Originally Posted by harlekin
Complexity reduction incoming.
All this is at its heart, is me finding a quote in a CineTechGeek interview with Steve Shaw from Lightspace, that for me finally made things click - and now being able to write in a few short paragraphs what HDR is about, where it is at - currently - and why. So I did.
bt2020 (color range) is currently not a target many folks are aiming for. That this has become a standard at all - has more to do with the intent of creating a "wrapper" (think of it as a coordination system) for "all possible perceivable colors".
bt2020 serves as a grid system where all potential visible colors can lie - and thats about it - and probably will be it, for some time (lets say 10 years minimum ).
HDR at a non marketing level is an extension of dynamic range for "digital film", that raises it on equal level with film stock (and even a bit above) - this is, at its core - why many people on the production side like it.
To make HDR usable on the production side, literally everyone (please correct if not.. ) in production, has gone for DCI-P3 colorspace (primaries) as defined in the bt2020 grid (again, bt2020 as a coordination system and not much more), when mastering their movie content.
This is not an actual standard - but it has become the defacto working standard - because it is the ONLY way, that studios can move material between facilities and make sure it gets handled correctly internally. This little sidenote in fact guarantees, that DCI-P3 is a thing. Its the only practical solution for intercompatibility, its the "thing" everyone has agreed upon, and its not going to go away anytime soon (otherwise all production facilities would have to agree on a different color space within bt2020 again - which is a pretty HARD problem. )
Nits is almost an entirely different beast. Nits (max. image brightness) are at the digression of different studios, using different grading screens, mastering at differing peak brightness levels.
The saving grace here is 1000 nits. Currently.
Here is why. Regardless what movies get mastered to in the range of 1000-4000 nits in mastering - the color artists master with consumer displays at 1000 nits in mind. Currently. Meaning - whatever higher nit values they pump into the master - they do it in specular highlights mostly, they dont master entire "scenes" at higher nit levels (sunrise, ...). 1000 nits as the "ideal minimum value" (on a display with a very good black level, because we are talking dynamic range), is talked about by some color correction artists as the level where "image characteristics - currently - shouldnt shift much, compared to their mastering screens".
This is not an exact guideline, because many TVs resort to dynamic screen dimming, if a bright HDR image (APL) stays on screen for more than a few seconds. But this is there the "around 1000nits is good" guideline originates (minimum required brightness level to see "most of the content" in "about the same image characteristics it was mastered in" ).
Below 1000nits - image characteristics are starkly different to the mastering material (because of color space compression that also hits "scenic elements" and not just specular highlights) than what gets currently pressed on discs or distributed digitally. (= OLEDs in HDR production environments, still suffer from that - which might be the main reason for why they usually arent seen as mastering monitors for HDR...)
Here comes the fun part. As "capabilities (nits) improve" its at the digression of the mastering studio to change its production guidelines. At will. As long as "touching up specular highlights" is all thats done in the "above 1000nits regime" - thats not hugely problematic - but, the base level (average picture brightness, or whats considered "ok" to master a bright scene in HDR (APL)) can change at some point in the future - which becomes a huge problem, when it does (do we give those Blurays yet another marketing label?) - but currently its not an issue. I will be though. At some point in the future.
Currently, if your consumer device has great black levels (whatever that means ) and supports around 1000 nits, you should be mostly fine. (If DCI-P3 dEs are reasonable) Metadata will scale the image towards your screens performance (if it works, which it doesnt in anything but DV ) - and the color range compression should hit mostly specualar highlights - again, if your TV supports 1000nits. If its a current OLED - color space compression results in more severe color level changes in a range where its more problematic. (700nits top, 500 if we are honest - which can affect general image characteristic, and not just specular highlights, on current masters.)
Now to the consumer realm, which is where the entire thing gets f*cked up grandiously..
I wondered quite a while about why - but then I got presented with one plausible answer in the video interview with Steve Shaw.
Basically - if every TV manufacturer agreed to do DCI-P3 at 1000nits - if possible (the closest (and actually pretty real ) thing we have as an HDR production/reproduction standard), there would be no race. Every Image on every TV would look the same. As intended.
TV manufacturers dont like this. The reason they are in the HDR game, mostly - is to find a replacement for the "contrast race" - which is over, because several current TVs reached "infinity grade" on that one. Thats why they are pushing nits like crazy - thats why they insist on "moving out of the HDR EOTF "differently" (when close to running out of brightness) , thats why they announce to go for "closer to bt 2020", which is not a forseeable goal anywhere in the medium future.
All of this is basically BS. And DCi-P3 at 1000nits, with very dark blacks, following the HDR EOTF (ST-2084) is where its at.
Metadata is a workaround - that doesnt work all that well, until displays hit around 1000nits (overall image characteristics get altered after color space compression), after they do - it works better.
No metadata - currently (below 1000 nits on many consumer screens) is even worse - as we see with HDR 10.
HDR 10 at the current stage is basically broken, where image characteristics get highly altered, depending on colorspace compression thats only defined by amount of peak brightness difference (between master and your TV), based on one value per movie. This results in large amounts of compression on most current screens, in all movies, in all scenes. Which is the cause for the "not the same color twice" effect.
DV at the current stage is a little better - because it only suffers from that problem in "some scenes". Other than that, both suffer from the same underlying issues.
HDR10+ is an attempt to do automatic color remapping based on algorithms taking into account what your TV "reports" as its capabilities, no color correction artist needed.
Which is - what LG currently does on their OLEDs in the Dynamic Contrast (?) low mode (soon to be its own menu item.. ). They basically analyze a HDR 10 signal scene by scnene (frame by frame), and interpret, if it is a high or low brightness (APL) scene, what the light sources might be - and adjust their color remapping dynamically. On low APL, some scenes could show up in the correct color, if you are lucky..
Summary
We are in a "no standard works" state - where there is actually a de facto standard on the production side, thats "mastered with in mind" (DCI-P3 (defined in bt2020 space) at 1000nits, with a "good" black level) - but that is not a commercial standard - basically because manufacturers want to continue the "contrast race" somehow. (Nits.. )
Also - there will be a jumping off point, where "mastering with 1000nits screens in mind" might be dropped, at which point we have more issues.
Dynamic Metadata is not the solution for any of this - but it can help, in some scenes in some movies.
What it now means, that Blade Runner 2049 showed up with 10.000nits metadata - is anybodys guess. Hopefully its still mastered with 1000nits displays in mind, and APL in bright sccenes didnt get raised accordingly.
TLDR;
1000 nits for the consumer space does matter, 2000-4000nits at the mastering level doesnt matter that much more (in comparison - its still nice to have - and more accurate towards the corresponding master ), if they put those higher peak output levels in specular highlights mostly (which they currently do), but thats subject to change, sometime in the future - as screens get brighter and colorists get bolder. At which point - we've got new problems..
The thing preventing them to do that all too quickly - apparently is, that their mastering displays also auto dim the brightness level at a high APL held for more than a few seconds - so they dont master "entire scenes" (sunrise) there, but mostly put it in specular highlights.
Boy - an actual standard would be mighty fine at some point... (Something that doesnt say bt2020 at 10.000nits - full frame white. )
Reference material:
h**ps://www.youtube.com/watch?v=RsgOwi3pRMg
and the NAB16 Panel videos at
h**ps://www.youtube.com/user/cinetechgeek
If you find an error anywhere in here, please share.
Apart from that, have fun now actually knowing what is going on with HDR.
|
All countries
Argentina
Austria
Belgium
Brazil
Bulgaria
Chile
China
Colombia
Cyprus
Czech Republic
Estonia
Denmark
Finland
Greece
Hong Kong
Hungary
Iceland
Indonesia
Ireland
Israel
India
Latvia
Lithuania
Malaysia
New Zealand
Norway
Philippines
Poland
Portugal
Romania
Russia
Singapore
South Africa
South Korea
Sweden
Switzerland
Taiwan
Thailand
Turkey
Ukraine
United Arab Emirates
....everyone is getting onboard the train…..
athlete……
, I’d imagine) 
, in the course of future February discussion, we might give a short shout out to Koala bars….
) 
Interesting post at AVS: Why HDR is where it is - currently
Linear Mode
