MadVR in use with LG OLED Thread

[SamuriHL]

Ah that's right I do remember reading that. Well that's unfortunate. But it does save me a lot of testing time. LOL Like I said, though, I'm quite happy with what madvr is giving me with the measurement files.

[chros]

One that has an LG OLED can be interested in this: III. section (https://forum.doom9.org/showthread.php?p=1871836#post1871836)

[SamuriHL]

C8 owners have something to smile about. Grab the 5.10.03 update as it seems to fix gamma black crush finally and seems to have improved dithering.

[QBhd]

Yes, it's a very nice boost to PQ even for someone like me that had the old but okay'ish x.x.15 firmware.

I have a quick question... in madVR with HDR tone mapping, what should I set "targer peak nits" to for a C8. I read it can do 820, but I don't have a meter to measure it.

QB

[SamuriHL]

I don't have a meter either. I typically set it to ~700-800 to make sure I'm under the max nits. Setting it slightly higher means that any frames that are slightly above the max nits of the TV won't be tone mapped by madvr, but, instead will be tone mapped by the TV. Which the C8 does a pretty good job at tone mapping. If you want to try to get as close as you can to the max theoretical nits you can try 800 and see how that looks to you. Right now I think I've got mine set ~750 or so. I keep messing with it to try to find that sweet spot without actually buying a meter, but, in the end ~700 works quite well and is most definitely under the max nit level for our TV.

[QBhd]

I was thinking the same thing. Good to hear another owner confirm it. Plus it always leaves the door open for another owner with proper gear to set us straight later on :P

BTW... I am stunned my ancient R9 270X can actually do any kind of 4k HDR. It took some work to get stable frames (especially on scenes with bitrates approaching 100k) but she does it! Even able to do tone mapping with NGU High for chroma! And to think this card was dead and I baked it in the oven to bring it back to life!

QB

[SamuriHL]

Yea, I suspect these panels vary in the actual max nits they can reach. IMO, better to be safe and keep it slightly under so that any tone mapping is handled by madvr. IMO it looks pretty damn awesome anywhere between 700 and 800.

That's awesome. I've no experience with the AMD cards. I'm loving my RTX 2070 right now. It works very well.

[j82k]

I think when using MadVRs tone mapping it is actually beneficial to send fake metadata to the TV instead of the real one. As you can see in the gif, the higher the maxCLL is in the metadata the more the TV will tone map.
But if you use MadVR for tone mapping you'd want the TV to do as little as possible which you can archive by sending the bogus metadata of either windows HDR or one of the older broken nvidia drivers.
This test was done on a C8, on other year models it might work differently.

https://i.imgur.com/717TZSf.gif

[SamuriHL]

That seems whacked though because those are sending, if memory serves, 1000 nits. What would be better is for madvr to send proper modified HDR metadata for the max target nits. You're saying it's not doing that?

[j82k]

I don't think madVR will modify the metadata. Not 100% sure on this though, maybe somebody else knows...

But when sending 1000 nits MDL metadata (like the fake one), the maxCLL value doesn't seem to matter. The TV will tone map the same as with MDL 4000/maxCLL 1000 in the gif, which is almost like clipping and I guess is what you'd want when letting madVR do the tone mapping.

[SamuriHL]

But if madvr were to modify the metadata to the max target nits, then, the TV wouldn't need to tone map at all. When madshi returns, it might be a good question to ask him and if it's not modifying the metadata, then maybe he can add an option to do so. I'd be curious to see what happens in that case.

[huhn]

madVR can/was already able to do that.

in in an older version you can even define the meta data.

[Warner306]

https://i.imgur.com/717TZSf.gif

That is a cool way to represent LG's tone mapping. Is the LG always brightening the highlights? It only seems to track 1:1 in the SDR range.

If madVR does the tone mapping, it is supposed to be sending both a mastering display peak value and an altered MaxCLL. Your graph implies any value 1,000 nits and under is virtually being represented with clipping. Which is ideal if you are using external tone mapping.

Someone with an HDFury should confirm madVR is actually sending those values to the display.

[j82k]

That is a cool way to represent LG's tone mapping. Is the LG always brightening the highlights? It only seems to track 1:1 in the SDR range.


This measurement was done using the technicolor preset at default settings. With it's custom white point it is quite a bit brighter than the more accurate warm2.
When I adjust color temperature to D65 the TV follows the EOTF more closely but also has a lower peak brightness.

As far as I know the C8s HDR EOTF tracking is designed for 700 nits peak so if the TV gets brighter than that it will go above the EOTF but you can adjust it using the 20 Point WB controls.

[SamuriHL]

madVR can/was already able to do that.

in in an older version you can even define the meta data.

That was my understanding, too, but, I'm far from any kind of expert in this HDR stuff. I would have thought tone mapping to a max nit level would change the metadata so the TV would know that it doesn't have anything to do.

[SamuriHL]

This measurement was done using the technicolor preset at default settings. With it's custom white point it is quite a bit brighter than the more accurate warm2.
When I adjust color temperature to D65 the TV follows the EOTF more closely but also has a lower peak brightness.

As far as I know the C8s HDR EOTF tracking is designed for 700 nits peak so if the TV gets brighter than that it will go above the EOTF but you can adjust it using the 20 Point WB controls.

That makes sense and may be where I got the recommendation for max target nit of 700 from originally. I think I'm running at 750 right now and it works well, but, I've not done any WB adjustments on mine.

[j82k]

madVR can/was already able to do that.

in in an older version you can even define the meta data.

That was my understanding, too, but, I'm far from any kind of expert in this HDR stuff. I would have thought tone mapping to a max nit level would change the metadata so the TV would know that it doesn't have anything to do.

So is that how it works? MadVR will send metadata according to whatever you specify in the "target peak nits" box? If so will it just use the specified value for both MDL and maxCLL?

[SamuriHL]

Partially. The MaxCLL is a little bit more complicated. I use measurement files and I *BELIEVE* madshi is using the 99% value for that. Don't quote me on this. I get very confused with all this HDR stuff at times. But yea, specifying the max target nits is SUPPOSED to change the metadata. Now the reason I was confused there is because we had very broken nVidia drivers for so long that it was always sending 1000 no matter what you set. Which on our OLED's would obviously tone map regardless of what you do. That's why your comment and graph raised an eyebrow cause I was like "waiiitttttt a minute that's not how I remember this working" LOL

[huhn]

who knowns what madVR is doing in term of HDR right now just wait for the next release and ask.

he changed to much to make this clear.

[SamuriHL]

I know. Hence my comment earlier that when he returns we should probably ask and get clarification. Right now I'm super happy with the latest beta build, the measurement files, and sending ~750 max target nits to the display. Now that LG finally fixed my C8 with the latest firmware, the result is freaking fantastic. Zero complaints here.

[j82k]

Now the reason I was confused there is because we had very broken nVidia drivers for so long that it was always sending 1000 no matter what you set. Which on our OLED's would obviously tone map regardless of what you do. That's why your comment and graph raised an eyebrow cause I was like "waiiitttttt a minute that's not how I remember this working" LOL

But the metadata only determines how the TV tone maps the HDR image and since there is no set standard for this, MadVR can't know what the TV will do with that. So the ideal case would be if the TV didn't do any tone mapping at all and would just follow the EOTF and clip at peak brightness so that MadVR is in full control of the tone mapping.

So the broken 1000 nits metadata wouldn't even matter when you use MadVRs tone mapping as with that metadata the TV will just clip at peak brightness like in my gif.

[Warner306]

This measurement was done using the technicolor preset at default settings. With it's custom white point it is quite a bit brighter than the more accurate warm2.
When I adjust color temperature to D65 the TV follows the EOTF more closely but also has a lower peak brightness.

As far as I know the C8s HDR EOTF tracking is designed for 700 nits peak so if the TV gets brighter than that it will go above the EOTF but you can adjust it using the 20 Point WB controls.

How do you create that graph? Is it not possible to do the same thing with pixel shader tone mapping enabled with HDR output checked?

madVR should set the MaxCLL and/or mastering display maximum luminance to the real display peak nits entered in madVR. Otherwise, the display won't be aware of the brightness of the source video after it has been rescaled. Every HDR video is tone mapped to the target display brightness you enter in madVR (~700 nits).

[SamuriHL]

I suppose, in theory, that's true. And yes we don't definitely know what the TV is doing with the tone mapped image from madvr. Madshi has recommended that we keep dynamic tone mapping enabled on the TV in hopes that it notices that it's already tone mapped. I'm not, however, sure I agree that the broken metadata didn't matter, though. It definitely seemed to make a difference in my testing.

[Warner306]

Dynamic tone mapping should ignore the metadata altogether and simply calculate the brightness of each incoming frame from the RGB values sent by madVR in real-time to determine how to tone map each scene.

According to j82k's graphs, you should get the most faithful representation of madVR's HDR to HDR tone mapping by disabling dynamic tone mapping in favor of using a static tone curve. Because the display chooses to follow the PQ curve 1:1 when it reads ~1,000 nit metadata, when it receives the RGB values from madVR, it should convert each RGB color to an equivalent PQ RGB value in the display panel without compressing any of them. If the RGB values are instead converted to YCbCr 4:2:2, the display would simply convert them back to the same PQ RGB value after it has processed the image. This means the input and displayed RGB values should stay the same, outside of consideration for very bright scenes that activate the display's ABL.

If there was additional compression or tone mapping at any point, the input RGB values would be converted to YCbCr and the display would compress some of the brighter pixels (compress Y) and apply additional desaturation to some of the brightest colors.

The static tone mapping in the C8 appears to display the SDR range (0-100 nits) 1:1 with the PQ EOTF and rolls off the remaining specular highlights later in the display curve. This would keep the image bright and accurate, but would most often clip some of the brightest scenes that have highlights near the video MaxCLL.

The dynamic tone mapping in the C8 and C9, on the other hand, is known to tone map the midtones of the SDR range at times, lowering the midtones for very bright scenes to preserve even more specular highlight detail and improve contrast and possibly even raising the midtones in other scenes to artificially brighten the picture. Outside of the brightest scenes, the midtones should stay neutral, so I don't know if the LG's are actually artificially boosting these scenes or not. If LG is always manipulating the midtones, I don't know if madVR's tone mapping can override this behavior, outside of improving specular highlight detail in some cases.

The tone curve used by madVR is BT.2390. BT.2390 avoids tone mapping the midtones completely if the display brightness is at least 480 nits and only rolls off the specular highlights. Compared to LG's existing static tone curve, this difference may only be visible in very bright scenes, where BT.2390 should retain more specular highlight detail and make the regions of the picture with highlights appear slightly dimmer from the more aggressive tone curve. BT.2390 recommends avoiding any tone mapping of the 0-100 nits SDR range, unless the display brightness is very limited. LG's dynamic tone mapping is not incorrect, but it chooses to tone map the SDR range (the midtones) in some scenes to better represent the source's contrast.

[SamuriHL]

Thank you for that very detailed explanation! That explains a LOT of what I've seen in testing over the last year. Indeed, dynamic tone mapping on the LG does brighten certain scenes as compared to having it turned off. I always thought this was "normal" and desired, but, if they are doing that in the SDR range then that's not desired at all given that the LG has more overall range in HDR than a projector. Therefore no need to compress the SDR range. It's been a while since I played with turning dynamic tone mapping off....maybe even before the nVidia driver fixes, and most certainly before I got my RTX and the latest LG firmware update. Hmmmmmmm. I think I'm going to have to do another round of testing!! This is potentially exciting. :)

[Warner306]

For what it's worth, Stacey Spears says he prefers to use the LG's dynamic tone mapping. I don't know what his reason is.

It would be nice if j82k could produce a graph of the tone curve produced by madVR to confirm it is working correctly and see what it looks like. Tone mapping to SDR gamma and PQ are completely different things and I don't know how much HDR output has been tested compared to SDR.

[SamuriHL]

Not a lot, sadly. I'm one of the few who seem to prefer to use madvr in that manner. And I really do prefer to use madvr in that manner. :D I would also love to see a curve if it's even possible to do so with madvr. That would be awesome cause it would give us more concrete data rather than "yea, it seems to look better or worse". And anecdotally, which of course I've only had a few minutes to test, I can see your point about disabling dynamic tone mapping. But, the other massive point here is that LG has finally fixed the freaking black crush so my detail is back. It's going to require some extensive testing but so far I like what i see with the LG disabled dynamic tone mapping, madvr set to 700 nits, are you nuts highlight recovery (cause, you know, I like the highlights :D), balanced fire, and obviously output in hdr set. I'm also using measurement file. Like I said, FAR more testing needed, probably Labor Day weekend I'll do some marathon viewing, but, this looks very promising so far!

[j82k]

It would be nice if j82k could produce a graph of the tone curve produced by madVR to confirm it is working correctly and see what it looks like.

MadVR's passthrough tone mapping doesn't seem to work with madTPG.

When I put the target peak nits to some low value like 400 nits I can see how the highlights get darker in movies, so madVR is tone mapping the picture.
But when measuring with madTPG it doesn't have any effect and will still measure up to 800+ nits.

[mini-moose]

Been trying to get madvr working on my setup for 4K HDR, but can't get it working without frame drops. Tried endless settings.

Granted, maybe my rig is too basic/old: i7-950 with 8gb ram (unused pc I decided to dedicate to it), GT 1030, LG OLED B6 (HDMI mode).
PC connected straight to TV with HDMI. Windows 10 x64 1903.

MPC-HC + madvr : almost works but occasionally stutters and drops frames.
DSPlayer+madvr: never even close to being smooth playback, choppy and drops frames like mad.

Plex and Emby seem to work well as far as I can tell.

Any suggestions would be highly appreciated!

[SamuriHL]

The 1030 is just too slow. You really need a 6gb 1060 minimum to drive uhd playback in madvr.

Sent from my SM-G975U using Tapatalk

[QBhd]

Yeah that's a pretty low end card, I was going to say that if my 2GB 270X can do it, anything can... but I just looked at a few comparisons of the two and the 1030 is seriously limited in what it can do

QB

[mini-moose]

Might be, yeah. I have seen several posts from people who do use it and it works ok.

[SamuriHL]

It depends on many things. Doing straight passthrough for HDR on 23.976fps movies with something like bicubic chroma and potentially a few trade performance options with non error diffusion dithering.... yeah it'll work. Maybe. There's a lot of factors that go into it.

Sent from my SM-G975U using Tapatalk

[mini-moose]

It depends on many things. Doing straight passthrough for HDR on 23.976fps movies with something like bicubic chroma and potentially a few trade performance options with non error diffusion dithering.... yeah it'll work. Maybe. There's a lot of factors that go into it.

Yeah, I tried to go down that route but still failing.

I wanted to use madvr for DSPlayer as I'm used to Kodi's interface and that's completely unusable.
No matter what I try, that won't even play somewhat smooth.
I always assumed DSP is using MPC in the background, so was very surprised by the big difference in performance.

I guess I'll have to use plex or emby for now.

[Warner306]

If you do a search for GT 1030 settings in the DSPlayer thread at the Kodi Windows support forum, you should be able to find stable settings that work with that GPU. I use DSPlayer regularly and the performance is identical to MPC. You need to lower the size of all queues and set image upscaling to DXVA2. Pixel shader tone mapping isn't possible.

[Warner306]

MadVR's passthrough tone mapping doesn't seem to work with madTPG.

When I put the target peak nits to some low value like 400 nits I can see how the highlights get darker in movies, so madVR is tone mapping the picture.
But when measuring with madTPG it doesn't have any effect and will still measure up to 800+ nits.

That's unfortunate. Can you post a graph showing where the roll-off starts when the C8 is calibrated to D6500 and 700 nits?

[j82k]

Well, this is the preset I currently use for watching movies. It's calibrated to D65 and tops out at 701 nits. I didn't do any adjustments for the highlights, only some changes in the SDR range and 2 Point High WB adjustments and it follows the EOTF much better than the brighter technicolor preset.
This was measured with 1000 nits metadata. I didn't do the metadata comparison for this one but I guess the roll-off would change similar like it did with my previous posted graph.

https://i.imgur.com/Tu14eSj.jpg

[Warner306]

That is an accurate curve, but it must clip some highlight detail. You prefer the extra brightness at the top end?

[SamuriHL]

I do. [emoji846] it's a fantastic experience.

Sent from my SM-G975U using Tapatalk

[j82k]

As a dark room viewer I'm usually much more concerned about the low-end than the highlights. I rarely see clipped highlights but I'm probably not really sensitive to it. I could get higher D65 calibrated peak brightness by increasing 2P high green instead of lowering red and blue but then the highlights would go above the EOTF and that would need major 20 Point adjustments and I'm not a brightness junkie so I didn't even bother to try that.


Regarding the metadata discussion when using MadVR tone mapping:

I just remembered that the C8 shows different 20 Point WB adjustments points when it receives 1000/4000 nits metadata. And when playing a 4000 nits mastered movie with MadVR tone mapping enabled it still shows the 4000 nits adjustment points. This means that MadVR does not change the metadata just like I previously suspected. So when playing a 4000 nits movie and using MadVR tone mapping you'll basically get double tone mapping and the highlights will be darker than they should because the TV uses a more aggressive roll-off which MadVR doesn't know about. I'm still of the opinion that using 1000 nits fake metadata is the way to go when using MadVR's tone mapping for watching 4000 nits mastered movies.

[mini-moose]

If you do a search for GT 1030 settings in the DSPlayer thread at the Kodi Windows support forum, you should be able to find stable settings that work with that GPU. I use DSPlayer regularly and the performance is identical to MPC. You need to lower the size of all queues and set image upscaling to DXVA2. Pixel shader tone mapping isn't possible.

I'm not sure where lies the issue. The only thing I can think of is that I have also a normal kodi 17.6 installed on the same system, maybe they are clashing in some way. I didn't remove the other one as I couldn't even get MPC to play properly. I don't need need upscaling as this is only for 4K HDR playback.
I will try and look into the forum again. thanks.

[SamuriHL]

Wouldn't it be even better to have madvr send the proper modified metadata? That way it should stay under the max target nit and not trigger the TV's tone mapping. Right?

Sent from my SM-G975U using Tapatalk

[Warner306]

I just remembered that the C8 shows different 20 Point WB adjustments points when it receives 1000/4000 nits metadata. And when playing a 4000 nits mastered movie with MadVR tone mapping enabled it still shows the 4000 nits adjustment points. This means that MadVR does not change the metadata just like I previously suspected. So when playing a 4000 nits movie and using MadVR tone mapping you'll basically get double tone mapping and the highlights will be darker than they should because the TV uses a more aggressive roll-off which MadVR doesn't know about.

If that's true, I would consider HDR to HDR to be broken with static tone mapping. I have heard reports that it was clipping and doing strange things, so I did wonder. Are you sure you are using a compatible driver and one of the most recent beta builds?

You should find a way to gather some tangible proof and post in the main forum for madshi to look at it. Without the metadata, the display won't know how bright the source is. That kind of defeats the purpose of tone mapping the source before it reaches the display and ruins madVR's dynamic tone mapping if the display thinks the source is brighter than it actually is because you won't save any brightness at all. The metadata is the only data the display has to use to determine which display curve to use.

Maybe you need to reach out to Manni to see if he can test this by capturing the metadata at the other end with his HDFury Vertex. This has been in use for a long time, so I don't know why there haven't been more reports of issues.

[chros]

I just remembered that the C8 shows different 20 Point WB adjustments points when it receives 1000/4000 nits metadata. And when playing a 4000 nits mastered movie with MadVR tone mapping enabled it still shows the 4000 nits adjustment points. This means that MadVR does not change the metadata just like I previously suspected.
Amazing find, @jk82, as always! I'll take a look at my B8 (maybe during the weekend).

That way it should stay under the max target nit and not trigger the TV's tone mapping.
But no matter what you do the TV's tonemapping will always work! (since you can't turn it off)
That's the point you guys (not just you) don't want to understand :)
That's why I use passthrough: because I don't see the point of doing something that can't be:
- repeated
- verified
- etc.

And the same applies to following the "right" EOTF curve, again, I don't see the point of talking about it, until our (literally) precious displays:
"have at least 3 types of ABL: ASBL, ABL (can't be turned off) and ANSD (can't be turned off)"
What's the point of talking about theory while in reality no one knows how these bloody displays behave :D

If we had a perfect display ... but we don't, far from it :)

[SamuriHL]

I never said you could "turn it off". What I said was that if you tone map to 700 nits and send the proper metadata with it, it SHOULDN'T trip the TV's tone mapping because it's already under the max nit level that the TV can display. That's not even close to the same thing as turning it off. I'd really like to see what Manni's HD Fury shows about the metadata that's being sent. I'm with Warner on this...if it's not changing the metadata, then it's highly broken. But since so few of us use it with HDR output, I'd not really be surprised to find out that's the case. I don't have any way of measuring it so I've no idea one way or the other.

[chros]

I never said you could "turn it off". What I said was that if you tone map to 700 nits and send the proper metadata with it, it SHOULDN'T trip the TV's tone mapping because it's already under the max nit level that the TV can display. That's not even close to the same thing as turning it off.
:) I know what you meant, but it's only a theory.
If you can't reproduce/test something in a reliable way then the "shouldn't" part stay as "shouldn't" but we will never know :)
Not to mention, if you turn on DTM (that you also did/do, not just me) then it will definitely touch everything :)

[SamuriHL]

Yes, Dynamic tone mapping very likely touches everything regardless of what the metadata says given how it works. madshi's hope for that was that it would measure the frame and minimally screw around with it if at all. However, I have noticed the changes in the SDR range before and thought those were "normal". They aren't (and I suppose it comes down to preference of whether you care that the SDR range is being messed with or not). A few days ago when we started this discussion I turned Dynamic tone mapping off and did some spot checking on a few of my favorite scenes and can definitely see the difference. That being said, if the metadata really isn't being sent as modified based on what you specify for max target nit (and especially if you're using a measurement file since that would be a more accurate representation of each frame than the real time algorithm) then it defeats the purpose of tone mapping with madvr pixel shader using HDR output and would be considered a bug. I don't have any way of measuring it, and it's only fairly recently that we even had drivers that weren't broken and sending 1000 nits. Hence why I'd like to see if Manni or someone else with the right equipment can verify what it's doing. Yes, I agree, that still doesn't mean we know what the TV is doing with its tone mapping, but, I'd say that if we can tone map to 700 nits and send the modified metadata that gives us the best chance at not doing double processing. Not guaranteed, I get it.

[Warner306]

Amazing find, @jk82, as always! I'll take a look at my B8 (maybe during the weekend).


But no matter what you do the TV's tonemapping will always work! (since you can't turn it off)

If you look at j82k's graphs, the C8 does appear to disable the display's tone mapping when 1,000 nit metadata is sent. The display follows the original EOTF curve all the way to the display peak and clips. That is considered a PQ EOTF response without any tone curve up to 800 nits. At least that is the case with the picture preset he had selected.

His second graph shows a very late roll-off at 700 nits. I don't think you can control the shape of the display's roll-off; that is likely fixed. But you can change where the roll-off point starts by manipulating the contrast and/or brightness of the display panel with the contrast and grayscale controls. I noticed he customized the EOTF a bit by moving reference white down a few nits.


And the same applies to following the "right" EOTF curve, again, I don't see the point of talking about it, until our (literally) precious displays:
"have at least 3 types of ABL: ASBL, ABL (can't be turned off) and ANSD (can't be turned off)"
What's the point of talking about theory while in reality no one knows how these bloody displays behave :D

If we had a perfect display ... but we don't, far from it :)

There is no standard for tone mapping, but almost all current HDR displays start the roll-off point at some point after 100 nits. More than 90% of an HDR image is within this range, as proven by numerous measurement files produced by madMeasureHDR.

I honestly think Panasonic is the best at this. Both its OLEDs and HDR Optimiser in its Blu-ray players do a good job of tone mapping the specular highlights alone without impacting the overall APL of the image.

The typical Panasonic tone curve for 4,000 nits+ sources looks similar to this:

http://upload.vstanced.com/images/2019/06/11/c64b7d16d0102875afa91a5a038f0808.png

The roll-off appears to start at around 200 nits and peaks at just under 700 nits.

As I said before, BT.2390 also recommends limiting tone mapping of HDR10 content to the specular highlights alone if the display has at least 500 nits of peak brightness. This is not best practice, but it is becoming closer to a consensus with display manufacturers and calibrators. Most people don't seem to notice clipped highlights as much as a loss of HDR brightness.

[Warner306]

One way to test if HDR to HDR is working correctly is to use a 1,000 nits source and set the real display peak nits in madVR to 1,000 nits. This should send the source through without any tone mapping as if it was a standard 1,000 nits source, save any scenes that may peak above 1,000 nits.

Then find a scene that is somewhere between 800-1,000 nits and see if there is any difference between straight passthrough and pixel shaders with HDR output checked. The brightness of the image should be identical in both cases because madVR should be reporting 1,000 nits to the display and not doing any tone mapping because the display is assumed to already be the same brightness as the source video. You could even create two profile groups and assign them keyboard shortcuts to quickly switch between the two.

Otherwise, you need to use some good test scenes with bright specular highlights and play these scenes in slow motion with MPC. Or use a camera with a high dynamic range and take some screenshots of the same frame (Ctrl + G in MPC). The screenshots would be even more useful to show any differences. Pixel shaders with HDR output should show a smoother transition into the highlights with less clipping compared to the LG's static tone curve without lowering the APL of the SDR range.

The dream sequence in Batman v Superman is good for this. There are bolts of electricity that usually clip with most displays and Ben Affleck's white shirt will lose all detail if any clipping is happening. If the display applies heavy tone mapping to the SDR range, Ben Affleck's face will appear more tanned. Max Max Fury Road also has some good test scenes involving fire, explosions and some lightning bolts that reach 10,000 nits.

The only example that I know of that explains how external tone mapping for HDR TVs works is the HDR Optimiser from the Panasonic UB9000/UB820 Blu-ray players. If you watch this video, you'll notice that the sending the correct metadata is critical to its usefulness:

https://www.youtube.com/watch?v=oTw_Toh0PzA

The first two examples use 4,000 nits sources where the original display curve is either tone mapping the SDR range too much and unnecessarily dimming the APL and/or not tone mapping enough and clipping the highlights. The Optimiser compresses the source to 1,000 nits, not allowing the display to select its 4,000 nits tone curve, and it does a better job of preserving the specular highlights with the correct hue without tone mapping the SDR range. The curve used by the HDR Optimiser is very similar to the one I posted in the last post and starts above 100 nits.

[j82k]

I'm currently not using a test build, probably should have tried one but as far as I know most of the work done in all the test builds have been about 'HDR to SDR' and not 'HDR passthrough' tone mapping. Actually I wanted to let the whole tone mapping thing rest and just use regular passthrough until a new final version of madVR releases.

Through the adjustment points check method on the C8 you can tell if the TV receives 4000 nits MDL metadata (second last adjustment point 713) or lower metadata (second last adjustment point 696).
What I can't know is what maxCLL madVR is sending so it might or might not modify that. I guess unless someone tests this with a device like the HD Fury we won't really know.


One other thing that might affect the tone mapping is that the points at which the C8 clips when using colors is different from white. I think red is the worst where it clips at like 540 or so...
I didn't measure this, only checked this on clipping test pattern.