Add informative note to tts:luminanceGain that tts:luminanceGain is not relative to graphics white of the display

[jmccrossan-fox]

Currently the definition of luminanceGain states:
"The tts:luminanceGain attribute is used to specify the absolute luminance of a region when its presentation requires a greater dynamic range than that provided by the [SRGB] color space, and when that presentation requires an absolute luminance value."

To provide further clarity, consider adding an informative note to the above to advise that when tts:luminanceGain is applied during presentation it defines an absolute color & brightness, and it does not reflect a luminance level which is in relation to the graphics white of the display.

[nigelmegitt]

@jmccrossan-fox Thank you for raising this issue!

Please could you explain more about the motivation here? For example is there a reading of the current text that suggests a relative luminance level? Has this text caused confusion for an implementer for example, that needs to be resolved?

My reason for pushing for this is to make sure we have a clear idea of the problem that you've found, and therefore can assess any potential solution against that problem.

[jmccrossan-fox]

Hi @nigelmegitt thanks for reviewing!

In technical discussions that we have participated in, some technical experts have interpreted that “luminanceGain” is a gain relative to graphics white – probably due to the use of the word “gain” in the attribute. The current normative text states “the tts:luminanceGain attribute is used to specify the absolute luminance of a region ….” which I think is clear as a normative statement, but given the confusion encountered we believe there is benefit to adding an informative note explicitly highlighting that tts:luminanceGain is not relative to graphics white.

[css-meeting-bot]

The Timed Text Working Group just discussed TTML2 luminance gain.

The full IRC log of that discussion

<cyril> Topic: TTML2 luminance gain
<nigel> scribe: nigel
<nigel> github: https://github.com//issues/1117
<nigel> Cyril: The context is that I started discussing internally at Netflix
<nigel> .. with teams who implement HDR on games consoles, asking how to render luminanceGain.
<nigel> .. People looking at the spec for the first time were confused by the term "gain".
<nigel> .. It's an absolute luminance but called a gain. Initially they thought it was gain relative to graphics white.
<nigel> .. You can discuss with the TV to know what its graphics white luminance is. They interpreted it
<nigel> .. as compared to that and not compared to 18 nits which the spec says.
<nigel> .. The spec is not unclear, but they had an assumption.
<nigel> .. We agree with what is written in the spec and Fox proposed this clarification. We think
<nigel> .. it is a good one.
<nigel> Nigel: Will you propose a pull request?
<nigel> Cyril: Yes, it will be a Note saying that the gain is relative to 18 nits and not to a relative value
<nigel> .. corresponding to the device graphic white, or something along those lines.
<nigel> Nigel: Ok, looking forward to seeing the pull request for that.

[skynavga]

This seems reasonable, I will create a PR to this effect. It is too bad that gain was included in the name of this property. I don't recall the particulars of the history of the naming but it was poorly made for certain. Names are important and should reflect their intended semantics.

[skynavga]

Upon re-reading the current specification text, I have come to the conclusion that luminance gain does indeed specify a gain. If it were an absolute luminance, then it would have the units of candelas per square meter (or equivalent). But it is in fact unit-less, and functions as a multiplier with respect to 80 cd/m2 with a reference color of SRGB white. I will express the note accordingly.