HDR heatmaps and torch mode UHDs

[Amano]

Just to make sure: Blade Runner 2049 was the US Warner Disc and not the Sony one?

[Geoff D]

Originally Posted by Amano

Just to make sure: Blade Runner 2049 was the US Lionsgate Disc and not the Sony one?

Warners released it in the US, but FWIW I reckon that any objective analysis of both discs will reveal that the grading is all but identical despite the discrepancies between the metadata on the two.

[Pyoko]

Originally Posted by Geoff D

Re: purple Joi, those kinds of magenta/purple hues are one area of the gamut where there's very little expansion when moving from 709 to P3 (and for blue there's practically none at all), so I'm not surprised that there's so little WCG there.

I guess one could have suspected it with 709 and P3 sharing the same blue primary. Also the analysis and the gamuts in the graphic are based on the CIE xy coordinates, which are not perceptually uniform. If comparing the gamuts on the perceptually uniform CIE u'v' diagram it looks like the greatest expansion of 2020 compared to 709 is towards the red primary and not the green one, and while blue still offers the smallest difference it's not quite as small as it looks like on CIE xy.

Originally Posted by Amano

Just to make sure: Blade Runner 2049 was the US Warner Disc and not the Sony one?

It's the UK Sony one, but with Deakins at the helm I can't really imagine the US one being much different.

[Pyoko]

It's one of the best film transfers on UHD with superb grain that looks completely untouched, but it's often way too bright. Highlights look great compared to the badly clipped BD though. Apart from the one shot I happened upon, very little use of the wider color gamut throughout.










Album

With a similar highlight roll-off the HDR version would on average be equivalent to the SDR version with peak white at 238 nits (Gamma at 2.40), based on 22 samples. (Min: 62, Max: 671, Median: 208)

[Geoff D]

God I'd love to see T2's UHD analysed like this.

[andreasy969]

I love those almost completely black and white gamut heatmaps.

[Amano]

The very high B/W presence shows that rec704 is already very good for real world photography and P3 is probably that what most people will ever need.

It is interesting to see that the greens are not extended at all in those shots. Even if the greens would be the most obvious use case for using rec2020 over P3 and P3 over rec704 - there is the biggest delta. But that might be due to the movie choice (desert/SF). At least Blade Runner '49, Mummy and Shape of Water seem to be extended mostly exclusively towards the yellow/red edge of the triangle.

I wonder how the Martian would fare? If the Mars photography was graded consciously outside of the "normal" rec704 spectrum for an "otherworldly" look? The BR49 desert shot seems to be made that way with nearly all the colours taken from the outer edge of the P3 triangle. That was obviously mastered in P3 and not in rec2020 and the cutoff line shows that. Is there any movie around with exclusive rec2020 colours? I guess not.

[Schwartzy]

King kong 2005 would be nice to see

[Geoff D]

Originally Posted by Amano

The very high B/W presence shows that rec704 is already very good for real world photography and P3 is probably that what most people will ever need.

It is interesting to see that the greens are seldomly extended because the greens would be the most obvious use case for using rec2020 over P3. But that might be due to the movie choice (desert/SF). At least Blade Runner '49, Mummy and Shape of Water seem to be extended mostly exclusively towards the yellow/red edge of the triangle.

I wonder how the Martian would fare? If the Mars photography was graded consciously outside of the "normal" rec704 spectrum for an "otherworldly" look? The BR49 desert shot seems to be made that way with nearly all the colours taken from the outer edge of the P3 triangle. That was obviously mastered in P3 and not in rec2020 and the cutoff line shows that. Is there any movie around with exclusive rec2020 colours? I guess not.

2020 colour is a unicorn anyway. There aren't even many professional monitors that can show that gamut, never mind consumer displays. There's an element of future proofing there, sure, but we're likely to get 4000 nit consumer displays well before we get 100% (or less) 2020 coverage.

Anyhoo, Vincent Teoh did some tests with the 'out of gamut' marker on the P3 Sony OLED mastering monitor, most 'real world' content with its conscious stylisation wouldn't fall outside of P3 or maybe even 709 but nature documentaries - shot to 'pop' as they do - showed some out of P3 colour and computer animated stuff - where they can program it to look however they like - like Spider-Verse and Lego Batman were constantly showing as being out of P3 range.

So I vote for Spider-Verse or Lego Batman next please Pyoko!

[Samcor_NZ]

I'd be interested to see the breakdown of 'A Few Good Men'.

[Fjodor2000]

I know there are discussions regarding whether OLED is bright enough to handle HDR without clipping, or if LCD has an advantage in this regard.

LCD has a disadvantage in that cannot handle brightness on a pixel level basis as OLED can.

OLED on the other hand has a disadvantage in that it cannot reach the same max brightness as LCD. But based on these heatmaps, is this a problem? For movies that are not unnaturally bright encodes, does OLED ever clip the brightness?

E.g the LG OLED CX has these max brightness levels in nits per percentage of screen area covered:
ANSI: 744
1%: 739
5%: 744
10%: 742
25%: 395
50% 217

Also, HDR10 has max 1000 nits brightness, HDR10+ 4000, and Dolby Vision 10.000. But are there any movies actually exceeding 1000 nits max brightness? Will it never occur for older movies that 1000 nits is exceeded, but might be possible to exceed for newer movies? If so, when is the breaking point roughly, e.g. year ~2010, ~2020, ...?

[Pyoko]

Originally Posted by andreasy969

I love those almost completely black and white gamut heatmaps.

I kind of want Sin City on UHD just to see if running it through the gamut analysis would result in the same image as the original input.

[Agent Kay]

Originally Posted by Pyoko

I kind of want Sin City on UHD just to see if running it through the gamut analysis would result in the same image as the original input.

Dame would be nuts on the whites I bet

[BrownianMotion]

Originally Posted by Fjodor2000

Also, HDR10 has max 1000 nits brightness, HDR10+ 4000, and Dolby Vision 10.000. But are there any movies actually exceeding 1000 nits max brightness? Will it never occur for older movies that 1000 nits is exceeded, but might be possible to exceed for newer movies? If so, when is the breaking point roughly, e.g. year ~2010, ~2020, ...?

That's not true. All support up to 10,000 nits. There are plenty of HDR10 titles that are mastered at 4000 nits, and some have MaxCLL that go up to 10,000 nits.

[Geoff D]

lol @ "are there any movies that exceed 1000 nits", "will it never occur for older movies that 1000 nits is exceeded", basically that whole post. Have you not actually looked at any of Pyoko's superb heat maps in this thread, guy? Check post #10 where you'll find an example of an HDR10-encoded 63-year-old film regularly exceeding 1000 nits, and post #17 for a 36yo film regularly blasting out above 1000 nits. Nits are not a function of the recording medium but the mastering that's been done to it, e.g. the MaxCLL on Wizard of Oz is like 2000 nits IIRC.

Originally Posted by BrownianMotion

That's not true. All support up to 10,000 nits. There are plenty of HDR10 titles that are mastered at 4000 nits, and some have MaxCLL that go up to 10,000 nits.

Correct. All three formats are based around the PQ EOTF residing in a 10,000 nit BT.2020 container. It's total, sheer, and complete and utter nonsense that any of them are capped at such levels.

[Fjodor2000]

Originally Posted by BrownianMotion

That's not true. All support up to 10,000 nits. There are plenty of HDR10 titles that are mastered at 4000 nits, and some have MaxCLL that go up to 10,000 nits.

Ok, I trust you are correct. The problem is there unfortunately is lots of contradictory information about this online. See examples below.

wikipedia:
"HDR10 most closely resembles Vesa's DisplayHDR 1000." Table then shows DisplayHDR 1000 having Max Brightness 1000 cd/m2.

digitaltrends:
"The HDR10 format allows for a maximum brightness of 1,000 nits, and a color-depth of 10 bits. [...] As the name suggests, HDR10+ takes all of the good parts of HDR10 and improves on them. It increases the maximum brightness to 4,000 nits [...] Dolby Vision provides for even greater brightness (up to 10,000 nits) and more colors too (12-bit depth, for a staggering 68 billion colors)."

rtings:
"HDR10
Mastered from 1000 to 4000 cd/m²

HDR10+
Mastered from 1000 to 4000 cd/m²

Dolby Vision
Currently mastered at 4000 cd/m², but supports up to 10000 cd/m²

[...]
Most Dolby Vision content is currently mastered at 4000 cd/m²; HDR10 and HDR10+ content are mastered at a variety of levels from 1000 to 4000 cd/m² depending on the title.

All three standards cater for images of up to 10,000 cd/m², although no display can currently reach that level. Therefore there is no real difference between the formats as they both top out at 4000 cd/m²."

[Geoff D]

I do understand that there's loads of nonsense out there from "trusted sites" but it really is as simple as what was described by Brownian. The digital trends piece is quite hilariously wrong. I mean, even when others get it correct they get it wrong, e.g. the rtings piece is quite right when it says that "all three standards cater for images of up to 10,000 cd/m2" but they muddle that with their previous claim of 10 and 10+ being "mastered from 1000 to 4000" for peak brightness with no mention of further "support". They're implying that all Dolby Vision content actually hits 4000 nits peak which is poppycock, sir.

If anyone has a Sony UHD disc then they already have a piece of full 10,000 nit content and they don't even know it: press 7669 on the main menu of any Sony UHD and you'll get some HDR10 test patterns, first a set of colour primaries and secondaries in 709, P3 and full 2020 and then a nit ramp from 0-10,000 nits, complete with 10,000 nit MaxCLL (which would be an excellent control test for pyoko's heat maps ):

[Pyoko]

Originally Posted by Geoff D

If anyone has a Sony UHD disc then they already have a piece of full 10,000 nit content and they don't even know it: press 7669 on the main menu of any Sony UHD and you'll get some HDR10 test patterns, first a set of colour primaries and secondaries in 709, P3 and full 2020 and then a nit ramp from 0-10,000 nits, complete with 10,000 nit MaxCLL (which would be an excellent control test for pyoko's heat maps ):

You mean this one?



One things has me confused though, the P3 blue in those charts differs from the one for Rec709 when everything I can find is that they should be the same, and I can't get it to line up on my gamut chart. The Sony P3 blue lands halfway between the 709 and 2020 blue points.

[andreasy969]

Originally Posted by Pyoko

You mean this one?



One things has me confused though, the P3 blue in those charts differs from the one for Rec709 when everything I can find is that they should be the same, and I can't get it to line up on my gamut chart. The Sony P3 blue lands halfway between the 709 and 2020 blue points.

I had a discussion with Deciazulado via pm a long time ago in regard to the Sony patterns. He was asking for some caps, because he noticed errors with the patterns on the Sony disc and wanted to make sure that it wasn't him i.e. his capturing process. He lost me with the colours , but in the end I got the same errors.

Long story short: The Sony pattern may very well come with errors. The grey scale contains colour for ex which you'll only see once you boost them.

[Pyoko]

Originally Posted by andreasy969

Long story short: The Sony pattern may very well come with errors. The grey scale contains colour for ex which you'll only see once you boost them.

I think that might be the case, the other 709/P3 colors are fairly close but some are still too much off for it to be conversion/rounding errors.