probably I’ll play with Cinema and Cinema Home: you can configure Cinema Home (in HDR10) to look identical to Cinema (so the 2 presets are beside of each other, easier to switch):
– color temp w50, gamut: extended ! (this is on my B8)

(That’s why DoVi is so hard, only the Game preset can be set near to Cinema, and we have a small lag during switching between them.)

Josh, can we switch to different picture preset with aiopylgtv? (I quickly went through the code but haven’t find it.) That would be the easiest: we wouldn’t have to bring up the OSD that just shines in HDR 🙂

With this fixed + the 1DLut which I use

So, dd you apply a DDC reset at first? Does this fix only require our custom 3dlut (and uploading a unity 3dlut at first) nothing else?

I think I might be able to use the C8 in HDR mode again. I’ve basically only used SDR (madVR tonemapping) for a very long time.

😀 Exactly! That would be awesome! We wouldn’t have to think to buy another newer model, hoping maybe that one is better 🙂

Now I’m wondering if only some LG C8 have this messed up factory HDR 3DLUT or is it all of them and nobody notices?

I’m 99% sure it’s all of them, and don’t forget you have C8 with different firmware now and I have a B8 with the good old 4.10.25 firmware. I’ll write a bit more about this later when I managed to try your solution…
Amazing!!! 🙂

• This reply was modified 3 years, 1 month ago by chros.

[chros]

So, dd you apply a DDC reset at first? Does this fix only require our custom 3dlut (and uploading a unity 3dlut at first) nothing else?

j82k, are you sure that wasn’t your 1dlut that fixed this? As we can see from the DoVi experiment, 1dlut can clearly produce this sh!t 🙂

[chros]

So, I tried it out, and indeed only the 3dlut modifies this. The following Rendering intents fix this perfectly, but introducing other anomaly:
– perceptual (ip), luminance preserving perceptual appearance (ilp), perceptual-appearance (ipa): all of them produce elevated shadows
– preserve-saturation (ims): this looks promising (sometimes it doesn’t have elevated issue, other times it has)

And here is other 2 that also interesting:
– relative colorimetric (ir): partly fixes cyan, magenta is the same, elevated shadows
– saturation (is): fixes cyan,  magenta is the same, oversaturated colors

Now I uploaded ims and I’ll start to use it to look for issues. Maybe the others also can be made darker a bit with different gamma settings.
I attached the 3 screenshots for the 2 displaycal apps.

Attachments:

You must be logged in to view attached files.

[Josh Bendavid]

So it should be clear that generating synthetic templates with DCI-P3 and BT.2020 is only an approximation.

You could do better generating a synthetic template mapping the measured wide mode color primaries onto BT.2020, however as previously said in a different context, this would have the ambiguity that you would have to decide whether to use primaries measured at the onset of desaturation (somewhere around a few hundred nits I guess) or at peak brightness.

I don’t think a proper XYZ LUT profile-based 3d lut is completely hopeless, but I haven’t had a chance to work through all the details yet.

An intermediate option would be to try and do some kind of interpolation between a Matrix based 3d lut (essentially what you are generating from the synethetic profile) with primaries measured at the desaturation point and at peak brightness.

[j82k]

So, I tried it out, and indeed only the 3dlut modifies this. The following Rendering intents fix this perfectly, but introducing other anomaly:
– perceptual (ip), luminance preserving perceptual appearance (ilp), perceptual-appearance (ipa): all of them produce elevated shadows
– preserve-saturation (ims): this looks promising (sometimes it doesn’t have elevated issue, other times it has)

And here is other 2 that also interesting:
– relative colorimetric (ir): partly fixes cyan, magenta is the same, elevated shadows
– saturation (is): fixes cyan,  magenta is the same, oversaturated colors

Now I uploaded ims and I’ll start to use it to look for issues. Maybe the others also can be made darker a bit with different gamma settings.
I attached the 3 screenshots for the 2 displaycal apps.

I didn’t notice any elevated shadows and I looked at quite a few scenes and compared to the factory 3DLUT. Also I didn’t modify the 1DLUT while testing, that’s why I had to factory reset the TV to get the default 1DLUT back… Did you use the factory 1DLUT? If not maybe that caused the elevated shadows…

Oh and in the 3DLUT maker I set tone curve to unmodified, not sure if that makes a difference…

So it should be clear that generating synthetic templates with DCI-P3 and BT.2020 is only an approximation.

You could do better generating a synthetic template mapping the measured wide mode color primaries onto BT.2020…

If you could write a small tutorial on how to do this I’d glady try it because I have no clue how..  🙂

I think the synthetic 3DLUT is very similar to the factory one on the LG, bright scenes look 100% identical, only in some scenes with darker colors I can see a difference. I even checked some HDR color clipping pattern and they also looked identical..

[chros]

Did you use the factory 1DLUT?

Yes, I only uploaded unity 3dlut first then our synthetic ones. And the previously mentioned 3 perceptual rendering (ip, ilp, ipa) result in desaturated colors, that’s why I was glad to find the 4th one (ims) that doesn’t.
Of you like I can upload a 2 minutes sample that I used, so we will see the same thing.

do some kind of interpolation between a Matrix based 3d lut (essentially what you are generating from the synethetic profile) with primaries measured at the desaturation point and at peak brightness.

I guess what Josh means is to measure our primaries and enter into the dci-p3 synthetic profile window?

Oh and in the 3DLUT maker I set tone curve to unmodified, not sure if that makes a difference…

OK, I’ll try out that one.
2 things I noticed in the synthetic profile window:
– bt2020 defaults to rec709 tone curve (not gamma 2.2)
— if you change it to g2.2 then profile class will be changed to Display from Input device
– dci-p3 d65 defaults to gamma 2.6 (!)

What do you think?

• This reply was modified 3 years, 1 month ago by chros.
• This reply was modified 3 years, 1 month ago by chros.

[j82k]

Yeah sure upload the sample and I’ll take a look..

I don’t know about those synthetic profile creator settings. I guess it’s worth a try to see if they even make a difference. I just fixed my HDR EOTF so I don’t feel like doing any more testing for today. 🙂

By the way, I don’t think I got desaturated colors with any of the 10 rendering intents.  Only the saturation ones boosted saturation…

[chros]

I can upload a 2 minutes sample

Here’s 1:30 minutes sample (up for a week):
– cyna magenta issue at multiple places (0:00-0:20, 0:51-0:59, etc)
– elevated issue (when present): 0:34-0:37, 1:10

I just fixed my HDR EOTF so I don’t feel like doing any more testing for today.

😀 Understandable. Is that a new cal file? If so then share it again please 🙂

• This reply was modified 3 years, 1 month ago by chros.
• This reply was modified 3 years, 1 month ago by chros.

[j82k]

I checked the sample, switching back and forth between presets and what I mostly see is that the synthetic 3DLUT desaturates cyan and in some parts that makes it a bit brighter. If I just watched it without comparing I wouldn’t say anything in that sample is elevated. Maybe it looks different on the B8 but I prefer the synth 3DLUT (perceptual) for pretty much the whole clip.

I wonder if there is a way to measure these colors in HCFR? Saturation sweeps don’t show anything unusual…

Here is the new 1DLUT with corrected EOTF but I really don’t think this will match your panel.  I also raised 0% as much as I could without inducing black glow, so set it to 0 if yours glows…

NUMBER_OF_SETS 13
BEGIN_DATA
0.0000 0.0003 0.0003 0.0003
0.0050 0.0122 0.0112 0.0112
0.0100 0.0162 0.0150 0.0144
0.0200 0.0234 0.0214 0.0197
0.0400 0.0440 0.0432 0.0377
0.1000 0.0988 0.0972 0.0857
0.1500 0.1485 0.1450 0.1266
0.2000 0.1942 0.1910 0.1666
0.3000 0.2984 0.2940 0.2560
0.5000 0.4920 0.4830 0.4185
0.7000 0.6763 0.6583 0.5723
0.9000 0.8560 0.8340 0.7250
1.0000 1.0000 0.9710 0.8430
END_DATA

[Author]

Posts