LAC 2009 - Paper: GROMA - Programming a permanent multichannel sound installation

2009·0.63 MB·English

by  Martin Rumori

GROMA
Programming a Permanent Multichannel Sound Installation
Martin RUMORI
Academy of Media Arts Cologne
Peter-Welter-Platz 2
50676 K¨oln
Germany
[email protected]
Abstract areainthecentreofCologne. Thesoundinstal-
lation therefore is intrinsically tied to this new
In this paper I shall be addressing some of the more
complex issues of developing the technical frame- Rheinauhafen neighbourhood. The parking lot
work for the permanent urban sound installation consists of only one level, which is entirely sub-
GROMA, which incorporates environmental sounds terraneous, but through its length of approxi-
from two partner cities of Cologne, Rotterdam and mately 1500 meters it is currently the longest
Li`ege, combined with ancient texts about urbani-
one in Europe (Figure 2).
sation. This installation has been inaugurated in
The title of the installation, GROMA, is de-
2008atthelocationofEurope’slargestunderground
parking lot in Cologne. The programming environ- rived from the ancient Roman instrument for
ment Supercollider was used for realising the al- topographical survey, Groma. It was used by
gorithmic rules of the score and a flexible routing the so called agrimensors for determining the
scheme for multichannel sound distribution. basic orientation of a new settlement.
In GROMA, sound situations from two part-
Keywords
ner cities of Cologne, Rotterdam and Li`ege, are
sound installation, algorithmic composition, Super-
combined with ancient texts about urbanisa-
collider, multichannel sound projection
tion in an algorithmic multichannel composi-
tion. Each staircase stands for one of the part-
ner cities. Texts used in GROMA include ex-
cerpts from Ten Books on Architecture of Ro-
man writer, architect and engineer Vitruvius.
Other texts are from Ancient Greece or Ancient
Near East and date back up to the 4th mil-
lenium before common era. All together they
form a canon of topics which are still relevant
to today’s urban structures and current devel-
opment.
The two staircases into which the installa-
tion is inscribed connect the underground park-
ing lot with the above ground area. They
are situated at the two main squares of the
new neighbourhood with a distance of approxi-
Figure 1: View from staircase TR5.02
mately 500 meters in between them, and there-
fore form the two main entrances of the parking
1 Introduction lot, which has 26 staircases in total.
GROMA [1] is a permanent urban sound instal- All the staircases are spaces which are nor-
lation by Cologne based artists Michael Scholz, mally only experienced in transition, and they
JudithNordbrock,andtheauthorofthispaper. are by no means comfortable places for a con-
This multichannel work has been inscribed into centrated listening situation. GROMA focuses
two staircases of an underground parking lot, on this volatile moment by providing an artifi-
which is situated in the area of Rheinauhafen cial third level between the beneath and above,
[2], Cologne’s former city harbour at the river aneternalalgorithmicsoundcompositionwhich
Rhine. Itisnowthelastbigurbandevelopment can never be experienced in its entirety.
Figure2: ParkinglotunderneathRheinauhafen Figure 3: Speaker designed for GROMA
2 Technical Setup water of the river Rhine, because otherwise the
fundament of the building would be raised by
2.1 Speakers
water pressure. Therefore, the racks of the au-
When developing the concept for the GROMA dio systems are on wheels and the complete
sound installation, it quickly became clear that wiring is realised with removable plugs.
the experience of passing through the stair- Currently, twomeasurementmicrophonesare
cases should be emphasised by a multichannel being installed in each staircase which will al-
sound projection, which would allow sounds to low for slight adjustments to the overall volume
travel, but also to appear from different per- of the installation according to the surrounding
spectives towards the building and the passer- mean sound pressure level.
by. ThebiggerstaircasenumberedTR5.02 was Thecomputersarerunningastableversionof
equipped with twelve speakers and plays the the 64Studio [4] Linux audio distribution. The
sounds from the city of Rotterdam. The speak- installation is controlled entirely by the Super-
ers cover the building up to the above ground collider [5] sound environment.
visitors’ platform which provides a nice view
2.3 Network
overthecityofCologne(Figure1). Thesmaller
staircasenumberedTR7.02, playingthesounds The network connection for synchronising the
of Li`ege, does not have such a platform, there- twocomputerswasrealisedwithanopticalcon-
fore eight speakers are sufficient. nection. Copper cable could not be used, be-
Upon request of the architects of the park- cause for the distance of more than 500 me-
ing lot, the speakers had to be as inconspicuous ters several repeaters would have been neces-
as possible while providing reliability and sta- sary,and,furthermore,therewouldhavestillre-
bility against weather and vandalism attempts. mained the risk of a ground potential difference
Therefore, especially designed speakers were between the two distant places in the building.
used for GROMA (Figure 3), which were built An uplink to the internet was also installed
by Johannes Heppenheimer from Audiance [3]. forNTPtimesynchronisationandremotemain-
tenance of the installation.
2.2 Audio System
3 Recordings and Mixing
The simplest and most reliable way of driving
two installations of twelve and eight channels InthepartnercitiesofCologne, Rotterdamand
respectively with an algorithmic, global, thus Li`ege, a series of field recordings was carried
distributed score was to use two separate net- out. The plan was to cover a certain amount
worked audio systems. They consist of a rack of the typical urban presence of both cities. A
containing one standard computer with a RME set of categories for urban functions helped to
multichannel sound card, ADAT to analogue structure the recordings according to the needs
audio converters and multichannel amplifiers. of the score rules.
The parking lot is constructed for being For the recordings, various recording tech-
flooded completely in case of an extremely high niquesandmicrophonetypeswereused, suchas
shotgunandM/Ssetsaswellasbinauralmicro- Inthestaircases, therearesomeloudspeakers
phones, contact microphones (accelerometers) reserved for text tracks. They are mounted at
and underwater microphones (hydrophones). specialplacesforanoptimalspeechcomprehen-
The recorded material then was catalogued sibility. The spatial location of a text track is
and used for mixing sound pieces of different also varied according to an algorithmic series.
kinds and lengths, as they were needed for the
4.1.2 Klanginseln
score. According to the facilities avaible at the
Klanginseln (“soundisles”)areusedforcombin-
Academy of Media Arts, mixing was done in a
ing them with the texts. Whenever a text track
multichannel Protools studio using the proto-
is running in one of the staircases, the other
type speakers of the installation. The last edit-
staircase would run a corresponding Klanginsel
ing and mixing steps where done on site on a
with the same length.
mobile Protools system, before the entire sound
The correspondance to the text is formed by
material was transferred to the Linux computer
the urban categories mentioned in section 3.
systems.
With every text, a set of three “allowed” cate-
4 Score Rules gories (or category combinations) is associated,
with decreasing probability of appearance. As
The aim when designing the score was to find
the Klanginseln are also categorised, for every
a simple structure of the composition which
text an appropriate file can be algorithmically
wouldproduceabigvarietyofpossiblearrange-
selected. There are 35 Klanginsel tracks for the
ments, while still keeping certain parts of the
Rotterdam staircase and 57 for Li`ege.
process well recognisable by the listener.
In order to always achieve the same playback
The variety is achieved by using an algorith-
length, Klanginseln are only used partially, or,
miccontrolwithaleatoricselectionprocessesfor
incasethetextislonger, arelooped. Therefore,
the tracks to be played next. However, full ran-
all Klanginseln have to be seemlessly loopable.
dom control especially of combinations of files
Playback of a Klanginsel does not necessarily
and of the spatialisation is also problematic, as
start from the beginning of the file. In fact,
it is difficult to assert a certain aesthetical level
playback may start at any point in the file,
of the sounding result.
which, over a longer period of time, would en-
Therefore, a combination of more restricted
sure that all parts of the Klanginsel appear
and more open situations has been designed,
equally often.
which led to the definition of several score el-
The spatial position of a Klanginsel in the
ements along with their functions.
staircase is also subject to algorithmic control.
4.1 Score Elements As described above, it was important to keep
a certain level of sound projection quality inde-
4.1.1 Texts
pendant from the position. Therefore, all Klan-
The texts are mono tracks which were recorded
ginsel tracks are mixed to two channel stereo
with professional speakers, each text with a
files. Intheroutingoftheplaybacksystem,sev-
male and a female voice. Thus, the same text
eral so called stereo slots would be selected ran-
may appear in two forms.
domly, which would then project the file onto
Inthescoreprocess,all29textsareformedto
one or more preconfigured speaker pairs. As
aseries. Thisseriesisthenplayed,onetextafter
opposed to the text loudspeakers, these stereo
the other, but alternating in one and the other
pairs are mostly mounted for diffuse sound ra-
staircase. The next cycle would then repeat
diation.
the same series, but starting in the other stair-
4.1.3 Trajekte
case and with the gender of the voices changed.
After the second cycle, a new series would be As opposed to a Klanginsel file, Trajekte (“tra-
formed and used for the next two cycles. jectories”) are sound compositions which use
Thepausebetweenonetextandthefollowing the full number of available channels, includ-
textisvariedeachtime, rangingfromapauseof ing the speakers reserved for text. They will be
up to 24 seconds until –8 seconds, which would played back always in their full length, from the
meanashortoverlapofbothtexts. Thebilinear beginning to the end. They are neither associ-
probability distribution being used, however, ated with categories nor text tracks and stand
concentrates the length of the pauses around for themselves.
zero seconds. Since a Trajekt is a static composition which
TRAJEKT SCHLAGWORT−EXPLOSION KLANGINSEL/TEXT CYCLE
1 ROT_KI_OER_024
2 SW_M_VIT_005_72
3 TE_F_ZOL_029
TR 5.0256784 OT_TJ_MUEHLE_01 SW_M_VIT_005_023 TE_M_VIT_005 ROT_KI_OER_024
9 R TE_M_PLU_009
10
11 ROT_KI_STI_017
12
1 SW_M_VIT_005_003 LUE_KI_NAR_041
TR 7.02567342 LUE_TJ_BAHNHOF_02 SW_M_VIT_005_081 LUE_KI_OER_045 TE_F_PLA_014 TE_M_VAR_007 LUE_KI_WIR_052
8
Figure 4: Score example of GROMA
is not further modified by the algorithmic con- 4.1.4 Schlagwort-Explosion
trol, it can make use of all spatial features pos-
ASchlagwort (“keyword”)listisaseparatever-
sible in the staircase, such as travelling sound
sion of each text, and consists only of some re-
portions or the spatial distribution of sound
markablewordsorwordgroupsappearinginthe
elements according to their narrative function
text. Itisthereforeacompletelyfragmentedin-
in the composition. For example, the “boat
carnation of the text, but still partially under-
trip” Trajekt of Rotterdam harbour incorpo-
standable through association, “human auto-
rates several vertical layers of recordings, from
completion” or prior knowledge of the text, for
hydrophone under water recordings up to the
example by having listened to it already in the
topmost point of the boat, which are reflected
installation.
in the spatialisation as well.
The Schlagwort lists were read and recorded
In both staircases, playback of Trajekt tracks separately, word by word, from both the male
would start simultaneously. Due to different and the female speakers. Depending on the
lengths of the files and depending on the actual length of the text, there are up to several hun-
combination, one staircase would be finished dred Schlagwort files belonging to it.
earlierthantheother. Inthiscase, therewillbe Schlagwort files are used in the composition
anintentionalpauseinthecomposition,waiting asatransitionbetweentheKlanginsel/Text and
fortheotherstaircaseto“becomeready”again. Trajekt cycles, and vice versa. After the last
text in a cycle has finished, its Schlagwort list
Every cycle of Klanginsel/Text combinations
will start to play the fragments in their order of
is followed by one Trajekt, before the next al-
appearance in the underlying text. Gradually,
gorithmic cycle begins. Therefore, the instal-
the order will become more and more chaotic
lation constantly oscillates between these two
and increasingly random. At the same time,
poles: On one hand the algorithmic triggering
this process also affects the spatial positions:
according to a strict ruleset, which entangles
starting from the reserved text speakers, the
theplaybackseriesofbothstaircases, butyields
fragments will be distributed more and more
the more varying and unpredictable results; on
overallavailablechannels. Afteracertaintime,
theotherhandfixedmultichannelcompositions,
the process will arrive at full randomness. The
but with a more narrative dramaturgy and a
fragments then stop and the Trajekt will start
more virtuosic way of dealing with space.
to play.
For each of the staircases, six Trajekt files In the other case of leading back from a Tra-
were composed. They are also organised in a jekt toaKlanginsel/Text cycle,thefragmentsof
series, which will determine their playback or- the first text in the following cycle will be used.
der for the duration of one week. Every week, The development of randomness will be back-
the series is rearranged. wards in this case, starting completely chaotic
both in terms of order and spatialisation and James McCartney. It has been a commercial,
going on towards the original order and more closedsourceapplicationforMacintoshcomput-
restricted to the text speakers. ers only. In 2002 and starting with version 3, it
This compositional element of using frag- became Open Source, Free Software, and was
mented clouds of words was given the name further developed by the community. This in-
Schlagwort-Explosion.1 It exemplifies the two cluded a port of Supercollider to Linux, which
aesthetical poles of the entire composition in a was done by Stefan Kersten.
subtle way, but also hints at fragmentation pro- Supercollider3 is designed after a client-
cesses in history and urban development. server model. The server scsynth is a realtime
audio synthesis engine, while the client sclang
4.2 Example Score Incarnation
implements a programming language for con-
Figure 4 shows a synthetic, yet possible exam-
trolling the server. sclang and scsynth commu-
ple incarnation of a score excerpt according to
nicate with Open Sound Control (OSC) [6].
the rules described above. Both staircases are
5.2.1 scsynth
shown with their speaker channels, time is ad-
vancing from left to right. At the beginning, a The Supercollider server follows a simple and
Trajekt isrunningonallchannels. Inthesecond modular, thus very powerful model. The actual
staircase, there is a pause due to a shorter file. soundproducingorsignalprocessingentitiesare
Then there follows the Schlagwort-Explosion, called synths. They can be dynamically created
followed by a Klanginsel/Text cycle. Typical and deleted at runtime. The layout of a synth
possible situations are shown, such as looping a is determined by a synth definition, or short
Klanginsel track twice, distributing a Klangin- synthdef,whichdescribesaninterconnectednet-
sel over two or four channels, and the gender work of unit generators (ugens).
change of the text tracks. This model dates back to the MusicN lan-
guage family and is similar to that of Pd or
5 Implementation
Csound, where a synth would be called an in-
5.1 Requirements strument and a ugen an opcode.
GROMA is a permanent sound installation. Synthsarereadingfromandwritingtobuses.
Therefore, its implementation has to be reliable Buses carry signals either with audio rate or
and stable. In case of a power failure, it has control rate. Multiple buses may form a mul-
to reboot automatically and continue to work tichannel bus.
seamlessly. Synthscanbecombinedtogroups. Allsynths
At the same time, GROMA is a site specific belonging to one group can be controlled to-
installation. Major parts of the score rules and gether. The order of execution of the synths is
almost all (fine) tuning settings could only be also determined by the order of groups. Groups
done on site. For most of the setup time of the can be nested into other groups.
installation in 2007 and 2008 this meant quasi- This design of scsynth allows for creating
outdoor working in winter. In order to allow highly flexible, complex sound synthesis or sig-
for a statisfying workflow of experiments and nal processing networks.
adjustments in a team of three artists, a high All tasks on the server, such as synth and
flexibility had to be maintained in the imple- group creation and destruction, setting control
mentation. values etc. are done with a set of OSC com-
For both reasons, a Linux audio distribution mands. Therefore, scsynth can be controlled by
astheoperationsystemandSupercolliderasthe any application implementing these commands,
sound framework were the tools of choice for not only sclang.
GROMA.
5.2.2 sclang
5.2 Supercollider The Supercollider language is the standard
Supercollider is a modern, versatile sound pro- clientforthescsynth application. Itimplements
gramming environment originally developed by a Smalltalk-inspired object oriented language
with a C-style syntax and an incremental, re-
1The term Schlagwort-Explosion is derived from the altime safe garbage collector. It also provides
radio play Die Maschine (“the machine”) by French
many features of functional programming lan-
writerGeorgesPerec,wheretowardstheendthefictious
guages, which are especially useful for algorith-
computer performs a Zitatenexplosion (“explosion of ci-
tations”). mic composition.
Trajekt player Klanginsel player Text player
Trajekt slot synth Klanginsel slot synths Text slot synths
Master synth
Loudspeaker synths
1 2 3 4 5 6 7 8 9 10 11 12
Figure 5: Routing layout of the GROMA system
5.3 Play GROMA slots (the latter have been omited in figure 5).
One major artistic decision for the score of All slot synths output their signals to the
GROMA was to only use preprocessed audio masterbus. Theyarereadbythemastersynth,
material as the score elements. Therefore, no which does the global processing, such as fixed
realtime sound synthesis (in a strong sense) is and dynamic global volume control.
involved. On the output side of the master, the split
The degrees of freedom for the realtime pro- mono signals are routed through one separate
cessarereducedtothesynchronisedselectionof synth for each speaker, before they are sent to
thesoundmaterialtobeplayedaccordingtothe the sound card.
score rules, and the algorithmic control of the
SynthDef(’gr_speaker_std’, {
spatialisation. Thus, the GROMA system ba- arg gain, out, in;
sically consists of a multichannel, dynamically
automated file player with an advanced routing Out.ar(out, gain * In.ar(in, 1));
system, all built with Supercollider. }).writeDefFile;
5.3.1 Routing Figure 6: Simple loudspeaker slot synthdef
Figure 5 shows the basic routing structure in
the GROMA system. Thisroutingallowsfortherequiredflexibility
The signal flow starts at the player synths, when setting up the installation: changes can
which are dynamically created depending on be made dynamically during runtime at every
which file is currently played back. Their out- stage in the routing process. It is possible to
puts connect to so called slot buses or simply fine tune the outputs on a per-speaker basis,
slots. Slots are mono or multichannel buses which most of the time applies to volume, but
which correspond to the elements of the score. as there are separate synths per speaker, they
The Trajekt slot therefore has twelve or eight could also have a completely different layout for
channels, depending on the staircase, while the e.g. applying a filter correction.
Klanginsel slotshavetwochannelstereoinputs. On the other end, adjustments can be made
The number of Klanginsel slots also shows the to each of the score elements and to each in-
multiple possible spatial locations of this score put slot separately. A Klanginsel slot with a
element. Thecorrespondingsynthsdotherout- two channel stereo output most probably would
ing two one or more output stereo pairs, as need another gain factor than one playing to
shown in the figure. three stereo pairs. For some of the text slots,
Likewise, there are text slots and Schlagwort some slight deessing turned out to be necessary.
SynthDef(’gr_slot_te_deesser’, { to be applied. This is handled automatically in
arg gain, out, in,
GrRouting by a careful layout of all the syn-
matrix = #[ 0, 0, 0, 0, 0, 0, 0, 0,
thdefs and a defined number of standard con-
0, 0, 0, 0, 0, 0, 0, 0 ],
trols per synth class.
preGain = 1.0, sideGain = 1.0,
sideFreq = 5000, compThresh = 0.5, 5.3.2 Saving the Routing Layout
compRatio = 0.25, compAttack = 0.002,
When the routing of the GROMA system is dy-
compRelease = 0.01;
namically adjusted and tuned, the actual state
Out.ar(out, matrix * Compander.ar( of the routing is only present on the Supercol-
in: preGain * In.ar(in, 1), lider server. When the server is quit, all the
control: HPF.ar(In.ar(in, 1), settings are lost. Therefore, the complete rout-
sideFreq, sideGain), ing settings need to be saved and reapplied on
thresh: compThresh, slopeBelow: 1.0,
restart or reboot.
slopeAbove: compRatio,
One solution to this would be to track all the
clampTime: compAttack,
client-side changes to the routing during the
relaxTime: compRelease,
setup period. The downside of this approach
mul: gain));
}).writeDefFile; is that it reduces the possible ways of manip-
ulating the routing. For the setup procedure
Figure 7: Text slot synthdef with deesser
of GROMA it was desirable to keep the full
flexibility for making adjustments, such as us-
ing the “low level” sclang methods for control-
The GrRouting class allows for comfortable
runtime creation and modification of all the lingsynths,conveniencemethodsof GrRouting,
custom functions, or even an external remote
synths in the routing. A synth may be replaced
application sending OSC commands directly to
byonewithadifferentunderlyingsynthdefcon-
scsynth, which would be hard to track.
figuration,orthesettingsofonesynthincluding
its synthdef may be applied to another synth. Therefore, a separate class GrSynthConfig
Forthedeessingexampleabove,thismechanism has been implemented, which traverses the
served for using the experimentally determined synth graph of GrRouting, queries the defini-
values for the first text slot as a basic setting tion and the control vector of all synths, and
for the other text slots. saves them to a configuration file. In the op-
posite direction, the values of the configuration
For implementing this convenient functional-
are applied to the routing. This could mean to
ity, some pitfalls had to be taken into account.
populate a completely empty GrRouting, e.g.
A running synth’s definition actually cannot
at system startup, or to apply only the differ-
be simply changed, rather the old synth has to
encesoftheconfigurationtothecurrentrouting
be deleted and a new synth using the new syn-
state. This allows for easily trying out different
thdef has to be created.
configurationsandforakindof“undo”function
Depending on their function in the rout-
in case of misconfiguration.
ing, synthdefs in GROMA have some standard
control parameters, such as out and in fields 5.3.3 Play Sound Material
or multichannel output matrices, whose values All sound content of GROMA has to be played
form the actual routing layout. Additionally, backdirectlyfromdisk, astheoverallsizeofthe
there may be “private” parameters to some of sound material would be over 55GiB in 32bit
the synths, such as filter parameters. floats.
In order to keep the routing layout intact InSupercollider,theDiskInugenstreamsau-
whenreplacingasynth, theroutinginformation dio data from disk. For the seemless loop play-
of an old synth has to be kept. In GrRouting, back of Klanginsel files, however, DiskIn had
the standard control values of the old synth are to be extended with a loop functionality, which
therefore queried from the server before delet- starting from version 3.2 was added to the stan-
ing it. They are then applied to the new synth dard distribution of Supercollider.
after its instantiation. The class GrTrack provides the client side
Whencopyingtheparametersofonesynthto means for instantiating a player synth. As
another, the routing channel information of the DiskIndoesnotprovideawaytosignaltheend
target synth has to be kept, while the new syn- ofafilebacktotheclient,anenvelopeisapplied
thdef and additional “private” parameters need to all files, which, after completion, deletes the
SynthDef(’gr_play_tj12_std’, {
[ 1, 2, 3 ].choose;
arg gain, out, bufnum,
env = #[0, 0, 0], loop = 0;
would select one of the elements of the array
with equal distribution, while
Out.ar(out, EnvGen.ar(Env.linen(
env[0], env[1], env[2], gain),
1.0, doneAction: 2) * [ 1, 2, 3 ].wchoose([ 0.6, 0.3, 0.1 ]);
DiskIn.ar(12, bufnum, loop));
}).writeDefFile; would do the selection based on the probabil-
ity weights for each element given as an argu-
Figure 8: Player synthdef for Trajekt
ment.
The lookup of Klanginsel tracks for a given
enclosingsynth. ForKlanginsel tracks,anenve-
text track is implemented as a combination of a
lope is necessary anyway, since it is looped until
weighted and an evenly distributed choose op-
the corresponding text track is finished.
erationonthedatabaseofcategoryassociations,
Another class, GrPlay, forms the high level
which have been described in section 4.1.2.
access to the tracks. It uses a database of all
The Schlagwort-Explosion uses both tech-
sound material of the installation in order to
niques: there are two streams, one for the or-
lookup the filename, a per-file gain factor, and
dered sequence of Schlagwort fragments of a
the score element class with its set of possible
text, the second for the shuffled sequence. A
playback slots for a given sound name.
weightedwchooseoperationonanarrayofboth
The algorithmic score control accesses only
streams with an increasing or decreasing weight
thissinglefunctiontotriggerthecompletescore
realises the transition from the ordered to the
unfolding of GROMA.
chaotic appearance of the fragments. The same
mechanism is used for the transition of spatial-
play { arg name, slot = 0, start = 0.0,
env = nil, duration = nil, isation from the reserved text speakers towards
fadeTime = 0.1, loop = 0, run = true, all speakers.
action = nil;
5.4.2 Realtime Score vs. Pregenerated
Figure 9: Head of the play method in GrPlay Score
ThepartsoftheGROMAscorewereprototyped
in a realtime version in order to test and adjust
5.4 Score Implementation
the score rules for the installation. The selec-
5.4.1 Algorithmic Principles
tion principles take place and immediately trig-
Supercollider provides excellent functionality
ger the playback using the GrPlay class.
for implementing the score rules described in
For the final production system, however,
section 4. One important concept is the mech-
some issues would need to be solved when using
anism of Patterns and Streams.
a realtime score.
A pattern is the template for forming a
First, the random seeds of the Supercollider
stream of values, which then can be accessed
threads’ random number generators would need
one by one with sending the next message to
tobesavedregularlyinordertobeabletoseem-
the stream object. The following code creates
lessly continue with the random sequence after
a stream with a random order of the numbers
a power failure or system crash.
from 1 to 10, which will repeat infinitely, and
Second, the synchronisation between the two
queries the first value:
staircases would be a non-trivial task. One pos-
sible solution would be to elect one machine the
p = Pshuf.new((1 .. 10), inf);
“score master”, which then controls both stair-
s = p.asStream;
s.next; cases. In case of a network or power failure, the
slave would need to dynamically become a mas-
Since Patterns can deal with objects of any ter. When the network connection is back, the
class as their values, the series of text and Tra- master would need to be negotiated again. For
jekt tracks in GROMA could be realised simply all serial operations, such as Trajekt and text
using this mechanism. sequences, the slave would need to track any
Another task when implementing the state change in order to be able to seamlessly
GROMA score rules are random selections of continue with the score in case the master dis-
tracks out of collections, such as arrays. appears.
Although there are standard solutions to all hofer Institute IAIS, Johannes Heppenheimer
of these problems and although it is possible to (Audiance),Netcologne,ManfredFoxEngineer-
implementtheseinSupercollider, themucheas- ing Berlin, Anselm G¨ortz (Audio & Acoustics
ierapproachofapregeneratedscorewaschosen. Consulting Aachen), Astrid Lutz (arts+org),
Several decades of score data have been gen- and Prof. Klaus Sch¨oning.
erated in advance and stored on both machines. All photos taken by Michael Scholz.
Each score event entry includes an absolute
References
point in time for its playback. Thus, both ma-
chines are completely independant from each [1] MichaelScholz,JudithNordbrock,andMar-
other and just need to be synchronised to a tin Rumori. Website of GROMA, 2008.
global NTP time server. The absolute time val- Available from World Wide Web: http:
ues of all score events ensures the exact syn- //www.groma-net.de [cited 2009–01–28].
chronicity of the distributed score in both stair-
[2] H¨afen und Gu¨terverkehr K¨oln AG. Rhein-
cases.
auhafen, 2008. Available from World Wide
As the uncompressed score data size is ap-
Web: http://www.rheinauhafen-koeln.
proximately 1.5 Gigabytes per decade, the data
de [cited 2009–01–28].
was split into several files which are loaded in-
crementally into memory on demand. [3] Johannes Heppenheimer. Audiance, 2008.
Of course, the decision of using a pregener- Available from World Wide Web: http:
ated score might be questioned as a conceptual //www.audiance.net [cited 2009–01–28].
renunciation from the “eternal realtime compo-
[4] 64 Studio Ltd. 64 studio, 2008. Avail-
sition” GROMA originally aimed at. On the
able from World Wide Web: http://www.
other hand, the audible result is exactly the
64studio.com [cited 2009–01–28].
same in both cases. The fact that hardware
failures, vandalism, high water, construction [5] James McCartney and et al. Homepage
changes or destruction of the buildings even to of supercollider, 2008. Available from
the point of war, however, are much more likely WorldWideWeb: http://supercollider.
to harm the eternity of GROMA than running sourceforge.net [cited 2009–01–28].
out of pregenerated score data opens an inter-
[6] Open sound control protocol, 2008. Avail-
esting area of thought.
able from World Wide Web: http://www.
opensoundcontrol.org [cited 2009–01–28].
6 Conclusions
GROMA is a permanent urban sound installa-
tion, being an intrinsic part of the new Rhein-
auhafen neighbourhood of the city of Cologne.
In this paper, it has been described how ur-
ban sound situations from two partner cities of
Cologne, Rotterdam and Li`ege, are algorithmi-
callycombinedwithancienttexts. Theaesthet-
ical considerations and the artistic process to-
wards a rule set for the score of the installation
have been exemplified. The technical details of
the realisation of GROMA and the balance be-
tween technically and artistically motivated de-
cisions have been shown.
7 Acknowledgements
My thanks go to my team members Michael
Scholz, who is the initiator of this project,
and Judith Nordbrock, sound engineer at the
Academy of Media Arts.
The team’s further thanks go to the project
partners HGK H¨afen und Gu¨terverkehr K¨oln,
modernesko¨lnStadtentwicklungGmbH,Fraun-