Ein Ereigniserfassungssystem für die automatische Transkription von Musik auf Idiophonen
Anwendung beim Gamelan Gender Wayang
An event recording system
for the automatic transcription of music played on idiophones
application to the gamelan gender wayang
Pierre Dutilleux and Thomas Ruoff
Ein System zur Aufnahme der Aufführung eines Ensembles aus Idiophonen wird beschrieben. Traditionelle Methoden zur transkribierten Erfassung waren bisher aufgrund der komplexen Struktur dieser Musik nicht möglich. Ein Kontaktmikrophon wird unter jede Metallplatte der Instrumente geklebt. Die Hüllkurve jedes Mikrophonsignals wird elektronisch errechnet und dann von einem Computer erfaßt und gespeichert. Um die Relevanz der erfaßten Daten zu überprüfen, wird eine Ereignisanalyse durchgeführt. Daraus werden MIDI Signale produziert, die zu einem Synthesizer geschickt werden. Die synthetisierte musikalische Sequenz kann mit der ursprünglichen Musik gehörsmäßig verglichen werden. Das System wurde in Bali eingesetzt um das Repertoire eines Gender-Orchesters aufzunehmen. Die gesammelten Daten sollen in Transkriptionen aufgearbeitet werden.
A system is described to record the performance of an orchestra of idiophones. A contact microphone is stuck under each of the metal plates of the instruments. The amplitude envelope of each of the microphones is electronically computed and then sampled and stored in a computer. As a way of controlling the validity of the acquired data, an event analysis is performed. It delivers MIDI information that is sent to a synthesizer. The synthesized sequence can then be compared to the recorded music. The system was used for recording the repertoire of a gender orchestra in Bali. The collected informations will be used to produce scores of this repertoire.
Before describing the data acquisition system, we must give a short introduction of the cultural domain for which the system was developed. Then we will focus on the technology, then comment about the field-operation of the system and the data that we have collected.
It is known that Bali is a very attractive place for tourism, that it's music is very lively and exotic, but who in the western countries has an in depth knowledge of the Balinese music ? The lay man would say "it sounds strange", the lecture of the musicologist would be clouded with exotic words, the musician, proficient in our "classical music", would be annoyed by the steadily mistuned instruments. Percussionists would be amazed by the original patterns and playing methods that have developed in Bali. Some people though are familiar with the music of Bali. These are musicians that got trained at playing along with the Balinese performers. They have learnt directly from a master, in a way that is unusual to us. They almost never use any paper or transcriptions for pedagogical purposes. The music is not written, it is learnt by hearing and imitating, again and again.
Among the various genres of the Balinese music, we are especially concerned by that played using genders. A gender is a "metallophone constructed so that the keys are suspended over bamboo resonators; more specifically the term is used to connote genders that are played with two mallets, one in each hand". They are used in the gamelan gender wayang: "a four-piece, slendro-tuned ensemble used mainly to accompany the shadow play" [Tenzer]. The four genders are: two larger genders and two smaller genders that sound one octave above the larger ones. Each gender covers 2 octaves. The slendro tuning has 5 equal intervals in each octave. The shadow play is a theatrical performance where the characters are puppets whos shadows are projected onto a screen.
In order to broaden the understanding and the knowledge of the gamelan gender wayang within western countries, a project led by the composer Johannes Goebel, intends to collect and tran-scribe the repertoire of a prominent master in Bali: I wayan Loceng.
Figure 1: A gamelan gender wayang.
Even the skilled European gender players have troubles at deciphering the different musical parts from a recording. To help them in the transcription process, a special recording equipment was desired. Such a system must be able to keep a separate track of each musical part of the interlocked performance. It is not the first time that the musicologists promote the development of an original investigation technique: see [Arom] in the realm of African ethnomusicology.
From the recording of the sound of an orchestra it is very difficult to trace back the various musical parts. We chose therefore a technique for recording the activity of each of the sounding elements in an individual recorder-channel. The system should be able to produce a note-list, indicating when and how which plate has been struck. This list would be comprised of records of the type:
<Plate number, Onset time, Onset intensity, Damping time>
Figure 2: A signal and the parameters that describe its envelope.
where "Onset time" and "Onset intensity" describe how the player has hit the "Plate", and "Damping time" tells when he has stopped the vibration of the plate with his thumb-ball.
The instruments of the gamelan gender wayang are idiophones, that means that they vibrate by themselves when they are struck. Idiophones are all those musical instruments which are in ge-neral struck or shaken, like bells, xylophones, shakers. With a contact microphone it is hence possible to get the information about when and how strong they are excited. So the principle of the technique is to stick a contact microphone under each of the sounding elements of the orchestra. In our case, we have to deal with a set of four genders, each of them having 10 plates. In addition we should be able to record also the mallet of the puppet player, the dalang. With his mallet, the dalang gives informations to the musicians; this enables a tight synchronisation between theatrical performance and music. That means that we need at least 41 channels of data-acquisition.
We did not find any suitable standard equipment to solve our problem. So we decided to build a recorder for the amplitude curves of a large amount of contact microphones. We named the device an Idiagograph. This new name stands for: "device for transcribing the music played on idiophones and taking into account the agogy ".
The system is designed for recording the amplitude curves of the signals from contact micro-phones onto a computer hard disk. The microphones should disturb neither the musician nor the sound of the instrument. The most practicable place for sticking the microphone is under the plate at the outer edge. Sticking the microphone damps the vibration of the plate, this effect should be kept very low. The suspension of the microphone must be very flexible not to introduce mechanical crosstalk between acquisition channels. Among the microphones that we tried, we retained the plain piezoelectric film because it damps the plates very little and it can be connected using a very flexible cable (Figure 3).
Figure 3: The data acquisition system with 12-channel modules.
Charge amplifiers convert the charges, developed into the piezoelectric film by the vibration of the plate, into usable low impedance voltages. Logarithmic root-mean-square converters (LOG-RMS) are used to compute the amplitude envelopes and express them in decibels. The conditioned signals are multiplexed and then digitized using a single 8-bit analog-to-digital flash converter (ADC). The data from the ADC are stored onto a computer disk. A data acquisition board was built for a PC-AT compatible. This board hosts the ADC, the clock and the adress generators for the multiplexers, and a memory to buffer the transfers between the ADC and the hard disk. The charge amplifiers, the LOG-RMS converters and the multiplexers are located in external conditionning modules.
To allow for flexibility in cabling the system within the gamelan gender wayang, a conditionning module was built for each gender. Each module can handle 12 channels, and the interfaces between modules and acquisition board are designed to allow for larger distances.
A special program was written for each of the operating modes of the idiagograph. Before recording a piece, it is necessary to make sure that each channel works properly and to establish reference values for a fortissimo and a pianissimo. For each acquisition run, it is possible to adjust the acquisition rate. In practice we have used rates from 100 Hz to 400 Hz. The lower rates are valid for larger groups or for longer pieces. The higher rates are preferred for recording solo or virtuoso performances.
A program is available for displaying the measured information, 2 channels at a time. It is most useful for appreciating the quality of the acquisition channel: noise floor, scaling, crosstalk, and to manually check the individual events. A program detects automatically all the onsets and dyingouts of notes and it outputs a sorted note-list. This program is most critical for the transcrip-tion. If it is too sensitive, it will generate a lot of false detections, if it is too coarse it will miss refinements of the performance. By comparing the note-lists and the acquired data with the audio and video recordings, it is possible to judge the quality of the detection and eventually to improve the detection program itself, hence the note-list.
The note-list contains absolute times, amplitudes, and channel numbers. A possibility to use this list is to convert it into a MIDI format. It can then be sent to a synthesizer. The sound produced by the synthesizer, although it is a very coarse representation of the gender performance, allows to check whether the data-acquisition was significant and consistent. Back to Europe, the note-lists in the MIDI format can be further processed by state of the art programs for musicians. The tran-scrip-tion of the note-lists is in progress. Additional software tools are under development by Sukandar Kartadinata.
Recording in Bali
In August 1994 the repertoire of the gender master I wayan Loceng was recorded in Bali, in the Sukawati village. Since, in a gamelan gender wayang, the 2 smaller genders merely double the musical parts played by the larger gender, only the 2 larger genders were recorded. 30 pieces have been recorded, amounting to about 2 hours of music. In addition a whole shadow play theater performance has been recorded, lasting 2 hours and 45 minutes.
Musical scale and spectrum of a gender plate
To compare the resynthesized music to the original one, it is important to know about the tuning and the spectra of genders. I wayan Loceng has several sets of genders available. Each set has its own tuning. To tune a set, one proceeds as follows: one of the larger genders is tuned first. It will be the "lower larger" gender. The second larger gender is tuned a few hertz above the lower larger gender: each plate sounds about 7 Hz higher than the corresponding plate of the lower gender (Table 1). When the plates are tuned, then the bamboo resonators are tuned to the fundamental frequency of the plate. Finally the smaller genders are tuned one octave above the larger ones, apart from a few herz, to allow for shimmering. We notice that the octaves are stretched (Table 2), we neither found this information in [McPhee] nor in [Tenzer]. The values here are derived from measurements made on a set of new genders that is rated "very good" by Loceng.
Table 1: Fundamental frequencies of the four genders of a set and beat-frequencies between same-sized genders.
Table 2: Fundamental frequencies, intervals and streched octaves, as measured on a larger and a smaller gender of the same set.
The spectra of several tones have been measured. The tones are highly inharmonic. As an example, let us consider the spectrum of the dung0 plate (Figure 4). The fundamental frequency (229 Hz as given in Table 2) appears here in the 232 Hz analysis bin, it is not the strongest partial. The higher partials often appear as doublets. The spectra are very dependent on how hard and where the plate is struck. Loceng uses this feature to produce subtle timbre variations.
Figure 4: Spectrum of the plate dung0, medium struck at the center, using a hard-wood mallet.
The note-lists collected in Bali are beeing processed to appear in a form that would be easy to use by western musicians. The critical eye and ear of a trained gender player, Andreas Herdy, is necessary to edit the notes detected by the computer. Notice that the pitches can be mapped onto the chromatic scale according to various criteria (Figure 5). One might believe that the best mapping is achieved when one assigns the chromatic pitch to a plate that has the closest frequency. Such a mapping will fail to provide the necessary musical properties. A better mapping enforces the scale to be pentatonic, but the definitive answer will be found by trial and error together with the performers.
Figure 5: How pitches are named, notated and played on the synthesizer.
A data acquisition system has been built and programmed to record the performance of a gamelan gender wayang. During one month in Bali, the repertoire of the master I wayan Loceng and a shadow play performance have been recorded. The collected data is of great value for musico-logists and musicians. We hope that it will contribute to a better understanding of the Balinese music and to its preservation.
The investigation of the scales and spectra shows that the octaves are streched and that the genders of the same gamelan are mistuned by about 7 Hz. Along with the doublets in the spectrum of each plate, these features contribute to a lively shimmering sound.
The idiagograph system has been tested with a gamelan gender wayang, but it could be used for other gamelans as well as for other idiophone orchestras. The 48-channel configuration that has been used in Bali could be extended to 360 channels, allowing to record large ensembles.
We are indebted to I wayan Loceng for his hospitality and his excellent temper. To our res-ynthe-sized musical sequences, he said: "It sounds as if you were playing like beginners!". Thanks to Ako, the Japanese student who played one part of the repertoire along with Loceng.
RUOFF Thomas, Meßwerterfassungs-System für die Automatische Transkription von Musik aus Idiophonen, Diplomarbeit, Fachhochschule Karlsruhe, Feinwerktechnik, 1993.
SADIE Stanley Ed., Gender, The new Grove dictionary of musical instruments, p.35. Macmillan 1984.
SADIE Stanley Ed., Indonesia §III, 1: Bali, The new Grove dictionary of music and musicians, p.179-189. Macmillan 1980.
I wayan LOCENG, Gender wayang music, a script in english and indonesian.
Banjar BABAKAN desa Sukawati, Wayan LOCENG, Wayan SARGA, Ketut BALIK, Wayan NARTHA, Gamelan gender wayang, music of Bali. Ricks records, Maharani, RCD-03.
TENZER Michael, Balinese Musik. Periplus Editions, Berkeley-Singapore, 1992.
McPHEE Colin, Music in Bali, a study in form and instrumental organization in Balinese orchestral music. Da Capo Press, New York, 1976.
ROSSING Thomas D. and SHEPHERD Robert B., Acoustics of Gamelan Instruments. Percussive Notes 19(3), 73-83, 1982.
JANSSEN Jos, KAEGI Heinerich, MIDIM-Duplication of a Central-Javanese Sound Concept, Interface, Vol. 15(1986), pp. 185-229, 1986.
AROM Simha, African polyphony and polyrhythm, Musical structure and methodology. Éditions de la Maison des Sciences de lHomme, CID, Paris, 1991.
AROM Simha, A Synthesizer in the Central African Bush: A Method of Interactive Exploration of Musical Scales. in: Für György Ligeti, Die Referate des Ligeti-Kongresses, Hamburg 1988, pp. 163-178. Laaber-Verlag, 1991.
TICHY J., GAUTSCHI G., Piezoelektrische Meßtechnik, Springer Verlag Berlin Heidelberg 1985.
© 2013 ZKM | Zentrum für Kunst und Medientechnologie Karlsruhe :: Impressum/Web Site Credits