A Reconstruction of Stria

John Chowning's computer-music composition Stria premiered on 13 October 1977 in Paris, during the "Perspectives du XX^e siècle" concerts. The work was commissioned in 1976 by Luciano Berio, who was then the director of IRCAM's electroacoustic music department. John Chowning, who began thinking about this project in 1972 but without the possibility to experiment, returned to the Stanford Artificial Intelligence Lab (SAIL) and realized the piece from July to October 1977 using a DEC PDP-10 computer and the computer languages SAIL (Stanford Artificial Intelligence Language) and MUSIC 10 (similar to Music IV).

The term "stria" derives from the striated aspect of the work that is so apparent in the sonogram representation. Stria is an historical piece for a number of reasons. Not least among these is its use of the intrinsic attributes of sound synthesis by frequency modulation, discovered by the composer ten years before, to structure inharmonic spectra in an ordered manner that is complementary to a non-traditional pitch space. Both the spectra and the pitch space are rooted in the Golden Ratio from antiquity (Bossis 2005).

Computer-synthesized music does not require, literally speaking, traditional scores or performers. Its existence depends on techniques and devices that are not dedicated to music and become quickly obsolete. As a result, this raises two essential problems: reproduction of works and analysis of sources. (See the article by Laura Zattra in this issue; Zattra 2007.) Therefore, there is a risk that the most "ancient" music (computer music celebrates its 50th anniversary this year) may not only freeze forever in its original and low-quality recording but also become baffling even for specialists. By reconstructing Stria, I sought this double purpose: to clone the original work in a high-quality format, as well as including the possibility to modify its parameters to analyze the process. I have therefore created a computer program for reconstructing Stria, one whose interface is easy to control and customize. The complete source code will be made available on the Web; see ccrma.stanford.edu/pieces/chowning/stria.

Sources

The piece was conceived by the composer with six parts called "sections." Each section is divided into "events," themselves automatically divided into "elements." I had for this transcription three bundles of photocopies of printed computer text files from John Chowning dated September and October 1977. The program in SAIL, COMP.SAI, consists of two parts: the event data entry and a set of functions that computes the parameters of each element. T*.MEM displays the output of the few previous computations and mostly the entry parameters of each event of the piece (here, * corresponds approximately to the events' entries in seconds: T0, T286, etc.). The program in Music 10, CPC4.INS, produces the digital sound from values computed by COMP.SAI.

The composer also provided me with any pertinent secondary documents and the original version of Stria on four tracks. Indeed, the commercial stereo version (Chowning 1988) contains a noticeable and unintended deviation from the four-channel version (all thoroughly explained in Zattra 2007) . Throughout this process, many points have been illuminated in numerous electronic mail exchanges with the composer starting in May 2006.

Transcription

This computer program, named Pystria, was realized on an Ubuntu-Edgy Linux platform, with the Python-2.4 and Csound-5.02 languages. I conceived the interface with Glade-3.0.2. I choose to use free software owing to the relative facility of the development in term of clearness, quickness of writing, and freedom. This article can provide only an overview of the transcription process; the reader interested in details is referred to the original and reconstructed source code.

From SAIL to Python

It is not easy to read the program of John Chowning, because SAIL is an obsolete language and the code is not very "clean." I had to establish correspondences between the SAIL keywords and their speculated functions in Python. Thus, I reduced several sets of recurrent lines to a single Python expression, which has shortened and clarified the code. But there was another blurred element: John Chowning left in his program some...