Spatial Orchestration Eric Lyon Sonic Arts Research Centre School of Music and Sonic Arts Queen’s University Belfast Belfast, UK e.lyon@qub.ac.uk Abstract — The emergence of multiple sites for the performance of multi-channel spatial music motivates a consideration of strategies for creating spatial music, and for making necessary adjustments to existing spatial works for performances in spaces with significantly different acoustic properties and speaker placement. Spatial orchestration is proposed as a conceptual framework for addressing these issues. I.INTRODUCTION Projection of electronic sound over multiple speakers has been an aspect of both live and fixed media electroacoustic music since the emergence of these artistic practices. The Telharmonium, perhaps the earliest electronic performance instrument, first publicly presented in 1906, produced its sound through acoustic horns distributed throughout the performance space [1]. The premieres of early electroacoustic masterpieces such as Karlheinz Stockhausen’s Gesang der Jünglinge in 1956 [2] and Edgard Varese’s Poème Électronique in 1958 [3] both surrounded the audience with speakers. Spatial projection is the central feature in performances at Audium, which have been ongoing since 1960 [4]. In the case of both Poème Électronique and Audium, the performances were site specific, so that the problem of transferring the piece to different spaces did not yet arise. II.MAPPING STRATEGIES Mapping of channels has been treated flexibly in both directions. For distribution purposes, both the five-channel tape of Gesang der Jünglinge and the four-channel tape of Poème Électronique were mixed down to stereo for commercial distribution on vinyl recordings. The four channels of Poème Électronique were performed into a 400 speaker space, the Philips Pavilion. The source tape for the 136 speaker space of Audium is also a four- channel recording. While mapping from multi-channel down to stereo is inevitably seen as a compromise, mapping from fewer to more speakers is an opportunity. We first consider existing mapping strategies before moving on to spatial orchestration schemes. III.DIFFUSION Diffusion is a key performance strategy in which relatively few channels of audio are mapped to a potentially large number of speakers in a given space. This has the advantage of separating the structure of the original audio from that of the performance space. Diffusion, properly done, can add a sense of liveliness to the composition. Irrespective of locative aspects of spatialization, a good diffusion performance can create the impression that the musical source is comprised of many more tracks than the stereo pair that often forms the basis for diffusion. This is often enhanced by interactions between the recorded audio materials and frequency- specific radiation properties of the performance space. IV.AMBISONICS Ambisonics encoding attempts to provide compositional access to spatial imaging throughout a given performance space [5]. Perception of localized images should be irrespective of listener position. However, as part of the ambisonics process, input sounds are filtered, which may go against the composer’s intentions. Speakers must be placed fairly precisely in the space in order to preserve the ambisonics effect. V.POINT SOURCE Point source composition requires advance knowledge of the locations of speakers in the space. The spatial location of each sound is then calculated as part of the composition process. This is a highly effective method for realizing trajectory patterns. Point source is essentially a panning scheme, without the psychoacoustic filtering in ambisonics. VI.CURRENT STANDARDS It is customary at electroacoustic music festivals to have access to at least an eight-channel (octophonic) playback system of good quality. However there are two common octophonic configurations, box and diamond, and neither configuration appears to be winning out over the other. One could compose with either point source or ambisonics for up to eight speakers, and expect a reasonable performance, though the ambisonics encoding would still be subject to vagaries of speaker placement to a greater degree than point source. Diffusion can be done to any number of speakers. However the more speakers available, the more difficult to control the performance with live diffusion, though this problem can be addressed with multiple performers, if a software system is prepared for merging control streams to the computer controlling outputs. But in general diffusion will still be limited to spatial movements that can be performed in realtime. Spatial aspects involving fragmentation of the sound, for