An introduction to Flow Acoustics

Mats bom

TRITA-AVE 2006:04 Copy right Mats bom 2006

ISSN 1651-7660 matsabom@kth.se
ISRN/KTH/AVE/N-06/04-SE
ii
Preface
This publication is based on lecture notes used for the course 4B1136 Flow
Acoustics offered at KTH to 4:th year students. The present English text is an
extended version of lecture notes in Swedish used by the author since the early 90:s.
It can be noticed that * marked sections are considered extra reading for the course
4B1136. Dr. Susann Boij is acknowledged for helping the author to correct errors and
improve the text in this English version.

KTH 2006-01-24
Mats bom

First revision
Minor errors have been corrected and more problems and solutions to the problems
have been added.

KTH 2007-01-19

Second revision
Dr. Susann Boij have made some improvements in the text, corrected a few errors
and added problems and solutions from exams given during 2007. From the year
2008, the course name is SD2155 Flow Acoustics.

KTH 2008-01-15

Third revision
A few errors have been corrected and problems and solutions from exams given
during 2008 have been added.

KTH 2009-01-12

Fourth revision
Problems and solutions from exams given during 2009 have been added.

KTH 2010-01-11

iii
iv
List of contents

Chapter 1: Introduction 1
1.1 Background 1

1.2 Some mathematical tools 2

1.21 The total (convective) time derivative
1.22 The gradient of a scalar field
1.23 The divergence of a vector field
1.24 Cartesian tensor notation

1.3 The fundamental equations of fluid mechanics 7

1.4 The classical wave equation 10

1.41 The speed of sound
1.42 Solutions to the classical wave equation
1.43 Acoustic impedance
1.44 Point sources and Greens functions
1.45* General solution
1.46 Acoustic energy and intensity

1.5 Multi-pole expansion of sound fields 25

Problems 31
Solutions 35

Chapter 2: Generation of sound in fluids 41

2.1 The inhomogeneous wave equation 41
2.11 Non uniqueness of the source term

2.2 Lighthills theory or acoustic analogy 44

2.21 Limitations in Lighthills theory
2.22 Interaction between flow and sound

2.3 The three source terms 47

2.31 Compact sources and dimensionless numbers
2.32 The monopole source term
2.33 The dipole source term
2.34 The quadrupole source term
2.35 Curles equation

2.4 Jet noise 56

2.41 Low Mach number jets

2.5* Noise from propellers 58

Problems 63
Solutions 71

v
Chapter 3: The effect of mean flow 79
3.1 The convective wave equation 79

3.2 Plane wave sound fields 81

3.21 Wave number and phase speed
3.22 Characteristic impedance
3.23 Scattering at a mean flow discontinuity (a vortex sheet)

3.3 Acoustic intensity 86

3.4 Sound in ducts 88

3.41 Rectangular and circular ducts
3.42* Sound from a ducted rotor

Problems 95
Solutions 109

Chapter 4: The effect of source motion 121

4.1 A moving point source 123

4.11 Doppler shift
4.12 Source moving along a straight path
4.13 The effect of a mean flow and the general Doppler shift formula

4.2 Application to a monopole, dipole and quadrupole source 129

4.3* The Ffowcs Williams&Hawkings equation 131

Problems 133
Solutions 135

Chapter 5: Self-sustained oscillations and whistles 141

5.1 Fluid driven whistles 144
5.11 Type 1 whistle
5.12 Type 2 whistle
5.13 Other types of whistles

5.2* The Rijke tube a thermo-acoustic whistle 149

Problems 151
Solutions 155

APPENDIX: On the use of multi-ports for flow ducts 157

vi
 List of symbols
c adiabatic speed of sound

d, D diameter

e unit vector

e energy per unit mass or volume

f frequency

f force per unit volume

F force

ht total enthalpy per unit mass

i imaginary unit or integer

I acoustic intensity

k wave number

L length

m mass production per unit volume

M Mach number

n normal vector for a surface

p pressure (total)

p acoustic pressure

p complex valued acoustic pressure or Fourier transform of p

Q acoustic volume flow

q heat flux vector

q acoustic volume flow per unit volume

q acoustic power per unit volume

r radial distance

r position vector source-receiver

vii
R acoustic reflection coefficient

S area of a surface or the 2-D set of points defining a surface or acoustic source
strength

s entropy per unit mass or acoustic source strength per unit volume

t time

T period length in time, characteristic time scale, acoustic transmission coefficient or

absolute temperature

TL Transmission loss (see Eq. 20 in the Appendix)

u velocity field in a fluid

u acoustic velocity field

U mean flow velocity, subscript 0 denotes a constant velocity

U flow speed

V volume of a region or the 3-D set of points defining a volume

x, y position vectors in space and (in the Appendix) state vectors

x, y the magnitude of x and y

xi , yi Cartesian co-ordinates in space (i = 1, 2, 3)

Greek

acoustic velocity potential

acoustic efficiency factor

dynamic viscosity

kinematic viscosity

density

angular frequency 2 f

displacement

time

viii
Mathematical

a complex valued amplitude for a signal a(t ) a exp(it )

a Fourier transform [ exp(it ) ] of the time-varying quantity a(t)

a the magnitude of a where a can be a vector

T
1
T 0
a time average of a time-varying quantity = limT a (t )dt

a the RMS value of a(t) defined as a2

a(t )
a time derivative =
t

a fluctuating (acoustic) part of a time varying quantity a (t )

Ai Cartesian components of the vector A ( A1 , A2 , A3 )

Re[..] real part of a complex valued quantity

(..)* complex conjugation

the nabla-operator which in a Cartesian system equals ( x1 , x2 , x3 )

2 the Laplace operator 2 x12 2 x22 2 x32

D
the absolute (total) or convective time derivative = U
Dt t

Subscripts

d dipole
e emission time
i, j, k denotes components of a Cartesian tensor
m monopole
ph phase speed
q quadrupole
x denotes the receiver or field domain
y denotes the source domain
0 constant (reference) value

ix
x
 Chapter One

Introduction

In this chapter some background to the subject is given both historically and from an
engineering point of view. A short summary of basic fluid mechanics, classical acoustics (the
wave equation) and mathematics - mostly vector analysis useful later in the course - is also
given.

1.1 Background
In this course the generation and propagation of sound waves in fluids (gases or liquids) is
studied. The origin of this subject goes back already to Newton who proposed a model for
sound waves in elastic media. This model however gave an incorrect value for the speed of
sound in a gas and a more correct model was later suggested by Laplace, who also gave the
first correct derivation of the classical wave equation. During the 19:th century works by
Kirchhoff, Helmholtz and Rayleigh laid the ground for the classical theory of sound in fluids
[1]. One limitation in these works is the lack of models for sound generation, i.e., there was
no source terms present in the wave equation. The only way to model sound generation in
classical acoustics is via a vibrating solid boundary, which means that sound generation by
combustion or other unsteady fluid processes cannot be handled. This was not a large issue
one hundred years ago but became a very important problem during the 1950s with the
development of commercial jet powered aircrafts. The need to reduce the noise from jet
engines put a focus on the lack of understanding of the underlying mechanisms. This
stimulated James Lighthill to develop his theory for aeroacoustic sound production [2,3] and
lay the foundation for a new field of research aeroacoustics.

1
The basic idea in Lighthills theory is that any unsteady flow will also produce sound a fact
confirmed by our everyday experience. For instance the noise or tone produced by wind
blowing on a string (Aeolian tones), e.g., a telephone wire or an antenna, has been known
since B.C. (wind harps). Many musical instruments are also based on the generation of
tones by a flow, e.g.; horns, flutes and organs. Also the human voice (the vocal cords) is a
fluid driven acoustic source. In our modern society fluid machines, i.e., devices delivering or
removing energy from a fluid, are one of the most common machine types and also sources of
noise. Examples are fans, pumps, turbines, compressors, IC-engines and jet engines.

Beside sound generation a flow field will also influence the sound propagation because the
speed of propagation will be higher in the downstream direction than in the upstream
direction. This will cause a downward shift of resonance frequencies in a pipe with a factor
(1 M 2 ) , where M is the Mach number, and will lead to curved propagation paths for sound
traveling through regions with mean flow gradients. To take this kind of effects into account it
is necessary to modify the operator in the classical wave equation.

1.2 Some mathematical tools

1.21 The total (convective) time derivative

In fluid mechanics it is often of interest to refer the change of a quantity, e.g., velocity or
temperature to the motion of a fluid particle1. Assume that a scalar field is given as a
function of the space x and time t co-ordinates. If we introduce a Cartesian co-ordinate system
x1, x2 and x3, then the path of a certain fluid particle can be written as
x(t ) x1 (t ), x 2 (t ), x3 (t ) . The time variation for the scalar field along this path will be

(t ) x1 (t ), x 2 (t ), x3 (t ), t . (1-1)

The rate of change with time for is obtained by calculating the total time derivative of (1-1)
using the chain rule

1
A fluid particle is a small volume over which the variation of a field quantity is negligible but still the volume
is much larger than the microscopic (atomic) scale.

2
 d dx1 dx2 dx3
. (1-2)
dt x1 dt x 2 dt x3 dt t

The velocity of the fluid particle is given by u dx1 dt , dx 2 dt , dx3 dt which implies that
equation (1-2) can be written

d
u , , , (1-3)
dt t x1 x 2 x3

where the . denotes a scalar product. This total or so called convective time derivative is
normally denoted D/Dt and, if we introduce the so called nabla-operator defined by

, , , (1-4)
x1 x 2 x3

we can define the convective time derivative by the following operator formula

D
u . (1-5)
Dt t

It can also be noted that since the nabla-operator can be interpreted as a vector, equation (1-5)
is given on vector form and is valid in any co-ordinate system.

1.22 The gradient of a scalar field

By operating with the nabla-operator on a scalar field (x,t) we get a vector field
/ x1 , / x 2 , / x3 or grad(. The interpretation of this vector field is the

direction of maximum growth of the scalar field. This is seen by studying the change in
when we move a small distance dx dx1 , dx 2 , dx3 for a fixed t. Differentiating gives

d dx1 dx 2 dx3 ,
x1 x 2 x3

which, using the definition of the scalar product and the nabla-operator, can be written as

d dx ds cos , (1-6)
dx

where ds dx and is the angle between dx and the vector . As can be seen from

equation (1-6) the direction of maximum growth is parallel with ( = 0). For = we
have maximum decrease and for / 2 there is no change.

1.23 The divergence of a vector field

If we form a scalar product between nabla and a vector field A(x,t) we obtain a scalar field
A A1 / x1 A2 / x 2 A3 / x3 (or div(A)), which is related to the flow out of a small
volume surrounding x. To prove this statement we first define the outflow produced by A
over a closed surface S in 3-D via the following integral

A ndS lim N A
k
k n k S k , (1-7)
S

where n is the outward unit normal, Sk is a division of S into smaller sub-surfaces with an
area that goes to zero when the number (N) of sub-surfaces goes to infinity, see Figure 1.1.

S
nk Sk

Figure 1.1 Definition of the surface integral, equation (1-7).

We will now calculate the surface integral of A over a small cubical volume V and for
simplicity first look on the contribution from the two surfaces with normal in the x1-direction,
see Figure 1.2,

A1
A ( x
x2 x3
1 1 x1 / 2, x 2 , x3 ) A1 ( x1 x1 / 2, x 2 , x3 ) dx 2 dx3
x1
V V ,

4
where V x1 x 2 x3 , goes to zero in the limit of a vanishing volume and the mean value

theorem has been applied. Using the same procedure for the x2 and x3 directions gives for the
complete surface integral over V

A1 A2 A3
A ndS x V V A V . (1-8)
S 1 x 2 x3

In the limit, this result implies that

1
A lim V0
V A ndS .
S
(1-9)

A result that sometimes is used as a definition of div(A) and which is valid independent of the
shape of V.

x3

x = (x1, x2, x3)

x3
x1
x2
x2

x1
Figure 1.2 The cubical volume used for the derivation of equation (1-8).

Example 1-1: If A is the velocity field u in a fluid then the integral in (1-9) can be interpreted
as the rate of change of volume of a fluid particle with volume V. For an incompressible
fluid the volume occupied by a fluid particle is constant which implies that u 0 .

V udt ndS

Outflow from the surface

element dS during dt
___________________________________________________________________________

5
Assume now a volume V bounded by a closed surface S and let us divide the volume into
small cubical volumes Vk. For each of the small volumes we can apply equation (1-8) and
add all the results

A ndS A
k S k k
k Vk .

As illustrated by Figure 1.3 the surface integral contributions from two surfaces that coincide
will cancel out. Therefore in the limit of vanishing volume sizeVk, the surface integral will
approach the integral over S. The sum on the right hand side will in the limit approach a
volume integral over V. We then arrive at the formula known as Gauss theorem

A ndS A dV ,
S V
(1-10)

S
V

Figure 1.3 A volume V divided into a number of small cubical volumes.

Example 1-2: If we apply Gauss theorem to a 1-D field (say) along the x1 direction we get

A
x2

S 0 A1 ( x 2 ) A1 ( x1 ) S 0 1 dx1 ,
x1
x1
i.e., it reduces to a well-known result from standard calculus.

Example 1-3: Gauss theorem can also be applied to a scalar field if we first multiply the
equation with a constant vector a

a ndS a ndS a dV a dV ,
S S V V

which since a is arbitrary implies: ndS dV .

S V

6
1.24 Cartesian tensor notation
To facilitate the writing of equations which are valid in any Cartesian co-ordinate system one
can introduce a compact notation known as Cartesian tensor notation. This means that all
quantities are written on component form relative a set of Cartesian axes x1, x2 and x3. The
equation will then have the same structure in any Cartesian system, i.e., its form will be
invariant (unchanged). For instance the vector field A is written Ai, where it is understood that
the index i always assumes the values 1 to 3. There is also a convention of summation which
says that if an index is repeated twice then summation from 1 to 3 is understood, e.g., div(A)
can be written as

Ai
div( A) . (1-11)
xi

A scalar product between two vectors can be written as A B Ai Bi . It can be noted that

combinations like xi xi Ai , i.e., more than two equal indices, are not allowed since they will
change form when transformed to a new Cartesian system.

1.3 The fundamental equations of fluid mechanics

The fundamental equations of fluid mechanics can be seen as conservation laws, i.e.,
conservation of mass, momentum and energy. If we start by looking at the conservation of
mass the following must hold for a control volume V fixed in space

t u ndS m dV
dV , (1-12)

V

S

V

rate of change of mass within V mass flow out from V mass produced in V

where is the fluid density, u is the velocity field and m is the rate of mass production per m3.
The mass production can for instance be used to represent injection of mass by a small pipe
opening (exhaust outlet) or the collapse of cavitation bubbles2 in a liquid. Using Gauss
theorem we can rewrite (1-12) as

t u m dV 0,
V

2
A cavitation bubble is a small bubble formed in a liquid when the local pressure is reduced below the vapor
pressure.

7
which since the control volume is arbitrary and the integrand a continuous function, implies
that

u m . (1-13)
t

Equation (1-13) is the equation for conservation of mass or the continuity equation.

For the conservation of momentum we start from Newtons law of motion, which when
applied to a fixed control volume V in a fluid gives

t
udV uu n dS f
dV f S dS ,
V (1-14)

V


S



V

S

rate of change of momentum in V flow of momentum out from V external force field internal force field

where the forces are split in external (volume) forces fV, e.g., electromagnetic fields in a
plasma or gravity, and surface forces fS associated with the interaction between the fluid
volume V and the surrounding fluid. The surface forces can be related to the state of the fluid
and various constitutive models exist. The most commonly used is the so called Newtonian
fluid model which states that

f s ,i pni ij n j , (1-15)

where p is the thermodynamic pressure and ij is the viscous stress tensor. This tensor is

related to the fluid kinematic state via

u u j 2 u k
ij i ij , (1-16)
x j xi 3 x k
where is the dynamic viscosity and ij is Kroneckers delta3. If we write equation (1-14) in

Cartesian tensor form and apply Gauss theorem, we again get a volume integral which this
time implies the following conservation equation

1 if i j
3
ij
0 if i j

8

u i u i u j p ij f V ,i . (1-17)
t x j xi x j

Example 1-4: For an ideal fluid there are no losses which mean that we can neglect the
viscous term in (1-17). Equation (1-17) then reduce to

u i ui u j p f V ,i .
t x j xi

The left hand of this equation can, by using the equation of continuity (1-13), be written as

u j

u i u i u j u i u i u j u i
t
mu i Du i ,

t x j t x j x j Dt

which when inserted in the equation gives

Du i p
f V ,i mu i . (1-18)
Dt xi

For the case of no mass injection and no external forces equation (1-18) reduces to the so
called Eulers equation for the flow of an ideal fluid.
___________________________________________________________________________

Using a similar procedure as above a conservation equation for the energy can also be
derived. In Cartesian tensor form this equation is

e u2 2

u i e u 2 2
pui qi ij ui f V ,i ui , (1-19)
t xi xi xi xi
where e is the internal energy per unit mass and qi is the heat flux vector. The number of
unknowns in the three conservation equations are 9 (p,, e, ui and qi), which is larger than the
number of equations (1+3+1=5) so more equations are needed to have a closed problem. The
missing equations are supplied by constitutive relationships for instance by relating the heat
flux vector to the temperature field T by Fouriers law:

9
 q i K T / xi , (1-20)

where K is the heat conductivity constant. Furthermore, for a single phase fluid with fixed
composition, i.e., no chemical reactions, an equation of state can be defined from, e.g.,
e e(T , ) . For the case of an ideal gas this reduces to: e CV T , where CV is the specific

heat at constant volume and unit mass.

For many cases a good approximation is that the change of state of a fluid particle is
adiabatic, i.e., the effect of heat conduction can be neglected. Based on the second law of
thermodynamics4 this is equivalent to stating that the entropy of a fluid particle stays the
same: Ds / Dt 0 . Using this a closed problem can be obtained from the conservation of
mass and momentum equations together with an equation of state in the form: p p ( , s ) .

A simplification of the energy equation will finally be written down for later use when we
discuss the definition of acoustic intensity in a flowing medium. If we neglect heat conduction
and viscosity and there are no external forces equation (1-19) simplifies to

ht p ht u i 0 , (1-21)
t xi

where ht e p u 2 2 is the total enthalpy per unit mass.

1.4 The classical wave equation

The classical wave equation is valid in an ideal (no losses) homogenous fluid at rest, i.e., the
pressure p0, density 0, etc. are constant and the flow field u is equal to zero everywhere5. We
will now study small disturbances of the fluid state around this equilibrium and introduce the
following notation: p(x, t ) p 0 p (x, t ) , (x, t ) 0 (x, t ) and u u (x, t ) , where the
primed quantities denote the disturbance (the acoustic field). Inserting these expressions in
the equations for conservation of mass and momentum, equations (1-13) and (1-17) keeping
only linear terms, i.e., neglecting all products of primed quantities, gives

4
Strictly speaking one must also assume that the process is reversible, i.e., that the fluid particle is in
thermodynamic equilibrium.
5
In the literature this is also referred to as a quiescent fluid.

10

t 0 u m

, (1-22)
u
0 p fV
t

where the source terms on the right hand side represent unsteady mass injection (m) or an
unsteady external force ( fV ). In the derivation of the classical wave equation these source
terms are not included, thus they will be left out in the following. The assumption of no
losses, in particular no heat conduction, implies that the entropy of a fluid particle is constant.
Therefore, an equation of state for the pressure can be written as: p p( , s0 ) . A Taylor
expansion of this expression around the equilibrium state gives
p
p p p 0 0 ... K 0 / 0 ..., (1-23)
0
where K 0 0 p 0 is the isentropic bulk modulus. Using this relationship to eliminate

the density in equation (1-22)1 and then performing the operation t 1 22 1 1 22 2

we obtain
1 2 p
2 p 0, (1-24)
c0 t
2 2

where 2 is the Laplace operator and c0 K 0 0 is the isentropic speed of sound6.

The classical wave equation neglects losses, a good approximation for gases and liquids like
water, except for propagation over long distances, high frequency sound and sound in narrow
pipes [4].

Example 1-5: The classical wave equation is obviously also satisfied by density disturbances
. From the linearized equation of motion (1-22)2 (with no source term !) it follows that

x

t x0
0 u dx p(x, t ) p(x 0 , t ), (1-25)

which implies that an acoustic velocity potential exists

6
Also referred to as the adiabatic speed of sound.

11
 x
(x, t ) (x 0 , t ) u dx .
x0

From (1-25) it follows that this velocity potential is related to the pressure via

0 p . (1-26)
t
From equation (1-26) it follows that this velocity potential satisfies the classical wave
equation. In more theoretical treatments of acoustics the velocity potential based version of
the wave equation is often preferred. From an experimental point of view the formulation
based of pressure is to prefer since this is the quantity we most easily can measure.
___________________________________________________________________________

1.41 The speed of sound

For an ideal gas the assumption of no heat conduction (adiabatic conditions) implies that [1]
the following equation of state is valid

p
, (1-27)
p 0 0

where is the specific heat ratio for the gas. From equation (1-27) it follows that the
adiabatic bulk modulus for an ideal gas is K 0 p 0 and that the adiabatic speed of sound is

p 0
c0 . (1-28)
0

Using the ideal gas law p RT , where R is the gas constant and T the absolute temperature
it follows that this can be written

c0 RT 0 . (1-29)

The gas constant can be related to the universal gas constant R0 = 8314 J/kmol.K via
R R0 / M , where M is the molar mass in kg.

12
Table 1.1 Data for some gases needed to calculate the speed of sound.
Gas Air (dry) N2 O2 CO2 CO He H2
1.40 1.40 1.40 1.30 1.40 1.67 1.41
M [kg/kmol] 28.96 28.01 32.00 44.01 28.01 4.00 2.02

For air the data in table 1.1 and equation (1-29) gives c0 = 331.2 m/s at 0 oC.

For water and other fluids no general formulas are available so measurements are needed. In
sea water the speed of sound can vary between approximately 1450 m/s to 1525 m/s
depending on pressure (depth), temperature and salt content. For sound propagation in water it
is also important to remember that a small portion of bubbles can drastically reduce the speed
of sound due the large change in the bulk modulus.

1.42 Solutions to the classical wave equation

First we can note that since the equation is linear the principle of superposition applies, i.e., if
p1 and p2 are two solutions then any linear combination p1 p2 , and constant, is

also a solution7.

The simplest solution is the 1-D or plane wave solution discovered by DAlembert which, if
we assume wave propagation in the x1 direction, can be written

p ( x1 , t ) p (t x1 / c0 ) p (t x1 / c0 ), (1-30)

where the sub-script + and denote propagation in the positive and negative x1 direction
respectively. As can be seen from equation (1-30) the general solution is a sum of two waves
that propagate, with the speed of sound and without change of shape, in the + and - x1
direction. Inserting (1-30) into the linearized equation of motion (1-22)2 gives for the x1
component of the acoustic particle velocity

u1 p t p t
0 ,
t t x1 t x1

7
Another property of the classical wave equation is that a solution can be reversed in time, i.e., if p(x,t) is a
solution then p(x,-t) is also a solution. This is a consequence of the time derivative being second order.

13
where t t x1 / c0 , t t x1 / c0 and the derivatives are t x1 1 / c0 and

t x1 1 / c0 . Using this we get after integration with respect to time

u1 ( x1 , t ) p (t x1 / c0 ) / 0 c0 p (t x1 / c0 ) / 0 c0 , (1-31)

where 0 c0 is called the characteristic impedance. For air at standard conditions (1.293 kg/m3
and 0 oC) this impedance has the value 428 Pa.s/m.

The result above can be generalized to plane waves propagating along an arbitrary direction
defined by a unit vector n

p (x, t ) p (t x n / c0 ) p (t x n / c0 ). (1-32)

Another important case where simple solutions exist to the wave equation is when spherical
symmetry is assumed, i.e., a wave field which only depends on the radius r from the origin. If
spherical symmetry is assumed the Laplace operator is reduced to

1 2
2 r . (1-33)
r 2 r r
If we insert this in the wave equation we obtain8
1 2 (rp ) 2 (rp )
0. (1-34)
c 2 t 2 r 2

In analogy with the plane wave solution it follows that this equation must have the solution

1 1
p ( r , t ) p (t r / c 0 ) p (t r / c 0 ) . (1-35)
r r

This solution is the sum of two waves, one that propagates out from the origin (+) and one
that propagates towards the origin (-). Both waves have an infinite amplitude at the origin,

1 2 p 1 2
8
r rp .
r 2 r r r r 2

14
i.e., exhibit a singularity. As shown in sub-section 1.44 this singularity can be related to a
point source at r = 0.

u r p
The linearized equation of motion in the radial direction: 0 , can be used to
t r
derive the corresponding velocity field. Inserting (1-35) we get after integration with respect
to time

1 1 1 1
u r p (t r / c0 ) p (t r / c0 ) P (t r / c0 ) P (t r / c0 ).
0 c0 r 0 c0 r 0r 2
0r 2
(1-36)
where P p (t ) dt and P p (t ) dt . In equation (1-36) the first two terms represent

acoustic waves propagating out from (+) and into (-) the origin while the last two terms
represent nearfield. In the limit of an incompressible fluid (c0 ) these last terms still exist
and represent kinetic energy in the fluid. The nearfields will not contribute to the acoustic
power radiated by a source, see sub-section 1.46. Note, since the wave equation is
homogenous the spherical solution can be referred to an arbitrary point x0 in space by putting
r x x0 .

Often in acoustics we study harmonic time dependence and assume that the wave field is
given on a complex form: p ( x, t ) p ( x) exp(it ) , where p is the complex valued amplitude
and ( 2 f ) the angular frequency. Since acoustics is based on a linear theory

superposition is allowed, which means that a general wave field can be built up by summation
of harmonic wave components, i.e., by a Fourier transform

1
p ( x, t )
2 p ( x, ) exp(it )d .

If we put a harmonic wave field into the wave equation (1-24) we get the Helmholtz equation

2 p k 2 p 0, (1-37)

15
where k / c0 is the wave number. The simplest way to include damping via, e.g., viscous
or thermal effects, in the classical wave equation, is to assume that the speed of sound is
complex valued c0 c0 1 i . Expressions for the damping can be found for instance in
Ref. [4].

A harmonic plane wave along x1 can be written as

p p expi (t x1 / c0 ) p expi (t kx1 ) p exp2i (t / T x1 / ) ,

where k 2 / and 2 /T . Here and T =1/f is the spatial and temporal period of the
wave. The spatial period is normally referred to as the wavelength. The wavelength can be
used to introduce a dimensionless frequency scale, the Helmholtz number. This number is
used to define acoustic similarity, e.g., to determine how the frequency scale should change if
we want to make a scale model. It can also be used to determine if a problem is of low,
mid or high frequency type (see Example 1-6). The Helmholtz number (He) is defined as

He 2 L / kL , (1-38)

where L is a characteristic length scale for the problem studied.

Example 1-6: In a cavity (room) with a typical cross dimension L the first acoustic resonance
(standing waves) can be expected when / 2 L , i.e., when He . For Helmholtz numbers
much less than there are no standing waves in the cavity and we can expect the acoustic
pressure to be more or less constant in the cavity (low frequency regime). For He of the
order of we have spatial variations of the acoustic pressure and a detailed analysis of the
field is needed (mid frequency regime). Finally when He >> the distinct standing wave
character is lost and the field can be regarded as a sum of randomly distributed and equally
strong propagating plane waves (high frequency regime). This is normally called a diffuse
sound field and leads to the energy based acoustic model used for large rooms [1].

16
1.43 Acoustic impedance
Since acoustics is a linear theory, the ratio between a complex valued pressure and a complex
valued velocity component at a certain field point is independent of the sources creating the
sound field. This ratio called the specific acoustic impedance is formally defined as

p (x, )
Z s (x, ) , (1-39)
u (x, ) n

where n is a unit vector, and it characterizes the sound field at the point in question. In some
applications, e.g., duct acoustics, acoustic impedance defined as the ratio of complex pressure
and acoustic volume flow over a surface with constant acoustic pressure is used.

In a field where a plane wave propagates along a certain direction, the specific acoustic
impedance in the direction of propagation equals the characteristic impedance for the plane
wave, i.e., Z s 0 c0 . For the case of an outgoing spherical and harmonic wave we have from
equations (1-35) and (1-36)

0 c0
Z s (r , ) . (1-40)
1
1
ikr

As can be seen from equation (1-40) for large r the specific acoustic impedance will
approach the same value as a propagating plane wave Z s 0 c0 . This result is generally true
for any field which propagates out from the origin into an infinite fluid, i.e., a free radiation
case [1,4].

1.44 Point sources and Greens functions

If we keep the source terms in equation (1-22) and then derive of the wave equation once
again, the following result is obtained,

1 2 p m
2 p fV . (1-41)
c0 t
2 2
t

17
This is the wave equation with source terms or the inhomogeneous wave equation. The source
terms represent unsteady injection of mass and fluctuating external forces. In general we will
write the source term on the right hand side as s(x,t) and in the next chapter we will
investigate how such a general source term can be related to an unsteady (turbulent) flow
field.

The simplest possible source is a point source that mathematically is represented by a delta-
function, which in 3-D is defined as

1, x 0 V
(x x
V
0 )dV
0, x 0 V
, (1-42)

the definitions for the 1- and 2-D cases being analogous. The solution to the wave equation
with a point source excitation is called a Greens function. We will now study the Greens
function G for a harmonic point source (with unit amplitude) in a 3-D infinite space, i.e., no
boundaries. This so called free field Greens function satisfies the equation

2

k 2 G x x 0 . (1-43)

A point source defined by a delta function as in (1-43) is also called a monopole source of
unit amplitude and for the free field case the resulting sound field is called a monopole field.

The solution to (1-43) must exhibit spherical symmetry and correspond to waves propagating
out from the origin. We therefore choose the outgoing wave of a harmonic spherical wave
solution in (1-35) which, if we leave out the time dependence, can be written as

A exp( ikr )
G ( , x, x 0 ) ,
r

where A is a complex valued amplitude and r x x 0 . To determine the amplitude we

substitute the solution into (1-43) and integrate over a volume containing x0. Since the
spherical solution satisfies the wave equation everywhere except at x0, it follows that the

18
integration volume around x0 can be chosen arbitrary so we will choose a sphere (with radius
) centered a x0
V

k 2 G dV 1.
2
er
V

Applying Gauss theorem to the first term on the left hand side implies

G e dS k G dV 1,
2
r (1-44)
S V

G A exp(ikr ) 1
where e r is a unit vector in the radial direction and G e r ik .
r r r
Inserting this and performing the integrals in (1-44) yields
A 4 1,
independent of . We have then shown that the free field harmonic Greens function in 3-D is
given by
exp( ikr )
G . (1-45)
4r
If we take the inverse Fourier transform of equation (1-43) with G defined by (1-45) we
obtain

1 2
2 2 2 G (t ) (x x0 ), (1-46)
c0 t

(t r / c0 )
where G (t , x, x 0 ) . This corresponds to an acoustic pulse propagating out from
4 r
the origin after a sudden volume change at the origin occurring at t = 0, for t < 0 the fluid is
at rest.

If we instead of the spherical outgoing wave had used a wave propagating towards the origin

A exp(ikr )
G ( , x, x 0 ) ,
r

19
it would not affect the derivation presented above. Mathematically this alternative is
equivalent, but from a physical point of view it represents an acoustic energy flow from
infinity towards the origin. This is the opposite of what we expect when we have a source at
the origin, it is the source that should feed the field will an outgoing energy flow. Therefore,
the outward propagating solution is chosen as the physically correct solution for a source at
the origin. This choice can also be related to the concept of causality. Since in the time
domain the solution corresponding to G is given by

(t r / c0 )
G (t , x, x 0 ) ,
4r

this solution implies a spherical wave which at t starts to propagate from infinity
towards the origin. At t = 0 it reaches the origin where its energy is absorbed by a sudden
volume change and for t > 0 the fluid is at rest.

1.45* General solution

In general we have a distribution of sources s(x,t) and also boundaries to consider when
solving the wave equation. An integral formulation of the solution for the general case can be
obtained by applying Gauss theorem. In the frequency domain (harmonic time variation) the
general problem can be formulated as

2 k 2 p ( , x) s( , x)

, (1-47)
2
k G (x, y ) (x y )
2

where y is the location of the point source. Multiplying (1-47)1 with G and (1-47)2 with p
then subtracting the two equations and integrating over a volume V bounded by an internal S
and external surface Sr gives9
Sr
V

G
p p 2x G dVx p ( , y ) sGdV
x,
2
x
V V S

9
Using the integration rule for -functions: F (x) (x x
V
0 )dV F (x 0 ).

20
where y is assumed to belong to V and the sub-script x denotes that the derivatives and
integrations are done with respect to x. Applying Gauss to the volume integral gives

G
S
x
p p x G ndS x G x p p x G ndS x p ( , y ) sG dV x .
Sr V

The internal surface S is normally chosen so that it coincides with some solid body and if
there are no other boundaries Sr can be moved to infinity. It can then be shown that in the
limit of r the contribution from the second surface integral goes to zero. This is
normally referred to as the Sommerfeld radiation condition [4]. If we for convenience
interchange x and y and introduce the outward normal on S ( n n out ) the final result is

p (x) p (y ) y G (y, x) G (y, x) y p (y ) n out dS y s(y )G (y , x)dV y , (1-48)
S V

where x is referred to as a field co-ordinate, y as a source co-ordinate and the dependence of

is suppressed. Equation (1-48) is known as the Kirchhoff-Helmholtz equation and
sometimes it is called the Rayleigh equation.

If the Greens function is exact, i.e., satisfies the boundary conditions on S, then the surface
integral in (1-48) gives no contribution and the solution is simply

p ( x) s(y )G (y , x)dV y . (1-49)

V

A result that also is evident based on the principle of superposition, since the contribution dp
to the pressure at x from a point (monopole) source at y with strength sdVy is:

r
sdV y
dp
dp ( x) s(y )G ( y , x) dV y . y
x

The exact Greens function in analytical form is only known for a limited number of cases.
Normally equation (1-48) is applied using the free field Greens function from equation

21
(1-45). This gives for a source distribution radiating into a free field, i.e., with no solid
boundaries that affect the radiation,

s( , y ) exp(ikr )
p ( , x) dVy , (1-50)
V
4r

where r x y . In the time domain this equation becomes

s(y, t e )
p (x, t ) dV y , (1-51)
V
4r

where te t r / c0 is the emission time (also called retarded time) for sound emitted from a
point y that reaches the observer at x at the time t.

1.46 Acoustic energy and intensity

A general conservation equation for energy can be written in the form
e
I q , (1-52)
t
where e is the energy density per m3, I is the energy flux or intensity vector [W/m2] and q is a
source term. We would now like to apply equation (1-52) to an acoustic field. To obtain
expressions for the energy density and intensity, one could start from the general energy
equation in fluid mechanics (1-19) and use the same procedure as in the derivation of the
classical wave equation. An alternative leading to the same result [1] is to start from the
linearized equations of fluid mechanics (1-22) and try to rewrite them as an energy equation

1 p
c 2 t 0 u m
0
.
u
0 p fV
t

To do this we multiply the mass conservation equation with p and the momentum equation
with u which after some rearrangement leads to

22
 1 p2 p u
mp

2 0c0 t 0
2

.
u u
0 u p fV u
2 t

Adding these two equations gives10

0u u p2 mp
pu
2
fV u . (1-53)
t 2 2 0c0 0

This is the sought energy conservation equation where

0u u p 2
e
2 2 0c02

I pu
. (1-54)

q mp f u
0
V

The energy density is a sum of two parts kinetic + potential where the potential is
associated with the density change of the fluid. The source terms show that maximum power
input from a fluctuating mass source is obtained at pressure maxima. For a fluctuating force
source maximum power input is obtained at velocity maxima but only if the force is parallel
with the velocity vector. If it is perpendicular the power input is zero.

For the case of a sound field stationary or periodic in time we can define the time average of
the acoustical field quantities. If f(t) is a field quantity at a point in space then the time
average is defined as

T
1
T 0
f lim T f (t )dt , (1-55)

10
p u p u p u

23
where for a periodic field the limit can be removed and the integration performed over a
single period. Applying this definition to (1-53) gives for stationary sound fields

pu q , (1-56)

since11 e / t 0 . Assume now that we have a source distribution q around the origin
radiating into an infinite space (free field). The power W radiated from this source region is

W q dV .
Vq

Using equation (1-56) and Gauss theorem this can be expressed as

W p u dV p u e r dS , (1-57)
Vr Sr

i.e., a surface integral over a sphere Sr centred at the origin which can have an arbitrary radius
as long as Vq Vr .

As stated in sub-section 1.43 for sufficiently large distances the field from a source
distribution around the origin will locally resemble a plane wave, i.e., p 0 c0 u r . If we use
this relationship equation (1-57) implies that

p 2
W p u r dS lim r dS . (1-58)
Sr Sr
0 c0

From equation (1-58) the conclusion can be drawn that, since the surface area of the sphere is
proportional to r2, the leading term in the sound field that produces the acoustic power must
be proportional to 1/r. Terms which decay faster are nearfield terms that do not produce any
net (time averaged) acoustic power output.

For a propagating plane wave equation (1-58) gives directly for the intensity in the direction
of propagation (x1)

f 1 f f (T ) f (0)
T
11
If f always has a finite maximum amplitude then dt 0, when T .
t T 0 t T

24
 p2
I x1 p u1 eq. (1-31) . (1-59)
0 c0

Example 1-7: For the case of harmonic and complex valued signals the time average of the
acoustic intensity can be written as I 12 Re p u * , where * denotes complex conjugation.

This result follows from the definition of time average and the observation that
a a exp(it ) a * exp(it ) / 2 .

1.5 Multipole expansion of sound fields

Given the monopole field

exp(ikr )
G ,
4 r
which is a solution to the Helmholtz-equation for all points r 0 . This means that

2

k 2 G 0, r 0.

If we operate on this equation with a spatial derivative (say) / x1 we obtain

2

k 2 G x1 0, r 0,

since we can change the order12 between the wave equation operator and / x1 . The result

implies that G x1 also is a new solution to the free space13 wave equation. Using this
procedure we can generate a whole family of new solutions to the wave equation. These
solutions are called multi-poles and are defined from

n exp(ikr )
G n l , (1-60)
x1 x2 x3 4 r
j k

where n j k l and Gn is called a multi-pole of order 2n. The first three multi-poles are
also called: monopole (n = 0), dipole (n = 1) and quadrupole (n = 2).

12
The wave equation operator has constant coefficients and G is also sufficiently regular (well behaved) for
r 0.
13
N.B. The new solution only satisfies the equation. If there are boundaries involved the new field normally does
not satisfy the boundary conditions any more.

25
Example 1-8: The order of the multi-poles corresponds to the number of monopoles needed
to generate the field Gn . For instance the dipole field G1 can be obtained by superposition of

two monopoles in anti-phase, i.e., G 0 (r ) G 0 (r h1 ) G 0 (r ) h1 . This implies that

G 0 G (r ) G 0 (r h1 )
ei 0 , where h 1 h 1e
xi h1
and which is exact in the limit of h1 0 . (r-h1) r
h1
From the left hand side we
(r)
see that G1 has three components
and actually can be regarded as a vector. We can now repeat this process on G1 and generate

G2 by superposition of two dipoles in anti-phase. Since each dipole can be seen as a

superposition of two monopoles the resulting quadrupole is a superposition of four
monopoles. Continuing this process it is realized that the n:th multi-pole is obtained by
superposition of 2n monopoles.

We will now demonstrate that a complicated sound field generated by a set of monopole
sources around the origin, can be expressed as a series expansion in terms of multi-poles. For
low frequencies the first non-zero term in this expansion will dominate the sound field and the
radiated power.

The sound field from one single monopole, m, is

S m exp( ikrm )
p m ,
rm

where rm x y m and the 4 factor has been included in the source amplitude S m . The total

field is obtained simply by superposition of all the source contributions

S m exp(ik x y m )
p (x) . (1-61)
m x ym

26
 x

S m

ym
S1

Figure 1.4 Radiation from a distribution of sources around the origin.

From equation (1-61) we see that p is a function of x y m and especially for cases when

x y m , a Taylor expansion around x could be an efficient way of rewriting the sum. The

general formula for a Taylor expansion of a function of several variables is

1 1
F ( x h) 1 (h x ) (h x ) 2 ... (h x ) n ... F ( x).
2! n!
Applying this formula to each term in (1-61) with h y m and F exp(ikx ) / x , where

x x , gives

1 (1)n exp(ikx)
p (x) Sm 1 (y m x ) (y m x )2 ... (y m x )n ... .
m 2! n! x

If we introduce Cartesian tensor notation and change the order of the summations we get

2 exp( ikx )
p ( x) S D i Q ij .... , (1-62)
xi xi x j x

1
where S S m , D i S m y mi and Q ij S m y mi y mj . Equation (1-62) represents the
m m 2 m
expansion of the resulting field in terms of multi-poles.

27
Example 1-9: Below two examples on multi-pole expansions are given.

xx2 A monopole placed outside the origin. Obviously

if it was positioned in the origin, a pure monopole
y1 = (a,a,0) field in x would result. But a monopole outside the
origin produces a field that is not spherically
S 0 symmetric in x and this is represented by the terms
x1 we get in the series or multi-pole expansion. The
first three terms are:

The monopole: S S 0 .

The dipole: D 1 D 2 S 0 a and D 3 0.

The quadrupole:
1 1
Q 11 Q 22 S 0 a 2 , Q 12 Q 21 S 0 a 2 and all
2 2
other terms, involving x3, are 0.
x2
Two monopoles with opposite phase are shown in the
S 0 y1 = (0,a,0) figure. Only multi-poles in the x2 direction will exist,
i.e., all multi-poles with an index 1 or 3 are zero. The
x1 first three terms are:

S 0 y2 = (0,-a,0) The monopole: S S 0 S0 0 .

The dipole: D 2 S 0 a S 0 (a) 2S 0 a.

1

The quadrupole: Q 22 S 0 a 2 S 0 a 2 0 .
2

If source 2 is in phase with source 1 we get: S 2S 0 , D 2 0 and Q 22 S0 a 2 . If we add a

monopole with strength 2S at the origin we eliminate the monopole term and the
0

leading term will be the quadrupole Q 22 .

28
References
1. H. Bodn, U. Carlsson, R. Glav, HP Wallin och M. bom (1999) LJUD OCH
VIBRATIONER. MWL, KTH.
2. M.J. Lighthill (1952) Proc. Royal Society A211, 564-587. On sound generated
aerodynamically. I General theory.
3. M.J. Lighthill (1954) Proc. Royal Society A222, 1-32. On sound generated
aerodynamically. II Turbulence as a source of sound.
4. A.D. Pierce (1989) ACOUSTICS An introduction to its physical principles and
applications. McGraw-Hill.

29
30
Problems

1. Given a scalar field (x) and a vector field A(x) which are functions of the space co-
ordinates x. Show that
2 2 2
a) 2 , in Cartesian co-ordinates x1, x2 and x3.
x12 x 22 x32

b) A A A .

c) Suppose that (r ) where r x x 0 and x0 is fixed (source) point. Show that

d
er ,
dr
where e r (x x 0 ) / r is a unit vector from the source point to the field point x.

d) Suppose that A A(r )e r . Show that

1 d
( Ae r ) 2
r dr
Ar 2 .

2*. Do the linearization of the fundamental fluid dynamic equations as described in the
text and compare the magnitude of the neglected second order term with the linear
terms. Show that the second order terms are small as long as / L 1, where is the
particle displacement in the acoustic field and L is a typical length scale for the spatial
derivatives.

3*. Assume that we have two different acoustic source distributions s1 and s 2 generating
two different acoustic fields p1 and p 2 . The fields satisfy the inhomogeneous
Helmholtz equation

2
k 2 p s.

Prove that p s
V
1 2 p 2 s1 dV 0 , where the integration is over all source points.

This result is called the reciprocity principle and is especially useful for the case of
point sources. The principle can be used to simplify measurements and derivations [1].

31
 N.B. The result is valid independent of any solid boundaries but for simplicity assume
free field conditions. This implies that in the far field the sound behaves locally as a
plane wave, i.e., p 0 c0 u r where r denotes the radial component.

4. A small source at the origin oscillates harmonically and produces a spherical sound
field.
a) Given the expression for the sound pressure, calculate the acoustic velocity field.

b) Study the velocity field in the limit when the wave number goes to zero (at fixed
frequency!). What is the physical interpretation of the velocity obtained in this limit?
Hint: Calculate the divergence.

c) Prove equation (1-40).

d) Calculate the acoustic power radiated by the source.

5. The field of a dipole source is given by

F exp(ikx)
p d .
4x

Show that the far field is given by

exp(ikx)
p d ikF cos ,
4x

and plot the directivity, i.e., p d ( ) for fixed x.

6. The field around a quadrupole with harmonic time variation can be written as

2 exp(ikx)
p q Q ij , x 0.
xi x j x

There exist two types of quadrupoles, longitudinal ( i = j ) and lateral ( i j ) .

a) Prove that there exist six different quadrupoles, three longitudinal and three lateral.

32
b) Given a quadrupole where the only component different from zero is Q11 . Show that
in the farfield where only terms proportional to 1/r is of importance, the field from this
quadrupole is given by

k 2 x12 Q 11 exp(ikx)
p q .
x2 x

c) Same as in b) but with Q 12 0.

k 2 x1 x 2 Q 12 exp(ikx)
p q .
x2 x

7. Show that in the far field the n:th term in the multi-pole expansion (eq. (1-61)) is
proportional to (kd ) n , where d is the diameter of the source region. What does this
imply for the radiated power ?

8. A moving piston in a duct

As a simple model for the pressure wave generated by a train traveling through a
(tight) tunnel we will assume a piston moving with constant speed U (< c) along a
duct. The piston will generate a plane wave with a wave front propagating at the speed
of sound.

a) The so called dAlemberts solution to the 1-D wave equation is given by

p ( x, t ) f (t x / c) g (t x / c),

33
 where f and g are arbitrary functions representing waves propagating in the + or
x-directions. These functions are determined by using the initial and boundary
conditions for the problem. For our problem the acoustic wave must satisfy the
condition of continuity of velocity at the piston. Assuming that the piston starts to
move from x = 0 at t = 0 we have:

U , t 0
u x ( x, t ) where x = Ut.
0, t 0

Assume a reflection free (semi-infinite) duct, that the medium is initially at rest and
use dAlemberts solution to show that the acoustic pressure in front of the piston is:

c U , Ut x ct
p ( x, t ) 0 0 for t >0.
0, x ct

b) Assume a train moving at 72 km/h calculate the resulting level of the pressure wave.
Data: 0 1.21 kg/m 3 and c0 = 340 m/s.

N.B. The same model can be used to estimate the sound generated by slamming a
door, i.e., not the structural part but the part associated with the abrupt stop. This is
like a piston brought to a sudden stop so the acoustic pulse generated is of the order
0 c0U .

34
Solutions

1. The nabla-operator can be seen as vector which also behaves an operator i.e. takes the
derivative of everything that stands to the right of the operator.

a) 2 the square of a vector is by definition Cartesian system

x1 , x2 , x3 x1 , x2 , x3 2 x12 2 x22 2 x32 .

This squared nabla operator is normally referred to as the Laplace operator.

b) A Cartesian system x1 , x2 , x3 A1 , A2 , A3
A1 x1 .... where the omitted terms are obtained by changing the index to 2 or 3
x1 A1.... A1 x1.... interpret this a a vector formula A A .

d
c) apply the chain rule + evaluate in Cartesian system r x1 , r x2 , r x3
dr

The gradient of r (distance source-receiver) is obtained by observing that:

r
x1 x10 x2 x20 x3 x30
2 2 2
r ,
x
x0
This implies that:

r 2 x1 x10 x x
1 10 . Inserting this result above gives:
x1 2 x x ...
2 r
1 10

d x1 x10 , x2 x20 , x3 x30 d x x 0 d r d

er .
dr r dr r dr r dr

dA 2 1 d 2
d) Ae r prbl. 1b A e r A e r prbl. 1c A 2
dr r r dr
r A ,
where we have used the relation:
er r r prbl. 1b+c r / r r er / r 2 r 3 2 / r .

4a) The sound pressure for a harmonic source [ exp(it ) ] at the origin is given by an
outgoing wave. Using eq. (1-35) and choosing the + solution we get:

35
 A exp(i (t r / c))
p(r , t ) . Omitting the harmonic time factor which will be the
r
A exp(ikr )
same for all field components we get: p (r ) . (1)
r
The acoustic velocity field can be obtained from the pressure field via the linearized
eq. of motion (1-22)2 (with no sources):

u
0 p . (2)
t
Since the pressure only depends on r we will only get a radial component (see prbl.
1c). Observing that / t i for a harmonic field one finds that eq. (2) implies:
dp
0iur . (3)
dr

Inserting eq. (1) in (3) gives:

A 1 ik A exp(ikr ) 1
ur 2 exp(ikr ) 1 ,(4),
0i r r 0 c0 r ikr

where the relation / k c0 has been used.

b) If the wave number k / c0 goes to zero (at constant frequency) it follows that the
speed of sound goes to infinity. Using this in eq. (4) we get:

A exp(ikr ) 1 A 1 c A
ur lim k 0 1 exp(ikr ) 1 lim c0 1 0 . (5)
0 c0 r ikr 0 r c0 i r i 0 r 2

If we integrate this over a spherical surface with radius r we get the outflow:

4 A
4 r 2 ur , i.e., a quantity independent of the radius which implies an
i0
incompressible flow. Alternatively one can use the formula for the divergence of a
spherically symmetrical field:

1 d
div(ur e r ) prbl. 1d 2
r dr
ur r 2 eq. (5) 0.

36
 It can also be observed that an infinite speed of sound implies zero density changes via
the relationship: p c02 . In other words the nearfield term ( 1/ r 2 ) in eq. (4)
represents in the limit of an infinite speed of sound a purely incompressible field.

c) Taking the ratio of eqs. (1) and (4) directly gives the answer.

d) The power radiated by a spherically symmetric field is given by:

W 4 r 2 I r , (6)

where the intensity for a harmonic field is found from the formula given in Ex. 1-7.

1 1 A A * exp(ikr ) 1
Ir r* eq.(1)+(4) Re exp(ikr )
Re pu 1
2 2 r 0c0 r ikr
2 (7)
A
.
2 0 c0 r 2

Inserting (7) in (6) gives the radiated power:

2
4 A
W .
2 0 c0

Note that the nearfield term in eq. (4) ( 1/ r 2 ) does not contribute to the radiated
power. This conclusion is generally valid for all terms that decay faster than 1/r.

5. The far field is obtained by taking the derivative of the numerator:

F exp(ikx) F (ik ) exp(ikx)x

p d , farfield probl. 1c x e x
4 x 4 x
exp(ikx) exp(ikx)
F e x ik ikF cos
4 x 4 x
F

This gives the directivity: p d , farfield cos , which defines two spherical surfaces as

shown in the figure.

37
6a) Each of the indices can take the value 1,2,3 this gives a total of 3x3 =9 quadrupoles.
Three of these correspond to identical indices or longitudinal quadrupoles. For the
remaining six mixed or lateral we observe that e.g.:

2 2
,
x1x2 x2 x2

since the monopole field inside the parenthesis is sufficiently regular for x 0 so we
can change the order of the derivatives. Furthermore the source strength satisfies:

1 1
Q12 Sm ym1 ym 2 Sm ym1 ym 2 Q 21 .
2 m 2 m

These results show that the mixed quadrupoles can be put together in pairs so they
only represent three different types corresponding to the index combinations: 12, 13
and 23. The conclusion is therefore that we have a total of six different quadrupoles,
three longitudinal (11, 22 and 33) and three lateral (12, 13 and 23).

2 exp(ikx)
b) Given the quadrupole field: p q Q11 2 , calculate the farfield. For a free
x1 x
field radiation case (see section 1-43) the farfield corresponds to the source component
that decays as 1/r. This implies that when performing the derivative we should only
operate on the numerator since otherwise nearfield terms are produced.

Taking a single derivative of the exponential gives:

ikx
exp(ikx) 1 exp(ikx) ,
x1 x

where we used the result: x / x1 x1 / x , see prbl. 1c. This result implies that the

exp(ikx) ikx1 exp(ikx)

farfield is obtained by using the formula: ,a
x1 x x x
result which obviously applies to the other co-ordinate directions if we change the
index to 2 or 3.

38
 Applying this formula twice gives:

Q11k 2 x12 exp(ikx)

2
ikx1 exp(ikx)
p q farfield Q11 ,
x x x2 x

where the first term describes the directivity of the source. If we introduce an angle
relative the x1-axis we get: cos 1 x1 / x , this implies

Q11 k 2 cos 2 1
p q farfield .
x

2 exp(ikx)
c) Same as b) for the field: p q Q12 .
x1 x2 x
Using the formula from b) implies:

ikx1 ikx2 exp(ikx) Q12 k 2 x1 x2 exp(ikx)

p q farfield Q12 .
x x x x2 x

This can be expressed using spherical co-ordinates with x3 as the polar axis:
x1 x sin 3 cos , x2 x sin 3 sin , as

Q12 k 2 sin 2 3 sin 2

p q farfield .
2x

1 y y .... exp(ikx)
n
7. The n:th term in eq. (1-61) is: p n
n! m
S

m mi mj
xi x j .... x
.

n terms

1
The source strength will be proportional to:
n! m
Sm ymi ymj .... d n , as the source

n terms

region has a diameter d. This implies that if we (say) double the diameter, i.e, double
the distance for each source to the origin, the sourced strength will increase with a
factor 2n . From prbl. 6 we know that the farfield is obtained by using the formula:

39
 n exp(ikx) n xi x j ...
ik ,
xi x j .... x xn

where the directivity factor is independent of the distance (x). These results imply that
the magnitude of the n:th term will be proportional to:

p n kd .
n

this result implies: Wn kd .

2 2n
Since the radiated power W is proportional to p n

So the relative radiated power from the multi-poles will be proportional to:

W0 : W1 : W2 :..... 1: kd : kd :.....
2 4

For the case of low frequencies when the wavelength is much larger than the source
region and kd 1 , this result implies that the first non-zero multi-pole will strongly
dominate the radiation. It is for this case a multi-pole expansion is mainly useful.

8a) Since we have a semi-infinite duct with the medium initially at rest, the piston will
create a wave in the + x-direction. This implies that: p ( x, t ) f (t x / c). This wave
creates a velocity field of: u x ( x, t ) f (t x / c) / 0 c0 . The continuity of velocity at
the piston implies:
U , t 0
f (t Ut / c) / 0 c0 . Here we can put: t t (1 M ) , M = U/c. Since M<1 t
0, t 0
c U , t 0
has the same sign as t. We can then draw the conclusion: f (t ) 0 0 .
0, t 0
Substituting t t x / c gives us the solution for any t >0 and x > Ut.

b) A speed of 72 km/h corresponds to U = 20 m/s, this gives a pressure wave level of

p = 1.21*340*20 = 8200 Pa.

40
 Chapter Two

Generation of sound in fluids

In this chapter we will discuss the physical processes that can generate sound waves in a fluid.
The starting point is the classical wave equation with a source term, i.e., the inhomogeneous
wave equation. The theory proposed by Lighthill (1952) on how to obtain this source term
from the fluctuating fluid state is described. In Lighthills theory sound is produced by three
basic source mechanisms: monopole, associated with fluctuating mass injection or volume
flow; dipole, associated with unsteady external forces or fluid pressure on a solid boundary;
and quadrupole, associated with an unsteady Reynolds stress (momentum transport). Free
field solutions for the three source mechanisms are presented and scaling laws are derived.
The modification of Lighthills theory when fixed solid surfaces are present as suggested by
Curle is also presented. At the end of the chapter two important applications are treated (jet
noise and noise from propellers).

2.1 The inhomogeneous wave equation

If we introduce a source term s(x,t) in the classical wave equation, the inhomogeneous wave
equation is obtained,

1 2 p
2 p s(x, t ). (2-1)
c0 t
2 2

41
The solution to this equation can be obtained by first studying a point source solution and
using the principle of superposition. A point source solution is called a Greens function and
is defined by

1 2 2
2 2 x G x, y , t , (t ) (x y ), (2-2)
c0 t

where x denotes a field point, y the source position and x is the nabla-operator with respect
to x. Using the property of the delta function we can represent any source distribution as a
superposition of point sources

s ( x, t ) s(y, ) (t ) (x y )ddV
V
y , (2-3)

where dVy denotes that the space integration is with respect to y. If we multiply equation

(2-2) with the source strength s(y,) and integrate over space and time we obtain

1 2
2 2 2x s (y , )G x, y , t , ddV y s (x, t ), (2-4)
c t
0 V
where the order of the wave equation operator and the integrations has been changed and
equation (2-3) is used. Equation (2-4) implies that the solution to the inhomogeneous wave
equation can be written as14

p (x, t ) s(y, )Gx, y, t , ddV .
V
y (2-5)

Equation (2-5) expresses the sound field as a superposition of point source solutions. The
equation is generally valid but the limitation is that we only have analytical expressions for
the Greens function for a few cases, e.g., the free field. For the case of a 3-D free field we
know from Chapter 1 (see eq. (1-46)) that the Greens function is given by

(t r / c 0 )
G ( x, y , t , ) , (2-6)
4r

14
For a 1- or 2-D case the spatial integral is reduced accordingly.

42
where r x y . Using equation (2-6) in (2-5) we obtain the 3-D free field solution to the

inhomogeneous wave equation,

s (y , t r / c 0 )
p (x, t ) dV y , (2-7)
V
4r

where t e t r / c0 is called the emission time. For the frequency domain version of this, i.e.,
the solution for a harmonic field, see equation (1-50).

r
s (y, t r / c0 )dV y
dp(x,t)
y V
x

Figure 2.1 Superposition of monopole contributions to build up the total sound field from a source region,
see eq. (2-7). The sound emitted from a point y at t-r/c0 reaches x at t.

In order to apply equation (2-5) to the 1- and 2-D free field case, the Greens function for
these cases must be known. For the 1-D case it can easily be derived (see prbl. 7a) and for the
2-D case it can be obtained by integrating the 3-D Greens function along a line. The time
domain versions of the 1- and 2-D Greens are presented in Table 2.1.

Table 2.1 The free field Greens functions for the 1- and 2-D cases. Notation: H is the Heaviside unit step

function15, x and y are 2-D vectors and 2 is the 2-D nabla-operator.

1 2 2 c0
1-D: 2 2 2 G ( x, y, t , ) ( x y ) (t ) G H (t x y / c0 )
c0 t x 2

1 2 1 H (t x y / c0 )
2-D: 22 G (x, y , t , ) (x y ) (t ) G
c 0 t
2 2
2 (t ) 2 x y 2 / c 2
0

1, x 0
15
H ( x)
0, x 0

43
2.11 Non uniqueness of the source term
From equation (2-7) it follows that if the source term s(x,t) is given in a finite volume V, then
the resulting sound field p (x, t ) is uniquely determined. The opposite is however not true,
i.e., given a field p (x, t ) that satisfies the inhomogeneous wave equation

s0
s0
1 2
s (x, t ), x V
2 2 2 p (x, t ) ,
c t 0, x
V
0 V

then an infinite number of source distributions will produce the same sound field outside V.
This is shown if we construct the field p1 p Ks , where K is a constant, then obviously

p1 and p are identical outside V. But p1 satisfies the inhomogeneous wave equation with

1 2
the source term: s1 s K 2 2 2 s. This important result implies that we have a
c0 t
considerable freedom in the process of defining acoustic source models.

2.2 Lighthills theory or acoustic analogy

The starting point for Lighthills theory is the inhomogeneous wave equation and the splitting
of space in two parts, the sound field and the source field. The sound field is defined as the
part of space where the fluctuating pressure (or density) field satisfies the homogeneous wave
equation. The part of space where we have an inhomogeneous wave equation is called the
source field. To obtain a model for the source field Lighthill rewrites the fundamental
equations of fluid mechanics in the form of a wave equation. This is the so called Lighthills
aeroacoustic analogy. From Chapter 1 we take the equations of conservation of mass and
momentum, (1-13) and (1-17),

u i
m
t xi
. (2-8)
u i u i u j p ij f V ,i
t x j xi x j

In the same way as we proceed when deriving the classical wave equation, we now perform
the operation: (2 8)1 / t (2 8) 2 / xi . This gives

44
 2 2 p m f V ,i 2
ui u j ij . (2-9)
t 2 xi xi t xi xi x j

So far no assumptions or simplifications have been introduced, the resulting equation is

simply a reformulation of the fundamental equations. We now assume a region V with
unsteady flow processes that is embedded in a surrounding homogeneous fluid with no mean
flow, characterized by p0, 0 and c0. If we substitute p p 0 p and 0 in (2-9) it

can be reformulated as a wave equation by adding either c02 2 / xi xi or 1 / c 02 2 p / t 2 .
The first alternative gives

2 2 m f V ,i 2

2
c 0 ( p c02 ) ij u i u j ij , (2-10)
t 2
xi xi t xi xi x j

and the second

1 2 p 2 p 1 f V ,i 2
m ( p c 2
) ui u j ij . (2-11)
xi xi t x
0
c02 t 2 c02 t i xi x j

Equation (2-10) is the aeroacoustic analogy first derived by Lighthill [Chap.1:2,3] and is
based on the density as the acoustic variable. In (2-11) the pressure is used instead which
produces a different acoustic analogy, i.e., the structure of the acoustic source terms differs.
The difference lies in the effect of the term ( p c02 ) . This term represents deviations from
adiabatic changes of state that can occur when we have heat release, for instance from
chemical reactions (combustion). For cases when this term can be neglected, e.g., the classical
model for jet noise, the two formulations are equivalent.

2.21 Limitations in Lighthills theory

The main limitation in Lighthills acoustic analogy is the assumption that the source field is
independent (uncoupled) to the acoustic field. The source field is assumed to be prescribed
based on measurements or from theoretical models. This means that situations where
feedback occurs and the acoustic field modifies the flow field cannot be treated. Examples of
such cases are found for instance when a periodic flow separation process couples to an

45
acoustic field. This can create a so called self sustained oscillator or a whistle, a phenomenon
which is studied in Chapter 5. Another limitation is that the effect of fluid motion and mean
flow gradients on the sound generation and propagation is not considered. As we will see in
Chapters 3 and 4, this can lead to curved propagation paths and to a frequency change
(Doppler shift) in the emitted sound. One way to remove this last limitation is to formulate the
wave equation operator in a more general way, i.e., for a fluid with a varying mean flow state.
Such formulations exist but will not be discussed here [1].

2.22 Interaction between flow and sound

There are three known mechanisms for the interaction between a flow field and an acoustic
field. The first is sound production by an unsteady flow embedded in a fluid volume

Flow Sound Flow Sound

Alt. 1 Alt. 2

Flow Sound

Alt. 3

Figure 2.2 Illustration of the interaction between a flow field and a sound field. 1) Production of sound via an
unsteady flow field; 2) Forced vortex separation caused by an acoustic field leading to dissipation of sound;
3) Fluid driven self sustained oscillation or whistle. Lighthills acoustic analogy only considers Alt. 1.

which is described by Lighthills theory. The second mechanism is forced vortex separation,
e.g., at a sharp edge, caused by an acoustic field. This will lead to a conversion of acoustic
energy into kinetic energy in the vortex, which will be converted into turbulence and
dissipated downstream. This mechanism therefore represents dissipation of acoustic energy
and can be important for low frequencies. An example of such dissipation of sound by a jet
from an open pipe termination is described by Bechert [2]. The third mechanism is when a
periodic flow separation couples to an acoustic field and forms a whistle, see Chapter 5.

46
Concerning mechanism 2 acoustic analogies exist that can be used to study this type of
interaction, for instance the vortex sound theory by Powell and Howe [3,4].

2.3 The three source terms

The source term in Lighthills acoustic analogy can be split into three parts which, if we
choose the pressure based formulation (2-11), have the form

1
s1 m 2 ( p c 0 )
2

t c0 t
f V ,i
s2 . (2-12)
xi
2
s 3 ui u j ij
x i x j

The general free field solution (2-7) can be used to write down the solution for these source
terms as a superposition of monopole fields. But for the source terms involving a spatial
derivative of order n, it is possible to rewrite the sound field and express it as a superposition
of multi-pole fields of order 2n. This is useful especially when the source region is compact,
i.e., small compared to the wavelength. Then the source distribution can be reduced to a single
multi-pole field, e.g., a dipole (n = 1) or a quadrupole (n = 2). To prove this statement we
simply operate on the inhomogeneous wave equation (2-1) with a spatial derivative

n 1 2 2 n s ( x, t )
x p ( x , t ) ,
x1j x2k x3l c02 t 2 x1
j
x k
2 x 3
l

where n = j+k+l and the pressure p is assumed to be given by the free field solution (2-7). If
p is assumed sufficiently regular, the order between the spatial derivative and the wave
equation operator can be interchanged. This implies that the free field solution to the wave
equation with the source term n s / x1j x 2k x3l can be written as

n s(y, t r / c0 )
p n (x, t ) j k l dV y , (2-13)
V x1 x 2 x3
4r
which proves the statement given above16.

16
Note the order of the partial derivatives and the integration can be interchanged since the integral is with
respect to y and the derivatives are with respect to x.

47
2.31 Compact sources and dimensionless numbers
For an aeroacoustic source process driven by an unsteady flow, a typical time scale for the
sound production can be taken as T L / U , where L is a typical length scale (diameter) of
the source region and U is a typical mean flow speed. The sound produced by this process has
a typical frequency of f 1 / T U / L , which corresponds to an acoustic wavelength of

c0 / f L / M , (2-14)

where M U / c0 is the Mach number. This implies that L / M , which means that for
small Mach numbers the source region is much smaller than the wavelength, the source region
is acoustically compact. For a compact source the variation of the emission time (or phase17)
over the source region can be neglected and we can put r x in (2-13). This gives

n S (t x / c0 )
p n (x, t ) , (2-15)
x1 x 2 x3
j k l
4x

where S (t ) s (y, t )dV y is the source strength.

V

The typical frequency scale for an unsteady flow process is often expressed using a
dimensionless number called the Strouhal number (St) defined by

L
St . (2-16)
U

In aeroacoustics there are three dimensionless numbers that can be used to characterize a
problem; the Strouhal number, the Mach number and the Helmholtz number (Eq. (1-38)).
These three numbers are not independent since

He St M . (2-17)

17
For an angular frequency the time variation is exp(i(t-kr)) which implies that the phase shift across the
source region is of the order kL, which for a compact source region is much smaller than 1.

48
2.32 The monopole source term
The monopole source term corresponds to s1 in equation (2-12). The resulting sound field is a
superposition of monopole fields

m t (y , t r / c0 )
p(x, t ) dVy , (2-18)
V
4 r

where m t mt / t and mt m (1 / c02 ) p c02 / t . The source mechanisms are mass
injection and deviation from adiabatic changes of state. As can be seen from the formula it is
only the unsteady parts that will contribute to the sound.

A classical example of an unsteady mass injection source ( m ) is a radially oscillating small

sphere. Obviously the sound produced by such an object is independent of the density of the
sphere, but will depend on the unsteady volume flow produced and on the density 0 of the
surrounding fluid. This is generally true and a mass injection process should be seen as a
volume flow source that can be represented as: m 0 q , where q is the unsteady volume
flow per m3. The sound generated by a compact mass injection source can then, using (2-15),
be written as

0 Q (t x / c0 )
p (x, t ) , (2-19)
4x
where Q Q / t and Q(t ) q (y, t )dV y is the total unsteady volume flow. The sound
V

power radiated from this type of monopole source, when it is stationary or periodic in time,
can be obtained from equation (1-58)

0 Q 2
Wm . (2-20)
4c0

The most important examples of mass injection or volume flow sources are:

1) A loudspeaker mounted in a box18 where a net volume flow is produced

when integrated around the box. For low frequencies when the wavelength
is much larger than the box size this will be a point monopole source.

18
This is a special case of the radiation from the surface of a vibrating solid body which can be seen as a
monopole source distribution.

49
2) The openings of inlet and outlet pipes from piston machines, e.g., internal combustion (IC)
engines, compressors and pumps. The volume flow injected to the surrounding fluid is
obtained by integrating the velocity over the duct opening as shown in Figure 2.3.

Figure 2.3 Pulsating volume flow Q from a pipe opening: Q u ndS , where S is the pipe cross-section.
S

For instance the sound radiated from an exhaust outlet on a car can be modeled as a monopole
source. Note that although the fluctuating exhaust gas flow at the outlet is hot and has a
different density than the surrounding air, it is the density of the surrounding air that should
be used to calculate the radiated sound in equation (2-19).

3) Collapsing cavitation bubbles in liquids. This kind of bubbles are formed when the local
static pressure is less than the vapor pressure. The bubbles are unstable and can suddenly
collapse and it is this sudden implosion which is the main cause of sound from cavitation.
Since the rapid change in volume creates a large time derivative, high sound levels can be
created.

Example 2-1: We will look at the scaling law for the monopole sound field produced by a
fluctuating volume flow of a pipe opening. This can also be called a pulsating jet. The
unsteady volume flow Q scales as: Q UD 2 , where U is the mean flow speed and D the pipe
diameter. If the process is periodic as for an IC-engine with a period time T, then (2-19)
implies that the radiated power scales as

0U 2 D 4 0 f 2U 2 D 4
Wm , (2-21)
c 0T 2 c0

where f = 1/T and a compact ( D ) source region is assumed.

Concerning the second type of monopole source process ( p c02 ) representing deviation
from adiabatic changes of state, there are two main possible mechanisms. One is chemical

50
reactions, for instance combustion, where an unsteady heat release is the cause of sound
production. This source type is also called a thermoacoustic source. It is important for
example in fluid machines where heat energy is converted into kinetic or pressure energy in a
fluid, e.g., IC-engines and gas turbines. Another mechnism is sound production from
acceleration of fluid particles through a region with mean flow gradients. This mechanism is
sometimes called acoustic Bremsstrahlung in analogy with a term used in high energy
particle physics. To describe this effect the density based acoustic analogy in equation (2-10)
is best suited.

2.33 The dipole source term

The dipole source term corresponds to s2 in equation (2-12), and the resulting sound field is a
superposition of dipole fields

fV ,i (y, t r / c0 )
p(x, t ) dVy , (2-22)
V
xi 4 r
where f V ,i is an unsteady external force acting on the fluid. For the case of a compact source

this reduces to
Fi (t x / c0 )
p (x, t ) , (2-23)
xi 4x

where Fi (t ) f V ,i (y, t )dV y is the total force acting on the fluid. The far field is obtained if
V

we only consider the field component proportional to 1/x, and it is given by

F (t x / c0 ) e x
p (x, t ) , (2-24)
4xc0

where F F / t and e x x / x . The sound power radiated from a dipole source can be
obtained from equation (1-58)

F 2
Wd , (2-25)
12 0 c 03

where F F .

51
Important examples of unsteady force sources are:

1) A loudspeaker with no box. This will produce no net volume flow

but an oscillating force on the fluid when it is accelerated from the
front to the back and vice versa. If the wave length is much larger then
the diameter of the loudspeaker, this will represent a piont dipole source.

2) Time varying or unsteady fluid forces (pressures) on solid bodies subject to a mean flow.

For a fixed non-moving body, flow separation can create

fluctuating forces. It is known, for instance, that a cylindrical
bar will produce a periodic flow separation with a characteristic
frequency (corresponding to a Strouhal number around 0.2 based on the diameter). The
periodic flow separation cause an oscillating force on the fluid, mainly perpendicular to the
flow direction, and this unsteady force generates sound. This type of sound can be generated
on vehicles at antennas and other solid objects such as mirrors.

Another example is propellers where the rotating blade forces

F
represent a time varying force distribution on the fluid. This
will create a sound field with harmonic components at
frequencies that are multiples of the rotational frequency.
Disturbances in the inflow will affect the time variation of the blade forces and increase the
sound production.

A more correct representation of this type of source term as a surface integral of unsteady
pressure rather than a volume force requires a reformulation of Lighthills analogy to include
solid boundaries. This is known as Curles equation, see section 2.35.

3) Unsteady electromagnetic forces in a plasma, for example in a discharge lamp.

Example 2-2: We will look at the scaling law for the dipole sound field produced by a
fluctuating volume flow out from a pipe opening. This can also be called a pulsating jet. The
unsteady force F (or momentum ejected) from the pipe opening scales as: F 0U 2 D 2 ,

52
where U is the mean flow speed and D the pipe diameter. If the process is periodic, as for an
IC-engine with a period time T, then (2-25) implies that the radiated power scales as

0U 4 D 4 0 f 2U 4 D 4
Wd , (2-26)
c03T 2 c03

where f = 1/T and a compact source region is assumed. Comparing this result with the
monopole power in example 2-1 we get

Wm : Wd 1 : M 2 . (2-27)

This means that for small Mach numbers the monopole radiation from a pulsating jet will be
much larger (20 dB at M = 0.1) than the dipole radiation.

Example 2-3: Another application is for an axial fan, when the speed should be taken as the
rotational speed that scales as U f D , where f is the rotational frequency and D the fan

diameter. Equation (2-26) then implies that: W fan D 2U 6 .

2.34 The quadrupole source term

The quadrupole source term corresponds to s3 in equation (2-12), and the resulting sound field
is a superposition of quadrupole fields

2 Tij (y , t r / c0 )
p( x, t ) dVy , (2-28)
V
xi x j 4 r
where Tij u i u j ij is called the Lighthill tensor. First, it can be noted that viscous forces

ij are mainly of importance for the damping of sound waves and their importance for sound
production is negligible19. For the case of a compact source (2-28) reduces to

2 Qij (t x / c0 )
p(x, t ) , (2-29)
xi x j 4 x

19
The ratio between the momentum transport term u i u j (see eq. (1-17)) and the viscous forces is of the order
of the Reynolds number, which for most cases of interest is very large.

53
where Qij (t ) Tij (y, t )dV y is the quadrupole source strength. The far field component of
V

this field is given by

(t x / c )e e
Q
p (x, t )
ij 0 x ,i x , j
, (2-30)
4xc02
2 Q / t 2 and e x / x . The sound power radiated from this quadrupole
where Q ij ij x ,i i

source can be obtained from equation (1-58)

2
ij Q
Wq
ij
, (2-31)
0 c05
where 1 / 20 when i = j and = 1/ 60 when i j .
The quadrupole source term corresponds to sound production due to the momentum transport
u i u j in a flow. The time average of this term is called the Reynolds stress and accounts for
the extra viscosity experienced in a turbulent flow. The unsteady part of the momentum
transport is the noise generating part in the quadrupole term. This source term is important in
a free turbulent flow and is the main source of noise from a high speed jet.

2.35 Curles equation

U n
D
S V

Figure 2.4 Sound produced by an unsteady flow interacting with a solid body, eq. (2-32). The surface integral is
taken over the surface S of the body and the volume integral over the part of space V where Lighthills
quadrupole term gives a contribution.

The extension of Lighthills analogy to include the effect of non-moving solid bodies is called
Curles equation. In the derivation a turbulent flow is assumed and only the quadrupole
source term is considered. In principle, the equation is obtained by rewriting the time domain
version of the Kirchhoff-Helmholtz equation (1-48), using the free field Greens function and
the equation of motion. The result is

54
 2 u i u j pni u i u j n j u i ni
p ( x, t ) dV y dS y dS y , (2-32)
V
xi x j 4 r te S
x i 4r te S
t 4 r te

where the brackets denote that the fields are evaluated at t = te.

The first surface integral represents dipole fields associated with the unsteady pressure p and
momentum transport at the body. It can be noted that the unsteady surface pressure20 should
be interpreted as pressure associated entirely with the flow field. The second surface integral
represents monopole fields associated with the volume flow created by fluid induced
vibrations of the body. In many cases the flow induced vibrations can be neglected, and if the
body is acoustically compact Curles equation can be reduced to

2 u i u j Fi (t x / c0 )
p (x, t ) dV y , (2-33)
V
xi x j 4r t xi 4x
e

where Fi (t ) p(y , t )ni dS y is the fluctuating force exerted by the body on the fluid.
S

To show the relative importance of the two terms in (2-33) we will derive a scaling law. The
spatial derivatives in a flow driven sound field will scale as

/ xi 1/ M / D U / Dc0 ,

see eq. (2-14). The quadrupole source term will scale as: u i u j 0U 2 and the whole

integral21 in (2-33) as
1 0U 2 3 0U 4 D
p q 2 D . (2-34)
x xc02
This implies that the radiated power from the quadrupole term scales as

p q 2 0U 8 D 2
Wq 4x 2 . (2-35)
0 c0 c 05
For the dipole contribution we can use (2-26) with f U / D which gives

20
The exception being cases when there is no mean flow then the field should be taken as the acoustic field.
21
Note, for M << 1 the source region is compact so the phase variations are possible to neglect which means that
the total source strength is proportional to the volume.

55
 0U 6 D 2
Wd . (2-36)
c 03
Comparing the two expressions gives

W d : Wq 1 : M 2 . (2-37)

This implies that for a small Mach number flow around a body the dipole sound associated
with fluctuating forces and flow separation will dominate over the free turbulence
quadrupoles.

2.4 Jet noise

Potential core

D U
Turbulent
mixing region

Figure 2.5 Noise from a non-pulsating jet leaving a pipe (or nozzle). The quadrupole source term or the jet
mixing noise originates from the turbulent mixing region 4-5 D downstream of the nozzle.

The first problem treated by Lighthill and the main motivation for the development of his
theory was the noise produced by aircraft jet engines. After World war II jet engines
developed for military aircraft began to be used in commercial aircrafts. Although these
engines delivered much more thrust for a given weight, their noise emission was a big
problem. To model jet noise we will assume a non-pulsating airflow and if we neglect that the
exhaust is hot only the quadrupole source term remains. For small Mach numbers the source
region will be compact (see eq. (2-14)) and the power can be calculated from equation (2-31)

2
ij Q
Wq
ij
,
0 c05

where Qij (t ) u i u j dV y is the quadrupole source strength. A detailed calculation of this

V

expression requires the use of turbulence models [1], but following Lighthill we can derive a
scaling law that is useful for understanding the parameters that control the jet noise. The
quadrupole source strength will scale as: Qij 0U 2 D 3 where the jet is assumed to have the

56
same density as the surrounding gas. The time scale for the sound production is: T D / U .
Using these relationships in (2-31) gives

0U 8 D 2
Wq , (2-38)
c05
which is Lighthills celebrated U8-law for jet noise. The U8-law is well confirmed by
experimental data. For a cold (same temperature and density as the surrounding air) circular
jet, experiments show that
D 2 0U 3
W jet , (2-39)
4 2
where 8 10 5 M 5 for subsonic ( M < 1 ) jets. The factor can be seen as an acoustic
efficiency factor corresponding to the fraction of the kinetic energy flux that is converted into
acoustic power.

As seen from equation (2-39) the jet noise is driven by the kinetic energy flux which increase
proportional to U 3 . This implies that the jet noise cannot grow as U 8 for ever, asymptotically
it most follow the speed exponent for the kinetic energy. In the limit of very high Mach
numbers for properly expanded jets (no shock waves), measurements show that 10 4 , i.e.,
a constant fraction of the kinetic energy is converted to sound.

2.41 Low Mach number jets

Equation (2-39) shows that quadrupole or jet noise is a very inefficient sound source for small
Mach numbers. Normally the source is important only for flows where the Mach number
approach 1. For a small Mach number jet, the monopole or dipole mechanisms will instead be
more important. For instance, as shown by Examples 2-1 and 2-2, for a pulsating jet, e.g., an
IC-engine exhaust pipe, the monopole source term will dominate the radiation. Experiments
also show that for a non-pulsating jet the radiated power is proportional to U 6 when M << 1.
This flow speed dependence is characteristic for a dipole type of source, as seen from
example 2-2 that can be applied to a non-pulsating jet by assuming the time scale to be:
T D / U . Equation (2-26) will then imply that the radiated power associated with unsteady
momentum outflow from a turbulent jet is proportional to U 6 . But what about the monopole
contribution? If we use equation (2-21) with T D / U we get a sound power that scales as
U 4 . This is, however, not the correct time scale since it corresponds to pressure fluctuations

57
associated with an incompressible (potential) flow. To create acoustic energy by a monopole
(volume flow) we must create density changes. The time scale for this is given by / U . If we
use this relation in (2-21) it turns out that the monopole power scales the same way as the
dipole, i.e., as U 6 . This is why for a low Mach number non-pulsating jet we observe a U 6
dependence for the radiated sound power.

2.5* Noise from propellers

The classical way to treat noise from propellers is due to Gutin [5] and dates back to 1936.
The Gutin model is based on modeling propeller sound as generated by a time varying force
field created by the rotating propeller. An alternative and more general theory based on
acoustic analogies was developed by Ffowcs Williams and Hawkings 1969 and will be
described later in Chapter 4.

Here, a version of the Gutin model for propeller or rotor noise is presented. We assume that
each blade produce a resulting force which is constant and equal for all blades. The N blades
are uniformly distributed in the x1-x2 plane and are rotating with the angular velocity , and
the resulting (drag) force points in the x3 direction, see Figure 2.6.

x2

x1
F0 / N
x3

Figure 2.6 Model of propeller noise. The resulting force on each blade are all positioned on a circle with radius
a. The resulting drag force from the N blades acting on the fluid is F F0 e 3 .

58
Mathematically the force distribution can be represented22 as

F0 1
fV ,3
N m R
( R a) ( t m ) ( x3 ), (2-40)

where m 2m N and m = 0, 1, 2,, (N-1). The delta function in is when seen as a

function of t , a periodic sequence of deltas separated by 2/N, which can be

expanded as a Fourier series

( m ) a n expinN . (2-41)
m n

The Fourier coefficients are given by

N
N N
an
2
N
( ) exp(inN )d
2
. (2-42)

Using equations (2-41) and (2-42) we can represent the force distribution as

F0 1
f V ,3
2
R ( R a) expinN ( t ) ( x ).
n
3 (2-43)

Equation (2-43) shows that the source signal consists of harmonics which are multiples of a
fundamental frequency corresponding to the blade passing frequency N .

The formula for a dipole field (2-22) gives, if we only consider the far field component

x3 f V ,3 (y, t r / c 0 )dV y
p (x, t ) , (2-44)
4x

22
The 3-D delta function in cylindrical co-ordinates ( R, , x3 ) is :
1
(x x 0 ) ( R R0 ) ( 0 ) ( x3 x0 ).
R

59
where r x y . The volume integral can be evaluated using (2-43)

2
F0
fV ,3 dV y 2 expinN ( t )d ,
n 0
e

where t e t r / c0 . The x3 derivative of this expression is

2
N F0 r

x3 fV ,3 dVy
2 c0 x3
in exp inN ( te ) d ,
n 0

where r / x3 x3 / x cos in the far field. For the emission time we have

xy xe ae x a
te t t x t (e x e) ... ,
c0 c0 c0 c0

where e x is the unit vector in the x direction, e is a unit vector in the x1-x2 plane and the
omitted terms vanish in the far field. The unit vectors can be expressed as

e x (sin cos 0 , sin sin 0 , cos )

,
e (cos , sin ,0)

which implies that e x e sin cos( 0 ) . Using these last results equation (2-44) becomes

iF0 cos


p (x, t ) ( i ) nN
nN J nN nNM r sin expinN ( 0 x / c 0 t ) , (2-45)
4xc0 n

where J nN is the Bessel function of order nN and M r a / c0 is the Mach number of the
rotational motion. To obtain (2-45) the following formula has been used

2
i N
2 exp(iN ) exp(iX cos )d J
0
N (X ).

60
Equation (2-45) shows that the sound field is composed of the blade passing frequency and its
harmonics and that it is proportional to the steady force. For small arguments the Bessel
function can be approximated by

( X / 2) N
J N ( X ) .
N !
This shows that the directivity of the field is proportional to: cos sin nN 1 which means

that the far field is zero for 0 , / 2 and .

In practice, the sound generated by a propeller will be much larger than the values predicted
by the above simple model. The main reason for this is the effect of steady and unsteady
inflow disturbances which will modulate the blade forces. The steady inflow disturbances
(deviations from a uniform inflow) will amplify the rotational harmonics, while the unsteady
flow (inflow turbulence and separation from the blades) will create a broad band spectrum.

61
References
1. M.E. Goldstein (1976) AEROACOUSTICS. McGraw-Hill.
2. D.W. Bechert (1980) Journal of Sound and Vib. 70, 389-405. Sound absorption
caused by vorticity shedding, demonstrated with a jet flow.
3. A.Powell (1990) ASME Journal of Vibration and Acoustics 112, 145-159. Some
aspects of aeroacoustics: From Rayleigh until today.
4. M.S. Howe (1975) Journal of Fluid Mech. 71, 625-673. Contributions to the theory of
aerodynamic sound, with applications to excess jet noise and the theory of the flute.
5. L. Gutin (1936) Phys. Z. Sovjetunion 9, 57-71. On the sound field of a rotating
airscrew.

62
Problems
1. Show that Lighthills equation
1 2 p 2 p f V ,i 2
m ( p c 2
0 ) ui u j ij ,
c02 t 2 xi xi t t xi xi x j

in a homogeneous ideal fluid (no heat conduction and viscosity) with no external
sources and zero mean flow, reduces to the classical wave equation if we neglect
second order terms.

2. Discuss and list aeroacoustic sound sources on a car that contribute to the exterior
sound. Group the sources in monopole, dipole and quadrupole types and order them
after their expected strength (sound power).

3. Which type of aero-acoustic source (monopole, dipole or quadrupole) is the dominant in the
following situations:
- hand clapping
- sound from a small insect (oscillating wings)
- pulsating jet (M<<1)
- cavitation
- flow separation around a car antenna

4. Jet noise

U Free
turbulence Free
D field

a) By using Lighthills U8 law, show that the ratio of the sound power from two jet
engines with different diameters but the same thrust is given by

63
 6
W2 D1
.
W1 D2

What is the reduction in dB if we increase the diameter with a factor 2 ?

b) Compare the kinetic energy flow (E ) from two engines with equal thrust. Show that

E kin , 2 D1
.
E kin ,1 D2

Since the kinetic energy is lost in turbulence this implies that a larger engine also is
more efficient, i.e., creates the desired thrust with smaller losses !

5. Flow separation (dipole) noise

Flow around an object sticking out (e.g. a side mirror) on a vehicle can create a
periodic flow separation. For objects with a regular shape the frequency of the flow
separation (Strouhal frequency) is approximately:
U
f s 0.2 ,
D
where U is the mean flow speed approaching the object and D the cross-dimension
perpendicular to the mean flow. The amplitude of the fluctuating force (RMS-value)
will typically be around 10 % of the total drag acting on the object:
1
Fd 0U 2 S ,
2
where S is the cross-sectional area perpendicular to the mean flow.

64
a) Compute the frequency and fluctuating force (RMS-value) for a car side mirror with:
D = 0.1 m and S = 0.02 m2 assuming a speed of 120 km/h (the max allowed on
Swedish roads). Compute also the wavelength of the sound and compare with the size
of the object. Can the source be considered compact, i.e., be treated as a point source?

b) Assuming a point force and a free field then the farfield sound pressure can, if we
neglect the effect of the source motion, be computed from (Eq. (2-24)) :
F (t x / c0 ) e x
p (x, t ) .
4xc0
Use this result to prove that the radiated sound power is given by:
F 2
Wd .
12 0 c 03
c) Compute the maximum sound pressure level at 1 m and the radiated acoustic power
level for the side mirror in a).

d) Estimate how much the sound pressure/power level will decrease if the speed is
reduced with 50 %.
Data: 0=1.21 kg/m3, c0=340 m/s.

6. Apply Curles equation to an acoustically compact rigid body performing an

oscillatory motion in a homogeneous fluid at rest (no mean flow) and show that the
resulting sound field becomes:

F 0Vb U
a) p (x, t ) b
, where F is the force created by the oscillating motion
4 x
on the fluid, Vb is the volume of the body and Ub(t) the velocity of the oscillation.

b) The result in a) can be used to analyse the sound generated by a flying object such as
an insect. For this case the force (generated by the wings) is related to the motion of
, where the density often is close to
the body by Newtons second law: F bVb U b

(but normally smaller than) water for most life forms. This result implies that when
flying in air the effect of the force completely dominates ( b 0 ) over the sound
due to the motion of the body. The same analysis can be applied to a swimming fish

65
 but in water the two densities are very close and there will be virtually no dipole sound
created. For insects the sound should therefore be given by

F
p (x, t ) .
4x
In order to fly, an insect must produce a time
averaged force in the vertical direction (say x3) F(t)

of Mg, where M is the mass of the insect and g the

acceleration due to gravity. If we assume that the
time variation of the force is harmonic with zero t
as the minimum value then: F (t ) Mg (1 cos(t )) .

With this information, calculate the sound produced by a mosquito with g = 10 m/s2,
M = 5x10-6 kg, and f = 1000 Hz at a distance of 0.3 m. Can the mosquito be heard ?

7. Compressed air is injected into a straight duct and creates a non-pulsating jet. The jet
is assumed to have the same temperature as the surrounding gas and does not interact
with any nearby solid objects. The source can therefore be taken as a pure quadrupole
source that is compact.

If the duct diameter is small, then only plane waves will exist. In that case Lighthills
equation is reduced to a 1-D case which gives in (the frequency domain)

2 2T11
2 k 2 p ,
x1 x12

66
 where x1 is parallel with the duct axis and T11 is the Lighthill quadrupole source term.
For a compact source region we have: T11 A ( x1 ) , where A is an amplitude factor
that determines the strength of the source and the source is assumed to be at the origin.

a) To solve the above problem we can first determine a Greens solution for the 1-D
wave equation. This Greens solution is defined by:

2
2 k 2 G ( x1 ) .
x1

Integration of this equation shows that G is continuous at the origin and that the
derivative has a jump that satisfy
G
G 1 .
x
1 0 x1 0
Show that the Greens function is given by

1 exp(ikx1 ), x1 0 1
G exp(ik x1 ) .
2ik exp(ikx1 ), x1 0 2ik

p exp(ikx1 ), x1 0
Hint: Assume that the field can be written as: G and
p exp(ikx1 ), x1 0
determine the amplitudes by using the behavior at the origin.

2
b) The solution for the given source term: s( x1 ) A 2 , can for the case of an infinite
x1
duct be obtained directly from the Greens solution. It is given by

2 G
p ( x1 ) A 2 .
x1

A k
Show that this implies: p ( x1 ) exp ik x1 .
2i
c) Show by scaling the result in b) in the usual way that the generated sound power
from a jet in a duct with only plane waves is proportional to:
2
W1Djet p U 6 .

67
d) Assume that the speed of the jet is doubled how much less will the power increase
compared to a jet in a 3-D space ?

8. A thermoacoustic source term due to unsteady combustion or heat release can be

derived using the equation of state for the fluid. Assuming an ideal gas and neglecting
heat conduction, the fluid pressure p can be seen as a function of density, and heat
added per unit volume, q, i.e., p = p(,q). Applying thermodynamics, linearizing and
assuming an ideal gas gives

p c02 ( 1)q ,

where the prim denotes a fluctuating quantity and is the specific heat ratio which is
assumed constant. The thermo-acoustic source term can then be written as

( 1) q
sthermo (x, t ) ,
c02 t

where q is the unsteady added heat power per unit volume and dot denotes a time
derivative.

Source Fluid at
region with rest
unsteady
combustion s(y,t)
p(x,t)

y r

a) An unsteady point heat source located at the origin varies harmonically around a
steady heating power. The unsteady part can be described by

q (x, t ) Q exp(it ) (x) ,

where Q is the complex valued amplitude. This unsteady heat source will give rise to

68
 an acoustic point source at x = 0. Derive an expression for the free field sound
pressure generated by this source from the known free field Greens function and the
above expression for the thermo-acoustic source term.

b) Show that the acoustic power radiated by this thermoacoustic source is given by

( 1) 2 2Q 2
Wac ,
40c05

where ~ denoted the R.M.S value.

Typically, only a small fraction of the heat power generated by a process (e.g.
combustion) is converted into acoustic power. As a measure of this, the concept of
acoustic efficiency is introduced as

Wac
,
Wheat

where the subscripts ac and heat refer to the acoustic power and steady state power of
the process, respectively.

~
c) Assume that the R.M.S value of the heat power in the process described above, Q , is a
fraction, , of the steady state value of the heat power, Wheat. Calculate the radiated
acoustic power in W and dB rel. 1 pW and the acoustic efficiency for the case:
Wheat 10 kW, 0.05, f 250 Hz, c0 340 m/s, 0 1.21 kg/m3 , 1.4.

69
70
Solutions

The first source term is: s1 m ( p c0 ) , since the fluid is source free we
2
1.
t t
can put m = 0. Under adiabatic conditions we have to the first order: p c02 , i.e.,

neglecting second order (quadratic) terms: p c0 0 . Therefore to the first order

2

s1 = 0.
fV ,i
The second term is: s2 , which is zero for a source free fluid.
xi

2
The third term is: s3
xi x j
uiu j ij , neglecting viscosity and second order terms
(velocity squared) this term is zero.

2. -

3. -
0 D 2U 8
4. Lighthills U8-law implies: W jet .
c05

a) Applying conservation of momentum for an engine gives that the thrust F (under
steady state) equals: F out SoutU out
2
in SinU in2 assume U out U in out SoutU out
2
.

Equal thrust for two engines implies:

out D1,2outU1,2out out D2,2 outU 2,2 out D1,outU1,out D2,outU 2,out , (1)

where we assume that the density of the outlet jet is unchanged. Comparing the sound
powers implies:
8 6
W2 D22U 28 D2 D D
2 8 equal thrust eq. (1) U 2 / U1 D1 / D2 22 1 1 .
W1 D1 U1 D1 D2 D2

An increase of a factor 2 in the diameter gives a reduction of the sound power with 26
or close to 18 dB.

71
b) The flux of kinetic energy is: E D 2U 3 . For two engines with the same thrust but
different diameters we then get:

E 2 D22U 23 D
2 3 equal thrust 1 .

E1 D1 U 1 D2

5 a) The given formula for the periodic (Strouhal) flow separation frequency gives:

insert given values U 120 km/h 33 m/s, D 0.1 m 67 Hz .

U
f s 0.2
D
1
The drag force is given by: Fd 0U 2 A . Ilnserting the given values gives:
2
Fd 13.4 N . The fluctuating RMS drag is 10 % of this i.e: 1.34 N.

The wavelength is: c / f s 340 / 67 5 m >> D, which implies that the source can
be considered to be acoustically compact i.e. it can be treated as a point source.

b) To obtain the power we must integrate the intensity over a large spherical surface in
the farfield (Eq. (1-58)) :
p 2
W p u r dS lim r dS . (1)
Sr Sr
0 c0

F (t x / c0 ) e x F (t x / c0 ) cos
The farfield pressure is here given by: p(x, t ) .
4 xc0 4 xc0

where is the angle between the force and the direction

F x
to the observer x. Inserting this expression in Eq. (1) gives:

F 2 cos 2 2 cos 2 sin

0 8 c3 d
F

W lim x
2
dS dS 2 x sin d
Sr
16 2 x 2 0c03 0 0

F 2 cos3 F 2
.
80c03 3 0 120c03

c) For a harmonic force we get: F F exp(it ) F iF exp(it ) , inserting this into

2
the expressions in b), noting that the RMS value is given by F 2 2 F / 2 2 F 2 ,

72
 gives:

F cos F 2 F 2 f 2 F 2
p(x, t ) p(x, t ) max p 2
4 xc0 4 xc0 16 2 x 2c02 4 x 2c02
inserting values 0.0174 Pa
2 F 2 f 2 F 2
and W insert given values 9.9 106 W .
120c03 30c03

This gives a sound max sound pressure level of (rel. 20 Pa):

L p 20log10 ( p / 2 105 ) 59 dB and a sound power level of

LW 10 log10 (W /1012 ) 70 dB.

d) The scaling for flow separation is obtained from eq. (2-36):

0U 6 d 2
W .
c3
A 50 % speed increase gives an increase in the sound power with 1.56 or
6 10 log10 1.5 11 dB . A 50 % increase in d the size of the object gives an increase of
1.52 or 3.5 dB.

6a) x
Ub(t)
y
d <<
Curles equation (2-32) implies:
n

2 uiu j
uiu j n j pni ui ni t e
p (x, t ) dVy dS y
te te
dS y ,
V
xi x j 4r S
xi 4r S
t 4r

where V is the region outside the body and S is the surface of the body. Since there are
no mean flow we only have acoustic fields contributing to the terms in the integral.
Keeping only linear contributions means that the terms which are encircled can be
neglected. A linear approximation of the last term gives: ui 0ui .

The sound field can therefore be written .

73
 p ni 0ui ni
p (x, t ) dS y dS y .
S
xi 4r S
t 4r

We will now make a Taylor expansion of the source integrals (compare multi-pole !).
Since the source is compact (d << ) the first term different from zero in the expansion
will dominate.

The Dipole term:

p (y , t x y / c0 ) p (y , t x / c0 )
Taylor expansion x 1 y x ..... ,
4 x y 4x

where x x1 , x2 , x3 and x = x . Inserting this in the integral gives:

1 F(t x / c0 )
pI p (y , t )ni dS x , where F is the force that the
xi 4x S 4x

Fi ( t )
t x / c0

body exerts on the fluid.

The Monopole term:

0ui (y , t x y / c0 ) u (y, t x / c0 )
1 y x ... 0 i , where u(y, t ) U b (t ) .
4 x y 4x

Inserting this in the integral gives:

1 1
pII 0 U b ndS y x y U b n dS y ... . Here the first integral is
t 4x S 4x S t x / c 0

zero: U
S
b ndS y div (U b )dV y 0 , since U b is independent of y. Concerning the
Vbody

second integral we have for the y1 component

74
 y1U b,1
y U
S
1 b n dS y
Vb
y1
dVy U b,1Vb and similar for the other y-components. This

implies that: yU
S
b n dS y U bVb . The first contribution different from zero in the

monopole term is therefore a field of dipole type.

(t x / c )V
0 U
pII x b 0 b
, where a dot denotes a time derivative.
4x

Finally if we add the two contributions together we get:

p pI pII x
F(t ) U (t )V
0 b b t x / c0
.
4x

Thus the sound produced from a small oscillating acoustically body will be of dipole
type with two contributions. One from the fluctuating force on the fluid created by the
motion and one associated with the acceleration of the fluid volume displaced by the
body.

b) Sound from insects

An insect can be regarded as a small oscillating object. Of course it is not a rigid body
but one can show that the any deformation (at constant volume) of the body will result
in higher order multi-poles, which for a compact body are weaker sources.

If the wings create a force F on the fluid then the reaction force F acts on the insect.
, where U
According to Newtons law this implies F bVb U should be
b b

interpreted as the acceleration of the centre of gravity. Inserting this in our formula for
small oscillating bodies gives.

b 0 Vb U b (t x / c0 )
p x .
4x

75
An insect as most living organisms consists mostly of water so the density lies
typically in interval 500-1000 kg/m3, say 750 kg/m3. Since the air density is around
1 kg/m3 the contribution from the displaced air volume can be neglected and sound is
only produced by the force created by the wings. A simple model for the force is: For
steady flight the time average of the force must equal the weight of the insect. Lets
assume that the time variation is harmonic and that the force only acts in the vertical
direction. If the force has zero as the minimum value it must then look as...

F(t)

Mg

F(t ) Mg 1 cost e g , M = the insect mass, g the acceleration of gravity, eg a unit

vector in the vertical direction and 2f , where f is the wing frequency. This gives
the sound field:
cos (t x / c0 ) Mge g sin (t x / c0 )
p Mge g x far field ex .
4x 4xc0

The maximum amplitude in the field is:

Mg
p max
4xc0

Typical data: A typical Swedish (?) mosquito M 5 106 kg, f 1000 Hz,
g 10 m/s2, c = 340 m/s. This gives at a distance of 1 m

p max 7 105 Pa or L p 11 dB rel. 20 Pa,

which theoretically is audible (0 dB at 1000 Hz is the limit !). But in a normal

environment with background noise levels between 25-30 dB is needed.

76
 At 0.3 m the level will be 21 dB and at 0.1 m 31 dB, which implies that typically a
mosquito would be detected on a range of 0.1-0.3 m.

7a) Using the given properties of G at the origin implies:

p p
(1)
ikp ikp 1

1
From eq. (1) it follows that: p , which gives the desired result.
2ik

b) Taking the derivative of G implies: G / x1 ikG 2G / x12 k 2G , where -/+

refers to the pos./neg. x1 direction. Putting in G from a) here gives:

Ak
p exp(ik x1 ) . (2)
2i

2
c) The sound power W in a plane wave is proportional to p . Using eq. (2) implies:
2
W p A k 2 . From aeroacoustic scaling laws is follows that: A U 2 and
2

k f / c U . Combining these results gives: W U 6 .

d) In the 3D case the speed exponent is 8 therefore the difference is: 286 22 4 or
6 dB less.

( 1)
8a) From the given information the source is found to be: sthermo i Q (x) . The
c02
known (Greens) solution for a unit amplitude harmonic point source at the origin is:

exp(ikx)
G .
4 x

( 1)i Q exp(ikx)
This implies that: p ( x) .
4 xc02

77
b) The power is obtained by integrating the intensity for the far-field component over a
(large) sphere. In this case where we only have a far-field component which only
2

p
2 ( 1) 2 2 12 Q 2

varies with x we simply get: W 4 x 2 , where Q 2 12 Q .
2 0 c0 40 c05

c) Inserting values gives: Wac 0.0014 W or 91.5 dB rel. 1 pW. The efficiency is:

1.43 107 .

78
 Chapter Three

The effect of mean flow

In this chapter we will discuss the effects of mean flow on sound generation and propagation.
A wave equation with source terms is derived for the case of sources embedded in a
homogeneous mean flow (constant mean flow velocity U0 and fluid state p0, 0 and c0).
Solutions to this convective wave equation with a source term will be discussed in Chapter 4.
Plane wave solutions to the convective wave equation without source terms is then treated and
applied to the scattering of sound at a sudden mean flow discontinuity (a vortex sheet). One
effect of flow is that it changes the equation for the acoustic energy balance. An expression
for the time averaged acoustic intensity when mean flow is present is therefore derived. An
important case where sound propagates in flow is pipe and duct systems. In such systems the
sound waves can only propagate along the duct axis and will form standing waves or modes
over the duct cross-section. A given mode can only propagate above a certain cut-on
frequency and in a hard (rigid) walled duct only the plane wave mode can propagate at all
frequencies.

3.1 The convective wave equation

It is possible to generalize the derivation of Lighthills equation for a source region embedded
in a homogeneous fluid flow [1]. Here, we will study a special case of this. We will derive a

79
modified wave equation starting from the conservation of mass (1-13)23 and the conservation
of momentum in Eulers form (1-18)

D
Dt u m
. (3-1)
Du p f mu
Dt V

Introducing small disturbances, the fields can be written as p p 0 p , 0 and

u U u and the source terms24 as m m and fV fV . Inserting this into (3-1) and

neglecting second order terms gives

D0
Dt 0 u m
, (3-2)
D0 u p f mU
0 Dt V 0

where D0 / Dt / t U is the linearized convective derivative, with the velocity U

equal to the mean flow velocity. Assuming adiabatic changes of state we have: p c02 , and

performing the operation D0 (3 2)1 / Dt (3 2) 2 we get

1 D02 D m
2 2 2 p 0 fV mU . (3-3)
c0 Dt Dt

This is the convective wave equation with source terms. The source term s(x,t), i.e., the right
hand side of (3-3), can also be written as
m
s (x, t ) fV 2Um . (3-4)
t
The first contribution to this source term can be seen as a monopole source and the remaining
as dipole sources. It is interesting to note that the fluid motion converts part of the unsteady

23
By using the relationship ( u) u u the time derivative can be rewritten as a total time
derivative.
24
Note that the assumption of constant mean flow implies that the constant source terms, m0 and f V , 0 are equal
to zero.

80
mass flow into a dipole source contribution. The solution to this convective wave equation
with source terms will be given in Chapter 4 when we discuss moving sources, as if we
observe the sound field from a co-ordinate system moving with the fluid the wave equation
reduces to the classical wave equation, but with sources moving with a constant velocity U0.

3.2 Plane wave fields

x
3.21 Wave number and phase speed

A harmonic plane wave in 3-D can be written as

e

p(x, t ) p expi (t k x) , (3-5)

U

where k ke is the wave number vector and e is a given unit vector along the direction of the
wave propagation. Inserting (3-5) in the homogeneous convective wave equation (3-3) gives,
if p 0 , a relationship between wave number and frequency

1
2
k U 2 k 2 0 .
c0
Solving this equation implies that

k0
k , (3-6)
1 M e
where k 0 / c 0 , the signs +/- denote wave propagating in the +e and in the e direction,
respectively25, and M = U/c0. The phase speed for the plane wave is

c ph c0 1 M cos ,
k
where is the angle between U and e.

3.22 Characteristic impedance

In order to calculate, e.g., the reflection and transmission of a plane wave incident on an
interface, it is necessary to relate the pressure field to the acoustic particle velocity in the

25
The interpretation of the direction of propagation is not obvious. To prove it one can for instance check the
direction in which the wave transports acoustic power. But this requires a definition of the acoustic intensity in a
medium with flow, see section 3.3.

81
plane wave. The relationship is found by substituting (3-5) into the linearized equation of
motion (3-2)1. This implies that the velocity field must have the same space-time variation
and that the pressure and particle velocity amplitudes satisfy

0 i ik U u ikp .

Using equation (3-6) this leads to

e
u p . (3-7)
o c 0

In other words, the characteristic impedance is unchanged from the no flow case, section 1.43.

3.23* Scattering at a mean flow discontinuity (a vortex sheet)

x3

U(x3)
Region b
Ub

Ua

Region a
Figure 3.1 Sound propagation across a shear flow profile. The transition in mean flow speed can be modelled as
a sudden jump between two regions a and b, a so called vortex sheet as long as << .

In order to model sound propagation in a shear flow profile, we will investigate sound
scattering at a sudden mean flow discontinuity. This method can be applied when the
thickness, , of the region where the mean flow changes is much smaller than the wavelength.
To calculate the scattering of plane waves incident on such a discontinuity, we need coupling
conditions relating the acoustic variables over the jump in mean velocity from region a to
region b. Assuming no mass flow across the jump, the conservation of momentum will imply
continuity of pressure, or pa pb . The assumption of no mass flow also implies that we can
define an impermeable surface (called a vortex sheet) between the two fluids. Let the
position of this moving surface be defined by S(x,t) = 0 at a certain time t. The normal
velocity, vn, for a point on the moving surface is found from:

82
 S dx S / t 0 v n dx / dt S / S S / t / S .

Since the surface is impermeable the normal velocity of the fluid next to the surface will also
be vn. This gives: u S / S S / t / S or

DS S
u S 0. (3-8)
Dt t

x3
We will now assume that the surface (vortex sheet) is a plane
defined by (say) x3 = 0 at equilibrium. Then the position of the Ub
surface can be described by: S x3 ( x1 , x 2 , t ) , where s
s
x1
Ua
is the displacement from the surface equilibrium. Inserting S in
(3-8), given a mean flow field U (U1 ,U 2 ,0) in the x1 - x2 plane and a superimposed acoustic

field u (u1 , u 2 , u 3 ) gives: (U 1 u1 ) (U 2 u 2 ) u 3 0 . In linearized form,

s s s

t x1 x 2

this expression becomes

D0 s
u3 . (3-9)
Dt
The fluid velocity component u 3 is related to the fluid displacement 3f , both in the a and b
region, by exactly the same (linearized !) relationship. This implies that

D0 s D0 3f
,
Dt Dt
which applies to both the a and b region. From this result it follows that the fluid
displacement in the x3 direction matches the displacement of the vortex sheet26. The result
also implies continuity of fluid displacement in the normal direction across a fluid interface

af,3 bf,3 , x3 0. (3-10)

We will now study sound scattering from a plane interface between two homogeneous fluid
regions a and b. Each fluid has a constant flow velocity U a (U a1 ,U a 2 ,0) and

U b (U b1 ,U b 2 ,0) parallel with the interface.

26
From the derivation it is clear that the result applies to any moving impermeable surface in contact with a
fluid, e.g., a vibrating solid boundary.

83
 x3
b , cb

Ub tr
x1
Ua
in re a , ca

Figure 3.2 Sound scattering at a plane interface between two fluids.

A harmonic plane wave is assumed to be incident from region a as illustrated in Figure 3.2

pin (x, t ) p in expi t k in x , x3 < 0, (3-11)

where k in k in e in and e in (sin in ,0, cos in ) . The incident plane wave will create a reflected
and a transmitted plane wave

pre (x, t ) p re expit k re x , x3 0

, (3-12)
ptr (x, t ) p tr expit k tr x , x3 0

where k re k re e re and k tr k tr e tr .The reflected and transmitted waves will share the wave
number with the incident wave along the interface. This principle of wave number matching is
equivalent to Snells law in optics. It implies that the reflected and transmitted waves has no
component in the x2-direction and that,

k in e in e1 k re e re e1 k tr e tr e1 . (3-13)

From the first equation in (3-13) and by using (3-6) for the wave numbers it follows that
in re , i.e., the angle of the reflected wave equals the angle of the incident wave, which
also implies that kin kre . From the last equation in (3-13) it follows that

kin sin in ktr sin tr , or

sin in sin tr
, (3-14)
c a U a1 sin in cb U b1 sin tr

from which the angle of transmission can be obtained. The amplitude of the reflected and the
transmitted waves can be related to the incident wave via the transmission, T p tr / p in , and

84
reflection coefficient, R p re / p in . To determine these coefficients we need the coupling

conditions at the interface. The continuity of pressure at the interface pa pb implies

1 R T. (3-15)

To use the other coupling condition, (3-10), stipulating continuity of normal displacement, we
must relate the displacement field to the pressure field. This is done via the linearized
equation of motion, (3-2)2, which when applied to a harmonic plane wave
gives: 0 i ik U u ikp , where the velocity amplitude is related to the fluid displacement

by: u D0 F / Dt . From this it follows, using (3-6), that

iep
F , (3-16)
k 0c02

where e is a given unit vector along the direction of propagation. Taking the projection of this
displacement in the x3-direction, the continuity of normal displacement at the plane x3 = 0 can
be formulated as

(ein e3 Re re e3 ) etr e3
T,
kin a ca2 ktr b cb2

where e in e 3 cos in , e re e3 cos in and etr e3 costr . This implies that

cosin (1 R) cos tr
T. (3-17)
kin a ca2 ktr b cb2

Using equation (3-15) and (3-17) we can now solve for the coefficients R and T

b cb2 sin 2 in a ca2 sin 2tr

R , (3-18)
b cb2 sin 2in a ca2 sin 2 tr

and T = 1+R. To obtain (3-18) the relationship: kin sin in ktr sin tr is used, see (3-14).

Example 3-1: For high frequencies, sound waves can locally be seen as plane waves or rays.
To track the propagation of rays in a fluid with a varying state (speed of sound and mean

85
flow), e.g., the atmosphere, we can consider a series of infinitesimal jumps which describe the
continuous variation. Equation (3-14) can then be formulated in differential form as

x3
Uh(x3)
sin
d 0, (3-19)
c U h sin

which applies to a fluid where, e.g., a vertical stratification (variation of sound speed and flow
speed) exists and the fluid motion is in the horizontal direction only. To get a differential
equation one can for instance introduce the distance along a ray path. Equation (3-19) is the
basis for so called ray tracing methods that are used to study sound propagation (refraction) in
the atmosphere or in the sea.

3.3 Acoustic intensity

An energy balance equation can be written as

e
I q, (3-20)
t

where e is the energy density, I the intensity vector and q the source term. There are two
strategies to derive an acoustic energy balance. One is to start from the linearized equations of
conservation of mass and momentum and manipulate them into the form of (3-20). An
example of this procedure is given in Chapter 1 for the case of a homogeneous fluid with no
mean flow. The other strategy is to start from the equation for conservation of energy and
introduce small disturbances around a mean flow state. It is not obvious that the two
procedures will produce the same result and a discussion of this subject can be found in
Goldstein [1]. Here we will start from the energy equation (1-21) and derive a general
expression for the time averaged acoustic intensity. A derivation along the same lines, which
also includes the energy density and the source term, is given by Myers [2].

The energy equation for an ideal (no viscosity and heat conduction) fluid is

ht p ht ui 0 , (3-21)
t xi

86
where ht p / u 2 / 2 e is the total enthalpy per unit mass (see (1-21)). The time average
of (3-21) is

ht u i 0 . (3-22)
xi

We now introduce the mass flux vector: mi u i . The mass flux and the enthalpy is then

split into a time averaged part and a fluctuating part: mi mi mi and ht ht ht , where

mi 0 and ht 0 . Substituting this in (3-22) gives

mi h
xi
mi ht miht 0
xi
ht mi t
xi xi
miht 0 . (3-23)

From the time average of the equation for conservation of mass (1-13) it follows that

mi / xi 0. (3-24)

We will now assume a flow with no vorticity, a potential flow, and for such a flow one can
introduce a velocity potential: u i / xi . Inserting this into the equation of motion for an
ideal fluid with no sources, equation (1-18), we get

1 p 1
1 p .
t xi x x x x t x 2 x x x xi
j j i i i i j j

From the first law of thermodynamics it follows that27: de pd (1 / ) Tds . The enthalpy is
defined by: h e p / , this implies: dh dp / Tds. For a fluid with constant entropy we
then have: dh dp / or h p / . Using this result in the equation above we get

u2
h 0. (3-25)
xi t 2

This is an equation which in fluid mechanics is known as Bernoullis equation. Taking the
time average of this equation implies that

27
It is assumed here that the same equation of state applies everywhere in the fluid.

87
 ht / xi 0 . (3-26)

Using (3-24) and (3-26) in equation (3-23) finally gives the relationship

miht 0. (3-27)
xi

This relation can be interpreted as a general definition for a time averaged acoustic intensity
vector. The vector field is source free for an ideal (loss free) fluid with constant entropy in
which there exists a potential flow (no vorticity). When these conditions are not satisfied,
there is a source term present which also can be negative, i.e., dissipate acoustic energy (see
Figure 2.2). To express (3-27) in terms of our normal acoustic field variables we observe that:
mi ( 0 )(U i u i ) 0U i u i 0 u i U i , which implies: mi 0 u i U i . The

fluctuating enthalpy is given by: ht p / 0 U i u i , when the entropy fluctuations are zero.
These last results lead to the following expression for the acoustic intensity

I 0 u U p / 0 U u . (3-28)

The time average of this is given by

p2 M i
I i p u i 0 c0 u iu j M j p u j M j M i , (3-29)
0 c0

where the relationship p c02 is used.

3.4 Sound in ducts

x3

x1

x2

Figure 3.3 A straight, cylindrical flow duct carrying a homogeneous fluid with a constant flow velocity
U = (U,0,0) parallel with the duct axis.

88
Assuming a harmonic time dependence, an ansatz for a propagating wave along the axis of a
duct can be written as

p (x, t ) p (y ) expi (t k1 x1 ) , (3-30)

where (y ) is a function describing the pressure distribution over the duct-cross section, y is
a position vector in x2-x3 plane defining a point on the duct cross-section, and k1 is the wave
number along the axis. Substituting equation (3-30) into the homogeneous wave equation,
(3-3), results in

2 k 2 0, (3-31)

2 2
where 2 2 , k 2 k 0 Mk1 k12 , M = U/c0 and k 0 / c 0 . Together with the
2 2

x 2 x3
appropriate boundary conditions at the duct wall, equation (3-31) defines an eigenvalue
problem for (y ) . A standard assumption is to assume rigid walls, i.e., that at the wall the
velocity component normal to the duct surface is zero. This assumption is normally a good
approximation for ducts filled with gases, but not for liquid filled systems. It can be shown [3]
that for rigid walls, equation (3-31) has an infinite sequence of solutions n (modes or

eigenmodes) and corresponding real-valued wave numbers k 2 ,n , n = 0, 1, 2,., that form a

growing sequence. For a rigid walled duct the smallest eigenvalue k 2 , 0 in the sequence is

always equal to 0 and this zero order mode corresponds to a plane wave. The non-plane
modes, corresponding to n > 0, are called higher order modes. For a rigid walled duct it can
also be shown that the eigenfunctions form an orthogonal set over the duct cross-section, i.e.,

S
m (y ) n (y )dS S mn , (3-32)

where S is the duct cross-section and mn is Kroneckers delta28.

For each mode n the corresponding axial wave number k1,n can be calculated from

k 0 Mk1,n k12,n k 2 ,n .
2

28
This implies that: 0 1.

89
The solution to this second order algebraic equation is

k 0 1 f nc / f
2
Mk 0
k1,n , (3-33)
1 M 2 1 M 2
c0 k ,n
where f nc 1 M 2 is called the cut-on frequency for mode n. The +/- sign in
2
equation (3-33) corresponds to propagation in the positive and negative x1 direction,
respectively. It can also be shown that for f f nc the n:th mode decays exponentially in the
direction of propagation and transmit no acoustic power. This type of field is called an
acoustic nearfield.

The phase speed for mode n is defined by

c ph ,n , f f nc . (3-34)
Rek1,n

The phase speed for the higher order modes will be infinite at cut-on and then asymptotically
approach c0 for high frequencies.

A general expression for a harmonic sound field in a duct is obtained from superposition of all
possible modes, and the total pressure can be expressed as

p (x, t ) p n n (y ) expi (t k1,n x1 ) p n n (y ) expi (t k1,n x1 ) . (3-35)

n

3.41 Rectangular and circular ducts

Ducts with rectangular or circular cross section are the two most common duct types used in
practice. For a rectangular duct, with dimensions 2a x 2b, and with rigid walls the
eigenfunctions and eigenvalues are given by [3]

2 2
m n
2

,mn
k
2a 2b

, (3-36)
nx3
mn ( x 2 , x3 ) m mx 2 n
2a 2b

90
where m, n = 0, 1, 2, 3, and j ( x) cos( x) when j is even and j ( x) sin( x) when j is

odd. As can be seen from equation (3-36) the modes are the plane wave and the mode shapes
that are standing waves over the duct cross-section, i.e., that satisfy k L m / 2 in both the
x2 and x3 directions.

Example 3-2: Calculate the cut-on frequency for the first higher order mode in a square duct
with cross-section 2a x 2a. From equation (3-36) it follows that the first higher order mode is
degenerate, i.e., there are two possibilities (m,n) = (1,0) and (m,n) = (0,1), giving the same
cut-on frequency. From equation (3-33) we find (assuming M = 0) that
c0
f1c .
4a

Thus, for a square duct with a side length of 2a = 0.1 m, the cut-on for the first higher mode
will occur at (c0 = 340 m/s) 1700 Hz. For frequencies smaller than this, only the plane wave
propagates in the duct.

For a circular duct with radius a and rigid walls the eigenfunctions and eigenvalues are given
by [3]
J m (k ,mn a) 0

, (3-37)
(r , ) exp(im ) J (k r )
mn m , mn

where m ,....,1, 0, 1,...., and n = 0, 1, 2,, Jm is the Bessel function of order m and
J m is its first derivative, and r and are the polar co-ordinates over the duct cross-section. In

the table below, the first eigenvalues for the case of a circular duct is given.

Table 3.1 The eigenvalues of the first five higher order mode for a circular duct. Note that k , mn k ,( m ) n .

k ,10 a 1.841 k , 20 a 3.054 k , 01 a 3.832 k ,30 a 4.201 k , 40 a 5.318

The solution in (3-37) is called a spinning mode solution, as for a fixed axial position each
exp i (m t ) , which implies
mode rotates (spins) in the circumferential direction: pmn

d / dt / m where is the phase. This type of solution is useful to describe the sound

91
field created by a rotor in a circular duct. For fixed non-rotating sources one can instead
rewrite the eigenmodes in terms of cos(m ) and sin(m ) .

3.42* Sound from a ducted rotor

We will now discuss the sound generated by a rotor in a circular duct, a case corresponding
to, e.g., an axial fan. An important result from the discussion is the so called Tyler-Sofrin rule
[4] for avoiding that the first harmonic from a rotor is cut-on.

Assume that we have a rotor located in an infinite circular duct. If the rotor has N equal blades
evenly distributed that rotates with an angular speed , the resulting pressure variation in the
rotor plane at t = 0 for a homogeneous inflow can be written as

N 1
p (r , ,0) g (r , 2j / N ) , (3-38)
j 0

where g is the field created by a single blade. This field is periodic in the circumferential
direction and can be expanded in a Fourier series

p (r , ,0) g
m
m (r ) exp(imN ).

Note now that the time variation is caused entirely by the rotation. This implies that the time
dependence of the field can be introduced by substituting: t , resulting in

p (r , , t ) g m (r ) expimN ( t ) . (3-39)
m

Equation (3-39) shows that the pressure field consists of spinning modes of order mN at
multiples, harmonics, of the blade passing frequency29. But to generate radiation of acoustic
power the mN:th harmonic must have a frequency which is higher than the cut-on frequency
for the lowest spinning mode of type mN, i.e.,

m k , m 0 c 0
f m f mc0 .
2 2

29
Note that the plane wave component corresponds to m=0 and a zero frequency so it is of no interest from an
acoustic point of view.

92
The cut-on is calculated assuming a negligible Mach number in the duct (see (3-33)) and
m m N . The lowest root xm0 to J m ( x) 0 has the property: x m 0 m . This implies

k ,m0 ac0 x m0 c 0
M tip a c0 . (3-40)
m m
where the rotor is assumed to have the same radius a as the duct. This result means that to
produce sound the Mach number of the rotor blade tips (Mtip) must be supersonic, i.e., larger
than 1.

In practice, interaction via flow separation between the rotor and fixed objects upstream or
downstream will affect the sound production. The effect of such steady mean flow
disturbances, i.e., deviation from a uniform inflow, is an increase in the sound production at
the blade passing harmonics through generation of supersonic spinning modes. An important
example of such interaction is the rotor-stator interaction that exists in axial flow
compressors, e.g., gas turbines and jet engines. The effect of rotor stator interaction on the
sound production can be represented as

p (r , , t ) g mn (r ) exp(inB ) expimN ( t ) , (3-41)
m n

where B is the number of fixed objects, e.g., guide vanes assumed evenly distributed. In
contrast to the case of expression (3-39) where all modes are spinning with the same angular
frequency , the modes in (3-41) are spinning with different angular frequencies

mN
mn . (3-42)
nB mN

This implies that one can generate supersonically spinning modes even if the rotor has a
subsonic tip speed. The reason is that the nominator in the expression can be much smaller
than the denominator, thereby creating modes spinning much faster than . To avoid that a
harmonic generates acoustic power one must assure that it can only produce subsonically
spinning modes. In order to block the blade passing frequency (m = 1) from an axial fan or a
turbo machine with fixed supports or guide vanes one must choose B 2 N . This condition is
called the Tyler-Sofrin rule. As seen from (3-42), fulfilling this rule will guarantee that for
m = 1 the modes produced will satisfy: 1n , and thus no supersonic modes are
generated.

93
References
1. M.E. Goldstein (1976) AEROACOUSTICS. McGraw-Hill.
2. M.K.Myers (1986) Journal of Sound and Vibration 109, 277-284. An exact energy
corollary for homentropic flow.
3. A.D.Pierce (1981) ACOUSTICS An introduction to its physical principles and
applications. McGraw-Hill.
4. J.M. Tyler and T.G. Sofrin (1962) SAE Transactions 70, 309-332. Axial flow
compressor studies.

94
Problems
1. Analyze the propagation of sound rays in a fluid where the speed of sound varies
linearly with the vertical direction.
a) Show that if there is no mean flow, the ray paths satisfy the equation
d sin
0,
ds c
where s is the path length along the ray.
b) Show that for the case in question rays will travel in circular paths with a radius of
curvature R given by
1 sin
c ,
R c
where c is the speed of sound gradient.

2. Plane wave sound fields in a duct.

a) Show that a harmonic plane wave field is given by:
p ( x1 , t ) p expi (t k x1 ) p expi (t k x1 ) ,
k0
where k .
1 M
b) A duct with flow has the length L and is terminated by two open ends. Show that the
eigenfrequencies for plane waves in the duct are given by

nc 0 (1 M 2 )
fn , n = 0, 1, 2, 3,.
2L
Hint: For low frequencies the boundary condition at an open end is p = 0.

3. The acoustic intensity vector in a medium with flow is defined by

I 0 u U p / 0 U u .

a) Show that in a plane wave field where adiabatic changes of state apply this equation
becomes
p2
I e e(M e) M M (M e) , M U / c0
0 c0
where e is a unit vector and +/- corresponds to propagation in the positive or negative
direction relative e.

95
b) Show that if U Ue , i.e., the flow field is parallel with the direction of propagation
for the plane wave, then
~
p2
I 1 M
2
e,
0 c0
where M = U/c0 is the Mach number and ~ denotes the RMS value.

4. A harmonic plane wave field with amplitudes p and p exists in a duct. Write down
the expression for the acoustic power transported along the duct. If the duct is
terminated by a passive (source free) termination in the positive direction then a
reflection coefficient can be introduced: R p / p . Use the expression for the power
transported along the duct and prove that
1 M
R .
1 M

5. Typical exhaust pipes on cars have a diameter of 0.05 m and on trucks 0.1 m.
Calculate the cut-on for the first higher order mode if the exhaust temperature is taken
to be 400 C and the medium air. The effect of mean flow is neglected.

6. A fluctuating point force, e.g., associated with flow separation from a small body exist
in a duct. Will this generate sound in the plane wave range ?
Hint: The force is assumed to be perpendicular to the flow.

7. y
p 0
y=h

y=0 x
p
0
y

For sound propagation in a water filled channel with depth h the boundary condition at
the water-air interface can be taken as p 0 . If the remaining boundaries are assumed

96
 rigid, the problem with a 2-D water filled channel can be modeled as shown in the
figure. An acoustic mode in this 2-D wave guide can be written as:

p A ( y ) exp i ( t k x x ) ,

where the mode shape satisfies

2
2 k y2 ( y ) 0 , with k y2 ( c ) 2 k x2 . .
y

a) Show that the mode shapes for the problem above is given by:

(2n 1)
n An cos(k y( n ) y )
with k y , n 0,1,2,...
(n)

2h

b) Show from the result in a) that the cut-on frequency for mode n is given by

(2n 1)c
f (n) , n 0,1,2,...
4h

c) The result in a) implies that no plane wave mode exists. Give a simple explanation
why.

d) Given a water filled channel with a depth of 0.1 m, calculate the lowest frequency
needed to excite a propagating sound wave in this channel. (c = 1500 m/s)

8. We will here analyse 1-D sound generation in a duct and investigate the effect of a
mean flow.

U
S x

a) The general 3-D convective wave equation with mass- and force sources included is
given by

1 D02 D m
2 2 2 p 0 fV mU ,
c0 Dt Dt

97
 where f denotes an external force and m a mass flow. Rewrite this equation for the 1-D
case (only plane waves along x) shown in the figure above.

0 Q 0
b) Given a harmonic [ exp(it ) ] monopole point source at the origin: m ( x) ( x) ,
S
where Q is the volume flow. Show that the equation in a) then can be written as

Q
1 M p
2
p
2ikM k 2 p 0 0 i 2U ( x) ,
2

x 2
x S x

where p ( x, t ) p ( x ) exp(it ) , M = U/c0 and k = /c0.

c) To solve the equation in b) we first look for a Greens solution to the equation

1 M xG 2ikM Gx k
2
2
2
2
G ( x) .

To find G start out from a plane wave ansatz around the source and choose the wave
amplitudes so that the boundary conditions at the source are satisfied. Show that this
leads to:

1 exp(ik x), x 0
G ( x) , where k k 1 M and k k 1 M .
2ik exp(ik x), x 0

Hint: G is continuous at the origin with a derivative that has a jump:

G
1 M Gx
2

x
1 .

0 0

d) Show, based on the Greens solution in c), that the solution to the equation in b) is
given by

1 M exp(ik x)
,x0

cQ
1 M
p ( x) 0 0 0 .
2S 1 M exp(ik x)
,x0
1 M

98
e) Show, based on the result in d), that the ratio between the up- and downstream
amplitudes are:
p

1 M 2 ,
p 1 M 2

and calculate the difference in sound pressure level for the case of M = 0.2.

9. Low frequency sound propagation in pipes with yielding walls

u w

In many cases it is not possible to assume that the walls of a pipe are rigid. If the walls
can be described by a locally reacting impedance, it is relatively straightforward to
derive a modified 1-D wave equation.

Assuming no mean flow the linearized 1-D equations of motion and mass
conservation are:
u x
t 0 m
x

,
u p
0 x 0
t x

where we have assumed no external forces but allowed unsteady sources of mass
injection. It can be shown that the wall vibrations in our problem ( u w ) can be

represented as a mass injection term: m 0 u w O / A , where O is the circumference

of the pipe and A the cross-sectional area. The assumption of local reaction further
implies that: p / u w Z w .

99
a) Show that the 1-D wave equation for sound in pipes with yielding walls is:

1 2 p 0 p 2 p
0,
c02 t 2 Z w d h t x 2

where d h A / O is the so called hydraulic diameter and the standard assumption of

adiabatic sound propagation have been used.

b) Assume a harmonic plane wave exp(i (t kx)) and prove that the wave number
must satisfy the dispersion relationship:

k c 0 0 i Z w d h .
2

c) Show that the phase speed is given by:

c0
c ph .

1 0 c02 id h Z w

d) For sufficiently low frequencies most structures are stiffness controlled which implies
that: Z w K w / i , where Kw is the spring constant. Show that for this case the phase
speed is always smaller than c0.

10. Low frequency sound radiation from an open pipe with flow

p
U
S x1

A plane wave is incident on the open end of a pipe. For low frequencies the boundary
condition at the opening is that the pressure fluctuations must be zero. At the opening
the sound wave is reflected but some acoustic power is also transmitted through the
opening.

100
a) Using the definition of acoustic intensity in a medium with mean flow the sound
power transmitted along x1 is given by

1 M 1 M
2 2 2 2
p p
I x1 ,
2 0 c0 2 0 c0

where M is the Mach number. Using the boundary condition at the opening show that
the acoustic power transmitted is

2
p S
Wx1 4M .
2 0 c0

The acoustic power radiated from the opening can be obtained by observing that at the
opening we have two type of sources. First, there is a monopole source with an
unsteady volume flow produced by the acoustic wave: Q uS
(u u ) S . Secondly,

the unsteady part of the momentum transport out of the opening:

0 (U u ) 2 S 0U 2 S 2 0Uu S 0u 2 S , represents a dipole type of source.

b) Show that the power radiated by the monopole contribution (assumed to be compact)
is given by
2
p S k 2 S
Wm .
2 0 c0

0Q (t x / c0 )
Hint: The sound field from a monopole is given by pm .
4 x

c) Show that the power radiated by the dipole contribution (assumed to be compact) is
given by
2
p S 4M 2 k 2 S
Wd .
2 0 c0 3

F(t x / c0 )
Hint: The sound field from a monopole is given by pd and for
4 x
this case the fluctuating force amplitude is given by: F1 2 0UuS
, if we neglect

second order terms.

101
d) The total radiated power is given by Wm Wd . Comparing this with the power
transmitted at the opening gives

Wrad Wm Wd k S 1 4 M / 3
2 2

.
Wx1 Wx1 4M

This result originally derived by Bechert (1980), JSV vol. 70, p. 389-405, shows that for low
frequencies a smaller and smaller fraction of the power transmitted through the opening
actually radiate as sound. Explain why!

11.

M S

x
-L 0

Given a straight duct with plane waves and a constant mean flow. For harmonic waves
the sound field can be written as

p ( x) p exp(ik x) p exp(ik x)

q ( x) p exp(ik x) p exp(ik x) S / 0 c0

where p is sound pressure, q is volume flow, S the cross-sectional area and the
harmonic time dependence is surpressed. The wave numbers are given by

k k /(1 M ) and k k /(1 M ) with k / c0 .

a) Show that the two-port (see the Appendix) on transfer matrix form for a straight duct
is given by

p ( L) cos( z ) iZ sin( z ) p (0)

exp(iMz ) ,
q ( L) (i / Z ) sin( z ) cos( z ) q (0)

102
 kL
where z and Z 0 c0 / S .
1 M 2

b) Calculate the transmission loss (TL) for the straight duct element and comment on the
result.

12. The basic reflective muffler consists of two area jumps after each other.

1 M
2

L
S S

a) Using the 2-port (see the Appendix) for a straight duct, derive the following
expression for the transmission loss for a single chamber muffler:

p 12 1Z Z
2

TL 10 log10
10 log10 1 c
sin 2 kL ,
p 22 4 Z Zc
p 2 0

where Z is the characteristic impedance in the inlet/outlet ducts (1/2), Zc is the

characteristic impedance in the chamber and p / are the pressure amplitudes for
waves propagating to the right/left in the inlet/outlet duct.

Hint: The 2-port for a straight duct with flow is given by

p a cos(kL) iZ sin( kL) p b /c

exp(iMkL) , where k and
q a (i / Z ) sin( kL) cos(kL) q b 1 M 2

Z c / S .

Apply this to the chamber plus the continuity of pressure and volume flow at the
chamber inlet/outlet.

103
b) For a given ratio Z c / Z sketch the behavior of TL as a function of kL. Mark out the
maximas and the minimas.

13. A harmonically oscillating monopole creating the volume flow Q0 is positioned in the
middle of an infinite 2-D duct with the height 2a and rigid walls. Neglecting mean
flow effects assuming small Mach numbers, the sound pressure satisfies the 2-D
Helmholtz equation:

2
xy k 2 p 0 iQ 0 ( x) ( y ), (1)

2 2
where 2xy and k / c0 .
x 2 y 2
To solve for the sound field we make the ansatz

p ( x, y ) An ( x) n ( y ) , (2)
n

where the duct eigenmodes n and eigenvalues kn satisfies the equation: 2 / y 2 kn2 n 0

and the rigid wall boundary condition. After substituting the ansatz (2) into the
Helmholtz equation (1) we get

A( x) k
n
n
2
x ,n An ( x) n ( y ) 0 iQ 0 ( x) ( y ), (3)

where k x2,n k 2 kn2 and denotes d/dx.

104
a) Multiply equation (3) with an arbitrary eigenmode m and integrate over the duct
cross-section, use the fact that the eigenmodes are orthogonal (rigid walls) and
assumed normalized30, and show that

Am ( x) k x2,m Am ( x) 0 i Q 0 ( x) m (0). (4)

b) For each mode m equation (4) represents a 1-D wave problem excited by a point
source at the origin. Show that the solution is given by

0 Q 0 m (0)
Am ( x) exp( ik x ,m x ).
2k x , m

Hint: Make an ansatz for a propagating wave on each side of the source and use the
fact that the field is continuous at the origin and that the derivate has a jump
satisfying:

Am (0 ) Am (0) 0 i Q 0 m (0) ,

which follows from integrating (4) across the origin.

c) Use the results above to write down the complete solution. Show that the plane wave
part (n = 0) of the solution equals

0 cQ 0
p 0 ( x) exp(ik x ).
4a

d) Up to which frequency will the plane wave part in c) represent the farfield in a duct
with the height 2a = 0.1 m? Assume c0 = 340 m/s.

a
1, m n
30

a
m ( y ) n ( y )
0, m n

105
14. Sound generation from fans.

M S
F

0 x
An acoustically compact, i.e., with kd << 1 where d is the duct/fan diameter, axial fan is
mounted in a duct. At the fundamental fan harmonic only plane waves propagate and the fan
can be modeled as a 1D dipole created by the resulting unsteady force from the blades in the
x-direction.

a) Show that for this case the 3D convected wave equation

1 D02 D m
2 2 2 p 0 fV mU ,
c0 Dt Dt

will reduce to

2 p p
(1 M ) 2 2ikM
2
k 2 p ( F / S ) ( x) , (1)
x x x

where p ( x, t ) p ( x) exp(it ) .

D0
Hints: i) i U for harmonic, 1D fields.
Dt x
ii) The force generated by the fan is: f V ( F / S ) exp(it ) ( x) .

b) It can be shown that the Greens function to Eq. (1)

2 G G
(1 M 2 ) 2ikM k 2 G ( x)
x 2
x

1 exp(ik x), x 0
is given by: G ( x) , where k k /(1 M ) and k k /(1 M ) . (2)
2ik exp(ik x), x 0

106
Use this result to show that the solution to Eq. (1) is given by:

F
exp(ik x), x 0
2(1 M ) S
p ( x) (3)
F
2(1 M ) S exp(ik x), x 0

Hint: Take the derivative of Eq. (2) to produce the Greens function for the source / x .

c) Show that the total acoustic power generated by the fan is given by:

2
F
W . (4)
4 0 c0 S

2 2
p (1 M ) 2 p (1 M ) 2
Hint: The 1D acoustic intensity in a duct is given by: I x .
2 0 c0 2 0 c0

d) Apply scaling to the result in c) and show that the power scales as:

0U 4 d 2
W , (5)
c0

where U is the flow speed and the characteristic length has been chosen as the fan/duct
diameter d.

15. Scattering at a sudden area expansion in a duct

2
Assume a low Mach number flow at a sudden duct expansion sometimes referred to (seen
from the flow) as a backward facing step. Neglecting compressibility and assuming low
frequencies and 1D flow we can make a so called quasi-stationary analysis. The governing
equations will be the conservation of mass and momentum. Because the flow separation at
the expansion will create irreversible losses and we cannot use for instance Bernoullis

107
equation. This gives for the fluid state at two cross-sections 1 and 2 assuming an ideal fluid
with no friction along the walls:

0U1S1 0U 2 S2
, (1)
p1 p2 S2 0U 2 S2 0U1 S1
2 2

where the pressure on the wall at the expansion is assumed to equal p1. This assumption is
equivalent with assuming that the flow line leaving pipe 1 is parallel to the wall in pipe 1.
This assumption means that the flow does not go around the sharp edge and is also called a
Kutta condition. For steady flow and low frequencies this assumption is known to be valid.

The equations in (1) are assumed to be valid both for the steady state (index 0) and a state
obtained by adding a low frequency perturbation. This can be written as:

p1 p01 p1

p2 p02 p2
, (2)
U1 U 01 u1
U 2 U 02 u2

where the (..) denotes a time varying perturbation (acoustic field) with a zero time average.

a) Insert Eq. (2) into (1) and simplify keeping only terms which are first order in (..), i.e.,
neglect all squared time-varying terms. Subtract the steady flow part from the resulting
equations and show that this can be written as the following 2-port matrix:

2 0 c0
p1 1 M 1 1 p2
S1
q , (3)
q1 0 1 2

with S1 / S2 , M 1 U1 / c0 and q S u is the fluctuating (acoustic) volume flow and it is

assumed that the temperature is constant (same speed of sound) along the duct.

b) Use the result in an a) to prove that the reflection coefficient for a wave coming from the
upstream side is given by:

p1 1 1 2M 1 1
R11 ,
p1 1 1 2M 1 1

assuming a reflection free termination downstream.

108
Solutions

1a) The refraction of a ray in the general case is given by:

x3 U(x3)

sin
U constant ,
c

which for the case of no flow and if we follow a ray along its path using the distance s
as variable implies

d sin
0, (1)
ds c

where and c now are considered as functions of s.

b) Introducing the radius of curvature R for the path at a certain point we have

Rd ds . (2) ds

Equation (1) implies that d

R

cos d sin dc
2 0. (3)
c ds c ds

Using (2) and noting that: dc / ds dc / dx3 dx3 / ds , where: dx3 / ds cos , equation
(3) can be written31

1 sin
c , (4)
R c

where c is the gradient of c. Note that for a given ray sin / c is constant (eq. 1).
Equation (4) then implies that for a linear velocity profile (c constant) rays will move
in circular paths.

31
Except when cos 0 which corresponds to horizontal rays which obviously will stay horizontal.

109
2a) In a straight duct with rigid walls only plane waves propgating along the axis exist at
sufficiently low frequencies. Assuming a uniform mean flow along the axis (x) the
wave numbers for plane waves along x are:
k0 k
k 0 ,
1 M e 1 M

where M = U/c0 and k 0 / c 0 . A harmonic plane wave field with waves in both the
positive and negative direction is then given by:

p ( x, t ) p expi (t k 0 x /(1 M )) p expi (t k 0 x /(1 M )) , or

p ( x, t ) p expi (t k x) p expi (t k x) , (1)

where k k 0 /(1 M ) and k k 0 /(1 M ).

b) Asssume the duct extends from x = 0 to L . The boundary conditions are:

p (0, t ) 0 and p ( L, t ) 0 . (2)

Using equation (1) gives

p p 0
(3)
p exp(ik L) p exp(ik L) 0

For non-trivial solutions the determinant of this linear system must be zero

exp(ik L) exp(ik L) 0 expi (k k ) L 1 (k k ) L n 2 , (4)

where n is an integer (0, 1, 2, 3,.). This relationship can also be written as:

nc 0 (1 M 2 )
fn .
2L

110
3. I 0 u U p / 0 U u . (1)

p e
a) For a plane wave field we have: u , where e is a unit vector in the direction
0 c0
of propagation. Inserting this into (1) and using the adiabatic relationship gives:

p e p U p p e p2
I 0 2 U e e(M e) M M (M e) .
0 c 0 c0 0 0 c0 0 c0

b) If M Me the result in a) becomes

p2 2 2
I
0 c0 0 c0
0 c0

e e(M e) M M (M e) p e 1 M M M 2 p e 1 M 2 .

4. If the duct extends along an x-axis the acoustic power transported is: W x I x S ,
where the overbar denotes time average and S is the duct cross-sectional area. The
intensity along x is32: I x I I where the intensity in the +/- x-direction can be
found from 3b. This gives the following expression for the power along x for a
harmonic sound field

Wx
S
2 0 c0
p 2

(1 M ) 2 p 2 (1 M ) 2 . (1)

If the positive x-axis points away from the sources in the duct then the following
relationship must be valid: W x 0. (2).

Also when there are no sources in the positive x-direction (passive termination) a
reflection coefficient exits: R p / p . (3)

Using (2) and (3) in equation (1) implies

32
Note we have here assumed that the net intensity is simply the difference between the +/- direction. This
seems reasonable and is, e.g., true for plane wave fields as can be proven from the definition of intensity.

111
 S
p 2

(1 M ) p (1 M ) 0 R
2 2 2 2 p 2

1 M 2 . (4)

2 0 c0 p 2 (1 M ) 2

1 M
A result which leads to: R .
1 M

This last equation is a consequence of power conservation or that the reflected power
stays smaller than the incident power. The equation implies that the amplitude of the
reflected field actually can be larger than the incident when flow is present.

5. For air at 0 C the speed of sound is 331 m/s and for a higher temperature we have:

(273 t )
c 331 (t 400C ) 520 m/s .
273

1.841c
The first cut-on in a circular duct occurs at: f1c , if we neglect M. Inserting
D
D = 0.05 m and 0.1 m gives: 3050 Hz and 6100 Hz. This is far beyond the engine
harmonics so for analysis of exhaust noise a plane wave analysis is sufficient.

6. -

7a) A suitable ansatz to solve the problem is given by: A cos(k y y ) B sin(k y y ) . The

boundary condition at y = 0 imply that B = 0. The boundary condition at y = h implies

that: A cos(k y h) 0 . Non-trivial solutions with A different from zero is obtained

when: k y( n ) h 2n 1 / 2 , n = 0,1,2,... This gives the mode shapes and the

corresponding eigenvalues.
2f ( n )
b) The cut-on frequency for mode n is given by: k (n)
x 0 which imply: k y( n ) .
c
Inserting the result from a) gives the requested result.

c) A plane wave has a constant pressure over the duct cross-section. Here the zero
boundary condition implies that such a mode must be zero over the entire cross-
section. Thus no plane wave mode exists!

112
d) To excite a propagating wave, the frequency of the source must at least correspond to
the cut-on for the first mode. Using the equation from b) with n = 0 gives: 3750 Hz.

D0 D2 2 2 2
8a) U 02 2 2U U 2 2 . Furthermore which
Dt t x Dt t tx x x
implies

1 2 2 2
2
2 p

2
2U U
2
p U m fV , x mU
c0 t 2
tx x x 2
t x x

1 2 2 2
2
2 p f

2
2U U
2
p 2U m V , x
c0 t 2
tx x x 2
t x x

b) Assuming a harmonic time variation we have: i also there is no force source.
t
Inserting the above and assuming a monopole at the origin in the result in a) gives

1 2
2
2 p 0Q 0
2
2i U U
p i 2U ( x) ,
c02 x x 2 x 2 x S

where the factor exp(it ) is omitted. This result can rewritten as

0Q 0
1 M xp 2ikM px k
2
p i 2U ( x) .
2 2
2
S x

c) A point source at the origin will create one wave in positive x-direction and one wave
in negative x-direction. We can therefore assume that

A exp(ik x), x 0
G ( x) .
A exp(ik x), x 0

To determine the wave amplitudes we use the regularity conditions for G at the
origin., G (0) G (0) A A , which means that the amplitudes are equal, and

113

denoted A . Taking the derivative we get: 1 M 2 ik A ik A 1 . Using the
relationships for the wave numbers in the +/- direction this becomes:
2ikA 1 A 1 / 2ik . This proves the result.

d) The Greens function can be used to solve the problem in b) if we multiply the

equation defining G with the operator: i 2U . Since the coefficients in the
x
equation are constants we change the order between this operator and the
x-derivatives, which implies that the solution is:

0Q 0 Q i 2Uik exp(ik x), x 0

p ( x) i 2U G ( x) 0 0 .
S x 2ikS i 2Uik exp(ik x), x 0

Inserting the relationships for the wave numbers in the +/- direction gives the result.

e) From the result in d) it follows that:

(1 M ) /(1 M ), x0
0c0Q 0
p ( x)
2S (1 M ) /(1 M ), x 0

which implies:
p

1 M
2
. For M = 0.2 this gives a level difference of
p 1 M 2
2 2
10 log10 p / p 7 dB.

9. a) Inserting m we get:

u x 0 p O
t 0 x AZ

w
, taking / t of the first equation and / x of
u x p 0
0 t x
the second and subtracting the second from the first, i.e., the standard procedure to
derive the wave equation, gives

114
 2 0 p 2 p
0. Inserting the adiabatic relationship: p c02 ,
t 2
Z w d h t x 2

gives the desired result.

b) Operating with / t on the harmonic field gives: i and with / x : ik .

Substitution this into the equation gives: ( / c0 ) 2 i 0 / Z w d h k 2 0, which
gives the result.

c) Use the definition c ph and the result from b).
k
c0
d) For a stiffness controlled wall we get: c ph c0 .
0c 2
1
K wdh

10a) The boundary condition implies that (x1 = 0 at the opening):

p p 0 p p . (1)

Using Eq.1 in the definition of power:

2 2

I x1
p
2 0 c0
1 M 1 M
2 2

p
2 0 c0
4 M . (2)

0Q (t x / c0 ) i0Q exp(ikx)
b) pm + Harmonic fields p m . (3)
4 x 4 x

u u S . Combining this with relationship

From the text we have: Q uS

between pressure and velocity in a plane wave we get:

S 2 p S
Q u u S p p Eq.1 . (4)
0 c0 0 c0

115
 To calculate the power we look at the far-field and integrate the plane wave intensity
over a large sphere. Since the field has constant amplitude we can simply multiply
with the area of the sphere:

2 2 2
p m 2 02 4S 2 p p S k 2 S
Wm 4 x Eqs.3&4
2
4 x 2
. (5)
2 0 c0 2 16 2 x 2 0 c0 2 0 c0
3

c)
F(t x / c0 )
pd + Harmonic fields and far-field
4 x , (6)

i F1 cos exp(ikx)
p d
4 xc0

where F1 2 0USu 2 0UQ . (7)

The power becomes:

2 2

p d 4 2 02U 2 4S 2 p cos 2

Wd 2 x sin d Eqs.4&6&7
2
2 x 2 sin d
0
2 0 c0 0
2 16 x c0 ( 0 c0 )
2 2 2 3

2 2
p S 2k 2 SM 2 p S 4 M 2 k 2 S
0
2
cos sin d . (8)
2 0 c0 2 0 c0 3

2 / 3

d) Combining Eqs. (2), (5) and (8) gives the desired result. The explanation that the
power transmitted through the opening does not reach the far-field is related to forced
vortex separation at the duct outlet. Here the sound field will be converted to vorticity
that eventually will add to the turbulence outside the opening and then gradually
dissipate (i.e. become heat).

11a) From the given expression for the sound field we can write the pressure and volume
flow at x = 0

p (0) p p p p (0) Zq (0) / 2

(1). From this we can solve: . (2)
q (0) p p / Z p p (0) Zq (0) / 2

116
 The pressure and volume flow at x = -L is given by:

p ( L) p exp(ik L) p exp(ik L)
(3). Inserting eq. (2) into (3) gives
q ( L) p exp(ik L) p exp(ik L) / Z

p ( L) 12 p (0) Zq (0) exp(ik L) 12 p (0) Zq (0) exp(ik L)

q ( L) 2 p (0) Zq (0) exp(ik L) 2 p (0) Zq (0) exp(ik L) / Z
1 1

p ( L) 12 p (0)exp(ik L) exp(ik L) 12 Zq (0)exp(ik L) exp(ik L)

. (4)
q ( L) 2 p (0)exp(ik L) exp(ik L) / Z 2 q (0)exp(ik L) exp(ik L)
1 1

If we use Eulers eq. exp(ix) cos x i sin x and the wave number relationships given
in the problem we get

exp(ik L) exp(ik L) 2 exp(iMz ) cos( zL)

, (5)
exp(ik L) exp(ik L) 2i exp(iMz ) sin( zL)

kL
where z . Inserting eq. (5) in (4) gives the sought result.
1 M 2

b) From the 4 matrix elements in a) we get directly

2
1
TL 10 log10 exp(iMz )cos( zL) i sin( zL) i sin( zL) cos( zL)
2
2
1
10 log10 exp(iMz )2 exp(izL) 10 log10 1 0
2

since M and z are real numbers. The result seems logical since there should be no
damping of sound in a straight duct ! At least if we neglect the effect of viscosity and
thermal conductivity that exists close to the walls which is the case here.

117
12a) For a straight duct (the chamber) we have:

p a cos(kL) iZ c sin(kL) p b
(1) exp(iMkL) .
qa (i / Z c ) sin(kL) cos(kL) qb

At the inlet and outlet of the chamber continuity of pressure and volume flow applies.
This implies:

p a p1 p b p 2
(2) and , where the cross-sections 1 and 2 are assumed to be
qa q1 qb q2

just before and after the inlet/outlet of the chamber. We also know that for a plane
wave:

p p p
(3) . Using eqs. (1-3) and the fact that we have a reflection free
Zq p p
termination at the outlet ( p 2 0) when we calculate the transmission loss, we obtain:

p1 p1 cos(kL) iZ c sin(kL) p 2
exp(iMkL) .
( p1 p1 ) / Z (i / Z c ) sin(kL) cos(kL) p 2 / Z

From this we can solve the ratio:

p1 Zc Z 1Z Z
2

exp(iMkL) cos(kL) i
sin(kL) TL 10 log10 1
c
sin 2 kL .
p 2 2Z 2Z c 4 Z Zc

c) When sin(kL) 0 we have minima with TL = 0. This corresponds to kL 0, , 2 ,....

When sin(kL) 1 we have maxima. This corresponds to kL / 2,3 / 2,....

118
13a)
a a

m ( y) An( x) k x,n An ( x) n ( y)dy 0 iQ 0 m ( y) ( x) ( y)dy

2

a n a
a a

n An( x) k An ( x) m ( y) n ( y)dy 0 iQ 0 ( x) m ( y) ( y)dy

2
x ,n

a

a
mn m (0)

Am ( x) k x2,m Am ( x) 0 i Q 0 ( x) m (0).

b) Ansatz to solve the equation in a): Am ( x) Bm exp(ik x ,m x ), which satisfies the

condition of continuity at the origin. To fix the amplitude, we use the jump condition:

Q (0)
ik x ,m Bm ik x ,m Bm 0 i Q 0 m (0) Bm 0 0 m .
2k x , m

c) The complete solution becomes: p ( x, y ) 0Q0 n (0) n ( y ) exp(ik x ,n x ) .
n 2k x , n

14. -

15.
a) Linearizing the given equations and subtracting the steady state case gives:

0U1S1 0U 2 S2 q1 q2
.
p1 p2 S2 0U 2 S2 0U1 S1 p1 p2 S2 2 0U 2 u2 S2 2 0U1 u1S1
2 2

Where the second eq. can be rewritten as:

p1 p2 S2 2 0U 2 u2 S2 2 0U1 u1S1 p1 p2 S2 2 0U1 S1 / S2 q2 2 0U1 q2

2 0 c0
p1 p2 S2 2 0c0 M1 1 q2 p1 p2 M 1 1 q2 .
S1

Where the underlined results lead to the sought 2-port relationship.

p 1 Z p2 2 0 c0
b) The given matrix can be written as: 1 , Z M 1 1 .
q1 0 1 q2 S1

119
 p p1 p1
At the inlet side we have: 1 ,
q1 ( p1 p1 ) S1 / 0 c0

and on the outlet side with a reflection free termination:

p2 p2
. Inserting these results into the 2-port eq. we obtain:
q2 p2 S 2 / 0 c0

p1 p1 p2 p2 p1 p1 p2 p2
, where 2 M 1 1 .
( p1 p1 ) S1 / 0 c0 p2 S2 / 0 c0 p1 p1 p2 /

Taking the ratio of the two last eqs. gives:

p1 p1 1 R 1 (1 )
(1 ) (1 ) R , which is the sought result.
p1 p1 1 R 1 (1 )

120
 Chapter Four

The effect of source motion

In this chapter we will discuss the effect of source motion on the radiated sound. It turns out
that motion will lead to directivity of the sound field and also to a frequency shift known as
the Doppler effect. In addition, the results show that acceleration of a steady source, e.g., a
force, can produce sound. The results can also, via a simple change of co-ordinate systems, be
applied to write down the solution for a point source in a steady mean flow. Finally, the so
called Ffowcs Williams&Hawkings equation for sound produced by arbitrary moving bodies
is discussed. This equation is the basis for example for propeller noise calculations.

4.1 A moving point source

e r V(te)

xs(te) r(te)

Figure 4.1 A moving point source in a homogeneous fluid at rest. The observer at x is assumed fixed. The
emission time te corresponds to the time when sound reaching the observer at x at time t to was emitted.

121
Assume a homogeneous fluid at rest containing a moving point source. The resulting sound
field will then satisfy

1 2
2 2 2 p (x, t ) S (t ) (x x s (t )) , (4-1)
c t
0
where S(t) is the source strength and xs(t) is the trajectory of the source. To solve (4-1) we use
equation (2-5) with the free field Greens function (2-6)

s (y , ) (t r / c0 )
p (x, t )
V
4r
ddV y , where r x y .

Inserting the source term from (4-1) in this equation gives

S ( ) (y x s ( )) (t x y / c0 )
p (x, t )
V
4 x y
ddV y .

If we change the order of integration and perform the volume integral first, the result is

S ( ) (t r ( ) / c 0 )
p (x, t )

4r ( )
d , (4-2)

where r (t ) x x s (t ) . The integral in (4-2) can be solved using the following formula for

delta functions

f ( n )
f ( ) ( g ( ))d g (
n n)
, (4-3)

where the summation is over all the zeros of g, and g is the derivative of g. From (4-2) it
follows that: g ( ) t r ( ) / c0 and: g ( ) 1 e r (V ) / c0 1 M r , where

M r M e r M cos is the projection of the source Mach number (M = V/c0) in the

direction towards the observer, and e r r / r.

The zeros of g define the emission times te from the source that contribute to the sound field
received at point x at time t, and are defined by

122
 c0 (t te ) r (te ) . (4-4)
Using the last results, the following expression for the sound field is obtained

S (t e )
p (x, t ) , (4-5)
te 4r (t e ) 1 M r (t e )

where the summation is over all emission times te .

4.11 Doppler shift

Assume a harmonic point source S (t ) S 0 exp(i 0 t ) , the sound field will then be given by

S 0 exp(i 0 t e )
p (x, t ) . (4-6)
te 4r (t e ) 1 M r (t e )

At the observer the time variation33 of the field is given by

p (x, t ) exp(i 0 t e ), (4-7)

and the instantaneous (angular) frequency of this signal can be defined as: / t , where
0 t e (x, t ). Following this, we have that 0 t e / t , where using (4-4):

t t t e 1
c0 1 e e r (t e ) V(t e ) e . (4-8)
t t t 1 M r (t e )
Thus, the instantaneous frequency is then given by

0
, (4-9)
1 M r (te )

where M r (t e ) M (t e ) cos (t e ) . Equation (4-9) is the so called Doppler shift formula which
describes the frequency shift for a moving source when measured by a fixed observer.

Example 4-1: A point source moves with a constant, subsonic speed along a straight line.
What is the Doppler shift for an observer standing close to the line ?

M
x

33
Note that there is also a time variation associated with the other terms in (4-6) but this is assumed to be much
slower and so it is neglected.

123
For large negative x-values 0 and for large positive x-values which implies that
the frequency varies between 0 /(1 M ) and 0 /(1 M ) , where M is the Mach number.
This means that the observer will experience a frequency transition when the source passes
by. This drop in the pitch can for instance be observed when an ambulance passes !

4.12 Source moving along a straight path

We will now study a source moving with constant speed along a straight path. To simplify the
discussion, we will assume a 1-D case but the results we find can be generalized to the 3-D
case. To investigate the emission times for the 1-D case it is useful to construct a space-time
diagram. Assume that the source moves along an x-axis and that it passes x = 0 at t = 0, then
the position at time t is: x Mc0 t . This trajectory has been plotted in Figure 4.2 for M < 1
and M > 1 together with the trajectories for sound waves in the positive and negative
x-direction.

x M>1

M<1

co t

Figure 4.2 Space-time diagram x-cot. Note the same scale is assumed on both axis so that the trajectories for the
sound waves (- - - - -) are always inclined at +/- 45 degrees.

Suppose that we have an observer at an arbitrary position (say) x = 0. As can be seen from the
figure for M < 1 only one unique emission time exists for any time t. For M > 1 there exist no
emission times before the source passes the observer (t < 0), and then for t > 0 there are two
solutions. One of the solutions corresponds to waves emitted in the positive x-direction before
the source reached the observer. These waves will arrive backwards, i.e., the waves emitted

124
last will arrive first. The other solution will correspond to waves emitted in the negative
x-direction after the source has passed the observer. Note that the two waves reaching the
observer after the source passage for M > 1 will have different Doppler shifts, i.e., two
different frequencies will be heard simultaneously from a harmonic source.

Example 4-2: From the Doppler shift, (4-9), it follows that the frequency for the wave
propagating in the positive x-direction is: 0 /(1 M ) . For M > 1 this frequency becomes
negative! The interpretation of this is that time is reversed34 or that the wave signal is played
backwards. For instance if we assume a loudspeaker playing a Mozart concerto moving
with M = 2 then: 0 /(1 M ) 0 , implying that the music is played backwards but at
perfect pitch !

For the 3-D case, the same conclusions are reached as for the 1-D case. The main difference is
that for M > 1, sound is only heard within a conical region called the Mach cone, see Figure
4.3.
c0
p=0

p 0

Mc0

Figure 4.3 Mach cone for a source moving with a constant supersonic speed (M > 1). On the Mach cone the
relationship: 1 M cos e 0 is valid implying an infinite acoustic pressure (a sonic boom). The angle of

emission and the angle of the cone, ,are complementary, i.e., / 2 e , thus: sin 1/ M , M >1.

4.13 The effect of a mean flow and the general Doppler shift formula
We will now assume that the homogeneous fluid surrounding a moving source (moving with
the velocity V) also has a constant mean flow velocity U relative to the observer. The
resulting sound field can be calculated from (4-5) if we move to a fluid fixed co-ordinate

34
Because exp(it ) exp(i ( )(t )) .

125
system (denoted x), where there is no mean flow relative to the observer. The
transformation35 between the x and xsystems is given by

x x x 0 Ut , (4-10)
where x is the system fixed relative to the observer, see Figure 4.1. From equation (4-5) it
follows that
S (t e )
p (x , t ) , where c0 (t t e ) x x s (t e ) .
te 4r (t e ) 1 M r (t e )
r ( te )

Transforming this to the x system, noting that V V U , gives

S (t e )
p (x, t ) , (4-11)
te
4r (t e ) 1 M r M f

r t
e

where M f U / c0 and the emission times are obtained from

c0 (t t e ) r (t e ) U(t t e ) , (4-12)

where r (te ) x x s (te ) . Equations (4-11) and (4-12) define the general solution for a point
source moving in a mean flow field. A special case is a fixed source embedded in a steady
mean flow, when (4-11) represents the point source solution to the convective wave equation,
see Chapter 3.

It is possible to assume in equation (4-12) (just as in (4-5)) that the observer is moving, i.e.,
that x = x(t). A general Doppler shift formula for this case is obtained from: 0 t e / t ,

t e t t
where from (4-12): c0 (1 ) e r V o (t ) V (t e ) e U1 e
t t t

t e 1 M ro M rf
. (4-13)
t 1 M r ,e M rf

Here V o dx / dt and M o V o / c 0 is the Mach number of the observer. All Mach numbers

are projected in the direction e r r / r and the sub-script e denotes a quantity evaluated at
the emission time te. This gives the general Doppler shift formula

35
Known as a Galilei transformation in classical mechanics.

126
 1 M ro M rf
0 . (4-14)
1 M r ,e M rf

Example 4-3: As seen from (4-14) it is the relative motion between a source and an observer
that creates a Doppler shift, because if M ro M r ,e there is no shift. Also note that only a

steady mean flow does not change the frequency.

4.2 Application to a monopole, dipole and quadrupole source

We will now use the result from the previous section and study the fields created by the three
basic source types in Lighthills theory.

Let us first consider sound created by a moving unsteady volume flow source (a monopole).
This problem is described by the equation

1 2
2 2 2 p (x, t ) 0 Q(t ) (x x s (t ) . (4-15)
c t t
0

The free field solution to this equation is obtained by putting: S 0 Q (x x s (t )) in (4-5)

and then taking the time derivative

0 Q(t e )
p (x, t ) . (4-16)
te t 4r (t e ) 1 M r (t e )

Using the relation: / t te / t / te , this can be rewritten as a derivative with respect to

te that can then be evaluated as the derivative of a product. Evaluating this and using equation
(4-8) gives

0Q e 0Qe e r M
0c0Qe M e cos e M e
p(x, t ) e
, (4-17)
4 re 1 M r ,e 4 re 1 M r ,e 4 re2 1 M r ,e
2 3 3

where the dot denotes a time derivative, M < 1 is assumed and only one emission time.
Equation (4-17) shows that sound is produced by the rate of change of the volume flow

127
(Q ) and from acceleration of the source motion ( M
e ) towards the observer. The last term is
r

a near field contribution that does not contribute to the radiated sound power.

The sound created by a moving unsteady force (a dipole) is described by the equation

1 2
2 2 2 p (x, t ) F(t ) (x x s (t ) . (4-18)
c t
0

The free field solution to this equation is obtained by putting: S Fi (x x s (t )) in (4-5) and
then taking the divergence

Fi (t e )
p (x, t ) . (4-19)
te xi 4r (t e ) 1 M r (t e )

Using the relation: / xi te / xi / te , this can be rewritten as a derivative with respect to

te which can then be evaluated as the derivative of a product. Evaluating this, using equation

(4-4) to calculate te / xi , gives

p(x, t )
F e
r e

F er e M er e F er e (1 M e2 ) F M e (1 M r ,e )
. (4-20)
4 c0 re (1 M r ,e ) 2 4 c0 re (1 M r ,e )3 4 re2 (1 M r ,e )3

Equation (4-20) shows that sound is produced by the rate of change of the force ( Fr ) and from
e ) towards the observer.
acceleration of the source motion ( M r

The above procedure can also be applied to a moving point quadrupole source:
S Tij 2 (x x s (t )) / xi x j . For this case we will only write down the far field contribution

for a source moving at constant speed (M < 1)

T e e
ij r ,i r , j e
p (x, t ) . (4-21)
4c02 re (1 M r ,e ) 3

Comparing equations (4-17), (4-20) and (4-21) we see that the far field intensity ( p 2 ) for a
source moving at constant speed varies as

128
 p 2 (monopole) 1 - M e cos e 4
2
p (dipole) 1 - M e cos e
4
, (4-22)
2
p (quadrupole) 1 - M e cos e
6

which shows that source motion for Mach numbers approaching 1 can lead to strong
directivity effects.

4.3* The Ffowcs Williams&Hawkings equation [1]

We will now derive an equation that is a generalization of Curles equation and applies to a
body performing an arbitrary motion through a turbulent flow. The derivation is based on the
idea of creating an acoustic analogy, i.e., trying to rewrite the exact equations of fluid
mechanics in a form that serves our purposes, in this case a representation of the source
distribution on a moving surface. As we know from Chapter 2, there is no unique solution to
this problem, instead it is important to create formulations that enable us to link the source
field to known geometric, kinematic and flow properties.

To describe a body performing an arbitrary motion we introduce a function f(x,t) defined by

0, x V

f (x, t ) 0, x S , (4-23)
0, x V

where V(t) denotes the body, S(t) its surface and the outward normal n is f / f f 0
. Using

the Heaviside function36, we can now elegantly write the fundamental equation of fluid
mechanics (see Chap. 1) so that they are valid everywhere

H ( f ) u j 0
t x j
, (4-24)

H ( f ) t u i x u i u j p ij 0
j

1, x 0
36
H (x) . The value at x=0 is not always specified but can be taken as .
0, x 0

129
where p p 0 p , 0 , the viscous stress is neglected and there are no external
sources. The next step is to rewrite (4-24) so that a wave equation can be derived. We first

observe that for a point on the moving surface: DS H ( f ) 0

Dt
H
V H 0 , (4-25)
t
where V(xS,t) is the surface velocity of the body. We also note that equation (4-25) is trivially
valid for points not on S since the derivatives of H then are zero. Using (4-25) we can then
rewrite equation (4-24) as

H H
t H x u j H 0V j x u j V j x
j j j
. (4-26)
u H u u H p H u u V H p H
t i
x j
i j ij i j j
x j xi

Performing the operation (4 26)1 / t (4 26) 2 / xi gives

2 H 2 pH ui u j H
2
H H
0V j (u j V j ) ui (u j V j ) p ij .
t 2
xi xi xi x j t x j xi x j

A wave equation is obtained from this result by assuming adiabatic changes of state and
putting p / c02 . This equation is often simplified by neglecting the first (quadrupole)

source term and by assuming a rigid body ( u n V n ). The resulting reduced form of the
Ffowcs Williams&Hawkings equation is

1 2 2
2 2 p 0 V n f ( f ) pn f ( f ) , (4-27)
c0 t t

where p p H , the relation H ( f )f has been used and f n f is evaluated on S.

Equation (4-27) is well suited for propeller or fan noise calculations. The first source term in
(4-27) corresponds to sound production by fluid displaced by the moving body. This term is
called thickness noise and it can be calculated once the geometry and motion is prescribed.
The second term is related to the pressure distribution over the moving surface and is called

130
loading or lift noise. This term can be obtained from aerodynamic models of propellers or
fans.

The solution to (4-27) can be obtained in a similar way as the moving point source case
treated earlier. The result is (for the subsonic case)

V (y , t e ) n e p (y, t e )n e
p (x, t ) 0 dS y 0 dS y , (4-28)
t Se 4re (1 M r ,e ) Se
4re (1 M r ,e )

where S e (x, t ) is the surface formed by the emission points (will differ from S !),

re x y (t e ) , c0 (t t e ) re (t e ) and the sub-script e denotes a quantity evaluated at te.

Care must be taken when evaluating the surface integrals in (4-28) since the emission time te
varies over the source region which is moving, i.e., changing with time. This implies that the
shape of S e can differ significantly from the surface of the moving body S. Assuming a
compact body, i.e., with a dimension much smaller than the wavelength, it is possible to
derive simplified versions of (4-28). One such version suggested by Farassat [2] is obtained
by using the relation: 0 V n f ( f ) 0 (1 H ( f )) / t , and noting that (1 H ( f )) 1

inside the body and 0 on the outside. Using this relation the first source term (the thickness
noise) in (4-27) becomes: 0 V n f ( f ) t 2 0 (1 H ( f )) / t 2 . This implies that

the contribution to the sound field is

2 0

pthickness 2
t 4 r (1 M
Ve e r ,e )
dVy ,

where in the compact limit Ve V . The second integral in (4-28) will in this limit approach

the total force on the body: F (t ) p ndS y . Therefore, in the compact limit the Ffowcs
S

Williams&Hawkings equation for a rigid body will reduce to

2 0V

Fe


p (x, t ) 2
t 4re (1 M r ,e ) 4r (1 M ) . (4-29)
e r ,e

131
References

1. J.E. Ffowcs Williams and D.L. Hawkings (1969) Phil. Trans. Royal Society.
A264, 321-342. Sound generated by turbulence and surfaces in arbitrary motion.

2. F. Farassat (1981) AIAA Journal 19(9), 1122-1130. Linear acoustic formulas for
calculation of rotating blade noise.

132
Problems
1. Discuss sound generation by a moving source and compare with a non-moving source.

2. Derive a general formula for the Doppler shift assuming a 1-D situation as shown in
the figure below, where we have a constant mean flow speed and both the source (s)
and the observer (o) is moving with constant speeds.

Vf
Vs
x

Vo

3.* Investigate the thickness noise term in Farassats formula (eq. (4-29)) and derive its
form in the far field.

4. A simple model for a propeller consists of assuming N equal blades, evenly distributed
around a circle, and assuming an equal rotating (steady) force on each blade. The force
corresponds to the lift created by each blade seen as an airfoil and can be obtained
from aerodynamic models. Using Farassats formula (4-29), neglecting the thickness
noise, derive a general expression for the sound field assuming free field conditions.
Calculate the maximum sound pressure at 100 m for the following data: Total lift
force 104 N, number of blades 2, rotational frequency 60 Hz, radius 0.6 m.

5.* Investigate the sound field from a point dipole at the origin embedded in a
homogeneous fluid with a constant flow velocity along the positive x1-direction.

133
134
Solutions
1. -

2. The wavelength created by the source will satisfy:

c0 V f T VsT , (1)

where T is the period for the oscillation frequency of the moving source. The time it
takes this wavelength to pass a moving observer T is determined by:

T . (2)
c
0 V f Vo

The frequency experienced by the observer is given by: f 1/ T inserting (1) in (2)
gives:

f
c 0 V f Vo
f
1 M f Mo
f.
c 0 V f Vs 1 M f M s

3. -

4. In the proposed model the pressure distribution on each blade is summed to a resulting
force F0. All these N point forces are assumed equal and evenly distributed around a
circle with radius a.

x2

x1
F r
x
xs
ex
x3

The position vector for the n:th force (source) is:

x s ,n a cos(t n ), sin(t n ),0 ae n ,

135
where = angular frequency for the rotation, n 2n , n = 0, 1, 2,...,N-1 and N is the
N
number blades. According to the lectures the far field from a dipole point source is
given by:

F e e r ,e e )
(Fe e r ,e )(M
p( x, t ) e r ,e
.
4c0 re (1 M r ,e ) 2 4c0 re (1 M r ,e )3

Here with constant blade forces the only contribution will come from the acceleration
term. The total sound field can therefore formally be written as:

e )
N 1 ( F e )( M
p ( x, t ) .
rn n rn

n 0 4c0 rn (1 M rn )
3
te , n

To evaluate this expression we investigate the geometric and kinematic parameters in

the far field:

rn xe x ae n x 1 2( a / x)e x e n a 2 / x 2 1/ 2
x1 O (1 / x ) ,

e r ,n rn / rn xe x ae n / xe x ae n (e x ae n / x)1 O (1 / x)

e x O(1 / x).

rn x
For large x we obtain:
e r , n e x

For an ideal fan/propeller the Force is entirely perpendicular to the rotor plane, i.e.,
there is no drag component. This implies:

Fr ,n F e r ,n (0, 0, F ) e x Fex ,3 .

The velocity for the n:th force is:

Vn x s ,n a sin(te n ), cos(te n ),0,

136
The acceleration then becomes:

a 2 cos(t ), sin(t ),0 a 2e ,

Vn e n e n n

which gives:

M r ,n V
e / c V
n r ,n
e / c (a 2 / c )e e
n x n x

In order to calculate the projection of the force, velocity and acceleration on the unit
x-vector we introduce spherical co-ordinates with respect to x3. This implies:
e x cos , sin cos , sin sin .

This gives for the expressions above:

Fr , n Fex ,3 F cos , where F = F0/N.

M r , n (a / c ) sin(te , n n ), cos(te , n n ),0 e x

( a / c) sin cos sin(te , n n ) sin cos(te , n n )
(a / c) sin sin(te , n n )

and

M r , n ( a 2 / c)e n e x
( a 2 / c) sin cos(te , n n ) cos sin(te , n n ) sin
( a 2 / c) sin cos(te , n n )

We can now introduce this into the sound field expression but before we do that we
look at the equation for the emission time: c (t te , n ) rn , where for large x :

rn x ae x e n which implies
t e,n t x / c (a / c) sin cos(t e,n n )

This equation must be solved by iteration for each blade (it converges nicely as long as
a / c 1 ).

Summarizing the results above the sound field for x >> a for a sub-sonic propeller
(neglecting thickness noise) can be written as:

137
 F0 ( a 2 / c ) N 1 cos sin cos(t e , n n )
p ( x, t ) ,
4cxN n 0 (1 ( a / c ) sin sin(t e , n n ))
3

where F0 is the total lift force. Obviously the field must have a rotational symmetry
around the x1 axis. As seen from the formula the pressure is zero at 0 and /2
and the magnitude of the pressure is symmetric around / 2.

Example: According to the text F0 = 104 N, = 2*60 Hz, a = 0.60 m, x =100 m,

= /4 and N = 2 (c = 340 m/s). This gives:

where it can be noted that the dominating harmonics are 120, 240, 360, Hz which
corresponds to multiples of the blade passing frequency.

If we change to four blades and keep the rest of the data fixed then.

138
 This means that the fundamental drops with 5-6 dB and shifts to 240 Hz. But to
evaluate the change in annoyance the fact that the ear is more sensitive for higher
frequencies must be remembered!

5. -

139
140
 Chapter Five
Self-sustained oscillations and whistles

In Lighthills theory it is assumed that the source term is known and unaffected by the
acoustic field. This is a good approximation when the source region is embedded in free space
and with no boundaries creating acoustic reflections and modes. For situations where acoustic
waves can be reflected back to the source region, a modulation of the source process is
possible, which can lead to a positive (unstable) feedback loop. This phenomenon, referred to
as a self-sustained oscillator, normally occurs at certain frequencies and results in the creation
of distinct tones, whistling. Self-sustained oscillators are used in many musical instruments
to produce sound, e.g., wind and bow instruments. In engineering practice a knowledge about
whistles are important as they can produce very high sound levels and result in high vibration
levels and risk for mechanical failure. In this chapter basic types of whistles will be discussed
and examples on how to predict and eliminate whistling will be given.

Flow Sound

Figure 5.1 A whistle is a self-sustained oscillator in a fluid, created by a positive (unstable) feedback loop,
where sound generated by a flow process is reflected back to the source region and modulates (affects) the
source process.

141
5.1 Fluid driven whistles
To create a fluid driven whistle an unstable flow process that can couple to an acoustic field is
necessary. An important example of such a process is periodic flow separation, either around
a compact body (type 1) or between an upstream and a downstream edge (type 2). In both
these cases there will exist natural oscillations, called Strouhal tones, in the flow. A whistle is
created when a Strouhal tone starts to interact with an available acoustic (or structural) mode
forming a positive feedback loop. If the feedback loop is created, the amplitude of the
oscillation will grow until it is limited by losses or non-linear effects. The energy stored in the
vibration is taken from the mean flow. Thus, in many fluid machine systems, e.g., gas
turbines, where the available energy is large there is a potential for very strong whistles.

Periodic flow
+ separation

Acoustic/
structural mode

Figure 5.2 A fluid driven whistles created by a positive feedback loop involving a periodic flow separation
phenomenon coupled to an acoustic/structural mode.

A necessary condition for a whistle is that the frequency f St of the Strouhal tone matches (or

is close to37) the resonance frequency f Mech of a mechanical mode (acoustic or structural), i.e.,

f St f Mech . (5-1)

The condition is not sufficient since even if the frequencies match, it is not certain that a
positive feedback loop will be created. Still very few complete models are available for the
prediction of whistles and in practice one is therefore limited to use equation (5-1), to try to
avoid matching between flow separation and known modes.

We will now discuss whistles created by flow separation around a body, which we will call a
type 1 whistle. From fluid mechanics [1,2] it is known that a cylinder (a rod) placed

37
Experience shows that when a periodic flow separation phenomenon starts to interact with a mode it can shift
or adjust its frequency to match the mode. Therefore an exact matching is not required to start a whistle.

142
perpendicular to a flow will produce periodic vortex shedding in a large Reynolds number
(Re) range, with a Strouhal frequency given by F

U d
f St , 40 Re 3x10 5 U (5-2)
d

where Re Ud , is the kinematic viscosity, and depends on the cross-section of the

body (close to 0.2 for a circular cross-section). The periodic flow separation will lead to an
oscillating lift force F on the rod, the force being mainly perpendicular to the mean flow. This
force can excite elastic waves in the rod, e.g., bending wave modes. The force F acts on the
surrounding fluid and represents an acoustical dipole source. This source can excite acoustic
modes if these modes have a velocity component parallel with F so that power can be fed into
the mode. Normally the flow separation occurring along a rod is not synchronized, i.e., it does
not occur simultaneously (in phase) along the direction perpendicular to the flow. However,
when the separation couples to a structural or acoustic mode, this will synchronize the flow
separation process. This means that the conversion of mean flow energy into a growing
oscillation becomes more efficient.

5.11 Type 1 whistles

To block a whistle of type 1, described above, one can either disturb the flow separation
process, or reduce or eliminate the coupling to the mechanical mode. One common procedure
to disturb the flow is to introduce irregularities on the surface of the body that destroy the
periodic flow separation. The best way to reduce the coupling to the mechanical mode is to
increase the damping in the mechanical system. If the flow speed is fixed so that the Strouhal
frequency does not vary, one can eliminate the coupling by changing the frequency of the
mode.

Example 5-1: A car radio antenna can be seen as a cylindrical rod. The periodic flow
separation around the antenna can produce a type 1 whistle if it couples to the first bending
wave resonance. The necessary condition for this is: f St f 1,bending , where the frequency for

the bending wave resonance is calculated assuming a free-clamped beam boundary

condition.

143
Example 5-2: In a common type of heat exchanger arrays of parallel tubes are surrounded by
a perpendicular mean flow. A type 1 whistle can be produced in two ways. The Strouhal
frequency may coincides with a bending wave resonance. The other possibility is that the
Strouhal frequency coincides with the cut-on for an acoustic cross-mode normally the first:
f St f1,cut on . In a rectangular duct with height h this implies: f St c 2h , if we neglect mean

flow effects on the cut-on frequency.

5.12 Type 2 whistles

Turning now to whistles created by periodic flow separation between an upstream and a
downstream edge, we will call this a type 2 whistle. For this type there is not a single Strouhal
frequency, instead a number of oscillation frequencies can exist. The basic flow configuration
is an unstable shear flow, for instance across an opening in a wall, see Figure 5.3. At the
upstream edge, a flow disturbance (vortex) is shed and then travels downstream with the mean
flow. As the flow is unstable the vortex will grow when it moves downstream. When the
vortex reaches the downstream edge it will create a pressure pulse which is propagated
upstream. This pulse will trig the shedding of a new vortex and the process is repeated. This
simple model for the periodic flow separation process is based on the work by Rossiter [3,4].

1 3
2

Figure 5.3 Periodic flow separation of type 2. The process is assumed to consist of four parts: 1) Separation
of a vortex from the upstream edge; 2) The vortex travels downstream and grows; 3) The vortex reaches the
downstream edge and creates a pressure pulse; 4) The pulse propagates upstream to start (trig) a new vortex
(grey arrow).

Following Rossiter one can calculate the frequencies of the periodic flow separation process.
We start by estimating the travel time T around one loop of the process. This is given by

d d
T r , (5-3)
U c c0

144
where U c is the convection speed of the vortex, it is assumed that the pressure pulse

propagates upstream as an acoustic wave (unaffected by flow) and r represents time delays
for the creation of a pressure pulse and a new vortex, respectively. The travel time must
correspond to a number of periods of the process which implies: St ,nT n 2 , where

n = 1, 2, 3, Using (5-3) this results in

n r / 2 U
f St ,n , (5-4)
1/ M d

where U is the mean flow speed (in the main flow), M = U/c0 is the Mach number,
U c / U , and r is a phase angle associated with the time delays. Equation (5-4) is referred
to as the Rossiter formula. Tables summarizing the , r parameters for different cases can
be found in Blake [1]. For the case of a cavity in a wall the constants are: 0.4 and
r / 2 .

Although (5-4) predicts an infinite number of Strouhal frequencies, only a finite number is
observed in practice, where typically n = 4-5 is the largest. The strongest oscillations can be
expected for the lowest Strouhal frequencies and the n = 1 case is the most important in
practice.

Concerning coupling to mechanical modes, the most important in practice are the acoustic
modes in a closed cavity or a side-branch duct. An important example of the closed cavity
problem is a Helmholtz resonator.

The resonator neck

Figure 5.4 A whistling Helmholtz resonator.

145
The well known tone produced by an empty bottle excited by blowing with the mouth is an
example of such a whistle. For whistling Helmholtz resonators, typical amplitudes u of the
oscillation (in the resonator neck) is of the order of: u / U O(10 1 ) [5].

To block a whistle of type 2, the same alternatives as discussed for type 1 applies, i.e., disturb
the flow separation process or eliminate the coupling to the mechanical mode.

5.13 Other types of whistles

Beside the fluid driven whistles discussed above there are some other types, in particular
related to jet oscillations [6,7]. Two important examples for the sub-sonic case are hole&ring
and edge tones. For the supersonic case there is also a phenomenon known as jet screech.
Hole&ring and edge tones are produced by the interaction between a jet and a downstream
hole or a sharp edge (wedge), see Figure 5.5. To predict these tones, Rossiters formula can be
used, see [1]. The edge tone mechanism is used in several wind instruments, e.g., an organ
pipe where the tones are amplified by coupling to longitudinal acoustic (plane wave) modes.

a) b) c)

Figure 5.5 Sub-sonic jet tone arrangements: a) Hole jet tones; b) ring jet tones; c) edge tones38. To produce a
strong tone, the separation L to the downstream object should not be too small or large, typically 1< L/d < 10
where d is the jet diameter.

Jet screech is created by a supersonic jet which is not properly expanded so that it contains
shock cells. These cells can oscillate and produce an acoustic disturbance that travels
upstream via the surrounding quiescent fluid to the jet outlet. At the outlet the acoustic
disturbance trigs a flow disturbance that causes a new oscillation of the shock cells, thereby
forming a feedback loop.

38
Here a rectangular jet with the same width as the edge (=wedge) is assumed.

146
5.2* The Rijke tube a thermo-acoustic whistle
Whistling can be created not only by unstable mean flows as discussed in the previous section
but also by thermal sources. A classical example of such a thermo-acoustic whistle is the
Rijke tube, see Figure 5.6.
hot grid

x
-L/2 L/2

Figure 5.6 The Rijke tube.

This classical experiment was first analyzed by Lord Rayleigh [8] and consists of an open
pipe with a localized (steady) heat source (a hot grid) and a mean flow. It is observed that the
fundamental plane wave mode c0/2L can be excited in this pipe when the hot grid is
positioned a distance L/4 from the upstream end. Normally the flow is produced via heat
convection simply by orientating the pipe vertically (like a chimney!).

We will now analyze this problem following the steps originally proposed by Rayleigh, but
with details added based on recent investigations [9]. Although the posed problem can seem
academic the possibility that a heat source, normally unsteady combustion, couples to an
acoustic mode is an important issue for instance for gas turbines and rocket engines.

An unsteady heat source is equivalent to a volume (monopole) type of source. The power
injected to a sound field by a harmonically oscillating monopole of strength Q is

Win 12 Re p Q * , (5-5)

where * denotes complex conjugate and a necessary condition for a growing acoustic field is

Win 0. (5-6)

Equation (5-6) is called the Rayleigh criterion. The heat power transferred from the grid to the
gas can be seen as a function of the surrounding gas temperature and the flow speed.
Assuming a linear process the unsteady heat power (here expressed as Q) can be expressed as

Q AT ( x) Bu x ( x), (5-7)

147
where T , u x are the temperature and velocity fluctuations at the grid, and A and B are
constants. The constant A is related to heat transfer by conduction and since the hot grid has a
high heat capacity and thermal conductivity (compared to the gas) it acts as a heat reservoir,
i.e., conducts heat without time delay. This implies that A can be taken as real-valued and
positive, since an increase in gas temperature means less transferred heat. The constant B is
related to heat transfer via convection and for this process there is a time delay, which implies
that B is complex-valued. For an adiabatic sound field temperature and pressure fluctuations
are proportional39, thus equation (5-7) can be written as

Q Ap ( x) Bu x ( x), A 0 . (5-8)

We will now investigate how this source can excite the first plane wave mode in the pipe. The
pressure and velocity fields for this mode can be written as

x
p ( x) p 0 cos L

. (5-9)
ip x
u x ( x) 0 sin
0 c0 L

Using equations (5-8), (5-9) in (5-5) we obtain an expression for the power injected by the hot
grid into the first mode

2
p 0 2 x
B 2x
Win A cos sin cos 2 , (5-10)
2 L 2 0 c0 L

where B B exp(i ) . The first term is negative for all x, has a minimum when the grid is

placed in the middle of the duct and is zero at the open ends, this term represents damping of
the mode. The sign of the second term will depend cos( / 2 ) Damping in lower half and
amplification in upper
of the phase shift (or time delay ) associated
with the convection of gas across the hot grid

( ) . The time delay can be expressed as:
Amplification in lower half
and damping in upper
39
p / T /( 1) , where is the specific heat ratio.

148
 d / U c , where Uc is the convection speed past the grid (< U) and d is the width of the grid
in the x-direction. This result implies that

c0 d d
, (5-11)
L Uc ML

where the frequency is set equal to c0 / 2 L , and U c / U . As seen from equation (5-11) we
can control the phase of the second term by adjusting the width of the grid and the flow speed.
Maximum amplification and power injection is obtained either for / 2 and x L / 4 or
for 3 / 2 and x L / 4 . In experiments only the first alternative is normally observed and
not the second that, according to (5-11) for a given grid, corresponds to a much lower flow
speed. The reason for this is that a lower speed also implies much lower levels for the
resulting self-sustained oscillation, since u x / U O(1) [9].

149
References

1. W.K. Blake (1986) MECHANICS OF FLOW-INDUCED SOUND AND

VIBRATION, Vol. 1. Academic Press.
2. R.D. Blevins (1990) FLOW-INDUCED VIBRATION, Van Nostrand Reinhold.
3. I.E. Rossiter (1966) Aeronautical Res. Council Reports and Memoranda 3438. Wind-
tunnel Experiments on the flow over rectangular cavities at subsonic and transonic
speeds.
4. H.H. Heller, D.G. Holmes and E.E. Covert (1971) Journal of Sound and Vib. 18, 545-
553. Flow-induced pressure oscillations in shallow cavities.
5. M.P. Verge, B. Fabre, A. Hirschberg and A.P.J. Wijnands (1997) Journal of the
Acoustical Soc. of Amer. 101, 2914-2924. Sound production in recorder-like
instruments I: Dimensionless amplitude of the internal acoustic field.
6. M.S. Howe (1997) Journal of Fluid Mech., 330, 61-84. Edge, cavity and aperture
tones at very low Mach numbers.
7. W.K. Blake and A. Powell (1986) In RECENT ADVANCES IN AEROACOUSTICS,
eds. A. Krothapali & C.A. Smith., New York: Springer. The development of
contemporary views of the edgetone, 247-345.
8. J.W.S. Rayleigh (1945) THE THEORY OF SOUND, vol I and II. Dover.
9. M.A. Heckl (1988) Journal of Sound and Vib., 124, 117-133. Active control of the
noise from a Rijke tube.

150
Problems
1. Describe how a fluid-driven whistle is created and give two practical examples.

2. Discuss different alternatives to eliminate a fluid driven whistle.

3.
y

H x
U

A common type of heat exchanger consists of a periodic array of tubes with a flow
field perpendicular to the tubes. Under certain operating conditions this heat
exchanger can start to whistle. One possibility for this is that the flow separation
around the pipes excites the first acoustic cross mode at its cut-on frequency.

a) Given that the Strouhal frequency for flow separation around the pipes is

U
f St 0.2 ,
d

where d is the pipe diameter, derive the following necessary condition for whistling

M , =2.5d / H ,
1 2

where M is the Mach number.

c0 k ,n
Hint: The cut-on frequency for a mode in a duct is given by f nc 1 M 2 .
2

b) Calculate the flow speed U where whistling can occur for the case40: d =10 cm, H =3
m with c0 = 450 m/s and0 =0.52 kg/m3.

c) The sound field in the first acoustic cross mode for a wave in the +x direction, i.e.,
along the duct axis, can be written as

40
The data given could represent a power plant application.

151
 y
p( x, y, t ) p 0 sin exp i (t k x x) , H / 2 y H / 2
H

where the origo for the y-axis is in the middle of the duct. Show that the corresponding
acoustic velocity field in the y direction is given by

p 0 y
u y ( x, y, t ) cos exp i (t k x x) .
H i 0 c0 (k Mk x ) H

Hint: Use the linearized equation of motion 0 D0u p , where D0 / Dt / t U

Dt
and assume that the variation along x, i.e., the wave propagation part, is the same.

d) Apply the result in c) at the cut-on frequency for the mode and show that the
maximum velocity amplitude in the y-direction is given by

p 0
max( uy ) .
H 0 c0 k

e) Typically in fluid driven whistles it is found that the velocity amplitude at steady state
is around 10 % of the mean flow speed. Assume therefore max( uy ) 0.1U in the
result in d) to estimate the maximum pressure amplitude that can be expected for a
whistle with the data given in a).

4. A whistling side branch-resonator

S Open
U
p p end

u
S
x
0 L

A side branch resonator mounted in a duct as shown in the figure is assumed to

whistle at a certain frequency. This will produce a fluctuating volume flow:
Q Su and a monopole source. It is known that the amplitude of the whistle will
scale with the mean flow speed and therefore can be written as: u U . For the
analysis below a compact source and plane waves are assumed and mean flow effects
are neglected since the Mach number is small.

a) Show that the sound pressure amplitude and power of the wave propagating in the +/-
direction for the case of an infinite duct, i.e., if we disregard the open end at x = L, is
given by

152
 0c0U

p
p
2
.
W W

2 2
c
0 0 SU 2

Hint: Conservation of momentum implies that the pressure is continuous across the
source. Use this and the conservation of mass i.e. volume flow.

b) Calculate the sound pressure level generated for the case: U = 20 m/s, 0 =1.21
kg/m3, c0 = 340 m/s, 1, 0.1 .

c) With the open end present the sound field (omitting the factor exp(it ) ) for x > 0 can
be written as

p ( x) p exp(ikx) p exp(ikx),

Determine the reflected wave amplitude by using the boundary condition p ( L) 0 .

Using the expression for the pressure show that the total radiated power is given by

1 2 2 0c0 SU 2 sin 2 (kL)

Wtot Re p (0) Q * 4W sin 2 (kL) .
def 2 2

d) Plot the result in c) and find the positions that minimize the whistling. Give a physical
interpretation of the result. Assume that the whistling frequency is 240 Hz and
calculate the shortest distance L away from the opening that produces a minimum
sound power ( c0 340 m/s ).

153
154
Solutions
1. -

2. -

3a) The cut-on for the first mode occurs when we have a wavelength over the cross-
section, i.e., when: k ,1 H k ,1 / H . This combined with the necessary
condition for whistling gives:

c0 U 2.5d
1 M 2 0.2 M 1 M 2 , from which the result stated in the exam
2H d H

is obtained.

b) U=37 m/s.

c) It follows from the given hint that: D0 / Dt / t U D0 / Dt i ik x U , the eq.

of motion then implies that: u y 1 p . Inserting the given expression
i 0 c0 k Mk x y
for the mode pressure into this gives:
p 0 y
u y ( x, y, t ) cos exp i (t k x x) .
H i 0 c0 (k Mk x ) H

d) By definition at cut-on k x 0 . Then observing that the max velocity amplitude is

obtained for y = 0 we obtain the given expression.

e) The result from d) implies:

41
max p 0 0.2 0 f St HU insert values f St 75 Hz 0.2 0.52 75 3 37=866 Pa .

4a) Conservation of momentum or in this case simply the symmetry of the problem
implies:

p p . (1)

Conservation of volume flow (same as mass for the no flow or small Mach number
case)

p S p S 2 p S

Q Eq.1 (2)
0 c0 0 c0 0 c0

From the text we have: Q SU (3). Eqs. 2 and 3 gives:

41
This corresponds to 153 dB and shows why this phenomenon is very serious when it occurs in practice.

155
 0c0U
p p . (4)
2

b) The acoustic power is given by:

2
p 2 2 0 c0 SU 2

W W


Eq.4 . (5)
2 0 c0 8

Inserting the values gives: p 411 Pa . This gives Lp = 143 dB.

c) The boundary condition implies:

p exp(ikL) p exp(ikL) 0 p p exp(2ikL) . (6)

This means that the pressure at x = 0 is: p (0) p 1 exp(2ikL) . (7)

The power input then becomes:

1 1
Wtot Re p (0) Q * Re p 1 exp(2ikL) . SU Eq.4
def 2 2
0c0 SU sin (kL)
2 2 2 2
4W sin 2 (kL).
2

d) Minimum power is obtained when sin kL 0 kL 0, , 2 ,.... i.e. when we have a

multiple of half a wavelength. These positions correspond to minima (zeros) in the
standing wave field for the acoustic velocity. The smallest distance away from the
opening we can choose to minimize the power at 240 Hz is:
/ 2 c0 / 2 f insert values 0.71 m .

156
 APPENDIX

On the use of multi-ports

for flow ducts

157
158
1. INTRODUCTION
This short note is aimed to give an introduction to the use of acoustical multi-ports for
modelling sound generation and transmission in flow ducts. After a definition of the concept
of multi-port, the application to acoustics is treated with a focus on the fluid-borne sound
case, but some remarks concerning the structure-borne case are also included. Various
definitions of state variables as well as symmetry and reciprocity are discussed. Section 2
treats the low frequency (plane wave) range and the use of 2-port models. First, the standard
transfer-matrix formulation is introduced and applied to cascade coupled systems. Arbitrary
coupled systems are briefly discussed and the scattering-matrix is defined. The 2-ports for a
few standard cases are presented, including a straight pipe, a side-branch, an axial fan and a
flow constriction. Section 3 focuses on the use of experimental methods for determining
multi-port data, including both passive (scattering) as well as active (source) properties.

1.1 The multi-port concept

In general a multi-port42 is a system where a casual relation exists between a set of input x and
output variables y. These variables or state vectors are assumed to be functions of the time t
and completely define the state at the input and output of the system. Furthermore, it is
normally assumed that x and y has the same dimensionality (N). The multi-port is then called
an N-port where N can be interpreted as the number of degrees of freedom. Mathematically
the existence of a causal relation implies the existence of an operator G such that

y G x . (1)

Often this type of approach is called a black-box model in the literature, since it can be
applied without a full knowledge of the inner properties of the studied system.

x G y

Figure 1 A multi-port of order N (i.e. N-port) is a system where a causal relation exists between an input x and
an output y, where x and y are state vectors with N-components. A simple example is an ideal spring with one
fixed end, which can be seen as a 1-port defined via F = -Kx, where F is the force, x the displacement and K the
spring constant.

For the case of acoustics x and y are often defined via pressure, force and displacement,
velocity, respectively. The choice of state variables is dictated by both physical (what can be
42
The word port originates from electrical engineering and denotes a terminal. In general it can be interpreted as
an opening or a channel through which a system interacts with its surrounding.

159
measured !) considerations as well as mathematical, since certain choices will lead to more
efficient formulations for some types of problems. This latter aspect is discussed further in
section 2.

The relation in equation (1) is completely general and for the case of acoustics some
simplifications are possible, since the fields normally can be seen as small perturbations
around a reference state. For these small perturbations (or acoustic variables) linearized
equations are derived. Applying this approach to equation (1) yields

y G x , (2)

where the prim denotes an acoustic variable or linearized operator. Another simplification
normally valid in acoustics is to assume that the problem is time-invariant. This implies that
the operator in equation (2) satisfies

d d
G x G x , (3)
dt dt

where d/dt denotes a time derivative. Applying a temporal Fourier transform to this equation
or assuming x to vary harmonically in time ( x x e i t ) plus expanding x relative a set of
base vectors b n

x x n b n , (4)
n

gives

d d
G xnb n ei t G xnb n ei t .
dt n dt n

Using the linearity of the operator this can be written

d i t
x
n
n
dt
i t
G b n e iG b n e 0 .

Since this equation must hold for any vector x it follows that

160
 d
G b n ei t iG b n ei t 0 , (5)
dt

which after introducing y n (t ) G b n ei t becomes

dy n
iy n 0.
dt

The solution to this equation is simply

y n (t ) g n ( )ei t , (6)

which shows that a harmonic input of unit amplitude along base vector n creates a harmonic
response with the same frequency and amplitude g n ( ) . This property to conserve the
excitation frequency is characteristic for linear and time-invariant systems.

Equation (6) can be used to obtain the frequency domain form of equation (2). For a harmonic
x, equation (2) can be written

y G xn b n ei t ,
n

using equation (4). From equation (6) it follows that y y e i t which leads to

y e i t x G b
n
n n e i t {eq .(6)} x g
n
n n ( ) e i t . (7)

From this the sought frequency domain relationship between input and output directly follows

y Gx , (8)

where the matrix G contains the vectors g n as columns.

Equation (8) is a description in the frequency domain of a linear and time-invariant multi-port
system. It can also be concluded that the model describes a passive system, i.e., with no
internal sources. Because if there is no input ( x 0 ) then equation (8) produces a zero output.
Of course when modelling acoustical systems there is an interest to also include source

161
processes. The question is therefore how can equation (8) be generalized to include active
properties or source mechanisms in a system ?
For a linear multi-port system where superposition is valid the answer is simple; an internal
source process can be included simply by adding a source strength vector ( y s ) to equation (8)

y Gx y s . (9)

In the next section this model will be applied to describe sound generation from a ducted axial
fan as well as a flow constriction.

1.2 Multi-ports in acoustics [1]

A classical choice of state variables for the fluid-borne case is to use pressure ( p ) and
volume velocity43 ( q ). Choosing pressure as the output and volume velocity as the input
variable, gives the so-called impedance matrix description

p Zq , (10a)

excluding any internal sources. The frequencies for free oscillations of the system
(eigenfrequencies) are obtained by putting ( p 0 ). The resulting homogeneous equation has
non-trivial solutions only when

det(Z ) 0 . (10b)

An alternative definition is to choose the volume velocity as the output variable and pressure
as the input variable. This leads to the so-called mobility matrix definition

q Mp , (11)

where M Z 1 . It can be noted that the mobility type of description often is preferred in the
structure-borne case, see e.g. [2]. The equation for free oscillations based on this definition is
obtained by changing Z to M in equation (10b). For a given system this will create two sets of
eigenfrequencies, corresponding to two types of boundary conditions (p = 0 or pressure
release and q = 0 or blocked/rigid end).

43
In this appendix q is used instead of Q.

162
 p1

q1

Figure 2 An example of an acoustical multi-port a volume with N connected ducts. If only plane waves exist in
the ducts this is an N-port. If each duct can carry M modes then this will be an NxM-port. The inlet and outlet
states are defined at the duct openings.

Assuming an N-port then the multi-port matrix G will be NxN. In many cases spatial
symmetries exist and the reciprocity principle [3,4] is valid, which leads to relationships
between the matrix elements. This will cut down the number of unknowns and can be used to
reduce the work both in numerical or experimental procedures. Concerning acoustical
reciprocity it implies that when source and receiver is interchanged the response (acoustic
pressure) is unchanged. This principle is valid for the classical wave equation and effects of
yielding boundaries and losses can be included [5]. However, strictly the principle is not valid
when a mean flow is present and care must be taken to apply it for such cases. If reciprocity
can be assumed, then it can be shown [3] that the impedance and mobility matrices are
symmetric or anti-symmetric (depending on the sign convention for q).

2. THE LOW FREQUENCY CASE - 2-PORT MODELS

Most duct elements, e.g., a silencer are a system with just two openings. If only plane waves
exist at these openings then the system can be described as an acoustical 2-port44. In practice
this condition is satisfied for sufficiently low frequencies, i.e., below the cut-on frequency for
higher modes. This assumes of course that the effect of coupled wall vibrations is negligible,
which normally is the case for circular gas filled ducts made of metal. For low frequencies
strong standing wave effects and coupling between a source and duct system is possible. This
makes noise control for low frequencies more difficult and not simply just a question of
adding damping to reduce noise levels. Instead a detailed analysis of wave interaction is
needed and this can be achieved by modelling a duct system as a network of 2-ports. In this
section the basic theory for this will be presented and applied to some standard cases. At the
end of the section some info on codes available for 2-ports analysis is given.

44
In electrical engineering this type of system is often called a four-pole and this term can also be found in
acoustic literature.

163
2.1 Modelling of cascade coupled networks [6]
The most commonly used formalism is obtained by choosing acoustic pressure and volume
velocity (p,q) to describe both the input and output state. This mixed choice leads to the so-
called transfer matrix formalism that is especially well suited for treating problems where all
duct elements are in a cascade (a chain). Examples of this are for instance exhaust systems on
automobiles. Denoting the input side a and the output b the transfer-matrix T for a passive
element can be defined via

p a p b
T . (12)
qa qb

This is called the backward transfer-matrix and if a and b are interchanged (=matrix
inversion) the so-called forward transfer-matrix results.

pa
qa qb
b
qa
pa T
qb pb
a
pb

Figure 3 Acoustical 2-port and circuit representation based on the electric-acoustic analogy [3]
(p =voltage and q =current).

The advantage with the transfer-matrix formalism for cascade coupled systems lies in that it
automatically fulfils the appropriate coupling conditions at the interface between 2-ports.
These conditions are a consequence of the conservation of momentum and mass applied over
a small control volume enclosing the interface and require continuity of pressure (p) and
volume velocity (q). This result is strictly valid in a straight duct with a homogeneous fluid
state. If there is an area jump the continuity of q still holds strictly for no flow [3] and for an
incompressible45 mean flow, but for p there is a jump

pb.1 pb.2 Z b qb , (13)

where Zb is the impedance associated with the interface and the notation is based on Figure 4.

45
This implies that the Mach number M is much smaller than 1, which in engineering practice means M < 0.3.

164
 pa

qa 1 b c
pb,1
pb,2 2
a q

Figure 4 Cascade coupled 2-ports.

The imaginary part of the impedance Zb is associated with non-propagating higher order
modes excited at b creating an acoustic near field. This term is equivalent to an acoustic
inertia (mass) and is proportional to the frequency. For no flow this part can be calculated
using the so-called Karal end correction [7]. Since the imaginary part grows with frequency,
the real part of Z will always dominate for sufficiently low frequencies. This part is associated
with losses and with flow and flow separation at the interface, the losses associated with
acoustically induced vortex shedding will dominate. These losses can be estimated in the zero
frequency limit via a quasi-stationary model, see e.g. [8], based on the energy loss across the
interface. To ensure that the continuity of p and q holds at element interfaces these should be
chosen where there are no jumps. Then any area jumps can be included by defining a special
coupling 2-port matrix Tb, since

pb.1 pb.2 Z b qb , 2

qb ,1 qb , 2

it follows that

1 Z b
Tb . (14)
0 1

From the discussion above it follows that the total transfer-matrix for a cascade coupled
system with M passive 2-ports, T1, T2,..,TM is

M
Ttot Tm , (15)
m 1

assuming the numbering to be from the input to the output of the cascade. In many problems
the cascade is excited at the inlet by a source, e.g., an IC-engine. The output is normally an
open duct or pipe radiating to a surrounding fluid (air). The input can assuming a linear and

165
time-invariant source be modelled as an active 1-port. With p and q as state variables, this 1-
port can be described via

p1 ps Z s q1 , (16)

where ps is the source strength and Zs is the source input impedance. Similarly the output can
be seen as a passive 1-port described via

pM 1 Z r qM 1 , (17)

where Zr is the radiation impedance.

Zs q1 q2 qM qM+1
ps p1 T1 pM TM pM+1 Zr
p2

Figure 5 A cascade with M passive 2-ports excited at the inlet by a 1-port source and terminated by an
impedance (a passive 1-port).

Using the model described above various measures of the transmission behaviour can be
calculated. One common is the so-called transmission loss (TL), which describes the passive
properties of the cascade, i.e., it is independent of the source at the input. This measure is
defined as the ratio (in dB) of the incident power (Win) to the power transmitted (Wtr) for a
given termination of the system. Normally this termination is chosen as a reflection free
(infinite) pipe.

Using the Transfer-matrix for the total cascade it follows that

p1 T11tot T12tot p M 1
tot , (18)
q1 T21 T22tot q M 1

splitting the waves at interface 1 and (M+1) into travelling waves in the +/- direction gives

p1 p1 p1
q1 p1 p1 / Z1

p (19)
M 1 p( M 1)


q M 1 p ( M 1) / Z M 1

166
where Z1 and ZM+1 are the characteristic impedances46 at 1 and (M+1), respectively.
Inserting (19) into equation (18) gives after a little algebra

2 p1 T11tot T12tot / Z M 1 T21tot Z1 T22tot Z1 / Z M 1 p ( M 1) .

This result together with the definition of acoustic power in a system with flow (eq. (3-29))
implies that

1 Z M 1 (1 M 1 ) 2 tot 2
TL 10 log10 T T12tot / Z M 1 T21tot Z1 T22tot Z1 / Z M 1 .
2 11
(20)
4 Z1 (1 M M 1 )

where M is the Mach number. Another measure, which takes the influence of the source into
account, is the insertion loss (IL). This can be defined as the ratio (in dB) between the
acoustic power radiated at the outlet of a reference system and the system investigated, with
both systems driven by the same source. Often the reference is taken as a straight pipe with
the same length as the investigated system. This measure is popular in practice since it is
much easier to measure IL than TL.

2.2 Modelling of arbitrary coupled networks

For systems with more complex connections than a cascade the transfer-matrix formalism
becomes less useful. As pointed out by Frid [2] a formalism based on mobility-matrices (see
equation (11)), analogous to the one used for structure-borne sound, is then better. Another
alternative presented by Glav and bom [9] is to use a scattering-matrix formalism.

a
pa+ b

pb+

Figure 6 Definition of positive directions for the scattering-matrix formalism.

This formalism is based on travelling pressure waves as the state vectors and a scattering
matrix S defined as

46
Z c / S , where is fluid density, c speed of sound and S cross-sectional area.

167
 p a p a p as
S s . (21)
pb pb p b

As argued by Glav and bom [9] this formalism is the most general and can be applied to
systems with both arbitrary couplings and sources anywhere in the network.

Finally, it can be mentioned that it is also possible to define a transfer-matrix formalism based
on travelling waves [10]. Again this leads to a formulation best suited for cascade coupled
systems.

2.3 Examples on 2-port models

In this section a few standard cases will be presented to illustrate how 2-port models can be
derived. The derivation will be based on the transfer-matrix (p,q) formalism. More examples
in particular for different muffler types, e.g., dissipative with porous lining and perforated,
can be found in the book by Munjal [6] and, e.g., in [11,12].

2.3.1 The straight duct

This is the simplest case and assuming a straight uniform duct containing a homogenous fluid
between an inlet section a and an outlet section b, the acoustic state is given by

p a p a p a p b p a e ik L p a eik L
,
qa p a p a / Z 0
qb p a e
ik L

p a eik L / Z 0
where Z0 is the characteristic impedance, k , k , M is the Mach number
c1 M c1 M
and L is the length of the duct. From the first equation one obtains

p a p a Z 0 qa / 2
,
p a p a Z 0 q a / 2

which when inserted in the second gives the desired transfer-matrix relationship in the
forward form. By inverting this result the backward form (b to a) is obtained

ikML / 1 M 2


cos kL / 1 M 2
iZ sin kL / 1 M 2
,
Te (22)

i / Z sin kL / 1 M 2

cos kL / 1 M
2

168
where k / c . It is possible to include damping by modifying this model and allowing k to
be complex valued, i.e., putting k / c 1 i . Expressions for the damping can be found
in references [3,4].

By putting together a series (cascade) of straight ducts with varying cross-section it is possible
to create an acoustic filter, which stops sound propagation in certain frequency regions. The
most well-known case being the so called expansion chamber muffler.

TL(dB)
TLmax increase
with S2/S1

S1 S2

c/4L 3c/4L 5c/4L f [Hz]

Figure 7 An expansion chamber muffler and its transmission loss curve exhibiting periodic stopbands with
maxima where the wavelength equals an odd multiple of L/4.

2.3.2 The side-branch

A side-branch element is in general an opening in a duct wall characterized by its input
impedance (Zin). This is a common element often used for noise control and a suitable (low)
impedance is created by mounting a resonator to the opening.

Zin Zin 3) Zin

1) 2)

a b a b a b

Figure 8 Side-branches. 1) General, 2) Helmholtz-resonator, 3) Quarter-wave resonator.

Close to resonance ( Z in 0 ) the side-branch will block (short circuit) sound propagation in
the duct by creating a pressure node. The most common type of resonators are the Helmholtz
and quarter-wave resonators [6].

169
Applying conservation of momentum and mass47 across the side-branch (with an opening
assumed to be much smaller than a wavelength) leads to
p a p b 0

qa qb qin
where the acoustic pressure is assumed to be approximately constant across the side-branch.
This implies that acoustic near fields and flow induced losses are neglected. Inserting the
definition of impedance ( qin p in / Z in ) into the above equations gives

1 0
T . (23)
1/ Z in 1

For the case of a quarter-wave resonator the input impedance can be obtained from the
straight pipe transfer-matrix (equation (22)), by assuming a rigid termination at b. This gives
the input impedance (M = 0)

Z in iZ 0 cot kL . (24)

2.3.3 The axial fan

An axial fan (a propeller) mounted in a duct will in the plane wave range behave as a
compact dipole source. Since an acoustic dipole is equivalent with an oscillating force, an
axial fan can be seen as a force source. Assuming further that the fan behaves as an ideal
(point) dipole, it follows that incident sound is transmitted without reflection and phase
shift. With these simplifications conservation of momentum and mass across the fan can be
written

p a p b F / S
,
qa qb

which in matrix form becomes

p a 1 0 p b F / S
. (25)
qa 0 1 qb 0

47
An incompressible mean flow is assumed.

170
 F
S

a b
Figure 9 Active 2-port model of an axial fan.

In the derivation of equation (25) the effects of flow on the dipole properties has been
neglected. A more detailed analysis can be found in reference [13]. To apply this model in
practice requires a model for the dipole force created by a fan. An example on how this can be
solved is given in reference [14], where noise from a so called jetfan is modelled.

2.3.3 The flow constriction

It is also possible to use 2-port models to describe noise generated by flow separation at
various flow duct constrictions. As demonstrated in reference [8,15], where noise from a
sharp edge (an orifice or a bend) is treated, the active part can be obtained by semi-empirical
scaling laws. These laws are based on the observation that for low Mach numbers the dipole
type of source mechanism, related to unsteady wall pressure, will dominate the sound
production [16]. The passive part can be modelled in the limit of low frequencies, by
assuming a quasi-stationary model based on the flow related losses. Assume a duct with a
constant cross-sectional area then the flow related loss over a constriction can be defined from

pa pb 12 CL 0U 2 , (26)

where CL is the loss coefficient, U the (average) mean flow speed and 1,2 denotes the up- and
downstream side of the constriction, respectively. For sufficiently low frequencies equation
(26) will still be valid and adding a small disturbance to the mean flow field gives:

pa pa pb pb 12 CL 0 U u
2
. Rewriting this keeping only the linear terms leads to

pa pb CL 0Uu , (27)

This equation combined with the assumption of an incompressible flow (valid for M << 1)
results in the following 2-port for the passive part

171
 1 0CLU
T . (28)
0 1

2.3.4 Codes for 2-port analysis

Today 2-port analysis is a standard tool for, e.g., designing automobile exhausts. Therefore a
number of specially written codes exist at manufacturers and consultancy companies active in
this field. One is SID or Sound In Ducts developed in Sweden at KTH. This code is based on
the theory in reference [9] and the present version, SID 3.0, is the result of the work of
Nygrd [15]. Another code is LAMPS [17] developed by Keith Peat at Loughborough
University, England. Then there is a French code called LACTUS [18] developed with
support from the French car industry (Peugeot&Citron). Also the two dominating
commercial codes, BOOST and WAVE, for 1-D gas dynamic analysis of IC-engines contain
linear 2-port modules. In the case of BOOST it is the SID 2.1 version which is behind this
module.

3.* EXPERIMENTAL DETERMINATION OF MULTI - PORT DATA

Here only a few general comments around this problem will be given. For a review of the
subject and in particular for discussions on how to determine source data for fluid machines,
e.g., fans, see reference [1].

The starting point is equation (9) describing a general multi-port of order N

y Gx y s .

In this equation the passive part described by the matrix G contains in general (no symmetries
or reciprocity assumed) N2 unknown elements. The source strength vector ys contains N
unknowns thus resulting in a total of N2+N unknowns. Assume now that it is possible to
measure the input and output state vectors. Then for each given state of the multi-port, i.e.,
known x and y, N equations would result. By testing the multi-port at N+1 different
(independent) states, a set of Nx(N+1) equations is obtained, from which in principle the
unknowns G and ys could be solved. In practice the problem to implement this scheme is
associated with the need for a reference signal coupled to the source process. Without such a
reference it is not possible to link different measurements and if the source process is broad

172
band, which is often the case, then a reference is normally not available. To avoid this need
for a reference it has been suggested [1] to split the measurement process in two steps. One
obvious way of doing this would be to ensure a zero input (x = 0), then y = ys. However, in
practice this is not easy to realize, in particular for a running fluid machine since zero input
typically implies no flow. A more practical procedure is to first excite the multi-port by an
external source, uncorrelated with the internal source process. By correlation techniques it is
then possible to extract the part of y that is correlated with the input. In this way the
contribution from ys is removed from the output and only the passive properties of the system
is tested. By repeating this with N different configurations for the external source (the
multiple source method) an equation system can be formed [1]

h 1
y/e . . h yN/ e
G h1x / e . . h xN/ e , (29)

where hy/e denotes a frequency response function between y and the signal e driving test state
n. This formulation will both suppress the contributions from internal sources (assumed
uncorrelated with e) and also, e.g., suppress flow noise (turbulence) at pressure probes used
for determining x and y.

It can be noted that to improve the measurement quality, equation (29) can also be formulated
as an overdetermined problem by adding more than N measurements to the matrices. By
inverting the matrix on the right side in equation (29) G can be solved. When G is known the
source strength vector ys can be obtained by testing the system state with no external sources.

A special case of the above is a passive N-port then the first step is sufficient. An alternative
to the the multiple source method, for determining the passive part, is to use a single
external source and create different x and y by changing the loads on different ports. As
shown in, e.g., reference [19] for the 2-port case, this multiple load method typically
performs less good than the multiple source method.

173
REFERENCES

1. H. Bodn and M. bom (1995) Acta Acustica 3, 549-560. Modelling of fluid

Machines as sources of sound in duct and pipe systems.
2. A. Frid (1989) J. Sound and Vib. 133, 423-438. Fluid vibration in piping systems a
structural mechanics approach.
3. A.D. Pierce, ACOUSTICS - An introduction to its physical principles and
applications. McGraw-Hill 1981.
4. S.W. Rienstra and A. Hirschberg, An introduction to acoustics. Report IWDE 01-03,
Technische Universiteit Eindhoven 2004.48
5. L.M. Lyamshev (1959) Sov. Phys. Dokl. 4, 405-409. A question in connection with
the principle of reciprocity in acoustics.
6. M.L. Munjal (1987), Acoustics of ducts and mufflers. Wiley-Interscience.
7. F. Karal (1953) J. Acoust. Soc. Am. 25, 327-334. The analogous acoustical impedance
for discontinuities and constrictions of circular cross-section.
8. H. Gijrath, S. Nygrd and M. bom (2000), ICSV8_paper no. 578. Modelling of flow
generated sound in ducts.
9. R. Glav and M. bom (1997) J. Sound Vib. 202, 739-747. A general formalism for
analyzing acoustic 2-port networks.
10. P.O.A.L Davies (1988) J. of Sound Vib. 124, 91-115. Practical flow duct acoustics.
11. M. bom (1990) J. of Sound Vib. 137, 403-418. Derivation of the four-pole
parameters including higher order mode effects for expansion chamber mufflers with
extended inlet and outlet.
12. R. Glav (2000) J. of Sound Vib. 236, 575-594. The transfer matrix for a dissipative
silencer of arbitrary cross-section.
13. M bom and H Bodn (1995) J. of Sound Vib. 186, 589-598. A note on the
aeroacoustic source character of in-duct axial fans.
14. J.G. Lalanne, H. Bodn, M. bom and R. Parchen (1999), Paper at Forum Acusticum
99, Berlin. Aeroacoustic response of a six-bladed axial fan to installation effects.
15. S. Nygrd (2000) Lic. Tech thesis KTH_Sweden. ISRN KTH/FKT/L-00/57-SE.
Modelling of low frequency sound in duct networks.
16. P.A Nelson and C.L Morfey, 1981. Aerodynamic sound production in low speed flow
ducts. Journal of Sound and Vibration 79(2), pp. 263-289.
17. K.S Peat (1995) Proc. of Euro-Noise `95, 791-796. LAMPS software for the acoustic
analysis of silencers.
18. F. Di Costanzo (2001) Proc. First Euro Forum Materials and Products for Noise and
Vibration Control at CETIM, France. Rduction des pulsations de pression dans un
rseau de tuyauteries industrielles.
19. M bom (1992) J. of Sound Vib. 155, 185-188. A note on the experimental
determination of acoustical two-port matrices.

48
Can be down loaded from: www.win.tue.nl/~sjoerdr/papers/boek.pdf

174