Abstract:

Requiem per una veu perduda (Requiem for a lost voice in Catalan) is a piece for mezzo-soprano with effects (pitch shift, reverb, etc.) and recorded electroacoustic material (either on DAT or CD). The latter is fixed, that is, it does not 'change' during performance, but the soprano has some flexibility for synchronising her singing with this material. The piece was commissioned by Phonos/Pompeu Fabra University's Audiovisual Institute (IUA), Barcelona, and it was completed in July 1997. So far, it has been performed live in Cuba, Scotland and Spain, and broadcast in Argentina by Radio de la Cuidad de Buenos Aires. The Requiem has four movements of approximately four minutes each: Kyrie, Gloria, Sanctus-Benedictus and Agnus Dei (only the first two were featured at SBC&M due to time constraints). The electroacoustic material was produced using three types of synthesis techniques: (a) physical modelling, (b) spectral modelling and (c) granular synthesis [Miranda 1998]. For granular synthesis, I used Chaosynth, a software of my own design [Miranda 1995]; the granular sounds (sic) can be heard at various sections in the piece. Spectral modelling was used to modify either or both the form and the content of the spectrum of various sampled sounds, mostly from Catalan poetry readings. The systems used here were the SMS package (Serra 1997) and CDP's Phase Vocoder (Fishman 1997). But it is the physical modelling technique that is the main focus of this short essay, as it was used to synthesise the Latin lyrics of the Gloria and the speech-like sounds of the entire piece. The pitch material for the mezzo-soprano part was defined based upon the analysis of the inner structure of the spectrum of vocal sounds. The relationship between the speech synthesis methods and the pitch material establishes the structural foundations of the piece. The spectral contour of vocal-like sounds has the appearance of a pattern of 'hills and valleys'. The central frequencies, the amplitudes and the bandwidths of the peaks define the colour (or timbre) of a vocal sound. For example, the central frequency of the first formant of a vowel /a/ (as in the word "sanctus", in Latin), sung by a tenor is approximately 331.12 Hz whereas the value of the first formant of a vowel /o/ (as in the word "eleison", in Latin), is approximately 196 Hz. Each vowel sound is associated with a specific formant configuration and by changing the shape of our vocal tract we change the lower formants as we speak or sing. The three lowest formants are the most important in making vowels recognisable. In male adults, the first formant can vary between 250 Hz and 1kHz, the second can vary between 600 Hz and 2.5 kHz and the third between 1.7 kHz and 3.5 kHz. I am particularly interested in composing with vowel-like sounds, as I believe that our ability to recognise vowels is closely related to our ability to recognise timbre. I find that this phenomenon is a good starting point for the definition of a framework to compose a piece. The pitches and, to a certain extent, the whole harmonic structure of the Requiem were defined according to the values of the first three formants' central frequencies for the vowels /a/, /e/, /i/, /o/, and /u/, as they are pronounced in Latin, sung by male and female voices. After much research and experimentation, I defined a note-based formant system by matching the frequencies of musical notes to formant values (based upon the standard Western 12-tone equal temperament system, with a tuning reference of A4 = 440 Hz). These notes were organised into a number of different pitch-sets (e.g. F0, F1, F2 and F3 sets, a set of vowel /a/ values, etc.). Then, the individual sets were used to produce the musical passages for the mezzo-soprano part. Figure 1 illustrates a typical excerpt from the Kyrie using notes from the F0 pitch-set.