Dan Dugan wrote:
> Mono recording piles everything together, and
> some sounds obscure others. That is not how
> humans hear--we hear in three dimensions.
This is confusing two issues but it will take me a bit to explain why. We h=
umans do indeed hear with ears that pick up sounds from all directions, but=
some directions are heard louder -- a lot more -- than others. And of cour=
se this directionality is frequency dependent, with higher frequencies bein=
g much more directional (see HRTF).
In the non-electro-acoustic world, sounds are (usually) emitted from point =
sources. Thus a point source microphone suffices to capture the sound. Gene=
rally one will want this to be as directional as possible to (help) exclude=
other sounds. Those who wish to approximate the sensation of "being there"=
will want to capture the ambiance as well as the sound source. For this th=
ey might resort to more complicated recording techniques.
Modelling one's microphone placement on human hearing makes sense in only o=
ne case: where you are planning on decoupling the listener's ears from the =
sound field during playback -- that is, for two-channel binaural recordings=
listened to on headphones. If instead you are planning on your sounds bein=
g heard over speakers, modelling human ears in the microphone placement int=
roduces an extra HRTF to the signal. This makes no sense to me.
Just as stereo speakers make no sense. No sound in the natural world is eve=
r emitted from two places at once in that way. It is an approximation only,=
not an ideal. The fact that we have two ears and there are two speakers is=
a non sequitur.
It is true that two recordings heard from one location might mask each othe=
r in cases where those same recordings heard from multiple locations will b=
e distinct. But this is a matter for playback not recording. That's why my =
last soundscape recording was for an 8.1 speaker system. I consider that a =
bare minimum to create an immersive sound field that can approximate the ac=
oustic world. (16.4 in a dome arrangement would be better.)
If I was fixated on verisimilitude I'd just use binaural and be done with i=
t.
-- robin
|