[opus] Antw: Re: Antw: [EXT] Opus merging streams

[Sampo Syreeni]

On 2022-04-06, Ulrich Windl wrote:

>> I cannot for the life of me understand why Atmos exists. Except for
>
> I don't know the theory behind, but I guess they want a format open to 
> future sound designs (i.e.: how many and where to place speakers).

Third order ambisonic is already nigh correct for a central, static 
listener. Especially when augmented with dynamic decoding for specular 
sources, akin to the higher order extensions of DirAC. It really does 
not *need* hundreds of separate parametric sources. It's just a good 
enough description of the central soundfield, taken from any angle.

If parallax was involved, such as in games where you don't stand still, 
I could understand the point. But as far as I understand it, neither 
Atmos nor AC-4 support anything of the kind. So where's the point? The 
selling point is adaptation to different speaker layouts, but high 
enough order ambisonic can do that already, 3rd order is well enough for 
regular speaker layouts, and no irregular layout is going to work even 
for re-rendered specular sources. That's just basic math: irregular 
enough meshes don't admit stable quadrature.

Plus in the meanwhile, Atmos is about synthetical, specular sources. You 
can't *mic* an Atmos field, but have to *compose* it. Third order 
ambisonics on the other hand *is* a technology you can so-and-so capture 
from the real world. Plug-and-play with spatial sound, to a degree.

To me it sounds just stupid to bring in all of the "spatial objects" 
nonsense, when all it does is to add complexity and weight to the 
abstraction.

> Similar why DCPs (Digital Cinema Packages) encode color in CIE XYZ 
> (you can encode invisible "colors").

Don't even get me started... XYZ is there because it's derived from the 
tristimulus theory at the physical level. Its weighting functions take 
on negative values only because the optimum transform to "color" call 
for them, and because we want the XYZ space to stay positive. Human 
vision then isn't quite like this, nonlinear as it is. So when charted 
out in the linear XYZ space, vision takes on a wonky, though convex, 
shape.

As such, the "invisible" colors are a mathematical artifact. They are 
not an intrinsic part of the color space, but rather the complex side 
conditions of which XYZ values you are allowed to use in order not to 
encode imaginary colors are the way we model the truth about how human 
vision works. If you went out of gamut, it's not that "an imaginary 
color has suddenly been discovered", but that you failed to respect the 
boundary conditions engendered by the -- rather well-defined -- 
restrictions of the gamut.


> I guess the idea also was not to restrict the color space to the 
> limits of any existing device.

Indeed it was not. The idea was to start with linear tristimulus theory 
which was already known to be true, and then to model the newer (still 
rather old) opponent process theory of color within it. CIE's various 
standards rather successfully do just that.

> Or maybe they just want to fight pirated copies by filling up the 
> disks faster ;-)

I think this is the most likely explanation. Because, you know, Dolby is 
Dolby. It's always gone the Intellectual Property route, throughout its 
existence. Hell, at one time they refused to let anybody licence their 
noise reduction algorithms except as implemented by their analogue 
chips. Fuck, I don't think they even now permit anybody to implement SR, 
A, B, C, S, god forbid Prologic-II(x) in software. And they just keep on 
"inventing" -- not unlike one of our Finnish national prides ABLOY (lit. 
"Aktie Bolaget Lukko Osakeyhtiö", roughly "plc. Lock Plc.") keeps 
churning out newer and newer designs of keys in order to keep them under 
trademark and design rights.
-- 
Sampo Syreeni, aka decoy - decoy at iki.fi, http://decoy.iki.fi/front
+358-40-3751464, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2