Infected by Immersion
The Satosphère dome at SAT Montréal
Deep Dive October 5, 2026 Spatial Audio

One Scene, Many Domes: The Satosphère Sound System and How to Take a SAT Soundscape on Tour

The SAT Satosphère in Montréal has one of the densest speaker arrays in any dome. Works made for it rarely travel well, because most of them are mixed straight to its speakers. Here's what SAT actually documents about the system, the spatial-audio science that limits portability, and a practical plan for keeping a SAT-grade soundscape intact on 16-speaker domes, 5.1 planetariums and headphones.

Contents
  1. What we know about the Satosphère audio system
  2. The science: why 93 channels don't simply shrink
  3. Strategy: one scene, many renders
  4. Venue tiers
  5. A step-by-step plan
  6. Open questions
  7. Sources
The short version: never treat the 93-channel speaker mix as your master. Master a layout-independent scene instead: point-source objects with movement data, a higher-order Ambisonics bed and a separate sub-bass stem. Then render that scene live, fresh for every venue, from a single speaker-layout file per room.

🔊 1. What we know about the Satosphère audio system

SAT's own pages describe the dome as 18 m in diameter and 11.5 m high, with 8 video projectors and "93.5 haut-parleurs": 93 full-range speakers and 5 low-frequency channels.[1][2] In a June 2024 article SAT adds that the Satosphère is a 210-degree, non-inclined hemispherical dome over a dance floor. The room is modular and takes up to 750 people, and the speakers are spread over 4 height levels. The system was entirely revamped thirteen years after the 2011 opening.[3]

Dome18 m diameter, 11.5 m high, 210°, not tilted[1][3]
Speakers (official)93 full-range + 5 low-frequency channels ("93.5")[2]
Speaker layout4 height levels[3]
AudienceUp to 750, modular room, flat floor[3]
Spatialisation in practiceSpatGRIS (GRIS / Université de Montréal with SAT), often driven from Ableton Live[7]
Alternative count in circulation"157 channels" (unresolved)[4][5]

93.5 or 157?

A RISD/Brown concert note says a piece was installed at SAT Fest 2024 "using the Satosphere's 157 channel loudspeaker array".[4] Électro-Acrylique's Satosphère page also lists 157 speakers.[5] SAT's main venue page still says 93.5. The gap could come from a different counting method, the post-revamp setup or an extended configuration. We haven't been able to resolve it. For anyone producing for the dome, the current speaker plan from SAT (positions and channel map) is the most important input there is.

Why "not tilted" matters for sound

Many planetariums are tilted domes with a front. The Satosphère is a level hemisphere over a flat floor where people stand, walk or lie down.[3] There's no front and no sweet spot. Every listener hears the dome from a different position, and that shapes which spatialisation methods work (see section 2).

The tools around the dome

🧪 2. The science: why 93 channels don't simply shrink

A speaker mix is a render, not a master

A 93-channel file is already rendered for one geometry. Fold it onto 16 speakers and each speaker carries several feeds that were meant for different directions. You get comb filtering, phantom images collapsing onto the nearest box, and a muddy upper hemisphere. The portable unit has to be the scene, meaning sources plus positions, and not the speaker feeds.

Three ways to render

Amplitude panning (VBAP / MBAP / DBAP)

Places each source between the 2–3 nearest speakers. Sharp, and robust for listeners far off-centre, which is ideal for a flat-floor room with no sweet spot. This is what SpatGRIS DOME mode does.[8] It works on any layout as long as you know the speaker positions.

Higher-Order Ambisonics (HOA)

Encodes the whole sound field as spherical harmonics: order N uses (N+1)² channels (3rd = 16, 5th = 36, 7th = 64). Irregular layouts and hemispheres are decoded with AllRAD. IEM's AllRADecoder builds decoders up to 7th order for any layout and fixes the missing lower half with an imaginary speaker below the listener, set to zero gain.[11] HOA is great for beds and ambiences, but at low orders the sweet spot is small.

Wave-field synthesis / beamforming

Rebuilds wavefronts physically. That needs dense, purpose-built arrays such as the Sphere's Holoplot system or RISD's custom WFS array.[4] It isn't something a touring dome show can rely on.

Order, speaker count and audience area

As a rule of thumb, a decoder of order N wants at least (N+1)² well-distributed speakers. So 93 speakers can in principle carry around 7th order, a 16-speaker dome around 2nd–3rd order, and 5.1 roughly first order in the horizontal plane only. The area that sounds right grows with order and shrinks with frequency. Listeners off-centre in a big dome also hear the nearest speaker first (the precedence effect), so low-order point sources get pulled towards the wall next to them. That's why big-dome practice uses amplitude-panned objects for anything that must localise and HOA for diffuse beds. Zotter & Frank's open-access Ambisonics is the standard reference for the theory.[12]

From our own bench: a 3rd-order, 16-channel HOA decode on a virtual hemisphere gave about 12.8° average direction error, close to 20° at the horizon and pulled upwards. That's the textbook behaviour of a low-order decode on a hemisphere. Keep sync-critical events as objects, not inside the bed.

Dome acoustics and bass

Concave domes focus reflections towards the centre. The results are flutter echoes, "whisper" spots and a boomy middle, and percussive, transient-heavy music exposes them hardest. Every venue needs its own delay and level alignment per speaker, plus decoder-output EQ. Below about 80–120 Hz bass is barely directional, so it should travel as its own LFE stem and get bass-managed to whatever subs the room has: five at SAT, four d&b E15X at CULTVR Cardiff, maybe one in a planetarium.

🧭 3. Strategy: one scene, many renders

The master: a scene package

  1. Objects: roughly 8–32 mono stems with azimuth/elevation/distance automation, for everything that has to localise (hits, voices, moving elements).
  2. HOA bed: 5th order (36 ch, ACN/SN3D) if possible, 3rd order (16 ch) as the minimum. Ambiences and reverbs go here. You can lower the order by simply dropping the higher channels.
  3. LFE stem(s), kept out of the scene.
  4. Metadata: the open EBU Audio Definition Model (ADM in BW64) can carry objects, HOA and channel beds together, and the EBU ADM Renderer (EAR) is an open-source baseline renderer.[13] For live use, the same trajectories can run as OSC into SpatGRIS.
  5. Reference renders: binaural, 5.1/7.1 and stereo.

Live: render on site

For a live A/V show the scene is rendered in real time. Ableton or SuperCollider feeds the objects and the HOA bus into a renderer on the show machine (SpatGRIS, IEM Plug-in Suite or SPARTA), which drives the venue's speakers over Dante or MADI. Positions travel as OSC, so the visual engine and the audio renderer share one coordinate bus, and sound and image stay together in any dome. The only thing that changes per venue is a layout file. It holds the speaker positions and channel map, and from it you generate the VBAP triangulation, the AllRAD decoder and the delay/gain/EQ preset.

🏛️ 4. Venue tiers

Tier S: Satosphère (93.5, possibly 157 channels)

Objects through SpatGRIS DOME (VBAP). The HOA bed is decoded at 5th order onto the same outputs. The LFE stem is bass-managed to the 5 subs. Prepare in the 97-speaker studio if SAT gives access.

Tier A: 3D domes with 16–40 speakers

Example: CULTVR Cardiff, 16 × d&b 5S plus 4 subs, either a 1:1 patch over Dante or d&b Soundscape objects. Objects go through VBAP on the venue layout, and the bed is decoded at 3rd order. Phantom sources are spaced wider here, so thin out dense object passages and let the bed carry the atmospheres.

Tier B: planetarium 5.1 / 7.1

Objects are panned onto the horizontal ring, with elevation suggested through level and spectral cues. The bed becomes a first-order horizontal decode or a premixed channel bed. Height is lost, so movement and dynamics have to carry the show.

Tier C: stereo and binaural

A binaural render for headphones, juries and press, and a stereo fold-down for streaming. Head-tracked binaural (IEM BinauralDecoder, SPARTA) makes remote rehearsal possible.

Rules that keep it portable

🗺️ 5. A step-by-step plan

0
Get the facts. Ask SAT for the current speaker count, positions (azimuth/elevation/distance per channel), sub positions, interface (Dante/MADI), house renderer and version, latency budget, and access to the studio. Don't optimise anything before this.
1
Scene format and toolchain. Fix the scene package (objects + 5th-order bed + LFE + OSC trajectories, with ADM as the archive format). Build a generator that turns one layout file per venue into an AllRAD decoder and a SpatGRIS speaker setup. First layouts to model: Satosphère, CULTVR 16+4, IEM CUBE Graz, generic 5.1 and a 32-speaker reference hemisphere.
2
Test renders. Render one 3–5 minute piece to every tier and listen binaurally through each virtual layout. Measure direction error per tier, then test in a real 3D room.
3
Tune in the Satosphère. During a residency, set delays and EQ, find the object/bed balance and check sub behaviour with a roaming audience. Feed the results back into the scene as metadata, never as baked audio, then re-derive the touring renders.
4
Touring kit. One show machine, a Dante interface, a layout file per venue and a 30-minute calibration: measure distances, set delay/gain, do a quick pink-noise EQ, then walk-test 8 reference objects.

❓ 6. Open questions

We'll update this article as answers come in.

Venue Sound Systems → Art in the Dome: SAT Montréal → Going Live in the Dome →

📚 Sources

  1. SAT – Satosphère (FR)
  2. SAT – Satosphere (EN)
  3. SAT – Fulldome immersive creation: A panorama (June 2024)
  4. Brown Arts Institute – Residual Noise: SRST Spatial Audio Concert
  5. Électro-Acrylique – EA+Satosphère
  6. CTVM – 30 ans pour la SAT (April 2026)
  7. SpatGRIS – GRIS-UdeM on GitHub
  8. GRIS – SpatGRIS & ControlGRIS
  9. SATIE – SAT GitLab
  10. IRCAM Forum – xp4l
  11. IEM – AllRADecoder guide
  12. Zotter & Frank – Ambisonics (Springer 2019, open access)
  13. EBU ADM Guidelines – Rendering / EAR

Published October 5, 2026 · Venue facts from SAT's own pages where available. Statements without a citation reflect established spatial-audio practice, not venue-confirmed specs.