Dolby Vision now possible through MP4 Mux.

[quietvoid]

There's a big different too when l2 is present but I think it's not as obvious as in the L1(no L2) scene. or maybe im dreaming
with the internal player, the L1 (no L2) scene is a lot brighter and more clipped than the x800m2. Roughly the same difference as in your first 1000nits pattern.

Okay, I don't really understand what the issue is anymore.
I thought there was little visible change on the Blu-ray player. Now it sounds more like the TV is just brighter.

I can't easily judge the brightness by eye, but the C2 does clip more than the FireTV Stick 4K Max for me.

The stick shows more blue at the edges of the water splash, while the TV internal player is almost just white.
So the TV is probably brighter too.

To me this isn't a significant issue, as I've always assumed the internal player to be brighter and "more accurate".

[RESET_9999]

could you re-do the first 1000nits pattern with only one 600nits trim set to 2048 for the L1 scenes?
maybe it just doesnt work properly when L1 is alone (which is very rare in UHD-BD)
edit: scratch that, i just checked your file and it has l2

[RESET_9999]

There's a big different too when l2 is present but I think it's not as obvious as in the L1(no L2) scene. or maybe im dreaming
with the internal player, the L1 (no L2) scene is a lot brighter and more clipped than the x800m2. Roughly the same difference as in your first 1000nits pattern.

Okay, I don't really understand what the issue is anymore.
I thought there was little visible change on the Blu-ray player. Now it sounds more like the TV is just brighter.

I can't easily judge the brightness by eye, but the C2 does clip more than the FireTV Stick 4K Max for me.

The stick shows more blue at the edges of the water splash, while the TV internal player is almost just white.
So the TV is probably brighter too.

To me this isn't a significant issue, as I've always assumed the internal player to be brighter and "more accurate".

I dont think we can say that one is better than the other with your patterns because you're just sending extreme value to some random 10 000 or 1000 image.
to me, this just shows that there's a difference and that one is incorrect. The real test is with real content, or madvr pixel measurements.
1- the shield behave exactly the same as the internal tv
2- same player same file behave the same on the C8
3-The CX is not affected.
it just doesnt add up

[chros]

I’m not sure I’m experiencing this bug. LG CX with Sony X700 and Shield Pro 2019. Testing the videos back and forth (ts on Sony and mp4 on Shield) I could easily see the changes on both. The shield might look slightly better but I easily see them as well on the x700.

yep, CX looks ok. lucky you

please try the cmv4.0 version on your Shield(or TV internal player) and tell us if you get L8 reaction.
https://drive.google.com/file/d/1tU3VLC ... sp=sharing

Both the Shield and TV show strong reaction in the L8 in the trims, chroma weight, saturation, clip trim, I don’t really notice anything happening in the vector fields.

Thanks for checking, so the CX is not affected that also has cmv4.0!

interestingly and kind of a good news, @manix said he got an L8 response on his CX and no L1 bug on the x700... mmm
so there's something going on with the C1/C2

That's a good news indeed! Meaning they (probably Dolby) screwed up something in the DoVi engine on C1/C2!

I'll do some tests with real movies.

That would be really useful, if you can provide a sample of the real bdmv (we can cut streams with CloneBD).

There's a big different too when l2 is present but I think it's not as obvious as in the L1(no L2) scene. or maybe im dreaming
with the internal player, the L1 (no L2) scene is a lot brighter and more clipped than the x800m2. Roughly the same difference as in your first 1000nits pattern.

Okay, I don't really understand what the issue is anymore.
I thought there was little visible change on the Blu-ray player. Now it sounds more like the TV is just brighter.

Checking other files (haven't checked the new test files yet), it seemed to me that the image is duller (less vibrant) via Oppo, compared to internal Plex client.

1- the shield behave exactly the same as the internal tv
2- same player same file behave the same on the C8
3-The CX is not affected.
it just doesnt add up

Yes, I think it's a DoVi engine bug only on C1/C2 (not CX) with cmv2.9 devices, simple as that
Btw, is profile 5 content is also affected? (not that I want to use them on the Oppo, just curious) I'll check again the lelabodejay samples...

[quietvoid]

Yes, I think it's a DoVi engine bug only on C1/C2 (not CX) with cmv2.9 devices, simple as that

Makes sense, I guess the firesticks are affected too. I just don't have another device to compare against to know for sure.

[RESET_9999]

I checked real content(3 movies) and I couldn't really spot a difference in the brightness changes. (i wonder if the difference could be L4 ?)
There's only a big difference when I generate DV from HDR10 or HDR10plus and with quietvoid's pattern so I don't know what's going on but i still think there's something wrong (real content or not).
I also noticed that the cmv2.9 and cmv4.0 have different L1 metadata. cmv4.0 clip a lot more in the "high contrast" scene on both the C2/C8 but on the x800m2 there's no difference.
Today I'll put back side by side my C8/C2 and make a more direct comparison.

the 3 movies test I did were profile 8 and one of them didn't have any L2 (ocean's eight). I did the test without splitting the file because I wanted to make sure that nothing was broken so I can't share samples.
Anyone can do the same test, just extract the RPU and delay it by 30 frames. It can be done in P8 or P7.

extract and convert to P8:

Code: Select all

dovi_tool.exe  -m 2 extract-rpu input.hevc 

or
extract untouched P7:

Code: Select all

dovi_tool.exe  -m 0 extract-rpu input.hevc 

edit rpu:

Code: Select all

dovi_tool.exe editor -i rpu.input.bin -j duplicate.json --rpu-out rpu_delayed.bin

duplicate json:

Code: Select all

{
	"duplicate": [
		{
			"source": 1,
			"offset": 1,
			"length": 30
		}
	]
}

inject:

Code: Select all

dovi_tool.exe  inject-rpu -i video_input.hevc --rpu-in rpu_delayed.bin

[RESET_9999]

Btw, is profile 5 content is also affected? (not that I want to use them on the Oppo, just curious) I'll check again the lelabodejay samples...

because of the dynamic color reshaping in P5, it can be hard to delay the rpu but I compared lelabodejay pattern again last night and didn't really see any difference.

same with the fixed luma version P5 of the "mutli_pattern" from original spears and munsil
https://drive.google.com/file/d/1GRWWow ... sp=sharing
I checked HDR10 too and it clips a lot more than DV and maybe there's something going on with the ST2084 tracking in HDR10?

[chros]

Yes, I think it's a DoVi engine bug only on C1/C2 (not CX) with cmv2.9 devices, simple as that

Makes sense, I guess the firesticks are affected too. I just don't have another device to compare against to know for sure.

You remember when I had the B8 and Firestick 4k (old one) and I told you guys that the resulting image was different compared to the Oppo/Plex, it was darker on the Firestick somehow.
Is the Max version cmv4.0? In other words is your cmv4.0 sample has any effect on it?

So, I've been thinking about this more, what we know is:
- L1 data is arrived to the TV: there's a slight change in the image
- this can mean what I thought is: DoVi engine deals with the old devices differently
- or: L1 data gets truncated somehow: but this scenario is very unlikely, since L2 is not?

I checked real content(3 movies) and I couldn't really spot a difference in the brightness changes. (i wonder if the difference could be L4 ?)

That's not a good news What do you mean about L4?

Today I'll put back side by side my C8/C2 and make a more direct comparison.

Cheers, that would be really helpful! We only have to find 1 example with true content where the issue is clearly present, after we can think it over by examining the RPU itslef.

Btw, is profile 5 content is also affected? (not that I want to use them on the Oppo, just curious) I'll check again the lelabodejay samples...

because of the dynamic color reshaping in P5, it can be hard to delay the rpu but I compared lelabodejay pattern again last night and didn't really see any difference.

same with the fixed luma version P5 of the "mutli_pattern" from original spears and munsil
https://drive.google.com/file/d/1GRWWow ... sp=sharing
I checked HDR10 too and it clips a lot more than DV and maybe there's something going on with the ST2084 tracking in HDR10?

Yes, I don't recall much difference with lelabodejay pattern either, that's why I asked about it. I'll try to recheck it soon. I'll also check that pattern.
Not sure about the HDR10 version (I don't have it), what is the metadata for it? (mdl, maxcll, etc)

There's only a big difference when I generate DV from HDR10 or HDR10plus and with quietvoid's pattern so I don't know what's going on

There's a only one more possibility in my book: that "only" profile 8.1 content is affected somehow
We will know this for sure when/if we find a true example.
The hunt is on!

[quietvoid]

Is the Max version cmv4.0? In other words is your cmv4.0 sample has any effect on it?

No, it only reacts to L2.

[chros]

Is the Max version cmv4.0? In other words is your cmv4.0 sample has any effect on it?

No, it only reacts to L2.

Nice, thanks, didn't know that! So at least it's similar than the rest of the problematic devices.

[RESET_9999]

That's not a good news What do you mean about L4?

@quietvoid, (i think) once said that it's supposed to help with the transition between shots. The pattern doesnt have L4, the madvr/hdr10plus generated DV too.

Not sure about the HDR10 version (I don't have it), what is the metadata for it? (mdl, maxcll, etc)

The HDR10 version I'm talking about is the original BL from SM. The one with strange LUMA that affects DV on the shield and our 2022 TV but strangely it doesnt affect HDR10.
it's encoded at 1000/400 but it's actually 10 000nits.

There's a only one more possibility in my book: that "only" profile 8.1 content is affected somehow

can't be, because the 3 movies I tested last night were all converted to P8.1 and didn't have any noticeable difference. I didn't test P7.

Also, I just learned that the C9 is supposed to have the newer cmv4.0 Dolby engine. So @deadphil, you should check it out.

We will know this for sure when/if we find a true example.

well, even without a real content example, I don't feel safe watching movies on my x800m2 anymore.
Patterns are not movies but they are always a good indication that something is going on, extremely (or barely) noticeable, it doesnt matter to me.

[chros]

Not sure about the HDR10 version (I don't have it), what is the metadata for it? (mdl, maxcll, etc)

The HDR10 version I'm talking about is the original BL from SM. The one with strange LUMA that affects DV on the shield and our 2022 TV but strangely it doesnt affect HDR10.
it's encoded at 1000/400 but it's actually 10 000nits.

So, another messed up encode from that disc

There's a only one more possibility in my book: that "only" profile 8.1 content is affected somehow

can't be, because the 3 movies I tested last night were all converted to P8.1 and didn't have any noticeable difference. I didn't test P7.

Hmm, indeed.

Also, I just learned that the C9 is supposed to have the newer cmv4.0 Dolby engine. So @deadphil, you should check it out.

I bet it doesn't have the issue, unless the DoVi engine was updated recently in the C9 firmware. Let's see...

We will know this for sure when/if we find a true example.

well, even without a real content example, I don't feel safe watching movies on my x800m2 anymore.
Patterns are not movies but they are always a good indication that something is going on, extremely (or barely) noticeable, it doesnt matter to me.

I agree with you completely!

[ragico]

Are these people (Dolby, Sony, Lg etc) joking with us?
This behaviour will discourage a lot of people.

[chros]

Here's a higher brightness L1 test video: https://0x0.st/ojA5.mkv
The image is a lot brighter than what the OLEDs are capable of, and the metadata reflects that.

So we have MDL at min 0.0050, max 4000 nits.
MaxCLL 5640 nits.

This works well on my FireTV Stick 4K Max + LG C2, every change in L1 has an effect similar to the internal player.
Let me know if it makes any difference for UHD BD players.

yep as soon as L1 is alone, big difference.

https://slow.pics/c/uF3dr3YN

DSLR camera settings locked

So it doesn't react when there's L2 present only?
The TV only looks dimmer in the photos, but the clipping is the same..

That's weird. The L2 is all set to 2048, though. I think usually real metadata has one of them at like 2049.
I doubt it would work, though.

Another thing might be that there's L2 up to 4000 nits, so maybe that does something too.

So it doesn't react when there's L2 present only?
The TV only looks dimmer in the photos, but the clipping is the same..

There's a big different too when l2 is present but I think it's not as obvious as in the L1(no L2) scene. or maybe im dreaming
with the internal player, the L1 (no L2) scene is a lot brighter and more clipped than the x800m2. Roughly the same difference as in your first 1000nits pattern.

I just checked this high brightness sample: all the trims are changing on the Oppo as well (!), just as you said, but as Reset said it looks different compared to internal Plex, not sure which is right ...

The previous test was at max 1000 nits, so it's possible some bypassing might happen if the player looks at the EDID, or the TV looks at some other metadata.
I'll have to test modifying the L6 metadata to see if it makes any difference.

In the end, it seems that HDMI input players may just be tracking PQ when *some* metadata is within the capabilities of the display, instead of clipping according to L1 like the internal player.

Now that's an interesting theory! But why the Shield is different?

[chros]

same with the fixed luma version P5 of the "mutli_pattern" from original spears and munsil
https://drive.google.com/file/d/1GRWWow ... sp=sharing

I checked this P5 one as well: bit more clipping on the Oppo and noticeably darker image! (Again, not sure which is right)