A CAP for graphic scores

Graphic notation and performance

Benny Sluchin Mikhail Malt

EIC-Ircam Ircam
sluchin@ircam.fr Mikhail.Malt@ircam.fr

ABSTRACT also accessed [4]. In all these cases, a computer interface

that assist the performer in preparing and performing the
Many graphic scores use the pitch versus time works were developed.
presentation, as a natural extension of the usual notation. The two main compositions we discuss here, present
In the general case, it displays discrete pitches, in a fixed performance notation difficulties. Conversely to previous
timeline. Nevertheless, graphic scores use a lot of work [5]1, our aim is to develop solutions, in a computer-
continuous lines, and the vertical dimension can be assisted performance study, for trombone version, and
adapted to a particular performance. In such a way, the offer aural guides to attain a performance, as close as
instrumentation is free, and the actual range of a possible to the graphical notations used.
particular instrument can be adapted according to the
notation. The present article is initiated by a search to
provide the performer with adequate tools to approach the
execution of such works. A computer assisted
3. TWO EXAMPLES, TWO PROBLEMS
performance approach helps the player in the preparation
process for both: the time and the pitch approximations. 3.1 Alvin Lucier’s Panorama (1993)
The simulation can enhance the performance in Scored for Trombone and piano, Panorama (1993) [6] is
approaching the graphical notation. a work whose shape is traced by the sliding trombone
. part (Figure 1).
"During the course of the work, the trombonist slides
1. INTRODUCTION continuously, outlining the shapes drawn
by the diagonal lines." [6]
The starting point of the present research was a
performance issue; works of the 20th century, aimed at
concerts and recordings, were approached seriously. The
player searched for tools that will help him attain the
composer’s intentions as proposed by the printed
(graphical) scores. Curiously enough, such tools did not
exist, and the precision of the performance was quite
lacking. Our work focuses on this category of works and
a CAP (Computer Assisted Performance) approach, that
is in fact help in preparing the works for performance, in Figure 1: Alvin Lucier’s Panorama, original manuscript
the same way a metronome is. score, first 25 seconds
The musician may stop sounding his instrument at any
time but his slide moves continuously. The pitches are
2. PREVIOUS WORK notated with great precision within microtonal tunings.
Since several years now, works having a generic Alvin Lucier uses a half, quarter, third and sixth note
presentation, that is to say, open form, has occupied us. notation he also adds a shift notation, in cycles on some
Three examples were discussed: Domaines (1968) by pitches (Figure 2) and give exact onset time (minutes and
Pierre Boulez, Duel (1959) and Strategy (1962) and seconds).
Linaia Agon by Iannis Xenakis and Concert for Piano As quoted by Volker Straebel [7], “Alvin Lucier mapped
and Orchestra (1958) by John Cage ([1], [2], [3]). Other the panorama of the Swiss Alps to the pitches of a slide
works by Cage, part of his late “number pieces” were trombone. He worked from a reproduction of a panorama
drawing by Fritz Morach after a landscape photo by
Copyright: © 2016 First author et al. This is an open-access article dis- Hermann Vögeli”.
tributed under the terms of the Creative Commons Attribution License
3.0 Unported, which permits unrestricted use, distribution, and 1
Enrico Francioni develops a technical solution for one-man
reproduction in any medium, provided the original author and source Ryoanji performance, in the contrabass, percussion and vocalize
are credited. version.
 in the flute version, J6 in the double bass version) to a
twelfth (J8 in the oboe version).
Trombone
t
#œ
t
œ
t
#œ
t
œ
t
œ
t
Figure 2: Alvin Lucier’s Panorama, microtonal notation #œ
t
œ
t
#œ
3.2 John Cage’s Ryoanji (1983-84) t
œ
Ryoanji (1983-4) [8] is a series of works based on the t
celebrated rock gardens in Kyoto. The contours of the œ
fifteen different rocks are the basis of the graphic t
bœ
notation. J. Cage’s instructions are typical; they give the t
poetic atmosphere for the performance: “Each two pages œ
t
are a ‘garden’ of sounds. The glissandi are to be played
#œ
smoothly and as much as is possible like sound events in t
nature rather than sounds in music. (…) The score is a œ
t
‘still’ photograph of mobile circumstances.” [8]
#œ
The solo instrument parts (there are versions for oboe, t
flute, trombone voice and double bass) appear all in the œ
t J1 J2 J3 J4 J5 J6 J7 J8
same form. The original score presents a frame where
continuous and complex gestures were drawn. For each
page, in the left upper part, we can remark a small stave Figure 4: Ryoanji Trombone Gardens Registers
with the page register (Figure 3). On the abscissa, the The form of the curves is very precise, but hard to
time (the length of a system) is represented on 24 cm. perform on the instruments. Both, the time factor
The height, the register used, is represented on 6 cm and (horizontal abscissa) and pitches (vertical) are to be
subdivided, according to the register, in semitones. approximated with maximal precision.

4. THE PERFORMANCE ISSUE

4.1 Pitch discrimination threshold

According to some authors ([9], [10]) the pitch interval
discrimination for trained musicians is around a semitone
tenth or 10 cents at 500 Hz (around B4) and increases to
lower pitches (100 Hz), until a half-tone. Two main
©

aspects present difficulties in these pieces. The first is the

difficulty to control slow glissando, where the
discrimination of pitch change is challenging, as in
Panorama. Lucier himself says:
Figure 3: Cage’s Ryoanji first trombone version page
“The player is not expected to be able to hear these finer
In Figure 3 the range is precise (from B0 to F#1), and the tunings, but may perceive them as beats per seconds
durations are to be determined by the performer. against the piano tones” [6].
Concerning the duration, only in the voice version Cage
was explicit (ca. 2 minutes for a "garden", composed of The second, mainly related to Ryoanji, is to evolve in a
four systems, engraved on two pages). For the very low instrumental register. Ryoanji, trombone
instrumental versions, musicians generally apply this version, evolves between F-1 (21,83 Hz) and G#0
same duration. Each "garden" is using a specific range, (51,91 Hz). In this register the pitch discrimination is far
which is shown in Figure 4. to enable the fine-tuning of the notated glissando curves.
The vertical span (which is the same for all the pieces) In addition, the performer physical control, lips vibration
indicates a variety of intervals going from a semi tone (J3 and slide position, is difficult and uncertain.
 5. COMPUTER SIMULATION
4.2 The Glissando As pointed previously the performance of Alvin’s Lucier
Panorama and Cage’s Ryoanji presents some difficulties
Defined as a glide from one pitch to another, the term mainly related to glissandi control, pitch discrimination
contains much more than the physical change of the and extreme low pitch control. The leading idea using
frequency. There are musical examples (ethno-musical simulation was to produce an exact and precise guide for
and contemporary) based exclusively on it. Its musical a performance study.
performance, even on very adapted instruments (i.e.
unfretted bowed strings, slide trombone), is far from 5.1 Alvin Lucier’s Panorama (1993
simple. Traditional orchestration books, that discuss
glissandi, miss a practical approach which therefore Before building the computer simulation a first step was
follows. intended. Panorama was transcribed in a Cartesian space,
A glissando has two primary features: the start and in a break point function presentation. Each page is
ending points, and the duration. A secondary character, having the same duration of one minute (Figure 5).
the form of the sliding is particularly important, can be
defined as the speed. A glissando can be linear (having a
constant speed) or not linear (having a varying speed).
Performing a glissando on a trombone2 (the instrument
related to this research) is strongly related to the register
involved. For middle and upper registers the execution is
determined by the slide position taken as a starting point.
Naturally, one will be unable to perform an ascending
glissando if the starting slide position has the shortest
slide length. Alternate slide positions for the same note
are then needed. Another problem, concerning the
doability of a glissando, is related to the fact that a
trombone slide can execute a maximum interval of an Figure 5: Graphic score containing pitches
augmented fourth (the span of seven positions a Bb
trombone slide has). In many cases, a glissando has to be This representation brought a better score awareness of
tiled from two or more simple ones. both pitch versus time evolution.
The slide length modification needed to perform a The second attempt, to improve and help performance,
glissando may be of some help to measure the speed of was to transcribe the original manuscript (Figure 1) in a
the glissando. The modification to perform a given time proportional musical notation score. In the original
interval changes according to the trombone position. For there is not a clear correlation between time and graphic
example, an interval of a half tone needs a greater slide space. Each system lasts for one minute (Figure 6) and
length modification in lower positions than in higher the dashed measure bars represent regular 10 seconds
Panorama
ones. time span. 1/2
Alvin Lucier

B w w #w
The slide is not the only way to control the pitch. Trombone
↑

Particularly in the low register, the lips can operate w w

0 10" 25"

Piano
? w #w
important frequency changes. These changes are though 1/2

accompanied by a loss of the particular timbre. As the B #w ↑w w w

1/2

#w #w
-2 +1 -2
↑

lips are vibrating in a different frequency from that

1' 1'14" 1'16" 1'32" 1'43" 1'50"

? #w # ww #w w
1'00"

w w w
corresponding to the instrument length.
When performing a glissando, the starting pitch is often ↓w
1/2

B # w↑ nw #↑w #-2w
1/2 1/2

just a floating moment. The pitch has to start moving 2'00"

?
w
2' 2'06"

b ww
nw
2'18" 2'29" 2'40"

#w # ww #w
immediately, though the ear won’t perceive a change
initially. The same holds for the ending point of a B #w # ↓w n ↓w
1/2

w
1/2

w
1/2
-3 +2 ↓

glissando. 3'00" #w
3'
#w
3'18" 3'33" 3'49"
w
3'51"

? nw ww #w
w
Another way to measure a glissando, lies on the notion of
differential beats. Having a reference pitch, close in
Figure 6: Alvin Lucier’s Panorama, transcribed score,
frequency to the moving one, will create this first page
phenomenon. Controlling the relationship between two
close frequencies, based on the number of differential Even if it could seem minor, it is not. This transcription
beats in time, is an approximate and quite effective improves the readability over the manuscript and a good
method. time versus graphic space proportionality helps the
performer to better evaluate the linear time span.
Controlling the slow pitch change over time is one main
difficulties for Panorama. For example, at 1’43”, the
trombone evolves over 7 seconds with a pitch change
2
from a C4 plus 1 Hz to a C4 minus 2 Hz. This change
It seems that some of these points may be generalized to string only can be made (as in Xenakis pieces) by referring to
instruments.
the differential beats produced between the trombone C4 Instr 1
fundamental and the piano C3 octave harmonic. But it k1 oscil1 0, 1.0 ,p3, 2
still presents a 0.4286 Hz/sec variation. Another moment, asig oscili p4,k1,1
from 14’02” to 14’14”, presents a 0.1667 Hz/sec out asig
variation on a Bb3. In general, the pitch slope is very Figure 9: Csound orchestra for Panorama synthesis
small, in relation to the pitch discrimination interval. In simulation
addition to these four crucial points (1’43”, 6’12”, 9’37”
The <f2> score function is built from the table data
and 14’02”), the piece presents other problematic pitch
(Figure 8). And the spectral content is set via a <f1> table
slope variations. For Example, the smallest slope is to be
in order to have a 5 partials harmonic sound (Figure 10)
found from 7’05” to 7’32”, where the performer is asked
to control a descending glissando (from a G4 to a G4
minus a sixth tone), over 27 seconds. A 0.59 cents/sec
variation.
These points lead us to provide a simulation to help the
performer with the tuning and the duration. Event if the
pitch changes is still below the pitch discrimination
hearing range. The performer, playing together with the
simulation, can refer to the beating as a way to adjust the
pitch. Figure 10: Alvin Lucier’s Panorama, first 3minutes, 5
From graphical data inferred in the precedent steps, a harmonic sound sonogram
three columns table with time (in seconds), MIDI pitch
As the aim of the simulation is to be a guide to a
and the shift in Hz (Figure 7) was created.
performance study, the use of a 5 partials harmonic sound
is intended to help the performer to tune and to follow
more precisely the glissando. Beating between higher
partials are more perceptible than beating between lower
partials.
Another simulation performance guide was built using a
sample of Eb4 ordinario trombone sound (from the SOL
sound, Ircam data base). This 4 seconds real sound was
stretched until to reach 16’40” using Audiosculpt
Figure 7: First seven trombone events (time in sec, (© Ircam)6. The stretched sound was then continuously
MIDI pitch and shift in Hertz) transposed using the data from Figure 7, with the
Plotting the table data from Figure 7 we obtain a SuperVP Kernel (© Ircam)7. As the mean pitch from all
glissando curve displaying the Alps landscape (Figure 8) the performance is 63.31 MIDI pitch and the median
[7]. 63.24 MIDI, we choose Eb4 to minimize the
transpositions of the original sound, keeping as possible
its concrete character.
A Max (© Cycling74)8 patch was built then, to allow the
performer to browse in the simulation to use the
transformed sound as a guide to study (Figure 11).

5.2 John Cage’s Ryoanji (1983–84)

Figure 8: Alvin Lucier’s Panorama glissando curve For Ryoanji, the main idea was, likely to Panorama, to
(Time x MIDI pitch)
build a simulation, a guide to be used in the study
The synthesis simulation was done with Csound3 in the performance phase, but the basic approach was
OpenMusic environment4. The Csound orchestra was completely different. The glissandi here are not linear but
simplest as possible (Figure 9) just a single oscillator freely drawn by Cage.
(<oscili>) where the frequency (k1), is driven by a <f2> The first step was to find a strategy to retrieve the pitch
score function5. versus time information from Cage score.
Some strategies were studied, like border detection in the
picture file, as a methodology to identify the curve. But
for all of them the preprocessing step (cleaning the
original score, erasing the frames, etc.) was to much time
consuming. Instead of that we adopted an old computer
3
http://www.csounds.com technic, “hand digitalization of plotted data” (as done in
4
With two external Libraries, OM2Csound and Chroma. See:
http://repmus.ircam.fr/openmusic/libraries. For OpenMusic
6
information: http://repmus.ircam.fr/openmusic/home http://anasynth.ircam.fr/home/english/software/audiosculpt
5 7
This table, <f2>, is indicated as being the last <oscili> http://anasynth.ircam.fr/home/english/software/supervp
8
parameter (2). https://cycling74.com/
the 70’s). This technic involves having a background
image and with another tool to identify points
(coordinates) in the background image. It was broadly
used in sea waves digitalization.

Figure 13: From first to second frame, Cage’s Ryoanji,

first trombone garden

Figure 14: Digitalized curves, inside OpenMusic <bpf>

Figure 11: Computer interface to browse Panorama object
simulation
Following the curves digitalized (Figure 15), it was easy
Using the OpenMusic (© Ircam) environment, we firstly to isolate them and use each one as a frequency data
reconstructed the scanned score to have each garden on curve to a Csound synthesizer as used previously for
one linear page to proceed afterwards to hand Panorama simulation.
digitalization.
As the original score was scanned from a manuscript, all
the frames, from the same garden, has not exactly the
same height. The bottom and upper border were not
exactly parallel and the semi-tone subdivisions were not
aligned across all the garden frames. To say that the
alignment was not so easy (Figure 12).

Figure 15: Extracted glissando curves

With this last figure one can imagine the stone outlines,
Cage used for the score generation [12].
While the registers are given, the time scale is up to the
performer. We choose to give to each garden a two-
minute duration.
All the gestures were synthesized as for Panorama, but as
the fundamental frequencies were to low we used a seven
harmonics spectra to help the tuning (Figure 16).
The final simulation sound file, would be imported in an
interface (as for Panorama), to allow the performer to
browse and use it for preparing his performance.

Figure 12: Tree intersections, from Cage’s Ryoanji first

trombone garden
To resolve this drawback, we chose to align the different
frames focusing on the continuity of the line describing
the gesture (Figure 13).
From the reconstructed garden, we proceeded to “hand
digitalization”, with the help of the <bpf> OpenMusic
object (Figure 14), trying to work as precisely the zoom
allow us to do. Compare with curves in Figure 3. Figure 16: Ryoanji, first trombone garden, first 16 sec
sonogram
 6. CONCLUSIONS le renouvellement des formes musicales dans la
musique contemporaine, Symposium, organized by
The two pieces treated here take advantage of graphical M. Grabócz, CDMC, Paris, 28 november 2014.
notation but present essential difficulties in performance. https://www.youtube.com/watch?v=ZixIYc7iL84&f
We propose here a computer strategy, simulation based,
eature=youtube_gdata
to assist the musician in the preparation and performance.
Concerning Panorama, while we found a way to bring a [3] B. Sluchin, “Linaia-Agon Towards an Interpretation
possible assistance for the glissando control and fine- Based on the Theory”, International Symposium
tuning, the optimization for the glissandi position changes Iannis Xenakis, M. Solomos, A. Georgaki, G. Zervos
still is to do. We look forward to deal with it mainly with (Ed.), Athens, May 2005, 1–16.
a constraint system. Concerning Ryoanji, the main http://Cicm.Mshparisnord.org/ColloqueXenakis/
problem (glissando control and fine-tuning) still
unsolved. As pointed previously, main parts from this [4] B. Sluchin, M. Malt, “Interpretation and Computer
work still impracticable, from cognitive and instrumental Assistance in John Cage’s Concert for Piano and
reasons. Ryoanji evolves in the extreme lower register Orchestra (1957-58)”, SMC - Sound and Music
what makes very difficult to control and listen to the Computing Conference, 2010, pp. 1 – 8.
small microtonal waves. For the first garden (where
[5] E. Francioni, “Csound for Cage's Ryoanji”, Csound
Cage’s register is from B1 to F#1) the gestures
Journal, vol. 18, 2014, 1 – 20.
(according with our hand digitalization) evolve between
C1 minus 19 cents (32.35 Hz) and a F#1 minus 14 cents http://www.csounds.com/journal/issue18/francioni.h
(45.88 Hz). tml
We have two pieces mainly based on glissandi, but with [6] A. Lucier, Panorama for trombone and piano,
two different performance goals. While in Panorama manuscript score, 1993.
Lucier expects for a precise pitch and time performance,
in Ryoanji, the graphical notation has not only a score [7] V. Straebel, “The Sonification Metaphor in
function, but also an inspirational and poetic purpose. As Instrumental Music and Sonification’s Romantic
stated in the Ryoanji instructions: Implications”, The 16th International Conference on
Auditory Display (ICAD-2010), 2010, pp. 1 – 8.
“The glissandi are to be played smoothly and much as is
[8] J. Cage, Ryoanji for Trombone Solo, Henmar Press
possible like sound events in nature
rather than sounds in music.” [8] Inc., 1985.
[9] J. H. McDermott, M. V. Keebler, C. Micheyl, A. J.
Composers, inspired by extra musical facts or concepts Oxenham, “Musical intervals and relative pitch:
can use sonification as an inspiration way. Translating Frequency resolution, not interval resolution, is
graphic or visual aspects in musical notation can be special” The Journal of the Acoustical Society of
source of novelty and challenge, but sometimes it can be America, vol. 128, no 4, 2010, pp. 1943 – 1951.
source of difficulty.
The notation graphic space does not have the same [10] F. Bianchi, S. Santurette, D. Wendt, T. Dau, “Pitch
properties as the basic pitch versus time real space. If in discrimination in musicians and non-musicians:
the graphic space, we can represent any pitch interval Effects of harmonic resolvability and processing
size, it is just a scale question dependence. In hearing effort” Journal of the Association for Research in
world, humans are bounded by their cognitive pitch Otolaryngology: JARO, vol. 17, 2016, pp. 69 – 79.
thresholds. While in the graphic space, we can navigate
in the time line, jumping from future to past and vice- [11] J. Cage, “Ryoanji: Solos for Oboe, Flute,
versa, in the real world the time appears to be Contrabass, Voice, Trombone with Percussion or
irreversible. Orchestral Obbligato (1983-85)”, PAJ: a Journal of
These two musical compositions were recorded based Performance and Art, vol. 31, no 3, 2009, pp. 58 –
upon the present research [14]. 64.
[12] M. Fowler, “Finding Cage at Ryōan-ji Through a
Re-Modelling of Variations II.” Perspectives of New
7. REFERENCES Music, vol. 47, no 1, Winter, 2009, pp. 174 – 191.

[1] B. Sluchin, M. Malt, “Open form and two [13] K. Schöning, “Ryoân-ji: oder die Gärten der Leere”,
combinatorial musical models: The cases of Neue Zeitschrift Für Musik (1991-), vol. 155, no 5,
Domaines and Duel”, 3rd International Conference 1994, pp. 24 – 27.
on Mathematics and Computation in Music, Paris, [14] J. Cage, Ryoanji, Benny Sluchin – trombone,
June 15-17, 2011, p. 211-217. Sylvain Rappaport – percussion, CD produced by
[2] M. Malt, B. Sluchin, “Théorie des jeux et structure ut_performance@hotmail.com, 2015, Paris, France.
formelle dans Duel (1959) et Stratégie (1962) de
Xenakis”, L’influence des théories scientifiques sur