A Bit of Everything: A Cherry-Picking Approach to Massively Multichannel Music Composition

A Bit of Everything: A Cherry-Picking Approach to Massively Multichannel Music Composition Pierre Alexandre Tremblay 1[0000-0002-3980-8811] 1 Conservatorio della Svizzera italiana, CH-6900 Lugano [email protected] Abstract. This paper describes a workflow for creating massively multichannel music, with a focus on its portability, to enable wider engagement with spatial composition. The proposition allows for the cohabitation of multiple complementary modes of encoding the space. After defining the motivations, the affordances of the setup are presented and illustrated in two pieces by the author and four other composers, successfully presented in various professional settings. Keywords: Spatial audio, computer composition, critical immersive listening 1 The Problem Space This paper is an aesthetically driven, critical technical note of autoethnographic flavour; it presents a modest contribution from compositional research fieldwork, on the affordances of the cohabitation of competing formats of multichannel music. It reflects on the entanglement of the underlying technology and the artistic approaches each favours, in the light of the growing literature on interface/outcomes entwinement [1]. Composing for a large loudspeaker array presents significant challenges of creative studio workflow, as well as of portability between the composition space and the various concert spaces. It also brings to the fore age-old issues of sonic imaging on recorded media. Some of these challenges are discussed in the next section, followed by a reflection on the most prominent approaches. Finally, a hybrid composition setup is presented where various methods were used concomitantly to their own strengths, followed by a critical discussion and proposed further research. 1.1 The aesthetic aim: high fidelity spatial music As a multichannel music listener, the author is often left wanting by how space is composed. Despite some ambitions catalogued in literature reviews [2-3], works often fall into two tropes of multichannel affordances: on the one hand, enveloping immersive soundscape-type environments; on the other, spinning objects that lead to massive, often granular, invasive masses. Both can be very impressive when well executed, but the author is left with a desire for a breadth of other possibilities, embracing the soundscape resolutions described by R Murray-Shaffer [4] - from high fidelity to low fidelity - not Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 423 2 P.A.Tremblay as two categories, but as a continuum. In other words, how can composing space embrace the full range of spatial creations? These questions, quite important in the literature of electronic music in stereo [5-8], seem to be dismissed once musicians invest in multichannel endeavours 1 . The author hopes that with more affordable and accessible multichannel music-making means, we can now reach a nuanced level of artistic use of these technologies, beyond the novelty factor, with a polyphony of critical discourses, a wider participation, and the elevation of critical creative questions to the fore. But what are the main technical problems that hinder such democratisation? 1.2 The three families of problems None of the following points of concern are new to anyone working in multichannel music yet stating them is essential to set the assessment criteria of the workflow presented in later sections. They do not apply to all multichannel musicking, but after decades of discussions with peers, masterclasses, conferences, and hints in the papers cited above, theses hurdles seem frequent enough to be worthwhile of the following review. Channel count matters. It is important to state that most classic studio manipulations (tools and thoughts) are stereo processes, when not monophonic. Despite the renewed commercial interests in immersive audio, equipment rarely supports more than two channels, and editing software, while supporting multichannel workflows, are not often agile beyond eight channels. When using the studio as an instrument [14] and attempting to use high-channel count sonic material - from various competing array of microphones and/or synthesised as part of a sound design sequence - the mobility and routing between tools and processes gets cumbersome to say the least. Moreover, most mixing and mastering tools do not support high channel count, adding yet again to the burden of working with such material towards high-level production values. This is not to say that there are no great examples of very successful workflows that exist, but they are far from being streamlined. For instance, Natasha Barrett’s keynote at the 2023 AES International Conference on Spatial and Immersive Audio (AESICSIA-23) gave the attendees a glimpse of how her impressive multichannel music gets produced, but calling these processes streamlined would be a misrepresentation of the incredible work she had to do over the years against the tools she was using to achieve such results, as documented in her publications of the last decades [9-10]. In all cases, dealing with mono and stereo material across tools is still incredibly simpler, agile, and fluid, but this affordance condemns the space-making as a secondary role, at the end of the sound creation tool chain; this position, in turn, is a limiting factor in the compositional workflow and the aesthetic questions it can investigate fluidly. Portability matters. The idea of composing music for a site-specific setup might appeal to some, but for others, the desire for a transferable sonic experience between different venues is essential. This priority is shared by the game and film industry, 1 With a few notable exceptions, for instance the work by Natasha Barrett, first hinting at such intentions in [9], and further in [10-11]; another example are speculative stemming tactics by BEAST members [12]; Roads’s ideas of spatial chords [13], as well as a few rich oppositions, are also noted, sadly with little development of the promised spatial theory. Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 424 A Bit of Everything 3 which has also been a driver of standardised multichannel formats in the last three decades. Furthermore, many high-density loudspeaker arrays have been built recently, but even for their apologists, such setups are still seen as privileged [15]; in all cases, they do not have a wide-enough reach to trigger thorough critical aesthetic investigations. Beyond the issue of standardisation, another major problem persists: very few people have the opportunity to compose for prolonged periods in high-count multichannel facilities [2]. Therefore, portability between composition studios with smaller loudspeaker setups is crucial in the workflow. For some, recent head-related transfer function-based improvements are convincing: head-tracking, headphone timbre compensation, and customisable pinnae filters helped, but in the author’s experience with the latest renderers, we are still nowhere near the quality of listening of a reference studio. This need for portability cannot be overstated, if spatial composition research is to grow beyond the special effect, towards critical affective musicking in an immersive space. Indeed, the impressive first encounter with high-count multichannel might explain the omnipresence of a few tropes; yet if we are to compose space beyond the exciting novelty factor, we need to generalise portable methods towards more subtle, slow, deep modes of engagement with various aesthetics of immersive music. Another advantage of such approach is the ability to dynamically test how the music folds down to a lesser spatial resolution. This process allows for a refresh of the listening, and to assess immediately what is gained by the composed spatial components and what overly relies on them. It also helps to assess multimodal perception of space [16] beyond relying on a specific setup. Obviously, the diminished spatial resolution makes for a denser and more confused sound stage, but the result can be reflected upon fluidly. Imaging matters. Even if multichannel endeavours were to address age-old problems of auditory scene creation in fixed media, the tension between localisation and envelopment is far from solved [3]. Indeed, the aesthetics of object making - their placement and interrelations, and their interaction with a background - is only further brought to the fore when channel count augments. Once these compositional positions are taken, the elusive size for an optimal listening experience, aka the sweet-spot issue, seems to be amplified with the multiplication of loudspeakers [2, 17], which brings back another age-old tension that is worth restating: how to make sure that most listeners can enjoy a musical experience in line with the intended composed space is a complex problem. 1.3 The various formats of potential solutions In this context of enmeshed, contradictory aims and tensions, and with the expensive infrastructures inaccessible to most, it is easy to see why most spatial composition papers are either detached from the technological and psychoacoustic issues to stay aspirational, or they tend to stick to technological advances, with little reflection on how the techne influences the art it makes possible. In both cases, they keep ideas and tools at a platonic distance 2 , which is problematic to say the least: it ignores the fact that the decisions at the core of the toolmaking are shown to bias much of their usage [1]. 2 A few notable exceptions bridge the gap between the intentions and the tools used for their rendering: the latest of Barrett’s aforementioned publications, and Enda Bates’s PhD [18]. Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 425 4 P.A.Tremblay In all cases, technical formats for multichannel musicking can be classified in a few format families (or topologies [9]) and are worth reiterating: The loudspeakers as fixed points. Stemming from the traditional stereo setup, the fixed points approach gives each loudspeaker a name and assigns one track per loudspeaker. Despite the many standards of track order and of loudspeaker positions, this proves a sturdy delivery format, when accompanied with a clear layout plan and test files, for up to 16 channels. It enables bespoke setups too. Positioning between loudspeakers is usually done via simple amplitude panning, between two or three adjacent points, with (and despite) all the psychoacoustic limits of this approach [16-17]. The encoded sound field. Another approach is to encode the whole sound field and expect the player to implement a decoder optimised to their listening environment. High-order ambisonics (HOA) is the leading format here, despite its combinatorics nightmare across various conventions and the baked-in assumptions of the format [19]. The high channel count required for accurate spatial rendering makes the workflow of natively HOA material far from seamless. As the format relies on multi-speaker rendering of phantom locations, limitations described above are also to be kept in mind. The augmented loudspeaker orchestra. Both previous solutions assume that sounds are placed on a sphere. The long tradition of the loudspeaker orchestra [5], on the other hand, embraces the depth of a stage-in-a-room with the acoustics in which the concert happens. This was historically done with stereo sources, and certain institutions have enabled the sound projection practice to support up to octophonic sources [12]. The article also mentions some mixed approaches, where composers have encoded multichannel material to fixed points, with an extra stereo pair of musical material to be performed via sound projection. Disciplined stemming is also mentioned and provides a degree of in-situ adaptation to the variability of acoustics and loudspeakers. Yet, these approaches soon bloat the final channel count of massively multi-channel works, and thus make it harder to move between compositional setups, let alone to distribute. The not-yet-a-solution: object-based encoding. This technique is gathering momentum, where low-channel-count objects - typically mono but sometimes up to octophonic - are accompanied by control tracks to position them in a virtual space. It relies on site-specific decoders, whatever their technology. It has an advantage to keep material at the lowest channel count needed to represent it, a recurrent idea in stemming, thus lowering the overall channel count in the composition workflow. It has the disadvantage of putting an even greater pressure on the decoding process, adding the various control channels to the array of potential issues to troubleshoot 3 . There are also a few wave field synthesis systems that could be included in this category of object-based encoding, with the advantage of not having any optimal sweetspot, but the inconvenience of the technology’s rendering limits [9], and being still incredibly hardware-heavy and thus expensive and inaccessible to most. 3 An academic implementation is the Spat-GRIS (gris.musique.umontreal.ca), a continuation of the Zirconium (zkm.de/en/zirkonium), both biased towards a dome. A commercial implementation is the recent Dolby Atmos (www.dolby.com/technologies/dolby-atmos), with various tracks with fixed positions, yet allowing some tracks for position-encoded objects, decoded at listening time. These systems are panning to a phantom location between the nearest loudspeakers to the desired location, with the expectations and limitations mentioned above. Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 426 A Bit of Everything 5 Some projects with multiple decoding algorithms have waxed and waned over the years. For instance, the SoundScape Renderer project 4 was proposing such stem-tospace decoding via different algorithms, but the non-standard distribution protocol made it cumbersome to share material. There was also an effort at standardising the spatial encoding format [20], but the latest EBU attempt, the Audio Definition Model 5 , seems to have superseded it. Other formats are based around IRCAM’s SPAT technology 6 , usually embedded in bespoke patches for ad hoc distribution streams. 2 Personal Explorations In Multichannel Electroacoustic composition in up to eight channels is now a mature practice. The author, mostly active in post-acousmatic [21] and mixed music and with professional experience in popular music production, started to work in such a setup in 1996. Despite the technological hardships and the various failings of the intended sound stages in concerts, composing dialogues of spaces kept developing in the musical works. When the 5.1 format became popular, further compositional ideas were explored, as the said format was considered as enhanced stereo [22]. Yet it raised different aesthetic questions and biases. The format solves minimally all three problems spelt out in section 1.2. It sounded great, especially when embracing the low-count of channels as a portability-enabling feature: in the sound design period, it enabled the use of most highquality production tools while clarifying the composed inner space of the music; in concert, it enabled a dynamic projection over a loudspeaker orchestra; in distribution, the format was commercially available on DVD, and there was high hopes that many would at last embrace domestic sound system enabling space to play a strong compositional part. The hardware and software also allowed for quick fold-down mixes. Then, this commercial format all-but-died; concomitantly, the author received a commission for the 100+ channel BEAST 7 . Here was an opportunity to go much further in terms of spatial compositional explorations, but access to the final sound stage would be limited to the day of the premiere. But taking half a year to write a piece that might not render well in its unique performance was not conducive to a slow, in-depth musical investment. In all cases, the uniqueness of the concert setup raised questions of methodology: where and how to compose for such a system in meaningful ways? The process started with comparing the various spatial encoding methods described above in the SPIRAL 8 . This allowed the author to test their respective workflows and feel their compositional affordances, while doing some ear-training to assess how they sounded in the sweet spot and off-centre. As they were largely equivalent in terms of (cumbersome) workflow, the most striking conclusion was the large discrepancy between the claims laid by the various communities around each technology, and what 4 http://spatialaudio.net/ssr/ 5 https://adm.ebu.io/ 6 https://forum.ircam.fr/projects/detail/spat/ 7 https://beast.cal.bham.ac.uk/ 8 https://www.cerenem.org/environment/facilities/studios Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 427 6 P.A.Tremblay was heard in the space in terms of spatial resolution 9 . For instance, after comparing multiple HOA decoders, we observed that 7th-order was a minimum needed to match the precision obtained by direct mapping, yet that 3rd-order offered uniformly sounding blurred points on the sphere, which was better than the unevenness of phantom panning. Yet neither offered a real distance to the loudspeakers like acousmatic ‘distant’ – even with the techniques creating illusions of distance through filtering and reverberation. With these considerations in mind - where each technique excels, where the others are mediocre - the author took this opportunity to experiment with a streamlined and portable workflow that would cherry-pick techniques according to their aural strength. 3 Beyond Allegiances: A First Cherry-Picking Attempt 10 As a single loudspeaker produces an artificially narrow, yet very precise point source, a group of these would be composed as is, with no use of phantom images in between. A 5:2:1.0 11 layout was adopted, as it has a frontal choir, allowing for a precise “stage” sound image for all attendees, independent of listening position. This enabled polyphonic dialogues between segregated sharp objects. The eighth point was placed above the stage, slightly left-of-centre, to give a sharp frontal elevated source. Limiting these point sources to eight has the benefit of staying compatible with professional mixing and mastering plugins, in line with their role of high-quality, impactful sound sources. As mentioned above, lesser than 7th-order ambisonics produces a blurry halo, even in the sweet spot. Nevertheless, with a good decoder, it offers the flexibility of discrete placement and downmixing with an even timbre all over the sphere. This set of features would be used to contrast with the point sources: a 3rd-order ambisonics (TOA) has a usable track count (16) and gives enough precision to locate, yet enough blurriness to enable diffuse material to shine. Moreover, trajectories between points could be filled, and the blurriness could be used to push the material in the background. Finally, as both the 5:2:1.0 and the TOA were placed on the sphere, a quadraphony of distant speakers - one in each corner of the room - was implemented to use one of the most salient affordances of the traditional loudspeaker orchestra tradition. Now, having three cohabitating setups - one for precision of space and timbre, the other for width, envelopment, and movement, the third for true distance - a crucial question remained: how to make sure their relation is consistent during the composition process and at the concert. Even more importantly, as they each add a different colouration to the playback: how to enable the exchange of musical material between them and ensure that their complementarity is consistent in various listening environments? In line with the author’s calibration work in mixed music [24], it was a matter of implementing a simple-yet-systematic process to match the colour and loudness of the 9 This is in line with the conclusive plea in [17], of which the author was unaware of at the time. 10 Early articulations of the ideas in the following sections have been discussed in an interview in 2020 with Dr Brona Martin, partially documented in [23]. 11 The notation of loudspeaker layout follows the commercial x:y:z.s where x is the number of frontal loudspeakers, y of surrounds, z of elevated, and s of low frequency effect channels. Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 428 A Bit of Everything 7 three cohabitating techniques. A two-step approach was adopted. First, a quick yet thorough setup check: making sure that all loudspeakers work where expected and are matched in volume and polarity. With so many loudspeakers, there are often discoveries at this stage. A simple pink noise going from channel to channel suffices to test routing and sound quality; a similar noise going through all adjacent loudspeakers is good to test the phase and expected phantom location. Second, the HOA decoder is tested and calibrated in volume and timbre by ear to match the point sources. To do so, sampled pink bursts are sent to the point source speakers, and similarly to their equivalent encoded position in HOA. This allows for a quantitative and qualitative matching between the two formats in an efficient way. The bursts must use the same sampled noise, as the subtle changes of spectrum between two different bursts would distract from the matching gain and equalisation task. Similarly, the distant loudspeakers were calibrated by ear. Because of their radically different colour, due to their positioning, the matching was quite approximate and more quantitative than qualitative; yet allowed for offensive timbral and level differences to be tamed. Thus was composed Bucolic & Broken [25] - locally in the author’s institutional 25.4 studios, with local portability tests done in 5.1 and 8.0, and a stereo home studio; two residencies in each of Hull’s spherical ambisonics studio, and NOTAM’s hemispherical one, have enabled further road testing of the portability of the calibration; more importantly, they provided quality time for the artistic research of composing high-fidelity spatial components in the music, in confidence on the aural judgement of their contributions. This is achieved with a total channel count of 28, significantly lower than a 7th-order ambisonics requirement (64) for a comparable point-source precision. As far as an evaluation of the result of this first protocol can be done, the piece has been presented successfully in various formats, from the optimal setup of the premiere to a 5.1 album release, in academic conference settings, as well as 13 professional multichannel gigs so far. The disciplined calibration in concert enabled, in one case, the discovery of an old angular HOA decoder error. At every performance, the music was lauded for its spatial composition and the sharpness of its sound design. 4 Road Testing: N > 1 Having concomitant access to multiple high-count multi-channel studios is a rare privilege; so is an opportunity to test this methodology. Such an occasion happened five years later, when the author held a guest professorship at CIRMMT, in the period when its staff explored their newly refurbished MMR 12 , fitted with a custom 62-channel Meyer Constellation system. They also fitted an almost-spherical 32-channel Genelec studio in its control room, the PeRL. Both spaces had great acoustics with dedicated people. Both presented challenges too: the MMR’s rectangular shape leads to loudspeakers not being equidistant to a listening point; this new space was also very much in demand, limiting the project time in situ to nine full days, spread over the year. Yet, this sort of access is still incredibly privileged; more importantly, there was also the 12 https://www.cirmmt.org/en/facilities Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 429 8 P.A.Tremblay opportunity to access the other world-class CIRMMT studios around those dates to test the portability of the methodology, for a total of four one-week-long residencies. Moreover, this allowed the author to share this methodology and see firsthand how it could be useful to other composers. A call for postgraduate composers-researchers was made to share this experience, with the agenda to teach and test the workflow; more importantly, this enabled a sharing of sounds and thoughts on compositional use of space in the music-in-progress, and to observe how each composer would use it in their respective musical aesthetics. All participants could also assess the portability of the workflow and flag any concerns arising from its transparent agenda. This time, the setup was augmented to 14-point sources, and a fifth-order ambisonics (5OA), without any real distant planned for the concert 13 . The points were a doublediamond equidistant octophony at ear level, where the front was augmented by a pair on each side of the central speakers, thus giving 10 channels with an oversampled stage. A quadraphony of elevated points was also added in a square, to enable precise articulations at that level, forming a 5:4:4.0 setup. During the four weeks of residencies, members of the team had sporadic access to the MMR and the PeRL, and full access to the 3:0:4.1 CLL, as well as to a 3:5:4.0 custom-built setup in another studio. A schedule was designed to enable equal access to every space for everyone. In between these residencies, the author had access to Huddersfield’s SPIRAL, while the Montreal-based composers had access to the studios of their various host institutions, namely the three domes at Université de Montréal, the 22.2 DCS at McGill, the 10-channel Milieux/FFAR, the 8-channel Matralab, UQAM’s dome, and Concordia’s 26.2. At the end of the year, the informal yet critical evaluation of the workflow was that it enabled diverse works, if a little cumbersome to set up, and a deeper engagement with the aesthetic of space in five very different ways 14 . This latter comment was repeated numerous times at the premiere by members of the public, ranging from candid listeners to professional multichannel composers. Many of the pieces have been performed since then in various settings; in the case of the author’s [26], in two academic conferences as well as five professional multichannel settings so far, on various setups from horizontal octophony upwards. Again, the work was lauded for the clarity of its spatial composition, the transparency of its form, and the quality of its sound design, especially with its counterpoints of articulated basses and the depth of its images. All of this was again achieved in a channel count lower than a single stem of 7th-order ambisonics, i.e. the minimum requirement for sharp point sources. 5 Conclusion And Future Work All in all, the proposed workflow has enabled the composition, critically and over time, for massively multichannel loudspeaker arrays, via a quick and flexible portability-focused setup, despite the difficulties of accessing such facilities for long periods. 13 The author still composed for a quadraphony of distant loudspeakers as described in Section 3, but it has always been downmixed in the 5OA for all studio and concert settings so far. 14 The composers were Kasey Pocius (eTu{d,b}e de Labo #1), Philippe Macnab-Séguin (Gone for Eggs), Nicola Giannini (Architecture éphémère) and Yulin Yan (Apropos of Thrownness) Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 430 A Bit of Everything 9 Using each format to its strengths, it allowed for multiple successful composition projects to be realised in various settings, most of which much smaller than the targeted concert setup. The author hopes that sharing this methodological proposal, which was resource-heavy to test, will provide an open opportunity for wider participation in massively multichannel music making and thinking, beyond the novelty effect. Three strands of improvement to this report are easy to envisage. On the technical level, the approach relies on good equalising skills in situ. The initial idea was to test if quick, subjective adjustment was good enough to match the complementary formats, thus making the setup portable. And it does, to the extent of the composer’s ability to correct the problems - matching a multi-speaker (comb-filtered) point to a real pointsource is not a trivial challenge. One avenue would be to automate the process, in line with the author’s previous projects [24, 27]. If streamlined, it could enable sonic correction of all formats at the same time, in a more objective and reproducible manner. On the artistic level, a more thorough ear training resource could be developed to highlight the entangled skills of auditory scene analysis and synthesis, of spatial scene assemblage via a mastery of tools and techniques, and of the musical hopes and wishes often falling short in between the two. Such a resource should be illustrated with highquality multichannel examples, as in some recent tonmeister papers [28]. Embracing genre-normative modes of listening [29] and tastes in various styles of electronic music could be enabled by more of these transparent aural bridges and discussions 15 . On the methodological level, this workflow has been tested on six works, all with the author somehow involved. A growing number of commissions around various HDLAs would enable us to strengthen the proposal; the cost of such an endeavour makes it prohibitive, hence proposing it in its current state to our communities. On a hopeful note, the author hopes that the current commercial interest in immersive sound will provide a diversity of options, maybe with the EBU’s open source ADM as a standard to guarantee the future playability of the works, and solidify the critical discourse around spatial composition in various genres of fixed media works… or at least, that it survives long enough to solidify DAW and plug-in routing capacity to support our rich artform for another generation! Acknowledgments. The author would like to thank Annie Mathani and Scott Wilson at BEAST; Jörn Nettingsmeier for the survival HOA guide; Asbjørn Blokkum Flø and Notto Thelle at NOTAM, and Rob Mackay and Matthew Barnard at Hull; Bob Hasawaka, Fabrice Marandola, Julien Boissinot, Sylvain Pohu, Yves Méthot, and Jackeline Bednard at CIRMMT; the four aesthetic co-researchers (Kasey, Philippe, Nicola, and Yulin) for the sharing of thoughts and works; Natasha Barrett for the discussions; CeReNeM and FluCoMa for the travel support. Disclosure of Interests. The authors have no competing interests to declare that are relevant to the content of this article. 15 For instance, at the AES-ICSIA-23, Andrew Scheps spoke of his love for the phantom centre, its diffuseness helping to widen the body of a lead vocal, by opposition to the narrowness of a centre loudspeaker as a point source. A similar comment, on totally different musical aesthetics, was done [9], a welcome confirmation of an aesthetic problem the author faced in the past with certain musical material. Such discussions need a platform to strive. Proc. of the 17th International Symposium on CMMR, London, UK, Nov. 3-7, 2025 431