MadVR in use with LG OLED Thread

[chros]

Seriously how can you tolerate that? :rolleyes:
:) I got used to it or I just favor (near) chroma 4:4:4 this much :D
You can sometimes see the banding/posterisation effect even on human skins (not just on walls/sky) but in the majority of cases I'm not bodered: it's not thaaat bad :)

But, in the beginning I created a wrong 23p custom timings with madvr (not just the back porch values (https://forum.doom9.org/showthread.php?p=1868998#post1868998) were changed), it took me for about a week to realize that something was off:
it screwed up the gamma response of the TV badly, resulting in every kind weird things (heavy near black banding, darkened picture, more posterisation). That was baaad. :)

I just tried pc-mode HDR again a few days ago. Played the UHD of Brightburn, skipped to some dark scene and it looked like some low bitrate TV broadcast.
I haven't noticed anything really bad with it, probably sometimes it was there, or maybe I just got used to it.

You are confused I'm afraid. [emoji846] Your question about what nit setting I used tells me so. You're talking about soulnights measurement modification tool. I am not. His tool is used to dynamically change the measurement file for those doing HDR to SDR. That's not what I want. I create the measurement file with madvr and do not touch it. This is very different then what the dynamic algo guys are doing.

Indeed, I thought about soulnights's tool. :) In this case it's not even DTM.

[chros]

More test:

- White windows max nits (compared to rtings's B8 result (https://www.rtings.com/tv/reviews/lg/b8-oled#comparison_140)):
100%: 145
10%: 700-750 (increasing quickly: maybe overheating the meter or display increases brighness?)

- checked how ASND behaves on B8 (https://www.avforums.com/threads/lg-oled-c8-owners-and-discussion.2162797/page-428#post-27335734)

[SamuriHL]

Indeed, I thought about soulnights's tool. :) In this case it's not even DTM.

That's also not accurate to say that it's not dynamic. Madvr is using the measured data for every frame to properly tone map for the max target nit. It's not real time dynamic tone mapping but it is the most accurate because it knows ahead if time the metadata for each frame. I don't need or necessarily want what the live algo or soulnights tool are doing because I'm not trying to compress the SDR range. In fact I'm trying to keep that untouched since I have room above it for the highlights. The straight measurement file untouched does what I want.

Sent from my SM-G975U using Tapatalk

[quietvoid]

Strange it works fine here, even when I switch between passthrough and pixelshader (I haven't noticed this before).
Which player do you use? I use MPC-BE / HC. I use MPC-BE as well, the WB 20 points just never change unless I use passthrough and check/uncheck "send HDR metadata to display".
Maybe LG ignores the peak nits, maybe madVR should be sending MaxCLL every time the nits change dynamically?

Indeed, I don't see any difference between 50 and 200 (not even in numbers in OSD), and there was a huge diff when I used madmeasure78 with SDR TV.
Maybe the bottom scene detection values, or did it get broken? Essentially I think it just tries to use higher target nits dynamically, and no difference with everything below the peak nits because LG applies its own PQ curve (and gamma, etc).
While with SDR it would also affect the overall brightness.

Does "apply dynamic clipping"? I haven't noticed it. Which clip(s) and when?
It should modify the protected 0-100 nits range.
Note that this option is broken since the beginning and madshi didn't fixed it yet, hence the lower 25 value is used. Maybe lowering it helps? 15/10? It's just something I have noticed playing back other titles, in scenes where the peak nits don't go over 300 nits, using LG's DTM increases overall brightness of everything over 100 nits it seems.
Pretty much like the current "dynamic clipping". The other one (in the madVR optimizer by Soulnight), has an extra algo to only apply on % max pixels to avoid that.
And for very bright scenes, with LG's DTM more highlight detail is resolved, but "dynamic clipping" might cause some clipping (haven't noticed any with 25, but it happens with 100).

[VBB]

:) I got used to it or I just favor (near) chroma 4:4:4 this much :D

And you're using 12-bit on top of that ;) Doesn't that make the banding even worse? It definitely does on my C7 in both normal and PC modes. If what Stacey Spears claims is right, then PC mode switches the TV into 8-bit only anyway.

I also recently tried PC mode, and while at first things appeared a bit sharper in video content, that could have very well been placebo. I sit about four feet from my 65", and I can't really tell the difference between 4:4:4 and 4:2:2 when watching something. Maybe I could in a side-by side comparison, but that's just not possible.

[Asmodian]

For most live video content that was originally 4:2:0 the difference between 4:4:4 and 4:2:2 is very minor even with madVR scaling the chroma.

This changes a lot when viewing content that was originally 4:4:4 or RGB, for that 4:2:2 on my C9 is pretty obvious and unacceptable.

[VBB]

Right, so the lack of proper 4:4:4 is not a huge deal, if all you do is watch commercial content. Still would be nice if the next generation of panels finally got past all the bandwidth and processing limitations, now that HDMI 2.1 is here. Come on LG! I wanna buy the C10!!!

[SamuriHL]

I'm not upgrading again until at LEAST the C13. LOL

[chros]

That's also not accurate to say that it's not dynamic. Madvr is using the measured data for every frame to properly tone map for the max target nit.
Thanks, yes, you are right. It was a long time ago for me ... :)

I use MPC-BE as well, the WB 20 points just never change unless I use passthrough and check/uncheck "send HDR metadata to display".
Maybe LG ignores the peak nits, maybe madVR should be sending MaxCLL every time the nits change dynamically?
Hmm. Maybe different firmware version of the TV or an AVR in the middle of the chain?

If what Stacey Spears claims is right, then PC mode switches the TV into 8-bit only anyway.
Well, there's a slight difference using "03. Grayscale\04. 10bit test" files of "Mehanik HDR10 test pattern" between 8bit and 10bit (12bit output is set in nvidia CP).
That means processing is not 8bit, but maybe like with chroma 4:4:4, it's not 10bit either :D I don't know.

And you're using 12-bit on top of that ;) Doesn't that make the banding even worse?
As I stated in this post (https://forum.doom9.org/showthread.php?p=1871836#post1871836) about PC mode:
- SDR content: there's banding in PC mode using both 8bit and 10bit input (difference is minimal between the 2)
- HDR10 content: there's heavy (!) banding in PC mode using both 8bit and 10bit input (difference is clearly visible between the 2)

Maybe I could in a side-by side comparison, but that's just not possible.
You don't have to: pause the playback and switch the label between PC and HDMI.
Before that use a higher chroma algo in madvr (e.g. NGU Sharp low at least) not just Bicubic60 :)

I also recently tried PC mode, and while at first things appeared a bit sharper in video content, that could have very well been placebo. I sit about four feet from my 65", and I can't really tell the difference between 4:4:4 and 4:2:2 when watching something.

For most live video content that was originally 4:2:0 the difference between 4:4:4 and 4:2:2 is very minor even with madVR scaling the chroma.
I *do* notice the difference during normal playback even from normal viewing distance (~3 meters).
My reasoning for using PC mode with HDR10 is this: I see banding/posterisation - let's say - 10% of the times watching regular content, but I see the effect of increased natural sharpness 90% of the times :D
And you can easily switch the label when you watch SDR or HDR10 content, if you like.
But don't get me wrong, I don't want to poison your minds with this :)

Still would be nice if the next generation of panels finally got past all the bandwidth and processing limitations, now that HDMI 2.1 is here.
Don't wait for that :D

From one of Ted's write up (https://www.avsforum.com/forum/139-display-calibration/2962814-2018-lg-oled-calibration-user-settings-no-price-talk-105.html#post58372750):
"LG will convert any HDMI input RGB-Video colorspace signal to YCbCr 4:4:4 for applying initially most of the UI (user interface) adjustments.
The UI user setting will be merged in an algorithm which will interpolate values between multiple user settings to manipulate the video signal.
Looks like that there additional features of Expert Controls Menu (Dynamic Contrast, White Balance, CMS etc.) which will require more complex processing so for the TV to significantly reduce the bandwidth and the required processing power it will compress the horizontal chroma resolution in half; chroma subsampling from YCbCr 4:4:4 -> YCbCr 4:2:2; as this will reduce by 33.3% the video data bandwidth.
I believe this is the reason why these OLED TV's will not display full chroma (YCbCr 4:4:4 or RGB-Video) with PC Icon enabled (of HDMI Input) from 24p/30p/50p input signal but only with 60p signal where all the Expert Control menu controls are disabled, so the video signal will bypass that 'Expert Controls Menu' processing step."

[huhn]

Well, there's a slight difference using "03. Grayscale\04. 10bit test" files of "Mehanik HDR10 test pattern" between 8bit and 10bit (12bit output is set in nvidia CP).
That means processing is not 8bit, but maybe like with chroma 4:4:4, it's not 10bit either :D I don't know.
you could show a difference between these using 6 bit so what's your point. that's why dithering exist to fake a bit deep that doesn't exist. the simples form is alternating pattern with for example 1 and 0 which would look like 0.5 ignoring gamma curves and such.

Don't wait for that :D

From one of Ted's write up (https://www.avsforum.com/forum/139-display-calibration/2962814-2018-lg-oled-calibration-user-settings-no-price-talk-105.html#post58372750):
"LG will convert any HDMI input RGB-Video colorspace signal to YCbCr 4:4:4 for applying initially most of the UI (user interface) adjustments.
The UI user setting will be merged in an algorithm which will interpolate values between multiple user settings to manipulate the video signal.
Looks like that there additional features of Expert Controls Menu (Dynamic Contrast, White Balance, CMS etc.) which will require more complex processing so for the TV to significantly reduce the bandwidth and the required processing power it will compress the horizontal chroma resolution in half; chroma subsampling from YCbCr 4:4:4 -> YCbCr 4:2:2; as this will reduce by 33.3% the video data bandwidth.
I believe this is the reason why these OLED TV's will not display full chroma (YCbCr 4:4:4 or RGB-Video) with PC Icon enabled (of HDMI Input) from 24p/30p/50p input signal but only with 60p signal where all the Expert Control menu controls are disabled, so the video signal will bypass that 'Expert Controls Menu' processing step."

sonys can do everything in 4:4:4 even entry level sonys...
and monitors anyway.

BTW. RGB is very useful for UI interfaces because you can alpha blend and such there is a reason computer and computer games use it.
doing colorcorrection and such is very easy even at 16 bit i would even say it is trivial simple quality scaling should cost more.
stuff like motion interpolation and such are processing intensive not the general work of a TV it's not much different from a monitor.

[chros]

Strange it works fine here, even when I switch between passthrough and pixelshader (I haven't noticed this before).
This works well, so it's easy to test: just pause the playback and switch back and forth.
Passthrough vs pixelshader (LG's DTM is off):
- pixelshader results just a bit brighter image with titles above 760 nits (e.g. JohnWick@01:04 , Alien@01:24 , Meg@00:07)
- pixelshader results just a bit darker image with titles below 760 nits (e.g. Mockingjay@00:06 , Alita@00:10)

Pixelshader is better with highlights
- Meg:
-- 00:03: sky is a bit whiter, blue leds on the girl's shoe is more vivid; waves/antennas getting more visible with highlight recovery very high
-- 00:15: girl's hair clip
- Godzilla:
-- 00:23: lights above the stove
-- 01:09: light is wither
- Mad Max:
-- 00:05: reflection om the edge of the car
-- 00:52: screw on the wheel looks more shiny
-- 02:15: highlights in the orange sky to the left (that's the biggest difference)

Pixelshader is different (worse?):
- JohnWick:
-- 00:04: blue lights to the left
-- 00:15: blue lights to the right

Notes:
- highlight recovery very high can help 4000's titles and above, it doesn't have any effect below (I've updated the settings picture)
- I haven't checked LG's DTM yet

[j82k]

Don't wait for that :D

From one of Ted's write up (https://www.avsforum.com/forum/139-display-calibration/2962814-2018-lg-oled-calibration-user-settings-no-price-talk-105.html#post58372750):
"LG will convert any HDMI input RGB-Video colorspace signal to YCbCr 4:4:4 for applying initially most of the UI (user interface) adjustments.
The UI user setting will be merged in an algorithm which will interpolate values between multiple user settings to manipulate the video signal.
Looks like that there additional features of Expert Controls Menu (Dynamic Contrast, White Balance, CMS etc.) which will require more complex processing so for the TV to significantly reduce the bandwidth and the required processing power it will compress the horizontal chroma resolution in half; chroma subsampling from YCbCr 4:4:4 -> YCbCr 4:2:2; as this will reduce by 33.3% the video data bandwidth.
I believe this is the reason why these OLED TV's will not display full chroma (YCbCr 4:4:4 or RGB-Video) with PC Icon enabled (of HDMI Input) from 24p/30p/50p input signal but only with 60p signal where all the Expert Control menu controls are disabled, so the video signal will bypass that 'Expert Controls Menu' processing step."

If the C8 internally works with YCbCr and only converts to RGB in the last step that would explain why I got better results outputting YCbCr in SDR pc-mode. Outputting YCbCr would mean additional color conversions happen on the pc side but less color conversions happen inside the TV.

And if the TV does all the internal processing with only 8-bit precision in pc-mode that would also explain the banding. I think madVR does such things in 16 or even 32-bit.

[SamuriHL]

It would make sense that the C8 would internally use YCbCr because most consumer equipment will be sending that signal to it. Those conversions are ok when they use proper dithering, but, given that we see banding on the C8, it is very possible that either its dithering sucks or is largely non-existent in those conversions. UGH.

[VBB]

And in the end we'll never know for sure (even the mighty Ted, for whom I have the utmost respect :)).

[ashlar42]

I still don’t own an OLED display but I’ve honed my Kuro to perfection in all these years. Reading all the caveats and quirks of these displays is kind of scary to be honest. :D
Banding, bitdepths, pc mode, it all seems really messy.

[huhn]

the list of issues with an kuro is even longer.
CR at all cost to make it short.

[VBB]

The benefits of OLED far outweigh any inherent issues, IMO. I can't watch anything on an LCD anymore :D

[QBhd]

Oh yeah... we bitch and moan about the littlest of things, but if ANYONE of us OLED owners had to go back to any other tech, I'm sure we all would rather have teeth pulled!

Maybe, MAYYYBEEE some of us would go back to Plasma, but even then, I doubt it.

QB

[VBB]

My only gripe is motion, really. There just isn't a good way to compensate for instant pixel response with sample-and-hold technology. This is where 24 fps really shows its age. The closer you sit, the more obvious the issue becomes. In the other thread I described how I now use madVR's smooth motion to play everything back at 60Hz. This actually works pretty well, but it adds a bit of blur to motion that would otherwise be stuttery on OLEDs (when played back without smooth motion and in its native frame rate). I do prefer it over the stutter, though. Playback at 120Hz should make this even better, when HDMI 2.1 is fully implemented and video cards are available.

[huhn]

at 120 HZ it would get "worse"(if you want the "ghosting") because you get less blended frames and they are shown shorter letting it look more like native 24 hz.

[VBB]

I'll still take it over no blending at all. Another benefit (for me anyway) of playback at 60Hz is no need for audio delay. Also with smooth motion there are no repeated frames, so custom resolutions become unnecessary.

[Asmodian]

Yep, smooth motion is a great feature for OLEDs. It works way better than I would have expected. :)

[chros]

If the C8 internally works with YCbCr and only converts to RGB in the last step that would explain why I got better results outputting YCbCr in SDR pc-mode.
In SDR mode? I haven't noticed anything like this with SDR content with RGB Full 12bit input.
If you have such a scene (preferably BD remux) then upload it somewhere.

Outputting YCbCr would mean additional color conversions happen on the pc side but less color conversions happen inside the TV.
Yes, although this seems the best option in PC mode (see below), from madshi's post (https://forum.doom9.org/showthread.php?p=1271416#post1271416):
"If you switch your GPU control panel to YCbCr, the GPU receives RGB from Windows, but you ask the GPU to send YCbCr to the TV. Consequently, the GPU has to convert the RGB pixels behind Windows' back to YCbCr. Some GPU drivers might do this in high bitdepth with dithering, which may produce acceptable results. But some GPU drivers definitely do this in 8bit without any dithering which will introduce lots of nasty banding artifacts into the image. Furthermore, there are various different RGB <-> YCbCr matrixes available. E.g. there's one each for BT.601, BT.709 and BT.2020. Now which of these will the GPU use for the conversion? And which will the TV use to convert back to RGB? If the GPU and the TV use different matrixes, color errors will be introduced. As a result I cannot recommend this configuration."

I notice a huge difference in banding with full range RGB input between PC mode and HDMI mode. I would rather have the banding than subsampled chroma, but I use the display for a lot more than watching 4:2:0 compressed video.

If I use YCbCr444 limited range input I get basically the same banding as when using HDMI mode, but while in PC mode with 4:4:4 chroma. This has the GPU converting madVR's output but it seems to do a better job than the display does, which probably works in limited range YCbCr internally somewhere given the banding differences. This seems like the best compromise I can get. I cannot stand subsampled chroma but the banding is still pretty bad with raw 8 bit RGB input on the C9. :(
Only good news is that YCbCr444 results in the exact same (near) chroma 4:4:4 as RGB Full.

And if the TV does all the internal processing with only 8-bit precision in pc-mode that would also explain the banding. I think madVR does such things in 16 or even 32-bit.
16bit internally. But there *is* "banding" (and other crazy sh*t) with 8bit RGB Full as well in HDR10 mode.

I just tried pc-mode HDR again a few days ago. Played the UHD of Brightburn, skipped to some dark scene and it looked like some low bitrate TV broadcast.
This. Good spot with HDR10 content! It's not even 8bit vs 10bit but RGB vs YCbCr 4:4:4.
It's like some basic processing is missing in low APL (darkish) scenes with LG's tonemapping (with and without LG's DTM) that destroys the image (madvr tonemapping can't help this). Interestingly I haven't spot wierd things in normal/high APL scenes.

Download this sample (~3.6GB) (https://mega.nz/#F!r9N0VK5R!WnnyYsGqwa7M-v1jckw6iQ) (01-2160p_23fps_hdr0686-thgm_banding.mkv).
I spotted at least 3 different problems (maybe 2nd and 3rd is the same), tested with passthrough + DTM + RBG full 12bit:
- banding: e.g. red light of torches around @03:20
- added noise: @00:32, around @01:16
- posterisation: clouds in dark scenes, @03:20, @03:25, etc.

Posterisation is just crazy sometimes! :) That's what you've noticed and I didn't so far. That's the biggest problem and this has nothing to do with 8bit vs 10bit as it's also present with 8bit.

8bit YCbCr 4:4:4 (limited) eliminates a lot of them, but there's still banding and the nasty posterisation still exist e.g. @03:25 (although removes it from @03:20).

None of this happens in HDMI mode even with RGB Full 12bit!

[iSeries]

Hey for those using/testing madVR tonemapping (with the latest test build) with HDR output to your OLED, what tonemapping settings are you using in madVR? There's a ton of options and I don't really know what most of them do and what values to use. Also, if using DTM, does using madVR tonempping even make a difference? Any help here appreciated.

I did just manage to disable Real Cinema in HDR mode and it only took me 10 minutes lol. Smooth Motion is definitely the way to go with these.

[ashlar42]

at 120 HZ it would get "worse"(if you want the "ghosting") because you get less blended frames and they are shown shorter letting it look more like native 24 hz.I’m not sure about this.
From Asmodian options guide:
The blurring is more visible when the refresh rate matches the source frame rate because all the frames will be blended frames. The blurring is less visible on higher refresh rate displays because fewer frames will be blended frames.
[CUT]
Note that smooth motion has less blurring when displaying 23.976 fps at 60 Hz than 23.976 fps at 24 Hz so this option is not recommended unless you cannot increase the display's refresh rate.
It’s rather confusing. I would say that displaying 23.976 material at 120Hz should lead to results similar to 23.976 at 24Hz. All frames would be blended frames, wouldn’t they?

[SamuriHL]

Hey for those using/testing madVR tonemapping (with the latest test build) with HDR output to your OLED, what tonemapping settings are you using in madVR? There's a ton of options and I don't really know what most of them do and what values to use. Also, if using DTM, does using madVR tonempping even make a difference? Any help here appreciated.

I did just manage to disable Real Cinema in HDR mode and it only took me 10 minutes lol. Smooth Motion is definitely the way to go with these.

This is a very complicated topic, I'm afraid. You're not going to get a single, concise answer. That being said, I can offer my personal opinion on this topic.

-I have settled on disabling DTM on the LG for now...I may change this opinion once I do more testing.
-I use unscrewed with measurement files. I don't try to get cute with them and modify them for more "dynamic" solutions. :D This is a very personal choice.

For HDR tone mapping settings:

-Target nit to 760
-Highlight recovery strength Are you Nuts? This is also a very personal choice as this can introduce ringing but damn it's nice IMO.
-fire set to balanced

I ignore the rest of it because I'm not using the live algo.

I have unchecked everything in trade performance options. I use error diffusion 2, color checked. I use ngu aa high chroma. That about covers it. FOR ME.

[huhn]

I’m not sure about this.
From Asmodian options guide:

It’s rather confusing. I would say that displaying 23.976 material at 120Hz should lead to results similar to 23.976 at 24Hz. All frames would be blended frames, wouldn’t they?

you are just wrongly interpret his post.
if you use SM 24 hz with 23p file you get blended frames that are shown for about 42 ms which is very blurry.

if you do SM at 120 hz with 23p the blurried frames are shown for 8.3 ms about every 42-8.3 ms and the far majority is not blended

[j82k]

Guess which one is pc-mode. :D

HDR RGB 8-bit (pc vs non-pc):
https://i.imgur.com/Kgk6IK0.gif


HDR ycbcr 8-bit (pc vs non-pc):
https://i.imgur.com/Kxr4NDa.gif


PC-Mode HDR completely fucks up the low end on my C8. Not only the banding but also the colors and it transitions into black too fast. I didn't even try to somehow calibrate this out because even then the banding would still remain. What's funny is when you use madVR's 'reduce banding artifacts' it sometimes makes the banding look even worse in pc-mode.

I also tried this with 12-bit output but that didn't make it any better.

[SamuriHL]

PC mode on these displays is completely broken IMO. I tried it before and had to turn it back off because the issues are just too much. It's nice to have chroma 4:4:4, no question, but, breaking so many other things defeats the purpose.

[QBhd]

I have been thinking the black crush on PC mode never got fixed with the latest C8 firmware. For SDR, I have bumped the luminance of the 5% 20-point to +10, but even then blacks seem to fall off way too much.

I have not watched enough HDR before or after the FW update to make a comparison.

QB

[Asmodian]

I also tried this with 12-bit output but that didn't make it any better.

How does PC mode look with YCbCr444 limited range input?

[j82k]

Banding wise I'd say for sdr pc-mode ycbcr444 8-bit is the best one to use, for hdr pc-mode everything is just terrible... :)

[chros]

Hey for those using/testing madVR tonemapping (with the latest test build) with HDR output to your OLED, what tonemapping settings are you using in madVR? Also, if using DTM, does using madVR tonempping even make a difference?
Use these for testing (https://forum.doom9.org/showthread.php?p=1882771#post1882771), and that's the purpose of the exercise :)

Guess which one is pc-mode. :D
:D These are perfect examples!

PC-Mode HDR completely fucks up the low end on my C8. Not only the banding but also the colors and it transitions into black too fast.
Interesting, not on mine (see below).

I didn't even try to somehow calibrate this out because even then the banding would still remain.
It's not "banding", at least not in the traditional sense, more like posterisation (due to lack of processing): and it sometimes (not all the time) destroys the image.

for hdr pc-mode everything is just terrible...
Just watched the new Men in Black HD10 passthrough in PC mode RGB Full 12bit + DTM: it was gorgeous! :D
I tried to count how many times I notice some anomaly: 5 times during the whole movie.
This is a very bright movie (9000 maxCLL, 4000 MDL, 942 FALL), probably that's why.

Banding wise I'd say for sdr pc-mode ycbcr444 8-bit is the best one to use
I never noticed this on my set (PC mode RGB Full 12bit), and the last 2 days I even tried to look for a sample (switching between PC and HDMI mode).
If you will come across such a sample upload it, please :)

I have been thinking the black crush on PC mode never got fixed with the latest C8 firmware. For SDR, I have bumped the luminance of the 5% 20-point to +10, but even then blacks seem to fall off way too much.
Are you sure? Have you switched between PC and HDMI mode to confirm black crush?
That's why I'm still on 4.10.25 B8.

Panel variance... :)

Just quoting Ted (https://www.avsforum.com/forum/139-display-calibration/3043520-lg-oled-s-3d-lut-profiling-using-lightspace-thread-10.html#post58376964) here:
"Some LG OLED's have passed the factory QC (Quality Control) with near black clipping issues and shipped to stores, for these panels it will be required to adjust the Brightness to 55-56 or up to 60-61 (in worst case scenario) from the default 50 value for the near black clipping issue to be resolved. "
I saw this as well, it's really said. I know at least one person who has a C8 like this (brand new set, no panel replacement, black crush at default settings with the built-in app).
Fortunately my B8 was factory calibrated, original panel hasn't been replaced yet, so it has almost perfect gamma tracking in every mode (SDR, HDR10, DV; both in HDMI and PC mode).

I know people don't want to believe that there can be such a huge panel variance with LG Oleds and but I've experienced this with my own C8. The original panel had a very bright low-end and then I got a replacement panel (due to vertical banding) which had an extremely dark/crushed low-end at default settings.
This means that the factory calibrated settings (White point in service menu, 1dlut, something else?) on the mainboard are not good for the new panel, and that's "normal".
The question is: can we do something about it by ourself?
- we can modify the RGB values of White points in service menu
- but I'm not sure about the screwed up 1dlut: can you correct it e.g. with Calman?
- what else needs to be done in this case? (whether it's possible or not)

[j82k]

I never noticed this on my set (PC mode RGB Full 12bit), and the last 2 days I even tried to look for a sample (switching between PC and HDMI mode).
If you will come across such a sample upload it, please :)


From what I can tell, the pc-mode banding is mostly in the lower range and the banding from 12-bit output I could often see in the mid or higher range, like gradients in the sky. I could look for scenes but right now I don't feel like messing around with that again. I'm just done with pc-mode and 12-bit for now. I just occasionally like to check if there had been any changes with a firmware/nvidia/windows update regarding those issues. In SDR it's not that bad but I don't want to constantly switch between pc-mode for SDR and non-pc for HDR.


This means that the factory calibrated settings (White point in service menu, 1dlut, something else?) on the mainboard are not good for the new panel, and that's "normal".
The question is: can we do something about it by ourself?
- we can modify the RGB values of White points in service menu
- but I'm not sure about the screwed up 1dlut: can you correct it e.g. with Calman?
- what else needs to be done in this case? (whether it's possible or not)


I think LG only does a very basic 2-Point Gain RGB adjustment in the service menu for each panel. They don't do anything that would affect the black levels. So just a lack of quality control I guess...
All you can do if you got such a crushed low-end panel is trying to find some combination of brightness/2P Low RGB/ 20-Point that fixes the low-end and doesn't raise true black...

[chros]

I could look for scenes but right now I don't feel like messing around with that again. I'm just done with pc-mode and 12-bit for now.
No worries, maybe others will :)

I think LG only does a very basic 2-Point Gain RGB adjustment in the service menu for each panel. They don't do anything that would affect the black levels.
Hmm, I might ask Ted :)

[SamuriHL]

Nope I've no desire to mess with the broken PC mode again either. 8 bit hdmi mode for me.

Sent from my SM-G975U using Tapatalk

[ashlar42]

One thing I was wondering about, considering my current status as a plasma owner: if one looks closely to the pixels, on a plasma, one sees "noise". I think it's PWM noise. For many it's something they wanted to eliminate but I'm wondering if it would be feasible to replicate it on an OLED, through a shader or something. I am not sure but I'm half convinced that it was part of the "organic" look that many liked with plasma.

Edit: maybe using Random Dithering/Error Diffusion 1, with color noise and change dither for every frame could suffice? Testing this, I would say no.

[huhn]

try 5-6 bit random dithering that should give you a similar results

[Gadgety]

I got this TV a couple years ago only to watch movies. In the store the TV was demoed with a 4k opera, and the image was breathtaking, and it looked like HDR. All reviews I saw highlighted the possibility for HDR images. Quite a while late I learned that the HDR was only possible when feeding the images via a USB stick (!) not via HDMI, and that it was caused by a software flaw that LG had no intention of correcting. HDMI input gets overridden, I understand. I've got two questions, first, does this influence the usage with MadVR? Second, is there a workaround, say via a service menu settings, to get HDR on this panel from the HDMI input?

[Klaus1189]

No, only when you want to passthrough HDR metadata to the TV, but let madVR do the tonemapping and it is fine.

[sat4all]

https://youtu.be/wxMQLzJvr4A?t=172
Both LG OLED 2018/2019 models use 700nits as default HDR peak luminance for internal gamma to PQ EOTF processing, so if you want accurate PQ EOTF tracking you should reduce oled light to achieve 700nits.

[Gadgety]

Thank you Klaus1189! That's great! Would you know, this being a 55 inch TV, any point in doing NGU Very high, or is there no discernible difference between that and NGU high? I have a lot of 720 material.

[SamuriHL]

If you're upscaling 720 or 1080 material to 2160, then try ngu sharp very high for luma and see what you think. For 720 content with my 2070 I have to drop chroma to ngu aa medium. For 1080 I use ngu aa high. That's while using ngu sharp very high for luma. The results are quite impressive but play around with it and see what you think. Ultimately the "best" quality is judged by you.

Sent from my SM-G975U using Tapatalk

[Gadgety]

Thank you, I'm just looking into what GPU to get. I see, it'll be exhaustive on performance. It'd be really interesting to see image comparisons for various chroma levels from ngu aa medium and going down the range, just to see what I can expect (and ultimately what kind of card to go for). Perhaps such a comparison already exists?

[SamuriHL]

It's only exhaustive on performance depending on your goals. I've told you what *I* do and for that I ended up with an RTX 2070. You may decide you can live with some compromises on quality for a cheaper card. Or you may want to go all out and get a heftier card. The choice is yours. Madvr will make use of any performance you can give it. The Envy Pro is going to drive a 2080ti under the hood. I know some people are enjoying the newest AMD cards, as well. I've no experience with them so I can't comment on how good they'd be for madvr. Depending on what you're doing, some people are perfectly content with a 1060. I was not satisfied with mine for UHD and opted to upgrade. The point I'm trying to make here is that madvr will take advantage of whatever card you give it, and it's up to you to determine what level of quality you're looking to get out of madvr and price a card accordingly. Hope this helps even if it's not a direct answer to your question. But at least you may now understand why there isn't one.

[Gadgety]

Yes, thank you SamuriHL! Would you know of any example images comparing the various levels of low, medium, high, very high for NGU?

[SamuriHL]

I'm sure someone's put some out there. I don't know of any off hand. But I can't imagine it'd be too hard to find some.

[Gadgety]

Which would be better AMD Vega RX 64 or the RTX2060 for MadVR? ATM they're priced about the same, RTX slightly more expensive.

[SamuriHL]

I don't know the answer. The RTX 2060 is ok. I'm not sure it'll let you get to very high for luma upscaling but I don't know for sure. I know nothing at all about the AMD cards and how well they work for madvr so maybe an AMD user will chime in here. Really, I'm not the one to help you on this decision as I just don't know enough about those cards.

[Gadgety]

Thank you for the reply, anyway!