Performance Practice
of Electroacoustic Music

Kaija Saariaho

Nuits, Adieux


Nuits, Adieux was commissioned by the Westdeutscher Rundfunk (WDR) and composed in 1991. The piece was first performed by Electric Phoenix on 11 May 1991 in Köln with Judith Rees, soprano, Meriel Dickinson, mezzo-soprano, Daryl Runswick, tenor, Terry Edwards, bass, and John Whiting, electronics. The electronics had been developed at IRCAM by the composer with the support of Jean-Baptiste Barrière.
The piece uses texts from the book Echanges de la lumière (1990) by French author Jacques Roubaud (*1932) and excerpts from the novel Séraphita (1834) by Honoré de Balzac. Texts from the two works appear alternatingly throughout the piece as sequences labelled “Nuit” and “Adieu”, respectively. In Échanges de la lumière six people meet on six nights to discuss the phenomenon of light, in the process touching on a wide range of aspects ranging from the scientific to the transcendental. The passages selected by Saariaho elaborate on the transformation of light in a nocturnal landscape. Séraphita, which belongs to the Études philosophiques of Balzac’s opus magnum La Comédie humaine develops around the androgyne figure of “Séraphitüs-Séraphîta”, an ideal being that unifies the opposites, thus transcending the human condition. [cf. Eliade 1962]
The setting of each text sequence is made up of five sections. These are then combined into a continuous form. (s. Fig. 1 below) The Nuit sections I, II, III and V contain soloistic passages for soprano, mezzo-soprano, tenor and bass, respectively, accompanied by single pitches and noise-like phonemes in the other voices. Nuit IV presents a “restless” polyrhythmic texture alternating breath in and out in the voices in combination with single phonemes and interjections “as if short of breath”, then gradually yielding to the “espressivo” quality typical of the other Nuit sections.












III – IV – V

III – IV – V

B  A  R  S    1  –  1  1  1

B  A  R  S    1 1 2  –  2 3 8


Fig. I. Formal scheme of Nuits, Adieux


The Adieu sections are based on permutations and re-combinations of the text elements, mirroring the structure of the Balzac’s fragment. The first is articulated along declamations of single sentences by each singer: “Farewell, granite, you will become a flower; farewell, woman, you will become pain; farewell, flower, you will become a dove; farewell, dove, you will be a woman.” The second Adieu section brings polyrhythmic layers with accelerating tempi based on permutations of single words and phonemes, a process clearly anticipating Nuit IV. The final Adieu sequence constitutes possibly the expressive core of the piece. It starts with a series of four chords culminating in clusters made up of different subsets of the spoken words “fleur, homme, femme, colombe, granit, souffrance, croyance” and separated by fermatas where the electronic sound resonates (Adieu III). Adieu IV is a static section combining single melodic gestures recalling the soloist Nuit passages with spoken phrases and fragments of Balzac’s text. Adieu V, “dolce, as if humming a lullaby”, slowly shifts from a quiet harmonic field into a smooth, fading and noise-like texture. The playful and dynamic interplays of the preceding sections gradually give way to intimacy and quietness.
The juxtaposition of two elements as the basis for formal developments can be observed in other pieces by Kaija Saariaho. In Verblendungen for orchestra and tape (1982–82) both layers follow inverted spectral transitions between instrumental sound and noise. Vers le blanc for tape (1982) is based on the slow transition from a start to a target chord. Nevertheless, an important difference in Nuits-Adieux is the absence of a gradual process informing the overall form. In Nuits-Adieux both text layers create a dialogue between expressive melodic gestures and the de-construction of text structures on a permutation basis along what the composer calls the “sound/noise axis”. [cf. Saariaho 1987] The harmonic grid common to both sequences serves as a unifying, “static” element. On a higher level, the two text layers might be understood as a combined symbolic field revolving around the ideas of humanity and the vanishing of light as a personal farewell gesture to the composer’s grandmother, to whose memory the piece is dedicated. [cf. Haselböck 2019]
The live electronic layer includes both amplified and live-processed voices. The processing was originally to be realised using custom digital hardware. A recent software-based system can be considered as a new, quadriphonic version, rather than a reconstruction of the original system. In both cases the most salient feature is the use of two microphones by each singer, each one routed to different transforming devices. The changes and transitions between microphones are precisely notated in the score. This singular interaction setting demands from the singers a high degree of awareness of the ongoing processes and very precise control of their movements.
The transformations consist mainly of different kinds of reverberation without substantially altering the time and pitch structure of the vocal parts. The most prominent, called “infinite reverb”, is linked to the second microphone and is used throughout the piece. It is a long decay reverb with a decay time inversely proportional to the input gain. The quieter the input, the longer the decay. The other transformations, linked to the first microphone, are used alternatingly, creating a different aura for each section.
Thus, compared to the rich vocal textures, the electronic part is quite calm. This is mainly due to the infinitive reverb, stretching the sound over long periods and resulting in a timeless sound quality that accompanies the singers throughout the piece. Saariaho comments:

“En général, ce temps est conçu pour être relativement long : le résultat auditif est celui d’une texture changeant continuellement, et qui forme une toile de fond mouvante pour les événements chantés dans le premier micro.“ [Saariaho, Note de Programme,]

This idea was further explored in the a capella version of Nuits Adieux for mixed choir with four soloists, completed five years later (1996). Here the other voices assume a function similar to that of the electronics by generating a “moving backdrop” to the four soloists. While the solo parts remained unchanged, the additional voices were the result of a new rendering of doublings, delayed imitations and “freezings” of elements of the solo parts.


1) Performance Materials:

a) Score: Edition Wilhelm Hansen KP 00011

The score contains the following information:

– Instructions on the interpretation of the text
– Explanations of certain phonemes that must be sung according to the pronunciation of the International Phonetic Alphabet
– Texts by Jacques Roubaud and Honoré de Balzac
– Performance instructions
– Technical specifications for live electronics and amplification
– Notation for the soloists, written on traditional staffs.

Notation for electronics:

The use of the microphones for each singer is notated with the corresponding mic number right under the staff. Transitions between microphones 1 and 2 are indicated with an arrow between the numbers.

A dedicated staff for the electronics at the bottom of the score contains the cues for the live electronic performers. Effect changes in the SPX1000 are given in a box. Changes in routings for mic 1 as well as changes in the state of mic 2 (on/off) for each singer are precisely indicated. The current state is given at the beginning of each page.

b) Other materials:

Patch, standalone

Source: (Download August 2019)
Date: December 2015
Author: Jean-Baptiste Barrière
Software: Max/MSP 7
File: Nuits-Adieux-v4.1

Editable Patch

Source: Jean-Baptiste Barrière
Date: February 2020
Author: Jean-Baptiste Barrière
Software: Max/MSP 8
File: Nuits-Adieux-v6.maxpat

2) Reference recording

Electric Phoenix, online: [last accessed 10 June 2020]


Auctorial instructions

There are two different options for realising the electronics: the original version using hardware devices as explained in the score (we will refer to it as the “hardware version”), and a more recent version implemented in Max (to which we will refer as the “software version”).

Hardware version

The electronics of the hardware version are conceived for “at least two loudspeakers”.
Three devices are needed: two digital multi-effect processor units, a Yamaha SPX1000, a Lexicon LXP-15 and a generic reverb unit.
As stated above, each singer requires two microphones. Microphone 1 is used to feed the following five transformation modules of the SPX1000:

-Delay (D): two independent delay lines, 100 ms left channel, 200 ms right channel
-Early reflections (ER): a cluster of 19 early reflections with a short, audible decay
-Harmonizer (H): two transpositions of +45 and -50 cent with short delays of 20 and 15 ms
-Gate reverb (G): a further “early reflections” effect activated above an input gain threshold value
-Flanger (F): a complex varying “comb filter” effect.

(s. SPX1000 manual for further details)

The harmoniser is only used once for the bass at bars 62–75 on the consonant sound ss (as in English mission). [cf. the instructions in the score] This sound has very little tonal content, so the resulting transpositions are very subtle and difficult to perceive.
Microphone 2 is used to feed the “Infinite reverb” of the Lexicon LXP-15. This is the only effect used in this device. Its decay length is inversely proportional to the amplitude of the incoming signal. If the signal level is low the reverb will tend to be infinite and performs a “freeze” of sorts; if the signal becomes stronger, the decay time is reduced.
The score introduction indicates the preset numbers and/or the parameters necessary to program the effects of both devices.
The additional reverb unit provides a “slight basic reverberation” to be set according to the room acoustics. No further information about this is given in the score.

Software version

This version is conceived for a quadriphonic speaker setup (with stereo as option). The application is written in Max and distributed as standalone. An editable version can be obtained upon request. It is either possible to use an audio mixer or to do everything in the patch itself. The audio interface should have at least eight inputs and four outputs. A midi pedal is required to trigger the cues.
The patch version was developed by Jean-Baptiste Barrière. It was intended as a more practical and portable solution. The singers still need to perform the movement between microphones 1 and 2. All the operations performed at the desk such as sending the signal to the different effects, muting or unmuting microphone 2 are fully automated.
The behaviour and sonic character of the different transformations are not strictly identical to those of the original devices. They must be considered merely an approximation and not a historical reproduction.
The standalone version retrieved from the website in August 2019 and the editable patch show some differences. The latter is a more recent version with some updates to the description and control of the parameters for the spatialisation units. We strongly recommend working with the editable patch.


The patch is based on a new cue list that supersedes the original score markings.


A new feature of the software version is the spatialisation. It is realised with IRCAM’s “Spat”. Three units are used:

-Choir spatialisation (on four channels)
-Effect spatialisation (on four channels)
-Infinite reverb (on the two front channels)

The parameters presence and azimuth are randomly automated. The result is not perceived as a clear sound movement, but rather as a feeling of being immersed in a sound field. The use of spatialisation also increases the sense of reverberation in the live amplification and the effects.

About the effects:

-p mod (flanger)
Frequency modulation based on the tapin tapout object.

-p rev (gate rev und ER)
Spat with stereo output with different settings. Azimuth und distance are randomly automated.

-p harm (Harmonizer) and p del (Delay)
Two identical patchers are used for harmoniser and delay. The patcher is based on the tapin tapout object.

All effects are then fed into the effect spatialisation patcher based on a spat object with quadra output.


There are some distinct differences between the hardware and software versions. On one hand, the six hardware-based transformation devices exhibit interesting peculiarities, such as the abrupt decay of the “early reflections”, the “unnatural” sustained length of the infinite reverb or the specific phasing of the flanger. On the other hand, the use of the spat with its smooth and natural sound for the different reverb-based transformations and the quadriphonic disposition in the software version leads to a more homogeneous, complexly reverberated and immersive sound image. The gain in naturality and immersion goes along with a certain loss of articulation and difference.

Additional indications concerning technique and amplification

There is some additional information relating to the patch version’s setup and amplification on the website. The most significant indications in terms of microphone technique and amplification esthetics are the comments on the recommended microphones and the performance amplification:

“Microphones for the four singers: four hand-held (e.g. Shure SM58), and four other microphones, preferably headsets (e.g. DPA:


The amount of amplification required naturally depends on the performance space, but it should never cover the acoustic sound of the voices.

The ideal sound is a clear and rich “close” sound. The microphones should be placed as close to the mouth as possible.

The general level should be rather loud, but not painfully so.”

[, last accessed on 19 May 2020]


Performance report

The concert took place on 17 January 2020 at ZHdK’s concert hall 1 (KS 1) with the Basle-based vocal ensemble SoloVoices (Svea Schildknecht, soprano, Francisca Näf, mezzo-soprano, Jean Knutti, tenor, and Jean-Christophe Groffe, bass), Germán Toro Perez and Carlos Hidalgo, sound projection and Leandro Gianini, sound engineer.
The hardware version was performed using the original hardware units Yamaha SPX1000 and Lexicon LXP-15. This report focuses mainly on this version. However, the team had the opportunity to test and rehearse with the software version as well.

Setup and routing

Two Kling & Freitag Gravis 15 speakers were used in a stereo configuration. For microphone 1 a Neumann KMS105 was used and for the infinite reverb Shure beta 58 microphones.
As mentioned above, microphone 1 is used to feed the SPX1000 and for the amplification of the singers. Some reverb will be added to the amplified signal using the generic reverb unit (s. below).
One aux channel of the mixer is used for routing the signals from the second microphones to the LXP-15. One live electronic performer opens or closes the send of each of the four channels according to the indications in the score.
There is no information about the panning of the signal. In this performance the signals were distributed from left to right (soprano, alto, tenor, bass).
The hardware units can be fed with stereo or mono signals. In this instance a stereo aux send was used, so the signals were sent to the effect unit according to the channel panning.
Much care was taken to find a practical disposition of the two music stands and the two microphones needed for each singer, allowing for a good communication between them and with the audience.


Fig. II

Fig. II. Setup.


Handling the microphones

When transitions are prescribed, the singers move smoothly from microphone 1 (SPX 1000) to microphone 2 (LXP-15 “Infinite Reverb”) and vice versa, performing an operation very similar to the opening and closing of an aux send on the desk sending the voice into the “Infinite reverb”. In some instances, the singers are asked to move to the other microphone immediately. During solo sections the live electronic performer is required to switch off the corresponding microphone 2, preventing any extra signal from being sent into the infinite reverb in view of higher transparence.
In order to perform movements and transitions between microphones in a smooth and effective way the singers need to be aware of the functioning of the microphones and the transformations they are assigned to at a given moment. The distance from the microphones is an essential factor in getting the right signal to noise ratio. Unwanted pops due to strong consonants as well as spill are an issue as well. Any uncontrolled input into microphone 2 can have significant effects due to the behavior of the infinite reverb. Accordingly, the position and direction of both microphones have to be actively minded and controlled by the singers throughout the piece. Taking into account further actions such as turning pages, using tuning devices and conducting, the piece becomes very demanding indeed in terms of movements.
It was decided to place both microphones on stands so that the singers would move from one microphone to the other, leaving the hands free for other actions. This setup proved the best compromise and helped to minimise additional noises. However, holding microphone 2 in the hand might offer more flexibility and better control of the “infinite reverb”.


One of the singers conducted and the two live electronic performers followed from FOH. No extra conductor was needed.

Sound direction

The following main tasks were divided among two live electronic performers controlling the effect units and the mixer:

1. Effect processors
– Change the effect program on the SPX1000 according to the score

2. Mixing desk
– Send the signal of each microphone no. 1 to the SPX1000 according to the score
– Unmute/mute the second microphones according to the score
– Balance the amplification and the electronic transformation

Additionally, the gain of spoken parts in “Adieux 1” was manually enhanced:

– Soprano: bar 26
– Alto: bar 34
– Tenor: bar 51
– Bass: bar 42


No audio monitoring was needed. The singers were close enough to the main speakers, and the space was not very large, allowing to clearly hear the sound transformations.

Hardware connections and external reverb

Both the SPX1000 and the LXP-15 have unbalanced analogue connections. It is important to connect them properly to the desk and to make sure that there is no sound degradation. Both units exhibited some noise in the signal, an expander was used to reduce this.

As external reverb unit we used a Bricasti M7 Stereo Reverb Processor with the following settings:

-Preset: Clear hall
-Reverb: Time 1.8s
-Pre-delay: 40ms

Only the direct signals from the no. 1 microphones used for amplification were reverberated.
To enhance clarity, the duration of the infinite reverb on the LXP-15 was reduced from infinite to seven seconds, which is still very long, longer than seven seconds.

Rehearsing the live electronics

It proved efficient to make a dry recording of the eight microphones’ signals on separate tracks in order to be able to rehearse the live electronics without the singers and work on the characteristics of the different sound transformations in advance.

Selected Bibliography

Eliade, Mircea (1962): Méphistophélès et l’androgyne. Paris: Ed. Gallimard.

Haselböck, Lukas (2020): Kaija Saariaho “Nuits, Adieux” (1991). Lecture held at ZHdK, 18 January 2020.

Saariaho, Kaija (2013): Nuits, Adieux. In: Saariaho, Kaja: Le Passage des Frontières. Ecrits sur la musique. Paris: Editions MF, p. 292.

Saariaho, Kaija (no year): Note de Programme. Online: [last accessed 27 April 2020]

Saariaho, Kaija (1987): Timbre and Harmony: Interpolations of Timbral Structures. In: Contemporary Music Review 2 (1), pp. 93–133.

Schematic Overview