You are on page 1of 133

The Pennsylvania State University

The Graduate School
Graduate Program in Acoustics



A Thesis in
Matthew E. Poese

c 2004 Matthew E. Poese

Submitted in Partial Fulfillment
of the Requirements
for the Degree of

Doctor of Philosophy

May 2004
The thesis of Matthew E. Poese was reviewed and approved∗ by the following:

Steven L. Garrett
Professor of Acoustics
Thesis Advisor, Chair of Committee

Thomas B. Gabrielson
Associate Professor of Acoustics

Horacio Perez-Blanco
Professor of Mechanical Engineering

Anthony A. Atchley
Professor of Acoustics
Chair, Intercollege Program in Acoustics

Signatures are on file in the Graduate School.

The design and the performance of regenerator-based thermoacoustic refrigerators
are described in this dissertation. The incorporation of a novel hybrid acousto-
mechanical resonator reduces the length of electrically driven thermoacoustic ma-
chines from about 1/7 of a wavelength to 1/100 of a wavelength. The design also
includes an acousto-mechanical feedback network that allows more efficient energy
feedback than previous thermoacoustic machines built with an energy recovery
mechanism. Through an evolution of three machines, this dissertation traces the
growth of these concepts and includes many of the relevant details of each machine.
Detailed models of these machines are also presented, and the measured perfor-
mance of the two prototypes match the model very well in second order quantities
such as cooling power and efficiency as well in first order measurable parameters
such as acoustic pressure.

Table of Contents

List of Figures vi

List of Tables viii

Acknowledgments ix

1 Introduction 1
1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Ideal heat engine cycles . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 How a regenerator is utilized to pump heat efficiently . . . . . . . . 4
1.3.1 The function of the regenerator and its advantages compared
to a stack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Stack-based Thermoacoustics . . . . . . . . . . . . . . . . . 10
Regenerator-based Thermoacoustics . . . . . . . . . . . . . . 12
1.3.2 Using a regenerator in an acoustic heat engine . . . . . . . . 12
1.3.3 The acoustic phasing network . . . . . . . . . . . . . . . . . . 14
1.4 Motivation for a regenerator-based refrigeration machine . . . . . . 19
1.4.1 Thermoacoustic-Stirling as a replacement to vapor compres-
sion refrigerators . . . . . . . . . . . . . . . . . . . . . . . . . 19

2 A Tabletop Regenerator-based Thermoacoustic Refrigerator 25
2.1 The Helmholtz Resonator as Acoustic Feedback Network . . . . . . 27
2.2 The Circuit Analogy to Lumped Acoustic Networks . . . . . . . . . 36
2.3 Summary of Tabletop Demonstration Device . . . . . . . . . . . . . 39

3 Design and Characterization of the Pre-Prototype 41
3.1 Design Considerations . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.1 Regenerator Design . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.2 Feedback Network . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.3 Heat Exchanger Considerations . . . . . . . . . . . . . . . . . 53
3.1.4 The Acousto-Mechanical “Bellows-Bounce”
Resonator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.5 Matching the Linear Motor to the Resonator Load . . . . . . 58

3.1.6 Detailed DeltaE Model . . . . . . . . . . . . . . . . . . . . . 61
3.2 Apparatus, Measurements and Results . . . . . . . . . . . . . . . . 78
3.2.1 Instrumentation and Measurement . . . . . . . . . . . . . . . 81
Pressures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Motor Dynamics . . . . . . . . . . . . . . . . . . . . . . . . 81
Temperature and Heat Flow . . . . . . . . . . . . . . . . . . 82
Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . 85
3.3 Measurement Results and Analysis . . . . . . . . . . . . . . . . . . 85

4 The Prototype Ice-Cream Storage Cabinet 92
4.1 New Aspects of the Prototype . . . . . . . . . . . . . . . . . . . . . 93
4.1.1 Vibromechanical Multiplier . . . . . . . . . . . . . . . . . . . 94
4.1.2 Heat Exchangers . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.1.3 Exhaust Heat Exchanger Design . . . . . . . . . . . . . . . . 105
4.2 Prototype Performance . . . . . . . . . . . . . . . . . . . . . . . . . 113

5 Conclusions 119

Bibliography 123

List of Figures

1.1 Polytropic state changes . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Stirling cycle p-v diagram . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Schematic of alpha-Stirling machine . . . . . . . . . . . . . . . . . . 6
1.4 Energy flow in a perfect regenerator . . . . . . . . . . . . . . . . . . 16
1.5 The abbreviated evolution of Stirling machines . . . . . . . . . . . . 18
1.6 Photographs of two kinematic Stirling machines . . . . . . . . . . . 22

2.1 Photograph of the table-top demonstration refrigerator . . . . . . . 25
2.2 Photograph of the thermal core in the demonstration refrigerator . 26
2.3 Resonance curve for demonstration device feedback network . . . . 30
2.4 Stepped illustration of the gas motion in the feedback network dur-
ing one acoustic cycle . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Lagrangian view of gas thermodynamics in the regenerator over an
acoustic cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.6 Three Stirling cycle p-v diagrams driven differently . . . . . . . . . 33
2.7 Circuit that represents the feedback network . . . . . . . . . . . . . 38

3.1 Photograph of regenerator used in the pre-prototype . . . . . . . . 49
3.2 Screenshot of the LabVIEW program created as a design tool . . . 51
3.3 Photograph of the shell-and-tube heat exchanger . . . . . . . . . . . 54
3.4 Photograph of the parts used to make the shell-and-tube heat ex-
changer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.5 Diagram of the entire pre-prototype refrigerator . . . . . . . . . . . 57
3.6 Lumped element circuit diagram of the pre-prototype refrigerator . 58
3.7 Photograph of the cylindrical spring used to augment the gas spring
in the pre-prototype resonator . . . . . . . . . . . . . . . . . . . . . 59
3.8 Schematic of the DeltaE model file . . . . . . . . . . . . . . . . . 62
3.9 Photograph of the pre-prototype ready to operate . . . . . . . . . . 79
3.10 Photograph of the thermal core ready to be installed . . . . . . . . 80
3.11 Schematic of the constant head exhaust circuit . . . . . . . . . . . . 84
3.12 Screen-shot of the LabVIEW data acquisition program . . . . . . . 86
3.13 Unmeasured heat flow in the pre-prototype machine . . . . . . . . . 88
3.14 Graph showing pre-prototype performance and model results . . . . 89

3.15 Graph showing pre-prototype performance and model results . . . . 91

4.1 Graph of losses in an inertance tube . . . . . . . . . . . . . . . . . . 95
4.2 Bellows bounce resonator . . . . . . . . . . . . . . . . . . . . . . . . 98
4.3 Bellows bounce with pre-prototype configuration . . . . . . . . . . . 99
4.4 Bellows bounce with alternate pre-prototype configuration . . . . . 100
4.5 Bellows bounce with inverted pre-prototype configuration . . . . . . 101
4.6 Prototype schematic . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.7 Solid model of the prototype . . . . . . . . . . . . . . . . . . . . . . 103
4.8 Schematic of the prototype with the cold head . . . . . . . . . . . . 106
4.9 Detail of the exhaust heat exchanger tube/fin . . . . . . . . . . . . 107
4.10 Photograph of exhaust heat exchanger . . . . . . . . . . . . . . . . 108
4.11 Sankey diagram of acoustic power flow in the prototype . . . . . . . 117

List of Tables

1.1 Chronological list of elementary heat engine cycles . . . . . . . . . . 5

4.1 Thermophysical properties of helium gas at 1 MPa . . . . . . . . . 110
4.2 Thermophysical properties of secondary exchange fluids . . . . . . . 111
4.3 Measured prototype performance . . . . . . . . . . . . . . . . . . . 115

5.1 Power density of other thermoacoustic research machines . . . . . . 120


The completion of this degree has been a goal of mine since I was about fifteen years
old; consequently, many people have indirectly contributed and had an influence in
the completion of this goal. These people include my parents who emphasized the
importance of education and in particular my father, who nurtured my curiosity
and inspired me toward science. My wife Kelly has been unfailing in her support
of this pursuit and, without exception, all of my interests and goals. I am blessed
with a vast “second family” through the Youngstown YMCA and elsewhere that
have provided generous encouragement and balance.

Professor Garrett has been all I could have hoped for in a mentor; I cannot thank
him enough for his teaching, coaching and cheerleading. Although he wasn’t on the
committee, Bob Smith has been a second thesis advisor throughout my graduate
career, patiently sharing his vast knowledge of science and expertise in measure-
ment technique. Although I get to include it in my dissertation, the final prototype
described here is the work of all three of us and in no way would it have been suc-
cessful were it not for the superposition of the talents of our three-person team.

I am also indebted to Professors Atchley and Gabrielson for serving on my thesis
committee and including me in their work with acoustic sensors in urban search
and rescue. The students and faculty in the Graduate Program provide a wonderful
place to live and learn; in particular, Ray Wakeland contributed to my education
significantly through enjoyable hours of discussion, debate and problem-solving.

Lastly, my graduate career has been generously underwritten by the Applied Re-
search Lab through the Enrichment and Foundation Program. The Office of Naval
Research supported this work in its early stages up through the design and con-
struction of the pre-prototype. Ben & Jerry’s Homemade has become the deus ex
machina for the technology, having been one of the best sponsors we’ve ever had
and allowing us to propel the technology to exciting new heights.

Chapter 1


1.1 Overview
This dissertation traces the evolution of the first series of low-lift, regenerator-based
thermoacoustic refrigerators. It is organized with an introduction that contains an
explanation of thermoacoustic refrigerators that use a regenerator and the short
history of such machines. Following this introduction, the dissertation contains
a chapter for for each of the three machines that were designed, constructed and
tested during the course of this research. A final chapter concludes the work by
summarizing relevant details and providing a road map for future development of
this style of thermoacoustic refrigerator.
The progression of the development of a regenerator-based low-lift thermo-
acoustic refrigerator (where low-lift denotes a modest temperature span from am-
bient temperature to cold temperature of about 200 K) began with the construction
of a simple lecture demonstration to gain some confidence and understanding of
the design principles. In this chapter, a simple model of an energy feedback net-
work is presented for use in a regenerator-based thermoacoustic refrigerator. If
this simple refrigerator, which contained no secondary heat exchange path failed
to work, it would have caused a reevaluation of the understanding at a conceptual
level. The demonstration model was driven by a commercial-off-the-shelf loud-
speaker and used a latex diaphragm to suppress the steady gas flows induced by
the transport of acoustic energy in the network (such steady gas flows are generi-
cally referred to as streaming) (Gedeon, 1997). Since the machine did create some
cooling, enough confidence was gained to move forward to a device that contained
a heat exchanger.

The second machine in the series, designed to work in air at atmospheric pres-
sure, employed a discrete shell-and-tube hot heat exchanger component and was
thermally loaded with a electrical resistance wire on the cold side of the machine.
This machine was also built to test the idea of eliminating the purely acoustic res-
onator and instead resonating the moving mass of the loudspeaker/motor against
the stiffness provided by the relatively short (compared to a wavelength) closed
cavity enclosing the thermal components. This concept has the nickname “bellows
bounce” and is the subject of a patent application that at the time of this writing
has been allowed in entirety (Poese et al., 2003). The resulting machine has a
much larger power density than one utilizing a longer resonator that contains gas
whose inertia effects and stiffness effects are of equivalent magnitude. The bellows
bounce machine is also simpler to control and design since acousto-mechanical res-
onator is essentially a single degree of freedom one rather than a tightly coupled
two-degree of freedom resonator that includes the mass of the motor and the mass
of the gas contained in the vessel.
During the construction of this second machine, our research group negotiated a
contract with Ben and Jerry’s Homemade (B&J) ice cream company headquartered
in Burlington, Vermont to build a thermoacoustic refrigerator that would power
an ice-cream sales cabinet. To execute this contract, a machine based on the
concepts being tested with the construction of the second machine was planned.
Consequently, the second machine took on the name “pre-prototype” since the real
“prototype” machine would be built to the B&J specifications.
The “prototype” machine, designed to specifications that will allow it to keep
the volume of an ice-cream sales cabinet (about 200 liters) at ice-cream storage
temperatures (-20±2 ◦ C) uses helium at 10 atmospheres, uses the concept of the
acousto-mechanical resonator and has a parallel plate exhaust exchanger that is
fluid-backed for secondary exhaust side heat exchanger. The load exchanger is also
a parallel plate design and since it can form the cap to the pressure vessel, it offers
great flexibility for secondary heat exchange. The prototype machine benefits sub-
stantially from a novel feedback network that also uses a hybrid acousto-mechanical
concept for the feedback network. This new configuration of the feedback network
is called a “vibromechanical multiplier” and uses an ordinary loudspeaker cone
as the inertial component instead of relying on the gas for inertia. It is also the

subject of an issued patent (Smith et al., 2003) and the approach allows a much
more compact geometry while still enjoying all of the efficiency gains of using
a regenerator. Since the vibromechanical multiplier concept doesn’t require an
open gas path around the regenerator, Gedeon streaming does not rob capacity
from the machine. However, intra-regenerator streaming and thermal buffer tube
instabilities must still be avoided.

1.2 Ideal heat engine cycles
A machine that either converts heat flow into mechanical work, or forces heat to
flow from a cold object to a hot one using mechanical work can be described in
terms of the thermodynamic steps that make up one cycle of operation. The state
of the working medium at the end of each step can be predicted (if the state before
the change is known) using the expression p0 V0 γ = pn Vn γ where γ is a variable
that depends upon the details of the thermodynamic process by which the system
changed state. Thermodynamic state change is commonly categorized in five ways
listed below and shown on a p-v diagram in Fig. 1.1:

1. isobaric (constant pressure, γ = 0)

2. isothermal (constant temperature, γ = 1)

3. polytropic (1 < γ < cp /cv )

4. adiabatic (constant entropy, γ = cp /cv )

5. isometric (constant volume, γ = ∞)

A heat engine cycle generally has four steps executed in succession. These four
special cases of a polytropic process (and the generic case labelled “polytropic”)
can be combined to realize ten different heat engine cycles since only two unique
processes can be included in a cycle.1 These five thermodynamic processes can be
combined into ten identifiable elementary heat engine cycles, which are listed in
Table 1.1 on page 5 (Kolin, 1995)
One reason for the requirement that each process must occur twice per cycle is that a machine
built to execute a heat engine cycle must repeat its motion; for example, if a machine executes




Figure 1.1. The four special cases of polytropic state change are shown on a p-v
diagram. For an ideal gas, γ is the ratio of specific heats for the adiabatic case. When
considering an ideal gas, there are no physically meaningful values of γ between the ratio
of specific heats and ∞ (the constant volume case).

Rev. Robert Stirling’s contribution in 1816 to the development of heat engines
was to recognize the importance of the regenerator (he called it the “economizer”)
to increase efficiency, which was important as the price of coal had begun to rise
steadily since the beginning of the Industrial Revolution (Urieli & Berchowitz,
1984; Sier, 1995).

1.3 How a regenerator is utilized to pump heat
Following Figure 1.1 the p-v diagram for the Stirling cycle is shown in Fig. 1.2. A
schematic of a machine that can execute this cycle is shown in Fig. 1.3.
The schematic diagram of the alpha-Stirling refrigerator shown in Fig. 1.3 con-
a constant volume process during one-quarter of the cycle, it will (usually) mechanically execute
another one to return to the initial state.

Type of State Change Credit Year

1 & 5, isobaric and isometric Papin 1690
1 & 3, isobaric and polytropic Cayley 1807
2 & 5, isothermal and isometric Stirling 1816
2 & 4, isothermal and adiabatic Carnot 1824
2 & 3, isothermal and polytropic Reitlinger 1843
1 & 4, isobaric and adiabatic Joule 1852
2 & 1, isobaric and isothermal Ericsson 1853
4 & 5, adiabatic and isometric Otto 1867
3 & 4, polytropic and adiabatic Lorenz 1894
3 & 5, polytropic and isometric Crossley 1896

Table 1.1. A chronological list of ten elementary heat engine cycles shows that the
concept of a regenerator occurred early in the development of heat engines.





Figure 1.2. A qualitative graph of pressure vs volume for an articulated Stirling cycle.
The oval represents the cycle driven with sinusoidal pressure and velocity oscillations
and the outer curves represent an articulated cycle.






Figure 1.3. This figure shows the piston motion in an alpha-Stirling refrigerator. The
letter designations correspond to the p-v diagram in Fig. 1.2 and the meters above each
picture display the relative magnitudes of the gas pressure and average gas temperature.

tains a regenerator sandwiched between heat exchangers with two pistons, one on
either side of the regenerator. (These will be referred to as the cold and hot side
piston in this discussion.) At steady state, the ideal regenerator has a linear tem-
perature gradient across its length. The pistons are connected by some linkage
that provides for the correct relative motion of the pistons to make the machine
function as a refrigerator or an engine (or nothing useful). One usual simplification
is to suppose that each piston has the range of motion to be able to sweep out the
entire volume of its cylinder, right up to the edge of the heat-exchanger/regenerator
edge. The simplest view of the operation of this machine as a refrigerator contains
four distinct strokes in the following order:

A–B Displacement Both pistons move in concert to move the gas from the hot
side piston cylinder through the regenerator to the cold piston cylinder. The
gas is cooled reversibly as it passes through the regenerator since the gas
is assumed to be in perfect thermal contact with the regenerator material
due to the porous design of the regenerator. Since the gas is colder, heat
was (reversibly) absorbed by the regenerator from the gas. This movement
through the regenerator is carried out so that the volume that the gas oc-
cupies is constant throughout the step. Since the temperature of the gas
decreases at constant volume, the pressure of the gas must decrease during
this step. Although no p∆v work is done on or by the gas (since the volume
didn’t change) the energy lost by the gas from thermal conduction can be
expressed by the decrease in pressure at constant volume.

B–C Cooling Now that the gas is at a low temperature, but at the starting vol-
ume, the cold side piston moves away from the hot side piston (the separation
distance increases) which decreases the pressure of the gas even further. This
expansion would cause the temperature of the gas to decrease to a tempera-
ture below the coldest temperature of the regenerator if the step were adia-
batic. In the isothermal limit, during each differential step of this expansion
stroke the gas gets a differential amount colder which allows a differential
amount of heat to move from heat-exchanger located at the cold end of the
regenerator to the gas. The cold (or load) heat-exchanger is assumed to
have the same temperature as the cold end of the regenerator and like the

regenerator, this heat exchanger is in perfect thermal contact with the gas.
Although the ideal Stirling refrigerator exhibits an isothermal expansion dur-
ing this step, in real machines, this step is closer to an adiabatic expansion,
as noted below. The job of the load heat exchanger is to move the heat from
the thing to be cooled (e.g. ice-cream, medicine, bologna) into the cold side
cylinder so that it can be transferred to the gas during this cooling step.

C–D Displacement Following the cooling stroke the pistons move together to
drive the gas back through the regenerator toward the hot side piston. Dur-
ing this constant volume displacement, the gas is reversibly warmed to the
hot side regenerator temperature. Since the warming occurs at a constant
volume, the pressure of the gas increases during this step. It is during this
step that the parcel of gas recovers the energy given to the regenerator from
step [A]–[C]. It is in this reversible recovery of energy that the regenerator
proves its worth; because of the regenerator, the expansion stroke only has to
move heat at the cold temperature and doesn’t have to cool the charge of gas
from ambient temperature to the cold temperature since the displacement
stroke through the regenerator took care of that part. It is this idea of pre-
cooling and pre-heating the gas in preparation for heat transfer to or from a
load that makes the Stirling cycle and others regenerative heat engine cycles
useful. The fact that the regenerator can, in principle, do this pre-cooling
and pre-heating reversibly makes it efficient.

D–A Exhaust Once all of the gas has been reversibly warmed to the hot side
temperature, the hot side piston moves toward the cold side piston com-
pressing the gas that has been warmed. If isothermal, the amount of heat
that was removed from the load exchanger plus the work done to remove it is
absorbed by the exhaust heat exchanger. The job of the exhaust exchanger
is to provide a path for that heat to leave the hot side cylinder.

After reading this simplified step-by-step description of the operation some
complications might be evident:

• A machine cannot produce such articulated motions to be able to easily mark,
for example, the end of the cooling step and the beginning of the displacement

step. In real machines, the pistons usually move in a sinusoidal motion (as
shown in Fig. 1.2 on page 5) which not only blur the distinction between
each of the four steps, but also reduce the power density of the machine
for a given operating pressure range. This reduction in power density can
easily be understood by appreciating that the area enclosed by the lines of
the p-v diagram is the amount of work required to move heat from material
at the cold temperature to the environment. If a machine was operating
at a constant efficiency, it would be able to move more heat for a specified
pressure and volume oscillation amplitude if it could operate on the red
curves in Fig. 1.2 on page 5 (articulated piston motion) than it could if the
pistons moved with sinusoidal motion as shown on the blue ellipse.

• It would be difficult to make a machine that could execute isothermal ex-
pansions and contractions in the cylinder spaces on the left and right of the
regenerator in Fig. 1.3. If the relative scale of the schematic shown in Fig. 1.3
were used to build a machine, that machine would be better modelled with
adiabatic compressions and expansions in place of the isothermal ones de-
scribed above. These adiabatic compressions decrease performance below the
Carnot limit since during the adiabatic pressure change, the gas gets measur-
able warmer or colder than the regenerator end-temperatures. Consequently,
when this hotter or colder gas enters the regenerator (or heat exchangers)
heat moves across a finite temperature difference, which is an irreversibil-
ity that decreases maximum theoretical performance below the Second Law
The Stirling cycle, as presented with isothermal compressions and expansions
is reversible (in the limit of zero viscosity) and therefore has the theoretical
potential to reach the Carnot limit of efficiency. In earlier days of Stirling
machine design, assuming isothermal expansions and compressions made the
analysis of potential designs tractable in closed form; the first published
Also called the Carnot limit, the Second Law of Thermodynamics sets an upper limit for
the amount of heat that can be moved from a mass at a cold temperature and exhausted to the
environment at ambient temperature for a given amount of input work. This limit depends only
upon the cold temperature TC and the ambient temperature T0 and is expressed as COPC =
QC /W = T0T−T C
where QC is the amount of heat removed from the cold object and W is the
amount of work required to move that heat.

isothermal analysis of the machine was in 1871 by Gustav Schmidt, a profes-
sor at the German Polytechnical Institute at Prague. It was Schmidt, fifty
years after the invention of the Stirling machine, who first represented the cy-
cle with isothermal expansions and contractions because that model yielded a
closed-form solution. Almost one hundred years later, Finkelstein published
the first analysis of a Stirling machine with non-isothermal expansions and
compressions (Urieli & Berchowitz, 1984).

1.3.1 The function of the regenerator and its advantages
compared to a stack
The purpose of the regenerator in Stirling machines is to provide for the temporary
storage of heat for the oscillating working gas. The gas must be brought to the cold
temperature before the cooling (expansion) stroke is executed and then warmed
before the exhaust (compression) is executed. The amount of heat removed to
provide the cool gas for the expansion stroke is exactly the same amount of heat
that must be returned to the gas in preparation for the exhaust stroke. The
regenerator provides a place for the gas to temporarily store this heat; to do so it
must have enough heat capacity to be able to store enough heat to get the gas to the
desired temperature and it also must be as thermally non-conducting as possible
to reduce heat flow along its length, a heat flow that reduces the effectiveness of
the regenerator. For this reason, many regenerators are made of stacked stainless
steel screens which provide plenty of heat capacity and a large array of pore size
possibilities. Since each layer of screen has a significant amount of thermal contact
resistance, heat conduction along the length is minimized to an acceptable level.

Stack-based Thermoacoustics

In the last twenty years of the 20th century, much research and design was dedi-
cated to so-called “thermoacoustic” machines that utilized an oscillating gas with
nominally standing wave phasing in the presence of a “stack” of closely spaced,
rather non-conducting plates (Swift, 1988). This stack of plates was designed with
a plate-to-plate spacing on the order of a few thermal penetration depths, or the
acoustic thermal boundary layer thickness. This acoustic boundary layer is called a

“penetration depth” because it is the distance that heat3 can diffuse in one acoustic
cycle. The penetration depths are defined as
δκ = (1.3.1)
ρm cp ω

δν = (1.3.2)
ρm ω

where ω is the angular frequency of the gas oscillation and ρm is the mean gas
density, cp is the gas specific heat at constant pressure, K is the thermal conduc-
tivity and µ is the shear viscosity of the gas. Since the stack plate-to-plate spacing
might be twice this thermal penetration depth, the rate of heat transfer between
the plates and a gas parcel translating in a plane equidistant from two plates is
not fast compared to an acoustic period, as it might be for a parcel of gas near to
the plates. The lag of heat transfer with respect to the motion of the “active” gas
parcels (those parcels that are not too near the plates) is what allows the gas to
execute a useful heat engine cycle – one that either spontaneously converts heat
input into sound energy or one that uses sound energy to move heat from a cold
reservoir to an ambient temperature one.
This natural phasing idea, utilized by designing the stack with the proper plate
spacing for a particular thermal penetration depth (which depends on working
gas properties and frequency of oscillation) is a great innovation since it rids a
machine of any number of complicated methods to impose the required phase
difference between heat transfer and gas motion. For example, in an automotive
engine, each cylinder has at least two valves that must be opened and closed at the
right time via the motion of rockers, cams and pushrods. Most Stirling machines,
as well as other air and gas engines, also have complicated linkages that impose
the phasing of gas motion, pressure and heat input and exhaust. However, the
mechanical simplicity that comes with the exploitation of natural phasing brings
with it a significant penalty in efficiency of the cycle. The fact that the thermal
contact between the stack and the gas in the stack is not perfect (in order to get
the right “natural” phasing) the heat that is exchanged between the stack and the
gas is not exchanged in a reversible way. Since this heat transfer that occurs each
or momentum, in which case it’s referred to as the viscous penetration depth

cycle happens over a finite temperature difference, the theoretical limit in efficiency
of the stack-based thermodynamic cycle is lower than the Second Law limit.

Regenerator-based Thermoacoustics

As opposed to using a stack as the second thermodynamic medium (Wheatley
et al., 1986), the use of a regenerator allows the theoretical thermodynamic effi-
ciency to reach the Carnot limit. The function of a regenerator is slightly different
than that of a stack. While both the stack and regenerator provide heat capacity
for the temporary storage of heat, the details of how this heat is exchanged with
the gas are different. In the stack, this exchange is irreversible because the active
gas is one or a few thermal penetration depths away from the stack material (as
explained above). It is this irreversibility that is the source of the natural phasing:
a simplicity that can be desirable and worth the efficiency penalty. In a regenera-
tor, the gas passage pores are much tighter. This insures that all of the gas in the
regenerator is well within a fraction of a thermal penetration depth so that the
thermal diffusion time is significantly shorter than an acoustic period. (A rule of
thumb might be that the pore sizes are a tenth the size of a thermal penetration
depth.) Consequently, the transfer of heat between the regenerator and the gas is
theoretically reversible which allows the efficiency to approach the Carnot limit.
This gain in potential efficiency comes with the loss of the elegant natural phasing
found within the pores of the stack. Therefore, the phasing of heat transfer and
gas motion in the regenerator must be imposed by some other mechanism.

1.3.2 Using a regenerator in an acoustic heat engine
Since the natural phasing found in the stack is lost when taking advantage of the
improved properties of the regenerator-based Stirling cycle, the correct phasing
of heat transfer and gas motion must be imposed in some other way. In the late
seventies, Ceperley (1979) recognized that a travelling acoustic wave exhibits the
correct phasing to force the oscillating gas to execute a Stirling cycle. This phas-
However, since the pore size is small, the COP of any real incarnation of a regenerator-based
heat engine will be limited by viscous loss in the regenerator. In the comparison of stack-based to
regenerator-based machines, it is somewhat arbitrary to compare the limiting case of a working
fluid with zero viscosity since it’s effect is much bigger in the regenerator than the stack.

ing, where oscillating pressure is in phase with the velocity of the gas is enforced in
conventional kinematic Stirling machines by way of the mechanical linkage. Ceper-
ley built an acoustic-Stirling engine apparatus that included a regenerator inside
of a pipe that was several wavelengths long but his engines failed to produce acous-
tic energy. Although the regenerator was providing the needed energy storage, it
also presented a significant viscous loss to the acoustic wave passing through the
regenerator matrix. This attenuation, proportional to the square of the acoustic
velocity, was much too large to allow any acoustic pressure oscillations to build in
the tube; as the pressure would increase from the Stirling cycle taking place in the
regenerator, the acoustic particle velocity would also increase causing increased
attenuation in the regenerator which robbed the wave of the power produced by
phased heat transfer in the regenerator.
This is best understood by considering that the impedance (the complex quo-
tient of oscillating pressure divided by the gas velocity) of the wave travelling along
the tube that is several wavelengths long is simply the characteristic impedance
of the gas in the tube: the product of density and sound speed. This impedance
is relatively low compared to the impedance found near the closed end of a tube
that is 1/2 of a wavelength long. In such a tube, the magnitude of the impedance
near the closed end is large compared to the characteristic impedance of that gas;
the oscillating pressure is large in magnitude compared to the relative magnitude
of the gas velocity which becomes zero where the gas meets the capped end of the
tube. This location would be a good place to locate a regenerator, except for the
fact that in such a standing wave tube, the pressure is almost 90 degrees out of
phase with the gas velocity.
This requirement for high impedance or ratio of driving potential to flow ve-
locity is also found in the electrical power transmission network. Electrical power
(alternating current) is the real part of the product of the complex voltage (analo-
gous to pressure) and complex current (analogous to acoustic velocity) in the form
Ėelec = 12 Re[ẽ · ĩ∗ ] where ∗
represents the complex conjugate. The energy lost to
Joule heating is the product of the square of the current and the electrical resistance
(Πloss = i2 R) of the transmission line. Therefore, electrical power is transmitted
at much higher voltage amplitudes than we like to use in our homes and at a sig-
nificantly reduced current amplitude to reduce the Joule loss. Since power is the

product of the potential and the flow, the amount of power transmitted can be
maintained while the potential is increased and flow decreased proportionally to
minimize losses.
In a regenerator, it is desirable to deliver a particular amount of power to
the regenerator with the acoustic wave in order to move the required heat for an
application. The delivery of this power must have pressure and velocity in phase
and at an acceptably large impedance so that viscous losses in the regenerator do
not make the machine so inefficient as to be of no practical use. The first apparatus
to successfully exploit this idea of an acoustic-Stirling machine with the regenerator
at a high impedance location was in an engine built at Los Alamos by Backhaus &
Swift (2000). Their pioneering concept for a “thermoacoustic-Stirling” engine led
the way for the thermoacoustic-Stirling refrigerators that are the subject of this

1.3.3 The acoustic phasing network
For successful utilization of a regenerator, several requirements must be met. First,
pressure oscillations and velocity oscillations must be substantially in-phase. The
frequency of this oscillation must be low enough that the acoustic boundary layer
thickness is several times larger than the average pore size of the regenerator
matrix. The magnitude of the complex ratio of oscillating pressure to oscillating
velocity must be relatively large in comparison to the characteristic impedance of
the working gas. In addition to these elements, the energy flow in the machine is
In a Stirling refrigerator, the regenerator is a sink for mechanical (or acoustical)
energy that passes from the exhaust (ambient temperature) side to the load (cold
temperature) side. Energy in the form of mechanical work is supplied to the
regenerator (see Fig. 1.4) and the regenerator absorbs some amount of this work
in order to move heat against a temperature gradient from the cold end to the
ambient (or hot) end of the regenerator. In a regenerator that transfers heat
reversibly using a gas with no viscosity (this is a “perfect” regenerator and could
also be thought of as one that generates no entropy) the remaining amount of
work that flows out of the cold end of the regenerator is equal to the amount of

heat removed from the load and pumped to ambient temperature. In this perfect
regenerator, the amount of total energy at any location along the regenerator’s axis
is zero. Since total energy, called enthalpy H2 , is the sum of the work energy and
the heat energy in the regenerator (and would include other types of energy if the
magnitude was changing within the regenerator), the heat moving to the left (“up”
the temperature gradient) is equivalent to the magnitude of work energy that is
flowing from to the right (the direction of propagation of the acoustic wave). In
a machine with a perfect regenerator, in order to move an amount of heat δQ
per cycle some amount of work, δW , is absorbed per cycle in the regenerator as
demanded by Carnot.
The work required by the Second Law of Thermodynamics to move an amount
of heat δQ is related only to the temperatures between which the heat is moved
T0 and TC : δW = δQ T0T−T
. In a real regenerator, the amount of work absorbed
in the regenerator due to viscosity is greater than the Carnot minimum; since
this work does not contribute to moving any useful heat, this extra work raises
the total power in the regenerator to a value greater than zero. Likewise, since a
real regenerator allows some conduction of heat from the ambient side to the cold
(load) side, and heat exchange in the real regenerator has irreversibility associated
with it, the net amount of heat moved from the load to the ambient reservoir will
not be as great as it could have been for the amount of work flowing into the
regenerator. This reduction in the useful amount of heat moved “up” the thermal
gradient within the regenerator also increases the magnitude of total power in the
regenerator above the ideal value of zero. The concept of total power allows a
quantitative way to judge the “effectiveness” of the regenerator and allows the
designer to know if a proposed change to the design of the machine or regenerator
parameters will increase efficiency.
Within the Stirling community, machines are commonly classified into two
major categories: kinematic and free-piston. In a kinematic machine, the two
moving elements (either a piston and a displacer or two pistons) are connected
together by a mechanical linkage and the motion of the two elements with respect
to one another is completely defined by the mechanics of the linkage. Free-piston
machines are a much more recent development, largely credited to Beale (1971),
but could also be attributed to Ringbom (Sier, 1995, page 197). The recognition

Win Wout

Energy Flow



Figure 1.4. This graph shows the energy flow through a regenerator that produces
refrigeration. The dotted line represents the heat and the solid line shows the mechanical
work. Note that everywhere in the regenerator the sum of the heat and work is zero.
The useful cooling capacity in an ideal regenerator is equal in magnitude to the amount
of work that flows out of the regenerator at the cold (load) side.

that the mass of the displacer could be made to resonate against the stiffness of
trapped gas in the machine created a renewed interest in Stirling machines in the
early 1970’s. Free-piston Stirling machines offer a more compact and mechanically
simpler alternative to their kinematic brothers, however, these benefits come at the
price of increased design complexity since the dynamics of the heat transfer directly
impact the motion of the two moving elements which create the pressures – the
pressures that dictate how much heat moves. In a kinematic machine, the motion
of the piston/displacer is set mechanically and the thermodynamic events taking
place in the machine cannot impact the displacement magnitude (or therefore, the
pressure amplitude).
In either category, the free-piston or the kinematic, the energy flow in the device
is the same; energy flows into the ambient temperature end (exhaust side) of the
regenerator and out of the cold (load side). This amount of “left-over” energy
that leaves the regenerator must be recycled to the front of the regenerator for the
device to be efficient – this is another job of the mechanical linkage and flywheel of
the kinematic Stirling machine and in a free-piston machine this energy is stored
and released by the piston that resonates against the sealed volume of gas – the
“free” piston of the free-piston machine.
As shown in Fig. 1.5 the innovation of Beale (or Ringbom) was to recognize
that the complicated and expensive linkages found in kinematic Stirling machines
could be removed and replaced with a spring: either a mechanical spring or the
stiffnesss formed by gas in a sealed chamber. The improvement made by Backhaus
and Swift was to eliminate the piston that Beale left in his device. Backhaus and
Swift substitute for the piston a “slug” of gas that has relatively more inertia than
the rest of the gas in the cylinder that contains the regenerator.
The crux of Backhaus and Swift’s 1999 engine is the concept of an acoustic
phasing network that both delivers acoustic power to the regenerator at a high
impedance and correct phase and then returns the power that is not utilized to
move heat. (In the case of an engine, the amount of power that is not taken from
the wave to do work on a load is returned to the regenerator.) The function of
the phasing network is twofold: condition the impedance of the incoming acoustic
energy so that it can be efficiently utilized by the regenerator and then return
some un-utilized energy back to the regenerator. This dissertation will focus on










Figure 1.5. The major steps in the evolution of mechanisms to support a Stirling cycle
are shown in this figure as refrigerators (These topologies are equivalent in the case of
an engine.) The kinematic machine, which was born in the early 19th century, has many
forms and derivatives, only the simplest alpha configuration is shown here. The free-
piston Stirling concept started to be developed in about 1970 and although the acoustic-
Stirling idea was conceived of in 1970, the first working prototypes were produced in the
late 1990’s by de Blok (1998) in the Netherlands and in the third version at the bottom
by Backhaus & Swift (2000) at Los Alamos National Laboratory.

the design considerations of the phasing/feedback network for low-lift refrigeration
applications and present a few network configurations.

1.4 Motivation for a regenerator-based refriger-
ation machine
Now that some historical and theoretical background has provided a context for
the research that will be described in this thesis, it is important to provide the
motivation for pursuing thermoacoustic-Stirling technology.

1.4.1 Thermoacoustic-Stirling as a replacement to vapor
compression refrigerators
The only real reason to look for a replacement to any established technology on
the market today is for cost reduction. If an alternative technology is cheaper,
then it is desirable. The converse is also true; if the alternative is more expensive
in some way, then it is undesirable. There are at least three considerations that
can influence cost:

Cost to manufacture The “commodity level” vapor-compression refrigerator that
is found in the kitchens of most homes in the industrialized world is inex-
pensive. However, since a thermoacoustic machine has few moving parts,
requires no exotic materials and has no inherently close dimensional toler-
ances, a commodity thermoacoustic machine stands some chance of being
manufactured as cheaply as a vapor compression machine. It is true that
thermoacoustic machines, as do all closed-cycle machines, require some sec-
ondary heat exchange mechanism on both the hot and cold sides. This
subsystem does increase complexity, but for some applications like fountain
beverage vending this fluid loop is integral since it’s the product delivery sys-
tem. In grocery store display cases and geothermal heat pumps, secondary
heat transfer systems are already in use with conventional compressor tech-

Cost to operate The cost of energy purchased to operate the refrigeration ma-
chine is minimized by increasing its efficiency. This cost is important for
some applications where the energy is hard to transport to the location of
the machine, or applications that demand significant amounts of cooling and
therefore reducing the energy demands of the machine saves a significant
amount of money. On the other hand, the efficiency of a cooling machine is
also important for some applications; in the world of commercial ice-cream
freezers, the company who makes/markets the ice-cream usually provides a
freezer to a grocery/convenience-store for no charge (as long as their prod-
uct continues to sell at the store) and the shop keeper pays the operating
costs. In this case, efficiency may be of secondary importance to acquisition
cost. For large capacity applications such as grocery stores and refrigerated
warehouses efficiency becomes more important since the energy and acquisi-
tion cost is borne usually by the same organization and the electrical power
consumption is large. Thermoacoustic machines, with a maximum possi-
ble efficiency equal to Carnot, can be made to be efficient and like Stirling
machines, may become competitive with vapor compression machines.

Cost to maintain/dispose Thermoacoustic machines use no environmentally
harmful working fluids/gases. Consequently, disposal is inexpensive com-
pared to vapor compression machines which currently utilize gases two to
three thousand times more potent as “greenhouse” gases than CO2 . Since
the machine uses no lubricants, maintenance is expected to be small. Com-
pared to large vapor compression machines which have compressors and ro-
tating machinery that require maintenance, thermoacoustic technology with
its linear motors can reduce life-cycle costs.

Pitting thermoacoustic technology head-to-head with the established vapor
compression technology (which is 100 years old) is daunting for thermoacoustics.
Vapor-compression is such a mature technology and the market has not yet found
vapor-compression to be lacking in the three categories above. However, if environ-
mental legislation were to be enacted, either banning certain refrigerants or forcing
consumers to pay a premium for buying and disposing of a machine that contains
a global warming gas, thermoacoustic technology could become more attractive in

the “cost to buy/manufacture” and the ”cost to dispose” categories. As the direct5
cost of energy increases, due to pollution “taxes” or carbon credits as specified by
the Kyoto Accords, thermoacoustics with its high inherent efficiency might also
be attractive in the “cost to operate” category. Support for technologies that do
not use global warming gases is much stronger in Europe at this time than it is in
North America.
However, direct competition with vapor compression technology may not be the
only path toward commercialization of thermoacoustic technology. There is surely
an application for a machine that is quiet, efficient, has low moving mass and can
easily take advantage of proportional control that could propel the technology to
prominence even at a cost premium. Consequently, putting research efforts into
increasing the power density, reducing cost and increasing efficiency are not wasted
– especially in light of the fact that thermoacoustic technology is functionally
reversible. This means that knowledge gained by research into the machine as a
refrigerator is directly transferable to the design of a machine that is built as an
Thermoacoustic machines that are built to convert heat energy to mechanical
energy (and possibly then to electrical energy) might be a attractive commercial-
ization target. In-home electrical generators that burn natural gas or fuel oil and
both create the electrical power for a home and heat the home and the water
supply might allow a household to reduce the number of subscribed utilities by at
least one and eliminate the significant transmission loss associated with electrical
power delivery that is wasteful of the global natural resources.
Now that a clear link has been established between the operation of thermo-
acoustic technology and Stirling machines, we should ask “Why spend money/time
developing thermoacoustic machines in favor of developing Stirling machines?” Al-
ternatively, this question could be asked of a developer of thermoacoustic technol-
ogy: “What makes you think that thermoacoustic machines will be commercially
feasible when Stirling machines haven’t reached a “commodity status” during the
nearly 200 year history of its development?6 ” There are several clear advantages
The direct cost of energy is exclusive of the indirect costs such as that associated with natural
resource depletion, military protection of oil prices, coal mine runoff and subsidence, etc.
Kinematic Stirling machines attained a brief “commodity status” at the start of the 20th
century as water pumps and compressors.

Figure 1.6. On the left is a photograph of a Rider-Ericson engine produces in the late
1800’s and early 1900’s. Notice the complicated linkage that connects the flywheel to
the piston rods that penetrate the top of the piston cylinder. On the right is a modern
Stirling machine made by Whisper Tech ( that is used
to generate electricity for residences, yachts and motor-homes. Hidden in the stainless
steel pressure vessel is a complicated linkage called a “wobble-yoke.” At the time of this
writing, the WhisperGen cost over ten thousand dollars.

and some disadvantages of thermoacoustics when compared to Stirling.

Mechanical Complexity Stirling machines, and kinematic Stirling machines es-
pecially, are mechanically complex. A kinematic machine requires some me-
chanical means to link the pistons in such a way to create the correct phasing
of pressure oscillation and gas displacement. In Fig. 1.6 two kinematic ma-
chines are shown that were made at the extremes of the twentieth century
and both are burdened with this linkage that requires lubrication, experi-
ences wear, dissipates energy and is costly to manufacture and maintain.
Many Stirling machines, both kinematic and free-piston, have a piston (to
create pressure oscillations) and a displacer (to move the gas through the
regenerator at a roughly constant volume) that must move in the same bore.
In many designs, this requires that the rod for the displacer penetrate the

piston and sometimes that both of the connectors for these mechanical com-
ponents penetrate the pressure vessel that contains the working fluid (usually
helium). These dynamic seals, sometimes in the presence of either hot or cold
temperatures, present a difficult engineering problem that is usually costly to
implement due to close-tolerance manufacturing requirements that provide
simultaneously low friction and low leakage. The most significant contribu-
tion of thermoacoustic technology to the development of regenerative heat
engines is simplicity. By eliminating the displacer or second piston (and in
a prime mover, by removing all moving metal) in favor of gas inertia and by
exchanging a mechanical phasing mechanism for one that is executed using
the compliance and inertia of the working fluid itself, a much simpler and
consequently cheaper machine can be made. Thus far, this reduction of com-
plexity has probably come at the price of power density. In the machines
that are the subject of this thesis, the sliding seals that dominate Stirling
machines have been replaced by flexure seals. These seals, while simple and
low maintenance require careful design for “infinite” fatigue life. At this
time, the metal bellows that are used are expensive.

Power Density Currently, the power density of thermoacoustic machines is not
as high as that of its Stirling “brothers” nor as high as either vapor compres-
sion refrigerators or internal combustion engines. The relatively low power
density comes as consequence of relying on the stiffness (compliance) of gas
trapped in the resonator to provide restoring force and moving gas to provide
inertia. Compared to a mechanical spring, gas springs are much bigger for a
given stiffness and the density of gas is much lower than the density of metal
or plastic used as pistons in traditional Stirling machines. That said, since
the compliance of trapped gas is a function of the mean pressure in the gas
(Cg = V /γpm ) and since density is also proportional to mean pressure, by
increasing the mean pressure in a thermoacoustic machine, the power density
can be improved although the cost/weight/complexity of a stronger pressure
vessel must be considered.

Moving Mass The moving mass of thermoacoustic machines is typically low for a
given amount of cooling capacity (or power output if operating as an engine)

compared to Stirling machines. This has advantages with respect to fatigue
life as well as transmitted noise/vibration levels and radiated noise levels.

Resonance One overarching advantage of thermoacoustic machines described
herein is that since the energy storage medium is the working gas itself and
the moving mass, there are no requirements for external energy storage ele-
ments (like flywheels) that are found in many traditional Stirling machines.
Chapter 2

A Tabletop Regenerator-based
Thermoacoustic Refrigerator

To gain some experience with the idea of a regenerator-based thermoacoustic refrig-
erator, a small table-top demonstration device was built (Poese & Garrett, 2001).
The resonator of this device is made from PVC plumbing parts and powered using
a commercial, off-the-shelf loudspeaker, Dynaudio Model 15W-751 . Photographs
of the tabletop demonstration unit are shown in Figs. 2.1 and 2.2.
This simple design establishes approximately a 20◦ C (36◦ F) temperature dif-
Dynaudio A/S, Sverigesvej 15, DK-8660 Skanderborg, Denmark

Figure 2.1. This photograph shows the table-top demonstration refrigerator with a
Dynaudio loudspeaker in a PVC enclosure at the left end, a standing wave tube with a
double-necked Helmholtz resonator termination at the right end of the tube. One “neck”
of the Helmholtz resonator houses the regenerator and the other “neck” constitutes the
inertial element of the conventional Helmholtz resonator.

Figure 2.2. This photograph shows the thermal core of the table-top demonstration
refrigerator. The regenerator is behind the copper screening that acts as an exhaust
heat sink. A copper ring holds the regenerator and separates it from the inertial portion
of the feedback network which are the three “windows” cut into the plastic ring that
suspends the copper ring in the resonator tube.

ference across the regenerator in a matter of seconds. It effectively validates the
concept of the acoustic feedback network and allows easy viewing of the parts in
a demonstration/lecture setting or in presentations to potential sponsors.
A limited thermoacoustic design effort went into the demonstration device un-
der the assumption that we wanted to “hurry up and get to our first mistake2 .”
The size of the speaker was chosen to fit inside commonly available PVC plumb-
ing parts and the resonator tube was cut from available stock. The rest of the
components were sized to fit these available parts.
The acoustical design that was done with this table top demonstration is in-
structive and provides an intuitive level of understanding about how the gas oscil-
lates in the various components and how energy flows within the device. It also
provides an appreciation of the number of adjustable parameters involved in the
design of a machine that uses this concept.

2.1 The Helmholtz Resonator as Acoustic Feed-
back Network
As shown in Fig. 2.1, the tabletop demonstration device can be thought of as
having three distinct parts: a loudspeaker, a λ/2 standing wave resonator and a
double-necked Helmholtz resonator at the end of the λ/2 resonator opposite the
loudspeaker. The loudspeaker is driven at the frequency f0 that corresponds to the
lowest longitudinal standing wave mode of the tube; since the tube is terminated
by a large impedance (a cap) the lowest mode supports nominally one-half of a
Referred to as “double-necked,” the Helmholtz resonator has two gas paths
into its compliant gas volume. One path, which is the inertial neck, has a diameter
much greater than a viscous penetration depth but substantially smaller than the
diameter of the λ/2 tube. This constriction forces the oscillating gas velocity
to increase in this region compared to the particle velocity in the adjacent λ/2
tube. The second neck contains the regenerator with a diameter of 1.75” made of
stainless steel woven cloth (Bekitherm NP 2503 ) with an average pore size that is a
J. C. Wheatley
Bekintex NV, Industriepark Kwatrecht, Neerhonderd 16, B-9230 Wetteren, Belgium

fraction of a viscous penetration depth. This fact makes the second neck primarily
an acoustically resistive element (Swift, 2002).
Although double-necked, the resonance frequency of the Helmholtz resonator
is predominantly determined by the size of the inertial neck and the size of the
volume behind the neck (if the resistance offered by the regenerator is large). This
frequency could be called the “Helmholtz frequency” and although it is not exactly
the damped natural frequency of this double-necked Helmholtz resonator, it is close
to the damped natural frequency defined as
r s
1 1 1 kgas
fH = = . (2.1.1)
2π LC 2π mgas

where L is the “inertance” of the hydrodynamic mass element and C is the compli-
ance of the gas “spring” formed by the trapped gas in the large open or “volume”
section of the resonator. In some situations it is more convenient to refer to the
stiffness of the gas which is simply the inverse of the compliance. These parameters
of a classic, single-necked Helmholtz resonator are defined as

ρm ln
L= (2.1.2)
C= (2.1.3)
ρm c2

where ρm is the mean density of the gas, c is the adiabatic sound speed in the gas,
ln is the length of the neck or the equivalent length of the part of the resonator
that contains gas whose response is dominantly inertial, and Vb is the volume of
gas whose response is dominantly compliant (or “springy”).
As with a classic, single-necked Helmholtz resonator driven near its resonance
frequency fH , the amplitude of the oscillating pressure inside of the “bulb” or
compliant element of the resonator (pH ) is greater than the pressure outside (p1 )
of the resonator by a ratio determined by the quality factor Q. The phase of the
pressure inside the compliant volume lags the phase of the pressure outside by 90◦
at resonance. (The quality factor, Q, of a resonant system is a non-dimensional
parameter that is proportional to the ratio of the energy stored in the reactive
components divided by the energy lost by the dissipative parts per cycle.) A system

with a high quality factor has a sharp resonance curve, while one with a more
moderate quality factor has a broader resonance curve. This well-known behavior
of a single degree-of-freedom oscillator is illustrated in Fig. 2.3 which shows the
pressure gain (ratio of the magnitude of the oscillating pressure inside the volume of
the Helmholtz resonator to the oscillating driving pressure) and phase difference of
the Helmholtz-resonator/feedback network in the tabletop demonstration device.
The network used in this tabletop regenerative thermoacoustic device has a
Q of around 10 (which is approximately equivalent to the pressure amplification
factor at resonance). Since the Helmholtz-resonator/feedback-network is driven at
a frequency much lower (in the tabletop device, the driving frequency is about 300
Hz) than its resonant frequency (about 600 Hz), the magnitude of the pressure os-
cillation “behind” the regenerator is greater than the pressure on the driver-side of
the regenerator by a factor of 1.3 (as seen in Fig. 2.3). Although the pressure gain
isn’t a factor of Q as it would be at resonance, by driving the feedback network well
below its natural resonance frequency, there is also no significant phase difference
between the pressure oscillation in front of the regenerator compared to the oscil-
lations behind it. This fact allows the pressure difference across the regenerator
to drive nearly in-phase gas flow through the regenerator. This phase relationship
is required for the gas to execute an efficient Stirling-like thermodynamic cycle
within the regenerator.
The drawing of Fig. 2.4 illustrates how the feedback network used in the demon-
stration device forces gas to flow through the regenerator in phase with the average
oscillating pressure (pH + p1 )/2 within the regenerator. Fig. 2.5 shows the step-
by-step process that a parcel of gas in the regenerator undergoes to complete a
modified Stirling cycle in order to pump heat against a temperature gradient.
Refer to the p-v diagram shown in Fig. 2.6 to follow the steps shown in Fig. 2.5
and to see how executing a Stirling cycle with a sinusoidal pressure wave is different
than the articulated cycle described in 1.3.
The following description of a parcel of gas as it traverses back and forth in-
side of the regenerator provides the reader with a Lagrangian description of how
the demonstration device moves heat from a cold place to a warm place. This
description will focus on the heat transfer between the regenerator and the parcel
of gas and the work done on or by that parcel without considering the effects of

16 R = 1000
15 R = 7000
14 R = 10000
Pressure gain (pH/p1)

5 Gain at driving frequency
250 300 350 400 450 500 550 600 650 700 750 800
Frequency (Hz)



R = 1000
Phase difference (degrees)

R = 7000
-60 R = 10000






250 300 350 400 450 500 550 600 650 700 750 800
Frequency (Hz)

Figure 2.3. This graph shows the magnitude of the pressure gain created in the
Helmholtz volume as a function of frequency as well as the phase difference between
the pressure inside the resonator compared to the phase of the driving pressure. The
R parameter represents the flow resistance of the neck of the double-necked Helmholtz
resonator that contains the regenerator material.







    % & ' ( 


" #$ 

Figure 2.4. This figure illustrates the gas motion in the feedback network in 1/8
increments of one acoustic cycle. Labelled on the Helmholtz resonator is the inertial
neck L, the compliance volume C and the resistive neck (the regenerator) labelled R. At
each step in the cycle, two pressure ”bar graphs” are located to the right of each drawing
that show the magnitude of the acoustic pressure outside of the resonator (p1 ) and the
acoustic pressure inside of the resonator (pH ) relative to the ambient pressure, at each
step of the acoustic cycle shown here (the first 3/4 of a cycle).

    ,     , 


) ) %&)+ + !"
%&'()* #$


 -%.+ , 

+ %&+ 

 2%.1 , 



, 2 %.1)  

345 /1

Figure 2.5. This companion figure to Fig. 2.4 shows how a net flow of heat is re-
alized from an oscillating flow of gas where the pressure and velocity of that flow are
substantially in phase. The letters that delineate various times in the acoustic cycle
correspond to the lettered states shown in Figs. 2.4–2.6. At the top of this figure, the
acoustic pressure is shown as a function of time both inside and outside of the Helmholtz




Figure 2.6. This graph of pressure vs volume shows a sinusoidally driven cycle as
an ellipse for the same pressure and volume oscillation magnitudes as the articulated
cycle shown in red. The shape of the cycle path approaches the isotherm lines as the
expansion and compressions become more isothermal. As the excitation becomes more
like a square-wave, the path becomes more parallel to the constant volume lines of the
articulated cycle. Notice that the enclosed area of the sinusoidally driven cycle is smaller
than the area enclosed by the articulated cycle. This indicates that driving the cycle
with a sinusoid sacrifices power density compared to a more square-wave like (articulated)
drive signal.

viscosity and imperfect thermal conduction. Energy entering the parcel is taken
to be positive. This means that work done on the parcel is also positive in sign.
Along the lines of the articulated cycle described in Section 1.3, the total back-
and-forth motion of the parcel over an acoustic period will be broken into four
traverses. However, the four points at which this break is chosen isn’t so clear when
the gas is driven sinusoidally. For that reason, two extra points are added to the p-
v diagram in Fig. 2.6 labelled a0 and e0 . These points represent the extremes of the
parcel’s volume fluctuation which do not occur exactly at the mean pressure (as it
would if there was no work being done on the gas). At the end of each traverse, the
parcel’s final state can be expressed as the superposition of a state change caused
by a temperature change imposed on the parcel and one caused by a pressure
change imposed on the parcel. For purposes of this explanation, the regenerator
is assumed to be constructed so that all of the gas within the regenerator is in
intimate thermal contact with the regenerator material (as expressed in Fig. 2.5
by the fact that the standoff distance of the parcel from the regenerator material
is nominally 1/10 the thickness of the thermal acoustic boundary layer δκ ) and
that the regenerator material has much greater accessible heat capacity than the
These are the desirable characteristics of a regenerator; to the extent that such a
thing can be fabricated, the parcel’s temperature as it moves along the regenerator
surface is prescribed by the temperature gradient along that surface. The flow of
heat between the parcel and the regenerator slab (motivated by the the parcel’s
movement to a regenerator location of a different temperature than the parcel)
can be reversible if as the parcel takes a differentially small step (for example from
[a] toward [c]) the temperature difference between the parcel and the regenerator
slab is also differentially small. As heat flows (in this case from the parcel to the
regenerator) that differentially small temperature difference is equalized, and then
the parcel takes another differentially-sized step toward [c]. This quasi-equilibrium,
reversible transfer of heat is assumed when describing the operation of a perfect
regenerator. The pressure variations experienced by the parcel are generated by the
loudspeaker and are denoted as p1 . The pressure inside the Helmholtz resonator,
pH , fluctuates in response to p1 .

a0 –c Cooling In Fig. 2.4 as the pressure p1 falls below ambient, gas rushes out of

the bulb through the inertial neck (labelled “L”) and pH falls always further
below ambient than p1 until the difference reaches a maximum at point [c]
(Fig. 2.4[c]). This pressure difference drives gas through the regenerator in
the opposite direction as the gas travelling through the inertial neck (sym-
bolized by the small arrow). Turning attention to Fig. 2.5, as a result of step
[g]–[a’] the parcel is at its smallest volume at [a’] and has been pre-cooled
to the coldest temperature of the regenerator. From [a’] to [c], an expan-
sion takes place and removes heat dQload + dQac from the regenerator. This
expansion of the parcel does work −dWa0 c on the wave.

c–e0 Pre-heat Continuing with Fig. 2.5, during the traverse from [c] to [e0 ] the
parcel must again increase in temperature, which causes heat dQce0 to flow
from the regenerator into the gas parcel. The acoustic wave is increasing in
pressure from its lowest value at [c] back to near-ambient. To undergo this
compression, the parcel requires work from acoustic wave dWce .

e0 –g Exhaust As seen in Fig. 2.4, the pressure across the regenerator has equal-
ized at point [e] (p1 = pH = pm ). This means that the parcel has stopped
its motion and is ready to change direction to move through the regenerator
away from the inside of the Helmholtz resonator. As gas rushes into the
Helmholtz volume through the inertial neck, the pressure pH becomes larger
than the pressure p1 . It is this pressure difference that drives the flow of gas
parcels through the regenerator. Looking back to Fig. 2.5 at point [e0 ], the
parcel occupies the largest volume that it ever will and it contains a sum
of heat energy that it acquired over its entire expansion traverse. However,
since the parcel is still a bit below mean pressure at [e0 ], the acoustic wave
must do some work on the parcel just to bring it to [e] at the mean pressure.
During this small compression, heat dQex is transferred to the slab. Then,
on the traverse from [e] to [g], the acoustic wave must do more work on the
parcel and heat dQe0 g is given to the slab to hold temporarily until the parcel
gets it back (dQac0 ) on the pre-heat step from [c]–[e0 ].

g–a0 Pre-cool During the traverse from [g] to [a0 ], the parcel must again decrease
in temperature and will deposit heat dQga0 to the slab for temporary storage
until the parcel picks it back up on the pre-heat traverse [c]–[e0 ] (dQce0 ). The

acoustic wave must do work on the parcel to compress it to the smallest
volume Va0 . However, at this volume, the pressure in the parcel is still a
little bit above ambient, and the relaxation that it does to get to ambient
removes an extra bit of heat dQload before the parcel finds itself at the mean
pressure again.

It should now be clear how the gas in the demonstration device is made to
have the correct phasing using the feedback network and how oscillating gas flow
with that correct phasing can move a net amount of heat from a cold temperature
to a warmer one. In an alpha-style Stirling machine parlance, the “compression”
space of the device is the volume of the Helmholtz resonator and the region outside
of the resonator is the “expansion” space. Consequently in the acoustic device,
the end of regenerator that faces into the bulb gets hot after some time passes
while the side of the regenerator that looks toward the outside of the Helmholtz
resonator gets cold. Of course, in a real device there are heat exchangers on either
side of the regenerator so that the heat removed from the cold side is replenished
by heat from the load and the heat rejected to the hot side is carried away to the
By looking at the elliptical path of the p-v diagram in Fig. 2.6 and comparing it
to the p-v diagram for the classic articulated Stirling cycle, the pre-cool/pre-heat
steps [g]–[a0 ]/[c]–[e0 ] can be seen to approach the constant volume displacement
[A]–[B]/[C]–[D] if the cycle were to approach the articulated cycle of Fig. 1.2.

2.2 The Circuit Analogy to Lumped Acoustic
An effective electrical analogy for quantitative design of thermoacoustic compo-
nents is the so-called “lumped element” circuit analysis. Brought into the field of
thermoacoustics formally with the publication of Swift’s textbook (Swift, 2002),
the lumped circuit analogy assigns common analog electrical circuit elements to
represent lumped acoustic elements. (A “lumped” acoustic element is much smaller
than the radian wavelength λ/2π.) Analogous to the voltage of an electrical cir-
cuit, the driving potential for the acoustic circuit is the acoustic pressure. This

choice forces the acoustic velocity (the volume velocity is most convenient, which
is the product of the acoustic particle velocity and the cross-sectional area of the
element) to be the flow variable analogous to electrical current in electrical cir-
cuits. For example, an acoustic gas spring like the one found in the open volume
of a Helmholtz resonator can be represented as a capacitor. As molecules of gas
rush into the cavity during one half of the acoustic cycle, pressure builds up and
this increase in pressure is 90◦ out of phase with the velocity. A constriction in a
resonator acts like an inductor that delays the velocity response by 90◦ compared
to the pressure difference across the section because of the momentum that the
gas gains speeding up through the resonator section. As seen in Fig. 2.7, the elec-
trical circuit analogy is applied to the classic single-necked Helmholtz resonator;
in Fig. 2.7(b) the circuit that represents the double-necked resonator used as the
feedback network is shown.
Fig. 2.7 also illustrates how the acoustic energy that exits the regenerator gets
recirculated through the inertance and back to the ambient temperature end of the
regenerator. Another way to look at this is that there is a lot of energy bouncing
back and forth between the inertance L and the compliance C. During each cycle
some fraction of that energy gets absorbed in the regenerator and moves some
heat from cold to hot (shown as TC and TH on in Fig. 2.7). Of course, this energy
distribution (a larger fraction of the total energy in the system being stored in L
and C) is only possible if the impedance of R is much larger than the impedance
of L. In fact, this gives the designer a good place to start — once a regenerator is
chosen, L must be an chosen so that R
is around 0.1.
This circuit analysis can be expanded to include a regenerator element that
not only dissipates power in proportion to the square of the volume velocity in the
regenerator, but also consumes the power required to move heat from cold to hot (in
the case of a refrigerator) or is an element that introduces oscillating velocity to the
circuit (in the case of a thermoacoustic prime mover). More will be presented about
regenerator design in Section 3.1.1. In the case of the demonstration device, the size
of the inertial and compliant elements were chosen so that the feedback network
would simply provide some overpressure behind the regenerator. Once these solely
geometry dependent parameters are known, the following simple expression gives
the overpressure ratio:


p C




p R  C

Figure 2.7. The upper circuit diagram (a) represents a classic single-necked pressure-
driven Helmholtz resonator while the lower diagram (b) represents the double-necked
Helmholtz resonator used as the feedback network.

pH 1
≈ ¡ f1 ¢ 2 . (2.2.1)
p1 1 − fH
Although convenient, this expression doesn’t take into consideration the usually
significant resistance of the regenerator nor does it reflect the compliance that the
non-occluded volume within the regenerator contributes to the system. But, Eq.
(2.2.1) does provide a simple first approximation of the sizes of the elements of
the feedback network required to generate a particular overpressure if f1 ¿ fH .
The magnitude of this overpressure is an important design consideration since it
is this pressure difference that drives the gas through the regenerator to move the
heat from the cold temperature to the exhaust temperature. The cooling capacity
of the device is proportional to the product of this velocity and the oscillating
pressure magnitude; however, the viscous drag force is also proportional to the
velocity through the regenerator. Hence, the rate of viscous energy dissipation is
proportional to the square of this flow rate.
In the demonstration device, this overpressure ratio can be easily adjusted
simply by changing the position of the regenerator relative the end of the λ/2 tube.
Since the compliance of the feedback network depends on the amount of volume
between the regenerator and the end of the λ/2 tube, changing the position of
the regenerator only changes the compliance, while leaving the inertial element
(the three “windows” cut into the plastic holder that suspends the regenerator in
the tube) unchanged. Several different overpressure ratios were explored with the
tabletop demonstration device.

2.3 Summary of Tabletop Demonstration Device
This demonstration device shows an easily measurable temperature difference as
large as 20◦ C at a cold side temperature of 10◦ C. However, since it has no means to
exhaust heat, the mean temperature continues to rise and steady state operation
isn’t attainable; a fact that makes detailed measurements difficult. For the first five
minutes of operation, no drift is noticeable since the exhaust heat is absorbed by
the heat capacity of the copper ring that holds the regenerator (the regenerator is
insulated from the copper ring by a plastic sleeve). Because this machine has such

a low cooling capacity (probably < 1W), and because there is no heat exchanger on
the cold side either, accurate measurement of the cooling capacity and consequently
device performance is impossible.
Since streaming (Gedeon, 1997) has been shown to be a problem (Backhaus &
Swift, 2000) when utilizing a feedback network, a 0.004” thick latex diaphragm was
installed over the cold side of the copper regenerator cartridge to enclose the cold
side of the regenerator and suppress the “Gedeon” streaming (Swift, 2002). Anec-
dotal evidence of it’s effectiveness was demonstrated by removing the diaphragm
to find that the cold side didn’t become cold at all.
The demonstration device was built to increase familiarity with the feedback
network concept, and to gain first-hand knowledge of what some of the important
design parameters are for such a device. Building a device takes much more time
than modelling it might take, and some disagree with that extra expenditure of
time. However, until something is built that works, mathematical models can give
a false sense of confidence and understanding. By constructing this demonstration
device and making some design mistakes along the way, real confidence was gained
as well as insight into what is important in the design of a device whose perfor-
mance can be measured. Of course, the additional pedagogical and marketing
opportunities provided by the existence of a visually transparent “working model”
has been additional compensation for this effort.
This chapter has laid out some of the fundamental concepts that will dominate
the rest of the thesis. The next chapter will describe a small refrigerator that has
an exhaust heat exchanger and electrical heat load so that some quantitative mea-
surements can be made. This so called ”pre-prototype” was created to serve as the
bridge between a theoretical understanding of a regenerator-based thermoacoustic
refrigerator and a working knowledge that would allow a confident and intelligent
design process for the Ben & Jerry’s prototype commercial ice-cream freezer that
is described in Chapter 4.
Chapter 3

Design and Characterization of the

The parameter space (i.e. regenerator diameter of about 1.75”, acoustic peak-to-
mean pressure ratio of about 5% and air as a working fluid at atmospheric pressure)
utilized in the table-top demonstration produced a measurable temperature dif-
ference and was intriguing enough to justify adding an exhaust heat exchanger to
the thermal core and instrumenting the system so that quantitative measurements
could be made. This chapter will detail some of the considerations given to the
more serious design of a thermoacoustic refrigerator that utilizes a feedback net-
work and a regenerator. It will also describe the atmospheric air machine that
was constructed, the measurement system surrounding it and the results of those
With the confidence gained from the demonstration device in both regenera-
tor and feedback performance, a novel approach for production of the required
acoustic pressure amplitudes was implemented (Poese et al., 2003). Instead of the
λ/2 standing wave resonator used in the table-top demonstration, a short (λ/60)
resonator that provides only compliance was used and the mass of the motor was
resonated against the compliance of the gas in the shortened resonator. The details
of this resonator, named “bellows bounce” are found in Sec. 3.1.4.

3.1 Design Considerations

3.1.1 Regenerator Design
Regenerators must have two somewhat conflicting thermal properties. First, it
must have enough heat capacity to anchor the temperature of the oscillating gas
to the regenerator temperature. Since only the gas within the thermal boundary
layer of the regenerator can be anchored to the regenerator temperature, the de-
signer must put enough solid material into the regenerator space so that all of the
gas is contained within a fraction of the working fluid’s thermal boundary layer.
However, the introduction of this much solid material into the high amplitude
acoustic wave causes viscous loss from the friction of the gas “scrubbing” along
the surface of the regenerator material. Secondly, it is important for the material
of the regenerator to be thermally non-conductive in order not to dissipate the
hot-to-cold temperature gradient across the regenerator.
In the textbook by Swift (2002), Equations (5.39) and (5.40) provide a good
quantitative and intuitive understanding of how to treat these separate effects.
Those equations are reproduced below and give approximate values for total power
Ḣ2 and acoustic power Ė2 flowing through a regenerator from the ambient tem-
perature end to the cold temperature end in the limit of rh ¿ δkappa and when the
velocity of the gas oscillations are in phase with the pressure oscillations.

17 ρm cp rh2 dTm ¡ ¢ dTm
Ḣ2 ≈ − 2
|U1 |2 − Agas kgas + Asolid ksolid (3.1.1)
35 Aω δκ dx dx
¡ Tload ¢ 3µ∆x 2
2 2
ω Arh ∆x
Ė2,in − Ė2,out = ∆Ė2 ≈ Ė2,in − 2
|U 1 | − |p1 |2 (3.1.2)
Tex − 1 2Arh 6kgas Tm

In these expressions, rh is the hydraulic radius of the pores in the screens of the re-
generator, kgas is the thermal conductivity of the working fluid, ksolid is the conduc-
tivity of the regenerator material, Tm is the zero-order, or mean gas temperature,
ω is the radian frequency of gas oscillation, ∆x is the length of the regenerator
and A is the cross-sectional area of the regenerator normal to the direction of gas
These approximations assume that pressure and velocity are in phase through-

out the length of the regenerator and that the regenerator is formed from a lay-up
of sheets of material of some thickness with regular gaps in between each sheet.
As mentioned before in section 1.3.3, the total power Ḣ2 in a perfect regenerator
is zero, or put another way, the heat being moved up the temperature gradient is
equal to the work required to move that heat. This statement is fundamental to
the Second Law of Thermodynamics; in the absence of any irreversible dissipation
(like imperfect heat transfer between the gas and the regenerator, or viscous losses
as the gas slides along the regenerator surface) the minimum amount of work
required to move an amount of heat depends only on the temperatures between
which the heat is moving. This is the first of three terms in (3.1.2), the sum of
which represents the change in acoustic power from the exhaust (hot) side of the
regenerator to the load (cold) side.
In a refrigerator, the first term is negative (in an engine, the regenerator is a
source of acoustic power, so the term is positive). In either case, the last two terms
always reduce the available acoustic power since the first of these represents the
power lost to the viscous “scrubbing” of the gas as it passes through the regenerator
(therefore proportional to the square of acoustic velocity) and the final term is the
irreversible thermal relaxation of the gas from the acoustic temperature oscillations
near a boundary (therefore proportional to the square of the acoustic pressure).
In a perfect regenerator, the amount of acoustic power that exits the regenerator,
Ė2 , is equivalent to the rate of heat removal from the load Q̇load . Losses of acoustic
power within the regenerator decrease Q̇load (or, for a constant Q̇load , the input
acoustic power must be larger than the Second Law minimum when the magnitude
of the last two terms of (3.1.2) are non-zero).
Total power, sometimes called enthalpy, is defined to be the sum of work energy
and heat energy at a location. In a regenerator, this sum is a constant and in a
perfect regenerator it is constantly zero. These facts mean that Eqs. (3.1.1)–(3.1.2)
are related in the following way:

Ḣ2 = Ė2,out + Q̇2,load (3.1.3)
Ḣ2 = Ė2,in + Q̇2,exhaust (3.1.4)

In (3.1.1) for total power, the first term is the most interesting since it represents

the non-ideality of the regenerator. The ratio δκ
is locally called the “Lautrec”
Number1 This ratio is useful to characterize a regenerator: rh is the hydraulic
radius of a regenerator pore and δκ is the thermal penetration depth. Since the
regenerator is only a thermodynamically reversible storage bank for energy during
an acoustic cycle if the temperature of a gas parcel can immediately assume the
temperature of the regenerator, it follows that the pore size should be small com-
pared to the distance that heat can diffuse in a 1/e fraction of an acoustic period.
In a perfect regenerator, Ḣ2 is zero and a vanishingly small Lautrec Number would
make the first term in (3.1.1) approach zero. Of course, as the Lautrec Number
gets small, viscous losses through the regenerator increase as seen in the second
term of (3.1.2)
The second term in (3.1.1) is simply the natural conduction of heat from the
exhaust (hot) side of the regenerator to the exhaust (cold) side through the metal
of the regenerator and through the gas in the regenerator. This conduction in-
creases Ḣ2 above zero since it reduces the net heat, Q̇2,load , moved from the load
temperature to the ambient, or hot temperature.
The two equations (3.1.1) and (3.1.2) provide a simple and insightful way to
evaluate some of the factors that must be considered in a regenerator design2 . For
example, the coefficient-of-performance (COP) for a refrigerator is defined as

COP = . (3.1.5)
Ė2,in − Ė2,out

Since Q̇load can be expressed in terms of Ḣ2 , ∆Ė2 and Ė2,in , a designer can use
these equations to get a sense of the maximum COP that the regenerator will allow.
For example, using these equations, the designer can get a sense of how small the
pore size in the regenerator should be. Although a small rh helps minimize Ḣ2 , it
increases the acoustic power dissipation in the regenerator as can be seen in the
second term of (3.1.2) where it appears in the denominator. These equations also
give the designer a sense of the desirable acoustic impedance of the regenerator,
or the ratio between acoustic pressure and acoustic velocity.
The Lautrec Number tells if a regenerator is “Toulouse.”
These equations leave out the performance degradation expected when the phase between
acoustic velocity and acoustic pressure is non-zero. In low-lift refrigerators, this degradation may
be small for phase differences as large as 30◦ .

In the long history of the design and construction of Stirling machines (both
engines and refrigerators), an accessible and simple three-way compromise between
low flow resistance, high heat capacity and low axial thermal conductivity has been
found with the use of a stacked pile of stainless steel screens. The stainless steel
is not highly thermally conductive, and the conductivity in the axial direction is
further reduced because of the large thermal contact resistance (point contact)
between each layer of woven screen. However, the stainless steel has adequate
specific heat capacity (relative to the gas) and does a good job of holding the
adjacent gas to the local regenerator temperature. Although the flow resistance is
not particularly low in a screen bed, the availability, low cost, and good thermal
properties make the stacked screen regenerator a convenient choice. Furthermore,
there are experimental correlations (Kays & London, 1998) for flow resistance and
convective heat transport properties as well as much experience with stacked screen
Equations (3.1.1) and (3.1.2) can’t be used directly for stacked screens since
they assume a parallel plate regenerator. To make calculations about regenerator
effectiveness for stacked screen beds, flow resistance correlations from Kays &
London (1998) implemented in the DeltaE software (Ward & Swift, 1994; Swift
& Ward, 1996) are used to decide which commercially available screen size to choose
for a regenerator. Detailed information about DeltaE software is available in the
User’s Manual (Ward & Swift, 1996). Later in this chapter, a segment-by-segment
annotation of the full DeltaE model that represents the pre-prototype is provided.
The DeltaE file reproduced below represents only the thermal core of a Stirling
machine: an exhaust (ambient temperature) heat exchanger, a regenerator and a
load (cold) heat exchanger. The program allows the user to quickly model the
performance of the commercially available screen sizes.

TITLE RETA Regenerator Design
!Created@18:54:38 04-Jan-02 with DeltaE Vers. 5.2b4 for the IBM/PC
!--------------------------------- 0 ---------------------------------
BEGIN the setup
9.7300E+04 a Mean P Pa 307.74 A T-beg G( 0c)
85.000 b Freq. Hz 4.8256E-03 B |U| G( 0f)

307.74 c T-beg K G 18.888 C Ph(U) G( 0g)
9000.0 d |p| Pa -14.837 D HeatIn G( 3e)
0.0000 e Ph(p) deg
4.8256E-03 f |U| m^3/s G
18.888 g Ph(U) deg G
air Gas type
ideal Solid type
!--------------------------------- 1 ---------------------------------
DUCT Dummy duct to get input power
1.0000 a Area m^2 9000.0 A |p| Pa
1.0000 b Perim m 0.0000 B Ph(p) deg
0.0000 c Length m 4.8256E-03 C |U| m^3/s
0.0000 d Srough 18.888 D Ph(U) deg
20.546 E Hdot W
sameas 0 Gas type 20.546 F Edot W
ideal Solid type 0.0000 G HeatIn W
!--------------------------------- 2 ---------------------------------
RPNTArg Peak Reynolds (A) gas p-p displacement (B) at HHX
0.0000 a Target (t) 1536.8 A RPNval
1.5366E-02 B RPNval
2 1C * 3a 3b * / # w / a<>b 4 * 3d * rho * mu /
!--------------------------------- 3 ---------------------------------
HX ambient heat exchanger
1.9600E-03 a Area m^2 9003.4 A |p| Pa
0.6000 b GasA/A -0.2676 B Ph(p) deg
sameas 2B c Length m 4.6003E-03 C |U| m^3/s
8.0000E-04 d y0 m 10.991 D Ph(U) deg
-14.837 e HeatIn W G 5.7083 E Hdot W
300.00 f Est-T K = 3H? 20.311 F Edot W
sameas 0 Gas type -14.837 G Heat W
copper Solid type 300.00 H MetalT K
!--------------------------------- 4 ---------------------------------
STKSCREEN the regenerator
sameas 3a a Area m^2 8440.0 A |p| Pa
0.8300 b VolPor -0.5650 B Ph(p) deg

2.3000E-02 c Length m 3.8290E-03 C |U| m^3/s
1.4400E-04 d r_H m -11.823 D Ph(U) deg
0.1500 e KsFrac 5.7083 E Hdot W
15.848 F Edot W
307.74 G T-beg K
sameas 0 Gas type 263.79 H T-end K
stainless Solid type -4.4631 I StkEdt W
!--------------------------------- 5 ---------------------------------
RPNTArg Back conduction in regen (watts)
0.0000 a Target (t) 1.3776 A RPNval
4G 4H - 4c / 4b k0 * 1 4b - ks 4e * * + 4a * *
!--------------------------------- 6 ---------------------------------
RPNTArg Resistance of regenerator
0.0000 a Target (t) 1.3366E+05 A RPNval
3A 4A - 3C 4C + 2 / /
!--------------------------------- 7 ---------------------------------
RPNTArg Peak Reynolds (A) gas p-p displacement (B) at CHX
0.0000 a Target (t) 1604.1 A RPNval
1.2193E-02 B RPNval
2 4C * 8a 8b * / # w / a<>b 4 * 8d * rho * mu /
!--------------------------------- 8 ---------------------------------
HX cold heat exchanger
sameas 3a a Area m^2 8427.2 A |p| Pa
sameas 3b b GasA/A -0.7609 B Ph(p) deg
sameas 7B c Length m 3.9287E-03 C |U| m^3/s
sameas 3d d y0 m -19.154 D Ph(U) deg
7.0000 e HeatIn W 15.708 E Hdot W
270.00 f Est-T K = 8H? 15.708 F Edot W
sameas 0 Gas type 10.000 G Heat W
ideal Solid type 270.00 H MetalT K
!--------------------------------- 9 ---------------------------------
RPNTArg Heat Pumping Power
10.000 a Target = 9A? 10.000 A RPNval
!--------------------------------- 10 ---------------------------------

RPNTARG keep phase real (on avg) in regenerator
0.0000 a Target =10A? 8.9407E-08 A RPNval
3B 4B + 2 / 3D 4D + 2 / -
!--------------------------------- 11 ---------------------------------
0.0000 a Target (t) 0.3444 A RPNval
2.0672 B RPNval
6.0021 C RPNval
4H 4G 4H - / sto 9A 1F 8F - / # rcl /

As noted above, the best screen size for a regenerator used in any design is
the one that minimizes the sum of three loss mechanisms: nuisance thermal con-
duction from the hot side to cold, viscous attenuation of the sound wave within
the regenerator matrix, and the extent to which the compression/expansion of the
gas is non-isothermal. Aside from screen size (wire diameter and wire ”pitch” or
spacing) several regenerator parameters need to be optimized simultaneously to
produce an acceptable design. The regenerator cross-sectional area affects power
density and the required pressure ratio. The acoustic resistance of the regenera-
tor affects the design of the phasing/feedback network and its length is inversely
proportional to thermal back-conduction loss for a fixed temperature span.
The regenerator used in the pre-prototype is shown in its holder in Fig. 3.1.
The screen was purchased from Universal Wire Cloth3 and is a square weave with
46 wires/inch. The wire size is 0.0045 inches in diameter. These are the parameters
usually used to specify screen from manufacturers, but for the purposes of calcu-
lating flow resistance, penetration depths and other thermoacoustic parameters,
volume porosity and hydraulic radius of the screen’s pores are more convenient.
The manufacturer’s specifications and the more convenient parameters are related
(Organ, 1992) with the following equations that map pitch m and wire diameter
dw into volume porosity φ and hydraulic radius rh .

1 − πmdw
φ= (3.1.6)
Universal Wire Cloth Company, 16 North Steel Road, Morrisville, PA 19067

Figure 3.1. This photograph shows the annular regenerator used in the pre-prototype,
which has a wire diameter of 0.0045 inches and is a square weave with 46 wires/inch. In
this photograph, the regenerator is shown in the process of stacking the screens in the
holder. The heat exchanger rests inside the plastic lip and against the wide rim of the
plastic holder. The brass bar that is blocking the inertance path is part of a jig used to
stack the screens and glue the diaphragm (not pictured).

dw ¡ φ ¢
rh = (3.1.7)
4 1−φ

The above “purchasing” parameters translate to a volume porosity of 83.4%
and a hydraulic radius of 144 µm. A photograph of the regenerator used in the
pre-prototype installed in its holder is shown in Fig. 3.1. The regenerator is made
from 88 layers of this screen and each screen layer is 2” in outer diameter with the
co-axial cutout for the inertance tube that is 0.75” in diameter.

3.1.2 Feedback Network
The design of machines such as the one described in this thesis involve a tangled
web of interrelated subsystems. One of the most important goals of this disserta-
tion is the creation of a design philosophy that can make the path to the design

of a practical machine more algorithmic. Since the parameters of the feedback
network dictate the pressure difference across the regenerator, the acoustic veloc-
ity through the regenerator is proportional to the resistance that the regenerator
offers. However, in the limit of a strongly non-resistive regenerator, the feedback
network won’t generate any pressure difference at all! Such an extreme might be
tempting for a designer trying to minimize viscous losses. A higher-order com-
plication is introduced by the fact that the pressure difference developed by the
feedback network also depends upon the resistance of the regenerator. These in-
terrelated factors make the design and optimization of the feedback network (i.e.
the size of the inertance and compliance of the Helmholtz resonator) a complicated
task that doesn’t have an easy closed-form solution but rather requires iteration.
While the simple DeltaE file shown in Section 3.1.1 is robust and provides
for a good design tool, trying to jump right into a DeltaE model of a feedback
network that includes a regenerator is complicated and convergence problems with
DeltaE usually do not permit the designer to develop a perspective on the avail-
able parameter space.
To accomplish a rough cut on the magnitudes of L, C, and to understand how
they interact with R a computer model was created in the process of this project
using LabVIEW4 software. A screenshot of the final version of the program is
shown in Fig. 3.2. This figure shows the program in its most complicated version,
taking into account such factors as compliance in the stacked-screen regenerator,
the Rott functions to approximate heat transfer coefficient within the regenerator
and the resistance in the inertance element. The real motivation for the creation
of this program was to allow the designer to input physical dimensions and prop-
erties of the inertance element (length, diameter), the compliance element (length,
diameter) and the regenerator (length, pore size, volume porosity), as well as oper-
ating frequency, static pressure and input acoustic pressure, and properties of the
working gas. The program quickly converts these physical parameters to equiva-
lent impedances of the various elements, total input impedance, velocities through
each element and the amount of over-pressure created by the network.
All of the inputs to the program are boxes that have small increment and
LabVIEW is made by National Instruments Corporation, 11500 N Mopac Expressway,
Austin, TX 78759-3504.

Figure 3.2. This screenshot shows the LabVIEW program created to efficiently explore
parameters for a design and gain intuition about the connections between parameters like
inertance, compliance and regenerator parameters. The program grew through several
versions; this is the final version with the most detail.

decrement arrows at the left side of each box. Most inputs are found on the left side
of the screen shown in Fig. 3.2. Boxes on the screen with no “adjustment arrows” in
them are the result of calculations made by the program. These outputs are simply
the result of many complex-number algebraic manipulations that would take the
designer a long time to complete by hand. Using this glorified calculator, results
of either a small or large perturbation can be instantly seen and evaluated. Of
course, this calculator doesn’t have nearly the accuracy of a program like DeltaE.
DeltaE integrates the wave equation with temperature gradients and adjusts
boundary conditions until a solution is found, but unlike DeltaE the lumped
element model doesn’t “get off the track” when the model parameters don’t make
physical sense.
The LabVIEW tool was created to prepare the designer to be able to build a
DeltaE model that had a reasonable probability of convergence – the purpose
of the LabVIEW tool is to reduce the algebraic tedium of trying to follow, for
example, a change to the length of the inertance element, all the way through to
its impact on the COP of the machine. The version of the calculator pictured
in Fig. 3.2 is the most complicated version created thus far, since it includes a
lumped model of a stacked-screen regenerator. In truth, this level of complexity
may be too high for the limited accuracy of a lumped regenerator model. DeltaE
models the regenerator in a more distributed way and in the spirit of using the
LabVIEW calculator as bridge to DeltaE, a simpler lumped regenerator model
(which was implemented in the version previous to the one shown) is probably all
that is justified.
The physical topology of the feedback network in the pre-prototype design
utilizes a annular regenerator with the inertance element formed by the co-axial
hole through the regenerator. Although compact, this is limits the minimum length
of the inertance tube to the length of the thermal core consisting of the regenerator
and heat exchangers. It also makes heat exchanger design more complicated since
it can add another interface that must be sealed against leakage of the working

3.1.3 Heat Exchanger Considerations
Because the pre-prototype was being constructed as a proof-of-concept tool, only
the exhaust heat exchanger was required to be fluid-backed so that heat could be
removed from the resonator and exhausted to the room. On the load side, an
electrical resistance heater was used to provide an easily measurable heat load for
the machine to move. Since others in the field (Backhaus & Swift, 2000) have had
good performance from shell-and-tube heat exchangers, it seemed like a reasonable
choice for use in the pre-prototype.
One of the perceived advantages to a shell-and-tube exchanger is the enhanced
convective heat transfer in the tubes through which the helium flows in response
to the acoustic fluctuations that could be turbulent (Hino et al., 1976). According
to conversations with Swift, the tubes can have a hydraulic radius as much as five
times the thermal penetration depth. Since a tube with a circular cross-section
has a hydraulic radius that is half of the physical radius, the tubes can have a
radius ten times the thermal penetration depth to take advantage of the enhanced
convection that comes with turbulence. (If the tube diameter is small, the flow will
need to have a higher velocity in order to transition to turbulence). A typical fluid-
backed parallel plate exchanger (Garrett et al., 1994) that we have made in the
past normally has a fin-to-fin spacing of about two or three thermal penetration
depths. From a fabrication standpoint then, shell-and-tube offered an easy-to-
fabricate design that would provide useful experience with a proven style. The
shell of the shell-and-tube exchanger also provides structural support for the tubes
and the manifold for the heat transfer liquid (water). Since the performance of
this machine was not critical to satisfy a sponsor, the pre-prototype provided a
good opportunity to test this new-to-us style of exchanger.
For the construction of this exchanger, 144 stainless steel tubes5 that are 1
cm in length, have an outer diameter of 1/8 inch and a wall thickness of only
0.004 inches are glued into the brass shell with Emerson and Cuming 1266 A/B
epoxy6 The corresponding hydraulic radius of these tubes is rh = A/Π where
A is the cross-sectional area and Π is the perimeter. For circular channels, this
Microgroup, Inc., 7 Industrial Parkway Road, Medway, MA 02053.
Emerson and Cuming, 46 Manning Road, Billerica MA.

Figure 3.3. This photograph depicts the shell-and-tube heat exchanger used in the
pre-prototype refrigerator.

is equivalent to half the physical radius; a tube with a 1/8 inch diameter has a
hydraulic radius of 0.031 inch or 0.8 mm. The thermal penetration depth in air at
atmospheric pressure is around 0.28 mm, so these tubes are right in the anecdotal
“zone” for turbulent enhancement of convective heat transport coefficient. The
completed exchanger and the parts are shown in Figs. 3.3 and 3.4.
This exchanger did not work well as evidenced by an unacceptably large tem-
perature difference between the metal and the oscillating air in the tubes. With
such a low power device, the gas flows were not fast enough to stimulate the gener-
ation of turbulent mixing, and consequently, the tubes were sized much too large
in diameter. If the flow is laminar, then the size of the tubes would have to be
fairly small, and the construction of the exchanger would have been difficult. The
exchanger worked for the pre-prototype experiment since it did move heat off of
the regenerator; however, it won’t be much use as designed for the prototype ma-
chine. Its use and flawed design in the pre-prototype certainly does not suggest
that it is a bad idea for use in thermoacoustics, just that more care must be given
to the sizing of the tubes than was exercised in the design of the pre-prototype.

Figure 3.4. This photograph depicts the parts used to build the shell-and-tube heat
exchanger used in the pre-prototype refrigerator.

3.1.4 The Acousto-Mechanical “Bellows-Bounce”
The pre-prototype provided a perfect opportunity to make a leap that has been
suggested by many for electrically driven thermoacoustic refrigerators: elimination
of the purely “acoustic” resonator. For a thermoacoustic engine, an acoustic res-
onator that has both a hydrodynamic mass (inertance) element and a gas spring
(compliance) is required for reactive energy storage7 An electrically-driven refrig-
erator also needs a way to store reactive power, and this was the motivation to
design stack-based machines usually with a 1/2 or 1/4 wavelength standing-wave
resonator that contained both a gas inertial element and a gas spring. This res-
onator has its own resonance frequency (as a lumped system, we could call it a
single-degree of freedom system like a mass-on-a-spring). The linear motor (or
loudspeaker) that is used to drive the such a system also has its own mass and
stiffness, so by itself it is a single degree of freedom oscillator. Of course, when
these two systems were coupled to power the refrigerator, a strongly coupled two-
Gas inertance and compliance are not necessary for operation, but some inertial element and
some compliant element are necessary.

degree of freedom system was the result and gas mixtures were used to tune the
acoustic resonator’s resonant response to overlap the motor’s resonant response.
However, this complication may have been unnecessary since it is possible to
use the moving mass attached to the motor as the only inertial element in the
system. The restoring force can either be a mechanical spring or the stiffness
of gas trapped in the part of the machine that houses the thermal core. This
hybrid acousto-mechanical resonator is the approach that was taken with the pre-
prototype in preparation for the design of the prototype for the B&J machine.
To seal the interface between the moving piston and the vessel that contains
the thermal core (the regenerator and heat exchangers) a metal bellows is used —
in fact, it’s a metal bellows that was designed for a different project, but happens
to be about the right size for the pre-prototype thermal core.
There are two factors influencing the size of the gas space in this acousto-
mechanical resonator. Since the pressure generated in the resonator is now only
a function of the ratio of the volume swept by the piston to the static volume of
the gas space, minimizing this space is desirable to minimize the stroke required
from the motor/bellows (or put another way, given a maximum stroke, minimizing
the gas space volume maximizes the pressure amplitude). The complex impedance
that the thermal core presents to the driver is a strong function of the size of the
gas volume in the acousto-mechanical resonator. So, minimization of this volume
may not be desirable if it makes the acoustic load presented to the motor out-
of-zone for the efficient power transmission to the resonator from the motor. A
cross-sectional diagram of the pre-prototype is shown in Fig. 3.5. A lumped circuit
representation of the system is shown in Fig. 3.6
The entire thermal core is located inside of the bellows volume (Poese et al.,
2003). This concept has been called “bellows-bounce” since the mass of the motor
bounces, or resonates against the gas contained inside of bellows. In the pre-
prototype, the thermal core was designed to use a conventional woofer and the
bellows was taken literally off the shelf and not designed for this specific applica-
tion. Consequently, some additional stiffness was required to increase the resonance
frequency to 80 Hz where the thermal core was designed to operate. This auxiliary
spring, or cylindrical spring (Garrett, 2003) is a novel idea that could prove to be
a replacement for the bellows if the gaps can be filled with a low-loss elastomer

Figure 3.5. This diagram shows the pre-prototype refrigerator with major parts la-
belled. The pressure vessel of the pre-prototype is about 12 inches tall and 8 inches in



p C  R  R  C

Figure 3.6. The lumped element circuit elements have been added to the diagram
shown in Fig. 2.7 to represent the acousto-mechanical resonator, which can be simply
represented as a compliance and parallel resistance.

of some type. The cylindrical spring is shown in Fig. 3.7 where it surrounds the
bellows without contacting it.

3.1.5 Matching the Linear Motor to the Resonator Load
Another important factor in the design of the thermal core and resonator is the
impedance that these components present to the driver. As both Wakeland (2000)
and Smith (Dec. 2001) have clearly demonstrated, presenting a linear motor with
the correct impedance is paramount to attaining the maximum electro-acoustic
efficiency of a particular motor. For any linear motor, this efficiency depends only
on the transduction coefficient (Bl product), the real (or dissipative) part of the
electrical impedance of the motor measured at the electrical terminals Re and the
mechanical resistance of the motor characterized by Rm . The optimum load to
present to the driver is <[Zload ] = Rload = Rm .
As can be seen when analyzing the circuit shown in Fig. 3.6, the compliance of
the gas in the resonator (the element labelled Cres ) has a big effect on the value of

Figure 3.7. This photograph shows the cylindrical spring used to augment the stiffness
provided by the gas trapped inside of the bellows. The spring is about 6 inches in
diameter and 4 inches tall. It is positioned so that the bellows just fits inside of it
without contact.

Rload . Rres is made up of thermal relaxation losses on the surface of the resonator
(bellows) wall and while it represents a large power loss mechanism, the volume of
the gas enclosed in the bellows represents a bigger lever to get Rload to match Rm .
Since the thermal core in the pre-prototype was designed for a moving coil, off-the-
shelf 5 1/4” woofer, adjusting the volume of the space enclosed by the bellows was
the way that the new moving magnet linear motor was matched to the existing
thermal core. Of course, since the stiffness of the gas enclosed in the bellows also
provides the restoring force for the driver’s moving mass, adjusting this volume
also sets the resonance frequency of the machine.
In the pre-prototype, most of the restoring force for the moving mass is provided
by the cylindrical spring; the stiffness of the spring was designed so that the system
is mechanically resonant at the intended frequency of operation, f = ω/2π =
k/m, where k is the combination of the gas stiffness, cylindrical spring stiffness
and driver suspension stiffness (1/k = 1/kgas + 1/kspring + 1/kdriver ) and m is the
sum of the driver’s moving mass, the piston mass and the moving mass of the
bellows. When driven at this resonance frequency, the maximum electro-acoustic
transduction efficiency ηmax , is defined after Wakeland (2000) in terms of two
dimensionless parameters β and σ:

ηmax = (3.1.8)
√ (Bl)2
where σ ≡ β + 1 and β ≡ Re Rm
These results can be derived for drivers with large β (≥ 5) by adjusting the load
to make the average power dissipated by Joule heating of the voice coil carrying
a peak current i (Πel = Re i2 /2) equal to the average power dissipated in the me-
chanical resistance of the driver moving with peak velocity v1 (Πmech = Rm v12 /2).
Wakeland’s analysis assumes that the Bl product is independent of the driver’s
amplitude of motion, x1 = v1 /ω. Although that is not true for the high-power,
high-efficiency moving-magnet electrodynamic motors that have been studied at
the Penn State Applied Research Laboratory (Smith, Dec. 2001; Smith et al., 1997;
Heake, Dec. 2001), for some motors it has been shown that the variation in Bl with
piston position has a small effect on the driver’s output power. This is because Bl
is largest around the equilibrium position, where the piston velocity is also largest.

3.1.6 Detailed DeltaE Model
In order to verify the understanding obtained from the lumped element model de-
tailed above, a DeltaE model was created that includes every aspect of the design
except for the driver. The program DeltaE is an acronym for “Design Environ-
ment for Low-Amplitude ThermoAcoustic Engines” and is a computer model using
a one-dimensional differential equation solver. A detailed description of the model
is available as a Users’ Manual for the program (Ward & Swift, 1996), currently
(Feb. 2004) included on a CD-ROM along with the program in Swift’s thermo-
acoustics textbook (Swift, 2002). For design purposes, DeltaE is particularly
useful because its execution is extremely fast. This high speed allows the designer
to examine many design variations in rapid succession and compare their relative
One product of the DeltaE model is an “output file.” The DeltaE output file
for our nominal design conditions (air at 1 atmosphere with acoustic peak-to-mean
pressure ratio, p1 /pm ∼ 6.5%) is included in this section. The output file provides
a listing of the segments that constitute the acoustical and thermal elements of
the model. The left column of each segment has the physical input parameters
required to define that segment. The type and number of those parameters vary
with the type of segment (e.g., duct, heat exchanger, regenerator, branch, target).
The right column of each segment contains output parameters calculated by the
The “solver” requires an equal number of “targets” and “guesses” to calculate
a solution to the model. The program obtains the closest possible match between
the “target” parameters and their targeted values by varying the parameters that
are chosen as the “guesses.” The goodness of the solution is measured by the
magnitude of the error vector constructed from the sum of squares of the differences
between the calculated target parameters and their targeted values. Much of the
“art” of thermoacoustic design using DeltaE is in the determination of what
parameters should be targets and what parameters should be guesses.
Among designers of thermoacoustic devices, the DeltaE file is the “language”
that provides a quantitative description of a device. A segment-by-segment nar-
rative that describes our chiller design follows the output file. One difficulty in
discussing the DeltaE file of a regenerator-based chiller with an acoustic feed-



Load HX Regen Exhaust HX

Figure 3.8. Presented above is a schematic diagram for the computational flow of the
DeltaE model. Global parameters such as mean pressure pm and frequency f are set in
the BEGIN segment. The program reaches the TBRANch segment and calculates through
the feedback network segments. That network branch ends at the SOFTEnd segment and
the computation resumes at the second BEGIN segment and computes through the load
(cold) heat exchanger, the STKSCREEN regenerator and the exhaust (hot) heat exchanger
until it reaches the UNION segment. At that point the program checks that the continuity
of complex pressure and volume velocity is preserved or changes some “guess” parameters
to improve the agreement. The final segment is the HARDEnd that ensures no power leaves
the end of the model.

back network is that the physical design has two parallel flow paths: one through
the regenerator and one through the feedback network produced by the Helmholtz
resonator. This configuration is shown schematically in Fig. 3.8. The DeltaE
file consists of a serial (not parallel) sequence of segments. This lack of topolog-
ical equivalence is corrected by employing a TBRANCh segment (#7) and a UNION
segment (#29).
The program receives its initial parameters in a BEGIN segment (#0). Next,
the program calculates the acoustic field through the volume of gas outside of the
bellows in the pressure vessel. An ENDCAp is placed in the model only to calculate
the thermal relaxation dissipation on the face of the power piston. After the
segment that models the bellows segment, the program starts calculating through
the feedback network after passing through a TBRANCh segment (#7). When it
reaches the SOFTEnd segment (#17), it starts again at a new BEGIN segment (#20)
and calculates it way through the cold heat exchanger, regenerator and hot heat
exchanger, ending at a UNION segment (#29) and a HARDEnd segment (#30).
The program enforces the acoustic pressure to be the same at the junction and

union and that volume velocity is conserved at the junction. This can be seen
by a check of output parameters 7A for the magnitude of acoustic pressure at
the junction, and 18A for the magnitude of pressure at the start of thermal core
branch. (You will find them to be equal - and since the phase of pressure at these
locations are the same, comparing magnitudes is all that is required.) Likewise,
check 6C and 6D for the complex volume velocity entering the junction and then
add the two complex quantities found in 7C/7D and 18C/18D to verify that the
sum is equivalent to what is found in 6C/6D. Following the list of the output file, a
segment-by-segment description will discuss some of the finer points of the model.

TITLE Complete Pre-Prototype, as built
!Created@22:41:09 02-Dec-02 with DeltaE Vers. 5.3b5 for the IBM/PC
!--------------------------------- 0 ---------------------------------
BEGIN Line ’em up
9.7300E+04 a Mean P Pa 1.8273E-02 A |U| G( 0f)
68.700 b Freq. Hz 88.202 B Ph(U) G( 0g)
sameas 35a c T-beg K 2.4411E+05 C Re(Zb) G( 7a)
6275.0 d |p| Pa -1.2929E+06 D Im(Zb) G( 7b)
0.0000 e Ph(p) deg 5.7310E-02 E Length G( 15c)
1.8273E-02 f |U| m^3/s G 277.29 F T-beg G( 20c)
88.202 g Ph(U) deg G 1.1935 G HeatIn G( 25e)
air Gas type
ideal Solid type
!--------------------------------- 1 ---------------------------------
CONDUCT Insulate resonator stub
!--------------------------------- 2 ---------------------------------
RPNTArg Driver stroke
0.0000 a Target (t) 6.9386 A RPNval
U1 mag 5a / w / 2 * 1000 *
!--------------------------------- 3 ---------------------------------
RPNTArg Input acoustic impedance
0.0000 a Target (t) 3.4340E+05 A RPNval
-88.202 B RPNval
( 1.0772E+04,-3.4324E+05) C

p1 U1 / # # arg a<>b mag
!--------------------------------- 4 ---------------------------------
DUCT Zero length duct to get input power
1.0000 a Area m^2 6275.0 A |p| Pa
1.0000 b Perim m 0.0000 B Ph(p) deg
0.0000 c Length m 1.8273E-02 C |U| m^3/s
0.0000 d Srough 88.202 D Ph(U) deg
1.7984 E Hdot W
sameas 0 Gas type 1.7984 F Edot W
ideal Solid type 0.0000 G HeatIn W
!--------------------------------- 5 ---------------------------------
ENDCAp Pusher Cone
1.2200E-02 a Area m^2 6275.0 A |p| Pa
0.0000 B Ph(p) deg
1.8272E-02 C |U| m^3/s
88.252 D Ph(U) deg
1.7484 E Hdot W
sameas 0 Gas type 1.7484 F Edot W
ideal Solid type -4.9958E-02 G HeatIn W
!--------------------------------- 6 ---------------------------------
DUCT Real estate within bellows
1.2000E-02 a Area m^2 6299.2 A |p| Pa
4.3000 b Perim m -1.8704E-02 B Ph(p) deg
4.4600E-02 c Length m 7.3541E-03 C |U| m^3/s
0.0000 d Srough 87.570 D Ph(U) deg
0.9746 E Hdot W
sameas 0 Gas type 0.9746 F Edot W
ideal Solid type -0.7738 G HeatIn W
!--------------------------------- 7 ---------------------------------
TBRANCh Junction of feedback network
2.4411E+05 a Re(Zb) Pa-s/m^3 G 6299.2 A |p| Pa
-1.2929E+06 b Im(Zb) Pa-s/m^3 G -1.8704E-02 B Ph(p) deg
4.7878E-03 C |U| m^3/s
79.289 D Ph(U) deg
2.7978 E Hdot W

air Gas type 2.7978 F Edot W
ideal Solid type -1.8233 G Edot_T W
!--------------------------------- 8 ---------------------------------
RPNTArg Capture P1 and U1
0.0000 a Target (t) ( 8.8985E-04, 4.7044E-03) A
( 6299.2 , -2.0563 ) B
P1 U1
!--------------------------------- 9 ---------------------------------
RPNTArg Calculate minor loss resistor
0.6500 a Target (t) 1.5465E+04 A RPNval
0.424 rho * 7C * 11a SQRD / 9a *
!--------------------------------- 10 ---------------------------------
IMPEDance Minor loss resistor
sameas 9A a Re(Zs) Pa-s/m^3 6285.9 A |p| Pa
0.0000 b Im(Zs) Pa-s/m^3 -0.6819 B Ph(p) deg
4.7878E-03 C |U| m^3/s
79.289 D Ph(U) deg
2.6206 E Hdot W
sameas 0 Gas type 2.6206 F Edot W
ideal Solid type -0.1773 G HeatIn W
!--------------------------------- 11 ---------------------------------
DUCT Feedback Inertance
3.1000E-04 a Area m^2 6580.5 A |p| Pa
6.2830E-02 b Perim m -1.2230 B Ph(p) deg
4.0000E-02 c Length m 4.5349E-03 C |U| m^3/s
0.0000 d Srough 78.782 D Ph(U) deg
2.5896 E Hdot W
sameas 0 Gas type 2.5896 F Edot W
ideal Solid type -3.0988E-02 G HeatIn W
!--------------------------------- 12 ---------------------------------
RPNTArg Calculate lumped inertance -- Z_l (A) and L (B)
0.0000 a Target (t) 6.6450E+04 A RPNval
153.94 B RPNval
( 8.8220E-04, 4.4482E-03) C
U1 P1 8B - U1 / IMAG W / ~ # W *

!--------------------------------- 13 ---------------------------------
RPNTArg Calculate minor loss resistor
0.6500 a Target (t) 1.4649E+04 A RPNval
0.424 rho * 11C * 11a SQRD / 13a *
!--------------------------------- 14 ---------------------------------
IMPEDance Minor loss resistor
sameas 13A a Re(Zs) Pa-s/m^3 6569.3 A |p| Pa
0.0000 b Im(Zs) Pa-s/m^3 -1.7936 B Ph(p) deg
4.5349E-03 C |U| m^3/s
78.782 D Ph(U) deg
2.4390 E Hdot W
sameas 0 Gas type 2.4390 F Edot W
ideal Solid type -0.1506 G HeatIn W
!--------------------------------- 15 ---------------------------------
DUCT Feedback Compliance
4.9010E-03 a Area m^2 6578.1 A |p| Pa
0.2481 b Perim m -1.8306 B Ph(p) deg
5.7310E-02 c Length m G 1.5740E-03 C |U| m^3/s
0.0000 d Srough -64.525 D Ph(U) deg
2.3749 E Hdot W
sameas 0 Gas type 2.3749 F Edot W
ideal Solid type -6.4069E-02 G HeatIn W
!--------------------------------- 16 ---------------------------------
RPNTArg Calculate f_H (A) and lumped compliance -- Z_c (B) and C (C)
0.0000 a Target (t) 282.06 A RPNval
1.1201E+06 B RPNval
2.0683E-09 C RPNval
P1 12C U1 - / IMAG W * INV ~ # # W * INV a<>b 12B * INV SQRT 2 pi * /
!--------------------------------- 17 ---------------------------------
SOFTEnd End of feedback leg at hot side of regenerator
0.0000 a Re(z) (t) 6578.1 A |p| Pa
0.0000 b Im(z) (t) -1.8306 B Ph(p) deg
1.5740E-03 C |U| m^3/s
-64.525 D Ph(U) deg
2.3749 E Hdot W

2.3749 F Edot W
23.988 G Re(z)
sameas 0 Gas type 46.463 H Im(z)
ideal Solid type 300.95 I T K
!--------------------------------- 18 ---------------------------------
DUCT Zero length duct to get acoustic variables
1.0000 a Area m^2 6299.2 A |p| Pa
1.0000 b Perim m -1.8704E-02 B Ph(p) deg
0.0000 c Length m 2.7056E-03 C |U| m^3/s
102.34 D Ph(U) deg
-1.8233 E Hdot W
sameas 0 Gas type -1.8233 F Edot W
ideal Solid type 6.6613E-16 G HeatIn W
!--------------------------------- 19 ---------------------------------
RPNTArg Total Energy Target
sameas 18F a Target (t) -1.8233 A RPNval
!--------------------------------- 20 ---------------------------------
BEGIN Reset temperature so feedback loop isn’t cold
sameas 0a a Mean P Pa
sameas 0b b Freq. Hz
277.29 c T-beg K G
sameas 18A d |p| Pa
sameas 18B e Ph(p) deg
sameas 18C f |U| m^3/s
sameas 18D g Ph(U) deg
sameas 0 Gas type
ideal Solid type
!--------------------------------- 21 ---------------------------------
CONDUCT Insulate thermal core
!--------------------------------- 22 ---------------------------------
IMPEDance Diaphragm impedance
0.0000 a Re(Zs) Pa-s/m^3 6461.9 A |p| Pa
6.1500E+04 b Im(Zs) Pa-s/m^3 0.2970 B Ph(p) deg
2.7056E-03 C |U| m^3/s

102.34 D Ph(U) deg
-1.8233 E Hdot W
sameas 0 Gas type -1.8233 F Edot W
ideal Solid type 0.0000 G HeatIn W
!--------------------------------- 23 ---------------------------------
DUCT Diaphragm volume
1.7060E-03 a Area m^2 6469.7 A |p| Pa
0.2230 b Perim m 0.3115 B Ph(p) deg
1.0000E-02 c Length m 2.3628E-03 C |U| m^3/s
0.0000 d Srough 104.18 D Ph(U) deg
-1.8324 E Hdot W
sameas 0 Gas type -1.8324 F Edot W
ideal Solid type -9.1446E-03 G HeatIn W
!--------------------------------- 24 ---------------------------------
RPNTArg Peak Reynolds (A), gas p-p displacement (B) at CHX
0.0000 a Target (t) 893.18 A RPNval
1.2247E-02 B RPNval
2 18C * 25a 25b * / # w / a<>b 4 * 25d * rho * mu /
!--------------------------------- 25 ---------------------------------
HX cold heat exchanger
1.7060E-03 a Area m^2 6471.2 A |p| Pa
0.6000 b GasA/A 0.3112 B Ph(p) deg
1.0000E-03 c Length m 2.3410E-03 C |U| m^3/s
6.0000E-04 d y0 m 104.37 D Ph(U) deg
1.1935 e HeatIn W G -0.6389 E Hdot W
270.00 f Est-T K (t) -1.8401 F Edot W
sameas 0 Gas type 1.1935 G Heat W
ideal Solid type 285.98 H MetalT K
!--------------------------------- 26 ---------------------------------
STKSCREEN the regenerator
sameas 25a a Area m^2 6568.1 A |p| Pa
0.8300 b VolPor -1.8532 B Ph(p) deg
2.0000E-02 c Length m 1.7589E-03 C |U| m^3/s
1.4400E-04 d r_H m 111.89 D Ph(U) deg
0.1500 e KsFrac -0.6389 E Hdot W

-2.3258 F Edot W
277.29 G T-beg K
air Gas type 300.95 H T-end K
stainless Solid type -0.4857 I StkEdt W
!--------------------------------- 27 ---------------------------------
RPNTArg Peak Reynolds (A), gas p-p displacement (B) at HHX
5.6200 a Target (t) 1449.0 A RPNval
8.6853E-03 B RPNval
2 26C * 28a 28b * / # w / a<>b 4 * 28d * rho * mu /
!--------------------------------- 28 ---------------------------------
TX ambient heat exchanger
sameas 25a a Area m^2 6578.1 A |p| Pa
0.5500 b GasA/A -1.8306 B Ph(p) deg
1.0000E-02 c Length m 1.5740E-03 C |U| m^3/s
1.5875E-03 d radius m 115.48 D Ph(U) deg
-18.487 e HeatIn W -2.3749 E Hdot W
300.00 f Est-T K (t) -2.3749 F Edot W
sameas 0 Gas type -1.7360 G Heat W
copper Solid type 298.59 H MetalT K
!--------------------------------- 29 ---------------------------------
UNION Join the regen leg with the feedback leg
17.000 a TendSg 6578.1 A |p| Pa
6578.1 b |p|End Pa = 29A? -1.8306 B Ph(p) deg
-1.8306 c Ph(p)E deg = 29B? 5.9019E-12 C |U| m^3/s
-10.907 D Ph(U) deg
1.9169E-08 E Hdot W
air Gas type 1.9169E-08 F Edot W
ideal Solid type 300.95 G End-T K
!--------------------------------- 30 ---------------------------------
HARDEnd End o’ the line
0.0000 a R(1/z) = 30G? 6578.1 A |p| Pa
0.0000 b I(1/z) = 30H? -1.8306 B Ph(p) deg
5.9019E-12 C |U| m^3/s
-10.907 D Ph(U) deg
1.9169E-08 E Hdot W

1.9169E-08 F Edot W
2.0342E-10 G R(1/z)
sameas 0 Gas type -3.2498E-11 H I(1/z)
stainless Solid type 300.95 I T K
!--------------------------------- 31 ---------------------------------
RPNTArg Nominal regenerator resistance
0.0000 a Target (t) 4.7311E+04 A RPNval
26A 25A - 26C 25C + 2 / /
!--------------------------------- 32 ---------------------------------
RPNTArg Back conduction in regenerator (W)
0.0000 a Target (t) 0.7855 A RPNval
26H 26G - 26c / 26b k0 * 1 26b - ks 26e * * + 26a * *
!--------------------------------- 33 ---------------------------------
RPNTArg Peak Reynolds (A) gas displacement (B) in regenerator
0.0000 a Target (t) 50.756 A RPNval
3.3538E-03 B RPNval
26C 25C + 2 / 26a 26b * / # w / a<>b 4 * 26d * rho * mu /
!--------------------------------- 34 ---------------------------------
RPNTArg Peak Reynolds (A) gas displacement (B) at HHX
0.0000 a Target (t) 648.37 A RPNval
3.8862E-03 B RPNval
28C 28a 28b * / # w / a<>b 4 * 28d * rho * mu /
!--------------------------------- 35 ---------------------------------
RPNTArg Regen Exhaust Side Temp
300.95 a Target = 35A? 300.95 A RPNval
!--------------------------------- 36 ---------------------------------
RPNTArg Regen Load Side Temp
277.29 a Target = 36A? 277.29 A RPNval
!--------------------------------- 37 ---------------------------------
RPNTArg Helmholtz Overpressure
4.8300 a Target = 37A? 4.8300 A RPNval
17A 0d / 1 - 100 *
!--------------------------------- 38 ---------------------------------

0.0000 a Target (t) 5.6626E-02 A RPNval
0.6636 B RPNval
11.720 C RPNval
26G 26H 26G - / sto 25G 4F / # rcl /

The output file corresponds to the nominal operating point: atmospheric pres-
sure air with a peak-to-mean pressure ratio, p1 /pm = 6.45%. Now that the general
computational flow of the program has been described, it is appropriate to examine
each segment.

Segment 0 The BEGIN statement initializes the acoustic field variables in the
model. In this case the mean pressure (0a) has been set to 97.3 kPa (atmo-
spheric pressure in the lab) and the gas set to air; the ambient temperature
set to mirror the exhaust exchanger temperature at 35a using the “sameas
35a” input in (0c). The acoustic pressure amplitude is set to about 6.45% of
the mean pressure, which is about what the motor could create at its stroke
limit. Since a motor isn’t included in the model, the volume velocity that
drives the gas in the model is allowed to be varied by DeltaE by designating
it’s magnitude (0f) and phase (0g) as GUESSes. Since the pressure magni-
tude and phase are fixed, the magnitude and phase of volume velocity must
be allowed to be varied by the model since the real and imaginary parts of
the input impedance vary with other parameters fixed by the modeler, like
temperatures, regenerator size, etc. In this way, the BEGIN segment becomes
the only source for acoustic power in the model. The right column of the
BEGIN segment lists the values of each GUESS parameter (along with their
location in the program) for the solution.

Segment 1 This segment allows the external surfaces of the model to be either
insulated, so that the only segments where heat can enter or exit the model
is at heat exchanger segments, or non-insulated, which allows the energy
dissipated each segment to be removed at that segment and not show up on
the heat exchangers. By entering the word CONDUCT for this segment, the
model is set to be non-insulating; for some unknown reason the INSULATE
function is problematic with this branched model.

Segment 2 This segment is labelled RPNTARg. An RPNTARget is a segment that
can be inserted to perform calculations and the results of that calculation
can be used in other segments or can be chosen as a TARGET. The calcu-
lation is specified by the input parameter (left side) 1b (which is an implicit
designation) and is coded in Reverse Polish Notation (RPN), brought to
prominence by Hewlett-Packard in their line of scientific calculators. It may
seem strange at first, but RPN is an efficient and logical way of organizing
algebra for a computer. In this segment we see U1 mag 5a / w / 2 * 1000
* indicating that the model value for the complex volume velocity U1 is re-
duced to a magnitude and divided by the the value found in 5a which is
the area of the piston or pusher cone. The result is divided by the model
value for angular frequency w and then multiplied by 2 and again by 1000
to produce the stroke in millimeters that a driver with a effective surface
area in 5a would produce. (A serious student of DeltaE would now try
to convert all of the calculations in the RPN Targets in this output file to
algebraic expressions using Tables VI.2 - VI.4 in the DeltaE User’s Manual
(Ward & Swift, 1996).)

Segment 3 Segment 3 is another RPNTARg that calculates the acoustic impedance
presented to the un-modelled driver and returns the magnitude, phase and
the complex value in rectangular x + iy form. Following along on line 3b,
first the acoustic impedance is calculated (p1 U1 /). The “#” symbol copies
this result to the next line of the segment’s stack (done twice in this case) and
then the arg command finds the phase angle of that complex number, swaps
its position in the stack with the value above it with the a<>b command and
then finds the magnitude of that value. All results of the RPNTARg calculation
are displayed in the output column and the 3a parameter is reserved for use
as a possible target (not used in this segment).

Segment 4 This “dummy” DUCT segment has zero length and is only in the model
to conveniently display the input power in 4E.

Segment 5 The ENDCAp segment models the thermal relaxation loss on the face
of the piston that seals the bottom end of the bellows and attaches to the
motor armature.

Segment 6 The next DUCT represents the bellows volume and surface area and
calculates the thermoviscous loss on that surface area. The dimensions in
this segment are not physical, except for bellows effective cross sectional
area. The length is calculated to reflect the amount of available gas volume
in the bellows, and the perimeter is calculated to reflect the actual amount
of surface area of the bellows convolutions and the surface area of the volume
excluder mounted to the bottom of the thermal core to increase the pressure
ratio that the motor could produce for a given stroke.

Segment 7 The TBRANch segment represents the fork in the road for energy in
this model. The energy that leaves the bellows can either flow into the
inertance or into the cold side of the regenerator. The impedance of the
branch is set by 7a and 7b. These are enabled as GUESSes so that DeltaE
can choose the correct impedance based on the geometry of the inertance
and compliance. The phasing of volume velocity is such that energy flows
into the inertance segment, into the regenerator at the ambient side and out
of the regenerator at the cold side. This is the desired phasing and is the
function of the network. Again, it is the size of the acoustic mass in the
inertance element and the Helmholtz compliance volume that enforce this
phasing, previously done with cranks, swash-plates, etc. in Stirling machines
of the past.

Segment 8 This RPNTArg segment just puts complex acoustic pressure and vol-
ume velocity at this point in the model into 8A and 8B so that it can be
easily referenced in calculations made later in the model (see Segment 12).

Segments 9, 13 and 14 Flow entering and leaving through abrupt changes in
cross-sectional area often exhibit reduced (irreversible) pressure recovery.
This means that some of the energy in the flow is carried away by vor-
tices that spin away from the main flow direction and get turned into heat
by the shear viscosity of the fluid. This is referred to as “minor loss” in
piping networks (Idelchik, 1996), and this term has been adopted by ther-
moacousticians (Wakeland, 2002), even though the magnitude of the loss can
be rather major. This RPNTArg computes the equivalent acoustic resistance
of the abrupt transition due to small constriction in the cold head, between

the bellows space and the inertial element; a factor of two could take care of
both the inlet and the outlet of the inertance tube, but for sake of clarity,
segments 13 and 14 take care of the minor loss at the outlet. This resistance
is scaled by the velocity through the tube, making this energy loss cubic
with respect to flow velocity (normal thermoviscous energy loss is quadratic
in velocity).

Segment 10 This IMPEDance segment applies the resistance calculated above to
the acoustic flow. Check 10G to see how much power is dissipated in this
’resistor’ and removed from the model.

Segment 11 This DUCT models the inertial element, or acoustic mass for the
feedback network. Notice the value in 11G that is labelled HeatIn, is a
negative number, which means that the heat is leaving the model at this
segment. This heat flow is a consequence of the attenuation of the acoustic
wave passing through the inertance tube due to thermoviscous loss on the
surface of the tube. The energy that is taken from the acoustic wave in this
segment and turned into heat leaves the model at this point. The attenuation
of the acoustic wave can be easily verified by checking the value of acoustic
power going into the TBRANch segment (7F) and subtracting the amount
of acoustic power Edot that leaves the DUCT segment (11F). The difference
should be the same as the sum of 10G + 11G.

Segment 12 The RPNTArg of segment 12 calculates the impedance presented by
the inertance tube and the equivalent acoustic “mass” of the tube.

Segment 15 This DUCT segment represents the so-called “Helmholtz” volume or
compliance of the multiplier sub-system. It acts like the spring that provides
the restoring force for the acoustic mass of the inertance tube. Check out
15G to see the thermoviscous loss associated with this element.

Segment 16 This RPNTArg calculates (reported in 16A) the resonant frequency
of the inertance-tube/compliance-volume system if it were uncoupled to any-
thing else. This is an important metric for design because it directly affects
the magnitude of the volume velocity that passes through the regenerator.
Additionally, it is directly proportional to the cooling capacity potential of

the machine. Acoustic impedance of this duct segment and effective com-
pliance, or inverse stiffness, is also calculated and reported in 16B and 16C

Segment 17 The SOFTEnd segment ends the leg of the feedback network. The
end is “SOFT” because acoustic power is allowed to flow through the End as
seen in 17E.

Segment 18 This dummy duct segment (zero length) starts the integration of the
acoustic field through the cold-exchanger, regenerator, ambient-exchanger
part of the feedback network. Its function here is to produce the acoustic
field variables that are referenced for calculation in Segment 19.

Segment 19 This RPNTArg is not used, but was an attempt to make the INSULATE
mode work correctly. It didn’t seem to create the desired condition.

Segment 20 This BEGIN segment is only here as a Band-Aid so that the gas
can be cold near the cold-exchanger and not cold everywhere else in the
model. Without it, the gas would be cold in the feedback sub-system, and
this isn’t a realistic condition. Notice the use of sameas x in many of the
input parameters - this is a way to reference input values at other segments of
the model. The sameas feature helps the model maintain consistency when
changes are made. The use of sameas here refers to the previous segment
and has the effect of passing all relevant acoustic variables except for the
temperature, which is allowed to be a GUESS, so that DeltaE can match
the TARGET found at 25f, the cold exchanger metal temperature.

Segment 21 This segment provides another option to insulate or un-insulate the
model from this segment to the end; it is not used.

Segment 22 This IMPEDance segment models the latex diaphragm used to mit-
igate so-called Gedeon streaming (Gedeon, 1997; Swift, 2002) through the
feedback-network and regenerator. Although DeltaE does not model this
second-order, non-zero time-averaged effect, the diaphragm does affect the
acoustic propagation of energy through the device. Since the diaphragm is

not streched tightly — in fact it hangs loose when there is no acoustic en-
ergy flowing in the device — it is modelled here as though it presents only a
mass-like (imaginary only, directly proportional to frequency) impedance to
the acoustic propagation.

Segment 23 Since there is a small volume enclosed between the diaphragm and
the cold side of the regeneraor, this DUCT segment takes care of the effect of
that gas space.

Segment 24 This RPNTArg calculates the peak-to-peak gas displacement as it
enters the regenerator and the Reynolds number. It is useful for design and
evaluation of the heat exchanger in the following segment.

Segment 25 The HX segment represents the thermally active portion of the cold-
side heat exchanger that is the electrical heater woven into a screen. The
MetalT is made a target (25f should equal 25H for a solution) and DeltaE
has been allowed to adjust the BEGIN temperature (20c) to hit this target.
Also, cooling capacity (25e) has been enabled as a GUESS - it’s a positive
quantity because the machine is a refrigerator and heat is moved into the
device through this heat exchanger. Although the real device doesn’t have a
real exchanger, just a serpentine weave of constantan wire, the model has to
have a heat exchanger segment in order to put a heat load on the regenerator.
The parameters of this notional parallel fin heat exchanger don’t try to model
the wire, although the length of the fins in the acoustic direction has been set
to only 1 mm in an attempt to minimize the amount of surface thermoviscous
loss that the model reports for the notional exchanger.

Segment 26 This is the regenerator, modeled as a stack of screens with some
cross-sectional area 26a, length 26c, volume porosity 26b (see 3.1.6) and
hydraulic radius 26d (see 3.1.7). The parameter 26e is a fudge factor to
allow for contact resistance between the layers of screen that is less than the
thermal conduction of solid metal. Notice that the screens have been specified
to be stainless steel. Valuable information appears on the right hand side of
this segment. The extreme temperatures of the gas in the regenerator are
shown in 26G and 26H. Notice that the exhaust temperature is a little higher

than the hot-side heat exchanger metal temperature, so that heat can flow
from the gas to the metal of the exhaust heat exchanger and likewise, the
cold-side gas temperature is lower than the load (cold) heat exchanger metal
temperature. The amount of acoustic energy dissipated in the regenerator
is shown in 26I. A large fraction of this is the minimum amount of energy
needed to pump 1.2 W (25G) up the specified temperature difference required
by the 1st and 2nd Laws of Thermodynamics while the rest of the energy
dissipated in the regenerator is dominated by the thermoviscous attenuation
in the screens.

Segment 27 This RPNTArg calculates the peak acoustic Reynolds number (27A)
and gas peak-to-peak displacement amplitude (27B) to compare to the length
of the hot exchanger tubes.

Segment 28 The ambient heat exchanger is modelled with the TX segment since
its a shell-and-tube type of exchanger. Heat exhausted is shown in 28G as a
negative quantity because it is leaving the model. DeltaE does not calculate
anything about the turbulent enhancement of the heat exchange coefficient.

Segment 29 The UNION segment tells DeltaE that the two branches should be
joined at this point in the acoustic field. DeltaE has two TARGETs in
this segment. These enforce equality between the complex pressure at this
location and the complex pressure at the SOFTEnd found at Segment 17.
DeltaE will get to adjust the impedance at the TBRANch in order to meet
these TARGETs.

Segment 30 The HARDEnd provides a rigid acoustic termination for the model
by way of the TARGETS that effectively force complex volume velocity to
be zero here. It’s also reassuring to note that the total power exiting the
model is zero (30E). This implies that all the energy put into the model at
segment 0, the BEGIN segment, has been dissipated or used to move heat and
consequently accounted for in the model.

Segment 31 This segment calculates the nominal regenerator acoustic resistance.
Although thinking of the regenerator as only a lumped resistance (or even a
two-terminal element) is not particularly accurate (since it is more accurately

a four-terminal element — both volume velocity and pressure change across
it), it can be a useful parameter when checking the design.

Segment 32 Thermal back conduction in the regenerator is an important loss to
consider in the design because it can be one of the dominant loss mechanisms.

Segment 33, 34 More calculations of Reynolds numbers and gas displacements
show up here.

Segment 35, 36 These segments allow the regenerator-end temperatures to be
specified as targets instead of the heat exchanger metal temperatures. Since
the regenerator-end temperatures are being measured in the experiment and
don’t depend on DeltaE’s naive heat exchanger model, it is more accurate
to use these temperatures to compare measurements to the DeltaE model.

Segment 37 This is another diagnostic calculation that calculates the Helmholtz
overpressure as a percentage increase in the pressure in the Helmholtz res-
onator compared to the pressure in the bellows space.

Segment 38 Lastly, the COP relative to Carnot COP is calculated in 38A, along
with COP in 38B and Carnot COP in 38C.

One of the reasons that the complete DeltaE model is reproduced here is
that it coveys the as built dimensions of the pre-prototype machine and several
operating parameters. The acoustic and thermal operating points shown in this
output file are a good representation of the operation of the machine in the testing
that will be described in the next section.

3.2 Apparatus, Measurements and Results
Shown in Fig. 3.9 is the exterior view of the complete pre-prototype machine.
Figure 3.10 shows the thermal core that includes the regenerator and hot exchanger
with the diaphragm glued in place over the cold side of the regenerator.
The labelled schematic cross-section shown in Fig. 3.5 should provide the reader
with a good understanding of how the pre-prototype subsystems fit together.

Figure 3.9. This photograph shows the exterior of pre-prototype apparatus ready for
operation. The pressure vessel shown has been designed for operation to 150 psi, but
the machine was run exclusively at 1 atmosphere for proof-of-concept.

Figure 3.10. This photograph shows the pre-prototype thermal core ready to be in-
stalled in the resonator. The loose latex diaphragm is shown glued in place and the
Lexan interface plate is shown at the bottom. This plate is sandwiched in the pressure
vessel bolt-up and accommodates the water circuit for the exhaust exchanger that enters
and exits the pressure vessel.

3.2.1 Instrumentation and Measurement
There are three major categories of measured quantities: static and acoustic pres-
sures, motor dynamics and electrical inputs, and temperatures. From these mea-
surements, both acoustic and thermal power flows can be calculated. This section
will address each of these categories and the calculations that are enabled by them.


There is interest in measuring pressure at two locations in the pre-prototype device:
within the bellows space and within the Helmholtz volume. The pressure at these
two locations was measured with an Endevco8 model 8510-B/C pressure transduc-
ers that have a 10-32 threaded interface. One of these transducers (8510B-200, SN
AG498) was used to measure the pressure in the bellows space for two reasons:
to provide a quantity to compare to the DeltaE model and to allow calculation
of the acoustic power flowing into the resonator from the motor. This calculation
involves knowing the velocity of the motor/piston; the phase difference between
the pressure and the velocity at the piston location; and the acoustically active
surface area of the piston (120 cm2 ). The measurement of velocity and phase will
be discussed in this section.
The other microphone (8510C-100, SN 10717; permanently mounted in the
machine and pressure-vessel feed-through constructed for the electrical wires) was
used to measure the pressure in the Helmholtz volume to provide data with which
to compare to design calculations about the pressure gain and phase shift provided
by the feedback network.

Motor Dynamics

Although the goal of the pre-prototype construction and measurement was to verify
understanding and prove design techniques, the motor used in the pre-prototype
will be used in the B&J prototype. Consequently, making some efficiency measure-
ments with this motor was both easy and useful. Using the Yokagawa9 WT-110
Power Meter, the current, voltage and the phase between them are measured.
Endevco Corporation, 30700 Rancho Viejo Road, San Juan Capistrano, CA 92675
Yokogawa, 2 Dart Road Newnan, GA 30265-1094

A Schaevitz10 model E300-0901 LVDT (linear variable differential transformer)
was used to measure the displacement response of the motor armature/piston
(Fraden, 1993). Because commercial LVDT demodulators often assume low fre-
quency operation, and commercial designs emphasize amplitude accuracy at the
expense of phase accuracy, a special lock-in detector circuit was developed to pro-
vide an LVDT excitation signal at 20 kHz and demodulate the signal to extract
the oscillating piston position x1 .
The calibration constant of the LVDT is proportional to the excitation voltage,
(typically 4.00 Vac ). Another custom circuit was developed that uses an error-
integrating feedback signal as input to an automatic gain control (AGC). This
stabilizes the excitation voltage by comparing its root-mean-square voltage value
to a high-stability voltage reference (National Instruments LM 136). The com-
bined lock-in detector and stabilizer circuit was tested and phase calibrated before
installation in the pre-prototype device.
The average acoustic power delivered by the driver Ė2 can be determined from
the product of the piston’s volumetric velocity (U1 ) and the pressure at the power
piston’s location (p1 ) times the cosine of the relative phase angle (θ) between p1
and U1 as shown in the following expression.

1 1
Ė2 = p1 U1 cos(θ) = p1 x1 ωApiston cos(θ) (3.2.1)
2 2

Temperature and Heat Flow

Temperature measurements were made at nine locations in the regenerator to
detect the presence of any streaming that would manifest as a non-linear tem-
perature gradient in the regenerator. These measurements were made with nine
Type-E (Ni-Cr/Cu-Ni) thermocouples, three of which were arrayed (separated by
120◦ in azimuth) at the cold side (one screen layer from the heater wire), three in
the middle of the regenerator (between the 45th and 46th screen) and three at the
exhaust side (one screen layer away from the hot exchanger). These thermocou-
ples also provide a set of data to input to the DeltaE model for the hot and cold
regenerator-end temperatures.
Measurement Specialties, Inc., 710 Route 46 East, Suite 206, Fairfield, NJ 07004.

Another set of temperature measurements were made to determine the exhaust
heat that flows from the heat exchanger to the water flowing within the shell of the
exhaust heat exchanger. Since the imposed heat load on the regenerator is provided
by an direct electrical current, it was easily measurable with a DC multimeter. On
the exhaust side, the inlet and outlet temperature of the water flowing in the
exhaust fluid circuit was measured with two thermistors (Model 4150-1/8-6-72-
TH55033-RPS). These are ∼2.25kΩ (at 20◦ C) probes from RDP Corp11 . If the
flow rate is known, heat flux into or out of the water stream can be calculated
with the following expression:

Q̇ex = ṁcp ∆T = U ρcp ∆T (3.2.2)

where U is the volume flow rate and ρ is the density of water which can be consid-
ered constant over the small (25◦ C —35◦ C) temperature range of the water in the
exchanger. The specific heat of the water is cp and ∆T is the difference between
outlet and inlet water temperatures.
Typically, making this measurement requires a flow meter of some kind, but
these have been found troublesome in the past; their calibration value is not con-
stant and they become fouled with even a small amount of suspended solids in the
water line. For this experiment, a constant head fluid delivery system was built
that is described in Fig. 3.11. The water is pumped to an inner chamber which
is allowed to overflow into the overflow tank. The water drains from this inner
chamber and through the heat exchanger. Since the water is always overflowing
the inner chamber, the static pressure on the supply line to the heat exchanger
is always constant, which guarantees a constant flow rate of water through the
heat exchanger. The system worked well; using the two thermistors, heat flow
measurements were calibrated to have an absolute error of less than 50 mW in 5
The calibration entailed building a calorimeter using a vacuum insulated bottle
and a resistance heater to supply a known heat flow into the water (in place of the
exchanger). The flow rate is measured with a stop watch and a scale by sliding
a small collection dish in front of the discharge pipe as the water falls into the
RDP Corp., 5877 Huberville Ave., Dayton, Ohio 45431.






Figure 3.11. This drawing shows a schematic representation of the constant head
thermal exhaust circuit. The box labelled EHX represents the shell-and-tube exhaust
heat exchanger that loads the circuit. The fact that the free surface of water overflowing
the inner chamber has a constant height means that the flow rate is constant through
the circuit. The added stability and measurement simplicity using this system allowed
heat flux measurements with absolute accuracy of 50 mW in 5W for the prototype and
worked equally as well for the 300 W exhaust stream of the B&J prototype.

collection tank. The amount of time that the collection dish is collecting water is
recorded and then the dish is weighed to obtain the mass flow rate. If the flow
circuit is changed (for example, if more tubing is added to the circuit) the flow rate
has to be remeasured. With consistent flow circuit configuration, variations in flow
rate were less than 0.5%. The stability of this measurement system is much better
than previous local methods of measuring heat flux in flowing liquids. Although
the pre-prototype only exhausts on the order of 10 W, this system was also used
to measure the exhaust heat flux on the factor-of-ten more powerful B&J machine.
The effort spent to install it was rewarded with consistently good performance.

Data Acquisition

The signals from the microphones, the LVDT and the thermistors were all mea-
sured using the multiplexed 6 1/2 digit digital multimeter (DMM) inside of an
HP 34970A Data Acquisition/Switch Unit. The linear motor was driven using
the oscillator output of an HP 4192A impedance analyzer. By using the internal
oscillator in the impedance analyzer (amplified by a Crown Power-Tech 2.1 linear
amplifier), the phase between the LVDT response and the response of the micro-
phone in the bellows space could be measured. This phase difference is required to
calculate the average acoustic power using (3.2.2). A Keithley model 740 thermo-
couple scanner provided the stable cold-junction compensation, micro-voltmeter
and non-linear regression of DC voltage response of the thermocouples mounted
in the regenerator. The HP 34970A, the impedance analyzer, the Yokogawa power
meter and the Keithley thermocouple scanner were all connected to a PC via a
GPIB bus. These instruments were polled using software written in LabVIEW.
A screen-shot of the data acquisition program created for this project is shown in
Fig. 3.12

3.3 Measurement Results and Analysis
Measurements were initiated by setting the acoustic power to be around 1.7 W
with no electrical heat load on the cold side and allowing the system to come to a
steady-state temperature; the system is judged to have reached steady-state when
the average of the cold side thermocouple readings stops showing a monotonic

Figure 3.12. This screen-shot shows the data acquisition program created for measure-
ment of the pre-prototype performance. The program allows control of the amplitude
and frequency of the oscillator in the HP 4192A, and in later versions allowed control of
the power to the resistance wire heat load as well. Performance metrics like COP and
input acoustic impedance are calculated in real time. The instruments completed a full
scan of all the sensors and instrumentation in about 30 seconds.

decrease. Once the system is judged to have come to steady-state, the electrical
current in the resistance heater wire is increased so that the electrical power dis-
sipation of the wire is increased by about 1/4 W. This process is continued until
the machine can no longer maintain a steady-state cold side temperature that is
below room temperature. At each steady-state point all pertinent data is recorded
Heat leaks from the room to the cold side of the regenerator and from the
hot side of the regenerator to the room present a measurement problem for any
system that isn’t at room temperature, especially when measuring low power flows
on the order of 1 W. However, since the measurement system allows independent
measurement of each term in the energy balance prescribed by the First Law of
Thermodynamics, these heat leaks can be estimated. The First Law energy balance

Q̇ex = ∆Ė2 + Q̇load (3.3.1)
Q̇meas,ex + Q̇leak,ex = ∆Ė2 + Q̇imposed,load + Q̇leak,load (3.3.2)

is shown schematically in Fig. 3.13.
The measurement system allows accurate knowledge of both heat fluxes with
the subscript “meas” and the amount of acoustic power delivered to the resonator.
Since most (if not all) of the dissipation of the acoustic wave winds up getting
moved up the regenerator, any imbalance of the equality in (3.3.2) can be attributed
to either Q̇leak,load or Q̇leak,ex . Without some complicated finite-element model to
determine what the conduction paths of these “leaks” might be, the easiest way
to correct for these two heat flows is to assume a single thermal conductance value
for each leak; one on the exhaust side and one on the load side, as follows:

Q̇leak,ex = (Tex − 20◦ C) (3.3.3)
Q̇leak,load = (20◦ C − Tload ). (3.3.4)

The thermal conductance 1/Rh and 1/Rc are inferred experimentally by adjusting
the values so that in a data set, the root-mean-square sum of the First Law imbal-
ance (the sum over each steady-state point) as described by (3.3.2) is minimized.
These values are found to be 1/Rh = 0.0594W/K and 1/Rc = 0.717W/K.



.  .  



Figure 3.13. This drawing illustrates the unmeasured heat flows that bypass the mea-
surement system. In a system with small heat flows, these “heat leaks” are important
to consider.


Cooling Load (W)



1.4 Modeled with DeltaE

6 8 10 12 14 16 18 20 22 24
Load Temperature ( C)

Figure 3.14. This graph shows the cooling capacity of the pre-prototype for a constant
input power of 1.75±0.05 W. The heat load on the cold side of the regenerator was
increased in 1/4 W increments to generate this curve.

After correcting for these nuisance heat flows, the data is ready to be compared
to DeltaE. This can be done by running DeltaE with experimental conditions
such as static pressure, acoustic pressure at the driver, regenerator-end tempera-
tures as inputs. Outputs from the DeltaE model to be compared to experimen-
tally measured quantities are the acoustic power input to the resonator (using the
volume velocity and phase since pressure is matched to experimental conditions),
cooling capacity Q̇load and exhaust heat flux Q̇ex . The graph in Fig. 3.14 shows how
the DeltaE model results compare to the corrected measured cooling capacity for
a constant input power of 1.75±0.05 W.
The only unknown geometric parameter in the model is the impedance that
the latex diaphragm (see Fig. 3.10) presents to the acoustic wave. Since this is a
difficult parameter to measure or calculate, the value for the active mass of the
annular diaphragm was adjusted in DeltaE using only one experimental point on
the graph in Fig. ?? (the second from the left at about 5.5◦ C) until the cooling
capacity returned by the model matched that measured at that operating point.
The value found to make the model agree with the data best is a reactance of 6.15
x 104 Pa sec/m3 .

To evaluate the physical reality of this value, the effective cross-sectional area
and moving mass of the diaphragm is required. This is difficult, since the di-
aphragm is glued to the fixture with an intentional bulge so it acts completely
mass-like and with no compliance for ease of modelling. As a consequence of the
bulge, the diaphragm doesn’t oscillate as a whole, but rather motion at one loca-
tion of the diaphragm is of a different phase than the motion at other places. (This
fact about the diaphragm motion is inferred about the pre-prototype, but directly
observed in the demonstration device.) Using a density of 0.1 kg/m2 and assuming
that the bulge increases the surface area of the diaphragm by 30% over the area of
the regenerator annulus (17.1 cm2 ), the physical mass of the diaphragm is about
0.22 g. The mass reactance, or inertance of the diaphragm is Id = ω md
. Assum-
ing that 100% of the physical diaphragm mass is moving and that the effective
cross-sectional area of the diaphragm is the physical cross-sectional area of 17.1
cm2 , the inertance would be 3.25 x 104 Pa sec/m3 , which is too small by a factor
of two. However, as a consequence of the “broken up” motion of the diaphragm
we might assume that effective cross-sectional area of the diaphragm is less than
the physical cross-sectional diameter and since that parameter is squared in the
denominator, an estimate of effective cross-sectional area of about 70% of the phys-
ical cross-sectional area makes the inertance about 6 x 104 Pa sec/m3 using the
physical mass.
As a consequence of this loose estimate of the effective diaphragm cross-sectional
area, it does seem that the value of 6.15 x 104 Pa sec/m3 used for the reactive part
of the impedance of the diaphragm represents the acoustic mass of the diaphragm
in a reasonable way.
This value for reactance was then fixed and the model was run on the other
steady-state operating points to generate the line shown in Fig. 3.14. The graph
shown in Fig. 3.15 shows the COP of the pre-prototype for the same steady-state
points as shown in Fig. 3.14 with the model results represented as a line.
This good agreement between measured results and the DeltaE model, al-
though qualified by an adjustment for inferred external heat leaks and plausible
adjustment for the impedance of the diaphragm, gives us confidence that these
measurement techniques are accurate, that our design technique and modelling
capabilities are adequate for low temperature span regenerator designs and that




0.8 Modeled with DeltaE

6 8 10 12 14 16 18 20 22 24
Load Temperature ( C)

Figure 3.15. This graph shows the coefficient-of-performance (COP) of the pre-
prototype for a constant input power. The heat load on the cold side of the regenerator
was increased in 1/4 W increments to generate this curve.

our understanding of the Helmholtz resonator feedback concept is correct. Follow-
ing these measurements we were confident that we could proceed to the design of
a pressurized system to refrigerate ice cream for Ben and Jerry’s.
Chapter 4

The Prototype Ice-Cream Storage

During the middle-stages of the design of what became the pre-prototype, our re-
search group (Steve Garrett, Bob Smith and the author) secured a contract with
Ben and Jerry’s Homemade Ice Cream of Burlington, Vermont (B&J) to design
and build a thermoacoustically refrigerated ice-cream storage cabinet. Two of
the goals of B&J for the project were to a) stimulate manufacturer interest in
thermoacoustic technology so that B&J and their parent company Unilever could
buy environmentally friendly refrigerated storage cabinets; and b) continue their
conversation with customers about reducing global warming, corporate environ-
mental impact and grassroots environmentalism. Since neither B&J nor Unilever
manufacture refrigerators/freezers but instead are huge consumers of commodity
refrigeration units, their sponsorship is forward-looking and altruistic.
As knowledgable consumers, B&J/Unilever had well-defined performance tar-
gets for the project. These included a continuous cooling capacity in excess of
70 W; a secondary load-side flow circuit temperature between -32 ◦ C and -37 ◦ C;
an exhaust-side flow circuit temperature above 45 ◦ C; and a net COP based on
electric input power at this operating point of better than 0.67. Additionally, the
prototype size is limited to 50 cm x 32 cm x 30 cm (l x w x h) and the weight is
limited to 25 kg. Although the target product temperature is -20 ±2 ◦ C and the
machine will likely exhaust to a controlled indoor temperature, the sponsors built
in a conservative temperature difference between the secondary fluid temperatures
and product/ambient temperatures.
A physically larger thermoacoustic chiller, with similar thermal performance

specifications, was produced and field-tested using a stack-based, standing wave
design (Garrett, 1991, 1997). Its overall COP, including heat exchanger losses
and the electroacoustic efficiency of its loudspeakers (η = 55%) would not have
been adequate to meet the specification outlined above. For that reason, and since
regenerator-based machines have the thermodynamic potential to be more efficient
than a stack for the reasons cited in Sec. 1.3.1, the design for B&J uses a regenerator
rather than a stack. Because the pre-prototype performance conformed so well to
the models, the investigators had confidence in their ability to design and build
hardware that could match performance predictions.
The prototype being described in this chapter is the incarnation of two patented
sets of ideas (Poese et al., 2003; Smith et al., 2003). The patentable ideas can be
broadly classified as a) the “bellows-bounce” concept where the thermal core (heat
exchangers, stack or regenerator, acoustic network and thermal buffer tubes) of the
refrigerator is located inside of the compliant cavity formed by the bellows (tested
in the pre-prototype and discussed in Sec. 3.1.4; and b) the “vibromechanical mul-
tiplier” (our term) which is essentially a acousto-mechanical Helmholtz resonator.
Using the vibromechanical multiplier, the prototype has a different topology com-
pared to the pre-prototype and other regenerative Stirling cycle devices made over
the past two centuries.
This chapter of the dissertation will first focus on the novel aspects of the B&J
prototype (called simply “the prototype”) and then describe the of the machine
that is able to keep ice-cream at required storage temperatures with a COP that
exceeds the project specification (and exceeds the COP of most of the current
cabinets used for ice cream sales used throughout the world today).

4.1 New Aspects of the Prototype
Although the pre-prototype was successful in the sense that its performance agreed
well with the predictive models, some aspects of the design were not efficient or
didn’t scale well to devices made to maximize efficiency, simplicity and power
density. First, although a round thermal core is the best geometry to maximize
power density, the thermal core in the prototype has a square aspect ratio. Wire
screens are usually woven into rectangular bolts and cutting them into disks (or

annuli) is fairly wasteful and expensive. Round heat exchangers are generally also
more complicated to fabricate and not prevalent in the heat exchanger fabrication
industry. Secondly, an inertance tube while simple, can (in the design process)
become rather long and lossy compared to the overall device size. These problems
motivated the invention of the vibromechanical multiplier subsystem that replaces
the feedback network as described in Sec. 3.1.2.
The prototype operates at 100 Hz and use helium at a static pressure of 10
atmospheres as the working gas. The 100 Hz operating frequency was dictated by
the available stroke (2x1 ), force factor (Bl-product) of the moving magnet linear
motor and the maximum allowable current in the coils of the motor. The model
of the design suggested that the motor would have to produce 200 W of acoustic
(mechanical) power, Ė2 = F v/2 = F x1 ω/2. Since F = Bli, the only way to
produce the required acoustic power while respecting the current limit is to increase
the operating frequency.
The regenerator is again a stack of stainless steel screens but is square in aspect
ratio measuring 8.94 cm (3.52”) on a side. The hydraulic radius is 39.2 µm and
the volume porosity is 74% open volume which corresponds to a screen of 145
wires/inch with a wire diameter of 0.0022”. The regenerator is 3 cm tall. These
regenerator parameters were chosen by using the process outlined in Chapter 3.
Another required element for a machine that is designed to span a significant
temperature is some way to isolate the cold gas from the ambient temperature
gas. This is typically part of the job of a displacer in a Stirling machine and
is sometimes called a “thermal buffer tube” in thermoacoustic literature (Swift,
2002). The thermal buffer in this machine is formed by two “windows” cut into
the thermally insulating plastic1 platforms that hold the regenerator and load heat

4.1.1 Vibromechanical Multiplier
Since the inertance of a passage is proportional to the ratio of the length to the
cross-sectional area, a long inertance tube can be shortened and narrowed to pre-
serve a required inertance value. This tradeoff comes at the price of increased minor
Ultem 1000, GE Plastics Structured Products, One Plastics Avenue, Pittsfield, MA 01201.

Tube length (cm)
10 20 30 40 50 60

Entry/Exit loss
Viscous loss
40 Thermal loss
Total dissipation
Dissipated Power (W)




2 3 4 5
Tube diameter (cm)

Figure 4.1. This graph shows the losses in an inertance tube that has an inertance
of 500 kg/m4 . The green curve represents the entry/exit losses and the black curve is
the sum of all of the losses in the tube. Notice that while the minimum amount of loss
possible is around 13 W, this requires an inertance tube that is almost 1/2 of a meter
long and four centimeters in diameter.

losses, the magnitude of which is a function of the ratio of the cross-sectional area
on one side of the interface compared to the area on the other side. The graph
in Fig. 4.1 shows the various loss mechanisms in the inertance tube that would be
required for the prototype. The figure shows the losses in an inertance tube that
has an inertance of about 500 kg/m4 , which is the value required in the feedback
network for the prototype. The inertance of the tube is expressed as L = ρm l/A
where l is the length of the tube, A is the cross-sectional area and ρm is the static
density of the working gas. For a constant value of inertance L, the ratio of length
and diameter is fixed if the inertance element is assumed to have a circular cross-
section of constant diameter. The expression for the viscous dissipation in the
inertance tube is
∆Ėv = U12 (4.1.1)
A2 δν

where U1 is the magnitude of the acoustic volume velocity in the inertance tube,
S is the perimeter and A is the area of the tube, µ is the shear viscosity of the
working gas and the δν is the viscous penetration depth. The expression for the
thermal relaxation dissipation in the inertance tube is proportional to the acoustic
pressure magnitude but is otherwise similar:

ω(γ − 1)Sδκ
∆Ėt = p21 (4.1.2)

where p1 is the acoustic pressure magnitude, γ = cp /cv is the ratio of specific heats
of the working fluid, ω is the radian frequency of oscillation and δκ is the thermal
penetration depth.
The exit/entry (or “minor”) loss at the interfaces of the inertance tube requires
the estimate of the minor loss coefficient, which can be between 0 and 1 depending
upon how sharp the change in cross-sectional area from the inertance tube to the
attached volume. The expression for the power dissipated by the entry/exit loss is

4 Kρm
∆Ėm = U12 |U1 | (4.1.3)
3π A2

where ρm is the static density of the gas and K is the minor loss coefficient. The
entry/exit loss is proportional to the cube of velocity and there is a location for
power loss at both ends of the inertance tube. Using the chart in Fox & McDonald
(1992) on page 355, the coefficient for each interface could be 0.85. For the minor
loss coefficient K in the equation above doubling the value of 0.85 to 1.70 is required
to account for both sides of the inertance neck.
Although the minimum loss is only about 13 W, such an inertance tube would
require a length of almost 1/2 of a meter and a four centimeter diameter. This
tube would be much too long to fit into a machine that has the size restrictions
mentioned above. Of course, a much shorter and narrower tube, while possible to
implement physically would create unacceptable large losses in the element for a
machine that is designed to move between 70 W and 140 W of heat.
Even if the graph in Fig. 4.1 overestimates the loss by a factor of two (which
might be true since entry/exit losses in acoustic flows are not well understood
(Wakeland, 2002)), any tube of reasonable compactness is still too lossy for this

application. The loss in the inertance tube represents an increase in the required
acoustic power, but also, all of the heat generated by this loss mechanism would
reduce the usable cooling capacity because the regenerator must pump this heat
at the expense of the amount of remaining useful heat removal from the load. A
solution to this problem is akin to the “bellows-bounce” concept; recognizing that
storing energy in the inertia of a physical mass has little loss associated with it
when compared to hydrodynamic inertia, an ordinary loudspeaker cone is used
as the inertance element. The cone is weighted to “tune” its inertance and it is
joined to the multiplier chamber with a Santoprene surround mounted to one end
of the cylinder that has the other end closed by the exhaust side of the thermal
core. This entire assembly is referred to as the “vibromechanical multiplier” since
as a feedback network it provides some multiplicative gain to the acoustic pressure
inside the multiplier volume. The following series of drawings in Figs. 4.2–4.6 show
how the topology validated in the pre-prototype evolved into the vibromechanical
multiplier concept used in the prototype.
The final design, shown in Fig. 4.6, has several advantages. The inertance ele-
ment is compact and the mass can be adjusted easily and independently by bolting
or removing washers. Although a second flexure seal is used for the multiplier cone,
the pressure difference across the seal is small. In the prototype design, the oscillat-
ing pressure magnitude is about 6% of the static pressure, and the “multiplication
factor,” or the gain of the feedback network is on the order of 10% of the oscillating
pressure; the pressure difference across the multiplier cone seal in the prototype is
then 10% of 6% or 0.6% of 150 psi or 0.9 psi.
A test rig was constructed to exercise the Santoprene surround at a stroke of
about 12 mm, which is the largest stroke that it could experience in the prototype.
After 95 million cycles (or 330 hours at 80 Hz) there were no signs of failure. (The
experiment was stopped after the actuator for the test rig failed!) If a machine
based on the prototype design were to be expected to have an 8-year lifetime the
total number of cycles required for both the bellows and the multiplier cone would
be 25 billion for continuous operation at a frequency of 100 Hz. While 95 million
cycles is two orders-of-magnitude less than the requirement for an 8 year lifetime,
the results of the fatigue test provided enough confidence to include the Santoprene
surround as the seal for the prototype. In the process of operating the machine,

Figure 4.2. This drawing shows the bellows bounce resonator used in the prototype.
The gas trapped inside the bellows and between the bellows and the pressure vessel
provides the dominant restoring force for the inertia of the moving mass of the system
(mostly the motor armature and power piston) that is forced to oscillate when electrical
current flows in the copper coils of the motor.

Figure 4.3. This drawing shows the bellows bounce resonator with a pre-prototype
configuration of the thermal core. The compliance element of the feedback network is on
the exhaust side, bolted to the top of the pressure vessel. The inertance tube penetrates
the center of the annular regenerator and the cold side of the regenerator is facing the

Figure 4.4. This drawing shows the bellows bounce resonator with an alternate pre-
prototype configuration where the inertance tube is an annulus that flanks the thermal
core. This is how the tabletop demonstration inertance tube was configured.

Figure 4.5. This drawing shows the bellows bounce resonator with an inverted pre-
prototype configuration of the thermal core. The regenerator has been turned upside
down. This allows the compliance element to be contained inside the bellows thus re-
ducing the volume of the entire machine. This is advantageous from the standpoint of
pressure ratio; since for a fixed motor stroke, the maximum attainable acoustic pres-
sure amplitude is function of the total volume in the resonator (among other things).
The inertance tube is independent of the length of the regenerator, which adds design

Figure 4.6. This drawing shows final prototype configuration with the multiplier cone
substituted for the hydrodynamic inertance tube. To increase thermal stability in the
resonator, the machine is operated in an inverted position compared to that shown here.
The pressure vessel is 10” in diameter and about 18” tall.

Figure 4.7. A cut-away solid model of the entire machine showing a cold heat exchanger
(dark grey) that is identical to the hot heat exchanger is contained within a thermally-
insulating Ultem plastic plate (brown). The cold heat exchanger plate is in contact with
the “platform” plate (yellow) that contains the regenerator and sensor signal lines. A
second thermally-insulating Ultem plastic plate (green) provides the contoured plenum
space that directs the oscillating cold helium gas in and out of the thermal buffer spaces
(“windows”) through the platform. A solid stainless steel plate (dark grey) is used to seal
the platform, cold heat exchanger plate, and plenum plate to the pressure vessel (green)
and provide the force necessary to resist the 10 atmospheres of internal helium gas
pressure. Enclosing the hot heat exchanger is the vibromechanical multiplier comprised
of the compliance volume within the multiplier’s cylinder (orange) that is terminated
by an ordinary loudspeaker cone (purple). Directly below the speaker cone is the power
piston cone (light green) that is attached to the bellows (gray). The moving-magnet
linear motor, which moves the power piston cone, is shown as a gray rectangle with
yellow straps. It is attached to the bottom plate (black) that forms the lower boundary
of the pressure vessel. The cylindrical portion of the pressure vessel is shown in green.

about another 50 hours has been accrued on the Santoprene seal. Based on the vast
experience accumulated by loudspeaker manufacturers, it appears that Santoprene
failures are attributed to exposure to ozone or ultraviolet light, neither of which
are present in a dark resonator filled with helium gas.
The total loss attributed to the inertance cone (see Fig. 4.11) is less than 3.5
W and most of that is due to damping (or hysteresis) loss in the Santoprene. This
amount of loss is acceptable, and is downright wonderful when compared to the
losses in an inertance tube of equivalent inertance as shown in Fig. 4.1 which would
be unacceptably large.
A second-order pressure difference ∆p2,0 will develop across the vibromechanical
multiplier cone due to the acoustic energy flux Ė2 through the regenerator. It can
be calculated approximately (Swift, 2002) using the low Reynolds Number limit of
the correlations for screens provided by Kays & London (1998).
∆p2,0 ' µm (x)Ė2 (x)dx (4.1.4)
Arh pm

An estimate for ∆p2,0 can be made simply by evaluating the integral using average
values of the acoustic power flow through the regenerator hĖ2 i = 230 W, and an
average viscosity of hµm i = 1.9 x 10−5 kg/sec-m of the helium gas. The area of the
regenerator, A = 80 cm2 , the hydraulic radius of the regenerator is rh = 39.2µm,
and the mean pressure at nominal conditions is pm = 1.0 MPa. Substitution into
(4.1.4) produces a pressure difference of only 64 Pa. Since the area of the rigid
mass element (e.g. loudspeaker cone) is only 130 cm2 , the total static force on the
element is less than one newton.

4.1.2 Heat Exchangers
The shell-and-tube exchanger used in the pre-prototype did not work well due to
an unacceptably large temperature difference between the metal and the oscillating
air. With such a low power device, the gas flows were not fast enough to stimulate
the generation of turbulent mixing; consequently, the tubes were sized much too
large in diameter for laminar flow. A large temperature difference between the
oscillating gas and the secondary heat exchange fluid reduces overall efficiency since
difference between load and exhaust temperatures must increase to accommodate

the heat exchanger temperature deficit on the load and the exhaust sides.
A standard parallel-fin heat exchanger was designed for the prototype for use
on the exhaust side. This exchanger is simple since it utilizes hollow tubes through
which the secondary exhaust fluid is pumped. Since helium at 10 atmospheres has
a fairly high thermal heat capacity and conductivity, fins between the tubes aren’t
required if the tubes are closely spaced.
One of the advantages of the prototype configuration, as shown in Fig. 4.6, is
that the cold exchanger is close to the flat wall of the cylindrical pressure vessel.
The co-location allows a design of an integrated cold heat exchanger and pressure
vessel wall, called a cold head (Chrsyler & Vader, 1994). The advantage of this
arrangement is that it can eliminate the need for a secondary fluid circuit to get
heat from the load to the helium in the machine. The atmospheric pressure side
of the exchanger could be finned and these might penetrate the walls of the ice-
cream storage volume. A fan could be used to circulate air past the external fins of
the cold head and the ice cream containers to increase the heat exchange rate. A
schematic of the machine with the cold head is shown in Fig. 4.8. Such a cold head
was created and tested with the prototype, but was abandoned in favor of a load
exchanger identical to the exhaust exchanger described in the following section.
Details of the fin geometry of the cold head were found to cause significant intra-
regenerator streaming: DC flows that circulate within the regenerator and move
heat from the hot side to the cold side. In the worst case, these intra-regenerator
streaming cells robbed all of the useful cooling that the machine produced. Since
a clear path exists to fabricate a new cold head that should work without induc-
ing streaming, an investigation of the intra-regenerator streaming problem could
become a useful research project.

4.1.3 Exhaust Heat Exchanger Design
The heat that is pumped from the cold end of the regenerator to the hot end
of the regenerator, along with the acoustic power absorbed by the regenerator
that is required to pump that heat, is deposited on the hot heat exchanger to be
exhausted from the system. This heat exchanger was constructed from a series of
parallel flat aluminum tubes that are extruded to include multiple parallel fluid

Figure 4.8. Cross-sectional drawing of the prototype with the integrated cold head
which serves as both the load heat exchanger and the cap of the pressure vessel. The
external atmospheric-side fins are normal to the page and the helium side fins are oriented
along the page.

Figure 4.9. Cross-sectional drawing of an extruded hot heat exchanger flat tube showing
ten rectangular internal channels that are 0.65 mm x 0.70 mm. The two channels at the
ends are “D” shaped with a radius of 0.35 mm and height of 0.70 mm. The tube is 1.3
mm thick and 12 mm wide. The tube is a custom extrusion produced by Thermalex,

flow channels within the tube2 . A cross-sectional drawing of one of these tubes is
shown in Fig. 4.9.
The flat tubes have a thickness, b = 1.30 mm, and are spaced by a gap, g =
700µm = 2.72δκ . The ratio of gas cross-sectional area to the total cross-sectional
area, Agas /A = g/(b + g) = 0.35. The regenerator area is 80.0 cm2 , so the length of
each side of the square regenerator is 8.94 cm (3.52 in). The number of tubes in an
array that will completely cover the regenerator surface is 45. The outer surface
area of each tube of length 8.94 cm is 22.51 cm2 so that the total surface area of
the tube array is 0.101 m2 .
The tubes were assembled to produce the hot heat exchanger by adding two
manifolds to the ends of the tubes that permit the heat transport fluid to enter and
exit the tubes. These manifolds were glued to the tubes using Stycast 2850 FT3 ,
a two-part room-temperature cure epoxy. A jig was fabricated from aluminum
that held each tube simultaneously at the correct spacing; the jig held the tubes
oriented vertically and the manifold was positioned. Then, the fairly thick epoxy
was poured into the manifold, just up to the level of the protruding tubes. A better
way to fabricate these exchangers is with a furnace braze operation, but since this
requires a custom brazing set-up, the time and budget of this project did not allow
it. A photograph of the completed exchanger is shown in Fig. 4.10.
The mean velocity of the gas hv1 i within the exchanger is related to the volume
Thermalex, Inc., 2758 Gunter Park Drive West, Montgomery, AL 36109; (334) 272-8270.
Emerson and Cuming, 46 Manning Road, Billerica MA.

Figure 4.10. This photograph shows the top view of the exhaust heat exchanger on the
left illustrating the 45 tubes and the gaps between them through which flows oscillating
helium gas. The half-oval shaped passages flanking the heat exchanger are part of the
thermal buffer region. The photograph on the right shows the end view of the tubes as
they are glued into the manifold.

velocity of the gas in the exchanger U1 and the gas-filled cross-sectional area of the
heat exchanger.
hv1 i = (4.1.5)
Ae x(Agas /A)
Again, assuming nominal conditions, U1 = 0.01027 m3 /sec (from the DeltaE
output file), and hv1 i = 3.37 m/sec. The peak-to-peak gas particle excursion 2ζ1
is simply related to hv1 i by the angular frequency ω = 2πf .

2hv1 i
2ζ1 = = 1.07 cm (4.1.6)

The kinematic viscosity of the gas ν and the viscous penetration depth δν
determine the acoustic Reynolds Number Reac for the gas in the tube.

hv1 iδν
Reac = (4.1.7)

For the nominal operating point, Reac = 51. This is well below the criterion
established by Merkli & Thomann (1975) for stability of the oscillatory viscous

boundary layer which is Reac ≤ 300. Viscous boundary layer stability suggests
that hgas ∼
= κ/δκ should provide a fairly accurate expression for the convective
heat transport coefficient hgas (Garrett et al., 1994).
The thermal conductance between the helium gas and the tube array depends
on the total tube surface area and the convective heat transport coefficient. The
thermophysical properties of the hot-side helium gas are reproduced in Table 4.1.
The estimated convective heat transport coefficient of the gas, hgas = 759 W/m2 -
K, is the quotient of κ/δκ and is also listed in Table 4.1. The thermal conductance,
1/Rth is taken to be the product of the total tube surface area and the convective
heat transport coefficient of the gas.

= hgas Atubes = 77 W/K (4.1.8)

Under nominal operating conditions, the exhaust heat that leaves the hot heat
exchanger is about 250 W. This heat exchanger model would suggest that the
difference between the gas temperature and the heat exchanger metal temperature
should be about 3.3 ◦ C. Measurements of hot-side regenerator temperature and
exhaust transport fluid temperature confirm that the difference is slightly less
than this theoretical estimate.
The performance of the fluid side of the hot heat exchanger can be calculated
by standard methods for convective heat transfer with internal flows under laminar
conditions (Incropera & DeWitt, 1996). Because the tubes are aluminum, a 50/50
mixture by volume of ethylene glycol (C2 H6 O2 ) and distilled water (H2 O) should
be used to prevent corrosion and fouling of the tube channels (see Table 4.2).
Pure water has far better heat transport properties (higher thermal conductivity
and lower viscosity), but anticorrosion compounds must be added for use in alu-
minum tubing. Since heat transfer data for water is well-known and anticorrosion
additives are used in negligible volumes compared to the amount of water in the
constant head flow system shown in Sec. 3.2.1, a mixture of water and anticorrosion
additives4 was used for initial laboratory performance measurements.
The internal surface area of the parallel channels in the array of 45 tubes (12
channels/tube) is 0.128 m2 . The hydraulic diameter Dh for each individual channel
these include a granular sodium nitrate powder and a commercial product called Cobratec
928 manufactured by PMC Specialties Group, 501 Murray Road, Cincinnati, OH 45217.
Property Symbol Units Warm Gas Cold Gas
Temperature Tm K/◦ C 323/50 238/-35
Pressure pm Pa 1 x 106 1 x 106
Density ρ kg/m3 1.490 2.039
Sound speed a m/sec 1057.5 904.1
Isobaric specific heat cp J/kg-K 5192 5192
Thermal conductivity κ W/m-K 0.1607 0.128
Thermal diffusivity α = κ/ρcp m2 /sec 2.077 x 10−5 1.213 x 10−5
Shear viscosity µ kg/m-sec 2.097 x 10−5 1.694 x 10−5
Kinematic viscosity ν = κ/ρ m2 /sec 1.407 x 10−5 8.310 x 10−6
Thermal expansion β 1/K 3.096 x 10−3 4.235 x 10−3
Thermal penetration depth (100 Hz) δκ µm 257 196
Viscous penetration depth (100 Hz) δν µm 211 163
Prandtl number σ 0.677 0.685
Polytropic coefficient γ 1.667 1.667
Acoustic convective coefficient hgas = κ/δκ W/m2 -K 759 790
Table 4.1. This table lists the thermophysical properties of helium gas at 1 MPa (145 psia).
Property Symbol Units Ethanol Ethanol 50/50 Glyc/Water Water
Temperature Tm K/◦ C 233/-40 273/0 316/43 320/47
Density ρ kg/m3 823 806 1052 989
Isobaric specific heat cp J/kg-K 2037 2249 3413 4176
Thermal conductivity κ W/m-K 0.186 0.174 0.245 0.637
Thermal diffusivity α = κ/ρcp m2 /sec 1.109 x 10−7 9.599 x 10−8 6.824 x 10−8 1.542 x 10−7
Shear viscosity µ kg/m-sec 4.810 x 10−3 1.773 x 10−3 2.000 x 10−3 5.790 x 10−4
Kinematic viscosity ν = κ/ρ m2 /sec 5.844 x 10−6 2.200 x 10−6 1.901 x 10−6 5.854 x 10−7
Prandtl number σ 52.4 22.9 28.4 3.8
Convective coefficient hgas = N uκ/Dh W/m2 -K 954 893 1200 3119
Nusselt number Nu 7.9 7.9 3.3 3.3
Table 4.2. This table lists the thermophysical properties of the secondary heat transport fluids: ethanol, glycol/water mix and
distilled water.

is given by four times the ratio of the channel cross-sectional area Achannel to the
channel perimeter P . For the channels in the Thermalex tubes, Dh = 674 µm.
The Nusselt Number N u relates the convective heat transport coefficient to the
transport fluid thermal conductivity κ and hydraulic diameter Dh :

κN u
hf luid = . (4.1.9)

For the channels of square cross-section, the value of the N u is 3.3±0.3, with the
uncertainty related to whether the heat transfer takes place at constant flux or
constant temperature (Incropera & DeWitt, 1996). The relevant thermophysical
properties for the heat transport fluids are listed in Table 4.2.
The thermal conductance for all of the channels 1/Rth is taken to be the product
of the total channel surface area and the convective heat transport coefficient of
the ethylene glycol.
= hf luid Achannels = 165 W/K (4.1.10)
Under nominal operating conditions, the exhaust heat that leaves the hot heat
exchanger is about 250 W. This heat exchanger model would suggest that the
difference between the ethylene glycol temperature and the heat exchanger metal
temperature should be 1.66 ◦ C.
The small channels within the flat tubes shown in Fig. 4.10 provide excellent
thermal contact to the ethylene glycol/water mixture, but they also produce vis-
cous resistance to the flow. Since the flow in the channels is assumed to be laminar,
the rate of heat transfer from the channels to the fluid is independent of flow rate
so the choice of flow rate is somewhat arbitrary. For the purposes of these design
calculations, the flow rate V̇ has been chosen to make the difference between the
inlet and outlet temperatures ∆T = 1.0 ◦ C. This difference is large enough to
provide adequate power resolution using a pair of thermistors to measure ∆T .

V̇ = (4.1.11)
ρcp ∆T

For the assumed exhaust heat and the thermophysical properties listed for the
ethylene glycol/water mixture in Table 4.2, the volume flow rate is V̇ = 71.3
cm3 /sec. The total cross-sectional area of the 540 channels in the 45 tubes is 2.41

x 10−4 m2 , so the average flow velocity of the ethylene glycol/water mixture hvi
= 0.296 m/sec. The Reynolds number for the flow is then based on the hydraulic
diameter (Dh = 674 µm) of each individual channel and the kinematic viscosity ν
of the ethylene glycol/water mixture.

ReD = = 105 (4.1.12)

That value is well below the Reynolds Number for the onset of turbulence, usually
taken to be ReD ∼= 2000, so the laminar flow assumptions are fully justified.
The pressure drop required to produce the required volumetric flow rate is (Fox
& McDonald, 1992)
4cf L ¡ 1 ¢
∆P = ρhvi2 (4.1.13)
ReD Dh 2
The length of the tubes L including the portions that protrude into the man-
ifolds is 95.4 mm. For laminar flow in channels of nearly square cross-section
cf ReD ∼= 14.4. This results in a pressure drop of ∆P = 3,580 Pa = 0.52 psid. The
required pumping power is the product of the pressure drop and the volumetric
flow rate: ∆P V̇ = 0.26 W. Although this pumping power does not include the
power required to move the ethylene glycol/water mixture through the connecting
plumbing it is sufficiently small that it is not a concern for this design.
Design of the load exchanger, which contains circulating ethanol, follows the
same logic as the design detailed above for the exhaust exchanger except that
ethanol is the transport fluid. The total electrical power consumed by the cold
side pump is under 5 W. Since this electrical consumption includes all of the
energy required to move the ethanol through 75 feet of 1/4” diameter tube in the
cabinet and the cold heat exchanger and interconnecting plumbing the cold side
fluid circuit does not present an unacceptable large efficiency degradation.

4.2 Prototype Performance
The strategies used to measure the performance of the prototype are similar to the
performance measurements used for the pre-prototype and described in Sec. 3.2.1.
The constant-head flow system was used for the exhaust side fluid circuit, that
circulated water with anti-corrosion additives. On the load side, the ethanol was

circulated with a model EGA152-0024 pump made by Micropump5 . The flow rate
of the ethanol was measured with a stopwatch and a scale. Since the heat capacity
of ethanol changes significantly with temperature, this was accounted for the in
LabVIEW data-acquisition/processing program when using (3.2.2). The correla-
tion used to determine the volumetric heat capacity is based on thermophysical
data of ethanol from White (1991) and given in units of [J/m3 -K] as

Cp = 17.70T 2 + 3899.6T + 1.8040x106 (4.2.1)

where T is in units of ◦ C. Like the exhaust circuit, two thermistor probes were
used to measure the inlet and outlet temperatures of the ethanol flow.
Because the LVDT discussed in Sec. 3.2.1 was mechanically problematic (the
core rubbed in the sleeve) it was abandoned in favor of a microphone that sensed the
pressure outside of the bellows but inside of the pressure vessel. This “backside”
pressure can be used to determine the stroke since the volume of that space is
known. There is also about a 0.5◦ phase offset (at 100 Hz) between the pressure
and the stroke due to thermoviscous attenuation on the surface in that backside
space. Due to the absence of flexing leads, a microphone is a good solution for
determination of piston motion. The back volume can be found experimentally at
low amplitude by mounting an accelerometer on the piston to measure the stroke
directly while the pressure in the backside is simultaneously measured with the
microphone. The phase correction is a calculated quantity, however, based on an
estimate of the surface area in the backside volume.
The LabVIEW program was modified from its pre-prototype configuration to
accommodate the ethanol flow circuit and other instrumentation changes. Nine
thermocouples were again placed in the regenerator at three azimuthal positions
and in three equally spaced planes along the acoustic axis. These were a big help
to diagnose the intra-regenerator streaming problems found with the use of the
cold head.
The measured performance of the prototype is presented in Table 4.3 and com-
pared to the predictions of a complete DeltaE model that is similar to the one
described in Sec. 3.1.6 but extended to include the linear motor, a detailed model
Micropump Inc., 1402 NE 136th Avenue, Vancouver, WA 98684-0818.

Quantity Unit Measured DeltaE

Tload C -24.6 ±0.5 input

Tex C 33.9 ± 0.5 input
Acoustic Pressure kPa 50.81± 0.05 input
Multiplication Ratio % 8.48 ± 0.25 8.14
Acoustic Power W 125 ± 3 123
Electrical Power to Motor W 147 ± 1.5 155
Useful Cooling Capacity W 119 ± 1.2 114
Exhaust Heat Flux W 266 ± 2.5 282
Overall COP 0.81 ± 0.02 0.73
Overall COP/COPc % 19 ± 1 17
Table 4.3. This table shows the measured performance of the prototype and a compar-
ison the performance to the DeltaE model. The first three entries in the table are given
to DeltaE as inputs, while the rest of the entries in the DeltaE column are results of
a convergent DeltaE model. DeltaE predicts poorer performance than is measured
mainly due to conservative motor parameters used in the model.

for the vibromechanical multiplier and several other additional geometrical details.

Measured performance of the prototype at this operating point is well within
the minimum performance targets specified by the sponsor for cooling capacity
and efficiency, although the temperature span measured at this operating point
is not as large as the sponsor specified. This reduced temperature span does not
decrease the specified product temperature since both the cold side helium/ethanol
exchanger and the ethanol/air heat exchange in the storage cabinet required much
smaller temperature differences than assumed by the specification.
During the course of the project, it was decided that a specific model of ice
cream storage cabinet, the Kelvinator BCS41, would be used to demonstrate the
prototype. After careful measurement of the heat load that the BCS41 presents to
the prototype machine (due to heat flux through the insulation), it was determined
that for a required product temperature of -20◦ C (an inflexible target when deal-
ing with ice cream!) the minimum ethanol temperature is -25◦ C and the capacity
requirement for a -20◦ C product temperature is about 80 W taking into consider-
ation heat flux through the ethanol tubing and the heat load that the pump adds
to the system via both p-v work and motor inefficiency (5 watts total). Likewise,
as the project has progressed, the demonstration venues have been restricted and

it seems clear that ambient temperatures will not exceed 30◦ C. This means that
an exhaust side fluid temperature of 34◦ C is not unreasonable.
Since the model agrees so well with the measurements, the details of the model
can indicate in what components the dominant losses occur. Shown in Fig. 4.11,
the flow of acoustic power through each component in the machine is illustrated
with the dissipation in those components represented by a scaled arrow away from
the main flow. Such flow diagrams are called “Sankey diagrams” and in this case
it is useful to illustrate where further effort should be expended to make this
prototype more efficient.
Starting at the left of Fig. 4.11, 160 W of electrical power is drawn from the
amplifier by the linear motor, of which 25.7 W are dissipated by electrical resis-
tance and eddy-currents in the iron core, mechanical hysteresis in the armature
supports and thermoviscous attenuation on the surfaces of the motor. From that
electrical input, the motor converts 134.4 W to acoustic power inside of the bel-
lows. The inner and outer surfaces of the bellows absorbs 14.4 W of this acoustic
power from thermoviscous (mostly thermo) attenuation on the surface. Because
of the convolutions of the bellows, it has about three times the surface area of a
right circular cylinder of the same dimensions, and hence has also three times the
acoustic power loss.
The amount of power delivered to the multiplier from the motor is 120.0 W
and this power is added to 205.7 W that flows out of the regenerator. The sum,
325.7 W, is managed by the multiplier with only 5.2 W of thermoviscous (surface)
loss and hysteresis loss in the Santoprene flexure that seals the multiplier cone.
The 1.6% loss “overhead” for the multiplier is a real indication of its value in this
design. Of the 320.5 W that are delivered to the thermal core, 10.2 W is absorbed
by the surface of the heat exchangers and 34 W is turned into heat on the surface
of the regenerator. This is especially bad, since this waste heat decreases available
cooling capacity. The first term of (3.1.2) is 63 W in this model of the prototype;
that is the work required by the Second Law of Thermodynamics to move 189.6
W of heat from the cold temperature (-24.6 ◦ C) to the exhaust temperature (29.0

C). In other words, a perfect regenerator would still dissipate 63 W of heat (for
the same temperature span) and would be operating at limiting Carnot COP.
Although the Sankey diagram of Fig. 4.11 shows only the flow of acoustic power,
Multiplier Components

Cone Compliance
14.4 W 2.4 W 0.9 W 1.9 W
Resonator Suspension Surface

160.1 W 325.7 W
Electric 120.0 W

1.2 W 63.0 W
Surface Carnot minimum
9.9 W 205.7 W
Suspension Regenerator
14.6 W
Joule loss 7.6 W
Surface 34.0 W 5.8 W
Linear Motor Surface Surface
4.4 W
320.5 W

189.6 W
Thermal load 296.6 W
57.3 W Exhaust
Exhaust Heat
10.4 W
Heat leak

121.9 W
Usable cooling
Load Heat Exchanger

Figure 4.11. This diagram shows the acoustic power flow in the prototype machine for the nominal operating point shown in
Table 4.3. The values shown in this diagram come exclusively from the DeltaE model which has been shown to agree well with
performance at this operating point. Values in red highlight the three largest loss mechanisms: the largest is the assumption
that all internal dissipation must be pumped by the regenerator which decreases net capacity, secondly, regenerator viscous losses

are half of the 2nd Law minimum energy requirement and the surface area of the bellows convolutions is the third largest loss

the arrows at the load heat exchanger show the various heat loads that the the
machine moves up the temperature gradient. The potential cooling capacity at
this operating point is 189.6 W. This value is reduces to 121.9 W by two nuisance
loads: the 57.3 W of internal surface dissipation described in the above paragraph
that must be pumped and 10.4 W of heat flow from the room to the cold side
of the machine. The estimate of heat leak was made by calculating the thermal
resistance of the plastic platforms and the bolts used to close the pressure vessel.
At last, the flow of acoustic energy that leaves the regenerator is fed back to
add to the stream of power that comes from the motor. The surface of the thermal
buffer tube formed by the “windows” in the regenerator platform and the flow
straightening screens dissipate 7.6 W of acoustic power before the feedback path
converges with the power supplied by the linear motor.
Chapter 5


This work describes the evolution and the current state of the art for low-lift
thermoacoustic refrigerators. The overall COP of the B&J prototype is the high-
est reported to date for an electrically driven thermoacoustic refrigerator. This
level of performance (overall COP = 0.8) satisfies the sponsor (a high-volume con-
sumer of such products), who through considerable research into the low efficiency,
commodity-level vapor compression compressors used in ice cream storage cabinets
claims that an overall COP of 0.7 is acceptable.
The research described in this dissertation reduced the size and weight of low-
lift thermoacoustic refrigerators considerably. The table below shows the power
density of various research machines, adjusted for their temperature lift as follows:

Q̇ ∆T
ρQ̇ = (5.0.1)

where V is the volume displaced by the entire machine, Q̇ is the cooling capacity,
∆T is the temperature lift and TC is the temperature of the load from which heat
is being removed by the machine.
As seen in Table 5.1 the B&J prototype is a factor of two more compact than
other electrically driven research prototypes built to date. The power density of
the commercially available vapor-compression system is about a factor of four bet-
ter than the B&J prototype, according to the Embraco1 catalog posted on their
web site. However, this estimate does not take into consideration that the B&J
prototype has over 25% more volume than would be required for it to function;
Embraco, Rua Rui Barbosa 1020 Cep: 89219-901, Joinville Brazil

Machine Capacity Volume ∆T TC ρQ̇

Triton 10 kW 1000 l 16.7 K 281 K 0.6 W/l
SETAC 294 W 47 l 37 K 276 K 0.6 W/l
B&J Prototype 125 W 24 l 60 K 248 K 1.3 W/l
Embraco Mod: EMI45HER 126 W 7l 70 K 248 K 5.1 W/l

Table 5.1. This table shows the normalized power density of three electrically driven
thermoacoustic refrigerators. TRITON and SETAC were both dual-stack machines
driven at the lowest acoustic mode of the resonator. The volume figure in the table
is the volume of the “envelope” of the resonator, which is significantly larger than the
space occupied by gas in the machine. Since these were research machines, their “enve-
lope” volume was not optimally minimized as would be done for a commercial product.

this extra space was included to instrument the prototype and allow for easy dis-
assembly and reassembly. The B&J prototype did meet the sponsor’s requirement
for overall volume and would fit into a standard ice cream freezer storage cabinet.
The considerable effort spent modelling the prototypes described in this dis-
sertation provide a clear path for efficiency improvement, as shown in Fig. 4.11.
Along with gains in efficiency, new regenerators and flexure seals must be created
that are cheaper than the screens and metal bellows used today. The opportunities
that still lie ahead for the technology of thermoacoustics are very exciting, both
scientifically and commercially.

Backhaus, S. & Swift, G. W. (2000), A thermoacoustic-Stirling heat engine:
Detailed study, J. Acoust. Soc. Am., 107(6), 3148–3166.

Beale, William T. (1971), Stirling cycle type thermal device. U.S. Patent No.

Ceperley, Peter H. (1979), A pistonless stirling engine – the traveling wave
heat engine, J. Acoust. Soc. Am., 66(5), 1508–1513.

Chrsyler, G. M. & Vader, D. T. (1994), Electronics package with improved
thermal management by thermoacoustic heat pumping. U.S. Patent No.

de Blok, Cornelis (1998), Thermo-acoustic system. International PCT Pub.
No. WO 99/20957.

Fox, Robert W. & McDonald, Alan T. (1992), Introduction to Fluid
Mechanics (John Wiley & Sons, New York), fourth ed.

Fraden, J. (1993), AIP Handbook of Modern Sensors: Physics, Design and
Applications (American Institute of Physics, New York).

Garrett, Steven L. (1991), Thermoacoustic Life Sciences Refrigerator,
Preliminary Design Study, Lyndon B. Johnson Space Center, NASA LS-10114,
Houston, Texas 77058.

Garrett, Steven L. (1997), High-power thermoacoustic refrigerator. U.S.
Patent No. 5,647,216.

Garrett, Steven L. (2003), Cylindrical spring with integral dynamic gas seal.
U.S. Patent App. No. 2003/0192322.

Garrett, Steven L., Perkins, David K., & Gopinath, Ashok (1994),
Thermoacoustic refrigerator heat exchangers: Design, analysis and fabrication,
in Proceedings of the Tenth International Heat Transfer Conference, vol. 4, pp.

Gedeon, David (1997), Dc gas flows in Stirling and pulse-tube cryocoolers, in
R. G. Ross (editor), Cryocoolers 9 (Plenum Press, New York), pp. 385–392.

Heake, John F. (Dec. 2001), Characterization of a 10-kW Linear
Motor/Alternator For Use In Thermoacoustic Refrigeration, Master’s thesis,
The Pennsylvania State University. Also ARL Technical Report No. TR-02-002.

Hino, M., Sawamoto, M., & Takasu, S. (1976), Experiments on transition to
turbulence in an oscillatory pipe flow, J. Fluid Mech., 75, 193–207.

Idelchik, I. E. (1996), Handbook of Hydraulic Resistance (Begell House, New
York), third ed.

Incropera, Frank P. & DeWitt, David P. (1996), Fundamentals of Heat and
Mass Transfer (Wiley, New York), fourth ed.

Kays, W. M. & London, A. L. (1998), Compact Heat Exchangers (Krieger,
Malibar, Florida), third ed. Third edition copyright 1984. Originally published
in 1955.

Kolin, Ivo (1995), Thermodynamic theory for stirling cycle machine designs:
Special lecture 1, in Proceedings of the Seventh International Conference on
Stirling Cycle Machines.

Merkli, P. & Thomann, H. (1975), Transistion to turbulence in oscillating
pipe flow, J. Fluid Mech., 68, 567–575.

Organ, A. J. (1992), Thermodynamics and Gas Dynamics of the Stirling Cycle
Machine (Cambridge University Press).

Poese, Matthew E. & Garrett, Steven L. (2001), Thermoacoustic-Stirling
model refrigerator, J. Acoust. Soc. Am., 109(5), 2404.

Poese, Matthew E., Smith, Robert W. M., Wakeland, Ray S., & Garrett,
Steven L. (2003), Compliant enclosure for thermoacoustic device. U.S. Patent
App. No. 2003/0192323.

Sier, Robert (1995), Rev. Robert Stirling D.D.: Inventor of the Heat
Economiser and Stirling Cycle Engine (L A Mair).

Smith, R. W. M., Garrett, S. L., Keolian, R. M., & Corey, J. A. (1997),
High efficiency 2-kW thermoacoustic driver, in Collected Papers, Proc. 137
Meeting ASA and 2nd Conv. European Acoustic Assn., pp. 385–392.

Smith, Robert W. M. (Dec. 2001), High Efficiency Two Kilowatt Acoustic
Source for a Thermoacoustic Refrigerator, Master’s thesis, The Pennsylvania
State University. Also ARL Technical Report No. TR-01-001.

Smith, Robert W. M., Poese, Matthew E., Wakeland, Ray S., & Garrett,
Steven L. (2003), Thermoacoustic device. U.S. Patent No. 6,725,670.

Swift, G. W. (1988), Thermoacoustic engines, J. Acoust. Soc. Am., 84(4),

Swift, G. W. & Ward, W. C. (1996), Simple hamonic analysis of regenerators,
J. Thermophys. Heat Transfer, 10(4), 652–662.

Swift, Greg (2002), Thermoacoustics: A Unifying Perspective for Some
Engines and Refrigerators (Acoustical Society of America []).

Urieli, Israel & Berchowitz, David (1984), Stirling cycle engine analysis
(Bristol : A. Hilger).

Wakeland, Ray Scott (2000), Use of electrodynamic drivers in thermoacoustic
refrigerators, J. Acoust. Soc. Am., 107(2), 827–832.

Wakeland, Ray Scott (2002), Influence of velocity profile nonuniformity on
minor losses for flow exiting thermoacoustic heat exchangers (L), J. Acoust. Soc.
Am., 112(4), 1249–1252.

Ward, Bill & Swift, Greg (1996), DeltaE: Design Environment for
Low-amplitude ThermoAcoustic Engines: User’s Guide and Tutorial, (Los
Alamos National Laboratory, LA-CC-93-8). Note: the DeltaE softaware
(Version 5.1) and the User’s Guide (Revision 6/1/2001 ) are now available on
the CD-ROM included with Swift’s text, Thermoacoustics: A Unifying
Perspective for Some Engines and Refrigerators (Acoustical Society of America
[], 2002), or online at

Ward, W. C. & Swift, G. W. (1994), Design environment for low amplitude
thermoacoustic engines (DeltaE), J. Acoust. Soc. Am., 95, 3671–3672.

Wheatley, John C., Swift, Gregory W., & Migliori, Albert (1986), The
natural heat engine, Los Alamos Science, 14, 2–33. LAUR 86–2699.

White, Frank M. (1991), Heat and Mass Transfer (Addison-Wesley, New York).
Matthew E. Poese
Matt Poese completed his Bachelor’s Degree in Mechanical Engineering at the University of
Cincinnati. Before coming to graduate school at Penn State, he worked with Deloitte and Touche
Management Consulting and at Bavarian Motor Works in Munich, Germany. Before embarking
on the work described in this dissertation, he completed a Master’s Degree in Acoustics making
measurements of high amplitude effects in a small stack-based thermoacoustic refrigerator.

Refereed Scientific Publications
• Gabrielson, T. B., Poese, M. E., Atchely, A. A., “Acoustic and vibration background noise
in the collapsed structure of the World Trade Center,” J. Acoust. Soc. Am., 113(1), 45
• Poese, M. E., Garrett, S. L., “Introduction to thermoacoustic engines and refrigerators,”
Proc. ASME Noise Control and Acoustics, R. Marboe ed., 28, #23517 (2001).
• Poese, M. E., Garrett, S. L., “Performance measurements on a thermoacoustic refrigerator
driven at high amplitudes,” J. Acoust. Soc. Am., 107(5), 2480-2486 (2000).
• Poese, M. E., Smith, R. W. M., Wakeland, R. S., Garrett, S. L. “Compliant Enclosure for
Thermoacoustic Device” U.S. Patent App. No. 2003/0192323.
• Smith, R. W. M, Poese, M. E., Garrett, S. L., Wakeland, R. S. “Thermoacoustic Device”
United States Patent No. 6,725,670
Invited Papers
• Poese, M. E., Gabrielson, T. B., “Acoustic and vibration systems for urban search and res-
cue at the World Trade Center collapse,” Given at the Joint Undersea Warfare Technology
Conference, San Diego, CA (2002).
• Poese, M. E., Garrett, S. L., “Introduction to thermoacoustic engines and refrigerators,”
Given at International Congress and Exposition on Mechanical Engineering 2001, New
York, NY (2001).
Contributed Papers
• Poese, M. E., Smith, R. W. M., Garrett, S. L., “Regenerator-based thermoacoustic re-
frigerator for ice-cream storage applications,” J. Acoust. Soc. Am., 114(4), Pt. 2, 2328
• Poese, M. E., Garrett, S.L., “Performance of a small, low-lift regenerator-based thermo-
acoustic refrigerator,” J. Acoust. Soc. Am., 112(5), Pt. 2, 15 (2002).
• Poese, M. E., Garrett, S. L., “Thermoacoustic-Stirling Model Refrigerator,” J. Acoust.
Soc. Am., 109(5), 2404 (2001).
• Poese, M. E., Garrett, S. L., “Effects of gas mixture on a thermoacoustic refrigerator driven
at high amplitudes.” J. Acoust. Soc. Am., 105(2), Pt. 2, 1012 (1999).
• Poese, M. E., Garrett, S. L., “Performance measurements on a thermoacoustic refrigerator
at high amplitudes,” Proc. 16th Int. Congress Acoust. and 135th Meeting Acoust. Soc.
Am., Kuhl and Crum, eds., 809 (1998).
• Poese, M. E., Garrett, S. L., “Magnetodynamic fluid pump,” J. Acoust. Soc. Am., 100(4),
2808 (1996).