Bridge hill/Woodhouse 1 May 25, 2004

On the bridge hill of the violin

J. Woodhouse
Cambridge University Engineering Department,
Trumpington St, Cambridge CB2 1PZ, U.K.
jw12@eng.cam.ac.uk
Summary
Many excellent violins show a broad peak of response in the vicinity of 2.5kHz, a feature
which has been called the bridge hill. It is demonstrated using simplified theoretical models
that this feature arises from a combination of an in-plane resonance of the bridge and an
averaged version of the response of the violin body at the bridge-foot positions. Using a
technique from statistical vibration analysis, it is possible to extract the skeleton of the
bridge hill in a very clear form. Parameter studies are then presented which reveal how the
bridge hill is affected, in some cases with great sensitivity, by the properties of the bridge and
body. The results seem to account for behaviour seen in earlier experimental studies, and they
have direct relevance to violin makers for guiding the adjustment of bridges to achieve desired
tonal quality.
PACS No. 43.75.De
1. Introduction
The characteristic high bridge of the violin and cello, and indeed of most bowed-string
instruments, presumably developed initially for ergonomic reasons. In order to provide the
range of angles needed to bow each string individually, the strings must be raised clear of the
instrument body. Also, the high bridge plays an essential role in rotating the transverse
force from the vibrating string into normal forces applied to the instrument body through the
bridge feet, which can then excite bending vibration of the body [1]. Compared to the low,
robust bridges of the piano or guitar, the violin bridge may seem to be a necessary evil: it is
fragile and requires regular attention to keep it straight and properly fitted. However, research
of recent years shows that this type of bridge has provided, perhaps somewhat fortuitously, a
crucial means for adjustment of the string-to-body impedance characteristics which has
allowed the violin family to acquire its familiar loudness and tonal colouration.
The oscillating force provided by the vibrating string can only excite vibration of the
instrument body by first passing through the bridge. The bridge thus acts as a filter, and it is
no surprise that the material properties and geometric configuration of the bridge can have a
significant influence on the sound of an instrument. The first systematic study of the
transmission properties of the violin bridge was made by Reinicke and Cremer [1,2]. They
showed that a normal violin bridge has internal resonances within the frequency range of
interest for the sound of the instrument, so that the filtering effect of the bridge has very
significant variation with frequency. The lowest bridge resonance is usually found around
3kHz when the bridge feet are held rigidly (for example in a vice), and the motion consists of
side-to-side rocking of the top portion of the bridge as sketched in Fig. 1.
Since the work of Reinicke and Cremer, several experimental studies have been carried out
which relate to the influence of this lowest bridge resonance. First, measurements have been
made by Dnnwald [3] and Jansson [4,5] of the frequency response of a wide variety of
violins. Both authors found that violins of high market value showed a strong tendency to
exhibit a broad peak of response in the vicinity of 23kHz, in a feature originally named the
bridge hill by Jansson. The name was given because he, and indeed Cremer [1], attributed
this feature to the filtering effect of the lowest bridge resonance just described. A typical
Bridge hill/Woodhouse 2 May 25, 2004
example is shown in Fig. 2: the plot shows the input admittance (velocity response to unit
force amplitude) in the direction of bowing, measured at the string position on a violin bridge
using a method similar to that described by Jansson [4]. The characteristics of the bridge hill
are seen clearly in Fig. 2: a rise in amplitude around the peak frequency of the hill, followed
by a steady drop in amplitude at higher frequencies, and over the same frequency range a
downward trend in phase towards 90.
The second relevant set of experimental studies is the extensive series reported by Jansson
and co-workers [6,7,8] in which various structural modifications were made to the bridge
and/or the top plate of the violin, and the effect on the input admittance measured. These
studies have shown that the appearance of a bridge hill, and its frequency, are not
determined by the bridge structure alone. In particular, they reveal that the feature can
sometimes still be seen when the standard bridge is replaced by a plate bridge with no cut-
outs, so that the vibration mode shown in Fig. 1 is no longer possible. Janssons
measurements also showed a significant influence arising from the spacing of the bridge feet
[7] and the stiffness/mass of the top plate in the vicinity of the bridge [6]. These results led
Jansson to become less satisfied by the name bridge hill, and he has advocated bridge-body
hill, or just hill [6]. However, it will be argued that the bridge is still the central defining
element in the phenomenon, and the original name bridge hill will be retained here.
The fact that the frequency of a bridge resonance will be significantly affected by the coupling
to the violin body was already stressed by Reinicke and Cremer. The results of Jansson have
fleshed this idea out with empirical data. The task of this paper is to explain the pattern of
behaviour and show how the frequency, height and shape of the bridge hill are influenced by
the constructional parameters of the bridge and violin. A first step has been taken by Beldie
[9], who has shown that a reasonable fit to Janssons results can be obtained if the behaviour
of the body beneath the bridge feet is approximated by simple springs. In particular, his idea
explains why there can still be a bridge hill with Janssons plate bridge without cut-outs: the
mode consists of side-to-side rocking motion of the entire bridge, with a restoring force
provided by the springs under the feet. However, Beldie gave no explanation for what
determines the stiffness of these effective springs, and more importantly, no-one appears to
have addressed the question of what determines the height and bandwidth of the bridge hill.
The approach here will be to explore the bridge hill using simplified theoretical models which
are violin-like but which remain simple enough to analyse without needing elaborate finite-
element calculations This simplification is convenient, but more importantly it can be
combined with recent developments in vibration theory to allow a calculation of the
underlying skeleton of the frequency response curve, not readily accessible from a detailed
prediction such as a finite-element model. Once it has been shown that the skeleton
calculation is reliable, this approach will allow attention to be focused on the parameters
which influence the hill without the distracting details of the individual body resonance
peaks. All the general features reported by Jansson can be demonstrated with these simplified
models. One particular conclusion of this study will be that the bridge hill observed by
Dnnwald [3] and Jansson [4,5] in many valuable instruments is probably not primarily a
property of the instruments as such, but a result of the fact that more care and attention has
been devoted to the fitting of appropriate bridges to these instruments. This suggestion is in
keeping with the experience of experts in violin set-up, who attribute great importance to
bridge adjustment.
2. Modelling the bridge and body
The amplitude and phase characteristics seen in Fig. 2 are precisely what would be expected
when a multi-modal system (the violin body) is driven through an intermediate system
exhibiting a resonance. The simplest example is shown schematically in Fig. 3. A force
Fe
it
is applied to a mass m, which in turn drives a system with input admittance Y
v
()
Bridge hill/Woodhouse 3 May 25, 2004
through a spring of stiffness k. (The subscript v indicates violin.) It is straightforward to
show that the input admittance of the combined system is
Y
b
()
kY
v
+ i
k m
2
+ ikmY
v
(1)
where the subscript b indicates bridge.
For a clear demonstration that this equation predicts behaviour similar to the observed bridge
hill, it is convenient to use a very idealised system Y
v
(): assume regularly spaced
resonances, all with the same amplitude and damping factor. Then Fig. 4a shows a plot of
Y
v
(), and Fig. 4b shows a typical plot of Y
b
(). Parameter values are given in the caption.
The dashed lines will be discussed in Section 3, and should be ignored for the moment. It is
immediately clear that Fig. 4b shows similarities to Fig. 2: the regular peaks of Y
v
() have
been modulated by a hill followed by a steady amplitude decline, while the phase plot
shows a downward trend towards 90. (Note that the phase of the input admittance cannot
go negative beyond 90, because the system is assumed to be dissipative at all frequencies,
rather than containing any energy source. This requirement is often used as a check on the
phase accuracy of measurements.)
This simple formulation is not quite sufficient to capture the behaviour of a violin bridge. It is
necessary to take into account that the bridge contacts the violin body at two points rather than
one. The natural model is the one put forward by Reinicke and Cremer [1,2], equivalent to
that sketched in Fig. 5. The portion of the bridge below the waist can be regarded as a rigid
body, coupled via a torsion spring to a mass-loaded rigid link representing the rotational
inertia of the upper part of the bridge. The vibrational force from the string drives this upper
mass transversely. This bridge model is parameterised by the foot spacing d, the length a of
the rotating link, the mass m and the torsional spring stiffness K. The resonant frequency of
the bridge with its feet rigidly clamped is then

b

1
a
K
m
. (2)
The admittance Y
b
() at the string notch, in other words at the mass m, is now to be
calculated in terms of these parameters, together with the 2 2 admittance matrix which
describes the properties of the violin body at the two foot positions. A useful intermediate
stage is to calculate the rotational admittance governing the motion of the lower part of the
bridge: this admittance R() is defined as the angular velocity of the bridge base when a
moment is applied to it of unit magnitude, oscillating sinusoidally at frequency . Assuming
that the mass of the bridge base is small enough to be neglected, it is easy to show that
R
Y
11
+ Y
22
2Y
12
d
2
(3)
where Y
jk
() denotes the velocity response of the violin body at the position of bridge foot k
to a harmonic force of unit amplitude applied at bridge foot j.
This admittance can also be written in terms of the modes of the violin body. The body
admittance matrix is given by the standard formula (see for example [10])
Bridge hill/Woodhouse 4 May 25, 2004
Y
jk
()
iu
n
(x
j
, y
j
)u
n
(x
k
, y
k
)

n
2
+ 2i
n

n

2
( ) n

(4)
where the nth mode of vibration of the body in the absence of the bridge has mode shape
u
n
(x, y) at position (x, y), natural frequency
n
and modal damping factor
n
(or
corresponding modal Q-factor Q
n
1 / 2
n
). The positions of the two bridge feet have
coordinates
(x
1
, y
1
), (x
2
, y
2
)
. The mode shapes are assumed to be normalised in the usual
way with respect to the system mass matrix or kinetic energy function [10]. Substituting in
eq. (3) yields
R()
i
d
2
u
n
(x
1
, y
1
) u
n
(x
2
, y
2
)
[ ]
2

n
2
+ 2i
n

n

2
( ) n

. (5)
In the limit d 0 this expression tends towards the point moment admittance of the plate,
involving the squared spatial derivative of the mode shapes.
Imposing moment balance around the torsion spring in the bridge model now yields an
expression for the input admittance at the top of the bridge which is analogous to eq. (1):
Y
b
()
KR + i ( )a
2
K ma
2

2
+ iKma
2
R
. (6)
The similarity is sufficiently close that bridge hill behaviour like that shown in Fig. 4 can
confidently be expected. To explore the consequences requires a more realistic choice of
model for the violin body than that used before, since to evaluate R requires mode shape
information.
There is only one vibration problem involving a two-dimensional bending plate which has a
simple closed-form solution, and it is natural to use this as a first approximation. This system
is a rectangular plate with hinged boundary conditions all around. If the plate has plan
dimensions L
1
L
2
, thickness h and density , then the normalised mode shapes ready for
substitution into eqs. (4) or (5) are
u
nm

2
L
1
L
2
h
sin
nx
L
1
sin
ny
L
2
(7)
where the term under the square root is simply the total mass of the plate, and x and y are
Cartesian coordinates in the plane of the plate, measured from one corner. The corresponding
natural frequencies satisfy

nm
2

h
2

D
1
n
L
1

_
,

4
+ D
3
m
L
2

_
,

4
+ D
2
+ D
4
( )
n
L
1

_
,

2
m
L
2

_
,

1
]
1
1
(8)
where D
1
D
4
are the elastic constants appropriate to a plate of orthotropic symmetry, as
discussed in detail by McIntyre and Woodhouse [11]. Suitable numerical values for these
elastic constants for instrument-like plates of spruce or maple, together with other numerical
values used in the violin models here, are given in Table 1.
Bridge hill/Woodhouse 5 May 25, 2004
Using this model, the rotational admittance R is plotted in Fig. 6 and the input admittance at
the bridge top is shown in Fig. 7. The dashed lines in these plots will be explained in the next
section. Figure 6 shows no very strong trend with frequency, somewhat similar to Fig. 4a,
while Fig. 7 shows a clear bridge hill. Also, the absolute level of the admittance in Fig. 7 is
similar to that seen in Fig. 2, confirming that this simple model has broadly violin-like
behaviour. The aspect of this model which is most obviously unrealistic is not immediately
apparent from these plots. Because the bridge has been placed symmetrically with respect
to the mid-line of the plate, many of the plate modes do not contribute to R. These modes have
motion that is symmetrical at the two bridge feet so that, from eq. (5), their contribution to R
is zero. In a real violin, the presence of the bass bar and soundpost destroy the symmetry of
the structure, so that potentially all modes could contribute to R.
It is not easy to incorporate a bass bar into the idealised model used here, but a representation
of the soundpost is quite simple to achieve. The model can be extended to include two
rectangular plates, representing the top and back of the violin. Both will have the same plan
geometry, and will have hinged boundaries along all edges. The two plates can then be
coupled together at a chosen point by a massless, rigid link representing the soundpost. Such
point-coupled systems are easily modelled using appropriate combinations of the admittances.
If the only requirement were the driving-point admittance Y
coup
of the coupled plates at the
soundpost position, it would be given simply by
1
Y
coup

1
Y
1
+
1
Y
2
(9)
where Y
1
, Y
2
are the input admittances of the two uncoupled systems at the same position.
This familiar formula expresses the fact that the displacements of the two coupled systems are
the same at the coupling point, while the total applied force is the sum of the forces applied to
the two separate subsystems.
A slightly more complicated version of this argument is needed in order to give the
admittances Y
jk
() relating to the positions of the bridge feet, neither of which is exactly at
the soundpost position. If the soundpost has coordinates (x
3
, y
3
) on both top and back plates,
then the required admittances can be obtained from the formula
Y
11
coup
Y
12
coup
Y
13
coup
Y
21
coup
Y
22
coup
Y
23
coup
Y
31
coup
Y
32
coup
Y
33
coup

1
]
1
1
1
1

Y
11
uncoup
Y
12
uncoup
Y
13
uncoup
Y
21
uncoup
Y
22
uncoup
Y
23
uncoup
Y
31
uncoup
Y
32
uncoup
Y
33
uncoup

1
]
1
1
1
1
+
0 0 0
0 0 0
0 0 1 / Y
33
back

1
]
1
1
1
(10)
where the superscripts coup and uncoup label the top-plate admittances from eq. (4) in the
coupled and uncoupled states respectively, in matrices relating to the three positions
(x
1
, y
1
), (x
2
, y
2
), (x
3
, y
3
)
. Y
33
back
is the input admittance at the soundpost position on the
uncoupled back plate. Equation (10) embodies the same physical argument as above, stating
that an impedance 1 / Y
33
back
has been added to the top plate at the soundpost position, while
nothing has been added at the positions of the bridge feet.
Using this extended model with the numerical values given in Table 1, the rotational
admittance is now as shown in Fig. 8 and the input admittance at the bridge top is as shown in
Fig. 9. These two figures are directly comparable with Figs. 6 and 7 respectively. Again, the
dashed lines in the plots will be discussed in the next section. Many more peaks are seen than
in the plots from the simpler model, both because the symmetry has been broken and because
Bridge hill/Woodhouse 6 May 25, 2004
the back plate has introduced additional resonances of its own. The trends are generally
similar, with a clear bridge hill seen in Fig. 9. Notice that the hill is rather higher and
narrower than in Fig. 7. This extended model will be used to investigate the influence on the
bridge hill of various parameters relating to the bridge and to the violin body.
It is interesting to compare Fig. 9 with the measured admittance in Fig. 2. This reveals that
the general levels are quite similar except near the hill, which is a little too prominent in this
particular simulation. However, it will be seen in section 4 that the parameter values of the
bridge model could readily be altered to achieve a closer match of the hill. The other obvious
difference between Figs. 2 and 9 is that the flat-plate violin has a higher density of
resonances at low frequencies than the real violin. The reason for this probably lies mainly in
the arched plates of the real violin: curved shells such as cylinders have low modal density
below the ring frequency [12], but at higher frequencies they tend towards the same modal
density as a flat plate of the same area. The ring frequency in Hz is given by c / 2R, where
R is the radius of the cylinder and c is the compressional wave speed. It is not influenced by
the thickness of the shell. There is no single ring frequency for the complex geometry of a
violin plate, but simple estimates based on typical axial or transverse radii of curvature of a
violin top, and the corresponding wave speeds of spruce, yield values of the order of 12kHz.
At frequencies of relevance to the bridge hill, the flat-plate model should have similar modal
density to the real violin. The model seems good enough that one might hope to obtain
plausible bridge-hill shapes with numerical values of bridge parameters close to those
measured from real bridges.
3. The response skeleton
Before looking at parameter studies, though, it is desirable to find a way to focus on the
underlying hill without the distracting details of the individual body modes. This can be done
readily for this simple model, by using an approach known in different guises as Skudrzyks
mean value method [13] or fuzzy structure theory [14]. Skudrzyks argument is the most
direct for the present purpose. The key insight is that, from eq. (4), the height of an isolated
modal peak is proportional to 1 /
n
, while the level at an antiresonance dip is proportional to

n
. It follows that the mean level of a logarithmic plot follows the geometric mean of these
two, and is thus independent of damping. If the damping were increased, the peaks and dips
would blur out and all admittance curves would tend towards smooth skeleton curves
representing the logarithmic mean of the original curves.
There is a physically appealing way to visualise the effect of increasing the damping. When a
force is applied at a point on the structure, it generates a direct field consisting of outward-
travelling waves. In time these will reflect from the various boundaries and return. Modal
peaks will occur at frequencies where the reflections combine in phase-coherent ways.
Antiresonances occur when the sum of reflected waves systematically cancels the original
direct field. But at an average frequency, where neither of these coherent phase effects
occurs, the reflected waves from the various boundaries tend to arrive in random phases and
to cancel each other out, leaving the direct field to dominate the response. If damping is
increased, the influence of reflections decreases. In the limit of high damping, the desired
skeleton of the admittance is given by the direct field alone.
The effect is thus the same as if the plate boundaries had been pushed further away until the
system becomes infinitely large. This gives a simple recipe to find skeleton curves for the
models discussed above: the rectangular top and back plates are replaced by infinite plates
with the same material properties and thicknesses. The vibration of a point-driven infinite
plate has a simple closed-form solution. For a plate of isotropic material of density ,
Youngs modulus E, Poissons ratio and thickness h the driving-point admittance is
Bridge hill/Woodhouse 7 May 25, 2004
Y

(, 0)
1
4h
2
3(1
2
)
E
, (11)
which represents a pure resistance: it is real and independent of frequency. For response at a
point a distance r from the driven point, the transfer admittance is
Y

(, r)
1
4h
2
3(1
2
)
E
H
0
(2)
(kr) H
0
(2)
(ikr)
[ ]
(12)
where the wavenumber k is given by
k
4

12(1
2
)
2
Eh
2
(13)
and H
0
(2)
is the Hankel function of the second kind of order zero [15].
For a plate of orthotropic material like wood the behaviour is somewhat more complicated,
but a standard approximation is sufficiently good for the present purpose. If it is assumed that
D
2
+ D
4
2 D
1
D
3
(14)
then the plate is equivalent to an isotropic one provided distance x along the grain is scaled
relative to distance y across the grain according to
x x D
3
/ D
1
( )
1/ 4
. (15)
The equivalent result to eq. (12) is then
Y

()
1
8h
2
D
1
D
3
H
0
(2)
(kr) H
0
(2)
(ikr)
[ ]
(16)
where
k
4

2
D
3
h
2
(17)
and r is calculated using x, y ( ). To illustrate the behaviour predicted by eq. (16), Fig. 10
shows the real and imaginary parts of Y

for the parameters of the spruce plate used in the

models in the previous section, and a value of r equal to the bridge-foot spacing assumed
there.
The calculations of the previous section can be readily repeated using these infinite-plate
admittances in place of the finite-plate results used before: in particular, eq. (10) can still be
used to give a model which allows for the soundpost, but treats both the top and back plates as
infinite in extent. The results are shown as dashed lines in Figs. 4, 69. It is clear that these
dashed lines do indeed follow accurately the mean trend of the logarithmic amplitude and the
phase, and also that they reveal the form of the bridge hill in Figs. 4b, 7 and 9. The skeleton
curves based on this infinite-plate modelling have been found to show similar accuracy over a
Bridge hill/Woodhouse 8 May 25, 2004
wide range of parameter values. It seems clear that, to study the bridge hill, it will be
sufficient to study these simple skeleton curves.
Some conclusions can be drawn immediately. The skeleton of the input admittance at the
bridge is given by eq. (6) with R replaced by the skeleton value R

. It is plausible, and
confirmed by the dashed lines in Figs. 6 and 8, that this value varies only slowly with
frequency. Substituting a constant value appropriate to the general vicinity of the bridge hill,
eq. (6) shows that the form of the hill is determined by complex poles which are the roots of

2
iR

K
b
2
0. (18)
If R

is real, this takes the familiar form of a damped harmonic oscillator which represents
the bridge resonance damped by radiation through the feet into the infinite plate system. The
effective damping factor
b
of this hill oscillator can be written in several equivalent forms:

b
R

K / 2
b
R

ma
2

b
/ 2 R

a Km / 2. (19)
If
b
is small the bridge hill will appear as a tall, narrow peak close to the clamped bridge
frequency
b
. As
b
increases, the bridge hill moves down somewhat in frequency and the
peak becomes broader and lower. If
b
reaches unity the bridge hill becomes critically
damped so that it ceases to be visible as a hill and simply becomes a low-pass filter.
Equation (19) thus shows how the various model parameters combine to determine the height
and bandwidth of the bridge hill, an aspect which has not been much discussed in the
literature of the subject.
If R

is complex the interpretation of eq. (18) is a little less obvious. An estimate of the
damping factor of the hill can probably be obtained from eq. (19) by replacing R

with
Re R

( ), but the imaginary part of

R

will contribute a reactive effect which will shift the

frequency of the hill away from the clamped frequency
b
. This is the physical origin of the
effect observed by Jansson [6,8], and idealised by Beldie in terms of effective springs beneath
the bridge feet [9]. Indeed, it is simple to obtain from eq. (18) an estimate of the hill
frequency when the flexible bridge is replaced with a rigid plate bridge as in Janssons
experiments [6]. Rewriting eq. (18) in the form

2
iR

K K / (ma
2
) 0 (20)
then allowing K , it is clear that the first term of eq. (20) becomes negligible in the
frequency range of interest, and the effective pole frequency is given by the other two terms as
i / ma
2
R
( )
. (21)
This expression clearly shows the hill frequency arising under these limiting conditions from a
balance between (rotational) stiffness provided by the violin body, and inertia provided by the
bridge.
To look a little more closely at the behaviour when R

is complex, the flat-plate model can be

used to suggest appropriate numerical values to explore. Figure 11 shows the real and
imaginary parts of R

corresponding to Fig. 8. For a bridge hill in the vicinity of 2.5kHz, it

seems reasonable to take Re R

( ) 100 rad m
-1
s
-1
N
-1
and to explore a range of Im R

( ) of
the same order of magnitude. Figure 12 shows the resulting variation of hill frequency and
Bridge hill/Woodhouse 9 May 25, 2004
loss factor, obtained by solving eq. (18) exactly. The frequency can be raised or lowered
relative to the clamped frequency (3kHz here), depending on the sign of Im R

( ): there
seems to be no physical reason why it should necessarily be positive (and hence spring-
like). The loss factor turns out to be influenced very little by Im R

( ): it remains close to the

value given by eq. (19) with R

replaced by Re R

( ), shown as the dashed line.

4. Parameter studies
4.1. Parameters of the bridge
The flat-plate model does not by any means represent all relevant features of the violin body,
but it contains enough detail that considerable insight of a semi-quantitative nature can be
gained by varying the parameters of the model. Sets of skeleton curves will be shown to
illustrate how the bridge hill is affected by various changes which have direct analogues in
violin-making practice. First, parameters relating to the bridge will be tested. Figure 13
shows the effect of varying the mass and stiffness of the bridge in such a way that the ratio,
and hence the clamped resonance frequency, is kept constant. The behaviour is as anticipated
from eq. (19): the main effect is that the damping factor of the hill resonance varies over a
wide range with relatively small changes in the mass and stiffness. The peak frequency of the
hill remains close to the clamped frequency throughout: for these particular parameter values,
as seen in Fig. 11, the reactive component of R

is small in the relevant frequency range.

From the perspective of a violin maker, this figure shows the effect of adjusting the bridge by
thinning the entire structure: the mass and the stiffness will both vary in proportion to the
thickness, and the ratio will remain constant. Such adjustment seems to give a very direct
way to change the height and bandwidth of the bridge hill.
Figures 14 and 15 show the effect of varying the bridge-top mass keeping the stiffness
constant, and conversely varying the stiffness keeping the mass constant. Both correspond to
realistic bridge adjustment procedures: the mass is determined by the thickness of the top
portion of the bridge (and by the bridge height and choice of wood), while the stiffness can be
varied independently by trimming around the cutouts. In addition, Fig. 14 illustrates the effect
of adding a mute to the violin bridge: the range of masses explored is sufficiently wide to
cover variations of bridge shaping and typical mutes. As anticipated, both figures show a
combination of varying hill frequency and bandwidth. Taken in combination with Figure 11,
it would appear that by judicious bridge adjustment the hill frequency and bandwidth could
both be placed wherever required on this simplified violin body. On a violin in which R

showed a bigger reactive component there might be limits on the range over which the
frequency and bandwidth could be adjusted (as shown by Janssons test with the plate bridge
[6]).
Finally, Jansson has shown an interesting series of measurements using bridges with different
foot spacings d [7]. This experiment is simulated, approximately, in Fig. 16. Results are seen
which qualitatively mirror the experimental findings. Decreasing the foot spacing reduces the
peak frequency of the hill, by changing R

. Interestingly, for this model at least, the hill

bandwidth varies at the same time in a non-obvious way. As the foot spacing is reduced from
its normal value, the bandwidth increases to a maximum with a foot spacing around 20mm,
then decreases again when the spacing is reduced further. Note that the clamped bridge
resonance remains at 3kHz throughout this series of simulations, so that the shift in hill
frequency is entirely due to changes in the reactive component of R

.
Bridge hill/Woodhouse 10 May 25, 2004
4.2. Parameters of the violin body
In a similar way, the model can be used to explore changes in the parameters of the violin
body, keeping the bridge unchanged. Figures 17 and 18 show the influence of the thickness
of the top and back plates respectively. The set of thicknesses tested were the same in both
cases, 24mm: in reality, a violin top plate would usually have a thickness in the bridge
region towards the lower end of this range [16], while the back plate near the soundpost
position would have a thickness at the upper end of the range. The results show similar tends
in both cases, but not surprisingly the hill is affected much more sensitively by the top
thickness than by the back thickness. A thinner top, and to a lesser extent a thinner back, has
the effect of reducing the hill frequency and increasing its bandwidth. It also has a significant
effect on the level of the skeleton curve at low frequencies, but it is an open question whether
this aspect of the results carries over to a real violin body with its arched plates, since the
frequencies in question are low enough that one would expect curvature to matter.
Finally, in Fig. 19 the effect of soundpost position is explored. The soundpost is moved along
a line behind the treble foot of the bridge, parallel to the grain of the top plate. The distance
between bridge foot and soundpost is varied in the range 530mm. Ordinarily, the soundpost
position would be near the lower end of this range. The results show only a rather slight
influence on the bridge hill, one which could easily be compensated by small adjustments to
the bridge. The well-known sensitivity of the sound of a violin to the position of the
soundpost does not seem to be associated to any great degree with changes to the bridge hill,
at least within this simplified model.
5. Conclusions and implications for violin makers
It has been shown that Reinickes model for the deformation of a violin bridge in its lowest in-
plane resonance [1,2] can be combined with a very simple model of violin body vibration to
give a system which can elucidate the various published measurements relating to the bridge
hill in the input admittance of a violin [6,7,8]. By replacing the finite plates in the body model
by infinite plates, the calculation can directly yield the skeleton curve which underlies the
bridge hill. Using the skeleton curve, the frequency and bandwidth of the hill were shown to
vary in a simple and predictable way with the parameters which determine the behaviour of
the bridge and the violin body model. There is obvious scope for experiments to test the
predictions of this study by making controlled adjustments to bridges and measuring the effect
on the input admittance and the bridge hill.
The results of this study are of direct interest to violin makers. The bridge model, and the
formula (19) for the effective loss factor of the hill, is quite robust. Although the body model
used here was highly schematic, the general conclusions about the effect of bridge adjustment
on the frequency, height and bandwidth of the hill should carry over directly to the behaviour
of a bridge on a real violin. If a normal bridge hill is desired (and one should not forget that
the evidence for the significance of the hill comes only from correlation studies, not from
psychoacoustical tests), then it should always be possible to create one on any reasonably
conventional violin by suitable bridge adjustment.
Some instruments place constraints on the potential for shaping the hill by bridge adjustment,
because the reactive contribution from the moment admittance of the body ( R

) has a strong
effect. Janssons test instrument seems to be like this, because it showed a fairly normal hill
even with the plate bridge with no cutouts [6]. But Dnnwalds data for master instruments
(see Fig. 3 of ref. [3]) suggests that many high-quality instruments are not like this: his plot
extends up to 7kHz with little obvious sign of the low-pass filtering effect of the bridge hill,
which should be present even when the hill loss factor exceeds unity so that there is no hill
as such. It seems likely from the present study that these instruments could all be modified by
adjusting their bridges, so that they behaved more like his set of old Italian violins (see the
same figure [3]).
Bridge hill/Woodhouse 11 May 25, 2004
If a scientific approach to bridge adjustment were wanted, a possible procedure would be to
start by measuring the input admittance of the violin in question using a Jansson plate bridge
in order to calibrate the body behaviour. This would reveal the constraints imposed by the
body behaviour, then it should be possible to adjust a bridge to bring the hill resonance to the
right frequency range. At the same time the maker should be careful to monitor the bridge
mass, to control the height and bandwidth of the hill. Note that the bridge height and foot
spacing also have an influence, although in practice there is only limited scope to vary these.
Another possible procedure might be to measure directly the 2 2 admittance matrix that
describes the properties of the violin body at the two foot positions, then calculate the moment
admittance R from eq. (3) and use it to optimise a virtual bridge design by computer. The
real bridge could then be cut while making regular comparisons with the computer model to
guide the adjustment process.
To understand why some violins seem to have a greater reactive component of R

than others
would require an extension of the modelling of the violin body. The idea of obtaining the
skeleton curve by allowing the top and back plates to become infinite is still valid, but certain
details of the violin structure are sufficiently close to the bridge that they should be included.
The arching and graduation pattern of both plates around the bridge/soundpost area, the
central portion of the bass bar and the free edges at the f-holes are all strong candidates for
inclusion [8]. However, the more remote regions of the plates could be allowed to extend to
infinity in some suitable way, perhaps by using absorbing boundaries in the computation so
that no reflections were generated. If such a model could be analysed, probably using finite-
element methods, it would allow the parameter study of section 4 to be extended to other
aspects of the violin structure. Such a model may be a worthwhile subject of future research.
Bridge hill/Woodhouse 12 May 25, 2004
Acknowledgements
The author is grateful to Erik Jansson, Claire Barlow and Robin Langley for useful
discussions on this work, to Martin Woodhouse for helping prepare Figure 1 and to Huw
Richards for the violin whose response is shown in Fig. 2.
References
[1] L. Cremer: The Physics of the Violin, MIT Press, Cambridge MA 1985. See chapter 9.
[2] W. Reinicke: Die bertragungseigenschaften des Streichinstrumentenstegs. Doctoral
dissertation, Technical University of Berlin, 1973.
[3] H. Dnnwald: Deduction of objective quality parameters on old and new violins. J.
Catgut Acoust. Soc. Series 2, 1, 7 (1991)15.
[4] E. V. Jansson: Admittance measurements of 25 high quality violins. Acustica Acta
Acustica 83 (1997) 337341.
[5] E. V. Jansson, B. K. Niewczyk: Admittance measurements of violins with high arching.
Acustica - Acta Acustica 83 (1997) 571574.
[6] E. V. Jansson, B. K. Niewczyk: On the acoustics of the violin: bridge or body hill. J.
Catgut Acoust. Soc. Series 2, 3, 7 (1999) 23-27.
[7] E. V. Jansson: Violin frequency response bridge mobility and bridge feet distance.
Submitted to Applied Acoustics (2004).
[8] F. Durup, E. V. Jansson: The quest of the violin bridge hill. Submitted to Acustica
Acta Acustica(2004).
[9] I. P. Beldie: About the bridge hill mystery. J. Catgut Acoust. Soc. Series 2, 4, 8 (2003)
913.
[10] C. H. Hodges, J. Woodhouse: Theories of noise and vibration transmission in complex
structures. Reports on Progress in Physics 49 (1986) 107-170.
[11] M. E. McIntyre, J.Woodhouse: On measuring the elastic and damping constants of
orthotropic sheet materials. Acta Metallurgica 36 (1988)13971416.
[12] E. Szechenyi: Modal densities and radiation efficiencies of unstiffened cylinders using
statistical methods. J. Sound Vib. 19 (1971) 6581.
[13] E. Skudrzyk: The mean-value method of predicting the dynamic response of complex
vibrators. J. Acoust. Soc. Amer. 67 (1980) 11051135.
[14] R. Ohayon, C. Soize: Structural Acoustics and Vibration, Academic Press, San Diego
CA, 1998. See chapter 15.
[15] L. Cremer, M. Heckl, E. E. Ungar: Structure-borne Sound, Springer, Heidelberg 1988.
See section IV, 3d.
[16] J. S. Loen: Reverse graduation in fine Cremonese violins. J. Catgut Acoust. Soc. Series
2, 4, 7 (2003) 27-39.
Bridge hill/Woodhouse 13 May 25, 2004
Table 1. Standard parameter values for the violin and bridge models. Wood properties
are typical of published data, except that no data are available for D
2
and D
4
for maple and
these values have been guessed. Calculations of skeleton curves assume eq. (14), and thus
disregard the values of D
2
and D
4
given here.
Plate property Symbol Unit Spruce value Maple value
Density kg m
-3
420 650
Elastic constants D
1
MPa 1100 860
D
2
MPa 67 140
D
3
MPa 84 170
D
4
MPa 230 230
Length L
1
mm 321 321
Width L
2
mm 204 204
Thickness h mm 2.9 4.0
Property Symbol Unit Value
Bridge foot positions x
1
, y
1
( ) mm (120,87)
x
2
, y
2
( ) mm (120,117)
Soundpost position x
3
, y
3
( ) mm (110,117)
Clamped frequency
b
/ 2 Hz 3000
Bridge mass m g 0.5
Bridge height a mm 20
Foot spacing d mm 30
Bridge hill/Woodhouse 14 May 25, 2004
Figure 1. A violin bridge with an indication of the motion in the lowest in-plane bridge
resonance.
200 500 1000 2000 5000
-50
-40
-30
-20
-10
Frequency (Hz)
A
d
m
i
t
t
a
n
c
e

(
d
B
)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 2. Input admittance of a violin, showing a typical bridge hill (in the frequency range
indicated by the dashed line in the upper plot). The upper plot shows the magnitude in dB re
1 m s
1
N
1
.
Figure 3. Sketch of a generic system driven through a resonator, as described in the text.
Bridge hill/Woodhouse 15 May 25, 2004
200 500 1000 2000 5000
-60
-40
-20
0
Frequency (Hz)
|
Y
v
|

(
d
B
)
(a)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
200 500 1000 2000 5000
-60
-40
-20
0
Frequency (Hz)
|
Y
b
|

(
d
B
)
(b)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 4. Frequency responses of an idealised system based on Fig. 3: (a) input admittance of
the base system; (b) input admittance at the position of forcing, including the effect of the
series oscillator and showing bridge hill behaviour. Dashed lines indicate the infinite
system response skeleton as explained in section 3. The body has modes equally spaced at
200Hz intervals, all with modal mass 0.1kg and Q-factor 50. The bridge has mass 1.5g
and clamped frequency 3kHz.
Bridge hill/Woodhouse 16 May 25, 2004
Figure 5. Idealised model of a violin bridge, including a single resonance to model the
deformation shown in Fig. 1.
200 500 1000 2000 5000
10
20
30
40
50
60
Frequency (Hz)
|
R
|

(
d
B
)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 6. Rotational input admittance R() of a rectangular spruce plate driven through the
rigid bridge base of Fig. 5. Parameter values are given in Table 1. Dashed lines indicate
the infinite system response skeleton R

() as explained in section 3.
Bridge hill/Woodhouse 17 May 25, 2004
200 500 1000 2000 5000
-60
-40
-20
0
Frequency (Hz)
|
Y
b
|

(
d
B
)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 7. Input admittance of the plate system of Fig. 6 driven through the bridge model of
Fig. 5. Parameter values are given in Table 1. Dashed lines indicate the infinite system
response skeleton as explained in section 3.
Bridge hill/Woodhouse 18 May 25, 2004
200 500 1000 2000 5000
10
20
30
40
50
60
Frequency (Hz)
|
R
|

(
d
B
)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 8. Rotational input admittance R() of the violin body model described in the text,
driven through the rigid bridge base of Fig. 5. Parameter values are given in Table 1.
Dashed lines indicate the infinite system response skeleton R

() as explained in section
3.
Bridge hill/Woodhouse 19 May 25, 2004
200 500 1000 2000 5000
-60
-40
-20
0
Frequency (Hz)
|
Y
b
|

(
d
B
)
200 500 1000 2000 5000
-90
-45
0
45
90
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
Figure 9. Input admittance of the violin body model of Fig. 8 driven through the bridge
model of Fig. 5. Parameter values are given in Table 1. Dashed lines indicate the infinite
system response skeleton as explained in section 3.
Bridge hill/Woodhouse 20 May 25, 2004
200 500 1000 2000 5000
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
Frequency (Hz)
Y

(
m

s
-
1
N
-
1
)
Figure 10. Real part (solid line) and imaginary part (dashed line) of the infinite-plate response
from eq. (16) using parameter values from Table 1 and a distance r equal to the bridge foot
spacing, 30mm.
Bridge hill/Woodhouse 21 May 25, 2004
200 500 1000 2000 5000
-40
-20
0
20
40
60
80
100
120
Frequency (Hz)
R

(
r
a
d

s
-
1
N
-
1
m
-
1
)
Figure 11. Real part (solid line) and imaginary part (dashed line) of the rotational input
admittance R

() of the skeleton model described in the text. Parameter values are given
in Table 1.
Bridge hill/Woodhouse 22 May 25, 2004
-100 -80 -60 -40 -20 0 20 40 60 80 100
2000
2500
3000
3500
4000
Im(R

) (rad s
- 1
N
- 1
m
- 1
)
H
i
l
l

f
r
e
q
u
e
n
c
y

(
H
z
)
(a)
-100 -80 -60 -40 -20 0 20 40 60 80 100
0.186
0.188
0.19
0.192
Im(R

) (rad s
- 1
N
- 1
m
- 1
)
H
i
l
l

l
o
s
s

f
a
c
t
o
r
(b)
Figure 12. Variation of the hill peak properties from Eq. (18) with the imaginary part
(assumed independent of frequency) of the skeleton rotational input admittance R

. The real
part of R

has the constant value 100 rad m

-1
s
-1
N
-1
. Plot (a) shows the frequency given by
Re ( ); plot (b) shows the loss factor
b
Im() / Re().
Bridge hill/Woodhouse 23 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 13. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the bridge mass and stiffness are varied
while keeping the bridge frequency
b
fixed. The curves show mass m=0.2, 0.3, 0.5, 1.0,
2.0g, the first of these values being indicated by a dashed line and the others following in an
obvious sequence.
Bridge hill/Woodhouse 24 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 14. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the bridge mass is varied while keeping
the bridge stiffness K fixed. The curves show mass m=0.2, 0.3, 0.5, 1.0, 2.0, 5.0, 10.0g, the
first of these values being indicated by a dashed line and the others following in an obvious
sequence. The lighter masses are intended to represent variations of bridge adjustment, while
the heavier ones represent the effect of adding a mute.
Bridge hill/Woodhouse 25 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 15. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the bridge stiffness K is varied while
keeping the bridge mass m fixed. The curves are most conveniently indexed by the bridge
frequency
b
/ 2, which takes the values 2.0, 2.5, 3.0, 4.0, 5.0kHz, the first of these values
being indicated by a dashed line and the others following in an obvious sequence.
Bridge hill/Woodhouse 26 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 16. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the bridge foot spacing d is varied. The
curves show d=5, 10, 20, 30, 40mm, the first of these values being indicated by a dashed
line and the others following in an obvious sequence.
Bridge hill/Woodhouse 27 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 17. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the thickness of the spruce top plate is
varied. The curves show thickness h=2.0, 2.5, 3.0, 3.5, 4.0mm, the first of these values
being indicated by a dashed line and the others following in an obvious sequence.
Bridge hill/Woodhouse 28 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 18. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the thickness of the maple back plate is
varied. The curves show thickness h=2.0, 2.5, 3.0, 3.5, 4.0mm, the first of these values
being indicated by a dashed line and the others following in an obvious sequence.
Bridge hill/Woodhouse 29 May 25, 2004
200 500 1000 2000 5000
-60
-50
-40
-30
-20
-10
0
10
Frequency (Hz)
|
Y
b
|

(
d
B
)
Figure 19. Skeleton curves for the input admittance showing variation of the bridge hill.
Parameter values are as given in Table 1, except that the position x
3
(parallel to the grain)
of the soundpost is varied. The curves show soundpost offset x
2
x
3
=5, 10, 15, 20, 30mm,
the first of these values being indicated by a dashed line and the others following in an
obvious sequence.