Preliminary Studies For Metamaterial-Based Audio Systems
Preliminary Studies For Metamaterial-Based Audio Systems
Abstract
Two of the key challenges in the design of professional audio systems are managing unwanted emissions and
directivity control at low frequencies. Traditional solutions often comprise massive cabinets. The result is a
static system, with limited freedom over the position of the listener for an optimal sound experience. In this
paper, we introduce the use of acoustic metamaterials as design tools and present preliminary measurements
where compact metasurfaces are used for both problems. In particular, with simulations and measurements,
we show how a metamaterial enclosure can be used to shape backward and side emissions with a weight
reduction of 60% compared to traditional systems. We also demonstrate, with measurements, how a system of
compact acoustic lenses can be used to deliver personalized audio messages. The optimization of our devices
could lead to headphone-less audio communication and to sound effects with a dynamic spatial component.
Keywords: acoustic metamaterials, metasurfaces, directivity control, loudspeaker systems, weight reduction.
1 Introduction
The development of loudspeaker systems is always a compromise between cost and performance [1].
Undesirable effects are often present at the lower end of the frequency range in which the individual transducer
operates and are potentially more audible and visible in lower cost loudspeakers, which in most cases consist
of sources coupled to cubic shaped cabinets. Equally noticeable is the presence of cabinet resonances, both
internally and on the front panel, typically at frequencies below 1 kHz [2]. These result in tonal components
and are typically damped with massive materials, used for the cabinet walls, or with absorbent materials inside
the cabinet. Wadding, however, is most effective at high frequencies and should be placed in the sound path,
whereas by necessity it is mounted on the cabinet walls, where its performance is further reduced. An additional
solution, incorporated into the cabinet where necessary, would be highly desirable.
A solution can be offered by acoustic metasurfaces [3], which are metamaterials with a thickness less than the
wavelength passing through them. In the past 10 years, metasurfaces have been successfully employed to
obtain acoustic devices with an optical equivalent, such as lenses and absorbers, but with practical dimensions.
Designed through computational modelling and micro-engineered into functional devices from classic
materials – such as wood, glass or plastic – metasurfaces have ushered a new era for acoustic design. In this
paper, we show how acoustic metasurfaces, shaped like a box (see Figure 1), can be used to increase the air
volume within a cubic loudspeaker enclosure and thus reduce the overall weight of the system while optimising
its performance.
Figure 1 – (left) The metamaterial cabinet used in this study (“Aurora” cabinet) and (right) a cabinet of
similar dimensions, but with smooth walls (“Smooth” cabinet).
Metasurfaces, however, are static: once the shape of the field is set, it cannot be changed, unless a hybrid
system is used [4]. The frontier in the research of acoustic metamaterials moved then to the design of dynamic
systems, with hybrid or mechanically implemented solutions [5]. These could offer the adaptable directivity
that is extremely desirable in audio systems, both for domestic listeners and studio monitors, but seldom
offered. In the second part of this paper, we will therefore show how a system of lenses, whose reciprocal
position is varied mechanically, can allow the directivity to be controlled even for a commercial audio
transducer.
2 Metasurface design
In 2017 Memoli et al. [6] demonstrated how metasurfaces can be assembled using only 16 prefabricated
metamaterial bricks (see Figure 2, left), each of which encodes a specific delay that is imposed on the sound
propagating through it. Similar geometries will be used in this work.
In the literature, the possibility of designing these bricks with a labyrinthine path has been exploited to
maximize the transmitted sound [14]. A further optimization, performed on the thickness of the bricks, resulted
in the realization of several metasurfaces (see bottom of Figure 2, left), each operating at different bandwidths
[5].
Figure 2 – (left) 3D printed metamaterial bricks in different assemblies. (right) Photograph of acoustic
telescope with auto-zoom lens used in this study, designed to deliver sound to a target listener.
where,
φ(x, y) is the local phase, assigned to a unit cell [°]; 𝐴 is a constant, related to the local curvature of the phase
profile [m-2]; λ0 is the design wavelength [m]; and φ0 is an arbitrary constant [°].
Figure 3 – Comparison of two different phase profiles, with A1<A2, highlighting how the unit cells encode
the phase distribution φ(x, y).
3 Numerical results
Figure 4 – (left) Geometry of the unit cell constituting the metamaterial cabinet. (centre) Complete geometry
of the Aurora cabinet. (right) Complete geometry of the Smooth cabinet. The two cabinets have similar
dimensions.
In the simulations, the performance of a metamaterial cabinet of the Aurora type was compared with a Smooth
cabinet of similar dimensions (both 21×21×21 cm, see Figure 4). The main difference between the two designs
is in their walls: while both cabinets have a wall thickness of 25 mm, the Smooth cabinet is whole, while the
Aurora cabinet is patterned so that there is only 1 mm of PLA between the inside and the outside.
The FEM analysis was run using the commercial software COMSOL Multiphysics, and in particular the
Acoustic Pressure Module (to simulate acoustic propagation) and the Solid Mechanics Module, to simulate
fluid-structure interactions. A study was conducted in the frequency domain, from 500 Hz to 5000 Hz, with an
interval of 1/12th octaves.
In order to simulate an infinite domain for the air outside the cabinets, a PML (Perfectly Matched Layer) was
assigned to an air box surrounding each cabinet, to mimic an open and non-reflecting infinite domain. In the
frequency domain analysis, the PML imposes a complex-valued coordinate transformation to the selected layer
which effectively makes it absorbing at a maintained wave impedance, and thus eliminating reflections.
To model a sound source with directivity in the range 500 Hz-5000 Hz, we employed a dipole point source
with a reference power equal to 3 W. The source was positioned at the centre of the open face of each cabinet.
Each cabinet was then closed with a 2.5 cm thick hardwood board, in order to simulate the surface supporting
the source used in the experiments (see later). The air inside the cabinet, beyond it and in the PML is
characterised by a density and a speed of sound at room temperature and the sound propagation neglects non-
linearities. Acrylic plastic was assigned to the cabinets and the rigid wall boundary condition was assigned on
the internal walls and on the face closing the Aurora cabinet. Thermo-viscous losses have not been taken into
account in this study.
The mesh size where sound propagation occurs is determined according to the FEM criterion, where at least
five nodes are used to simulate a wavelength in air. Thus, a maximum element size is assigned as a sixth of
the minimum study wavelength.
Due to the computational weight of the numerical model of the entire Aurora cabinet, it was decided to exploit
the C4 symmetry of the cabinet. Therefore, only a quarter of each cabinet was used, and symmetry boundary
conditions have been assigned to its rigid walls and to the air inside and outside it. The entire geometry of the
cabinet was later used for the numerical analysis of the eigenmodes, which will be discussed in the next section.
In order to observe the acoustic insulation provided by the metamaterial cabinet, when compared with the
Smooth cabinet, we observed the rear sound pressure level for the two configurations, at the same distance of
0.15 m. As can be seen in Figure 5, the FEM analysis predicted a good sound insulation from both the Aurora
cabinet and the Smooth one. However, the Smooth cabinet insulation performances are mainly due to the fact
that its walls are more massive, as will be more clearly explained later in this work. A more thorough
investigation on the Aurora cabinet, discussed in next section, was run to evaluate the effect of internal cabinet
resonances on its overall sound insulation.
Figure 5 – Rear total sound pressure level measured at 0.15 m; two configurations compared: (orange)
Aurora cabinet (1 mm effective walls), (grey) Smooth cabinet (25 mm walls), (black) Without cabinet.
A similar numerical analysis in the frequency domain was performed to study the sound radiation of the
metamaterial cabinet when coupled with a generic tweeter. This study was also conducted in the frequency
range from 500 Hz to 5000 Hz, with an interval of 1/12th octave, and a dipole was used as source. In this
simulation, the loudspeaker system was positioned inside an air-filled sphere, and only the latter was
surrounded by an external PML layer. Simulations were repeated using alternatively the Aurora cabinet and
the Smooth cabinet, and assessing the radiation emitted at a distance of 1 m by the two loudspeaker systems
(Figure 6).
A comparison of the directivity graphs in Figure 6 shows that although no significant change in directivity at
lower frequencies can be observed, a major change of the systems directivity in the horizontal plane occurs at
1900 Hz (black dashed line in Fig. 6). So around this frequency we will investigate in more detail what
influence the metamaterial cabinet walls have on this behaviour.
Figure 6 – Directivity graphs of the loudspeaker systems realized with (left) Aurora cabinet and (right)
Smooth cabinet. The black dashed line in both graphs highlights directivity at 1900 Hz.
Figure 7 shows that the region between 1000 Hz and 3000 Hz is modal dense, with some of the internal
eigenmodes of the standing waves inside the cabinet highlighted. A more detailed analysis of these modes can
be found in Figure 8, which shows the pressure field obtained on three evaluation planes placed at different
heights of the metamaterial cabinet around the frequency of 2000 Hz i.e. where one of the resonant behaviours
of the system response was noted in the simulations.
Looking at the graphs in Figure 8, sound pressure minima and cancellations generated by the geometry of the
metamaterial cabinet are evident. If one of these modes is excited by the source, the acoustic energy transforms
into vibrations of the structure. Thus, vibration of the cabinet due to reaction forces of the electrodynamic
transducer occurs. One solution to avoid this effect would require changing the dimensions of the cabinet in
such a way that the vertical and horizontal axial modes due to the side walls do not coincide with those due to
the top and bottom walls. Thus, reducing the total power in the antinodes of the resonant mode.
4 Experimental results
The experimental campaign was conducted at Contralto Audio's laboratories. The aim was to study the acoustic
response of two prototype devices: (1) the system obtained coupling the previously described Aurora cabinet
to a commercial tweeter and (2) the adjustable system obtained coupling the metamaterial loudspeaker to an
auto-zoom lens.
Figure 9 – Frequency response of the back of the system (-180°), measured at 1 m. Three configurations
compared.
Figure 9 shows the frequency response of the back of the systems, measured at 1 m from the source. It can be
seen that the A cabinet produces the same performance as the reinforced S cabinet, but using less mass. Around
1.9 kHz there is a strong loss of acoustic energy, probably due to the cooperation of the metamaterial cabinet’s
vibrational modes, resulting in a cancellation which is not observed in the case of the Smooth cabinet.
Figure 10 – Directivity graphs of the two loudspeaker systems. (left) cabinet S with a wooden board and
(right) cabinet A (VACS Viewer).
The effectiveness of the Aurora cabinet (“A” in Fig. 10) is confirmed by the response as a function of the angle
on the horizontal plane (Fig. 10). The graphs are normalised with respect to the maximum level value for each
frequency, as was done in the numerical analysis. It can be observed that, with the same frontal emission, in
the case of A there are less low-frequency back-emissions.
The graph in Figure 11 represents the on-axis response measured at 1 m. It can be clearly seen that the
metamaterial cabinet guarantees the same performance as a traditional massive enclosure when coupled with
the same sound source. Moreover, it can be noticed the previously discussed resonant behaviour around 2 kHz.
Figure 11 – Frequency response of the front of the system, measured at 1 m. Three configurations compared.
Rotating base
Figure 12 – Experimental set-up. Acoustic telescope with Aurora cabinet (A) + twitter.
For this study, we used two convex lenses for sound having the same focal length (f = 53 mm) which we
mounted in front of the source system + cabinet A. Measuring on axis at 1 m and 2 m distance from the second
lens (Figure 12) we focused the sound on the first microphone and then on the second. Figure 13 shows the
response of the system, measured as above. The directivity increased (comparison with Fig. 10), consistently
from 1.5 kHz onwards. Numerous peaks and valleys are observed in the response, probably due to the physical
presence of the auto-zoom objective, and at some frequencies the decay is greater than expected at doubling
distance.
Figure 13 – Directivity graphs of the acoustic telescope. Focus at (left) 1 m and (right) 2 m.
5 Applications
Centuries of development in the field of optics have provided us passive devices such as lenses or mirrors to
enrich the immersivity of the audience with light effects, but there is nothing similar for sound.
Possible applications of the systems described above include the reception (or emission) of a signal exclusively
for a single person in the crowd. Other applications concern acoustic displays (e.g. a group of people dancing
and following a specific acoustic focus in a discotheque) or the reproduction of music with a dynamic spatial
component.
Proof of concept that a metamaterial telescope can be used to follow a moving source in the field of view has
been shown by Rajguru et al. [15]. In the future, similar systems could be used to deliver sound effects to
multiple users having the same immersive experience, without a headset.
In the near future, we also expect to improve the audio reproduction performance of low-cost loudspeaker
systems or ceiling and in-wall speakers. The latter are often directly mounted into the wall because frequently
there is no space for a cabinet to be coupled to them. Reducing the size of the cabinet by exploiting and
optimising the geometry of the metasurfaces that make up the enclosure can be highly desirable.
In this work we have shown that it is possible to reduce the weight of a loudspeaker system by 60% by using
a metamaterial cabinet and that this allows us to shape backward and side emissions of a system without
deteriorating its acoustical performances. This was done both by numerical analysis and by experimentally
testing the system in its different configurations. The source used to perform the comparative analysis
presented in this study has a major influence on the results, both in simulations and experiments. In fact,
depending on the frequency reproduced, the vibrating transducer components can also play a key role in the
system emission. A more accurate validation of the conclusions will therefore require a complete numerical
vibroacoustic analysis of the source. In parallel, measurements with different transducers will give us insights
into the potential effects.
Numerical analysis was employed to investigate the resonant behaviours observed in the simulated response,
highlighting that a more accurate spectroscopy of the system is required, as these resonances may “colour” the
emission of the system.
Finally, in this study we used two metasurfaces, arranged in a telescope, to control the directivity of audio
systems. In the experimental measurements, directivity control at low frequencies was observed which was
not investigated in the numerical analysis and which probably also depends on the type of source used.
Our study is set in a context, where acoustic lenses and metamaterial systems [11,12,13] are used in
professional audio. Optimisation of the prototype devices described above could lead to a personalised form
of audio communication, but it will be very important to carry out experiments in the telephony band in order
to demonstrate that the intelligibility is guaranteed.
The key limitation of metamaterials, in fact, is still their bandwidth of operation. While in optics one octave in
frequency is sufficient to cover the visible spectrum, metasurfaces cannot currently transmit the 11 octaves
that make up our audible range. As many studies are showing, however, the bandwidth of a metasurface is as
much a design parameter as its spatial footprint: acoustic metamaterials auto-zoom lens, incorporating lenses
that can focus a basic melody, are not far away.
Acknowledgements
This work has been done in collaboration with the University of Sussex, which acknowledges funding through
UKRI grant EP/S001832/1. LC would like to thank all the other members of the Aurora project team for the
useful discussions.
References
[1] Murphy, D. J. Design Considerations for Loudspeaker Systems. 6th Australian Audio Eng. Soc.
Convention, 1996.
[2] Magalotti R. A Multiphysics Approach to the Design of Loudspeaker Drivers, Proceedings of the
COMSOL Conference (Grenoble), 2015.
[3] Assouar B.; Liang, Y B.; Wu Y.; Li Y., Cheng J.; Y. Jing Y. Acoustic Metasurfaces, Nature Reviews
Materials, 2018.
[4] Norasikin M.A.; Martinez Plasencia D.; Polychronopoulos S.; Memoli G.;, Y. Tokuda Y.; Subramanian
S. SoundBender: Dynamic Acoustic Control Behind Obstacles, Proceedings of the ACM UIST, 2018, pp
247-259.
[5] Memoli G.; Chisari L.; Eccles J.P.; Caleap M.; Drinkwater B.W.; Subramanian S. VARI-SOUND: A
Varifocal Lens for Sound, Proceedings of ACM CHI ’19, New York USA, Article 483, 2019.
[6] Memoli G.; Caleap M.; Asakawa M.; Sahoo D.R.; Drinkwater B.W.; Subramanian S. Metamaterial bricks
and quantization of meta-surfaces, Nature Communications 8, 2017.
[7] Li Y.; Liang B.; Gu Z.; Zou X.; Cheng J. Reflected wavefront manipulation based on ultrathin planar
acoustic metasurfaces, Scientific Rep. 3, 2013.
[8] Chisari L.; Bonoldi L.; Memoli G. Metamaterial Solutions for noise management in hospital settings, ASA
Journal, Vol. 148, 2020, pp. 2754.
[9] Pesce G.; Volpe G.; Maragó O.M.; Jones P.H.; Gigan S.; Sasso A. Step-by-step guide to the realization of
advanced optical tweezers, J. Opt. Soc. Am. B 32, 5, 2015, B84–B98
[10] Memoli G.; Graham T. J.; Kybett J. T.; Pouryazdan A. From light to sound: prisms and auto-zoom
lenses, ACM SIGGRAPH Talks, 2019, pp. 1,2.
[11] Berstis V. 3D Printed Acoustic Lens for Dispersing Sound, AES Journal, vol. 66, no. 12, 2018.
[12] Chisari L.; Di Cola M.; Martignon P. Acoustic Metamaterials in Loudspeaker Systems Design, 147th
Audio Eng. Soc. Convention, 2019.
[13] Degraeve S.; Oclee-Brown J. Metamaterial Absorber for Loudspeaker Enclosures,148th Audio Eng.
Soc. Convention, 2020.
[14] El Ouahabi A. A.; Chisari L.; Memoli G. Analytical and experimental investigation of the
transmission/reflection coefficient from labyrinthine metamaterials, Proc. Mtgs. Acoust. 42, 065003,
2020.
[15] Rajguru C.; Blaszczak D.; PourYazdan A.; Graham T. J.; Memoli G. AUDIOZOOM: Location Based
Sound Delivery system, SIGGRAPH Asia Posters, Article no. 35, 2019.