Audio Engineering Society

Convention Paper
Presented at the 117th Convention
2004 October 28–31 San Francisco, CA, USA

This convention paper has been reproduced from the author's advance manuscript, without editing, corrections, or consideration
by the Review Board. The AES takes no responsibility for the contents. Additional papers may be obtained by sending request
and remittance to Audio Engineering Society, 60 East 42nd Street, New York, New York 10165-2520, USA; also see www.aes.org.
All rights reserved. Reproduction of this paper, or any portion thereof, is not permitted without direct permission from the
Journal of the Audio Engineering Society.

Voice Coil Impedance as a Function of Frequency and

Displacement
1 2 3
Mark Dodd , Wolfgang Klippel , and Jack Oclee-Brown.

1
KEF Audio UK (Ltd),Maidstone,Kent,ME15 6QP United Kingdom.
Mark.dodd@KEF.com
2
Klippel GmbH, Dresden, 02177, Germany.
wklippel@klippel.de
3
KEF Audio UK (Ltd),Maidstone,Kent,ME15 6QP United Kingdom.
Jack.oclee-brown@KEF.com

ABSTRACT
Recent work by Klippel [1] and Voishvillo et al.[2] has shown the significance of voice coil inductance in respect to
the non-linear behaviour of loudspeakers. In such work the methods used to derive distortion require the inductance
to be represented by an equivalent circuit rather than the frequency domain models of Wright and Leach. A new
technique for measurement of displacement and frequency dependant impedance has been introduced. The complex
relationship between coil impedance, frequency and displacement has been both measured and modelled, using FE,
with exceptional agreement. Results show that the impedance model requires that its parameters vary independently
with x to satisfactorily describe all cases. Distortion induced by the variation of impedance with coil displacement is
predicted using a lumped parameter method, this prediction is compared to measurements of the actual distortion.
The possibility of using an FE method to predict distortion is demonstrated. The nature of the distortion is discussed.
Dodd et al. Voice coil impedance

also been predicted using FEM analysis. A new

1. INTRODUCTION impedance model with displacement varying parameters
is discussed and fitted to the measured data. Finally,
In order to describe the performance of a loudspeaker distortion induced by the variation of impedance with
precisely, one must consider the electrical input coil displacement is predicted using both lumped
impedance at higher frequencies. The voice coil does parameter and FE methods, these predictions are
not operate in free air but close to conducting and compared to measurements of the actual distortion. The
magnetically permeable structures: the pole tips, the nature of this distortion is discussed.
magnet, the voice coil former, copper rings etc. The
impedance can be only roughly modelled as a resistor 1.1. Glossary
Re and an ideal inductance Le, the occurrence of eddy
currents in these structures usually decreases the Bl(x) force factor (Bl product)
inductance of the coil and increases losses at higher Cms(x) = 1 / Kms(x) mechanical
frequencies [3]. compliance of loudspeaker suspension
The nature of the voice coil impedance is dependent Exm exponent of imaginary part in WRIGHT
upon the geometry and physical properties of the model
magnet and surrounding assemblies. There are three Erm exponent of real part in WRIGHT
prominent methods of deducing the coil (blocked) model
impedance for a particular loudspeaker. Fm(x,i) reluctance force
- The blocked impedance may be measured by Krm factor of real part in WRIGHT model
clamping the voice coil stationary within the Kxm factor of imaginary part in WRIGHT
magnet assembly to remove the motional part of the model
impedance. Blocked impedance may then be
K factor in LEACH model
measured as with conventional impedance
measurements. Kms(x) mechanical stiffness of loudspeaker
suspension
- Various impedance models have been developed in
Le(x) inductance used (cascaded model)
order to describe the electrical behaviour of the
coil, these models may be fitted to conventional L2(x) inductance (cascaded model)
unblocked impedance measurements. Leff(f,x) effective inductance depending on
- Recently magnetic transient FEM has been Mms mechanical mass of loudspeaker
successfully applied to deduce the blocked diaphragm assembly including air load
impedance [4]. and voice coil
n exponent in LEACH model
At large amplitudes the impedance also varies
significantly with voice coil displacement. Whereas the Rms mechanical resistance of total-
frequency dependence of the impedance is a linear loudspeaker
phenomenon effecting the amplitude response, but not Re electrical voice coil DC-resistance
generating distortion, variation of impedance with R2(x) resistance (cascaded model)
displacement generates significant harmonic and inter-
Reff(f,x) effective losses due to eddy currents
modulation distortion.
depending on frequency and
The nature of the impedance generated distortion is a displacement
dependent upon the coil impedance, or more precisely
ZL(j ω,x) complex excess impedance representing
how the coil impedance varies with frequency &
the effect of the lossy inductance
displacement of the coil.
(motional impedance and Re is
This paper investigates the voice coil impedance as a removed)
function of frequency and of displacement.
Measurements of frequency & displacement dependent angular frequency = 2 f
impedance have been made using a quasi-static method
using the LPM module of the Klippel analyser [5].
Frequency & displacement dependent impedance has

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 2 of 19
Dodd et al. Voice coil impedance

2. MODELLING electrical equivalent circuit (as shown in Figure 2 b) or

as a digital IIR filter.
2.1. Lumped Parameter Model ZL(jω,x)
a)
Re Cms(x) Mms Rms Fm(x,I)
ZL(jω,x) L2(x)
v
I Re Le(x)

U Bl(x)v Bl(x) Bl(x)I

b)

R2(x)

Figure 1 Electro-mechanical equivalent circuit of the loudspeaker

Leff(f, x) Reff(f, x) c)

Figure 2 Representation of the electrical impedance

Figure 1 shows a simple equivalent circuit of a
loudspeaker system. The dominant nonlinearities are the
variation of the force factor Bl(x), the mechanical 2.1.3. WRIGHT model
compliance Cms(x) and the electrical impedance
Wright [7] proposed a model using separate weighted
ZL(j ω,x) with voice coil displacement x. The DC
resistance Re and the motional impedance are not power functions in ω for both the real and imaginary
part of impedance.
considered in the electrical impedance ZL(jω,x). Thus
the impedance ZL(j ω,x) may be measured at the ZL(j ω)= Krm· ωErm + j·(Kxm· ωExm ) (3)
electrical terminals by blocking the movement of the
coil and subtracting the resistance measured at a very This model uses four free parameters and normally gives a
low frequency. better fit than the other models with less parameters.
Unfortunately, this function can not be directly realised as
Different linear models have been developed to describe an analogue or digital system.
the frequency dependency of ZL(jω,x) with a minimal
2.1.4. Effective inductance
number of free parameters:
2.1.1. LEACH model ZL(j ω) = Leff·(f)j ω + Reff(f)
(4)

M. Leach [6] proposed a weighted power function of the
complex frequency as an approximation for ZL
ZL(j ω)= K·(j ω)n ; ω= 2 f
Although using only two free parameters this function
can sometimes give a very good fit over a wide
frequency range. Unfortunately, this function can not be
represented by an electrical equivalent circuit nor a
simple digital system.
2.1.2. LR-2 Model
This model uses a series inductance Le connected to a
second inductance L2 shunted by resistance R2.
ZL(j ω) = Le·j ω + (R2·L2·j ω ) / (R2 + L2·jω)
Although this model uses three free parameters it often
provides a worse fit to measured ZL than the LEACH
model. However, this model may be realised as an
M. Leach also proposed normalising the imaginary part
of the electrical impedance ZL(jω) to the frequency jω
(1) and introducing an effective inductance Leff(f) which
varies with frequency. The real part of ZL(jω) may be
considered as a frequency depending resistance Reff(f)
describing the losses due to eddy currents as shown in
Figure 2c . Though the number of parameters is very
high , two parameters for each frequency point, both
parameters are easy to interpret and convenient for
graphical representation.
2.1.5. Large signal modelling
The linear models may be easily expanded to higher
(2) amplitudes by allowing each parameter to be dependent
upon the displacement x.
For example considering the LR-2 model, the three
parameters Le(x), R2(x) and L2(x) are functions of the
displacement x and may be approximated by a truncated
power series expansion such as

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 3 of 19
Dodd et al.
N (5)
Le ( x) = li x i
i =0
N (6)
R2 ( x) = ri xi
i =0
N (7)
L2 ( x) = λi xi
i=0
where the coefficients li, ri and i are the free parameters
of the model.
3. MEASUREMENT
The application of a model to a particular real object
usually requires an estimation of the free model
parameters in such a way that the model describes the
real object with maximal accuracy.
With the linear models straightforward techniques are
available which may be applied for loudspeakers at
small amplitudes.
Non-linear models require special techniques for the
parameter identification. Static, quasi-static and full
dynamic techniques have been developed to measure
the force factor Bl(x), compliance Cms(x) and
inductance parameter Le(x). The dynamic techniques
have the advantage that an audio-like ac signal is used
for excitation and the loudspeaker is operated under
working conditions.
The current version of the LSI module in the Klippel
analyser performs a dynamic measurement using a noise
stimulus [8]. The free model parameters are optimised
to give the best fit between measured and modelled
current and displacement. The LR-2 model is currently
constrained so that the three lumped parameters vary
with the same shape in x.
Le ( x) R2 ( x) L2 ( x) (8)
≈ ≈ . professional loudspeaker stand as shown in Figure 3.
Le (0) R2 (0) L2 (0)
Although this assumption is a good approximation for
most loudspeakers without shorting rings it is a purpose
of the paper to investigate the validity of this
approximation for more elaborate loudspeakers using
copper cups and aluminium rings for reducing the
inductance nonlinearity.
Clearly the measurement of suspension stiffness using a
dc offset gives significantly different results from a
dynamic measurement due to creep, relaxation and other
AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 4 of 19
Voice coil impedance
time dependent properties. Additionally, due to flux
modulation, the force factor Bl(x) and the inductance
Le(x) are dependent on the current i. Thus using an
audio-like signal will produce far more meaningful data
than measurement with extremely small input current
(coil offset generated by external force or pressure) or
extremely large currents (coil offset generated by a dc
current).
Despite these limitations a quasi-static technique is a
useful method for investigating the variation of the
impedance with both frequency and displacement.
Measurements have been performed on two test
loudspeakers. Both loudspeakers share the following
specifications:
Voice coil diameter: 2” nominal (51.30mm ID)
Voice coil DCR: 6.72 Ohms
Turns in Voice coil: 126 Turns
Ferrite ring magnet
Annular low carbon steel top-plate
Pole/plate assembly type yoke.
The magnet assembly of loudspeaker 1 has an
aluminium shorting ring placed above the magnetic gap.
The magnetic assembly of loudspeaker 2 has an
aluminium shorting ring placed below the magnetic gap.
The geometry of the two loudspeakers is shown in
Figure 26 & Figure 27. The pair serve to illustrate the
influence of the effect of aluminium rings on the
variation of the voice coil inductance with displacement.
3.1. Mechanical Setup
A method for measurement of the displacement
dependent impedance was developed at GP Acoustics
(UK). The method is a simple modification to the
standard Linear Parameter Measurement (LPM) of the
KLIPPEL Analyser system. The measurements
presented in this paper were performed by Klippel
GmbH.
The loudspeaker is clamped in vertical position in the
An additional spider is attached to the diaphragm.
The spider holds an inner clamping part made of
aluminium which is secured to the lower rod (usually
used for holding the microphone).
By shifting the lower rod a displacement may be
imposed to the coil position. Clearly displacing the coil
will also change the other parameters such as Bl(x),
Cms(x) and also loss factors.
Dodd et al. Voice coil impedance

The Distortion Analyser also provides a Displacement
Meter ([5] Hardware) this is used for measuring the
original rest position of the cone and to measure the
imposed static displacement. Throughout this paper the
convention is that a positive displacement of the coil
refers to movement out of the magnet assembly.
3.2. Small signal measurements
The additional spider increases the stiffness of the total
suspension and the resonance frequency. However, the
modified loudspeaker is still a second-order system and
can be represented by the equivalent circuit in
Figure 1.
The module Linear Parameter Measurement (LPM) of
the KLIPPEL Analyser system is used to measure the
linear parameter at each prescribed displacement.
The loudspeaker is excited by a multitone signal of 0.5
V rms at the terminals. Since the voltage, current and
displacement are measured simultaneously all of the
linear parameters can be identified instantaneously. An
additional measurement with a mechanical perturbation
(additional mass or measurement in a test enclosure) is
not required.
A sparse multitone signal used as excitation signal
allows assessment of the distortion generated by the
loudspeaker. During the small signal measurements the
maximum distortion occurred 20 dB below the
fundamental lines in the current spectrum. This shows
sufficiently linear operation of the loudspeaker [9].

3.3. Fitting of the inductance model

Figure 3 Measurement Setup.
The loudspeaker under test is connected to Distortion
Analyser 2 allowing a simultaneous measurement of
voltage, current and displacement signal.
At first the linear parameters are measured at the rest
position (x=0) and the different inductance models (LR-
2, WRIGHT, LEACH) are used to describe the
impedance response, measured up to 18 kHz.
110

100

90

80
Impedance [Ohm]

70

60

50 LR-2

40
Measured
30
WRIGHT
20

10 LEACH
Figure 4 Generating an additional DC offset by the lower rod
connected by an addition spider to the diaphragm. 2 5 10 20 50 100 200 500 1k 2k 5k 10k
Frequency [Hz]

Figure 5 Magnitude of electrical impedance of loudspeaker 1

measured and fitted by LR-2, Wright and LEACH model up to 18
kHz.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 5 of 19
Dodd et al. Voice coil impedance

Figure 5 shows a measured curve and the fitted curves
using the three models. The LEACH and WRIGHT are
able to describe this particular impedance curve very
well. The LR-2 causes minor deviations about 500 Hz
and 5 kHz. Although the WRIGHT usually gives the
best fit there are cases where the other models have
provided a superior fit.
Since the test loudspeaker is based on a woofer intended
impedance below 100 Hz impossible. The calculated
phase of the excess impedance is about 68 degrees with
a small decay to higher frequencies. The LEACH model
assumes a constant phase, which proves to be a good
approximation for this particular loudspeaker.
50
Phase (fitted)

for frequencies below 200 Hz the models are have also 101
Phase (measured)
been fitted using data only up to 2 kHz, Figure 6. In this 0

instance all models are able to give a good fit to the

[deg]
measured curve.

[Ohm]
-50
100
110
-100
100 Magnitude (measured)

90 Magnitude (fitted)
10-1 -150
80

70 2 5 10 20 50 100 200 500 1k 2k 5k 10k

Frequency
60
Figure 7 Magnitude and phase of the electrical impedance of
Impedance [Ohm]

50
Measured loudspeaker 1 ZL(jω) measured (solid lines) and fitted by using the
40 LEACH model (dotted lines).
WRIGHT
30 LEACH
LR-2 Phase (fitted) 100
20

10
101
2 5 10 20 50 100 200 500 1k 2k 50
Frequency [Hz]
Phase (measured)
Figure 6 Magnitude of electrical impedance of loudspeaker 1 0
measured and fitted by LR-2, WRIGHT and LEACH model up to 2

[deg]
100
[Ohm]

kHz.
Magnitude (fitted) -50

Magnitude (measured)
3.4. Excess impedance ZL 10-1 -100

The LPM module also calculates the amplitude and

phase response of the excess impedance ZL(jω). The -150

measured and the fitted curves are shown in Figure 7 10-2

using the LEACH model. The measured curves are
calculated by subtracting the estimated dc resistance Re
and the motional impedance (calculated from the
estimated parameters Bl, MMS, RMS, and CMS) from the
total electrical input impedance. The magnitude
increases with frequency, normally with a slope usually
less than 6 dB per octave. The LEACH model uses a
constant slope corresponding with the exponent n in
Equation (1). Close to loudspeaker resonance the
magnitude of the calculated excess impedance varies
significantly. In this region the motional impedance is
very high (~100 Ohm), measurement and modelling
error (in the order of 1 %) will be assigned to the excess
impedance and makes accurate estimation of the excess
2 5 10 20 50 100 200 500 1k 2k 5k 10k
Frequency
Figure 8 Magnitude and phase of the electrical impedance ZL(jω) of
loudspeaker 1 measured (solid lines) and fitted by using the LR-2
model (dotted lines).
Figure 8 shows the excess impedance match using the
LR-2 model. While the fitting above 1 kHz is good, at
lower frequencies there are significant differences in
both phase and amplitude. The LR-2 model, and also
other shunted models using more Li and Ri elements (i >
2), behaves as an ideal inductance at very low
frequencies giving a 6dB per octave slope and a phase
shift of 90 degrees. This property corresponds with the
observation that the eddy currents are frequency

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 6 of 19
Dodd et al. Voice coil impedance

dependent and will vanish at very low frequencies.
Though a small increase in the measured phase at low
frequencies supports this observation, the phase shift of
the LR-2 model begins at a higher frequency than the
measured shift. Thus the LR-2 is usually limited to use
over a frequency band of two decades. Using an
additional shunted section (R3 and L3) improves the fit
significantly and results in a good description over the
whole audio band (three decades). The cascade of
shunted inductances is a minimum-phase system and
can be realised in the analogue or digital domain.
50
Phase (fitted)
101
0
hardware unit, Distortion Analyser 2, was used to
measure the offset. The LPM module was then again
used to measure the linear parameters. The excess
impedance ZL is displayed for loudspeaker 1 in Figure
10 for a negative offset of -8mm (coil in) and a positive
offset of 7.5mm (coil out), the impedance at the rest
position is also shown. The variation of the impedance
with displacement may be clearly seen. It may be
observed that the inductance of the coil is effectively
reduced in the region near to the aluminium shorting
rings, above the gap in this case. Conversely as the coil
moved into the magnetic assembly, the effective
inductance is increased as the coil moves away from the
pole piece and top plate.
70
60

Phase (measured)

[deg]
50
[Ohm]

-50
Magnitude (measured)
100 40

[Ohm]
7.5 mm
-100 30 0 mm

Magnitude (fitted) 20 - 8 mm

10-1 -150
10

2 5 10 20 50 100 200 500 1k 2k 5k 10k

0
Frequency [Hz] 2 5 10 20 50 100 200 500 1k 2k 5k 10k
Frequency [Hz]
Figure 9 Magnitude and phase of the electrical impedance ZL(jω) of
loudspeaker 1 measured (solid lines) and fitted by using the WRIGHT Figure 10 Magnitude of electrical impedance ZL(jω) of Loudspeaker 1
model (dotted lines). measured at rest position (solid line), at –8 mm (dotted line) and 7.5
mm (dashed line)
The excess impedance fitted by the WRIGHT model is
shown in Figure 9. The magnitude response can be The phase of the excess impedance ZL was also
approximated by smooth line having a different slope at calculated and is shown in Figure 11 at the same coil
low and high frequencies (corresponding with positions. Whereas the impedance magnitude varies by
exponents Erm and Exm). Contrary to the LEACH up to 40 % at higher frequencies the phase stays almost
model the phase is not constant but depends on all four constant at 70 degrees.
parameters. The WRIGHT model also considers the 150
response below 100 Hz, where measurement errors and 7.5 mm
noise have corrupted the measurement, and generates a 100

decrease of phase shift at very low frequencies. Since 50

the WRIGHT model is not bounded to be minimum
-8 mm
phase and not composed from a system lumped 0
0 mm
[deg]

electrical elements it may be deceived by measurement

-50
artefacts when used to represent a measured curve. Thus
a good match with the measured impedance curve does -100
not guarantee that the parameters are meaningful.
-150

3.5. Impedance versus displacement 2 5 10 20 50 100 200 500 1k 2k 5k 10k

Frequency [Hz]

Once measurements had been performed at rest position Figure 11 Phase of electrical impedance ZL(jω) of Loudspeaker 1
(x=0) a dc offset was imposed upon the coil using the measured at rest position (solid line), at –8 mm (dotted line) and 7.5
mm (dashed line)
lower rod in Figure 4. The Displacement Meter at the

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 7 of 19
Dodd et al. Voice coil impedance

3.6. Non-linear Parameters to our observations regarding the measured excess

Having identified the linear parameters for different
values of coil offset, the displacement dependency of
the parameters were be calculated using the Math
Processing Software (MAT), a free programmable
(SCILAB or MATLAB) module [10] for the KLIPPEL
Analyser. This module imports all of the results
measured by the LPM for all measured voice coil
offsets and calculates the coefficients li, ri and i in
equations (5) - (7).
impedance, it can be seen that the fitted model also
identifies the effective inductance as reducing when the
coil is close the aluminium shorting ring above the
magnetic gap. It is interesting to observe that the shape
of the parameter functions; Le(x), R2(x) and L2(x); is
very similar in this instance. In this case the assumption
of (8) appears to be valid.
1.2 -7.0 mm
0.0 mm
1.1

1.0
0.70
0.9
0.65
L2(X)
0.8

[mH]
0.60 7.0 mm

0.7
0.55
Le(X)
[mH]

0.6
0.50
0.5
0.45
0.4
0.40

0.35 100 101 102 103 104

Frequency [Hz]

Figure 14 Effective inductance Leff(f,x) versus frequency f of

-5.0 -2.5 0.0 2.5 5.0 loudspeaker 1 plotted for the rest position (solid line) and –7 and +
Displacement X [mm] 7mm displacement (dotted and dashed line, respectively).
Figure 12 Inductance Le(x) and L2(x) of loudspeaker 1 versus
displacement x.

20.0
-7.0 mm

22.5 17.5
R2(X)
15.0 -0.0 mm
20.0
12.5
[Ohm]

17.5 7.0 mm
10.0
[Ohm]

15.0 7.5

5.0
12.5
2.5

10.0 0.0
100 101 102 103 104
Frequency [Hz]
-5.0 -2.5 0.0 2.5 5.0
Displacement X [mm] Figure 15 Effective resistance Reff(f,x) versus frequency f of
Figure 13 Resistance R2(x) of loudspeaker 1 versus displacement x. loudspeaker 1 plotted for the rest position (solid line) and –7 and +
7mm displacement (dotted and dashed line, respectively).
Figure 12 and Figure 13 show the parameters Le(x),
L2(x) and R2(x) of the LR-2 model versus displacement Use of the effective resistance Reff(f,x) and the effective
x for loudspeaker 1 (ring above the gap). The LR-2 inductance Leff(f,x), as defined in equation (4) and
model is used as the parameters have an analogue illustrated in Figure 2 c, simplifies interpretation of the
representation and are easy to interpret. Corresponding excess impedance. Figure 14 shows the effective

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 8 of 19
Dodd et al. Voice coil impedance

inductance Leff(f,x) for three different voice coil
displacements based on the LR-2 model. It is clearly
shown that the voice coil inductance decreases if the
coil moves outwards and increases as the coil moves in.
At low frequencies the effective inductance Leff(f,x) is
equal to the sum of Le(x) and L2(x) and the effective
resistance Reff(f,x), Figure 15, is close to zero. At high
frequencies the Leff(f,x) is equal to Le(x) only and the
Clearly all the curves have a distinct asymmetry and
decrease with positive displacement.
A second loudspeaker has been made using the same
suspension and motor structure but with the aluminium
ring located below the magnetic gap.
80

effective resistance Reff(f,x) becomes equal to R2(x). 70

7 mm
60

50 0 mm
1.2

[Ohm]
40
133 Hz -7.5 mm
1.1
30
1.0 1546 Hz 20
0.9
10
0.8
[mH]

0
2 5 10 20 50 100 200 500 1k 2k 5k 10k
0.7 Frequency [Hz]
0.6 18000 Hz Figure 18 Magnitude of electrical impedance ZL(jw) of loudspeaker 2
(ring below) measured at rest position (solid line), at –7.5 mm (dotted
0.5 line) and 7 mm (dashed line)
0.4

-5.0 -2.5 0.0 2.5 5.0 150

Displacement [mm] 7 mm
0 mm
100
Figure 16 Effective inductance Leff(f,x) versus displacement x of
loudspeaker 1 plotted for frequencies 133 Hz, 1545 Hz and 18 kHz. 50

-7. 5 mm
0
[deg]

-50
20.0
18000 Hz
-100
17.5
-150
15.0
2 5 10 20 50 100 200 500 1k 2k 5k 10k
Frequency [Hz]
12.5
Figure 19 Phase of electrical impedance ZL(jw) of Loudspeaker 2
[Ohm]

10.0 measured at rest position (solid line), at –7.5 mm (dotted line) and 7
mm (dashed line)
7.5
1546 Hz
5.0
133 Hz The magnitude and phase of the electrical impedance
2.5 ZL(j ω) of the second loudspeaker are shown in
0.0 Figure 18 and
-5.0 -2.5 0.0 2.5 5.0
Displacement [mm] Figure 19 for three coil displacements.
Figure 17 Effective resistance Reff(f,x) versus displacement x of
loudspeaker 1 plotted for frequencies 133 Hz, 1545 Hz and 18 kHz.

Figure 16 and Figure 17 shows the variation of effective

inductance and resistance versus displacement x for
selected frequencies 133Hz, 1545Hz and 18kHz.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 9 of 19
Dodd et al. Voice coil impedance

-0.0 mm
1.1 1.1 133 Hz
7.0 mm 1546 Hz
1.0 1.0

0.9 0.9

0.8 0.8

[mH]
[mH]

-7.0 mm
0.7 0.7

0.6 18000 Hz
0.6

0.5
0.5

0.4
0.4

100 101 102 103 104 -5.0 -2.5 0.0 2.5 5.0
Frequency [Hz] Displacement [mm]
Figure 20 Effective inductance Leff(f,x) versus frequency f of
loudspeaker 2 (ring below) plotted for the rest position (solid line) and Figure 22 Effective inductance Leff(f,x) versus displacement x of
loudspeaker 2 (ring below) plotted for frequencies 133 Hz, 1545 Hz
–7 and + 7mm displacement (dotted and dashed line, respectively).
and 18 kHz.

17.5
17.5
-7.0 mm
15.0
15.0 18000 Hz
-0.0 mm
12.5
12.5
7.0 mm
[Ohm]

10.0
[Ohm]

10.0
7.5
7.5
1546 Hz
5.0
5.0
133 Hz
2.5 2.5

0.0 0.0
100 101 102 103 104
Frequency [Hz] -5.0 -2.5 0.0 2.5 5.0
Figure 21 Effective resistance Reff(f,x) versus frequency f of Displacement [mm]
loudspeaker 2 (ring below) plotted for the rest position (solid line) and Figure 23 Effective resistance Reff(f,x) versus displacement x of
–7 and + 7mm displacement (dotted and dashed line, respectively). loudspeaker 2 (ring below) plotted for frequencies 133 Hz, 1545 Hz
and 18 kHz.
The effect of changing the location of the shorting ring
can be seen most clearly in Figure 20 & Figure 21.
Loudspeaker 1 exhibited an effective inductance which
decreased as the coil moved out of the magnetic gap.
This trend is close to the reverse for loudspeaker 2, the
effective inductance of the coil remains almost constant
as the coil moves out the gap and into free air. When the
coil moves inward toward the now internally located
ring, the effective inductance is seen to fall.
Additionally, the phase varies significantly more with
the displacement.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 10 of 19
Dodd et al. Voice coil impedance

4. FEM MODELING
0.65 L2(X)

0.60
4.1. The Use of FEM

0.55 The most obvious application of FEM is for literal

Le(X) modelling in which the aim is to produce a result
[mH]

0.50 equivalent to a measurement of some particular aspect

of performance such as distortion spectra or frequency
0.45 response. FEM allows the geometry and material
properties to determine the behaviour by applying the
0.40 appropriate physical laws. However, while this gives a
wealth of information about how a loudspeaker behaves
0.35 under particular circumstances it does not explain the
behaviour or help to improve it.
-5.0 -2.5 0.0 2.5 5.0
Displacement X [mm]
A different approach is to represent the device with a
Figure 24 Inductance Le(x) and L2(x) of the loudspeaker 2 (ring system of analytic equations having variables
below) versus displacement x.
representing simplified physical aspects. This
parameterised approach allows the engineer to work
21 with manageable data and to clearly identify how
R2(X) improvements to a design may be realised in terms of
20
the simplified parameters. However, while this gives a
19 clear set of specifications & targets for design or
redesign it does not tell the engineer whether or how
18 these may be achieved.
[Ohm]

17 It is clearly most effective to use FEM and lumped

parameter modelling together along with measurements
16
to allow the reduction, as far as is possible, of the
15 assumptions within the lumped parameter models
through, for example, use of various FEM derived or
14 measured transfer functions & parameters. Use of FE as
a design tool has forced the development of this strategy
-5.0 -2.5 0.0 2.5 5.0 [4]. The process essentially allows the engineer to
Displacement X [mm] consider a particular FEM result not alone but within
Figure 25 Inductance R2(x) of the loudspeaker 2 (ring below) versus the context of other physical systems & parameters and
displacement x. in a way to which they are accustomed. Secondly the
use of the methods together allows the engineer a route
Figure 24 and Figure 25 show the estimated parameters from the parameters to the physical systems. Any
Le(x), R2(x) and L2(x) of the LR-2 model. Contrary to identified parameter goals may be readily researched
the parameters of loudspeaker 1 both the inductance L2 using the relevant FE model to investigate how or
and the shunt R2 decay symmetrically for positive and whether improvement may be achieved, and this is
negative displacement. The inductance Le(x) is still performed in terms of the actual physical geometry and
asymmetric but increases in an unusual way if the coil materials of the design not comparatively arbitrary
moves outwards. This is a very interesting result as it parameters.
demonstrates a case were the assumption of (8) would
not be valid. Until FEM evolves to the extent that a model is able to
fully represent all the physical systems of a loudspeaker
and hardware allows physical changes to be computed
in real time, linking various results & measurements
using assumptions and relationships from lumped
parameter modelling is the most puissant method
currently available.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 11 of 19
Dodd et al. Voice coil impedance

4.2. Modelling Method

A discretized model of the magnet assembly was

produced using the Flux2D a programme supplied by
CEDRAT. The model domain was axisymmetric with
second order elements. The model boundary is defined
with an outer annular ‘infinite region’. In this region the
elements have a modified co-ordinate system in which
the outer nodes are at an infinite distance. This
technique avoids the errors due to modelling only a
small region of space [11][12].
Figure 27 Loudspeaker 2. Lines of constant power density at 31.25Hz,
Evaluating the impedance requires the current and x=-5.75mm
voltage through the voice coil. This has been achieved
by means of Transient Magnetic FEM in which a In both loudspeakers the lines of equal power density
voltage source has been coupled to the voice coil region are dense in the steel pole surface and more widely
[11]. The result we are seeking is the current at each spaced in the aluminium rings. This is largely due to
time step. permeability of the steel being very much higher. With
In the coil region of the FEM model the current is increasing frequency both the extent and thickness of
restricted to flow uniformly since the coil is a stranded the skin reduce.
conductor. In the other conductive regions the magnetic The current waveform flowing through the coil is
forces produced by the current are allowed to force the extracted by evaluating the current through the coil at
current flow into a skin on the surface of the conductor. each timestep. A typical result is shown in Figure 28.
The discretization is critical for this analysis and a
suitable ‘skin’ of quadrilateral elements must be formed
on the outer surfaces of conductive regions [11]. After
sufficient time-steps for the starting transient to settle, a
steady-state waveform of current versus time may be
extracted from the solution files, along with the driving
voltage. This analysis includes the effect of eddy
currents induced in the pole and the aluminium ring;
these may be seen in Figure 27 and Figure 26. The
effect of the eddy currents is to produce a field opposing
the voice coil flux. This has the effect of reducing the
voice coil impedance. Figure 28 Current waveform for Loudspeaker 2 at 31.25Hz, x=
-5.75mm.

Subsequently the voice coil’s ‘blocked’ electrical

impedance may be calculated by applying Ohm’s law to
the fundamental components of the waveforms. Each
loudspeaker was solved for seven frequencies in five
positions. The positions were chosen to be coincident
with the measurement positions.

4.3. FEM results

Figure 29 shows the FE calculated impedance compared

Figure 26 Loudspeaker 1. Lines of constant power density at 31.25Hz,
x=4.
to the measured data for loudspeaker 1 with the coil in
the rest position. At low frequencies the small errors in
the modelled mechanical impedance cause some
significant artefacts in the measured data. At higher
frequencies the agreement is very good.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 12 of 19
Dodd et al. Voice coil impedance

! " #$ % & "$ ) # ! "#$ & "$

'

) #

'

Figure 29 Example FE impedance calculation (dotted) measured data (

(solid). x=0 loudspeaker 1.

At around 1kHz the measured value is consistently

below the FE value. To determine the cause of this
discrepancy a conventional impedance measurement of
loudspeaker 2 was made. The loudspeaker voice coil
was then glued in position and a direct measurement
made of the blocked impedance. The result in Figure 30 Figure 31 upper graph FEM impedance magnitude loudspeaker 1.
clearly shows that at 1kHz the blocked impedance is Lower graph measured data of loudspeaker 1.
higher than the free impedance. This curious difference ) # ! "#$ & "$

is thought to be an artefact resulting from the motional '

impedance due to non-pistonic diaphragm motion. (

! " #$ & " $

) # "#$

'

Figure 30 Conventional impedance measurement (solid) directly

measured blocked impedance (dashed) loudspeaker 2.

Figure 32 upper graph FEM impedance magnitude of loudspeaker 2.

The impedance results are illustrated opposite together
with the equivalent data from the measured values using
the method described in Impedance versus displacement
section. To allow the large number of results to be
compared contour plots have been used.
Lower graph measured data of loudspeaker 2.
It is evident that in both cases the FEM yields the same
trend of impedance variation as the measurements. The
measured data exhibits some noise around the
fundamental resonance as mentioned in Excess
impedance. There is also evidence of non-pistonic
modal behaviour which has resulted in non-inductive
impedance variations.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 13 of 19
Dodd et al.
) # ! & "$
) # ! & %
Figure 33 Loudspeaker 2 upper graph real part of FEM impedance.
lower graph imaginary part of FEM impedance.
Figure 33 shows the real and imaginary parts of the
FEM derived impedance for loudspeaker 2. The
difference in symmetry of the impedance mentioned in
Non-linear Parameters can be clearly seen.
5. PREDICTION OF DISTORTION
5.1. Lumped parameter-based prediction
If non-linear parameters of the equivalent circuit in
Figure 1 are measured then the state signal (current,
displacement) and the sound pressure output can be
predicted for any excitation signal. This technique is
useful for several applications: diagnostic testing of a
loudspeaker; assessment or improvement of a new
design; auralization of large signal performance with
music or other test signals [1]. The Simulation module
(SIM) [13] has been used to predict the distortion based
upon the measured non-linear parameters Le(x), R2(x)
and L2(x) above.
The results of these simulations may be directly
compared with the results of the 3D Distortion
measurement module (DIS) [14] which provide a direct
measure of the actual distortion of the loudspeaker. In
this way we are able to validate the measured non-linear
parameters of the LR-2 model by comparing predicted
and measured distortion generated in the input current.
Measurement of distortion in the electrical input current
reveals the distortion generated by the varying input
AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Voice coil impedance
impedance almost directly (provided the loudspeaker is
connected to an amplifier with a suitably low output
impedance). The distortion generated in the current will
also appear in the displacement, velocity and sound
pressure output. However, the distortion generated by
displacement varying force factor Bl(x) and compliance
CMS(x) appear in the current only close to the resonance
where the velocity and back EMF are high. Doppler
distortion and any other radiation distortion will not be
detected in the input current.
& "$
A very important aspect is the selection of the test
stimulus. A single tone reveals only harmonic distortion
which is, for displacement varying input impedance
ZL(jω,x), relatively low. This is because at low
frequencies the variation of ZL(jω,x) with displacement
x is relatively small, seen in Figure 10 and
Figure 18 for both loudspeakers. In addition at high
frequencies the voice coil displacement becomes very
small and the variation of ZL(jω,x) with displacement is
minimal.
However, a two-tone signal comprising a bass tone at
frequency f2 and a probe tone at higher frequency f1 is a
much more revealing signal [15]. The bass tone is set to
a fixed frequency below resonance to produce a large
displacement of ~5mm peak. The second ‘probe’ tone is
varied from 200Hz to 18 kHz to represent any audio
signal in the pass-band of the loudspeaker. The varying
impedance ZL generates not only harmonics of both
tones but additionally difference and summed-tone
intermodulation components, which may exceed the
harmonics significantly.
There is a simple relationship between shape of a non-
linear parameter and the order of the resulting distortion
component. A distinct asymmetry of the parameter,
such as in Figure 16 and Figure 17, will generate
dominant 2ndorder distortion, which will outweigh
3rdorder and higher distortion. Conversely a symmetrical
curve, such as in L2(x) and R2(x) in Figure 24 and
Figure 25, will generate strong 3rd-order and other odd-
order distortion components.
5.2. FEM prediction
The FE provides an alternative means of predicting the
distortion. Using the FEM software a full kinematic
solution for an arbitrary input signal may be computed.
This solution can again be used to predict the distortion,
this time from FE models adapted from those developed
in order to model the voice coil impedance. For the
purpose of this paper, this type of FEM solution appears
Page 14 of 19
Dodd et al. Voice coil impedance

for only one frequency; the solution time is very long 5.3. Results
and use of this type of modelling is thus restrictive.
The two-tone intermodulation stimulus must be solved 5.3.1. Lumped Parameter Model results
with a small time step defined by the higher frequency
over two periods of the lower frequency, provided that 20.0
the frequencies are integral multiples. It is estimated
predicted
that solving with a two-tone intermodulation stimulus
17.5 measured
for the seven octave spaced frequencies up to 2kHz
would take 180 hours of processor time on a 2GHz PC.
15.0
As we have seen, the results of the FE correlate well

[Percent]
with measured ZL(jω,x). The use of the measured non- 12.5
linear parameters to calculate the distortion is dependent
upon the LR-2 fitting and the assumption that the model 10.0
is adequate to describe the behaviour of the system.
Additionally, as previously discussed, the measurement 7.5
method (quasi-static) may have a bearing upon the
deduction of the LR-2 parameters and indeed the
5.0
measured ZL(jω,x). The full kinematic analysis does not
have these limitations as it returns directly to the 103 104
Frequency f1 [Hz]
fundamental physical relationships in order to calculate
Figure 34 Second –order intermodulation distortion in the input
the system output. It is also able to account for more current measured and predicted by using the lumped parameter
complex phenomena such as the effect of current method (SIM) for loudspeaker 1
magnitude on the impedance response. The FE allows
application of specific laws of physics to model the 4.0
behaviour whereas the lumped parameter model predicted
matches specific effects in such a way that the resulting 3.5
non-linear system of equations behaves in a closely
3.0
similar way to the loudspeakers measured behaviour.
The FE method used here is also able to represent other measured
2.5
[Percent]

non-linear relationships using user prescribed power

series. This facility could be used to model the Cms(x) 2.0
non-linearity for example.
1.5
It would be possible to use the ZL(jω,x) results from
FEM to determine the LR2 parameters and predict the 1.0
distortion with the SIM module. This has not been
done here since the results would be derived from 0.5
almost identical data. In practice the time saving of this
103 104
method is substantial and it is anticipated that further Frequency f1 [Hz]
work will be done using this method. Where further
Figure 35 Third –order intermodulation distortion in the input current
understanding of the physics is an aim the detailed measured and predicted by using the lumped parameter method (SIM)
results of a direct FEM approach are likely to outweigh for loudspeaker 1
the time cost.
Figure 34 and Figure 35 show the measured and
predicted 2nd-order and 3rd-order distortion for
loudspeaker 1. The 2nd-order distortion is dominant and
is caused by asymmetry of the parameters Le(x), L2(x)
and R2(x).

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 15 of 19
Dodd et al. Voice coil impedance

22.5 5.3.2. FEM results

$* +,

20.0
Measured
17.5

15.0
[Percent]

12.5

10.0
predicted
7.5 (

5.0
'

2.5
Figure 38 Acceleration response spectra of kinematic model with two-
103 104 tone excitation at 12.5Hz & 2kHz. Level of 2kHz 2.83v, 12.5Hz level
Frequency f1 [Hz] adjusted to give same excursion as lumped parameter model.
Figure 36 Second–order intermodulation distortion in the input current Loudspeaker 2.
and predicted by using the lumped parameter method (SIM) for
loudspeaker 2 (ring below) The discretized geometry of loudspeaker 2 was solved
using a full kinematic solution with a two tone input
4.0 signal in order to reveal intermodulation distortion
measured components as previously discussed. The model was
3.5
excited with an input signal comprising a 12.5 Hz tone
and a 2Khz tone.
3.0 predicted
Figure 38 shows, by means of an FFT, the acceleration
[Percent]

2.5 response spectra of the model to this signal. The two

excitation tones may be clearly seen on the spectra at
2.0 12.5Hz and 2kHz. In addition, due to the non-linearities
of the model, distortion artefacts have been produced.
1.5
These artefacts are both harmonic products of the two
tones, occurring at multiples of the two base
1.0
frequencies, and intermodulation products resulting
from the interaction of the two excitation signals,
103 104 occurring at frequencies of the sum and difference
Frequency f1 [Hz] between the base frequencies and their multiples.
$* + ,
Figure 37 Third–order intermodulation distortion in the input current
measured and predicted by using the lumped parameter method (SIM)
for loudspeaker 2 (ring below)

Loudspeaker 2, with shorting ring below the gap, gives

a slightly better performance than loudspeaker 1 for
frequencies below 1 kHz. Here the 2nd-order distortion
components are smaller but still dominant as the
asymmetry remains in the non-linear parameters. At
high frequencies the loudspeaker 2 generates similar (

distortion levels to loudspeaker 1 which corresponds

with the increasing asymmetry of the effective '

inductance Leff(x,jω) in Figure 22.

Figure 39 Acceleration response spectra of kinematic model with two-
Note that the distortion increases with frequency tone excitation at 12.5Hz & 2kHz – detail of 2kHz region.
similarly to the increase in the variation of the Loudspeaker 2.
impedance ZL with x shown in Figure 10.
An enlarged view of the spectra is shown in Figure 39.
This view clearly shows that the product visible at 4kHz

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 16 of 19
Dodd et al.
is composed from three distinct peaks. These peaks
would appear to be the results of additional distortion
products as well as the intuitive second harmonic of the
2kHz tone.
# A comparison of the FEM predicted levels of distortion
Figure 40 Current flow in aluminium ring inside magnet, excited with
two-tone signal at 12.5Hz & 2kHz. Loudspeaker 2.
The FEM solution allows generation of additional data
such as the current flow in the aluminium ring as shown
in Figure 40. It is difficult to think how else this type of
detailed information may be generated. FEM is
excellent at providing data for situations that would be
extremely difficult to measure or otherwise predict.
A number of observations may be made from Figure 40.
Firstly the induced current in the aluminium ring is
predominantly at high frequencies, the eddy currents &
skin depth increase in severity with frequency as their
induction is proportional to current flux. Additionally
the modulation of the current with voice coil position is
clearly shown.
#
Figure 41 Current flow in poles, excited with two-tone signal at
12.5Hz & 2kHz. Loudspeaker 2.
Figure 41 shows a similar result for the current in the
pole. In this case a far greater proportion of low
frequency current flows. The overall level of the current
is significantly higher than that of the aluminium ring
even though the resistivity of steel is higher. This is
AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 17 of 19
Voice coil impedance
thought to be a consequence of the close proximity of
coil and poles as well as the high magnetic permeability
in the poles.
and the measured and lumped parameter predicted
levels of distortion has not been performed. The type of
analysis performed here is extremely time consuming
both in terms of model setup and solution. The data
generated & shown in the previous figures is not
sufficient in quantity to provide accurate & comparable
values for the relative levels of the intemodulation
products.
5.4. Discussion of modelling
Throughout this paper FE and lumped parameter models
have been applied to model static voice coil impedance
at a constant temperature. This leaves only the
magnetic and electrical domains to determine the coil
impedance. Helpfully FE considering both these two
domains is available. All that is required is the
appropriate material properties and geometry.
The FE path followed to its logical conclusion would
provide results equivalent to measurements. A complete
FEM approach to modelling the full behaviour of a
loudspeaker would require a large number of physical
domains to be considered; mechanical, acoustic,
magnetic, electrical & thermal. Solving in the time
domain allows non-linear effects to occur. Fully
applied, this approach could result in the ultimate
simulation of reality that we are currently able to
conceive. However, coupling FEM models results in a
large increase in computation time compared to solving
the models separately; this is further exacerbated by the
long thermal time-constants found in loudspeakers. The
resulting FE model would be impracticably slow to
solve even in 2D and would not give the engineer the
necessary data to improve the design.
Even with the relatively simple case considered in this
paper the disadvantages of an exclusively FE method
are clear. Calculation of the intermodulation distortion
using the SIM module can be performed in ~30 seconds,
faster than an equivalent measurement, the kinematic
FE solution for the seven frequencies used for the
ZL(jω,x) modelling is estimated to take in the region of
100 hours solving time on an equivalent PC. The
lumped parameter method is faster by factor of around
10,000. Applying Moores’ law [16], the FEM
computation is likely to be able to match the current
lumped parameter computation speed in approximately
35 years.
Dodd et al.
With lumped parameter models the question, ' which
physical domains must we consider?' , is replaced by a
requirement to understand the physics involved so that
appropriate simplifications may be made. This allows
development of a simplified system with appropriate
parameters for representation of an effect. The system
and parameters may be defined in numerous ways to
suit the users requirements in order to provide the
required accuracy, complexity, practicality etc.
The ability of FE to allow display of the fields
concerned may well prove enlightening to the engineer.
However, while FE provides extensive information, the
overall principles by which a loudspeaker operates are
best illustrated using lumped parameter models. Using
FE to produce data for a lumped parameter model
combines these viewpoints to give an engineer the most
enlightening information in a reasonable time. A
number of FE models may be used to provide values for
the various model parameters. Furthermore these FE
results show how the fields concerned change for the
parameter modelled. Even without modelling the result
of the non-linearities such as the intermodulation
distortion the FEM ZL(jω,x) is extremely useful to the
engineer. Another very significant result if that if one
aspect of the design is altered only the appropriate
model need be resolved.
As shown here for the voice coil impedance, with
careful consideration and investigation, the most
important information can be compressed into a few
meaningful parameters and an appropriate lumped
model. The benefits of an integrated modelling
approach, such as that described in [4], cannot be
emphasised too strongly. This approach becomes even
more appropriate when nonlinearities are to be
modelled.
6. CONCLUSION
Detailed knowledge of the voice coil inductive
behaviour is essential for the design of loudspeakers
with minimal distortion.
In this paper, the variation of voice coil impedance with
position and frequency has been both measured, using a
new quasi static technique, and modelled, using FE,
with exceptional agreement. It is possible to use this
information to investigate the effect of the shorting rings
and to derive other indications for improvements.
The measured or modelled voice coil impedance may be
expressed in terms of a few meaningful parameters
using Wright, Leach or LR-2 models. The LR-2 model
allows the expression of the variation with displacement
AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 18 of 19
Voice coil impedance
in a form that may be incorporated in a non-linear
lumped parameter model. An extended model with more
RL cascades (LR-3) can provide a better fit when
required. Loudspeaker 2 impedance results show that
all LR-2 parameters must vary independently with
position to satisfactorily describe the non-linear
behaviour of some loudspeakers over the full frequency
range. A non-linear lumped parameter model using the
new LR-2 model has been successfully used to predict
the intermodulation distortion of a loudspeaker.
Using modern (full kinematic FE) numerical methods it
is beginning to become possible to deduce the final
behaviour of a loudspeaker (linear and non-linear) based
upon the geometry and the material properties. FE
allows additional data and visualisations to be computed
that it are not possible to deduce using measurement or
other simulations. Most usefully numerical models may
also be applied to conceptual loudspeakers. However,
describing the loudspeaker using one numerical model
requires massive computation, furthermore the results
will inevitably be complex and difficult to interpret.
This paper has demonstrated the equivalency of
measurement and appropriate careful numerical
modelling. The results of the two techniques can both be
used to generate non-linear lumped parameter models.
An alternate route is to simply use the kinematic FE
method to model the full case and miss out the
intermediate impedance results. This alternative is,
surprisingly, a less powerful technique as it is greatly
less efficient and does not provide the highly insightful
ZL(jω,x) results. After further consideration it is
believed that similar findings will emerge with the
increasing application of numerical techniques to
loudspeaker design. An approach for the use of FE and
other numerical techniques is presented. Using
numerical methods to provide data for lumped
parameter models the geometry & performance may be
linked in an efficient way which conveys the maximum
understanding.
6.1. Acknowledgements
Many thanks to the Klippel GmbH & GPAcoustics
engineering teams for their help and support for this
paper. Thanks to Joerg Panzer for use of his
visualisation program VACS.
7. REFERENCES
[1] W. Klippel, “Prediction of Speaker Performance at
High Amplitudes”, Presented at the 111th
Dodd et al. Voice coil impedance

Convention of the Audio Engineering Society, [14] “3D Distortion Measurement (DIS)”, Specification
preprint 5418, November 2001. J. Audio Eng. Soc., of the KLIPPEL Analyzer module, Klippel GmbH,
Vol. 49, No. 12, p. 1216, December 2001. www.klippel.de, 2003.
[2] A. Voishvillo, V. Mazin, “Finite-Element Method [15] W. Klippel, “Assessment of voice-coil peak
of Modeling of Eddy Currents and Their Influence displacement Xmax”, J. Audio Eng. Soc., Vol. 51,
on Nonlinear Distortion in Electrodynamic No. 5, pp. 307-324, 2003.
Loudspeakers”, Presented at the 99th Convention of
[16] G. Moore “Cramming more components onto
the Audio Engineering Society, preprint 4085,
integrated circuits,” Electronics, Vol. 38, No. 8,
September 1995.
April 19, 1965.
[3] J.Vanderkooy, “A Model of Loudspeaker Driver
Impedance Incorporating Eddy Currents in the Pole
Structure” J. Audio Eng. Soc., Vol. 37 No 3 pp.
119-128; March 1989.
[4] M. Dodd, “The Development of a Forward
Radiating Compression Driver by the Application
of Acoustic, Magnetic and Thermal Finite Element
Methods,” Presented at the 115th Convention of the
Audio Engineering Society, preprint 5886,
September 2003.
[5] “Manual of the KLIPPEL Analyzer System”,
Klippel GmbH, www.klippel.de, 2004.
[6] W.M. Leach, “Loudspeaker voice-coil inductance
losses: circuit models, parameter estimation, and
effect on frequency response”, J. Audio Eng. Soc.,
Vol. 50, No. 6, 2002.
[7] J.R. Wright, “An empirical model for loudspeaker
motor impedance”, J. Audio Eng. Soc., Vol. 38,
No. 10, 1990.
[8] “Large Signal Identification (LSI)”, Specification
of the KLIPPEL Analyzer module, Klippel GmbH,
www.klippel.de, 2003.
[9] “Maximizing LPM Accuracy”, Application note
AN 25 of the KLIPPEL Analyzer System, Klippel
GmbH, www.klippel.de, 2004.
[10] “Processing Software (MAT)”, Specification of the
KLIPPEL Analyzer module, Klippel GmbH,
www.klippel.de, 2003.
[11] M.A. Dodd, “The Transient Magnetic Behaviour of
Loudspeaker Motors” Presented at the 111th
Convention of the Audio Engineering Society,
preprint 5410, November 2001.
[12] FLUX2D 7.60 User Manual, CEDRAT, Meylan
France, 2002.
[13] “Simulation (SIM)”, Specification of the KLIPPEL
Analyzer module, Klippel GmbH, www.klippel.de,
2003.

AES 117th Convention, San Francisco, CA, USA, 2004 October 28–31
Page 19 of 19