Bilbao Acustica r2

ACTA ACUSTICA UNITED WITH
Vol. 98 (2012)
ACUSTICA
Numerical Modeling of Collisions in Musical

Instruments
S. Bilbao1) , A. Torin1) , V. Chatziioannou2)
1)
Acoustics and Audio Group, Kings Buildings, University of Edinburgh, Edinburgh, United Kingdom.
sbilbao@staffmail.ed.ac.uk
2)
Institute of Music Acoustics, University of Music and Performing Arts, Vienna, Austria
Summary
Collisions play an important role in many aspects of the physics of musical instruments. The striking
action of a hammer or mallet in keyboard and percussion instruments is perhaps the most important
example, but others include reed-beating effects in wind instruments, the string/neck interaction in
fretted instruments such as the guitar as well as in the sitar and the wire/membrane interaction in
the snare drum. From a simulation perspective, whether the eventual goal is the validation of musical
instrument models or sound synthesis, such highly nonlinear problems pose various difficulties, not
the least of which is the risk of numerical instability. In this article, a novel finite difference time
domain simulation framework for such collision problems is developed, where numerical stability
follows from strict numerical energy conservation or dissipation, and where a power law formulation
for collisions is employed, as a potential function within a passive formulation. The power law serves
both as a model of deformable collision, and as a mathematical penalty under perfectly rigid, nondeformable collision. Various numerical examples, illustrating the unifying features of such methods
across a wide variety of systems in musical acoustics are presented, including numerical stability and
energy conservation/dissipation, bounds on spurious penetration in the case of rigid collisions, as well
as various aspects of musical instrument physics.
PACS no. 43.75.Zz
1. Introduction
Various mechanisms of sound production in musical
instruments rely on collisions; the obvious examples
are the interaction of a striking object, such as a hammer with a string, or a drum stick or mallet with a percussive instrument, but more subtle examples include
reed-beating effects in wind instruments, string/neck
interactions along the fretboard of a guitar, and also
in the sitar and tambura, and most dramatically, the
wire/membrane interaction in a snare drum. In all
cases, the collision interaction is necessarily strongly
nonlinear, and simulation design becomes a challenging problem.
Such collisions may be grouped into two types. In
the first, one of the two objects involved in the collision is modelled as lumpedi.e., it is characterized
by a single position/velocity pair. Such is the case for
most models of the hammer-string [1, 2], mallet-bar
[3] and mallet-membrane interactions [4], even if the
colliding object occupies a finite interaction region.
Received 27 October 2012,
accepted 6 December 2012.
S. Hirzel Verlag EAA
The perceptual effects of such a nonlinear interaction

are major, particularly with regard to the spectral
content of the resulting sound. In the second, both
objects must be considered to be fully distributed,
and the region of contact will vary in a non-trivial
manner. Examples include the snare drum [5], sitar
[6] and string fret interactions [7] mentioned above.
Collision modeling is a branch of the large area of
contact mechanics with applications across a variety
of disciplines, and especially robotics [8] and computer
graphics [9]see Wriggers et al. [10] for a recent review. In musical acoustics and applications in sound
synthesis, various techniques have been employed, including digital waveguides [11, 7], modal methods [6]
and time stepping methods in the lumped setting
[12, 13, 14], in modeling lumped/distributed collisions
[15, 16, 2, 17], the interaction of a distributed object
with a rigid barrier [11, 18, 19], and in the collision
of deformable objects for synthesis applications [20].
Finite element methods are often used to model collisions of complex deformable bodies in mainstream
applications [21]; in the present case of musical acoustics and sound synthesis, where geometries are often
simple, finite difference time domain methods [22] are

Vol. 98 (2012)
ACUSTICA
Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions
an efficient alternative, and will be employed here.

Such methods serve as an alternative to other techniques which can in some instances be more efficient
(such as, e.g., digital waveguides [23] or modal methods [24]) because they generalize more easily to handle distributed interactions, as in the present case of
collisions.
In the setting of musical acoustics, previous work
on particular cases of collisions using finite difference
methods, such as, for example, the snare membrane
interaction [5] and the reed/lay interaction [17] has
generally made use of ad hoc techniquesthat is to
say, there is not any attempt at proving numerical stability under such strongly nonlinear conditions. As in
the case of other types of inherent distributed nonlinearities in musical instruments (as, e.g., in strings [25]
or curved shells [26]), an approach based on energy
principles [27] is of great utility; energetic methods
are widely used in elastodynamics, whether in a variational setting [28] or when strict energy conservation
or dissipation is enforced [29]. The general aim of this
paper, then, is to incorporate the nonlinear collision
mechanism into the energy conservation framework,
and use such a formulation in order to arrive at sufficient stability conditions for such methods, and show
its application to a wide variety of systems, spanning
the range of musical instruments. The mechanism by
which this is accomplished is through the introduction
of a potential energy term corresponding to the collision interaction; such a mechanism has indeed been
proposed, in the case of a lumped collision in the semidiscrete case by Rhaouti et al. in [4] and in the fully
discrete case by Chabassier in [15] and Chabassier et
al. [16]. The case of distributed collisions, however,
which forms a large part of the present article, has not
been treated previously within an energy-conserving
framework to the knowledge of these authors.
If the collision is accompanied by some deformation
of the colliding object (as for, e.g., the piano hammer), then the additional potential has the interpretation of energy stored in the deformed object. In the
case of rigid collisions, however, the potential must be
interpreted as a penaltysuch a potential results in
a strong opposing force, penalizing penetration, but
some limited interpenetration of the colliding bodies is permitted. Such penalty methods [30] can be
viewed in contrast with methods based on hard nonpenetrative constraints [31]. As long as contact velocities are low (which is generally the case in musical
acoustics applications), the latter approach is justifiable. Indeed, a second benefit of such an energy-based
formulation, beyond proving stability, is a means of
determining bounds on such spurious penetrationas
will be seen, the amount of penetration can be made
as small as desired (and particular, small enough to
be negligible in applications in acoustics). Numerical
existence and uniqueness results in the simple case
of a lumped collision with a rigid barrier have been
presented recently [14] and will be extended to fully

distributed systems here. The issue of existence and
uniqueness for the underlying systems in the presence
of collision mechanisms is a difficult oneit has been
approached, in the case of lumped collisions with a
rigid barrier and with a vibrating string, by Joly and
Rhaouti [32], and is out of the scope of this work.
In Section 2, the collision of a mass with a rigid barrier is used as a test case in the construction of stable
energy conserving methods, accompanied by an extension to the case of a lossy collision. Collisions of
a lumped object with a distributed system are considered next. To this end, some background material
on discrete representations of distributed systems is
presented in Section 3. The case of the hammer string
interaction is treated first in Section 4, followed by
reed beating effects, in a simple model of a wind instrument in Section 5. A fully distributed collision is
considered next, in the case of a string in contact with
a rigid barrier in Section 6. In 2D, the simplest system of interest is the mallet/membrane interaction,
described in Section 7. Finally, the case of a wire in
contact with a membrane, modeling the snare interaction is described in Section 8. As a nontrivial example, in this last section, a complete 3D model of a
snare drum, including the cavity, membranes, acoustic field and snares is outlined. As the schemes presented here all require the solution of nonlinear equations, some existence/uniqueness results are provided
in Appendix A, along with some comments on the use
of iterative methods.
2. Prelude: Collision of a Mass with a

Rigid Barrier
A starting point in many numerical studies of collisions [10, 13, 14] is the case of a lumped deformable
object approaching a rigid barrier from below, under a
nonlinear interaction force. The system may be written as
M
d2 u
= f
dt2
f=
d
d du
=
/
du
dt dt
(1)
Here, u = u(t) is the position of the object at time t

R+ , and M is its mass. For this second order equation,
two initial conditions, namely u(0) and du/dt(0) must
be supplied. f = f (t) is the interaction force, written
here in terms of a potential (u) 0. For collisions,
with a barrier at u = 0, f is zero for u 0; when
u > 0, a deformable object undergoes compression,
and a rigid object can be viewed as penetrating the
barrier. A power law is a useful general choice of the
potential:
= K, =
K[u]+1
+
0
+1
f = K[u]
+ . (2)
ACTA ACUSTICA
It depends on a stiffness parameter K 0 and an

exponent > 1; in this article, the notation []+ indicates the positive part of, i.e., [u]+ = 12 (u + |u|).
Such power law nonlinearities are common in models of collisions in many settings, including not just
the present case of the lumped collision, but also in
models of the hammer string interaction [2], where
and K are empirically determined (though for simpler
systems, such as the contact of two spheres, Hertzian
models allow a direct calculation [33]). Such models
may be viewed as permitting a certain deformation of
the colliding object when in contact with the barrier.
If the colliding object is perfectly rigid, then such a
model is obviously unphysical, in that some penetration is permittedin this case, the potential may
be interpreted as a penalty, and u > 0 is itself the
penetration; in numerical treatments, = 1 is often
chosen [30]. Most of what follows here does not depend
on a particular choice of the potentialbut existence
and uniqueness in implementation do depend on the
form of , and the power law has various advantages
which will be highlighted.
UNITED WITH ACUSTICA

Vol. 98 (2012)
2.2. Time Series and Difference Operators

Moving to a discrete time simulation setting, let u =
un represent an approximation to u(t) at t = nk, for
integer n, and for a given time step k.
Unit shifts et+ and et are defined as
et+ un = un+1
et un = un1 .
(7)
Forward, backward and centered difference approximations to a first time derivative may thus be defined
as
t+ =
et+ 1
k
1 et
k
t =
t =
et+ et
,(8)
2k
and an approximation to a second time derivative as

tt =
et+ 2 + et
.
k2
(9)
Various averaging operators may be defined as

t+ =
et+ + 1
2
t =
1 + et
2
t =
et+ + et
.(10)
2
2.1. Energy Balance
2.3. Finite Difference Scheme
The collision model presented above is energy conserving [13]. Multiplying (1) by du/dt leads to
An approximation to (1), at time step n is then
du
du d
d
du d2 u
= f =
=
,
M
2
dt dt
dt
dt du
dt
(3)
and thus to the energy balance, in terms of total energy H:

2
dH
M du
= 0 , H(t) =
+ .
(4)
dt
2
dt
Thus H(t), corresponding to the total energy of the
object at time t, is conserved and non-negative:
H(t) = H (0) 0 .
(5)
Note that the procedure above, involving multiplication by the velocity, leads to a construction of the
energy conservation law, and can be viewed as the reverse of the procedure employed in the derivation of
equations of motion from expressions for the energy
using Hamiltons equations [34].
For the particular choice of the power law potential
function = K, from (2), the conservation law
implies bounds on both u(t) and du/dt at any time
t, in terms of the initial energy H(0). These may be
arrived at by noting that, as the individual terms of
H in (4) are both non-negative, both are bounded
individually by H(0), leading to:

1
du r 2H(0)
( + 1) H(0) +1

, u(t)
. (6)

dt
M
K
The first bound holds for any non-negative potential

, and the second employs monotonicity of the power
law K, in u 0.
M tt un = f n ,
(11a)
where tt is a second difference operator, as defined in

(9), and where f n is a time series defined by
1
fn =
t n+ 2
t un
n+ 2 = t+ (un )
(11b)
in terms of a discrete potential n+ 2 , itself defined

in terms of the potential of the model problem.
Note here that the definition of f n here mirrors that
of the continuous time case, from (1); simpler, fully
explicit forms are available, but can lead to stability
problemssee the end of Section 4.4 for an example
of numerical instability, in the case of a hammer in
contact with a string. The backwards and centered
difference operators t and t employed in (11b) are
defined in (8), and the averaging operator t+ is de1
fined in (10). The time series n+ 2 is interleaved with
respect to the time series un itself; thus scheme (11)
is centered about time step n and, if stable, is second
order accurate [22] (note in particular the application
1
of the backward difference operator to n+ 2 , which
centers the approximation about time step n).
Multiplying (11a) by t un , and employing (11b)
gives
1
M t un tt un = t un f n = t n+ 2 .
(12)
Employing the identity

t un tt un =
1
2
t (t+ un )
2
(13)

Vol. 98 (2012)
ACUSTICA
leads to the discrete energy balance

n+ 12
t h
=0
n+ 12
2.4. Nonlinear Losses
= const. 0 , (14)
where the discrete energy of the system is defined as

n+ 12
f=
M
1
2
=
(t+ un ) + n+ 2 0 .
2
(15)
Bounds on the solution size follow as in the continuous case, under the choice of a power law nonlinearity
K, :
|t+ un |
2h
M
un
2 ( + 1) h
K
1
+1
,(16)
and the scheme is thus unconditionally stable. Notice

in particular that the amount of penetration may be
controlled through the choice of K in this case.
The scheme (11) requires the solution of a nonlinear
equation at each time step,
G(r) = r +
m
((r + a) (a)) + b = 0 , (17)
r
in terms of an unknown r, defined in terms of un as

r = un+1 un1 ,
(18)
m = k /M a = u
n1
b = 2u + 2u
n1
(19)
It is worth contrasting this energy-conserving

method with that presented in [14], which is also
energy-conserving, and ultimately distinct from that
presented here. In this paper, the energy balance is derived from the dynamical system; in [14], Hamiltons
equations are taken as a point of departure. Whether
these two approaches can be unified is an interesting question. The key feature, though, that of a nonnegative conserved quantity used to bound solutions,
is the same in both cases.
(20)
where is defined as in (2), and for some function

(u) 0; in particular, the choice of
K, (u) = K[u]
+
(21)
for some 0 has been employed in previous studies

of collisions. (Other models used in musical acoustics, particularly in the case of hammer felt are similar [36].) In this case, the energy balance (4) can be
generalized to
2
dH
du
= Q , Q =
(u) 0 ,
(22)
dt
dt
and thus energy is monotonically decreasing, and the
bounds on the velocity and displacement hold as before.
Scheme (11) may be generalized using
1
fn =
t n+ 2
+ t un n ,
t un
(23)
where n = (un ). Now the energy balance (14) may

be generalized to
1
At each time step, once r is determined, the next displacement un+1 may be determined from r and un1 .
The function G(r), as defined in (17), as well as certain generalized forms, play a central role in all the
algorithms to be described subsequently here in the
distributed case. It has recently been shown [14] that
G(r) = 0 possesses a unique solution for one-sided potentials of power law form. An iterative method such
as the Newton-Raphson algorithm may be employed,
but conditions for global convergence are not obvious.
See Appendix A for more discussion.
d du
+
(u) ,
du
dt
t hn+ 2 = qn
where m, a and b are given by

2
In some more refined collision models [13], a nonlinear

damping term is included. From the model of Hunt
and Crossley [35] the force f in (1) may be augmented
to
, qn = (t un ) n 0 , (24)
and thus energy is monotonically decreasing, and solution bounds hold as before. The nonlinear equation
to be solved at each time step must be generalized to
G(r) = (1 + c) r +
m
((r + a) (a)) + b = 0 (25)
r
where r, the unknown, as well as the constants m, a

and b are as before, and where c = k(un )/2M .
2.5. Simulations
Consider first the collision of a mass with a rigid barrier under different choices of the exponent , using
scheme (11), as illustrated in Figure 1, and with a
typical value of K (here 108 ) which is in the range of
values used typically in studies of piano hammers [37].
The trajectories of the mass are illustrated at top; the
obvious effect of the exponent is on the duration of the
contact time, but as the scheme is exactly lossless, the
initial and exiting speeds of the mass are identical to
machine accuracy. As a further indication of the exact
energy conservation property of the scheme, the normalized energy variation is plotted against time step
in Figure 1, at bottom; single bit variation of the energy (at double precision floating point) is evident as
quantization.
u (mm)
= 1.5
=2
= 2.5
3.1. Continuous Inner Products and Norms
10
20
t (ms)
15
x 10
2
0
2
0
50
100
150
time step n
200
250

Vol. 98 (2012)
time domain schemes in 1D and 2D is presented, with

a focus on energy techniques.
10
0
ACTA ACUSTICA
300
Figure 1. Top: trajectory of a mass M = 10 g, colliding

with a rigid barrier, under different values of the exponent
, as indicated. Here the mass has initial velocity 10 m/s,
and K = 108 . Scheme (11) is used, with a sample rate
of 44.1 kHz.
Bottom: normalized energy variation n =

n+ 1
1/2
2
h
h
/h1/2 , when = 2.5.
As a further example, the behaviour of the same

system, under the lossy collision model given in Section 2.4 is plotted in Figure 2, under difference choices
of the loss parameter . Energy is monotonically decreasing, illustrating a complex non-exponential loss
characteristic during the collision itself.
Consider real-valued functions (x, t) and (x, t) defined over a d-dimensional domain, with x D Rd ,
and for time t R+ . Particular domains of interest
in the present setting are, for a given length L, a 1D
interval DL and the 2D square domain DL,L , defined
as
DL = {x R|0 x L} , DL,L = DL DL ,(26)
where indicates a Cartesian product. The L2 spatial
inner product and norm are defined as
Z
p
h, iD =
dx , kkD = h, iD 0 , (27)
D
where dx is a d-dimensional differential element.

For such an inner product, it is true that
d 1
kk2D = h, t iD ,
(28)
dt 2
where t and d/dt represent partial and total differentiation with respect to t, respectively.
Integration by parts (or Greens first identity) may
be written as
Z
h, iD = h, iD +
n d (29)
D
u (mm)
=0
= 0.5
=1
0
5
10
Energy (J)
t (ms)
3.2. Discrete Domains and Grid Functions

An N + 1 point discrete 1D spatial domain dN corresponding to the line segment DL may be defined
as
0.5
where is the Laplacian operator in d dimensions,

is the d-dimensional gradient, and where D is the
boundary of the domain D, with outward normal n.
50
100
150
time step n
200
250
300
Figure 2. Top: trajectory of a mass colliding with a rigid

barrier, under different values of the loss coefficient , as
indicated, with = 2.3. All other conditions are as in the
1
caption to Figure 1. Bottom: energy hn+ 2 , when = 1.
dN = {l Z, 0 l N }
for some integer N such that the grid spacing h =

L/N . In a finite difference setting, other domains
which differ from dN through the removal of endpoints
are also of interest:
dN = dN {N } , dN = dN {0, N } .
In the case of a completely rigid collision, some spurious penetration (here u > 0) is inevitable. Considering again the system under the choice of parameters given in Figure 1. Using values of K = 1016 and
= 1.2, approximating such a rigid collision, the penetration is under 8108 m, which is negligible in any
acoustics application.
3. Background: Distributed Systems

and Grid Functions
In this section, some background material on systems
and grid functions suitable for use in finite difference
(30)
(31)
unl ,
A 1D grid function
defined over l d, for some
1D domain d, and for integer n represents an approximation to u(x, t) at t = nk and x = lh.
Similarly, a 2D discrete domain dN,N corresponding
to the square region DL,L may be written as
dN,N = dN dN ,
(32)
and truncated domains follow as

dN ,N = dN dN , dN,N = dN dN .
(33)
A 2D grid function unl,m , defined over (l, m) d, for

some 2D domain d, and for integer n represents an
approximation to u(x, y, t) at t = nk, x = lh and
y = mh.

Vol. 98 (2012)
ACUSTICA
3.3. Difference Operators

The time difference and averaging operators presented
in Section 2.2 are unchanged in their application to
grid functions.
In 1D, unit rightward and leftward spatial shifts ex+
and ex as applied to a grid function unl are defined
as
ex+ unl = unl+1 , ex unl = unl1 .
(34)
Here and henceforth, a reference to a grid point outside of the domain dN or dN,N indicates a "ghost
point," the value of which will be set subsequently
in terms of values over the interior through an appropriate boundary condition.
Forward, backward and centered difference approximations to a first spatial derivative may be defined
in terms of these shifts as
x+ =
ex+ 1
1 ex
ex+ ex
, x =
, x =
,(35)
h
h
2h
and centered approximations to second and fourth

spatial derivatives as
ex+ 2 + ex
h2
2
e 4ex+ + 6 4ex + e2x
= x+
.
h4
xx =
xxxx
(36a)
(36b)
Averaging operators may be defined as

x+ =
ex+ + 1
1 + ex
, x =
, xx = x+ x . (37)
2
2
In both cases, other choices of inner products are

possible (perhaps scaling the boundary grid values by
an appropriate amount, such as 1/2 in the 1D case).
Distinct choices of inner product lead to distinct numerical boundary conditions, and the choice made
here is for simplicity only. See Section 5.2.9 of [17]
for more on this point.
3.5. Identities and Bounds
The following identities hold for grid functions u defined over a domain d:
1
hu, t uid = t hu, et uid
2
1
ht u, tt uid = t kt+ uk2d ,
2
ex unl,m = unl1,m
ey+ unl,m
ey unl,m
unl,m+1
unl,m1
(38)
, (39)
and first difference operators in both x and y may

be defined in analogy with (35) and second difference
operators xx and yy with (36a). A simple approximation to the Laplacian operator follows as
= xx + yy .
(40)
3.4. Inner Products

For 1D grid functions nl and ln , defined over a domain d, an L2 inner product and norm may be defined
as
X
p
h, id =
hnl ln kkd = h, id 0 . (41)
ld
n
For 2D grid functions nl,m and l,m
, defined over a
domain d, the inner product and norm may be defined
as
X
p
n
h, id =
h2 nl,m l,m
, kkd = h, id 0 .(42)
(l,m)d
(44)
where here, the spatial and temporal indices have

been suppressed. Furthermore, the following inequality holds:
hu, et uid
k2
kt uk2d .
4
(45)
Various discrete counterparts to integration by parts

(29) are available; here are various forms of interest,
written in terms of two grid functions nl and ln . In
1D, one has:
hn , xx n idN = hx+ n , x+ n idN
n
n0 x 0n + nN x+ N
hn , xxxx n idN = hxx n , xx n idN
(46a)
(46b)
n0 x xx 0n + x+ n0 xx 0n
n
n
+nN x+ xx N
x nN xx N
In 2D, unit shifts ex+ , ex , ey+ , and ey as applied

to a grid function unl,m are defined as
ex+ unl,m = unl+1,m
(43)
In 1D, the action of spatial difference operators may

be bounded as
2
4
kx+ ukdN 2 kukdN .
h
h
kxx ukdN
(47)
Similarly, in 2D, summation by parts in the two directions x and y, may be written as
hn , xx n idN,N = hx+ n , x+ n idN ,N (48a)
+
N
X
m=0
n
n
h2 nN,m x+ N,m
n0,m x 0,m
hn , yy n idN,N = hy+ n , y+ n idN,N (48b)

+
N
X
l=0

n
n
h2 nl,N y+ l,N
nl,0 y l,0
,
and the following bounds hold:

kx+ ukdN,N
ky+ ukdN,N
2
kukdN,N . (49)
h
See Chapter 5 of [17] for a fuller discussion of such

identities and bounds.
ACTA ACUSTICA
4. The Hammer-Linear String Interaction

The nonlinear interaction of a striking hammer with a
string has seen a good deal of investigation [1], and an
energy conserving method (distinct from that which
follows here) has been presented recently [15].
Consider a stiff string in contact with a hammer
striking from below, defined by
d2 uh
s tt u = Ls u + gf , M 2 = f .
dt
(50)
Here, u(x, t) is the transverse displacement of the

string in a single polarization, as a function of time
t R+ and x DL as defined in (26), where L is
the string length. s is linear density in kg/m, and tt
represents second partial differentiation with respect
to time. For both members of (50), which are second
order in time, initial conditions must specified (i.e.,
u(x, 0), t u(x, 0), uh (0) and duh /dt(0)).
f = f (t) is the force imparted to the string by a
colliding hammer, of mass M and at vertical height
uh = uh (t), and where g = g(x) is a distribution selecting the region of impact of the hammer over the
string. For a pointwise impact at a location x = x0 ,
one may use a Dirac distribution g(x) = (x x0 ),
but for a hammer
of finite width, a normalized distriR
bution with DL gdx = 1 may be employed.
The linear operator Ls in (50) is defined by
Ls = Ts xx Es Is xxxx 2s,0 s t +2s,1 s txx ,(51)
where Ts is string tension in N , Es is Youngs modulus in Pa, Is is the moment of inertia of the string in
m4 (and equal to rs4 /2 for a string of circular cross
section, where rs is the string radius), and s,0 0
and s,1 0 are parameters allowing for frequency
dependent losswhen s,0 = s,1 = 0, the system is
lossless. Such a system [38] is similar to that which
has been used in models of lossy string vibration
[39, 2], but without recourse to higher time derivatives. Boundary conditions of various types may be
considered, but for the present investigation clamped
conditions of the type
u=0 ,
x u = 0
(52)
at the domain endpoints x = 0 and x = L are sufficient.

The force f depends on a measure of distance
between the string and the hammer:
f=
d
, = uh hg, uiDL .
d
(53)
The notation h, iDL represents an L2 inner product

over DL , as defined in (27). (Note that if g represents
a distribution such as a Dirac delta function, which
is not in L2 , then the inner product above is to be

Vol. 98 (2012)
interpreted as a duality pairing between a distribution

and a test function which reduces to an L2 when g
is square integrable.) Here again, () is a potential
function, typically modeled as a power law of the form
of K, [2], with K and set from experiment; lossy
models, using a force term of the form given in (20)
are also used in studies of hysteresis of the felt in piano
hammers.
4.1. Energy Balance
The time derivative of the total energy of the combined string/hammer system may be derived by taking an inner product of the first of (50) with t u over
DL :
s ht u, tt uiDL = ht u, Ls uiDL + ht u, giDL f . (54)
Using identities (28) and integration by parts (29),
and noting from (53) that, for a fixed distribution g,
ht u, giDL =
d
d
hu, giDL =
(uh ) ,
dt
dt
(55)
then the following energy balance results:

L
dH

= Qs + Bs
dt
0
, H = Hs + Hh .
(56)
Here, the string energy Hs , hammer energy Hh , power

dissipated in the string Qs and supplied at the boundary Bs are given by
s
Ts
Es Is
kt uk2DL + kx uk2DL +
kxx uk2D(57a)
L
2
2
2

2
M duh
Hh =
+
(57b)
2
dt
Hs =
Qs = 2s,0 s kt uk2DL + 2s,1 s ktx uk2DL

Bs = Ts t ux u Es Is t uxxx u
(57c)
(57d)
+Es Is tx uxx u + s,1 s t utx u ,
where Hs , Hh , Qs 0. Under the clamped conditions

(52), Bs vanishes, and the system is strictly dissipative, i.e., dH/dt 0. As the individual terms in the
energy balance are non-negative, it is again possible
to arrive at bounds on uh and u in terms of the total
energy.
Employing the difference operators mentioned in the
previous section, an approximation to (50) over the
grid dN , of spacing h, is then
s tt unl = ls unl + gl f n , M tt unh = f n ,
(58)
where unl is a grid function approximating the string

displacement, and where unh and f n are time series approximating the hammer displacement and force respectively. gl is an approximation to the spatial distribution of the hammer. As in the continuous case,

Vol. 98 (2012)
ACUSTICA
the
PN grid function gl is chosen normalized such that
l=0 hgl = 1.
The operator ls is an approximation to Ls :
ls = Ts xx Es Is xxxx2s,0 s t +2s,1 s t xx (59)

and the force f may be written in terms of a discrete
1
potential n+ 2 as
where hn+ 2 , the total numerical energy is defined as

h = hs + hh , and
s
Ts
kt+ un k2dN + hx+ un , x+ un+1 idN (65a)
2
2
Es Is
s,1 ks
n
n+1
+
hxx u , xx u
idN
kt+ x+ un k2dN
2
2
M
1
n+ 1
2
hh 2 =
(t+ uh ) + n+ 2
(65b)
2

n+ 12
hs
qns = 2s s,0 kt un k2dN + s,1 kt x+ un k2dN
(65c)
bns,0
fn =
t n+ 2
, n = unh hg, un idN ,
t n
(60)
= Ts t un0 x un0 Es Is t un0 x xx un0

(65d)
n
n
n
n
+Es Is t x+ u0 xx u0 + 2s,1 s t u0 t x u0
n
bs,N = Ts t unN x+ unN Es Is t unN x+ xx unN
(65e)
n
n
n
n
+Es Is t x uN xx uN + 2s,1 s t uN t x+ uN .
where n+ 2 = t+ ( n ), and where h, idN is a discrete 1D inner product, as defined in (42).

Scheme (58) accompanied by (60), as in the case of
the lumped collision, requires the solution of a scalar
nonlinear equation of the form (17), in r = n+1
n1 , where
m=k
kgk2dN
1
+
s (1 + s,0 k) M
b = 2(unh un1
)+
h
where
a = n1
k2
hg, n idN ,
1 + s,0 k
(61)
(62)
Here, under (for example) discrete clamped boundary conditions

un0 = x+ un0 = 0 , unN = x unN = 0 ,
(66)
then the system is numerically dissipative because

(65d) and (65e) vanish. For numerical stability, all
that remains is to find a condition under which 1the
n+
energy function is non-negativenotice that hs 2 is
of indeterminate sign. To this end, using the inequalities (45) and (47), it may be bounded as

s Ts k 2 4Es Is k 2 2s,1 s k
n+ 12
hs
kt+ un k2dN .(67)

2
h
h4
h2
For a given time step k, this term is non-negative

under the condition h hmin , where

2
Ts
Es Is
n
n
s
l =
t + xx
xxxx +2s,1 t xx ul ,(63)
2
k
s
s
16Es Is
Ts k 2 ks,1 k
Ts k
2
hmin =
+
+
+4s,1 +
,(68)
2s
2
2
s
s
which again possesses a unique solution. Notice in parwhich serves as a stability condition for the complete
ticular that, given values of u and uh through time
schemeit is equivalent to the condition arrived at
step n, a and b may be computed explicitly, and thus
through frequency domain (von Neumann) analysis
only a single scalar nonlinear equation must be solved
[22] for the string alone, though now under nonlinin order to solve for r; once r is known, f n may be calear conditions, and reduces to the familiar Courant
culated, and the scheme (58) may be advanced to time
Friedrichs Lewy condition [41] in the case of an ideal
step n + 1. Such methods are similar in terms of comlossless string.
putational expense to other methods which have been
proposed in the context of nonlinear acoustic simula4.4. Simulations
tion (such as, e.g. the K method [40], which also reAs a simple illustration of this scheme, the case of a C4
quire the solution of a nonlinear algebraic equation),
piano string subject to a piano hammer striking action
but now with an energy-based stability condition.
is considered here, with parameters as drawn from
the article by Chaigne and Askenfelt [37]. In Figure
4.3. Numerical Energy and Stability Condi3, the force history of such a strike is plotted, under
tion
hammer velocities corresponding to piano, mezzoforte
and forte strikes, illustrating the main features of a
An energy balance for scheme (58) follows from an ingeneral decrease in contact time with increasing strike
ner product with t u, and using summation by parts
velocity, and the appearance of secondary humps in
identities (46), as
the force history, due to the reflection of waves from
the string termination.
More interesting, in the present context, are more
1
(64)
delicate features due to hammer width, which can be
t hn+ 2 = qns bns,0 + bns,N ,

ACTA ACUSTICA

Vol. 98 (2012)
20
hammer velocity 0.5 m/s
hammer velocity 1.5 m/s
hammer velocity 4 m/s
0.03
Energy (J)
f (N)
15
10
0.01
0
0
0.5
1.5
2.5
t (ms)
Figure 3. Hammer force history f in N, for a struck string,

under striking velocities, as indicated. Parameters corresponding to a C4 piano string are chosen: s = 0.0063
kg/m, Ts = 670 N, Es = 2 1011 Pa, rs = 5 104 m,
s,0 = 0.5 and s,1 = 0.5, and L = 0.62 m. The hammer,
of mass M = 0.0029 kg strikes pointwise at a position 0.12
of the way along the string. The collision potential K,
is used, with K = 4.5 109 and = 2.5. The sample rate
is 44.1 kHz.
modeled through a distributed contact region g; such

a choice has little impact on the computational cost of
the algorithm, and subtle variations in the force history can be observed with increasing contact width,
as shown in Figure 4, where g is a simple rectangular
window. One issue which emerges here is of the representation of such a function on a relatively coarse
gridone solution is to operate at a high sample rate
(here 176 kHz), leading to a sufficiently fine spatial
grid resolution; another is to employ a high order approximation to such a finite width distribution over
a coarse grid. The former approach has been adopted
here.
20
hammer width 3 mm
hammer width 10 mm
hammer width 20 mm
f (N)
15
10
0.5
1.5
t (ms)
2.5
0.5
1.5
t (ms)
2.5
15
x 10
Energy variation
string energy
hammer energy
0.02
0
5
10
0
Figure 5. Top: time evolution of the energy partition between the hammer and string, for the system as given
in the caption to Figure 3, under lossless conditions (i.e.,
with s,0 = s,1 = 0). Bottom: normalized variation in
numerical energy for the combined system.
a calculation of f n from previously computed values

as
f n = K[unh hun , gidN ]
+.
(69)
Even under mild hammer excitation conditions, the

non-conservative scheme can exhibit severe spurious
oscillations, as illustrated in Figure 6 under lossless
and non-stiff conditions. The introduction of losses
(or the reduction of the Courant number) always has
an ameliorating effect on such spurious oscillations,
so the example described here is to be viewed as a
worst case. Conservative methods represent a cautious
approach to such designs (perhaps too much so in
the present case, but definitely of interest under more
complex settings such as that of the snare drum, as
described in Section 8).
30
0
0.5
1.5
2.5
t (ms)
Figure 4. Variations in hammer force history with hammer

width. Here, the hammer interaction region g is modelled
as a rectangular window, of width as indicated, and the
string is as in the caption to Figure 3, and the hammer
strikes with a velocity of 4 m/s. The sample rate is 176
kHz.
Numerical energy conservation is illustrated in Figure 5, under lossless conditions; as in the case of the
lumped collision, numerical energy is conserved to machine accuracy (here double precision floating point
arithmetic).
As a demonstration of the risk of not using a conservative method in modeling such collisions, consider a
comparison between the results of the scheme (58) using the numerically conservative force definition from
(60), with a simpler non-conservative (and fully explicit) scheme. The simplest possible design employs
f (N)
conservative
nonconservative
20
10
0
0.5
1.5
2.5
t (ms)
Figure 6. Force history, for the system as given in the caption to Figure 3, under lossless and non-stiff conditions,
and under a hammer velocity of 1.5 m/s, using a conservative scheme, and a nonconservative scheme.
5. The Single Reed Wind Instrument

The single reed instrument is an interesting case
of an excitation mechanism incorporating distinct
nonlinearitiesone is the usual nonlinear pressure/flow relation in the mouthpiece [42], but effects
of collision of the reed with the lay also play a role.
Reed-beating effects have been modelled by various

Vol. 98 (2012)
ACUSTICA
authorsin some cases, a non-penetrative constraint

is employed [43], and in others, some penetration is
allowed [44]. A complete model requires a distributed
model of the reed itself [45, 46]; only the lumped case
will be treated in this short section. A finite difference
model of the reed/lay collision has been presented previously (see Section 9.3.1 of [17]); in that case, however, the collision was treated using ad hoc methods,
and no stability condition was available. Here, the formalism presented in the previous section, making use
of a penalty potential is employed, leading to strictly
dissipative numerical scheme.
Consider an acoustic tube, of length L, and crosssection S(x), defined again over x DL . Here, x = 0
corresponds to the mouthpiece termination, and x =
L to the bell. Lossless wave propagation in a planar
mode in the tube can be described by Websters equation [47]:
S
tt = Lw , Lw = x (Sx ) ,
c2
(70)
where and c are air density in kg/m3 and sound

speed in m/s, respectively, and where here, (x, t),
for x DL , and for t R+ , is the velocity potential
in the tube. Acoustic pressure deviation p and volume
velocity u may be derived from as
p = t , u = Sx .
(71)
In particular, pin = p(0, t), uin = u(0, t), pb = p(L, t),

ub = u(L, t) are the pressure/volume velocity pairs at
the excitation and bell termination, respectively. In
a more realistic model, viscothermal boundary layer
loss effects must be included, but will be neglected
here.
The reed is modelled as a second order oscillator in
displacement z(t), for t R+ , as
d2 z
dz
+ 2Mr r
+ Mr r2 z = f Sr p . (72)
dt2
dt
Here, Mr is reed mass, Sr is effective reed area, r a
loss parameter, and r is reed angular frequency. The
force term f is due to collision with the lay, located at
z = H, where H is the equilibrium distance between
the reed and lay, assumed constant here. It may be
written as
d d
f=
/
, = z H ,
(73)
dt dt
Mr
for some potential (), such as = K, , which

may be viewed as a penalty. p is a pressure difference
across the reed, and may be written as
p = pm pin ,
(74)
where pm is the mouth pressure supplied by the

player. Furthermore, from Bernoullis law, one has for
the flow um through the mouthpiece,
s
2|p |
um = w[]+
sign(p ) ,
(75)
10
where w is the width of the reed channel, assumed

constant here, and where as before, []+ is the positive
clipping operator as described after (2). By conservation of flow,
ur = um uin ,
(76)
where ur , the flow induced by the reed is given by

ur = Sr dz/dt
(77)
for an effective surface area Sr . This is distinct from

Sr in general, but here, for brevity, we will take
Sr = Sr (if they are distinct, all the subsequent energy and stability analysis remains unchanged if (72)
is multiplied by the scaling factor Sr /Sr indeed, recent work indicates that Sr and Sr are identical [48]).
The system as whole, for which the unknowns to be
solved for are (x, t) and z(t), must be complemented
by two initial conditions for each of these variables,
namely (x, 0), t (x, 0), z(0) and dz/dt(0). Boundary conditions for at the domain endpoints will be
specified shortly.
5.1. Energy Balance
Through an inner product of Websters equation (70)
with t , and employing integration by parts (29), an
energy balance for the acoustic tube may be written
as
dHw
= B w |L
(78)
0 = pin uin pb ub ,
dt
where

1
Hw =
k St k2DL + k Sx k2DL . (79)
2
2c
2
In this simple study, effects of radiation at the bell are
ignored, so pb = 0; they may easily be incorporated
into the analysis presented here (see Section 9.1.5 of
[17]).
Similarly, by multiplying (72) by dz/dt, and using
(73) to (77), one may arrive at an energy balance for
the reed:
dHr
= Qr Qm pin uin + pm uin ,
(80)
dt
where Hr , Qr and Qm , all non-negative, are given by
2
Mr dz
Mr r2 2
Hr =
+
z +
(81)
2
dt
2
r
2
dz
2
Qr = 2Mr r
Qm = w[]+
|p |3/2 (. 82)
dt
Combining (78) and (80) leads to the total energy

balance
dH
= Qr Qm + pm uin ,
(83)
dt
where H = Hw + Hr is the total energy, again nonnegative. Under undriven conditions, it is monotonically decreasing; under usual playing conditions (i.e.,
for pm 6= 0), there is a term corresponding to power
supply.
ACTA ACUSTICA

For a grid function nl defined over l dN , a finite
difference approximation to (70) may be written as
Sl
tt nl = lw nl , lw = x (x+ Sx+ ) ,
c2
(84)
where Sl is derived from the continuous bore profile

S(x), through sampling at locations x = lh, and using the averaging operation as defined in (37), and
where Sl = xx Sl . Pressure and velocity pairs at the
mouthpiece and bell may be defined as
pnin = t n0
pnb = t nN
unin = x S0 x n0
(85a)
n
n
ub = x+ SN x+ N , (85b)
where the operations in the + and directions for the

velocity terms indicate approximations to derivatives
normal to the endpoints [17]. The assumption of pb =
0 is enforced by setting nN = 0.
The reed system (72) is approximated as
Mr tt z n + 2Mr r t z n + Mr r2 t z n = f n Sr pn (86)
in terms of time series z n , pn , and f n , defined by
where here, the constants m, v1 and v2 are all positive,

and the values bn , an and v0n can be computed directly
given known values of the state through time step n.
As previously, under a choice of penalty potential such
as = K, , this pair of equations admits a unique
solution, as shown in Appendix A. At time step n,
once rn and pn are determined, z n+1 and n+1
, may
0
be calculated directly, along with the remaining values
of nl .
An energy balance follows as in the continuous case
as
1
t hn+ 2 = qnr qnm + pnm unin ,

1
n+ 1
(92)
n+ 1
where hn+ 2 = hr 2 + hw 2 and where

1 p
n+ 1
hw 2 =
k St+ k2dN
(93a)
2c2
+ hx+ Sx+ , et+ x+ idN

2
1
Mr
Mr r2
n+ 12
2
2
(t+ z n ) +
t+ (z n ) + n+ 2(93b)
hr
=
2
2
r
2 n 3
2
2 .
qnr = 2Mr r (t z n )
qnm = w[ n ]+
|p | (93c)

n+ 1
1
t n+ 2
fn =
, n+ 2 = t+ ( n ) ,
t
(87)
where n = z n H. Equations (74) to (76) remain

as written, for time series pnin , unm and pnm , which is assumed sampled from a given mouth pressure function
pm (t). Equation (77) may be approximated as
unr = Sr t z n .
(88)
In contrast with the case of the hammer string interaction, due to the presence of two distinct nonlinearities, namely the Bernoulli effect and collision,
there is now a pair of nonlinear equations to be solved
simultaneously. To this end, note that when evaluated
at l = 0, the scheme (84) leads to an instantaneous
relationship between pnin and unin :
pnin = cn0 + c1 unin ,
(89)
where c1 > 0 is a constant, and cn0 can be computed

from values of through time step n.
Using this relation, in conjunction with (86), (87),
(88), as well as (74), (75) and (76), when viewed as
relations among time series, one arrives at the pair of
equations:
G(rn ) gpn = 0 , rn R(pn ) = 0 ,
(90)
with rn = n+1 n1 , for a constant g > 0, and

where
m
G(rn ) = rn + bn + n ((rn + an ) (an )) (91a)
r
q
R(pn ) = v0n v1 pn v2 [ n ]+

Vol. 98 (2012)
|pn |sign(pn ) ,(91b)
Only hw 2 is of indeterminate signas in the case

of the hammer/string interaction, however, it may be
shown to be non-negative (see Section 9.1.5 of [17])
under the condition
h ck ,
(94)
which is again the Courant Friedrichs Lewy condition,

now generalized to the case of Websters equation
notice in particular that it is independent of the bore
profile S, and is thus convenient to use in practice.
5.3. Simulations
As a simple test of this algorithm, consider a clarinetlike bore profile, as illustrated in the top panel of Figure 7. In this case, the reed parameters are chosen as
Sr = 1.46 104 m2 , Mr = 3.37 106 kg, r = 1500
s1 , r = 23250 s1 , and where the equilibrium distance H of the reed from the lay is 4 104 m and
the reed channel width is w = 0.01 m. Parameters for
the acoustic field are chosen as = 1.2 kg/m3 and
c = 340 m/s. For the penalty potential, a choice of
= K, is used, with = 1.3 and K = 1013 . Both
non-beating, and beating behaviour are shown; under
the beating conditions, and at a sample rate of 88.2
kHz (this higher sample rate is used here for plot resolution), the maximum penetration of the reed into the
barrier over the duration of the simulation is 1.5108
m. As a check on the use of such a penalty potential
for a rigid collision, plots of reed displacement under
different choices of the penalty potential are shown in
Figure 8. Rather complex multiple bounce patterns
are evident, the character of which is retained regardless of the choice of the exponent in the potential.
11
bore radius (mm)

Vol. 98 (2012)
ACUSTICA
6. The String in Contact with a Distributed Rigid Barrier
50
50
0
100
200
300
400
x (mm)
500
600
y (mm)
0
0.2
0.4
335
340
345
350
t (ms)
355
360
365
340
345
350
t (ms)
355
360
365
y (mm)
0
0.2
0.4
335
Figure 7. Top: clarinet-like bore profile. Middle: steadystate oscillation of the reed position y, under a mouth
pressure of pm = 2000 Pa, for which the reed oscillates
without beating against the lay (illustrated as a grey region at y = 0.4 mm). Bottom: reed position under a
higher pressure of pm = 2500 Pa, illustrating beating effects. The sample rate is 88.2 kHz.
y (mm)
= 1.3
0.4
338
339
340
t (ms)
341
342
y (mm)
= 1.8
0.38
0.4
337
338
339
340
t (ms)
341
342
difference (mm)
s tt u = Ls u + F ,
x 10
(96)
A choice of of the form of a one-sided power law

such as = K, is a convenient choice.
Through an inner product with t u, an energy balance
L
dH

= Qs + Bs , H = Hs + Hb
(97)
dt
0
results, where Hs , Qs and Bs are as given in the case

of the hammer string interaction in (57), and where
the interaction energy Hb is given by
Z
Hb =
dx 0 ,
(98)
DL
and the system is thus strictly dissipative.

0
1
337

338
339
340
t (ms)
341
342
Figure 8. Detail of the collision interaction of the reed under beating conditions, using a penalty potential K, , for
K = 1013 and for two different values of the exponent , as
indicated, at top and middle. Bottom: difference between
the two simulation results, in mm.
A finite difference scheme, again defined over l dN ,

follows for this system as
s tt unl = ls unl + Fln ,
(99)
where the difference operator ls is as defined in (59),

and where the force density Fln is defined as
n+ 1
t l 2
n
Fl =
t ln
12
(95)
where string parameters and operator Ls are as for the

system described in Section 4, and where F represents
the interaction force/unit length with the barrier. As
before, two initial conditions for u must be supplied,
namely u(x, 0) and t u(x, 0). If the barrier is assumed
to be perfectly rigid, F may be written in terms of a
penalty potential density () as
F = t /t , = b u .
0.38
337
In the previous examples, the colliding object, though

occupying a finite region, is best considered as
lumpedthe nonlinear interaction force is a scalar
function of time alone. In some settings, the interaction force must also be modeled as distributed. A
simple example is the case of the string in contact with
a rigid barrier. Though it is possible to perform a geometric analysis for the ideal string, defined by the 1D
wave equation, in contact with a barrier of specified
shape [49, 50], for more realistic systems, time stepping methods (including not just the FD schemes described here, but also methods such as digital waveguides [19]) are probably a necessity.
Consider a stiff and lossy string, defined again over
x DL , and for t R+ , and in contact with a barrier
below, located at height b(x). A model of this system
is
n+ 12
, l
= t+ (ln ) , ln = bl unl .(100)
ACTA ACUSTICA
(102)
where 1" indicates a grid function consisting of ones.

Under lossless boundary conditions, the scheme is
stable under the same condition as previously, namely
(68).
The scheme (101) now requires the solution of nonlinear equations along the length of the string. If
rl = ln+1 ln1 , then one must solve a set of equations of the form of (17), i.e. Gl = 0, l dN , where
where ln is as defined in (63). In this case, existence and uniqueness follow immediately from the
scalar case, as the nonlinear equations to be solved
are uncoupledwhich is not the case in more complex scenarios where two distributed objects are in
contact. See Section 8.
In this case, where the string and barrier are assumed perfectly rigid, the penalty formulation allows
some spurious penetration. This may be bounded, numerically, by noting that, because the barrier poten1
n+ 1
tial hb satisfies 0 hb 2 hn+ 2 when the scheme is
stable (i.e., under condition (68)), the penetration ln
may be bounded, at all times, by
2 ( + 1) h1/2
Kh
1
+1
(105)
Thus the penetration is bounded in terms of the initial

energy, and, furthermore, can be made as small as
desired through the choice of K (within reasonsee
comments in the Conclusion). In practice, this bound
is in fact very conservativesee the comments at the
end of Section 6.2.
6.2. Simulations
The collision of a string with a rigid barrier gives rise
to a wide variety of complex phenomena. As a simple
example, consider a string vibrating against a rigid
barrier of parabolic shape, as illustrated in Figure 9.
(The parabolic shape has been used in previous studies of string collision [49, 19] as a rough approximation
to a bridge termination in various instruments such as
the sitar or tambura; an exaggerated profile has been
0.2
0.4
x (m)
t =1ms
10
0.2
0.4
x (m)
t =2ms
10
0.2
0.4
x (m)
t =3ms
0.2
0.4
x (m)
0.6
0.2
0.4
x (m)
t =2.5ms
0.6
0.2
0.4
x (m)
t =3.5ms
0.6
0.2
0.4
x (m)
0.6
10
0
10
0.6
10
0
10
0.2
0.4
x (m)
t =1.5ms
0
10
0.6
0
10
10
0
10
0
10
0.6
u (mm)
10
k
k2
, al = et ln , bl =
n , (104)
s (1+s,0 k)
1+s,0 k l
ln
(103)
with
m=
0
10
t =0.5ms
10
u (mm)
m
((rl + al ) (al )) + bl ,
Gl = rl +
rl
t =0ms
10
u (mm)
= hn+ 2 , 1idN 0 ,
u (mm)
n+ 12
hb
u (mm)
n+ 1
where hn+ 2 = hs 2 + hb 2 with qns , bns,0 and bns,N

are as in the hammer string interaction in (65), and
where
u (mm)
n+ 1
(101)
u (mm)
t hn+ 2 = qns bns,0 + bns,N ,
chosen, for illustration). In this case, the string is initialized using a triangular distribution. The potential
K, is employed here, with K = 1013 and = 1.3.
Plots of the time evolution of the string are shown
in Figure 9. Notice in particular the intermittent contact/recontact phenomena in evidence over the collision region, signaling that analysis or synthesis approaches based on a model of such a colliding string in
terms of a moving end point may pose some difficulties, due to the non-contiguity of the freely vibrating
portions of the string [49, 11].
u (mm)
The scheme satisfies an energy balance of the form

Vol. 98 (2012)
0.6
0
10
Figure 9. Snapshots of the time evolution of a vibrating

string in partial contact with a barrier of parabolic shape
(in grey). The string is considered to be non-stiff, of length
L = 0.62 m, with s = 6.3 g/m, Ts = 670 N, s,0 = 0 and
1,0 = 5 104 , and is initialized in a triangular shape
corresponding roughly to a plucked excitation. The sample
rate is 88.2 kHz, so that fine features of the interaction may
be observed.
One perceptual effect of termination against such a

smooth obstacle is an effective change in pitch with
amplitude. This is illustrated in Figure 10, showing
output spectra under triangular initial conditions of
different amplitudes. As amplitude is increased, the
spectral peaks migrate towards the higher frequencies, reflecting an effective shortening of the vibrating
portion of the string, accompanied by a broadening
of the peaks due to the nonlinear interaction. Under
lossy conditions, one may expect pitch glide phenomena to occur.
Interesting from a numerical perspective, given the
use of a penalty based approach to rigid collision as
13

Vol. 98 (2012)
ACUSTICA
40
umax = 0.1 mm
umax = 1 mm
umax = 10 mm
|
uo | (dB)
20
0
20
40
200
400
600
f (Hz)
800
1000
1200
Figure 10. Spectrum of string response, for the string collision as in Figure 9, under lossless conditions, and under
a triangular excitation of amplitude as indicated.
u (mm)
employed here are questions of the amount of penetration, and also of convergence to an exact solution
across a range of different choices of the penalty potential itself. See Figure 11, showing the preservation
of fine features of the string profile, after undergoing a
collision with the barrier for different values of ; the
difference between the two simulations is also shown,
which is small, and has an oscillatory character (as is
to be expected) concentrated near the barrier interaction region. For = 1.3, and under these conditions,
the maximum penetration of the string into the obstacle can be bounded, from (105), by ln 2.6105 m.
In fact, this bound is rather conservativefor the simulation results shown in Figure 9, the maximum penetration over all grid locations, and over the length of
the simulation is under 3 microns. It should be noted
that it is not claimed here that the existence of an exact solution to the model problem has been proved
rather, it is illustrated that differences among solutions become small, independently of the parameters
which define the collision, for sufficiently large values
of the stiffness K when a rigid barrier is considered.
=1.3
1
0
1
2
u (mm)
difference (mm)
0.1
0.2
0.3
x (m)
0.4
0.5
0.6
0.2
0.3
x (m)
0.4
0.5
0.6
=1.5
1
0
1
2
0
0.1
d2 wh
=f.
(106)
dt2
Here, w = w(x, y, t) is the transverse displacement of
a square membrane, defined for coordinates (x, y)
DL,L , and for t R+ . m is membrane density, in
kg/m2 , and g = g(x, y) represents the region of contact between the mallet
and the membrane, again norR
malized such that DL,L gdxdy = 1. wh represents the
vertical position of the mallet, of mass M . The operator Lm , incorporating effects of tension and simple
frequency-independent loss, is defined as
0
0.05
Lm = Tm 2m,0 m t ,
where here, is the Laplacian operator, defined by

=
2
2
+ 2,
2
x
y
(108)
and where Tm is the membrane tension/unit length

and m,0 is a loss parameter. The operator in (107)
may easily be extended, as in the case of the string,
to include effects of stiffness and frequency-dependent
loss; these terms will be neglected here for the sake of
brevity. Boundary conditions are assumed to be of
fixed type:
w=0
over
DL,L ,
0.1
0.2
0.3
x (m)
(109)
where DL,L is the boundary of DL,L . As in previous

cases, the system (106) must be complemented by two
initial conditions each for w and wh , namely w(x, y, 0),
t w(x, y, 0), wh (0) and dwh /dt(0).
The mallet force f acting from above is defined as:
d
f=
, = hg, wiDL,L wh ,
(110)
d
where h, iDL,L represents a 2D inner product over the
domain DL,L , and where 0 is again a one-sided
potential function, sometimes chosen [4] as a power
law nonlinearity of the form K, .
The expression for conserved energy is directly generalized from that of the hammer string system, as:
Z
dH
= Qm +
Bm , H = Hm + Hh . (111)
dt
DL,L
m
Tm
kt wk2DL,L +
kwk2DL,L
2
2

2
M dwh
Hh =
+
2
dt
Hm =
0
(107)
Here,
0.05
0.4
0.5
0.6
Figure 11. String profile, for the string collision as in Figure 9, after 2.5 ms, under a penalty potential K, with
K = 1013 , and for two different values of as indicated,
as well as the difference between the two simulations.
14
The interaction of a mallet with a membrane is a generalization to 2D of the hammer string interaction described in Section 4, and may be written as
m tt w = Lm w gf , M
60
80
7. The Mallet-Membrane Interaction
Qm = 2m,0 m kt wk2DL,L
Bm = Tm t wn w ,
(112a)
(112b)
(112c)
(112d)
where n represents a vector normal to the boundary.

Under lossless boundary conditions (such as the fixed
condition given in (109)), the system is dissipative.
ACTA ACUSTICA

Vol. 98 (2012)
7.2. Simulations
The scheme, which is a direct extension of that of the

hammer/string system, can be presented briefly here:
In this section, the results of a simulation for a square

membrane are shown, corresponding roughly to a typical drum configuration. Force histories are shown in
Figure 12 at top, illustrating again the decrease in
contact duration with mallet velocity, and at middle
and bottom, the energy partition between the mallet and membrane, as well as the normalized energy
variation are shown. The variation in the energy is
compounded by the number of degrees of freedom of
the system, which is considerably larger than in the
previous cases.
n
n
m tt wl,m
= lm wl,m
gl,m f n
M tt whn = f n , (113)
n
where wl,m
, whn and f n are approximations to w (now
defined over the 2D discrete domain dN,N ), wh and
f , respectively, and where lm is an approximation to
Lm :
lm = Tm 2m,0 m t
(114)
and the force f may be written in terms of a discrete

1
potential n+ 2 as
1
fn =
t n+ 2
, n = hg, wn idN,N whn ,(115)
t n
150
(x ,y )dN,N
where dN,N indicates the set of grid points lying on

1
the boundary of dN,N . hn+ 2 , the total numerical energy is defined as h = hm + hh , where
m
n+ 1
hm 2 =
kt+ wn k2dN,N +
(118a)
2

Tm
hx+ wn , x+ wn+1 idN ,N + hy+ wn , y+ wn+1 idN,N
2
M
1
n+ 1
2
hh 2 =
(t+ wh ) + n+ 2
(118b)
2
qnm = 2m,0 m kt wn k2dN,N
(118c)
bnm,x ,y = Tm t wnx ,y b+ wnx ,y .
mallet velocity 1 m/s

100
50
0
6
t (ms)
10
12
Energy (J)
3
membrane energy
mallet energy
2
1
0
6
t (ms)
10
12
6
t (ms)
10
12
14
x 10
Energy variation
The energy balance is now, after taking an inner

product over dN,N with t w, and using (48),
X
1
t hn+ 2 = qnm +
bnm,x ,y , (117)
f (N)
where n+ 2 = t+ ( n ).
As previously, updating scheme (113) requires the
solution of a nonlinear equation of the form (17),
where
!
kgk2dN,N
1
2
m=k
+
a = et n (116)
m (1 + m,0 k) M

kTm
2k
n
b = 2kt wh
hg, t+
wn idN,N .
1+m,0 k
2m
2
0
2
0
Figure 12. Top: Force experienced by a mallet, of mass

0.028 kg, and with stiffness parameters K = 1.6 108 and
= 2.54, striking a membrane under different velocities,
as indicated. The membrane, of dimensions 0.6 0.6 m,
with m = 0.26 kg/m2 and Tm = 3325 N/m, is struck at
a location 0.1 m from a corner. Middle: Energy partition.
Bottom: normalized energy variation. The sample rate is
22.05 kHz.
(118d)
Here, the notation b+ indicates a spatial first difference operation in the direction normal to the boundary. At corner points in the domain dN,N , it is to be
applied in both directions and summed. Under the
n
condition that wl,m
is zero at the boundary, corresponding to a fixed termination (109), the system is
again dissipative.
A stability condition for scheme (113) follows again
from the non-negativity of hm . Using bounds (45) and
(49) leads to the condition
s
2Tm
h hmin = k
,
(119)
m
which is the same as the bound obtained using frequency domain techniques for the membrane scheme
in isolation.
8. Fully Distributed Interaction

The methods described in the previous sections can
also be applied to more complex scenarios involving
distributed coupling between multiple dynamic objects. In this section, the collision of an ideal string
with a membrane is described, with an eye towards
full simulation of the snare drum, which will be briefly
outlined.
8.1. The String/Membrane Collision
Consider the system of an interacting string and membrane, defined over the regions DLs and x =
(x, y) DLm ,Lm , and with transverse displacements
15

Vol. 98 (2012)
ACUSTICA
u = u(, t) and w = w(x, t), for t R+ . The equations of motion are:

m tt w = Lm w + hg, F iDLs s tt u = Ls u F .(120)
The operators Lm and Ls have been introduced in
(107) and (51), respectively. As previously, two initial conditions must be supplied for each of u and w,
namely w(x, y, 0), t w(x, y, 0), u(, 0) and t u(, 0).
The collision force density F is again a distributed
function F = F (), and the contribution of F to
the membrane equation must be integrated along the
domain of interaction with the string with a suitable
distribution function. The simplest possible choice is
probably g = (x ()), where : DLs DLm ,Lm
is the projection from each point of the string to the
corresponding point on the membrane and is a 2D
Dirac delta function. As before, F may be related to a
potential = (()), which depends on the distance
between the string and membrane over the region
of interaction:
F=
t
, = u hg, wiDLm ,Lm .
t
(121)
Once again, one can derive an expression for energy

which is conserved in the lossless case.
A finite difference approximation for (120) and (121)
can be written as follows:
n
n
m tt wl,m
= lm wl,m
+ (ism F n )l,m
s tt unl
ls unl
Fln
(122a)
(122b)
in terms of grid functions unl , defined over dNs , and

n
wl,m
over dNm ,Nm , where Fln , may be written as
n+ 12
Fln
t l
=
t ln
ln = unl (ims wn )l .
(123)
Here, integrals have been expressed as linear operators:

hg, idNs = ism ,
hg, idNm ,Nm = ims . (124)
Ultimately, the finite difference scheme can be updated by solving a non-linear equation in the vector
r:
G(r) = r + M + b = 0 ,
(125)
with the elements of r and given by

rl = ln+1 ln1 ,
l =
et ln
(rl + al ) (al )
. (126)
rl
As before, al =
and b depend only on known
values of w and u. In contrast with the previous cases,
16
values of the solution r are now coupled by the presence of a square matrix M defined as:

k2
k2
M=
1+
Ims Ism ,(127)
s (1+s,0 k)
m (1+m,0 k)
where 1 represents the identity matrix. Ism and Ims
are the matrix forms of the operators ism and ims ,
respectively, and are required to be the transposes of
one another for energy conservation reasons (see Section 5.2.4 of [17]). Therefore, M is positive definite,
which guarantees existence and uniqueness of the solution (see Appendix A). ) The size of M is the same
as the number of points in the snare involved in the
collisions with the membrane. For example, a string of
25 cm with a typical grid spacing of 5 mm requires 51
points. In a snare drum simulation (see below), this
number must be multiplied by the total number of
snares (generally 10-20).
8.3. Boundary conditions
Boundary conditions for the membrane have been discussed in Section 7. For the string, one possibility is
to use fixed termination at both ends. In real snare
drums, however, the snares are held in contact with
the membrane by two bridges attached to the rim of
the drumhead. In a numerical simulation, this complex termination can be modelled, as a first crude
approximation, by attaching the strings directly to
the membrane. In this case, additional force terms resolved at the boundaries of the string must be added
to the equation for the membrane:
(0)
(L )
n
n
m tt wl,m
= lm wl,m
+ gl,m f (0) + gl,ms f (Ls ) (128)
+ (ism F n )l,m ,
where g (0) and g (Ls ) are suitable distribution functions (e.g., 2D Dirac delta functions) for = 0, Ls .
Energy analysis can be applied to find stable boundary conditions at = 0 (and similarly at = Ls ):
D
E
(129)
u0 = g (0) , w
dNm ,Nm
f (0) = (Ts Es Is + 2s,1 s t )u0 .
The first condition is that displacements are equal at

the snare endpoint and the point on the membrane to
which is connected, and the second that the force exerted on the membrane is equal and opposite to that
exerted on the end of the snare. Under these conditions, the scheme as a whole is dissipative, and stable
under the separate conditions (68) and (119). The expression for matrix M in (127) is slightly altered, but
remains positive definite.
8.4. Case Study: The Snare Drum
In this section, an application to the simulation of
the snare drum is briefly outlined. The snare drum
ACTA ACUSTICA

Vol. 98 (2012)
is double-headed, with the distinguishing feature of a

set of snares (metal wires) in contact with the lower
membrane, which gives this instrument its characteristic rattling sound.
Finite difference time domain simulation of the
snare drum has been presented previously [5]; in that
case, however, the collision of the snares with the
membrane was treated using ad hoc methods, and
no stability condition was available. Here, again, the
penalty potential formalism described in the previous
sections is extended to handle this more complex case,
including the interaction between the mallet and the
upper membrane, leading to a provably stable algorithm.
A simple model of the snare drum consists of two
circular membranes with fixed boundary conditions,
coupled by a rigid cavity and immersed in a box of
air, with a set of snares in contact with the lower
membrane (see Figure 13). The system can be excited through a mallet acting on the upper membrane.
Absorbing conditions are applied at the walls of the
enclosure, ideally simulating an anechoic space.
Figure 14. Snapshots of the evolution of the snare drum
system at times as indicated. Displacements have been
scaled for illustration purposes. The sample rate is 48 kHz.
10
E ne rgy (J )
The approach described above may be applied to

the mallet-membrane interaction and to the snares
colliding against the membrane, in order to obtain a
fully energy conserving model. Figure 14 shows some
snapshots of the time evolution of the system, subject
to an initial strike on the upper membrane. The delay
in the excitation caused by the air inside the cavity on the lower membrane is apparent. The snares
are launched by the contact with the lower membrane; though the movement is at first coherent, it is
rapidly randomized through multiple collisions. Figure 15 shows the contributions to total energy of the
various components, with normalized energy variations on the order of machine accuracy.
10
10
10
10
E ne rgy vari ati on
x 10
t (ms)
10
t (ms)
10
14
4
0
Figure 15. Top: energy partition among the various components of the snare drum (solid line: upper membrane,
bold dashed: mallet, dotted: lower membrane, dashed:
air, dot-dashed: snares). Bottom: normalized energy variations.
Figure 13. Diagram of snare drum geometry, illustrating
the various components. A finite enclosure with absorbing
conditions at the walls surrounds the system.
9. Concluding Remarks
This article has explored basic features of time domain numerical simulation of collision interactions
in musical instruments, with a focus on an energy-
17

Vol. 98 (2012)
ACUSTICA
conserving (and more generally passive) formulations,

within which the collision mechanism may be included
through an added potential; such a formulation allows for the stable simulation of a wide range of collision interactions in musical instruments. It is hoped
that such techniques will aid in the investigation of
finer features in musical instrument acoustics. From a
sound synthesis perspective, one issue which has not
been touched upon here is aliasingalways present in
nonlinear models, but somewhat alleviated by physical damping.
When the colliding objects are assumed to be rigid,
the potential has the interpretation of a penalty. As
has been shown at various instances, a penalty formulation of rigid collisions does lead to spurious penetration, but generally this is very small by the standard of acoustics applicationsfurthermore, there is
some degree of control over the amount of penetration, which may be conveniently bounded, provided
a numerical energetic framework is available. Under
continuously forced conditions (as in, e.g., the case of
the lip/reed mechanism), however, a bound is not immediately available; though a bound on penetration
at a given instant can surely be written in terms of
the total power supplied up until that instant, a tight
bound is not immediately apparent, and it would be
of interest to find such a bound. Another aspect worthy of further study is the particular choice of such
a penalty; a power law has been used here, for simplicity, but many others are available, perhaps with
superior properties in terms of the reduction of spurious penetration.
The formulations presented here require the solution of nonlinear equations in the main update. If
the colliding object is lumped, as in the case of the
hammer/string, mallet/membrane and reed interactions then a single equation to be solved results. For
fully distributed collisions between two deformable
objects, the solution to a system of coupled equations is required. In all cases examined here, however,
the system to be solved possesses a unique solution
see Appendix A. The Newton Raphson method has
been employed here in order to solve such nonlinear equations. One aspect of such iterative methods
which has not been addressed here is convergence,
even though a unique solution exists. The Newton
Raphson method employed here has in all cases led to
convergence (easily observed by energy conservation
to machine accuracy), but as yet its convergence remains unproven for the systems examined here. Some
partial results, however, are availablesee Appendix
A. Newton Raphson, however, is but one method of
solutionmany others are available [51]. Useful also,
here, would be bounds on the number of required iterations. This is particularly important in the case of
rigid collisionthough penetration bounds have been
repeatedly shown here, these depend on the collision
stiffness Kand indeed, there is a link between in-
18
creased stiffness and and increase in the number of

required iterations, though it is not fully clear at
present.
Though the energy-conserving framework presented
here has been applied to the case of finite difference
time domain methods, one might suspect that it applies more generally in different formulations which ultimately reduce to time stepping methods (including
time domain finite element methods, spectral methods, as well as modal techniques), though over more
general basis functions. Various references are available [52, 53] giving an overview of such families of
numerical techniques. Another interesting choice in
the interest of efficiency, particularly for synthesis
applications, might involve digital waveguides [23]
interesting recent work by Kartofelev et al. [19], dealing with the string/barrier problem indicates at least
a partial affirmative answer in the case of an ideal
non-stiff string. Such choices will form the basis of
future investigations.
Finally, it should be stated clearly that such energyconserving formulations do come at a costnamely
the use of an iterative solver; though relatively minor in the case of lumped collisions, where there is
a single nonlinear equation to be solved (or two, in
the case of the beating reed), in the distributed case,
and for rigid collisions, the expense incurred is large
leading to perhaps a 10-20 fold increase in the operation count, if a rigid collision interaction is the dominant cost. Simpler, fully explicit ad hoc methods are
available, and can perform very well, again particularly in the case of lumped collisions. In the case of
fully distributed collisions, however, it appears to be
the case that the greater the complexity of the system,
the greater the risk of instability using such methods,
with the snare drum presented here as an extreme
example. For such systems, the observation of an energy balance would appear to be at least one way of
arriving at well-behaved designs.
A. Nonlinear Equations: Existence/Uniqueness and Convergence

The solution of the nonlinear equation G(r) = 0, for
G(r) as defined in (17), as well as several variants,
plays a key role in all of the algorithms described here.
It is thus worth outlining here some properties of such
equations, especially with regard to the existence and
uniqueness of solutions, as well as the convergence of
the Newton Raphson method.
It has recently been shown [14] that the equation
G(r) = 0, for G(r) as defined in (17) has a unique solution in r. The essence of the proof is that for a differentiable and convex potential , G (r) is non-negative,
and bounded away from zero, and thus G(r) = 0 possesses a unique solution. In particular, G (r) 1. The
power law potential K, is convex and differentiable,
and thus the numerical methods here that rely on the

solution of such an equation, namely the collisions described in Sections 2, 4, 6 and 7, admit a unique update.
The Newton Raphson method can be shown to
be globally convergent, independent of the starting
point r , if G(r) is itself convex [54]. Indeed, 2,
G (r) = rm3 P (r) where P (r) = r2 (r + a) 2r (r +
a)+2(r +a)2(a). Now, P (r) = r2 (r +a) 0,
since K, is convex, hence P (r) is a monotonically
increasing function passing through the origin, which
results in G (r) 0 (which also holds in the limit
r 0). When 1 < < 2, ceases to exist for
certain values of r and a. In this case only local convergence can be guaranteed, subject to a good initial
guess r , which is ensured in practice by choosing the
previous value of r.
The case of the beating reed requires the solution
of a pair of nonlinear equations, as given in (90), with
the functions G and R as defined in (91). First, note
that R is one-to-one, thus possessing an inverse R1 ,
so system (90) may be condensed to
J(rn ) = G(rn ) gR1 (rn ) = 0 .
(130)
Note that R is negative, and bounded between v1

and . and thus (R1 ) is also negative, and
bounded between 0 and 1/v1 . Because g > 0, and
G 1, J 1, and the system possesses a unique
solution. It is not, however, convex, and a simple conclusion regarding the convergence of Newton Raphson
is not easily arrived at.
Finally, consider the vector system of nonlinear equations arising in the case of distributed/distributed interaction in (125). Because
M > 0, the system to be solved is equivalent to
G(r)
= M1 r + + M1 b = 0 .
(131)
The system possesses a unique solution if the Jacobian

is positive definite. The Jacobian is M1 + ,
of G
where is a diagonal matrix, the diagonal entries
of which are the derivatives of the components of
which are individually non-negative, and is thus positive semi-definite. Thus the Jacobian is positive definite.
Acknowledgment
This work was supported by the European Research
Council, under grant number StG-2011-279068-NESS.
References
[1] X. Boutillon: Model for piano hammers: Experimental determination and digital simulation. J. Acoust.
Soc. Am. 83 (1988) 746754.
[2] A. Chaigne, A. Askenfelt: Numerical simulations of
struck strings. I. A physical model for a struck string
using finite difference methods. J. Acoust. Soc. Am.
95 (1994) 11121118.
ACTA ACUSTICA

Vol. 98 (2012)
[3] A. Chaigne, V. Doutaut: Numerical simulations of

xylophones. I. Time domain modeling of vibrating
bars. J. Acoust. Soc. Am. 101 (1997) 539557.
[4] L. Rhaouti, A. Chaigne, P. Joly: Time-domain modeling and numerical simulation of a kettledrum. J.
Acoust. Soc. Am. 105 (1999) 35453562.
[5] S. Bilbao: Time domain simulation of the snare drum.
J. Acoust. Soc. Am. 131 (2012) 914925.
[6] C. Vyasarayani, S. Birkett, J. McPhee: Modeling
the dynamics of a vibrating string with a finite
distributed unilateral constraint: Application to the
sitar. J. Acoust. Soc. Am. 125 (2010) 36733682.
[7] G. Evangelista: Physical model of the string fret interaction. Proc. Int. Conf. Digital Audio Effects, Paris,
France, Sept. 2011, 345351.
[8] D. Marhefka, D. Orin: A compliant contact model
with nonlinear damping for simulation of robotic systems. IEEE Trans. Systems, Man and Cybernetics
Part A: Systems and Humans 29 (1999) 566572.
[9] D. Baraff: Fast contact force computation for nonpenetrating rigid bodies. Computer Graphics Proc.,
Orlando, Florida, July 1994, 2334.
[10] P. Wriggers, T. Laursen: Computational contact mechanics. Springer-Verlag Vienna, 2008, 1255.
[11] A. Krishnaswamy, J. O. Smith III: Methods for simulating string collisions with rigid spatial obstacles.
IEEE Workshop on Appl. of Siqnal Processing to Audio and Acoust., New Paltz, New York, October 2003,
233236.
[12] F. Avanzini, D. Rocchesso: Modeling collision sounds:
Nonlinear contact force. Proc. Int. Conf. on Digital
Audio Effects, Limerick, Ireland, December 2001, 61
66.
[13] S. Papetti, F. Avanzini, D. Rocchesso: Numerical
methods for a nonlinear impact model: A comparative study with closed-form corrections. IEEE Trans.
on Audio Speech and Language Processing 19 (2011)
2933.
[14] V. Chatziioannou, M. van Walstijn: An energy conserving finite difference scheme for simulation of collisions. Proc. Stockholm Musical Acoust. Conf., Stockholm, Sweden, 2013.
[15] J. Chabassier: Modeling and numerical simulation of
the piano through physical modeling. Dissertation.
Ecole Polytechnique, Paris, France, 2012.
[16] J. Chabassier, A. Chaigne, P. Joly: Modeling and simulation of a grand piano. J. Acoust. Soc. Am. 134
(2013) 648665. DOI: 10.1121/1.4809649.
[17] S. Bilbao: Numerical sound synthesis. John Wiley
and Sons, Chichester, UK, 2009, 1456.
[18] S. Siddiq: A physical model of the nonlinear sitar
string. Arch. Acoust. 37 (2012) 7379.
[19] D. Kartofelev, A. Stulov, H.-M. Lehtonen, V.
Vlimki: Modeling a vibrating string terminated
against a bridge with arbitrary geometry. Proc.
Stockholm Musical Acoust. Conf., Stockholm, Sweden, August 2013.
[20] J. Bensoam: A reciprocal variational approach to the
two-body frictionless contact problem in elastodynamics. Int. J. of Mech. Sci. 45 (2010) 15301535.
[21] T. Hughes, R. Taylor, J. Sackman, A. Curnier, W.
Kanoknukulchai: A finite element method for a class
of contact-impact problems. Comp. Meth. Appl.
Mech. Eng. 8 (1976) 249276.
19

Vol. 98 (2012)
ACUSTICA
[22] J. Strikwerda: Finite difference schemes and partial

differential equations. SIAM, Philadelphia, 2004, 1
435.
[41] R. Courant, K. Friedrichs, H. Lewy: On the partial

differential equations of mathematical physics. Mathematische Annalen 100 (1928) 3274. In German.
[23] J. O. Smith III: Physical audio signal procesing. Stanford, CA, 2004. Draft version. Available online at
http://ccrma.stanford.edu/jos/pasp04/.
[42] T. Wilson, G. Beavers: Operating modes of the clarinet. J. Acoust. Soc. Am. 64 (1974) 15661569.
[24] J.-M. Adrien: The missing link: Modal synthesis.

In: Representations of Musical Signals. G. DePoli,
A. Picialli, C. Roads (eds.). MIT Press, Cambridge,
Massachusetts, 1991, 269297.
[25] B. Bank, L. Sujbert: Generation of longitudinal vibrations in piano strings: From physics to sound synthesis. J. Acoust. Soc. Am. 117 (2005) 539557.
[26] O. Thomas, C. Touz, A. Chaigne: Non-linear vibrations of free-edge thin spherical shells: Modal interaction rules and 1:1:2 internal resonance. Int. J. Solids
and Structures 42 (2005) 33393373.
[27] D. Greenspan: Conservative numerical methods for
x
= f (x). Journal of Computational Physics 56
(1984) 2841.
[28] E. Vouga, D. Harmon, R. Tamstori, E. Grinspan:
Asynchronous variational contact mechanics. Comp.
Meth. Appl. Mech. Eng. 8 (2010) 249276.
[29] J. Simo, N. Tarnow, K. Wong: Exact energymomentum conserving algorithms for symplectic
schemes for nonlinear dynamics. Comp. Meth. Appl.
Mech. Eng. 100 (1992) 63116.
[30] D. Harmon: Robust, efficient, and accurate contact
algorithms. Dissertation. Columbia University, 2010.
[31] R. Fetecau, J. Marsden, M. Ortiz, M. West: Nonsmooth lagrangian mechanics and variational collision
integrators. SIAM J. Appl. Dyn. Sys. 2 (2003) 249
276.
[32] P. Joly, L. Rhaouti: A qualitative analysis of a simplified model for the nonlinear membrane-mallet interaction. Tech. Rept. RR-3234, INRIA, 1997.
[33] G. Horvay, A. Veluswami: Hertzian impact of two
elastic spheres in the presence of surface damping.
Acta Mechanica 35 (1980) 285290.
[43] P. Guillemain: A digital synthesis model of doublereed wind instruments. EURASIP J. Appl. Sig. Proc.
2004 (2004) 9901000.
[44] S. Karkar, C. Vergez, B. Cochelin: Numerical tools
for musical instruments acoustics: Analysing nonlinear physical models using continuation of periodic solutions. Proc. Acoustics 12, Nantes, France, April
2012.
[45] S. Stewart, W. Strong: Functional model of a simplified clarinet. J. Acoust. Soc. Am. 68 (1980) 109120.
[46] F. Avanzini, M. van Walstijn: Modeling the mechanical response of the reed-mouthpiece-lip system of a
clarinet. Part I. A one-dimensional distributed model.
Acta Acustica united with Acustica 90 (2004) 537
547.
[47] P. Morse, U. Ingard: Theoretical acoustics. Princeton
University Press, Princeton, New Jersey, 1968, 1927.
[48] M. van Walstijn, F. Avanzini: Modeling the mechanical response of the reed-mouthpiece-lip system of a
clarinet. Part II: A lumped model approximation.
Acta Acustica united with Acustica 93 (2007) 435
446.
[49] J. K. R. Burridge, C. Morshedi: The sitar string, a
vibrating string with a one-sided inelastic constraint.
SIAM J. Appl. Math. 42 (1982) 12311251.
[50] M. Schatzman: A hyperbolic problem of second order
with unilateral constraints: The vibrating string with
a concave obstacle. J. Math. Anal. Appl. 73 (1980)
138191.
[51] C. Kelley: Iterative methods for linear and nonlinear
equations. SIAM, Philadelphia, USA, 1995, 1166.
[52] G. Evans, J. Blackledge, P. Yardley: Numerical methods for partial differential equations. Springer, London, UK, 1999, 1290.
[34] H. Goldstein: Classical mechanics. Pearson Education, 2002.
[53] A. Quarteroni, A. Valli: Numerical approximation of

partial differential equations. Springer, Berlin, 1997,
1544.
[35] K. Hunt, F. Crossley: Coefficient of restitution interpreted as damping in vibroimpact. ASME J. Appl.
Mech. (June 1975) 4405.
[54] J. Ortega, W. Rheinboldt: Iterative solution of nonlinear equations in several variables. SIAM, New
York, 1970, 1572.
[36] A. Stulov: Dynamic behavior and mechanical features

of wool felt. Acta Mechanica 169 (2004) 1321.
[37] A. Chaigne, A. Askenfelt: Numerical simulations of
struck strings. II. Comparisons with measurements
and systematic exploration of some hammer-string
parameters. J. Acoust. Soc. Am. 95 (1994) 163140.
[38] J. Bensa, S. Bilbao, R. Kronland-Martinet, J. O.
Smith III: The simulation of piano string vibration:
From physical models to finite difference schemes and
digital waveguides. J. Acoust. Soc. Am. 114 (2003)
10951107.
[39] P. Ruiz: A technique for simulating the vibrations
of strings with a digital computer. Diploma Thesis.
University of Illinois, 1969.
[40] G. Borin, G. D. Poli, D. Rochesso: Elimination of
delay-free loops in discrete-time models of nonlinear
acoustic systems. IEEE Transactions on Speech and
Audio Processing 8 (2000) 597605.
20

Bilbao Acustica r2

Uploaded by

Copyright:

Available Formats

Bilbao Acustica r2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bilbao Acustica r2

Uploaded by

Copyright:

Available Formats

ACTA ACUSTICA UNITED WITH

Numerical Modeling of Collisions in Musical

S. Hirzel Verlag EAA

The perceptual effects of such a nonlinear interaction

ACTA ACUSTICA UNITED WITH

Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions

an efficient alternative, and will be employed here.

presented recently [14] and will be extended to fully

2. Prelude: Collision of a Mass with a

Here, u = u(t) is the position of the object at time t

Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions

It depends on a stiffness parameter K 0 and an

UNITED WITH ACUSTICA

2.2. Time Series and Difference Operators

and an approximation to a second time derivative as

Various averaging operators may be defined as

2.1. Energy Balance

2.3. Finite Difference Scheme

An approximation to (1), at time step n is then

and thus to the energy balance, in terms of total energy H:

The first bound holds for any non-negative potential

where tt is a second difference operator, as defined in

in terms of a discrete potential n+ 2 , itself defined

Employing the identity

ACTA ACUSTICA UNITED WITH

Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions

leads to the discrete energy balance

2.4. Nonlinear Losses

where the discrete energy of the system is defined as

and the scheme is thus unconditionally stable. Notice

in terms of an unknown r, defined in terms of un as

It is worth contrasting this energy-conserving

where is defined as in (2), and for some function

for some 0 has been employed in previous studies

where n = (un ). Now the energy balance (14) may

where m, a and b are given by

In some more refined collision models [13], a nonlinear

where r, the unknown, as well as the constants m, a

Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions

3.1. Continuous Inner Products and Norms

UNITED WITH ACUSTICA

time domain schemes in 1D and 2D is presented, with

Figure 1. Top: trajectory of a mass M = 10 g, colliding

As a further example, the behaviour of the same

where dx is a d-dimensional differential element.

3.2. Discrete Domains and Grid Functions

where is the Laplacian operator in d dimensions,

Figure 2. Top: trajectory of a mass colliding with a rigid

for some integer N such that the grid spacing h =

3. Background: Distributed Systems

and truncated domains follow as

A 2D grid function unl,m , defined over (l, m) d, for

ACTA ACUSTICA UNITED WITH

Bilbao, Torin, Chatziioannou: Numerical Modeling of Collisions

3.3. Difference Operators

and centered approximations to second and fourth

Averaging operators may be defined as

In both cases, other choices of inner products are

and first difference operators in both x and y may

3.4. Inner Products