On 2012-09-12, Augustine Leudar wrote:
thanks Fons
Adding to Fons, who is pretty much always on the mark...
another question is - is our auditory system able to take account of
the curvature of a wavefront to estimate the distance of a source or
does it only use other cues such as the spektral content of the
sound etc etc
A stationary listener will only sense the curvature if the source is quite
near, less than half a meter or so, depending on the spectral content.
The reason this can be heard is because it necessarily gets reflected
into HRTFs and any other free field description of how people hear
sound. Including both WFS and ambisonic. In the free field, HOA once
again reduces to a spherical version of WFS, so that they give the
similar kinds of clues to a single central listener.
The area where they do so best differs, as do the precise errors they
produce outside of the sweet area: for WFS the sweet area is a narrow
tube at any angle, at a fixed distance, from the line array, and if more
lines are used to break the symmetry, then in a single point. In the HOA
case, you always have just a single sweet point from the start even in
pantophony, but the errors around it are circularly symmetric, and much
more better behaved. It's just that while ambisonic is easier to do at
low order, it's suddenly much more computationally and physically
challenging to do to orders which approach true holophony; even just
pantophonically.
For larger distances a moving listener can detect the curvature, but
only indirectlty as the apparent direction of the source changes.
As I said earlier, that is usually dubbed "auditory parallax". In
analogy with "visual parallax" where you can infer lots of stuff by
moving and looking at stuff from different angles. The analogy is
perfect: you can't infer that stuff even with your eyesight simply by
rotating your head; you actually have to move.
Also note that 2-D WFS systems are optimized for a particular speaker
line to listener distance, it's not possible to get exact wave
synthesis every- where (it is possible for 3-D systems).
I don't remember who it was who once lectured about these systems here
in Finland, on TKK, but I think I've talked about this problem in the
past: he claimed that it was a basic problem of ambisonic that there is
a distance term in the equation. But that then only goes for pantophony,
not periphony. It's one of those funky little low dimensional topology
thingies which snipes you in the back: in 3D, if you do the line array
of WFS or the circle of pantophonic ambisonic, you will necessarily
suffer from that precise 1/r attenuation in distance, even in the
horizontal plane (WTF becomes a line system, ambisonic becomes a point).
But if you add the third dimension, as ambisonic easily -- even
canonically -- does and WFS does not, that problem goes away not only in
the horizontal plane but in a full sphere around the sweet spot.
And anyway the things will not work correctly if you are too close,
less than 5 times the distance between the speakers or so.
Exactly as Fons says: there are going to be interference problems there
even at the lowest frequencies. Though I should add, there are going to
be interference problems anyway, which you will have to
psychoacoustically optimize away using a ton of signal processing. As
Ralph or somebody once suggested, a well-temporally-shaped 20ms burst of
white noise convolved with the frequencies above the temporal Nyquist
limit of the right ought to do the job.
Not really sure if Ville Pulkki's research ever got to this one, but it
too could be relevant, once again.
--
Sampo Syreeni, aka decoy - [email protected], http://decoy.iki.fi/front
+358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2
_______________________________________________
Sursound mailing list
[email protected]
https://mail.music.vt.edu/mailman/listinfo/sursound
