On Wed, Dec 12, 2012 at 07:20:31PM +0100, Natasha Barrett wrote:
> 1. Proximity illusions on WFS at IRCAM: all four sides WFS arrays
> (BTW flat along the walls), sounds appear beside 'you', but then
> 'you' have to be in the right place. Kind of like one sweet spot
> for each sound point (and of course you have to be on the correct
> side of the traveling wavefront). Therefore there are mixing and
> composition tactics to get the best out of WFS proximity illusion.
> It sounds super, but as all spatialisation is in real-time you loose
> the advantage of HOA's encoding-decoding separation.
The 'having to be in the right place' or at least the right side, is
probably unavoidable, whatever the system used. Focussing, no matter
how it's done, does not create a source.
> 2. Proximity illusions on HOA at IRCAM: I worked with 7th order 3D
> over 75 loudspeakers and 12th order 2D over 25 loudspeakers. HOA
> alone will not, and is not supposed to, make *focused* sounds inside
> the array. HOWEVER, with the orders I worked with there are tricks to
> with sound motion, mixing, contrast, sound sources etc that gives the
> illusion of 'something' inside the array. But remember that this is
> to do with artistic use of the sound rather than embedded in the
> technology itself. And of course these high orders gave an enormous
> sweet spot / stable listening area! Completely different sound to WFS
> and it was great to use HOA and WFS at the same time! In fact quite
> depressing to return to my own lower order composition studio.
>
> I also made practical tests with implementations of published NFC-HOA
> theory (near-field compensated HOA). There is definitely something
> interesting perceptually - a 'kind of' focused source inside the array
> effect, not as spooky clear as WFS, but heard for the complete audience.
> With NFC the bass boost is still an issue even with compensation filters
> that end up removing the NFC effect (and so the technology still has some
> way to go).
Indeed. HOA can actually produce focussed sources just as WFS can, but in
practice it is not an easy thing to do.
The theory you'll find in Moreau's PhD thesis, dated 2006.
A practical implementation is quite another issue.
Let's trace the complete processing chain:
1. Mono source signal
A. AMB panner with near field encoding of distance (optionally
compensated for the size of the reproduction array).
2. HOA 'B-format' signal
B. AMB decoder
3. Speaker signals
C. Speakers
4. Reconstructed sound field.
(where 1,2,3,4 are 'signals' and A,B,C are 'processes' that convert
signals into some other form)
The problem is that for focussed sources (2) and (3) will have very high gain
resp. levels at low and mid frequencies, essentially the response rises by 6 dB
per octave down per order. So for example a 10th order AMB signal for a focussed
source will have components that rise in gain by 60 dB per octave as frequency
goes down. Compensating for the reproduction array size in (A) instead of doing
this in (B) mitigates the problem, but does not remove it.
In practice that means that for focussed sources the speaker signals will be at
extremely high levels, half of them in antiphase w.r.t to the rest, and most of
the acoustic signals produced that way will cancel out in the reconstructed
field.
That cancelling out means that most of the energy produced by the speakers is
actually useless. To avoid this problem the frequency range of the higher order
AMB signals (2) must be limited at the low end. And that must be done in a way
that preserves their relative phase relationships. While the theoretical near
field encoding filters are quite simple, adding the required high-pass component
complicates them by some orders of magnitude, in particular if such filters have
to be variable in real time as you would expect from a 'panner with distance'.
As shown by Moreau, such filters would actually limit the otherwise extreme
gains/levels to less than +6 dB, quite harmless in practice, while fully
preserving the focussing effect. Most of it is actually the result of the
_phase_ response of the filters, and not of the gain.
I'm so far not aware of any existing implementation that actually achieves
this, except 'research' code for a fixed encoded distance and reproduction
array size.
Some people (including me) are working on this, so maybe some day we'll have
something that can be used for actual production work.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound
