Re: [LAU] ambisonics: fons ambdec and muse.demon.co.uk ambidec

From: <fons@email-addr-hidden>
Date: Fri Jan 07 2011 - 17:15:39 EET

On Fri, Jan 07, 2011 at 01:21:37PM +0100, Giso Grimm wrote:

For setting up a _convinving_ demo system the best advice I can
give ATM is:

* Use ambdec with one of the regular layouts for which a configuration
  is provided.
* Use a set of identical speakers, of reasonable quality.
* Carefully check and adjust levels.

> What is your literature recommendation for those who want to understand
> before they are setting up unconvincing demo systems? Can you give
> specific recommendations for
>
> - overview on theory

If you really want to fully understand things there is no way besides
diving into the maths. This is in particular the case for higher order.
The PhD theses of Daniel and Moreau are good starting places in that
case, and understanding what is explained there is more or less a
requirement before you can go on reading e.g. some of the proceedings
of the AMB symposia, the work of other researchers, etc.

What does not exist AFAIK are publications that translate the maths
into more intuitive language. For the good reason that it is *quite
difficult* to do that without going astray, much more difficult than
for e.g. general DSP theory. I've tried to do it many times in posts on
this list and others, and in private communication. Once people start
asking the right questions it can be very hard to avoid the maths, or
to present them in more consumable format.

For me, any intuitive, 'common language' understanding I may have is
not the result of just knowing the theory but of working with it -
using it to calculate or simulate things and trying to understand the
results in practical terms. That takes time, but I don't think there
is any alternative for it.

Compare it to your knowledge of ancient instruments and of the
interpretation and performance issues of the music composed for
them - it's not something you can acquire in a few weeks. And it
can't be explained without referring to lots of other knowledge
including in this case music theory and history.

> - finding decoding matrix for regular layout

This more or less follows directly from the theory.

> - the same for irregular layout (probably the mentioned one?)

This is *black art* at the moment. There are no general purpose
algorithms to do it. That does not mean you can't find a decoder
for a given layout, but it will be a compromise far below what it
*could* be. The methods I use are a combination of semi-automatic
procedures and manual tweaking with feedback from simulations.
The latter becomes quite hard as order goes up, I wouldn't even
try anything above 3rd order that way.
There is some work being done to develop automated systems, e.g
by Bruce Wiggins (univ. of Derby) and Franz Zotter (IEM, Graz),
but it remains quite experimental at the moment.
 
One thing that is clear to me is that irregular layouts can be
optimised only if you have higher order input. The information
to do it just isn't present in e.g. first order recordings.

For example, you can reproduce first order on a 5.0 layout,
and it works, but result are *much* better for second order.
And 5.0 is more or less the simplest case to consider.

> - subjective performance compared to other systems (localization,
> quality, sweet spot etc)
> - practical issues (maybe the many LAC papers?)

The best reference here IMHO are the publication of Joern
Nettingsmeier who has probably more experience with real
life (in particular large) systems than any practising
sound engineer I know. I expect he will present some more
at the next LAC.

Ciao,

-- 
FA
There are three of them, and Alleline.
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Fri Jan 7 20:15:03 2011

This archive was generated by hypermail 2.1.8 : Fri Jan 07 2011 - 20:15:03 EET