This action might not be possible to undo. Are you sure you want to continue?
Flow Interacting with Solid
Boundaries
Copyright c _2010 by P.C.H. Aben, Eindhoven, The Netherlands.
All rights are reserved. No part of this publication may be reproduced, stored in a
retrieval system, or transmitted, in any form or by any means, electronic, mechanical,
photocopying, recording or otherwise, without prior permission of the author.
Printed by Print Service Technische Universiteit Eindhoven
Cover design by Paul Verspaget
A catalogue record is available from the Eindhoven University of Technology Library
Aben, Paul
HighAmplitude Thermoacoustic Flow Interacting with Solid Boundaries /
by Paul Aben. 
Eindhoven: Technische Universiteit Eindhoven, 2010.
Proefschrift.  ISBN 9789038623993
NUR 928
This research was ﬁnancially supported by the Technology Foundation (STW), grant
number ETTF.6668.
HighAmplitude Thermoacoustic
Flow Interacting with Solid
Boundaries
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de
Technische Universiteit Eindhoven, op gezag van de
rector magniﬁcus, prof.dr.ir. C.J. van Duijn, voor een
commissie aangewezen door het College
voor Promoties in het openbaar te verdedigen
op woensdag 8 december 2010 om 16.00 uur
door
Paul Cornelis Hubertus Aben
geboren te Sittard
Dit proefschrift is goedgekeurd door de promotor:
prof.dr. A.T.A.M. de Waele
Copromotor:
dr.ir. J.C.H. Zeegers
Nomenclature
List of symbols with a short description and the page on which it is introduced for the ﬁrst time.
Symbol Unit Description Page
α kineticenergy correction factor 67
α W/(m
2
K) heat transfer coefﬁcient 88
β momentum correction factor 66
β K
−1
volume expansivity 131
γ ratio of speciﬁc heats 1
δ
ν
m viscous penetration depth 16
δ
κ
m thermal penetration depth 18
ζ m displacement 24
η efﬁciency 23
η
C
Carnot efﬁciency 24
θ rad phase angle 25
κ m
3
/s thermal diffusivity 18
λ m wavelength 4
µ Pas dynamic viscosity 86
ν m
3
/s kinematic viscosity 14
ξ m
−1
complex wave number 21
Π m wetted perimeter 13
ρ kg/m
3
density 14
σ
s
−1
viscous stress tensor 14
ψ porosity 54
ω rad/s angular frequency 15
ω
z
s
−1
vorticity 108
A m
2
crosssectional area 13
A
c
m
2
cone area 36
b
µ
viscosity exponent 86
b
k
thermal conductivity exponent 86
B
l
N/A motor force factor 36
B
o
Boussinesq number 133
vi
c
p
J/(kgK) speciﬁc isobaric heat capacity 14
c
so
J/(kgK) solid heat capacity 86
C
c
contraction coefﬁcient 68
C
+
, D
+
Pa amplitude of rightward traveling wave 55
C
−
, D
−
Pa amplitude of leftward traveling wave 55
COP coefﬁcient of performance 24
COP
C
Carnot coefﬁcient of performance 24
d m pore wall thickness 53
D m diameter 38
D m pore size 53
D
p
m plate separation 86
D
r
drive ratio 15
˙
E W total energy ﬂow 19
∆
˙
E
ml
W energy loss due to minor losses 66
f Hz frequency 41
f
0
Hz characteristic frequency 54
f
ν
viscous Rott function 17
f
κ
thermal Rott function 18
F N force 36
g m/s
2
acceleration due to gravity 131
h J/kg speciﬁc enthalpy 19
h
ν
, h
κ
channelgeometry dependant function 17
I A current 36
k W/(Km) thermal conductivity 14
k N/m spring constant 36
k
/
m
−1
real component of ξ 55
k
//
m
−1
imaginary component of ξ 55
K minorloss coefﬁcient 67
K
c
minorloss coefﬁcient for contraction 68
K
e
minorloss coefﬁcient for expansion 67
KC
D
KeuleganCarpenter number based on D 113
KC
L
KeuleganCarpenter number based on L 111
l
0
m half the plate thickness 16
L
el
Vs/A selfinductance 36
L m length 38
vii
˙ m kg/(m
2
s) mass streaming 130
M kg mass 36
N
u
Nusselt number 133
p Pa pressure 14
P
r
Prandtl number 18
˙
Q W cooling/heating power 23
R
s
J/(kgK) speciﬁc gas constant 14
R
e
Reynolds number 107
R
eδ
Reynolds number based on δ
ν
113
R
el
Ω electrical resistance 36
R
me
Ns/m friction constant 36
R
a
Rayleigh number 133
R
H
m hydraulic radius 13
R
r
reﬂection coefﬁcient 57
s J/(kgK) entropy per unit mass 14
S
t
Strouhal number 107
S surface 65
t s time 14
T K or
◦
C temperature 14
T transfer matrix 21
u m/s xcomponent of v 14
u
c
m/s coil velocity 37
U m
3
/s volume ﬂow rate 15
v m/s velocity vector 14
v m/s ycomponent of v 14
[v
/
[ m/s rootmeansquare value of v 129
V V voltage 36
V
b
m
3
back volume 36
V
ind
V induced voltage 36
w m/s zcomponent of v 14
˙
W W Work ﬂow 23
˙
W W acoustic energy ﬂow 60
˙
W
i j
W acoustic energy ﬂow at position (x
i
+ x
j
)/2 62
W
p
m plate width 86
x m position along tube axis 14
y m 14
y
0
m half the plate distance 16
z m 14
Z impedance 36
Z
ac
N m
−5
s acoustical impedance 36
Z
el
V/A electrical impedance 37
(∇T)
crit
K/m critical temperature gradient 18
viii
List of sub and superscripts and special operators.
Symbol Description Page
a
ν
viscous 17
a
κ
thermal 18
a
a
ambient 2
a
ac
acoustical 36
a
C
cold 2
a
el
electrical 36
a
H
hot 2
a
in
(power) into the device 24
a
L
lefthand side 21
a
me
mechanical 36
a
R
righthand side 21
a
out
(power) out of the device 24
a
p
stack plate 86
a
so
solid 19
a
0
zeroth order (average over one period) 15
a
1
ﬁrst harmonic 15
a
2
second harmonic 15
¸a¸ crosssectional average 17
a
∗
complex conjugate 19
˙ a per unit time 19
a
i j
matrix element of a with row i and column j 38
a
+
MoorPenrose generalized inverse of K 57
a
H
Hermitian matrix of K 57
Contents
1 Introduction 1
1.1 Thermoacoustics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 History and applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Standingwave or stackbased devices . . . . . . . . . . . . . . . . . 2
1.2.2 Travelingwave or regeneratorbased devices . . . . . . . . . . . . . 3
1.2.3 Special applications and phenomena . . . . . . . . . . . . . . . . . . 4
1.3 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Theory and model 13
2.1 General thermoacoustic theory . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.1 Basic equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.3 Velocity proﬁle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.4 Thermoacoustic continuity equation . . . . . . . . . . . . . . . . . . 17
2.1.5 Total energy ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.1 Transfer matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2.2 Energy equation and temperature . . . . . . . . . . . . . . . . . . . 22
2.2.3 Iteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Thermoacoustic devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.2 Stackbased devices . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.3 Regeneratorbased devices . . . . . . . . . . . . . . . . . . . . . . . 30
3 Electroacoustics 35
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Theory and model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2.1 Speaker equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2.2 Acoustic impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.1 Empty resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4.2 Fixed speaker voltage . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.3 Fixed frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
x CONTENTS
4 Acoustic measurements 51
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.1 Microphones and speaker . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3 Multimicrophone method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Acoustic energy losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.4.1 Acoustic energy ﬂow . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.4.2 Fusco method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.4.3 Travelingwaves method . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4.4 Acousticenergyﬂow measurements . . . . . . . . . . . . . . . . . . 62
4.4.5 Acoustic energy losses in a stack . . . . . . . . . . . . . . . . . . . . 63
4.4.6 Minor losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.5 Minorloss correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.6 Singlestackposition method . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.7 Multiplestackpositions method . . . . . . . . . . . . . . . . . . . . . . . . 75
4.8 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5 Temperature proﬁle in a stack 81
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3.1 Steady State model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.3.2 Time Dependent model . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.4 Heat losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.5 Temperature proﬁle measurement . . . . . . . . . . . . . . . . . . . . . . . 91
5.5.1 Time dependent measurements . . . . . . . . . . . . . . . . . . . . . 91
5.5.2 Amplitude sweep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.5.3 Frequency sweep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.6 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6 Flow visualization in and around a stack 99
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2.1 Perspex setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.2.2 Aluminium setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.2.3 Particle Image Velocimetry . . . . . . . . . . . . . . . . . . . . . . . 105
6.3 A typical measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.4 Different categories of vortex patterns and dimensionless numbers . . . . 108
6.5 The inﬂuence of the plateend shape . . . . . . . . . . . . . . . . . . . . . . 113
6.6 Numerical simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.6.1 CFD model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.6.2 Subsequent cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.6.3 Minor losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.7 Vortex street evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.8 Velocity proﬁle inbetween two plates . . . . . . . . . . . . . . . . . . . . . 125
6.8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.8.2 A single measurement of the velocity proﬁles . . . . . . . . . . . . 125
CONTENTS xi
6.8.3 Fitting procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.8.4 First harmonic velocity and the transition to turbulence . . . . . . . 126
6.8.5 Velocity ﬂuctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
6.8.6 Streaming velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.9 Natural Convection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.9.1 Scale analysis on a vertical wall . . . . . . . . . . . . . . . . . . . . . 130
6.9.2 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.10 Streaming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.11 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7 Flow measurements in coaxial regeneratorbased devices 143
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.3.1 Instantaneous measurements . . . . . . . . . . . . . . . . . . . . . . 145
7.3.2 Streaming in window 2 . . . . . . . . . . . . . . . . . . . . . . . . . 147
7.3.3 Streaming in window 3 . . . . . . . . . . . . . . . . . . . . . . . . . 149
7.4 Discussion and conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Chapter 1
Introduction
1.1 Thermoacoustics
Thermoacoustics concerns phenomena in which an interaction of acoustics with ther
modynamics takes place. In an acoustic wave the gas parcels always undergo tempera
ture variations, which is a consequence of their compression and expansion. When the
acoustic wave is adiabatic, the temperature variations to ﬁrst order T
1
are given by
T
1
T
0
=
γ −1
γ
p
1
p
0
, (1.1)
in which T
0
is the average temperature, γ the ratio of speciﬁc heats, p
1
the pressure
variation, and p
0
the average pressure. In adiabatic sound waves these temperature
variations go unnoticed. However, when a solid is present near the acoustic wave, the
wave interacts with the solid and can cause a transfer of heat from one location in the
solid to another. This is called the thermoacoustic heatpumping effect. This effect is the
driving mechanism in stackbased coolers and heat pumps. Vice versa, when a sound
wave interacts with a solid with a temperature gradient above a certain critical value,
the temperature gradient enhances the sound wave.
Lord Rayleigh was the ﬁrst to give a qualitative description of thermoacoustic phe
nomena. In his work ”The Theory of Sound” [1], published in 1887, he discussed the
ability to generate temperature differences using acoustic waves. The subject remained
largely untouched for over eighty years, until in 1969 Nicholas Rott began a series of
publications [2–7] that initiated a revival in thermoacoustic research. Rotts work forms
the theoretical basis of most of modern standingwave thermoacoustic research. Later,
Rott’s publications have been reviewed by Swift [8]. Once the theory for parallel plates
was established, the thermoacoustic theory for different pore shapes (triangular, rectan
gular, circular and pin arrays [9] instead of parallel plates) was developed by Arnott,
Bass, and Raspet [10]. Also other authors contributed to the understanding of ther
moacoustics, including Wheatley and Atchley. Wheatley [11] introduced the thermoa
coustic couple and compared his measurement results quantitatively with Rott’s theory.
Atchley worked on thermoacoustically generated temperature gradients in a thermoa
coustic couple [12], on the analysis of thermoacoustic prime movers and their onset of
2 Introduction
selfoscillation [13–16], and together with Gaitan on the energy dissipation of higher
harmonics and how to avoid this by detuning the resonance tube [17].
In ’t panhuis [18] rederived the thermoacoustic theory in a very systematic way.
At the Los Alamos National laboratory a ﬂexible code, called Design Environment for
Lowamplitude ThermoAcoustic Engines (DeltaE) [19], was developed for the numeri
cal simulation of thermoacoustic devices. In 2007 a version with a better user interface
was developed, called DeltaEC [20]. This code has been used by many groups to de
velop and test new devices.
1.2 History and applications
We distinguish heatdriven devices, called prime movers, and sounddriven devices,
called refrigerators or heat pumps. In a prime mover the absorbed heat at a higher tem
perature T
H
is partially converted into work and the remaining energy is released at a
lower temperature T
C
< T
H
. A refrigerator or heat pump absorbs heat at a lower tem
perature and requires input of work to release or pump heat to a higher temperature,
where the released heat is the sum of the absorbed heat and work. The difference be
tween a heat pump and a refrigerator is that in a heat pump, the heat is transferred from
ambient temperature T
a
to a higher temperature T
H
> T
a
, whereas in a refrigerator heat
is transferred to ambient temperature T
a
from a lower temperature T
C
< T
a
.
1.2.1 Standingwave or stackbased devices
Already in the 19th century, glass blowers noticed that when a hot glass bulb was at
tached to a cool glass tube, it sometimes emitted sound, and Sondhauss [21] in 1850
quantitatively investigated the relation between the pitch of the sound and the dimen
sions of the apparatus.
Standingwave devices are called this way, since their operation is based on a stand
ing wave that is created in a resonator tube. In case of a refrigerator one side of the
resonator is closed and at the other side an acoustic power source is located, e.g. a loud
speaker, linear motor, or thermoacoustic prime mover. In case of a prime mover, one
end of the resonator is closed and at the other end an acoustic load is located, e.g. a
piston or a thermoacoustic refrigerator. For the operation of a standingwave device
an imperfect heat transfer of the gas with a solid is essential. To optimize the gaswall
interaction, a solid object, called a stack, is installed in the resonator. A stack consists
of pores whose size is of the same order of magnitude as the thermal penetration depth
in the gas. For a good performance the stack is located between a pressure node and a
pressure antinode, since the thermoacoustic effect requires both pressure and displace
ment oscillations. The waves in a thermoacoustic device are never pure standing waves,
since standing waves cannot transport energy. The phase angle between pressure and
velocity is close to 90
◦
, but never exactly 90
◦
.
In the Sondhauss tube the thermoacoustic effect occurs in a single pore. In standing
wave engines, however, the process occurs in a stack, which consists of many pores in
parallel, all of which contribute to the acoustic power generation. Such a stack was not
added to a Sondhauss tube until the 1960s. This important development allowed ﬁll
ing a largediameter tube with relatively small pores, creating a large volume of strong
1.2 History and applications 3
thermoacoustic power production, while leaving the rest of the resonator open and rela
tively low in dissipation. Heat exchangers spanning the ends of the stack are needed for
efﬁcient delivery and extraction of the large amounts of heat needed by a stack. Early
use of such heat exchangers was described by Feldman and Carter [22] and by Wheatley
et al. [23].
The ﬁrst strandingwave refrigerator was developed by Hoﬂer [24] in 1986. He used
a loudspeaker to drive a closed resonator tube with a stack positioned near the loud
speaker. At the other part of the tube a volume was attached to simulate an open end
ing. The tube is effectively a quarterwave resonator. At both sides of the stack a heat
exchanger is located, one at low temperature and the other at ambient temperature. Ti
jani et al. [25, 26] optimized this standingwave refrigerator, using the back volume of
the loudspeaker as a gasspring system to adept the effective resonance frequency of the
loudspeaker to its load. They managed to cool down to 70
◦
C. They also demonstrated
the advantage of using gas mixtures, in order to get a low Prandtl number.
One application of stackbased devices is an icecream freezer that is developed by
Garret [27] from the Pennsylvania State University in 2004 in cooperation with Ben &
Jerrys Homemade. It is 25 cm in diameter, 48 cm tall, is ﬁlled with 10 bar helium, and
is driven by a movingmagnet linear motor at around 100 Hz. The cooling capacity is
119 W at 24.6
◦
C and the coefﬁcient of performance is 19% of Carnot. Although this
refrigerator is more expensive than conventional refrigerators, no chemicals or gases
are used that can be harmful to the environment. Another advantage of thermoacous
tic devices, including this refrigerator, is that they have only a few moving parts, and
even more important, no moving parts in the cold, which makes them very reliable
and relatively simple and ﬂexible to build. This makes them also interesting for space
applications [28].
When pumping natural gas through pipes to the surface, sensors are located far
below the ground. To power these sensors a reliable electrical power supply is required.
One way to provide this power is to create a side branch to the main ﬂow, which was
studied by Slaton [29–32]. At the edges of the side branch with the main pipe, vortices
will be shed. When the vortexshedding frequency corresponds with the resonance
frequency of the side branch, a standing wave is created. By installing a stack in the
side branch the acoustic power can be used to generate a temperature difference, that
can produce electric power with thermoelectric elements.
1.2.2 Travelingwave or regeneratorbased devices
In Stirling engines and socalled travelingwave engines, the conversion of heat to acous
tic power occurs in the regenerator, which smoothly spans the temperature difference
between the hot heat exchanger and the ambient heat exchanger and contains small
pores through which the gas oscillates. A regenerator differs from a stack by a much
smaller pore size.
The pores in a regenerator are small enough that the gas in them is in excellent local
thermal contact with their walls. Asolid matrix such as a pile of ﬁnemesh metal screens
or spherical particles is often used. Proper design causes the gas in the channels to move
toward the hot heat exchanger while the pressure is high, and toward the ambient heat
exchanger while the pressure is low. The time phasing described above is that of a
traveling acoustic wave, which carries acoustic power from ambient to hot. In contrast
4 Introduction
to standingwave engines, acoustic power must be injected into the ambient end of a
regenerator in order to create more acoustic power; the regenerator is an ampliﬁer of
acoustic power. Yazaki et al. [8] demonstrated a travelingwave engine very similar
to that ﬁrst conceived by Ceperley [33–35], with the path length around the toroidal
waveguide nearly equal to 2λ. At about the same time, De Blok [36] and the Los Alamos
group [37, 38] invented a travelingwave engine with the heat exchangers imbedded in
a lumpedacoustic impedance torus much shorter than the wavelength λ.
Backhaus and Swift [37–39], at the Los Alamos National Laboratory, have developed
a heatdriven thermoacoustic refrigerator to liquify natural gas. It is a combination of a
thermoacoustic engine that drives a oriﬁce pulse tube. The engine is a regenator based
on toroidal geometry.
In the process industry large quantities of waste heat are released to the environment
that cannot be reused, mostly because the temperature level is too low. In order to
reuse part of this waste heat, a heat pump is necessary that can provide a temperature
lift to the required temperature levels. A thermoacoustic device can accomplish this
goal. The energy research center of the Netherlands (ECN) [40, 41] develops traveling
wave devices to upgrade waste heat to either process heat or to generate cooling. Their
thermoacoustic devices are a combination of an engine and a heat pump. The waste
heat is used to create a temperature gradient and is converted into acoustic power. The
acoustic power is used to drive a heat pump that upgrades part of the waste heat to a
required temperature level.
The traditional Stirling engine has high efﬁciency, but it has moving parts. The
thermoacousticStirling hybrid engine has reasonably high efﬁciency and very high re
liability, but the toroidal topology needed is responsible for high fabrication costs. Fi
nally, the stackbased standingwave thermoacoustic engine is reliable and costs little to
fabricate, but its efﬁciency is only about 2/3 that of a regeneratorbased system.
1.2.3 Special applications and phenomena
A Rijke tube is similar to the Sondhauss tube, in the sense that it turns heat into sound,
by creating a selfamplifying standing wave. Rijke [42] discovered in 1859 a way of
using heat to sustain a sound in a cylindrical tube, open at both ends. The tube was
oriented vertically and a hot wire gauze was located at the lower half of the tube. The
ﬂow of air past the gauze is a combination of two motions. There is a uniform upwards
motion of the air due to a convection current resulting from the gauze heating up the
air. Superimposed on this is the motion due to the sound wave. Rijke oscillations only
occur in a pipe having both ends open, since the upward motion, due to natural convec
tion past the hot gauze is an essential part of the operation of the Rijke tube. Feldman
produced a literature overview of both the Rijke [43] and the Sondhauss [44] thermoa
coustic phenomena.
A completely different use of thermoacoustics is the separation of gas mixtures by
thermoacoustic waves, which was discovered by Swift and Spoor [45–47]. The super
position of nonzero timeaveraged mole ﬂux on a thermoacoustic wave in a binary gas
mixture in a tube produces continuous mixture separation.
A cascade thermoacoustic engine, developed by Gardner and Swift [48], is a device
in which one standingwave engine and two travelingwave engines are cascaded in
series. Most of the acoustic power is produced in the efﬁcient travelingwave stages.
1.3 Objectives 5
The straightline series conﬁguration is easy to build and allows no Gedeon streaming.
The engine delivers up to 2 kWof acoustic power, with an efﬁciency (the ratio of acoustic
power to heater power) of up to 20%.
Luo et al. [49, 50] have been working on a novel cascade thermoacoustic prime
mover. And currently Hu et al. [51–53] are still working on cascade thermoacoustic
engines. They use much higher frequencies than Gardner and Swift did.
Bauwens et al. have worked on numerical simulations of the ﬂow and heat transfer
in a stack [54, 55] and on a transient theory [56].
1.3 Objectives
The thermoacoustic theory, as developed by Rott and Swift [57], is a linear theory. How
ever, in many situations the amplitudes are so high that the use of the linear theory is
not justiﬁed [38, 58–61]. Therefore we decided to study the effects that appear at high
amplitudes. The nonlinear effects that we are interested in, are:
• vortex shedding at the end of a parallelplate stack,
• dissipation at the ends of a stack due to the sudden change in cross section,
• transition to turbulence inbetween plates,
• streaming.
In order to study these effects we have built a setup to do detailed measurements of
pressure, stack temperature, and local velocity ﬁelds. Since nonlinear effects that we
want to study occur at high amplitudes (≈ 3% [62–67]), we need to build a setup that
can produce such high amplitudes. We want to emphasize that the goal is not to build
a device that is optimized for efﬁciency, but rather to do measurements to gain a better
understanding of different phenomena that are occurring in thermoacoustic devices.
The work is a dual PhDproject, in which Peter in ’t panhuis focuses on the theoretical
and mathematical aspects and we focus more on the practical and experimental part.
An additional goal of this work is to help several Dutch companies, including ECN,
Aster Thermoacoustics, and Shell, that work on thermoacoustics and also partially
funded this project. By getting more insight in thermoacoustics we hope that they will
be able to improve their thermoacoustic devices.
1.4 Thesis outline
In this thesis we will ﬁrst derive a wellknown set of three differential equations that
give a mathematical description of thermoacoustics for low amplitudes. These equa
tions can be applied to many different geometries, including parallelplate stacks, pores,
even various crosssection shapes, and resonator tubes. The equations include viscous
and thermal dissipation, and thermal conduction. Losses due to entrance effects in a
stack, changes in cross section, and streaming are not included. Since the time depen
dency is assumed to be harmonic, higher harmonics are also not included. The set of
equations can only be solved analytically for special cases. For all the other cases we
6 Introduction
wrote a code to solve the equations numerically. The thermoacoustic theory, the code,
and the different types of thermoacoustic devices are discussed in chapter 2.
In chapter 3 we present the experimental setup that is used, with a focus on the
electroacoustics, i.e. the acoustic output of the loudspeaker with respect to the electrical
input. By including the speaker equations in our code we can calculate the pressure
and velocity in the resonator as a function of the speaker voltage and frequency. We
have measured the current through the loudspeaker, its excursion, and also the pres
sure at different positions in the resonator, using microphones. The measurements are
compared with the model.
In chapter 4 we describe the use of a multimicrophone method to determine the
transfer matrix of various stacks. From the transfer matrix elements the Rott functions
can be determined as functions of the frequency. The Rott functions are ﬁtted to an
analytical solution with the pore size and porosity as ﬁtting parameters. Furthermore,
the multimicrophone method is used to determine energy ﬂuxes at both ends of a stack.
This way the dissipated energy in the stack is measured and compared with our model.
The temperature proﬁle in a parallelplate is registered as a function of time, using
32 thermometers that are inserted in the center plate of the stack. Since the stack is not
isolated from the environment, the stack temperature is not uniform over the cross sec
tion. The radial temperature gradients are included in the model, which is discussed in
chapter 5. Temperature measurements in time are compared with our timedependent
model.
In chapter 6 we visualize the ﬂow using Particle Imaging Velocimetry (PIV). A large
part of the chapter discusses the vortex shedding that occurs behind the plate ends.
The ﬂowvisualization measurements are compared with a numerical simulation. In
the center of the stack inbetween two plates the velocity is measured. The velocity
proﬁle is compared with an analytical solution. At sufﬁciently high Reynolds numbers
the proﬁles are expected to deviate from this analytical solution, due to turbulence. We
will measure the transition of laminar to turbulent. The PIV technique is also used to
measure jet streaming and natural convection.
Regeneratorbased devices, consisting of a coaxial loop, are designed by Aster Ther
moacoustics and ECN. It is valuable to know how an oscillatory ﬂow behaves in such a
coaxial loop, which is studied using PIV in chapter 7. Especially, the ﬂow at the sharp
corners is of interest.
We conclude this thesis with the most important conclusions and recommendations.
Bibliography
[1] L. Rayleigh, Theory of Sound, Vol. II. New York: Dover, 1945.
[2] N. Rott, “Damped and thermally driven acoustic oscillations in wide and narrow
tubes,” Journal of Applied Mathematics and Physics, vol. 20, pp. 230–243, 1969.
[3] N. Rott, “Thermally driven acoustic oscillations. part II: Stability limit for helium,”
Journal of Applied Mathematics and Physics, vol. 24, pp. 54–72, 1973.
[4] N. Rott, “Thermally driven acoustic oscillations. part III: Secondorder heat ﬂux,”
Journal of Applied Mathematics and Physics, vol. 26, pp. 43–49, 1975.
[5] N. Rott and G. Zouzoulas, “Thermally driven acoustic oscillations. part IV: Tubes
with variable crosssection,” Journal of Applied Mathematics and Physics, vol. 27,
pp. 197–224, 1976.
[6] N. Rott, “Thermoacoustics,” Advances in Applied Mechanics, vol. 20, pp. 135–175,
1980.
[7] N. Rott, “The inﬂuence of heat conduction on acoustic streaming,” Journal of Ap
plied Mathematics and Physics, vol. 25, pp. 417–421, 1974.
[8] G. Swift, “Thermoacoustic engines,” Journal of the Acoustical Society of America,
vol. 84, pp. 1146–1180, 1988.
[9] G. W. Swift and R. M. Keolian, “Thermoacoustics in pinarray stacks,” Journal of the
Acoustical Society of America, vol. 94, no. 2, pp. 941–943, 1993.
[10] W. Arnott, H. E. Bass, and R. Raspet, “General formulation of thermoacoustics for
stacks having arbitrarily shaped pore cross sections,” Journal of the Acoustical Society
of America, vol. 90, no. 6, pp. 3228–3237, 1991.
[11] J. Wheatley, T. Hoﬂer, G. Swift, and A. Migliori, “An intrinsically irreversible ther
moacoustic heat engine,” Journal of the Acoustical Society of America, vol. 74, pp. 153–
170, 1983.
[12] A. Atchley, T. Hoﬂer, M. Muzzerall, M. Kite, and C. Ao, “Acoustically generated
temperature gradients in short plates,” Journal of the Acoustical Society of America,
vol. 88, no. 1, pp. 251–263, 1990.
8 BIBLIOGRAPHY
[13] A. Atchley, “Analysis of the initial buildup of oscillations in a thermoacoustic
prime mover,” Journal of the Acoustical Society of America, vol. 95, no. 3, pp. 1661–
1664, 1994.
[14] A. Atchley, “Standing wave analysis of a thermoacoustic prime mover below onset
of selfoscillation,” Journal of the Acoustical Society of America, vol. 92, no. 5, pp. 2907–
2914, 1992.
[15] A. Atchley, H. Bass, T. Hoﬂer, and H.T. Lin, “Study of a thermoacoustic prime
mover below onset of selfoscillation,” Journal of the Acoustical Society of America,
vol. 91, no. 2, pp. 734–743, 1992.
[16] A. Atchley and F.M. Kuo, “Stability curves for a thermoacoustic prime mover,”
Journal of the Acoustical Society of America, vol. 95, no. 3, pp. 1401–1404, 1994.
[17] D. Gaitan and A. Atchley, “Finite amplitude standing waves in harmonic and an
harmonic tubes,” Journal of the Acoustical Society of America, vol. 93, no. 5, pp. 2489–
2495, 1993.
[18] P. H. M. W. in ’t Panhuis, S. W. Rienstra, J. Molenaar, and J. J. M. Slot, “Weakly non
linear thermoacoustics for stacks with slowly varying pore crosssections,” Journal
of Fluid Mechanics, vol. 618, pp. 41–70, 2009.
[19] B. Ward and G. Swift, “Design environment for lowamplitude thermoacoustic en
gines,” Journal of the Acoustical Society of America, vol. 95, pp. 3671–3672, 1996.
[20] W. Ward, J. Clark, and G. Swift, Design Environment for Lowamplitude
Thermoacoustic Energy Conversion. Los Alamos National Laboratory,
http://www.lanl.gov/thermoacoustics/DeltaEC.html, 2008.
[21] C. Sondhauss, “Ueber die schallschwingungen der luft in erhitzten glasr¨ ohren un
in gedeckten pfeifen von ungleicher weite,” Annalen der Physik, vol. 79, p. 1, 1850.
[22] K. T. Feldman and R. L. Carter, “A study of heat driven pressure oscillations in a
gas,” Journal of Heat Transfer, vol. 92, pp. 536–541, 1970.
[23] J. C. Wheatley, T. Hoﬂer, G. W. Swift, and A. Migliori, “Understanding some sim
ple phenomena in thermoacoustics with applications to acoustical heat engines,”
American Journal of Physics, vol. 53, pp. 147–162, 1985.
[24] T. J. Hoﬂer, Thermoacoustic refrigerator design and performance. Phd thesis, Physics
Dept., University of California, San Diego, 1986.
[25] M. E. H. Tijani, J. C. H. Zeegers, and A. T. A. M. De Waele, “A gasspring sys
tem for optimizing loudspeakers in thermoacoustic refrigerators,” Journal of Ap
plied Physics, vol. 92, no. 4, p. 2159, 2002.
[26] M. Tijani, Loudspeakerdriven Thermoacoustic Refrigeration. Eindhoven: Universiteits
drukkerij Eindhoven, 2001.
[27] M. E. Poese, R. W. M. Smith, S. L. Garrett, R. van Gerwen, and P. Gosselin,
“Thermoacoustic refrigeration for ice cream sales,” in Proceedings of the 6th Gustav
Lorentzen Natural Working Fluids Conference, 2004.
BIBLIOGRAPHY 9
[28] S. L. Garrett, J. A. Adeff, and T. J. Hoﬂer, “Thermoacoustic refrigeration for space
application,” Journal of Thermophysics and Heat Transfer, vol. 7, no. 4, pp. 595–599,
1993.
[29] W. Slaton and J. Zeegers, “An aeroacoustically driven thermoacoustic heat pump,”
Journal of the Acoustical Society of America, vol. 117, no. 6, pp. 3628–3635, 2005.
[30] W. Slaton and J. Zeegers, “Acoustic power measurements of a damped aeroacous
tically driven resonator,” Journal of the Acoustical Society of America, vol. 118, no. 1,
pp. 83–91, 2005.
[31] W. Slaton and J. Zeegers, “Thermoelectric power generation in a thermoacoustic
refrigerator,” Applied Acoustics, vol. 67, no. 5, pp. 450–460, 2006.
[32] J. Zeegers and W. Slaton, “Aeroacoustic thermoacoustics,” in AIP Conference Pro
ceedings, vol. 838, pp. 416–423, 2006.
[33] P. H. Ceperley, “Apistonless stirling enginethe traveling wave heat engine,” Journal
of the Acoustical Society of America, vol. 66, pp. 1508–1513, 1979.
[34] P. H. Ceperley, “Gresonant travelling wave heat engine,” US Patent 4, pp. 355,517,
1982.
[35] P. H. Ceperley, “Gain and efﬁciency of a short traveling wave heat engine,” Journal
of the Acoustical Society of America, vol. 77, pp. 1239–1244, 1985.
[36] C. M. de Blok, “Thermoacoustic system,” Dutch Patent. International Application
Number PCT/NL98/00515, 1998.
[37] S. Backhaus and G. W. Swift, “A thermoacousticstirling heat engine,” Nature,
vol. 399, pp. 335–338, 1999.
[38] S. Backhaus and G. W. Swift, “A thermoacousticstirling heat engine: Detailed
study,” Journal of the Acoustical Society of America, vol. 107, pp. 3148–3166, 2000.
[39] B. Arman, J. Wollan, G. Swift, and S. Backhaus, “Thermoacoustic natural gas liq
ueﬁers and recent developments,” in Proceedings of the 2003 International Conference
on Cryogenics and Refrigeration, pp. 123–127, 2003.
[40] L. A. Nijeholt, T. J.A., and S. M.E.H., Spoelstra, “Simulation of a travelingwave
thermoacoustic engine using computational ﬂuid dynamics,” Journal of the Acous
tical Society of America, vol. 118, no. 4, pp. 2265–2270, 2005.
[41] M. Tijani and S. Spoelstra, “Study of a coaxial thermoacousticstirling cooler,” Cryo
genics, vol. 48, no. 12, pp. 77–82, 2008.
[42] P. L. Rijke, “On the vibration of the air in a tube open at both ends,” Philosophical
Magazine, vol. 17, pp. 419–422, 1859.
[43] K. T. Feldman, “Reviewof the literature on rijke thermoacoustic phenomena,” Jour
nal of Sound and Vibration, vol. 7, pp. 83–89, 1968.
10 BIBLIOGRAPHY
[44] K. T. Feldman, “Review of the literature on sondhauss thermoacoustic phenom
ena,” Journal of Sound and Vibration, vol. 7, pp. 71–82, 1968.
[45] P. Spoor and G. Swift, “Modelocking of acoustic resonators and its application to
vibration cancellation in acoustic heat engines,” Journal of the Acoustical Society of
America, vol. 106, no. 3 I, pp. 1353–1362, 1999.
[46] P. Spoor and G. Swift, “Thermoacoustic separation of a hear mixture,” Physical
Review Letters, vol. 85, no. 8, pp. 1646–1649, 2000.
[47] G. Swift and P. Spoor, “Thermal diffusion and mixture separation in the acous
tic boundary layer,” Journal of the Acoustical Society of America, vol. 106, no. 4 I,
pp. 1794–1800, 1999.
[48] D. Gardner and G. Swift, “A cascade thermoacoustic engine,” Journal of the Acous
tical Society of America, vol. 114, no. 4 I, pp. 1905–1919, 2003.
[49] H. Ling, E. Luo, W. Dai, and X. Li, “Numerical simulation on a novel cascade ther
moacoustic prime mover,” in Proceedings of the Twentieth International Cryogenic En
gineering Conference, ICEC 20, pp. 333–336, 2005.
[50] E. Luo, E. Dai, and H. Ling, “Basic operation principle of a novel cascade thermoa
coustic prime mover,” in Proceedings of the Twentieth International Cryogenic Engi
neering Conference, ICEC 20, pp. 317–320, 2005.
[51] Z. Hu, Q. Li, X. Xie, G. Zhou, and Q. Li, “Design and experiment on a mini cascade
thermoacoustic engine,” Ultrasonics, vol. 44, no. SUPPL., pp. e1515–e1517, 2006.
[52] Z. Hu, Q. Li, Q. Li, and Z. Li, “A high frequency cascade thermoacoustic engine,”
Cryogenics, vol. 46, no. 11, pp. 771–777, 2006.
[53] Z. Hu, Q. Li, Z. Li, and Q. Li, “Orthogonal experimental study on high frequency
cascade thermoacoustic engine,” Energy Conversion and Management, vol. 49, no. 5,
pp. 1211–1217, 2008.
[54] L. Bauwens, “Oscillating ﬂow of a heatconducting ﬂuid in a narrow tube,” Journal
of Fluid Mechanics, vol. 324, pp. 135–161, 1996.
[55] O. Hireche, C. Weisman, D. BalteanCarls, P. L. Qu´ er´ e, M.X. Franc¸ois, and
L. Bauwens, “Numerical model of a thermoacoustic engine,” Comptes Rendus
Mcanique, vol. 338, no. 1, pp. 18 – 23, 2010.
[56] L. Bauwens, “Thermoacoustics: Transient regimes and singular temperature pro
ﬁles,” Physics of Fluids, vol. 10, no. 4, pp. 807–818, 1998.
[57] G. W. Swift, “Thermoacoustic engines,” Journal of the Acoustical Society of America,
vol. 84, no. 4, pp. 1145–1180, 1988.
[58] G. Swift, A Unifying Perspective for Some Engines and Refrigerators. Melville: Acous
tical Society of America, 2002.
[59] G. Swift, “Thermo acoustic engines and refrigerators,” Physics Today, vol. 48, no. 7,
pp. 22–28, 1995.
BIBLIOGRAPHY 11
[60] M. E. Poese and S. L. Garrett, “Performance measurements on a thermoacoustic
refrigerator driven at high amplitudes,” Journal of the Acoustical Society of America,
vol. 107, no. 5, pp. 2480–2486, 2000.
[61] H. S. Roh, R. Raspet, and H. E. Bass, “Parallel capillarytubebased extension of
thermoacoustic theory for random porous media,” Journal of the Acoustical Society
of America, vol. 121, no. 3, pp. 1413–1422, 2007.
[62] A. Berson and P. BlancBenon, “Nonperiodicity of the ﬂow within the gap of a ther
moacoustic couple at high amplitudes,” Journal of the Acoustical Society of America,
vol. 122, no. 4, pp. EL122–EL127, 2007.
[63] A. Berson, P. BlancBenon, and G. ComteBellot, “A strategy to eliminate all non
linear effects in constantvoltage hotwire anemometry,” Review of Scientiﬁc Instru
ments, vol. 80, no. 4, 2009.
[64] A. Berson, G. Poignand, P. BlancBenon, and G. ComteBellot, “Capture of instan
taneous temperature in oscillating ﬂows: Use of constantvoltage anemometry to
correct the thermal lag of cold wires operated by constantcurrent anemometry,”
Review of Scientiﬁc Instruments, vol. 81, no. 1, 2010.
[65] D. Marx and P. BlancBenon, “Numerical calculation of the temperature difference
between the extremities of a thermoacoustic stack plate,” Cryogenics, vol. 45, no. 3,
pp. 163–172, 2005.
[66] D. Marx and P. BlancBenon, “Computation of the mean velocity ﬁeld above a stack
plate in a thermoacoustic refrigerator,” Comptes Rendus M´ ecanique, vol. 332, no. 11,
pp. 867–874, 2004.
[67] D. Marx and P. BlancBenon, “Numerical simulation of stackheat exchangers cou
pling in a thermoacoustic refrigerator,” AIAA Journal, vol. 42, no. 7, pp. 1338–1347,
2004.
12 BIBLIOGRAPHY
Chapter 2
Theory and model
2.1 General thermoacoustic theory
We derive three basic thermoacoustic differential equations. Two of them relate the
pressure and volumeﬂow rate for a given temperature gradient and geometry. To
gether they form Rott’s wave equations. The third equation is the energy equation,
from which a temperature proﬁle can be determined for a given pressure, volumeﬂow
rate, and geometry. The theory was developed in the seventies by Rott et al. [1–6] and
was later reviewed by Swift [7], for the application of parallel plates. The theory is gen
eral in the sense that it applies to a whole range of different geometries. The size of
the ﬂow channels can vary from smaller than the thermal penetration depth, in case
of regenerators, to a few times the thermal penetration depth, in case of a stack, to big
resonator tubes. The hydraulic radius R
H
, deﬁned as the ratio of the crosssectional
area of the ﬂow A to the wetted perimeter Π, is a measure of the ﬂow channel size. The
shape of the geometry can be parallel plates, circular cylinders, rectangular cylinders, or
any other cylindrical regular shape. The generalization to arbitrary pore geometry was
developed by Arnott et al. [8] using the assumption that the temperature at the bound
aries is constant. In ’t panhuis et al. [9] repeated this generalization, without using this
assumption, by taking into account the heat equation in the solid, and also allowed for
gradual changes of the hydraulic radius of the pore.
In the derivation of the three basic thermoacoustic equations we make the following
major assumptions:
• The relevant parameters, pressure, density, velocity, and temperature can be lin
earized, i.e. can be written as the sum of a mean value and a ﬁrstorder variation.
The assumption that secondorder effects are ignored yields to a major limitation
of this ’linear’ theory to the amplitudes. For highpressure amplitudes (higher
than 10% of the mean pressure, according to literature) the secondorder terms
cannot be neglected.
• The acoustic wave has a single, ﬁxed frequency and is propagating in only one
direction that is along the tube axis. This implies that the hydraulic radius is re
quired to be smaller than the wavelength, to avoid cutoff modes.
• The system is in steady state, i.e. at the end of a cycle the system returns to the
14 Theory and model
same state as in the beginning of the cycle.
• The local timeaverage mean velocity is equal to zero.
In addition to these major assumptions, some additional assumptions that are com
mon in ﬂuid dynamics and thermodynamics, like perfect heat contact between the solid
and gas, noslip condition at wall, the gas is assumed to be a Newtonian ﬂuid and a
perfect gas, the only relevant viscous force to be shear viscosity, and the kinetic energy
to be negligible in the energy equation.
2.1.1 Basic equations
Here we will derive the three basic equations that follow from conservation of mass,
momentum, and energy respectively. The second law of thermodynamics and the ideal
gas law are two additional basic equations. From the conservation of mass the well
known continuity equation follows
∂ρ
∂t
+∇ (ρv) = 0, (2.1)
where ρ is the density, t the time, and v the velocity. From the conservation of mo
mentum and approximating the viscous term by ν∇
2
v, the wellknown NavierStokes
equation
∂v
∂t
+ (v ∇)v = −
∇p
ρ
+ν∇
2
v, (2.2)
can be derived. Here p is the pressure and ν the kinematic viscosity. By combining the
ﬁrst law of thermodynamics with the Eqs. 2.1 and 2.2, Landau and Lifshitz [10] derived
a general equation of heat transfer for ﬂuids
ρT
_
∂s
∂t
+v ∇s
_
= ∇ k∇T + (σ
∇) v, (2.3)
where s represents the entropy per unit mass, k the thermal conductivity, T the temper
ature, and σ
the viscous stress tensor. Using ρTds = −dp +ρc
p
dT (for ideal gases),
with c
p
the speciﬁc isobaric heat capacity, we can rewrite the heat equation to
ρc
p
_
∂T
∂t
+v ∇T
_
−
_
∂p
∂t
+v ∇p
_
= ∇ k∇T + (σ
∇) v. (2.4)
The ideal gas law yields
p = ρR
s
T, (2.5)
where R
s
is the speciﬁc gas constant.
2.1.2 Linearization
The coordinate system for a ﬂow channel with arbitrary cross section is shown in ﬁgure
2.1. The velocity in Cartesian coordinates v = (u, v, w), with u the xcomponent of v.
2.1 General thermoacoustic theory 15
x
z
y
gas area
solid
Figure 2.1: A geometry consisting of a homogeneous ﬂow channel with an arbitrarily shaped
crosssection. The ﬂowchannel length (in x direction) is much larger than its perimeter.
The basic equations can be simpliﬁed by linearizing the relevant variables:
p(x, t) = p
0
+Re¦p
1
(x)e
iωt
¦, (2.6a)
ρ(x, y, z, t) = ρ
0
(x) +Re¦ρ
1
(x, y, z)e
iωt
¦, (2.6b)
T(x, y, z, t) = T
0
(x) +Re¦T
1
(x, y, z)e
iωt
¦, (2.6c)
u(x, y, z, t) = Re¦u
1
(x, y, z)e
iωt
¦, (2.6d)
U(x, t) = Re¦U
1
(x)e
iωt
¦, (2.6e)
where ω is the angular frequency and U
1
is the volume ﬂow rate, i.e. the sectional
integrated velocity over a surface ⊥x. The indexes denote the order. In case of the
pressure, for instance, p
0
is the zerothorder term, which is the time mean value, and
p
1
is the ﬁrstorder term, also called acoustic velocity. The drive ratio, D
r
, is deﬁned as
the ratio of [p
1
[ and p
0
. Note that p
1
is not a function of y and z and that p
0
is constant,
as was shown with smallparameter asymptotics by In ’t panhuis [11]. He also showed
that T
0
and ρ
0
are only x dependent. We will now substitute the linearized variables
into the three basic equations and will then gather all ﬁrstorder terms.
Substituting the linearized density and velocity into Eq. (2.1) yields
iωρ
1
+
∂
∂x
(ρ
0
u
1
) +ρ
0
∂v
∂y
+ρ
0
∂w
∂z
= 0. (2.7)
Using a similar concept for the xcomponent of the momentum equation results in
iωρ
0
u
1
= −
dp
1
dx
+ρ
0
ν
_
∂
2
u
1
∂y
2
+
∂
2
u
1
∂z
2
_
. (2.8)
A linearization of the heat equation, Eq. 2.4, yields
ρ
0
c
p
_
iωT
1
+ u
1
dT
0
dx
_
−iωp
1
= k
_
∂
2
T
1
∂y
2
+
∂
2
T
1
∂z
2
_
, (2.9)
where we assume that k is temperature independent.
16 Theory and model
2.1.3 Velocity proﬁle
First we will derive the velocity proﬁle for the special case of a parallelplate channel
and then for the general case of a ﬂow channel with an arbitrary cross section.
Parallelplate channel:
x
z
y
2y0
l0
gas
solid
center of plate
center of plate
Figure 2.2: Parallelplate geometry. The center of the gas channel is at y = 0 and the boundaries
are at y = ±y
0
. At the centers of the plates a periodic boundary condition in the ydirection is
applied. The plate thickness is 2l
0
.
We will solve the momentum equation, Eq. 2.8, ﬁrst for a parallelplate geometry
(ﬁgure 2.2), and then for an arbitrary geometry. For parallel plates, separated by 2y
0
with y = 0 in the center, the following boundary conditions apply:
u
1
(x, −y
0
) = 0, (2.10a)
u
1
(x, y
0
) = 0. (2.10b)
Since the problem is twodimensional, it is independent of the zposition: u
1
= u
1
(x, y).
We assume to be sufﬁciently far away, i.e. more than twice the displacement length,
from the plate ends. Solving the differential equation, Eq. 2.8, leads to
u
1
=
i
ωρ
0
dp
1
dx
+ C
1
cosh[(1 + i)y/δ
ν
] + C
2
sinh[(1 + i)y/δ
ν
], (2.11)
where C
1
and C
2
are complex constants and δ
ν
the viscous penetration depth, deﬁned
as
δ
ν
=
_
2ν
ω
. (2.12)
From the boundary conditions, Eqs. 2.10a and 2.10b, it follows that
C
1
=
i
ωρ
0
dp
1
dx
1
cosh[(1 + i)y
0
/δ
ν
]
, (2.13a)
C
2
= 0. (2.13b)
This makes the solution for parallel plates
u
1
=
i
ωρ
0
_
1 −
cosh[(1 + i)y/δ
ν
]
cosh[(1 + i)y
0
/δ
ν
]
_
dp
1
dx
. (2.14)
2.1 General thermoacoustic theory 17
General solution for a ﬂow channel:
Other channel geometries (ﬁgure 2.1) lead to different solutions for u
1
, but all of them
can be written as a general solution
u
1
=
i
ωρ
0
[1 −h
ν
(y, z)]
dp
1
dx
, (2.15)
with h
ν
(y, z) a complex function, depending on the channel geometry, the frequency,
and weakly on x due to the temperature dependence of the viscosity. The volume ﬂow
rate is the crosssectional integrated velocity
U
1
=
__
A
u
1
dA =
iA
ωρ
0
[1 − f
ν
]
dp
1
dx
, (2.16)
where A is the crosssectional area and f
ν
is the spatial average of h
ν
and is called the
viscous Rott function
f
ν
=
1
A
__
A
h
ν
(y, z)dA. (2.17)
Eq. 2.16 is an important equation, as it is one of the three equations that is used by our
model that will be discussed in section 2.2. For parallel plates
h
ν
=
cosh[(1 + i)y/δ
ν
]
cosh[(1 + i)y
0
/δ
ν
]
, (2.18a)
f
ν
=
tanh[(1 + i)y
0
/δ
ν
]
(1 + i)y
0
/δ
ν
. (2.18b)
The h
ν
and f
ν
functions of other geometries, cylindrical, triangular, rectangular, and pin
arrays, are derived by Arnott et al. [8].
2.1.4 Thermoacoustic continuity equation
Here we will derive a second equation, the thermoacoustic continuity equation, for a
general pore geometry. T
1
is derived from the linearized heat equation and is used to
determine ρ
1
, using the ideal gas law. ρ
1
is substituted into the linearized continuity
equation, which is integrated over the cross section.
By using the divergence theorem and that v = 0 at the pore boundaries, it can be
shown that
__
A
∂v/∂ydA =
__
A
∂w/∂zdA = 0. This equation can be used to integrate Eq.
2.7 over the cross section, resulting in
iAω¸ρ
1
¸ +
d
dx
(ρ
0
U
1
) = 0, (2.19)
where ¸¸ denotes the crosssectional average. Solving the heat equation (Eq. 2.9) in the
gas, with the boundary condition that the temperature at the solid is constant (T
1
= 0),
results in
T
1
=
(1 −h
κ
)p
1
ρ
0
c
p
−
1
iωA
dT
0
dx
(1 −h
κ
) −P
r
(1 −h
ν
)
(1 − f
ν
)(1 −P
r
)
U
1
, (2.20)
18 Theory and model
where P
r
= c
p
µ/k is the Prandtl number and h
κ
and f
κ
are channelgeometry dependant
functions that are very similar to h
ν
and f
ν
, with the only difference that δ
ν
is replaced
with δ
κ
. For instance, for parallel plates
h
κ
=
cosh[(1 + i)y/δ
κ
]
cosh[(1 + i)y
0
/δ
κ
]
, (2.21a)
f
κ
=
tanh[(1 + i)y
0
/δ
κ
]
(1 + i)y
0
/δ
κ
. (2.21b)
The thermal penetration depth δ
κ
is given by
δ
κ
=
_
2κ
ω
, (2.22)
with κ = k/ρ
0
c
p
the thermal diffusivity. The viscous and thermal penetration depths
are related by the Prandtl number
P
r
=
δ
2
ν
δ
2
κ
. (2.23)
If we take the crosssectional average of Eq. 2.20
¸T
1
¸ =
(1 − f
κ
)p
1
ρ
0
c
p
−
1
iωA
dT
0
dx
(1 − f
κ
) −P
r
(1 − f
ν
)
(1 − f
ν
)(1 −P
r
)
U
1
. (2.24)
For ﬂow channels with a hydraulic radius much larger than both the thermal and
viscous penetration depth (R
H
¸δ
ν
, R
H
¸δ
κ
) the Rott functions approach zero f
ν
→
0, f
κ
→0. In case of a parallelplate stack this can be easily seen by taking y
0
→∞. For
gas parcels sufﬁciently far away from the ﬂowchannel wall h
ν
→ 0, h
κ
→ 0, Eq. 2.20
can be simpliﬁed to
T
1
=
p
1
ρ
0
c
p
−
1
iωA
dT
0
dx
U
1
. (2.25)
From this equation it is easily seen that T
1
=0 if,
dT
0
dx
=
iωA
ρ
0
c
p
p
1
U
1
. (2.26)
This equation can be fulﬁlled in case of a standing wave with a critical temperature
gradient
(∇T)
crit
=
ωA[p
1
[
ρ
0
c
p
[U
1
[
. (2.27)
If [
dT
0
dx
[ = (∇T)
crit
, the gas parcels are at the same temperature as the plate during the
whole cycle, and consequently no heat is exchanged between the plates and the gas.
This would be the case for an inviscid thermoacoustic couple in steady state. Inviscid
standingwave engines have [
dT
0
dx
[ > (∇T)
crit
and inviscid standingwave refrigerators
or heat pumps have [
dT
0
dx
[ < (∇T)
crit
.
2.1 General thermoacoustic theory 19
To solve Eq. 2.24 we need boundary conditions for T
1
at the gas channel walls. For
parallel plates Swift [7] takes into account the heat equation in both the gas and the
solid. For ﬂow channels with arbitrary cross sections this is more difﬁcult. Arnott et
al. [8] assume that the temperature on the boundary of the ﬂow channel is constant.
When the thermal capacity and conductivity of the solid is much larger than that of
the gas, this is a reasonable assumption. In ’t panhuis et al. [9] showed how to solve
differential equation for arbitrary pores without using this assumption. Just like Swift
did for parallel plates, they take into account the heat equation in both the gas and the
solid.
Using Eq. 2.5 in combination with Eq. 2.24, substituting it in Eq. 2.19, and using Eq.
2.16 results in
iAω(
ρ
0
p
0
p
1
−
ρ
0
T
0
¸T
1
¸) + U
1
d
dx
(
p
0
RT
0
) +ρ
0
dU
1
dx
= 0. (2.28)
Substituting Eq. 2.24 for ¸T
1
¸ leads to
iAω(
ρ
0
p
0
p
1
−
(1 − f
κ
)p
1
c
p
T
0
) +
dT
0
dx
(1 − f
κ
) −P
r
(1 − f
ν
)
(1 − f
ν
)(1 −P
r
)
ρ
0
U
1
T
0
−
ρ
0
U
1
T
0
dT
0
dx
+ρ
0
dU
1
dx
= 0.
(2.29)
Using c
p
=
γ
γ−1
p
0
ρ
0
T
0
, combining terms, and dividing by ρ
0
leads to the ﬁnal differential
equation
dU
1
dx
= −
iωA[1 + (γ −1) f
κ
]
γp
0
p
1
+
f
κ
− f
ν
(1 − f
ν
)(1 −P
r
)
dT
0
dx
U
1
T
0
. (2.30)
2.1.5 Total energy ﬂow
The total energy ﬂow
˙
E is the sum of the enthalpy ﬂow and heat ﬂow:
˙
E =
1
2
ρ
0
__
A
Re[h
1
u
∗
1
]dA −(Ak + A
so
k
so
)
dT
0
dx
, (2.31)
with h the speciﬁc enthalpy, A
so
the solid area, k and k
so
, the thermal conduction coefﬁ
cient of the gas and the solid, and
∗
denoting the complex conjugate. For ideal gases
h
1
= c
p
T
1
. (2.32)
A substitution of this into Eq. 2.31 leads to
˙
E =
1
2
ρ
0
c
p
__
A
Re[T
1
u
∗
1
]dA −(Ak + A
so
k
so
)
dT
0
dx
. (2.33)
Substituting Eq. 2.20 and u
∗
1
=
1−h
∗
ν
1−f
∗
ν
U
∗
1
A
into this equation and performing the integra
tion results in
20 Theory and model
˙
E =
1
2
Re
_
p
1
U
∗
1
_
1 −
f
κ
− f
∗
ν
(1 + P
r
)(1 − f
∗
ν
)
__
+
ρ
0
c
p
[U
1
[
2
2Aω(1 −P
2
r
)[1 − f
ν
[
2
Im( f
κ
+ P
r
f
∗
ν
)
dT
0
dx
−(Ak + A
so
k
so
)
dT
0
dx
. (2.34)
This is the third important equation that is used in the model.
2.2 Model
In section 2.1 three 1D differential equations were derived (Eq. 2.16, 2.30, and 2.34),
with three variables (U
1
(x), p
1
(x), and T
0
(x)), and one unknown constant (
˙
E). When the
boundary conditions are known, the system can be solved. In general, the set of three
differential equations cannot be solved analytically. For this reason a numerical code is
developed in Matlab. A number of thermoacoustic numerical codes exist, among which
the famous DeltaE, which was developed by Swift [12]. We have chosen to write our
own code, in collaboration with Wei Dai from the Chinese Academy of Sciences, for the
following reasons:
Flexibility: Having a code of our own provides ﬂexibility. Measured data can be used
as input for the code.
Better insight: We want to gain more insight in the equations and algorithms behind
the code. This way we also learn about the limitations of the code.
Independence: Not being dependent on others.
Easy to use: In Matlab it is easy to write scripts for making good plots and connect
them to the model.
2.2.1 Transfer matrix
A stack with a ﬁxed temperature gradient can be described by the ﬁrst two equations,
Eqs. 2.16 and 2.30,
dp
1
dx
+ R
2
U
1
= 0, (2.35a)
dU
1
dx
+ R
1
p
1
−R
3
U
1
= 0. (2.35b)
Note that p
1
and U
1
are explicit functions of x, whereas the parameters
R
1
= iωA(1 + (γ −1) f
κ
)/γp
0
, (2.36a)
R
2
= iωρ
0
/(1 − f
ν
)A, (2.36b)
R
3
=
( f
κ
− f
ν
)dT
0
/dx
(1 − f
ν
)(1 −P
r
)T
0
, (2.36c)
2.2 Model 21
depend only implicitly on x by the T dependency. A substitution of (2.35a) into (2.35b)
leads to the ’wave’ equation of Rott:
d
2
p
1
dx
2
−R
3
dp
1
dx
−R
1
R
2
p
1
= 0. (2.37)
For constant R
1
, R
2
, and R
3
this differential equation has as general solution
p
1
= (C
1
cosh(R
4
x) + C
2
sinh(R
4
x)) exp(R
3
x/2), (2.38a)
U
1
= −
_
2R
4
C
2
+ R
3
C
1
2R
2
cosh(R
4
x) +
2R
4
C
1
+ R
3
C
2
2R
2
sinh(R
4
x)
_
exp(R
3
x/2),
(2.38b)
where R
4
=
_
R
2
3
/4 + R
1
R
2
and C
1
and C
2
are complex constants. The values of C
1
and
C
2
depend on the boundary conditions:
p
1
= p
L
, U
1
= U
L
, for x = x
L
= 0, (2.39)
where p
L
and U
L
are the pressure and volume ﬂow rate at the left side of the compo
nent. For C
1
= p
L
and C
2
= −
R
3
p
L
+2R
2
U
L
2R
4
both boundary conditions are fulﬁlled. The
pressure and volume ﬂow rate at the righthand side of the component, at x = x
R
, p
R
and U
R
, are linear combinations of p
L
and U
L
and can be written as
_
p
R
U
R
_
= T
_
p
L
U
L
_
. (2.40)
Here T is a transfer matrix given by
T = exp(R
3
x
R
/2)
_
_
cosh(R
4
x
R
) −
R
3
sinh(R
4
x
R
)
2R
4
−
R
2
R
4
sinh(R
4
x
R
)
−
R
1
R
4
sinh(R
4
x
R
) cosh(R
4
x
R
) +
R
3
sinh(R
4
x
R
)
2R
4
_
_
.
(2.41)
When R
3
x
R
¸1 the transfer matrix simpliﬁes to
T =
_
_
cosh(
_
R
1
R
2
x
R
) −
_
R
2
R
1
sinh(
_
R
1
R
2
x
R
)
−
_
R
1
R
2
sinh(
_
R
1
R
2
x
R
) cosh(
_
R
1
R
2
x
R
)
_
_
. (2.42)
Since in this case cosh(R
4
x
R
) = cos(
ω
c
_
1+(γ−1) f
k
1−f
ν
x
R
), the pressure waves (using 2.38a)
can be written as
p
1
(x) = C
+
e
−iξx
+ C
−
e
iξx
, (2.43)
with ξ the complex wave number given by
ξ =
ω
c
¸
1 + (γ −1) f
k
1 − f
ν
. (2.44)
22 Theory and model
2.2.2 Energy equation and temperature
When p
1
, U
1
, and two boundary conditions for the temperature are known, T
0
(x) can
be determined from the energy equation. This equation is solved using a discretization,
as is shown in ﬁgure 2.3. The geometry is equidistant divided in n sections with size ∆x.
In the section i the temperature is T
0,i
, the pressure is p
1,i
, and so on. When considering
the energy ﬂow equation at the dashed contour, it follows that
˙
E(x
L
) =
˙
Q
i
+
˙
E(x
R
), (2.45)
where
˙
Q
i
is the heat ﬂow from the environment.
˙
E(x
L
) and
˙
E(x
R
) are computed from
Eq. 2.30. When ∆x is sufﬁciently small, we can use the approximations dT
0
/dx[
x
L
=
(T
0,i
−T
0,i−1
)/∆x and dT
0
/dx[
x
R
= (T
0,i+1
−T
0,i
)/∆x. We also use the approximations
T
0,x
L
= (T
0,i−1
+ T
0,i
)/2 and T
0,x
R
= (T
0,i
+ T
0,i+1
)/2 for the temperature, and the
pressure and volume ﬂow rate at positions x
L
and x
R
are approximated similarly. Four
T0,i1 T0,i T0,i+1
xL xR
T0,i+2 T0,i2
∆x
Q
i
Figure 2.3: An overview of the discretization for the xdependence that is used for the energy
ﬂow equation.
˙
Q
i
is the heat exchange with the environment and T
0,i
is the discretized tempera
ture.
different boundary conditions are possible: the energy ﬂow at the left hand side of a
component; the energy ﬂow at the right hand side; the temperature at the left hand
side; and the temperature at the right hand side. Out of these four different possible
conditions, two are required to solve the energy ﬂowequation. Eq. 2.45 for 2 ≤ i ≤ n−1
in addition to the two boundary conditions, give us n equations for n unknown T
0
’s.
2.2.3 Iteration
We have a method to determine p
1
and U
1
when T
0
is known and a method to determine
T
0
when p
1
and U
1
are known. But in practice none of them are known. For this reason
an iteration technique is used (ﬁgure 2.4).
step 1 We make a ﬁrst guess of the temperature T
0
(x)
step 2 We calculate p
1
(x) and U
1
(x) using the transfer matrix (Eq. 2.41)
2.3 Thermoacoustic devices 23
step 3 From the calculated p
1
(x) and U
1
(x) we compute T
0
(x) as explained in section
2.2.2
step 4 We compare the computed T
0
(x) with the proﬁle from the previous calculation
step. If the difference is sufﬁciently small, the iteration process is completed, oth
erwise we repeat steps 2 and 3.
3. calculate new T0(x)
2. calculate p1(x), U1(x)
1. initial guess of T0(x)
4. is
T0
new
T0
old

< ?
ε
no
yes
end
Figure 2.4: The schematic of the iteration algorithm.
2.3 Thermoacoustic devices
2.3.1 Introduction
Two types of thermoacoustic devices that make use of the thermoacoustic effect can be
distinguished:
Engines Atemperature difference, T
H
−T
C
, is used to generate power
˙
W
out
, as is shown
in ﬁgure 2.5(a).
Refrigerators or heat pumps Power
˙
W
in
is used to extract heat from one location and
release heat at another location at a higher temperature, as is shown in ﬁgure
2.5(b). In case of a refrigerator the objective is to create cooling power
˙
Q
C
by
keeping T
H
at the environmental temperature and in case of a heat pump the
objective is to generate heating power
˙
Q
H
by keeping T
C
at the environmental
temperature.
From the ﬁrst law of thermodynamics it follows that in steady state
˙
W
in,out
=
˙
Q
H
−
˙
Q
C
.
For engines the efﬁciency is deﬁned as η =
˙
W
out
/
˙
Q
H
. The maximum efﬁciency, which
can only be reached in an ideal device, is called Carnot efﬁciency η
C
, and η
C
= (T
H
−
24 Theory and model
T
C
)/T
H
. For refrigerators the coefﬁcient of performance is deﬁned as COP =
˙
Q
C
/
˙
W
in
and the Carnot coefﬁcient of performance is deﬁned as COP
C
= T
C
/(T
H
−T
C
).
TH
TC
engine Wout
QC
QH
(a) engine
TH
TC
refrigerator
or heat pump
Win
QC
QH
(b) refrigerator or heat
pump
Figure 2.5: Schematic drawing of the working principle of two different types of thermodynamic
devices. The squares represent thermal reservoirs. The two top reservoirs are at temperature T
H
and the two bottom reservoirs at T
C
. The arrows show the energyﬂow directions.
a) An engine converts the heating power
˙
Q
H
at temperature T
H
partially into power
˙
W
out
and
transfer the waste heat
˙
Q
L
to the cold reservoir.
b) The power
˙
W
in
is used to generate a cooling power
˙
Q
L
in case of a refrigerator and a heating
power
˙
Q
H
in case of a heat pump.
The devices can also be categorized in a different way: stackbased devices (section
2.3.2) and regeneratorbased devices (2.3.3). We deﬁne a stack as a geometry for which
the hydraulic radius of the pores is of similar size as the thermal penetration depth,
R
H
· δ
κ
, and a regenerator as a geometry for which the hydraulic radius of the pores
is much smaller than the thermal penetration depth, R
H
¸ δ
κ
. In literature these cat
egories are often referred to as standingwave and travelingwave devices. Since, in
practice, the waves are never purely standing nor purely traveling, this nomenclature
can lead to confusion.
2.3.2 Stackbased devices
We consider a standing wave in a straight resonator, closed at both ends, at the ﬁrst
resonance frequency (ﬁgure 2.6). The pressure and velocity are out of phase. If we
neglect the interaction of the wave with the wall, the wave is a perfect standing wave.
When we follow a gas parcel during one cycle, we can draw a pressuredisplacement,
p −ζ, plot of it (ﬁgure 2.7(a)). The displacement ζ is the position of one gas parcel
during one cycle with respect to its average position x
0
. We write the pressure as
p = p
0
+[p
1
[ cos(ωt +θ
p
), (2.46)
and the displacement as
ζ = [ζ
1
[ cos(ωt +θ
ζ
), (2.47)
2.3 Thermoacoustic devices 25
ωt = 0
ωt = π
pp0
x
ωt = π/2; 3π/2
(a) pressure
ωt = 3π/2
ωt = π/2
u
x
ωt = 0; π
(b) velocity
Figure 2.6: Acoustic pressure and velocity as functions of the position in a straight resonator at
the ﬁrst resonance frequency, at four different phase angles ωt.
with θ a phase angle. A substitution from 2.47 into 2.46 results in two equations for the
pressure as a function of the displacement
p = p
0
+[p
1
[ cos
_
arccos(
ζ
ζ
1
) ±(θ
p
−θ
ζ
)
_
, −1 ≤
ζ
[ζ
1
[
≤ 1. (2.48)
The two equations in (2.48) together describe an ellipse. If we normalize the pressure
p
/
= (p − p
0
)/[p
1
[ and displacement ζ
/
= ζ/[ζ
1
[, the ellipse axis is at an angle of 45
◦
with the ζaxis for [θ
p
−θ
ζ
[ < π/2 and 45
◦
for π/2 < [θ
p
−θ
ζ
[ < π. In case of
[θ
p
−θ
ζ
[ = π/2 the ellipse is a circle and in case of [θ
p
−θ
ζ
[ = π or [θ
p
−θ
ζ
[ = 0 it is a
straight line. The eccentricity of the ellipse yields e = cos(θ
p
−θ
ζ
).
p
0
p0
(a) pζ
p
x0
(b) pζ at various positions
Figure 2.7: Pressuredisplacement plots at different positions, x, in the resonator.
In ﬁgure 2.7(b) the p −ζ plots of parcels at different positions are shown for a stand
ing wave. Since the pressure and displacement are in phase, i.e. θ
p
= θ
ζ
, it follows
that
p = p
0
+
[p
1
[
[ζ
1
[
ζ, (2.49)
26 Theory and model
which is a straight line with steepness [p
1
[/[ζ
1
[ in the pressuredisplacement plot. The
T −ζ plots have a similar shape as the p −ζ plots. For this reason the enthalpy ﬂow
(Eq. 2.31) is zero in a standing wave.
If the pressure and displacement are out of phase, i.e. [θ
p
−θ
ζ
[ = π/2, if follows
that
(
p − p
0
[p
1
[
)
2
+ (
ζ
[ζ
1
[
)
2
= 1, (2.50)
which is the description of a circle in a normalized pressuredisplacement plot.
For the thermoacoustic effect to take place, we need an interaction of the sound wave
with a solid boundary. This is accomplished by installing a stack in the resonator, as is
shown in ﬁgure 2.8.
Even in an ideal engine, performing at Carnot efﬁciency, a nonzero power
˙
W =
(T
H
/T
C
−1)
˙
Q
C
is required. This power is delivered by a driver, which is a loudspeaker
in our conﬁguration (ﬁgure 2.8a), delivering a power
˙
W
in
. The power is converted to an
acoustic energy ﬂow
˙
E
in
in the resonator. As a consequence a travelingwave compo
nent is added to the sound wave. The cold part of the stack is connected to a reservoir
at a temperature T
C
using a heat exchanger and the hot part is connected to a reservoir
at temperature T
H
. The sound wave transfers an energy ﬂow
˙
Q
C
from the cold to the
hot reservoir.
For the engine (ﬁgure 2.8b), the temperature gradient is used to generate an acoustic
power
˙
W
out
. A heating power
˙
Q
H
from the hot reservoir is required to sustain this
temperature gradient.
˙
Q
H
is partially converted to acoustic power and ﬂows partially
to the cold (environmental) reservoir. The acoustic power
˙
W
out
can be used e.g. to drive
a piston.
Now we will follow a gas parcel that is oscillating near a stack plate, at a distance
2δ
κ
. The displacement and pressure as functions of time during one cycle are shown
in ﬁgure 2.9. In a standing wave the pressure is out of phase with the velocity and in
phase with the displacement. We will explain the working of a stackbased device from
a Lagrange point of view. In ﬁgure 2.9 four moments during a cycle are indicated by A,
B, C, and D. In ﬁgure 2.10 the thermodynamic processes occurring inbetween the four
succeeding moments in time are schematically shown. In the left column a gas parcel in
a refrigerator and in the right column a parcel in an engine is shown. In both devices a
positive temperature gradient is present in the solid. For a refrigerator or a heat pump
the temperature gradient is lower than the critical temperature gradient, which is given
by Eq. 2.27, and in the engine the temperature gradient is lower than (∇T)
crit
. The gas
parcel is continuously moving and exchanging heat with the solid, but for simplicity we
neglect the parcel movement in steps B→C and D→A, and neglect the heat exchange in
steps A→B and C→D.
We will ﬁrst discuss the four steps in the refrigerator or heat pump (left column):
• Step A→B: The parcel moves to the left while the pressure decreases, which causes
the parcel to expand and cool down.
• Step B→C: The parcel is at its leftmost position and the right half of the resonator
is at its minimum pressure. The temperature of the solid at the same xposition is
higher than that of the parcel, resulting in heat ﬂow δq from the solid to the gas
parcel.
2.3 Thermoacoustic devices 27
TH
TC
Ein
QC
QH
E
hot heat
exchanger
cold heat
exchanger
stack speaker
Win
(a) Stackbased refrigerator/heat pump
TH
TC
Eout
QC
QH
E
hot heat
exchanger
cold heat
exchanger
stack
piston
Wout
(b) Stackbased engine
Figure 2.8: A schematic overview of a loudspeakerdriven refrigerator or heat pump (a) and a
thermoacoustic engine driving a piston.
A B C D
pp0
t
pp0
Figure 2.9: The displacement and pressure as functions of time during one cycle.
28 Theory and model
• Step C→D: The parcel moves to the right while the pressure increases, which
causes the parcel to compress and heat up.
• Step D→A: The parcel is at its rightmost position and the right half of the resonator
is at its maximum pressure. The temperature of the solid at the same xposition is
lower than that of the parcel, resulting in heat ﬂow δq from the gas parcel to the
solid.
Now we will give a stepwise description of the processes in an engine (right col
umn):
• Step A→B: The parcel moves to the left while the pressure decreases, which causes
the parcel to expand and cool down.
• Step B→C: The parcel is at its leftmost position and the left half of the resonator
is at its maximum pressure. The temperature of the solid at the same xposition is
lower than that of the parcel, resulting in heat ﬂowδq from the parcel to the solid.
• Step C→D: The parcel moves to the right while the pressure increases, which
causes the parcel to compress and heat up.
• Step D→A: The parcel is at its rightmost position and the right half of the resonator
is at its minimum pressure. The temperature of the solid at the same xposition
is lower than that of the parcel, resulting in heat ﬂow δq from the solid to the gas
parcel.
2.3 Thermoacoustic devices 29
column 1: column 2:
refrigerator or heat pump engine
A→B: movement, pressure decrease A→B: movement, pressure decrease
δq δq
B→C: heat absorption, expansion B→C: heat rejection, compression
C→D: movement, pressure increase C→D: movement, pressure increase
δq δq
D→A: heat rejection, compression D→A: heat absorption, expansion
Figure 2.10: Moving with a gas parcel during one cycle in a refrigerator (column 1) and an
engine (column 2).
30 Theory and model
2.3.3 Regeneratorbased devices
Aregeneratorbased engine is shown in ﬁgure 2.11. The right openend of the Tjunction
is connected to a large resonator. The size of the loop is much smaller than the wave
length. The regenerator acts like an ampliﬁer in the ideal case. The volume ﬂow rate at
the hot side of an ideal regenerator is ampliﬁed by a factor T
H
/T
C
in comparison with
the cold side. Since the pores in the regenerator are relatively small (R
H
¸ δ
κ
), the
ﬂow through the regenerator could lead to a signiﬁcant viscous dissipation. To reduce
this effect the loop geometry is designed in such a way that the volume ﬂow rate U
1
is
relatively small at the position of the regenerator. In many practical regeneratorbased
devices [p
1
/U
1
ρ
0
c[ is inbetween 10 and 20 at the position of the regenerator [13]. Since
only the travelingwave component of the sound wave have a useful effect on the per
formance of a regeneratorbased device, in practice p
1
and U
1
are in phase at the location
of the regenerator. To fulﬁll these requirements a change in crosssection is included in
the feedback loop (inertance to compliance).
The heating power
˙
Q
H
is converted to an acoustic energy ﬂow
˙
E
out
, which is split
up in the Tjunction. One part of
˙
E
out
ﬂows into the resonance tube, where it is used to
drive a piston, producing power
˙
W
out
. The other part of
˙
E
out
ﬂows into the inertance,
through the compliance, and is rejected as wasted heat
˙
Q
C
into the heat sink T
C
. In
an ideal regenerator the enthalpy ﬂow is zero and
˙
E in the regenerator only consists of
conduction.
QC
QH
E
hot heat
exchanger
cold heat
exchanger
regenerator
TH
TC
piston
intertance
compliance
Tjunction
environmental
heat exchanger,
flow straightener
thermal
buffer tube
resonance tube
Ein
Eout
Wout
Figure 2.11: A schematic drawing of a regeneratorbased engine.
Different geometries can be used for regeneratorbased engines. In chapter 6 a coax
ial design, developed by De Blok [14], is discussed.
2.3 Thermoacoustic devices 31
T
x0
cold heat
exchanger
hot heat
exchanger
regenerator
thermal
buffer tube
compliance
Figure 2.12: The temperature T of gas parcels as a function of their displacement, during a
cycle, at different positions x
0
. In the compliance the mean temperature T
0
as function of x
0
is constant, in the regenerator it increases linearly, and in the thermal buffer tube it decreases
slowly.
Regeneratorbased devices are often referred to as travelingwave devices since the
pressure and velocity are ideally in phase in the loop. As a consequence the waves
transfer energy and lead to an enthalpy ﬂow. In ﬁgure 2.12 the Tζ plots in and nearby
the regenerator are shown. In the compliance T and ζ are out of phase, resulting in
ellipses in the Tζ plots. Since the parcels move towards the regenerator with higher
temperature and enthalpy than when they move away from the regenerator, this leads
to a positive enthalpy ﬂow in the compliance.
In an ideal regenerator the gas parcels have the same temperature as the solid. In the
Tζ plot this results in straight lines with steepness equal to ∂T/∂x of the solid. Since
the gas parcels have the same temperature when moving from the hot to the cold side
as from the cold to the hot side, the enthalpy ﬂow in an ideal regenerator is zero. The
difference in the energy ﬂow between compliance and regenerator is wasted to the cold
reservoir at temperature T
C
in the cold heat exchanger. In the thermal buffer tube T and
ζ are out of phase, just as in the compliance, which corresponds to positive enthalpy
ﬂow
˙
W
out
, that is taken from the hot reservoir. The heating power
˙
Q
H
is larger than the
cooling power
˙
Q
C
and the difference is converted to power to drive a piston at the end
of the resonator tube.
In a regeneratorbased device the enthalpy ﬂow is zero inside the regenerator and
nonzero in all other parts of the device. Compare this to a stackbased device, in which
the enthalpy ﬂow out of the stack is close to zero, whereas the enthalpy ﬂow into the
stack is signiﬁcant. For both devices the difference in enthalpy ﬂow between into and
out of the stack or regenerator leads to heat ﬂows in and out of the heat exchangers.
The heat transfer between solid and gas is imperfect in a stack, inevitably leading to an
entropy production. In an ideal regenerator the heat transfer between solid and gas is
adiabatic. By avoiding the entropy production due to heat transfer, regeneratorbased
devices potentially have a higher efﬁciency.
32 Theory and model
Bibliography
[1] N. Rott, “Damped and thermally driven acoustic oscillations in wide and narrow
tubes,” Journal of Applied Mathematics and Physics, vol. 20, pp. 230–243, 1969.
[2] N. Rott, “Thermally driven acoustic oscillations. part II: Stability limit for helium,”
Journal of Applied Mathematics and Physics, vol. 24, pp. 54–72, 1973.
[3] N. Rott, “Thermally driven acoustic oscillations. part III: Secondorder heat ﬂux,”
Journal of Applied Mathematics and Physics, vol. 26, pp. 43–49, 1975.
[4] N. Rott and G. Zouzoulas, “Thermally driven acoustic oscillations. part IV: Tubes
with variable crosssection,” Journal of Applied Mathematics and Physics, vol. 27,
pp. 197–224, 1976.
[5] N. Rott, “Thermoacoustics,” Advances in Applied Mechanics, vol. 20, pp. 135–175,
1980.
[6] N. Rott, “The inﬂuence of heat conduction on acoustic streaming,” Journal of Ap
plied Mathematics and Physics, vol. 25, pp. 417–421, 1974.
[7] G. Swift, “Thermoacoustic engines,” Journal of the Acoustical Society of America,
vol. 84, pp. 1146–1180, 1988.
[8] W. P. Arnott, H. E. Bass, and R. Raspet, “General formulation of thermoacoustics
for stacks having arbitrarily shaped pore cross sections,” Journal of the Acoustical
Society of America, vol. 90, no. 6, pp. 3228–3237, 1991.
[9] P. H. M. W. in ’t panhuis, S. W. Rienstra, J. Molenaar, and J. J. M. Slot, “Weakly non
linear thermoacoustics for stacks with slowly varying pore crosssections,” Journal
of Fluid Mechanics, vol. 618, pp. 41–70, 2009.
[10] L. Landau and E. Lifshitz, Fluid Mechanics. Pergamon, 1982.
[11] P. in ’t panhuis, Mathematical aspects of Thermoacoustics. PhD thesis, Einhoven Uni
versity of Technology, 2009.
[12] B. Ward and G. Swift, “Design environment for lowamplitude thermoacoustic en
gines,” Journal of the Acoustical Society of America, vol. 95, pp. 3671–3672, 1996.
[13] S. Backhaus and G. W. Swift, “A thermoacousticstirling heat engine: Detailed
study,” Journal of the Acoustical Society of America, vol. 107, pp. 3148–3166, 2000.
[14] de Blok and van Rijt, “(wo/1999/020957) thermoacoustic system,” 1999.
34 BIBLIOGRAPHY
Chapter 3
Electroacoustics
3.1 Introduction
Here we will study the coupling of a loudspeaker to an acoustic setup. First we will
model the loudspeaker behavior by coupling two electric variables, the voltage over
and current through the coil, to two mechanical variables, the coil movement and the
Lorentz force, which are coupled to two acoustic variables, the pressure and the volume
ﬂow rate. The electroacoustical model in addition to the thermoacoustic model (chapter
3.2) allows us to model a complete setup.
To study nonlinearities we had to build an experimental setup in which high drive
ratios (up to 20%) could be reached. We choose the, in thermoacoustics commonly used,
setup of a loudspeaker connected to a resonator. In the development of this setup the
model was used to predict the attainable drive ratios. The setup was to be used for
acoustic measurement as well as ﬂow visualization measurements and had to fulﬁll the
following requirements:
• sufﬁciently high drive ratios, up to 20%,
• suitable for acoustic measurement,
• suitable for ﬂow visualization measurements,
• ﬂexible in frequency and amplitude,
• ﬁt into a lab of 4.2 m by 3.5 m.
The experimental setup is described and explained in section 3.3. Once the setup was
completed we validated our model (section 3.4). Besides empty resonators also res
onators with parallelplate stacks are studied.
In stackbased thermoacoustics heatpumps or refrigerators are often driven by
movingcoil loudspeakers. Therefore it is important to have a complete model of a
driver coupled to a geometry. Bailliet et al. [1] used linear speaker equations to cal
culate the speaker amplitudes (current, displacement, and force) for known frequency,
input voltage and acoustic impedance of the geometry coupled to the speaker. Tijani et
al. [2] introduced a concept of using an adjustable back volume to match the mechani
cal resonance frequency with the acoustical resonance frequency. Marx et al. [3] took a
36 Electroacoustics
closer look at the agreement between the speaker model and experimental results, but
only for an empty resonator. We performed a similar study, but with a stack enclosed in
the resonator, to test a model based on transfer matrices.
3.2 Theory and model
3.2.1 Speaker equations
We consider a loudspeaker, driven with a constant angular frequency ω at a voltage V.
The behavior of the loudspeaker depends on the acoustical impedance of the system it
is driving. The speaker is connected to a thermoacoustic device with a known acoustic
impedance. The acoustic impedance is deﬁned as the ratio of pressure to volume ﬂow
rate at the speaker membrane position [4]
Z
ac
=
p
1
U
1
. (3.1)
To reduce sound waves leaking to the environment, the loudspeaker is enclosed in a
back enclosure.
By using Newton’s second law and adding all the forces acting on the loudspeaker
coil we ﬁnd the mechanical equation describing the motion of the coil
M
me
d
2
ζ
dt
2
= F −R
me
dζ
dt
−[k
me
+ k
b
]ζ − A
2
c
Z
ac
dζ
dt
, (3.2)
where M
me
is the moving mass, ζ the coil displacement, F the Lorentz force, R
me
the
friction constant, k
me
the spring constant of the speaker, k
b
the gas spring constant of
the back enclosure, A
c
the cone area, and Z
ac
the acoustical impedance of the geometry
connected to the speaker. The Lorentz force is given by
F = B
l
I, (3.3)
with B
l
the motor force factor and I the current through the coil. The spring constant of
the back enclosure is given by
k
b
=
γp
0
A
2
c
V
b
, (3.4)
where V
b
is the average volume of the back enclosure.
For determining an electric equation the speaker is modeled as a resistance and coil
in series, and the total voltage is a superposition of the speaker voltage V and the in
duced voltage V
ind
V + V
ind
= R
el
I + L
el
dI
dt
, (3.5)
with R
el
the DC resistance and L
el
the selfinductance of the coil. The induced voltage
due to the coil movement yields
V
ind
= −B
l
dζ
dt
, (3.6)
3.2 Theory and model 37
with u
c
= dζ/dx the coil velocity.
We linearize the speaker variables analogously to Eqs. 2.6
V(t) = Re¦V
1
e
iωt
¦, (3.7a)
I(t) = Re¦I
1
e
iωt
¦, (3.7b)
F(t) = Re¦F
1
e
iωt
¦, (3.7c)
ζ(t) = Re¦ζ
1
e
iωt
¦, (3.7d)
u
c
(t) = Re¦u
c1
e
iωt
¦. (3.7e)
We deﬁne the mechanical impedance as the Lorentz force to velocity ratio
Z
me
=
F
1
u
1
. (3.8)
The mechanical impedance can be found by substituting the linearized variables from
Eq. 3.7
Z
me
= R
me
+ i (M
me
ω−[k
me
+ k
b
]/ω) + A
2
c
Z
ac
. (3.9)
The total effective electrical impedance is deﬁned as the voltage to current ratio
Z
el
=
V
1
I
1
. (3.10)
A substitution of Eqs. 3.7 into Eq. 3.5 results in
Z
el
= R
el
+ iωL
el
+ B
l
u
c1
I
1
. (3.11)
Using Eqs. 3.8 and 3.3 the last term of 3.11 can be rewritten
Z
el
= R
el
+ iωL
el
+
B
2
l
Z
me
. (3.12)
For a given voltage using Eq. 3.10 the current can be calculated. Next, using Eq.
3.3, the Lorentz force can be determined and by using Eq. 3.8 the coil velocity can be
computed. Multiplying the coil velocity with the cone area results in the volume ﬂow
rate and Eq. 3.1 gives us the pressure.
3.2.2 Acoustic impedance
If all parameters are known the speaker equations can be solved. The only unknown
parameter is Z
ac
, which depends on the geometry that the speaker is coupled to. This
geometry can consist of multiple components varying in diameter, length, shape and
temperature distribution. To determine the behavior of the system we need to deter
mine the acoustic impedance of the complete geometry. Our method to do this is to
calculate the transfer matrix of each individual component and then by a matrix multi
plication determining the total transfer matrix.
38 Electroacoustics
The transfer matrix T relates the p −u vectors at both sides of a component (x
L
and
x
R
) and is given by Eq. 2.41. In this chapter the inﬂuence of the temperature gradient
on the acoustic equations is neglected. In this case the simpliﬁed transfer matrix of Eq.
2.42 can be used. The total transfer matrix of a system of n components is
T
total
= T
1
T
2
. . . T
n
. (3.13)
If the right end of the geometry is closed (U
R
= 0), the pressure and volume ﬂow rate
have to fulﬁll the condition
T
21
p
L
+ T
22
U
L
= 0, (3.14)
which results in the impedance at the speaker side of the geometry
Z
ac
=
p
1
U
1
¸
¸
¸
¸
x=0
= −T
total,22
/T
total,21
. (3.15)
3.3 Experimental Setup
The setup consists of a movingcoil loudspeaker
1
enclosed in a cylindrical back volume,
V
b
, as is shown in ﬁgure 3.1. The speaker parameters are presented in table 3.1. The back
volume is connected to a horn by a ﬂexible rubber membrane to reduced the inﬂuence
of mechanical speaker vibrations on the resonator. The inner diameter of the speaker
side of the horn D
0
= 0.356 m, the inner diameter at the resonatortube side D
end
= 60
mm, and the length L
horn
= 1.01 m. The resonator tube has an inner diameter, D
tube
, of
60 mm and a length of 1.707 m.
Table 3.1: Speaker parameters.
Resistance R
el
2.95 Ω
Self inductance @ 1 kHz L
el
1.06 mH
Motor force factor B
l
15.39 Tm
Moving mass M
me
248.15 g
Spring constant k
me
6.535 N/mm
Friction constant R
me
5.73 Ns/m
Resonance frequency f
res
25.2 Hz
Cone area A
c
0.078 m
2
Maximum excursion ζ
max
20.32 mm
Back volume V
b
60.3 L
At different locations in the tube a parallelplate stack can be placed. On both sides
of the stack three microphones are mounted with their measurement surface at the inner
tube wall. These piezoresistive pressure transducers
2
are connected to an Endevco DC
differential voltage ampliﬁer
3
. The microphone positions, stack position and endplate
1
JBL W15GTi 15
//
woofer
2
Endevco 8510B2
3
Endevco 136 DC
3.3 Experimental Setup 39
V
I
0
Lhorn
xend
Vback
p1 p2 p4 p5 p6
x2 x1 x6 x5 x3 x4 xL xR
Dend
D0
p3
Figure 3.1: A schematic drawing of the experimental setup (not to scale). A loudspeaker,
enclosed in a back enclosure, is connected to a conical horn. Inside the resonator tube a parallel
plate stack is located. Six microphones, p
1
. . . p
6
, are mounted in the tube wall. The voltage over
and current through the speaker coil are registered. The conical horn starts with diameter D
0
and ends with the tube diameter D
end
.
position are presented in table 3.2.
Table 3.2: Microphone and stackend positions.
x
1
1.210 m
x
2
1.410 m
x
3
1.860 m
x
L
2.049 m
x
R
2.251 m
x
4
2.367 m
x
5
2.517 m
x
6
2.667 m
x
end
2.717 m
The horn diameter varies as function of the axial position according to four different
curves:
D
0
, 0 ≤ x < a
0
(3.16)
a
1
cosh([x −a
0
]/a
2
) + a
3
, a
0
≤ x < 0.14 m (3.17)
tan(18
o
)(a
4
+ a
0
−x) + a
5
, 0.14 m ≤ x < 0.29 m (3.18)
D
end
cosh((L
horn
−x)/a
6
), 0.29 m ≤ x ≤ L
horn
(3.19)
40 Electroacoustics
with D
0
= 0.356 m, a
0
= 0.050 m, a
1
= 0.0080 m, a
2
= 0.04487 m, a
3
= 0.364 m, a
4
=
0.240 m, a
5
= 0.236 m, a
6
= 0.351 m, and L
horn
= 1.01 m. Both D(x) and dD/dx have no
discontinuities, in order to avoid boundarylayer separation.
The displacement of the speaker membrane is measured using a laser triangulation
sensor
4
. The current through the speaker is recorded as well as the voltage, using a
voltage divider. Using the temperature dependance of the DCresistance of the coil, we
can estimate the coil temperature. The coil resistance is determined by applying a small
DC current through the coil.
A sinusoid with a certain amplitude and frequency is generated by a function gen
erator and, after ampliﬁcation, used as the input for the loudspeaker. All measured
variables (six microphones, voltage, current, displacement) are recorded using a data
acquisition card at a frequency of 200 kHz. The amplitudes and phase angles (in relation
to a reference signal) of all signals are determined using a digital lockin function, with
the functiongenerator signal as a reference signal.
Inside the resonator tube a stack of parallel plates made of perspex is placed. The
plate distance is 1.0 mm, the plate thickness is 1.0 mm, and the stack length x
R
− x
L
=
206.0 mm.
3.4 Results
3.4.1 Empty resonator
As the acoustics department of Philips have experience with loudspeakers, we have
cooperated in the modeling of the speaker and resonator. Together with Ouweltjes from
Philips [5] we have calculated the pressures and velocities as function of the position in
the resonator and as function of the frequency.
We use Eq. 2.42 to determine the transfer matrix of the resonator tube. For the f
ν
and f
κ
functions we use the boundary layer approximation. The transfer matrix of the
horn is more difﬁcult to calculate, since its diameter is not constant. For this reason
we divided the horn into 200 sections with length L
horn
/200. The sections are modeled
as tubes with constant diameters. The diameters of the sections are chosen in such a
way that the horn shape is approximated. By multiplying the transfer matrices of the
individual sections, the transfer matrix of the whole horn is determined.
This method of discretization of the horn diameter is tested for an exponential cone,
for which the transfer matrix is known analytically and works very well (within 0.1%
accuracy).
By multiplying all individual transfer matrices, we calculate the transfer matrix of
the whole resonator T
total
which can be used to determine the acoustical impedance Z
ac
from Eq. 3.15. By coupling the acoustic impedance of the geometry to the electroacous
tical model, we can calculate the pressure and velocity amplitudes at every position in
the resonator for a speciﬁc frequency range (10 to 400 Hz in this case). We use a color
plot to show both the frequency and position dependency in one ﬁgure 3.2. In ﬁgure
3.2a the pressure amplitude [p
1
[ is shown and in ﬁgure 3.2b the velocity amplitude. The
colors are a measure of the amplitude, as is shown by the color bars to the right of the
plots.
4
LMI LDSc90/45
3.4 Results 41
x [m]
f
[
H
z
]
0 1 2
10
1
10
2
p
1
[
k
P
a
]
10
10
10
10
3
5
1
1
(a) pressure amplitude [p
1
[
x [m]
f
[
H
z
]
0 1 2
10
1
10
2
u
1
[
m
/
s
]
10
10
10
10
3
5
1
1
(b) velocity amplitude [u
1
[
Figure 3.2: Numerical calculation of the pressure amplitude (a) and velocity amplitude (b)
as function of the frequency f , logarithmically on the vertical axis, and of the position in the
resonator x, linearly on the horizontal axis. The colors are a measure of the amplitude on a
logarithmical scale, as is shown by the color bars to the right of the plots. The white dashed
vertical lines represent the position x = L
horn
.
42 Electroacoustics
The ﬁrst two resonance frequencies can be easily recognized in both ﬁgure 3.2a and
3.2b by the horizontal lines of high intensity (45 Hz and 133 Hz). The exponential horn
is located inbetween x = 0 m and x = 1.01 m. At the resonance frequencies the nodes
and antinodes in both the pressure and velocity can be clearly observed. From these
plots we can also easily observe for which positions and frequencies we can reach the
highest velocities and the highest pressures.
3.4.2 Fixed speaker voltage
First we consider a resonator with a parallelplate stack enclosed in it with a plate thick
ness of 1 mm and a plate distance of 3 mm. In ﬁgures 3.3 and 3.4 the measured data are
compared with the model. In ﬁgures 3.3(a)3.3(d) the amplitudes of the speaker voltage,
speaker current, voltagetocurrent ratio, speaker excursion and in ﬁgures 3.4 the pres
sure amplitudes at four positions are shown as functions of the frequency. The dashed
plots represent the measured data and the solid curves the model.
0 50 100 150 200
0
2
4
6
8
f [Hz]

V
1

[
V
]
(a) [V
1
[
0 50 100 150 200
0
0.5
1
1.5
2
2.5
f [Hz]

I
1

[
A
]
measurement
model
(b) [I
1
[
0 50 100 150 200
0
10
20
30
40
50
f [Hz]

V
1
/
I
1

[
V
/
A
]
(c) [V
1
/I
1
[
0 50 100 150 200
0
0.5
1
1.5
f [Hz]

ζ
1

[
m
m
]
(d) excursion [ζ
1
[
Figure 3.3: Amplitudes of (a) voltage, (b) current, (c) voltagetocurrent ratio, and (d) cone
excursion as functions of the frequency. Measurements are plotted dashed and the model solid.
3.4 Results 43
0 50 100 150 200
0
0.5
1
1.5
2
f [Hz]

p
1
,
1

[
k
P
a
]
measurement
model
(a) pressure p
1,1
0 50 100 150 200
0
0.5
1
1.5
2
2.5
3
f [Hz]

p
1
,
2

[
k
P
a
]
(b) pressure p
1,2
0 50 100 150 200
0
0.5
1
1.5
2
2.5
f [Hz]

p
1
,
3

[
k
P
a
]
(c) pressure p
1,3
0 50 100 150 200
0
1
2
3
4
5
f [Hz]

p
1
,
4

[
k
P
a
]
(d) pressure p
1,4
Figure 3.4: The pressure amplitudes [p
1
[ measured with microphones 1 to 4 at positions x
1
to
x
4
as functions of the frequency. Measurements are plotted dashed and the model solid.
44 Electroacoustics
In the frequency range 13 to 24 Hz, the setup vibrated heavily. Therefore no mea
surements were made in this frequency range. The ampliﬁer driving the speaker is un
able to maintain a constant V
1
of 6.8 V for low frequencies (ﬁgure 3.3(a)). The measured
voltage is used as an input parameter for the model. The measured [V
1
/I
1
[ (ﬁgure 3.3(c))
is in good agreement with the model, especially for relatively low frequencies (smaller
than 120 Hz). At relatively high frequencies (higher than 150 Hz) the calculated value
is 20% lower than the measured one. The measured displacement corresponds well
with the model, except in the frequency range 25 to 75 Hz. The discrepancy is probably
caused by the vibrations of the setup. The measured pressures agree very well with the
model, with the exception that the measured peak values are lower than the computed
ones. The deviations at the peak values are probably caused by nonlinear effects. We
will look into the measurement at the resonance frequencies in more detail in the next
subsection.
3.4.3 Fixed frequency
The ﬁrst resonance frequency is 44 Hz. We performed an amplitude sweep from 1 to
52 V at 44 Hz (ﬁgure 3.5). For the ﬁxedfrequency measurements a different stack, with
plate thickness and plate distance both 2 mm, is used.
According to the theory (section 3.2.1), all variables are expected to be linear with the
speaker voltage. The speaker current (ﬁgure 3.5a) is linear with the speaker voltage, but
the other variables (displacement and pressures, ﬁgure 3.5bd) are strongly nonlinear,
which is in agreement with Marx et al. [3]. The nonlinearity can be caused by various
effects:
1. nonlinearities in the electric equation of the speaker, e.g. the heating of the coil,
leading to a higher electric resistance.
2. nonlinearities in the mechanical equation of the speaker, e.g. an increase of the
friction coefﬁcient as the cone excursion increases.
3. nonlinearities of the sound wave at high drive ratios effecting the acoustical
impedance.
4. small changes in the resonance frequency, due to the heating up of the gas.
Since the current is very linear with the speaker voltage, nonlinearities in the electric
equation can not be the cause of the nonlinearity of the pressure and cone excursion
with the speaker voltage. Also from measurement of the DC resistance, it was found
not to increase signiﬁcantly. Therefore we can conclude that the effect of the coil heating
can be neglected.
Small changes in the resonance frequency can lead to large changes in the measured
pressures, when measuring at a peak that is very steep. When the resonance frequency
changes only a little bit (1 Hz for instance) the system is not measuring at its resonance
frequency, resulting in a much lower pressure amplitude. When we increase the speaker
voltage, more power goes into the system, resulting in a higher gas temperature, leading
to a higher resonance frequency. While the speaker frequency remains at 44 Hz and
the resonance frequency grows above 44 Hz, the difference between measurement and
calculation increases. In theory this effect could be an explanation of the discrepancy
3.4 Results 45
0 20 40 60
0
2
4
6
8
10
V
1
 [V]

I
1

[
A
]
measurement
model
(a) current [I
1
[
0 20 40 60
0
2
4
6
8
V
1
 [V]

ζ
1

[
m
m
]
(b) excursion [ζ
1
[
0 20 40 60
0
0.05
0.1
0.15
0.2
V
1
 [V]

p
2

/
p
0
(c) pressure [p
1,2
[
0 20 40 60
0
0.1
0.2
0.3
0.4
V
1
 [V]

p
6

/
p
0
(d) pressure [p
1,6
[
Figure 3.5: The current, membrane excursion and pressures as functions of the speaker voltage.
The plate thickness and plate distance of the stack are both 2 mm. Measurements are plotted
dashed and the model solid.
46 Electroacoustics
between measurements and theory at high amplitudes in ﬁgure 3.5. To verify whether
the resonance frequency truly changes at higher amplitudes we have compared two
different measurements of the pressure amplitudes as function of the frequency at two
different but ﬁxed speaker voltages. We do not see any changes in resonance frequency
of the two measurements, thus we conclude that this effect can not be the explanation
of the discrepancy in ﬁgure 3.5.
Two options that can explain the discrepancy remain. Additional measurements at
100 Hz are performed to determine which of these two nonlinear mechanisms is domi
nating. In ﬁgure 3.6a the pressure is shown as functions of the speaker voltage for 100
Hz. Since the resonator is not in resonance at this frequency, the pressure amplitudes
are signiﬁcantly lower than at 44 Hz. We expect the nonlinear acoustic effects to reduce.
The measured pressure is still nonlinear with the speaker voltage, so we can conclude
that the nonlinear acoustic effects are not dominant here. The nonlinearities of the mea
sured pressures to the speaker voltage are caused by nonlinearities in the mechanical
equation of the speaker.
Now we will look at the relative deviation of the measured pressure to the calcu
lated pressure (according to our linear model). The relative deviations of the pressures
are plotted as functions of the modeled cone excursion (ﬁgure 3.6b) and modeled pres
sure (ﬁgure 3.6c), at 44 Hz and at 100 Hz. These plots show that the relative deviations
are linear with ζ
1
(ﬁgure 3.6b) and that both plots have the same steepness. The steep
ness of the two plots in ﬁgure 3.6c are very different. From this we conclude that the
nonlinearities are probably caused by speaker nonlinearities due to high cone excur
sions. To strengthen this conclusion, we have plotted the relative pressure deviations
for three different microphones for both frequencies as functions of the cone excursion
in ﬁgure 3.6d. The steepness of the six plots are consistent.
3.5 Discussion and conclusion
We have succeeded in building a setup that can reach drive ratios of more than 20%,
even in the presence of a stack in the resonator. The decoupling of speaker and resonator
with an elastic membrane works well for frequencies above 30 Hz. A recommendation
for future studies is to install the membrane inbetween the horn end and the resonator
tube, instead of between the speaker and the beginning of the horn. This reduces the
forces between the two decoupled parts, as the crosssectional area is smaller at this
position.
The results of the electroacoustical model are in good agreement with the measure
ments. The peak frequencies are predicted within an accuracy of 1 Hz. The cone ex
cursion shows the highest deviation (20%) at 24 Hz, due to mechanical vibrations of
the setup. The calculated pressures, peaks excluded, are within 5% accuracy of the
measurements. At the ﬁrst resonance frequency the discrepancy between pressure cal
culations and measurements grows with the amplitude up to 50%, at a drive ratio of
20%. The discrepancy between theory and measurement may have several causes:
• Nonlinearities within the loudspeaker, due to a large cone excursion.
• Acoustic nonlinearities within the resonator.
• Small changes in the resonance frequency. Since the peaks are very narrow, small
3.5 Discussion and conclusion 47
0 20 40 60
0
0.005
0.01
0.015
0.02
0.025
V
1
 [V]

p
6

/
p
0
measurement
model
(a) [p
1,6
[ versus [V
1
[ at 100 Hz
0 2 4 6 8
−0.4
−0.3
−0.2
−0.1
0
ζ
1


∆
p
1
,
6

/

p
1
,
6

44 Hz
100 Hz
(b) [∆p
1,6
[ versus [ζ
1
[
0 0.1 0.2 0.3 0.4
−0.4
−0.3
−0.2
−0.1
0
p
1,6
/p
0

∆
p
1
,
6

/

p
1
,
6

44 Hz
100 Hz
(c) [∆p
1,6
[ versus [p
1,6
/p
0
[
0 2 4 6 8
−0.4
−0.3
−0.2
−0.1
0
0.1
ζ
1


∆
p
1

/

p
1

mic 1, 44 Hz
mic 1, 100 Hz
mic 2, 44 Hz
mic 2, 100 Hz
mic 3, 44 Hz
mic 3, 100 Hz
(d) p
1,6
, p
4,6
, and p
6,6
Figure 3.6:
(a) The pressure at the end of the resonator p
6
as a function of the speaker voltage at 100 Hz.
Measurements are plotted dashed and the model solid.
(b) The relative pressure deviations as functions of the cone excursion, for two different frequen
cies (dashed at 44 Hz and solid at 100 Hz).
(c) The relative pressure deviations as functions of the drive ratio, for two different frequencies
(dashed at 44 Hz and solid at 100 Hz).
(d) The relative pressure deviations for microphones 1 (dashed), 4 (dashdotted), and 6 (solid),
measured at two different frequencies (grey at 44 Hz and black at 100 Hz).
48 Electroacoustics
deviations of the resonance frequency, for instance due to the heating up of the
resonator gas, can lead to large changes in the pressure amplitudes.
We have explained that the nonlinearities within the loudspeaker, due to a large cone
excursion, are the most sensible explanation for the discrepancy.
The good agreement of the calculation with the measurements, for low amplitudes,
for a wide frequency range (5 to 200 Hz) and a wide amplitude range, does not only
validate the electroacoustical model and the acoustical model of the resonator and the
stack, but also gives conﬁdence on our measurement techniques.
Bibliography
[1] H. Bailliet, P. Lotton, M. Bruneau, and V. Gusev, “Coupling between electrody
namic loudspeakers and thermoacoustic cavities,” Acta Acustica united with Acustica,
vol. 86, no. 2, pp. 363–373, 2000.
[2] M. E. H. Tijani, J. C. H. Zeegers, and A. T. A. M. De Waele, “A gasspring system for
optimizing loudspeakers in thermoacoustic refrigerators,” Journal of Applied Physics,
vol. 92, no. 4, p. 2159, 2002.
[3] D. Marx, X. Mao, and A. J. Jaworski, “Acoustic coupling between the loudspeaker
and the resonator in a standingwave thermoacoustic device,” Applied Acoustics,
vol. 67, no. 5, pp. 402–419, 2006.
[4] L. E. Kinsler, A. R. Frey, A. B. Coppens, and J. V. Sanders, Fundamentals of acoustics,
vol. 4th edition. Wiley, 1999.
[5] R. Aarts, J. Nieuwendijk, and O. Ouweltjes, “Efﬁcient resonant loudspeakers with
large formfactor design freedom,” Journal of the Audio Engineering Society, vol. 54,
no. 10, pp. 940–953, 2006.
50 BIBLIOGRAPHY
Chapter 4
Acoustic measurements
4.1 Introduction
Seybert and Ross [1] proposed a new method for impedance tube measurements of
sound absorbtion using the pressure at two positions, called the twomicrophone
method, which was much faster than the conventional standingwaveratio method.
Chung and Blaser [2–4] further developed this method and later Chu [5, 6] also in
cluded the tubeattenuation effect, allowing the microphones to be placed farther away
from the sample. Bod´ en and
˚
Abom [7] showed by numerical simulations that the two
microphone method has its lowest sensitivity when the two microphones are separated
by a quarter wavelength. Chu [8] has shown experimentally that for accurate measure
ments one of the microphones has to be located close to a pressure minimum and the
choice of the other microphone is not as critical as long as the separation is not close
to half a wavelength. Fujimori et al., Pope, Chu [9], Jones and Parrot [10] describe a
multiplemicrophone method: a least square method on the pressure measurements at
more than three positions. This way it is possible to cover a much wider frequency
range. Jang and Ih [11] theoretically and experimentally studied the inﬂuence of the
microphone positions on the accuracy of the multiplemicrophone method. They have
shown that the equidistant positioning of sensors yields the smallest error within the ef
fective frequency range. In addition they showed that the measurement accuracy can be
increased and the frequency range can be widened by increasing the number of equidis
tant sensors. The transfermatrix approach is a wellknown method for characterizing
porous materials [11]. Also in thermoacoustics the transfermatrix approach has proven
to be useful. Penelet et al. [12] used this approach to make an analytical model of the
acoustic ﬁeld for an arbitrary temperature proﬁle. This model was used to study the
inﬂuence of the temperature proﬁle on the thermoacoustic ampliﬁcation in an annular
thermoacoustic prime mover.
We use the multimicrophone method to determine the transfer matrix elements of
a thermoacoustic sample. From the transfer matrix elements the viscous and thermal
Rott functions can be determined. The theoretical Rott functions of pores of various
shapes are wellknown [13]. Not much literature is available about experimental mea
surements of the Rott functions of small pores. Wilen measured the thermoviscous Rott
functions of a single pore [14, 15] and Petculescu and Wilen also studied the inﬂuence of
52 Acoustic measurements
a temperature gradient [16] and of high amplitudes [17] on the Rott functions of a sin
gle pore. In this paper instead of a single pore we study practical thermoacoustic stacks.
We present a method to determine the acoustic properties of a stack in a resonator only
with the use of microphones.
In chapter 2 the linear theory of thermoacoustics is derived, which resulted in three
1D differential equations (Eq. 2.16, 2.30, and 2.34). In this chapter we focus on the
ﬁrst two equations, the momentum equation and the continuity equation. For a given
temperature gradient dT
0
/dx the two differential equations can be solved for p
1
and
U
1
. The p
1
and U
1
at the right side of a domain can be related to the p
1
and U
1
at the
left side, by a transfer matrix. This is shown in subsection 2.2.1 and the transfer matrix
is given by Eq. 2.41.
If we assume that
dT
0
/dx
T
0
/L
¸1, the inﬂuence of the temperature gradient on the trans
fer matrix can be ignored. The transfer matrix without the presence of a temperature
gradient is given by Eq. 2.42. Later we will verify this assumption using our numerical
model.
We want to determine the transfer matrix of a stack experimentally. A multimicro
phone method (section 4.3) is used to measure the leftward and rightward traveling
wave amplitudes at both sides of the stack. From the amplitudes the p
1
and U
1
at
both sides of the stack can be calculated. This gives us two equations for the transfer
matrix elements. Since the transfer matrix has four elements, two additional equations
are required. Two different methods are explored to do this: the singlestackposition
method and the multiplestackpositions method.
For the singlestackposition method (section 4.6) we use the symmetry of the ge
ometry to obtain two additional equations for the transfer matrix. This method is only
valid for a symmetrical geometry in which the inﬂuence of the temperature gradient is
negligible.
For the multiplestackpositions method (section 4.7) the two additional equations
are obtained by using different p
1
U
1
vectors at one side of the stack. This is accom
plished by putting the stack at different positions in the acoustic wave. Two different
stack positions yield four equations for the transfer matrix. When more than two dif
ferent positions are used, the system is overdetermined, and a least squares method is
used to determine the transfer matrix elements. Since no symmetry equations are used,
this method can in principle also be used for stacks with a nonzero temperature gra
dient. This is possible only when the temperature gradient is the same at the different
locations.
Once the transfer matrix elements are determined, the viscous and thermal Rott
functions, f
ν
and f
κ
, can be calculated using Eq. 2.42. The objective of the experi
ments is to measure the Rott function of different stack geometries as a function of the
frequency. We hope to verify the linear model and also to study under which conditions
(below which drive ratio) the results are in agreement with the linear theory, as well as
where differences occur with the linear model.
In addition to the transfer matrix, we also study the acousticenergy dissipation of a
stack by using only microphones. We will show the importance of minor losses.
4.2 Experimental Setup 53
4.2 Experimental Setup
To determine different properties of stacks and to see wether the transfer matrix method
holds for large amplitudes, an experimental setup (ﬁg. 4.1) is built.
Stack Damper Speaker
Elastic
connection
Mic. 1 Mic. 2 Mic. 3 Mic. 4 Mic. 5 Mic. 6
DC Amplifier DC Amplifier
Lresonator
Dresonator
Closed back volume
Figure 4.1: Schematic drawing of the multimicrophone setup.
The whole construction is placed on a 5 m long aluminium rail table
1
. The resonator
tube is a long aluminiumcylindrical tube ﬁlled with air at ambient pressure and at room
temperature. The left end of the tube is connected to a linear horn with a loudspeaker
and a back volume. The right end of the tube is closed by an end plate and a damper
made of foam. The tube consists of separate parts that can be connected together. This
allows us to change stacks, change the stack position, and change the tube length. To re
duce the vibration from the speaker to the tube, an elastic connection is placed between
the resonator tube and the linear horn. This connection consists of two different parts
that can slide over each other. Between them, a rubber ring is located, so no gas can
escape out of the system. This way, the vibrations generated in the speaker housing can
not transfer to the resonator tube wall.
For our measurement different ceramic stacks with square pores are used. Aschematic
drawing is shown in ﬁgure 4.2 The different parameters of the stack are the pore size D
d
D
L
D
Figure 4.2: Schematic drawing of a ceramic stack. The image on the right is a closeup of the
front of the stack. The pore size D and the pore wall thickness d vary for the different stacks.
and the pore wall thickness d. Another parameter which is related to the pore diameter
1
Item
54 Acoustic measurements
and pore wall thickness is the porosity. It is given by:
ψ =
D
2
(d + D)
2
. (4.1)
In this experiment ﬁve different ceramic stacks are used. The diameter of each stack is
38.1 mm. The stack properties can be found in table 4.1.
Table 4.1: Stack properties: the pore size D; the pore wall thickness d; the porosity; and the stack
length L.
D [mm] d [mm] porosity L [cm]
Ceramic 1 0.36 0.76 0.68 2.6
Ceramic 2 0.36 0.76 0.68 5.0
Ceramic 3 0.62 0.50 0.85 16.0
Ceramic 4 1.30 0.20 0.75 2.6
Ceramic 5 0.98 0.76 0.86 2.6
4.2.1 Microphones and speaker
In our setup the signal coming from a function generator, is ampliﬁed using a 21200
watt linear ampliﬁer
2
. This ampliﬁed signal is sent to the loudspeaker
3
. The speaker
properties are shown in table 4.2.
To measure the acoustics in the resonator, six microphones
4
are used. These micro
phones generate an electrical signal which is sent to an ampliﬁer
5
.
Table 4.2: Speaker properties.
Diameter 25.4 cm
f
s
26 Hz
x
max
(one way) 3.8 cm
RMS power handling 750 W
Peak power handling 1500 W
The Rott functions show interesting behavior when the viscous penetration depth is
of the same size as the hydraulic diameter of the pores. The characteristic frequency f
0
at which this happens depends on the hydraulic radius. We normalize the frequency by
2
Dynacord L2400 21200 watt linear ampliﬁer
3
10
//
Eclipse SW9102
4
Endevco 8510B2
5
3channel Endevco DC ampliﬁer, model 136 DC
4.3 Multimicrophone method 55
dividing by the characteristic frequency, which is deﬁned as
f
0
=
ν
πR
2
H
. (4.2)
A normalized frequency of unity ( f /f
0
= 1) corresponds to δ
ν
= R
H
.
Another step, for which Matlab is used, is calculating the values of f
ν
and f
κ
. From
these calculations, a ﬁt is made. From this ﬁt, the values of the hydraulic radius of the
pores R
H
and the porosity are determined.
4.3 Multimicrophone method
The complex pressure in a tube is the sum of two traveling waves:
p
1
(x) = (C
+
e
−iξ(x−x
L
)
+ C
−
e
iξ(x−x
L
)
), (0 ≤ x ≤ x
L
), (4.3a)
where C
+
and C
−
are the amplitudes of the rightward and leftward traveling waves
respectively, x
L
is the reﬂection position andξ is the complex wave number. It is conve
nient to chose x
L
equal to zero.
The complex wave number was derived in chapter 2 (Eq. 2.44)
ξ =
ω
c
¸
1 + (γ −1) f
k
1 − f
ν
. (4.4)
For cylindrical resonator ducts, with radius R = 2R
H
, the f
ν
and f
κ
functions are given
by [13]
f
ν
=
2J
1
[(i −1)R/δ
ν
]
J
0
[(i −1)R/δ
ν
](i −1)R/δ
ν
, (4.5a)
f
κ
=
2J
1
[(i −1)R/δ
κ
]
J
0
[(i −1)R/δ
κ
](i −1)R/δ
κ
, (4.5b)
where J
0
and J
1
are Bessel functions of the ﬁrst kind. If R
H
¸δ
ν
the boundary approx
imation can be used, yielding [13]
f
ν
=
(1 −i)δ
ν
2R
H
(4.6a)
f
κ
=
(1 −i)δ
κ
2R
H
. (4.6b)
The complex wave number can be written as the sum of a real and imaginary compo
nent:
ξ = k
/
−ik
//
. (4.7)
To determine the complex amplitudes of both traveling waves, C
+
and C
−
, it is nec
essary to measure the pressure at two or more different locations. For now we will as
56 Acoustic measurements
sume that two microphones are used (twomicrophone method), but later we will show
that it is advantageous to use more than two (multimicrophone method). By using a
lockin function the complex amplitudes of the microphone pressures p
x1
and p
x2
are
determined. The complex sound pressure amplitudes in terms of traveling waves are:
p
x1
= C
+
e
−iξx
1
+ C
−
e
iξx
1
, (4.8a)
p
x2
= C
+
e
−iξx
2
+ C
−
e
iξx
2
, (4.8b)
where x
1
and x
2
are the microphones positions (x
1
< x
2
< 0). Using equations (4.8) the
amplitudes of the traveling waves can be determined:
C
+
=
p
x2
e
iξx
1
−p
x1
e
iξx
2
2i sinξ(x
1
−x
2
)
, (4.9a)
C
−
=
p
x1
e
−iξx
2
−p
x2
e
−iξx
1
2i sinξ(x
1
−x
2
)
, (4.9b)
x
x
2
x
n
x
3
x
1
C
+
C

Figure 4.3: A schematic drawing of the multimicrophone method. To the left of the sample n
microphones (n ≥ 2) are located at distances x
1
, x
2
, . . . , x
n
from the left side of the sample.
When using the twomicrophone method to determine the left and rightward trav
eling waves, the distance between the two microphones is very critical. We will focus
on the region to the left of the sample as is shown in ﬁgure 4.3. The two microphone
method fails if
sinξ(x
1
−x
2
) = 0. (4.10)
And even when sinξ(x
1
−x
2
) is not equal but close to zero the method is very sensitive
to errors. When the measurements are done at a ﬁxed frequency the situation of (4.10)
can be easily avoided by choosing the microphone positions wisely, but for a whole
frequency range this is a real problem. A way to overcome this problem is to use more
than two microphones: the multiple microphone method. When n microphones are
used, at positions x
1
, x
2
, . . . , x
n
measuring pressures p
1
, p
2
, . . . , p
n
, the traveling wave
components C
+
and C
−
are described by the following overdetermined (for n > 2)
4.3 Multimicrophone method 57
linear matrix equation
Kx = p, (4.11)
with
K =
_
_
_
_
_
exp(−iξx
1
) exp(iξx
1
)
exp(−iξx
2
) exp(iξx
2
)
.
.
.
.
.
.
exp(−iξx
n
) exp(iξx
n
)
_
_
_
_
_
, x =
_
C
+
C
−
_
, p =
_
_
_
_
_
p
1
p
2
.
.
.
p
n
_
_
_
_
_
. (4.12)
One can compute the optimal solution for x using the leastsquares method. The best
approximate solution for x (p−Kx is minimal) is obtained by using the MoorPenrose
generalized inverse K
+
[18]
x = K
+
p = (K
H
K)
−1
K
H
p, (4.13)
where K
H
denotes the Hermitian matrix of K. Using the matrix deﬁnition of K in Eq.
4.12
K
H
K =
_
∑
n
j=1
exp(2k
//
x
j
) ∑
n
j=1
exp(2ik
/
x
j
)
∑
n
j=1
exp(2k
//
x
j
) ∑
n
j=1
exp(2ik
/
x
j
)
_
. (4.14)
By applying Cramer’s rule to Eq. 4.13 and substituting Eq. 4.14 we ﬁnd for x:
x =
1
det(K
H
K)
_
_
∑
n
j=1
p
j
e
iξ
∗
x
j
∑
n
l=1
e
−2k
//
x
l
−∑
n
j=1
p
j
e
−iξ
∗
x
j
∑
n
l=1
e
2ik
/
x
l
∑
n
j=1
p
j
e
−iξ
∗
x
j
∑
n
l=1
e
2k
//
x
l
−∑
n
j=1
p
j
e
iξ
∗
x
j
∑
n
l=1
e
−2ik
/
x
l
_
_
, (4.15)
where (
∗
) denotes the complex conjugate. The denominator in 4.15 is equal to zero
when det(K
H
K) = 0. As
det(K
H
K) =
_
n
∑
j=1
e
2iξx
j
__
n
∑
l=1
e
−2iξx
l
_
−n
2
, (4.16)
it follows for n = 2 (the two microphone method) that det(K
H
K) = 0 if cos 2ξ(x
1
−
x
2
) = 1.
To verify our experimental setup and to determine the accuracy of the multimicro
phone method we measure the reﬂection coefﬁcient R
r
of a solid end plate. The reﬂec
tion coefﬁcient is deﬁned as
R
r
=
C
−
C
+
, (4.17)
and should be equal to unity in our setup. In ﬁgure 4.4 the reﬂection coefﬁcient, mea
sured by the twomicrophone method, is shown as a function of the frequency. The two
microphones are 1 m apart from each other. Better results could be achieved by putting
the two microphones closer to each other, but we deliberately choose to put them at a
large distance to demonstrate the peaks in the reﬂection coefﬁcient that occur when Eq.
4.10 is satisﬁed.
The reﬂection coefﬁcient measured by the multimicrophone method with ﬁve mi
58 Acoustic measurements
10
2
10
3
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
frequency [Hz]

R

Figure 4.4: The reﬂection coefﬁcient as a function of the frequency. The vertical dashed lines are
drawn at all frequencies that are multitudes of 171.5 Hz.
crophones is shown in ﬁgure 4.5. In ﬁgure 4.5(a) the reﬂection coefﬁcient is plotted
linearly and in 4.5(b) the difference with a unity reﬂection coefﬁcient is plotted logarith
mically. The microphones are at distances to the resonator end: 1.600 m, 1.300 m, 1.000
m, 0.940 m, 0.600 m, and 0.300 m. When comparing ﬁgure 4.4 with ﬁgure 4.5(a) the
peaks have disappeared. The multimicrophone method is working well at almost the
whole frequency range. The only big peaks are located at 6.3 10
2
Hz and in between
1.0 10
3
Hz and 1.1 10
3
Hz. These peaks excluded, the error in the reﬂection coefﬁcient
is smaller than 0.03.
4.3 Multimicrophone method 59
10
2
10
3
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
frequency [Hz]

R

(a) linear scale
10
2
10
3
10
−3
10
−2
10
−1
10
0


R

−
1

frequency [Hz]
(b) logarithmic scale
Figure 4.5: The reﬂection coefﬁcient measured by the multimicrophone method as a function of
the frequency. In subﬁgure (a) it is plotted linearly and in (b) logarithmically.
60 Acoustic measurements
4.4 Acoustic energy losses
4.4.1 Acoustic energy ﬂow
We want to study the acoustic energy ∆
˙
W that is dissipated in a stack. The acoustic
energy loss is determined by measuring the acoustic energy ﬂows at both sides of the
stack. The acoustic energy ﬂow
˙
W is deﬁned as
˙
W =
ω
2π
_
p(t)U(t)dt =
1
2
Re¦p
1
U
∗
1
¦. (4.18)
The acoustic energy ﬂow in the resonator is determined by microphone measurements.
Two methods will be presented: the Fusco method and the travelingwaves method.
These methods are based on measuring the pressure at two or more different positions.
A more direct approach is to measure both the velocity and the pressure at a single po
sition. Among others, this approach was used by Bailliet et al. [19]. They measured the
acoustic power ﬂow in a thermoacoustic resonator by means of laser Doppler anemom
etry together with microphonic measurement and found their results in good agreement
with analytical calculations.
4.4.2 Fusco method
This method is called after Fusco, who introduced a technique that determines acoustic
energy ﬂow in a wide duct from measurements of pressure based on the boundary
layer approximation [20]. He achieved an accuracy of 5% with this method, even at
high amplitudes where the acoustic ﬂow is turbulent. Biwa et al. [21] validated this
technique experimentally also for ducts with a radius smaller than the viscous boundary
layer thickness.
In the Fusco method the relation between the velocity and the pressure gradient Eq.
2.8 is used to determine the volume ﬂow rate. Since this method is used in a resonator
with R
H
¸δ
ν
, we can neglect the viscous term in Eq. 2.8, resulting in
iωρ
0
u
1
= −
dp
1
dx
. (4.19)
Two microphones, separated by ∆x, are used, as is shown in ﬁgure 4.6. If the separation
between the microphones is sufﬁciently small (∆x ¸ λ), the pressure gradient can be
approximated by
dp
1
dx
=
p
1B
− p
1A
∆x
, (4.20)
where p
1A
and p
1B
are the pressures measured with microphones A and B respectively.
The pressure midway between the microphones can be approximated as the average of
the two pressures
p
1
=
p
1A
+ p
1B
2
. (4.21)
4.4 Acoustic energy losses 61
A
xA xB xQ
∆x
A B
p1A p1B
WQ
Figure 4.6: A schematic drawing of the Fusco method for energy ﬂows.
By substituting Eqs. 4.204.21 into Eq. 4.18 and using u
1
= U
1
/A it follows that
˙
W =
iA
4ωρ
0
∆x
Im¦(p
1A
+ p
1B
)(p
1A
− p
1B
)
∗
¦ = −
iA
4ωρ
0
∆x
Im¦p
1A
p
∗
1B
− p
∗
1A
p
1B
¦.
(4.22)
Using z −z
∗
= 2Im¦z¦ yields
˙
W =
iA
2ωρ
0
∆x
Im¦p
1A
p
∗
1B
¦ =
iA
2ωρ
0
∆x
[p
1A
[[p
1B
[ sin(θ), (4.23)
with θ the phase angle by which p
1A
leads p
1B
. Fusco et al. [20] extended Eq. 4.23 to
an equation which is also valid if the requirement ∆x ¸ λ is not fulﬁlled, by including
attenuation in the laminar boundarylayer approximation
˙
W =
A
2ρ
0
c
0
sin(ω∆x/c
0
)
_
Im¦p
1A
p
∗
1B
¦
_
1 −
δ
ν
4R
H
_
1 −
γ −1
_
P
r
+
_
1 +
γ −1
_
P
r
_
ω∆x
c
0
cot
_
ω∆x
c
0
_
__
+
δ
ν
8R
h
([p
1A
[
2
−[p
1B
[
2
)
_
1 −
γ −1
_
P
r
+
_
1 +
γ −1
_
P
r
_
ω∆x
c
0
csc
_
ω∆x
c
0
_
__
.
(4.24)
4.4.3 Travelingwaves method
The complex pressure in a tube is the sum of two traveling waves:
p
1
(x) = (C
+
e
−ik(x−x
Q
)
+ C
−
e
ik(x−x
Q
)
), (4.25)
62 Acoustic measurements
with x
Q
the position at which we want to determine
˙
W and the complex wave number
k given by Eq. 2.44. The travelingwave amplitudes C
+
and C
−
can be determined
experimentally with either a twomicrophone method or a multimicrophone method
as is explained in section 4.3. Using Eq. 4.25 we can determine pressure at position
x = x
Q
and using Eq. 2.16 we can also determine the volume ﬂow rate
p
1Q
= C
+
+ C
−
, (4.26a)
U
1Q
=
Ak(1 − f
ν
)
ρ
0
ω
(C
+
−C
−
), (4.26b)
where f
ν
is given by Eq. 4.6a. Substituting Eqs. 4.26 into Eq. 4.18 yields
˙
W
Q
=
A
2ρ
0
ω
Re¦(C
+
+ C
−
)(C
+
−C
−
)
∗
k
∗
(1 − f
∗
ν
)¦. (4.27)
If the imaginary part of k(1 − f
ν
) can be neglected in Eq. 4.27, it can be simpliﬁed to
˙
W
Q
=
A
2ρ
0
c
0
([C
+
[
2
−[C
−
[
2
). (4.28)
This equation is intuitively understandable, as
A
2ρc
0
[C
+
[
2
is the energy ﬂow of the right
ward traveling wave C
+
and
A
2ρ
0
c
0
[C
−
[
2
is the energy ﬂow of the oppositedirected
wave.
The travelingwaves method has two advantages to the Fusco method. The position
x
Q
, at which we want to measure the acoustic energy ﬂow, is not limited to the position
inbetween the microphones. When multiple microphones are present in a tube, this
method can be used as a multimicrophone method, instead of a Fusco method, result
ing in a higher accuracy. The disadvantage of this method is that it is sensitive to errors
in ξ, as the gas temperature is often not exactly known.
4.4.4 Acousticenergyﬂow measurements
At both sides of the stack, three microphones are positioned. The microphone positions
and the stack position are shown in table 4.3. For the Fusco method, at both sides of
the stack three different p
1A
− p
1B
combinations can be used. We use the following con
vention:
˙
W
i j
is the acoustic energy ﬂow at position (x
i
+ x
j
)/2 determined by using Eq.
4.24 with p
1A
the pressure measured with microphone i and p
1B
the pressure measured
with microphone j.
To compare the different methods for determining the acoustic energy ﬂow, we mea
sure the pressures at six different positions in the empty resonator tube as functions of
the frequency. For practical reasons the measurement is performed at constant speaker
voltage. In ﬁgure 4.7 the dashed plots represent the Fusco method at positions x
12
. . . x
56
and the solid plots represent the travelingwaves method. All the plots are normalized
by dividing by
˙
W
12
determined by the travelingwaves method. The complex wave
amplitudes C
+
and C
−
are here determined by using all six microphones.
For f < 120 Hz the plots show a lot of peaks. We are not sure what is causing
those peaks, but they are also present in a measurement of the reﬂection coefﬁcient.
4.4 Acoustic energy losses 63
Table 4.3: Microphone positions x
i
and acousticenergyﬂow positions x
i j
.
x
i
[cm] x
i j
[cm]
x
1
210.0 x
12
195.0
x
2
180.0 x
23
168.0
x
3
156.0 x
34
125.0
x
4
94.0 x
45
77.0
x
5
60.0 x
56
45.0
x
6
30.0
The reason of the peak in
˙
W
34
at 278 Hz is that at this frequency λ/2 ≈ x
4
− x
3
. This
effect has been discussed in section 4.3. Apart from these peaks the two methods are in
good agreement with each other. The results from the travelingwaves method are more
smooth, since they are determined by the pressures at six positions. What the traveling
waves method in this case is effectively doing is determining the total dissipation in the
resonator and then using the complex wave number to determine the energy ﬂow at
positions x
i j
.
Now we install a stack (sample 1) in the resonator at position x
S
= 100.3 cm. At
the left side of the stack, microphones 1 and 2 give us
˙
W
12
using the Fusco method and
microphones 2 and 3 give us
˙
W
23
. Using microphones 1, 2, and 3 the traveling wave
amplitudes C
+
and C
−
at the left side of the stack are determined. With the traveling
wave method from C
+
and C
−
we determine
˙
W
12
and
˙
W
23
to compare both methods.
At the right side of the stack we do the same for
˙
W
45
and
˙
W
56
. In ﬁgure 4.8 the
˙
W
i j
for
four different microphone combinations are plotted as function of the frequency. The
solid plots represent the Fusco method and the dashed ones the traveling wave method.
The black dashed plots are the acoustic energy ﬂows determined by using the traveling
wave method at the positions of the two stack ends.
The acoustic energy dissipated in the stack ∆
˙
W
S
will be determined as the difference
between the two black dashed plots in ﬁgure 4.8.
4.4.5 Acoustic energy losses in a stack
Using the deﬁnition of acoustic work ﬂow (Eq. 4.18) we ﬁnd the acoustic energy
˙
W that
is dissipated per unit length
d
˙
W
dx
=
1
2
Re
_
p
1
dU
∗
1
dx
+ U
∗
1
dp
1
dx
_
. (4.29)
For dU
∗
1
/dx and dp
1
/dx we use Eqs. 2.35(ab)
d
˙
W
dx
=
1
2
Re
_
p
1
(−R
∗
1
+ R
∗
3
U
1
) −U
∗
1
R
2
U
1
_
= −
1
2
Re¦R
1
¦[U
1
[
2
−
1
2
Re¦R
2
¦[p
1
[
2
+
1
2
Re¦R
∗
3
p
1
U
∗
1
¦, (4.30)
64 Acoustic measurements
100 200 300 400
0
0.2
0.4
0.6
0.8
1
˙
W
i
j
/
˙
W
1
2
f [Hz]
12
23
34
45
56
12
23
34
45
56
Figure 4.7: The normalized acoustic energy ﬂows
˙
W
i j
as functions of the frequency at different
positions x
i j
(as shown in the legend) in the empty resonator. The dashed plots represent the
Fusco method and the solid ones the travelingwaves method. All the plots are normalized by
dividing by
˙
W
12
determined by the travelingwaves method. The complex wave amplitudes C
+
and C
−
for the travelingwaves method are here determined using all six microphones.
with
Re¦R
1
¦ = −
ωρ
0
A
Im( f
ν
)
[1 − f
ν
[
2
, (4.31a)
Re¦R
2
¦ = −
(γ −1)ωA
γp
0
Im( f
κ
). (4.31b)
The total acoustic energy loss ∆
˙
W in the stack is found by integrating d
˙
W/dx over the
whole stack length.
The pressure p
1
and volume ﬂow rate U
1
as function of the position in the stack are
determined by solving the thermoacoustic equations. As the boundary conditions for
the these equations we use the pressure and volume ﬂow rate at the left end of the stack,
that are measured using the travelingwave method.
4.4.6 Minor losses
The major losses in a stack are caused by viscous dissipation and thermal relaxation.
These effects are included in the linear theory. At high amplitudes minor losses will
become a signiﬁcant factor. In case of a stack these minor losses occur at the stack ends.
The sudden change of the crosssectional area of the gas causes losses. For instance,
4.4 Acoustic energy losses 65
50 100 150 200 250 300 350 400
0
0.2
0.4
0.6
0.8
1
˙
W
i
j
/
˙
W
1
2
f [Hz]
12
23
45
56
12
23
45
56
stack end
stack end
Figure 4.8: The acoustic energy ﬂows as functions of the frequency at different positions in a
resonator with a stack installed inside. The solid plots represent the Fusco method and the dashed
ones the travelingwaves method. All the plots are normalized by dividing by
˙
W
12
determined
by the Fusco method. The black dashed plots are the acoustic energy ﬂows determined by using
the travelingwave method at the positions of the two stack ends.
when the ﬂow is directed out of the stack vortices are formed.
We will derive an equation to describe these minor losses quantitatively. First we
will do this for a steady ﬂow, since the minor losses in steady ﬂows are wellknown in
literature. Unfortunately the equations are very empirical. Then we will extend this
steady theory to oscillatory ﬂows, determine the energy loss due to minor losses, and
verify it with measurements.
Steady ﬂow
First we study a sudden increase of cross section (also called expansion), from A
1
to A
2
,
in a steady ﬂow (ﬁgure 4.9). Before the expansion, the ﬂow is developed with velocity
u
1
(y, z), pressure p
1
, and density ρ
1
. The dashed rectangle represents a control volume.
Its surface at the left, S
1
, is directly behind the expansion and its surface at the right, S
2
,
is chosen at the position where the ﬂow is developed again. At surface S
2
the velocity
is u
2
(y, z), the pressure is p
2
, and the density is ρ
2
. We apply the integral law of mass
conservation to the control volume,
¸u
1
¸ρ
1
A
1
= ¸u
2
¸ρ
2
A
2
, (4.32)
66 Acoustic measurements
S2
S1
A
2
u2
p2 p1
A
1
u1
Figure 4.9: A schematic drawing of sudden increase in cross section in a steady ﬂow.
where ¸¸ denotes the crosssectional average. We apply the integral momentum equa
tion to the control volume
__
S
1
(ρu
2
+ p)dA =
__
S
2
(ρu
2
+ p)dA, (4.33)
and assume that the pressures and densities are uniform over the cross section
β
1
ρ
1
¸u
1
¸
2
A
1
+ p
1
A
2
= β
2
ρ
2
¸u
2
¸
2
A
2
+ p
2
A
2
, (4.34)
with β
i
the momentum correction factor deﬁned as
β
i
=
¸u
2
i
¸
¸u
i
¸
2
. (4.35)
We consider the integral energy equation
∆
˙
E
ml
=
__
A
(h +
1
2
[v[
2
)(ρv) dA, (4.36)
with ∆
˙
E
ml
the energy loss due to minor losses and h the enthalpy per unit mass, which
is given by
h =
p
ρ
+ c
v
T =
c
p
R
s
p
ρ
, (4.37)
for ideal gases. We apply the energy equation to the control volume
∆
˙
E
ml
=
__
S
1
h
1
ρ
1
u
1
dA −
__
S
2
h
2
ρ
2
u
2
dA. (4.38)
In the enthalpy per unit mass (Eq. 4.37) we write the c
p
/R
s
term as 1/(1 −γ
−1
). If the
4.4 Acoustic energy losses 67
temperature increases during the expansion due to losses, we consider this as a loss,
which is included in ∆
˙
E
ml
. After substituting Eq. 4.37 into Eq. 4.38 and integrating over
the cross section
∆
˙
E
ml
=
1
2
α
1
ρ
1
¸u
1
¸
3
A
1
+
p
1
¸u
1
¸A
1
1 −γ
−1
−
1
2
α
2
ρ
2
¸u
2
¸
3
A
2
+
p
2
¸u
2
¸A
2
1 −γ
−1
, (4.39)
with α the kineticenergy correction factor deﬁned as
α =
¸u
3
¸
¸u¸
3
. (4.40)
If we assume that ρ
1
= ρ
2
= ρ and use Eq. 4.32 it can be shown that
∆
˙
E
ml
=
1
2
α
1
ρ¸u
1
¸
3
A
1
−
1
2
α
2
ρ¸u
1
¸
3
A
1
A
2
1
A
2
2
+
(p
1
− p
2
)¸u
1
¸A
1
1 −γ
−1
. (4.41)
Substituting the p
1
− p
2
of Eq. 4.34 into Eq. 4.41 results in
∆
˙
E
ml
= K
e
1
2
A
1
ρ¸u
1
¸
3
, (4.42)
with minorloss coefﬁcient K for an expansion given by
K
e
= α
1
−
2β
1
ψ
1 −γ
−1
+ (
2β
2
1 −γ
−1
−α
2
)ψ
2
, (4.43)
with the porosity ψ = A
1
/A
2
. The pressure drop due to minor losses is given by
∆p =
∆
˙
E
ml
A
1
¸u
1
¸
= K
e
1
2
ρ¸u
1
¸
2
. (4.44)
Note that ∆p differs from p
1
− p
2
. The pressure difference p
1
− p
2
is from S
1
to S
2
,
whereas ∆p is from the location just before the contraction to S
2
.
If the velocity is uniform over the cross section (i.e. α
1
= β
1
= 1 and α
2
= β
2
= 1)
K
e
= (1 −ψ)
2
. (4.45)
For a developed laminar ﬂow inbetween two parallel plates
u =
3
2
¸u¸
_
1 −
y
2
y
2
0
_
, for − y
0
≤ y ≤ y
0
, (4.46)
it follows that α = 54/35 and β = 6/5. In case of a developed laminar ﬂow through a
cylinder
u = 2¸u¸
_
1 −
r
2
R
2
_
, for 0 ≤ r ≤ R, (4.47)
68 Acoustic measurements
it is found that α = 2 and β = 4/3.
In case of a contraction only empirical data is available. According to Streeter et
al. [22] the energy loss and pressure drop for a contraction can also by approximated by
Eqs. 4.42 and 4.44, with the minorloss coefﬁcient for a contraction
K
c
=
_
1
C
c
−1
_
2
. (4.48)
The contraction coefﬁcient C
c
for water was determined by Weisbach [22] for different
A
2
/A
1
ratios, which we ﬁtted according to a third order polynomial
C
c
= C
c3
ψ
3
+ C
c2
ψ
2
+ C
c1
ψ+ C
c0
, (4.49)
with C
c3
= 0.55, C
c2
= −0.34, C
c1
= 0.17, and C
c0
= 0.61.
Now that we have discussed the theory of minor losses through a sudden expansion
and contraction in the cross section for a steady ﬂow, we will extent the analysis to the
more complicated oscillatory ﬂow.
Oscillatory ﬂow
We use Eq. 4.44 for a timedependent ﬂow
∆p(t) = −
K(t)ρ(t)
2A
2
[U(t)[U(t), (4.50)
with the pressure p and U real and time dependent. The pressure difference due to
minor losses is 180 degrees out of phase with the velocity and its amplitude proportional
to the velocity squared. For the ﬁrst order term of the pressure variation we can ignore
the density variations: ρ(t) = ρ
0
. The minor loss coefﬁcient K(t) is time dependent
since its value is different for contraction phase than for the expansion phase. As an
approximation we use a timeindependent average value K(t) = (K
e
+ K
c
)/2. We write
the volume ﬂow rate as
U(t) = [U
1
[ cos(ωt), (4.51)
with [U
1
[ the amplitude. The acoustic energy loss due to minor losses we ﬁnd with
∆
˙
W =
ω
2π
_
2π/ω
0
∆p(t)U(t)dt =
2
3π
Kρ
0
A
2
[U
1
[
3
. (4.52)
The drop in pressure ∆p
1
due to minor losses, can be approached by the ﬁrst harmonic
[ cos(ωt)[ cos(ωt) ≈ a
1
cos(ωt), (4.53)
where a
1
follows from taking the ﬁrstorder amplitude of the Fourier series
a
1
=
ω
π
_
π/ω
−π/ω
[ cos(ωt)[ cos
2
(ωt)dt =
8
3π
. (4.54)
4.4 Acoustic energy losses 69
In ﬁgure 4.10 [ cos(ωt)[ cos(ωt) and a
1
cos(ωt) are shown as function of t. By substi
tuting Eq. 4.53 into Eq. 4.50 the minorloss pressuredrop amplitude yields
∆p
1
= −
4
3π
Kρ
0
A
2
[U
1
[U
1
. (4.55)
Eq. 4.55 is consistent with 4.52, since
1
2
Re¦∆p
1
U
∗
1
¦ =
2
3π
Kρ
0
A
2
[U
1
[
3
. (4.56)
−1
−0.5
0
0.5
1
ωt
0
π 2π 3π 4π
cos(ωt) cos(ωt)
8/3π cos(ωt)
Figure 4.10: The [ cos(ωt)[ cos(ωt) dependency of the minor losses is approximated by
8/3π cos(ωt).
Nowlet us discuss the value of K
e
for a transition of a stack into a resonator tube. The
diameter of the resonator tube is much larger than the viscous penetration depth, thus
α
2
≈ 1 and β
2
≈ 1. In ﬁgure 4.11a the values of α and β are shown for a parallel plate
stack (solid curve), circular pores (dashdotted curve), and rectangular pores (dotted
curve) as a function of the hydraulic radius. For ψ = 0 the parallel plates have α =
54/35 and β = 6/5, which are the same as for steady ﬂow. This resemblance with the
steady ﬂow also holds for circular pores, where α = 54/35 and β = 6/5, for ψ = 0. In
the limit of an inﬁnitely large resonator diameter both α and β approach unity.
In ﬁgure 4.11b the minorloss coefﬁcient K
e
is plotted as function of the porosity
for a stack with squared pores with three different sizes: D
h
/δ
ν
→ 0, D
h
/δ
ν
= 1,
and D
h
/δ
ν
→ ∞. The minorloss coefﬁcient K
c
for a contraction is also shown, using
the empirical data of Weisbach. For an expansion the α and β correction factors for the
velocity proﬁle have a huge effect on the minor loss coefﬁcient. The K coefﬁcient is twice
70 Acoustic measurements
0 1 2 3 4 5
1
1.2
1.4
1.6
1.8
2
2.2
D
H
/δ
ν
α
,
β
circular pores
squared pores
parallel plates
α
β
(a)
0 0.2 0.4 0.6 0.8 1
0
0.5
1
1.5
2
2.5
ψ
K
K
e
, D/δ
ν
→0
K
e
, D/δ
ν
= 1
K
e
, D/δ
ν
→∞
K
c
(b)
Figure 4.11:
(a) The kineticenergy correction factor α and momentum correction factor β as function of the
D
H
/δ
ν
ratio for a developed oscillatory ﬂow through three different stack geometries: circular
pores, squared pores, and parallel plates.
(b) The minorloss coefﬁcient K as function of the porosity ψ. For an expansion of squared pores
into a resonator tube, K
e
is plotted for three different pore sizes and for a contraction K
c
is plotted
according to the empirical data of Weisbach (Eq. 4.48).
as big as for a uniform velocity proﬁle (D
h
/δ
ν
→ ∞), even for zero porosity. For large
porosities the effect of the velocity proﬁle is relatively much larger. For a porosity of 1,
there is no expansion, but still the difference in velocity distribution, leads to K
e
= 0.40
for D
h
/δ
ν
→ 0 and K
e
= 0.32 for D
h
/δ
ν
= 1. The difference between D
h
/δ
ν
→ 0 and
D
h
/δ
ν
= 1 is very small for all porosities.
4.5 Minorloss correction
In ﬁgure 4.12 the acoustic power losses of sample 4 are shown as a function of the vol
ume ﬂow rate. In ﬁgure 4.12a the frequency is 36.5 Hz and in ﬁgure 4.12b 105.5 Hz. The
dashdotted curve represents measured data. The dashed curve is a numerical calcula
tion without taking into account minor losses, the solid curve is with taking into account
minor losses and the black dotted curve is with also including a K
p
[p
1
[
4
term. Both fre
quencies (36.5 Hz and 105.5 Hz) are resonance frequencies. The pore size D = 1.1 mm,
the porosity ψ = 0.72, and K = 0.41. This value for K = 0.41 is determined by a ﬁt
in ﬁgures 4.12. The minor loss coefﬁcient K agrees very well with the theoretical value:
0.39.
In ﬁgure 4.13 we show similar plots, but this time for sample 5. The pore size D =
0.83 mm, the porosity ψ = 0.85, and K = 0.35. This value for K is determined by a
ﬁt in ﬁgures 4.13. The minor loss coefﬁcient K = 0.35 is higher than the theoretical
value: 0.28. This can be explained by inaccuracies in the porosity and pore size, which
4.5 Minorloss correction 71
0 0.02 0.04 0.06
0
20
40
60
80
∆
˙
W
S
[
W
]
U
1
 [m
3
/s]
K = 0
K = 0.41
K = 0.41, K
p
p
1

4
measurement
(a) f = 36.5 Hz, D
r,max
= 19.3%
0 0.01 0.02 0.03 0.04
0
5
10
15
20
25
30
∆
˙
W
S
[
W
]
U
1
 [m
3
/s]
K = 0
K = 0.41
K = 0.41, K
p
p
1

4
measurement
(b) f = 105.5 Hz, D
r,max
= 5.9%
Figure 4.12: The acousticpower losses ∆
˙
W
S
as functions of the volume ﬂowrate at two different
frequency and drive ratio combinations. The dashdotted curve represents the measurement and
the other three curves are analytical calculations. The dashed curves are the theoretical value
without correcting for minor losses (K = 0). The theoretical functions corrected for minor losses
(K = 0.41) are shown by the solid curves. An additional correction term K
p
[p
1
[
4
term is added
for the dotted curves.
are known with a 5% accuracy, and also inﬂuence the energy dissipation. Additionally
the pores are not perfectly squared, with exactly the same pore size along the whole
stack length, which can also increase the dissipation. And also in the derivation of the
minor loss theory some assumptions are made which leave room for betterment. The
ﬂow at the entrance of the pore is assumed to be developed. Other nonlinear effects like
turbulence and higher harmonics are not taken into account. Also, for determining the
acoustic energy ﬂows by the travelingwaves method, it is required that no temperature
gradients are present in the resonator tube. Although the tube wall is very thick and is
heat conducting, at high drive ratios, up to 19.5%, temperature gradients can occur. So
it is not surprising that the Kvalue determined from the ﬁt is higher than the theoretical
value.
72 Acoustic measurements
0 0.02 0.04 0.06
0
20
40
60
80
∆
˙
W
S
[
W
]
U
1
 [m
3
/s]
K = 0
K = 0.35
K = 0.35, K
p
p
1

4
measurement
(a) f = 36.5 Hz, D
r,max
= 19.5%
0 0.01 0.02 0.03 0.04
0
5
10
15
20
25
∆
˙
W
S
[
W
]
U
1
 [m
3
/s]
K = 0
K = 0.35
K = 0.35, K
p
p
1

4
measurement
(b) f = 105.5 Hz, D
r,max
= 5.7%
Figure 4.13: The acousticpower losses ∆
˙
W
S
as functions of the volume ﬂowrate at two different
frequency and drive ratio combinations. The dashdotted curve represents the measurement and
the other three curves are analytical calculations. The dashed curves are the theoretical value
without correcting for minor losses (K = 0). The theoretical functions corrected for minor losses
(K = 0.35) are shown by the solid curves. An additional correction term K
p
[p
1
[
4
term is added
for the dotted curves.
4.6 Singlestackposition method 73
4.6 Singlestackposition method
0
xL xR xend
x
x1
x2
x4
x3
uL
pL
uR
pR
Rend
C+
C
D+
D
Figure 4.14: A schematic drawing of the experimental setup for an acoustic characterization
of a sample using the twomicrophone method. A loudspeaker connected to a tube which is
terminated by a reﬂection coefﬁcient R
end
. Between x
L
and x
R
the tube is ﬁlled with the acoustic
sample and at both sides of the sample two microphones are located.
To study the acoustical properties of a thermoacoustic sample, e.g. a stack or a re
generator, a multimicrophone setup is built, which is shown in ﬁgure 4.14. At the left
end of the tube a loudspeaker is used as a sound source and the right end of the tube
is closed by a plate with a reﬂection coefﬁcient R
end
, which will be close to unity. The
advantage of a closed tube compared to an anechoic termination, which is commonly
used, is that it allows higher pressure amplitudes, especially in resonance. The complex
pressure in the tube is the sum of two traveling waves:
p(x, t) = (C
+
e
−iξ(x−x
L
)
+ C
−
e
iξ(x−x
L
)
)e
iωt
, (4.57)
for (0 ≤ x ≤ x
L
) and
p(x, t) = (D
+
e
−iξ(x−x
R
)
+ D
−
e
iξ(x−x
R
)
)e
iωt
, (4.58)
for (x
R
≤ x ≤ x
end
).
The complex amplitudes of the traveling waves C
+
, C
−
, D
+
and D
−
are found by
using the multimicrophone method (Eq. 4.12). At each side of the stack three micro
phones are installed in the resonator, but in ﬁgure 4.14 only two of them are shown.
The p
1
u
1
vectors at the left and the right side of the sample can be expressed in
terms of the traveling wave amplitudes:
_
p
L
u
L
_
=
_
C
+
+ C
−
(C
+
−C
−
)/(ρ
0
c)
_
, (4.59a)
_
p
R
u
R
_
=
_
D
+
+ D
−
(D
+
−C
−
)/(ρ
0
c)
_
. (4.59b)
74 Acoustic measurements
Using Eq. 4.11 the traveling wave amplitudes can be determined, after which Eq. 4.59
provides us with two equations for determining the transfer matrix components. Two
additional equations are required. We make use of the reciprocity of the linear acoustic
equations. Reciprocity refers to situations for which a magnitude associated with an
”effect” at a point is unchanged when the locations of ”cause” and ”point of observa
tion” are interchanged. The transfer matrix has to obey the reciprocity condition for all
pu vectors. From this it follows that the determinant of the transfer matrix is unity [23].
In case of a symmetrical sample it can be shown that also the condition T
11
= T
22
ap
plies [23]. Now we have a total of four equations for the transfer matrix:
p
L
= T
11
p
R
+ T
12
u
R
, (4.60a)
u
L
= T
21
p
R
+ T
22
u
R
, (4.60b)
T
11
T
22
−T
12
T
21
= 1, (4.60c)
T
11
= T
22
. (4.60d)
Combining these four equations the following expressions for the transfer matrix ele
ments can be found:
T
11
= T
22
=
p
L
u
L
+ p
R
u
R
p
R
u
L
+ p
L
u
R
, (4.61a)
T
12
=
p
2
L
− p
2
R
p
R
u
L
+ p
L
u
R
, (4.61b)
T
21
=
u
2
L
−u
2
R
p
R
u
L
+ p
L
u
R
. (4.61c)
The transfer matrix elements as functions of the frequency are difﬁcult to interpret.
As Eq. 2.42 relates the Rott functions to the transfer matrix elements, we can calculate
the Rott functions from the measured transfer matrix elements. In ﬁgure 4.15 the real
and imaginary components of the viscous Rott function f
ν
are shown as a function of
the frequency. The reference frequency f
0
is 174.9 Hz for this stack. The data points are
the measurement data and the solid curve is a ﬁt according to [24]:
f
ν
= 1 −
64
π
4
∞
∑
m=0
∞
∑
n=0
_
(2m + 1)
2
(2n + 1)
2
(1 + i
2π
2
δ
2
ν
D
2
((2m + 1)
2
+ (2n + 1)
2
)
_
−1
(4.62)
for rectangular pores with the pore size D. Pore size D and the porosity ψ are ﬁtting
parameters.
The peaks at f /f
0
= 0.97 and f /f
0
= 1.95 are caused by velocity nodes located at the
stack position at these frequencies. Since viscous effects are a consequence of velocity,
the viscous effects are small when the stack is in a velocity node, which makes them
difﬁcult to measure and more sensible to errors.
To study the performance of the singlestackposition method at high driver ratios
we have increased the maximum D
r
in the resonator from 1% in ﬁgure 4.15 to 8% in
ﬁgure 4.16. Both ﬁgures are almost identical. Since the f
ν
functions have not deformed
at D
r
= 8%, the ﬂow in the stack is probably still laminar. Also we can conclude that
the singlestackposition method still works well at this D
r
.
4.7 Multiplestackpositions method 75
0 0.5 1 1.5 2
0
0.05
0.1
0.15
0.2
f/f
0
ψ
R
e
[
1
−
f
ν
]
0 0.5 1 1.5 2
0
0.1
0.2
0.3
0.4
f/f
0
ψ
I
m
[
1
−
f
ν
]
Figure 4.15: Plot of the real and imaginary part of f
ν
as functions of the normalized frequency.
The data points are measurements and the solid curve is a ﬁt. The maximum D
r
reached in the
tube is 1 %.
0 0.5 1 1.5 2
0
0.05
0.1
0.15
0.2
f/f
0
ψ
R
e
[
1
−
f
ν
]
0 0.5 1 1.5 2
0
0.1
0.2
0.3
0.4
f/f
0
ψ
I
m
[
1
−
f
ν
]
Figure 4.16: Plot of the real and imaginary part of f
ν
as functions of the normalized frequency.
The data points are measurements and the solid curve is a ﬁt. The maximum D
r
reached in the
tube is 8 %.
4.7 Multiplestackpositions method
To determine the different elements of the transfer matrix, two extra equations were
used, which came from the reciprocity of the linear acoustic equations and from the
76 Acoustic measurements
symmetry of the stack (Eqs. 4.60c and 4.60d).
To determine if the use of these equations is valid, the overdetermined transfer ma
trix method is used. In this method frequency sweep measurements are performed at
two or more stack positions. Between these measurements the sample is moved inside
the tube. This way extra equations for the transfermatrix are derived. For example, The
ﬁrst measurement is performed at x = −1.00 m and the second at x = −1.05 m. For ev
ery measurement the pressure and velocity left and right from the stack are determined.
From these pressures and velocities the transfer matrix can be determined:
_
p
L1
p
L2
p
Ln
u
L1
u
L2
u
Ln
_
= T
_
p
R1
p
R2
p
Rn
u
R1
u
R2
u
Rn
_
. (4.63)
For two stack positions, the transfermatrix can be determined exactly. In case of more
than two positions, the system of equations is overdetermined. In that case, the best
values for the transfer matrix elements, T
11
, T
12
, T
21
, and T
22
, are determined using a
leastsquares method.
The measurement using the singlestackposition method, from which the results
were shown in ﬁgure 4.15, is repeated using the multiplestackposition method. In ﬁg
ure 4.17 the real and imaginary components of the viscous Rott function, f
ν
, are shown
as a function of the frequency. The reference frequency f
0
is 174.9 Hz for this stack.
0 0.5 1 1.5 2
0
0.05
0.1
0.15
f/f
0
ψ
R
e
[
1
−
f
ν
]
0 0.5 1 1.5 2
0
0.05
0.1
0.15
0.2
f/f
0
ψ
I
m
[
1
−
f
ν
]
Figure 4.17: Plot of the real and imaginary part of f
ν
as a function of the normalized frequency.
The data points are measurements and the solid curve is a ﬁt. The maximum D
r
reached in the
tube is 1 %.
The multiplestackposition results are in excellent agreement with the singlestack
position results. This conﬁrms the validity of both methods. Since we do the mea
surements for ﬁve different stack positions there is no frequency at which the stack is
always in a velocity node. For this reason we do not measure the peaks at f /f
0
= 0.97
and f /f
0
= 1.95 as we saw in the singlestackposition measurements.
4.8 Discussion and conclusion 77
4.8 Discussion and conclusion
The multimicrophone method proved to be a valuable technique for determining
acousticenergy ﬂows in a resonator tube as well as the transfer matrix of a stack.
The twomicrophone and multimicrophone methods are wellknown techniques
for determining the transfer matrix of porous media. This is the ﬁrst time the multi
microphone method is used in case of a stack as porous medium. Also, as far as we
know, the travelingwave method has not been used before in determining the acoustic
energy dissipation of a stack.
The multimicrophone method is more accurate than the twomicrophone method.
Especially for frequencies at which the distances between the two microphones are
multiples of a half wavelength, the twomicrophone method fails, whereas the multi
microphone method works well for a whole range of frequencies (100 Hz to 2000 Hz).
For determining the acousticenergy ﬂow two methods have been tested: the Fusco
method and the travelingwave method. The two methods are in good agreement with
the exception of the frequencies at which the microphone are a multitude of λ/2 sepa
rated from each other. At these frequencies the Fusco method gives unreliable results.
The travelingwave method does not show this problem.
The acousticenergy dissipation at low amplitudes is in agreement with the linear
theory. At higher amplitudes the linear theory starts to deviate from the measurements.
These deviations can partially be explained by minor losses at the stack ends, especially
for high velocity amplitudes. For high pressure amplitudes the deviations are probably
caused by higher harmonics, but this has not been veriﬁed.
For the ﬁrst time the multimicrophone method has been used to determine the
transfer matrix elements of a stack. From the transfer matrix elements the f
ν
functions
of a stack can be determined as functions of the frequency. Two methods have been
tested: the singlestackposition method and the multiplestackpositions method. Both
methods are in excellent agreement with each other. The singlestackposition method
only shows a few small deviations from the analytical f
ν
function. The multiplestack
positions method is even in better agreement with the analytical plot of f
ν
as function
of the frequency.
78 Acoustic measurements
Bibliography
[1] A. F. Seybert and D. F. Ross, “Experimentaldetermination of acoustic properties
using a 2microphone randomexcitation technique,” Journal of the Acoustical Society
of America, vol. 61, no. 5, pp. 1362–1370, 1977.
[2] J. Y. Chung and D. A. Blaser, “Transferfunction method of measuring induct
acoustic properties .1. theory,” Journal of the Acoustical Society of America, vol. 68,
no. 3, pp. 907–913, 1980.
[3] J. Y. Chung and D. A. Blaser, “Transferfunction method of measuring induct
acoustic properties .2. experiment,” Journal of the Acoustical Society of America,
vol. 68, no. 3, pp. 914–921, 1980.
[4] J. Y. Chung and D. A. Blaser, “Transferfunction method of measuring acoustic
intensity in a duct system with ﬂow,” Journal of the Acoustical Society of America,
vol. 68, no. 6, pp. 1570–1577, 1980.
[5] W. T. Chu, “Extension of the 2microphone transferfunction method for
impedance tube measurements,” Journal of the Acoustical Society of America, vol. 80,
pp. 347–348, July 1986.
[6] W. T. Chu, “Transferfunction technique for impedance and absorption
measurements in an impedance tube using a single microphone,” Journal of the
Acoustical Society of America, vol. 80, pp. 555–560, Aug. 1986.
[7] H. Bod´ en and M.
˚
Abom, “Inﬂuence of errors on the 2microphone method for
measuring acoustic properties in ducts,” Journal of the Acoustical Society of Amer
ica, vol. 79, pp. 541–549, Feb. 1986.
[8] W. T. Chu, “Further experimental studies on the transferfunction technique for
impedance tube measurements,” Journal of the Acoustical Society of America, vol. 83,
pp. 2255–2260, June 1988.
[9] W. T. Chu, “Impedance tube measurements  a comparativestudy of current prac
tices,” Noise Control Engineering Journal, vol. 37, pp. 37–44, July 1991.
[10] M. G. Jones and T. L. Parrott, “Evaluation of a multipoint method for determining
acousticimpedance,” Mechanical Systems and Signal Processing, vol. 3, pp. 15–35,
Jan. 1989.
80 BIBLIOGRAPHY
[11] S. H. Jang and J. G. Ih, “On the multiple microphone method for measuring induct
acoustic properties in the presence of mean ﬂow,” Journal of the Acoustical Society of
America, vol. 103, pp. 1520–1526, Mar. 1998.
[12] G. Penelet, S. Job, V. Gusev, P. Lotton, and M. Bruneau, “Dependence of sound am
pliﬁcation on temperature distribution in annular thermoacoustic engines,” Acta
Acustica united with Acustica, vol. 91, no. 3, pp. 567–577, 2005.
[13] W. P. Arnott, H. E. Bass, and R. Raspet, “General formulation of thermoacoustics
for stacks having arbitrarily shaped pore cross sections,” Journal of the Acoustical
Society of America, vol. 90, no. 6, pp. 3228–3237, 1991.
[14] L. A. Wilen, “Measurements of scaling properties for acoustic propagation in a
single pore,” Journal of the Acoustical Society of America, vol. 101, no. 3, pp. 1388–
1397, 1997.
[15] L. A. Wilen, “Measurements of thermoacoustic functions for single pores,” Journal
of the Acoustical Society of America, vol. 103, no. 3, pp. 1406–1412, 1998.
[16] G. Petculescu and L. A. Wilen, “Thermoacoustics in a single pore with an applied
temperature gradient,” Journal of the Acoustical Society of America, vol. 106, no. 2,
pp. 688–694, 1999.
[17] G. Petculescu and L. A. Wilen, “Highamplitude thermoacoustic effects in a single
pore,” Journal of the Acoustical Society of America, vol. 109, no. 3, pp. 942–948, 2001.
[18] A. F. Seybert and B. Soenarko, “Error analysis of spectral estimates with application
to the measurement of acoustic parameters using random sound ﬁelds in ducts,”
Journal of the Acoustical Society of America, vol. 69, no. 4, pp. 1190–1199, 1981.
[19] H. Bailliet, P. Lotton, M. Bruneau, V. Gusev, J. Vali` ere, and B. Gazengel, “Acoustic
power ﬂow measurement in a thermoacoustic resonator by means of laser doppler
anemometry (l.d.a.) and microphonic measurement,” Applied Acoustics, vol. 60,
no. 1, pp. 1–11, 2000.
[20] A. Fusco, W. Ward, and G. Swift, “Twosensor power measurements in lossy
ducts,” Journal of the Acoustical Society of America, vol. 91, no. 4 I, pp. 2229–2235,
1992.
[21] T. Biwa, Y. Tashiro, H. Nomura, Y. Ueda, and T. Yazaki, “Experimental veriﬁca
tion of a twosensor acoustic intensity measurement in lossy ducts,” Journal of the
Acoustical Society of America, vol. 124, no. 3, pp. 1584–1590, 2008.
[22] V. L. Streeter, E. B. Wylie, and K. W. Bedford, Fluid Mechanics, vol. 9th edition.
McGrawHill Companies, 1998.
[23] B. H. Song and J. S. Bolton, “A transfermatrix approach for estimating the charac
teristic impedance and wave numbers of limp and rigid porous materials,” Journal
of the Acoustical Society of America, vol. 107, pp. 1131–1152, Mar. 2000.
[24] G. Swift, A Unifying Perspective for Some Engines and Refrigerators. Melville: Acous
tical Society of America, 2002.
Chapter 5
Temperature proﬁle in a stack
5.1 Introduction
In this chapter we study the temperature proﬁle in a parallelplate stack, experimentally
as well as numerically. Both the steady state and the time dependency are studied. The
experimental setup consists of a stack without heat exchangers in a resonator tube,
which is driven by a loudspeaker. Such a stack without heat exchangers is also called
a thermoacoustic couple [1]. Were the thermoacoustic couple devoid of thermal contact
with the environment, it would keep heating up. In practice, a steady state will be
reached, due to thermal contact with the environment.
In the steady state, the total heat ﬂow to the environment is equal to the absorbed
acoustic power. For a good understanding of a thermoacoustic stack it is essential to
know how the heat ﬂows to the environment. To study the heat losses we studied the
passive cooling down of a stack. First we used the loudspeaker to create a temperature
difference in the stack. Once a steady state is reached, the loudspeaker is turned off
and the temperature proﬁle over the stack is measured as function of time. The pas
sive cooling down of the stack gives us valuable information of the heat losses to the
environment. It will be shown that the conduction through the plates in radial direction
(from the the middle of the plates to the tube wall) is dominating the heat losses.
Most thermoacoustic models [1–4] assume that the total power in a stack
˙
E is con
stant in axial direction. This works well in a thermally isolated stack with heat exchang
ers at both sides. In case of a thermoacoustic couple this assumption does not hold, since
the energy ﬂow at the front of the stack is almost equal to the acoustic loudspeaker out
put, whereas the energy ﬂow at the end of the stack is almost zero. The difference is
caused by heat losses. One way to solve the energy equation for the temperature pro
ﬁle, is to assume a linear energy proﬁle
˙
E(x). This is a rough approximation, since the
heat losses are not constant over the length of the stack, but depend on the temperature
difference with the environment, T(x) −T
a
. As a consequence the results of the steady
state model are not satisfying.
For this reason we developed a new, timedependent, model. This model includes a
more realistic approximation of the heat losses.
Detailed measurement on the temperature ﬁelds in the gas are done by Wetzel and
Herman [5–8]. They used a combination of holographic interferometry and highspeed
82 Temperature proﬁle in a stack
cinematography to measure the local temperature and the heat transfer qualitatively.
Zoontjens et al. [9, 10] have done detailed numerical calculations of the ﬂow and energy
ﬁelds in thermoacoustic couples and have also studied the inﬂuence of different plate
edges on the energy ﬂow.
5.2 Experimental setup
A parallelplate stack is installed in the resonator tube. A cross section of the stack is
shown in ﬁgure 5.1. The plate distance D = 1.0 mm and the plate thickness d = 1.0 mm.
The inner tube diameter 2R
in
= 59.1 mm and the outer diameter 2R
out
= 70.0 mm. The
stack length is 202.0 mm and the porosity is 0.52. The tube wall and the plates are made
of transparent perspex for ﬂow visualization (chapter 6).
59.1 mm
70.0 mm
Figure 5.1: Schematic drawing of a cross section of the stack. The plate distance is 1.0 mm as
is the plate thickness. The center plate, depicted in lighter gray, has a printed circuit board with
temperature sensors on top of it.
To measure the plate temperature T
0
at the different positions in the stack 32 tem
perature sensors are installed on top of the center plate. The temperature sensors are
platinum resistance thermometers
1
. The temperature sensors are connected to a data
acquisition system by a thin ﬂexible printed circuit board, with copper lanes towards a
connector outside of the tube, as is shown in ﬁgure 5.2. To keep the top of the plate ﬂat
and its thickness at 1.0 mm a thin layer is milled out of the plate, such that the printed
circuit board ﬁts in it. At the positions of the temperature sensors, small cavities are
made in the plate to ﬁt the sensors in. The printed circuit board is glued to the plate and
the cavities are ﬁlled with thermallyconducting glue around the temperature sensors.
1
PT1000
5.2 Experimental setup 83
1 32
32
1
∆x
∆x0
∆xend
8.9 mm 61.0 mm
205.6 mm
64 pin connector
Figure 5.2: Schematic drawing of the center plate, viewed from above, with a printed circuit
board with temperature sensors glued on top of it. The sensor are separated by ∆x = 6.5 mm.
The distances to the boundaries, ∆x
0
= 1.8 mm and ∆x
end
= 2.3 mm. Only two measuring leads
are shown, schematically.
84 Temperature proﬁle in a stack
Having the printed circuit board installed on the stack, the temperature sensors are
calibrated again. The temperature plate has one branch with a connector on it. The
printed circuit board is connected to a dataacquisition system, which monitors the re
sistance of the temperature sensors. The resistance of all thermometers is measured at a
sample frequency of 1 kHz, averaged over 1 sec and stored as data.
The resistance of a thermometer depends linearly on the temperature
R(T) = R
0
+β(T −T
r
), (5.1)
with R
0
and β empirical constants and T
r
a reference temperature of 300 K. The ther
mometers have been calibrated by measuring the resistance at room temperature and
after being in a oven at 80
◦
Cfor 6 hours . This way the R
0
and β of all 32 thermometers
are determined. This accuracy of the measured temperature is 0.1
◦
C. The inaccuracy is
mainly caused by small errors in the offset resistance R
0
.
5.3 Model
In section 2.1 three 1D differential equations were derived (Eq. 2.16, 2.30, and 2.34),
dp
1
dx
= −
iωρ
0
A(1 − f
ν
)
U
1
, (5.2a)
dU
1
dx
= −
iωA[1 + (γ −1) f
κ
]
γp
0
p
1
+
f
κ
− f
ν
(1 − f
ν
)(1 −σ)
dT
0
dx
1
T
0
U
1
, (5.2b)
˙
E =
1
2
Re
_
p
1
U
∗
1
_
1 −
f
κ
− f
∗
ν
(1 + P
r
)(1 − f
∗
ν
)
__
+
ρ
0
c
p
[U
1
[
2
2Aω(1 −P
2
r
)[1 − f
ν
[
2
Im( f
κ
+ P
r
f
∗
ν
)
dT
0
dx
−(Ak + A
so
k
so
)
dT
0
dx
, (5.2c)
with three variables (U
1
(x), p
1
(x), and T
0
(x)) and the total energy (
˙
E(x)).
In the Steady State model (SS model) we assume that
˙
E(x) decreases linearly with x
in the stack and solve the system of equations 5.2 with a Runge–Kutta method. Later an
improved model is developed, which includes the time dependency of the temperature
and also heat losses to the environment. This is the Time Dependent model (TD model).
5.3.1 Steady State model
To solve the set of equations 5.2 we need three boundary conditions and we need to
know
˙
E(x). As boundary conditions we use
p
1
(x
L
) = p
L
, (5.3a)
U
1
(x
L
) = U
L
, (5.3b)
T
0
(x
L
) = T
L
, (5.3c)
5.3 Model 85
as is also indicated in ﬁgure 5.3. At position x = x
L
the acoustic energy ﬂow into the
stack is
˙
W
L
=
1
2
Re[p
L
U
∗
L
], (5.4)
and at position x = x
R
the acoustic energy ﬂow out of the stack is
˙
W
R
=
1
2
Re[p
1
(x
R
)U
∗
1
(x
R
)]. (5.5)
TL
pL
UL
WL WR
xL=0 xR=Lp
p1(x) T0(x) E(x) U1(x)
Figure 5.3: Schematic drawing of the SS model. A parallelplate stack is enclosed in a resonator.
x is the position in the stack, with x
L
denoting the left side and x
R
the right side of the stack.
The boundary conditions on the left side of the stack are the pressure p
L
, the volume ﬂow rate
U
L
, and the temperature T
L
. The total energy ﬂow in the stack,
˙
E(x), is assumed to decrease
linearly with x, with
˙
W
L
the incoming energy ﬂow at x = x
L
and
˙
W
R
the outgoing energy ﬂow
at x = x
R
. By solving the set of three differential equations the pressure p
1
(x), the volume ﬂow
rate U
1
(x), and temperature T
0
(x) can be found as functions of the position in the stack.
Unfortunately, we do not know the x dependency of
˙
E(x). But we do know the total
energy at the stack boundaries:
˙
E(x
L
) =
˙
W
L
and
˙
E(x
R
) =
˙
W
R
, since no heat exchangers
are present. Since
˙
W
L
>
˙
W
R
the total power
˙
E(x) must decrease along the stack. We
have to ask ourselves, where this power is going to. Initially, the stack temperature is
close to the environmental temperature and the dissipated acoustic power is completely
used to heat up the stack. Later, when the stack is heated, part of the dissipated acoustic
power ﬂows to the environment. Once the system is in steady state, the acoustic energy
that is put into the stack is equal to the heat losses to the environment. In the SS model
both the heat losses and power required to heat up the stack are not included. The total
power
˙
E(x) is assumed to decrease linearly with the position in the stack x:
˙
E(x) =
˙
W
L
−
˙
W
L
−
˙
W
R
L
p
x. (5.6)
This is a rough assumption, since the heat losses are expected to be proportional with
the temperature difference with the environment, which is not constant over the stack.
The set of three ordinary differential equations 5.2 is solved with a threestage Lo
batto IIIa method, which is an implicit Runge–Kutta method and is a standard function
included in Matlab.
As we have discussed how the equations are to be solved, let us have a look at
the parameters in Eqs. 5.2 and their dependencies on the temperature. Since the stack
86 Temperature proﬁle in a stack
consists of parallel plates, we can use Eq. 2.18b for f
ν
and f
κ
:
f
ν
=
tanh[(1 + i)D
p
/2δ
ν
]
(1 + i)D
p
/2δ
ν
, (5.7)
f
κ
=
tanh[(1 + i)D
p
/2δ
κ
]
(1 + i)D
p
/2δ
κ
, (5.8)
with D
p
= 2y
0
the plate separation and δ
ν
and δ
κ
the boundary layer thicknesses, given
by
δ
ν
=
¸
2µ
ωρ
0
, (5.9)
δ
κ
=
¸
2k
ωρ
0
c
p
. (5.10)
For the dynamic viscosity µ and the thermal conduction k we use the following empiri
cal temperature dependencies:
µ = µ
0
_
T
0
T
r
_
b
µ
, (5.11)
k = k
0
_
T
0
T
r
_
b
k
, (5.12)
where T
r
is a reference temperature at which µ = µ
0
and k = k
0
and b
µ
and b
k
are two
empirical constants. The dimensionless Prandtl number, deﬁned as P
r
= c
p
µ/k, is also
temperature dependant. The ratio of speciﬁc heats is calculated by γ = c
p
/(c
p
− R
s
).
The density is determined by the ideal gas law ρ
0
= p
0
/R
s
T
0
. In our experimental
setup we use atmospheric air. In table 5.1 the constants that are used in the model are
presented.
5.3.2 Time Dependent model
In the Time Dependent (TD) model both the heat losses to the environment and the
heating up of the stack as function of time are included. Another difference with the SS
model is that the temperature over the stack cross section is not assumed to be uniform.
We focus on the middle plate, as that is the plate where the thermometers are located.
The middle plate is shown in ﬁgure 5.4. The plate has thickness d
p
, width W
p
, and
length L
p
. Note that the stack consists of 17 parallel plates, but only the middle one is
shown here, as that is the one we use in the modeling of the energy equation. The plates
are attached to a cylindrical stack holder (dashed circles) with inner radius R
i
and outer
radius R
o
. The stack plates and the stack holder are made of perspex.
The TD model is based on a number of assumptions. The time dependence is as
sumed to be so small that it does not inﬂuence the wave equations. Only the temper
ature is modeled as function of time. But since the pressure and volume ﬂow rate are
5.3 Model 87
Table 5.1: List of constants used in our model.
Gas area A 14.3 cm
2
Solid area A
so
13.1 cm
2
Stack length L
p
205.6 mm
Pressure p
0
1.0110
5
Pa
Viscosity µ
0
1.85 10
−5
Pas
b
µ
0.76
Gas thermal conduction k
0
2.610
−2
W/Km
b
k
0.89
Reference temperature T
r
300 K
Solid conduction k
so
0.155 W/Km
Speciﬁc gas constant R
s
286.7 J/(kgK)
Gas heat capacity c
p
1.003 kJ/(kgK)
Solid heat capacity c
so
1.17 kJ/(kgK)
Plate thickness d
p
1 mm
Plate separation D
p
1 mm
Width of center plate W
p
59.1 mm
R
i
y
z
x
d
p
R
o
L
p
W
p
Figure 5.4: Schematic 3D drawing of the center plate in the stack. The plate has thickness d
p
,
width W
P
, and length L
p
. The plate is located in a cylindrical plate holder, with inner diameter
2R
i
and outer diameter 2R
o
. Note that the other plates in the stack are not shown here and that
the plate dimensions are not on scale.
88 Temperature proﬁle in a stack
x
i
x1
x2
x
i1 x
i+1 x
n
Ti
Ti1 Ti+1
Tn T2
T1
y
x
S1
Si1
Si
Te
Te
WL
WR
Figure 5.5: Schematic drawing of the numerical grid on the center plate of the stack. The plate
is divided in n −1 segments numbered S
1
to S
n−1
. The grid points are located at the borders of
the segments. In total n grid points are located on the dashed center line, indicated with dots.
The x positions of the grid points are x
1
. . . x
n
and the temperatures are T
1
. . . T
n
. The edges of
the plate that connect with the stack holder are assumed to be at environmental temperature T
e
.
The acoustic energy ﬂow on the left side of the plate is
˙
W
L
and on the right side
˙
W
R
.
temperature dependent, they also slowly change in time. This time dependency of p
1
and U
1
is assumed to be much smaller than the acoustic variations. We assume that the
heat losses to the environment can be approximated by
˙ q
e
(x) = α(T(x) −T
e
), (5.13)
where ˙ q
e
(x) is the heat ﬂow, α a constant heat transfer coefﬁcient, T(x) the temperature,
and T
e
the environmental temperature (T
e
= 24
◦
C).
To solve the energy equation for T(x, t) we use a discretization of the plate in the
x direction, which is shown in ﬁgure 5.7. The middle stack plate is in the xdirection
(along the tube axis) divided into segments S
1
. . . S
n−1
, which all have a heat capacity.
At every edge of a segment a grid point for the temperature is located. The x positions of
the grid points are x
1
. . . x
n
and the discreticized temperatures are T
1
. . . T
n
. Each plate
segment has its own energy balance. The total net power of the plate segment divided
by the heat capacity determines the temperature increase dT
i
/dt of the segment. The
edges of the plate that are attached to the stack holder are assumed to be at environ
mental temperature T
e
.
Using the energy equation (Eq. 5.2c) the total power at the two borders of each
segment can be calculated. The ﬁrst term of the energy equation we deﬁne as Y(x),
Y(x) =
1
2
Re
_
p
1
U
∗
1
_
1 −
f
κ
− f
∗
ν
(1 + P
r
)(1 − f
∗
ν
)
__
, (5.14)
is an energy ﬂow and the second term,
Z(x) =
ρ
0
c
p
[U
1
[
2
2Aω(1 −P
2
r
)[1 − f
ν
[
2
Im( f
κ
+ P
r
f
∗
ν
)
dT
0
dx
, (5.15)
is the thermoacoustic pumping effect. The pumping effect can either enhance the thermal
conduction term or compensate for it, depending on its sign. Now let us look at the total
5.3 Model 89
energy balance of segment i. The energy ﬂow into the segment at the left side is Y(x
i
)
and the energy ﬂow out of the segment is Y(x
i+1
). The contribution of the pumping
term on the left side is Z(x
i
)(T
i
−T
i−1
)/∆x and analogous on the right side. The ther
mal conduction on the left side is (Ak + A
so
k
so
)(T
i
− T
i−1
)/∆x and analogous on the
right side. And the last contribution is of the heat losses. The change of temperature as
function of time ∂T/∂t time the heat capacity is equal to the net power:
ρc
p
V
∂T
∂t
= Y(x
i
) −Y(x
i+1
) + (Z(x
i
) + Ak + A
so
k
so
)(T
i
−T
i−1
)/∆x −
(Z(x
i+1
) + Ak + A
so
k
so
)(T
i+1
−T
i
)/∆x −
˙
Q
e
, (5.16)
with V the volume of the segment.
The model of the plate is shown as an electrical network in ﬁgure 5.6.
Wi
Ci
1/(α∆xWp)
Wi+1
Ti
Wi+1
Ci+1
1/(α∆xWp)
Wi+2
Ti+1
Zi
Figure 5.6: Electrical network of the TD model. The voltages V
i
correspond to temperatures
T
i
and the currents correspond to energy ﬂows. The capacitance to earth is the equivalent of
heat capacity. The heat losses are a resistance to earth. The conduction and pumping effect are
integrated in the resistance Z
i
. The energy ﬂows Y
i
are included by current sources.
The heat ﬂows are approximated by Eq. 5.13. We need to determine α. The heat
losses are dominated by conduction in the y direction of the plate towards the stack
holder, which is at temperature T
e
. The temperature proﬁle in the y direction of the
plate is shown in ﬁgure 5.7. The heat ﬂow due to losses is
˙ q = k
so
_
_
∂T
∂y
¸
¸
¸
¸
y=−W
p
/2
−
∂T
∂y
¸
¸
¸
¸
y=W
p
/2
_
_
. (5.17)
For a quadratic proﬁle it follows that α = 8k
so
/W
p
.
The initial temperature proﬁle that is used for the model is uniformly room temper
ature. At this temperature proﬁle T(x) we calculated p(x) and U(x) using Rott’s wave
equations (Eqs. 2.16 and 2.30). Consequently
∂T
∂t
is calculated, from which T(x) one
timestep later can be determined. The new determined T(x) can be used for calculating
the new p(x) and U(x) and the whole process is repeated again.
This whole process is repeated timestep after timestep until a steady state is reached.
90 Temperature proﬁle in a stack
W
p
/2 W
p
/2
0
T
T
e
T
c
y
Figure 5.7: Model of the plate temperature as function the y position, with y = 0 the center of
the plate, W
p
the plate width, T
e
the temperature at the stack holder, and T
c
the temperature of
the center of the plate.
5.4 Heat losses
When performing a longterm measurement, after a while the system reaches a steady
state. The amount of acoustic power dissipated in the stack is equal to the heat loss
of the stack to the environment. We will study how this energy is transferred to the
environment. We ask ourselves, how much is lost by radiation, conduction, forced con
vection, and free convection? We consider the resonator tube wall as a heat sink at
environmental temperature. Due to its thickness and good heat conductivity, the tube
wall remains a at environmental temperature.
To study which part of the heat loss can be accounted to each heat loss effect, we
monitor the temperature in the stack when the system is in steady state and the speaker
is suddenly turned off. Two things can be noticed. The temperature gradient in the
stack will slowly decay, due to the absence the thermoacoustic effect. Additionally, the
internal energy in the stack as a whole, will decay due to heat ﬂow to the surrounding.
The heat losses in the system with the speaker turned off, are due to radiation, free
convection and conduction.
With static heat losses in a stack we mean the heat losses of the stack plates to the
environment with absence of an acoustic ﬂowthrough the stack. The x direction is along
the tube axis, the z direction is perpendicular to the stack plates, and the y direction is
parallel to the plates and perpendicular to the tube axis.
The temperature in the plate as a function of time is described by the Fourier equa
tion:
ρc
∂T
∂t
= ∇ (k∇T), (5.18)
As the plate thickness (1 mm) is much smaller than the length, the temperature is as
sumed to be uniform in the zdirection and ∇ (k∇T) = k(∂
2
T/∂x
2
+ ∂
2
T/∂y
2
). The
k∂
2
T/∂y
2
is the main cause of the heat losses to the environment. The tube wall is close
to environmental temperature and is effectively a heat sink.
In ﬁgure 5.8 two cooling down measurements are presented. The temperature pro
ﬁle, i.e. the temperature as function of the position, in the stack is shown at six different
moments in time. At time t = 0 the loudspeaker is turned off and the temperature
measurement is initiated. The difference between ﬁgure 5.8a and 5.8b is the initial tem
5.5 Temperature proﬁle measurement 91
perature proﬁle. The initial temperature difference is 30
◦
C in ﬁgure 5.8a and 60
◦
C in
5.8b.
0 0.05 0.1 0.15 0.2
10
20
30
40
50
x [m]
T
[
o
C
]
t =0 s
t =365 s
t =731 s
t =1463 s
t =2927 s
t =7818 s
t =0 s
365 s
731 s
1463 s
2927 s
7818 s
(a)
0 0.05 0.1 0.15 0.2
0
20
40
60
80
x [m]
T
[
o
C
]
t =0 s
t =365 s
t =731 s
t =1463 s
t =2927 s
t =7818 s
t =0 s
365 s
731 s
1463 s
2927 s
7818 s
(b)
Figure 5.8: The temperature as function of the position, measured at 32 positions, at six different
moments in time after the loudspeaker is turned off. The solid curves represent measured data
and the dashed curves are calculated numerically. The initial temperature difference is 30
◦
C in
ﬁgure (a) and 60
◦
C in (b).
5.5 Temperature proﬁle measurement
5.5.1 Time dependent measurements
We recorded the temperature proﬁles as functions of time during a period of 1600 s. The
measured temperature proﬁles are shown at ﬁve moments in time by the solid curves
in ﬁgure 5.9a. The initial temperature proﬁle is almost constant around 23
◦
C. The
stack is located in a region with a positive temperature gradient. When the speaker is
turned on, the temperature ﬁrst starts to deviate from the initial temperature proﬁle at
the edges of the stack, as can be seen after 25 s. At the edges the difference with the
critical temperature gradients is the largest. After 100 s the temperature proﬁle is still
ﬂat in the middle of the plate (50 mm< x <180 mm). At t = 400 s a nonzero temperature
gradient starts to develop in the middle of the plate and at t = 1400 s the proﬁle is almost
linear.
The measured initial temperature is used as the initial temperature for the TDmodel.
The calculated temperature proﬁle is shown by the dashed curves. The model corre
sponds quite well with the experiments. This justiﬁes the assumptions made in the
model, regarding heat losses. Especially at t = 25 and t = 100 the model and experiments
are in good agreement with each other. At t = 400 and t = 1400 the calculated tempera
ture gradients at the edges are signiﬁcantly higher than the measured ones. This can be
explained by extra heat losses at the stack edges, mainly due to convection, which are
not included in the TD model.
92 Temperature proﬁle in a stack
0 0.05 0.1 0.15 0.2
0
10
20
30
40
50
x [m]
T
[
o
C
]
t =0 s
t =25 s
t =100 s
t =400 s
t =1400 s
0 s
25 s
100 s
400 s
1400 s
(a)
0 0.05 0.1 0.15 0.2
0
20
40
60
80
100
x [m]
1
T
[
o
C
]
1
SS model
1
TD model
1
initial
1
ﬁnal
1
(b)
Figure 5.9: Measurements and calculations of the temperature proﬁle (T(x)) in a stack at a
frequency of 50 Hz.
a) The temperature proﬁle determined by measurement (solid curves) and for the TD model
(dashed curves) at different moments in time.
b) The ﬁnal temperature proﬁle determined by measurement and for both models.
The measured and calculated proﬁle share a certain point (or small region) where
the temperature stays at (or close to) the environmental temperature during the whole
measurement. This point is located at x = 80 mm. When the thermacoustic effect would
only involve the pumping of heat, this point is expected to be in the middle of the plate
(x = 103 mm). Since the thermoacoustic effect also leads to the heating of the whole
plate, this point is shifted to the left.
In ﬁgure 5.9b the ﬁnal (t=1660 s) temperature proﬁles are shown. The solid curve
is calculated by the SS model, the dashed curve by the TD model (after 1660 s), and
the dashdotted curve is measured. For the SS model we use the measured ﬁnal tem
perature at x = 0 as the T
L
boundary condition. The SS model predicts a much higher
temperature difference over the stack than the TD model and measurement. The system
is in steady state after 1660 s, which was veriﬁed by continuing the TD simulation until
t = 5000 and noticing no signiﬁcant changes in the temperature proﬁle. As the system is
in steady state, the difference between the SS model and the measurements can only be
explained by heat losses to the environment which are not included in the SS model.
5.5.2 Amplitude sweep
Here we study the ﬁnal temperature difference ∆T = T
R
− T
L
over the stack. The
measurement is continued until ∆T converged and this measurement is repeated at
different drive ratios. In ﬁgure 5.11a the temperature difference is shown as function of
the pressure amplitude at the left side of the stack, at a frequency of 55 Hz. The circular
dots are measurement points, the solid curve represents the SS model and the dashed
curve the TD model. The dashdotted line is the critical temperature difference. In
5.5 Temperature proﬁle measurement 93
0 0.05 0.1 0.15 0.2
18
20
22
24
26
28
x [m]
T
[
o
C
]
t =0 s
t =25 s
t =100 s
t =400 s
t =1400 s
0 s
25 s
100 s
400 s
1400 s
(a)
0 0.05 0.1 0.15 0.2
10
20
30
40
50
60
70
x [m]
1
T
[
o
C
]
1
SS model
1
TD model
1
initial
1
ﬁnal
1
(b)
Figure 5.10: Measurements and calculations of the temperature proﬁle (T(x)) in a stack at a
frequency of 80 Hz.
a) The temperature proﬁle determined by measurement (solid curves) and for the TD model
(dashed curves) at different moments in time.
b) The ﬁnal temperature proﬁle determined by measurement and for both models.
ﬁgure 5.12a the acoustic energy ﬂow into the stack W
in
(at the left axis) and the velocity
amplitude [u
L
[ (on the right axis) are shown as a function of the pressure amplitude [p
L
[,
at 55 Hz. The acoustic energy ﬂow is proportional with [p
L
[
2
and [u
L
[ is proportional
with [p
L
[.
A similar measurement is performed at 158 Hz. The temperature differences at 158
Hz are shown in ﬁgure 5.11b and the acoustic energy ﬂow and velocity amplitude are
shown in ﬁgure 5.12b.
The critical temperature is determined by integrating the critical temperature gradi
ent over the stack
(∆T)
c
=
_
x
R
x
L
_
dT
dx
_
c
dx, (5.19)
with (
dT
dx
)
c
given by Eq. 2.27. The [p
1
[ and [U
1
[ that are used in Eq. 2.27 are calculated
with the SS model, at the highest [p
L
[. It has been veriﬁed that when [p
L
[ is increased
further the temperature difference predicted with the SS model approaches (∆T)
c
.
5.5.3 Frequency sweep
In ﬁgure 5.13a the temperature difference over the stack is plotted versus the frequency.
The stack is located at 69 mm from the closed end of the resonator. In ﬁgure 5.13b the
drive ratio and velocity are shown. During this measurement the loudspeaker voltage is
constant. The resonance frequencies of the resonator are at 55 Hz and at 165 Hz, which
can be clearly seen from the peaks in the pressure and the velocity. The temperature
difference also shows a peak at these frequencies, as the acoustic power input to the
94 Temperature proﬁle in a stack
0 5 10
0
20
40
60
80
100
120
p
L
[kPa]
∆
T
[
o
C
]
SS model
TD model
measurement
(∆T)
c
(a) f =55 Hz
0 0.5 1 1.5 2
0
20
40
60
80
100
p
L
[kPa]
∆
T
[
o
C
]
SS model
TD model
measurement
(∆T)
c
(b) f =158 Hz
Figure 5.11: The temperature difference as function of the pressure amplitude [p
L
[ at the left
side of the stack. The measurement is performed at 55 Hz and at 158 Hz. The temperature
difference is determined from a measurement as well as from both models.
0 5 10 15
0
0.2
0.4
W
i
n
[
W
/
c
m
2
]
p
L
 [kPa]
0 5 10 15
0
5
10

u
L

[
m
/
s
]
(a) f =55 Hz
0 1 2 3
0
0.05
0.1
W
i
n
[
W
/
c
m
2
]
p
L
 [kPa]
0 1 2 3
0
2
4

u
L

[
m
/
s
]
(b) f =158 Hz
Figure 5.12: The energy ﬂow
˙
E
0
that is applied to the stack and the velocity amplitude [u
L
[
as functions of the pressure amplitude [p
L
[ at the left side of the stack. The measurement is
performed at 55 Hz and at 158 Hz.
stack is higher and the stack has not reached the critical temperature yet. The TD model
is in very good agreement with the measurements. The SS model overestimates the
temperature difference, as it does not include heat losses. The shape of the temperature
plots is determined heavily by the pressure and velocity dependence of the frequency.
In ﬁgure 5.14a the temperature difference over the stack is plotted versus the fre
5.6 Discussion and conclusion 95
50 100 150
0
50
100
150
f [Hz]
∆
T
[
o
C
]
SS model
TD model
measurement
(∆T)
c
(a)
50 100 150
0
0.05
0.1
D
R
f [Hz]
50 100 150
0
0.02
0.04

u
L

/
c
0
DR
u
L
/c
0
(b)
Figure 5.13: a) The temperature difference T
R
−T
L
in steady state as function of the frequency,
determined by the SS model, by the TD model and from measurements. The calculated critical
temperature difference (∆T)
c
is also plotted. b) The drive ratio D
r
= [p
L
[/p
0
and the Mach
number [u
L
[/c
0
at the left side of the stack as function of the frequency. The stack is located at
69 mm from the end of the resonator.
quency. In ﬁgure 5.14b the normalized pressure and velocity are shown. The stack is
located at 469 mm from the closed end of the resonator. During this measurement the
loudspeaker voltage is constant. The resonance frequencies of the resonator is at 43 Hz.
This resonance frequency is lower than in the previous measurement, because in this
case the tube is 40 cm longer. The temperature difference shows a peak at the resonance
frequency, but this peak is ﬂatter than in the previous measurement. The reason for
this is that the temperature is close to the critical temperature and therefore can not go
higher. The critical temperature difference is much lower this time, because the stack is
much farther away from the resonator end. The temperature difference goes towards
zero at a frequency of 120 Hz, as a knot in the pressure amplitude reaches the stack.
At one part of the stack the temperature gradient is positive and at the other part it
is negative. The critical temperature gradient does not go to zero because the critical
temperature gradient is always positive, by deﬁnition. At even higher frequencies the
temperature difference becomes negative.
The TD model is in very good agreement with the measurements, except at the res
onance frequency, where it approaches the SS model.
5.6 Discussion and conclusion
Literature shows many to have measured the temperature difference between the two
stack ends, or between two heat exchangers, but, as far as we know, this is the ﬁrst time
that the temperature proﬁle in a stack is measured. We have measured the temperature
at 32 positions in the stack solid. Wetzel and Herman have visualized and quantiﬁed
96 Temperature proﬁle in a stack
0 50 100 150 200
−20
0
20
40
60
f [Hz]
∆
T
[
o
C
]
SS model
TD model
measurement
(a)
50 100 150 200
0
0.01
0.02
0.03
0.04
0.05
0.06
f [Hz]
D
R
,

u
L

/
c
0
DR
u
L
/c
0
(b)
Figure 5.14: a) The temperature difference T
R
−T
L
in steady state as function of the frequency,
determined by the SS model, by the TD model and from measurements. The calculated critical
temperature difference (∆T)
c
is also plotted. b) The drive ratio D
r
= [p
L
[/p
0
and the Mach
number [u
L
[/c
0
at the left side of the stack as function of the frequency. The stack is located at
469 mm from the end of the resonator.
the temperature ﬁelds in the neighborhood of a single plate with a combination of holo
graphic interferometry and highspeed cinematography.
Not only did we measure the steadystate temperature proﬁle, but also the dynamic
effects. We made a model of the energy balance in the center plate of the stack. We
have included the heat capacity of the plate, to model the timedependency. For the
temperature in the radial direction a parabolic proﬁle in the plates is assumed.
The modeled temperature proﬁles as functions of time are consistent with the mea
surements. So the model works well for a stack that is not thermally isolated.
For further research it would be interesting to use thermometers at different radial
positions to verify the parabolic temperature proﬁle in this direction. The heat losses
to the environment could be reduced signiﬁcantly by reducing the thermal contact be
tween the stack and the resonator tube.
Bibliography
[1] J. Wheatley, T. Hoﬂer, G. Swift, and A. Migliori, “Intrinsically irreversible ther
moacoustic heat engine.,” Journal of the Acoustical Society of America, vol. 74, no. 1,
pp. 153–170, 1983.
[2] A. Atchley, T. Hoﬂer, M. Muzzerall, M. Kite, and C. Ao, “Acoustically generated
temperature gradients in short plates,” Journal of the Acoustical Society of America,
vol. 88, no. 1, pp. 251–263, 1990.
[3] A. Piccolo and G. Cannistraro, “Convective heat transport along a thermoacoustic
couple in the transient regime,” International Journal of Thermal Sciences, vol. 41,
no. 11, pp. 1067–1075, 2002.
[4] P. in ’t panhuis. PhD thesis.
[5] M. Wetzel and C. Herman, “Experimental study of thermoacoustic effects on a
single plate. part i: Temperature ﬁelds,” Warme und Stoffubertragung Zeitschrift,
vol. 36, no. 1, pp. 7–20, 2000.
[6] M. Wetzel and C. Herman, “Experimental study of thermoacoustic effects on a
single plate. part ii: Heat transfer,” Warme und Stoffubertragung Zeitschrift, vol. 35,
no. 6, pp. 433–441, 1999.
[7] M. Wetzel and C. Herman, “Limitations of temperature measurements with holo
graphic interferometry in the presence of pressure variations,” Experimental Ther
mal and Fluid Science, vol. 17, no. 4, pp. 294–308, 1998.
[8] M. Wetzel and C. Herman, “Accurate measurement of highspeed, unsteady tem
perature ﬁelds by holographic interferometry in the presence of periodic pressure
variations,” Measurement Science and Technology, vol. 9, no. 6, pp. 939–951, 1998.
[9] L. Zoontjens, C. Howard, A. Zander, and B. Cazzolato, “Numerical study of ﬂow
and energy ﬁelds in thermoacoustic couples of nonzero thickness,” International
Journal of Thermal Sciences, vol. 48, no. 4, pp. 733–746, 2009.
[10] L. Zoontjens, C. Howard, A. Zander, and B. Cazzolato, “Numerical comparison of
thermoacoustic couples with modiﬁed stack plate edges,” International Journal of
Heat and Mass Transfer, vol. 51, no. 1920, pp. 4829–4840, 2008.
98 BIBLIOGRAPHY
Chapter 6
Flow visualization in and around
a stack
6.1 Introduction
The last decades a vast amount of research in the ﬁeld of thermoacoustics has been
conducted [1]. Rott [2–7] and Swift [8, 9] have derived analytical equations describing
thermoacoustics. This linear theory, however, is only valid at a relatively small drive
ratio, which is deﬁned as the pressure amplitude to average pressure ratio. The linear
theory is in good agreement with measurement for drive ratios smaller than 10% [10].
The theory is used as basis for designing thermoacoustic applications. In many appli
cations, however, large drive ratios (more than 10%) are used. At larger drive ratios,
unfortunately, the coefﬁcient of performance (COP), which is deﬁned as the cooling
power to work ratio, decreases signiﬁcantly. This might be attributed to nonlinear ef
fects occurring at these high drive ratios, like turbulence, vortex formation, streaming,
and higher harmonics. Therefore the study of nonlinear effects has become an impor
tant issue in thermoacoustics. This chapter focusses on the vortex formation at the end
of a parallelplate stack and the velocity proﬁle in the middle of the stack.
Steady ﬂow around obstacles, especially circular cylinders, is wellknown [11]. In
case of R
e
< 1 (the Reynolds number is deﬁned here as R
e
=
V
∞
D
ν
, with V
∞
the velocity
far away from the cylinder, D the cylinder diameter, and ν the kinematic viscosity) the
Stokes approximation can be used. The vorticity, which is created close to the cylinder
surface due to the noslip boundary condition, is not advected, resulting in a symmetri
cal ﬂow. At increasing Reynolds numbers twin vortices appear (4 < R
e
< 40), a vortex
street develops (40 < R
e
< 200), which ﬁnally becomes unstable (R
e
> 200), and even
turbulent (R
e
> 5000).
Oscillatory ﬂows around single circular cylinders have been studied recently by Ne
hari et al. [12], Anagnostopoulo and Minear [13], Wybrowet al. [14], and others. Parallel
plates have a rectangular shape instead of circular. Ozgoren [15] studied the ﬂow struc
ture behind both square and circular cylinders using DPIV. A stack of parallel plates
is much different from a single circular cylinder. Not much is known about oscillatory
ﬂow around rectangular cylinders and certainly not about oscillatory ﬂow around a
100 Flow visualization in and around a stack
stack of parallel plates. Because the length of the plates (130 mm) is much larger than
the acoustic displacement (< 7 mm), the two stack edges can be studied separately. A
stack edge can be modeled as an oscillating ﬂowthrough an array of channels into a vol
ume. This geometry is very similar to a channel with a sudden change in cross section.
Oscillating ﬂow in channels with a sudden change in crosssection has been studied by
Ibrahim and Hashim [16] and Morris et al. [17] numerically and by Smith [18, 19] and
Skulina [20, 21] experimentally, using PIV and microphone measurements. Oscillatory
ﬂow through a rapid expansion was also studied by Smith [22, 23]. This geometry is
similar geometry to a jet pump, which was studied by Petculescu and Wilen [24]. Al
ready in 1979 Disselhorst and Van Wijngaarden [25, 26] studied oscillatory ﬂow near
tube entrances both experimentally and theoretically. They used schlieren to study the
vortex shedding near both sharp and round edges and showed that different vortex
shedding modes exist, depending on the initial state. We will show both numerically
and experimentally that a similar behavior occurs for the ﬂow at the end of a stack plate.
Not much research has been conducted on the ﬂow around a stack of parallel plates
speciﬁcally. Stoltenkamp studied steady ﬂow around stacked plates at an angle con
cerning a rotor [27, 28]. Oscillating ﬂow around stacked plates has been studied exper
imentally using PIV and also numerically by BlancBenon et al. [29] at relatively low
amplitudes, where a vortex pair originates behind the stack plates and no vortex streets
appear. Berson et al. [30–32] and Mao et al. [33] recently have measured a vortex street
behind parallel plates at higher amplitudes using PIV. The interaction of the ﬂow with
a stack and heat exchangers is studied numerically by Besnoin and Knio [34]. The heat
exchange, velocity ﬁeld, vorticity and temperature distribution at the end of a stack
are calculated. However, until now, no systematic experimental research has been un
dertaken on the ﬂow patterns behind stack plates, in which all the relevant parameters
(plate thickness and separation, velocity amplitude, frequency, and plateend shape) are
varied and their inﬂuence on the ﬂow structure is studied.
This chapter focusses on oscillatory ﬂow through a stack. In section 6.2 the experi
mental setups are described. PIV measurements of the vortex shedding behind a stack
plate are discussed in section 6.3, the inﬂuence of different dimensionless numbers in
section 6.4, different plateend shapes in section 6.5, numerical simulations of the vor
tex shedding in section 6.6, and the evolution of a vortex street in section 6.7. In section
6.8 the velocity proﬁle inbetween two parallel plates is discussed, including a laminar
turbulent transition. The natural convection at the hot end of a stack is discussed in
section 6.9 and the streaming velocity at the end of a stack is discussed in section 6.10.
We conclude with section 6.11.
6.2 Experimental Setup
For the ﬂow visualization we use a method called Particle Image Velocimetry (PIV). For
this method we inject small oil droplets (with a mean particle diameter of 0.2 µm) into
the setup, which follow the acoustic ﬂow and are used as tracer particles to visualize
the ﬂow. The movement of the particles is recorded by a combination of a camera and a
laser. The details of the PIV method and hardware are discussed in section 6.2.3.
Two different setups were built for PIV measurements. The ﬁrst setup consisted of
a perspex resonator tube and is discussed in detail in section 6.2.1. This setup was a
6.2 Experimental Setup 101
relatively simple one, in the sense that it was easy and cheap to build. The idea of this
setup was mainly to test and familiarize ourselves with the PIVhardware and software,
and also the particle generator and particle behavior in an acoustic ﬂow. Although this
setup was relatively simple, it enabled us to obtain many results of the vortex shedding
behind plates.
With the experience gained with the perspex setup, we designed a second, more
advanced, PIV setup. The Aluminium setup was designed to have a number of im
provements in comparison to the perspex setup:
• Thick aluminium resonator walls, to allow for accurate acoustic measurements,
which were not possible in the perspex setup.
• Obtain higher drive ratios and velocities, to study nonlinear effects.
• Longer resonator tube, to use lower frequencies and therefore larger boundary
layers.
• Larger inner diameter of resonator tube, to increase the number of plates in a stack.
• More ﬂexibly in stack position, tube length, plate thickness, and separation.
The geometry, loudspeaker, and microphone details of this setup were already dis
cussed in section 3.3 and the ﬂow visualization part of this setup is described in section
6.2.2.
6.2.1 Perspex setup
The perspex setup consists of a loudspeakerdriven standingwave resonator. The res
onator is a cylindrical perspex tube, with an inner diameter of 25 mm and a total length
of 1400 mm (ﬁgure 6.1). The tube is closed with a transparent end plate on the left end,
connected to a Dynaudio D54AF loudspeaker on the right end, and is ﬁlled with atmo
spheric air at room temperature. The ﬁrst resonance frequency in the setup occurs at
125 Hz. At different locations microphones are inserted in the tube wall to measure the
pressure. A stack of parallel plates is placed 350 mm in front of the endplate. The ﬂow
ﬁeld measurements occur at the left end of the stack. A closeup of this measurement
area is shown at the bottom of the ﬁgure. At the measurement window (ﬁeld of view)
the tube has a smaller wall thickness in order to increase its optical transparency. The
observation window is still thick enough not to act as a membrane and start vibrating.
The measurement window (ﬁeld of view) is in the center of the tube and perpendicular
to the stack plates.
In most thermoacoustic devices, in general, the ﬁrst resonance frequency is of the
order 0.1  1 kHz. In our experiments we have chosen a relatively low resonance fre
quency: 125 Hz. A low frequency has two advantages for visualization:
1. it lowers the required time resolution of the laser and camera;
2. we can increase the plate separation of the stack as a consequence of larger pene
tration depths.
3. due to larger scales, visualization is easier.
102 Flow visualization in and around a stack
Figure 6.1: Schematic drawing of the experimental setup. At the top a resonance tube connected
to a loudspeaker. (a) Transparent endplate, (b) injection nozzles for seeding, (c) stack of parallel
plates, (d) loudspeaker. At the bottom a closeup of the measurement area. The dotted area is the
measurement window, which is 20 mm 15 mm in size.
6.2 Experimental Setup 103
The 2nd advantage follows from the penetration depths being proportional to ω
−1/2
(Eq. 2.12). The stacks that are used in this setup consist of ﬁve parallel perspex plates.
The plate thickness d = 1 mm and the plate distance D = 4 mm. To study the inﬂuence of
the plate thickness in some measurements stacks with a plate thickness of 0.5 mm and a
plate distance of 4.5 mm are used, as well as single plates with a thickness of 3 mm. The
stack length is 130 mm. To reduce laser light reﬂections all stack plates are coated black.
6.2.2 Aluminium setup
The geometry of the Aluminium setup, the positions of the microphones, and the
speaker details are discussed in section 3.3. Here we will focus on the PIV Lavision
part of the setup. In ﬁgure 6.2 a part of the resonator tube and the laser optics are
shown. Two different conﬁgurations are used to create a light sheet at the measurement
window. In conﬁguration (a) the light sheet enters the resonator through the transpar
ent end plate, whereas in conﬁguration (b) the light sheet enters the resonator tube from
above. For most measurements both conﬁgurations can be used, but for measurements
in the middle of the stack conﬁguration (a) is required and for measurements in the
regeneratorbased setup (chapter 7) conﬁguration (b) is required. The laser and the
ﬁrst 45
o
mirror are standing on an horizontal rail, beneath the table. The positions of
the mirrors as well as their angle can be modiﬁed for a good alignment of the laser beam
and laser sheet. The laser beam passes through a hole in the table and is then reﬂected
by a second mirror. In conﬁguration (a) this mirror is located at the tube axis and in
conﬁguration (b) above the tube axis. The laser beam is turned into a light sheet by the
light sheet optics. The focal point and the angle of the light sheet can be manually set.
In conﬁguration (a) the light sheet passes the transparent end plate. In conﬁguration
(b) the light sheet is reﬂected by a third mirror and enters the resonator tube, which
is transparent at this location. The focal point of the light sheet is at the measurement
window and the angle is such that at least the whole measurement window is covered.
The CCD camera is located perpendicular to the plane of the ﬁgure and is directed at
the measurement window. In front of the camera zoom lens a ﬁlter is placed, to reduce
the inﬂuence of background light. Only light in a wavelength range around that of the
laser can pass the ﬁlter. To reduce laser reﬂections we use ﬂuorescent paint. Fluorescent
paint converts the wavelength of the original laser light into a different wavelength that
cannot pass through the ﬁlter in front of the camera.
The camera is located on a horizontal rail, so we can easily change the position of
the measurement window along the tube. The light sheet optics and third mirror of
conﬁguration (a) are attached to a frame that moves along with the camera over the
horizontal rail. A photo of this frame, the camera, the horizontal rail, and the stack
is shown in ﬁgure 6.3. A second horizontal rail is used to change the distance of the
camera to the resonator tube, in order to vary the size of the measurement window.
104 Flow visualization in and around a stack
45
o
mirror
45
o
mirror
lightsheet optics
measurement window
doublepulsed laser
stack
transparent window
(a) a light sheet through the transparent end plate
45
o
mirror
45
o
mirror
lightsheet optics
measurement window
doublepulsed laser
stack
transparent window
45
o
mirror
(b) a light sheet from above
Figure 6.2: Schematic drawing of the Aluminium setup.
(a) The laser beam from a doublepulsed laser is reﬂected by two 45
o
mirrors. Here the beam
goes through the lightsheet optics and a light sheet is formed that enters the tube through a
transparent end plate and illuminates the particles in the measuring window.
(b) The beam goes through the lightsheet optics above the tube, a light sheet is formed that is
reﬂected by a 45
o
mirror and enters the tube from above.
In both conﬁgurations the CCD camera is perpendicular to the plane of picture aimed at the
measurement window.
6.2 Experimental Setup 105
6.2.3 Particle Image Velocimetry
Particle Image Velocimetry (PIV) is an optical method of ﬂow visualization, that can
provide an accurate quantitative measure of the instantaneous ﬂow ﬁeld across a planar
area of a ﬂow ﬁeld. It is used to obtain an instantaneous 2D velocity vector ﬁeld in and
around a stack. The ﬂuid is seeded with tiny tracer particles, oil droplets smaller than
1 µm, which are assumed to faithfully follow the ﬂow dynamics. This method is to a
large degree nonintrusive, as the added oil droplets cause negligible distortion of the
ﬂuid ﬂow. It is the motion of these seeding particles that is used to calculate velocity
information of the ﬂow being studied. During PIV, the particle concentration is such
that it is possible to identify individual particles in an image, but not with certainty to
track it between images. Therefore, instead of following individual particles, particle
clusters are followed.
Using a light sheet, formed by passing a double pulsed laser beam through light
sheet optics, the particles in the ﬂow are illuminated twice with a small time separation
∆t. One picture is taken at t
1
= t − ∆t/2 and a second at t
2
= t + ∆t/2. The time
between the pictures, ∆t, is chosen such that the maximum particle displacement is
roughly 5 pixels, which is 1/6 of the size of an interrogation window [35]. With the
illumination of these two short duration laser ﬂashes, two pictures of the illuminated
particles are captured by a chargecoupled device (CCD) camera and are stored in two
separate frames. The frames are split into a large number of interrogation windows,
as is shown in ﬁgure 6.4. It is then possible to calculate a displacement vector for each
window with the help of signal processing and crosscorrelation techniques. The spatial
displacement that produces the maximum crosscorrelation statistically approximates
the average displacement of the particles in the interrogation cell. Velocity associated
with each interrogation spot is the displacement divided by the time between the laser
pulses ∆t. The size of the interrogation window is chosen to have at least 6 particles per
window on average.
With PIV a subpixel accuracy can be reached. But since the correlation function is
determined on the discrete pixels, the determination of its maximum is biased toward
integer values. This is called the peaklocking effect, which can be problematic, espe
cially if the particles are smaller than one pixel and the particle concentration is low. In
our measurements the probability density function of the displacement did not show
a bias towards integer values, so we concluded that the peaklocking effect was not an
issue for us.
In our setups we use a 15 Hz PIV System from LAVISION, including a double laser,
CCD, optics and software. For illumination a New Wave NDYAG laser is used, which
produces 2 x 50 mJ of energy with a pulse length of 5 ns. Laser sheet optics in front of the
laser create a laser sheet that coincides with the measurement area and has a thickness
of 0.5 mm. The optics consist of a spherical lens and cylindrical lens combination. The
cylindrical lens expands the laser into a plane while the spherical lens compresses the
plane into a thin sheet. Images are recorded with a 1600x1200 pixels
2
Imager Pro CCD
camera.
The PIV system is calibrated by placing a calibration grid at the ﬁeld of view in the
plane of the light sheet. This calibration is used to convert the displacement in pixels to
physical length scales and also to correct for spherical abberations due to the cylindrical
shape of the resonator tube.
106 Flow visualization in and around a stack
Figure 6.3: Photo of the PIV part of the aluminium setup. (a) resonator tube, (b) 45
o
mirror,
(c) 2 mm parallelplate setup, (d) camera lens, (e) horizontal rail, (f) bellows, (g) CCD camera.
Figure 6.4: Schematic view of the crosscorrelation method that is used to determine the dis
placement vectors within an interrogation window. Two frames time ∆t apart are shown. Both
frames are divided in a number of interrogation windows (16 in this picture, but much more
in the actual experiment). A closeup of one interrogation window of each frame is shown. We
determined the crosscorrelation function of the two frames, by varying the displacement ∆x and
∆y. By taking the highest peak, we ﬁnd the best displacement vector (∆x, ∆y).
6.3 A typical measurement 107
Diethyl hexyl sebacate (DEHS) seeding particles are created by a Palas AGF2.0 seed
ing generator. The mean particle size, distributed by the seeding generater, is 200 nm,
with a maximum particle size of 1 µm. An airﬂow through the resonance tube is gen
erated during the preparation of each measurement, to create a homogeneous aerosol
distribution over our measuring volume. The particles are injected through a nozzle in
front of the measurement section and are released through a nozzle behind the mea
surement section. After closing the nozzles, the seeding density in the measurement
section is sufﬁciently high. When the injection nozzles are closed, the injection ﬂow has
to be decayed sufﬁciently, before measurements can be performed. This is important for
streaming measurements in particular. From test measurements it was concluded that
1 minute after closing the injection nozzles, no signiﬁcant ﬂow (smaller than 0.2 mm/s)
was present in the measurement area.
Before starting the PIV algorithm we use an image preprocessing method, called
”subtract a sliding background”. With this method large intensity ﬂuctuations in the
background due to laser reﬂections are ﬁltered out while small intensity ﬂuctuations
of the particle signal will pass through. This way we receive an image with a con
stant background level without affecting the particle signal we correlate on. The cross
correlation algorithm is used iteratively, with decreasing the interrogationwindow size
each time. In the ﬁnal crosscorrelation step the interrogation windows are 32 by 32
pixels
2
.
6.3 A typical measurement
The vortex formation is studied using PIV at the left end side of a stack. In this typical
measurement the stack consists of 5 rectangular plates with the following geometry
parameters: d = 1.0 mm, D = 4.0 mm, and L = 130 mm. The Reynolds and Strouhal
numbers are here deﬁned as:
R
e
=
[u
1
[d
ν
, (6.1)
S
t
=
f d
[u
1
[
, (6.2)
with [u
1
[ the peak velocity of the main ﬂow just outside the stack. The peak velocity
inside the stack is higher than V due to a smaller gas area inside the stack. The ﬂow
can be characterized by f = 125 Hz, [p
1
[ = 3.7 kPa, [u
1
[ = 5 m/s. This corresponds to
R
e
= 330 and S
t
= 0.025. The viscous penetration depth δ
ν
= 0.2 mm. The phase angle
at which the measurement has been performed ωt
0
= π. This is the moment at which
the main ﬂow is at rest and the particles are at their leftmost excursion. During the mea
surement one picture is taken at t
1
= t
0
− ∆t/2 and a second at t
2
= t
0
+ ∆t/2. The
time between the pictures, ∆t, is chosen such that the maximum particle displacement
is roughly 5 pixels, which is 1/6 of the size of an interrogation window [35]. By a cross
correlation algorithm the two pictures are compared and the particle displacement at
different positions is determined. The particle displacement is divided by the time dif
ference ∆t, resulting in a velocity ﬁeld that is shown in the upper plot of ﬁgure 6.5. The
vector lengths and colors are a measure of the magnitude of the velocity vectors. The
color bar to the right of the picture indicates which vector color corresponds to which
108 Flow visualization in and around a stack
velocity. The black rectangles represent a part of the stack plates. In total 10075 vec
tors are determined, which corresponds to a resolution of 14 vectors/mm
2
. However, to
keep the individual vectors visible, the resolution of the images in this paper are lower.
The relative errors of the measured velocities are typically 1% of the maximum velocity.
When studying vortices originating at the end of stack plates the vorticity
ω
z
=
∂v
∂x
−
∂u
∂y
, (6.3)
is a useful parameter. A vorticity plot of the same measurement is shown in the bottom
plot of ﬁgure 6.5. The color bar at the right side of the plot shows the correspondence
between color and vorticity. At the end of each plate a street of alternating vortices is
present. Note that there is a slight phase shift between the forming of vortices behind
the different plates. Sufﬁciently far away from the stack (more than ﬁve times the dis
placement length) the ﬂow in the stack is axisymmetrical. The asymmetry of the ﬂow
patterns can be caused by small asymmetries in the geometry. In section 6.6 we will
show that the same asymmetries are also present in numerical simulations with a sym
metrical geometry.
6.4 Different categories of vortex patterns and dimension
less numbers
Figure 6.7 is a simpliﬁed drawing of the different vortex patterns that occur at the end
of a plate when the amplitude is increased. We can divide the vorticity plots at the stack
end into four categories:
A two vortices: Two vortices of opposite sign (a dipole) are formed behind each
stack plate, as is shown in ﬁgures 6.6(a) and 6.7(a).
B four vortices: Behind each stack plate four vortices are formed, as is shown in
ﬁgures 6.6(b) and 6.7(b).
C transition area: A category inbetween four vortices and a vortex street, as is
shown in ﬁgure (6.6(c)). The distinction between this category and categories B
and D is somewhat arbitrary. A vortex pattern is placed in category B when the
number of vortices is four and the vortices are located directly behind the stack
in a symmetrical way. In the transition area the number of vortices is larger than
four or the vortices are not symmetrically located in the extension of the plate.
D vortex street: A row of at least seven vortices is formed behind each plate (ﬁgures
6.6(d) and 6.7(c)). A vortex pattern is called a vortex street when at least four
alternating vortices can be distinguished.
The objective is to ﬁnd a correlation between the category of vortex formation (A
to D) and relevant dimensionless parameters (R
e
and S
t
). At higher Reynolds numbers
the instationary term of the NavierStokes equation becomes more important. Therefore
we expect that at higher Reynolds numbers the ﬂow is more inclined towards a vortex
street than to two or four vortices. The Strouhal number is a ratio of the plate thickness
6.4 Different categories of vortex patterns and dimensionless numbers 109
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
x/d
y
/
d
u
[
m
/
s
]
0
1
2
3
4
5
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
Figure 6.5: A typical ﬂow visualization measurement at the end of a stack. The top ﬁgure shows
the velocity vector ﬁeld and the bottom picture its corresponding vorticity plot. The three black
rectangles represent (a part of) the stack plates, which are 1 mm thick and at a distance of 4 mm
from each other. The frequency is 125 Hz and the velocity amplitude is 5.0 m/s.
110 Flow visualization in and around a stack
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
ω
z
[
1
0
4
s
−
1
]
−0.06
−0.04
−0.02
0
0.02
0.04
0.06
(a) v=0.50 m/s
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
ω
z
[
1
0
4
s
−
1
]
−0.1
−0.05
0
0.05
0.1
(b) v=0.83 m/s
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
ω
z
[
1
0
4
s
−
1
]
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
(c) v=2.5 m/s
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(d) v=5.0 m/s
Figure 6.6: The vorticity behind plates for four different velocity amplitudes. The phase angle
θ = π. The velocity amplitude inbetween the stacks [u
1
[ = 5 m/s and the frequency is 125 Hz.
The plates, represented by black rectangles, have a thickness of 1 mm and are 4 mm apart from
each other.
+

(a) cat A: two vortices
+
 +

(b) cat B: four vortices
+

+



+
+
(c) cat D: vortex street
Figure 6.7: Different categories of vortex formation at the end of a stack plate.
6.4 Different categories of vortex patterns and dimensionless numbers 111
and the displacement amplitude. The plate thickness is a measure for the vortex size in
the direction perpendicular to the plate and the displacement amplitude determines the
number of vortices in the direction parallel to the plate. The vortex size is based on the
optical appearance in the vorticity plots. At low Strouhal numbers we expect the vortex
to be more elongated and farther away from the plate and is therefore expected to be
more inclined towards a vortex street.
In ﬁgure 6.8 the Strouhal number is plotted versus the Reynolds number for a set of
25 measurements with different frequencies (5, 25, 45, 114, 125, 240, and 372 Hz) and
different velocity amplitudes (0.1 to 10.0 m/s). For each measurement the vorticity pat
tern is studied and is placed in one of the four categories. Depending on the category
the data points in ﬁgure 6.8 have different symbols. There is a clear distinct correlation
between the two relevant dimensionless parameters and the different categories. Each
category has its own area in the ReynoldsStrouhal space. The two black lines indi
cate the boundaries between these areas. The lines can be described by the following
function:
S
t
/R
e
= f ν/V
2
= α, (6.4)
in which α is a constant. For the border line between category B and C it follows that
α = (1.9 ±0.4) 10
−4
and for the border line between C and D it follows that α = (8 ±
1) 10
−4
. For the range of R
e
and S
t
it seems that only the ratio S
t
/R
e
is determinative
for the category of vortex patterns occurring at the end of a plate.
1/S
t
is a measure for the number of vortices that can be formed. For 1/S
t
∼ 2π, i.e.
V/2π f ∼ d, only two vortices are expected, since the displacement length is approx
imately of the same size as the vortices. For a vortex street to be formed, two criteria
need to be fulﬁlled:
A 1/S
t
> 2π, otherwise the displacement length is not long enough for more than
two vortices to be formed,
B the R
e
number needs to be high enough for vortices to be formed.
Another dimensionless parameter that is often used in oscillating ﬂows is the
KeuleganCarpenter number, KC, also called the period number. It is the ratio of a
displacement amplitude (V/ω) to a characteristic length scale.
When this characteristic length scale is the length of the stack, the Keulegan
Carpenter number is deﬁned as
KC
L
=
V
ωL
. (6.5)
In our setup KC
L
¸1. This implies that the stack is much longer than the displacement
amplitude. The stack can be considered as semiinﬁnite. Sufﬁciently far away (more
than 2V/ω) from the plate edges the ﬂow is fully developed in the stack. There is no
interaction between the ﬂow effects at the two ends of the stack (apart from streaming
effects). The situation is different when KC
L
≈ 1, in which case the stack plates are of the
same size as the displacement amplitude. This can occur with very small stacks or very
high amplitudes. Gas parcels can move from one stack end to the other. In the middle
of the stack the ﬂow is not fully developed and there is a strong interaction between the
ﬂow effects at the two stack ends.
112 Flow visualization in and around a stack
0 50 100 150 200 250 300 350 400 450 500 550 600 650
0.0
0.1
0.2
0.3
0.4
0.5
S
t
Re
two vortices
four vortices
transition
vortex street
Figure 6.8: The Strouhal number plotted versus the Reynolds number for different data points.
For each data point it is determined in which category it belongs by studying the vorticity plots.
The different categories are indicated by different colors and shapes, as is indicated in the legend.
The groups of data points can be separated by straight lines trough the origin.
6.5 The inﬂuence of the plateend shape 113
When the characteristic length scale of the KC number is the plate separation D, we
get the following deﬁnition
KC
D
=
V
ωD
. (6.6)
This number is related to the Strouhal number as
KC
D
=
1
2πS
t
. (6.7)
Another important parameter is the Reynolds number based on the viscous pene
tration depth
R
eδ
=
Vδ
ν
ν
. (6.8)
This number is important in the transition of laminar to turbulent for oscillatory ﬂows.
This number is related to α
R
eδ
=
_
2
α
. (6.9)
The two border lines in ﬁgure 6.8 correspond to R
eδ
= 50 and R
eδ
= 102.
6.5 The inﬂuence of the plateend shape
So far all measurements were performed on rectangular plates with a ﬂat end (ﬁgure
6.9(a)). In this subsection we discuss the inﬂuence of the shape of the plate ends on the
vortex shedding at the end of the plates. We measured the ﬂow around single plates
with a thickness of 3 mm. In ﬁgure 6.9(b) the ﬂow around a circular end is shown,
in ﬁgure 6.9(c) around a 90
o
triangle and in ﬁgure 6.9(d) around a 25
o
triangle. Due
to the absence of sharp corners for the circular end the vorticity plot is more smooth
and the two vortices are almost circles. The 90
o
triangle does have sharp corners and
is very similar to a rectangular end. At the end of a 25
o
triangle however, two thin
layers of vorticity are present behind the sharp corner. These two layers do not roll
up. The vortex on top of the triangle is caused by the transition of a straight plate
towards a triangle. A sharp triangular shape is clearly giving different results as the
other geometries. The choice of a shape can inﬂuence the dissipation due to the vortices
that are shed off and the heat transfer in the heat exchanger. At triangular ends with
even sharper edges (12
o
or smaller) vortices completely disappear and the shear layer
dissolves smoothly in the main stream.
6.6 Numerical simulations
6.6.1 CFD model
We have used commercial CFD packages to run a numerical simulation of a oscillating
ﬂow at the end of a stack. We have used the same geometry as in the experiment which
is discussed in section 6.3, with the same frequency f =125 Hz, and the same velocity
amplitude [u
1
[ = 5 m/s. We are only interested in ﬂow effects and therefore do not
account for the compressibility of the gas.
114 Flow visualization in and around a stack
x/d
y
/
d
−2 0 2
−2
−1
0
1
2
ω
z
[
1
0
3
s
−
1
]
−5
0
5
(a) circle
x/d
y
/
d
−2 0 2
−2
−1
0
1
2
ω
z
[
1
0
3
s
−
1
]
−5
0
5
(b) rectangle
x/d
y
/
d
−2 0 2
−2
−1
0
1
2
ω
z
[
1
0
3
s
−
1
]
−5
0
5
(c) triangle
x/d
y
/
d
−2 0 2
−2
−1
0
1
2
ω
z
[
1
0
3
s
−
1
]
−5
0
5
(d) knife
Figure 6.9: The vorticity behind plates with different end shapes. The plate thickness d = 3 mm,
the velocity amplitude of the main ﬂow is 5 m/s, f = 125 Hz. The angle of the triangle (c) is 90
o
and the angle of the knife (d) is 25
o
.
6.6 Numerical simulations 115
The geometry is two dimensional and is shown in ﬁgure 6.10. The modeled region
is indicated with a gray color and the black regions represent the wall (stack plates and
resonator wall). In total we have four stack plates in our model, with thickness d = 1
mm and with 4 mm inbetween them. At the interface with the walls a noslip condition
is applied. We have used two additional boundary conditions:
u(t) = [u
1
[ sin(ωt), for x/d = −30, (6.10)
∂u
∂x
= 0, for x/d = 30. (6.11)
We have created a multi block structured mesh, with 113,120 rectangular elements.
The mesh is reﬁned at the boundary layers, where the velocity gradient is highest. The
differential equations are solved using the so called SIMPLE (SemiImplicit Method
for PressureLinked Equations) algorithm. A secondorder implicit time discretization
scheme together with a secondorder upwind space discretization for convective terms
was chosen. No turbulence modeling was applied. We use 200 time steps for each pe
riod and run the simulation for 10 periods. The ﬂuid is atmospheric air at 300 K and
incompressible.
To verify the algorithm the simulation is performed two times, with different codes.
First the simulation was ran in CFX and subsequently in FLUENT, with a different mesh
and a different time step. Since the results were very similar, we conclude that the
algorithm works well. But this does not imply that the model is correct. Therefore we
will compare the simulation results with the PIV measurements.
x/d
0 30 30
y/d
0
5
d
9.5
14.5
u(t)
0 = x / u
Figure 6.10: A schematic drawing of the twodimensional geometry that used for the numerical
simulations. The dimensions are scaled by the plate thickness d = 1 mm. The black rectangles
represent walls, at which a noslip condition is applied.
In ﬁgure 6.11 a typical result of the numerical simulation at θ = 3π/2 + 2πn is
shown. This result can be compared to the PIV measurement in ﬁgure 6.5.
When comparing the numerical simulation with the measurement we see that qual
itatively both results are very similar. The size of the vortices (inbetween 1 and 2 mm
in diameter) and distance inbetween the vortices (approximately 2 mm) are in good
agreement. Differences can be caused by averaging over 50 samples, which is executed
for the measurement, whereas the numerical results are instantaneous measurements.
116 Flow visualization in and around a stack
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
x/d
y
/
d
u
[
m
/
s
]
0.5
1
1.5
2
2.5
3
3.5
4
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
Figure 6.11: Numerical simulation. A plot of the velocity ﬁeld (top) and of the vorticity (bot
tom).
6.6 Numerical simulations 117
6.6.2 Subsequent cycles
From the PIV measurements we observed that the measurements are not perfectly re
producible. When we repeat the measurement at the same phase angle the vorticity plot
differs. We expect this to be caused by statistical ﬂuctuations. This can be observed most
clearly at the phase angle θ = 3π/2. Based on symmetry one would expect the vorticity
pattern to be completely symmetrical with the centers of the plates the symmetry axes.
The measurements show a pattern in which the vortices are created alternately, which is
similar to a Von K´ arm´ an vortex street. This seems to be energetically favorable. The vor
tex street can start either with a positive or with a negative vortex. This is determined
by statistical ﬂuctuations or by the previous cycle. In this subsection we will study the
reproducibility of the numerical simulations. We will look into the vorticity pattern at
θ = 3π/2 at ﬁve subsequent cycles, which are shown in ﬁgure 6.12. In ﬁgure 6.12a
behind the top plate, a negative vortex is attached to the plate end, followed by a vortex
with a positive sign and then negative and positive vortices alternatingly. The second
vortex resembles much more a circular shape then the third and fourth vortices, which
are elongated ellipses. To the left of the ﬁfth vortex the vorticity is not uniform, but the
vortices can not be distinguished clearly. The vortex street behind the middle plate is
very similar to the vortex street behind the top plate. The bottom plate has a different
vortex pattern behind it. Here the vortex attached to the plate is positive in sign. This
vortex street is almost a mirror image of the other two.
In ﬁgure 6.12a the vortex pattern is similar to that of 6.12b, with the major difference
being that at the bottom plate the ﬁrst and third vortex (both positive in sign) start to
attach to each other. In ﬁgure 6.12c we see that the attachment of the ﬁrst and third
vortex continues and that they have uniﬁed into one vortex. The uniﬁed vortex and the
second vortex are next to each other. And also the third and fourth vortex are next to
each other instead of being behind each other.
In ﬁgure 6.12d an interesting vortex pattern is created behind the top plate. One
could even argue that this is not a vortex street, but instead two vortices next to each
other, both attached to the plate and both of them with a ﬁlament of vorticity behind
them.
In ﬁgure 6.12e all three plates show a very similar vortex pattern.
We see that different modes are possible for the vortex shedding, depending on his
tory. We see the same behavior in our measurements. Also small asymmetries in the
geometry can have a signiﬁcant inﬂuence on the vortex shedding mode.
6.6.3 Minor losses
In chapter 3 we discussed the minor losses occurring at the ends of a stack, due to
a sudden change in cross section. In this section we will determine the minor losses
by a numerical simulation. Near the edge of the stack, the ﬂow is heavily distorted
due to vortices. Sufﬁciently far away from the stack edge (more than two times the
displacement amplitude) the velocity proﬁle is laminar. The numerical code does not
include the energy equation, only the NavierStokes and the continuity equation. The
gas is considered by the code as an incompressible ﬂuid with a temperature that is
118 Flow visualization in and around a stack
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
Figure 6.12: The vorticity plots at θ = 3π/2 for ﬁve subsequent cycles, determined by a nu
merical simulation. The three black rectangles represent (a part of) the stack plates which have
thickness d=1 mm and are at a distance of 4 mm from each other. The frequency is 125 Hz and
the velocity amplitude in the stack is 5.0 m/s.
6.6 Numerical simulations 119
constant in time and position. In the laminar area the energy ﬂow can be determined by
˙
W =
__
1
τ
_
τ
0
ρ(h +[v[
2
/2)udtdA, (6.12)
with the enthalpy per unit mass given by h = + p/ρ. The internal energy per unit
mass is not relevant here, as the numerical simulation is at constant temperature. The
energy ﬂow per unit of area can be written as
˙
W(x)/A =
1
τ
_
τ
0
_
(p(t, x, y) +ρ[v(t, x, y)[
2
/2)u(t, x, y)
_
dt, (6.13)
with τ = 2π/ω the time of one period, and ¸¸ denoting the crosssectional average. In
the laminar area the viscous losses per unit of length are given by
d
˙
W
dx
= −
1
2
Re¦R
1
¦[U
1
[
2
, (6.14)
with R
1
given by Eq. 4.31:
Re¦R
1
¦ =
ωρ
0
A
Im¦−f
ν
¦
[1 − f
ν
[
2
. (6.15)
Note that thermal relaxation is not included in this simulation, because the gas is at
constant temperature in the simulation.
In ﬁgure 6.13
˙
W(x)/A is plotted as function of x. The left ends of the stack plates are
located at x = 30 mm. In this plot the slope, d
˙
W/dx, is always negative, as is expected as
a result of a decrease in the energy ﬂow due to viscous dissipation. The kinetic energy
has a signiﬁcant contribution to the energy ﬂow. At x/d = 60 the energy ﬂow is zero.
For 10 < x/d < 20 (out of the stack) the energy ﬂow
˙
W(x)/A is expected to decrease
linearly by 0.07 Wm
−2
mm
−1
according to Eq. 6.14. The numerical data are ﬁtted lin
early for 10 < x/d < 20 to a linear curve with a slope of 0.07 Wm
−2
mm
−1
. The ﬁt is
shown by the dashed line. For 25 < x/d < 35 the ﬂow is distorted due to vortices. For
x/d > 35 (in the stack) the ﬂow is laminar again and according to Eq. 6.14 the energy
ﬂow is expected to decrease by 0.50 Wm
−2
mm
−1
. The numerical data are ﬁtted linearly
for 40 < x/d < 55 to a linear curve with a slope of 0.50 Wm
−2
mm
−1
. The ﬁt is shown
by the dashdotted line. At the x/d = 30 the two ﬁt lines would intersect if there would
be no minor losses. The difference between the two ﬁt lines is the minor loss ∆
˙
W/A =
(2.3 ± 0.5) W/m
2
, indicated by the arrow. Using the relation between minor losses and
K, which was derived in section 4.4,
∆
˙
W =
2
3π
Kρ
0
A
2
[U
1
[
3
(6.16)
we ﬁnd that the corresponding minor loss coefﬁcient K = 0.07 ± 0.02. The analytical
value of minor loss coefﬁcient for a sudden expansion K
e
= 0.04. The minor losses
120 Flow visualization in and around a stack
during the contraction are determined by the vena contracta
K
c
=
_
1
C
c
−1
_
2
, (6.17)
with C
c
the ratio between the area of the jet at the vena contracta to the area between
the plates. According to Weisbach, for ψ = 0.8 this C
c
ratio is 0.81, which results in
K
c
= 0.05. But this value of C
c
is measured in a circular geometry instead of parallel
plates. In section 6.7 we will show that it is difﬁcult to determine the vena contracta
from our numerical simulations.
15 20 25 30 35 40 45 50 55
5
10
15
20
25
x/d
˙
W
/
A
[
W
/
m
2
]
∆
˙
W/A
Figure 6.13: The energy ﬂow
˙
W/A including kinetic energy as function of the position. Four
parallel plates are present in the geometry for x ≥ 30 mm. The plates are 1 mm thick and at
a distance of 4 mm from each other. The frequency is 125 Hz and the velocity amplitude is 5.0
m/s. The dashed line represents the energy ﬂow outside of the stack if no change in cross section
would occur and the dashdotted line is the energy ﬂow between the plates sufﬁciently far away
from the crosssection change. The difference between the two ﬁts, at x = 30 mm, is the minor
loss ∆
˙
W/A (indicated by the double arrow).
6.7 Vortex street evolution
The velocity as function of the phase angle at the x/d = −30 is shown in ﬁgure 6.14.
The velocity u(t) is normalized by its peak value u
L
= 5 m/s and the phase angle θ is
normalized by 2π. The squares indicated by A to F are the phase angles at which we
will show plots of the vorticity. The plots corresponding to A, B, and C are shown in
ﬁgure 6.15 and the plots corresponding to D, E, and F are shown in ﬁgure 6.16. In the
6.7 Vortex street evolution 121
left columns the PIV measurements are shown and in the right columns the numerical
simulations.
0 0.25 0.5 0.75 1
−1
−0.5
0
0.5
1
θ/2π
u
(
t
)
/
u
L
A
B
C
D
E
F
Figure 6.14: The velocity at x/d = −30 as function of the phase angle. The squares A to F
indicate at which phase angle a vorticity plot is shown in the next two ﬁgures.
As a starting point (A) we choose θ = 0, the moment at which the gas is at its
rightmost position. The velocity is zero at this moment. In the plot of the numerical
simulation the boundary layer near the plates is not developed. The disturbances in the
boundary layer are caused by vortices, that were formed during the previous period,
interfering with the boundary layer. In the measurement plot these disturbances in the
boundary layer are less clear. As the spatial resolution (the number of velocity vector
per area) is lower, the disturbances are smoothed out. The vorticity pattern behind the
top plate is a mirror of the one behind the middle plate. This is a consequence of a
symmetry axis in the geometry at y/d = 2.5. The disturbances in the boundary at the
start of the period will have an effect on the vortex shedding during the period. In the
measurements plots, at the top and bottom of the plate a small white region is present,
in which no measurements could be done due to laser reﬂections.
At θ = π/2 (B) the velocity is at its peak, so the vortices formed at this phase are
the strongest. When we focus on the center plate (y/d = 0) the positive vortex is elon
gated and is almost split by the negative vortex. Once this positive vortex is split into
two parts, a vortex is shed off. Not only in the numerical simulation, but also in the
measurements, we see some vortices that have already been shed off and located now
around x/d = 5. The displacement amplitude, given by
[ζ
1
[ = [u
1
[/ω, (6.18)
is in this case equal to ﬁgure 6.4d. This is the distance the gas has traveled from moment
A to moment B. For x/d <6.4 the vorticity is close to zero ([ω
z
[ < 0.2 10
4
s
−1
) in both
the measurement and the numerical simulation.
At moment (C), θ = 3π/4, the vortex streets have grown in length and in number
122 Flow visualization in and around a stack
measurement numerical simulation
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(A) θ = 0
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(B) θ = π/2
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(C) θ = 3π/4
Figure 6.15: The vorticity behind plates as function of time during the ﬁrst half of a period.
In the left column the PIV measurements are shown and in the right column the numerical
simulations. The velocity amplitude inbetween the stacks [u
1
[ = 5 m/s and the frequency is 125
Hz. The plates, represented by black rectangles, have a thickness of 1 mm and are 4 mm apart
from each other.
6.7 Vortex street evolution 123
measurement numerical simulation
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(D) θ = π
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(E) θ = 5π/4
x/d
y
/
d
−10 −5 0 5
−5
0
5
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
x/d
y
/
d
−10 −5 0 5
−6
−4
−2
0
2
4
6
8
ω
z
[
1
0
4
s
−
1
]
−1
−0.5
0
0.5
1
(F) θ = 3π/2
Figure 6.16: The vorticity behind plates as function of time during the second half of a period.
In the left column the PIV measurements are shown and in the right column the numerical
simulations. The velocity amplitude inbetween the stacks [u
1
[ = 5 m/s and the frequency is 125
Hz. The plates, represented by black rectangles, have a thickness of 1 mm and are 4 mm apart
from each other.
124 Flow visualization in and around a stack
of vortices. Also notice that the vortices in numerical simulation have tilted a little
towards a more vertical orientation. In the measurement plot the vortices are still more
horizontally oriented.
The phase angle θ = π is shown in ﬁgure 6.16(D). At this phase the gas is at its
leftmost position. The vortex streets are at its maximum length and can therefore be
identiﬁed most clearly at this moment of the period. The length of the vortex streets
divided by d is (12 ± 1) in the numerical simulation and (11 ± 1) in the measurement.
The displacement inbetween the plates 2[ζ
1
[/d equals 12.8. Since the jet out of the stack
spread out, the displacement value is smaller here.
At θ = 5π/8, in ﬁgure 6.16(E), the ﬂow is into the stack. The vortices move with the
ﬂow into the stack. The vortices that are away from the plates get a circular shape. The
strongest vortices are the ones that are going into the stack but are still connected with
the corners of the plates. We can recognize the vortices also in the measurement plot.
At θ = 3π/4, in ﬁgure 6.16(F), the vortices disappear into the boundary layers, cre
ating disturbances in these boundary layers. These disturbances will still be present a
the start of the next period and will inﬂuence the vortex pattern of the next period.
6.8 Velocity proﬁle inbetween two plates 125
6.8 Velocity proﬁle inbetween two plates
6.8.1 Introduction
The velocity parallel to the plates is given by [9]
u(y, t) = Re
_
i
ωρ
m
(1 −
cosh[(1 + i)y/δ
ν
]
cosh[(1 + i)y
0
/δ
ν
]
)
dp
dx
e
iωt
_
, (6.19)
with y = 0 in the center of the channel, ρ
m
the mean density, 2y
0
the plate distance and
Re¦x¦ representing the real part of x. Eq. 6.19 can also be written as
u(y, t) = Re
_
C(1 −
cosh[(1 + i)y/δ
ν
]
cosh[(1 + i)y
0
/δ
ν
]
)e
iωt
_
, (6.20)
with C a complex constant.
At higher R
e
or higher Wo the ﬂow becomes turbulent [9]. The transition occurs at
a critical Reynolds number based on the viscous penetration depth:
Re
δ
√
2
=
V
√
νω
≈ 400
[36].
6.8.2 A single measurement of the velocity proﬁles
A velocity proﬁle is the velocity parallel to the plates, u, as a function of the distance to
the bottom plate y. The velocity proﬁle between two plates, at a distance of 2 mm from
each other, is measured using PIV. First we will consider measurements at a frequency
of 55 Hz and at a drive ratio of 1.3%. The velocity is measured at 21 different phase
angles, with a time difference of 1 ms between them, covering a whole period (ﬁgure
6.17).
In the next section we will show a good method to compare the measurements with
the analytical solution.
6.8.3 Fitting procedure
Since the velocity is a function of two variables (position and time), it is difﬁcult to ﬁnd
a good way to plot it and compare it with the analytical solution (Eq. 6.19). When we
look at the velocity function for a ﬁxed position, y = y
A
, it is a harmonic function of t
u(y
A
, t) = C
1
sin(ωt +θ
1
), (6.21)
with C
1
and θ
1
, the amplitude and phase angle respectively, as ﬁtting parameters. Eq.
6.21 is only valid in an ideal case in which
A no streaming is present,
B no higher harmonics are present,
C the ﬂow is laminar and
D no entrance effects are present.
126 Flow visualization in and around a stack
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
−10
−8
−6
−4
−2
0
2
4
6
8
10
y [mm]
v
e
l
o
c
i
t
y
u
[
m
/
s
]
Figure 6.17: The velocity proﬁle at 21 different phase angles at a frequency of 55 Hz and a drive
ratio of 1.3%.
We can correct for the effects of streaming and the harmonic by ﬁtting according to
u(y
A
, t) = u
0
+[u
1
[ sin(ωt +θ
1
) +[u
2
[ sin(2ωt +θ
2
), (6.22)
with u
0
, [u
1
[, θ
1
, [u
2
[ and θ
2
as ﬁtting parameters. For every position y
A
we can ﬁt the
measured data according to Eq. 6.22. This way we ﬁnd the ﬁve ﬁtting parameters as a
function of y.
In ﬁgure 6.18 the time dependence of the velocity at position y = y
0
is shown. The
red curve is a ﬁt using Eq. 6.22.
Edge effects are avoided by measuring sufﬁciently far away from the plate ends
(more than 2 times the displacement amplitude away). Fluctuations in time due to tur
bulence are reduced averaging over 50 measurements. If the presence of turbulence
effects the average velocity proﬁle, we can observe this by comparing the ﬁtting param
eter C
1
with the amplitude of Eq. 6.20.
6.8.4 First harmonic velocity and the transition to turbulence
The measured ﬁtting parameters [u
1
[(y) and θ
1
(y) are the amplitude and phase angle
acoustic velocity proﬁle. We can compare the ﬁtting parameters with the amplitude
and phase angle of the analytical solution (Eq. 6.20). In ﬁgure 6.19 the amplitudes
and in ﬁgure 6.20 the phase angle of the acoustic velocities are shown at three different
drive ratios. We can add an arbitrary offset constant to the phase angle. The phase
angle θ
1
is therefore deﬁned as such that the θ
1
=0 for y = 0. The velocity amplitude
6.8 Velocity proﬁle inbetween two plates 127
0 5 10 15 20 25
−10
−8
−6
−4
−2
0
2
4
6
8
10
t [ms]
v
e
l
o
c
i
t
y
[
m
/
s
]
Figure 6.18: The velocity as a function of time at position y = y
0
. The solid curve is a ﬁt.
[u
1
[(y) is normalized by its value at y = 0. The viscous penetration depth is 0.29 mm.
The analytical solution is very consistent with the measurements. The linear theory
−1 −0.5 0 0.5 1
0
0.2
0.4
0.6
0.8
1
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(a) DR = 0.1%, [u
1
(0)[=0.66 m/s
−1 −0.5 0 0.5 1
0
0.2
0.4
0.6
0.8
1
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(b) DR = 0.4%, [u
1
(0)[=3.45 m/s
−1 −0.5 0 0.5 1
0
0.2
0.4
0.6
0.8
1
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(c) DR = 1.3%, [u
1
(0)[=12.4 m/s
Figure 6.19: The velocity amplitude [u
1
[ as function of the yposition inbetween the plates.
DR is the drive ratio at the end of the resonator tube.
produces good results for these amplitudes. We can conclude that the ﬂow is linear.
We are interested in what happens at higher amplitudes. Therefore we have performed
measurements at the resonance frequency (44 Hz). These measurements are shown in
ﬁgures 6.21 and 6.22. The viscous penetration depth is 0.33 mm.
At DR = 1.7% the ﬂow is still laminar, as no signiﬁcant deviations from the analytical
solution can be seen. At DR = 3.4% the measured phase angle starts to deviate signiﬁ
cantly from the analytical solution. At DR = 6.8% the measured phase angle is constant
over the cross section. In a laminar ﬂow pattern the ﬂow at the boundary layers is lag
ging in correspondence to the main ﬂow in the center of the plates. This is no longer
true at DR = 6.8%. The ﬂow has become turbulent and has turned into a plug ﬂow, with
128 Flow visualization in and around a stack
−1 −0.5 0 0.5 1
0
0.05
0.1
0.15
0.2
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(a) DR = 0.1%
−1 −0.5 0 0.5 1
−0.05
0
0.05
0.1
0.15
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(b) DR = 0.4%
−1 −0.5 0 0.5 1
−0.05
0
0.05
0.1
0.15
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(c) DR = 1.3%
Figure 6.20: The phase angle θ
1
of the velocity as function of the yposition inbetween the
plates. DR is the drive ratio at the end of the resonator tube.
−1 −0.5 0 0.5 1
0
0.5
1
1.5
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(a) DR = 1.7%, [u
1
(0)[=10.7 m/s
−1 −0.5 0 0.5 1
0
0.2
0.4
0.6
0.8
1
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(b) DR = 3.4%, [u
1
(0)[=21.3 m/s
−1 −0.5 0 0.5 1
0
0.5
1
1.5
y/y
0

u
1
(
y
)

/

u
1
(
0
)

experimental
analytical
(c) DR = 6.8%, [u
1
(0)[=42.7 m/s
Figure 6.21: The velocity amplitude [u
1
[ as function of the yposition inbetween the plates.
DR is the drive ratio at the end of the resonator tube.
−1 −0.5 0 0.5 1
0
0.02
0.04
0.06
0.08
0.1
0.12
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(a) DR = 1.7%
−1 −0.5 0 0.5 1
−0.05
0
0.05
0.1
0.15
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(b) DR = 3.4%
−1 −0.5 0 0.5 1
0
0.05
0.1
0.15
0.2
y/y0

θ
1
(
y
)

/
2
π
experimental
analytical
(c) DR = 6.8%
Figure 6.22: The phase angle θ
1
of the velocity as function of the yposition inbetween the
plates. DR is the drive ratio at the end of the resonator tube.
6.8 Velocity proﬁle inbetween two plates 129
a constant phase angle over the cross section. The amplitude dependency of the y posi
tion only shows small deviations from the analytical solution. At DR = 3.4% the phase
angle proﬁle is inbetween that of a laminar and a turbulent ﬂow. This is a transition
point. The Reynolds number based on the viscous boundary thickness is deﬁned as
R
eδ
=
[u
1
(0)[δ
ν
ν
. (6.23)
For a drive ratio of 3.4% the Reynolds number R
eδ
= 468.
6.8.5 Velocity ﬂuctuations
The velocity ﬂuctuations are shown in ﬁgure 6.23. To determine the ﬂuctuations in
the boundary layer the rootmeansquare values (rms) are averaged over the region
0.88 < y < 0.97, which are indicated with squares. The ﬂuctuations in the bulk ﬂow,
indicated by diamonds, are averaged over the region 0 < y < 0.88. The rms values are
determined by:
[v
/
[ =
¸
1
n −1
n
∑
k=1
[v
k
−v[
2
. (6.24)
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(a) DR = 0.2%, [u
1
(0)[=2.06 m/s
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(b) DR = 0.4%, [u
1
(0)[=3.45 m/s
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(c) DR = 1.3%, [u
1
(0)[=12.4 m/s
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(d) DR = 1.7%, [u
1
(0)[=10.7 m/s
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(e) DR = 3.4%, [u
1
(0)[=21.3 m/s
0 5 10 15 20
0
0.02
0.04
0.06
0.08
0.1
t [ms]

R
M
S
(
v
)

/

u
1

(f) DR = 6.8%, [u
1
(0)[=42.7 m/s
Figure 6.23: The velocity ﬂuctuations [v’[ as function of time. The squares represent the ﬂuctu
ations in the boundary layer and the diamonds in the bulk ﬂow. The ﬂuctuations are normalized
by the peak value of the velocity [u
1
[. DR is the drive ratio at the end of the resonator tube.
130 Flow visualization in and around a stack
6.8.6 Streaming velocity
From the time dependent ﬁt we can also determine the DCcomponent of the velocity,
also called the streaming velocity. We deﬁne the streaming velocity as the time average
velocity,
u
0
=
1
τ
_
τ
0
u(t)dt, (6.25)
not to be confused with the mass streaming, which is given by
˙ m =
1
τ
_
τ
0
ρ(t)u(t)dt, (6.26)
The ratio of the streaming velocity u
0
(y) to the acoustic velocity in the center [u
1
(0)[ is
shown in ﬁgure 6.24 as function of the position inbetween the plates.
0 0.5 1
−5
0
5
10
x 10
−3
y/y
0
u
0
(
y
)
/

u
1
(
0
)

(a) DR = 0.4%, [u
1
(0)[=3.45 m/s
0 0.5 1
−0.04
−0.02
0
0.02
0.04
y/y
0
u
0
(
y
)
/

u
1
(
0
)

(b) DR = 1.3%, [u
1
(0)[=12.4 m/s
Figure 6.24: The streaming velocity [u
1
[ as function of the yposition inbetween the plates.
DR is the drive ratio at the end of the resonator tube. The streaming velocity is normalized by
the acoustic velocity in middle of the plates, [u
1
(0)[.
The ﬂuctuations of temperature and density in the boundary layers cause a net
streaming velocity in the boundary layers. The mass streaming integrated over the tube
cross section should always be zero, since the system is closed. To compensate for the
net streaming in the boundary layers, a streaming velocity in the opposite direction is
generated in the bulk of the gas. The streaming in the bulk of the gas is expected to have
a parabolic shape.
6.9 Natural Convection
6.9.1 Scale analysis on a vertical wall
We want to do a scale analysis of the natural convection in order to compare the order
of magnitude of the convection velocity with the measurements and to approximate the
heat losses due to convection. One simpliﬁcation we make is that we replace the stack
6.9 Natural Convection 131
x
Ta
Tw
y
0
0
δκ
δν
ρ
u=0
g
0
v
x
δν
0
T
x
δκ
Figure 6.25: A schematic view of natural convection near a solid vertical wall at temperature
T
w
, which is higher than the ambient temperature T
a
. Due to buoyancy the gas moves up and
both a viscous and a thermal boundary layer are formed. The boundary layer thicknesses, δ
ν
and
δ
κ
respectively, grow with the yposition along the wall. Schematic plots of the two boundary
layer thicknesses as functions of x are also shown.
end, which consists of 1 mm thick plates, separated 1 mm from each other, with a solid
vertical wall. This wall is at temperature T
w
and has height H.
First we want to consider the situation of a ﬁnite vertical wall in an inﬁnitely large
space at temperature T
w
higher than the ambient temperature T
a
, as is shown in ﬁg
ure 6.25. Both a viscous and a thermal boundary layer develop at the wall and their
thicknesses, δ
ν
and δ
κ
, grow with the vertical position y.
In steady state (∂ρ/∂t = 0) the mass conservation law yields
∂u
∂x
+
∂v
∂y
= 0, (6.27)
and the momentum conservation law in y direction yields
ρ
_
u
∂v
∂x
+ v
∂v
∂y
_
= −
∂p
∂y
−ρg +µ
_
∂
2
v
∂x
2
+
∂
2
v
∂y
2
_
, (6.28)
with g the acceleration due to gravity. The energy conservation law yields
u
∂T
∂x
+ v
∂T
∂y
= κ
_
∂
2
T
∂x
2
+
∂
2
T
∂y
2
_
. (6.29)
According to Bejan [37], the density variation due to temperature changes can be ap
proximated by
ρ ≈ ρ
∞
−βρ
∞
(T −T
a
), (6.30)
where β is the volume expansivity (β = 1/T in ideal gases) and T
a
is the ambient
132 Flow visualization in and around a stack
temperature. Since the pressure distribution is hydrostatic
∂p
∂y
= −ρ
∞
g. (6.31)
Substituting Eqs. 6.31 and 6.30 into Eq. 6.28 and dividing by ρ
∞
results in
u
∂v
∂x
+ v
∂v
∂y
ρ
∞
[1 −β(T −T
a
)] = βg(T −T
a
)ρ
∞
+ν
_
∂
2
v
∂x
2
+
∂
2
v
∂y
2
_
ρ
∞
[1 −β(T −T
a
)].
(6.32)
Since T −T
a
is 40 K at maximum, β(T −T
a
) ∼ 10
−1
. We only want to make a rough
model of the natural convection and therefore neglect β(T −T
a
) in comparison to 1. The
∂
2
v/∂x
2
term is of the order v/δ
2
κ
and ∂
2
v/∂y
2
∼ v/y
2
. Since y ¸ δ
κ
the ∂
2
v/∂y
2
term
can be neglected in comparison to ∂
2
v/∂x
2
. Eq. 6.32 simpliﬁes to
u
∂v
∂x
+ v
∂v
∂y
= βg(T −T
a
) +ν
∂
2
v
∂x
2
. (6.33)
The two terms at the left hand side are the inertia terms, and the ﬁrst term at the right
hand side is the friction and the second term is the Buoyancy. The buoyancy is the
driving force of the natural convection and is caused by the decrease in density due to
the heating up of the gas by the hot wall.
To determine whether the inertia or the friction is balancing the buoyancy force, we
will now make an order approximation of the three different terms.
v
2
y
∼ βg(T −T
a
) +ν
v
δ
2
κ
. (6.34)
Applying a scale analysis to the mass and energy conservation:
u
δ
κ
∼
v
y
, (6.35a)
v
∆T
y
∼κ
∆T
δ
2
κ
. (6.35b)
By using Eq. 6.35b we ﬁnd for the friction term in Eq. 6.34 that νv/δ
2
κ
∼ P
r
v
2
/y.
Thus for P
r
¸ 1 the friction is dominating the inertia, whereas for P
r
¸ 1 the inertia is
dominating the friction.
When the buoyancy is balanced by friction (P
r
¸ 1) the momentum equation in y
direction gives
ν
v
δ
2
κ
∼ gβ∆T. (6.36)
We have three scale equations (Eqs. 6.36a,b and Eq. 6.34) and three unknowns (v, δ
κ
, u).
6.9 Natural Convection 133
This set of equations can be solved easily resulting in
v ∼
κ
y
R
1/2
a
, (6.37a)
δ
κ
∼ yR
−1/4
a
, (6.37b)
u ∼
κ
y
R
1/4
a
, (6.37c)
with the Rayleigh number
R
a
=
gβ∆Ty
3
κν
. (6.38)
When the buoyancy is balanced by inertia (P
r
¸ 1) by solving the scale equations
we ﬁnd very similar Eqs. as in Eqs. 6.37, but now the R
a
number is replaced by the
Boussinesq number, B
o
:
v ∼
κ
y
B
1/2
o
, (6.39a)
δ
κ
∼ yB
−1/4
o
, (6.39b)
u ∼
κ
y
B
1/4
o
, (6.39c)
in which the Boussinesq number is given by
B
o
= R
a
P
r
. (6.40)
For P
r
≈ 1, as is the case in our setup (P
r
= 0.72), both the inertia and the friction
need to be taken into account. Eqs. 6.39 and 6.37 converge and either of them can be
used.
The Nusselt number is deﬁned as
N
u
=
αy
k
, (6.41)
where α is the heat transfer coefﬁcient, deﬁned as
α =
q
w
T
w
−T
a
=
−k(∂T/∂x)
x=0
T
w
−T
a
. (6.42)
By using the approximation −(∂T/∂x)
x=0
∼ (T
w
−T
a
)/δ
κ
, we ﬁnd for the Nusselt num
ber
N
u
∼ R
1/4
a
, for P
r
¸1, (6.43a)
N
u
∼ B
1/4
o
, for P
r
¸1. (6.43b)
Although these equations are based only on scale analysis they are in good agreement
134 Flow visualization in and around a stack
with detailed numerical calculations, which resulted in:
N
u
= 0.503R
1/4
a
, for P
r
¸1, (6.44a)
N
u
= 0.600B
1/4
o
, for P
r
¸1, (6.44b)
and for the whole range of P
r
numbers:
N
u
= 0.503
_
P
r
Pr + 0.986P
1/2
r
+ 0.492
_
1/4
R
1/4
a
. (6.45)
Using Eq. 6.39 for our setup, with atmospheric air, ∆T = 40
◦
C, and H = 60 mm, we
ﬁnd that v ≈ 0.3 m/s, δ
κ
≈ 2 mm, and δ
ν
≈ 1.7 mm, at position y = H.
6.9.2 Measurements
When the hot end of the stack heats up, this induces natural convection. The ﬂow
around the hot end of the stack is measured using PIV. We use thermoacoustics to heat
up the hot end of the stack. The loudspeaker is used as the source of acoustic power.
Once the hot end reaches 60
◦
C, the loudspeaker is turned off. The reason for turning
off the loudspeaker is that the absence of acoustics makes it much easier to measure the
convective ﬂow. An additional advantage is that we exclude other forms of streaming,
e.g. acoustic streaming. The temperature of the stack is registered by thermometers,
as is described in chapter 5. Due to heat losses to the environment and due to conduc
tion, the hotend temperature decreases over time. Every time the temperature drops
approximately 5
◦
C we do a PIV measurement of the ﬂow. In ﬁgure 6.26 the ﬂow ﬁeld
behind the stack is shown, for two different hotend temperatures. The colors used in
the two plots are on the same scale, whereas the vector sizes in the plot are normalized
by the maximum velocity. The measurement plane is vertical and intersects with the
tube axis. The hot end of the stack is located 63 mm from the resonator end. The tube
diameter is 59.1 mm.
It can be seen in ﬁgure 6.26(a) that near the stack end (x/d = 0) a viscous boundary
develops, that grows with the y position. The velocity v
y
increases with the y position,
as expected, but for y/d ≥ 10 it decreases again. The reason for this is that the gas is
conﬁned in a closed volume, so the gas has to move to the right, due to mass conser
vation. A circulation within the stack due to streaming is also a possibility (to the left
at the upper side of the stack and to the right at the bottom side), but from the results
it seems that this effect is negligible. The viscous resistance within the stack is much
higher than outside of the stack.
The ﬂow structure in ﬁgure 6.26(b) has changed. The boundary layers are much
thicker than in ﬁgure 6.26(a). This cannot be explained by the lower temperature dif
ference T
w
− T
a
. A possible cause for this difference is that the ﬂow in ﬁgure 6.26(a)
is not fully developed yet. As the measurements are done during one cooling down
process, the ﬂow in ﬁgure 6.26(b) had more time to develop and is therefore close to
the stationary state. In a stationary state the buoyancy ﬂow can be modeled by using
a closed volume, as is shown in ﬁgure 6.28. Figure 6.26(a) resembles the vertical wall
model more than ﬁgure 6.26(b).
6.10 Streaming 135
0 10 20 30 40
−30
−20
−10
0
10
20
30
x/d
y
/
d
u
[
m
/
s
]
0
0.01
0.02
0.03
0.04
0.05
(a) T = 60
◦
C
0 10 20 30 40
−30
−20
−10
0
10
20
30
x/d
y
/
d
u
[
m
/
s
]
0
0.01
0.02
0.03
0.04
0.05
(b) T = 30
◦
C
Figure 6.26: A velocity vector ﬁeld of the convection ﬂow behind the hot end of the stack at two
different stackend temperatures T. The black rectangles represent the stack plates, which are 1
mm in thickness and 1 mm apart from each other. The stack end is located at 63 mm from the
end of the resonator.
In ﬁgure 6.27 a velocity proﬁle of v
y
as function of x at height y/d = 10. The solid
plot is at temperature T
w
= 60
◦
C and the dashed plot is at T
w
= 30
◦
C. The maximum
velocities are v
m
= 0.046 m/s and v
m
= 0.014 m/s, respectively. The boundary layer
thickness is here deﬁned as the position at which the velocity is at its maximum value.
We ﬁnd that δ
ν
/d = 3 and δ
ν
/d = 5, respectively. For T
w
= 60
◦
C the found δ
ν
/d of 3
is higher than the expected 1.7. As the scale analysis can only determine the order of
magnitude, this result is still quite good.
We can use the Nusseltnumber approximation to approximate the heat losses due
to buoyancy. For T
w
= 60
◦
Cand H = 0.06 m we ﬁnd that Nu ≈ 11 and that q ≈17 W/m
2
.
As a comparison the heat losses to the environment due to radial conduction in the stack
are 10 times a high. Thus, in this setup the effects of buoyancy on the energy balance of
the stack are relatively small. The heat ﬂow due to buoyancy can increase signiﬁcantly
when the ﬂow becomes turbulent, which takes place at R
a
/P
r
≈ 10
9
, according to Bejan
and Lage [38]. Since the R
a
number grows with H
3
, this can be an important effect for
large devices.
6.10 Streaming
Streaming is a net velocity ﬁeld that is obtained from averaging the velocity over a
whole period. When the oscillation is a sine wave the streaming velocity is zero. In
the presence of objects of wall the sine wave is disturbed locally, resulting in a local
streaming. The local streaming behind a sudden expansion is called jet streaming. It
is cause be a difference in ﬂow pattern during the expansion phase, when the ﬂow
behaves like a jet and vortices are formed, and the suction phase, when the ﬂow goes
into the stack. The measured streaming velocity around a single plate with d = 1 mm,
136 Flow visualization in and around a stack
0 5 10 15
−0.01
0
0.01
0.02
0.03
0.04
0.05
x/d
v
y
δ
ν
v
m
Figure 6.27: The vertical velocity v
y
, due to buoyancy, as function of the distance from the stack
x measured at height y/d = 10. The solid plot is at temperature T
w
= 60
◦
C and the dashed plot
is at T
w
= 30
◦
C. The maximum velocities are v
m
= 0.046 m/s and v
m
= 0.014 m/s, respectively.
x
T
a
T
w
y
0
0
g
W
H
insulated
Figure 6.28: A schematic view of natural convection in a rectangular volume with height H and
width W. The left wall is at temperature T
w
, the right wall is at temperature T
a
(< T
w
), and the
top and bottom walls are adiabatic. Due to buoyancy the gas moves up at the hot wall on the left
and down at the cold wall on the right.
6.11 Discussion and conclusion 137
[u
1
[ outside the stack is 0.5 m/s, and f =100 Hz is shown in ﬁgure 6.29. Behind the
plate one vortex pair can be distinguished. Two other vortices are more difﬁcult to
distinguish, one is located above and the other beneath the plate, with the two vortex
centers near x/d ≈ 1.
−3 −2 −1 0 1 2
−2.5
−2
−1.5
−1
−0.5
0
0.5
1
1.5
2
2.5
x/d
y
/
d
u
[
m
/
s
]
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Figure 6.29: The measured velocity vector ﬁeld of the streaming at the the end of a stack plate.
The black rectangle represent the stack plate, which is 1 mm in thickness. [u
1
[ outside the stack
is 0.5 m/s, f =100 Hz.
In ﬁgure 6.30 the measured streaming velocity of a parallelplate stack is shown, in
which the plate ends are not perfectly parallel. Since the plate ends are not tight up, the
plates can bend a little at their ends. This result in asymmetries of the stack at its ends,
resulting in asymmetrical ﬂow patterns. The ﬂow channels that have a smaller outlet
show a stronger jet ﬂow. As a consequence the channels with a small outlet have a
higher dissipation and are less preferred when the ﬂow is directed out of the stack. This
difference in preference of the different channels leads to an internal convection, which
is found to be approximately 1% of the acoustic ﬂow, for [u
1
[ is 5 m/s. This shows the
importance of a careful stack design. In all other measurements these asymmetries in
the geometry are avoided by keeping the plate ends at the same distance with the use
of a comb.
6.11 Discussion and conclusion
PIV has proven to be a useful measurement technique in the study of oscillating ﬂow
around stacks of parallel plates.
It is shown that different vortex patterns (one vortex pair, two vortex pairs, vortex
street) can originate behind a parallel plate. The Reynolds and Strouhal numbers, and
their ratio in particular, are the relevant dimensionless numbers for determining which
vortex pattern occurs. The inﬂuence of porosity and plateend shape are also studied.
The vorticity pattern behind sharp edged plate ends (25
◦
or smaller) differs fromsquared
or circular plate ends: instead of a vortex pair or a vortex street, two thin layers of
vorticity are formed.
Next we have shown that the numerical simulations, which are in good agreement
138 Flow visualization in and around a stack
−2 −1 0 1 2 3 4
−12
−11
−10
−9
−8
−7
−6
−5
x/d
y
/
d
u
[
m
/
s
]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 6.30: The measure streaming velocity at the end of a stack for which the plate ends are
not perfectly parallel. The black rectangles represent the ends of the stack plates, which are 1 mm
in thickness. [u
1
[ outside the stack is 1 m/s, f = 50 Hz.
with each other, are in good agreement with the PIV measurement. Not only does this
verify the numerical simulations but also the PIV measurement principle. We have
shown that the vortex pattern behind the plates show huge differences when compar
ing the vorticity plot at the same phase angle for different periods. These ﬂuctuations
are also be found in the numerical calculations, even without including any turbulence
modeling. These ﬂuctuations are not a startup effect, as after calculating for ten periods,
the ﬂuctuations persist. Apparently the vorticity pattern is sensitive to small changes
in the velocity ﬁeld at the start of the cycle, which is typical for nonlinear effects. This
shows that the vorticity pattern is strongly determined by its prehistory.
Furthermore we have measured the velocity inbetween two plates in the middle of
the stack, where no entrance effects takes place, as functions of the position inbetween
the plates y and the phase angle. By ﬁtting the velocity as function of the phase angle
at a single position to a sine wave, we can determine the amplitude and phase shift. We
have studied this amplitude and phase angle as function of the position y and compared
them with an analytical solution. For sufﬁciently low Reynolds numbers R
eδ
200 the
measurements were very consistent with the analytical solution. For R
eδ
≈ 500 the
phase shift differs from the analytical solution. And at R
eδ
≈ 1000 the measured phase
shift is constant over the cross section. This is caused by turbulence. To study the
turbulence in more detail, we have studied the instantaneous velocity proﬁles instead
of averages.
The types of streaming that can be present in a standingwave device are jet stream
ing, internal streaming, natural convection, and Rayleigh streaming. Of these four
streaming types, we have shown the ﬁrst three types in PIV measurements.
Bibliography
[1] S. L. Garrett, “Resource letter: Ta1: Thermoacoustic engines and refrigerators,”
American Journal Of Physics, vol. 72, pp. 11–17, Jan. 2004.
[2] N. Rott, “Damped and thermally driven acoustic oscillations in wide and narrow
tubes,” Journal of Applied Mathematics and Physics, vol. 20, pp. 230–243, 1969.
[3] N. Rott, “Thermally driven acoustic oscillations. part II: Stability limit for helium,”
Journal of Applied Mathematics and Physics, vol. 24, pp. 54–72, 1973.
[4] N. Rott, “Thermally driven acoustic oscillations. part III: Secondorder heat ﬂux,”
Journal of Applied Mathematics and Physics, vol. 26, pp. 43–49, 1975.
[5] N. Rott and G. Zouzoulas, “Thermally driven acoustic oscillations. part IV: Tubes
with variable crosssection,” Journal of Applied Mathematics and Physics, vol. 27,
pp. 197–224, 1976.
[6] N. Rott, “Thermoacoustics,” Advances in Applied Mechanics, vol. 20, pp. 135–175,
1980.
[7] N. Rott, “The inﬂuence of heat conduction on acoustic streaming,” Journal of Ap
plied Mathematics and Physics, vol. 25, pp. 417–421, 1974.
[8] G. Swift, “Thermoacoustic engines,” Journal of Acoustical Society of America, vol. 84,
pp. 1146–1180, 1988.
[9] G. Swift, Thermoacoustics: A Unifying Perspective for Some Engines and Refrigerators.
Acoustical Society of America, 2002.
[10] G. Swift, “Thermoacoustic engines and refrigerators,” Physics Today, pp. 22–28,
1995.
[11] P. Kundu and I. Cohen, Fluid Mechanics. Academic Press, San Diego, CA., 3th
edition ed., 2004.
[12] D. Nehari, V. Armenio, and F. Ballio, “Threedimensional analysis of the unidirec
tional oscillatory ﬂow around a circular cylinder at low keulegancarpenter and
beta numbers,” Journal of Fluid Mechanics, vol. 520, pp. 157–186, Dec. 2004.
[13] P. Anagnostopoulos and R. Minear, “Blockage effect of oscillatory ﬂow past a ﬁxed
cylinder,” Applied Ocean Research, vol. 26, pp. 147–153, May 2004.
140 BIBLIOGRAPHY
[14] M. F. Wybrow, B. Yan, and N. Riley, “Oscillatory ﬂow over a circular cylinder close
to a plane boundary,” Fluid Dynamics Research, vol. 18, pp. 269–288, Oct. 1996.
[15] M. Ozgoren, “Flow structure in the downstream of square and circular cylinders,”
Flow Measurement And Instrumentation, vol. 17, pp. 225–235, Aug. 2006.
[16] M. Ibrahim and W. Hashim, “Oscillating ﬂow in channels with a sudden change in
crosssection,” Computers & Fluids, vol. 23, pp. 211–224, Jan. 1994.
[17] P. Morris, S. Boluriaan, and C. Shieh, “Numerical simulation of minor losses due
to a sudden contraction and expansion in high amplitude acoustic resonators,”
Journal of the Acoustical Society of America, vol. 90, pp. 393–409, 2004.
[18] B. L. Smith, “Pressure recovery in a radiused sudden expansion,” Experiments in
Fluids, vol. 36(6), pp. 901–7, 2004.
[19] B. L. Smith and G. W. Swift, “Power dissipation and timeaveraged pressure in
oscillating ﬂow through a sudden area change,” Journal of the Acoustical Society of
America, vol. 113(5), pp. 2455–63, 2003.
[20] D. Skulina, A study of nonlinear ﬂows at the open end of a tube using particle imaging
velocimetry. PhD thesis, University of Edinburgh, 2005.
[21] D. Skulina, “A study of nonlinear acoustic ﬂows at the open end of a tube using
particle image velocimetry,” Journal of the Acoustical Society of America, vol. 120(5),
p. 3363, 2006.
[22] B. L. Smith and C. V. King, “Timeresolved piv and pressure measurements of os
cillating and pulsating ﬂow in a rapid expansion,” in 2007 5th Joint AMSE/JSME
Fluids Engineering Conference, 2007.
[23] B. L. Smith, K. V. Mortensen, and S. Wendel, “Oscillating ﬂow in adverse pressure
gradients,” in 2005 AMSE Fluids Engineering Summer Conference, 2005.
[24] A. Petculescu and L. A. Wilen, “Oscillatory ﬂowin jet pumps: nonlinear effects and
minor losses,” Journal of the Acoustical Society of America, vol. 113(3), pp. 1282–92,
2003.
[25] J. Disselhorst and L. Van Wijngaarden, “Flow in the exit of open pipes during
acoustic resonance,” Journal of Fluid Mechanics, vol. 99, no. pt 2, pp. 293–319, 1980.
[26] J. Disselhorst and L. v. Wijngaarden, “Oscillatory ﬂow around the edge of a ﬂat
plate,” Journal of Engineering Mathematics, vol. 13, no. 3, pp. 271–287, 1979.
[27] P. Stoltenkamp, S. Araujo, H. Riezebos, J. Mulder, and A. Hirschberg, “Spurious
counts in gas volume ﬂow measurements by means of turbine meters,” Journal of
Fluids and Structures, vol. 18, no. 6, pp. 771–781, 2003.
[28] P. Stoltenkamp, Dynamics of turbine ﬂow meters. PhD thesis, Eindhoven University
of Technology, 2007.
BIBLIOGRAPHY 141
[29] Ph. BlancBenon, E. Besnoin, and O. Knio, “Experimental and computational vi
sualization of the ﬂow ﬁeld in a thermoacoustic stack,” Comptes Rendus M´ ecanique,
vol. 331, pp. 17–24, Jan. 2003.
[30] A. Berson, Ph. BlancBenon, and M. Michard, “Ecoulements secondaires aux
extr´ emit´ es du stack d’un r´ efrig´ erateur thermoacoustique : mesure des champs de
vitesse oscillante ` a l’aide de la piv,” in 8` eme Congr` es Fran¸ cais d’Acoustique, pp. 749–
752, 2006.
[31] A. Berson, M. Michard, and Ph. BlancBenon, “Measurement of acoustic velocity in
the stack of a thermoacoustic refrigerator using particle image velocimetry,” Heat
and Mass Transfer, vol. 44, pp. 1015–1023, 2008.
[32] A. Berson and Ph. BlancBenon, “Nonperiodicity of the ﬂow within the gap of a
thermoacoustic couple at high amplitudes,” Journal of the Acoustical Society of Amer
ica, vol. 122(4), pp. EL122–127, 2007.
[33] X. Mao, D. Marx, and A. J. Jaworski, “Piv measurement of coherent structures and
turbulence created by an oscillating ﬂow at the end of a thermoacoustic stack,” in
Progress in Turbulence II Proceedings iTi Conference in Turbulence (M. Oberlack, ed.),
vol. 109, pp. 99–102, 2007.
[34] E. Besnoin and O. M. Knio, “Numerical study of thermoacoustic heat exchangers,”
Acta Acustica united with Acustica, vol. 90, pp. 432–444, May 2004.
[35] M. Raffel, C. Willert, and J. Kompenhans, Particle Image Velocimetry. A Practical
Guide. Springer, 1998.
[36] P. Merkli and H. Thomann, “Transition to turbulence in oscillating pipeﬂow,” Jour
nal of Fluid Mechanics, vol. 68, pp. 567–575, 1975.
[37] A. Bejan, Heat Transfer. John Wiley & Sons, 1993.
[38] A. Bejan and J. Lage, “Prandtl number effect on the transition in natural convection
along a vertical surface,” Journal of Heat Transfer, vol. 112, no. 3, pp. 787–790, 1990.
142 BIBLIOGRAPHY
Chapter 7
Flow measurements in coaxial
regeneratorbased devices
7.1 Introduction
An important type of regeneratorbased engines is depicted in Fig. 7.1. It consists of a
coaxial loop, that is designed by Aster Thermoacoustics and ECN. The advantage of the
coaxial design in comparison to the loop design (ﬁgure 2.11) is that it is compact and
easy to manufacture. This coaxial loop is created by inserting an open cylinder into the
resonator tube. Inside of the cylinder a regenerator and thermal buffer zone (TBZ) are
located. Inbetween the cylinder and the resonator wall the inertance is located and to
the right of the cylinder the compliance.
It is valuable to know how an oscillatory ﬂow behaves in such a coaxial loop. Espe
cially the ﬂow at the sharp corners, at both sides of the inertance, are of interest. In front
of the TBZ one or more ﬂow straighteners can be placed. It is known that the presence
of ﬂow straighteners improves the performance. We expect that the ﬂow straighteners
reduce the turbulence in the TBZ.
7.2 Experimental setup
The experimental setup is shown in ﬁgure 7.1. The concentric coaxial regenerator
based loop is enclosed in a transparent cylindrical tube (c), which is closed on the right
side and is connected to the subwoofer setup (which is described in section 3.3) on
the left side. A transparent cylindrical open tube (d) separates the intertance (b), also
called bypass, from the thermal buffer zone (TBZ) (e). At the right end of the TBZ a
stackedscreen regenerator (f), is located. The volume on the right of the regenerator
is the compliance volume (g). In front of the TBZ one or more wiregauze screens are
installed which act as straighteners. The setup is designed to work as an regenerator
based thermoacoustic couple at a frequency of 50 Hz. The system is ﬁlled with air at
atmospheric pressure and drive ratios up to 15 percent can be reached. The dimensions
of the various components are shown in table 7.1.
From the top of the setup a laser sheet is projected into the setup, in the plane of
144 Flow measurements in coaxial regeneratorbased devices
0
D
i
D
o
Dr Dt
x1 x2 x3x4
x5
b
e f g
d
a
c
Figure 7.1: A schematic drawing of the experimental setup of a regeneratorbased loop that can
be used for ﬂow visualization. (a) one or more ﬂow straighteners, (b) intertance, (c) outer tube
wall, (d) cylindrical loop wall, (e) thermal buffer zone (TBZ), (f) regenerator, (g) compliance.
Table 7.1: The dimensions of the geometry in mm.
x
1
8
x
2
104
x
3
112
x
4
128
x
5
220
D
r
40
D
t
50
D
i
60
D
o
70
the picture, as described in section 4.2. The camera, which is directed perpendicular
to the plane of the picture, is focussed on the light sheet. Three different interrogation
windows are used (ﬁgure 7.2). To reduce laser reﬂections, large parts of the outside and
inside cylindrical tubes are covered with ﬂuorescent paint.
(2) (1)
(3)
Figure 7.2: A schematic drawing of the three interrogation windows (1)(3) in the experimental
setup.
As the tube is cylindrical it is difﬁcult to visualize the ﬂowusing PIV. The tube acts as
7.3 Results 145
a lens in vertical direction and therefore changes the focal point, whereas in the horizon
tal direction the focal point remains undisturbed. As a consequence, it is not possible to
have both the horizontal and the vertical direction in focus. A small point in the tube
is projected as a vertical line or a horizontal line on the CCD of the camera (depending
on which direction is in focus). When performing measurements in the TBZ this effect
is even stronger, as both the tube wall and the loop wall act as lenses. Measurements in
the intertance are also difﬁcult because they are located very close to the wall.
It is very difﬁcult to measure the Gedeon streaming in this geometry using PIV. This
is due to the fact that the Gedeon streaming velocity is small in relation to the veloc
ity amplitude and that it is not possible to do measurements close to the wall due to
reﬂections.
7.3 Results
First we will show instantaneous PIV results through window 1. One cycle is divided
into 20 equidistant phase angles, at which the measurements are done. In section 7.3.1
the results at four different phase angles are shown. The streaming velocity is deter
mined by averaging the velocity over a complete cycle. We approximate the streaming
velocity by averaging over the 20 phase angles. In sections 7.3.2 and 7.3.3 the ﬂow
around the sharp edges is studied. These are indicated as interrogation windows 2 and
3. In these sections we will restrict ourselves to show only the streaming results.
7.3.1 Instantaneous measurements
In ﬁgures 7.3 the instantaneous velocity ﬁelds through window 1 are shown. One cycle
is divided into 20 equidistant phase angles, at which the velocity ﬁelds are measured,
but here only four of them are shown.
At ωt = 0 the ﬂowin the resonator is directed towards the loop. Close to the loop the
velocity at the top and bottom is higher than in the middle. The gas has a preference to
ﬂow into the inertance, due to a lower ﬂow resistance than in the TBZ. The regenerator
in the TBZ has a high ﬂow resistance.
At ωt = π/2 the ﬂow velocity in the resonator is approximately zero. Because the
acoustic velocities are small, disturbances have a larger effect. At the bottom of the
ﬁgure the ﬂow is directed to the right. This is a consequence of a background ﬂow, as
will be shown in the streaming results. At this moment there is a small ﬂow out of the
TBZ and into the inertance. The velocity in the TBZ is ahead in phase to the ﬂow in the
resonator and the ﬂow in the inertance is behind in phase.
At ωt = π the ﬂow is directed out of the loop. Analogous to ωt = 0 the ﬂow out of
the inertance is higher than out of the TBZ.
At ωt = 3π/2 the ﬂow velocity in the resonator is approximately zero again. At the
top right and the bottom right of the ﬁgure, two vortices originate. The phase difference
between the velocity in the inertance and TBZ can not explain these vortices. The phase
difference has only a small effect, as was shown at ωt = π/2. The two vortices are
caused by a jet ﬂow out of the inertance.
By averaging all 20 phase angles, we determine the timeaverage velocity, called
streaming velocity. In ﬁgures 7.4 the streaming velocity through window 1 is shown at
146 Flow measurements in coaxial regeneratorbased devices
−60 −40 −20 0
−20
−10
0
10
20
0.2
0.4
0.6
0.8
1
1.2
1.4
(a) ωt = 0
−60 −40 −20 0
−20
−10
0
10
20
0.1
0.2
0.3
0.4
0.5
(b) ωt = π/2
−60 −40 −20 0
−20
−10
0
10
20
0.2
0.4
0.6
0.8
1
1.2
1.4
(c) ωt = π
−60 −40 −20 0
−20
−10
0
10
20
0.2
0.4
0.6
0.8
1
1.2
(d) ωt = 3π/2
Figure 7.3: The instantaneous velocity ﬁeld in window 1 at different moments during the cycle,
at D
r
= 5.23%. The positions at the axes are in mm. The color and the vectors lengths are
measures of the velocity. The velocities to the right of the color bars are in m/s.
7.3 Results 147
two different D
r
.
−60 −40 −20 0
−20
−10
0
10
20
0.1
0.2
0.3
0.4
0.5
0.6
(a) D
r
= 5.23%
−60 −40 −20 0
−20
−10
0
10
20
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
(b) D
r
= 10.5%
Figure 7.4: Vector plots of the streaming velocity ﬁeld in window 1. The positions at the axes
are in mm. The color and the vectors lengths are measures of the velocity. The velocities to the
right of the color bars are in m/s.
At D
r
= 5.23% the streaming velocity is a combination of jet streaming (the two vor
tices) and a background ﬂow. The background ﬂow is probably a remaining ﬂow from
the oil particles that were injected before the measurement. This could be tested by
repeating the experiment with a longer pause between the particle injection and the
measurement. At D
r
= 10.5% the jet streaming dominates the background ﬂow.
7.3.2 Streaming in window 2
Window 2 is located at the Tjunction of the device. Here the inertance, TBZ and res
onator tube are all connected. At a D
r
of 2.60% (ﬁgure 7.5a) a vortex originates in the
streaming velocity ﬁeld at the sharp angle at the start of the loop (indicated with the
dashed ellipse). This vortex is caused by jet streaming. The velocity in the bypass is
much higher than in the tube, due to the high resistance of the regenerator. When the
ﬂow is leftwards, a jet ﬂow is going from the bypass into the tube, which has a much
larger cross section, causing a vortex.
When the D
r
is increased (ﬁgures 7.5 b and c), the vortex increases in strength and
dimensions. Also the center of the vortex shifts. In the D
r
range from 2.60% to 5.23% it
mainly shifts leftwards. In the range from 5.23% to 10.5% it mainly shifts upwards.
From the streaming plots in ﬁgure 7.5 it seems that the measured velocity vectors do
not obey the law of mass conservation. This can be explained by 3D effects. As we only
measure the velocity in one plane, we do not know the velocity in axial direction.
Another effect that can be relevant is that the measurements in this particular
travelingwave setup have a higher uncertainty, due to the cylindrical wall of the res
onator, as was discussed in section 7.2. When summing over 20 measurements that
compensate each other, resulting in a small quantity, the uncertainty in the individual
measurements become more clearly expressed.
148 Flow measurements in coaxial regeneratorbased devices
Now that we have shown the presence of the jet effect, we want to discuss its conse
quences:
A Due to the jet effect the resistance at the inertance end is asymmetric, resulting in
Gedeon streaming.
B Jet streaming, which can enhance the heat losses to the environment.
C The extra resistance at the edges of the inertance increases the dissipation.
Gedeon streaming is a net timeaverage mass ﬂow through the regenerator [1] that
can be present in thermoacoustic devices containing a loop, as is the case in traveling
wave devices, including the setup we are studying. The timeaverage mass ﬂow
˙
M
causes an unwanted enthalpy ﬂow,
˙
H =
˙
Mc
p
∆T trough the regenerator from hot to
cold. This effect can have a severe negative effect on the efﬁciency of a device. This
phenomenon is wellknown in the literature about doubleinlet pulsetube refrigerators
[2]. Gedeon streaming is caused by second order effects in the regenerator and can be
enhanced or reduced by an asymmetric resistance at another location within the loop.
Swift et al. [3] used a jet pump to reduce streaming. This is similar to the jetpumping
effect that is present at the inertance entrance in this setup. The resistance of the ﬂow
into the inertance differs from resistance of the ﬂow out of the inertance. According to
Swift [4] the timeaverage pressure drop is
∆p = −
ρ
0
[U
1
[
2
(K
out
−K
in
)
8A
2
In
, (7.1)
where U
1
is the volumeﬂow rate in the inertance, A
In
= π/4(D
2
i
− D
2
r
) is the cross
sectional area, and K
in
and K
out
are the minorloss coefﬁcients of the ﬂow into and out
of the inertance respectively. The minor loss coefﬁcient is deﬁned by Eq. 4.42 for steady
ﬂows. This can reduce the Gedeon streaming. To completely eliminate the Gedeon
streaming, in many regeneratorbased devices an elastic membrane is installed in the
loop, that is transparent for acoustic waves, but not for timeaverage mass ﬂows.
The vortices we have seen in ﬁgures 7.4 are a consequence of jet streaming. Because
this streaming is local, it does not contribute to Gedeon streaming. This vortex can
enhance the heat ﬂow from the TBZ wall to the resonator wall, due to forced convection.
Because the location of the streaming vortex is relatively far away from the regenerator,
this effect is relatively small.
The dissipation due to a sudden change in crosssection has been studied in section
4.4. The energy loss is given by Eq. 4.52
∆
˙
W =
2
3π
Kρ
0
A
2
In
[U
1
[
3
. (7.2)
Because the inertance size D
i
− R
r
is much smaller than the viscous penetration depth,
the minor loss coefﬁcient can be approximated by Eq. 4.45
K = (1 −ψ)
2
, (7.3)
7.4 Discussion and conclusion 149
with ψ the ratio of cross sections:
ψ =
A
In
A
Re
= 1 −
D
2
r
D
2
i
, (7.4)
where A
Re
= πD
2
i
/4 is the crosssectional area of the resonator. For this approximation
of ψ the ﬂow to the TBZ is neglected, thus this approximation of ψ is a little too low, but
it is good enough to make a rough approximation of the minor losses.
At D
r
= 5.23% and f = 50 Hz in the inertance [u
1
[ = 4 m/s. From Eq. 7.2 it
follows that ∆
˙
W = 8 mW and for D
r
= 10.5 this is 66 mW. As a comparison, the viscous
dissipation at the inertance wall is 10 mW for D
r
= 5.23% and 40 mW for D
r
= 10.5%.
7.3.3 Streaming in window 3
In ﬁgure 7.6 the streaming in window 3 is shown. In ﬁgure 7.6a the vortex, due to jet
streaming, is difﬁcult to distinguish. In ﬁgures 7.6b and 7.6c the jet streaming vortex is
clearly visible. The vortices at this location are much stronger than in front of the loop,
whereas the velocity amplitude is smaller. The reason for this is that at the end of the
bypass the change in cross section is very abrupt. In front of the loop the change in cross
section is much more gradually. Similar vortices were calculated by a CFD simulation
by Lycklama ` a Nijeholt et. al. [5].
7.4 Discussion and conclusion
We have measured the ﬂow at both ends of the inertance. Due to the jet effect vortices
originate at these locations. The major consequences of the jet effect on the performance
of a device are the dissipation due to minor losses, a locally enhanced heat transfer
due to forced convection caused by jet streaming, and Gedeon streaming due to an
asymmetry in ﬂow resistance for the in and outgoing ﬂow at the inertance entrance.
Unfortunately, in this setup, it was not possible to measure the Gedeon stream
ing. First of all the Gedeonstreaming velocity is small in comparison to the oscillatory
ﬂow. Secondly, the thick cylindrical walls make it very difﬁcult to performPIVmeasure
ments. Since the wall acts as a lens in vertical direction and not in horizontal direction,
the vertical and horizontal directions have different focus points. A recommendation
for future measurements is to use thinner walls or even better is to install plane mea
surement windows into the wall. It is also interesting to study the inﬂuence of different
shapes of the inertance entrances on the ﬂow.
150 Flow measurements in coaxial regeneratorbased devices
−10 −8 −6 −4 −2 0
0
2
4
6
8
10
0.05
0.1
0.15
0.2
0.25
0.3
0.35
(a) D
r
= 2.60%
−10 −8 −6 −4 −2 0
0
2
4
6
8
10
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
(b) D
r
= 5.23%
−10 −8 −6 −4 −2 0
0
2
4
6
8
10
0.5
1
1.5
2
2.5
(c) D
r
= 10.5%
Figure 7.5: Vector plot of the streaming velocity ﬁeld in window 2. The positions at the axes are
in mm. The color and the vectors lengths are measures of the velocity. The velocities to the right
of the color bars are in m/s.
−4 −2 0 2 4 6
0
2
4
6
8
10
0.1
0.2
0.3
0.4
0.5
0.6
(a) D
r
= 2.60%
−4 −2 0 2 4 6
0
2
4
6
8
10
0.5
1
1.5
(b) D
r
= 5.23%
−4 −2 0 2 4 6
0
2
4
6
8
10
0.5
1
1.5
2
2.5
3
3.5
(c) D
r
= 10.5%
Figure 7.6: Vector plot of the streaming velocity ﬁeld behind the bypass.
Bibliography
[1] D. Gedeon, “Dc gas ﬂows in stirling and pulsetube cryocoolers,” in Cryocoolers 9
(R. G. Ross, ed.), pp. 385–392, 1997.
[2] S. Zhu, P. Wu, and Z. Chen, “Double inlet pulse tube refrigerators: An important
improvement,” Cryogenics, vol. 30, pp. 514–520, 1990.
[3] S. Backhaus and G. W. Swift, “A thermoacousticstirling heat engine: Detailed
study,” Journal of the Acoustical Society of America, vol. 107, pp. 3148–3166, 2000.
[4] G. Swift, A Unifying Perspective for Some Engines and Refrigerators. Melville: Acousti
cal Society of America, 2002.
[5] J. A. Lycklama ` a Nijeholt, M. E. H. Tijani, and S. Spoelstra, “Simulation of a
travelingwave thermoacoustic engine using computational ﬂuid dynamics,” Jour
nal of the Acoustical Society of America, vol. 118, pp. 2265–2270, 2005.
Summary
Thermoacoustics concerns phenomena in which an interaction of acoustics with ther
modynamics takes place. In an acoustic wave the gas parcels always undergo temper
ature variations, which is a consequence of their compression and expansion. In adia
batic sound waves these temperature variations go unnoticed. However, when a solid is
present near the acoustic wave, the wave interacts with the solid and can cause a trans
fer of heat from one location in the solid to another. This is called the thermoacoustic
heatpumping effect. This effect is the driving mechanism in stackbased coolers and
heat pumps. Vice versa, when a sound wave interacts with a solid with a temperature
gradient above a certain critical value, the temperature gradient enhances the sound
wave.
One of the major disadvantages of thermoacoustics is that the power density of
acoustic waves is relatively low. One way to increase the power density is to use higher
drive ratios. Unfortunately the linear theory of thermoacoustics is only valid at low am
plitudes (drive ratios up to 3%). Nonlinear effects that are not taken into account by the
linear theory include vortex shedding at the end of a stack, dissipation at the ends of
a stack due to the sudden change in cross section, transition to turbulence inbetween
plates, and streaming. The objective of this PhD work is to gain a better understanding
of different phenomena that are occurring in thermoacoustic devices with an emphasize
on nonlinear effect, that occur at high amplitudes.
Thermoacoustics is a complicated but also very interesting phenomena, since three
ﬁelds of research, i.e. acoustics, ﬂow dynamics, and thermodynamics all come together
and all of them bring different key quantities, i.e. pressure, velocity, and temperature.
To gain a better understanding of different phenomena that are occurring in thermoa
coustic devices we have measured, calculated, and studied all three quantities.
In order to measure the pressure, three microphones were installed in the resonator
on each side of the stack. Following the multimicrophone method, we determined
the acousticenergy ﬂows at both sides of the stack. The found difference between the
acousticenergy ﬂows at both sides of the stack equals the energy absorbed by the stack.
We have been able to show that at low amplitudes the absorbed energy is in good agree
ment with the linear theory of thermoacoustics. Unfortunately at high drive ratios the
linear theory underestimates the energy losses in the stack, which is mainly caused by
minor losses at the two stack ends. Furthermore, the multimicrophone method is used
to determine the transfer matrix of a stack. These transfermatrix elements are employed
to determine the Rott functions as functions of the frequency, which we found to be in
good agreement with analytical ﬁts.
A ﬂow visualization technique called PIV is used to measure the velocity. By de
termining the displacement of very small oil droplets (typically 1 µm in diameter) that
move along with the gas during a short time interval (typically 1 µs), a velocity vector
ﬁeld is determined. Even using a measurement window as small as 3 mm 2 mm, a
velocity vector ﬁeld of 100 75 vectors can be obtained, during a time interval as small
as 0.1 µs. Moreover, the powerful PIV technique is used to study the vortex shedding
behind the plates of a parallelplate stack, the velocity proﬁle inbetween two parallel
plates, and a timeaverage velocity, called streaming velocity. The velocity proﬁles in
between two plates showa change in phase dependency when above a critical Reynolds
number, indicating a transition from laminar to turbulent ﬂow.
In order to register the stack temperature proﬁle as a function of time, we installed
32 thermometers in a stack plate. Furthermore we made a model of the energy balance
of the stack. The calculated temperature proﬁles as functions of time were found to be
in good agreement with the temperature measurements.
We have used different measurement techniques, microphones, thermometers, and
PIV, to get a more complete view of different thermoacoustic phenomena. We have
studied various nonlinear effects: minor losses at the stack ends, streaming, and a tran
sition to turbulence in a parallelplate stack. Also we have made a complete model
of linear thermoacoustics, based on the established linear theory, that can predict the
temperature proﬁle in a stack, even dynamically. We have gained more insight in ther
moacoustics and nonlinear effects in particular and we hope that other researchers and
the industry will beneﬁt from this.
Samenvatting
Thermoakoestiek betreft fenomenen waarin een interactie van akoestiek met thermo
dynamica plaatsvindt. In een akoestische golf ondergaan de gaspakketten altijd tem
peratuurvariaties, ten gevolge van compressie en expansie. In adiabatische geluids
golven blijven deze temperatuurvariaties in het algemeen onopgemerkt. Wanneer de
geluidsgolf echter in thermisch contact komt met een vast lichaam, vindt er een inter
actie plaats die ertoe kan leiden dat warmte wordt verplaatst van de ene locatie naar
een andere. Dit wordt het thermoakoestisch warmtepompeffect genoemd. Op dit ef
fect is het mechanisme in staandegolfkoelers en warmtepompen gebaseerd. Vice versa,
wanneer een geluidsgolf in wisselwerking staat met een vast lichaam, waarvan de tem
peratuursgradi¨ ent een kritieke waarde overschreidt, versterkt de temperatuurgradi¨ ent
de geluidsgolf.
E´ en van de belangrijkste nadelen van thermoakoestiek is dat de energiedichtheid
van de akoestische golven vrij laag is. De energiedichtheid kan vergroot worden door
hogere drukamplituden (drive ratio’s) te gebruiken. Helaas is de lineaire theorie van
thermoakoestiek slechts geldig bij relatief lage drukamplituden (drive ratio’s tot 3%).
Nietlineaire effecten die niet worden meegenomen in deze lineaire theorie zijn o.a.
wervelafschudding aan het uiteinde van stackplaten en de hierdoor veroorzaakte dissi
patie aan het uiteinde van de stack, een overgang naar turbulentie tussen de stackplaten,
en een DCcomponent in de snelheid, beter bekend als streaming. De doelstelling van
dit promotieonderzoek is het verkrijgen van beter inzicht in de verschillende fenome
nen  in thermoakoestische machines, waarbij de nadruk ligt op nietlineaire effecten 
die optreden bij relatief hoge amplituden.
Thermoakoestiek omvat gecompliceerde maar ook zeer interessante fenomenen, om
dat het zich afspeelt daar waar de drie vakgebieden akoestiek, stromingsleer en thermo
dynamica elkaar overlappen. Deze drie vakgebieden brengen elk hun eigen grootheden
met zich mee, te weten druk, snelheid en temperatuur. Om beter inzicht te krijgen in
de verschillende fenomenen die voorkomen in thermokoestische apparaten, zijn al deze
drie grootheden gemeten, berekend en bestudeerd.
Voor de drukmetingen zijn zes microfoons in de resonator geplaatst, aan ieder stack
uiteinde drie. De akoestische energiestroom aan beide stackuiteinden is bepaald met
de multimicrofoonmethode. Het gemeten verschil tussen de twee energiestromen aan
beide uiteinden is gelijk aan het vermogen geabsorbeerd door de stack. We hebben
laten zien dat voor lage drukamplituden het geabsorbeerde vermogen consistent is met
de lineaire theorie van thermoakoestiek. Voor relatief hoge drukamplituden onderschat
deze theorie het geabsorbeerde vermogen, hetgeen voornamelijk te wijten is aan minor
losses aan de stackuiteinden. Daarnaast is deze multimicrofoonmethode gebruikt voor
het bepalen van de overdrachtsmatrix van een stack. De overdrachtsmatrixelementen
zijn gebruikt voor het bepalen van de Rottfuncties als functie van de frequentie, welke
consistent met analytische vergelijkingen bleken te zijn.
Voor het bepalen van snelheid is de stromingsvisualisatietechniek PIV gebruikt. Het
bepalen van de verplaatsing van kleine oliedruppeltjes (typisch 1 µm in diameter), die
meebewegen met het gas, tijdens een kort tijdinterval (typisch 1 µs), resulteert in een
snelheidsvectorveld. Zelfs gebruikmakend van een meetvenster zo klein als 3 mm bij
2 mm, kan een snelheidsvectorveld van 100 bij 75 vectoren worden gemeten tijdens
een tijdsinterval van 0.1 µs klein. Voorts wordt de krachtige PIV techniek gebruikt om
wervelafschudding te bestuderen aan de uiteinden van een parallelleplaatstack, het
snelheidsproﬁel tussen twee parallele platen en een tijdsgemiddelde snelheid, ook wel
streaming genoemd. De snelheidsproﬁelen tussen twee platen vertonen een verander
ing in faseafhankelijkheid wanneer het Reynoldsgetal een kritische waarde overschrijdt,
hetgeen op een overgang van laminair naar turbulent duidt.
Om het proﬁel van de stacktemperatuur als functie van tijd te registreren, zijn 32
thermometers in een stackplaat ingebouwd. Om de metingen te kunnen verklaren is
de energiebalans van de stack gemodelleerd. De berekende temperatuurproﬁelen als
functies van tijd bleken consistent te zijn met de temperatuurmetingen.
We hebben verschillende meettechnieken, microfoons, thermometers, en PIV ge
bruikt, om een vollediger beeld te krijgen van verschillende thermoakoestische fenome
nen. We hebben diverse nietlineaire effecten bekeken: minor losses aan de stackuitein
den, streaming en een overgang van laminair naar turbulent tussen de stackplaten.
Ook hebben we een volledig model gemaakt van lineaire thermoakoestiek, gebaseerd
op de gevestigde lineaire theorie, dat het temperatuurproﬁel in een stack kan voor
spellen, zelfs tijdsafhankelijk. We hebben op deze manier meer inzicht gekregen in
thermoakoestiek en nietlineaire effecten in het bijzonder en we hopen dat andere on
derzoekers en de industrie hiervan kan proﬁteren.
Dankwoord
Een groot project, zoals een promotieonderzoek, doe je niet alleen. Er zijn talloze mensen
aan wie ik dank verschuldigd ben, omdat ze elk op hun eigen wijze hebben bijgedragen
aan mijn promotieonderzoek en aan het tot standkomen van dit proefschrift. Enkele
mensen wil ik in het bijzonder vermelden.
Allereerst natuurlijk mijn promotor, Fons de Waele, vanwege zijn grote kennis van
thermodynamica en pulsbuiskoelers, zijn didactische en theoretische kwaliteiten en
omdat hij altijd een originele en verfrissende kijk had op mijn resultaten. Ook voor
de adviseur in dit project, Mico Hirschberg, was een belangrijke rol weggelegd. Als
akoestisch specialist had hij altijd goede adviezen, over het bouwen van de opstelling,
meetmethodes, analyses, numerieke simulaties, interpretatie van meetresultaten en het
aanbevelen van relevante literatuur.
Met de technici Leo van Hout, Paul Bloemen, Paul Ni¨ el en Peter Helfferich heb
ik altijd zeer plezierig samengewerkt, veel dank voor al jullie technische ondersteun
ing. Daarnaast ben ik de medewerkers van de werkplaats, Marius Bogers, Henk van
Helvoirt, Frank van Hoof, Han den Dekker en Ginny FransenTer Plegt, veel dank ver
schuldigd voor de vervaardiging van onderdelen en stacks. Zij stonden altijd voor
me klaar voor het doen van technische klussen. In het tot stand komen van de grote
akoestische opstelling gaat in het bijzonder mijn dank uit naar Jaap Hoevenaar, die alle
hoornsegmenten heeft gedraaid op de grote freesbanken bij de GTD.
Met het wiskundeteam binnen dit project heb ik altijd heel prettig samengewerkt.
Allereerst wil ik Peter in ’t panhuis bedanken voor de goede samenwerking en vriend
schap en verder ook Sjoerd Rienstra, Jaap Molenaar en Han Slot.
Ik wil gaarne de stichting STW bedanken voor hun ﬁnanci¨ ele bijdrage, en in het bij
zonder Corine Meuleman voor de begeleiding. Voorts wil ik de medewerkers van ECN,
Aster Thermoacoustics en Shell bedanken voor de interessante discussies. In het bijzon
der wil ik noemen Okke Ouweltjes van Philips voor zijn bijdrage aan de electroakoestis
che berekeningen en Cees de Blok voor het maken van het coaxiale koppelstuk voor de
lopende golf metingen.
Gunes Nakiboglu en Wenqing Liang ben ik dank verschuldigd voor hun assistentie
bij de numerieke simulaties en mijn stagiaires Jan van Kemenade en Wim Weltjens voor
alle metingen, analyses en discussies.
Met Yan Li heb ik als kamergenote en collega een goede band opgebouwd. Tijdens
de lunchpauzes zorgden Paul Ni¨ el, Yan Li en Wenqing Liang voor gezelschap en ver
maak, mede door de vele spelletjes die gespeeld zijn in de kofﬁeruimte.
Gezelligheid en steun buiten het werk is ook erg belangrijk om een dergelijk groot
project af te ronden. Allereerst wil ik mijn ouders bedanken, omdat ze me altijd zijn blij
ven steunen. Mijn vriendin, Nicole, die ik pas in de laatste fase van mijn promotie heb
leren kennen, heeft mij ook altijd gesteund en bovendien geholpen met het corrigeren
van spelling en grammatica. Heel belangrijk waren ook mijn makkers Coen Oosse, Joost
Heltzel, Prahlada Belle en Raoul Lemmen. Verder wil ik mijn vrienden van schaakclub
’t Pionneke en mijn oudmedestudenten, Niels, Jeroen, Leon, Rob, Bart en Joost hier
graag vermelden.
In de periode van het schrijven van mijn proefschrift heeft Sonja FeinerValkier me
geholpen om door een moeilijke fase heen te komen.
Degene die ik verreweg het meeste dankbaar ben is Jos Zeegers. Hij stond altijd voor
me klaar, was altijd erg betrokken en wist me steeds te motiveren. Vanaf het aanvragen
van project, het ontwerpen van de opstelling, het leggen van contacten met univer
siteiten, de bezoeken aan universiteiten en conferenties in Frankrijk tot het schrijven van
papers en dit proefschrift. Hij heeft op vele manieren geholpen en een betere begeleider
had ik me niet kunnen wensen.
Curriculum Vitae
Paul Aben was born the ﬁrst of September 1981 in Sittard, the Netherlands. After ﬁnish
ing his preuniversity education (Atheneum) in 1999 at Bisschoppelijk College Broekhin
in Roermond, he started his studies in Technical Physics at Eindhoven University of
Technology in Eindhoven that same year. Within in one year he ﬁnished his ﬁrstyear
diploma (propedeuse) and won the CIVI encouragement award for being the student
with the highest grades in the Netherlands. During his studies he did an internship at
Oc´ e, Venlo, entitled “Warmtehuishouding in digitale kopieerapparaten”. In 2005 he gradu
ated within the Elementary Processes in Gas discharges (EPG) group in Eindhoven on
“Optical Study of Breakdown Phenomenology in Metal Halide Lamps”, under the supervision
of dr. Freddy Manders and prof.dr.ir. Marco Haverlag.
In 2005 he started a PhD project at Eindhoven University of Technology, within the
LowTemperature Physics (LTE) group. This project, under the supervision of dr.ir. Jos
Zeegers and prof.dr. Fons de Waele, was sponsored by the Dutch Technology Foun
dation (STW). He defended his thesis, entitled “HighAmplitude Thermoacoustic Flow In
teracting with Solid Boundaries” on the eight of December 2010. As of July 2010 he is
employed at ASML in Veldhoven.