Ambisonics is a full-sphere spatial audio technique that represents a sound field, including height, as a set of spherical harmonic components rather than as signals for specific loudspeakers. In spatial audio, it is the main scene-based format: the recording or mix describes the sound around a listening point, and a decoder later renders it for any speaker layout or for headphones.
The approach was developed in the 1970s, largely through the work of Michael Gerzon and colleagues in the United Kingdom. It saw limited commercial use for decades and then became central to 360-degree video, virtual reality and immersive production, because a sound field in ambisonic form can be rotated easily to follow head movement.
How ambisonics encodes a sound field
Ambisonics decomposes the sound arriving at a point into spherical harmonics, a set of directional patterns of increasing detail. The zeroth-order component is omnidirectional and captures overall pressure. The three first-order components are figure-8 patterns pointing front-back, left-right and up-down. Higher orders add patterns with more lobes, which describe directions more sharply.
The first-order set is called B-format and consists of four channels traditionally labeled W, X, Y and Z. A source placed in the mix is encoded by weighting its signal into each channel according to its direction, so direction is stored as amplitude relationships between channels, not as a choice of speakers.
The number of channels grows with the ambisonic order N as (N+1) squared. Higher orders improve localization precision and enlarge the area where the reproduction is accurate, at the cost of more channels, more processing and, for recording, more complex microphones.
| Order | Channels | Spatial resolution | Typical use |
|---|---|---|---|
| First order (FOA) | 4 | Broad, soft localization | 360-degree video, field recording |
| Second order | 9 | Noticeably sharper images | Games and VR audio engines |
| Third order | 16 | Good precision for most listeners | Immersive mixes, binaural rendering |
| Higher orders (4 to 7) | 25 to 64 | High precision, larger sweet area | Research, installations, large arrays |
Recording, channel conventions and decoding
First-order recordings are usually made with a tetrahedral microphone of four cardioid or subcardioid capsules. Its raw output, called A-format, is converted to B-format by matrixing and filtering. Higher-order recording requires spherical microphone arrays with dozens of capsules, and even these suffer from spatial aliasing at high frequencies.
Two channel conventions are in use, and mixing them up is a frequent source of errors. They differ in channel ordering and normalization, which changes the relative levels of the components.
| Convention | Channel order | Normalization | Status |
|---|---|---|---|
| AmbiX | ACN, ordered by degree and index | SN3D | Current standard for most tools and platforms |
| FuMa | Traditional W, X, Y, Z order | MaxN with W reduced by 3 dB | Legacy first-order and older material |
| N3D variants | ACN | Full three-dimensional normalization | Some research and higher-order tools |
Playback requires a decoder, which converts the ambisonic channels into loudspeaker feeds for a specific layout, ideally a regular arrangement around the listener. For headphones, a binaural decoder renders virtual speakers through head-related transfer functions, and head tracking applies a matching rotation of the sound field before rendering.
A loudspeaker decoder works best with at least as many speakers as ambisonic channels, so a first-order signal needs four or more speakers and a third-order signal sixteen or more for full benefit. Decoders also apply dual-band optimization: a velocity-oriented decode at low frequencies and an energy-oriented decode at higher frequencies, which matches how the ear localizes sound in each range.
In a mixing workflow, mono or stereo sources are placed with an ambisonic panner that encodes them at a chosen azimuth and elevation. The whole scene can then be rotated, mirrored, focused or widened with matrix tools before a single final decode, which keeps the master independent of any particular playback system.
Ambisonics compared with other spatial formats
Spatial audio formats fall into channel-based, object-based and scene-based families. Ambisonics is the main scene-based format, and its strengths and weaknesses follow from that model.
| Approach | What is stored | Playback flexibility | Main limitation |
|---|---|---|---|
| Stereo and surround sound | Signals for fixed speaker positions | Tied to the intended layout | No height or rotation without remixing |
| Object-based audio | Sound objects with position metadata | Renders to many layouts | Many objects require complex rendering |
| Ambisonics | Spherical harmonic sound field | Any layout or headphones, easy rotation | Limited sharpness at low orders |
| Binaural audio | Two ear signals for headphones | Headphones only | Fixed to one head orientation and HRTF |
Formats such as Dolby Atmos combine channel beds with objects, while some production systems include an ambisonic bed for background ambience. In practice, ambisonics excels at diffuse, enveloping ambience, and object-based audio excels at precisely placed individual sources.
Practical use and limitations
Ambisonics is the natural choice when the listener’s viewing direction can change, as in 360-degree video, virtual reality and games, because rotation is a simple matrix operation applied to all channels at once.
- First-order recordings sound spacious but vague; individual voices and instruments are often added as separate encoded sources for sharper placement.
- Loudspeaker decoding depends on layout regularity; irregular room layouts need specialized decoders to avoid image shifts.
- Listeners outside the central sweet spot perceive collapsed or distorted images, especially at low orders.
- Rendering to headphones uses generic HRTFs, so front-back confusion and inside-the-head localization still occur for some listeners.
Distribution platforms for 360-degree video commonly accept first-order AmbiX, often paired with a head-locked stereo track for music and narration that should not rotate with the picture. Higher orders deliver better precision, but their channel count makes storage, streaming bandwidth and real-time decoding noticeably more demanding, so many projects mix at third order and deliver at first order.
The short version
Ambisonics records sound from every direction at once, including above and below. It does not save sound for particular speakers. Instead, it saves a description of the whole sound bubble around the listener.
Later, that bubble can be played on almost any speaker setup or on headphones. It can also be turned as you move your head. That is why virtual reality and 360-degree videos use it: the sound stays in place while you look around.
Where it makes a difference
A documentary team films a 360-degree video in a busy market. Alongside the camera, the sound recordist places a first-order tetrahedral microphone at head height, so that the ambience of stalls, footsteps and conversations is captured in all directions.
In post-production, the mixer converts the recording to AmbiX B-format, aligns it with the picture orientation and adds the presenter’s lavalier microphone as a separately encoded source at the presenter’s on-screen position. Viewers wearing a headset hear the vendors stay fixed in their real locations as they turn their heads, while the narration track remains centered.