[KH]

On 6/13/2012 4:47 PM, Sebastian Kaliszewski wrote:
KH wrote:
On 6/12/2012 8:20 PM, Sebastian Kaliszewski wrote:
KH wrote:

snip

And phase as Audio Empire points out.

Well yes, but what is phase except a temporal shift?

You're conflating phase and wavefront. You can have 180deg off phase
signals coming at the same moment.

What I meant to say was "phase differences". Two identical waves 180deg
off phase are different only in an f/2 temporal shift right?


in conjunction with the HRTF of the listener, create a spacial image.
That information was not, however, encoded into the recording except
as temporal and level information.

And possibly phase as well.

Ditto

See above. Phase is a property different from timing.

With respect to the specific discussion, I don't see how they can be
considered separate properties. For example, if we were to take two
instruments that could produce a single pure tone (hypothetically) and
place them on a stage, one 3M from the mic, and one 5M from the mic, and
shifted laterally 1M. If both instruments were started simultaneously,
calibrated to provide equal signal levels at the microphone, then the
wavefront at the microphone position would, from a direct perspective,
comprise two out of phase waves right? Yet, the only difference in the
signals is arrival times of the respective peaks and troughs, no?

So add a second microphone, and you have the same signals recorded from
a different position in space. As long as you know the microphone
positions, its easy to determine the relative positions of the two
instruments, aurally or mathematically. Yet when you add in the effects
of the reverberant sound field you have a whole new set of signals of
varying strengths and arrival times, and thus phase differences. As a
listener, in place of the microphones, even minor head movements allow
you to localize the instruments by sampling different angular
presentations (i.e. the HRTF effect) and analyzing multiple wave fronts.
This depth of information is simply not captured in a stereo recording.
That is the information that is missing; that's the information that
allows us to establish accurate positional data.


snip

Well, no, the point is it still must be an illusion, because the
information is not in the recording.

PArt of the infomation is.

Part, yes. That's the point. Is there enough to make a pretty
convincing reproduction? Clearly yes for a vast number of folks.

You're making an error 180deg from Gary's
error, but stil an error.
Due to projection from higher to lower number of dimansions part of the
information is lost but part is still retained.

The part that is lost is the part that allows us to "make sense" of the
reverberant sound field. Similar to the failings of binaural; the
reverberant field is there, but from a single fixed perspective which
defeats a lot of our ability to localize sounds as we do normally.

I agree, it only need be "sufficient" to fool the listener. But it
simply cannot be "the same".

But "the same" is not needed. Our ears have finite resolution and our
brains are sesnsitive obly to some parts of the signal.

Absolutely true. Now, if we were talking about using that information
and tailoring the reproduction to accentuate the parts to which are most
sensitive (e.g. perceptual coding), no problem. We're not though.
We're talking about a brute force approach that adds a lot of spacial
information that is not related to the spatial environment of the venue.
Hence the whole concept of the reproduction as a "separate work of art".


Reflecting the sound cannot, except in the context of listener
preference, ameliorate this constraint.

It's not staright out prooven either way. But I'd say it's rather
improbable. But I'm open to be shown otherwise. That's why I wanted to
see a theory not a nice trick. Theory which would explain that the
needed clues are in the reproduced signal and distractions are either
masked or attenuated enough.

I agree that a real theory would be nice. I think, however, that it's
pretty clear that taking the entire signal, direct and reverberant, and
adding additional phase shifts by reflecting off the front wall, is a
totally indiscriminate approach.


All subject to individual listener response however. Since the
reproduction *must* be different, and must present a different HRTF
than the original, it has to be listener dependent.

Only to a point. Listeners are humans not superbeings and all have their
limitations.

I disagree. Listener limitations are the issue. If we were superbeings,
and all heard "perfectly", then we could, indeed, create some
universally recognized paradigm for perfect reproduction (might not get
there physically, but it's theoretically feasible).

If you have an device capable of running 30Mph you can
outrun any human going on his feet. IOW listener dependence has its bounds.

I don't think that's true, in this context. Aural limitations that *I*
have are more likely, IMO, to make me prefer a different presentation
than a 20 year old with perfect hearing. IOW, I don't think there can be
a universal standard of sufficiency, precisely because of listener
limitations or preferences.


Can you create a model that is statistically "better"? Certainly. But
then, you must understand, that statistically, *low* bit-rate MP3's
are sonically fine.

And high bitrate Oggs, Musepacks or AC3s are sonically indistinguishable.

That's not the point though. Even ignoring that there are likely those
who would disagree with that characterization. The point is, yes, you
can do lot's of things to improve the reproduction, things that may,
statistically be considered better - possibly by a large margin - but
that does not a "paradigm" make.


snip

In fact real properly[*] recorded events are miked at a distance closer
than a typical listener is. Moreover mikes are typically high in the
air, so they get early reflections primarily just from the floor and not
from all the close surroundings of typical listener (as there aren't any
up there).

Yes, and how does reflecting these floor reflections - that arrive at
the listener from a specific incident angle - from a totally different
incident angle, during replay, provide an accurate representation?

First, one have to ascertain what acuracy of incident angles is really
needed. That's in fact a part of what I miss from what Gary presented.

Yes, but using his approach, it doesn't matter how much accuracy is
*needed* because there is no selective filtering of any kind being applied.

Second, those floor reflections in real life listener position (i.e. not
10 feet allmost above conductors head) get rereflected as well.


Stereo recordings recorderd from a typical listener position
do not sound too spectacularily. This is (partly) because that sound is
then replayed at listener venue where there are additional reflections
(nobody listens in anechoic chamber). So good recording already take
into account those additional reflections. Thus additional reflections
are often 'unnatural' -- they contain peaks due to room shape and
dimensions (the incorporate replay room info), in case of box speakers
they are much damped in the highs, etc...

I would basically agree. But, "taking into account those additional
reflections" means altering the recording to account for the playback
medium and venue. Such tailoring must make the information on the
recording inaccurate relative to the acoustic in the original venue, no?

Well, this information is accurate at the miking position(s). Only that
position is choosen

OK but then you are not talking about "taking into account" the
additional reflections in the context of adjusting the recording to
compensate, right?


Almost analogous to an RIAA pre-emphasis and de-emphasis without a
reference standard.

Somewhat, I agree. But even without pre-emphasis standard no preemphasis
at all was worse.
Similarily, recordings miked from typical listener position tend to
present unimpressive audio scene rendidtion.

I read your "taking into account" description above as meaning you
adjust the recording to compensate, hence my analogy of RIAA. It appears
I misunderstood you.

Keith