9. Spatial Audio & Localization · Concept 12 of 12
VBAP
A simple panning method that places a sound between two or three nearby speakers by balancing their loudness.
VBAP steers a phantom image around a ring by trading loudness between the two nearest speakers — constant-power, level-only, no delay.
What it is
VBAP places a sound between the nearest 2 (line) or 3 (dome) speakers by balancing only their loudness levels.
Key facts
VBAP = Vector Base Amplitude Panning; invented by Ville Pulkki, Helsinki, 1997
2D (rings) = pairwise panning between 2 speakers; 3D (domes) = triplet-wise between 3 speakers
Gains solved by g = p * L^-1, where g = speaker gains, p = direction vector to the source, L = matrix of the active speaker direction vectors
Gains normalised to constant power: g1^2 + g2^2 + ... = 1, so loudness stays even
Constant-power pan law: dead centre each speaker sits at -3 dB (0.707x), not 0 dB
Speed of sound = 343 m/s in air at 20 C; rises ~0.6 m/s per +1 C
+6 dB = double amplitude; +10 dB = roughly double perceived loudness; +3 dB = double power; -6 dB per doubling of distance (inverse-square law)
Amplitude-only: VBAP never uses delay, so it stays phase-coherent and is very cheep on CPU
Active speaker count is capped: max 2 (2D) or 3 (3D) fire at once, no matter how big the rig
Tight sweet spot: off-centre the phantom image collapses onto the nearest box; widen it with the 'spread' parameter at the cost of focus
How it works
Map every speaker as a unit direction vector from the listening centre.
Take the target direction you want the sound to come from.
Pick the 2 (ring) or 3 (dome) speakers whose arc encloses that target.
Solve for the loudness (gain) of each so their vectors sum to the target.
Normalise the gains to constant power (squares add to 1).
Move the target and the active speakers and gains update smoothly.
Real examples
Dolby Atmos-style ceiling array steering a helicopter overhead across a cinema dome.
A 360-degree festival ring panning a synth sweep smoothly around the crowd.
Theatre show following an actor's voice from stage-left to stage-right speakers.
Planetarium / dome show placing thunder at a precise compass bearing.
Museum installation parking ambient birdsong in one corner of the room.
How it helps in live sound
Measure real speaker angles from the mix position; VBAP assumes accurate vectors, guesswork smears the image.
Keep speaker spacing even (aim < ~30-45 deg gaps) so phantom images stay stable between boxes.
Time-align/EQ every box to match level first; VBAP only balances loudness, mismatched boxes break the illusion.
Warn the crowd it's a sweet-spot effect: front rows hear the pan, edge seats hear the nearest speaker.
Add a touch of 'spread' for fuller coverage when the audience is wide, accept a softer, less pin-point image.
Pick VBAP for low-latency live steering; choose Ambisonics or wavefield (WFS) when you need a wide listening area.
Everyday analogy
It's stereo's left/right balance knob, but wrapped around a full circle of speakers so any two neighbours can host the sound between them.
Watch out
Myth: more speakers = more in use at once. Reality: VBAP only ever drives 2 (2D) or 3 (3D) at a time, so the image hard-pins to one box once you leave the sweet spot.
Fun fact
Pulkki's original 1997 VBAP paper has been cited thousands of times and the same maths now steers sounds in Dolby Atmos, MPEG-H and most VR/game audio engines.
Key takeaways
Level-only panning, no delay, so it's phase-safe and ultra light on CPU.
Uses just 2 or 3 nearest speakers; gains found by inverting their direction vectors.
Constant-power law keeps loudness even (centre = -3 dB per speaker).
Works on almost any layout, from a stereo pair to a full dome.
Sharp images but a tight sweet spot, off-centre it snaps to the nearest box.