HDR Discussion Thread

[Geoff D]

Originally Posted by Fendergopher

One could at least argue that the potential benefits of 4K/HDR outweigh a lot of the issues that we see with implementation, as opposed to if it was simply a higher resolution. 10(12)Bit, WCG, HDR & 4K offers a lot more than simply a sharper image with less compression, and considering how every new home video format has had problems it should come as no surprise that this new generation will hit quite a few bumps in the road to smooth operation.

Still doesn't excuse issues like the metadata farce going on, or how displays are all over the place as far as their HDR capabilities go, with manufactorers vying to jump on 8K before they've even made watching movies with dynamic metadata easy and convenient. Basically the technology is outpacing both the market and the studios at a rapid pace, workflows tend to change painfully slowly.

Very good point, it was hard enough for them to roll out a simple resolution increase so rolling out moar rezolushuns PLUS an entirely new electro optic transfer function was no doubt going to come with even more fun and games. But the frustration is there, and somewhat justified IMO, because the powers-that-be should've realised how much of a crapshoot HDR was going to be, and to be fair the Samsung mouthpiece provided a mea culpa of sorts in the article that Dan quoted. In lieu of an open-source dynamic metadata solution they should've at least decided on a set tone mapping method, but just getting all these companies around a table to agree the barest of HDR specs was hard enough, never mind all of them agreeing on one tone mapping solution when the entire TV market is predicated upon them all trying to one-up the other with all kinds of unique special sauce. As is often the case, the industry is its own worst enemy.

I'm not sure that 8K is going to be some vast jump in terms of what we're actually seeing though, as we're still going to be using the PQ EOTF first and foremost, still using P3 mastering for movies, still going to be watching movies at 24p, so all this other stuff like full 2020 coverage and hyper HFR frame rates is so much window dressing IMO. (Yes, it could be used more advantageously for live TV events 'n' stuff but strictly speaking for UK terrestrial TV we're still in the ****ing stone age.) Still, if 8K gives them an opportunity to properly unify HDR deployment then that's not a bad thing.

[alexanderg823]

I'm going to illustrate a little story I've had below with my Q9FN, which really highlights the benefit of a standardized HDR format, like Dolby Vision and why, as others point out along with me, HDR10 in its current state should have never been released.

The Q9FN doesn't follow the EOTF curve correctly, as per Samsung's decision. It shifts the curve upwards (while maintaining a similar slope) for 1,000 nit content, which results in a much brighter than intended imagery. This by itself isn't so bad really, as the Q9FN is so much brighter than 1,000 nit so it has the headroom. For 4,000 nit mastered content, the Q9FN actually does a terrific job by adhering close to the EOTF curve and then rolling off to preserve highlight detail all the way up to 4,000 nit. When you've got 2,000+ nit to work with, it works very very well.

However, one thing I noticed when watching Warner HDR10 content (which is mastered at 4,000 nit) on the Vudu app is that it always looks flat and blown out when there's specular detail in the image. What I discovered is that the Vudu app is sending the data as a 1,000 nit signal, sending some wonky metadata to the Samsung Q9FN, which interprets the 4,000 nit content as a 1,000 nit signal, then applies the jacked up custom EOTF curve to it, making the image look flat and blown out.

I was messing around with my Apple TV and loaded up the Vudu App for shits in giggles and then played around with a few warner titles such as Fantastic Beasts and Tomb Raider. What would you know? It's no longer blown out, as its sending the proper 4,000 nit signal for the TV to interpret and it looks absolutely fantastic. I couldn't believe my eyes.

The moral of this story isn't to talk about my Samsung Q9FN, or its Vudu App performance. The moral of the story here is that HDR10 leaves so much room for screwing something up that, surprise surprise, something probably will get screwed up, having some pretty dramatic effects on the PQ. With a proprietary end to end solution like Dolby Vision, issues like this are removed from the equation as Dolby controls the data flow.

So, at the end of the day, it makes sense why companies like Sony would go with DV on their high nit LCDs or why any company really would go to Dolby to turnover processing to a proprietary third party software like DV. Because the HDR10 system allows so much room for error and misinterpretation that it's just not worth having something along the chain screw up the signal and destroy the picture quality of the content being watched. Most people wouldn't be able to explain what was going on with the image, all they'd really be able to know is that inaccurate EOTF tracking can just end up with an image that looks "not good" and "less realistic" than one that's not having issues.

[Geoff D]

The metadata itself is at the root of all of this. I know you don't rate the "ignore metadata" method but I wonder if PQ10 (HDR10 w/out metadata) was the way to go (as is potentially the case for PQ-based live broadcasts), perhaps leaving in the colour information (primaries, white point) but having no brightness metadata whatsoever, letting the TV just map out the content according to where it's set to clip.

Then again it wouldn't have helped lower-nit displays in some respects, as although this approach would stop your Goodfellas and your BR2049s from looking too dark and murky it would've gone to extremes at the other end, blowing out certain brighter movies to shit. Yep, HDR has been a quite majestic balls-up thus far.

[alexanderg823]

Originally Posted by Geoff D

The metadata itself is at the root of all of this. I know you don't rate the "ignore metadata" method but I wonder if PQ10 (HDR10 w/out metadata) was the way to go (as is potentially the case for PQ-based live broadcasts), perhaps leaving in the colour information (primaries, white point) but having no brightness metadata whatsoever, letting the TV just map out the content according to where it's set to clip.

Then again it wouldn't have helped lower-nit displays in some respects, as although this approach would stop your Goodfellas and your BR2049s from looking too dark and murky it would've gone to extremes at the other end, blowing out certain brighter movies to shit. Yep, HDR has been a quite majestic balls-up thus far.

Yes. If you must use fixed brightness points, which I guess the studios really want so home releases get as close standardized as possible when viewing at home.

But to be quite frank I don't get what's really "wrong" with the HLG relative brightness approach. Some of the best HDR I've seen is HLG, such as the link I'm posting below. Just looks incredible.

https://www.youtube.com/watch?v=Uucd...ature=youtu.be

[Geoff D]

What's "wrong" with HLG is that it wasn't adopted by the BDA

[Robert Zohn]

I love HLG as it's adaptable for all TV Broadcasters to implement into their existing FHD system and is easily intergrated when they upgrade to OTT and ATSC 3.0 IP and OTA.

HLG is an analog HDR scheme and does not require a HDR shader or any infrastructure upgrade. HLG HDR has been included in professional ENG and studio cameras for the past few years.

[alexanderg823]

seems like HLG has less "potential" for image quality than PQ EOTF/HDR10, but the sheer amount of foolery and nonsense surrounding getting PQ to work right (along with all the potential opportunities to screw it up as I described above) just make HLG look much better - both literally and metaphorically - at the end of the day.

But, correct me if I'm wrong, I think that was kind of the point of HLG in the first place, to solve complexity and compatibility issues, so it should be no surprise it works very well and provides fantastic image quality without issue. That's what it was well thought out and designed for.

[Robert Zohn]

Yes in a way, but HLG was developed for on the fly applications that do not have the luxury of pre and post production. So it's a perfect match for TV Broadcasters and especially live sports.

[DisplayCalNoob]

Originally Posted by Geoff D

The metadata itself is at the root of all of this. I know you don't rate the "ignore metadata" method but I wonder if PQ10 (HDR10 w/out metadata) was the way to go (as is potentially the case for PQ-based live broadcasts), perhaps leaving in the colour information (primaries, white point) but having no brightness metadata whatsoever, letting the TV just map out the content according to where it's set to clip.

Then again it wouldn't have helped lower-nit displays in some respects, as although this approach would stop your Goodfellas and your BR2049s from looking too dark and murky it would've gone to extremes at the other end, blowing out certain brighter movies to shit. Yep, HDR has been a quite majestic balls-up thus far.

Isn't that what Disney does and some of the early releases like Life of PI, there's no metadata. All you get is better colors, better highlight detail and enhanced shadow detail. Metadata gives the colorist and cinematographer more time to refine depth.

Personally, I think Sony and Samsung both made the most mistakes. The commitment to edge lit displays, denying the benefits of dynamic metadata, Sony clipping instead using a roll off.

[Geoff D]

Originally Posted by DisplayCalNoob

Isn't that what Disney does and some of the early releases like Life of PI, there's no metadata. All you get is better colors, better highlight detail and enhanced shadow detail. Metadata gives the colorist and cinematographer more time to refine depth.

Personally, I think Sony and Samsung both made the most mistakes. The commitment to edge lit displays, denying the benefits of dynamic metadata, Sony clipping instead using a roll off.

There's no MaxCLL/MaxFALL metadata on Fox or Disney titles but there is still max/min mastering display metadata that practically every single UHD title has, apart from Pi (which was one of Fox's earliest titles) and T2 (which is from StudioCanal, 'nuff said), and it's that maximum mastering figure that some displays map to which is totally the wrong way of going about it IMO.

As for metadata itself I think there's a bit of confusion here: its primary function is to control how the content is downconverted to a display that cannot handle 100% of the signal, the actual underlying HDR grade is first completed without any of this stuff. Once the grading is done, scene by scene - yes, even in HDR10 - then the finished grade is analysed so that the static MaxCLL/MaxFALL metadata can be generated (if used) and the same is true for generating dynamic metadata, which is first created in an automated sweep even for Dolby Vision. The colourist then looks at the downconverted version mapped to whatever nits target is being aimed for (100, 600, 1000, whatever) and tweaks the metadata accordingly.

[alexanderg823]

Originally Posted by DisplayCalNoob

Isn't that what Disney does and some of the early releases like Life of PI, there's no metadata. All you get is better colors, better highlight detail and enhanced shadow detail. Metadata gives the colorist and cinematographer more time to refine depth.

Personally, I think Sony and Samsung both made the most mistakes. The commitment to edge lit displays, denying the benefits of dynamic metadata, Sony clipping instead using a roll off.

There's metadata, just not information on the mastering light level.

Three pieces of metadata are the mastering light level, max content light level, and max frame average light level.

I guess it's up to the display to decide how to tone map if they want to use the max content light level. if no mastering info exists.

[Geoff D]

[alexanderg823]

Originally Posted by Geoff D

lol. sums up the disaster that is HDR quite well.

[DisplayCalNoob]

Originally Posted by Geoff D

There's no MaxCLL/MaxFALL metadata on Fox or Disney titles but there is still max/min mastering display metadata that practically every single UHD title has, apart from Pi (which was one of Fox's earliest titles) and T2 (which is from StudioCanal, 'nuff said), and it's that maximum mastering figure that some displays map to which is totally the wrong way of going about it IMO.

As for metadata itself I think there's a bit of confusion here: its primary function is to control how the content is downconverted to a display that cannot handle 100% of the signal, the actual underlying HDR grade is first completed without any of this stuff. Once the grading is done, scene by scene - yes, even in HDR10 - then the finished grade is analysed so that the static MaxCLL/MaxFALL metadata can be generated (if used) and the same is true for generating dynamic metadata, which is first created in an automated sweep even for Dolby Vision. The colourist then looks at the downconverted version mapped to whatever nits target is being aimed for (100, 600, 1000, whatever) and tweaks the metadata accordingly.

The article I read, points to Da Vinci Resolve allowing several peak nit targets that are for consumer display targets. 600 nits, 1000 nits, and 2000 nits. I wonder if that is limited to DV, and Dolby patent also suggest that dynamic tone mapping and dynamic metadata can be separate functions of DV., tone mapping is disabled when the target display matches the mastering monitor 1:1, could this also be DV specific.

[DisplayCalNoob]

Originally Posted by alexanderg823

There's metadata, just not information on the mastering light level.

Three pieces of metadata are the mastering light level, max content light level, and max frame average light level.

I guess it's up to the display to decide how to tone map if they want to use the max content light level. if no mastering info exists.

HDR10 specific, limited data compared to DV.

[DisplayCalNoob]

Originally Posted by Geoff D

HDR10 specific, doesn't mention the added data that DV has, not consumer display data but actual content data.

I don't think it's a all that apply deal. There's more pixel, nit, and color data in DV.

[Staying Salty]

One-Blue Announces Memory-Tech to join Licensing Program for Blu-ray Disc™ and Ultra HD Blu-ray™ Products

https://www.prnewswire.com/news-rele...300682158.html

"Memory-Tech Holdings is pleased to join the One-Blue product licensing program as a licensor. We are sure of rapid growth of 4K and HDR through UHD BD, and hope to contribute to developments of this industry as one of the leading Optical Disc Replicators through the One-Blue pool," said Yutaka Ueda, President of Memory-Tech Holdings.

Got to be good news for Us Discsers™.

[vincentric]

Originally Posted by Geoff D

[Show spoiler]

Am no military man but that there is a charlie foxtrot of the highest magnitude. I think going forward the studios should just pay Sony to do their UHD disc encodes and call it a day.

[gkolb]

Originally Posted by vincentric

Am no military man but that there is a charlie foxtrot of the highest magnitude. I think going forward the studios should just pay Sony to do their UHD disc encodes and call it a day.

But, that chart has nothing to do with different discs, it’s how different manufacturers displays handle the SAME disc (theoretically).
They’re all over the cornfield each doing something else , to the same content.

[LordoftheRings]

Originally Posted by gkolb

But, that chart has nothing to do with different discs, it’s how different manufacturers displays handle the SAME disc (theoretically).
They’re all over the cornfield each doing something else , to the same content.

Sony should lead the way, and make their own HDR OLED panels, plus teach other movie studios how to make HDR 4K Blu-ray movie transfers.
Sony should buy Dolby, and LG too, and Disney.