An Introduction To Aeroacoustics

An introduction to aeroacoustics
A. Hirschberg∗ and S.W. Rienstra∗∗

Eindhoven University of Technology,
∗ Dept. of App. Physics and ∗∗ Dept. of Mathematics and Comp. Science,
P.O. Box 513, 5600 MB Eindhoven, The Netherlands.
Email: A.Hirschberg@tue.nl and S.W.Rienstra@tue.nl
18 Jul 2004
20:23 18 Jul 2004 1 version: 18-07-2004

1 Introduction
Due to the nonlinearity of the governing equations it is very difficult to predict the sound production
of fluid flows. This sound production occurs typically at high speed flows, for which nonlinear inertial
terms in the equation of motion are much larger than the viscous terms (high Reynolds numbers). As
sound production represents only a very minute fraction of the energy in the flow the direct prediction
of sound generation is very difficult. This is particularly dramatic in free space and at low subsonic
speeds. The fact that the sound field is in some sense a small perturbation of the flow, can, however,
be used to obtain approximate solutions.
Aero-acoustics provides such approximations and at the same time a definition of the acoustical
field as an extrapolation of an ideal reference flow. The difference between the actual flow and the
reference flow is identified as a source of sound. This idea was introduced by Lighthill [68, 69]
who called this an analogy. A second key idea of Lighthill [69] is the use of integral equations as a
formal solution. The sound field is obtained as a convolution of the Green’s function and the sound
source. The Green’s function is the linear response of the reference flow, used to define the acoustical
field, to an impulsive point source. A great advantage of this formulation is that random errors in
the sound source are averaged out by the integration. As the source also depends on the sound field
this expression is not yet a solution of the problem. However, under free field conditions one can
often neglect this feedback from the acoustical field to the flow. In that case the integral formulation
provides a solution.
When the flow is confined, the acoustical energy can accumulate into resonant modes. Since the
acoustical particle displacement velocity can become of the same order of magnitude as the main flow
velocity, the feedback from the acoustical field to the sound sources can be very significant. This leads
to the occurrence of self-sustained oscillations which we call whistling. In spite of the back-reaction,
the ideas of the analogy will appear to remain useful.
As linear acoustics is used to determine a suitable Green’s function, it is important to obtain basic
insight into properties of elementary solutions of the wave equation. We will focus here on the wave
equation describing the propagation of pressure perturbations in a uniform stagnant (quiescent) fluid.
While in acoustics of quiescent media it is rather indifferent whether we consider a wave equation
for the pressure or the density we will see that in aero-acoustics the choice of a different variable
corresponds to a different choice of the reference flow and hence to another analogy. It seems para-
doxical that analogies are not equivalent, since they are all reformulations of the basic equations of
fluid dynamics. The reason is that the analogy is used as an approximation. Such an approximation
is based on some intuition and usually empirical observations. An example of such an approximation
was already quoted above. In free-field conditions we often neglect the influence of the acoustical
feedback on the sound sources.
While Lighthill’s analogy is very general and useful for order of magnitude estimate, it is less
convenient when used to predict sound production by numerical simulations. One of the problems is
that the sound source deduced from Lighthill’s analogy is spatially rather extended, leading to slowly
converging integrals. For low Mach number isothermal flow we will see that aerodynamic sound
production is entirely due to mean flow velocity fluctuations, which may be described directly in terms
of the underlying vortex dynamics. This is more convenient because vorticity is in general limited to
a much smaller region in space than the corresponding velocity field (Lighthill’s sound sources). This
leads to the idea of using an irrotational flow as reference flow. The result is called Vortex Sound
20:23 18 Jul 2004 1 version: 18-07-2004

Theory. Vortex Sound Theory is not only numerically efficient, it also allows us to translate the very
efficient vortex-dynamical description of elementary flows directly into sound production properties
of these flows.
We present here only a short summary of elements of acoustics and aero-acoustics. The structure
of this chapter is inspired by the books of Dowling and Ffowcs Williams [25] and Crighton et. al [16].
A more advanced discussion is provided in text books [102, 132, 82, 35, 6, 16, 49, 47, 48, 119]. The
influence of wall vibration is discussed in among others [14, 58, 93].
In the following sections of this chapter we will consider:
– Some fluid dynamics (section 2),
– Free space acoustics (section 3),
– Aero-acoustical analogies (section 4),
– Aero-acoustics of confined flows (section 5),
2 Fluid dynamics
2.1 Mass, momentum and energy equations
We consider the motion of fluids in the continuum approximation. This means that quantities such as
the velocity v and the density ρ are smooth functions of space and time coordinates (x, t) [105, 3,
63, 126, 62, 99, 27]. We consider the fundamental equations of mass, momentum and energy applied
to an infinitesimally small fluid particle of volume V . We call this a material element. We define
the density of the material element equal to ρ, and the mass is therefore simply ρV . As the mass is
conserved, i.e.
d(ρV ) = ρdV + V dρ = 0,
the rate of change of the density, observed while moving with the fluid velocity v, is equal to minus
the dilatation rate:
1 Dρ 1 DV
=− = −∇·v
ρ Dt V Dt
where the Lagrangian time derivative Dρ/Dt is related to the Eulerian time derivative ∂ρ/∂t by:
Dρ ∂ρ
= + (v ·∇)ρ. (1)
Dt ∂t
For a cartesian coordinate system x = (x1 , x2 , x3 ) we can write this in the index notation:
Dρ ∂ρ ∂ρ ∂ρ ∂ρ ∂ρ ∂ρ
= + vi where vi = v1 + v2 + v3 . (2)
Dt ∂t ∂ xi ∂ xi ∂ x1 ∂ x2 ∂ x3
According to the convention of Einstein, the repetition of the index i implies a summation over this
dead index. Substitution of definition (2) into equation (1) yields the mass conservation law applied
to a fixed infinitesimal volume element:
∂ρ ∂ρ ∂ρvi
+ ∇·(ρv) = 0, or + = 0. (3)
∂t ∂t ∂ xi
20:23 18 Jul 2004 2 version: 18-07-2004

2.1 Mass, momentum and energy equations
We call this the conservation form of the mass equation. For convenience one can introduce a mass
source term Qm in this equation:
∂ρ ∂ρvi
+ = Qm . (4)
∂t ∂ xi
In a non-relativistic approximation such a mass source term is of course zero, and only introduced
to represent the influence on the flow of a complex phenomenon (such as combustion) within the
framework of a model that ignore the details of this process. Therefore, there is some ambiguity in the
definition of Qm . We should actually specify whether the injected mass has momentum and whether
it has a different thermodynamic state than the surrounding fluid.
In agreement with the non-relativistic approximation we apply the second law of Newton to a fluid
particle:
Dv
ρ = −∇·P + f (5)
Dt
where f is the density of the force field acting on the bulk of the fluid, while −∇·P is the net force
acting on the surface of the infinitesimal volume element. This force is expressed in terms of a stress
tensor P. Using the mass-conservation law (3) without mass source term (Qm = 0) we obtain the
momentum equation in conservation form:
∂ρv ∂ρvi ∂ρvi v j ∂Pi j
+ ∇·(P + ρvv) = f , or + =− + fi . (6)
∂t ∂t ∂x j ∂x j
The isotropic part pδi j of this tensor corresponds to the effect of the hydrodynamic pressure p =
Pii /3:
Pi j = pδi j − σi j (7)
where δi j = 0 for i = j and δi j = 1 for i = j . The deviation σi j from the hydrostatic behaviour
corresponds in a simple fluid to the effect of viscosity. We define a simple fluid as a fluid for which
σi j is symmetrical [3].
The energy equation applied to a material element is:
D
ρ (e + 12 v 2 ) = −∇· q − ∇·(P ·v) + f ·v + Q w (8)
Dt
where e is the internal energy per unit of mass, v = v, q the heat flux and Qw is the heat production
per unit of volume. In conservation form this equation becomes:
∂
ρ(e + 12 v 2 ) + ∇·[ρv(e + 12 v 2 )] = −∇· q − ∇·(P ·v) + f ·v + Q w , (9a)
∂t
or in index notation
∂ ∂ ∂qi ∂Pi j v j
ρ(e + 12 v 2 ) + ρvi (e + 12 v 2 ) = − − + f i vi + Q w . (9b)
∂t ∂ xi ∂ xi ∂ xi
The mass, momentum and energy conservation laws in differential form are only valid when the
derivatives of the flow variables are defined. When those laws are applied to a finite volume V one
20:23 18 Jul 2004 3 version: 18-07-2004

2.2 Constitutive equations
obtains integral formulations which are also valid in the presence of discontinuities such as shock
waves. For an arbitrary volume V , enclosed by a surface S with outer normal n, we have:

d
ρ dV + ρ(v − b)· n dS = 0, (10a)
dt V
S

d
ρv dV + ρv(v − b)· n dS = − P · n dS + f dV (10b)
dt V
S
S V
d
ρ(e + 12 v 2 ) dV + ρ(e + 12 v 2 )(v − b)· n dS = − q · n dS − (P ·v)· n dS + f ·v dV
dt V S S S V
(10c)
where b is the velocity of the control surface S. For a material control volume we have v · n = b· n.
For a fixed control volume we have b = 0.
The mass, momentum and energy equations (3), (6) and (9) involve much more unknowns than equa-
tions. The additional information needed to obtain a complete set of equations is provided by empirical
information in the form of constitutive equations. An excellent approximation is obtained by assum-
ing the fluid to be locally in thermodynamic equilibrium, i.e. within a material element [61]. This
implies for a homogeneous fluid that two intrinsic state variables fully determine the state of the fluid.
For acoustics it is convenient to choose the density of mass ρ and the specific entropy (i.e. per unit
of mass) s as variables. All other intrinsic state variables are function of ρ and s. Hence the specific
energy e is completely defined by a relation
e = e(ρ, s). (11)
This is what we call a thermal equation of state. This equation is determined empirically. Variations
of e may therefore be written as
∂e ∂e
de = dρ + ds. (12)
∂ρ s ∂s ρ
Comparison with the fundamental equation of thermodynamics,
de = T ds − pdρ −1 , (13)
provides thermodynamic equations for the temperature T and the pressure p:

∂e
T = (14)
∂s ρ
and: ∂e
p = ρ2 . (15)
∂ρ s
As p is also a function of ρ and s we have:

∂p ∂p
dp = dρ + ds. (16)
∂ρ s ∂s ρ
20:23 18 Jul 2004 4 version: 18-07-2004

As sound is defined as isentropic (ds = 0) pressure-density perturbations, the speed of sound c =

c(ρ, s) is defined by:
∂p
c= . (17)
∂ρ s
An extensive discussion of the speed of sound in air and water is provided by Pierce [102]. In many
applications the fluid considered is air at ambient pressure and temperature. Under such conditions
we can assume the ideal gas law to be valid:
p = ρ RT (18)
where R is the specific gas constant, the ratio R = kB /m w of the constant of Boltzmann kB and of the
mass of a molecule mw . By definition, for such an ideal gas the energy density only depends on T ,
e = e(T ), and we have:
p
c = γ = γ RT (19)
ρ
where γ = c p /cv is the Poisson ratio of the specific (i.e. per unit of mass) heat capacities at respec-
tively constant volume:
∂e
cv = (20)
∂T ρ
and constant pressure:
∂i
cp = (21)
∂T p
where i is the enthalpy per unit of mass defined by:
p
i =e+ . (22)
ρ
For an ideal gas we have cp − cv = R. An ideal gas with constant specific heats is called a perfect
gas.
As we consider local thermodynamic equilibrium, it is reasonable [61, 126] to assume that trans-
port processes are determined by linear functions of the gradients of the flow state variables. This
corresponds to a Newtonian fluid behaviour:

σi j = 2η Di j − 13 Dkk δi j + µv Dkk δi j (23)
where the rate of strain tensor Di j is defined by:
1 ∂vi ∂v j
Di j = + . (24)
2 ∂x j ∂ xi
Note that Dkk = ∇·v takes into account the effect of dilatation. In thermodynamic equilibrium,
according to the hypothesis of Stokes, one assumes that the bulk viscosity µv vanishes. The dynamic
viscosity η is a function of the thermodynamic state of the fluid. For an ideal gas η is a function of
the temperature only. While the assumption of vanishing bulk viscosity µv is initially an excellent
approximation, one observes significant effects of the bulk viscosity in acoustical applications such as
propagation over large distances [102]. This deviation from local thermodynamic equilibrium is due
20:23 18 Jul 2004 5 version: 18-07-2004

2.3 Approximations and alternative forms of the basic equations
in air to the finite relaxation time of rotational degrees of freedom of molecules. The corresponding
approximation for the heat flux q is the law of Fourier:
∂T
qi = −K (25)
∂ xi
where K is the heat conductivity. For an ideal gas, K is a function of the temperature only. It is
convenient to introduce the kinematic viscosity ν and the heat diffusivity a:
η
ν= (26)
ρ
and
K
a= . (27)
ρc p
The kinematic viscosity and the heat diffusivity are diffusion coefficients for respectively momentum
and heat transfer. For an ideal gas both transfer processes are determined by the same molecular
velocities and similar free molecular path. This explains why the Prandtl number Pr = ν/a is of
order unity. For air at ambient pressure and temperature Pr = 0.72.
Starting from the energy equation (9) and using the thermodynamic law (13) one can derive an equa-
tion for the entropy:
Ds
ρT = −∇· q + σ :∇v + Q w . (28)
Dt
If heat transfer and viscous dissipation are negligible and there are no heat sources, the entropy equa-
tion reduces to:
Ds
= 0. (29)
Dt
Hence the entropy of a material element remains constant and the flow is isentropic. When the entropy
is uniform we call the flow homentropic, so ∇s = 0. An isentropic flow originating from a reservoir
with uniform state is homentropic.
When there is no source of entropy, the sound generation is dominated by the fluctuations of the
Reynolds stress ρvi v j . Sound generation corresponds therefore often to conditions for which the term
|∂ρvi v j /∂ x j | in the momentum equation (6) is large compared to |∂σi j /∂ x j |. Assuming that both
gradients scale with the same length D while the velocity scales with U0 (a “ main flow velocity” )we
find: Re = U0 D/ν 1, where Re is the Reynolds number. In such case one can also show that
the dissipation is limited to thin boundary layers near the wall and that for time scales of the order
of U0 /D the bulk of the flow can be considered as isentropic. Note that the demonstration of this
statement in aero-acoustics has been subject of research for a long time [79, 80, 94, 137]. It is not
a trivial statement. Actually, a turbulent flow is essentially dissipative. On the time scales relevant
to sound production dissipation is negligible outside the viscous boundary layers at walls [80]. We
further often assume that heat transfer is limited to thin boundary layers at the wall and that the bulk of
the flow is essentially isothermal. We will see later, that, when the entropy of the flow is not uniform,
the convection of those inhomogeneities is an important source of sound. We have now discussed
the problem of dissipation and heat transfer in the source region. We will later consider the effect of
friction and heat transfer on wave propagation.
20:23 18 Jul 2004 6 version: 18-07-2004

In a frictionless flow the momentum equation (6) reduces to the equation of Euler:
Dv
ρ = −∇ p + f . (30)
Dt
Using the definition of enthalpy (22): i = e + p/ρ combined with the fundamental equation (13):
T ds = de + pd(1/ρ) we find:
Dv f
= −∇i + T ∇s + . (31)
Dt ρ
The acceleration Dv/Dt can be split up into an effect of the time dependence of the flow ∂v/∂t, an
acceleration in the direction of the streamlines ∇(12 v 2 ) and a Corriolis acceleration due to the rotation
ω = ∇×v of the fluid as follows:
Dv ∂v
= + ∇ 12 v 2 + ω × v. (32)
Dt ∂t
Substitution of (32) and (31) in Euler’s equation (30) yields:
∂v f
+ ∇ B = −ω × v + T ∇s + (33)
∂t ρ
where the total enthalpy or Bernoulli constant B is defined by:
B = i + 12 v 2 . (34)
In general, the flow velocity field v can be expressed in terms of a scalar potential φ and a vector
stream-function ψ:
v = ∇φ + ∇×ψ. (35)
There is an ambiguity in this definition, which may be removed by some additional condition. One
can for example impose ∇·ψ = 0. In most of the problems considered the ambiguity is removed by
boundary conditions imposed on φ and ψ. While the scalar potential φ is related to the dilatation rate
∇·v = ∇ 2 φ (36)
because ∇·(∇×ψ) = 0, the vector stream function ψ is related to the vorticity:
ω = ∇×v = ∇×(∇×ψ) (37)
because ∇×∇φ = 0. This will be used as an argument to introduce the unsteady component of the
potential velocity field as a definition for the acoustical field within the framework of Vortex Sound
Theory (see section 4.5).
For a homentropic (∇s = 0) potential flow (v = ∇φ) without external forces ( f = 0), the
momentum equation 30) can be integrated to obtain the equation of Bernoulli:
∂φ
+ B = g(t) (38)
∂t
where the function g(t) can be absorbed into the definition of the potential φ without any loss of
generality.
20:23 18 Jul 2004 7 version: 18-07-2004

In the Vortex Sound Theory (4.5), where the reference flow (acoustic field) corresponds to the
unsteady component of the potential flow ∇φ, the source is the difference between the actual flow and
the potential flow. Therefore, the sources are directly related to the vorticity ω. In a homentropic flow
the density is a function ρ = ρ( p) of the pressure p only (barotropic fluid). In such a case we can
eliminate the pressure from the equation of Euler by taking the curl of this equation (30). We obtain
an equation for the vorticity [124]:
Dω f
= ω ·∇v − ω∇·v + ∇× . (39)
Dt ρ
In the absence of external forces, the equation reduces to a purely kinematic equation. Solving
this equation yields, with ω = ∇×v, the velocity field. This approach is most effective for two-
dimensional plane flows v = (v1 (x1 , x2 ), v2 (x1 , x2 ), 0). In that case the vorticity equation reduces
to Dω3 /Dt = 0. The study of such flows provides much insight into the behaviour of vorticity near
sharp edges.
The assumed absent viscosity yields mathematically a set of equations and boundary conditions
that have no unique solution. By adding the empirically observed condition that no vorticity is pro-
duced anywhere, we have again a unique solution. This, however, is not exactly true near sharp
edges. Depending on the Reynolds number and the (dimensionless) frequency and amplitude, a cer-
tain amount of vorticity is shed from a sharp edge. For high enough Reynolds number and low enough
frequency and amplitude, the amount of shed vorticity is just enough to remove the singularity of the
potential flow around the edge. This is the so-called Kutta condition [63, 105, 99, 15, 110].
When the flow is nearly incompressible (such as in acoustical waves), we can approximate the
enthalpy by:
dp p
i= . (40)
ρ ρ0
Under these circumstances the equation of Bernoulli (38) reduces to:
∂φ 1 2 p
+ 2v + = 0. (41)
∂t ρ0
When considering acoustical waves propagating in a uniform stagnant medium we may neglect the
quadratically small term 12 v 2 , which yields the linearized equation of Bernoulli:
∂φ p
+ = 0. (42)
∂t ρ0
3 Free space acoustics of a quiescent fluid
3.1 Order of magnitudes
In acoustics one considers small perturbations of a flow. This will allow us to linearize the conser-
vation laws and constitutive equations described in the previous section (2). We will focus here on
acoustic perturbations of a uniform stagnant (quiescent) fluid. For that particular case we will now
discuss order of magnitudes of various effects. This will justify the approximations which we use
further on.
20:23 18 Jul 2004 8 version: 18-07-2004

3.1 Order of magnitudes
We will focus on the pressure perturbations p which propagate as waves and which can be de-
tected by the human ear. For harmonic pressure fluctuations the audio range is:
20 Hz ≤ f ≤ 20 kHz. (43)
The Sound Pressure Level (SPL) measured in decibel (dB) is defined by:
p
SPL = 20 log10 rms
(44)
pref
where pref = 2 × 10−5 Pa for sound propagating in gasses and pref = 10−6 Pa for propagation in other
media. The sound intensity I = I · n is defined as the time averaged energy flux associated to the
acoustic wave, propagating in direction n. The intensity level (IL) measured in decibel (dB) is given
by:
I
IL = 10 log10 (45)
Iref
where in air Iref = 10−12 Wm−2 . The reference intensity level Iref is related to the reference pressure
pref by the relationship valid for propagating plane waves:
p2
I = , (46)
ρ0 c0
because in air at ambient conditions ρ0 c0 400 kg m−2 s−1 . The time averaged power P generated
by a sound source is the flux integral of the intensity I over a surface enclosing the source. The
Sound Power Level (PWL) measured in decibel (dB) is defined by:
P
PWL = 10 log10 (47)
Pref
where Pref = 10−12 W corresponds to the power flowing through a surface of 1 m2 surface area with
an intensity I = Iref .
The threshold of hearing (for good ears) at 1 kHz is typically around SPL = 0 dB. This corre-
sponds physically to the thermal fluctuations in the flux of molecules colliding with our eardrum. In
order to detect 1 kHz we can at most integrate the signal over about 0.5 ms. At ambient conditions this
corresponds to the collision of N 1020 molecules√with our eardrum. The thermal fluctuations in the
measured pressure is therefore of the order of p0 / N = 10−5 Pa, with p0 the atmospheric pressure.
The maximum sensitivity of the ear is around 3 kHz (pitch of a police man whistle). Which is due to
the quarter-wave-length resonance of our outer ear, a channel of about 2.5 cm depth. The threshold of
pain is around SPL=140 dB. Even at such high levels we have pressure fluctuations only of the order
p / p0 = O(10−3 ). The corresponding density fluctuations are:
ρ p
= (48)
ρ0 ρ0 c02
also of the order of 10−3 , because in air ρ0 c02 / p0 = γ = c p /cv 1.4. This justifies the linearisation
of the equations. Note that in a liquid the condition for linearisation ρ /ρ0 1 does not imply
a small value of the pressure fluctuations because p / p0 = (ρ0 c02 / p0 )(ρ /ρ0 ) while ρ0 c02 p0 .
In water ρ0 c02 = 2 × 109 Pa. We should note, that when considering wave propagation over large
20:23 18 Jul 2004 9 version: 18-07-2004

3.2 Wave equation and sources of sound
distances nonlinear wave steepening will play a significant role. In a pipe this can easily result into
the formation of shock waves. This explains the occurrence of brassy sound in trombones at fortissimo
levels [38, 40]. Also in sound generated by aircraft nonlinear wave distortion significantly contributes
to the spectral distribution [16].
For a propagating acoustic plane wave the pressure fluctuations p are associated to the velocity
u of fluid particles in the direction of propagation. We will see later that:
p
u = . (49)
ρ0 c0
The amplitude δ of the fluid particle displacement is for a harmonic wave with circular frequency ω
given by: δ = |u |/ω. At f = 1 kHz the threshold of hearing (0 dB) corresponds with δ = 10−11 m.
At the threshold of pain we find δ = 10−4 m. Such small displacements also justify the use of a
linearized theory. When the acoustical displacement δ becomes of the same order of magnitude as the
radius of curvature of a wall, one will observe acoustical flow separation and the formation of vortices.
In a pipe when δ approaches the pipe cross-sectional radius one will observe acoustical streaming. At
the pipe outlet this will result into periodic vortex shedding [52, 53, 22, 101, 100, 18]. In woodwind-
musical instruments and bas-reflex ports of loudspeaker boxes this is a common phenomenon [38, 18,
123].
When deriving a wave equation in the next section, we will not only linearize the basic equations,
but we will also neglect friction and heat-transfer. This corresponds to the assumption that in an
acoustical wave, with wave-length λ = c/ f , the unsteady Reynolds number:
λ2 f |ρ ∂u |
∂t
Re unst = =O , (50)
ν ∂2u
|η 2 | ∂x
is very large. For air ν = 1.5 × 10−5 m2 s−1 so that for f = 1 kHz we find Re unst = O(107 ). We
therefore expect that viscosity only plays a role on very large distances. As the Prandtl number is of
order unity Pr = O(1) in a gas, we expect heat transfer to be also negligible. At high frequencies we
however observe a much stronger attenuation due to non-equilibrium effects (bulk-viscosity). This
results in a strong absorption of these high frequencies when we listen to aircraft at large distances.
Furthermore, in the presence of walls visco-thermal dissipation will also be much larger. The ampli-
tude of a plane wave travelling along a tube of cross-sectional radius R will attenuate exponentially
exp(−αx) with the distance x. The attenuation coefficient is given for typical audio-conditions by
[102, 125]: √
πfν γ − 1
α= 1+ √ . (51)
Rc0 Pr
In most woodwind musical instruments at low pitches the visco-thermal dissipation losses are larger
than the sound radiation power [34].
3.2 Wave equation and sources of sound
We consider the propagation of pressure perturbations p in an otherwise quiescent fluid. The pertur-
bation of the uniform constant reference state p0 , ρ0 , s0 , v 0 are defined by:
p = p − p0 , ρ = ρ − ρ0 , s = s − s0 , v = v − v0, (52)
20:23 18 Jul 2004 10 version: 18-07-2004

3.3 Green’s function and integral formulation
where for a quiescent fluid v0 = 0. We assume that f , Qw and the perturbations p / p0 , ρ /ρ0 , . . . are
small so that we can linearize the basic equations. We neglect furthermore heat transfer and viscous
effects. The equations of motion (3, 6 and 28)reduce to:
∂ρ ∂v ∂s
+ ρ0 ∇·v = 0, ρ0 + ∇ p = f , ρ0 T0 = Qw, (53)
∂t ∂t ∂t
and the constitutive equation (17) becomes:
∂p
p = c02 ρ + s. (54)
∂s ρ
Subtracting the divergence of the linearized momentum equation from the time derivative of the lin-
earized mass-conservation law yields:
∂ 2ρ
− ∇ 2 p = −∇· f . (55)
∂t 2
Combining the entropy equation with the constitutive equation yields:
∂ 2 p 2
2∂ ρ (∂ p/∂s)ρ ∂ Q w
= c + . (56)
∂t 2 0
∂t 2 ρ0 T0 ∂t
Elimination of the density fluctuations from equations (55) and (56) yields a non-homogeneous wave-
equation:
1 ∂ 2 p
− ∇ 2 p = q, (57)
c02 ∂t 2
(∂ p/∂s)ρ ∂ Q w
q= − ∇· f .
ρ0 c02 T0 ∂t
The first source term corresponds to the dilatation of the fluid as a result of heat production in processes
such as unsteady combustion or condensation. This type of sound generation mechanism have been
discussed in detail by Morfey [78] and Dowling [16]. The second term describes the sound production
by a non-uniform unsteady external force field.When considering a moving body, the reaction of the
body to the force exerted by the fluid can be represented by such a force field. An example of this is
a model of the sound radiated by a rotor calculated by concentrating the lift force of each wing into a
point force. This model will be discussed later and corresponds to the first theory of sound generation
of propellers as formulated by Gutin [36] and commonly used in many applications [9, 122].
We introduced q(x, t) as shorthand notation for the source term in the wave equation. In the
absence of a source term, q = 0, the sound field is due to initial perturbations or boundary conditions.
In the next section we present a general solution of the wave equation.
Using Green’s theorem [81] we can obtain an integral equation which includes the effects of the
sources, the boundary conditions and the initial conditions on the acoustic field. The Green’s function
G(x, t| y, τ ) is defined as the response of the flow to a impulsive point source represented by delta
functions of space and time:
1 ∂ 2G
− ∇ 2 G = δ(x − y)δ(t − τ ) (58)
c02 ∂t 2
20:23 18 Jul 2004 11 version: 18-07-2004

where δ(x − y) = δ(x1 − y1 )δ(x2 − y2 )δ(x3 − y3 ). The delta function δ(t) is not a common function
with a pointwise meaning, but a generalised function [16] formally defined by its filter property:
∞
F(t)δ(t) dt = F(0) (59)
−∞
for any well-behaving function F(t).

The definition of the Green’s function G is completed by specifying boundary conditions at a
surface S with outer normal n which encloses both the source placed at position y and the observer
placed at position x. As we consider here an acoustical phenomenon we follow Crighton’s 1 sugges-
tion to call the observer a listener. A quite general linear boundary condition is a linear relationship
between the value of the Green’s function G at the surface S and the (history of the) gradient n ·∇G
at the same point. If this relationship is a property of the surface and independent of G, we call the
surface locally reacting. Such a boundary condition is usually expressed in Fourier domain in terms of
an impedance Z(ω) of the surface S, i.e. the ratio between pressure and normal velocity component,
as follows:
i Ĝ
Z(ω) = (60)
ωρ0 n ·∇x Ĝ
where Ĝ is the Fourier transformed Green’s function defined by
∞
1
Ĝ(x, ω| y, τ ) = G(x, t| y, τ ) e−iωt dt (61)
2π −∞
and its inverse: ∞
G(x, t| y, τ ) = Ĝ(x, ω| y, τ ) e iωt dω. (62)
−∞
(Always check the sign convention in the exponential! Here, we used exp(+iωt). This is not essential
as long as the same convention is used throughout!) A problem when using Fourier analysis is that
the causality of the solution is not self-evident. We need to impose restrictions on the functional
dependence of Z and 1/Z on the frequency ω [115, 119].
Causality implies that there is no response before the pulse δ(x − y)δ(t − τ ) has been released, so
G(x, t| y, τ ) = 0 for t < τ. (63)
Consider a Green’s function G, not necessarily satisfying the actual boundary condition prevailing on
S and a source q, not necessarily vanishing before some time t0 . For the wave equation (57) we find
then the formal solution:
t

p (x, t) = q( y, τ )G(x, t| y, τ ) dVy dτ
t0 V
t
+ G(x, t| y, τ )∇y p − p ( y, t)∇y G · n dS y dτ
t0 S

1 ∂ p ∂G
+ 2 G(x, t| y, τ ) − p ( y, τ ) dVy (64)
c0 V ∂τ ∂τ τ =t0
1 Crighton D.G., private communication (1992).
20:23 18 Jul 2004 12 version: 18-07-2004

3.4 Inverse problem and uniqueness of source
where dVy = dy1 dy2 dy3 . The first integral is the convolution of the source q with the pulse response
G, the Green’s function. The second integral represents the effect of differences between the actual
physical boundary conditions on the surface S and the conditions applied to the Green’s function.
When the Green’s function satisfies the same locally reacting linear boundary conditions as the actual
field, this surface integral vanishes. In that case we say that the Green’s function is “tailored”. The
last integral represents the contribution of the initial conditions at t0 to the acoustic field. If q = 0 and
p = 0 before some time, we can choose t0 = −∞ and leave this term out.
Note that in the derivation of the integral equation (64) we have make use of the reciprocity relation
for the Green’s function [81]:
G(x, t| y, τ ) = G( y, −τ |x, −t). (65)
Due to the symmetry of the wave operator considered, the acoustical response measured in x at time
t of a source placed in y fired at time τ is equal to the response measured in y at time −τ of a source
placed in x fired at time −t. The change of sign of the time t → −τ and τ → −t is necessary
to respect causality. The reciprocity relation will be used later to determine the low frequency ap-
proximation of a tailored Green’s function. This method is extensively used by Howe [47, 48]. It
is a particularly powerful method for flow near a discontinuity at a wall. In many cases, however,
it is more convenient to use a very simple Green’s function such as the free-space Green’s function
G 0 . We will introduce this Green’s function after we have obtained some elementary solutions of the
homogeneous wave equation in free space.
3.4 Inverse problem and uniqueness of source
It can be shown that for given boundary conditions and sources q(x, t) the wave equation has a unique
solution [81]. However, different sources can produce the same acoustical field. A good audio system
is able to produce a music performance that is just as realistic as the original. Mathematically the non-
uniqueness of the source is demonstrated by the following enlightening example of Ffowcs Williams
[25]. Let us assume that p (x, t) is a solution of the non-homogeneous wave equation:
1 ∂ 2 p
− ∇ 2 p = q(x, t) (66)
c02 ∂t 2
in which q(x, t) = 0 in a limited volume V . Outside V the source vanishes, so q(x, t) = 0. As a
result, p + q = p for any x ∈/ V . However, p + q satisfies the equation:
1 ∂ 2 ( p + q) 1 ∂ 2q
− ∇ 2
( p + q) = q(x, t) + − ∇2 q (67)
c02 ∂t 2 c02 ∂t 2
which has in general a different source term than equation (66).
In order to determine the source from any measured acoustical field outside the source region, we
need a physical model of the source. This is typical of any inverse problem in which the solution is not
unique. When using microphone arrays to determine the sound sources responsible for aircraft noise
one usually assumes that the sound field is built up of so-called monopole sound sources [129]. We
will see later that the sound sources are more accurately described in terms of dipoles or quadrupoles
(section 4.1). Under such circumstances it is hazardous to extrapolate such a monopole model to
angles outside the measuring range of the microphone array or to the field from flow Mach numbers
other than used in the experiments.
20:23 18 Jul 2004 13 version: 18-07-2004

3.5 Elementary solutions of the wave equation
We consider two elementary solutions of the homogeneous wave equation (q = 0):
1 ∂ 2 p
− ∇ 2 p = 0 (68)
c02 ∂t 2
which will be used as building blocks to obtain more complex solutions:

– the plane wave
– the spherical symmetric wave.
We assume in both cases that these waves have been generated by some boundary condition or initial
condition. We consider their propagation through an in all directions infinitely large quiescent fluid,
which we call “free space”.
We first consider plane waves. These are uniform in any plane normal to the direction of propa-
gation. Let us assume that the waves propagate in the x1 -direction, in which case p = p (x1 , t) and
the wave equation reduces to:
1 ∂ 2 p ∂ 2 p
− = 0. (69)
c02 ∂t 2 ∂ x12
This 1D wave equation has the solution of d’Alembert:
x1 x1
p = F t − +G t + , (70)
c0 c0
where F represents a wave travelling in the positive x1 direction and G travels in the opposite direc-
tion. This result is easily verified by applying the chain rule for differentiation. The functions F and
G are determined by the initial and boundary conditions.
Consider for example the acoustic field generated by an infinite plane wall oscillating around
x1 = 0 with a velocity u0 (t) in the x1 direction. In linear approximation v1 (0, t) = u 0 (t), i.e. the
acoustical velocity at x1 = 0 is assumed to be equal to the wall velocity. It is furthermore implicitly
assumed in the definition of “free-space” that no waves are generated at infinity. Therefore we have
for x1 > 0 that G = 0. Using the linearized equation of motion (53) in the absence of external force
field f = 0:
∂v ∂ p
ρ0 1 = − 1 (71)
∂t ∂ x1
we find:
p = ρ0 c0 v1 . (72)
We call ρ0 c0 the specific acoustical impedance of the fluid. Using the boundary condition v1 (0, t) =
u 0 (t) and p (x1 , t) = F(t − x1 /c0 ) we find as solution:
p = ρ0 c0 u 0 (t − x1 /c0 ) (73)
for x1 > 0. This equation states that perturbations, observed at time t at position x1 , are generated at
the wall x1 = 0 at time t − x1 /c0 . The time te = t − x1 /c0 is called the emission time or retarded time.
In a similar way we find:
p = −ρ0 c0 u 0 (t + x1 /c0 ) (74)
20:23 18 Jul 2004 14 version: 18-07-2004

for x1 < 0 if the wall is of zero thickness and perturbs the fluid at either side.
By analogy of (70), we easily find for a plane-wave solution propagating in a direction given by
the unit vector n the most general form
n· x
p = F t − . (75)
c0
For the particular case of harmonic waves the plane wave solution is written in complex notation as:
p = p̂ e iωt−ik· x (76)
where k = kn is the wave vector, k = ω/c0 is the wave number and p̂ is the amplitude. The complex
notation is a shorthand notation for:
p = Re( p̂ e iωt−ik· x ) = Re( p̂) cos(ωt − k · x) − Im( p̂) sin(ωt − k· x). (77)
By means of Fourier analysis in time, an arbitrary time dependence can be represented by a sum or
integral of harmonics functions. In a similar way general spatial distributions can be developed in
terms of plane waves.
Another important elementary solution of the homogeneous wave equation (68) is the spherically
symmetric wave. In that case the pressure is only a function p (r, t) of time and the distance r to the
origin.
∂
2 ∂ F
1 ∂ 2r F
By identifying ∇2 F(r) = r12 ∂r r ∂r = r ∂r 2 , the wave equation (68) reduces for r > 0 to:
1 ∂ 2 pr ∂ 2 pr
− = 0. (78)
c02 ∂t 2 ∂r 2
Note that at r = 0 the equation is singular. As we will see this will correspond with a possible point
source. Equation (78) implies that the product pr of the pressure p and the radius r, satisfies the 1D
wave equation, and may be expressed as a solution of d’Alembert:
1
p = F(t − r/c0 ) + G(t + r/c0 ) (79)
r
in which F represents outgoing waves and G represents incoming waves. In many applications we
will assume that there are no incoming waves G = 0. We call this free-field conditions. We now focus
on the behaviour of outgoing harmonic waves:
A iωt−ikr
p = e (80)
r
where A is the amplitude and k = ω/c0 the wave number. The radial fluid particle velocity vr
associated with the wave can be calculated by using the radial component of the momentum equation
(53):
∂v ∂ p
ρ0 r = − . (81)
∂t ∂r
We find:
p i
vr = 1− . (82)
ρ0 c0 kr
20:23 18 Jul 2004 15 version: 18-07-2004

At distances r large compared to the wave length λ = 2π/k (kr = 2πr/λ 1) we find the same
behaviour as for a plane wave (72). The spatial variation due to the harmonic wave motion dominates
over the effect of the radial expansion. We call this the far-field behaviour. In contrast to this, we have
for kr 1 the near field behaviour in which the velocity vr is inversely proportional to the square
of the distance r. This is indeed the expected incompressible flow behaviour. Over small distances
the speed of sound is effectively infinite because any perturbation arrives without delay in time. As a
result, the mass flux is conserved and vr r 2 is constant. All this can be understood by the observation
that |(∂ 2 p /∂t 2 )/[c02 (∂ 2 p /∂ 2r 2 )] ∼ (kr)2 so that the wave equation reduces to the equation of Laplace
∇ 2 p = 0 for kr → 0.
Outgoing spherical symmetric waves correspond to what is commonly called a monopole sound
field. Such a field can be generated by a harmonically pulsating rigid sphere with radius a:
a = a0 + â e iωt . (83)
In linear approximation (in â/a0 ) we have:
v̂r (a0 ) = iωâ. (84)
Combining this boundary condition with equations (80) and (82) we obtain the amplitude A of the
wave:
ρ0 ω2 a0 â a0 iωt−ik(r−a0 )
p = − e . (85)
1 + ika0 r
In the low frequency limit ka0 1 we see that the amplitude of the radiated sound field decreases
with the frequency. If the volume flux V = 4π a02 vr (a0 ) = 4π ia02 ωâ, generated at the surface of the
sphere, is kept fixed the sound pressure p decreases linearly with decreasing frequency:
iωρ0 V iωt−ik(r−a0 )
p = e . (86)
4πr
A monopole field can for example be generated by unsteady combustion, which corresponds to the
entropy source term in the wave equation. This will occur in particular for a spherically symmetric
combustion. In general the monopole field will be dominant when the source region is small compared
to the acoustic wave length ka0 1. We call a region which is small compared to the wave length a
compact region. We have seen that a compact pulsating sphere is a rather inefficient source of sound
under free-field conditions. More formally, a monopole source corresponds to a localized volume
source or point source placed at position y:
∂V
q(x, t) = δ(x − y). (87)
∂t
We will discuss this approach more in detail later. Note the time derivative in the source term of
equation (87): it reflects the fact that a steady flow does not produce any sound.
Using the monopole solution (80) we can build more complex solutions. If p0 is a solution of
the wave equation (68), any spatial derivatives ∂ p0 /∂ xi are also solutions because the wave equation
has constant coefficients and the derivatives may be interchanged. A first order spatial derivative of
the monopole field is called a dipole field. Second order spatial derivatives correspond to quadrupole
fields.
20:23 18 Jul 2004 16 version: 18-07-2004

An example of a dipole field is the acoustic field generated by a rigid sphere translating harmoni-
cally in a certain direction x1 with a velocity vs = v̂s e iωt . The radial velocity vr (a0 , θ) on the surface
of the sphere is given by:
v̂r (a0 , θ) = v̂s cos θ (88)
where θ is the angle between the position vector on the sphere and the translation direction x1 . Since
we have the identity
∂r ∂ x1
= x12 + x22 + x32 = = cos θ, (89)
∂ x1 ∂ x1 r
we can write for the dipole field:
∂ e−ikr ∂ e−ikr
p̂ = A = A cos θ . (90)
∂ x1 r ∂r r
Substitution of (90) into the momentum equation (81) yields:
∂ 2 e−ikr
iωρ0v̂r = − A cos θ . (91)
∂r 2 r
We apply this equation at r = a0 . Comparison with equation (88) yields:
iωρ0v̂s a0 cos θ a 2
e−ik(r−a0 ) .
0
p̂ = (1 + ikr) (92)
2 + 2ika0 − (ka0 ) 2 r
Another example is the calculation of the field p generated by an unsteady non-uniform force
field f = ( f 1 , f 2 , f 3 ). Following equation (57) we have
1 ∂ 2 p
− ∇ 2 p = −∇· f . (93)
c02 ∂t 2
Let us assume that we have obtained a solution F1 of the wave equation in free space, thus satisfying
1 ∂ 2 F1
− ∇ 2 F1 = − f 1 . (94)
c02 ∂t 2
Then we may find the solution p of equation (93) in free space by taking the space derivative of F1
∂ F1
p = . (95)
∂ x1
This indicates that the dipole field is related to forces exerted on the flow.
Another way to deduce the relationship between forces and dipole fields is to consider the dipole
as the field obtained by placing two opposite monopole source of amplitude V at a distance y1
from each other. Taking the limit of y1 → 0 while we keep V y1 constant yields a dipole
field. As in free space changes in source position y are equivalent to changes in listener position
x, it is obvious that this limit relates to the spatial derivative of the monopole field. If we con-
sider now the two oscillating volume sources forming the dipole, there will be a mass flow V from
one source to the other. Such a unsteady mass flow is associated with an unsteady momentum flux.
This unsteady momentum flux must, following Newton, be produced by an external force acting on
the flow [105, 27]. Hence we see that a dipole is not possible without the action of a force. This
idea is illustrated in figure 1 in which we consider waves generated by a boat on the water surface.
20:23 18 Jul 2004 17 version: 18-07-2004

3.6 Acoustic energy and impedance
When a person jumps up and down in the boat, he produces an

unsteady volume injection and this generates a monopole wave
field around the boat. When two persons on the boat play with
a ball, they will exert a force on the boat each time they throw
or catch the ball. Exchanging the ball results into an oscillat-
ing force on the boat. This will make the boat translate and this
generates a dipole wave field. We could say that two individuals
fighting with each other is a reasonable model for a quadrupole.
This indicates that quadrupoles are in general much less effi-
cient in producing waves then monopoles or dipoles. This will
indeed appear to be the case.
It is often stated that Lighthill [69] has demonstrated that Figure 1: Monopole, dipole and
the sound produced by a free turbulent isentropic flow has the quadrupole generating waves on the
character of a quadrupole. A better way of putting it is that surface of the water around a boat.
since in such flows there is no net volume injection due to en-
tropy production nor any external force field, the sound field can at most be a quadrupole field [41].
Therefore, Lighthill’s statement is actually that we should ignore any monopole or dipole emerging
from a poor description of the flow. We will consider this later more in detail.
The definition of acoustical energy is not obvious when we define the acoustic field on the basis of
linearized equations. The energy is essentially quadratic in the perturbations. We may anticipate
therefore that there is some arbitrariness in the definition of acoustical energy. This problem has been
the subject of many discussions in the literature [63, 102, 35, 76, 88, 89, 90, 55]. In the particular
case of the acoustics of a quiescent fluid the approach proposed by Kirchhoff [63], starting from the
linearized equations (53), appears to be equivalent to the result obtained by expanding the energy
equation (9) up to the second order [63]. After elimination of the density by using the constitutive
equation we can write the linearized mass conservation in the form:
1 ∂ p 1 ∂ p ∂s
+ ρ 0 ∇·v = (96)
c02 ∂t c02 ∂s ρ ∂t
and the momentum equation in the form:
∂v
ρ0 + ∇ p = f . (97)
∂t
We multiply the first equation (96) by p /ρ0 and add the result to the scalar product of the second
equation (97) with v to obtain the acoustic energy equation
∂E
+ ∇· I = −D, (98)
∂t
where we have defined the acoustic energy E by:
1 1 p2
E= ρ0 v 2 + . (99)
2 2 ρ0 c02
20:23 18 Jul 2004 18 version: 18-07-2004

The intensity I, defined as

I = p v , (100)
is identified as the flux of acoustic energy. The dissipation D is the power per unit volume delivered
by the acoustical field to the sources
1 ∂ p ∂s
D=− p − f ·v . (101)
ρ0 c02 ∂s ρ ∂t
From the mass conservation law (96) we see that the source term (∂ p/∂s)ρ /(ρ0 c02 )(∂s /∂t) in the
dissipation, corresponds to the dilatation rate induced by the source. This allows us to relate the first
term in the dissipation to the work of the acoustical field due to the change in volume (dW = p dV ).
For harmonically oscillating fields p = p̂ e iωt , v = v̂ e iωt the time averaged E of the acoustic
energy is (of course) independent of time
2π/ω
ω
E = E dt, (102)
2π 0
hence the energy equation (98) reduces to:
∇·I = −D . (103)
By integration of this equation over a volume enclosing the sources we find the source power

P = − D dV = I · n dS. (104)
V S
where n is the outer normal to the control surface S. If we assume an impedance boundary condition
on the surface S:
p̂
Z(ω) = (105)
v̂ · n
we have:
I · n = 12 Re(Z)|v̂ · n|2 . (106)
We see that the real part Re(Z) of the impedance Z is associated to the transport of acoustic energy
through the surface S. The imaginary part is associated to pressure differences induced by the inertia
of the flow.
We can now easily verify by using equation (104) that the spherically symmetric wave solution
(80) satisfies the acoustic energy conservation law. The r−1 dependence of the pressure (79) in a
simple outgoing wave results into a conserved value of 4πr2 I · n.
To illustrate this we consider the impedance of a pulsating sphere of radius a0 . From equation
(85) we find for the impedance Z of the surface of the sphere:
p̂ ρ0 c0 i
Z= = 1 + . (107)
v̂r 1 + (ka1 )2 ka0
0
The real part is given by

(ka0 )2
Re(Z) = ρ0 c0 . (108)
1 + (ka0 )2
20:23 18 Jul 2004 19 version: 18-07-2004

3.7 Free space Green’s function
We see that for a large sphere ka0 1 the impedance is equal to ρ0 c0 , the impedance experienced
by a plane wave (72) of any plane control surface. For a compact sphere ka0 1 we see that
Re(Z) ρ0 c0 (ka0 )2 which implies very little energy transfer and so a very inefficient sound source.
The imaginary part Im(Z) of the impedance of the sphere, given by
ka0
Im(Z) = ρ0 c0 , (109)
1 + (ka0 )2
vanishes for ka0 → ∞. For a compact sphere, ka0 1, it corresponds to the pressure calculated
by means of the linearized equation of Bernoulli (42) if we assume an incompressible flow v̂r =
∞
iωâ(a0 /r) around the sphere. (Note that φ∞ − φ(a0 ) = a0 vr dr = iωâa0 .)
Furthermore, we note that in order to deliver acoustical energy a volume source needs to be sur-
rounded by a field of high pressure. This occurs when it is surrounded by a surface of which the
real part of the impedance is large. A force field needs a large velocity fluctuation in order to pro-
duce acoustical energy efficiently. This corresponds to a large real part of the acoustical admittance
Y = 1/Z.
3.7 Free space Green’s function
The free space Green’s function G0 is the acoustical field generated at the observer’s position x at
time t by a pulse δ(x − y)δ(t − τ ) released in y at time τ . In order to calculate the free space Green’s
function G0 we will make use of the Fourier transform (61, 62). We seek a spherically symmetric
wave solution (80) of the form
A
Ĝ 0 = e−ikr where r = x − y. (110)
r
In order to determine the amplitude A we integrate the wave equation (58) over a compact sphere of
radius a0 around y. Making use of the properties of the delta function we find:

e−iωτ ∂ Ĝ 0 2 ∂ Ĝ 0
− = (k Ĝ 0 + ∇ Ĝ 0 ) dV
2 2
∇ Ĝ 0 dV =
2
dS = 4π a0 . (111)
2π V V S ∂r ∂r r=a0
Using the near field approximation (∂Ĝ 0 /∂r)r=a0 − A/a02 we can calculate the amplitude A and we
find
1
Ĝ 0 = 2
e−iωτ −ir/c0 (112)
8π r
which leads by (generalised) inverse Fourier transformation to
1
G0 = δ(t − τ − r/c0 ). (113)
4πr
We observe at time t at a distance r from the source a pulse corresponding to the impulsion delivered
at the emission time
r
te = t − . (114)
c0
As G 0 depends only on r = x − y rather than on the individual values of x and y, the free space
Green’s function does not only satisfy the reciprocity relation (65) but also the symmetry relation:
∂ G0 ∂ G 0 ∂r ∂ G 0 ∂r ∂ G0
= =− =− . (115)
∂ xi ∂r ∂ xi ∂r ∂ yi ∂ yi
Approaching the source by the listener has the same effect as approaching the listener by the source
∂r/∂ xi = −∂r/∂ yi .
20:23 18 Jul 2004 20 version: 18-07-2004

3.8 Multipole expansion
3.8 Multipole expansion
We can use the free space Green’s function G0 to obtain a more formal definition of monopoles,
dipoles, quadrupoles, etc. As we will see, this corresponds to the use of a Taylor expansion of the free
space Green’s function. We will consider the far field p in free space of a compact source distribution
q(x, t). In order to derive the general multipole expansion we will first consider the field at a single
frequency. By using the free-field Green’s function
e−ikr
Ĝ 0 (x| y) =
4πr
we find the acoustic field for a given time-harmonic source distribution q̂(x)e iωt in a finite volume V
to be given by
e−ikr
p̂ = q̂( y)Ĝ 0 (x| y) dVy = q̂( y) dVy (116)
V V 4πr
Suppose the origin is chosen inside V . We are interested in the far field, i.e. x is large, and a
compact source, i.e. k L is small where L is the typical diameter of V . This double limit can be taken
in several ways. As we are interested in the radiation properties of the source, which corresponds with
kx ≥ O(1), we will keep k x fixed. In that case the limit of small k is the same as small y, and we
can expand in a Taylor series around y = 0

1/2
x· y y2 (x · y)2
r = x2 − 2(x · y) + y2 = x 1 − x 2 + 2x2
− 2x4
+ . . .

= x 1 − x cos θ + 12 x
2
2 sin θ + . . .
y y 2
where θ is the angle between x and y, and
e−ikx
1
3
e−ikr
= 1 + 1 + ikx x j y j + . . .
r x x2 j =1
∞ l+m+n
y1l y2m y3n ∂ e−ikr
= . (117)
l,m,n=0
l! m! n! ∂ y1l ∂ y2m ∂ y3n r y1 =y2 =y3 =0
Utilising the symmetry of r as a function of x and y, this is equivalent to

∞ −ikx
e−ikr (−1)l+m+n l m n ∂ l+m+n e
= y1 y2 y3 l m n . (118)
r l,m,n=0
l! m! n! ∂ x1 ∂ x2 ∂ x3 x
The acoustic field is then given by

∞ −ikx
1 (−1)l+m+n ∂ l+m+n e
p̂ = l m n
y1 y2 y3 q̂( y) d y l m n . (119)
4π l,m,n=0 l! m! n! V ∂ x1 ∂ x2 ∂ x3 x
As each term in the expansion is by itself a solution of the reduced wave equation, this series yields
a representation in which the source is replaced by a sum of elementary sources (monopole, dipoles,
quadrupoles, in other words, multipoles) placed at the origin ( y = 0). Expression (119) is the multi-
pole expansion of a field from a finite source in Fourier domain. From this result we can obtain the
corresponding expansion in time domain.
20:23 18 Jul 2004 21 version: 18-07-2004

3.9 Doppler effect
From the integral formulation (64) we have the acoustic field from a source q(x, t)
∞
δ(t − τ − r/c0 ) q( y, t − r/c0 )
p = q( y, τ ) d ydτ = dy (120)
−∞ V 4πr V 4πr
If the dominating frequencies in the spectrum of q(x, t) are low, such that ωL/c0 is small, we obtain
by Fourier synthesis of (119) the multipole expansion in time domain (see Goldstein [35])
∞
1 (−1)l+m+n ∂ l+m+n 1
p = y y y q( y, te ) d y
l m n
4π l,m,n=0 l! m! n! ∂ x1l ∂ x2m ∂ x3n x V 1 2 3
∞
∂ l+m+n (−1)l+m+n
= µlmn (te ) (121)
l,m,n=0
∂ x1l ∂ x2m ∂ x3n 4π x
where te = t − x/c0 is the emission time and µlmn (t) is defined by:
l m n
y1 y2 y3
µlmn (t) = q( y, t) d y. (122)
V l! m! n!
The (lmn)-th term of the expansion (121) is called a multipole of order 2l+m+n . The 20 -order term
corresponds to a monopole, a concentrated volume source at y = 0 with source strength µ000 =
V q( y, t)dVy , which is called the monopole strength.
Since each term is a function of x only, the partial derivatives to xi can be rewritten into expres-
sions containing derivatives to x. In general, these expressions are rather complicated, so we will
not try to give the general formulas here.
For very large x each multipole further simplifies because
∂ 1 µ (t ) µ(t ) x
e e l µ (te ) xl xl ∂
µ(te ) = − − − =− µ(te ). (123)
∂ xl x c0 x x x
2 c0 x x c0 x2 ∂t
This leads to
∞

x1l x2m x3n ∂ l+m+n
p µlmn (te ), (x → ∞). (124)
l,m,n=0
4π(c0 x)l+m+n x ∂t l+m+n
Most results below will be presented in this far-field approximation.
3.9 Doppler effect
We can use the Green’s function formalism to determine the effect of the movement of a source on
the radiated sound field. We consider a point source localized at the point xs (t):
q(x, t) = Q(t)δ(x − x s (t)). (125)
For free-field conditions, using equation (113), we find:

∞
Q(τ )δ( y − x s (τ )) y − x
p (x, t) = δ t −τ − dVy dτ. (126)
−∞ V 4π y − x c0
20:23 18 Jul 2004 22 version: 18-07-2004

3.9 Doppler effect
After integration over space, using the property (59) of the delta function, we obtain:
∞
Q(τ ) R
p (x, t) = δ t −τ − dτ. (127)
−∞ 4π R c0
where
R(τ, x) = x − x s (τ ), R = R.
The contributions of this integral are limited to the zeros of the argument of the δ-function. In other
words, this is an integral of the type
∞ tn +ε F(tn )
F(τ )δ(g(τ )) dτ = F(τ )δ((τ − tn ) dτ
d
g(tn )) dτ = (128)
−∞ n tn −ε n
| dτ
d
g(tn )|
where τ = tn correspond to the roots of g(τ ). In the present application we have
R(τ, x)
g(τ ) = t − τ − (129)
c0
and so
dg R ·v s dx s
= −1 + = −1 + Mr , where vs = . (130)
dτ Rc0 dτ
and Mr is the component of the source velocity vs in the direction of the listener scaled by the sound
speed c0 . We call this the relative Mach number of the source. It is positive for a source approaching
the observer and negative for a source receding the observer. It can be shown that for subsonic source
velocities |Mr | < 1, the equation g(te ) = 0 or
c0 (t − te ) = R(te , x) (131)
has a single root, which is to be identified as the emission time te . Hence, we find for the acoustic field
the Liénard-Wiechert potential [54]
Q(te )
p (x, t) = . (132)
4π R(1 − Mr )
When the source moves supersonically along a curve multiple solution te can occur. This may lead to
a focussing of the sound into certain region of space, leading to the so-called super-bang phenomenon.
The increase (when approaching) or decrease (when receding) of the amplitude is called Doppler
amplification, and the factor (1 − Mr )−1 is called Doppler factor. This Doppler factor is best known
from its occurrence in the increase or decrease of pitch of the sound experienced by the listener. For a
sound source, harmonically oscillating with frequency ω which is high compared to the typical sound
source velocity variations, the listener experiences at time t a frequency
d(ωte ) ω
= . (133)
dt 1 − Mr
The right-hand side is obtained by implicit differentiation of (131). Hence the observed frequency is
the emitted one, multiplied by the Doppler factor.
20:23 18 Jul 2004 23 version: 18-07-2004

3.9 Doppler effect
In this discussion we ignored the physical character of the source. If for example we consider
monopole source with a volume injection rate V (t) the source is given by:
∂
q(x, t) = ρ0 V (t)δ(x − x s (t)) . (134)

∂t
The corresponding sound field is:
∂ ρ0 V (te )
p (x, t) =
∂t 4π R(te , x)(1 − Mr (te ))
1 ∂ ρ0 V (te )
= . (135)
1 − Mr (te ) ∂te 4π R(te , x)(1 − Mr (te ))
Although this is for an arbitrary source V and path xs an extremely complex solution, it is interesting
to note that even for a constant volume flux V there is sound production when the source velocity vs
is non-uniform.
In a similar way we may consider the sound field generated by a point force F(t).
q(x, t) = −∇·[F(t) δ(x − x s (t))]. (136)
The produced sound field is given in a far-field approximation by:

R ∂ F
p (x, t) = · . (137)
c0 R(1 − Mr ) ∂te 4π R(1 − Mr )
Even when the source flies at constant velocity dtd x s this solution involves high powers of the Doppler
factor.
An interesting application of this theory is the sound production by a rotating blade. The blade
can be represented by a point force (mainly the lift force concentrated in a point) and a compact
moving body of constant volume Vb (the blade). The lift noise (the contribution of the lift force) can
be calculated by means of equation (137). Note that in practice this lift is not only the steady thrust
of the rotating blade but contains also the unsteady component due to interaction of the blades with
obstacles like supports or with a turbulent or non-uniform inflow. For the effect of the volume of the
blade it can be shown that it is given by the second time derivative
∂2 ρ0 Vb
p (x, t) = . (138)
∂t 2 4π R(1 − Mr )
Even though the blade volume remains constant, the displacement of air by the rotating blade induces
a sound production. This so-called thickness noise depends on a higher power of the Doppler factor
than the lift-noise. This implies that at low Mach numbers such as prevails for a ventilation fan, the
lift-noise will be dominant. At high source Mach numbers, such as aircraft propellers, the lift noise
will still dominate at take off when the required thrust is high, but at cruise conditions the thickness
noise becomes comparable, as the tip Mach number is close to unity (or even higher).
We will see in the next section that the sound produced by turbulence in a free jet has the character
of the field of a quadrupole distribution. As the vortices that produce the sound are convected with
the main flow, there will be a very significant Doppler effect at high Mach numbers. This results
into a radiation field mainly directed about the flow direction. Due to convective effects on the wave
propagation the sound is, however, deflected in the shear layers of the jet. This explains that along the
jet axis there is a so-called cone of silence [49, 35].
20:23 18 Jul 2004 24 version: 18-07-2004

3.10 Uniform mean flow, plane waves and edge diffraction
The problem of a source, observer and scattering objects moving together steadily in a uniform stag-
nant medium is the same as the problem of a fixed source, observer and objects in a uniform mean
flow. If the mean flow is in x direction and the perturbations are small and irrotational we have for
potential φ, pressure p, density ρ and velocity v the problem given by
∂ 2φ ∂ 2φ ∂ 2φ 1 ∂ ∂ 2
+ + − + U 0 φ = 0,
∂x2 ∂ y2 ∂ z2 c02 ∂t ∂x
∂ (139)
∂
p = −ρ0 + U0 φ, p = c02 ρ, v = ∇φ
∂t ∂x
where U0 , ρ0 and c0 denote the mean flow velocity, density and sound speed, respectively. We assume
in the following that |U0 | < c0 . The equation for φ is known as the convected wave equation.
3.10.1 Lorentz or Prandtl-Glauert transformation
By the following transformation (in aerodynamic context named after Prandtl and Glauert, but qua
form originally due to Lorentz)
x M U0
X= , T = βt + X, M= , β= 1 − M 2, (140)
β c0 c0
the convected wave equation may be associated to a stationary problem with solution φ(x, y, z, t) =
ψ(X, y, z, T ) satisfying
∂ 2ψ ∂ 2ψ ∂ 2ψ 1 ∂ 2ψ ρ0 ∂ ∂
+ + − = 0, p=− + U0 ψ. (141)
∂ X2 ∂ y2 ∂ z2 c02 ∂ T 2 β ∂T ∂X
For a time-harmonic field eiωt φ(x, y, z) = e iT ψ(X, y, z) or φ(x, y, z) = e iK M X ψ(X, y, z), where
= ω/β, k = ω/c0 and K = k/β, we have
∂ 2ψ ∂ 2ψ ∂ 2ψ
+ + + K 2 ψ = 0. (142)
∂ X2 ∂ y2 ∂ z2
The pressure may be obtained from ψ, but since p satisfies the convected wave equation too, we may
also associate the pressure field directly by the same transformation with a corresponding stationary
pressure field. The results are not equivalent, however, and especially when the field contains singu-
larities some care is in order. The pressure obtained directly is no more singular than the pressure of
the stationary problem, but the pressure obtained via the potential is one order more singular due to
the convected derivative. See the example below.
3.10.2 Plane waves
A plane wave (in x, y-plane) may be given by

x cos θn + y sin θn cos(θ − θn )
φi = exp −ik = exp −ikr (143)
1 + M cos θn 1 + M cos θn
20:23 18 Jul 2004 25 version: 18-07-2004

where θn is the direction of the normal to the phase plane and x = r cos θ, y = r sin θ. This is
physically not the most natural form, however, because θn is due to the mean flow not the direction of
propagation. By comparison with a point source field far away, or from the intensity vector
I = (ρ0 v + ρv 0 )(c02 ρ/ρ0 + v 0 ·v) ∼ (β 2 ∂∂x φ − ik Mφ)e x + ∂

∂y
φe y ∼ (M + cos θn )ex + sin θn e y
we can learn that θs , the direction of propagation (the direction of any shadows), is given by
M + cos θn sin θn
cos θs = , sin θs = . (144)
1 + 2M cos θn + M 2 1 + 2M cos θn + M 2
By introducing the transformed angle s
cos θs M + cos θn
cos s = = , (145)
1 − M 2 sin2 θs 1 + M cos θn
β sin θs β sin θn
sin s = = (146)
1 − M 2 sin2 θs 1 + M cos θn
and the transformed polar coordinates X = R cos , y = R sin , we obtain the plane wave

φi = exp i K M X − i K R cos( − s ) . (147)
3.10.3 Half-plane diffraction problem
By using the foregoing transformation, we obtain from the classical Sommerfeld solution for the half-
plane diffraction problem (see Jones [56]) of a plane wave (147), incident on a solid half plane along
y = 0, x < 0, the following solution (see Rienstra [110])

φ(x, y) = exp i K M X − i K R F(s ) + F( s ) (148)
where ∞
eiπ/4 2
e−it dt.
2
F(z) = √ e iz (149)
π z
and
s , s = (2K R)1/2 sin 12 ( ∓ s ). (150)
An interesting feature of this solution is the following. When we derive the corresponding pressure

p(x, y) = exp i K M X − i K R F(s ) + F( s )

e−iπ/4 M cos 12 s
2 1/2
+ √ exp i K M X − i K R sin 12 , (151)
π 1 − M cos s KR
we see immediately that the first part is just a multiple of the solution of the potential, so the second
part has to be a solution too. Furthermore, the first part is regular like φ, while this second part is
singular at the scattering edge. As the second part decays for any R → ∞, it does not describe the
incident plane wave, so it may be dropped if we do not accept the singularity in p at the edge. By
considering this solution

sin 1
pv (x, y) = exp i K M X − i K R √ 2 , (152)
KR
20:23 18 Jul 2004 26 version: 18-07-2004

θs
Figure 2: Sketch of scattered plane wave with mean flow
a bit deeper, it transpires that this solution has no continuous potential that decays to zero for large |y|
(see [57, 110]). This solution corresponds to the field of vorticity (in the form of a vortex sheet) that
is being shed from the edge. This may be more clear if we construct the corresponding potential φv
for large x, which is ω ω
φv ∼ sign(y) exp − |y| − i x , pv ∼ 0. (153)
U0 U0
In conclusion: we obtain the continuous-potential, singular solution by transforming the no-flow so-
lution in potential form, and the discontinuous-potential, regular solution from the no-flow solution in
pressure form. The difference between both is the field of the shed vortex sheet.
The assumption that just as much vorticity is shed that the pressure field is not singular anymore,
is known as the unsteady Kutta condition. Physically, the amount of vortex shedding is controlled by
the viscous boundary layer thickness compared to the acoustic wave length and the amplitude (and
the Mach number for high speed flow). These effects are not included in the present acoustic model,
therefore they have to be included by an additional edge condition, for example the Kutta condition.
As vorticity can only be shed from a trailing edge, a regular solution is only possible if M > 0. If
M < 0 the edge is a leading edge and we have to leave the singular behaviour as it is.
It may be noted that the same physical phenomenon seems to occur at a transition from a hard to
a soft wall in the presence of mean flow. Normally, at the edge there will be a singularity. When the
soft wall allows a surface wave of particular type accounting for a modulated vortex sheet along the
line surface, a Kutta condition can be applied that removes the singularity, possibly at the expense of
a a (spatial) instability [106, 114, 117]. It should be emphasized that this results from a linear model
and a too severe instability may be not acceptable in a fully nonlinear model.
4 Aero-acoustic analogies
4.1 Lighthill’s analogy
Until now we have considered the acoustic field generated in a quiescent fluid by an imposed external
force field f or by heat production Qw . We have furthermore assumed that the sources induce linear
perturbations of the reference quiescent fluid state. Lighthill [68] proposed a generalisation of this
approach to the case of an arbitrary source region surrounded by a quiescent fluid. Hence we no
20:23 18 Jul 2004 27 version: 18-07-2004

more assume that the flow in the source region is a linear perturbation of the reference state. We only
assume that the listener is surrounded by a quiescent reference fluid ( p0 , ρ0 , s0 , c0 uniformly constant
and v0 = 0) in which the small acoustic perturbations are accurately described by the homogeneous
linear wave equation (68). The key idea of Lighthill is to derive from the exact equations of mass
conservation (4) with Qm = 0 and momentum conservation (6) a non-homogeneous wave equation
which reduces to the homogeneous wave equation (68) in a region surrounding the listener.
By taking the time derivative of the mass conservation law (4) and subtracting the divergence of
the momentum equation (6) we obtain:
∂ 2ρ ∂2 ∂ fi
= (Pi j + ρvi v j ) − . (154)
∂t 2 ∂ xi ∂ x j ∂ xi
By adding the term c0−2 ∂t∂ 2 p to both sides and by making use of definition (7), i.e. σi j = pδi j − Pi j ,
2
we can write (154) as

1 ∂ 2 p ∂ 2 p ∂2 ∂ fi ∂ 2 p

− = (ρvi v j − σ i j ) − + − ρ (155)
c02 ∂t 2 ∂ xi2 ∂ xi ∂ x j ∂ xi ∂t 2 c02
in which the perturbations p and ρ are defined by:
p = p − p0 and ρ = ρ − ρ0 . (156)
This equation is called the analogy of Lighthill. Note that neither ρ /ρ0 nor p / p0 are necessarily
small in the so-called source region. In fact, equation (155) is exact. Furthermore, as we obtained
this equation by adding the term c0−2 ∂t∂ 2 p to both sides, this equation is valid for any value of the
2
velocity c0 . In fact, we could have chosen c0 = 3 × 108 ms−1 (the speed of light in vacuum), or
c0 = 1 mm/century. Of course, the equation would be quite meaningless then. By choosing for c0 and
p0 the values of the reference quiescent fluid surrounding the listener, we recover the homogeneous
wave equation (68) whenever the right-hand side of equation (155) is negligible. Hence equation
(155) is a generalisation of equation (57) which was derived for linear perturbations of a quiescent
fluid.
We should now realize that we did not introduce any approximation to equation (155), so this is
exact. Therefore, this equation does not provide any new information which was not already contained
in the equations of mass conservation (4) and of momentum (6). In fact, we have lost some informa-
tion. We started with four exact equations (4, 6) and eleven unknowns (vi , p, ρ, σi j ). We are now left
with one equation (155) and still eleven unknowns. Obviously, without additional information and
approximations we haven’t got any closer to a solution for the acoustic flow.
The first step in making Lighthill’s analogy useful was already described above. We have identi-
fied a listener around which the flow behaves like linear acoustic perturbations and is described by the
homogenous wave equation (68). This is an assumption valid in many applications. When we listen,
under normal circumstances, to a flute player we have conditions that are quite reasonably close to
these assumptions. At this stage the most important contribution of Lighthill’s analogy is that it gener-
alises (156), the equations for the fluctuations ρ and p to the entire space, even in a highly nonlinear
source region. The next steps will be that we introduce approximations to estimate the source terms,
i.e. the right-hand side of equation (155).

∂ 2

2 ( c2 − ρ ) which is a generalisation of the
p
We recognize in the right-hand side of (155) the term ∂t
0
entropy production term in equation (57).
20:23 18 Jul 2004 28 version: 18-07-2004

As shown by Morfey [78, 16] this term includes complex effects due to the convection of entropy
non-uniformities. The effect of external forces f is the same in both equations (68) and Lighthill’s
analogy (155). We have, however, now removed the condition that the force should only induce a
small perturbation to the reference state. An arbitrary force is allowed, as long as we take into account
any additional effects it may have to the other terms at the right-hand side of (155).
We observe additional terms due to the viscous stress σi j and the Reynolds stress ρvi v j . The
viscous stress σi j is induced by molecular transport of momentum while ρvi v j takes into account the
nonlinear convection of momentum. One of the key idea of Lighthill [68] is that when the entropy
term and the external forces are negligible the flow will only produce sound at high velocities, cor-
responding to high Reynolds numbers. He therefore assumed that viscous effects are negligible and
reduce the sound source to the nonlinear convective effects (∂2 ρvi v j /∂ xi ∂ x j ). It is worth noting that
a confirmation of this assumption being reasonable was provided qualitatively by Morfey [79, 80] and
Obermeier [94] only about thirty years after Lighthill’s original publication. A quantitative discussion
for noise produced by vortex pairing was provided by Verzicco [137]. An additional assumption, com-
monly used, is that feedback from the acoustic field to the source is negligible. Hence we can calculate
the source term from a numerical simulation that ignores any acoustic wave propagation and subse-
quently predict the sound production outside the flow. In extreme cases of low Mach number flow,
a locally incompressible flow simulation of the source region can be used to predict the (essentially
compressible!) sound field.
Equation (155) can be formally solved by an integral formulation of the type (64). This will have
the additional benefit to reduce the effect of random errors in the source flow on the predicted acoustic
field. One can state that such an integral formulation combined with Lighthill’s analogy allows to
obtain a maximum of information concerning the sound production for a given information about
the flow field. A spectacular example of this is Lighthill’s prediction that the power radiated to free
space by a free turbulent isothermal jet scales as the eight power U08 of the jet velocity. This result is
obtained from the formal solution for free space conditions:
∞

δ − τ − ρvi v j
r
∂ 2 t c ∂2
p (x, t) = ρvi v j 0
dVy dτ = dVy (157)
∂ xi ∂ x j −∞ V 4πr ∂ xi ∂ x j V 4πr τ =te
where r = x − y and te = t − r/c0 . Assuming that the sound is produced mainly by the large
turbulent structures with a typical length scale of the width D of the jet, we estimate the dominating
frequency to be f = U0 /D, where U0 is the jet velocity at the exit of the nozzle. Hence the ratio of
jet diameter to acoustical wave length D f /c0 = U0 /c0 = M. This implies that at low Mach numbers
we can neglect variations of the retarded time te if the source region is limited to a few pipe diameters.
As the acoustic power decreases very quickly with decreasing flow velocity we have indeed a source
volume of the order of D3 . As v ∼ U0 and ρ ∼ ρ0 , we may assume that ρvi v j ∼ ρ0U02 . From the far
field approximation ∂/∂ xi = −∂/(c0 ∂t) ∼ 2π f /c0 we find:
D
p (x, t) ∼ ρ0U02 M 2 , (158)
x
where we ignored the effect of convection on the sound production [35, 49]. The radiated power is
thus found to be
I 4π x2 p vr ∼ ρ0U03 M 5 D 2 ∼ U08 . (159)
This scaling law appears indeed to be quite accurate for subsonic isothermal free jets. Note that this
law implies that we can achieve a dramatic reduction of aircraft jet noise production by reducing the
20:23 18 Jul 2004 29 version: 18-07-2004

4.2 Curle’s formulation
flow Mach number. In order to retain the necessary thrust the jet cross section has to be increased.
This is is exactly what happened in the 60’s and 70’s when the high-bypass turbofan aeroengines
replaced the older turbofan engines without or with little by-pass flow. This is known as the “turbofan
revolution” [130, 49].
As stated by Crighton [16] and Powell [104], Lighthill’s theory is rather unique in its predicting
a physical phenomenon before experiments were accurate enough to verify it. This made Lighthill’s
analogy famous. Note that we have actually discarded any contribution from entropy fluctuations or
external forces. This means that if we use, as input for the analogy, data obtained from a numerical
simulation that includes significant viscous dissipation and spurious forces, we still would predict the
same scaling law.
As the amplitude of the acoustic pressure generated by a compact monopole and dipole would
scale in free field conditions as respectively M2 and M 3 , these spurious sources easily dominate the
predicted sound amplitude. This is one reason for which most of the direct numerical simulations of
sound production by a subsonic flow in free field conditions are carried out at high Mach numbers
(M 0.9).
In the presence of walls the sound radiation by turbulence can be dramatically enhanced. In the
next section, we will see that compact bodies will radiate a dipole sound field associated with the
force which they exert on the flow as a reaction to the hydrodynamic force of the flow applied to
them. Sharp edges are particularly efficient radiators. This corresponds to our common experience
with the production of sounds like the consonant /s/. The interaction of the sharp edges of our teeth
is essential. In free field conditions much attention has been devoted to the so-called trailing edge
noise of aircraft wings [35, 49, 6, 47]. The scaling rule for this sound field is p ∼ M 5/2 , which is
just in-between a monopole and a dipole. For confined subsonic flows at low frequencies the scaling
rules are quite different. A compact turbulent flow in an infinitely long straight duct will produce an
acoustic field, that scales according to p ∼ M 6 [32, 73, 119]. The sound field from a monopole and
an axially aligned dipole will both scale with M2 , while transversal dipoles will not radiate any sound.
We will consider elements of the acoustics and aero-acoustics of confined flows in section (5). The
scaling rules for supersonic free jets is obscured by the temperature difference between the flow and
the environment [131]. Globally, however, one expects a power that is limited by a U03 -law, because
an extrapolation of the scaling rule ∼ U08 would imply that the acoustical power generated by the flow
would soon become larger than the kinetic energy flux of the flow, which scales as 12 ρ0U03 . At high
Mach numbers, the theory has to be modified in order to take intrinsic temperature differences into
account.
The integral formulation of Lighthill’s analogy can be generalised for flows in the presence of wall’s.
We will later discuss the use of tailored Green’s functions (section 5.3). We now consider the approach
of Curle [17]. We use the free space Green’s function G0 (114). We use instead of the pressure p
as aeroacoustical variable the density ρ . Subtracting from both sides of equation (154) the term
c02 (∂ 2 ρ /∂ xi2 ) we obtain the analogy of Lighthill for the density:
∂ 2ρ 2
2∂ ρ ∂ 2 Ti j ∂ fi
− c = − (160)
∂t 2 0
∂ xi
2 ∂ xi ∂ x j ∂ xi
20:23 18 Jul 2004 30 version: 18-07-2004

in which the stress tensor of Lighthill Ti j is defined by:
Ti j = Pi j + ρvi v j − c02 ρδi j . (161)
We further assume that f = 0 and focus on the other sound sources. We have selected here the density
as the dependent variable, because this was Lighthill’s [68] original choice. We will later discuss the
implications of this choice (section 4.4).
We consider a fixed surface S with outer normal n and we apply Green’s theorem (64) to the
volume V outside of S. Note that n is chosen towards the interior of V , so the sign convention of
n is opposite to the sign convention used in (64). By means of partial integration and utilising the
symmetry properties ∂ G0 /∂ xi = −∂ G 0 /∂ yi (120) and ∂ G0 /∂τ = −∂ G 0 /∂t of the Green’s function
G 0 , we obtain [35, 119]:

∂2 Ti j ∂ ρvi
p (x, t) = c02 ρ (x, t) = dVy + n i dS
∂ xi ∂ x j V 4πr τ =te ∂t S 4πr τ =te

∂ Pi j + ρvi v j
− n i dS (162)
∂x j S 4πr τ =te
which is clearly a generalisation of (157). In this equation we used the assumption that at the listener’s
position p = c02 ρ . In the far-field approximation (123) we find:

xi x j ∂2 Ti j 1 ∂ ρvi
p (x, t) dVy + n i dS
4π x2 c02 ∂t 2 V r τ =te
4π ∂t
S r τ =te

xj ∂ Pi j + ρvi v j
+ n i dS. (163)
4π xc0 ∂t S r τ =te
For a compact body we can neglect the variations of te over the surface and we can write r = x if
we chose the origin y = 0 inside or near the body. Assuming that Ti j decays fast enough, we have in
that case:

xi x j ∂2 1 ∂
p (x, t) 2 ∂t 2
Ti j τ =tedVy + ρvi τ =ten i dS
4π x c0 3
V 4π x ∂t S

xj ∂
+ Pi j + ρv i v j n i dS. (164)
4π x2 ∂t S τ =te
and te = t − x/c0 . The second integral corresponds to the monopole sound field generated by the
mass-flux through the surface S. The third integral corresponds to the dipole sound field generated by
−Fj of the surface to the surrounding fluid. This is the reaction of the surface
the instantaneous force
to the force Fj = − S (Pi j + ρvi v j )n i dSof the flow on the surface. This result is a generalisation of
Gutin’s principle [36, 35].
Using this theory we understand easily that a rotor blade moving in a non-uniform flow field will
generate sound due to the unsteady hydrodynamic forces on the blade. At low Mach number this will
easily dominate the Doppler effect due to the rotation. Wind rotors placed downwind of the supporting
mast are cheap because they are hydrodynamically stable. There is no need for a feedback system to
keep them in the wind. However, the interaction of the wake of the mast with the rotor blades causes
dramatic noise problems [50, 138].
20:23 18 Jul 2004 31 version: 18-07-2004

4.3 Ffowcs Williams-Hawkings formulation
4.3 Ffowcs Williams-Hawkings formulation
While Curle’s formulation discussed in the previous section assumes a fixed control surface S, the
formulation of Ffowcs Williams and Hawkings allows the use of a moving control surface S(t). The
key idea is to include the effect of the surface in the differential equation (160) [33, 35, 16]. This is
achieved by a most elegant and efficient utilisation of generalised functions (viz. so-called surface
distributions).
We assume that the volume B(t) enclosed by the surface S(t) and this surface are sufficiently
smooth to allow the definition of a smooth function h(x, t) such that

h(x, t) < 0 if x ∈ B(t)  
h(x, t) = 0 if x ∈ S(t) (165)


h(x, t) > 0 outside B(t).
Consider any physical quantity, like ρ , defined outside B(t), and extend its definition to all space by
giving it a value equal to zero inside B(t). This is efficiently done by assuming ρ smoothly defined
everywhere. Then by multiplying it by the Heaviside function H (h) we create a new variable ρ H (h)
which vanishes within the volume B(t) (where H (h) ≡ 0) and is equal to ρ outside B(t) (where
H (h) ≡ 1). The next step will be to extend the prevailing equations to the whole space by adding
suitable surface sources. To achieve this we need the normal n to the surface S(t), given by

∇h
n= . (166)
∇h h=0
We assume that the surface S(t) is parameterized2 in time and space by the coordinates (t; λ, µ). A
point x s (t) ∈ S(t) with fixed parameters λ and µ is moving with the velocity b. Hence we have
h(x s , t) = 0 and
∂h
= −b ·∇h = −(b· n)∇h. (167)
∂t
After multiplying the mass conservation equation (4) and the momentum equation (6) by H (h) and
reordering terms, we obtain the following equations valid everywhere
∂
[ρ H ] + ∇·[ρv H ] = [ρ0 b + ρ(v − b)] ·∇ H, (168a)
∂t
∂
[ρv H ] + ∇· (P + ρvv)H = [P + ρv(v − b)] ·∇ H, (168b)
∂t
where H stands for H (h). As ∇ H = δ(h)∇h, the equations can be interpreted as generalisations of
the mass and momentum equations with surface sources at S. Using the above relations and following
Lighthill’s procedure for acoustic variable p = p − p0 , we find:

1 ∂2 2
∂ 2 p
[ p H ] − ∇ [ p H ] = ∇· ∇·[(ρvv − σ )H ] − ∇·[ f H ] + 2 −ρ H
c02 ∂t 2 ∂t c02
∂

+ ρ0 b + ρ(v − b) ·∇ H − ∇· p I − σ + ρv(v − b) ·∇ H , (169)
∂t
2 When S(t) is the surface of a solid and undeformable body, it is natural to assume a spatial parametrisation which is
materially attached to the surface. Like the auxiliary function h, this parametrisation is not unique.
20:23 18 Jul 2004 32 version: 18-07-2004

4.4 Choice of aero-acoustic variable
where (I)i j = δi j and p0 is the uniform reference value of the pressure. Note that ∇·(∇·( p0 I H )) =
∇·( p0 I ·∇ H ). For a solid surface v · n = b· n. In that case by applying Green’s theorem and using
the free-space Green’s function we find:

∂2 (ρvi v j − σi j )H ∂ fH
p (x, t) = dVy − dVy
∂ xi ∂ x j R3 4πr τ =te
∂ xi R3 4πr τ =te
2
∂2 ( p /c0 − ρ )H ∂ ρ0 b· n
+ 2 dVy + dS
∂t R3 4πr τ =te
∂t S(te ) 4πr(1 − Mr ) τ =te

∂ p n i − σi j n j
− dS, (170)
∂ xi S(te ) 4πr(1 − Mr ) τ =te
where r = x − y and Mr = b·(x − y)/rc0 and we used the following generalisations of equation
(128)

g(x)
g(x)δ(h(x)) dx = dS, (171a)
R3
|∇h|
S
g ·∇h
g(x) ·∇ H (h(x)) dx = 3 (g ·∇h)δ(h) dx = dS = g(x)·n(x) dS. (171b)
R R S |∇h|
3
S
The first three integrals correspond to the contribution of the flow around the surface S(t) while the
last two integrals represent generalisations of the thickness noise and sound generated by the surface
forces which we have discussed earlier. A reduced form, widely used for subsonic propeller and fan
noise when volume sources and surface stresses are negligible, is thus [31]

1 ∂ ρ0 b· n 1 ∂ p n i
p (x, t) = dS − dS. (172)
4π ∂t S(te ) r(1 − Mr ) τ =te 4π ∂ xi S(te ) r(1 − Mr ) τ =te
In the previous discussion we used p as the dependent aero-acoustical variable to introduce the anal-
ogy of Lighthill (section 4.1). We then used ρ to introduce the formulation of Curle describing the
effects of stationary boundaries (section 4.2). Finally, for the formulation of Ffowcs Williams and
Hawking, describing the effect of moving boundaries, we retuned to p . In acoustics this would have
been indifferent as the two variables are related by the equation of state p = c02 ρ . As we assume
linear acoustics of quiescent fluids to be valid around the listener, we made use of this relationship
in equation (162). Actually, in aero-acoustics there is a subtle difference which appears when we
compare the source terms of the two wave equations:
1 ∂ 2 p ∂ 2 p ∂2 ∂ fi ∂ 2 p

− = (ρvi v j − σ i j ) − + − ρ (155)
c02 ∂t 2 ∂ xi2 ∂ xi ∂ x j ∂ xi ∂t 2 c02
and:
∂ 2ρ 2
2∂ ρ ∂2 ∂ fi ∂2
− c = (ρvi v j − σ i j ) − + ( p − c02 ρ ). (173)
∂t 2 0
∂ xi
2 ∂ x i ∂ x j ∂ x i ∂ xi
2
Without further approximation, these two forms are both equivalent. However, considered as an
analogy the right-hand sides are assumed to be known and act as given source distribution. In that
case we see that when p is used as the aero-acoustical variable the effect of entropy fluctuations
20:23 18 Jul 2004 33 version: 18-07-2004

∂ 2 (( p /c02 ) − ρ )/∂t 2 has the character of a monopole sound source. On the other hand, when ρ
is used the apparently same effects produce a quadrupole distribution ∂2 ( p − c02 ρ )/∂ xi2 which is
qualitatively different. Of course, there is no difference if we consider the exact equations, but if we
do not introduce any approximation the analogy is just a reformulation of basic equations without
much use. Clearly, we have to be careful in selecting the aero-acoustic variable.
When considering sound production by subsonic flames we should choose p as aero-acoustical
variable because most of the sound is produced by the volume changes associated with the combus-
tion. When we neglect convection effects we have exactly the source term ∂2 (( p /c02 ) − ρ )/∂t 2 . As
shown by Morfey [78, 16] this term includes complex effects due to the convection of entropy non-
uniformities. They become explicit when we use the equation of state (15) as applied to a material
element:
Dp Dρ ∂ p Ds
= c2 + (174)
Dt Dt ∂s ρ Dt
in combination with (12):
Ds De D 1
T = +p . (175)
Dt Dt Dt ρ
Morfey [78] obtains the following result:

∂ 2 ρe ∂ c2 ρe Dρ ρ 2 ∂ T Ds
− 2 = −1+ + 2 + ∇·(ρe v) (176)
∂t ∂t c02 ρ Dt c0 ∂ρ s Dt
where the excess density ρe is defined by ρe = ρ − ( p /c02 ). The first term vanishes in a subsonic free
jet of an ideal gas with constant heat capacity. The second term is the entropy production (combustion)
term corrected for convective effects. The last term corresponds to the force exerted by a patch of
fluid with a different density on it surroundings in an accelerating flow. This may be compared with a
buoyancy (Archimedes) effect. It is induced by the fact that due to the difference of density between
the particle and its surroundings, the pressure gradient imposed by the surroundings of a particle does
not match its acceleration.
The question arises whether ρ could be a useful choice too. If we consider a turbulent isentropic
flow in a region with a speed of sound c which differs strongly from the speed of sound c0 at the
listener position, the source term can be rewritten as ∂2 p (1 − (c0 /c)2 )/∂ xi2 . We expect in a subsonic
turbulent flow the local pressure fluctuations p to scale with 12 ρU02 . Hence the analogy indicates that
when the speed of sound c0 at the listener position is much higher than the speed of sound c in the
source region, there will be a strong enhancement of the sound production compared to a similar flow
in a uniform fluid. Such a spectacular effect does indeed occur when we consider a flow in a water-air
mixture such as obtained by directing the nozzle of the shower towards the surface of the water. Due
to air entrainment there will be a volume fraction β of air in the water flow. The density of the mixture
will be:
ρ = (1 − β)ρwater + βρair . (177)
If we assume a quasi-static response of the bubbles and neglect dissolution of air in the water, the
compressibility 1/(ρc2 ) of the mixture will be the sum of the compressibilities of both phases [141,
16]:
1 1−β β
= + . (178)
ρc 2 ρwater cwater
2
ρair cair
2
20:23 18 Jul 2004 34 version: 18-07-2004

4.5 Vortex Sound
In the case of water-air mixtures for not too small nor too large values of β the density is mainly
determined by the water phase, while the air determines the compressibility. Hence we find for the
speed of sound:
ρair cair
2
c2 . (179)
β(1 − β)ρwater
Typical values are β = 0.5, ρair cair
2
= 1.4 × 105 Pa , ρwater = 103 kg m−3 and cwater = 1.5 × 103 ms−1 .
−1
We find c 2 × 10 ms and if we have our head underwater (c0 /c)2 5 × 103 . This should result
1
into an enhancement in SPL of the order of 60 dB. Indeed the flow is much noisier than when we
avoid air entrainment by putting the shower nozzle underwater.
Other choices of aero-acoustic variables lead to different analogies. In many cases such analogies
tend to avoid the problem induced by the fact that the analogy of Lighthill does not distinguish between
propagation and production of sound waves in a strongly non-uniform flow, which induces refraction.
This becomes very important in supersonic flow. In such cases the source is not compact. Simple
results like equation (161) are not valid any more. Alternative analogies which attempt to overcome
such problems are described in the literature [35, 49, 70].
4.5 Vortex Sound
One of the drawbacks of the analogy of Lighthill (155) is that the source term is spatially quite ex-
tended. As observed by Powell [103] the sound production in subsonic homentropic flows is associ-
ated to the dynamics of vortices. Vorticity ω = ∇×v appears to be spatially less extended than the
Reynolds stress ρvi v j . The reason for this is that around vortices there is a large region of potential
flow which actually does not produce any sound. In its original form the Vortex Sound Theory [103]
was applied to free-field conditions at low Mach numbers. It was only a special form of Lighthill’s
analogy stressing the role of vorticity. In the case of free field conditions the Vortex Sound Theory
enhances very much the predicted sound field. Various modifications of the theory of Powell have
been proposed which impose more explicitly the conservation of momentum and energy to the flow
in the source region [75, 41, 127]. This also improves the performance of the theory.
Howe [44, 47, 48] has generalised the theory of Powell [103] to allow its application to confined
flows and conditions in which the listener is placed into a potential flow rather than a quiescent fluid.
In this theory the fluctuations of the total enthalpy B = ( p /ρ0 ) + v v0 appears as a natural aero-
acoustical variable [44, 23, 24, 86]. In its general form the Vortex Sound Theory can be applied to
arbitrary Mach numbers. A similar analogy was derived by Möhring [76] on the basis of acoustical
energy considerations. However, such analogies become quite obscure. They do not provide much
intuitive insights. They can only be used numerically as proposed by Schröder [30].
We consider now the case of low subsonic flows in which Howe [46] proposed a very nice energy
corollary which provides much insight into the role of vorticity in sound production. We propose here
an intuitive approach to this theory. The key idea of the theory is that a potential flow is silent. This
is illustrated in figure 3 in which we consider a sketch of our vocal folds. The oscillation of the vocal
folds result into a variable volume flow from our lungs into the vocal tract. This variable volume
flux is the source of sound. Seeking a simplified model for the flow through the glottis, we consider
the Reynolds number Re and the Strouhal number St . For a typical flow velocity U0 = 30 ms−1 ,
a length scale of the glottis of the order of L = 2 mm , a frequency f = 102 Hz and a kinematic
viscosity ν = 1.5 × 10−5 m2 s−1 we have: Re = U0 L/ν = O(103 ) and St = f L/U0 = O(10−2 ). A
20:23 18 Jul 2004 35 version: 18-07-2004

4.5 Vortex Sound
quasi-steady frictionless approximation seems promising. As the Mach number is low M = O(10−1 )
and the flow is compact L/λ = MSt = O(10−3 ) we assume that the flow is locally incompressible.
Under such circumstances the equation of Bernoulli in the form p + 12 ρ0 v 2 = pt with pt a constant,
should be valid. About one diameter upstream and downstream of the vocal folds we expect a uniform
flow velocity. Assuming the flow channel upstream of the vocal folds (trachea) to have the same cross
section as the channel downstream (vocal tract) we find that the velocities should be equal. As a
consequence we conclude that there is no pressure difference
across the vocal folds. The volume flux is therefore not con-
trolled by the opening of the folds! This implies that they
cannot produce any sound. This corresponds to the paradox
of d’Alembert. This paradox is solved by realizing that we
can never neglect the effect of friction at the walls. Even at
high Reynolds numbers there are always thin viscous bound-
ary layers near the walls. In these boundary layers the pressure
is essentially equal to that of the main flow but the velocity
decreases to zero at the wall. This implies that the stagnation
pressure pt is lower than its value in the main flow. Hence, the
fluid in the boundary layers cannot flow against the strong ad-
verse pressure gradient in the diverging part of the glottis. This
results into separation of the boundary layers and the formation
Figure 3: A potential flow through
of a free jet. Turbulent dissipation in the free jet explains the
the vocal folds is silent. Sound is
flow control by the oscillating vocal folds and the pressure dif-
produced by the volume flow con-
ference across the glottis. Of course, the fluid from the bound-
trol associated to flow separation
ary layers injected into the main flow has vorticity. Hence we
and formation of a free jet. This im-
see that vorticity injection into the main flow is associated with
plies vorticity injection in the main
the production of sound. This illustrates the statement of Müller
flow.
and Obermeier [83]: “Vortices are the voice of the flow”.
As explained in section (2.3) the velocity field can be separated into a potential and a vortical flow.
The potential part ∇φ of the flow is associated with the dilatation rate ∇·v = ∇2 φ of fluid particles
in the flow. As the acoustical flow is essentially compressible and unsteady, Howe [46] proposed to
define the acoustic field as the unsteady component of the potential flow uac :
uac = ∇φ , (180)
in which φ = φ − φ0 is the time dependent part of the potential φ. For an homentropic flow we can
write the equation of Euler (33) in the form:
∂v f
+ ∇B = c (181)
∂t ρ
in which f c = −ρ(ω × v) is the density of the Corriolis force associated with the vorticity of the
flow. When ω = 0 we have a potential flow. Hence we identify f c /ρ as the source of sound. At low
Mach number for compact flows we can neglect the variation of the density so that fc is the source
of sound. Under such circumstances we can in first approximation apply the energy equation (105) in
the form:
P = f c · uac dV. (182)
V
20:23 18 Jul 2004 36 version: 18-07-2004

where P is the time average acoustical power generated by the vortices and V is the volume in
which the vorticity ω is confined.
The success of this equation is due to the fact that even when using a highly simplified vortex
model it provides a fair prediction of the sound production. A first explanation for this is that it is
an integral formulation which smoothes out random errors. A second aspect is that when considering
models of isolated vortices with time dependent circulation it ignores the effect of the spurious
pressure difference −ρ ∂t∂ induced by the model along the “feeding line” of the vortex. This corre-
sponds to the force necessary for the transport of vorticity towards the “free” vortex. Hence a model
which does not satisfy the momentum equation still gives reasonable predictions because we ignore
the effect of spurious forces.
The theory of Howe provides much insight. In particular, equation (182) clarifies the essential role
of sharp edges at which the potential flow is singular and where vortex shedding occurs. In section
(5.4) we will consider examples of the use of this theory.
For the sake of completeness we provide now a more formal form of the analogy of Howe [48].
Starting from the divergence of the momentum conservation equation (181):
∂∇·v
+ ∇ 2 B = −∇·(ω × v) (183)
∂t
and eliminating ∇·v by means of the mass conservation law (3):
1 Dρ
∇·v = − (184)
ρ Dt
we obtain for an isentropic flow:
1 D20 B 2 1 D20 B ∂ 1 Di
− ∇ B = ∇·(ω × v) + − (185)
2
c Dt 2 2
c Dt 2 ∂t c2 Dt
where:
D0 B ∂ B
= + (∇φ0 ) ·∇ B (186)
Dt ∂t
and we made use of the equation of state for isentropic flows:
1 Dρ 1 Dp 1 Di
= 2 = 2 . (187)
ρ Dt c ρ Dt c Dt
Following Howe [47] the first source term ∇·(ω × v) in (185) is dominant at low Mach numbers.
While the resulting wave equation (185) looks relatively simple, for an arbitrary steady potential flow
∇φ0 it can only be solved numerically.
5 Confined flows
5.1 Wave propagation in a duct
For a narrow hard-walled duct the only waves that propagate are of the form given by equation (70),
which is for a time-harmonic sound field in complex notation the pair of plane waves
p (x, t) = A e iωt−ikx +B e iωt+ikx , (188)
20:23 18 Jul 2004 37 version: 18-07-2004

where k = ω/c0 and the medium is uniform and stagnant. Note that each plane wave is self-similar in
x apart from a phase change. This solution can be generalised for wider ducts or higher frequencies
as follows.
The time-harmonic sound field in a duct of constant cross section with linear boundary conditions
that are independent of the axial coordinate may be described by an infinite sum of special solutions
– modes – that retain their shape when travelling down the duct. They consist of an exponential term
multiplied by an eigenfunction of the Laplace eigenvalue problem on a duct cross section.
Consider the two-dimensional area A in (y, z)-plane with a smooth boundary ∂A and an externally
directed unit normal n. By shifting A in x-direction we obtain the duct D given by
D = {(x, y, z)|(0, y, z) ∈ A} (189)
with axial cross sections being copies of A and where the normal vectors n are the same for all x
(figure 4). In the usual complex notation (with +iωt–sign convention), the acoustic field
A
D
∂A
x
z
Figure 4: Straight duct of arbitrary cross section
p (x, t) ≡ p (x, ω) e iωt , v (x, t) ≡ v (x, ω) eiωt (190)
satisfies in the duct (x ∈ D) the equations
∇ 2 p + k 2 p = 0, iωρ0v + ∇ p = 0. (191)
At the duct wall we assume the impedance boundary condition
p = Z(v · n) for x ∈ ∂D. (192)
Hard walls correspond with Z = ∞. The solution of this problem may be given by
∞

p (x, y, z) = Cn ψn (y, z) e−iκn x (193)
n=1
where ψn are the eigenfunctions of the Laplace operator reduced to A, i.e. solutions of
∂2 ∂2
− +
ψ = α 2 ψ for (y, z) ∈ A,
∂ y2 ∂ z2 (194)
−iωρ0 ψ = (n ·∇ψ)Z for (y, z) ∈ ∂A,
20:23 18 Jul 2004 38 version: 18-07-2004

where α 2 is the
corresponding eigenvalue. The axial wave number κn is given by one of the square
roots κn = ± k 2 − αn2 , + for right and − for left running, while the branch of the squareroot is to be
taken such that Re(κn ) ≥ 0 and Im(κn ) ≤ 0. Therefore, the left-running and right-running modes are
usually explicitly given as
∞

p (x, y, z) = ψn (y, z) An e−iκn x +Bn e iκn x . (195)

n=1
Each term in the series expansion, i.e. ψn (y, z) e−iκn x , is called a duct mode. For hard walls, the
eigenvalues αn2 are real and positive, except for the first one, which is α1 = 0. In the hard-wall case
we have the important distinction between k > αn where κn is real and the mode is propagating, and
k < αn where κn is imaginary and the mode is evanescent, i.e. exponentially decaying. Propagating
modes are called “cut on”, and evanescent modes are “cut off”. The frequency ω = c0 αn is called the
“cut-off frequency” of the mode. For low frequencies, the only cut-on mode is the plane wave (with
cut-off frequency zero)
ψ1 = 1, α1 = 0, κ1 = ±k.
The next eigenvalue α2 is typically of the order of a number between 3 and 4 divided
√ by the duct
diameter D. As a result, any higher modes of a sound field with frequency ω < 10 c0 /D decay
faster than exp(−x/D). At any distance more than, say, two diameters away from a source, the field
is well described by just plane waves.
If the duct cross section is circular or rectangular and the boundary condition is uniform every-
where, the solutions of the eigenvalue problem are relatively simple and may be found by separation
of variables. These eigensolutions consist of combinations of trigonometric and Bessel functions in
the circular case or combinations of trigonometric functions in the rectangular case. In particular for
a cylindrical duct we have in polar coordinates (r, θ) the spiralling modes
ψ = Jm (αmµ r) e−imθ , m ∈ Z, (196)
where Jm is a Bessel function of integer order m. Note that J−m = (−)m Jm . Positive m correspond
with counter-clockwise rotating modes (phase ωt − mθ = constant) and negative m with clockwise
rotating modes. With a fixed source it is sometimes clearer to distinguish between symmetric and
anti-symmetric modes by using sin(mθ) and cos(mθ) rather than e−imθ . For a rectangular duct we
have
sin(βµ y) sin(γν y)
ψ= (197)
cos(βµ y) cos(γν y)
where αµν2
= βµ2 + γν2 . Note that due to the symmetry, in both geometries the eigenvalue has multi-
plicity 2: for each eigenvalue α there are 2 eigenfunctions. This is not the case for an arbitrary cross
section. Some other geometries, like ellipses, do also allow explicit solutions, but only in special
cases such as with hard walls. For other geometries one has to fall back on numerical methods for the
eigenvalue problem.
Without mean flow the problem is symmetric, and to each eigenvalue there corresponds a right-
running and a left-running mode, as both κn and −κn can occur. The modes form a complete set of
basis functions for the solutions to the wave equation. These modes are not exactly orthogonal to each
other, but the complex conjugated modes (more precisely: the solutions of the adjoint – which is here
20:23 18 Jul 2004 39 version: 18-07-2004

the complex conjugated – problem) are mutually orthogonal or “bi-orthogonal” with ψn . It follows
that if we denote ψ̂n = ψn∗ (the asterisk denotes the complex conjugate), we have

(ψn , ψ̂m ) = ψn ψ̂m∗ dσ = δnm ψn2 dσ = δnm (ψn , ψ̂n ),
def
(198)
A A
δnm = 1 if n = m and is zero otherwise. Note that hard-walled modes are real, so ψ̂n = ψn and thus
hard-walled modes are orthogonal. The bi-orthogonality can be used to determine the coefficients of
an expansion. Assume we have the pressure given along cross section x = 0 at the right side of a
source, such that only right running modes occur. We have
∞

p (0, y, z) = F(y, z) = An ψn (y, z). (199)
n=1
After multiplication of left-hand and right-hand side by ψ̂m and integration across A we get
(F, ψ̂m )
Am = . (200)
(ψm , ψ̂m )
In a similar way we can determine an expansion of the Green’s function Ĝ for a duct [81].
∞
ψn (y, z)ψn (η, ζ )
Ĝ(x, y, z|ξ, η, ζ ) = 1
2
i e−iκn |x−ξ | . (201)
n=1 κn (ψn , ψ̂n )
This result can also be obtained, albeit in a more laborious way, by means of Fourier transformation
to x.
Solutions within more complex geometries, consisting of piecewise constant pipe elements, can
be solved by matching two series of expansions in modes of adjacent pipe segments. This procedure
involves an approximation by truncation of the mode expansion. This is quite successfully applied
to stepwise changes in pipe cross-section [81, 60, 119], the diaphragm [81, 119], the elbow [74], the
bend [12], the T-joint [107], the tone-hole of woodwinds [26]. A subtle point is the fact that the
physical singularity of the sound field at the sharp edges is related to the convergence rate of the series
expansion. It is therefore important, when using such approximations, to carefully tune the number of
modes in the two series expansions [119].
With mean flow, the problem is not symmetric anymore. Only for uniform mean flow in a hard-
walled duct this is only a minor problem, because the problem can be transformed to an equivalent
no-flow problem, like we described in section 3.10.1. This is not possible in the combination of mean
flow with soft walls, at least when we use the Ingard-Myers boundary condition [51, 87] for inviscid
flow along an impedance wall, where the eigenvalue appears in the boundary condition. We have
different eigenvalues for the left- and for the right-running modes, and we write instead of (195) now
∞
+ −
p (x, y, z) = An ψn+ (y, z) e−iκn x +Bn ψn− (y, z) e−iκn x . (202)
n=1
Orthogonality cannot be used to determine the amplitudes, but a Greens functions can still be deter-
mined by Fourier transformation to the axial coordinate. Furthermore, there are indications that the
20:23 18 Jul 2004 40 version: 18-07-2004

5.2 Low frequency Green’s function in an infinitely long uniform duct
system may be unstable for certain impedances, with the result that one seemingly upstream-running
mode is to be interpreted as really a downstream running instability[117].
Modes can also be obtained in more complex situations such as pipes with a main flow [29, 117,
98] or non-rigid wall’s [113]. In many applications the prediction of mode propagation allows a
significant insight into sound radiation problems. Example of such an application is the Tyler and
Sofrin rules [134] for the design of aircraft turbines, minimizing the number of excited cut-on modes,
and so the radiation of sound generated by rotor stator interaction.
When the duct and its possible mean flow vary in axial direction, modes that are strictly self-
similar in x are in general not possible. However, if the duct varies slowly, like for example in an
aero-engine flow duct, we can still identify approximate, “almost” modes, which retain their shape but
with a slowly varying amplitude, eigenvalue and axial wave number [116, 118]. (The approximation
is known as a variant of the WKB approximation.) If the duct and mean flow variations scale on a,
so-called, slow variable X = εx where ε is small, we can describe a slowly varying mode by
x
p (x, y, z) = Nn (X)ψn (y, z; X) e−i κn (εξ )dξ
(203)
where ψn with κn is a mode with wave number at cross section A(X). The variable X acts as no
more than a parameter. An adiabatic invariant can be identified yielding the varying amplitude. If the
modes are normalised such that (ψn , ψ̂n ) = 1 and the mean flow is nearly uniform, i.e. modulated by
the duct, the amplitude Nn is given for hard-walled ducts by

1/2
ρ0 Nn2 k 2 − (1 − M 2 )αn2 = constant (204)
where ρ0 , k = ω/c0 and M = U0 /c0 depend on X, and eigenvalue αn corresponds to the mode
ψn at position X. For soft walls the expression is similar. When the mode passes a turning point,
i.e. where k = (1 − M 2 )1/2 αn , the solution breaks down because the incident mode couples with its
back-running counterpart (the mode “reflects”). A local analysis is possible to describe this effect
[118, 95, 96, 97].
At frequencies below the cut-off frequency for higher order modes the acoustical field in a duct is, at
some distance from the source, dominated by the plane wave mode. We expect therefore the Green’s
function to be independent of the transverse coordinate of the listener’s position. Using the principle
of reciprocity (65), this implies also that the Green’s function G should not depend on the transverse
coordinate of the source. Hence we can use a one dimensional Green’s function g defined by:
1 ∂ 2g ∂ 2g
− 2 = δ(x3 − y3 )δ(t − τ ). (205)
c02 ∂t 2 ∂ x3
A solution of this equation will be nonzero only for t > τ and of the form f (t − x3 /c0 ) for x3 > y3
and g(t + x3 /c0 ) for x3 < y3 . So we try (or see [59]) the auxiliary function

ϕ(x, t) = H (x) f t − xc + H (−x)g t + xc ,
20:23 18 Jul 2004 41 version: 18-07-2004

Figure 5: a) Method of images applied to a source at y = (y1 , y2 , y3 ) at a distance y3 from a hard wall
x3 = 0 has a Green’s function: G(x, t| y, τ ) = G0 (x, t| y, τ )+G 0 (x, t| y∗ , τ ) with y∗ = (y1 , y2 , −y3 ).
b) A source between two parallel hard walls generates an infinite row of images. c) A source in a
rectangular duct generates an array of sources.
where H denotes the Heaviside stepfunction, and note that it has the properties
∂ 2ϕ

= H (x) f t − xc + H (−x)g t + xc ,
∂t 2
∂ϕ

= δ(x) f (t) − δ(x)g(t) − 1c H (x) f t − xc + 1c H (−x)g t + xc ,

∂x
where we introduced δ(x) f (t − x/c) = δ(x) f (t) and similarly for g,
∂ 2ϕ

= δ (x)( f (t) − g(t)) − 1c δ(x)( f (t) + g (t)) + 1

H (x) f t − xc + 1
H (−x)g t + xc ,
∂x2 c2 c2
and hence satisfies

1 ∂ 2ϕ ∂ 2ϕ
− 2 = −δ (x)( f (t) − g(t)) + 1c δ(x)( f (t) + g (t)) = δ(t)δ(x).
c ∂t
2 2 ∂x
if f (t) = g(t) = 12 cH (t). By a simple coordinate transformation we find henceforth that
|x3 − y3 |
g(x3 , t|y3 , τ ) = 12 c0 H t − τ − . (206)
c0
A more generic approach may be the use of Fourier transformation in t and x (see [119]).
This result can also be understood by using the method of images [102]. The acoustical field of a
point source placed at a distance d from an infinite plane hard wall will generate the same acoustical
field as the original source and it’s image placed in free space (figure 5). The image’s position is
along the normal to the plane at the distance d from the plane, opposite to the source. The image
takes into account the effect of the reflection of the waves on the plane. A source placed between two
hard plane walls will generate an infinite row of images (figure 5). A source placed in a rectangular
duct at y = (y1 , y2 , y3 ) has the same effect as an array of sources in the (y1 , y2 ) plane as shown in
figure 5. Each source of the array generates a pulse G0 = δ(t − te )/(4πr) with te = t − r/c0 . For a
given listener coordinate (0, 0, x3 ) in the duct we have r 2 = y12 + y22 + (x3 − y3 )2 and dte = −dr/c0 .
The number of images corresponding to an emission time between te and te + dte scales for fixed t, x
and y3 as π d(y12 + y22 ) = π dr 2 = 2πrdr = 2π c0rdte . As G 0 scales as r −1 , the increase in number of
sources with r exactly compensates the decrease of G0 . This results into an Heaviside step function
behaviour for a delta function source behaviour.
20:23 18 Jul 2004 42 version: 18-07-2004

5.3 Low frequency Green’s function in a duct with a discontinuity
Note that for a source placed between two rigid planes we obtain a two dimensional response
which is intermediate between the delta pulse and the Heaviside step. One observes a long decay
scaling with t −1/2 . Thunder acts qualitatively as a line source with a two dimensional acoustical
field. This phenomenon explains the long decay time of thunder sound. This makes acoustics in two
dimensions quite complex and different from three dimensional and one dimensional acoustics [25].
A two dimensional acoustical field does not have a simple near-field behaviour as a three dimensional
acoustical field. This essential difference between three dimensional and two dimensional acoustics
is a major problem when considering the sound production by a two dimensional model of a three
dimensional flow. Extending the two dimensional model to the acoustical far field will dramatically
exaggerate radiation losses. Placing a radiation condition at a finite distance will provide results which
depend on the distance between the boundary and the flow.
In an analogous way we may find that in the presence of a uniform subsonic main flow U0 in the
duct the Green’s function, satisfying
1 ∂ ∂ 2 ∂2g
+ U 0 g − = δ(x3 − y3 )δ(t − τ ), (207)
c02 ∂t ∂x ∂x2
is given by
x3 − y3 1 x3 − y3
g(x3 , t|y3 , τ ) = 12 c0 H (x 3 − y3 )H t − τ − + 2 c0 H (y3 − x3 )H t − τ + . (208)
c0 + U0 c0 − U0
Note that this Green’s function satisfies the reverse-flow reciprocity principle [44] rather than (65).
When we exchange the source and listener’s positions, we should also reverse the main flow in order
to keep the travel time r/(c0 ± U0 ) of the waves between the source and the listener constant.
We want to introduce the concept of low frequency Green’s function by considering the effect of a
discontinuity in an infinitely long duct. We will use the reciprocity principle (65) in order to determine
the Green’s function. This implies that for a source placed at the discontinuity, we can deduce the
Green’s function by considering the sound field generated by a source placed at the position of the
listener. For a source placed far from the discontinuity and a listener placed at the same side of
the discontinuity as the source, the Green’s function will be build up of the waves generated at the
source plus the reflection of one of the waves at the discontinuity. A listener placed beyond the
discontinuity will only be reached by the waves transmitted through the discontinuity. In other words
the problem reduces to the determination of the reflected and transmitted waves at the discontinuity.
We limit ourselves to the case of a compact transition region xI ≤ x3 ≤ x I I of cross sectional area
S(x3 ) between two semi-infinite ducts of cross section SI at the side of the source and SI I at the
opposite side. In the transition region we will now for simplicity assume a quasi-one dimensional
incompressible potential flow v3 (x3 , t). Applying the integral (10a) mass conservation law across the
discontinuity we have for xI ≤ x3 ≤ x I I :
S(x3 )v3 (x3 , t) = S I v3 (x I , t) = S I I v3 (x I I , t). (209)
20:23 18 Jul 2004 43 version: 18-07-2004

The equation of Bernoulli (42) applied in xI ≤ x3 ≤ x I I between x I and x3 yields:

∂ ∂ x3
p I − p (x3 , t) = ρ0 φ(x3 , t) − φ I = ρ0 v3 dx3
∂t ∂t x I
x3 S ∂v (x , t)
I 3 I
= ρ0 dx3 . (210)
x I S(x 3 ) ∂t
For harmonic waves p̂I = p+ I e

−ikx 3
+ p− I e
ikx 3
and p̂I I = p+ II e
−ikx 3
we have:

S I p+I e
−ikx I
− p− I e
ikx I
= S I I p+
II e
−ikx I I
and:

p+
I e
−ikx I
+ p−I e
ikx I
− p+
II e
−ikx I I
= ik0 L eff p+I e
−ikx I
− p−
I e
ikx I
(211)
with the effective length Leff defined by:

xI I
SI
L eff = dx3 . (212)
xI S(x3 )
We find for the reflection coefficient R and the transmission coefficient T :

p−
I e
ikx I
S I − S I I (1 − ik L eff )
R= + −ikx I =
PI e S I + S I I (1 + ik L eff )
and:
p+
II e
−ikx I I
T = . (213)
p+I e
−ikx I
This results reduces to the well known result R = [(SI − S I I )/(S I + S I I )] and T = 2SI /(S I + S I I )
in the limit of k L eff → 0. For an ideal open pipe end at xI = 0 we find R = −1 by taking the limit
S I /S I I → 0 and x I = 0 [25, 82]. For a closed pipe at xI = 0 we find R = 1 by taking the limit
S I /S I I → ∞. When SI = S I I we recover the result for a diaphragm in a pipe R = ik Leff /(2+ik L eff )
and T = 2/(2 + ik L eff ).
In the limit k L eff → 0 the Green’s function g which we are looking for is for xI > y3 :
c0 |x3 − y3 | c0 x3 + y3 − 2x I
g= H t −τ + + R H t −τ + for x3 < x I ,
2 c0 2 c0
and:
c0 x3 − y3
g = T H t −τ − for x I < x3 . (214)
2 c0
Considering the case xI > y3 , while we neglect the effect of reflections R = 0 and T = 1, we can
write for k L eff 1 along x I < x3 < x I I
c0 x I + xeff − y3
g = H t −τ − , (215)
2 c0
where: x3
SI
xeff = dx3 . (216)
xI S(x3 )
The effective coordinate xeff corresponds to the potential difference induced between x3 and x I by a
flow having at x I a unit velocity vI (x I ) = 1.
20:23 18 Jul 2004 44 version: 18-07-2004

5.4 Aero-acoustics of an open pipe termination
We can use the same equations to the problem of an arbitrarily shaped discontinuity in a pipe, if
we replace the definition (216) of the effective position, xeff , by the more general definition:
φ(x3 ) − φ I
xeff = , (217)
v I
where we do not assume a quasi-one dimensional flow at the discontinuity. This corresponds to the
low frequency Green’s function proposed by Howe [44] for a discontinuity in a pipe. This generalisa-
tion allows to calculate the effect of a thin diaphragm in a pipe [82, 102] or the reflection at an elbow
[10] by using potential theory for incompressible flows. This approach is limited to compact regions,
but has the great advantage of providing a detailed model of the acoustical flow at such a discontinuity
even in the neighborhood of sharp edges. Note that near a sharp edge an expansion of the solution in
pipe modes fails to converge. Such theories are first order approximations in a Matched Asymptotic
Expansion procedure [64]. The Match Asymptotic Expansion procedure allows a systematic approx-
imation which can be extended to higher order in the small parameter k Leff . In the present case we
have neglected terms of order (k Leff )2 . We take the inertia of the flow at the discontinuity into account
but we neglect the effect of compressibility.
The same kind of approximation leads to the low frequency Green’s function for a compact rigid
body in free-space proposed by Howe [47, 48].
5.4.1 Introduction
Modelling of the aero-acoustical behaviour of internal flow involves in many cases an open pipe
termination. This is not only a boundary condition for the calculations of the internal acoustical field,
the pipe termination is also a source of sound for the external acoustical field. In the presence of a
main flow or at high amplitudes this involves a complex unsteady flow at the pipe termination. This
problem is often oversimplified and underestimated. At low frequencies for example the boundary
condition p = 0 is often used. In the presence of a main flow this corresponds to a quasi-steady
response of the free jet outside the pipe. This model is quite reasonable to describe the reflection
of acoustical waves. When, however, we consider the convection of vorticity, such a model induces
spurious sounds and flow phenomena due to the fact that vortices cannot be transported through such
a boundary. Also such a model does not predict the spectacular effect of the shape of the nozzle
edges on the aero-acoustical behaviour of the pipe termination. Under particular flow conditions and
for specific nozzle shapes, the pipe termination can be a source of sound. Coupling with acoustical
resonances of the pipe system results into whistling. We give here a short survey of the aero-acoustical
behaviour of open pipe terminations.
5.4.2 Low frequency linear behaviour without main flow
We consider plane harmonic waves propagating in the x1 -direction along a duct of uniform cross
sectional area A:
p = p̂ e iωt = ( p+ e−ikx1 + p− e ikx1 ) e iωt . (218)
20:23 18 Jul 2004 45 version: 18-07-2004

These waves induce a volume flow at the open pipe termination in x1 with an amplitude:
p+ − p−
φV = Av̂1 = A . (219)
ρ0 c0
The boundary condition at the open pipe termination x1 = 0 for the internal acoustical field can
be expressed in terms of a radiation impedance:
p̂ p+ + p−
Zp = = ρ0 c0 + . (220)
v̂1 x1 =0 p − p−
The acoustical power < P > radiated at the pipe termination is given by:
P = ( p v1 )x1 =0 A = 12 v̂1 v̂1∗ A Re(Z p ) (221)
where the asterisk denotes the complex conjugate. The real part of this impedance can be determined
by applying the conservation of acoustical energy (equation 99) between the pipe exit and the far field
outside the pipe. We consider first a thin walled, unflanged pipe, emerging into free space. In that
case the far field will be dominated by the monopole radiation (82):
iωρ0φV iωt−ikr iωρ0Av̂1 iωt−ikr

p e = e . (222)
4πr 4πr
The corresponding radial velocity becomes for kr 1 equal to vr p /(ρ0 c0 ). Together with
eqaution (222) this gives for the power P radiated in the far field
1 1
P = p vr 4πr 2 p2 4πr 2 = ρ0 c0 k 2 A2 v̂1 v̂1∗ . (223)
ρ0 c0 8π
Combination of equations (221) and (223)with the conservation of acoustical energy yields:
k 2A
Re(Z p ) = ρ0 c0 . (224)
4π
When we consider a pipe emerging from a hard wall into a half-infinite free space, flanged pipe,
the radiation energy is distributed over a surface 2πr2 rather than 4πr 2 . Furthermore, the acoustical
amplitude is increased by a factor two due to reflection of waves at the wall. The combination of
both effects results into an increase of Re(Zp ) by a factor two compared to the unflanged pipe termi-
nation case (224). Further confinement of the pipe outlet will further increase the radiation impedance.
The imaginary part Im(Z p ) of the radiation impedance takes the inertia of the acoustical flow
outside the pipe into account. This effect is often expressed in terms of an end correction δ:
Im(Z p ) = ρ0 c0 kδ. (225)
This is the length of a pipe segment which

√ has the same inertia as the outer flow. In first order
approximation in the small parameter k A/π, the reflected wave p− seems to be generated from a
reflection of the incoming wave p+ , without phase shift, from an ideal open pipe termination with
p = 0 placed at x1 = δ. The pipe behaves acoustically if it were longer by a length δ if we assume
that it is terminated by ideal open ends. In contrast with Re(Zp ) which is determined by global flow
20:23 18 Jul 2004 46 version: 18-07-2004

properties, the end correction δ of an open pipe termination is sensitive to the details of the local flow
around at the pipe outlet. For a pipe with circular cross-section A = π a2 and for infinitely thin walls,
Levine and Schwinger [66] found in the low frequency limit δ 0.61a. For thick walls the end
correction increases gradually from δ = 0.61a up to the flanged pipe limit δ = 0.82 [1, 101, 91, 19].
Please note that the order of magnitude of the radiation impedance of an unflanged pipe is found by
considering the low frequency limit of the radiation impedance Zs of a compact sphere of radius a.
Following equations (80) and (81) we have:
ika + (ka)2
Z s = ρ0 c0 ρ 0 c0 ika + (ka)2
+ . . . . (226)
1 + (ka)2
The simple behaviour discussed above is typical for a radiation into a three dimensional free field.
Confinement will drastically affect the radiation impedance. If we consider for example the radiation
of a pipe termination emerging between two closely spaced parallel hard walls, the radiation field
will be two dimensional. The radiation impedance has in such a case a complex behaviour [65]. One
cannot for example identify a locally incompressible near field of the pipe termination. In general
the radiation impedance will be much larger than in a three dimensional case. It is quite tempting to
use a two dimensional flow simulation to describe the oscillating grazing flow along a wall cavity.
It is however essential to realize that with this two dimensional calculation, we will dramatically
overestimate the radiation losses of the flow. The results of the calculations will actually depend on
the size of the calculation domain.
5.4.3 High frequency linear behaviour without main flow
For high frequencies any incident mode of amplitude A, frequency ω, circumferential order m and
radial order µ reflects in several propagating radial modes. Due to cylindrical symmetry, no mode
reflects in another circumferential order. Outside the pipe we have in the far field
e−ik
pmµ (x, r) ρ0 c02 ADmµ (ξ ) (k → ∞), (227)
k
where x = cos ξ , r = sin ξ , and Dmµ (ξ ) is called the directivity function, and |Dmµ (ξ )| is the radi-
ation pattern. As each mode has its own spiralling phase plane orientation, the radiated pattern consists
of lobes interlaced by zeros. Each zero is found at the propagation direction of a reflected mode, i.e.
at ξ = arcsin(αmν /k), except for the direction of the incident mode, i.e. at ξmµ = arcsin(αmµ /k). As
may be expected, here we find the radiation maximum or main lobe.
The field inside may be written as
∞

p(x, r, θ, t) = ρ0 c02 A e iωt−imθ Jm (αmµr) e−iκmµ x + Rmµν Jm (αmν r) eiκmν x (228)
ν=1
where the branch of the square root

κmµ
2
= k 2 − αmµ
2
is chosen such that κmµ is either positive real or negative imaginary. The matrix Rm = {Rmµν } is
called the reflection matrix. Explicit analytical expressions (in the form of complex contour integrals)
are found by application of the Wiener-Hopf method [66, 16].
20:23 18 Jul 2004 47 version: 18-07-2004

A subtlety in the solution of the scattering problem is the duct edge. Here, the boundary condition
is, technically speaking, not applicable, and without further condition we have no unique solution due
to possible point or line sources “hiding” at the edge. So we have to add the (so-called) edge condition
of finite energy in any neighbourhood of the edge. This amounts to an integrable squared velocity field
∼ |∇ p|2 . It transpires that p varies near the edge like the squareroot of the distance.
5.4.4 Influence of main flow on linear behaviour at low frequencies
Flow has a dramatic effect on the radiation impedance Zp of an open pipe termination. We consider
the effect of a uniform subsonic flow velocity U0 in the pipe. In this section we limit our discussion
to low frequencies. By low frequencies we do not only mean that we limit ourselves to plane-wave
propagation in the pipe, we also assume that the flow at the pipe termination is locally quasi-steady.
This corresponds to a low Strouhal number limit St = ωa/U0 1. As the Strouhal number is related
to the Mach number M = U0 /c0 by: St = ka/M , the combination of St 1 with M < 1 implies
that we consider very long wave length ka 1.
The acoustic field in the pipe is described by the convective wave equation:
∂ ∂ 2 ∂ 2 p
+ U0 p − c02 2 = 0. (229)
∂t ∂ x1 ∂ x1
The solution of this wave equation is:

+x −x
p = p+ e iωt−ik 1
+ p− e iωt+ik 1
(230)
with:
ω
k± = . (231)
c0 ± U0
The fluctuations v1 in the flow velocity are obtained by applying the linearized momentum equation
∂
ρ0 ( ∂t + U0 ∂∂x1 )v1 = − ∂∂x1 p :
1 + iωt−ik + x1 −
v1 = p e − p− e iωt+ik x1 (232)

ρ0 c0
Following the Vortex Sound Theory [48], the acoustic field is most efficiently described in terms of
fluctuation B = U0 v1 + p /ρ0 of the total enthalpy B = i + v12 /2:
+x −x
B = B + e iωt−ik 1
+B − eiωt+ik 1
(233)
with: U0
B ± = p± 1 ± . (234)
c0
We consider a flow leaving the pipe. Due to flow separation at the pipe outlet, a free jet will be formed.
It can be demonstrated that the only possible subsonic jet flow is a jet flow in which the pressure is
uniform and equal to the pressure in the surroundings p = patm [128]. Following this model we have
the boundary condition p = 0 at the open pipe termination x1 = 0. This implies a pressure reflection
coefficient:
p−
= −1 (235)
p+
20:23 18 Jul 2004 48 version: 18-07-2004

and a total enthalpy reflection coefficient:
B− p− c0 − U0 c0 − U0
= =− . (236)
B + p c0 + U0
+ c0 + U0
This result implies a loss of acoustical energy. In the free jet the kinetic energy of the flow will
be dissipated without any pressure recovery. As an acoustical modulation of the flow at the pipe
termination implies a modulation of the kinetic energy in the jet, one can expect an absorption of
acoustical energy. The vorticity in the shear layers of the free jet act as a source of sound. The end-
correction appears to be determined by details of the flow just outside the pipe. As we will see in
the next section in the limit of very low Strouhal numbers for an unflanged circular pipe δ 0.2a
[109, 111, 101]. A rather unexpected result.
The above model corresponds to the very simple boundary condition p = 0 at the pipe outlet
x1 = 0. While this model appears to be quite accurate in the limit St 1, it should be used with
care. A vortex with a dimension of the order of the pipe diameter correspond to a perturbation with a
Strouhal number of order unity. A boundary condition of uniform pressure at a pipe termination will
not allow such a vortex to flow out of the pipe! This is of course in contradiction with experimental
observations, that vortices do leave the pipe through an open pipe termination.
In the case of inflow, we can have a potential flow. In such a case the radiation impedance for low
frequencies will not be strongly affected by the flow. If flow separation occurs one can consider the
use of a quasi-steady model [140].
It is interesting to note that a diaphragm placed at the pipe outlet can be used as a non-reflecting
pipe termination at a critical Mach number M = U0 /c0 . This anechoic pipe termination behaviour
was predicted by Bechert [4] using a quasi-steady flow model. Applying the quasi-steady equation of
Bernoulli for an incompressible flow, between the pipe termination x1 = 0 with cross section A and
the free jet of cross section Aj formed at the diaphragm we have:
A
1
ρ (U0
2 0
+ v1 )2 + p(0, t) = 12 ρ0 (U0 + v1 )2 , (237)
Aj
where we assume that in the jet the pressure remains constant, pj = 0. Substitution of p = p+ + p−
and v1 = ( p+ − p− )/(ρ0 c0 ) yields after linearisation:

U0 A2

1− −1
p+ c0 A2j
= −
(238)
p− 1+ U0 A2
−1
c0 A2j
which vanishes for U0 /c0 = ((A/A j )2 − 1)−1 . Generalisation of this result to arbitrary subsonic
Mach numbers is discussed by Hofmans [42]. Application to the design of silencers is discussed by
Durrieu [28].
When the Mach number is increased, we will eventually reach a critical flow at the diaphragm
(choked flow, unit Mach number at the diaphragm). In this case the Mach number of a steady flow in
the pipe is imposed by the ratio A/Aj of pipe and diaphragm cross sections. It will in a quasi-steady
theory remain constant, independent of any acoustical perturbation. The condition:
v1 c
= (239)
U0 c0
20:23 18 Jul 2004 49 version: 18-07-2004

combined with the assumption of an ideal gas behaviour ( p / p0 = γρ /ρ0 and c02 = γ p0 /ρ0 ) yields
[2]:
p 2
= ρ0 c0 . (240)
v1 M(γ − 1)
A more elaborated discussion of the acoustical response of a chocked nozzle is provided by Marbel
and Candel [72].
5.4.5 Influence of main flow on linear behaviour at high frequencies
The available most complete model of sound radiation from a flow duct with jet, that is analytically
tractable, is the following. It is a semi-infinite hard-walled duct with walls of zero thickness, contain-
ing inviscid uniform mean flow, density and sound speed, of which the values may differ inside and
outside the duct or jet. The interface between jet and outer medium is modelled as a vortex sheet.
Its analytically exact but mathematically rather involved solution of Wiener-Hopf type was given by
Munt in [84, 85]. The solution contains all the elements of the no-flow solution, like reflection into
radial modes and a radiation pattern with mode-related lobes, but the presence of flow adds a number
of particular features. Of course, there are geometrical effects like a redirecting of the radiation pattern
and refraction across the jet interface, but there are also two effects due to coupling with the mean
flow.
First, for a velocity different across the vortex sheet, the jet is unstable. This instability is mathe-
matically to be recognized by means of a causality analysis in the complex frequency plane. Second,
there is the edge condition which is even more subtle than without flow (see section 3.10.3.) The con-
dition of finite energy is still necessary to select physically possible solutions, but is now not enough
for a unique solution. We still have a choice, corresponding to the amount of acoustic vorticity shed
from the trailing edge. The possibility of vortex shedding is included in the model but its amplitude is
not yet fixed, as it is determined by viscous and nonlinear processes which are not included. We have
to add an extra condition. One such a condition is full regularity of the field at the trailing edge, the
Kutta condition, which corresponds physically to the maximum amount of vortex shedding possible.
Any other, “non-Kutta condition”, solution will be singular at the edge but only one corresponds to no
vortex shedding. Since it is the shed vorticity that excites the jet instability, the strict absence of the
instability is a way to apply the condition of no vortex shedding for jets. If the mean flow is uniform
everywhere, the absence of vortex shedding is easiest typified by a continuous potential. An alterna-
tive way to characterise the Kutta condition is via the streamline emanating from he edge, given by
r = a +Re(h(x) exp(iωt −imθ)). Without Kutta condition its shape for x ↓ 0 is like h(x) = O(x1/2 ).
With Kutta condition it is like h(x) = O(x3/2 ).
The Kutta condition seems to be the relevant condition for relatively low frequencies and acoustic
amplitudes and high Reynolds numbers [15]. Therefore it will be the condition that is supposed here
throughout. Depending on the mean flow Reynolds number, dimensionless frequency and amplitude,
other conditions (generalised Kutta conditions) are also possible [110, 112].
Physically, the shed vorticity, while moving near edge of the solid duct wall, produces some
additional sound and thus adds some acoustic energy to the sound field. At the same time there is a
certain amount of acoustic energy and a certain amount of mean flow energy needed for the creation
of vorticity. Usually the net profit of energy is negative (acoustic energy is lost into the hydrodynamic
vorticity) but this is not necessary [110].
20:23 18 Jul 2004 50 version: 18-07-2004

Very interesting effects of vortex shedding occur for low frequencies, so we will consider results
of Munt’s solution in more detail in the next section.
5.4.6 Frequency dependence of the effect of flow on the radiation impedance
The experiments by Bechert, Michel and Pfizenmaier [5] in 1977 showed for the first time a dramatic
loss of acoustic energy when a long plane wave from upstream a jet exhaust was partially reflected and
partially transmitted at the exit. Although only a part was reflected, practically nothing was recovered
in the radiated field outside. This was explained by Howe [45] in 1979 for low Mach numbers, by
showing that due to the presence of mean flow the sound field sheds vorticity from the edge, in such a
way that acoustic energy is converted into the acoustically undetected hydrodynamic energy. Cargill
then showed in [13] that the Munt model [84] includes all the reported effects for any Mach number.
This makes the Munt model extremely interesting for studying various aspects of sound field, mean
flow coupling.
Let us consider the predictions of the Munt model for plane waves of relatively low wave number
k in a duct of radius a with Mach jet number Mj . The feature that is directly related to the energy
absorption is the modulus of the pressure reflection coefficient |R|. With Kutta condition, it tends
to unity for low Helmholtz number ka, which yields for the transmitted power in the duct PT ∼
(1 + M j )2 − |R|2 (1 − M j )2 , a finite value, which contrasts the vanishing radiated value of O((ka)2 ).
The difference is the energy lost in the vortices.
Apart from the modulus reflection coefficient [85], little results are reported from the Munt solu-
tion pertaining to the low frequency range. Therefore, we present here some recent results obtained
by P. in’t Panhuis [71].
0.65
M = 0.01 + Kutta c.
0.6 M=0
0.55
0.5
0.45
δ/a
0.4
0.35
0.3
0.25
0 0.5 1 1.5
ka
Figure 6: The endcorrection for no flow (Mj = 0) and a little flow (M j = 0.01). Note the nonuniform
behaviour for ka → 0, M j → 0.
A rather remarkable feature is the behaviour of the end correction for a jet without co-flow. We
define the end correction as the virtual point just outside the duct exit against which the plane wave
appears to reflect with a free field boundary condition of: | p| attains a minimum. If we write R =
−r e−iθ , then the plane wave pressure is given by
kx
−i kx
i 1+M −iθ
p(x) ∼ e 1+M j −r e j ,
20:23 18 Jul 2004 51 version: 18-07-2004

and so 2k M x
| p(x)|2 ∼ 1 + r 2 − 2r cos 1−Mj 2 − θ

j
which attains its minimum for x = δ where
δ (1 − M 2j )θ
= .
a 2ka
A marked difference is the nonuniform limit of the end correction for ka → 0 and Mj → 0. As
long as the Strouhal number ka/Mj is large, it converges to the no flow value ∼ 0.6127, but once
ka M j , the Strouhal number becomes small and the behaviour changes completely: the final value
will be ∼ 0.2554. Therefore, results for low Mach number and low Helmholtz number are usually
better presented depending on the Strouhal number. See figure 6 where Munt’s solution of very low
Mach number is compared with the Levine and Schwinger [66] results for no flow. This non-uniform
limit was predicted in [109, 111] and actually experimentally confirmed more than 10 years later by
Peters et al. [101].
A pressure reflection coefficient larger than one for ka between 0 and, say, 1 was already reported
by Munt [85]. In figure 7 the reflection coefficients together with the corresponding end corrections
are given for various Mach numbers (no flow outside, no density or sound speed difference). Note that
the limiting values for ka → 0 of |R| are unity for any Mach number; the endcorrections converge to
0.2554(1 − M 2j )1/2 .
1.2
1.1 0.01
0.6
|R| 1
0.01
0.1 0.2 0.3 0.4
0.5 0.6
0.1
0.5
0.9
δ/a
0.2
0.3
0.8 0.4 0.4
0.5
0.7 0.6
0.3
0.6
0.5 0.2
0.4
0 0.5 1 1.5 0.1
ka 0 0.5
ka 1 1.5
Figure 7: Plane wave reflection coefficient |R| and endcorrection δ at jet exhaust without co-flow for
M j = 0.01, 0.1, . . . , 0.6.
Finally, in figure 8 the effect of co-flow is shown. When the outer Mach number Mo varies from
zero to the jet Mach number M j , both reflection coefficient and end correction return to a behaviour
similar to the no-flow case.
5.4.7 Whistling
Depending on the shape of the nozzle, a pipe termination can become a source of sound. This is
a familiar phenomenon. By adjusting our lips and blowing at a critical velocity we can whistle.
20:23 18 Jul 2004 52 version: 18-07-2004

1.2
0.65
1.1 0 0.6
0.25 1
0.5 0.55
1
0.75 0.75
1 0.5
0.9
0.50
0.45
0.8
|R| 0.4 0.25
0.7 δ/a
0.35
0
0.6 0.3
0.5 0.25
0.4 0.2
0 0.5 1 1.5
0 0.5 1 1.5 ka
ka
Figure 8: Plane wave reflection coefficient |R| and end correction δ at jet exhaust with Mj = 0.3 and
co-flow velocities Mo /M j = 0, 0.25, 0.5, 0.75, 1.
Sound production by such a smoothly curved pipe outlet has been first studied in laboratory ex-
periments by Powell [7] and by Wilson et. al [142]. We propose a qualitative model of this phe-
nomenon in terms of Vortex Sound Theory. The same model has been used to explain the behaviour
of whistler-nozzle [37, 39, 47]. The model has also been applied to other configurations in which
self-sustained oscillations of the flow can be described in terms of discrete vortex shedding. Ex-
amples of such phenomena are the edge-tone [43], the Helmholtz resonator [92, 20, 77, 108], the
splitter plate [139], the closed side branches [11, 143], the flute [135, 136, 21], the wall cavities
[47, 133], the diffusors [67] and the solid propellant rocket engine [2]. Other examples of related
self-sustained flow oscillations are described in many review papers and textbooks [120, 121, 6, 8].
Let us start our discussion by reconsidering the acoustical

response of an unflanged pipe termination in the case of a uni-
form main flow U0 directed towards the pipe outlet. The pipe
cross section has a radius a. Flow separation at the sharp edges
of the pipe outlet results into the formation of a free jet. We con-
sider the reflection of harmonic plane waves with a frequency f
travelling down the pipe, in the direction of the main flow. The
acoustical flow uac is defined following the Vortex Sound The-
ory [46] as the unsteady part of the potential flow component of
the flow field. This potential flow bends around the sharp edges
of the pipe outlet as illustrated in figure 9.
In a potential flow the centripetal force on the fluid parti-
cles, allowing a curved stream line, is provided by the pressure Figure 9: Acoustic flow at a pipe
gradient. The corresponding acceleration is −|uac |/R 2 where outlet a) for an unflanged pipe ter-
R is the radius of curvature of the stream line. This implies that mination and b) for a horn.
the pressure decreases towards the interior of the bend. As a
consequence the velocity should increase. This is easily verified in a quasi-steady approximation by
applying the equation of Bernoulli. At a sharp edge the radius of curvature of the stream lines in the
potential flow vanish. This implies a singular behaviour. The pressure becomes infinitely low and the
velocity infinitely large.
20:23 18 Jul 2004 53 version: 18-07-2004

The singular behaviour of the flow at a sharp edge implies that viscous forces can never be ne-
glected at such an edge. The result is that flow separation occurs which induces the shedding of vor-
ticity. The vorticity is such that it compensates the potential flow singularity. Within the frame work
of a frictionless theory this is the so-called Kutta condition [15]. At moderate acoustical amplitudes
|uac |/U0 1, the amount of vorticity generated by the acoustic field will be negligible compared to
the vorticity shedding induced by the main flow. The acoustic field does however trigger an instability
of the shear layers of the free jet at the pipe exit. This instability results into the concentration of the
vorticity into coherent structures which can be described as vortex rings. From experiments it appears
that a new vortex start to be formed at the pipe outlet at the beginning of each acoustical oscillation
period, when the acoustical velocity turns from pipe inwards to pipe outwards. We call this t = 0.
The vortices travel along the pipe axis with a convective velocity Uc which is about half the main flow
velocity U0 . The circulation of the vortex ring increases almost linearly in time as it accumulates the
vorticity shed at the edge [92, 10, 21]. After one oscillation period t = T a new vortex is shed an
the vortex ring travels further downstream. Following the theory of Howe [46] the acoustical power
generated by the vorticity is given by:
f c ·uac = −ρ(ω×v)·uac . (241)
This theory predicts that the vorticity field will initially absorb energy from the acoustic field. This
seems logical as the acoustic field is perturbing the shear layer. This initial absorption is strong
because, near the edge of the pipe exit, the acoustical velocity is normal to the convective velocity
v = (Uc , 0, 0). Furthermore the acoustical velocity is large because of the singular behaviour near
the sharp edge. The theory of Howe [46] predicts that after half an oscillation period T /2 < t < T
the vortex will produce sound because the acoustic flow is reversed while the signs of the convective
velocity and the vorticity do not change. In spite of the growth of the vortex, the sound production
will be weaker than the initial absorption which occurred for 0 < t < T /2. This is due to the fact that
the vortex moves away from the edge singularity. Both the magnitude of the acoustical velocity and
the angle between the vortex path and the acoustical streamlines decrease. This discussion provides a
qualitative understanding of the sound absorption predicted by linear theory in the previous sections.
An important feature of this discussion is that it stresses that the sound absorption is a balance
between the initial absorption for 0 < t < T /2 and the sound production for T /2 < t < T . This
implies that it should be possible to obtain a net sound production by reducing the initial absorption.
This is obtained by using a pipe with rounded edges like our lips or a horn. The singular behaviour at
the flow separation point is suppressed. Furthermore the convective velocity near the flow separation
point is almost parallel to the potential flow lines of the acoustic field. This reduces so much the initial
absorption that at critical flow velocities sound production is observed. Typically one finds an optimal
sound production when the travel time of the vortices along the lips corresponds to an oscillation
period of the acoustic field. Experimentally one observes this for a Strouhal number based on the
radius of curvature R of the lips given by: St R = f R/U0 = 0.2.
When acoustical energy can accumulate in a resonant mode of the pipe system upstream of the
pipe termination one can observe self-sustained oscillation which are referred to as whistling. This
occurs when the sound production is large enough to compensate for visco-thermal losses and radi-
ation losses of the resonator. It is important to realize that in such a case the oscillation amplitude
is limited by non-linear effects. This can be an increase of losses at high amplitudes or a saturation
of the source. In the particular case of whistling the main non-linear amplitude limiting effect is the
saturation of the circulation of the vortex rings. Once the vortex has accumulated all the vorticity
20:23 18 Jul 2004 54 version: 18-07-2004

REFERENCES
available in the shear layer its circulation has reached a maximum value. In the case of a flute the
main non-linear amplitude limiting mechanism is additional shedding of vorticity at sharp edges such
as the labium [135]. The present example also illustrates that sound production does not necessarily
involve impingement of a vortex or shear layer on a sharp edge.
This qualitative discussion indicates that for Strouhal number of order unity, a numerical flow
simulation of a pipe system should include a model of the dynamical response of the free jet formed
at a pipe outlet. Simple boundary conditions such as assuming a constant outlet pressure are only
reasonable at low Strouhal numbers.
References
[1] Y. Ando. On the sound radiation from semi-infinite circular pipe of certain wall thickness.
Acustica, 22:219–225, 1969.
[2] J. Anthoine, J-M. Buchlin, and Hirschberg A. Effect of nozzle on resonance in large srm:
theoretical modeling. Journal of Propulsion and Power, 18:304–311, 2002.
[3] G. Batchelor. An introduction to fluid dynamics. Cambridge University Press, Cambridge, UK,
1976.
[4] D.W. Bechert. Sound absorption caused by vorticity shedding, demonstrated with a jet flow.
Journal of Sound and Vibration, 70:389–405, 1980.
[5] D.W. Bechert, U. Michel, and E. Pfizenmaier. Experiments on the transmission on sound
through jets. AIAA Journal, 16:873–874, 1978. AIAA paper 77-1278, AIAA 4th Aeroacoustics
Conference, Atalanta, Georgia, 3-5 October 1977.
[6] W.K. Blake. Mechanics of Flow-induced Sound and Vibration, Volume I. Academic Press,
Orlando, 1986.
[7] W.K. Blake and A. Powell. The development of contemporary views of flow-tone generation.
In Recent Advances in Aeroacoustics, New York, 1983. Springer-Verlag.
[8] R.D. Blevins. Flow-induced Vibration. Van Nostrand Reinhold, New York, 2nd edition, 1990.
[9] H.H. Brouwer. On the use of the method of matched asymptotic expansions in propeller aero-
dynamics and acoustics. Journal of Fluid Mechanics, 242:117–143, 1992.
[10] J.C. Bruggeman. The propagation of low-frequency sound in a two-dimensional duct system
with t joints and right angle bends: Theory and experiment. Journal of Acoustical Society of
America, 82:1045–1051, 1987.
[11] J.C. Bruggeman, A. Hirschberg, M.E.H. van Dongen, A.P.J. Wijnands, and J. Gorter. Self-
sustained aero-acoustic pulsations in gas transport systems: experimental study of the influence
of closed side branches. Journal of Sound and Vibration, 150:371–393, 1991.
[12] A. Cabelli. The acoustic characteristics of duct bends. Journal of Sound and Vibration, 68:369–
388, 1980.
20:23 18 Jul 2004 55 version: 18-07-2004

REFERENCES
[13] A.M. Cargill. Low-frequency sound radiation and generation due to the interaction of unsteady
flow with a jet pipe. Journal of Fluid Mechanics, 121:59–105, 1982.
[14] L. Cremer and M. Heckl. Structure-Borne Sound. Springer-Verlag, Berlin, 2nd edition, 1988.
Translated and revised by E.E. Ungar.
[15] D.G. Crighton. The Kutta condition in unsteady flow. Annual Review of Fluid Mechanics,
17:411–445, 1985.
[16] D.G. Crighton, A.P. Dowling, J.E. Ffowcs Williams, M. Heckl, and F.G. Leppington. Modern
Methods in Analytical Acoustics. Lecture Notes. Springer–Verlag, London, 1992.
[17] N. Curle. The influence of solid boundaries upon aerodynamic sound. Proceedings of the Royal
Society,, Series A 231:505–514, 1955.
[18] J.P. Dalmont, C.J. Nederveen, V. Dubos, S. Ollivier, V. Méserette, and E. te Sligte. Experimen-
tal determination of the equivalent circuit of an open side hole: linear and non-linear behaviour.
EAA, 46:567–575, 2002.
[19] J.P. Dalmont, C.J. Nederveen, and N. Joly. Radiation impedance of tubes with different flanges:
numerical and experimental investigations. Journal of Sound and Vibration, 244:505–534,
2001.
[20] S. Dequand, X. Luo, J. Willems, and A. Hirschberg. Helmholtz-like resonator self-sustained

oscillations, part 1: Acoustical measurements and analytical models. AIAA Journal, 41:408–
415, 2003.
[21] S. Dequand, J.F.H. Willems, M. Leroux, R. Vullings, M. van Weert, C. Thieulot, and
A. Hirschberg. Simplified models of flue instruments: Influence of mouth geometry on the
sound source. Journal of the Acoustical Society of America, 113:1724–1735, 2003.
[22] J.H.M. Disselhorst and L. van Wijngaarden. Flow in the exit of open pipes during acoustic
resonance. Journal of Fluid Mechanics, 99:293–319, 1980.
[23] P.E. Doak. Momentum potential theory of energy flux carried by momentum fluctuations.
Journal of Sound and Vibration, 131:67–90, 1989.
[24] P.E. Doak. Fluctuating total enthalpy as a generalized field. Acoust. Phys., 44:677–685, 1995.
[25] A.P. Dowling and J.E. Ffowcs Williams. Sound and Sources of Sound. Ellis Horwood Publish-
ers, Chichester, UK, 1983.
[26] V. Dubos, D. Keefe, J. Kergomard, J.P. Dalmont, and C.J. Nederveen. Theory of sound prop-
agation in a duct with a branched tube using modal decomposition. Acustica-acta acustica,
85:153–169, 1999.
[27] W.F. Durand. Aerodynamic Theory. Dover, NY, 1963.
[28] P.P.J.M. Durrieu, G.C.J. Hofmans, G. Ajello, R.J.J. Boot, Y. Auregan, A. Hirschberg, and
M.C.A.M. Peters. Quasi-steady aero-acoustic response of orifices. Journal of Acoustical Soci-
ety of America, 110:1859–1872, 2001.
20:23 18 Jul 2004 56 version: 18-07-2004

REFERENCES
[29] W. Eversman. Theoretical models for duct acoustic propagation and radiation. In Harvey H.
Hubbard, editor, Aeroacoustics of Flight Vehicles, Theory and Practice–Volume 2: Noise Con-
trol. the Acoustical Society of America, November 1993.
[30] R. Ewert and Schröder W. On the simulation of trailing edge noise with a hybrid les/ape
method. Journal of Sound and Vibration, 270:509–524, 2004.
[31] F. Farassat. Linear acoustic formulas for calculation of rotating blade noise. AIAA Journal,
19(9):1122–1130, 1981.
[32] J.E. Ffowcs Williams. Hydrodynamic noise. Annual Review of Fluid Mechanics, 1:197–222,
1969.
[33] J.E. Ffowcs Williams and D.L. Hawkings. Sound generated by turbulence and surfaces in
arbitrary motion. Philosophical Transactions of the Royal Society of London, A264:321–342,
1969.
[34] N.H. Fletcher and T.D. Rossing. The Physics of Musical Instruments. Springer–Verlag, New
York, second edition, 1998.
[35] M.E. Goldstein. Aeroacoustics. McGraw-Hill Book Company, Inc., New York, 1976.
[36] L. Gutin. On the sound field of a rotating propeller. Zhurnal Tekhnicheskoi Fiziki, 12:76–83,
1936. Original in Russian, translation: NACA TM 1195, 1948.
[37] W.G. Hill and P.R. Greene. Increased turbulent jet mixing rates by self-excited acoustic oscil-
lations. Transactions of the ASME, Journal of Fluids Engineering, page 520, 1977.
[38] A. Hirschberg. Aero-acoustics of wind instruments. In A. Hirschberg, J. Kergomard, and

G. Weinreich, editors, Mechanics of Musical Instruments, pages 291–369, Wien New York,
1995. International Centre for Mechanical Sciences, Udine, It., Springer-Verlag.
[39] A. Hirschberg, J.C. Bruggeman, A.P.J. Wijnands, and N. Smits. The whistler-nozzle and horn
as aero-acoustic sound sources in pipe systems. Acustica, 68:157–160, 1989.
[40] A. Hirschberg, J. Gilbert, R. Msallam, and A.P.J. Wijnands. Shock waves in trombones. Journal
of the Acoustical Society of America, 99(3):1754–1758, 1996.
[41] A. Hirschberg and C. Schram. A primitive approach to aeroacoustics. In Y. Auregan, A. Maurel,

and V. Pagneux, editors, Waves in fluids, pages 291–369, Wien New York, 1995. Cargese, fr,
Springer-Verlag.
[42] G.C.J. Hofmans, R.J.J. Boot, P.P.J. Durrieu, Y. Auregan, and A. Hirschberg. Aeroacoustic
response of a slit-shaped diaphragm in a pipe at low helmholtz number, part 1: quasi-steady
results. Journal of Sound and Vibration, 244:35–56, 2000.
[43] D.K. Holger, T.A. Wilson, and G.S. Beavers. Fluid mechanics of the edge tone. Journal of the
Acoustical Society of America, 62:1116–1128, 1977.
[44] M.S. Howe. Contributions to the theory of aerodynamic sound, with application to excess jet
noise and the theory of the flute. Journal of Fluid Mechanics, 71:625–673, 1975.
20:23 18 Jul 2004 57 version: 18-07-2004

REFERENCES
[45] M.S. Howe. Attenuation of sound in a low Mach number nozzle flow. Journal of Fluid Me-
chanics, 91:209–229, 1979.
[46] M.S. Howe. The dissipation of sound at an edge. Journal of Sound and Vibration, 70:407–411,
1980.
[47] M.S. Howe. Acoustics of Fluid-Structure Interactions. Cambridge University Press, Cam-
bridge, UK, 1998.
[48] M.S. Howe. Theory of Vortex Sound. Cambridge University Press, Cambridge, UK, 2002.
[49] H.H. Hubbard, editor. Aeroacoustics of Flight Vehicles: Theory and Practice. Volume 1 Noise
Sources; Volume 2 Noise Control (Nasa Reference Publication 1258). Acoustical Society of
America, 1995.
[50] H.H. Hubbard and K.P. Shepherd. Aeroacoustics of large wind turbines. Journal of the Acous-
tical Society of America, 89:2495–2508, 1991.
[51] K.U. Ingard. Influence of fluid motion past a plane boundary on sound reflection, absorption,
and transmission. Journal of the Acoustical Society of America, 31(7):1035–1036, 1959.
[52] K.U. Ingard and S. Labate. Acoustic circulation effects and the non-linear impedance of ori-
fices. Journal of the Acoustical Society of America, 22:211–218, 1950.
[53] U. Ingard and H. Ising. Acoustic non-linearity of an orifice. Journal of the Acoustical Society
of America, 42:6–17, 1967.
[54] J.D. Jackson. Classical Electrodynamics. John Wiley & Sons, New York, 3nd edition, 1999.
[55] P.L. Jenvey. The sound power from turbulence: a theory of the exchange of energy between the
acoustic and non-acoustic fields. Journal of Sound and Vibration, 131:37–66, 1989.
[56] D.S. Jones. The Theory of Electromagnetism. Pergamon Press, Oxford, 1964.
[57] D.S. Jones. Aerodynamic sound due to a source near a half plane. Journal of the Institute for
Mathematics and Applications, 9:114–122, 1972.
[58] M.C. Junger and D. Feit. Sound, Structures and their Interaction. MIT Press, Cambridge,
Massachusetts, 2nd edition, 1986.
[59] R.P. Kanwal. Generalized Functions: Theory and Technique. Birkhäuser, Boston, 2nd edition,
1998.
[60] J. Kergomard, A. Garcia, and G. Taguy. Analysis of higher order mode effects in an expansion
chamber using modal theory and equivalent electrical circuits. Journal of Sound and Vibration,
129:457–475, 1989.
[61] G.D.C. van Kuiken. Thermodynamics of irreversible processes: Applications to diffusion and
Rheology. J. Wiley & Sons, Chichester, UK, 1995.
[62] P.K. Kundu. Fluid Mechanics. Academic Press, San Diego, USA, 1990.
20:23 18 Jul 2004 58 version: 18-07-2004

REFERENCES
[63] L.D. Landau and E.M. Lifshitz. Fluid Mechanics. Pergamon Press, Oxford, 2nd edition, 1987.
[64] M.B. Lesser and D.G. Crighton. Physical acoustics and the method of matched asymptotic ex-
pansions. In W.P. Mason and R.N. Thurston, editors, Physical Acoustics Volume XI. Academic
Press, New York, 1975.
[65] M.B. Lesser and J.A. Lewis. Applications of matched asymptotic expansions methods to acous-
tics i. the webster horn equation and the stepped duct. Journal of the Acoustical Society of
America, 51:1664–1669, 1972.
[66] H. Levine and J. Schwinger. On the radiation of sound from an unflanged circular pipe. Physical
Review, 73:383–406, 1948.
[67] L. van Lier, S. Dequand, A. Hirschberg, and J. Gorter. Aeroacoustics of diffusers: An ex-
perimental study of typical industrial diffusers at reynolds numbers of O(105 ). Journal of the
[68] M.J. Lighthill. On sound generated aerodynamically I. Proceedings of the Royal Society of
London, Series A211:564–587, 1952.
[69] M.J. Lighthill. On sound generated aerodynamically II. Proceedings of the Royal Society of
London, Series A222:1–32, 1954.
[70] M.G. Lilley. The source of aerodynamic noise. International Journal of Aeroacoustics, 2:241–
253, 2004.
[71] P. in ’t panhuis. Calculations of the sound pressure reflection coefficient and the acoustic
end correction of a semi-infinite circular pipe issuing a subsonic cold or hot jet with co-flow.
Internal report, MWL, Aeronautical and Vehicle Engineering, Project EU-ARTEMIS GRD1-
2000-25507, KTH, Stockholm, 2003. Supervisors H. Bodén, S.W. Rienstra.
[72] F.E. Marble and S.M. Candel. Acoustic disturbance from gas non-uniformities convected
through a nozzle. Journal of Sound and Vibration, 55(2):225–243, 1977.
[73] A. Michalke. On the propagation of sound generated in a pipe of circular cross-section with
uniform mean flow. Journal of Sound and Vibration, 134:203–234, 1989.
[74] J.W. Miles. The diffraction of sound due to right-angle jointts in rectangular tubes. Journal
[75] W. Möhring. On vortex sound at low mach number. Journal of Fluid Mechanics, 85:685–691,
1978.
[76] W. Möhring. Energy conservation, time-reversal invariance and reciprocity in ducts with flow.
Journal of Fluid Mechanics, 431:223–237, 2001.
[77] L. Mongeau, H. Kook, D.V. Brown, and S.I. Zorea. Analysis of interior pressure oscillations
induced by flow over vehicle openings. Noise Control Engineering Journal, 45:223–234, 1997.
[78] C.L. Morfey. Amplification of aerodynamic noise by convected flow inhomogeneities. Journal
of Sound and Vibration, 31:391–397, 1973.
20:23 18 Jul 2004 59 version: 18-07-2004

REFERENCES
[79] C.L. Morfey. Sound radiation due to unsteady dissipation in turbulent flows. Journal of Sound
and Vibration, 48:95–111, 1976.
[80] C.L. Morfey. The role of viscosity in aerodynamic sound generation. International Journal of
Aeroacoustics, 2:230–240, 2003.
[81] P.M. Morse and H. Feshbach. Methods of Theoretical Physics Volume I. McGraw-Hill Book
Company, Inc., New York, 1953.
[82] P.M. Morse and K.U. Ingard. Theoretical Acoustics. McGraw-Hill Book Company, Inc., New
York, 1968.
[83] E.A. Müller and F. Obermeier. Vortex sound. Fluid Dynamics Research, 3:43–51, 1988.
[84] R.M. Munt. The interaction of sound with a subsonic jet issuing from a semi-infinite cylindrical
pipe. Journal of Fluid Mechanics, 83(4):609–640, 1977.
[85] R.M. Munt. Acoustic radiation properties of a jet pipe with subsonic jet flow: I. the cold jet
reflection coefficient. Journal of Sound and Vibration, 142(3):413–436, 1990.
[86] R.E. Musafir. A discussion on the structure of aeroacoustic wave equations. In Proceedings of
the 4th Congress on Acoustics, Marseille, France, pages 923–926, Toulouse, Fr., April 1997.
Teknea edition.
[87] M.K. Myers. On the acoustic boundary condition in the presence of flow. Journal of Sound
and Vibration, 71(3):429–434, 1980.
[88] M.K. Myers. An exact energy corollary for homentropic flow. Journal of Sound and Vibration,
109:277–284, 1986.
[89] M.K. Myers. Generalization and extension of the law of acoustic energy conservation in a
nonuniform flow. In 24th Aerospace Sciences Meeting, Reno, Nevada, January 6–9 1986.
AIAA. AIAA 86–0471.
[90] M.K. Myers. Transport of energy by disturbances in arbitrary flows. Journal of Fluid Mechan-
ics, 226:383–400, 1991.
[91] C.J. Nederveen. Acoustical aspects of Woodwind instruments. Northern Illinois University
Press, DeKalb, USA, 2nd edition, 1998.
[92] P.A. Nelson, N.A. Halliwell, and P.E. Doak. Fluid dynamics of flow excited resonance, part 2:
Flow acoustic interaction. Journal of Sound and Vibration, 91:375–402, 1983.
[93] M.P. Norton. Fundamentals of Noise and Vibration Analysis for Engineers. Cambridge Uni-
versity Press, Cambridge, 1989.
[94] F. Obermeier. Aerodynamic sound generation caused by viscous processes. Journal of Sound
and Vibration, 99:111–120, 1985.
[95] N.C. Ovenden. Near cut-on/cut-off transitions in lined ducts with flow. In 8th AIAA/CEAS
Aeroacoustics Conference, Breckenridge, CO, June 17-19 2002. AIAA 2002-2445.
20:23 18 Jul 2004 60 version: 18-07-2004

REFERENCES
[96] N.C. Ovenden. A uniformly valid multiple scales solution for sound in flow ducts with cut-on
cut-off transition. Journal of Sound and Vibration, Accepted for publication, 2004.
[97] N.C. Ovenden, W. Eversman, and S.W. Rienstra. Cut-on cut-off transition in flow ducts: com-
paring multiple-scales and finite-element solutions. In 10th AIAA/CEAS Aeroacoustics Confer-
ence, Manchester, UK, May 10-12 2004. AIAA 2004-2945.
[98] V. Pagneux and B. Froelich. Influence of low mach number shearflow on acoustic propagation
in ducts. Journal of Sound and Vibration, 246:137–155, 2001.
[99] A.R. Paterson. A First Course in Fluid Dynamics. Cambridge University Press, Cambridge,
1983.
[100] M.C.A.M. Peters and A. Hirschberg. Acoustically induced periodic vortex shedding at sharp
edged open channel ends: simple vortex models. Journal of Sound and Vibration, 161:281–299,
1993.
[101] M.C.A.M. Peters, A. Hirschberg, A.J. Reijnen, and A.P.J. Wijnands. Damping and reflection
coefficient measurements for an open pipe at low Mach and low Helmholtz numbers. Journal
of Fluid Mechanics, 256:499–534, 1993.
[102] A.D. Pierce. Acoustics: an Introduction to its Physical Principles and Applications. McGraw-
Hill Book Company, Inc., New York, 1981. Also available from the Acoustical Society of
America.
[103] A. Powell. Theory of vortex sound. Journal of Acoustical Society of America, 36:177–195,
1964.
[104] A. Powell. Some aspects of aeroacoustics: From Rayleigh until today. Journal of Vibration
and Acoustics, 112:145–159, 1990.
[105] L. Prandtl and O.G. Tietjens. Fundamental of Hydro- and Aeromechanics. Dover, NY, 1957.
[106] M.C. Quinn and M.S. Howe. On the production and absorption of sound by lossless liners in
the presence of mean flow. Journal of Sound and Vibration, 97(1):1–9, 1984.
[107] T.L. Redmore and K.A. Mulholland. The application of mode coupling theory to the trans-
mission of sound in sidebranch of a rectangular duct system. Journal of Sound and Vibration,
85:323–331, 1982.
[108] D. Ricot, V. Maillard, C. Bailly, and D. Juve. Numerical simulation of the unsteady flow past a
cavity and application to sun roof-buffeting. In 7th AIAA/CEAS Aeroacoustics Conference and
Exhibit, Maastricht, The Netherlands, 28-30 May 2001. AIAA paper No 2001-2112.
[109] S.W. Rienstra. On the acoustical implications of vortex shedding from an exhaust pipe. Trans-
actions of the ASME, Journal of Engineering for Industry, 103(4):378–384, 1981.
[110] S.W. Rienstra. Sound diffraction at a trailing edge. Journal of Fluid Mechanics, 108:443–460,
1981.
[111] S.W. Rienstra. A small Strouhal number analysis for acoustic wave–jet flow–pipe interaction.
Journal of Sound and Vibration, 86(4):539–556, 1983.
20:23 18 Jul 2004 61 version: 18-07-2004

REFERENCES
[112] S.W. Rienstra. Acoustic radiation from a semi-infinite annular duct in a uniform subsonic mean
flow. Journal of Sound and Vibration, 94(2):267–288, 1984.
[113] S.W. Rienstra. Contributions to the theory of sound propagation in ducts with bulk-reacting
lining. Journal of the Acoustical Society of America, 77(5):1681–1685, 1985.
[114] S.W. Rienstra. Hydrodynamic instabilities and surface waves in a flow over an impedance wall.
In G. Comte-Bellot and J.E. Ffowcs Williams, editors, Proceedings IUTAM Symposium ’Aero-
and Hydro-Acoustics’ 1985 Lyon, pages 483–490, Heidelberg, 1986. Springer-Verlag.
[115] S.W. Rienstra. 1-D reflection at an impedance wall. Journal of Sound and Vibration, 125:43–
51, 1988.
[116] S.W. Rienstra. Sound transmission in slowly varying circular and annular ducts with flow.
[117] S.W. Rienstra. A classification of duct modes based on surface waves. Wave Motion, 37(2):119–
135, 2002.
[118] S.W. Rienstra. Sound propagation in slowly varying lined flow ducts of arbitrary cross section.
[119] S.W. Rienstra and Hirschberg A. An Introduction to Acoustics. Report IWDE 01-03 May 2001,
Eindhoven University of Technology, Eindhoven, NL, 2001.
[120] D. Rockwell. Oscillations of impinging shear layers. AIAA Journal, 21:645–664, 1983.
[121] D. Rockwell and E. Naudascher. Self sustaining oscillations of flow past cavities. Journal of
Fluid Engineering, Trans. ASME, 100:152–165, 1978.
[122] M. Roger. Noise from moving surfaces. In J. Anthoine, editor, Advances in Aeroacoustics and
Applications, Brussels, 2004. Von Karman Institute. Lecture Notes Series 2004-5.
[123] N.B. Roozen, M. Bockholts, P. van Eck, and A. Hirschberg. Vortex sound in bass-reflex ports
of loudspeakers: part i & ii. Journal of the Acoustical Society of America, 104:1914–1924,
1998.
[124] P.G. Saffman. Vortex Dynamics. Cambridge Monograph on Mechanics and Applied Mathe-
matics. Cambridge University Press, Cambridge, UK, 1992.
[125] S. Scheichl. On the calculation of the transmission line parameters for long tubes using the
method of multiple scales. The Journal of the Acoustical Society of America, 115:534–555,
2004.
[126] H. Schlichting. Boundary-layer Theory. McGraw-Hill Book Company, Inc., New York, 6th
edition, 1968.
[127] C. Schram and A. Hirschberg. Application of vortex sound theory to vortex-pairing noise:
sensitivity to errors in flow data. Journal of Sound and Vibration, 266:1079–1098, 2003.
[128] A.H. Shapiro. The dynamics and thermodynamics of compressible flow, volume I. The Ronald
Press Comp., NY, 1953.
20:23 18 Jul 2004 62 version: 18-07-2004

REFERENCES
[129] P. Sijtsma, S. Oerlemans, and H. Holthusen. Location of rotating sources by phased array
measurements. In 24th Aerospace Sciences Meeting, Maastricht, NL, 2001. AIAA. AIAA
2001–2169.
[130] M.J.T. Smith. Aircraft Noise. Cambridge University Press, Cambridge, 1989.
[131] C.K.W. Tam. Noise from high speed jets. In J. Anthoine and C. Schram, editors, Advances in
Aeroacoustics, VKI Lecture Series 2001-2, pages 1–34, Brussels, 2001. Von Karman Institute
for Fluid Dynamics.
[132] S. Temkin. Elements of Acoustics. Acoustical Society of America, New York, 2001. Reprint.
[133] M.C. Thompson, K. Hourigan, and M.C. Welsh. Acoustic sources in a tripped flow past a
resonator tube. AIAA Journal, 30:1484–1491, 1992.
[134] J.M. Tyler and T.G. Sofrin. Axial flow compressor noise studies. Transactions of the Society
of Automotive Engineers, 70:309–332, 1962.
[135] M.P. Verge, B. Fabre, A. Hirschberg, and A.P.J. Wijnands. Sound production in recorderlike
instruments, I: Dimensionless amplitude of the internal acoustic field. Journal of the Acoustical
Society of America, 101:2914–2924, 1997.
[136] M.P. Verge, A. Hirschberg, and R.R. Caussé. Sound production in recorderlike instruments, II:
A simulation model. Journal of the Acoustical Society of America, 101:2925–2939, 1997.
[137] R. Verzicco, A. Iafrati, G. Riccardi, and M. Fatica. Analysis of the sound generated by the
pairing of two axisymmetric co-rotating vortex rings. Journal of Sound and Vibration, 200:347–
358, 1997.
[138] S. Wagner, R. Bareiss, and G. Guidati. Wind Turbine Noise. Springer Verlag, Berlin, 1996.
[139] M.C. Welsh and A.N. Stokes. Flow-resonant sound interaction in a duct containing a plate, part
i: Semi-circular leading edge. Journal of Sound and Vibration, 95:305–323, 1984.
[140] L. van Wijngaarden. On the oscillations near and at resonance in open pipes. Journal of
Engineering Mathematics, 11:225–240, 1968.
[141] L. van Wijngaarden. One-dimensional flow of liquids containing small gas bubbles. Annual
Review of Fluid Mechanics, 4:369–396, 1972.
[142] T.A. Wilson, G.S. Beavers, M.A. De Coster, D.K. Holger, and D. Regenfuss. Experiments on
the fluid mechanics of whistling. Journal of the Acoustical Society of America, 50:366–372,
1971.
[143] S. Ziada and E.T. Bühlmann. Self-excited resonances of two side-branches in close proximity.
Journal of Fluids and Structures, 6:583–601, 1992.
20:23 18 Jul 2004 63 version: 18-07-2004

An Introduction To Aeroacoustics

Uploaded by

Copyright:

Available Formats

An Introduction To Aeroacoustics

Uploaded by

Document Information

Original Title

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

An Introduction To Aeroacoustics

Uploaded by

Copyright:

Available Formats

An introduction to aeroacoustics

A. Hirschberg∗ and S.W. Rienstra∗∗

20:23 18 Jul 2004 1 version: 18-07-2004

20:23 18 Jul 2004 1 version: 18-07-2004

2.1 Mass, momentum and energy equations

20:23 18 Jul 2004 2 version: 18-07-2004

20:23 18 Jul 2004 3 version: 18-07-2004

2.2 Constitutive equations

e = e(ρ, s). (11)

Comparison with the fundamental equation of thermodynamics,

provides thermodynamic equations for the temperature T and the pressure p:

As p is also a function of ρ and s we have:

20:23 18 Jul 2004 4 version: 18-07-2004

As sound is defined as isentropic (ds = 0) pressure-density perturbations, the speed of sound c =

σi j = 2η Di j − 13 Dkk δi j + µv Dkk δi j (23)

where the rate of strain tensor Di j is defined by:

20:23 18 Jul 2004 5 version: 18-07-2004

2.3 Approximations and alternative forms of the basic equations

20:23 18 Jul 2004 6 version: 18-07-2004

because ∇·(∇×ψ) = 0, the vector stream function ψ is related to the vorticity:

ω = ∇×v = ∇×(∇×ψ) (37)

20:23 18 Jul 2004 7 version: 18-07-2004

3 Free space acoustics of a quiescent fluid

3.1 Order of magnitudes

20:23 18 Jul 2004 8 version: 18-07-2004

20:23 18 Jul 2004 9 version: 18-07-2004

3.2 Wave equation and sources of sound

20:23 18 Jul 2004 10 version: 18-07-2004

3.3 Green’s function and integral formulation

20:23 18 Jul 2004 11 version: 18-07-2004

for any well-behaving function F(t).

G(x, t| y, τ ) = 0 for t < τ. (63)

20:23 18 Jul 2004 12 version: 18-07-2004

3.4 Inverse problem and uniqueness of source

20:23 18 Jul 2004 13 version: 18-07-2004

3.5 Elementary solutions of the wave equation

We consider two elementary solutions of the homogeneous wave equation (q = 0):

which will be used as building blocks to obtain more complex solutions:

20:23 18 Jul 2004 14 version: 18-07-2004

20:23 18 Jul 2004 15 version: 18-07-2004

In linear approximation (in â/a0 ) we have:

v̂r (a0 ) = iωâ. (84)

20:23 18 Jul 2004 16 version: 18-07-2004

20:23 18 Jul 2004 17 version: 18-07-2004

When a person jumps up and down in the boat, he produces an

3.6 Acoustic energy and impedance

and the momentum equation in the form:

20:23 18 Jul 2004 18 version: 18-07-2004

The intensity I, defined as

∇·I = −D . (103)

The real part is given by

20:23 18 Jul 2004 19 version: 18-07-2004

3.7 Free space Green’s function

20:23 18 Jul 2004 20 version: 18-07-2004

3.8 Multipole expansion

where θ is the angle between x and y, and

Utilising the symmetry of r as a function of x and y, this is equivalent to

The acoustic field is then given by

20:23 18 Jul 2004 21 version: 18-07-2004

∇·I = −D . (103)

q(x, t) = ρ0 V (t)δ(x − x s (t)) . (134)

φi = exp i K M X − i K R cos( − s ) . (147)

φ(x, y) = exp i K M X − i K R F(s ) + F( s ) (148)

p(x, y) = exp i K M X − i K R F(s ) + F( s )

in which the perturbations p and ρ are defined by:

p (x, t) = A e iωt−ikx +B e iωt+ikx , (188)

p (x, t) ≡ p (x, ω) e iωt , v (x, t) ≡ v (x, ω) eiωt (190)

p = Z(v · n) for x ∈ ∂D. (192)

p (x, y, z) = ψn (y, z) An e−iκn x +Bn e iκn x . (195)

= δ(x) f (t) − δ(x)g(t) − 1c H (x) f t − xc + 1c H (−x)g t + xc ,

= δ (x)( f (t) − g(t)) − 1c δ(x)( f (t) + g (t)) + 1

S(x3 )v3 (x3 , t) = S I v3 (x I , t) = S I I v3 (x I I , t). (209)