Dolby Vision

[DaMacFunkin]

No one knows for sure what DVDFAB is doing, I personally seem to think it discards all HDR10 data from layer 1 and interleaves meta data and extended Color information from layer 2, thus creating a compatible lower latency stream.

I have played back single and dual layer rips of the same files on my LG 55” C7 and 65” C9 and I would argue to the death there is no noticeable difference, obviously with the dual layer stream none DV compatible media players can still playback in HDR10, with the single layer you don’t get any picture, just audio.

[wesk05]

DVDFab is not producing true profile 5 Dolby Vision stream. It has been tested out by us (Sorry, I can't give the details).

I personally seem to think it discards all HDR10 data from layer 1 and interleaves meta data and extended Color information from layer 2, thus creating a compatible lower latency stream.

It does nothing of that sort.

[noggin]

Reading between the lines it looks to me like DVDFab may have found a way of muxing the Dolby Vision RPU data (but not any additional video content in the Enhancement Layer?) into the HDR10 stream to create something that is enough to convince DV displays that they are getting a single layer DV stream and process that metadata (if indeed they are)?

This could mean you get some of the DV benefits of dynamic DV metadata, but not the benefits of the video content in the enhancement layer (which could add 10->12 bit enhancement in some cases?)

[DaMacFunkin]

DVDFab is not producing true profile 5 Dolby Vision stream. It has been tested out by us (Sorry, I can't give the details).

I personally seem to think it discards all HDR10 data from layer 1 and interleaves meta data and extended Color information from layer 2, thus creating a compatible lower latency stream.

It does nothing of that sort.

You can’t just shoot it down like that, you either have proof or you don’t, if you have knowledge and are not sharing it, it is no wonder people are collating things they see and things they read and interpreting it as it it seems to be, I’m not the only person who has come to that conclusion.

Waiting for the clever people to educate me.

[noggin]

No one knows for sure what DVDFAB is doing, I personally seem to think it discards all HDR10 data from layer 1 and interleaves meta data and extended Color information from layer 2, thus creating a compatible lower latency stream.

I'm not sure what you mean by 'discards all HDR10 data from layer 1'? Layer 1 is a clean HDR10 stream isn't it? If you discard all HDR10 data - then you are literally deleting that stream?

HDR10 is just a fancy way of saying 10-bit HEVC 4:2:0 video with an ST.2084 PQ EOTF and ST.2086 static mastering metadata (i.e. what the MaxCLL/MaxFALL and mastering display specs).

My understanding is that the Layer 2 DV track in a dual-layer disc can either just be an MEL (Minimal Extension Layer) which adds the DV RPU which allows Dolby Vision Scene Metadata (ST.2094-10) to be provided alongside the Layer 1 ST.2086 title metadata (and not a huge amount else) or it can be an FEL (a Full Extension Later) which adds both RPU Metadata AND additional information that allows the 10-bit HDR10 track in Layer 1 to be expanded to 12-bits (effectively the Layer 2 track contains a 10-12 bit difference signal - or something that allows it to be reconstructed)

I suspect the RPU data (which will have a significant impact on how some DV content is presented on a given display) is more visible to most of us than the 10-12 bit difference data (the former will have a lot more visual impact than the latter, which will be more to do with reduced quantisation artefacts - i.e. banding?)

However wesk05 (who REALLY knows his stuff when it comes to real world mastering and video processing) is suggesting that what DVDFab is producing isn't a fully DV Profile 5 stream (whether this is because it's not handling metadata properly, just not interleaving RPU and Extension layer stuff properly, discarding 10-12 bit extension layer stuff, or just sending bogus DV metadata enough to trigger a DV logo and not much else, and nobody has spotted this, who knows!)

AIUI DV does some pretty HEVC NAL-level low level stuff in HEVC to carry the DV-specific stuff - so it's quite low-level at the codec level (which is presumably why it still flagged as HEVC)

I suspect if wesk05 is employed in his day job to do work for specific companies, he can't share this openly as it would be confidential.

[davilla]

wesk05 has much video decoding fu

[zalman]

Hi, need same help please!

[Importer] Cannot load filter for input file "H:\dtshd.mp4"

What i'm doing wrong?

Thanks

Solved

Next step, mux srt and truehd atmos

[wesk05]

Those who have read by past posts would know that the statements that I make are almost always substantiated. In this particular case, the tests were done at my work place. So I can't really provide the details. @noggin's assumption is actually correct: the elementary stream is "manipulated" to trigger Dolby Vision. The RPU data itself has constants (header, VdrDmDataPayload). The CM data in RpuDataMapping needs to be generated, also profile 5 stream must be in IPTPQc2/IPT color space. That's not the case with the streams produced by DVDFab. This is all the information that I can provide now.

[davilla]

Those who have read by past posts would know that the statements that I make are almost always substantiated. In this particular case, the tests were done at my work place. So I can't really provide the details. @noggin's assumption is actually correct: the elementary stream is "manipulated" to trigger Dolby Vision. The RPU data itself has constants (header, VdrDmDataPayload). The CM data in RpuDataMapping needs to be generated, also profile 5 stream must be in IPTPQc2/IPT color space. That's not the case with the streams produced by DVDFab. This is all the information that I can provide now.

Thx wesk05 for this info, I'm sure everyone understand you are limited in what details you can provide. I've been in similar situations in the past.

[noggin]

Those who have read by past posts would know that the statements that I make are almost always substantiated. In this particular case, the tests were done at my work place. So I can't really provide the details.

Totally understand. I'm often in a similar situation. Thanks for sharing what you can.

@noggin's assumption is actually correct: the elementary stream is "manipulated" to trigger Dolby Vision. The RPU data itself has constants (header, VdrDmDataPayload).

When you say constants - does this mean the RPU data is fixed/constant and totally unrelated to the source content, and all DVD Fab is doing is saying 'Look - I'm DolbyVision, I've got some RPUs in this stream - please put up a DV logo to fool people into thinking this is proper DV'?

The CM data in RpuDataMapping needs to be generated, also profile 5 stream must be in IPTPQc2/IPT color space. That's not the case with the streams produced by DVDFab. This is all the information that I can provide now.

Ah - so Profile 5 can't be in Rec 2020 YCrCb and has to be in a totally different colourspace ? (AIUI IPT is VERY different to YCrCb - and is based around intensity, blue-yellow and red-green components)

Using an ITP space rather than YCrCb makes sense - it's allowed under Rec.2100 and is a favoured Dolby approach (presumably Dolby are a major reason it's in Rec.2100?)?

So the DVDFab stuff is presumably YCrCb with either real or bogus RPUs?

Worst case - it's just HDR10 flagged just enough to trigger a DV logo?
Better case - it's HDR10 + useful DV RPUs that trigger DV and improve tone mapping on DV displays?
I'm going to guess it does nothing with a Full Extension Layer to take it to 12 bit from 10...