[openal] AL_SOFT_source_spatialize proposal

[Lubomir I. Ivanov]

On 7 May 2017 at 02:57, Chris Robinson <chris.kcat at gmail.com> wrote:
> On 05/06/2017 03:20 PM, Lubomir I. Ivanov wrote:
>>
>> automatic 1/channels attenuation when downmixing to mono is very
>> common on IRL consoles and DAWs.
>> it's user friendly and IMHO prevents a complication over giving more
>> manual control.
>
>
> Right. Though I've also seen plenty of recommendations to use
> sqrt(1/channels) attenuation too, which better preserves the apparent
> loudness. What's the benefit of using 1/channels instead of
> sqrt(1/channels), aside from clipping prevention?
>

i guess it becomes a question if one wants to amend in terms of
perceived loudness or in terms of precision.
from what i've seen on consoles and DAW mixers used by engineers the
multiplier is strictly 1/N.

for such a bus it's crucial to amend by the precise multiplier as
there is also observation of magnitude meters involved.

for "ambisonics" and things like that sqrt(1/N) does a better job,
they say. but this will clip more as more channels are added, thus i
would probably go with: 1 - sqrt(1/N).

to really preserve the loudness all inputs need to be amended based on
an adaptive impulse response:
http://www.tubepreamp.net/images/human-hearing-frequency-range-curve-graph.gif

because the linear multiplier sqrt(1/N) is not really sufficient - at
least that's how i see it.

lubomir
--