AURA

Overview

THE AURA (Atmosphere Using Reconstructed wave field with Array) is a 192-channel Wave Field Synthesis (WFS) array. It is built as a dense, single-layer aperture that implements a 2.5D WFS scheme: closely spaced full-range loudspeakers act as secondary sources, driven to reconstruct a target wavefront within the audience plane. Rather than optimising sound for a single sweet spot, the array reconstructs the sound field itself, so that spatial relationships and localisation cues hold across the listening area for both seated and freely moving audiences.

Design philosophy

The array is intended to read as an architectural element rather than a scenic intrusion. Its design follows Dieter Rams' principles of good design: the enclosures are low-profile and visually recessive, and operation is reduced to the essential gesture — connect power, connect signal. There are deliberately no volume controls or power switches on the modules. The transport is kept license-free and future-proof so that the line-level signal can be provided by any present or future interface.

What this manual covers

• Safety practices for electrical operation, rigging, and exposure.
• The physical module: drivers, enclosure, mounting.
• Full technical specifications.
• Assembly into full, partial, and curved configurations.
• The MADI signal backbone and digital-to-analog conversion chain.
• Powering, and a step-by-step setup & installation procedure.
• PanWFS — the SuperCollider rendering software — install, coordinate system, the four UGens, parameters, and worked examples.
• Maintenance, service, and troubleshooting.

Safety

Electrical

• Each module is self-powered by an internal 120 W laptop-style supply. Use the supplied mains cabling and a correctly earthed outlet.
• Inspect cabling for damage before each deployment. Replace, do not repair in the field.
• Heat is conducted into the aluminium chassis, which acts as a passive heatsink. Do not obstruct the enclosure faces or stack heat sources against them. The system is ventless and requires no active cooling.

Rigging and mounting

• Each module weighs approximately 20 lb (9 kg). Use the rated VESA mount or the light-stand spigot adapter; never suspend a module by its cabling.
• For temporary rigs on light-stand spigots, confirm the stand's load rating and secure with a safety tie where overhead.
• Verify all modules are mechanically captive before applying signal.

Hearing and exposure

System Architecture

The module

THE AURA is composed of 24 modular speaker units. Each module is one metre long, 12.5 cm high, weighs about 20 lb, and holds eight drivers — eight individual mono emitters. A single module therefore functions as an 8-channel WFS array in its own right. Placing all 24 modules side by side yields the full 192-channel configuration spanning 24 metres.

Geometry

The array is an equidistant linear array: all modules sit on a common axis with uniform spacing, and all drivers are spaced 12.5 cm apart. This spacing fixes the spatial sampling of the aperture and therefore the alias-free band of the 2.5D WFS reconstruction — approximately 1.37 kHz — which preserves the interaural time and level cues the auditory system uses for localisation while retaining usable low-frequency response.

Drivers

The chosen driver is the 3-inch Dayton Audio RS75-4, a full-range unit with a near-flat response below 10 kHz. Coupled to the aluminium chassis, the low-frequency output is reinforced to compensate for roll-off below 100 Hz, aiding reproduction of sub-bass content without a separate subwoofer in many rooms.

Enclosure

Enclosures are folded from quarter-inch aluminium and welded. Aluminium is unconventional as a loudspeaker material because of its resonant character, but it is chosen here for durability, dimensional stability, and its function as a heat sink for the embedded amplifiers — eliminating the need for vents or active cooling. The low profile lets the array mount along stage edges, risers, or wall reveals and recede visually at distance.

Technical Specifications

Assembly & Mounting

Spacing and alignment

Modules sit on a single axis. Maintain the 12.5 cm driver spacing across module joints as well as within modules so the aperture samples uniformly. Aim modules toward the listening regions and install around 1.7 m for standing audiences.

Mounting options

• VESA mount. The bottom of each module carries a VESA pattern for fixed installation along stage edges, risers, or wall reveals.
• Light-stand spigot. An adapter accepts standard light-stand spigots for temporary rigs.

Configurations

The array is modular by design. Beyond the full 24 m line, subsets of modules can be deployed, and curved or segmented layouts can follow a room's architecture. See Configurations for recommended layouts.

Signal Flow & Routing

The array accepts line-level signal from any multichannel interface; the number of modules used can be matched to the channel count an interface provides. The reference backbone is MADI, chosen over Dante/AoIP for fixed low latency, stable clocking, license-free operation, and a simple signal-flow schematic.

Reference chain

1. RME MADIface XT II acts as the master, distributing three discrete MADI streams (two optical, one BNC), each carrying 64 channels.
2. Each 64-channel stream feeds two RME M-32 DA converters daisy-chained — the first set to channels 1–32, the second to 33–64.
3. Six M-32 DA units therefore cover all 192 D/A paths.
4. From each M-32 DA, four DB-25 connectors carry 8 line outs each, routed to the modules (8 drivers per module).

The shared clock is carried on the MADI stream, keeping the whole system in a single sync domain with known, low latency.

Scaling down

The architecture scales cleanly. A single M-32 DA drives up to four modules; a single module can be driven from any 8-channel interface using a DB-25-to-TRS or DB-25-to-XLR snake for small-scale tests or partial deployments.

Power

Each module is self-contained: an internal 120 W laptop-style supply feeds the eight on-board amplifier channels. Operation is plug-and-play — connect mains and connect the DB-25 signal cable; there is nothing to switch or set on the module itself. Heat is conducted into the chassis, so the system runs silent without fans.

Setup & Installation

This is the end-to-end procedure for bringing THE AURA up in a room: mounting the speakers, running the cabling, connecting the converter chain to a computer, and verifying the system before performance. The conceptual topology lives in Signal Flow & Routing; this chapter is the practical checklist.

1. Plan the layout

• Decide the configuration and module count (full 24 m line, a partial line, or a curved/segmented layout — see Configurations).
• Mark the array line on the floor or rig, keeping a continuous, uniform 12.5 cm driver spacing across every module joint.
• Confirm enough mains circuits for the module count, and plan cable runs from the converter position to each module.

2. Mount the speakers

• Fix each module via its VESA pattern, or seat it on a light-stand spigot adapter for a temporary rig.
• Set the array at roughly 1.7 m (standing ear height) and aim it toward the listening area.
• Confirm every module is mechanically captive and the line is level before applying power or signal.

3. Power the modules

• Connect each module's internal supply to mains, distributing modules across circuits so no single circuit is overloaded.
• There are no switches: a module is live once mains is connected. Verify the line is mounted and safe first.

4. Run the signal cabling

• From each RME M-32 DA, run its four DB-25 outputs to the modules — eight balanced line channels per DB-25, one DB-25 per module.
• Use DB-25 snakes or DB-25-to-TRS / DB-25-to-XLR breakouts that follow the TASCAM (AES59) analog pinout (Appendix A).
• Label every run by module and channel block so the mapping is unambiguous at both ends.

5. Connect the converters and the computer

• Connect the RME MADIface XT II to the host computer (USB).
• Carry the three MADI streams (two optical, one BNC) to the six M-32 DA units — two units daisy-chained per stream, the first on channels 1–32 and the second on 33–64.
• Establish a single clock domain: set one device as clock master and slave the rest to the MADI stream.
• Set the sample rate to 48 kHz for the full 192-channel deployment.

6. Configure routing in TotalMix FX

• In RME TotalMix FX, map the host's software-playback channels to the MADI outputs feeding each M-32 DA.
• Confirm the channel order so that software output n reaches the intended driver. A mis-mapped matrix is the usual cause of "channels routed wrong."

7. Verify and calibrate

• Send a test signal to one channel at a time and confirm the correct driver sounds, walking the full count to validate the mapping.
• Set conservative levels, then load a PanWFS scene (Software: PanWFS) and walk the room to check stable localisation and parallax.

Software: PanWFS

PanWFS is an open-source SuperCollider extension developed alongside THE AURA and compatible with any 2.5D equidistant linear WFS array. It treats WFS authoring as another panning algorithm: define a virtual source position (and, for directional sources, an orientation) and PanWFS synthesises the wavefront by computing per-loudspeaker delay and amplitude from the Rayleigh/Huygens formulation. Doppler shift emerges naturally from changing propagation delays — no separate algorithm is required.

Installation

Copy the PanWFS folder into your SuperCollider Extensions directory and recompile the class library:

SuperCollider/Extensions/PanWFS/Classes/*.sc

The extension provides four classes — two that render onto the array and two for headphone monitoring:

• PanWFS_Monopole and PanWFS_Cardioid — render virtual sources onto the loudspeaker array.
• PanWFS_Headphone and PanWFS_Binaural — preview the same scene on headphones, sharing the array's coordinate system.

Confirm the install by evaluating PanWFS_Monopole in the interpreter.

Coordinate system

Positions are given in metres on a Cartesian grid, with the centre of the array as the origin (computed from the driver count and spacing):

• virtual_x: negative to the left of centre, positive to the right.
• virtual_z: positive behind the array (non-focused sources); negative in front (focused sources).

Rendering UGens

PanWFS_Monopole

Renders a single virtual point source for monopole (omnidirectional) sources — radiating equally in all directions — over a straight-line loudspeaker array using 2.5-D WFS. It outputs one signal per channel, applying the per-speaker delay and amplitude from the Rayleigh/Huygens formulation, and can place both non-focused (behind the array) and focused (in front of the array) sources.

SynthDef(\WFS, {
    arg out = 0, amp = 0.5, virtual_x = 0.0, virtual_z = 2.0;
    var sig;
    sig = WhiteNoise.ar * EnvGen.kr(Env.perc(0.01, 0.2), Impulse.kr(2));
    sig = PanWFS_Monopole.ar(sig, virtual_x, virtual_z, 8, 0.127);
    Out.ar(out, sig);
}).add;

PanWFS_Cardioid

Renders a single virtual point source for directional sources (such as a voice or a human head) over a straight-line loudspeaker array using 2.5-D WFS. It behaves like PanWFS_Monopole with an added yaw that sets the source's facing direction, and likewise supports both non-focused and focused sources.

SynthDef(\WFS, {
    arg out = 0, amp = 0.5, virtual_x = 0.0, virtual_z = 2.0;
    var sig;
    sig = WhiteNoise.ar * EnvGen.kr(Env.perc(0.01, 0.2), Impulse.kr(2));
    // yaw = -pi/2 faces the audience
    sig = PanWFS_Cardioid.ar(sig, virtual_x, virtual_z, -pi/2, 8, 0.127);
    Out.ar(out, sig);
}).add;

Headphone-monitoring UGens

These let a spatial scene be sketched and previewed on ordinary headphones using the same Cartesian source coordinates as the rendering UGens, so a composition can be drafted away from the room and then rendered on the array unchanged.

PanWFS_Headphone

A headphone panner using a simplified binaural model based on ITD, ILD, and head-shadow cues. It computes source-to-ear distance, derives interaural time differences from geometric delay, applies level differences through distance- and side-dependent attenuation, and adds a simple frequency-dependent head-shadow on the far ear. Because it avoids full HRTF filtering it is low-CPU and portable — ideal for rapid sketching, testing trajectories, and composing in PanWFS coordinates — though it is not a substitute for individualised binaural realism. With Apple AirPods, the Headitude app can route head-tracking data to headyaw for active listening through natural head movement.

SynthDef(\WFS, {
    arg out = 0, amp = 0.5, virtual_x = 0.0, virtual_z = 2.0,
        listener_x = 0.0, listener_z = (-2.0), headyaw = 0;
    var sig;
    sig = WhiteNoise.ar * EnvGen.kr(Env.perc(0.01, 0.2), Impulse.kr(2));
    sig = PanWFS_Headphone.ar(sig, virtual_x, virtual_z, listener_x, listener_z, headyaw);
    Out.ar(out, sig);
}).add;

PanWFS_Binaural

A headphone panner using a lightweight parametric HRTF-style model. Instead of full measured-HRIR convolution it derives interaural arrival-time differences from source-to-ear geometry, applies distance-dependent attenuation, and shapes each ear with a small set of filters approximating head shadow and basic pinna-related spectral cues. It is a translation-oriented preview tool that preserves the spatial intent of a PanWFS scene during composition — not a replacement for the array or an individualised HRTF renderer. As with PanWFS_Headphone, Headitude can drive headyaw from AirPods head-tracking.

// HRTF parameter set (pass as the hrtf argument, or use nil for defaults)
(
~paramHRTF = (
    headWidth: 0.18,        speedOfSound: 343.0,   maxDelay: 0.05,
    lagTime: 0.02,          minDistance: 0.125,    distRef: 1.0,
    shadowFc: 3500,         shadowRs: 1.0,         shadowDb: -10.0,
    lowShelfFc: 250,        lowShelfRs: 1.0,       nearLowBoostDb: 2.0,
    nearRefDistance: 0.6,
    pinnaPeakFcFront: 3200, pinnaPeakFcSide: 4200, pinnaPeakRq: 0.8,
    pinnaPeakDb: 4.0,
    notchFcFront: 8500,     notchFcRear: 6000,     notchRq: 0.45,
    notchDb: -8.0,
    rearHiShelfFc: 5000,    rearHiShelfRs: 1.0,    rearHiShelfDb: -4.0,
    outputGainDb: 0.0
);
)

SynthDef(\WFS, {
    arg out = 0, amp = 0.5, virtual_x = 0.0, virtual_z = 2.0,
        listener_x = 0.0, listener_z = (-2.0), headyaw = 0;
    var sig;
    sig = WhiteNoise.ar * EnvGen.kr(Env.perc(0.01, 0.2), Impulse.kr(2));
    sig = PanWFS_Binaural.ar(sig, virtual_x, virtual_z, listener_x, listener_z, ~paramHRTF, headyaw);
    Out.ar(out, sig);
}).add;

Choosing zRef

Reflections and reverb

For an open-field character, a bare PanWFS source has no room cues. To place a source in an acousmatic room, early reflections can be built with the image-source method — mirroring the source across each wall and feeding each image back through PanWFS_Monopole as an additional virtual source, with a low-pass filter approximating wall absorption. A diffuse late tail can be spread across all loudspeakers with small random per-channel delays and gentle filtering, keeping the tail enveloping and non-localisable while the direct and early field preserve precise localisation.

Configurations

• Full line (24 m / 192 ch). 24 modules along a stage edge for room-scale field making and wide, communal movement of sources.
• Partial line. Any subset of modules matched to the available interface channel count.
• Single module (8 ch). One module driven from any 8-channel interface for tests, teaching, or small rooms.
• Curved / segmented. Layouts that follow the architecture; keep uniform driver spacing across segments.

Maintenance & Service

Serviceability is built in. The rear cover is removable, exposing the input and power board. Drivers and amplifiers connect on screw terminals rather than solder, so a failed part is swapped in minutes and any fault localises to a single card or driver.

• Replace an amplifier card. Remove rear cover; release the card from its terminal block; fit the replacement; re-seat terminals.
• Replace a driver. Disconnect the driver's screw-terminal leads; unbolt; fit replacement; re-torque and reconnect.
• Periodic. Inspect terminal tightness and cabling; verify chassis is clean for heat dissipation.

Troubleshooting

DB-25 (TASCAM) Pinout

The module input is a DB-25 connector wired to the TASCAM (AES59) analog standard for eight balanced line-level channels. This is the widely adopted analog DB-25 pinout used by RME, Avid, MOTU, SSL, and others. Pin 13 is not connected.

Further Reading

• Kerem Ergener, “Acousmatic Spaces, Lived Places: Holophony and the Embodied Experience of Shared Atmospheres,” Computer Music Journal (MIT Press). doi:10.1162/COMJ.a.693
• Kerem Ergener, Phenomenology of Atmosphere and Sonic Embodiment of Architectural Spaces and Places (Ph.D. dissertation, Louisiana State University, 2026). repository.lsu.edu/gradschool_dissertations/7015

Photographs

THE AURA in the studio and in performance. Select any image to view it larger; use the arrow keys or the on-screen controls to move between photographs.

Glossary

Holophonic audio — Reproduction that reconstructs the sound pressure field itself, so virtual sources behave as if in physical space and remain consistent for listeners across the area — communal rather than personalised.

2.5D WFS — A practical reduction of full 3D WFS to a horizontal line of sources, amplitude-correct along a chosen reference line (zRef).

Secondary source — A loudspeaker driven so the superposition of all speakers reconstructs the target (primary) wavefront.

Focused source — A virtual source rendered in front of the array (negative virtual_z), synthesised with a time-reversed phase.

Spatial aliasing limit — The frequency above which driver spacing can no longer sample the wavefront cleanly (≈ 1.37 kHz here).