Spatial Sound


Mathis Bonotto

The technology of 4DSOUND has been based on an approach to installing equally distributed grids of loudspeakers vertically and horizontally through space, which equals drawing cubes in space. With the use of omnidirectional speakers “OmniDrive”, this configuration allows the reproduction and synthesis of sound sources of any size and position in space. The 4DSOUND system is described as a “non-conventional sound system that removes the localizability of the speaker from the equation regardless of the listener's individual hearing properties'', and these speakers are described as “phantom loudspeaker”, which means non-localizable. Nevertheless, even if the spatialisation with 4DSOUND is efficient and completely unique, the listener can still perceive the sound coming from the speaker when one is physically close to it. This phenomenon of localizability of the speaker creates a distortion of the projected sound image, which damages the perception of a coherent sound source spatialisation.

Figure 1 - An octahedron     
Figure 2 -Actual SSI position configuration

Figure 3 - 3-2-3 Octahedral position configuration

Figure 4 - Visualization of the two
configurations shape compared in the experiment

During the internship at the Spatial Sound Institute for his acoustical engineering degree, Mathis Bonotto challenged the 4DSOUND system in order to investigate possible improvements to the listening conditions within the area of the Spatial Sound Institute’s studio space. To this extent he proposed a new position configuration of loudspeakers in the studio based on an octahedron shape, instead of the original cube, in order to reduce the phenomenon that he described as “polarization of listening to the closest speaker”, which can also be described as the “non-transparence” or “reduced coherence” of the projected sound image, or the “localizability of the speaker in space”. The idea behind the use of an octahedron shape for speaker positions was to distribute more homogeneously the sound energy in space to reduce this phenomenon.

The project postulates for 4DSOUND a first step into the exploration of the use of a new geometry paradigm for speaker positions to spatially project sound sources, and was carried out in three different phases: research, modelling and experiment.

After studying the different variables that can influence the perception of the localizability phenomenon of the speaker in space and modelling a final theoretical speaker position configuration based on octahedron shape named “3-2-3 octahedral position configuration”, a test of the performance of the octahedron shape was conducted through a perceptual experiment executed with the participation of different listeners. (see figure 3)

This experiment was designed to investigate the following question: Does the effect of polarization of the speaker decrease in an octahedron setup compared to a cube setup?

During this experiment, the octahedron shape was compared to the cube shape (as normally used in the 4DSOUND system at the Spatial Sound Institute) by blending the two configurations in the same position in the studio. This setup was chosen in order to evaluate the performance of the proposed octahedral configuration. Attendees to the experiment were asked to localise the position and height of a sound source into a specific area by moving freely through the studio. The data of the positions perceived by the listener compared to the projected position of the sound source reveals the configuration shapes projection accuracy and performance of one or another configuration shape (octahedron vs. cube). (see figure 4)

Different tools have been designed to develop this experiment.

With the help of Poul Holleman, Lead Audio Engineer at 4DSOUND, a specific tool "research panner” was designed in Max/MSP to project sound sources with either the cube or the octahedron shape in an A-B manner. This software was used as an interface for the experiment and to switch fluently between configuration shapes used to project sound sources during the experiment.

Furthermore, the experiment was designed in order to collect some information on the influence of the nature of sound on the perception of the phenomenon of polarization of listening to the closest speaker. To this end, a sound classification was realized based on two different theories of sound analysis:
  • the typo-morphology approach from Pierre Schaeffer: the sound material itself is analysed and categorised

  • the referential approach from Raymond Murray Schafer: the cause of the sound is analysed and categorised
More than 80 samples have been chosen and then analysed following Pierre Couprie's “Liste des critères morphologiques pour la constitution d’une grille d’analyse” from his thesis “La musique électroacoustique, analyse Morphologique et représentation analytique” and have been classified in fine into the Typo-Morphologic table designed by Pierre Schaeffer in his essay “Traité des objets musicaux essai interdisciplines”. Despite the effort of the data gathered, the analysis didn’t reveal any correlation of the nature of the sound on the perception of the phenomenon of polarization on the closest speaker. (see figure 5 & 6)

Figure 6 - The final typo-morphological table of the sound bank analysed samples

The experiment revealed that the octahedron shape is less precise and accurate to project sound sources spatially than the original cube configuration used in the Spatial Sound Institute’s studio. The octahedron shape configuration didn’t improve the listening conditions in the studio. This is due to the fact that the use of one octahedron leads to an increased concentration of the energy of the sound in the center of the shape, i.e. a statistically higher chance for a higher amount of virtual positions within the shape, and therefore the perception of the location of the source is disrupted, and hence biased. (see figure 7)

The observed difference in the effect can easily be explained by the geometry of the octahedron: there is less loudspeakers used in an octahedron shape than a
cube shape and most of all there is only one loudspeaker at the top and at the bottom of the octahedron shape placed on the center - compared to four loudspeakers for the cube shape placed on top and bottom. 

The researcher also concluded that the design of the experiment itself had a huge influence on the results and its quality. On one hand the size of the position configuration of the speakers in space was bigger than the modelling (“3-2-3 octahedral position configuration”), and this had an effect on the sound source projection quality: it increased the encounter with the phenomenon of the polarization of the closest speaker, which disrupted the projected sound image. On the other hand, due to practical constraints, only one octahedron shape has been tested in this experiment and not several ones configured together as in the original modelling, and this is the cause of the effect of the source leaning toward the center of the octahedron.

This contribution opened the road for exploration of alternative loudspeaker position organisation in space with the use of different geometrical paradigms.

A possible future for this project would be the extension of the performance test to the complete “3-2-3 octahedral position configuration” modelled during this project, which could avoid the problem of the perception of the sound source leaning toward the center of the shape encountered when using only one octahedron.

              figure 7