Sampo Syreeni wrote:
- Backward-compatible (to stereo) forms of Ambisonics are very
probably possible.
They are not "probably possible". That's what the BHJ version of UHJ
*is*.
Yes. But you still have to put L/R + 1 or 2 ext. channels into some <
stereo > MP3 or AAC file. Our aim would be to be backward compatibility
to common file and stream stereo formats. This is not completely
trivial, but can be done. (Extension channels "buried" in file or
container format. We don't want to break existing stereo decoders...)
It's as far psychoacoustically optimized for that role, at least given
it's underlying Makita theory, as any passive, linear matrix *can* be.
UHJ was always a hierarchy scheme. Even so you have to find some space
for T and Q channels. It is all just about some "practical space
problems", not Makita theory itself... Just IT and standard related
stuff, not math! ;-)
I have proposed this idea some time ago. (At 1st order the proposal
is based on 3/4 channel UHJ, so on some LR-TQ channel "stereo core +
extension" scheme. This concept could probably be extended to HOA, as
well.)
Yes, it'd be interesting to see how the compatibility format idea
might be extended to more modern settings. I think I've talked about
that for a bit in the past, myself. But of course, I've just jotted
out something; it'd be nice to see it fleshed out for real. Especially
in the original Gerzon vein of building out a complete, mutually
compatible, extensible hierarchy of both channel encodings, and
speaker playback formats. :)
Agreed! (What?!) :-)
Isn't it a bit late for 2-channel UHJ decoders anyway? Some radio
broadcaster already could try to apply some < multiformat > surround
to binaural decoders on the "customer" side.
Well, do they do anything as nice as BHJ right now? Do they even try?
Do they even have something like this on their future map? I don't
think so. As such, I for one -- as a newbie on-list, and perhaps even
a pure theoretical buff with nigh to none practical experience --
still think the ambisonic way remains useful. Highly so, even if the
basic framework could and does benefit from newer innovation. (Say, my
alltime favourite decoder, DirAC, and its newer, higher order, and
more principled derivatives.)
DirAC has the problem that it currebtly can be used "just as research"
tool. (I won't discuss this here. But you know this as well as I do. )
5.1 and Ambisonics should/could be both supported, and maybe more
formats.
That is an interesting line of thought for me as well. Undoubtedly for
many on-list. How do you really bridge common discrete formats such as
5.1 or 7.1 with the ambisonic framework? Also, in the high audiophile,
research mindset which engenders HOA and (as shown by Daniel and
Nichols to be in the limit equivalent, I seem to remember) WFS.
I think the proper formulation would start from idealized far field
(planewave) response, angularly evened out to whatever ambisonic order
we choose to work with. In fact I believe when I and Martin Leese did
our little thingy with OggPCM channel formats, we were working on that
precice if implicit assumption: that the directionality inherent to
most discrete formats could be modelled within the ambisonic framework
as distant, infinitely precise in direction, sources, to be folded
down into an ambisonic soundscape, and from there on to be (optimally)
decoded as usual.
Or maybe so that knowing some source is tagged as a point source, it
might be separately decoded via something like VBAP, while the rest of
the field is reproduced by ambisonic principles. Dunno.
Maybe. But before we "just" would need some good binaural decoder for
some standard surround formats; 5.1 and FOA surely included.
Until 2morrow, and many thanks for your thoughtful and constructive
posting! (I have to think about this quite a bit more....)
Stefan
-----------------------
Stil, be as it may, the idea of folding all of the current spatial
audio into one big theory isn't new. I certainly didn't invent it, nor
did anyone on this list. There's a lot of history behind the idea,
perhaps not starting even with *Gerzon* himself. We'd all like to see
the theory developed more fully, but I think we can all agree, it's
not exactly a new invention, or ours, even if we were to be the ones
to bring it to its fruition.
(If format wars are not necessary from a modern perspective, we
should not try to continue some partially imagined conflict.
(Software) format wars go away once there is simplistic, free, open
sourced software to "just do the thing any which way and convert". Cf.
Eric de Castro Lopo's libsndfile; it pretty much took away any problem
with differing sound file formats, once it reached its current maturity.
So, I'd say that the same would happen with ambisonic, if we had a
free, comprehensive library to deal with our favourite sound system.
Well documented, proven in practice, and made easy to integrate.
I believe we can all agree on what that sort of thing should look
like. What the interfaces should be, and so forth. But to date
nobody's actually implemented the thing. Parts of it, sure. Even
multiple times over, for fame and other individual reasons. But I
think it still remains the fact that we don't have "that one
definitive, all-encompassing, one, ambisonic library", which everybody
*just* *likes* *to* *use*, and then *does*.
I believe that's what'd settle the matter in the end. Not abstract
theoretisising, or even particular efforts to implement some forms of
the system, but a concerted effort to put all of the knowledge the
ambisonic community has, into one easily used, well-generalized, and
well-maintained, piece of code.
Ambisonics has some clear advantages in the areas of AR/VR and 360º
video. So some natural application cases finally exist.)
Yes. If you want in-the-wild examples there, take a look at Peter
Hajba's work (aka Skaven, of FutureCrew, in the demo circles). And of
course then Ville Pulkki and his research group's work at Aalto
University. Headtracked, simultaneous audio and "VR" is *awesome*! :D
Even HTML5 surround decoders are possible by now:
http://hyperradio.radiofrance.fr/son-3d/
Nice. Will look into that.
Marc Lavallée presented one online player/decoder for XYW recordings.
Now somebody would "just" have to do some integration.
Yes! Who would be that someone then? ;)
funding from the European Commission under the Horizon 2020 programme
Hmm. An interesting idea, at least for me... Would it be possible to
get that "one, definitive, free and opensource ambisonic library" I
alluded to above, done with public money? Some invention grant from
the EU or whatnot?
Because, I mean, such funding sources like demos. The ambisonic
community has some darn impressive demos already, to just give out.
So, such funding -- even if we all know where it once went with the
NRDC -- might be rather readily available. If only somebody bothered
to jot down a well-thought out grant proposal.
This really oughtn't be rocket surgery...
ORPHEUS will deliver a sustainable solution, ensuring that workflows
and components for object-based audio scale up to enable
cost-effective commercial production, storage, re-purposing,
play-out and distribution.
"Object based audio." Yes, well, it might be that nowadays with all of
the power of modern processors, you can actually do mixed and
synthetic audio at what amounts to "infinite order" in ambisonic
terms. Pure point sources, panned around.
But in at least two respects the ambisonic theory/framework still
reigns supreme.
First, you can't get past the first, most basic theorem of ambisonic
decoding: if you want to decode *anything*, including those point
sources of today, into a sparse speaker array, you'll have to
psychoacoustically optimize what you're doing. Then even the most
basic Makita theory Gerzon started out with tells you, you simply
cannot do things like VBAP without losing isotropy.
Fine, maybe you don't want that every time. Maybe you want to do stuff
which risks exposing speaker locations, as the price for more spatial
definition. That's certainly what Dolby does right now with Atmos. But
then we have point two:
Secondly, and most forcefully, there is absolutely no other way to
capture a live soundfield faithfully other than some version of
ambisonic. All with its unique soundfield mic designs. Even if you
somehow conventionally miced and mixed stuff just right, so as to
mimic a live soundfield, there *still* is no extant framework within
which to fully process directional sound, absent ambisonic. It's all
you've got -- and because of its underpinnings, all you'll ever have,
with complete mathematical certainty.
As such, if this shit was to be packaged in a neat, free, open
library, in its most general and generous form, I believe it'd easily
become the new norm. Because there just wouldn't be any alternative to
the comprehensive theory and practice which goes along with it.
I think the only question is "whodunnit". ;)
Of course they could (or should?) start to implement 5.1 and
Ambisonics decoders firstly. But < who > would fund < this > stuff?!
Then, does anybody actually need to fund this, even? All of the stuff
we need is already there. It just needs to be packaged right. Even the
hardest part, the nonlinear decoder optimizer code, already exists, at
least in Benjamin, Lee and Heller's Tabu search code (which is general
as fuck). And all of the movers and shakers pretty much are on this
list, or once removed.
I don't think it's even about funding. It's about simple organization.
Perhaps with a few egos soothed via visible credits, and such. But not
much more.
"Wait until 2020. All will be good by then." (It is also some
classical Ircam strategy to point to the benefits of the < next
generation >. We never will be < there > , but never mind! :-D )
As dyed-in-the-wool individualist and one nasty libertarian, I'd say
*still* fuck IRCAM. Seriously. it's a piece of age-old French
dirigisme, before it ever was a place of learning and innovation. I
feel sorry for the (grantedly many) excellent, sonically and signal
processing minded individuals who had the unfortune to land in said
hellhole. Had they ever seen a community like that of MIT or Caltech,
they'dd have shat on IRCAM's very name, and after that done *much*
more for themselves and humanity than they now have.
Now the modern likes of Serre et al. are doomed to basically live in a
black hole. An insular island removed from the scientific and
technical mainstream. People kind of fall into IRCAM, because of its
attractions. Once they do, they fall beyond that particular,
institutional, event horizon. Nothing ever comes out from there, we
don't know what the fuck goes there beyond their (rather impressive)
toys, and what little comes out is...pardon me, garbled French.
So: We don't have to be < too > conservative. (We also don't have to
wait 'til "2020" to start with something.)
Well agreed.
P.S.: And quadraphonic formats are maybe interesting - but quite
obviously surround history. Some recordings still exist, but there
won't be anything recent or new "stuff" around.
Yes. But history is still worth preserving. And I don't just mean on
the recording side. I also think the theory and the practice of
quadraphonic technology is worth preservation; if only because it
serves as a counter-example to the more developed theory which is
ambisonic.
"Alice's Adventures In (3D Audio?) Wonderland".
I can remember listening to this play on BBC Radio 3/4 several times,
great stuff.
You're privileged in having (had?) access to this sort of stuff.
Around here, and in my age cohort, none of that experimental radio
stuff has *ever* been around. I grew up thinking stereo is all there
could possibly be.
I still have some off-air recordings of the BBC's Matrix H system.
Never tried to decode them but they have a pleasing surround
sensation when listened to on headphones. Must have been the mic.
placements ??
Richard, mind digitizing and sharing them? I'd be rather interested in
the stuff.
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound - unsubscribe here, edit
account or options, view archives and so on.