Distributed-Mode Loudspeaker

Radiation Simulation

Joerg Panzer and Neil Harris

New Transducer Ltd, Huntingdon, England

Presented at the AES 105 San Francisco 1998

A radiation model of the Distributed Mode Loudspeaker (DML) is investigated and compared to
measurements. The approach makes use of the bending wave eigen-functions and Fourier
transformation to describe the acoustic coupling. The model is implemented into a lumped element
simulator, which helps to display the complete system response including exciter and other
components.

Introduction

The Distributed Mode Loudspeaker is an electro-acoustic transducer based on the direct radiation of
modal bending waves [1-3]. The DML mechanics are described through partial differential equations and
with the help of boundary conditions for the structure and fluid. The fluid boundary is the interface
between the vibrating panel surface and the pressure of the reacting medium in which the DML is
embedded.

The radiation of the DML is governed by the wave-equation, which is again a partial differential
equation. Acoustic boundary conditions determine the solution. The acoustic field must match the
vibrating surface of the panel and other surfaces in the radiation space. In this paper we investigate the
far-field radiation pattern in free space.

Bending Wave Radiation

Assuming linearity, the radiating bending wave can be expressed as the superposition of plane transversal
velocity waves. A single plane wave can be regarded as a sub-module in order to describe complex field
patterns. Therefore let us first investigate the radiation of these sub-modules. In one dimension and
normal to the panel surface a single plane bending wave may have the form (time factor exp(jωt)
suppressed) [4]:

v( x) = A ⋅ e − j ⋅kb ⋅ x (1)
with

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105

 k b = 4 ω2 ⋅ µ / B Bending wave number [m-1]
B Bending stiffness [Nm]
µ Mass surface density [kg/m2]
A Arbitrarily constant
The generated sound pressure satisfies the wave equation and the boundary conditions of continuity at the
air-panel interface. A plane transversal surface wave can only radiate if it matches to the trace of a
longitudinal acoustic wave. The trace of the acoustical wave is the projection onto the panel surface.

Ai r k ϑ Ai r

z z

Panel
x Panel x

Figure 1 Plane transversal velocity wave in x direction.

Left: No radiation: λb < λo
Right: Radiation: λb > λo

Figure 1 displays a section through the panel where a bending wave travels in the x direction. Above, the
grid displays the sound pressure variation. The left picture of Figure 1 demonstrates the situation when
the transversal bending wavelength is smaller than the wavelength in air. The air wave cannot project a
trace on the bending wave. There is only a fluctuation very near to the surface exponentionally decaying
in the z-direction. Hence no real sound power is radiated.

The right picture displays the sound field above coincidence, where the bending wavelength is longer
than the wavelength in air. The direction of radiation is called coincidence angle ϑc and is related to the
speed of sound in air, co:

λo c µ
sin(ϑc ) = = o ⋅4 (2)
λb ω B

The frequency where the wavelength is the same for the bending wave and for the sound wave in air is
called coincidence frequency, fc:

c 2o µ
fc = ⋅ (3)
2⋅π B

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 2

Far-Field Response

The far-field response of any vibrating area in an infinite baffle is proportional to the velocity
wavenumber spectrum [5].

e − j⋅k o ⋅R (
p(R,ko , ϑ,ϕ ) = j ⋅ k o ⋅ Z o ⋅ ⋅ v (ko , ϑ, ϕ ) R→∞ (4)
2⋅ π ⋅R
with

p (R,ko,ϑ,ϕ) Sound pressure at frequency f = k o ⋅ c o / 2π ,

listening angle ϑ, ϕ and distance R.
Zo Specific plane wave impedance Zo = ρo co
ρo, co Density and sound velocity of air
(
v (ko ,ϑ,ϕ) Spatial Fourier-transform of panel velocity.
Arguments mapped to listening angles ϑ and ϕ.
(
The velocity wavenumber spectrum, v (k x ,k y ) , is the two dimensional Fourier transform of the radiating

(
∫
velocity field due to a bending wave in the xy-plane: v (k x ,k y ) = v ( x, y ) ⋅ e j⋅ (k x ⋅ x + k y ⋅ y ) dx dy . Thus the

radiating velocity profile of the plate is described with the help of plane waves.

z P1 P1

ϑ y
ϑ ky

y
ko
Panel ko
ϕ

Ko
x
x Kx

Figure 2 Left: Listening angles in space. P1: receiving point

Right: Mapping of listening angles onto k-plane

From this spatial spectrum only those wavenumber components whose k-factor is smaller than the
wavenumber, ko, radiate into the far-field, as explained above. The wavenumbers are related to the
listening angles ϑ and ϕ by (Figure 2)

k ox = k o ⋅ sin ϑ ⋅ cos ϕ and k oy = k o ⋅ sin ϑ ⋅ sin ϕ (5)

On axis response gives kox = 0 and koy = 0. The maximum range which can be covered is given by the
radius ko as illustrated in the right diagram of Figure 2.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 3

Aperture Function

Let us call any formulation for the

mechanical velocity profile normal to
the panel, vo(x,y). On the baffle, outside
the vibrating plate, the normal
component of velocity is zero, since Mechani cal
we assume a reflecting, rigid and boundar y
condi t i ons
infinitely large baffle. Thus any
Figure 3 Aperture function A(x,y) on top of panel velocity function vo(x,y)
formulation of the acoustic velocity
profile is strictly limited to the
vibrating surface. The acoustically active velocity function is therefore: v(x,y) = vo(x,y)⋅A(x,y). We call A(x,y)
the aperture function. A(x,y) = 1 for x,y inside the radiating area and zero elsewhere as illustrated in Figure
3. The spatial Fourier transform of v(x,y) = vo(x,y)⋅A(x,y) is therefore the convolution of v( o ( k x ,k y ) with
(
A (k x ,k y ) , which is the spatial Fourier transform of A(x,y):

( ( (
v (k x , k y ) = v o (k x , k y ) ∗ A (k x , k y ) (6)

This result is important since it enables the radiation of subsonic plane bending waves. As soon
as we apply any limits to the velocity field the acoustic field is distorted, as illustrated in Figure
3. The cancellation process below coincidence is now leaky and some of the energy can be
transduced into the far-field. Above coincidence the directivity is smoothed by the aperture
(
spectrum A (k x ,k y ) and radiation occurs in all directions.

Figure 4 Magnitude of near-field sound pressure in front of an aperture placed over a plane bending wave.
Left: below coincidence. Right: above coincidence.

Figure 4 displays the near field of the sound pressure, which is created by a single plane bending wave
according to equation (1) but now limited to the size of a panel. The plots display the linearly scaled
sound pressure contours. The distortion of the sound field contours is clearly visible. The parameters used
are the same as for the following examples of the large panel. At the bottom of the diagrams the finite
vibrating panel area is symbolized with the help of an aperture covering an infinite plane bending wave.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 4

Examples

In order to further demonstrate the principles of bending wave radiation, let us investigate an example of
the radiation of an infinitely large panel with and without an aperture. This intermediate step has the
advantage of enabling the study of the bending wave radiation of a finite aperture without the involved
details of modal vibration of the finite panel. The material properties of the sample panel shall be such
that the bending stiffness is B = 10 Nm and the mass surface density is µ = 1.015 kg/m3. With the above
values for B and µ the coincidence frequency is fc = 6 kHz corresponding to kc = 109.6 m-1.

On the left hand side of the following series of diagrams the wavenumber spectrum is displayed. For the
sake of clarity we display the properties for one dimension only. The abscissa is therefore the
wavenumber kx in the x-direction. The ordinate is the sound pressure arbitrarily scaled since the actual
level is not of interest here. On the right hand side the associated polar plot according to equation (5) is
displayed. The ordinate of the polar plot is arbitrarily log scaled.

1. No aperture - below coincidence

Figure 5 No aperture - below coincidence, f = 4 kHz

In this first example let us assume a free bending wave cos(kb⋅x) of f = 4 kHz, which is below
coincidence. kb is marked in Figure 5 with the help of two arrows, because the Fourier transform of a
single plane wave is a Dirac function. No radiation can occur, which an empty directivity plot on the right
hand side symbolizes (see also Figure 1, right diagram). According to the mapping of (5) the listening
angle covers the range of kx = -73 m-1 ... +73 m-1 while the angle ϑ runs from -90° ... +90°. The range is
marked with the help of dotted lines. The given frequency, f, of our free bending wave corresponds to a
wave number of kb = 89.5 m-1, which is outside of the listening angle range where the limit is given by ko
= 73 m-1.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 5

2. No aperture - above coincidence

Figure 6 No aperture - above coincidence, f = 8 kHz

In this example the free bending wave cos(kb⋅x) is above coincidence, at f = 8 kHz. The listening angle
range now includes the two Dirac functions (Figure 6). Hence a strong radiation occurs in a direction
where the trace of the acoustic wave matches the bending wave of the panel (see also left diagram of
Figure 1).

3. Large aperture - below coincidence

Figure 7 Large aperture - below coincidence, f = 4 kHz

As illustrated in Figure 3 and Figure 4 the infinite panel is now covered by an opaque screen containing a
slit of width Ax = 0.5 m. We assume a plane bending wave cos(kb⋅x) travelling in the x-direction. The
origin is placed in the center of the aperture. In spite of driving below coincidence a far-field radiation is
possible. The reason is the distorted acoustic field as sketched in Figure 3. At some point, generated
sound pressure can not be totally cancelled by an inverted pressure at some other point. The residual
forms the radiated sound power. The power is related to the grayed zone under the curve in Figure 7, i.e.
exactly the range, which can be covered by the listening angle. On the right hand side of Figure 7 this
range is plotted again in form of the directivity plot. Compared to Figure 4 it can be seen how the far-field
pattern evolves from the aperture.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 6

4. Large aperture - above coincidence

Figure 8 Large aperture - above coincidence; Ax = 0.5 m, f = 8 kHz

We have the same situation as in the previous example but our bending wave corresponds now to a
frequency of f = 8 kHz, i.e. well above coincidence of fc = 6 kHz. The 'coincidence beams' are within the
range, which the listening angle is able to cover (grayed zone in Figure 8). The Dirac function of example
2 is now convolved with the aperture spectrum. The polar plot displays the strongly emphasized radiation
at the coincidence angle ϑc. The right diagram of Figure 4 displays the corresponding near-field at the
aperture.

Exactly at coincidence (f = fc = 6 kHz) the range of the listening angle would end at the peaks of Figure 8,
i.e. the coincidence angle is in this particular case ϑc = ±90°.

5. Small aperture - below coincidence

Figure 9 Small aperture - below coincidence; Ax = 0.25 m, f = 3 kHz

This example demonstrates the effect of the aperture-size below coincidence. The situation is the same as
in example 3 but the width of the slit is halved (Ax = 0.25 m). The smaller aperture causes a stronger
distortion of the acoustic field. Hence a stronger radiation can be observed as demonstrated in Figure 9.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 7

6. Small aperture - above coincidence

Figure 10 Small aperture - above coincidence; A x = 0.25 m, f = 8 kHz

Above coincidence the effect of halving the slit width of the aperture (Ax = 0.25 m) increases the width of
the coincidence beams as shown in Figure 10.

By comparing the results of the examples 1 to 6 it is obvious that the aperture size controls the directivity.
Below coincidence a large aperture (in comparison to the wavelength) diminishes the acoustic coupling.
Above coincidence the radiated energy is focused into a certain direction called coincidence angle. Below
coincidence a small aperture increases the radiation. Above coincidence the radiation in the
neighbourhood of the coincidence angle is broadened.

In two dimensions similar results are obtained because a plane wave travelling in any direction can
always be separated into the components relating to the x and y directions.

The Finite Panel

So far we have investigated the interaction of a single plane bending wave radiating through an
aperture. Hence there are no reflecting bending waves and no boundary conditions to be
matched. In a finite plate on the other hand, the bending wave field has to satisfy boundary
conditions. For example for a finite free vibrating panel, there are no external forces and
moments applied at its edges. Of course, any panel of finite size comes automatically with an
'in-built' aperture function A(x,y), the size of which is usually the same as that of the vibrating area
of the panel. The velocity profile of a finite panel can be described with the help of a series
expansion in a linear independent function system. Well-known examples are the Fourier or
Hankel series where the function system consists of sin-functions or Bessel-functions,
respectively. For rectangular panels an approach, which uses so-called beam-functions, has
certain advantages. The beam-functions are themselves a series of sin-functions and those
properties are closely related to plate mechanics. Without going into details here, we can thus
always write our panel velocity profile in the form

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 8

 v o ( x, y ) = ∑v ⋅ φ
i
i i ( x, y ) (7)

with
vo(x,y) Bending wave velocity at point x, y
i Mode number
vi i-th mode coefficient
φi (x,y) i-th mode shape or eigen function

The velocity profile vo(x,y) has to be a solution of the governing plate differential equation and has to
match the boundary conditions at the edge of the panel as displayed in Figure 3. Both criteria determine
the choice of eigen-functions φi (x,y) and their corresponding eigen-values for each mode, i. Together with
these, the mode coefficients vi are formed through the location of driving points, material parameters and
radiation impedance. For the purpose of this paper the involved details of vi and φi (x,y) are not necessary.

As mentioned above the acoustical active velocity is v(x,y) = vo(x,y)⋅A(x,y). Since the mode coefficients vi are
independent of x and y the only step we have to do is to apply the spatial Fourier transformation to the
mode shape functions φi (x,y), i.e.

( (
∑v ⋅ φ
(
v (k x ,k y ) = i i (k x , k y ) ∗ A ( k x ,k y ) (8)
i

Thus inserting (8) into the far-field Rayleigh formula (4), the far-field response of the finite panel in a
baffle is the convolution of the Fourier spectrum of eigen-functions with the spectrum of the aperture.

Examples

As already mentioned we are in some sense free to choose an eigen-function system. The proper choice
depends mainly on numerical reasons and on the way the input parameters are given. As an example we
implement the in-vacuum beam-model [6]. Since the eigen values of this model do not take into account
the radiation impedance there is a coupling between the mode coefficients vi, which has to be calculated
separately. Further the aperture function A(x,y) is not included in the beam eigen-functions and must be
applied explicitly as given by equation (8).

The following comparisons between simulation and measurement display the response of (8) in principle.
These should be regarded as an intermediate report of development of the DML radiation model. The
beam-mode-model for panels is implemented in a lumped element network simulator [7]. Some parts are
missing in the following simulation, for example the mode model for the driving force and the diffraction.
The electro-mechanical part is the same as described in [8], i.e. the driving force is acting on the
impedance of an infinite panel, which is mainly a constant. In reality the modal impedance fluctuates
strongly around this constant. Thus the driving power for each mode fluctuates accordingly. Since the
sum of all modes controls the directivity, the frequency response of the input has an effect on the

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 9

directivity. Further the current model does not include diffraction. In order to take into account some
aspects of two-sided radiation (no in-plane pressure) a model for an absorbing baffle is used.

The points on the polar plots are measured in 5° steps in a reflection free environment. The measured
panels are unbaffled and the response may include diffraction from the rearward side. The absolute
scaling of the diagrams is ignored but the range is maintained and is 50 dB.

The used panels are not particularly selected to be good ones for loudspeaker design. The selection
ensured a light, stiff and low loss material.

1. Small panel

Panel parameter
Size: Lx = 260 mm, Ly = 230 mm (x - horizontal)
Material: Sandwich, plastic skin on foam, isotropic, B = 1.3 Nm, µ = 0.72 kg/m2
Exciter: Single, at x = 150 mm, y = 100 mm, voice coil diameter dVc = 25.5 mm
Edges: Free
Coincidence: fc = 14 kHz

The directivity plots of this panel are displayed in Figure 11. Clearly visible is the slightly asymmetric
radiation pattern because the exciter is mounted off-center. Below coincidence the directivity is almost
omni-directional. At higher frequencies coincidence lobes start to form the typical radiation pattern of
DMLs at and above coincidence. The plot at 2 kHz is affected by interference between the rearward
radiated sound with the direct one. The mean pathlength between the center of the rearward side to the
front is half a wavelength at 2.8 kHz, where the strongest cancellation occurs on-axis.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 10

 Figure 11 Directivity: small panel, simulation - measurement

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 11

2. Large panel

Panel parameter
Size: Lx = 600 mm, Ly = 600 mm (x - horizontal)
Material: Sandwich, plastic skin on honeycomb,
orthotropic, mean B = 7 Nm, µ = 1.2 kg/m2
Exciter: Single, at x = 265 mm, y = 330 mm, voice coil diameter dVc = 25.5 mm
Edges: Free
Coincidence: fc = 7.8 kHz

Figure 12 displays the directivity plots of the larger panel. The lower coincidence frequency together with
the relative large size leads to stronger side radiation, which already starts above 2 kHz. Clearly visible
are the 'coincidence beams' at high frequencies. Again interference caused by pathlength difference
between rear and front side causes the indentation at the 2 kHz and 5 kHz measurement plots.

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 12

 Figure 12 Directivity: large panel, simulation - measurement

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 13

Conclusion

The principles of bending wave radiation are explained step by step. Starting with the radiation of a single
one dimensional plane wave below and above coincidence. The effect of an opaque screen with an
aperture on the radiation is demonstrated. The aperture effect is used for explanation of the general
baffled far-field radiation formula. Several examples are given. A simulation model for a finite panel is
demonstrated and discussed in comparison with measurements.

Further research is focusing on a complete modal transmission system for the distributed mode
loudspeaker, i.e. also including the mechanical part. An important area of investigation is what effect
diffraction has on a radiating panel.

Acknowledgement

The authors would like to thank Mr. Vladimir Gontcharov for the measurements and their colleagues for
help and inspiring discussions.

References

[1] Azima, H.; NXT White Paper; New Transducer Ltd. Huntingdon PE18 6ED

[2] Harris, N.; Hawksford, M.; The Distributed-Mode Loudspeaker (DML) as a Broad-Band Acoustic
Radiator; AES 103rd Convention September 1997

[3] Azima, H.; Harris, N.; Boundary Interaction of Diffuse Field Distributed-Mode Radiators; AES
103rd Convention September 1997

[4] Cremer, L.; Heckl, M.; Ungar, E: Structure-Borne Sound, Springer Verlag (1973)

[5] Morse, P; Ingard U.; Theoretical Acoustics; McGraw Hill 1968

[6] Blevins, R. D.; Formulas for Natural Frequency and Mode Shape; Krieger Publ., Malabar 1984

[7] Panzer, J.; Campbell, R.; Multiple Driver Modeling with a Modern Lumped Element Program;
AES 102rd Convention March 1997

[8] Panzer, J. W.; Harris, N.; Distributed Mode Loudspeaker Simulation Model; AES 104th
Convention May 1998

Panzer, Harris: Distributed-Mode Loudspeaker Radiation Simulation, AES 105 14