Panning Algorithms

There have been a lot of scientific and artistic works to discover new methods for the electronic reproduction of audio signals to achieve the sensory impression that sounds are positioned in the space between speakers, rather than only coming from a single speaker. One of the underlying mechanisms that make it possible to position sounds or make it seem like sounds are moving around the listener, is conventionally referred to as panning.

Broadly speaking, panning is the result of an algorithm - sometimes called ‘the panning law’ - that is used to calculate the amplitudes of signals sent to two or more identical loudspeakers (or virtual sources) arranged in a particular spatial configuration. In stereo - which we could say is the lowest resolution of the speaker array - a simple panning algorithm can make the sound of a voice, for example, seem to be placed in the air between two speakers - as if there were an invisible speaker there. This illusion works up to a point, and breaks down if the speakers are too far apart. With only two speakers in the system, the panning law is a simple graph that has become quite standardized over the years. Things get more complicated when panning over 2D or 3D speaker setups and methods for spatializing on multichannel systems are not standardized yet. Many algorithms and standards have been advanced by the audio engineering industry, composers and academics.

Essentially, when we start working with multiple speakers in different configurations, things get more complex for the panning algorithms - goals have also expanded over time, with spatial sound design able to reach beyond simple point source panning into fully immersive audio experiences that can be hyperrealistic or totally synthesized. Many practitioners will have their preferred methods, but really it is up to the spatial sound designer to choose methods that seem to give the best experience for the listener on any given type of setup. There’s never been just one way to spatialize sound. And with a powerful system like SPAT Revolution, panning methods can even be combined in novel ways to achieve high-quality immersive audio experiences.

SPAT Revolution lets you explore some of the most advanced panning algorithms for surround, immersive 3D or ad-hoc sound systems.

In SPAT Revolution, you will be able to explore some of the best panning algorithms for multi-speaker setups. You can apply them in real time and identify their characteristic differences by ear. Trying them out in real time on a setup will help you select the panning algorithm that is best suited for your particular project and material.

Although there are technical aspects to be interested in and aware of, you are still invited to be creative and use your ears when deciding which are the right panning types for your project and intended audience.

Note

Try using more than one SPAT Revolution virtual Room to use different panning algorithms for sound material that has different sonic qualities (Ultimate license only).

Note

Some academic papers about some following panning laws could be found here at the end of this section.

Stereo-exclusive Panning Laws

Stereopan Law

This mode reproduces the basic experience of a pan pot. It comes with some options:

• The pan law can follow a sin/cos approach or a square root. This produces a subtle difference in how the sound travel between the left and right speaker.
• It is possible to change the center attenuation. By default, a sound placed at the center of the two speakers is played back with an attenuation of 3dB (considering acoustic summing). Other possible attenuation is: -4.5dB, -6dB.
• The PMAP, Perceptually Motivated Amplitude Panning, aims at improving sound localization on stereo systems and on any pair speaker system with an arbitrary base angle.

XY and AB

These two Panning Types will only become available when a Virtual Room is set to be virtualizing a stereo speaker arrangement. They are pan laws that are derived from widely used dual microphone techniques for rendering stereo imaging from an omnidirectional scene.

AB panning simulates the recording of the sound scene by a pair of spaced cardioid microphones, pointing laterally at azimuths +/- 55 deg. (elevation 0), with a distance of 17 cm between the two capsules. Also known as ORTF.

XY panning simulates the recording of the sound scene by a pair of microphones in an XY coincident configuration.

Note

The aim is to get the same stereo flavor as these dual microphone tracking techniques. Try them on close miked sources or any mono source, to get a realistic stereo image.

Vector Base Amplitude panning (VBAP)

Vector Base Amplitude Panning has become one of the more standardized methods for multichannel spatialization. It can reproduce on a 2D or 3D configuration. Its sound is characterized by clearly localizable virtual sound sources. Multiple moving or stationary sounds can be positioned in any direction over the speaker array using this method. In theory, VBAP can be used on an unlimited number of loudspeakers and can even be reliable on relatively asymmetric setups.

How does it work?

Traditional VBAP works by manipulating the gain of the signals being routed to the two (in 2D), or three (in 3D), the closest speakers to a virtual sound source. VBAP relies heavily on an accurate speaker arrangement model to do this. It triangulates gain vectors mathematically in order to render a virtual object in the physical space and achieves its characteristic ‘sharp’ focus by using only a few speakers closest to the virtual source location. Additionally, it is possible to uniformly extend the traditional VBAP pair-wise (or triplet-wise) speaker picking and activate more of the sound system, effectively diffusing the relationship between individual speakers and sounds using spread.

Note

Widen a VBAP point source by increasing the Spread source parameter.

Three important dependencies to consider when using VBAP:

1. Speakers must be placed around the listening position.
2. Speakers ideally should be equidistant from the listening position*.
3. 2D Speakers should be on the same horizontal plane as the ears.
4. VBAP works best when the listening room is not very reverberant.

Note

See the Alignment 18.4.1 section: the speaker alignment feature provides the impression that the actual speakers are equidistant even when they are not.

Vector Base Intensity (VBIP)

Vector Base Intensity Panning is a similar variation to the VBAP technique. It can also reproduce a 2D or 3D immersive sound field with sharply localized virtual sound sources.

How does it work?

VBIP was designed to improve on VBAP when calculating the high-frequency (above 700 Hz) localization criteria. The selection of which speakers to use to render a virtual sound source is similar to VBAP, only the gain calculations differ.

The same four dependencies mentioned for VBAP, also apply to VBIP. You will need to listen for quite a nuanced difference between these two panning algorithms. Try to compare how each panning type handles the higher frequency content of your material.

Dual Band Vector Based Panning (VBP Dual-Band)

Both Intensity and Amplitude Vector Based panning have an ideal frequency range of action:

• Localization of low frequencies is better with Amplitude Panning.
• Localization of high frequencies is better with Intensity Panning.

A hybrid approach of vector-based panning has been developed in this way: the Dual Band Vector Based Panning. This panning type merges the two approaches in order to combine the best of both worlds and to reach a better localization. Amplitude panning is applied below the crossover frequency, while intensity panning is applied above. The crossover frequency has been defined to 700 Hz by default.

The dependencies mentioned in the VBAP section also apply to Dual Band Vector Based Panning.

Layer-based amplitude panning (LBAP)

Layer-based amplitude panning can be explained as multiple 2D VBAP layers: The speaker setup is split into several layers, depending on the speaker elevation. The panning used between speakers on the same layer is the VBAP 2D. Between these layers, a crossfade is applied between the two nearest layers.

Note

The difference between VBAP 3D and LBAP is the number of speakers that will be active between the layers: three in VBAP versus four in LBAP.

Distance Base Angular panning (DBAP)

As we mentioned earlier, there are a few panning algorithms that are not Sweet Spot dependent. Distance-base amplitude panning is one of them. DBAP is useful in a number of practical situations such as concerts, stage productions, installations and museum sound design where the predefined geometric speaker layouts which immersive sound fields rely on, are not possible to establish.

DBAP was constructed with two assumptions: - All speakers are always active, independently of the source position. - The resulting level is independent of the source position.

Only level differences are used with this panning method.

How does it work?

DBAP localizes sounds towards arbitrarily positioned speakers in a space using a matrix-based technique. It calculates signal amplitudes according to the actual positions of the speakers in a space, while making no assumptions as to where the listeners are situated. Speaker tuning and interesting acoustic features in a room should be more utilized when working with DBAP.

Note

Speakers can be freely positioned when using DBAP - look for reflections and reverberations in a room to enhance spatial aspects.

K-Nearest Neighbor (KNN)

KNN is another panning type that does not depend on a Sweet Spot to be perceived correctly. It is a version of a ‘Nearest-neighbor’ interpolation algorithm. This family of algorithms is also used in the fields of complex systems, 3D graphics and network science to name a few. In SPAT Revolution, you can sonically explore a network of loudspeakers using this panning type and some virtual sound sources.

How does it work?

An interesting parameter of KNN is that the user gets manual control over one of the main coefficients in the underlying algorithm. The parameter is called Nearest Neighbor spreading. It sets a maximum limit to speakers’ number that the algorithm can use as neighbors. The parameter becomes available as a continuously variable percentage for each virtual source in a SPAT Revolution room.

Source Spreading

What makes this particularly interesting is that different sources can activate less or more of the sound system dynamically and in a very smooth way. For example, one virtual sound source might seem to pop in and out of individual speakers because its Nearest Neighbors Spread parameter is set at a low percentage, while another sound source could seem to diffuse over the entire sound system because its spread variable is set to 100%. For example, on a 10-speaker arrangement: 1-10% will use 1 speaker, 11% to 20% 2, and so on.

Note

Try automating the Nearest Neighbors Spread in a relationship with another source property of the same sound sources such as room presence.

Speaker-Placement Correction Amplitude (SPCAP)

SPCAP is a 3D panning algorithm that takes its inspiration from VBAP. SPCAP selects not just 2 or 3, but any number of speakers to render a virtual source and weights signal gains according to how much each selected speaker is actually contributing to the overall power output of the speaker configuration. Using this method SPCAP guarantees conservation of loudspeakers power output across any speaker arrangement. Its strengths lie in the down-mixing and up-mixing of virtual scenes from very different channel-based speaker arrangements, and in being able to render wider sound sources by smartly using more speakers.

How does it work?

The result will still be Sweet Spot dependent, although it will be a wider listening area. SPCAP inherits some dependencies of VBAP to get successful spatial imaging.

1. Speakers must be placed around the listening position.
2. 2D speakers should be on the same horizontal plane as the ears.

Note

★ SPCAP panning can do a good job of translating surround audio mixes from one speaker configuration to another.

Ambisonic Equivalent Panning (AEP)

In common with the channel-based panning types we have covered so far, Ambisonics is a technology that also distributes virtual sound sources in space. Yet it achieves this in a fundamentally different way. Ambisonics rely on a two-step process.

1. Encoding Audio sources along with their positional information are wrapped up together using signal mathematics to create encoded Ambisonic audio. Ambisonic scenes are always carried on at least three channels of audio. They are not intended to be listened to directly they are intended to be decoded.
2. Decoding Ambisonic audio signals are unwrapped and the positional information contained within them is decoded specifically for one type of speaker configuration. What we get is an immersive sound field that should accurately render the original spatial composition in 2D or 3D on the specified speaker configuration.

Keeping these two steps separate has a number of advantages. Primarily, that of being able to record the encoded Ambisonic audio signals independently of any fixed speaker arrangement. On the other hand, it is possible to “fuse” the two stages of the process together resulting in what appears to be the output of a generalized channel-based type of panning. That is the AEP panning type in a nutshell.

How does it work?

AEP has certain computational and ambisonic mixing advantages and exhibits very different behavior from the VBAP/VBIP pairwise approaches. It is up to you to decide whether to work with pure Ambisonic rooms (more about that in the later section) or to use AEP as a channel-based panning law. Both approaches are valid and could be useful. As we have mentioned a few times already, the choice of panning type depends on what sounds best in the context of your material, your compositional goals and the acoustics of the system you are working with.

Angular and PanR

These are legacy 2D pan pot laws from the original IRCAM Spat library. They only become available when using 2D channel-based streams and are primarily included for backward compatibility.

How does it work?

Angular and PanR are pairwise amplitude panning essentially the same as VBAP 2D described on the next page. There is a subtle difference, however, in the way the panning law changes when moving the source from one speaker to another.

Continuous Surround Panning (CSP)

This Panning Type is available in Virtual Room with 5.0 speakers arrangements. It optimizes the render into this arrangement, using circular harmonics. This leads to a continuous law, independent of the angle.

You can find more explanation about it in the relative paper.

Related papers

VBAP 2D/3D (Vector Base Amplitude Panning)

“[…] Using the method, vector base amplitude panning (VBAP), it is possible to create two- or three-dimensional sound fields where any number of loudspeakers can be placed arbitrarily. The method produces virtual sound sources that are as sharp as is possible with current loudspeaker configuration and amplitude panning methods.”

“[…] the approach enables the use of an unlimited number of loudspeakers in an arbitrary two- or three-dimensional placement around the listener. The loudspeakers are required to be nearly equidistant from the listener, and the listening room is assumed to be not very reverberant. Multiple moving or stationary sounds can be positioned in any direction in the sound field spanned by the loudspeaker.”

DBAP

“[…] Most common techniques for spatialization require the lis-tener to be positioned at a Sweet Spot surrounded by loudspeakers. For practical concert, stage, and installation appli-cations such layouts may not be desirable. Distance-based amplitude panning (DBAP) offers an alternative panning-based spatialization method where no assumptions are made concerning the layout of the speaker array nor the position of the listener.”

“[…] Distance-based amplitude panning (DBAP) is a matrix-based spatialization technique that takes the actual positions of the speakers in space as the point of departure while making no assumptions as to where the listeners are situated. This makes DBAP useful for a number of real-world situations such as concerts, stage productions, installations, and museum sound design where predefined geometric speaker layouts may not apply”

AEP

“[…] A further advantage of AEP is the possibility to use an arbitrary order of directivity for each individual sound source. It becomes possible to mix pre-recorded low order ambisonic B-format, medium order ambient sounds, high order precise localizable sound and sounds with changing localizability. How the individual sounds are perceived if different orders are used at the same time is an open question that can be answered only by experience.”

PMAP

“This paper proposes and evaluates a new constant-power amplitude-panning law named ‘Perceptually Motivated Amplitude Panning (PMAP)’. The method is based on novel image shift functions that were derived from previous psychoacoustic experiments. The PMAP is also optimised for a loudspeaker setup with an arbitrary base angle using a novel phantom image localisation model. Listening tests conducted using various sound sources suggest that, for the 60° base angle, the PMAP provides a significantly better panning accuracy than the tangent law. For the 90° base angle, on the other hand, both panning methods perform equally good. The PMAP is considered to be useful for intelligent sound engineering applications, where an accurate matching between the target and perceived positions is important.”