A “Home based” Holophony approach based on WFS
Helmut Oellers Mike Mazarick
Erfurt, Germany Raleigh, NC USA
oellers@syntheticwave.de mazarick@bellsouth.net
application. This experimental prototype
will try to overcome some of the
Abstract obstacles.
Linux - free of license costs and with a We have seen the steady march of
freedom to modify and customize limited improvements while pursuing this dream
only by the imagination and - from mono recordings, to stereo, to
determination of the experimenters, quadraphonic sound of the 70s, to 5.1
opens new opportunities for regular sounds, and to 7.1 recordings of today.
people to have the possibility of But the spatial structure of the recording
developing astonishingly sophisticated room sonic field is too complex for
projects. This paper tells of one such an transmitting the spatial information in a
attempt. A German sound engineer limited number of channels.
together with a U.S. Linux hobbyist
makes an attempt to change the audio
world by an experimental setup with
cheap PC speakers and some Gadget
Labs sound cards (inexpensive semi-pro
cards made by a long bankrupt
company). The efforts are directed by a
collection of downloaded Linux programs
and glued together by a few simple, self
developed scripts. It seems like a funny
attempt at first glance, but let us
describes the underlying idea. You may
change your opinion.
1. The goal of the experiment
The goal of the experiment is the proof Mirror sources in the recording room
of concept for an idea, which arises by produce a complex reflection pattern.
the authors since more as fife years; Because of individual filter system of our
restore the spatial sound field in all three head and body, we cannot simply
room dimensions alone by a loudspeaker reproduce those reflections from a new
screen in front of the listener. The location. The only possibility for true
listener shall be acoustically transported spatial audio seems to be synthesizing it
to the virtual recording room. from the pure audio signal and recording
room properties.
The desire for synthesizing loudspeaker
sound, indistinguishable from the In principle the WFS takes this direction.
genuine sonic field has long been a goal, The spatial structure of the sonic field
especially with the continual and fast becomes reconstructed in the same
developments of a wide range of audio manner as the genuine wave fronts and
hardware. The principle of wave field its reflections in the recording room.
synthesis offers just such a possibility of
reaching this goal today, even though But such attempt would require tightly
there are existing practical constraints spaced loudspeakers all around the
which are preventing its widespread listener. Besides, the underlying
1
Kirchhoff- Helmholtz- Integral needs a
source free volume - an anechoic
chamber. Planning on constructing this
is not a good start for achieving the “wife
acceptance factor”.
A different idea as presented in this
paper provides a more acceptable
solution for the home. This is by taking a
different systems approach. The goal of
the procedure isn‟t based on providing a
perfectly exact loudspeaker signal, but
instead, providing congruent signals
between a defined position of a virtual
listener in the recording room and a
dedicated listener position in the home
cinema. By using such different But there is also another commonly
approach, the playback room acoustics known “political” problem. Many
isn‟t a disturbing component in the unorthodox ideas have almost no chance
transmission chain, but an appreciated for realization as long as the source of
element in a common system. the solution isn‟t a famous research
institute, a global player with “deep
pockets”, or a known name brand that is
The inclusion of playback room acoustics
acknowledged in the field. That problem
not only eliminates the need for the
increases when there is an increased
playback room to resemble an anechoic
amount of development to realize the
chamber, it allows speakers mounted in
solution; all the really simple solutions
a two dimensional plane to accurately re-
are already been done today. Many
create a three dimensional sonic space
formidable and comprehensive privately
surrounding the listener in all positions in
developed ideas today starve on the way
the room. Virtual sound sources on
to their realization.
either side of the speaker plane could be
reproduced by nearby accurate positions
in all room dimensions, so that a sound Fortunately, this may not be a problem
field envelops the listener from all for this idea. Like most good ideas,
directions. there was an element of „chance‟ and
„circumstance‟ that makes this whole
idea feasible.
It would also get rid of the familiar
limitation that the reproduced sources
originate only in the horizontal plane of 2. What do we hope to accomplish?
the listener. This is one of the most
audible disadvantages of other sound We want by the Linux based experiment
reproduction methods, and is also a to show the principal verification of the
problem with the wave field synthesis way to build up a complete Holophonic
that have horizontal loudspeaker rows. reproduction of the genuine sound field
by the Wave Field Synthesis.
The solution of that problem was
accentuated often in the scientific world In principle the Kirchhoff- Helmholtz
in recent year‟s i.e. DAFx^04 discovered Integral describes such a possibility in
that problem as the main disadvantage case that the sound pressure and
of the WFS procedure: velocities are known at the surface of a
source free volume.
By Raileigh II that requirement can be
reduced to knowing only the pressure, if
we allow sound pressure outside of the
volume. But this approach also would
need the loudspeakers to be placed all
around the listener.
2
So the underlying idea of our experiment After the XP driver was developed and
was, substitute the loudspeakers all open sourced, was done, it became
around by the controlled reflections of obvious that an Alsa Linux driver was
the waves on the playback room walls. feasible. A small project was started on
SourceForge, and two really strong
Such a way is in use for a long time by developers were recruited to help the
the known sound projectors to simulate effort – Dave Marion in Utah and Rick
fake loudspeakers. But they never take Macri in Melbourne, Australia. Dave
the attempt to fake the sound sources Marion had come up with a novel way to
including its first reflections in the write/debug the driver, which involved
recording room until now. That‟s the core breaking it down into sections or
of our Linux based experiment. modules with a definitive purpose.
However, it was discovered later that the
coding standards didn‟t match the other
drivers in the Alsa project (such as, all
code should be in one module and all
3. Thumbnail History of the Gadget
comments have to be removed from the
Labs card, the very small Linux
code for starters). Since there were so
project to create an Alsa driver for
few people who could test this card (six
it, and how this all led to a home
people to date have attempted the Linux
Wave Field Synthesis system
driver and all have been successful),
there hasn‟t been the critical mass or
Gadget Labs was a small startup founded groundswell of support for converting the
in 1998 in Washington State by former GL Driver code (which now works well)
employees of Intel. They sold „state of to be part of the regular Alsa tree and
the art‟ audio systems over the internet included in the Linux kernel. During the
using Win98. Their flagship product has testing of the driver with a cheap set of
8 channels and operated at 24 bits, 44.1 PC speakers, it was noticed that it could
and 48 kHz. When the VLSI chip that only support about 16 inputs in one
drives the Delta 1010 and Hoontek cards system (microphones, instruments,
arrived, they decided that there were etc.), but it appeared to be able to
many other solutions available to this support virtually unlimited outputs. The
problem and could not sustain the Linux driver also supported multiple
company. Gadget Labs went bankrupt cards and a control panel for each card.
in 2001. A mailing list was started on Now, what could possibly be done with a
Yahoo Groups allowing the owners of GL large number of outputs? The German
hardware to provide „self help‟ and author was invited to take part in the
support. very small list of people following the
development of the driver, and everyone
Out of this effort, a „mystery man‟ responded well to his efforts with
(actually a student at the time) familiar advanced audio systems, even though
with the „black art‟ of reverse no one was familiar with WFS.
engineering hardware/software arrived
on the scene. He went by the This is the reason for the American
pseudonym „Mostek‟. He was able to contributor to this project proposed
reverse engineer the drivers that were developing a Linux based low cost
left behind, making them rock solid for solution. By using some old GL cards
audio. Eventually, he completely reverse and cheap PC speakers in order to prove
engineered the entire card and the Altera the concepts, along with a special
firmware and created a Windows XP arrangement of the speaker placement
driver. He released the code and the in the room (set up to take advantage of
new firmware to public domain. This is the room acoustics), the number of
described at this link and the speakers required for WFS was greatly
software/firmware is still available: reduced. The goal of the prototype isn‟t
the tonal accuracy. The goal of the
http://mostek.subcultural.com/ prototype is to study the true spatial
distribution of the sound sources and its
mirror sources in the playback room in
3
all three dimensions using wave field room impose a new acoustic behaviour
synthesis with a two dimensional (in a on the sound that was originally
plane, on one wall) frontal loudspeaker recorded in another room with different
arrangement. properties.
6. Near field solution
5. The problems of the horizontal The only possible solution for avoiding
WFS- loudspeaker rows the unwanted impact of the living room
acoustics (also called the playback room)
Wave field synthesis today is commonly is near field reproduction. Two
achieved by a horizontal Loudspeaker possibilities exist for that, either the
row surrounding the listener. Such speakers are placed near the listener or
solutions are possible today with some the radiating speaker surface must be
hundreds of loudspeakers without very large.
unbearable problems. But a broad based
breakthrough seems not to be within One of the members of our Linux user
reach. Besides of the acceptance factor, network group, we have mentioned his
some additional problems haven‟t been pseudonym „Mostek‟ above, had
solved yet. The most audible effect is the described such near field reproduction
horizontal limitation. Other procedures, vividly:
like Ambisonics or Vector Based
Amplitude Panning, have shown that a “I've read the stuff you have written on
real three dimensional reproduction of the www.syntheticwave.de and have to
the sound event has been elusive using tell you that around 1995 I've
these techniques. They are also missing been in Munich rooming around big
an advantage that WFS offers - a large sound equipment store.
„stereo sweet spot‟. The Ambisonic and And there was a curved wall made out of
Vector Based Amplitude Panning around 1000 small speakers (1.5m wide
techniques can only be experienced in a and 2m high). There was also a mark on
more limited subset of the room the floor that defined where you must
compared to WFS. stand. So now if you didn't stand on the
exact spot -> lets say you stood 1meter
Though additional problems arise, the away nothing special was happening ->
playback room imposes its own acoustic just normal music playing.
behaviour for reproduction. Loudspeaker But if you stood on the spot you walked
rows around the listener cannot solve "into the sound". It was like you stood
this problem. In order to produce the into a different matter. Like that the air
acoustics of another room, the playback wasn't the air anymore, but the most
room acoustics must become completely fascinating sound matter; as soon the
suppressed. That‟s hardly possible in a music wasn't the music anymore, but a
normal living room. Such a system can fluctuating matter that was all around
be delivered in experimental setups, but you.
without broad market acceptance. I simply couldn't believe what I was
Additionally, horizontal rows of speakers hearing. So that definitely changed my
don‟t really produce parallel wavefronts perspective of sound. And was one of the
which can be steered easily (which is greatest experiences in my life - it's
often what is desired). They radiate burned into my brains for ever.
cylindrical waves. The radiated cylinder
waves lose 3 dB of volume every time So hope your quest will be successful at
the distance is doubled. As far as the the end. And welcome to the group.”
listener is near the speakers, the
doubled distance is a short distance, so In normal living room environments the
the sonic impact is minimized. distance separating the speakers is
around three to four meters. We would
What is worse is that the sound energy need a diaphragm of adapted diameter
doesn‟t immediately dissipate. The for accurate near field reproduction. This
ceilings, floors, and walls of the playback is hardly possible for a single speaker,
4
but by using the wave field synthesis develop a “speaking silver screen” last
principle, all loudspeakers work together year.
as a single unit. If we use a sufficient
number of speakers aligned in a plane The directive effect of such huge
level in front of the listener, for example, resulting diaphragm could hardly feed
behind a picture screen, the different the mirror sources of the playback room
WFS signals from each speaker could because its unwanted influence is
reproduce a curved wave front the same broken. We want to show a way to
as speakers in the curved wall: include reflections on purpose in the next
chapter. This is a key difference with
practically all other wave field synthesis
approaches to date.
7. Including the playback room
properties for Holophonic approach
Holophonic means all wave fronts arrive
in the same temporal and spatial
behaviour as would be heard by the
listener in the original recording room.
The conventional wave field synthesis
approach doesn‟t have a dedicated
listener position. Without this, we cannot
This Graphic diagram shows how much compare between the sound impression
equivalent speaker excursion would be of the playback room and the recording
provide, if the source of the sound was a room. The main goal of the described
point about 1 meter (~3„) behind a large Holophonic approach doesn‟t aim to fake
array of 48x27 speakers the acoustic environment in the
recording room in the playback room.
If a two inch distance between the single The main goal is to restore coincident
speakers was used, such a near field signals on the ears of the listener in
would have a size of be 2.43m (~8‟) x home cinema as like by a virtual listener
1.38 m (~4.5‟), therefore, the near field in the recording room. That would create
condition in a normal home could be the same perception.
reached. The spatial aliasing effects
would occur, dependent from the ingoing (END OF EDITS BY MIKE MAZARICK)
and outgoing angles regarding the
radiating surface, above this frequency: By such different system approach we
don‟t want to depress the playback room
acoustics. The reflections of the playback
room become an adjuvant element in the
transmitting chain.
As an example, a 30 degree angle
difference would result in a 13.5 kHz. It As mentioned above, we want to
is known that our perception isn‟t very substitute the loudspeakers all around,
keen regarding spatial aliasing. which would be need for a complete
Holophony system, by means of the
Technically, such a field would be able to reflections of the playback room walls. It
be produced today; some speaker rows would be hopeless take such attempt for
around the listener have comparably the correct origin of all reflections by the
numbers of speakers. In [1] Prof. Diemer current available computing power. But
de Vries from the University of Delft, one there is no need for that.
of the fathers of the wave field synthesis
principle has actually announced plans to
5
The reverberation contains very
important information‟s regarding the
recording room properties. All the
reflective behaviour, the fine structure of
the surfaces, the timbre of the room
becomes submitted by the reverberation.
But there is proven by Wittek et al. [2]
that is sufficient for the reverberation tail
a reproduction from some different
directions. We cannot assign the third
and later reflection mirror sources any
starting point in the recording room.
On the other hand, the direction of the The outer room 3a is the recording
direct sound source and its first room, inside the playback room 3c.
reflections in the recording room hardly Centre of the geometrically system is the
changing the timbre, but contains the common listener Position in recording
most important cues regarding the and playback and recording room.
spatial impression.
The source 3c becomes restore by wave
Therefore isn‟t reasonable concerning field synthesis in the Speaker field nearly
early reflections and reverberation in perfect by the acoustic curtain. But the
common manner, as broadly done in the problem arise, we must produce the
prevalent wave field synthesis ceiling mirror source in the recording
approaches. For the reverberation tail room far outside of the speaker field.
the convolution into the impulse
response evidently deliver perfect For fake that first reflection source by
results. Though, restore the exact the playback room reflection we must
position of each mirror source, as shift the recording room mirror position
practiced often by the wave field in the common geometrically system on
synthesis causing hardly bearable effort a circular path into the range of the
and computing power. ceiling mirror source of the playback
room speaker field. Simply mirroring
For recreate the spatial impression we that position on the playback room
can more easily use the model based ceiling geometry deliver the final starting
approach. We can simplify the source point 4b for the ceiling reflection, which
positions to a simply calculable model of now arrive relatively correct in time and
the most important reflections. Such direction by the listener.
model deliver the possibility for calculate
the runtimes and levels for each speaker
very easily. Moreover we can in such
manner assembly the recording and
playback room acoustics into the
common approach.
That should become exemplify by that
little sketch by the example of the ceiling
reflection of the recording room:
The procedure fakes now the recording
room altitude. The genuine playback
room ceiling mirror source hardly
becomes feed by energy because of the
directive effect of the huge speaker field.
6
Therefore, if execute the same principle divergence problem
for all reflection walls the procedure
fakes the spacing of the recording room. transformation factor
The timbre of the room then becomes source turn
restore by the convolution of the impulse
response. But we no longer need spatial 10. Remaining problems
impulse responses recorded by a team of
technicians on help of expensive
the way for reproduce conventionally
microphone arrays, recorded in multi
records
track and interpolated during the
playback process for completely restore
all reflection origins. Normal recorded Jumping front source
Responses, as offers by waves and other
less sophisticated sources fully adequate. 11 How the sound tone quality will
be balanced
If we have in addition a roughly
adequate geometrically model of the There is a Linux system for providing
dedicated recording room we can fake whole room correction/equalization
the source almost in the most attractive called DRC. This program provides the
acoustically environments of the world complex equalization parameters that
still. can be used in downstream digital filters
like Brute FIR and set up a method for
providing Digital Room (and speaker)
=============
correction. This same procedure can be
used to determine the optimal cut off
End of editing. THIS IS WHAT HAS BEEN frequencies, since the design calls for 2
DONE TO DATE...
separate channels (actually 3 separate
channels, two tweeter stereo and one for
========================= bass).
8. Simplifications for the “Home
based” approach
12. How the system will be tested
…. Sometimes restrictions a chance, we for the ability to spatialize a sound
must search speaker alignment by a or group of sounds.
smaller number of speakers.
We plan to use a pair of moving head LED
Only first reflections lights to shine on the calculated sound
source location. This has the advantage
……Phantom sources for treble because that there are several ways to determine
of aliasing this value (from the number of „0‟ samples
sent to each speaker, for instance. It
Randomised placing for aliasing would be possible to take two or three
sources and look at the differences in the
amount of delay for a given source and
9. New possibilities for WFS calculate it‟s position accurately. The lights
are DMX controlled, and standard DJ
..virtually walking across the record lighting control software can make them
room point to an arbitrary point. This will be
checked by using two blind people to point
transformation factor to where they perceive the sound is
originating.
Sources:
[1] http://www.dradio.de/dkultur/sendungen/ewelten/664051/
7
[2] http://www.hauptmikrofon.de/HW/Wittek_thesis_201207.pdf
thumbnail for H.: Dipl. wegemann disturbing reverberation portion
10. What will be tested:
Interaural Cross-Correlation Coefficient (IACC) value for demo pieces of music
played in normal stereo format on the WFS setup
Interaural Cross-Correlation Coefficient (IACC) value for music that has been
processed for WFS
Interaural Cross-Correlation Coefficient (IACC) value for music that has been
processed for WFS and has had DRC correction applied (to see if tone correction
impacts spatial perception
A subjective opinion on what each one sounds like and the impact of each method
In addition, A video will be presented of the results using moving head lights to point
at the virtual sound source, and what two blind people perceive as the location of the
sound (they will demonstrate by pointing at it’s location).
8
11. What’s next?
We will need to determine how regular stereo pieces can be played back and enhanced
on a home WFS system.
We will need to investigate if room sized noise cancellation is feasible and how much
noise can be cancelled along with what types of noise.
We invite submissions of papers addressing all areas of audio processing based on Linux and
open source software. Papers can focus on technical, artistic or scientific issues and can target
developers or users. This includes (but is not limited to) the following categories:
Music production tools
Audio and DSP languages
Drivers, system and sound architecture
Standards and interfaces
Audio distributions
Documentation and educational issues
Projects realised using Linux Audio
Papers should be written and presented in English.
In general talks should take 20–30 minutes followed by 5 minutes discussion.
The length of a paper should be 4 to 8 pages. Papers should have up to 5 keywords and
include an abstract of 50 to 100 words. The abstract will be published on the conference
website once the paper has been accepted. The full paper will
be available on the website after the conference.
All papers are peer reviewed by a committee of experts in different disciplines. The reviewers
may suggest improvements to the author(s), or require changes in order for the paper to be
accepted.
How to submit
The deadline for paper submission is Thursday, January 15, 2009, 24:00 UTC (This
is equal to Friday, January 16, 2009, 0:00 UTC).
Do not send papers via email as with past conferences, instead please use the paper
upload form in our conference management system.
9
File format is PDF, formatted for A4 paper. Make use of the templates for paper
formatting available on the download page.
Authors of papers selected to be included in the printed conference proceedings may
also have to supply supplemental material like illustrations needed to layout the
printed proceedings separately.
Please notify us if you need a special technical setup for your presentation. The standard setup
will be:
microphone or head set
projector with XVGA input (resolution 1024×768)
stereo playback with mini jack input
Also, if you are not able to bring your laptop along with you, please notify us in advance.
Important Dates:
15.January 2009: Paper submission deadline
9.February 2009: Notification of acceptance
20.February 2009: Camera ready version deadline
10