Re: On 8 speakers


Subject: Re: On 8 speakers
From: DG Malham (dgm102@york.ac.uk)
Date: Tue Jan 26 1999 - 07:00:58 EST


On Mon, 25 Jan 1999, Chris Rolfe wrote:

> I wrote the software for the system Darren and co. are using, so maybe a
> word or two here about design philosophy wouldn't be out of place.
>
> The user can specify angle placement for any given input. The speaker map
> is edited separately, the idea being that you might want to pan a sound
> from 10 degrees to 190 degrees, and not be worried about speaker location.
> Speaker maps are simple graphic documents that record the current venue
> speaker locations.
>
> On the other hand, sometimes the allocation of an input channel to an
> output channel is important, say an assign to global reverb. In that case,
> you're free to override the speaker map and assign directly to the output.
>
> The idea was to liberate the diffusion composition from the particular
> placement and number of speakers.
>

This is one possible approach to the separation, but it appears that it
would still recquire either prior knowledge of the replay speaker layout
or re-synthesis of the piece at the concert venue. As mention by Joe
Anderson, Ambisonics provides a methodology for completely separating
composition and performance speaker layouts, recquire few channels on the
storage medium (only four in total for full 3-d images of arbitrary
complexity in first order systems). Although, in its first order format it
is not possible to send a signal to a single speaker channel only (all
speakers cooperate to form the soundfield) in compensation, apart from the
composition/performance separation, you get simpler tools for doing
complex things like field rotations and, because all speakers cooperate,
you get more acoustic power out of any given rig. Of course, individual
speaker feeds can be incorporated - even on a single Adat tape you have
four spare channels - and if you go to higher orders, for instance third
order which needs 15 channels, by exploiting various characteristics of
the system you can get sufficient separation between adjacent speakers in
an eight speaker circular 2-d rig (or 20 speaker spherical 3-d rig) for
them to appear to have individual feeds if necessary.
 
> The biggest trick is to get composers to work in 8 channels. We're all so
> used to reducing to stereo that we fail to compose in space. The greatest
> mistake I see is that composers tend to reduce everything to a "flying
> object", when in fact there are very few sounds that pan successfully. 8
> discrete channels are, however, a wonderful thing to hear.
>

Although composers have been working (conciously or unconciously) with
spatial elements probably for as long as we have had music, it is only in
this century that we have acquired the technical ability to explore many
aspects of this. The development of the musical language appropriate to
the concious compositional use of spatial elements is still in the
formative stage. However, once you start becoming fully aware of the parts
of that language that are already in place, you will find it appearing in
many unexpected places. Next time you go to a non-ea concert, try
listening not to what you might now regard as the music, the
timbres/phrases/notes whatever but instead listen, as Vareses did in the
1920's, to the movements, the physical structures in space - no matter
what piece it is, you may well be suprised.

> As far as the speaker layouts proposed (circular vs. stereo side/side) I
> would suggest that the difference is trivial. While there are some
> spatialization techniques that rely upon equidistant speakers (favouring
> the circle approach), 99% of the audience is outside the sweet-spot, and so
> a few degrees here or there doesn't matter. In general, the equidistant
> circle is preferable, but the rectangular-shape doesn't irreparably damage
> a well-diffused piece.
>

Oddly enough, we find that getting it right increases the sweetspot area -
because the actual 'spot' is sweeter it can be traded off against improved
performance for those not in it.

> Remapping DVD (5.1 and relatives) is a different question, of course.
> Generally, DVD is "super-stereo", that is, stereo tracks are de-correlated
> to provide max bang-fer-buck for the high-end home theatre. The
> de-correlated signals are, however, usually manufactured from the original
> stereo+sub track.
>

This certainly applies to the video version of DVD we have now, but the
audio only version (DVD-A) will (hopefully) support Meridian Lossless
Packing as a recquired part of it's specification. This will allow many
more channels, higher sampling rates and more bits (typically this might
allow 96KHz, 24 bit and eight channels as opposed to only 2 uncompressed
channels at 96KHz 24 bit on current DVD) with the audio signals possibly
being a decode, speaker feed signal version of Ambisonics called G format
which means the ordinary user will get excellent surround sound over 5
speakers without needing a decoder, but the original B format signals will
be recoverable for feeding to a dcoder which can then feed a higher number
of speakers for better performance. For more info see

http://www.ambisonic.net/dvda.html

        Dave

/**************************************************************************/
/* Dave Malham "http://www.york.ac.uk/depts/music/dgm.htm" */
/* Music Technology Group "http://www.york.ac.uk/inst/mustech/welcome.htm"*/
/* Department of Music "http://www.york.ac.uk/depts/music/welcome.htm" */
/* The University of York Phone 01904 432448 */
/* Heslington Fax 01904 432450 */
/* York YO1 5DD */
/* UK 'Ambisonics - Component Imaging for Audio' */
/* "http://www.york.ac.uk/inst/mustech/3d_audio/ambison.htm" */
/**************************************************************************/



This archive was generated by hypermail 2b27 : Wed Jun 11 2003 - 13:08:50 EDT