ICST Ambisonics Workshop: Working With Ambisonics

Learning Objectives

Contents

Ambisonics is a channel-format-independent, full-sphere audio system. Every production follows the same chain — regardless of whether the target is headphones, an 8-channel ring, or a 24-channel dome.

The Ambisonics Signal Flow

The golden rule: Render the B-format master track — never the decoder output. The decoder is for monitoring only. One and the same B-format master works for headphones, loudspeakers, and streaming.

Monitoring Setup

Before recording starts, monitoring must be in place:

• Binaural (headphones): Load ICST AmbiDecoder with HRTF preset → direct spatial feedback without loudspeakers
• Check decoder preset: Does the layout (channel count, speaker geometry) match the actual setup?
• Never run binaural and loudspeakers in parallel: Only one decoder active — otherwise phase errors in the mix
• Separate monitoring branch: Binaural monitor track on a dedicated bus, not looped into the B-format master

Formats & Normalisation

Rule of thumb: Always work and archive in AmbiX (ACN/SN3D). Use FuMa only for compatibility with legacy material.

Delivery Formats

Recording 3D Audio: The Recording Workflow

Workflow (Hands-on with Zylia → Main Reaper)

1. Short Recording Session — Set up Zylia ZM-1, check levels (−18 dBFS), record 30–60 seconds → 19-channel A-format PCM
2. A to B Converting — Zylia App: import A-format, export as 3rd Order HOA → 16-channel B-format AmbiX WAV
   A-format = PCM audio (regular audio data, raw capsules)
   B-format = Mathematical format (spherical harmonics, not directly listenable)
3. Listening B-Format — Load in Reaper, listen binaurally via HOA bus + AmbiDecoder: does "front" actually sound in front?
4. Upsampling & FX — EQ, reverb on HOA bus, distance automation on the encoder
5. Mastering B-Format — Render HOA Render Bus: 16 ch, 48 kHz, 32-bit float WAV → B-format master

Microphone Specifications

• 19-capsule array (omnidirectional)
• Output: 19 raw channels → encoded to 3rd Order HOA (16 channels, AmbiX) by the Zylia software
• Connection: USB-C, no external power required

Microphone Placement

The position of the microphone determines the listening perspective — this is not a technical but a compositional decision.

• Height: approx. 1.5 m for a natural ear-level perspective; lower for a “ground level” effect, higher for an overview perspective
• Orientation: The microphone has a preferred direction (marker). “Front” in B-format corresponds to this direction — document it so that decoding is correct
• Distance to source: Closer = more direct sound, more spatial detail; further = more diffuse field, less localisation
• Reflections: Hard walls, floors, and ceilings are very prominent in an HOA microphone — listen carefully to the placement location before recording
• Wind protection: Always use a windshield outdoors (blimp or fur). Activate LF roll-off below 80 Hz

Gain Management

• Target level: approx. −18 dBFS (check all 19 channels individually)
• Headroom: at least 12 dB — transients in HOA microphones can peak unexpectedly
• Monitor all raw channels separately: a single clipping channel corrupts the entire B-format
• Set the gain structure in the Zylia app, not in Reaper after the fact

A-to-B Conversion

The Zylia ZM-1 delivers A-format (19 raw channels). The Zylia software converts these to 3rd Order HOA (16 channels, AmbiX):

• When to convert? Immediately after recording, before importing into Reaper
• Quality control: Test phase and channel assignment — a mono test source from the front should sound clearly in front in the binaural monitor
• Archive raw data: Always keep the A-format (19 channels) — this allows re-conversion later with better software

Field Recording Checklist

Before recording:

• Acoustic survey: reflections, noise sources, wind direction
• Decide on microphone placement and document it (height, orientation, distance)
• Level check: all 19 channels at approx. −18 dBFS
• Binaural monitoring active and verified
• Metadata template prepared (location, date, take number, weather)

After recording:

• A-to-B conversion in Zylia app
• Load HOA file (16 ch AmbiX) in Reaper and verify binaurally
• Archive raw channels (A-format)

Reaper Session (pre-configured)

Common Recording Errors and Solutions

Composers Workflow — The Three Stages

Every Ambisonics production — whether field recording, synthesiser, or live performance — passes through three stages:

Encoding with ICST AmbiEncoder_64

Plugin: ICST AmbiEncoder_64 (VST3) — pre-installed on all workshop computers.

Automation: Moving Sound Sources

Physik-basierte Trajektorie: atan2 für Azimut, √(x²+y²) für Distanz.

Processing

Distance

In the ICST AmbiEncoder, Param 13 (Distance) simulates depth in space. The plugin attenuates level and high-frequency content with increasing distance — similar to air absorption in real space:

• Near sources (distance ~0): present, direct, high HF content
• Distant sources (distance ~1): quieter, duller, more diffuse-field content
• Combined with reverb: The greater the distance, the higher the wet amount in the room plugin on the HOA bus

Room / Reverb

The reverb sits on the HOA bus track — not on the source track. Only then does the reverb remain anchored in Ambisonics space and get decoded correctly with the B-format.

• ReaVerb or FX reverb on the HOA bus (after all encoders, before the decoder)
• Dry/wet ratio: start sparingly (~10–20% wet), as HOA reverb sounds very prominent
• Room size and pre-delay determine the size of the space
• Separate reverb sends per source allow different room distances

Example: Helicopter Flyover

The script calculates position, azimuth, and elevation directly from 3D coordinates (atan2, Pythagoras) — no manual angle estimation.

Decoding

• Binaural: ICST AmbiDecoder with HRTF file → 2-channel headphones
• Loudspeakers: Decoder with .spk configuration file
• Screensets: Ctrl+Alt+1 (Recording) · Ctrl+Alt+2 (Mixing) · Ctrl+Alt+3 (Decoding)

FuMa → AmbiX Conversion (FOA)

For legacy material in FuMa format (e.g. SoundField, older productions):

Reaper is not the only environment for Ambisonics. These tools are used by composers and researchers — each with a different approach and different strengths.

Workflow with MaxMSP: ICST Ambisonics Externals (ambicontrol, ambiencode, ambidecode). Real-time spatialisation, OSC control of the ICST AmbiEncoder in Reaper. Strength: generative and algorithmic spatialisation, live electronics.

Workflow with Csound (Cabbage.app): Csound opcodes bformenc1 / bformdec2. Cabbage wraps Csound as a VST3 plugin. Strength: precise mathematical control of the sound field, additive synthesis, granular synthesis.

Workflow with Max for Live: MaxMSP integrated in Ableton Live. Combines Ableton’s production workflow with Ambisonics spatialisation. Strength: live performance with Ambisonics, studio production.

Csound is a programmable audio synthesis language. Cabbage wraps Csound instruments as VST3 plugins and makes them usable in Reaper.

Cabbage → Reaper (DAW)

Key Csound opcodes for Ambisonics:

• bformenc1 — encodes a mono signal into B-format (1st–3rd order). Input: audio, azimuth, elevation. Output: HOA channels.
• bformdec2 — decodes B-format to a loudspeaker array.

Example: Additive Synth & Random Spatialisation (Cabbage)

A Cabbage example demonstrates: an additive synthesiser generates partials (sine waves) in real time. Each partial is independently spatialised with bformenc1 and a randomised azimuth/elevation. The position drifts randomly — the result is a moving cloud of sound in 3D space, then decoded binaurally.

MaxMSP with the ICST Ambisonics Externals provides a complete real-time Ambisonics environment. Unlike a DAW, Max allows algorithmic and generative control of spatialisation — positions can be driven by data, gestures, sensors, or other audio signals.

MaxMSP → Reaper (DAW)

ICST Ambisonics Externals:

• ambiencode~ — encodes mono audio into B-format. Message input for azimuth, elevation, distance.
• ambidecode~ — decodes B-format to loudspeakers or binaural.
• ambicontrol — GUI panel for manual spatialisation control.

ACT-Tool Example:

The ACT-Tool (Ambisonics Composition Tool) is a MaxMSP patch developed at ICST:

• 3D visual interface for placing and animating sound sources
• Sources can be moved manually (mouse/trackpad) or algorithmically (LFOs, envelopes, data streams)
• OSC output: sends azimuth, elevation, and distance to any Ambisonics encoder — including the ICST AmbiEncoder in Reaper
• The ACT-Tool can be used as a “spatial score”: define movement visually, record OSC output as automation in Reaper

Hands-on Exercise

Guided exercise with the prepared sessions:

1. Open the Zylia session, verify microphone routing and decoder preset
2. Route a mono test source to the HOA bus, listen binaurally
3. Run the encoding script: position source at azimuth −45°, elevation +20°
4. Start the train pass-by automation and follow the trajectory binaurally
5. Encode your own source and move it in space (automation or manually)
6. Render and verify the B-format master (16 ch, not 2 ch!)

A guided discussion to consolidate the workshop experience. Not a lecture — participants share their own observations and develop ideas together.

Topics:

• Space as compositional material: Can space itself be the instrument? What happens when the space moves and the source stays still?
• The composer’s perspective: How does working in Ambisonics change compositional decisions compared to stereo?
• Elevation and expressivity: When does height feel natural, when artificial? What does the vertical axis enable that stereo never could?
• Distance and reverb: How do distance and reverb interact? When does a source sound “inside the space” rather than “outside”?
• Outlook — follow-up workshop: MaxMSP + Csound in depth (Blocks D, E, F).

Preparation & Downloads

• Reaper is pre-installed on all workshop computers (v7+, incl. ICST plugins)
• Bring headphones (binaural monitoring)
• Download the pre-configured Reaper session in advance