See discussions, stats, and author profiles for this publication at: [Link]

net/publication/354518996

SIMULATING LOW FREQUENCY SOUND TRANSMISSION LOSS OF MOUNTED

PANELS

Conference Paper · January 2021

DOI: 10.7712/120121.8861.18701

CITATIONS READS

0 155

3 authors:

Moris Kalderon Paradeisiotis Andreas

National Technical University of Athens University of Cyprus
24 PUBLICATIONS 87 CITATIONS 27 PUBLICATIONS 112 CITATIONS

SEE PROFILE SEE PROFILE

Ioannis Antonios Antoniadis

National Technical University of Athens
190 PUBLICATIONS 3,383 CITATIONS

SEE PROFILE

All content following this page was uploaded by Moris Kalderon on 14 September 2021.

The user has requested enhancement of the downloaded file.

 COMPDYN
2021
8th
ECCOMAS
Thematic
Conference
on
Computational
Methods
in
Structural
Dynamics
and
Earthquake
Engineering

M.
Papadrakakis,
M.
Fragiadakis
(eds.)
Streamed
from
Athens,
Greece,
230
June
2021

SIMULATING LOW FREQUENCY SOUND TRANSMISSION LOSS OF

MOUNTED PANELS

Moris Kalderon1 , Andreas Paradeisiotis1 and Ioannis Antoniadis1

1 National Technical University of Athens Dynamics and Structures Laboratory, School of Mechanical
Engineering
Heroon Polytechniou 9, 15780 Zografou Athens, Greece
e-mail: moriska@[Link] , aparadis@[Link]

Keywords: low frequency, panel, sound transmission loss, elastic mounts, reverberation room

Abstract. Lightweight panels are very common in the building, aerospace and automotive
industry, and their acoustic insulation properties are well studied; nevertheless, low frequency
noise is inherently difficult to deal with, mainly due to its high penetration. To encounter this
issue, several treatments have been proposed in the form of perforated plates, meta-surfaces
and acoustic/elastic meta-materials among others. Despite, the continuous attempts, none of
the available solutions has indicated an outstanding performance in the low frequency range.
On the other hand, boundary conditions of a vibrating panel are known to have an influence
on the sound transmission loss (STL). Driven by the recently extended standards to include
frequencies as low as 50 Hz, and the academic community debate about the reliability of low-
frequency measurements in reverberant chambers, STL improvement in the region of the first
panel resonance is investigated in this paper, considering a panel supported on elastic mounts.
Following a general vibration isolation approach, the fundamental eigenfrequency of the panel-
mount system is reduced and is explicitly selected below the threshold of human hearing. STL
predictions are accomplished utilizing a Finite Element (FE) vibroacoustic model that simulates
real test room conditions. Compared to the commonly used models of a baffled plate radiating
to infinity, this detailed approach allows not only a more accurate calculation of STL but also
reveals the potential issues emerging during laboratory measurements.

5214
 Kalderon M., Paradeisiotis A. and Antoniadis I.

1 INTRODUCTION
When considering room-to-room noise transmission inside buildings, sources such as elec-
trical appliances, music and movies in neighboring apartments etc., have notable low frequency
content. Another significant component of low frequency noise in residential areas is road
noise emitted by vehicles, since their engines operate typically at speeds up to 6000 RPM (100
Hz). Therefore, there is an ever increasing demand for noise mitigation solutions in this range,
providing an improvement on the residents quality of life but also for compliance with legal
requirements [1].
However, the low frequency range approximately below 200 Hz, has been overwhelmingly
overlooked in the past when considering applications and we presently lack understanding to
properly model the processes governing the sound transfer in this frequency range. Almost
exclusively, the focus is on the region of the coincidence frequency of finite partitions. The lack
of complete industrial standardization and guidances for such cases, especially below 100 Hz is
also a factor. Yet, a large research program on low frequency sound transmission into buildings
was conducted by NGI (the Norwegian Geotechnical Institute) between 2010-2016, comprising
of laboratory, full scale and numerical testing [2]. The disseminating research findings of the
program have been proven very useful, but there are still uncharted areas in this field.
When it comes to acoustic treatment in rooms there are also a lot of misconceptions among
the general public. Full scale tests reveal that when noise transmission occurs inside a building,
the peaks of the low frequency sound spectrum are of the same order as the fundamental acoustic
room-modes [3], leading to a collection of resonances, which compromise the measurements.
International standards describe procedures to extend the measurement range down to the 50
Hz one-third octave band [4, 5, 6]. Prescriptions include use of specific source and receiver po-
sitions, however the procedure is complex, with the worst case scenario to propose the use of a
sufficiently large number of source-receiver combinations as a means to smooth the large spatial
variations in measured levels. Moreover, regarding the laboratory geometry, large rooms with
volumes V > 200m3 and specific room size ratios are suggested when low frequency measure-
ments are conducted. Recently, Ayr et al. [7] presented a detailed low frequency qualification
procedure for a typical reverberant test room in order to perform sound power measurements,
by taking advantage of a room finite element model. Similarly, Løvholt et al. [8] developed
a finite element methodology that incorporates a two-way coupled fluid-structure interaction
and they compared their numerical results with experimental measurements including low fre-
quency sound transmission of plain walls, and walls with windows. Vorländer and Aretz also
worked on room to room acoustic simulations, focusing on the room boundary conditions [9]
and the uncertainties of the computational models [10].
As far as the analytic methods are concerned, the acoustic radiation from plates has long
been an important subject in structural dynamics and acoustics, thus numerous mathematical
models describing the structural response to noise, acoustic fatigue and sound transmission of
partitions have been developed. Beranek and Ver, Cremer et al., Ordubadi and Lyon and London
[11, 12, 13, 14] derived equations for the unbounded/infinite partitions (”infinite-panel theory”)
simplifying the problem and later ad hoc corrections were added to enhance the accuracy of the
Sound Transmission Loss (STL) formulas. Although, the infinite panel theory can approximate
pretty accurately the panel response when high frequency noise mitigation is concerned, it is
known that bounded plates vibrating at frequencies below the critical frequency are very much
influenced by the presence of their boundaries [15, 16]. Similarly, Kim [17] approached sound
transmission through partitions from the standpoint of impedance mismatch. Despite, the merits

5215
 Kalderon M., Paradeisiotis A. and Antoniadis I.

of the aforementioned approaches, in low frequency the modal response of the structure affects
the transmission properties.
The aim of this contribution is to investigate the effects of elastic mounting in the STL
performance, considering a much relevant specific case of single-leaf plasterboard partitions
and show how the uncertainties due to practical factors can be diminished with this approach.
Namely, the ways that elastic mounting may be incorporated in already existing panels and set
circumstances regarding their spatial configuration, as well as, the effectiveness of the damping
of the mounts in the mitigation of the first panel resonance. The analysis is conducted via both
simple mathematical models and a coupled finite element vibroacoustic model where the room
to room sound transmission is described based on the laboratory set-up in SINTEF Building
laboratory in Oslo, Norway [8].
The idea of utilizing ”soft mounts” to reduce the fundamental frequency of the system delves
into the investigation of what constitutes soft mounting. Therefore, a modeling procedure based
on certain approximations [18] is formulated in order to provide a projection of the expected
level of stiffness that satisfies this terminology and serves the intended purpose. Intuitively,
following the general concepts of vibration isolation for the reduction of the eigenfrequency
through the reduction of stiffness, supporting an existing panel on elastic mounts, the ’mass
law’ behavior may be dragged into lower frequencies and the resonant region of the new system
panel-mount will be moved below a critical frequency, for example at 20 Hz considering the
threshold of human hearing.
To sum up, in Section 2 the theoretical formulation of the three applied methods, are provided
together with the FE model describing the room to room transmission. In Section 3 a parametric
investigation is presented to reveal the parameters that affect STL of a rectangular panel and
finally in Section 4 the conclusions of this work are briefly discussed.

2 MATERIALS & METHODS

By defining the transmission coefficient as
Πt
τ= (1)
Πi
where Πi is the incident and Πt the transmitted (radiated) sound power, the STL is expressed as
1
ST L = 10 log10 (2)
τ
A typical form of the STL curve for sound propagation through a thin panel is decribed in
[19]. In the ’mass law’ region, the STL is controlled by the mass per unit area of the panel.
Specifically, for a thin panel, neglecting stiffness and damping (limp wall):
 Ω2 m̄2 
ST L = 10 log10 (3)
4(ρ0 c0 )2 / cos2 θi

where Ω is the cyclical frequency of the propagated incident sound waves, θi is the angle of
incidence, c0 is the speed of sound in air, ρ0 is the air density and m̄ [kg/m2 ] is the mass per
unit area of the panel. Although the mass law is very simple and compact, it is only valid for
large simple plates because it is derived from the infinite and rigid plate assumption. Therefore,
attention should be given when utilized for STL predictions of general finite panels.

5216
 Kalderon M., Paradeisiotis A. and Antoniadis I.


 
 



  

 






 













  






 

 










 


Figure 1: (α) Coordinate systems and transmission geometry of a rectangular plate in a rigid baffle. (b) Transmis-
sion geometry of the free finite rigid panel.

2.1 Sound radiation from a rectangular panel in an infinite rigid baffle

The fluid domain is assumed homogeneous and compressible. The mass density of air and
sound speed are noted as ρ0 and c0 , respectively. The infinitely large acoustic rigid baffle divides
the space into two domains i.e. the excitation domain V − and the receiver domain V + . The
panel is subjected to the incident pressure pi , with incidence angle θi and azimuth angle φi .
Therefore, the excitation domain consists in the incident pressure pi , the reflected pressure pr ,
and the re-radiated pressure p− rad due to the motion of the finite panel. The receiver domain
consists only in the transmitted pressure pt , namely the radiated pressure due to the motion of
the panel p+rad . Assuming that the incident sound pressure pi is a harmonic function, we can
express it as
pi (x, y, z, t) = Pi e−j(κx x+κy y+κz z) ejΩt = p̃i (x, y, z) ejΩt (4)
where p̃i (x, y, z) the time invariable complex amplitude and κ = Ω/c0 is the wavenumber with

κx = κ sin θ cos φ , κy = κ sin θ sin φ , κz = κ cos θ (5)

The reflected pressure by the surface of the panel is then expressed as

pr (x, y, z, t) = Pr e−j(κx x+κy y−κz z) ejΩt = p̃r (x, y, z) ejΩt (6)

denoting the opposite direction from the incident wave. Lord Rayleigh [20] determined a
relationship expressing the radiated (transmitted) pressure in terms of the structural velocity
based on Green’s function,
 
+ jΩρ0 jΩt lx /2 ly /2 e−jκR ∂w(ξ , η  )  
prad (r, t) = pt (r, t) = e dη dξ (7)
2π −lx /2 −ly /2 R ∂t

where R referring to Fig.1, is defined as

%
R = |r − rP | = (x − ξ )2 + (y − η  )2 + z 2 (8)

and r = (x, y, z) is the position vector of a point in the receiver domain while rP = (ξ , η  , 0) is
the position vector of the center of an elemental radiator with surface δS on the panel, having a
normal velocity amplitude ẇ(ξ  , η  ). For the local coordinates holds that ξ  = ξ − l2x , η  = η − l2y .
Following the calculation of the expression for the transmitted pressure, there are two dif-
ferent ways used to determine the radiated (transmitted) power. The first is to integrate the

5217
 Kalderon M., Paradeisiotis A. and Antoniadis I.

intensity on a hemisphere in the far-field enclosing the plate which is followed in this paper,
and the latter is to integrate the acoustic intensity over the surface of the vibrating plate. Both
approaches require the knowledge of the distribution of the velocity over the plate and they
assume a weak coupling between the vibrating structure and the radiated sound field.
Applying the far-field approximation as described in Section A.1, the transmitted intensity
It is calculated as
|pt (r, θ, φ)|2
It (r, θ, φ) = (9)
2ρ0 c0
Then, the transmitted power Πt comes by integrating the transmitted intensity on a hemi-
sphere in the far field enclosing the plate as
 2π  π/2
Πt = It r 2 sin θdθdφ (10)
0 0

while the incident power on the panel is defined as

|pi |2 lx ly cos θ
Πi = (11)
2ρ0 c0
2.2 Vibration of a panel subject to external pressure
With reference to Fig.1α, the finite-sized flat panel partition is assumed to be rectangular
and baffled, with lengths lx and ly along the x and y axes respectively. The panel of thickness
h, is considered homogeneous and isotropic and is modeled as a classical thin plate, implying
that the effects of both the rotary inertia and the transverse shear deformation can be neglected.
Hereinafter, the equation of motion governing the bending vibration of the plate is given by
 ∂ 4 w(ξ, η, t) ∂ 4 w(ξ, η, t) ∂ 4 w(ξ, η, t)  ∂ 2 w(ξ, η, t)
D +2 + + ρh =
∂ξ 4 ∂ξ 2 ∂η 2 ∂η 4 ∂t2 (12)
= pi (ξ, η, t) + pr (ξ, η, t) + p-rad − p+rad

where D = Ẽh3 /12(1 − ν 2 ) is the bending stiffness of the plate, E, h, ρ and ν are the Young’s
modulus, thickness, mass density and Poisson’s ratio of the plate, respectively. w(ξ, η, t) is
the instantaneous transverse displacement. In order to account for energy dissipation due to
structural damping, a complex modulus of elasticity is introduced Ẽ = E(1 + jn), where n is
the loss factor.
The right hand side must satisfy the velocity continuity at the surface of the panel (z = 0).
In that case, it holds that pr = pi and also since pt = p+rad = −p-rad the right hand side becomes

pi (ξ, η, t) + pr (ξ, η, t) + p-rad − p+rad = 2pi (ξ, η, t) − 2pt (ξ, η, t) (13)

The so-called ”blocked pressure” [17] pb = 2 pi , is the pressure when the incident wave
meets a rigid wall.

2.3 Free Finite Rigid (FFR) Panel Approximation

In order to replace the infinite panel approximation (limp wall) with a more appropriate
reference for finite panels, the free finite rigid (FFR) panel approximation is formulated. In this
case it is assumed that the panel is rigid, consisting only in mass and oscillating freely in an
infinite rigid baffle but instead has finite dimensions as demonstrated in Fig.1b.

5218
 Kalderon M., Paradeisiotis A. and Antoniadis I.

Since the panel is considered rigid, the spatial derivatives of Eq.(12) are eliminated. Also,
considering only the blocked pressure pb = 2pi as the forcing pressure and that pi are incident
plane waves, the equation of motion becomes:

ρhẅ = 2Pi ejΩt (14)

The forcing pressure is a harmonic function, therefore the steady-state transverse displace-
ment of the panel can be expressed as

w = W (ξ, η)q(t) (15)

where
q(t) = CejΩt (16)
For the rigid assumption, the shape function is independent of the position (ξ, η) on the surface
of the panel, namely W (ξ, η) = 1. Substituting the solution into Eq.(14) and solving for C
gives:
2Pi
CFFR = − (17)
ρhΩ2
The transmission coefficient then comes as
1  ρ0 2 1
τFFR = Iθφ (18)
l x ly π m2

where m = ρh is the mass density (mass per surface area) of the panel, and Iθφ is the double
integral over a hemisphere in the receiver domain as given in Eq.(A.10).

2.4 Simply Supported Plate First Mode Approximation

Analytic solutions for the problem can be found via modal superposition methods, for ex-
ample as formulated by Roussos [21]. However, in these analytic approaches, past researchers
usually avoided to include in their calculations the re-radiated pressure 2pt (ξ, η, t) (often re-
ferred to as ”fluid loading”) of Eq.(13) in order to reduce complexity. According to Roussos
[21] neglecting the effect of the re-radiated pressure allows an accurate solution to be obtained
over large frequency range, and gives invalid answers only for frequencies near the panel’s
fundamental resonant frequency.
Since the low frequency range of interest in the present case is exactly the range near the
fundamental resonance of the panel, an attempt for a more accurate and usable approximation
is formulated in this section, including the effect of fluid loading.
Considering only the 1st vibrational mode of the simply supported plate, the spatially de-
pendent amplitude of the transverse displacement w(ξ, η) = W1 (ξ, η, t)e−jΩt is assumed by the
shape function    
πξ πη
W1 = C1 sin sin (19)
lx ly
the resulting expression for the transmitted (radiated) power Πt is given by Eq.(A.25). The
detailed analysis is presented in Section A.3.
Assuming normal incident waves (θi = 0), the incident power is given by

Pi2 lx ly
Πi = (20)
2ρ0 c0

5219
 Kalderon M., Paradeisiotis A. and Antoniadis I.

leading to the transmission coefficient τ as

1 (16Ω2 ρ0 )2
τ1 = Iθφ (21)
lx ly |π 3 ρh(ω12 − Ω2 ) − 16Ω2 ρ0 e−jkr
r
|2

where distance r needs to be large enough to satisfy the far-field approximation and ω1 is the
natural frequency of the 1st mode resulting from the homogeneous form of Eq. (12), as
 2
Dπ 4 1 1
ω12 = 2
+ 2 (22)
ρh lx ly

Expressing the dynamic behavior of the continuous system of the simply supported panel
approximated by the 1st mode, as a single degree of freedom (SDoF) lumped parameter model
(LPM), the corresponding generalized mass and stiffness [22] come as

ρhlx ly mlx ly m
m∗1 = = = (23)
4 4 4
Dπ 4
l l 1 1 2
k1∗ =
x y
+ (24)
4 lx2 ly2

The transmission coefficient is then calculated as

 2
4Ω2 ρ0 1
τLPM = lx ly 4 52 Iθφ (25)
π3 m∗1 |ω12 − Ω2 |

2.5 Lamped parameter model for elastic mounts

The goal is to formulate a straightforward procedure for the investigation of how the bend-
ing stiffness of the panel combines with the mounting stiffness in the case where the panel is
supported on elastic mounts and how the acoustic performance is affected. In order to do that,
the main assumption is that the bending stiffness of the deformable plate is in a way in series
with the stiffness of the mounting. Therefore, the resulting stiffness is

k1∗ k0
ktot = (26)
k1∗ + k0

When the mounting is very stiff (k0 >> k1∗ ) leads to ktot = k1∗ , namely, the case of the simply
supported plate is approached. When k0 → 0, the plate is essentially free floating. Natu-
rally, when the eigenfrequency of the system is reduced along with the total stiffness, the mass
considered in the SDoF model increases. In this case this variation is neglected for reasons of
simplification. As a result, when the considered mounting stiffness k0 is of comparative order of
magnitude with k1∗ , there is some deviation between the resulting eigenfrequency of the model
and the actual eigenfrequency. However, in cases when k0 >> k1∗ or k0 << k1 ∗ the deviation is
nullified. Thus the model is used for an estimation of the required stiffness level and guidance
for the design of the mounts.
Fig.2, shows the equivalent dynamic SDOF model of the deformable thin panel when it
is supported on elastic mounts according to the ’in-series’ assumption. This is achieved by
utilizing the generalized values of the structure, namely generalized mass and the corresponding

5220
 Kalderon M., Paradeisiotis A. and Antoniadis I.

 
 

 



Figure 2: Modelling of a deformable plate on elastic mounts with stiffness k0 .

generalized stiffness for the appropriate modes, as calculated from Eqs.(23),(24) for the 1st
mode approximation of the simply supported panel.
Since the loss factor (n) can represent more accurately the dynamic response of nonlinear
systems compared with the damping ratio which is defined on the grounds of the linear single
degree of freedom (SDOF) viscous model [23, 24], hysteretic damping is introduced indirectly
considering complex stiffness elements as
k0 = k0 (1 + jn) (27)
(28)
1
The STL of the model is then calculated as ST L = 10log10 τ
, where the transmission coeffi-
cient τ comes as
 2
4Ω2 ρ0 1
τ = lx ly Iθφ (29)
π3 |− Ω2 m∗1 + ktot )|2
.

2.6 3D vibro-acoustic FE model

In addition to the aforementioned analytical methodologies, a 3-D FE coupled vibro-acoustic
model was developed on the commercial finite element software package ABAQUS [25]. In the
present simulations, the geometry of the SINTEF laboratory [8] is utilized as the computational
domain, as shown in Fig.3 α and b. It is noted that the laboratory geometry and the measurement
procedure do not follow accurately the proposed from the standards guidance, however, the
developed model can be used for room to room sound transmission predictions. The panel is
excited by the source located in the left corner of the source room and the incidence and radiated
acoustic pressure is extracted from the fluid-structure interface nodes. An arithmetic average is
performed by
1
n
Pt = |pt,i | (30)
n i=1

1
n
Pi = |pi,i | (31)
n i=1
where n denotes the total number of measuring nodes; i denotes the node number; pi denotes
the incidence pressure and pt denotes the transmitted pressure. Then the averaged pressures
Pi , Pt are used to compute the simulated STL by
|Pi |
ST L = 20 log10 (32)
Pt

5221
 Kalderon M., Paradeisiotis A. and Antoniadis I.

Source room, h=3.0m Wall Receiving room, h=3.9m

Incidence
Transmitted
4.2m pressure
pressure
measurement
measurement
Loudspeaker

6.3m
1.0m

6.2m
1.0m

8.0m
(b)
(
)

(c)

Figure 3: (α) Plan view of the sound laboratory setup used in ABAQUS model. (b) Depiction of the vibro-acoustic
model. (c) Positioning of the twelve (12) mounts on the surface of the panel.

The panel is discretized by 20-node quadratic solid hexahedral elements, while the fluid do-
main is discretized by 20-node quadratic acoustic hexahedral elements. Tie constraints are used
to simulate the coupling at the fluid-structure interface, while reflecting boundary conditions
were specified to simulate the room conditions. In cases that the plate is supported on elas-
tic mounts, connectors are installed on the required locations with the appropriate properties.
Lastly, the discretization of the fluid domain has more than three quadratic elements across the
wavelength of interest to increase the accuracy of the computational results [26].

3 Numerical Example - Parametric investigation of STL

In this section, the elements that affect STL of a rectangular plate are investigated, especially
in case of a single leaf plasterboard partition. Initially, the analytic models proposed in the
previous section A 0.625 x 1.6m rectangular plasterboard is considered, according to real-life
masonry applications. The width corresponds to the horizontal distance between the upright
supporting beams providing the stable frame of the plasterboard. Typically, the height of such a
plasterboard is around 3.2 m, meaning that it can be covered by two pieces with the dimensions
consider here. The selected dimensions correspond also to a frontal surface of 1 m2 . The
relevant properties of the plasterboard are summarized in Table 1.

Table 1: Plasterboard properties.

kg
ρ [m3] Lx [m] Ly [m] h [m] E [M P a] ν [−]
668 0.625 1.6 12.5 10−3 2900 0.31

The red dashed line in Fig.4, represents the FFR panel approximation as given by Eq.(18).
Intuitively, this approximation should constitute the ideal case regarding the STL of a panel

5222
 Kalderon M., Paradeisiotis A. and Antoniadis I.

with finite dimensions and may act as a reference curve for the various comparisons instead
of the infinite panel approximation which is invalid in the lower frequency range that is being
examined. The yellow line corresponds to the 1st mode approximation of the simply supported
panel while the blue dotted line is the lumped parameter model as derived from the 1st mode
approximation. The difference between the two methods is essentially the effect of fluid loading
(FL), whereas in the latter model is not included. It is observed that the fluid loading has very
little effect on the damping of the first resonance. Thus, the elimination of its participation in
the respective models is justified and consists a safe approximation. Lastly, the black dashed
line represents the modal superposition method as formulated by Roussos and the grey line
with markers shows the STL as calculated from the finite element model. In general a good
agreement is observed among all the applied methods.

3.1 Mathematical vs Numerical models

Figure 4: Comparison of 1st mode, rigid approximations and numerical model.

The objective of the simplified models is to provide a more straightforward and fast pre-
diction of the acoustic performance of the panel along with the ability to extract intuitive and
revealing expressions of the main dynamic parameters that govern its frequency response. Since
the higher modes have significantly lower modal participation factors in the frequency response
than the fundamental mode, they are relatively easily damped. With that thinking, the approx-
imation using only the 1st mode is considered appropriate in order to study this low frequency
region, and following, the effect of elastic mounting and damping.

3.2 Effect of support conditions in STL

A case for the resulting STL frequency response compared to the simply-supported and
clamped configurations are presented in Fig.5α. The total stiffness of the mounting is consid-
ered equal to the generalized stiffness of the simply supported panel (k = k1∗ = 1.1 105 N/m),

5223
 Kalderon M., Paradeisiotis A. and Antoniadis I.



Figure 5: (α) Effect of support conditions in Sound transmission loss (ST L). (b) Effect of damping in Sound
transmission loss (ST L).

while, no damping is considered for the mounts. This choice for the mounting stiffness is based
on the Lamped parameter model, described in Section 2.5. Specifically, the total mounting
stiffness is selected to drop the 1st resonance below 20 [Hz] and then the estimated stiffness is
allocated to the 12 springs of the FE model.
Following the logic of classic vibration isolation, by lowering this frequency enough it can
be moved outside the frequency range of interest, e.g. above 20 Hz. However, significantly de-
creasing this frequency can have certain implications. On one hand, it leads to slightly reduced
STL frequency response and on other hand, it may present practical problems including exces-
sive vibration magnitude, possible resonances with other supportive structural elements such as
bolts or the frame, or even inadequate rigidity for impact loads. The optimal solution requires
a combination that leads to a low enough fundamental eigenfrequency that can be effectively
dampend without the need of excessive damping while at the same time is high enough that the
overall STL is not significantly reduced and no vibrational implications come into play when
the mounting is very soft.
Another purpose of the comparison in Fig.5α is to highlight the deviation of the eigenfre-
quencies between the simply supported and clamped panels in regards to the uncertainty intro-
duced by this fact in real world installations. Depending on the technique of the installation
the exact support conditions may vary between these two cases. However, the inclusion of the
soft elastic mounts, eliminates this uncertainty as it produces an explicit, calculated frequency
response by design. Additionally, the reduction of the eigenfrequencies alone almost eliminates
the participation of the higher modes even without the consideration of any energy diffusive ele-
ments besides the structural damping of the panel itself which holds for all three case compared
here.

3.3 Effect of damping

Further examination considers the effect of damping of the elastic mounts. In order to inves-
tigate the effects of damping 12 linear springs were attached to the panel as in Fig.3c. Assuming
the same stiffness for the mounts as in the previous case, a comparison is presented in Fig. 5b
damping ratios ζ = 0.3 and ζ = 0.7.
The results presented in Fig.5b are meant to demonstrate the expected level of damping in

5224
 Kalderon M., Paradeisiotis A. and Antoniadis I.

order to effectively damp the fundamental frequency. As already mentioned, even in cases of
low damping the STL response is flattened as the higher eigenfrequencies are very effectively
damped while only the first eigenfrequency requires significant damping. How realistic are the
actual values of the assumed viscous dampers for high damping ratios is surely a matter of
discussion and very much depends on practical factors like the material used for the mounts and
spatial considerations of the real world implementation.
Certainly, the positions of the mounts have a significant role on their damping effectiveness.
Especially in this case of viscous damping, the velocity amplitude is maximum at the center of
the panel and minimum along the edges, therefore a mount positioned at the center would most
effectively dampen the fundamental resonance. However, this case would change the dynamic
behaviour of the panel and would also require an additional supporting frame along the middle
of the panel for example, which is something that is preferably avoided. For this reasons such a
modification is not investigated in the current contribution.

4 Conclusions
The variation of the support conditions of the panel, between simply supported and clamped,
in a real-world implementation affects significantly the eigenfrequencies and therefore the fre-
quency response of the STL.
The term ”soft mounts” is approximately defined as the case in which the total stiffness of
the mounts is lower than the generalized stiffness of the simply supported panel. When the
panel is supported regionally on soft elastic mounts, the values of the eigenfrequencies of the
structure are reduced. This has two main consequences:

• The value of the fundamental eigenfrequency can be explicitly defined, depending on

the stiffnes of the mounts, and be moved outside the frequency range of interest, namely
below 20 Hz.

• The contribution of the higher modes is almost nullified without the need of any signifi-
cant damping in the mounts.

Therefore, the utilization of soft mounts displaces the STL curve towards lower frequencies,
providing smooth performance in the entire low frequency range. Additionally, this practice
diminishes any uncertainties introduced in the acoustic performance either due to practical sup-
port conditions or variations in the rigidity of the panel.
Finally, the analytic models showed good agreement with the developed FE model, indicat-
ing that they can be used as a quick tool for the design of soft mounts.

5 Acknowledgments
Moris Kalderon has been financed by the European Union’s Horizon 2020 research and in-
novation programme under the Marie Sklodowska-Curie grant (grant agreement No INSPIRE-
813424, ”INSPIRE - Innovative Ground Interface Concepts for Structure Protection”).

Appendix
A.1 Far-field approximation
The distance R can be written as
%
R= r 2 − 2xξ  − 2yη  + ξ 2 + η 2 (A.1)

5225
 Kalderon M., Paradeisiotis A. and Antoniadis I.

%
where r = |r| = x2 + y 2 + z 2 . In spherical coordinates R becomes

sin θ  ξ 2 + η 2
R = r 1−2 (ξ cos φ + η  sin φ) + (A.2)
r r
The solution of Eq.(7) is not trivial and has been calculated analytically only for a few cases
of boundary conditions. A way around this obstacle comes by utilizing the far field approxima-
tion. In the far field we can ignore the weak dependence of the amplitude of the integrand on
the position but not the dependence of the phase, namely it holds that
1 1
≈ (A.3)
 R r
−jκr 1− sinr θ (ξ  cos φ+η sin φ)  cos φ+η sin φ)
e−jκR ≈ e = e−jκr ejκ sin θ(ξ (A.4)

Since r is independent of ξ  and η  , the corresponding terms in Eq.(7) can be moved outside the
double integral, leading to
 
jΩρ0 j(Ωt−κr) lx /2 ly /2
∂w(ξ  , η ) jκ(ξ sin θ cos φ+η sin θ sin φ)  
pt (r, t) = e e dη dξ (A.5)
2πr −lx /2 −ly /2 ∂t

which may be analytically calculated.

A.2 Free Finite Rigid (FFR) Panel Approximation

Then, the transmitted pressure by considering the far-field approximation, is calculated from
the expression
 lx  ly
−Ω2 ρ0 W e−jkr jΩt 2 2  sin θ cos φ+η  sin θ sin φ)
pt (r, t) = e ejκ(ξ dη  dξ  (A.6)
2π r − l2x
l
− 2y

where the surface integral is calculated

8 sin(σx ) sin(σy )
Iξ η = (A.7)
k 2 sin2 θ sin 2φ
and
lx
σx (θ, φ) = k sin θ cos φ (A.8)
2
ly
σy (θ, φ) = k sin θ sin φ (A.9)
2
The transmitted intensity It and consequently the transmitted power, are calculated from Eqs.(9),(10)
where Iθφ is defined as
 2π  π/2 6 62
6 6
Iθφ = 6Iξ η 6 sin θ dθ dφ (A.10)
0 0

5226
 Kalderon M., Paradeisiotis A. and Antoniadis I.

A.3 Simply Supported Plate - First Mode Approximation

Considering only the 1st mode for interpolation of the transverse displacement the derivatives
to be substituted in the equation of motion are calculated as

∂ 2 w(ξ, η)
2
= −Ω2 W1 (ξ, η)e−jΩt
∂t
 4  4
∂ 4 W1 (ξ, η) π πξ πη π
= C 1 sin( ) sin( ) = W1 (ξ, η)
∂ξ 4 lx lx ly lx
 4  4
∂ 4 W1 (ξ, η) π πξ πη π (A.11)
4
= C1 sin( ) sin( ) = W1 (ξ, η)
∂η ly lx ly ly
 2 2  2 2
∂ 4 W1 (ξ, η) π πξ πη π
2 2
= C1 sin( ) sin( ) = W1 (ξ, η)
∂ξ ∂η lx l y lx ly l x ly

Considering the far field approximation for the transmitted pressure and normal incident
waves (θi = 0), the equation of motion becomes
   
4 1 1 1 2
Dπ + 2 2 2 + 2 − Ω ρh W1 (ξ, η) =
lx2 lx l y ly
(A.12)
Ω2 ρ0 −jkr
2Pi + e Iξ  η 
πr
where Iξ η is the double integral on the surface of the plate
 lx /2  ly /2
 sin θ cos φ+η 
Iξ  η  = W1 (ξ  , η  )ejk(ξ sin θ sin φ)
dη  dξ  (A.13)
−lx /2 −ly /2

and the local coordinates ξ  = ξ − l2x , η  = η − l2y . For ξ  , η  the shape function from Eq.(19)
becomes

πξ  πη 
W1 (ξ  , η ) = C1 cos( ) cos( ) (A.14)
lx ly
The surface integral is then calculated as

Iξ η = C1 Iξη (A.15)

where

cos(σx ) cos(σy )
Iξη = 4π 2 lx ly (A.16)
(lx k sin θ cos φ) − π (ly k sin θ sin φ)2 − π 2
2 2

and σx , σy are the same as in Eq.(A.8). Solving (19) for C1 , the surface integral can be written
as
W1 (ξ, η)
Iξ  η = Iξη (A.17)
sin( πξ
lx
) sin( πη
ly
)

5227
 Kalderon M., Paradeisiotis A. and Antoniadis I.

Substituting to equation (A.12), multiplying with W1 and integrating over the panel surface
gives
 lx  ly   2 
4 1 1 Ω2 ρ0 e−jκr Iξη (θ, φ) 2
Dπ + − − Ω ρh ...
0 0 lx2 ly2 πr sin( πξ
lx
) sin( πη
ly
)
 lx  ly (A.18)
... W12 (ξ, η) dη dξ = 2Pi W1 (ξ, η) dη dξ
0 0

and inserting ω1 from Eq.(22) leads to

 lx  ly
2 2
ρh(ω1 − Ω ) W12 (ξ, η) dη dξ =
0 0
    (A.19)
Ω2 ρ0 e−jκr lx ly
= 2Pi + C1 Iξη (θ, φ) W1 (ξ, η) dη dξ
π r 0 0

The two integrals are calculated as

 lx  ly
l x ly 2
W12 (ξ, η) dη dξ =
C (A.20)
0 0 4 1
 lx  ly
4lx ly
W1 (ξ, η) dη dξ = 2 C1 (A.21)
0 0 π
Substituting in Eq.(A.19) and solving the quadratic equation for C1 leads to:
8Pi 32Pi + pt
C1 = = (A.22)
π 2 [ρ0 h
(ω12 −Ω2 )
− 4Ω2 ρ0 e−jkr
Iξη (θ, φ)] π ρh(ω12 − Ω2 )
2
4 π3 r

Therefore, substituting in Eq.(A.5) for the transmitted pressure, comes:

Ω2 ρ0 e−jkr
pt (r, t) = C1 Iξη (θ, φ)ejΩt
2πr
16Ω2 ρ0 Iξη (θ, φ)Pi (A.23)
⇒ pt (r, t) = −jkr e−jkr ejΩt
r[π 3 ρh(ω12 − Ω2 ) − 16Ω2 ρ0 e r
]
The transmitted intensity It is calculated from Eq.(9) as
 2
1 16ω 2|Iξη |Pi
It (r, θ, φ) = (A.24)
2ρ0 c0 r 2 |π 3 ρh(ω12 − Ω2 ) − 16Ω2 ρ0 e−jκr
r
|

and the transmitted power comes from Eq.(10) as

Pi2 (16Ω2 ρ0 )2
Πt = Iθφ (A.25)
2ρ0 c0 |π 3 ρh(ω12 − Ω2 ) − 16Ω2 ρ0 e−jκr
r
| 2

where Iθφ is the double integral over a hemispheric surface in the receiver domain
 2π  π/2
Iθφ = |Iξη |2 sin θdθdφ (A.26)
0 0

and Iξη is defined in Eq.(A.16).

5228
 Kalderon M., Paradeisiotis A. and Antoniadis I.

A.3.1 Lumped Parameter Model

Let us rewrite the transverse displacement as
   
πξ πη
w(ξ, η, t) = sin sin q(t) (A.27)
lx ly
where
q(t) = C1 ejΩt (A.28)
The kinetic energy of the system is defined as
 lx  ly  2
1 ∂w
T = ρh dξdη
2 0 0 ∂t
1 ρhlx ly 2 (A.29)
= q̇ (t)
2 4
1
= m∗1 q̇ 2 (t)
2
and the potential energy as
 lx  ly  2
1 2
U= D ∇ w(ξ, η, t) dξdη
2 0 0
 2
1 Dπ 4 lx ly 1 1
= + q 2 (t) (A.30)
2 4 lx2 ly2
1
= k1∗ q 2 (t)
2
where m∗1 and k1∗ are the generalized mass and stiffness of the fundamental, 1st mode.
The equation of motion of the equivalent SDOF model is

m∗1 q̈ + k1∗ q = f ∗ (t) (A.31)

where the generalized excitation, considering only the blocked pressure, comes as
 lx  ly
∗ 8Pi lx ly jΩt
f (t) = 2pi W1 (ξ, η)dξdη = e (A.32)
0 0 π2
Consequently, by substituting in Eq.(A.31), and since k1∗ = m∗1 ω12 , the factor C1 is calculated
as:
8Pi lx ly 1
C1 = ∗ 2
(A.33)
π 2 m1 (ωi − Ω2 )

REFERENCES
[1] Vigran E T. Building Acoustics. Taylor & Francis Limited, 2019, 2019.

[2] Norén-Cosgriff K, Løvholt F, Brekke A et al. Countermeasures against noise and vibra-
tions in lightweight wooden buildings caused by outdoor sources with strong low fre-
quency components. Noise Control Engineering Journal 2016; 64: 737–752.

5229
 Kalderon M., Paradeisiotis A. and Antoniadis I.

[3] Løvholt F, Madshus C and Norén-Cosgriff K. Analysis of low frequency sound and sound
induced vibration in a norwegian wooden building. Noise Control Engineering Journal
2011; 59(4): 383–396.

[4] ISO 717, acoustics — rating of sound insulation in buildings and of building elements.
Standard.

[5] ISO 10140, acoustics — laboratory measurement of sound insulation of building elements.
Standard.

[6] ISO 15186, acoustics —measurement of sound insulation in buildings and of building
elements using sound intensity. Standard.

[7] Ayr U, Martellotta F and Rospi G. A method for the low frequency qualifica-
tion of reverberation test rooms using a validated finite element model. Applied
Acoustics 2017; 116: 33–42. DOI:[Link] URL
[Link]

[8] Løvholt F, Norèn-Cosgriff K, Madshus C et al. Simulating low frequency sound transmis-
sion through walls and windows by a two-way coupled fluid structure interaction model.
Journal of Sound and Vibration 2017; 396: 203 – 216. DOI:[Link]
2017.02.026.

[9] Aretz M and Vorländer M. Efficient modelling of absorbing boundaries in room acoustic
fe simulations. Acta Acustica United With Acustica 2010; 96: 1042–1050.

[10] Vorländer M. Computer simulations in room acoustics: Concepts and uncertainties. The
Journal of the Acoustical Society of America 2013; 133(3): 1203–1213. DOI:10.1121/1.
4788978.

[11] Beranek LL and Vér IL. Noise and vibration control engineering: Principles and applica-
tions. John Wiley & Sons, New York.

[12] Cremer L, Heckl M and Petersson BA. Structure-borne sound: structural vibrations and
sound radiation at audio frequencies. Springer Science & Business Media, 2005.

[13] Ordubadi A and Lyon RH. Effect of orthotropy on the sound transmission through ply-
wood panels. The Journal of the Acoustical Society of America 1979; 65(1): 133–139.
DOI:10.1121/1.382255.

[14] London A. Transmission of reverberant sound through double walls. The Journal of the
Acoustical Society of America 1950; 22(2): 270–279. DOI:10.1121/1.1906601.

[15] Callister J, George A and Freeman G. An empyrical scheme to predict the sound transmis-
sion loss of single-thickness panels. Journal of Sound and Vibration 1999; 222(1): 145 –
151. DOI:[Link]

[16] Pellicier A and Trompette N. A review of analytical methods, based on the wave approach,
to compute partitions transmission loss. Applied Acoustics 2007; 68(10): 1192 – 1212.
DOI:[Link]

[17] Kim YH. Sound propagation: an impedance based approach. John Wiley & Sons, 2010.

5230
 Kalderon M., Paradeisiotis A. and Antoniadis I.

[18] Paradeisiotis A, Kalderon M, Antoniadis I et al. Acoustic Performance Evaluation of

a panel utilizing negative stifffness mounting for low frequency noise control. In Pro-
ceedings of EURODYN 2020. Athens, Greece, 23–26 November: EASD Procedia, pp.
4093–4110.

[19] Cowan AJ. Sound Transmission Loss of Composite Sandwich Panels. PhD Thesis, De-
partment of Mechanical Engineering, University of Canterbury, 2013.

[20] Rayleigh JWSB. The Theory of Sound, volume 1. Macmillan, 1894, 1896. DOI:10.1017/
CBO9781139058087.

[21] Roussos L. Noise transmission loss of a rectangular plate in an infinite baffle. The Journal
of the Acoustical Society of America 1985; 75. DOI:10.1121/1.2021367.

[22] Yang Y, Lam NTK and Zhang L. Estimation of response of plate structure subject to
low velocity impact by a solid object. International Journal of Structural Stability and
Dynamics 2012; 12(06): 1250053.

[23] Koblar D and Boltežar M. Evaluation of the frequency-dependent young’s modulus and
damping factor of rubber from experiment and their implementation in a finite-element
analysis. Experimental Techniques 2013; .

[24] Carfagni M, Lenzi E and Pierini M. The loss factor as a measure of mechanical damping.
In Proc. SPIE, Proceedings of the 16th International Modal Analysis Conference, volume
3243. p. 580.

[25] Smith M. ABAQUS/Standard User’s Manual, Version 6.9. United States: Dassault
Systèmes Simulia Corp, 2009.

[26] Marburg S. Six boundary elements per wavelength: Is that enough? Journal of Computa-
tional Acoustics 2002; 10(01): 25–51. DOI:10.1142/S0218396X02001401.

5231

View publication stats