AVS Forum banner

121 - 140 of 2439 Posts

·
Registered
Joined
·
1,773 Posts
Yes, when switching the Vetex DV LLDV EDID in and out, the source device alters the signal thru the chain between DV and HDR10.

However, a question I still have, is there a difference in the actual streamed source content, depending on how its initially launched? ie... is there a DV and HDR10 version depending on how the initial handshake takes place? For instance with Disney+, you will see either format displayed depending on how one is connected to the site.

Or, does D+ only have a single 4k UHD DV/HDR source that is sent and is decoded DV or HDR10 depending on the streaming box capability?

If the latter, yes it might be as simple as turning DV LLDV on/off like with content on disc. I'd love to confirm what specifically happens when streaming with a host. I have given deference to the possibilty of streaming being different from playback from disc, but welcome clarification.
While I can't tell you the intricacies of streaming profiles and how they adapt, I can tell you that a Dolby Vision Profile 5 source (aka LLDV) is a single layer video in a Dolby specific colour space they call IPT which is very similar to BT.2100 which in turn has the same colour primaries as BT.2020. The EOTF is an ST.2084 PQ curve (Starts to sound familiar doesn't it?) and scenes are analysed in post for CLL to provide dynamic but predetermined tone mapping criteria.

So although DV Profile 5 isn't HDR10 compatible in the way that Profile 7 is (HDR10 base layer + enhancement layer combined to reproduce the mezzanine layer) it's not a million miles away.

And this is what you're seeing when you squirt a Profile 5 source at an HDR10 monitor.

Goodness knows if it's all clever enough to adapt profiles if you switch EDID mid stream though...

Have to admit that although it's a bit bonkers and the guy who came up with this has previous, it is intriguing.. Definitely not what the DOP intended, but intriguing nonetheless.

Cue 70's guitar riff with a heavy Gretsch back beat....
 
  • Like
Reactions: Bytehoven

·
Registered
Joined
·
4,904 Posts
Although "hotswapping" between HDR10 and LLDV works fine on the OPPO 203, i am having trouble doing the same on the ATV4K and XB1X, probably because it is live streaming, and strougles to adjust to another profile on the fly. So it is a bit difficult comparing while watching.
 

·
Registered
Joined
·
6,965 Posts
Discussion Starter #123 (Edited)
While I can't tell you the intricacies of streaming profiles and how they adapt, I can tell you that a Dolby Vision Profile 5 source (aka LLDV) is a single layer video in a Dolby specific colour space they call LPT which is very similar to BT.2100 which in turn has the same colour primaries as BT.2020. The EOTF is an ST.2084 PQ curve (Starts to sound familiar doesn't it?) and scenes are analysed in post for CLL to provide dynamic but predetermined tone mapping criteria.

So although DV Profile 5 isn't HDR10 compatible in the way that Profile 7 is (HDR10 base layer + enhancement layer combined to reproduce the mezzanine layer) it's not a million miles away.

And this is what you're seeing when you squirt a Profile 5 source at an HDR10 monitor.

Goodness knows if it's all clever enough to adapt profiles if you switch EDID mid stream though...

Have to admit that although it's a bit bonkers and the guy who came up with this has previous, it is intriguing.. Definitely not what the DOP intended, but intriguing nonetheless.

Cue 70's guitar riff with a heavy Gretsch back beat....
Thank you Mark.

- Given the unintended nature, could you comment regarding the range of pitfalls related to what adjustments we might be making on various displays or projectors?

Color space seems to be one of the areas where folks are observing different image quality performance of saturation, tracking.

Overly dark shadow detail performance or exaggerated upper midtones and highlights are two things I have noticed on a couple of HDR10 titles.

- Out of your own intrigue or curiosity, where might you look next to improve upon this unintended use?

You might have noted one member asking about editing the EDID or DV data blocks to provide offset adjustments for our unique display/projector.

Or how about experimental custom gamma curve calibrations to build a better basket to receive the DV LLDV signal?
 

·
Registered
Joined
·
6,965 Posts
Discussion Starter #124
Although "hotswapping" between HDR10 and LLDV works fine on the OPPO 203, i am having trouble doing the same on the ATV4K and XB1X, probably because it is live streaming, and strougles to adjust to another profile on the fly. So it is a bit difficult comparing while watching.
Thank you for that observation...

I noticed using switching DV on/off on the ATV4k and relaunching the test movie, sometimes resulted in sound dropping out, but I thought it might have been server loads associated with the launch and popularity of the title.
 

·
Registered
Joined
·
1,773 Posts
Thank you Mark.

- Given the unintended nature, would you venture further regarding the pitfalls related to what compensation might be required on various displays or projectors?

Color space seems to be one of the areas where folks are observing different image quality performance.

Overly dark shadow detail performance or exaggerated upper midtones and highlights are two things I have noticed on a couple of HDR10 titles.

- Out of your own intrigue or curiosity, where might you look next to improve upon this unintended use?

You might have noted one member asking about editing the EDID or DV data blocks to provide offset adjustments for our unique display/projector.

Or how about experimental custom gamma curve calibrations to build a better basket for the DV LLDV signal?

I guess this assumes DV LLDV with some EDIDs might not expect the display to do any further DV processing, which I think the Sony A1 might a prime example.
So issues would seem to be the disparity of the proprietary colour space being interpreted as presumably either BT.2020 or P3 in a non DV display. Also you have the PQ transfer function which then would require tone mapping for a particular display (one of the key elements of DV) by a different and notably non DV display.

Then there's the issue of the tone map - and this is exactly not DTM (ie calculated on the fly) - the Dolby meta data is transported either in an SEI message as it is for HDR10 or NAL dependent on delivery medium and there's no way an HDR10 display will know what to do with it either way.

I honestly think this is a bit of a curiosity that will produce results that are "different" and that's not to say you can't interpret that as better than what you had before but it certainly isn't what the guy sitting in the swivel chair wiggling his trackball at DaVinci at 2AM with a jug full of coffee next to him intended.

I hesitate to use the word hack since it has acquired a slightly more apocryphal meaning in these "enlightened" times but as with any hack, it needs experimentation and fiddling rather than following a rule book - so I think definitely a custom colour profile with IPT primaries (and I'm only talking JVC language here) and then some kind of one size fits all gamma curve, but I think that will be much more of a compromise than standard custom curves applied to HDR10...

I'm afraid I certainly can't see this as a substitute for decent processing, whether that's internal tone mapping in say a Panny 820 or Oppo 203 or the external, true the DTM processing of a lumagen, but then all of these have heavy compromises in trying to crowbar HDR video into the very much LDR capabilities of projectors.

YMMV ;)
 
  • Like
Reactions: Bytehoven

·
Registered
Joined
·
6,965 Posts
Discussion Starter #126 (Edited)
... I honestly think this is a bit of a curiosity that will produce results that are "different" and that's not to say you can't interpret that as better than what you had before but it certainly isn't what the guy sitting in the swivel chair wiggling his trackball at DaVinci at 2AM with a jug full of coffee next to him intended.

... but then all of these have heavy compromises in trying to crowbar HDR video into the very much LDR capabilities of projectors.

YMMV ;)
Thank you for that perspective, and may I say I appreciate the late nite wiggling. :grin:
 

·
Registered
Joined
·
9,066 Posts
Then there's the issue of the tone map - and this is exactly not DTM (ie calculated on the fly) - the Dolby meta data is transported either in an SEI message as it is for HDR10 or NAL dependent on delivery medium and there's no way an HDR10 display will know what to do with it either way.
How much of the “heavy lifting” is done by the player in the case of LLDV? Maybe the name “Player-Led” has been giving me false hopes.:(
 

·
Registered
Joined
·
1,205 Posts
Could someone explain in detail the difference between Integral2 and Vertex2 for LLDV playback specifically? I have a JVC X590 if that makes any difference. What features do I miss with the cheaper Integral2?
 

·
Registered
Joined
·
1,773 Posts
How much of the “heavy lifting” is done by the player in the case of LLDV? Maybe the name “Player-Led” has been giving me false hopes.:(
If we're talking Profile 7 Dolby Vision UHD Blu-Ray discs, then yes, the player is doing the processing that would usually be handed off to the display, but only to a certain extent because the display still has to handle Profile 5.

I see a lot of posts (not in this thread I hasten to add) of people touting so-called LLDV (not an official acronym!) as a "new" version of DV, implying that it's somehow an improvement where in fact it was a late capitulation by Dolby to appease perhaps their biggest partner, Sony, after they sprung out of the gate early with inadequate chipsets unable to process full-fat DV (Profile 7).

There are 5 DV profiles with different degrees of cross compatibility but the one most people are familiar with is Profile 7 which is cross compatible with HDR10 (others are cross compatible with HLG and even BT.709).

The reason for this cross compatibility, and the cleverness with this, stems from the use of 2 separate video streams,

The base layer, in other words the movie itself, is a standard UHD 10-bit HEVC 3840x2160 HDR10 grade, albeit theoretically graded to Dolby's standards. (I should note here that Profile 7 also allows for a 1920x1080 base layer too)

There is then a 10-bit HEVC 1920x1080 Enhancement Layer, a separate FHD video stream, which contains the "difference" information between the Dolby Master or the Mezzanine (A mezzanine is a compressed but indistinguishable version of the master used for deliverables).

I love the fact that someone at Dolby decided to call the software to create it The Mezzinator :)

Combining the two provides the full 12-bit Dolby Vision stream and it's at this point where someone at Sony called someone at Dolby and said "er... about that......"

So Profile 5 was created - single layer, ie just one 10-bit HEVC video stream, with a proprietry colour space and PQ based ST.2084 EOTF and no cross compatibility. This pleased Sony.

But here's the thing:

HDR was hijacked from the get go by rabid marketeers and touted as "contrast this" and "brightness that" and "ooh it's so vivid", when in actual fact HDR is exactly not that - the APL of an HDR grade should be very similar to that of the LDR grade. The luminance graph should be the same up to diffuse white at 100nits and only specular highlights (an oft toted but generally misunderstood term) sit beyond.

But that doesn't sell TVs. In the old days, "showroom Mode" sold TVs (brightness full, contrast full) but now they have HDR, and since HDR10 is an open standard, there's nothing to stop content creators from using this range badly.

Which in the early days they did. A lot.

Things are calming down now and grades are getting better but it's all still a bit of a mess - The one thing Dolby tried to do right from the start was to control the process and keep standards where they should be.

They provide training and workflows and certification so theoretically a Dolby grade should be correct every time which means maybe, just maybe, if you take a properly graded Profile 5 stream, create a colour profile to match IPT and figure out the right transfer function (gamma) you may, just may, be onto a winner.

And that's really what this hobby is all about - fiddling around with stuff like this and anything that adds interest has to be good.
 

·
Registered
Joined
·
392 Posts
According to Light Illusion IPT is not a colour space but 'signal encoding'. Dolby Vision uses Rec.2020 colour space.
 

·
Registered
Joined
·
1,773 Posts
According to Light Illusion IPT is not a colour space but 'signal encoding'. Dolby Vision uses Rec.2020 colour space.
True for Profile 7 but not profile 5.

You may be getting mixed up with ITP which is another name for ICtCp colour encoding which is derived from IPT.

Far too many acronymns in this business!
 

·
Registered
Joined
·
1,773 Posts
Could someone explain in detail the difference between Integral2 and Vertex2 for LLDV playback specifically? I have a JVC X590 if that makes any difference. What features do I miss with the cheaper Integral2?
All these devices need to do in this context is to spoof the EDID, so a Vertex, an Integral2 or a Vertex2 are equally capable.
 
  • Like
Reactions: Bytehoven

·
Registered
Joined
·
6,965 Posts
Discussion Starter #133
... so theoretically a Dolby grade should be correct every time which means maybe, just maybe, if you take a properly graded Profile 5 stream, create a colour profile to match IPT and figure out the right transfer function (gamma) you may, just may, be onto a winner.
Now I know what to put on my holiday wish list.

It also explains why this Profile 5 exploit looks great with some films and less so with some?
 

·
Registered
Joined
·
392 Posts
True for Profile 7 but not profile 5.

You may be getting mixed up with ITP which is another name for ICtCp colour encoding which is derived from IPT.

Far too many acronymns in this business!
It was the answer I received after I explicitly mentioned Profile 5 and pointed to posting #121 .
 

·
Registered
Joined
·
9,066 Posts
Now I know what to put on my holiday wish list.

It also explains why this Profile 5 exploit looks great with some films and less so with some?
Presumably if one has a Priofile 5 compatible TV some measurements can be made to find out the colour space and the EOTF?

Why would this kind of information not be published by Dolby?
 

·
Registered
Joined
·
6,965 Posts
Discussion Starter #136
Presumably if one has a Priofile 5 compatible TV some measurements can be made to find out the colour space and the EOTF?

Why would this kind of information not be published by Dolby?
The Sony A1 OLED seems to be Ground Zero.
 

·
Registered
Joined
·
1,773 Posts
Presumably if one has a Priofile 5 compatible TV some measurements can be made to find out the colour space and the EOTF?

Why would this kind of information not be published by Dolby?
Dolby are quite rightly careful about exposing their IPR and although they publish white papers, they are understandably kind of vague.

Basically it's because unlike HDR10 and 10+, DV is not open source and is licensed, so you can't really blame them.

I think a good starting point is actually BT.2100 and since the primaries are the same, the JVC BT.2020 colour profile should be as near as dammit.

As for gamma - well, I'd start with the standard ST.2084 and experiment in Excel from there - that's how I came up with my current gamma curves for HDR10 - a more empirical approach rather than mathematical.
 

·
Registered
Joined
·
1,773 Posts

·
Premium Member
Joined
·
3,883 Posts
Could someone explain in detail the difference between Integral2 and Vertex2 for LLDV playback specifically? I have a JVC X590 if that makes any difference. What features do I miss with the cheaper Integral2?
All these devices need to do in this context is to spoof the EDID, so a Vertex, an Integral2 or a Vertex2 are equally capable.
For those with JVC projectors, HDfury has agreed to my request to add a JVC Macro that is executed when the input signal is LLDV. This will allow automated selection of the JVC user mode you want to use for LLDV. This also eliminates any reason to inject fake HDR metadata values.

I expect this macro will be added to the Vertex2, Diva, and Maestro. It may or may not be added to the Integral2 depending upon whether there is memory still available for the additional macro.
 
121 - 140 of 2439 Posts
Top