Users Guide
Users Guide
LA-CC-01-13, LA-CC-16-053
December 4, 2017
For the latest version, visit www.lanl.gov/thermoacoustics
(moving to www.lanl.gov/org/padste/adeps/materials-physics-
applications/condensed-matter-magnet-science/thermoacoustics/
Contents
Contents 4
1 Introduction 1
1.1 What DeltaEC does . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Users Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Obtaining DeltaEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Installing DeltaEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.6 Copyright . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
PART I: TUTORIAL 9
4
3.1.2 Legs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.1.3 Phasor plotting options . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2 Incremental plotting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.1 One-dimensional incremental plots . . . . . . . . . . . . . . . . . . . 45
3.2.2 Two-dimensional incremental plots . . . . . . . . . . . . . . . . . . . 47
3.2.3 Rewinding incremental plots . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.4 Incremental plots when more than one model is run . . . . . . . . . . 49
4 RPN math 50
4.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.2 Non-standard results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3 List Linkages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.4 Non-standard inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.5 Non-standard targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.6 Changing an integration variable . . . . . . . . . . . . . . . . . . . . . . . . 56
4.7 Two examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.8 Non-standard guesses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.9 Linking multiple models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.10 Other math segments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
13 Solids 256
13.1 Ideal solid (ideal) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.2 Copper (copper) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.3 Nickel (nickel) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.4 Stainless steel (stainless) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.5 Molybdenum (molybdenum) . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
13.6 Tungsten (tungsten) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
13.7 Kapton (kapton) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
13.8 Mylar (mylar) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
13.9 Celcor (celcor) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
13.10User-de…ned solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
14 Derivations 261
15 Symbols 270
Bibliography 274
i
1. Introduction
1
Figure 1.1: Driven, lossy plane-wave resonator.
2
Figure 1.5: Standing-wave thermoacoustic refrigerator (TALSR style [3]).
Figure 1.6: John Wheatley’s heat-driven refrigerator, a¤ectionately called “the beer cooler”in the
early 1980s [4].
3
Figure 1.7: Thermoacoustic-Stirling hybrid engine [5].
4
Figure 1.9: Resonant self-circulating heat exchanger with one gas diode. [8, 9]
5
of the components of the mixture, inverted to …nd the light mole fraction as a function of
position. Modeling of acoustics superimposed on steady ‡ows is also supported.
With its multi-parameter shooting method to satisfy a variety of mixed boundary con-
ditions, DeltaEC gives the user considerable freedom in choosing which variables are com-
puted as “solutions.”For example, in a simple acoustic resonator (the …rst example below),
DeltaEC can compute the input impedance as a function of frequency, or the resonance
frequency for a given geometry and gas, or the length required to give a desired resonance
frequency, or even the concentration in a binary gas mixture required to give a desired
resonance frequency in a given geometry.
DeltaEC includes few of the nonlinear e¤ects that arise at high amplitudes, so be cau-
tious using it when Mach numbers or Reynolds numbers are high. The principal exceptions
to this rule are the turbulence algorithm in ducts and cones (introduced brie‡y in Chap-
ter 2 and described in detail in Section 10.1.1) and the minor-loss segment, which models
lumped-element dissipation where pressure di¤erence is proportional to the square of the
velocity (introduced brie‡y in Chapter 6 and described in detail in Section 10.2.4).
6
DeltaEC, so there may be some formatting irregularities. In Chapter 2, screen shots of the
user interface are often shown, but in later chapters most examples are displayed as they
appear in a text editor after DeltaEC has saved them to the disk. This keeps the Users
Guide <.pdf> …le size smaller.
We assume that the reader of the Users Guide is very comfortable with basic linear
acoustics [11, 12] and reasonably familiar with thermoacoustics. The choices of which vari-
ables to regard as independent and integrable re‡ects the philosophy of Ref. [13], extended
here to accommodate mixture separation and superimposed steady ‡ow. We use variables
as de…ned, for example, in the lists of symbols in Refs. [13] and [4], and Chapter 15.
7
DeltaEC can be uninstalled via the Windows Control Panel, Add or Remove Pro-
grams; or by running <d:nProgram Files(x86)nDeltaECnD-uninst.exe>. To install a
new version of DeltaEC, it should not be necessary to un-install a previous version. If a
re-installation of DeltaEC does not work as it should, even after uninstalling the previous
version, look for ghost processes named <DeltaEC.exe> in the Windows task manager, and
kill them; or reboot your computer to kill such ghosts. (If you can reproducibly create such
ghosts, please let us know how, so we can prevent them in future versions. We have never
yet seen a ghost in versions 6.3 or 6.4, so we hope we’ve exorcised them forever.)
1.5. Acknowledgments
The development of DeltaE and DeltaEC has been supported in part by many agen-
cies and entities: Tektronix Corporation, SPAWAR, the Naval Postgraduate School, ONR,
Praxair, and, most importantly, by several o¢ ces of the Department of Energy: Advanced
Industrial Concepts, Materials Science (a part of Basic Energy Sciences in the O¢ ce of
Science), the Technology Transfer Initiative O¢ ce, and the O¢ ce of Fossil Energy. Lo-
cal support at Los Alamos has also been provided, via our Industrial Partnership O¢ ce
(later called the Technology Transfer Division; now the Feynman Center for Innovation),
our Locally Directed R&D program, and our local management in the Materials Physics
and Applications Division and the Advanced Engineering Technology Division here at Los
Alamos National Laboratory.
In the 1990s, a long discussion with Pat Arnott helped us de…ne the initial scope of this
work, and comparisons with the results of parallel-plate-stack codes (written by Al Migliori
and Dick Martin) were useful in the early stages. Suggestions by Kim Godshalk, Charles
Jin, Tom Ho‡er, Je¤ Olson, Scott Backhaus, Vince Kotsubo, and Jalal Zia led to signi…cant
improvements in capability and usability through the years. The TRITON project-kicko¤
meeting at Penn State motivated the development of the RPN segment. Charles Jin, Ray
Radebaugh, and the code regen3.1 were indispensable in development of the stacked-
screen algorithm. Scott Backhaus, David Gardner, Matt Poese, and Steve Garrett have
helped shape the DeltaEC user interface.
1.6. Copyright
This Software was produced under a U.S. Government contract (DE-AC52-06NA25396) by
Los Alamos National Laboratory, which is operated by the Los Alamos National Security,
LLC (LANS) for the U.S. Department of Energy’s National Nuclear Security Administra-
tion. The U.S. Government is licensed to use, reproduce, and distribute this Software.
Permission is granted to the public to copy and use this Software without charge, provided
that this Notice and any statement of authorship are reproduced on all copies. Neither the
Government nor the LANS makes any warranty, express or implied, or assumes any liability
or responsibility for the user of this Software.
8
Part I: Tutorial
9
2. Acoustics & user interface
In this Chapter we use the simplest acoustic segments, especially ducts and cones, to intro-
duce DeltaEC’s most basic acoustics and user-interface features.
dp1 i! m
= U1 ; (2.2)
dx A
dU1 i!A
= p:
2 1
(2.3)
dx ma
This point of view is taken in Ref. [13]. The dp1 =dx equation is derived from the momentum
equation of ‡uid mechanics and the dU1 =dx equation is derived from the continuity equa-
tion of ‡uid mechanics. In this form, the equations are ready for simultaneous numerical
integration along the axial position coordinate x to generate solutions p1 (x) and U1 (x):
DeltaEC uses more complicated momentum and continuity equations that include ad-
ditional e¤ects such as dissipation of acoustic power along the sides of ducts. It uses di¤erent
equations in di¤erent segments to suit local circumstances. For example, in boundary-layer
10
approximation in large-diameter ducts and shallow cones, the governing equations are
1
dp1 i! m 1 i
= 1 U1 ; (2.4)
dx A 2 A
dU1 i!A 1 i 1
= 2
1+ p1 : (2.5)
dx ma 2 A1+ s
where A is the cross-sectional area, is the perimeter, and are the viscous and thermal
penetration depths, respectively, is the ratio of isobaric and isochoric speci…c heats, and
s is a correction for thermal properties of the solid wall that is usually negligible. The
equations for each DeltaEC segment are given in Chapters 10 and 11.
DeltaEC uses continuity of p1 and U1 to pass from the end of one segment to the
beginning of the next. Within each segment, wave propagation depends on local parameters
such as area and perimeter as well as on global parameters such as frequency. Although
DeltaEC uses analytic solutions to the governing equations for some of the simplest segment
types, it usually must integrate the equations numerically, so it is generally correct to imagine
DeltaEC beginning at the beginning of a series of segments and numerically integrating
“a momentum equation”Fmomentum ; such as Eq. (2.2) or (2.4), and “a continuity equation”
Fcontinuity ; such as Eq. (2.3) or (2.5), through each segment, sequentially, to the end of the
series of segments, always using the local values of variables such as area, perimeter, and p1
and U1 themselves:
dp1
= Fmomentum (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, solid properties, etc.),
dx
(2.6)
dU1
= Fcontinuity (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, solid properties, etc.),
dx
(2.7)
where most variables on the right-hand sides of these equations are functions of x:
The solution p1 (x), U1 (x) is only determined uniquely if four real boundary conditions
are imposed, because the governing equations can be expressed as two coupled …rst-order
equations in two complex variables or as four coupled …rst-order equations in four real
variables. This is true whether considering a single segment or a one-dimensional series of
segments with each joined to its neighbor(s) by continuity of p1 and U1 . If all four boundary
conditions are given at the initial end of the apparatus (i.e., if we know the magnitude and
phase of p1 and the magnitude and phase of U1 at the initial end) then the integration is
straightforward, proceeding from the initial end to the …nal end. But usually one or more
boundary conditions are instead given at the …nal end. In such circumstances DeltaEC
uses a shooting method,1 by guessing any unknowns among the four numbers de…ning p1
1
Precisely speaking, DeltaEC forms a system of nonlinear equations from the model using the guesses
and targets that the user selects, and manipulates the guesses to make the integration results match
the targets. The routine incorporated in the code is called DNSQ, and it is part of the SLATEC
Common Mathematical Library, which is freely available through the internet software repository at
“http://www.netlib.org.” Its algorithm is a modi…cation of the Powell hybrid method. See Chapter 8 for
details.
11
Figure 2.1: A simple open-topped bottle for the …rst example, <bottle0.out>. (a) Conventional
representation. (b) DeltaEC representation.
and U1 at the initial end of the integration, integrating to the other end, comparing the
results with the target boundary conditions imposed at that other end or elsewhere, and
adjusting its guesses until the integration results meet the targets.
One of DeltaEC’s most powerful features is that the guesses are not limited to the
conventional choices consisting of real and imaginary parts of p1 and U1 (or, equivalently,
magnitudes and phases) at the initial end of the numerical integration along x: Any variables
that have an e¤ect on the downstream target variables can be used. This enables DeltaEC
to calculate a resonance frequency, a geometrical dimension, a temperature, or even the
concentration in a binary gas mixture to satisfy given boundary conditions.
12
1.7780E-02 c Length m 0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 2 ---------------------------------
CONE Transition from neck to volume
2.1410E-04 a AreaI m^2 0.0000 A |p| Pa
5.1870E-02 b PerimI m 0.0000 B Ph(p) deg
0.1003 c Length m 0.0000 C |U| m^3/s
1.8680E-03 d AreaF m^2 0.0000 D Ph(U) deg
0.1532 e PerimF m 0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 3 ---------------------------------
DUCT Bottle volume
1.8680E-03 a Area m^2 0.0000 A |p| Pa
0.1532 b Perim m 0.0000 B Ph(p) deg
0.1270 c Length m 0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 4 ---------------------------------
SURFACE Bottom End
1.8680E-03 a Area m^2 0.0000 A |p| Pa
0.0000 B Ph(p) deg
0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
Several features of DeltaEC model …les are illustrated here. Model …les should be named
<something.out>. These …les consist of a set of “segments” whose order and format are
important. After a TITLE (with an optional heading where the user can type notes), the
initial (or “zeroth”) segment is always a BEGIN segment. This is one of the “logistical”
segments of DeltaEC, usually having no actual geometrical correspondence to parts of the
hardware being modeled. Subsequent segments describe the geometry and other properties
of the “physical”hardware parts of the acoustic system— here, DUCTs, a CONE, and a SURFACE.
All units are MKS. The number and order of data in each segment is crucial. Within each
line, the …rst number (e.g., “1.e5” or “100.”) or word (e.g., “helium” or “BEGIN”) is the
most important to DeltaEC as input.2 Subsequent …elds on each line give the units or
name of the variable whose value appears. Lines that begin with “!” are comments with
no impact on the integrations. Numbers can be in …xed or exponential format.3 Segment
names are stored uppercase, and only the …rst …ve characters are interpreted.
The column of zeroes on the right side shows where DeltaEC’s calculated results will
appear. At the moment, they are all zero because DeltaEC has not yet calculated anything.
BEGIN sets the stage, in this case, with 1-bar room-temperature air being driven at 300
Hz with a pressure amplitude of 1 Pa and a volume ‡ow rate amplitude of 0.0001 m3 /s,
in phase with the pressure oscillation. Some of these variables, such as frequency, extend
unchanged through all subsequent segments. Other variables, such as pressure amplitude,
are initial conditions for DeltaEC’s numerical integration through subsequent segments.
p A lossy duct segment, DUCT, comes next. Here, we have made the perimeter equal to
4 area, to make this a circular duct. A CONE and another DUCT come next, completing
the shape of the bottle. The …le ends with a SURFACE to account for oscillatory-pressure
thermal-hysteresis dissipation at the bottom.
2
Exceptions: The …le reader looks beyond the …rst …eld of characters to resolve ambiguity about optional
parameters, described later.
3
In some computers, integers must be followed by a decimal point. Also, some computers require the last
line in the …le to be followed by an end-of-line character, before the end-of-…le character occurs.
13
2.3. Running DeltaEC
As children, most of us learn that blowing across the mouth of a bottle produces a pleasant
audible tone. With this example, the new user of DeltaEC will obtain the frequency of
that tone, while exploring the user interface and learning to “run”the numerical integration.
Open the …le <bottle0.out>, using whichever of these three methods you prefer:
1. Open the folder <...nExamplesnAcoustics> wherever your installation has created
it, and double click on the …le <bottle0.out>. If this folder is write-protected from
changes, SaveAs to an unprotected folder immediately.
2. In Windows, …nd DeltaEC through the Start button at the bottom-left corner of
your desktop, via Programs and the DeltaEC group. Double click on the program
to launch it. Then open the …le <bottle0.out> via the DeltaEC Files pulldown
menu or by clicking the icon that looks like a manila folder. In either case, …nd your
way to the <ExamplesnAcoustics> folder in the usual way.
3. Double click on the Desktop DeltaEC icon to launch DeltaEC, and open the …le
<bottle0.out> as described in method 2.
After opening <bottle0.out>, DeltaEC displays the list of segments, like this:
Clicking on the “+” next to any segment’s name exposes that segment’s details. For
example, to see the dimensions of the neck, click on the “+”next to 1 DUCT:
14
Blue text (numbers or words) means you can double click or right click to make a change
or see more options. Italic red numbers indicate what will be a result, as soon as the
calculation is run, but right now the red color indicates that these numbers are probably
inconsistent with the blue input numbers. Green numbers that you will see shortly indicate
good results: calculated by DeltaEC and consistent with all the other numbers in the
model.
Now show all the segments in the model, by clicking on the “+”for each segment or on
the big “+= ”on the toolbar. What you see should look like this:4
TITLE Model of 1982 Penn State Championship Bottle
!--------------------------------- 0 ---------------------------------
BEGIN The mouth
1.0000E+05 a Mean P Pa
300.00 b Freq Hz
300.00 c TBeg K
1.0000 d |p| Pa
0.0000 e Ph(p) deg
1.0000E-04 f |U| m^3/s
0.0000 g Ph(U) deg
air Gas type
!--------------------------------- 1 ---------------------------------
DUCT the neck
2.1410E-04 a Area m^2 0.0000 A |p| Pa
5.1870E-02 b Perim m 0.0000 B Ph(p) deg
1.7780E-02 c Length m 0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Hdot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 2 ---------------------------------
CONE Transition from neck to volume
2.1410E-04 a AreaI m^2 0.0000 A |p| Pa
5.1870E-02 b PerimI m 0.0000 B Ph(p) deg
0.1003 c Length m 0.0000 C |U| m^3/s
1.8680E-03 d AreaF m^2 0.0000 D Ph(U) deg
0.1532 e PerimF m 0.0000 E Hdot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 3 ---------------------------------
DUCT Bottle volume
1.8680E-03 a Area m^2 0.0000 A |p| Pa
0.1532 b Perim m 0.0000 B Ph(p) deg
0.1270 c Length m 0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Hdot W
ideal Solid type 0.0000 F Edot W
(etc.)
The BEGIN segment establishes that the bottle is full of air at atmospheric pressure and
room temperature, and it launches a 300-Hz wave into the bottle with a pressure amplitude
of 1 Pa and a volume-‡ow-rate amplitude of 10 4 m3 /s. (We will examine its “optional”
parameters in later chapters, when we encounter them.)
Now, run the …le, by typing ‘r’, by clicking the blue “Run”button, or by accessing “Run”
in the Tools pulldown menu. (To encourage users to get in the habit of using the keyboard
shortcuts, the Users Guide summarizes all such options in the briefest possible way, with
the shortcut letter underlined, e.g., run. Infrequently used operations whose shortcuts the
user has not memorized can be found by browsing the pulldown menus.)
Upon running, the Run Monitor appears at the bottom of the display, indicating with
its cheerful green color and the word “Success” that the run has succeeded, in this case
4
Please ignore minor formatting di¤erences between what you see on your computer screen and what is
present in the Users Guide. Much of the Users Guide was written before the user interface was complete,
and it’s much easier for us to let text-formatted results page-break automatically in the Users Guide than
to manage large …gures that show the actual computer-screen display. Text-formatted results also minimize
the Users Guide …le size.
15
taking only a fraction of a second. (Typing “b”toggles the Run Monitor display o¤ and on,
and resets it to default size.) Now the results in the right column of the model display are
nonzero, and they have changed from red italic to green, indicating that they are consistent
with one another and with the blue input numbers:
DeltaEC has integrated through the model and calculated some results, in particular p1 (x)
and U1 (x): The results in the right column of the <.out> …le show values for p1 (x) and
U1 (x) and other results at the end of each segment. If a few numbers are all you want to
see, you are done!
One of the other results is the acoustic power Edot,
_ 1 h ei
E = Re p1 U1 ; (2.8)
2
16
2.3.1. Font augmentation
You can change the size of the font in DeltaEC’s main display and many secondary win-
dows by spinning your mouse wheel while holding down the Ctrl key, or by using Ctrl n,
Ctrl b, and Ctrl /. (Be sure the computer’s focus is on the DeltaEC window— click it if
necessary.) Resizing the DeltaEC window by dragging its perimeter with the mouse, or
making DeltaEC …ll the entire screen by clicking the usual Windows “Maximize” button
in the top-right corner, makes more room for a bigger font.
The shades of blue, green, and red used to indicate di¤erent types of variables seem to
show up well for color-blind users.
Numbers are lined up in columns. The initial columns are leg number (relevant when
TBRANCH and/or multiple BEGIN segments are used; described in Section 11.2.1), seg-
ment number, and x.
DeltaEC writes Nint + 1 lines of data for each segment that it integrates (e.g., ducts,
cones, and stacks). Nint is the number of Runge-Kutta integration steps— see Section
8.1 for details.
17
Figure 2.2: State plots of p1 and U1 for <bottle2>.
DeltaEC writes two lines of data for lumped elements (e.g., SURFACE) and other short
segments that do direct end-to-end calculations (e.g., most heat exchangers): one line
before and one line after the segment is computed.
Logistical segments (e.g., BEGIN) that do not correspond to any hardware parts gen-
erate nothing in the <.sp> …le.
The …le <bottle2.ssv> instructed DeltaEC’s plotter that we would like to see graph-
ical plots of p1 (x) and U1 (x) after each run. These graphs, shown in Fig. 2.2, will automat-
ically refresh after each new run, as the user makes changes to the model.
The real parts of p1 and U1 show the values of oscillating pressure and oscillating volume
‡ow at times !t = 0; 2 ; 4 ; ... and the imaginary parts show their values at times
!t = =2; 2 =2; 4 =2; ...
DeltaEC’s plotter lets the user select other plots easily. The time-averaged product
of p1 and U1 is acoustic power E, _ which is also interesting. To change the display for this
_
model to E vs x; check the y box under Edot, and uncheck the Re[p], Im[p], Re[U], and
Im[U] boxes, to obtain Fig. 2.3. Chapter 3 explains much more about DeltaEC’s plot
capabilities.
Now use Save Copy As to save this …le with a new name, <bottle3.out>. Examine
the resulting …le with a text editor. Note that the output results giving p1 and U1 at the
ends of the DUCT, CONE, DUCT, and SURFACE are now part of the …le. The fact that we want
to plot E_ vs x from now on has also been included in the associated <bottle3.ssv> …le.
18
Figure 2.3: E_ vs x for <bottle2>.
bottle that is supposed to be sealed at the bottom. In this section we introduce DeltaEC’s
shooting method to resolve this issue, allowing boundary conditions at the end of a model
(or anywhere else in a model) to be “targeted”to desired values.
Start DeltaEC again, opening …le <bottle2.out>. As a matter of good habit, run it,
and check some plots for sanity (your own sanity, as well as DeltaEC’s). Now open the
“Guesses and Targets”summary display box. You can open this box through the pull-down
menus by clicking on Display, then Guesses Targets. Alternatively, just type “g”(the “g”
stands for “guesses”). The guess–target summary should look like this:
The table is blank: At this point DeltaEC has no guesses and no targets. Each time we
have run DeltaEC thus far in the tutorial, it has integrated along x only once, starting
from the values of p1 and U1 that we gave it in the BEGIN segment.
To set up this model with appropriate targets, we need to insert an additional segment
at the end of the model. To insert a segment, pull down the Edit menu and click on Insert
Seg, or type “i”, or right click below the last segment and select Append. Whichever way
19
Figure 2.4: The bottle. (a) HARDEND untargeted, so U1 can be nonzero at the bottom. (b) HARDEND
targeted to (0.00, 0.00), to seal the bottom.
you choose to do it, insert a HARDEND, at the end of the model, after the SURFACE.
Now establish 5a and 5b in the HARDEND as targets by clicking on Possible targets in
the HARDEND itself or through Add Targ in the guess–target summary. Inspect your work
thus far by examining the guess–target summary:
Also notice how the targets are ‡agged with yellow markers in the main display.
(We could have programmed HARDEND segments with the real and imaginary parts of U1
as our targets, but experience has shown that the DeltaEC shooting-method algorithm
usually works without extra attention from us if we normalize the targeted U1 by dividing
by area and p1 : The rare occasions when more attention is needed are described in Section
8.3.2.)
We have two targets, which will soon enforce complex U1 = 0 at the bottom end of
the model, appropriate for the sealed end of the bottle. We must choose two guesses that
20
DeltaEC can adjust to meet those targets.5 Basically, DeltaEC integrates from BEGIN
to HARDEND. DeltaEC will re…ne whatever two guess variables we allow it to vary, to …nd
a solution to this acoustics problem that arrives at the HARDEND with zero complex volume
‡ow rate. What to choose as these guesses depends on exactly what situation you are trying
to model. Do you have in mind a resonator driven by a …xed volume ‡ow rate amplitude,
or one that is driven by a …xed pressure amplitude? Neither answer is wrong; it depends on
the circumstances.
In the present example, we arbitrarily choose to keep jU1 j …xed at 0.0001 m3 /s, and let
DeltaEC adjust p1 : Begin to accomplish this by adding 0d and 0e as guesses, by double
clicking or right clicking on their values and selecting “guess,”or through Add Guess in the
guess–target summary. Inspect the resulting situation by looking at the summary.
The target values are red italic, indicating that DeltaEC does not know whether results
are consistent with input variables; it knows that it has not yet run with the most recent
changes of input variables. The guesses (1 Pa, 0 degrees) are still just the values that are
left over from before. DeltaEC has not yet adjusted the guesses to try to hit the targets,
but it is ready to do so, because there are equal numbers of guesses and targets: two each
5
Users sometimes ask why we need two targets and two guesses when we are really only trying to impose
one constraint at the HARDEND, namely jU1 j = 0: The answer to this question can be appreciated by imagining
a blindfolded archer trying to hit a bull’s-eye. If an observer tells the archer “Your last attempt missed by
2.5 meters,”which way is the archer supposed to revise her aim for her next attempt to hit the target? More
useful information is conveyed if the observer tells the archer “Your last attempt struck 1.5 meters above
the bull’s-eye and 2.0 meters to the left of the bull’s-eye.” Like the blindfolded archer’s target, DeltaEC’s
HARDEND is a point target in a two-dimensional space, namely the complex plane of U1 . DeltaEC needs to
adjust its aim (its guesses) in two dimensions to hit the target, and DeltaEC can hit the target in a small
number of attempts if it gets two-dimensional information about how far o¤-target it is.
21
in this case.
With an equal number of guesses and targets, let’s run. After running, DeltaEC’s
Run Monitor at the bottom of the display reports that it integrated along x 28 times while
adjusting guesses and coming ever closer to the targets, and it spent 0.08 seconds doing so.
(The numbers that are reported by the Run Monitor may vary from computer to computer
and version to version. How close to the targets is judged to be “close enough”is described
in Section 8.3.2.)
Inspect the results:
The value of the phase shown in 0e, 7288.6 ; might seem strange until we realize that
DeltaEC does not understand that integer multiples of 360 can be subtracted from any
phase angle without consequence. With the initial guessed value of 0d so far from reality,
DeltaEC’s shooting method wandered far, looking for a solution; under normal circum-
stances, we might have given it a more plausible starting point (such as the result of a
previous run). Subtracting 7200 from the value of 0e for the sake of our human sanity and
running again yields
The targets are accurately met, because DeltaEC has adjusted the two guesses to be 31.870
Pa, 88.622 degrees. The state plots in Fig. 2.5 also show that the targets have been met,
because they show that jU1 j = 0 at the end.
Save the <.out> …le with a new …lename of your choosing, and examine it with a text
22
Figure 2.5: State plots for the bottle, with HARDEND targets met so U1 = 0 at the bottom.
editor to see how DeltaEC has indicated which variables are guesses and targets, next to
the guesses and targets themselves and in the restart table at the bottom of the …le.
Most of the thought required to successfully use DeltaEC occurs while trying to …gure
out which of the variables are appropriate targets and guesses, i.e., what constraints must
be imposed on results of a single DeltaEC integration pass, and what variables are needed
as inputs by a single DeltaEC integration pass but are in fact unknown to us when we
ask DeltaEC to run. While many choices are possible, as long as the number of guesses
equals the number of targets, some choices are physically nonsensical and cannot succeed.
For example, asking DeltaEC to try to achieve resonance in a given bottle by guessing
the mean pressure of the air would be futile, because the speed of sound in an ideal gas is
independent of pressure. In the examples in subsequent Chapters, and in advanced use of
DeltaEC, the choice of good target and guess variables is seldom as easy as it is in this
simple example.
Notice the use of columnar location and color in the display: Columns organize variables
according to DeltaEC’s most short-sighted internal needs, while colors organize variables
according to users’needs. The numbers in the left column (ordinary blue numbers, and the
guesses, and the bull’s-eye target values) are variables needed by DeltaEC in each pass of
its integration, as it integrates repeatedly, adjusting the guesses to try to hit the bull’s-eyes.
Variables in the right column are computed as results in each pass, and are displayed green
when targets are met. We users, however, often wish to forget that DeltaEC must adjust
guesses and hit targets. From our perspective, blue numbers are what we give DeltaEC as
known quantities, and green numbers are the results that we want DeltaEC to give us. If
the green numbers are indeed green, DeltaEC feels con…dent of its results, but when those
numbers are red DeltaEC is not displaying a consistent set of results, for any of a number
of reasons— usually because we have not yet run it or because it has missed the targets.
23
Figure 2.6: State plot with HARDEND targeted and resonance achieved, so U1 = 0 at the bottom
and p1 and U1 are in phase at the mouth.
Now let’s …nd the resonance frequency f , which is the frequency at which the bottle
sings when we blow across its top. Look at the guess–target summary to get oriented.
Guess-Target Summary:
GUESS 0d 0e
name BEGIN: |p| BEGIN:Ph(p)
units Pa deg
value 31.870 88.587
TARGET 5a 5b
name HARDE:R(1/z HARDE:I(1/z
value 0.0000 0.0000
result -5.6504E-13 4.1801E-14
First, think about the physics long enough to realize that having p1 and U1 in phase at
the mouth is a suitable, approximate condition for resonance.6 Now make the corresponding
changes in the DeltaEC model: modify 0e to be 0 instead of 41:5 and remove 0e from
the guess list so DeltaEC cannot change it any more. Next, add 0b, the frequency, in the
guess list. Inspect the guess–target summary again. Now, “run” the calculation. Inspect
the results in the main display or the guess–target summary, …nding that the resonance
frequency is 194.99 Hz. Inspect the state plots in Fig. 2.6, to see that the desired boundary
conditions at both ends are met. Save this version as bottle5, for use later in this Chapter.
As acousticians, we know that this resonator has more than one resonance mode. To …nd
another one, modify 0b to be 600 Hz, and run again. With this starting guess, DeltaEC
converges on the 2nd mode, resembling 3/4 of a wavelength from end to end, at 1029.6 Hz.
6
That short sentence is an example of where novice users of DeltaEC who have insu¢ cient acoustics
knowledge can have a great deal of trouble. Please do not expect to learn “everything you need to know”
about acoustics from DeltaEC. Take a university acoustics course or study an acoustics textbook! Re-
member that DeltaEC cannot “understand” physical acoustics better than you do. It merely calculates,
accurately and quickly. N.b.: we said “approximate condition for resonance”because some users might want
to include radiation impedance. See Section 10.4.1.
24
DeltaEC can use any physically appropriate input variable as a guess. We could have
learned what temperature makes the system resonate at 180.00 Hz, by using 0c instead of
0b as a guess. (The answer is 255 Kelvin.) Or, by using 1c as a guess, we could have found
out what length the neck should be to put the resonance at 180 Hz at 300 K.
DeltaEC’s gas library contains three binary mixtures of gases: He-Xe, He-Ar, and
Ne-Xe. These introduce another real variable in BEGIN: the mole fraction of helium in the
mixture (for example, HeAr with nL = 0:889 indicates 88.9% helium and 11.1% argon).
The helium fraction in one of these mixtures can be used as a guess variable for resonance.
Simply add 0j as a guess. Thus, we could learn the helium concentration that would have
to be added to argon in the bottle to make the resonance equal 180.00 Hz.
BEGIN Required immediately after TITLE. The “zeroth”segment must be BEGIN. It contains
global variables such as mean pressure, gas type, and frequency, and initial conditions
for variables that can vary with x; including p1 , U1 ; and mean temperature (whose x
dependence is introduced in Chapter 5). See details in Section 11.1.2.
SURFACE A surface area with pressure-hysteresis loss in its thermal penetration depth. Usu-
ally used at ends of ducts. See details in Section 10.2.2.
DUCT A duct, with viscous and thermal losses at the wall if the perimeter is realistically
nonzero. Separate entry of area and perimeter accommodates ducts of any cross-
sectional shape. Laminar or turbulent. See details in Section 10.1.1.
CONE A cone, often used to adapt between ducts of di¤erent sizes. Uses lossy Webster horn
equation. Laminar or turbulent. See details in Section 10.1.2.
25
IMPEDANCE A lumped-parameter, series impedance. Resistance and inertial reactance are
speci…ed. See details in Section 10.2.3.
IDUCER and VDUCER Current-driven and voltage-driven transducers, with parameters inde-
pendent of frequency, connected “on the side” of the DUCTs or other segments. See
Fig. 2.7. See details in Section 10.4.2.
ISPEAKER and VSPEAKER Current-driven and voltage-driven electrodynamic transducers, pa-
rameterized by mass, B-l product, etc., so impedance coe¢ cients depend on frequency.
Connected “on the side” of a sequence of DUCTs or other segments. See Fig. 2.7. See
details in Section 10.4.2.
IEDUCER and VEDUCER, IESPEAKER and VESPEAKER The four transducers that were described
above, which do not have “E” as the second letter in their names, are connected in
side-branched con…gurations, where the “back side”of the transducer hangs outside of
DeltaEC’s computation space. The four Enclosed versions are their series counter-
parts, with one side of the transducer facing the previous segment and the other side
facing the subsequent segment, so the volume ‡ow rate remains constant across the
segment (except for thermal-hysteresis e¤ects on the surfaces) and p1 usually changes
signi…cantly. See Fig. 2.8. See details in Section 10.3.1.
Figure 2.7: IDUCER, VDUCER, ISPEAKER, and VSPEAKER are attached as side branches to the series
of DeltaEC segments.
Figure 2.8: IEDUCER, VEDUCER, IESPEAKER, and VESPEAKER are acoustically in series with the
adjacent DeltaEC segments.
26
HARDEND A logistical …nal segment, used to enforce U1 = 0 through use of the inverse of the
acoustic impedance in the target list. See details in Section 11.1.3.
SOFTEND A logistical …nal segment. Can be used to set p1 = 0 through use of the acoustic
impedance in the target list. Useful for de…ning a mirror-image plane in symmetric
apparatus with a pressure node at a center of symmetry. Also used for loops and
networks of segments, to mark a location that will be reconnected elsewhere in the
model. See details in Section 11.1.3.
27
natural-gas combustion products (i.e., ‡ue gas), liquid sodium, and eutectic liquid sodium-
potassium. The ideal-gas equation of state is used for most of the gases, while carbon
dioxide, sodium, and eutectic sodium-potassium have their own equations of state.
Solids, described in detail in Chapter 13, include Kapton, Mylar, stainless steel, molyb-
denum, tungsten, copper, nickel, Celcor, and ideal. An ideal solid has extremely high heat
capacity, density, and thermal conductivity.
DeltaEC also allows users to specify “external,” user-de…ned gases or solids that are
not part of its own internal library. Properties are calculated, according to local Tm and pm ,
from coe¢ cients read from a user-written text …le. Up to …ve user-de…ned gases and …ve
user-de…ned solids can be used at one time. Details of user-de…ned pure gases and binary
mixtures are described in Sections 12.15.1 and 12.15.2, respectively. Details of user-de…ned
solids are described in Section 13.10.
To illustrate the use of a single-component user-de…ned gas, consider the example below.
To replace DeltaEC’s internal “ideal” helium gas (see Section 12.1) with a more accu-
rate representation that calculates density and sound speed including the …rst coe¢ cient
of the virial expansion for helium, we can create the following …le in a text editor, call it
<helium.tpf>, and put it in the same folder as our model:
! user-defined gas; He with first virial coeff (B=12cc/mole)
! Equation is in UsersGuide Section 12.15:
! C0 + C1*pm/(T+C2*pm) + C3*T + C4*T^2 + C5*T^C6 + pm^2 *C7*T^C8 + pm*C9
! Density, rho (m^3):
0. 4.814e-4 1.44e-6
! isobaric heat capacity, cp (J/kg-K):
5192.
! Thermal conductivity, k (W/m-K):
0. 0. 0. 0. 0. 0.0025672 0.716
! Square of sound speed, a^2 (m^2/s^2):
0. 0. 0. 3461.92 0. 0. 0. 0. 0. .0100
! Viscosity, mu (kg/m-s):
0. 0. 0. 0. 0. 0.412e-6 0.68014
The coe¢ cients for density in this …le are used by DeltaEC in Eq. (12.1) and were
determined by a user who knew that
pm m
m = ; (2.9)
Runiv (Tm + Bpm =Runiv )
where Runiv = 8:314 J/mole-K, the molar mass m = 0:0040026 kg/mole, and the …rst virial
coe¢ cient B = 1:2 10 5 m3 /mole. Thus, the user set C1 = m=Runiv and C2 = B=Runiv in
the density line in the <.tpf> …le. For squared sound speed, the user knew that
Runiv T Bpm
a2 = 1+2 ; (2.10)
m Runiv Tm
28
2.6.4. Turbulence
A turbulence algorithm can be enabled in DUCT and CONE, by double clicking on “Optional
parameters.” Checking the Turbulence box brings up parameter d in DUCT and parameter f
in CONE, which specify the relative roughness (roughness height divided by pipe diameter).
Set the roughness equal to a small value greater than zero for rough walls. (We usually use
5 10 4 ; even if this is larger than the actual relative roughness of the channel.) To ensure
a laminar calculation, check the laminar box for the segment (which causes the roughness
parameter to be hidden from view in the <.out> …le).
The turbulence algorithm, which is described in detail in Section 10.1.1, follows the quasi-
steady approximation, (the spirit of the assumptions of Iguchi et al., Ref. [15]), assuming
that oscillatory-‡ow losses can be calculated by using the steady-‡ow Moody friction factor
at each instant of time during the oscillatory ‡ow, with the surface roughness used as an
experimentally determined …tting parameter. The quasi-steady assumption has only a little
experimental validation for DUCTs and CONEs in the range of Reynolds number and R= of
interest in thermoacoustics (where R is radius), but we believe it provides a useful estimate,
much better than no estimate at all.
When you add a parameter speci…ed by sameas to the guess list, you must …rst sever
the sameas relationship. This is required because the value at this point will be controlled
by DeltaEC’s shooting method. But if a variable that is the root of one or several sameas
29
references is caused to change by DeltaEC itself in any way, e.g., as a guess, all sameas
references to this root within the model will change with it.
to update the perimeter based on the area, thereby maintaining the cross-sectional shape.
Thus, as area is changed, this link keeps a circular duct circular and maintains the aspect
ratio of a rectangular duct.
Figure 2.9: When the area of a DUCT is changed and perimeter–area linking is in use, the perimeter
is automatically recalculated to maintain the DUCT’s shape.
30
Figure 2.10: Illustration of some master–slave linking. (a) Two lengths can be linked so the total
length remains constant when one length is changed. (b), (c), (d) Perimeters can be linked to
areas so shapes remain constant when areas are changed. (e) In STKSLAB, porosity can be linked to
pore size and solid thickness, to keep the fraction of the stack that’s devoted to support structure
constant. The plane of this illustration is perpendicular to x: (f) In STKCIRC and TX, porosity can
be linked to hole radius, to keep the number of holes constant. The plane of this illustration is
perpendicular to x: (g), (h) In segments with impedances, the imaginary part can be linked to
the real part so either phase or magnitude remains constant when the real part is changed. (i)
In STKDUCT, maintain shape and volume when length is changed. (j) In CONE, adjust length to
maintain taper angle when areas are changed.
Below is a list of all parameter-link options and the segment types for which they are
available, including some for segments that have not yet been introduced in this tutorial.
The numbers in parentheses at the end indicate how the link is recorded in the restart table
at the bottom of the <.out> …le. Figure 2.10 illustrates these links.
Adjust length in segment n when the length here changes, to keep their sum constant: All
segments with length. (In the restart table, this is indicated as n > 0:)
Adjust porosity (i.e., GasA/A) when gas gap or solid thickness changes in STKSLAB, keeping
constant area and constant fraction of the area devoted to support ribs. Adjust porosity
(i.e., GasA/A) when hole radius or area changes in STKCIRC or TX, keeping constant
number of holes. Set porosity (i.e., GasA/A) equal to ab=(a + l)(b + l) in STKRECT. For
details, see Section 10.5.1. (In the restart table, this is indicated as 1:)
Adjust perimeter when area changes, to maintain shape: DUCT and STKDUCT. Adjust initial
perimeter when initial area changes, to maintain initial end’s shape: CONE and STKCONE.
(In the restart table, this is indicated as 2:)
31
Adjust …nal perimeter when …nal area changes, to maintain shape of …nal end: CONE and
STKCONE. (In the restart table, this is indicated as 3:)
Adjust length and both perimeters, to maintain end shapes and taper angle: CONE and
STKCONE. (In the restart table, this is indicated as 4:)
Adjust surface area when volume changes, to maintain shape: COMPLIANCE. (In the restart
table, this is indicated as 5:)
Adjust area and perimeter when length changes, to maintain constant volume and cross-
sectional shape: STKDUCT. (In the restart table, this is indicated as 6:)
Adjust imaginary part when real part changes, to maintain constant magnitude (when
possible): IMPEDANCE, BRANCH, and TBRANCH. (In the restart table, this is indicated as
7:)
Adjust imaginary part when real part changes, to maintain constant phase angle: IMPEDANCE,
BRANCH, and TBRANCH. (In the restart table, this is indicated as 8:)
Adjust initial and …nal perimeters in CONE and STKCONE as initial and …nal areas change,
maintaining the shapes on both ends. (In the restart table, this is indicated as 9:)
Master–slave links can be tracked by status …elds in the center of the <.out> …le,
between the input and output columns. The master parameter— the one through which
the mode is controlled— is labeled Mstr. The slave parameter, i.e., the one that cannot be
modi…ed independently when it is controlled by a link, has a status indicator of the form
nnp to indicate its master’s address (e.g., 2a).
Newly inserted DUCT and CONE segments automatically initialize with their perimeters
linked to their areas on the basis of a circular shape, with a radius of 1 meter.
With each of the <bottle> model’s perimeters slaved to its local area, and adjacent
areas that must be identical connected by sameas, the <bottle> model looks like this:
32
Figure 2.11: (a) The schematic display of the <bottle> model. Dimensions are to scale, assuming
that areas are circular. Right-clicking on any segment brings up a menu of options, similar to
right-clicking on a segment in the main model display. (b) Some phasor diagrams added to the
schematic.
so it is apparent that the bottle has only …ve independent geometrical dimensions— neck area,
bulb area, and the lengths of the neck, cone, and bulb— whose independence is indicated
by their blue color. All other dimensions depend on those …ve. The user who is interested
in studying the resonance frequency as a function of geometry can now change those …ve
dimensions at will, con…dent that self-consistent geometry will be maintained by the master–
slave and sameas links.
33
for U1 . Drag the phasor circle or its anchor point with the mouse to reposition them. In
integrable segments like DUCTs, Nint unique x locations are available; nonintegrable segments
like COMPLIANCE have only one location.
To delete a phasor circle, move the mouse over the anchor point and press the Alt or
Delete key. To delete all phasor circles, close and reopen the schematic window.
DeltaEC models often have many dozens of segments. The user is usually not interested
in examining every numerical result after every run— typically, we examine the guesses (and
perhaps targets) to ensure that DeltaEC has not gotten completely lost, and we examine a
small number of other results of particular interest. The highlights display shows a compact,
user-customized list of such variables of particular interest, which is updated after every run.
Any input or output variables can be added to the highlight list, by double clicking or
right clicking.
Returning to the bottle as an example, we might be particularly interested in the pressure
amplitude at the mouth of the bottle and at the bottom of the bottle, and the resonance
frequency, as we explore di¤erent dimensions of the neck by modifying its area and/or length
and running. A convenient setup for such an investigation looks something like the …gure
below, just after lengthening the neck and running. In ordinary use, we would type ‘m’to
modify one or more parameters, then ‘r’to run, examine the results, and repeat.
34
Figure 2.12: TBRANCH allows DeltaEC to integrate in branched geometries like this.
35
Figure 2.13: A driven, parallel LRC network. (a) Schematic. (b) One way to model the LRC
network in DeltaEC, with the L and C in the TBRANCH and the R in the trunk, rejoined at the
UNION, whose downstream end must be sealed by a targeted HARDEND. (c) Another way to model the
LRC, with the L in the TBRANCH and the R in the trunk, rejoined at the UNION, whose downstream
end feeds the C, whose exit must be sealed by a targeted HARDEND. (d) A third way, with the L,
R, and C all in the TBRANCH. Returning to the trunk, DeltaEC encounters no segment before the
UNION, so the UNION e¤ectively connects the initial and …nal ends of the branch to each other. The
UNION’s exit must be sealed by a targeted HARDEND.
impedance usually determines how the (complex) volume ‡ow rate splits up at the TBRANCH.
Volume ‡ow rates are summed at the UNION. (Note that the UNION targets are somewhat
di¤erent from HARDEND targets discussed above, because “target”numerical values in UNION
are not speci…ed by the user— they vary from run to run, depending on what is happening
at the associated SOFTEND.)
Figure 2.13 and the DeltaEC models below show three ways to use TBRANCH and UNION
to model a parallel LRC network. The reader is encouraged to invent one or more additional
ways to model this network, and try them.
Here is the DeltaEC model corresponding to Fig. 2.13(b):
TITLE Example LRC1
! lrc1.out
!Created@19:30:01 9-Jul-06 with DeltaE Vers. 5.5b6 for the IBM/PC-Compatible
!--------------------------------- 0 ---------------------------------
BEGIN Initialize things
1.0000E+05 a Mean P Pa 6.1460E-03 A |U| G( 0f) P
60.000 b Freq Hz 80.940 B Ph(U) G( 0g) P
300.00 c TBeg K 3.7260E+05 C Re(Zb) G( 1a) P
2000.0 d |p| Pa -2.7012E+05 D Im(Zb) G( 1b) P
0.0000 e Ph(p) deg
6.1460E-03 f |U| m^3/s G
80.940 g Ph(U) deg G
air Gas type
!--------------------------------- 1 ---------------------------------
TBRANCH the split
3.7260E+05 a Re(Zb) Pa-s/m^3 G 2000.0 A |p| Pa
-2.7012E+05 b Im(Zb) Pa-s/m^3 G 0.0000 B Ph(p) deg
4.3459E-03 C |U| m^3/s
35.940 D Ph(U) deg
3.5185 E Hdot W
3.5185 F Edot W
-2.5508 G EdotTr W
!--------------------------------- 2 ---------------------------------
IMPEDANCE pure inertance
0.0000 a Re(Zs) Pa-s/m^3 2282.4 A |p| Pa
1.0000E+05 b Im(Zs) Pa-s/m^3 -8.8682 B Ph(p) deg
4.3459E-03 C |U| m^3/s
35.940 D Ph(U) deg
3.5185 E Hdot W
36
3.5185 F Edot W
!--------------------------------- 3 ---------------------------------
COMPLIANCE a compliance tank
4.8360E-02 a SurfAr m^2 Fn( 3b) 2282.4 A |p| Pa
1.0000E-03 b Volume m^3 S= -5 -8.8682 B Ph(p) deg
4.3459E-03 C |U| m^3/s
-54.060 D Ph(U) deg
3.4951 E Hdot W
ideal Solid type 3.4951 F Edot W
!--------------------------------- 4 ---------------------------------
SOFTEND this will reconnect at the UNION
0.0000 a Re(z) (t) 2282.4 A |p| Pa
0.0000 b Im(z) (t) -8.8682 B Ph(p) deg
4.3459E-03 C |U| m^3/s
-54.060 D Ph(U) deg
3.4951 E Hdot W
3.4951 F Edot W
44.392 G Re(z)
44.690 H Im(z)
!--------------------------------- 5 ---------------------------------
IMPEDANCE pure resistance
1.0000E+05 a Re(Zs) Pa-s/m^3 2282.4 A |p| Pa
0.0000 b Im(Zs) Pa-s/m^3 -8.8682 B Ph(p) deg
4.3459E-03 C |U| m^3/s
125.94 D Ph(U) deg
-3.4951 E Hdot W
-3.4951 F Edot W
!--------------------------------- 6 ---------------------------------
UNION close the loop here
4.0000 a SegNum 2282.4 A |p| Pa
2282.4 b |p|End Pa = 6A? -8.8682 B Ph(p) deg
-8.8682 c Ph(p)E deg = 6B? 7.0901E-17 C |U| m^3/s
-82.266 D Ph(U) deg
2.3118E-14 E Hdot W
!--------------------------------- 7 ---------------------------------
HARDEND seal the final end of the resonator
0.0000 a R(1/z) = 7G? 2282.4 A |p| Pa
0.0000 b I(1/z) = 7H? -8.8682 B Ph(p) deg
7.0901E-17 C |U| m^3/s
-82.266 D Ph(U) deg
2.3118E-14 E Hdot W
2.3118E-14 F Edot W
7.4001E-17 G R(1/z)
37
44.690 G Re(z)
!--------------------------------- 4 ---------------------------------
IMPEDANCE pure resistance
1.0000E+05 a Re(Zs) Pa-s/m^3 2282.4 A |p| Pa
0.0000 b Im(Zs) Pa-s/m^3 -8.8682 B Ph(p) deg
4.3459E-03 C |U| m^3/s
125.94 D Ph(U) deg
-3.4951 E Hdot W
-3.4951 F Edot W
!--------------------------------- 5 ---------------------------------
UNION close the loop here
3.0000 a SegNum 2282.4 A |p| Pa
2282.4 b |p|End Pa = 5A? -8.8682 B Ph(p) deg
-8.8682 c Ph(p)E deg = 5B? 6.1460E-03 C |U| m^3/s
80.940 D Ph(U) deg
2.3429E-02 E Hdot W
2.3429E-02 F Edot W
300.00 G End-T K
!--------------------------------- 6 ---------------------------------
COMPLIANCE a compliance tank
4.8360E-02 a SurfAr m^2 Fn( 6b) 2282.4 A |p| Pa
1.0000E-03 b Volume m^3 S= -5 -8.8682 B Ph(p) deg
2.8686E-14 C |U| m^3/s
119.89 D Ph(U) deg
-2.0495E-11 E Hdot W
ideal Solid type -2.0495E-11 F Edot W
!--------------------------------- 7 ---------------------------------
HARDEND seal the final end of the resonator
0.0000 a R(1/z) = 7G? 2282.4 A |p| Pa
0.0000 b I(1/z) = 7H? -8.8682 B Ph(p) deg
2.8686E-14 C |U| m^3/s
119.89 D Ph(U) deg
-2.0495E-11 E Hdot W
-2.0495E-11 F Edot W
-6.5606E-14 G R(1/z)
38
2.5508 F Edot W
!--------------------------------- 5 ---------------------------------
SOFTEND this will reconnect at the UNION
0.0000 a Re(z) (t) 2000.0 A |p| Pa
0.0000 b Im(z) (t) -3.4849E-13 B Ph(p) deg
4.3459E-03 C |U| m^3/s
-54.060 D Ph(U) deg
2.5508 E Hdot W
2.5508 F Edot W
32.398 G Re(z)
!--------------------------------- 6 ---------------------------------
UNION close the loop here
5.0000 a SegNum 2000.0 A |p| Pa
2000.0 b |p|End Pa = 6A? 0.0000 B Ph(p) deg
-3.4849E-13 c Ph(p)E deg = 6B? 1.3999E-16 C |U| m^3/s
-73.811 D Ph(U) deg
3.9031E-14 E Hdot W
!--------------------------------- 7 ---------------------------------
HARDEND seal the final end of the resonator
0.0000 a R(1/z) = 7G? 2000.0 A |p| Pa
0.0000 b I(1/z) = 7H? 0.0000 B Ph(p) deg
1.3999E-16 C |U| m^3/s
-73.811 D Ph(U) deg
3.9031E-14 E Hdot W
3.9031E-14 F Edot W
39
3. More about plotting
40
by the kinkiness of the plot, because the fourth-order Runge-Kutta integration technique
[16] essentially simulates the physics on a …ner length scale than the <.sp> …le actually
tabulates.
3.1.1. Cloning
The experienced user will soon want to display more than one plot simultaneously. “Cloning”
DeltaEC plots lets this occur. Return to <bottle6> of Chapter 2, which awakens displaying
p1 vs x: Format this curve for visual appeal, and then “clone” it twice from the plotter’s
File pulldown menu, editing each clone to arrive at something that looks like this layout:
All …ve curves displayed here (on three graphs) are updated every time the model runs
successfully, and all our e¤ort of formatting and cloning is saved in the <.ssv> …le if we
save when closing the model, so these formatted graphs will reappear if we return to the
<bottle6.out> model weeks later.
Sometimes it is useful to include Agas vs x in one or more of the state plots, to have a
sort of “drawing”of the apparatus lined up with the data graphs. The SegEndMarks option
in the plotter’s Options menu puts segment numbers on the graphs, at the end of each
segment, as another indication of apparatus locations on the data graphs. The schematic
41
view, available on DeltaEC’s Display pulldown menu, also gives a picture of the geometry
(see Fig. 2.11 and the Users Guide’s title page for examples).
3.1.2. Legs
Models with one or more TBRANCHes and/or two or more BEGINs are topologically compli-
cated, so DeltaEC’s <.sp> …le breaks them down into “legs,”with each leg starting over at
x = 0 for purposes of displaying graphs of data. For example, Fig. 3.2 shows a branched res-
onator, which we modeled with <legs.out> (available in the <ExamplesnPlotting> folder).
Figure 3.3 shows the standing-wave part of the pressure in that resonator. The side-branch
parts of the data form one curve, and the trunk parts of the data form another curve.
Using clone plots and the Legs pulldown menu, the user can show one variable per graph,
with all legs plotted in each graph; or can show one leg per graph with all variables of interest
plotted in each graph.
Figure 3.1: The bottle, used as an example in both Chapter 2 and the present chapter.
42
Figure 3.2: A branched resonator. In the model of this resonator, p1 and U1 have the same phase
at BEGIN, and the frequency is a guess. The magnitude of p1 in BEGIN is …xed and the magnitude
of U1 in BEGIN is a guess. Both TBRANCH impedances are guesses, and the four targets in the two
HARDENDs are used.
Figure 3.3: The standing-wave part of the pressure oscillation in the branched resonator of Fig.
3.2. There are two curves because the resonator has two legs and both legs have been selected in
the Legs pulldown menu. The kink in the upper curve occurs at the same pressure as the start of
the lower curve, because these are the same location: the TBRANCH.
43
Figure 3.4: A traveling wave with jp1 j nearly constant at 100 Pa and the phase of p1 starting at
zero and going down by 360 ; resetting from 180 to +180 near x = 5 m. The little menus to
convert the real and imaginary parts of p1 to the magnitude and phase are indicated here by the
red boxes. Similar features are available for U1 and Z:
the real part plotted along the horizontal axis and the imaginary part plotted along the
vertical axis. Conventionally, arrows from the origin point to each such complex value. One
can imagine all of the phasors on a phasor diagram rotating counterclockwise around the
origin at radian frequency !: Then the actual variables represented by the phasors are the
time-dependent projections of the phasors on the horizontal axis. Thus a phasor diagram
shows the relative magnitudes and time-phase relationships of a number of phasors. The
di¤erence between phasors at adjacent places in an apparatus also o¤ers insight into the
impedance of what lies between. The di¤erence between pressure phasors is related to the
volume ‡ow rate phasor via the momentum equation, which can include resistance and
inertance; the di¤erence between volume-‡ow-rate phasors is related to the pressure phasor
via the continuity equation and the compressibility of the intervening gas [13].
Figure 3.5 shows the U1 phasors for the bottle.
Another way to look at phasors, based on mouse cursor position in the schematic view,
is described in Section 2.7.3.
44
Figure 3.5: The phasor plot displays complex p1 and/or U1 on the real–imaginary plane. With
SegEndMarks selected in the Options pulldown menu, segment numbers are displayed at the end
of each segment.
<.ip>. Users can then manipulate and/or plot the <.ip> …le with DeltaEC’s built-in plot-
ter (or with their favorite graphics or spreadsheet software). We illustrate these features by
continuing two previous examples.
A driven, parallel LRC network was used as an example of structured branches and unions
in Section 2.7.5. Here, we use the same example to illustrate a simple incremental plot
in DeltaEC. The system is illustrated in the …gure below, and the …le we start with
is <lrc1.out> from the <ExamplesnAcoustics> folder. Open that …le, and save it as
<lrc1b.out> so subsequent changes don’t a¤ect the original. Review the model, to un-
derstand that this LRC network is driven at 60 Hz and 2 kPa. Under these circumstances,
the pressure amplitude in the COMPLIANCE is 2.28 kPa— higher than at the driver.
Let’s see how the pressure amplitude in the COMPLIANCE depends on frequency, using an
incremental plot. Unroll the BEGIN segment, double click on the frequency, and check the
Plot box. Let the frequency incrementing start from 40 Hz, end at 800 Hz, and do so in
77 steps so the increments are 10 Hz. Now unroll the COMPLIANCE segment, double click on
45
output A— the pressure amplitude— and select it as a dependent plot variable. Note that 0b
and 3A are ‡agged in the main display as plot variables now. Open the Incremental Plot
Summary from the Tools pulldown menu, and see what information is presented. (Note that,
in addition to our selection of 3A, all four guesses are listed as dependent plot variables— to
let the user return to any incrementally plotted point easily, as described in Section 3.2.3.)
Now start the repetitive run. DeltaEC begins at the initial value of the frequency,
40 Hz, and performs an ordinary run, adjusting the guesses until the targets are met. It
writes a line to the incremental-plot …le <lrc1b.ip>, recording on one line the independent
variable (f ), the guesses, and the chosen dependent variables (jp1 j in the COMPLIANCE in the
present case). Then DeltaEC increments the independent variable by its step (10 Hz in
this case), performs another ordinary run, and records the results on the next line in the
<.ip> …le...repeating until the …nal value of the independent variable is reached. In a text
editor, the <.ip> …le shows this:
BEGIN:Freq BEGIN:|U| BEGIN:Ph(U) TBRAN:Re(Zb TBRAN:Im(Zb COMPL:|p|
Hz m^3/s deg Pa-s/m^3 Pa-s/m^3 Pa
0b 0A 0B 0C 0D 3A
40.00 3.9237E-03 84.11 5.5931E+05 -4.5475E+05 2186.
50.00 5.0127E-03 82.56 4.4726E+05 -3.4400E+05 2234.
60.00 6.1460E-03 80.94 3.7260E+05 -2.7012E+05 2282.
...
The pressure amplitude starts at 2 kPa at low frequency, and rises with frequency at …rst,
peaking near 200 Hz. Could it be a resonance? The user could now change a parameter
such as the resistance in segment 5, repeat the repetitive run, and examine the e¤ects of
such a change on the resonance frequency and quality factor. (Don’t forget that a frequency-
independent Im[Zs ] in segment 2 is not the same as a frequency-independent inertance!)
46
3.2.2. Two-dimensional incremental plots
We use the bottle example from Chapter 2 again, starting from <bottle6.out> in the
<ExamplesnAcoustics> folder. Make the guess–target summary look like this:
Guess-Target Summary:
GUESS 0f 0g
name BEGIN:|U| BEGIN:Ph(U)
value 1.00E+03 90.
units m^3/s deg
TARGET 5a 5b
name HARDEND :R(1/z HARDEND :I(1/z
value .00 .00
Keep these parameters as two of the dependent variables to be plotted (they are copies
of the guesses).
We will watch these variables as functions of two independent variables: frequency f and
mean temperature Tm : To set up these two independent variables to be incremented, modify
each of them, checking the Plot box in the Parameter Edit dialog and then entering initial
and …nal values, and the number of steps. Use only three values of mean temperature in the
“outer”loop— 300, 400, and 500 K— by letting Tm go from 300 to 500 K in three steps of 100
K, and let f go from 100 Hz to 1200 Hz in 551 steps to get exactly 2-Hz increments in the
“inner” loop. Also add the pressure in the bottom of the bottle, 5A, as another dependent
plot variable, by double clicking on it and selecting it for plotting. After these actions, check
the Incremental Plot Summary again:
Outer: Inner: Dependent:
addr desc unit addr desc unit addr desc unit
0c TBeg K 0b Freq Hz 0f |U| m^3/s
Start: 300.0 Start: 100.0 0g Ph(U) deg
End: 500.0 End: 1200.0 5A |p| Pa
Step: 100.0 Step: 2.0
Points: 3 Points: 551
Now start a repetitive run. DeltaEC steps through the frequency and mean tempera-
ture, taking several seconds. For each value of frequency and mean temperature, it adjusts
guesses until targets are met, and then writes one line to the <.ip> …le, recording the values
of the independent and dependent plot variables, including all guesses. When it has …nished,
use the pc’s operating system to …nd the new …le, <bottle6.ip>:
BEGIN:TBeg BEGIN:Freq BEGIN:|U| BEGIN:Ph(U) HARDEN:|p|
K Hz m^3/s deg Pa
0c 0b 0f 0g 5A
300.000 100.000 2.0250E-06 -8.8471E+04 1.38690
300.000 102.000 2.0953E-06 -8.8471E+04 1.40830
300.000 104.000 2.1682E-06 -8.8471E+04 1.43090
.
.
.
500.000 1198.00 1.6259E-07 -8.8481E+04 0.61493
500.000 1200.00 1.7641E-07 -8.8480E+04 0.62388
The results are displayed by DeltaEC’s plotter as shown in Fig. 3.6. The plotter’s
pulldown menu must be used to get all three curves to show on the plot.
47
Figure 3.6: Pressure amplitude at the bottom of the bottle as a function of drive frequency, with
the pressure amplitude at the mouth held constant at 1 Pa, for three di¤erent temperatures.
Selecting a new operating point after using incremental plotting to explore a broad range
of behaviors. For example, the user could select a point at the top of one of the peaks
in Fig. 3.6.
Selecting a recent well-converged operating point from an incremental plot that contains
both converged and unconverged operating points and/or from a long incremental plot
that seemed to go awry and was canceled by the user.
The rewind feature is available whenever an <.ip> …le exists. It is accessible in the
Tools pulldown menu or through a button on the toolbar near the top of the main display.
When a point is selected, DeltaEC restores the independent plot variable(s) and the values
of guesses at that point. It also restores any output parameters that it has in the <.ip>
…le, which it might have selected in addition to the users choices because it caught the
user inadvertently or boldly using a sameas or RPN pointer to an output parameter in a
segment downstream of that sameas or RPN, despite DeltaEC’s “Forward Reference Con-
dition” warning about the unstable shooting-method behavior that can result from using
such downstream results. (See Section 5.4 for an example.)
48
3.2.4. Incremental plots when more than one model is run
Two or more models can be open in DeltaEC at the same time. If these models are run
separately, any incremental plots in any of those models will run separately, just as if they
were open in separate DeltaEC windows.
Linking two or more models together in a single DeltaEC window and running them
together with Run All is described in Section 4.9. The interaction between incremental plots
and multi-model running is described near the end of that section.
Rewinding from an incremental plot of inter-linked models behaves similarly to a single-
model plot rewinding as described above in Section 3.2.3.
49
4. RPN math
4.1. Summary
The RPN segment lets the user create nonstandard guesses, targets, and simple algebraic
calculations anywhere in a DeltaEC model. The format of the segment is shown in this
example:
!--------------------------------- 10 ---------------------------
! The result(s) of the second line’s calculations show up as output(s) in the right column.
RPN Dimensionless magnitude of specific acoustic impedance
0.000 a 7.8540E-05 A Znorm
p1 U1 / mag 9a * rho / a /
Input “a” is a real number, which can be a guess, a target, and/or a number used
subsequently in calculations via sameas links that point to it. The second line (line “b”) is a
formula expressing a user-de…ned algebraic procedure in Reverse Polish Notation (RPN), a
parenthesis-free algebra encoding technique. Variables such as density rho are re-evaluated
at this RPN segment’s location in the model (using, for example, Tm at this location) during
each numerical integration through the model.
A complete lesson in Reverse Polish Notation (also called post…x) can be found in in-
struction manuals for most HP (Hewlett-Packard) pocket calculators and on many websites,
including Refs. [17] and [18]. DeltaEC users who are unfamiliar with RPN notation are
encouraged to consult such a resource. Brie‡y, as numbers (including pointers to numbers
elsewhere in the model, and including variables introduced below) are encountered in the
instruction line, these numbers are pushed down onto a stack. Any numbers already present
on the stack are pushed deeper into the stack. When a unary operator such as cos, log, or
sqrt is encountered in the formula, it takes a single number o¤ the stack (“pops” it), acts
on it, and pushes the result back onto the stack. When a binary operator such as + or * is
encountered, it pops two numbers o¤ the stack, combines them appropriately, and pushes
the result back onto the stack. Other numbers that may be on the stack but are not involved
in the current operation simply move up and down during the operation so no gaps appear
in the stack and no uninvolved numbers get lost. (Some of us …nd it helpful to visualize a
spring-loaded stack of dishes at a cafeteria, which adjusts itself so the top dish is always at
the most convenient height no matter how many dishes are hidden underneath it.)
After a simple formula has been processed, there may be only one number remaining on
the stack; if so, this number is the “result” of the RPN segment and is displayed as a single
output, “A.” However, it is often desirable to leave additional results on the stack to be
accessed elsewhere or viewed by the user. Whatever is left on the stack after the instruction
line has been processed is displayed as outputs A, B, C,... For example, if the instruction line
is “1 2 3 4 + 45 cos”, the segment’s output values will be A = 0.7071, B = 7, C = 2, D = 1.
To assist in human readability, DeltaEC recognizes the semicolon as a punctuation mark
50
that the user can insert anywhere without a¤ecting the calculation. Thus, the instruction
line given above could also be written “1 ; 2 ; 3 4 + ; 45 cos”. (Note that blanks must both
precede and follow the semicolon.) Three shades of colored highlighting in the main display
of RPN segments also help human readability.
As a further aid to human readability, clicking on an output parameter in an RPN segment
and then clicking on “List Linkages”brings up a display showing how that parameter is cal-
culated, among other things. There, an RPN calculation can be displayed with parentheses
if desired. For example, outputs A and B in the last paragraph can be displayed as cos(45)
and 3+4, respectively.
DeltaEC’s RPN segment recognizes the common trigonometric, logarithmic, exponential,
and algebraic functions present on most pocket calculators, as well as hyperbolic and Bessel
functions that are not available on most calculators. Most of these functions accept complex
arguments. Other functions convert between complex numbers and real numbers in the usual
ways (e.g., real, imaginary, magnitude, phase). For complete lists and short descriptions of
DeltaEC’s RPN operations, see the tables in Section 11.4.1.
Whenever the command string of an RPN is being modi…ed, clicking the ShowOps button
displays a reminder of available operations and variables similar to the following:
Valid RPN commands listed below.
(It is OK to type lower case, except Bessel functions.)
Basic: + - * / ^ ~ inv
log log10 exp tenx sqrt sqrd abs
Trig: cos acos sin asin tan atan atanr pi
cosr acosr sinr asinr tanr atan2 atan2r
sinh cosh tanh
Bessel: J0 J1 Y0 Y1
RPN stack: # lstx a<>b sto rcl min max avg inp
Complex: mag real imag arg argr conj i cmplx cmplxr
Gas: gamma a rho cp k mu beta enth dk dn m
Mixture: nL kT D12 mL mH NLdot nLzero
Solid: rhos cs ks ds
State: Tm w f pm p1 U1
Thermodyn: Edot Htot H2k Xdot Tzero
Mean flow: Ndot p20HL p20tot
Cumulative: F1 vol x
Change: =Tm =p1 =U1 =H2k =Ndot =NLdot
=pm =f =nL =F1 =vol =p20HL
=Tzero =nLzro
Clicking on one of these reminders will save the user a little typing.
Where capital letters appear in the preferred spelling of an RPN operation, DeltaEC
will usually accept lower-case typing.
The rest of this chapter introduces many ways that RPN segments can be used.
51
The “List Linkages, Show with parentheses” display shows this as {[(5C/2)/pi]/0b}/5a/.
The “0.00”in line a is not used in this example. 5C points to jU1 j in segment 5, 0b points to
the frequency in segment 0, and 5a points to the duct area in segment 5. Alternative ways
to arrive at the same result in DeltaEC RPN formulas include
5C 2 pi * 0b * 5a * /
or
5C 2 pi 0b 5a * * * /
or
U1 mag w / 5a /
These examples show how DeltaEC recognizes numbers (e.g., “2”), pointers to input and
output numbers (e.g., “0b”and “5C”), special constants (here “pi”), and local complex and
real variables (here U1 and !):
As a second example, the user could display the local values of several thermophysical
properties of the gas, namely m ; cp ; a; and ; by inserting this RPN segment into a model:
RPN A=rho, B=cp, C=sound speed, D=viscosity (MKS units)
0.00
mu a cp rho
DeltaEC can perform complex arithmetic in RPNs. For example, it can calculate
tanh [(1 + i) y0 = ]
fk = (4.2)
(1 + i) y0 =
for a parallel-plate geometry [e.g., see the …rst of Eqs. (10.106)]. We let the input parameter,
line “a”, be equal to the plate half-spacing y0 so we can change it easily later:
RPN Thermal function for parallel plate geom. "a"=y0 (meters)
2.0000E-04 a (t) ( 0.9244 , -0.2392 ) A
(1,1) inp * dk / tanh lstx /
Here (1,1) illustrates the entry of a complex constant, (1 + i). This complex constant
could just as well have been made by entering 1 i + . The input variable of the segment,
line a, is referenced via inp in the RPN command string, and the thermal penetration depth
is accessed via dk. The “List Linkages, Show with parentheses”display shows this result
as tanh([(1,1)*2a]/dk)/{[(1,1)*2a]/dk}.
For complicated algebra that won’t …t in one RPN, the user can cascade two or more RPNs,
each relying on the output(s) of the previous one(s).
It can be dangerous to link the master of a master–slave set to the result of an RPN
segment through a sameas link. Such an arrangement works …ne until the model is saved
to the disk and later re-opened by DeltaEC. While opening such a model, DeltaEC
calculates the constant associated with the master–slave link during …le parsing, before the
model is run and thus before the numerical value of the master is calculated with Fortran
double-precision accuracy. Thus, the constant might be set inaccurately. If the <.out> …le
has been modi…ed in a text editor, values of the output parameters read from the disk, which
are usually irrelevant to subsequent model behavior, can a¤ect such master–slave linkages
severely.
52
4.3. List Linkages
When cascaded RPNs also involve parameters that are slaved to, or sameas, other para-
meters, and input parameters in other, non-RPN segments are sameas to the RPN results,
the dependencies can become very complicated. The List Linkages display, available in the
modify-parameter dialog for both input and output variables (with RPN parameter b being
the only exception— click on the associated output variable(s) instead) can be very useful for
avoiding mistakes, diagnosing in…nite loops of pointers, diagnosing Forward Reference Con-
ditions, etc. At Los Alamos, we often …nd ourselves with complicated situations, illustrated
with this example for the Linkages List from variable 8A, segment 8 being an RPN:
8A k(m-1) is an RPN result, calculated by 8b, via
the formula (sqrt[{1+[(gamma-1)*7B]}/(1-7A)]*w/a,
which needs 7B and 7A.
7A fnu is an RPN result, calculated by 7b, via
the formula ({[(1-i)*8a]*dn}/2)/7a,
which needs 8a and 7a.
7B fkappa is an RPN result, calculated by 7b, via
the formula ({[(1-i)*8a]*dk}/2)/7a,
which needs 8a and 7a.
8a G_or_T does not need other parameters.
7a G_or_T does not need other parameters.
9A needs 8A via RPN calc.
9B needs 8A via RPN calc.
9C needs 8A via RPN calc.
10a needs 9A via sameas.
10b needs 9B via sameas.
If you were planning to delete this segment, the …nal lines in the List Linkages display
would inform you to look …rst at segments 9 and 10 for unintended consequences, and it
would suggest that you could investigate whether any other segments need segment 7, which
might then become another candidate for deletion.
53
3.1372E-02 B Perim
inp 100 / sto pi * rcl sqrd pi * 4 /
!--------------------------------- 2 ---------------------------------
RPN a=volume (cubic cm). A=length (meters)
10.000 a G or T (t) 0.1273 A length
inp 1E6 / 1A /
!--------------------------------- 3 ---------------------------------
SURFACE top end cap of resonator (imagine it’s a moving driver)
sameas 1A a Area m^2 1000.0 A |p| Pa
0.0000 B Ph(p) deg
3.9203E-05 C |U| m^3/s
0.0000 D Ph(U) deg
1.9602E-02 E Hdot W
ideal Solid type 1.9602E-02 F Edot W
!--------------------------------- 4 ---------------------------------
DUCT the most important piece of this resonator
sameas 1A a Area m^2 996.65 A |p| Pa
sameas 1B b Perim m 5.8368E-04 B Ph(p) deg
sameas 2A c Length m 6.6737E-07 C |U| m^3/s
5.8368E-04 D Ph(U) deg
3.3257E-04 E Hdot W
ideal Solid type 3.3257E-04 F Edot W
!--------------------------------- 5 ---------------------------------
SURFACE End of the pipe
sameas 1A a Area m^2 996.65 A |p| Pa
5.8368E-04 B Ph(p) deg
2.2261E-19 C |U| m^3/s
-104.32 D Ph(U) deg
-2.7438E-17 E Hdot W
ideal Solid type -2.7438E-17 F Edot W
!--------------------------------- 6 ---------------------------------
HARDEND seal the final end of the resonator
0.0000 a R(1/z) = 6G? 996.65 A |p| Pa
0.0000 b I(1/z) = 6H? 5.8368E-04 B Ph(p) deg
2.2261E-19 C |U| m^3/s
-104.32 D Ph(U) deg
-2.7438E-17 E Hdot W
-2.7438E-17 F Edot W
-1.1503E-16 G R(1/z)
-4.5062E-16 H I(1/z)
With this …le, the user can modify the diameter (in cm) and the volume (in cm3 ) conve-
niently in 1a and 2a, and could use those as independent plot variables or as guesses. RPN
instruction 1b calculates the area (in m2 ) and puts it in output 1A for use in segments 3,
4, and 5. RPN instruction 2b calculates the length from the given diameter and volume,
putting it in output 2A for use in segment 4.
Note that segments 3, 4, and 5 show how sameas can point to DeltaEC outputs (capital
letters) as well as inputs (small letters).
54
For example, the stroke of a linear motor is often limited by the ‡exure characteristics
of its suspension. Here is an example using the RPN segment to …nd out how much current
must be delivered to a motor to keep the motor’s stroke targeted at 2 mm while driving a
resonator. The usual HARDEND targets are also used, with complex U1 in BEGIN being the
corresponding guesses. The RPN segment’s input, 2a, which is set by the user at 2.000, is the
user’s desired target, and the RPN formula yields the motor stroke, which is written in 2A.
DeltaEC has adjusted guess 1h, the motor current, until the motor stroke met its target.
TITLE Example 1: Plane-wave resonator
!Created@14:22:59 8-Jul-06 with DeltaE Vers. 5.5b6 for the IBM/PC-Compatible
!--------------------------------- 0 ---------------------------------
BEGIN Initialize things
1.0000E+05 a Mean P Pa
150.00 b Freq Hz
300.00 c TBeg K
0.0000 d |p| Pa
0.0000 e Ph(p) deg
9.4483E-03 f |U| m^3/s G
147.39 g Ph(U) deg G
helium Gas type
!--------------------------------- 1 ---------------------------------
IESPEAKER a linear motor and piston
sameas 3a a Area m^2 2443.2 A |p| Pa
1.0000 b R ohms -177.67 B Ph(p) deg
3.0000E-03 c L H 9.4248E-03 C |U| m^3/s
5.0000 d BLProd T-m 147.29 D Ph(U) deg
1.0000E-02 e M kg 9.4267 E Hdot W
1.0000E+04 f K N/m 9.4267 F Edot W
0.1000 g Rm N-s/m 20.919 G WorkIn W
4.7776 h |I| A G 14.031 H Volts V
0.0000 i Ph(I) deg 4.7776 I Amps A
51.381 J Ph(V/I) deg
2443.2 K |Px| Pa
ideal Solid type -177.67 L Ph(Px) deg
!--------------------------------- 2 ---------------------------------
RPN peak to peak stroke of the motor (mm)
2.0000 a G or T = 2A? 2.0000 A stroke
U1 mag 1a / w / 2 * 1000 *
!--------------------------------- 3 ---------------------------------
DUCT the most important piece of this resonator
1.0000E-02 a Area m^2 S= -2 2441.6 A |p| Pa
0.3540 b Perim m Fn( 3a) -177.56 B Ph(p) deg
6.7000 c Length m 2.8677E-05 C |U| m^3/s
-177.56 D Ph(U) deg
3.5009E-02 E Hdot W
ideal Solid type 3.5009E-02 F Edot W
!--------------------------------- 4 ---------------------------------
SURFACE End of the pipe
sameas 3a a Area m^2 2441.6 A |p| Pa
-177.56 B Ph(p) deg
8.8927E-15 C |U| m^3/s
-36.711 D Ph(U) deg
-8.4187E-12 E Hdot W
ideal Solid type -8.4187E-12 F Edot W
!--------------------------------- 5 ---------------------------------
HARDEND seal the final end of the resonator
0.0000 a R(1/z) = 5G? 2441.6 A |p| Pa
0.0000 b I(1/z) = 5H? -177.56 B Ph(p) deg
8.8927E-15 C |U| m^3/s
-36.711 D Ph(U) deg
-8.4187E-12 E Hdot W
-8.4187E-12 F Edot W
-4.6190E-14 G R(1/z)
3.7606E-14 H I(1/z)
! The restart information below was generated by a previous run
! You may wish to delete this information before starting a run
! where you will (interactively) specify a different iteration
! mode. Edit this table only if you really know your model!
guessz 0f 0g 1h
xprecn -1.1737E-08 4.2373E-03 -2.9847E-05
targs 2a 5a 5b
Mstr-slave 1 3 -2
The target parameter of an RPN, like any other target, can be used as the independent
variable in a plotting loop.
55
An RPN segment should usually be placed downstream of all the results and guesses
that it depends on, so the RPN will have current values during each DeltaEC integration
sequentially through the segments— especially if the RPN is used as a target or if its results
are used elsewhere in the model.
or
RPN
320.0
inp =Tm
or
RPN
20.0
Tm inp + =Tm
Used in this way, RPNs can essentially act like a user-created lumped-element segment,
changing the integration variables in ways not available in DeltaEC’s built-in segments.
For a complicated example of a “custom segment”implemented with RPNs that change three
integration variables, see “pulsed combustion”in Chapter 14.
56
verifying that the Reynolds number is low enough that the laminar calculation of the ‡ow
impedance should be accurate. Another RPN calculates the piston stroke. A few more
RPNs allow the IESPEAKER’s complex current to be guessed, to target a speci…ed electrical
impedance: A 100-Ohm resistor in series with a 39- F capacitor, chosen to cancel the e¤ect
of the IESPEAKER’s internal inductance. Final RPNs and other segments con…rm that all
incoming acoustic power is accounted for as the sum of electrical power delivered, electrical
power dissipated in the IESPEAKER’s internal resistance, mechanical power dissipated in the
IESPEAKER’s Rm , and thermal-hysteresis losses on all the surface areas in the model.
Figure 4.1: Two examples using many RPN segments. (a) A linear alternator with speci…ed load
impedance and piston leakage. (b) A loop of tubing …lled with air and driven at full-wave resonance
by two speakers separated by a quarter wavelength.
<closedLoop.out> shows a loop of tubing 2 meters long, with two side-branch drivers a
half meter apart being electrically driven 90 out of phase at a frequency intended to create
a full-wavelength traveling wave in the air in the loop. This example illustrates the use of
four RPNs as targets, to ensure that DeltaEC “connects”the end of the model back to the
beginning, by making p1 and U1 identical at the end and at the beginning.
TITLE A loop of tubing, with two side-branch drivers
0 BEGIN Start in the middle of nowhere, and integrate around a loop
1.0000E+5 a Mean P Pa
178.50 b Freq Hz
300.00 c TBeg K
688.67 d |p| Pa G
-0.92599 e Ph(p) deg G
1.7189E-3 f |U| m^3/s G
2.3837 g Ph(U) deg G
air Gas type
1 RPN Edot passing this point
2 DUCT FIRST QUARTER WAVELENGTH
3 TBRANCH
-2.5622E+5 a Re(Zb) Pa-s/m^3 G
1.7433E+6 b Im(Zb) Pa-s/m^3 G
4 VESPEAKER FIRST DRIVER
5 RPN driver resonance frequency
6 SOFTEND backside of driver open to the room
0.0000 a Re(z) =6G
0.0000 b Im(z) =6H
7 DUCT SECOND QUARTER WAVELENGTH
8 RPN miscellaneous calculations, some needed later
9 TBRANCH
-4.5837E+5 a Re(Zb) Pa-s/m^3 G 700.64 A |p| Pa
1.7440E+6 b Im(Zb) Pa-s/m^3 G -174.43 B Ph(p) deg
10 VESPEAKER SECOND DRIVER
57
11 SOFTEND backside of driver open to the room
0.0000 a Re(z) =11G
0.0000 b Im(z) =11H
12 DUCT THIRD AND FOURTH QUARTER WAVELENGTH
13 RPN close the loop: |p1|
sameas 0d a =A? 688.67 A Pa
p1 mag
14 RPN close the loop: phase(p1)
sameas 0e a =A? -0.92599 A deg
p1 arg
15 RPN close the loop: |U1|
sameas 0f a =A? 1.7189E-3 A m3/s
U1 mag
16 RPN close the loop: phase(U1)
sameas 0g a =A? 2.3837 A deg
U1 arg
B3 5B 2 + 3 = Tm ; (4.3)
58
sharing a common resonator, requiring so many segments and so many guesses and targets
that several models are needed to hold them all. How can such multiple models communicate
with one another automatically? This is accomplished with an extended RPN pointer of the
form filename:addr in an RPN formula.
For example, suppose that three models, named <engine1.out>, <engine2.out>,
and <engine3.out>, which represent a cascade engine [19], are simultaneously open in
DeltaEC. Suppose that hot-heat-exchanger heats are given at address 12e in all three
models. An RPN segment near the end of <engine3.out> could calculate the total heat
required by the set of three engines with the following formula which brings results from
<engine1.out> and <engine2.out> into <engine3.out>:
engine1:12e engine2:12e + 12e +
(The notation “ % ”in an interlinked RPN formula signi…es “the previously named model.”)
To run all of the models open in DeltaEC together, use Run All, with capital-R as its
keyboard shortcut. The order of running will be left to right in the sequence in which the
model tabs are laid out. Use the mouse to drag tabs from side to side to change that order.
(Drag a tab o¤ the side, top, or bottom of the panel to split the panel into two or more
displays so two or more models can be seen simultaneously.) Choose the order of running
carefully, so each model has up-to-date values of any numbers that it gets from other models.
For convenience, choose …le names so the default alphabetical opening of multiple models
puts them in the desired tab order.
If two or more models are run separately, any incremental plots in any of those models
will run separately, just as if they were open in separate DeltaEC windows. However,
Run All can optionally change this behavior dramatically, disabling any incremental plots
in the separate models and following the choices established by the user in the Multi-Model
Incremental Plot Control dialog, which is available in the Display pulldown menu. The
Multi-Model Plot Control lets the user choose one or two of the independent plot variables
that exist in the individual models as the overall independent plot variable (1-dimensional
plot) or variables (2-dimensional plot), and choose from among the established dependent
plot variables (including guesses) in the individual models too. A Run All will then run all
59
the models, increment an independent plot variable, run all the models again, etc., creating
a table of results in a …le called <MultiMod.ip>. For example, a set of linked models often
share the same frequency, linked together via RPN formulas similar to firstmodelname:0b
=f . If parameter 0b in model firstmodelname is chosen as the independent Multi-Model
Plot variable, then a Run All will sequentially increment the frequency in all of the linked
models.
To view graphs of <MultiMod.ip> after such a Run All, click the appropriate button in
the Multi-Model Plot Control. DeltaEC tries to include variables in the <MultiMod.ip>
…le that assist Rewinding when the user has ignored Forward-Reference-Condition warnings,
so don’t be surprised to …nd a few variables there that you don’t remember including yourself.
Please do not create more than one set of such inter-linked models with Multi-Model
Plotting in a single folder, because DeltaEC can have only one <MultiMod.ip> …le per
folder.
60
5. Engines & refrigerators
The examples given in the previous Chapters were relatively simple cases of linear acoustics.
In this Chapter, we introduce the full thermoacoustic computing power of DeltaEC with
some simple engines and refrigerators. After discussing the principles and assumptions that
are built into the computation, we present example calculations.
2
( 1)f ma d 1 f dp1 a2 (f f ) dTm dp1
1+ p1 + = 0; (5.1)
1+ s !2 dx m dx 2
! (1 )(1 + s ) dx dx
dp1 i! m
= U1 ;
dx A (1 f )
dU1 i!A ( 1)f (f f ) dTm
= 2
1+ p1 + U1 : (5.2)
dx ma 1+ s (1 )(1 + s ) dx
These equations are used in most of the STK** segments introduced below, which accommo-
date dTm =dx 6= 0 in engines and refrigerators. The complex variables f and f depend on
the size and shape of the pores and on the gas properties. Of course the right sides of these
61
equations depend implicitly on the local value of Tm ; because gas properties such as m and
a depend on Tm :
To complete this picture for many thermoacoustic engines and refrigerators, only two
more equations are required, i.e., those for the temperature Tm (x) and the total power
H_ tot (x): As for p1 (x) and U1 (x), the equations for Tm (x) and H_ tot (x) depend on the type of
segment, and Tm and H_ tot are continuous across the junctions between segments.
In DeltaEC, the total power is treated as the sum of two terms,
The …rst term, H_ 2;k ; represents power carried in the x direction by second-order thermo-
acoustic e¤ects, generally proportional to the lateral spatial average of Re[T1 u e1 ] ; and by
_
zeroth-order thermal conduction. The second term, HN_ ; represents power carried by nonzero
time-averaged ‡ow of gas in the x direction. In the present chapter, we consider only zero
time-averaged ‡ow, so H_ tot = H_ 2;k : (Nonzero H_ N_ is introduced in Chapter 6.)
Most segments, such as DUCTs and CONEs described in Chapter 2, obey simply dTm =dx = 0
and dH_ tot =dx = 0:
STK** segments have a more complicated di¤erential equation for Tm (x), which is derived
from Rott’s energy equation [22],
" !#
1 f f~
H_ 2;k = Re p1 U~1 1
2 (1 + s )(1 + )(1 f~ )
" #
c jU
m p 1 j2
dT m (f f~ )(1 + s f =f )
+ 2
Im f~ +
2Agas !(1 )j1 f j dx (1 + s )(1 + )
dTm
(Agas k + Asolid ks ) ; (5.4)
dx
by solving it algebraically for dTm =dx:
The total power ‡ow H_ tot through a STK** segment is assumed to be independent of x;
because the side walls are assumed to be rigid and thermally insulated. The total power
H_ tot changes with x in heat exchanger segments, according to
dH_ tot
= q;
_ (5.5)
dx
where q_ is the heat added per unit length in the heat exchanger.
(There is also a “thermally anchored”mode in which dH_ 2;k =dx = dE=dx _ in DUCTs, CONEs,
etc. to re‡ect the assumption that such segments are immersed in heat sinks; this exception
to the normal circumstances will be described in Section 11.3.1.)
So, for thermoacoustic calculations, DeltaEC integrates from the BEGINning through
the sequence of subsequent segments, with respect to two complex and two real variables:
p1 (x), U1 (x); Tm (x), and H_ tot (x). DeltaEC uses continuity of p1 ; U1 ; Tm ; and H_ tot to pass
from the end of one segment to the beginning of the next. Within each segment, it uses
equations appropriate to that segment type, governed by local parameters, such as area
and perimeter, by global parameters, such as frequency and mean pressure, and by evolving
62
parameters such as p1 ; U1 , Tm ; and H_ tot themselves and the properties of the gas and solid.
Mathematically, the DeltaEC integration can be regarded as
dp1
= Fmomentum (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
(5.6)
dU1
= Fcontinuity (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
(5.7)
dTm
= Fx-energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
(5.8)
dH_ tot
= Flateral energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.).
dx
(5.9)
In reality, as documented here and in Chapter 10, most of the equations for speci…c
segments are simpler than what’s suggested by the fully general forms of Eqs. (5.6)–(5.9).
For example, the right-hand side of the momentum equation never depends explicitly on
H_ tot ; many segment types have simply dU1 =dx = 0 or dTm =dx = 0 or can calculate p1
algebraically instead of with a numerical integration; and dHtot =dx is usually zero or as
simple as Eq. (5.5). Nevertheless, it is useful to keep in mind the fully general picture of
Eqs. (5.6)–(5.9), to help remember what DeltaEC needs to know at each location x as it
proceeds with its numerical integration.
Tom Ho‡er’s thermoacoustic refrigerator was described in detail in his Ph.D. thesis [1] and
was brie‡y summarized in Ref. [2]. We use this case to further illustrate capabilities of
DeltaEC, generating curves similar to Figs. 16 and 17 in Ref. [1] (Figs. 5 and 6 in Ref. [2]).
The apparatus is shown here in Fig. 5.1. We begin with a …le (<hofler1.out>, in the
<ExamplesnEnginesAndRefr> folder) whose geometry is that of Ho‡er’s “long”apparatus.
(We originally typed this …le in a Notepad text editor while studying Ref. [1].) The list of
segments serves as an outline of the model:
0 BEGIN:
1 SURFACE: driver end
2 DUCT : ambient temperature duct
3 HX : ambient temperature heat exchanger
4 STKSLAB: Stack
5 HX : Cold heat exchanger
6 DUCT : Cold Duct
7 CONE :
8 COMPLIANCE: end bulb
63
Figure 5.1: Ho‡er’s thermoacoustic refrigerator.
64
!--------------------------------- 6 ---------------------------------
DUCT Cold Duct
3.8400E-04 a Area m^2 0.0000 A |p| Pa
6.9400E-02 b Perim m 0.0000 B Ph(p) deg
0.1670 c Length m 0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 7 ---------------------------------
CONE
sameas 6a a AreaI m^2 0.0000 A |p| Pa
sameas 6b b PerimI m 0.0000 B Ph(p) deg
6.6800E-02 c Length m 0.0000 C |U| m^3/s
1.1600E-03 d AreaF m^2 0.0000 D Ph(U) deg
0.1210 e PerimF m 0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
!--------------------------------- 8 ---------------------------------
COMPLIANCE end bulb
4.9000E-02 a SurfAr m^2 0.0000 A |p| Pa
1.0600E-03 b Volume m^3 0.0000 B Ph(p) deg
0.0000 C |U| m^3/s
0.0000 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 0.0000 F Edot W
The segments that were not introduced in Chapter 2 are HX and STKSLAB.
HX’s have parallel-plate geometry, with plate spacing 2yo . Wave propagation through
heat exchangers is computed using the momentum and continuity equations, including both
viscous and thermal dissipation, much like the calculations in DUCTs. There is one additional
input: the heat added to the thermoacoustic wave from an external thermal reservoir (or, if
negative, the heat rejected to the external reservoir). The heat ‡ow determines the change
in total energy ‡ow H_ tot in the heat exchanger. As an input variable, heat ‡ow can either
be speci…ed by the user (and, optionally, an independent plot variable) or it can be a guess.
There is also a potential target: the temperature of the solid in the heat exchanger, which
is assumed to be spatially uniform and which di¤ers from the temperature in the gas in
proportion to the heat, as described in Chapter 10. The temperature di¤erence between gas
and solid can presently be computed only with an accuracy of about a factor of 2, even in
the acoustic approximation; nevertheless, it is included in DeltaEC, to prevent naive users
from being led to designs with heat exchangers of negligible surface area that have negligible
losses and that would appear to have no disadvantages if this temperature di¤erence were not
included. We hope that future revisions of DeltaEC will have a more accurate calculation
algorithm for this e¤ect. Meanwhile, however, if you prefer not to use this feature, watch or
target the gas mean temperature instead of the solid temperature by using RPN (see Chapter
4) or observing the gas temperature in the adjacent stack segment (parameter G or H).
STKSLAB also has parallel-plate geometry. This is the …rst segment we have encountered
in the Tutorial in which Tm can depend on x: DeltaEC obtains the x dependence by
integrating the thermoacoustic energy equation, schematically given by Eq. (5.8) and in
detail in Eqs. (5.4) and the equations given in Section 10.5.1, with H_ tot independent of x:
Using the lessons we learned in Chapter 2, we recognize that we do not know all the
initial conditions at BEGIN, so we must introduce some necessary guesses and targets. We
know that zero U1 must ‡ow through the bottom of the refrigerator, so we insert a HARDEND
there and use its two impedance targets. Corresponding to those targets, we’ll guess two
variables in BEGIN. These might be complex U1 ; but Ho‡er always ran his refrigerator on
resonance, so we’ll guess jU1 j and frequency f as guesses to get started, leaving the phases
of p1 and U1 equal (both zero) as the condition for resonance at the driver. After adding
these variables, the guess–target list looks like
65
GUESS 0b 0f
name BEGIN:Freq BEGIN:|U|
units Hz m^3/s
value 5.00E+02 5.00E-04
TARGET 9a 9b
name HARDE:R(1/z HARDE:I(1/z
value .00 .00
Running this case yields f = 498 Hz and jU1 j = 5:50 10 4 m3 /s. We are grati…ed that
none of the results thus far appear crazy. If you are running this …le for yourself, note also
that the solid is cooler than the gas in the ambient heat exchanger, where the (negative)
heat ‡ows from gas to solid, and that the solid is warmer than the gas in the cold heat
exchanger, where the (positive) heat ‡ow is from solid to gas.
Now consider what guesses and targets might be required by the heat and temperature
aspects of this model. We know that the bottom of the apparatus was thermally insulated,
so no energy can ‡ow in or out there. We also note from Ref. [1] that Ho‡er maintained the
solid temperature in the ambient heat exchanger at 300 K by vigorous ‡ow of temperature-
regulated water through it. These two constraints, not at the BEGINning, call for the use
of two targets: 9c and 3f. Ho‡er applied a known electrical heater power to the cold heat
exchanger, so we need not guess that; we leave its value at 3.00 W for now. However, he did
not control the heat removed at the ambient heat exchanger, nor the temperature of the gas
at the ambient end; so DeltaEC must guess those two variables: 0c and 3e. After making
the necessary changes, the guess–target summary looks like
GUESS 0b 0c 0f 3e
name BEGIN:Freq BEGIN:TBeg BEGIN:|U| HX:HeatI
units Hz K m^3/s W
value 498.02 300.0 5.054E-04 -9.640
TARGET 3f 9a 9b 9c
name HX:Est-T HARDE:R(1/z HARDE:I(1/z HARDE: Hdot
units K W
value 300.0 0.00 0.00 0.00
after running. By examining segment 5 in the <.out> …le, we see that DeltaEC predicts
a cold temperature of 223.76 K with this 3.00 W heat load on the cold heat exchanger.
All of these results are reasonable, so now we examine Ref. [1] very closely to con-
sider what else we should do to compare DeltaEC results directly to the published results
of Ho‡er’s experiments. Ho‡er added the calculated dissipation in the DUCT, CONE, and
COMPLIANCE to his electrically applied heat load, to plot the gross cooling power instead of
the net cooling power. So we do the same, with an RPN segment at the end of the model. We
add another RPN to display another of his plotted variables, the temperature ratio. On the
basis of the gross cooling power and the measured acoustic power delivered to the resonator
by the driver, Ho‡er also plotted the coe¢ cient of performance, COP = Q_ C;gross =E;
_ relative
to Carnot’s COP, COPCarnot = TC = (T0 TC ), so we calculate COP/COPCarnot with a third
RPN at the end. These additions look like
66
+ 0 BEGIN
+ 1 SURFACE driver end
+ 2 DUCT ambient temperature duct
+ 3 HX ambient temperature heat exchanger
+ 4 STKSLAB Stack
+ 5 HX Cold heat exchanger
+ 6 DUCT Cold Duct
+ 7 CONE
+ 8 COMPLIANCE end bulb
+ 9 HARDEND target this to seal the end
!--------------------------------- 10 ---------------------------------
RPN gross cooling power is net power plus tail dissipation
0.0000 a G or T (t) 3.8487 A Qg (W)
5e 5F +
!--------------------------------- 11 ---------------------------------
RPN temperature ratio, Tc/T0
0.0000 a G or T (t) 0.7459 A ratio
5H 3H /
!--------------------------------- 12 ---------------------------------
RPN COP / Carnot’s COP
0.0000 a G or T (t) 0.1614 A COP rel
10A 1F / 5H 3H 5H - / /
We let the net cooling power, i.e., heat 5e at the cold heat exchanger, be the independent
plot variable in DeltaEC, ranging from 1 to 7 W in 0.5 W steps. To obtain the variables
we want to display, we add 10A, 11A, and 12A to the list of dependent plot variables. After
running this case, we changed jp1 j to 0.015 of pm , and changed 5e to 0.5 W, making those
changes by hand in a couple of steps to make sure DeltaEC did not get lost. Then we
again used 5e as the independent plot variable, over the range of Q_ C to 0.5–3.5 W in steps
of 0.25 W, and ran it again, appending the results to the previous plot run. We imported
the resulting <.ip> …le into a spreadsheet so we could display it together with Ho‡er’s
experimental data. The results are shown in Figs. 5.2. The DeltaEC calculations come
reasonably close to the measurements presented in Figs. 16 and 17 of Ho‡er’s thesis [1].
Let’s modify this example a little more, using it to illustrate some points about side-
branch and series transducers and about DeltaEC’s consistency with the …rst law of ther-
modynamics. First, let’s add a more realistic driver to the system, using VSPEAKER. We
insert VSPEAKER at the beginning, using Ref. [1] for approximate numerical values for the
mass, resistance, and force constant for the speaker, and we delete the SURFACE segment
that was previously near the beginning because VSPEAKER accounts for the oscillatory pres-
surization losses on its surface area. We use the drive voltage as a guess, so we are asking
DeltaEC what voltage we must supply to maintain jp1 j = 30 kPa in front of the speaker;
we estimate it will take about 20 V to drive it, so we use that as our starting value for the
guess. We’ll also add an RPN target after the VSPEAKER, to enforce resonance by ensuring
that the phases of p1 and U1 are equal at the driver. We do this by forcing their di¤erence,
computed by subtracting the values appearing as results 1B and 1L, to be zero, which is the
value given in line 2a.
With this change, BEGIN is no longer at a position of nonzero jU1 j; BEGIN is now at a
place where U1 = 0 and p1 is nonzero, and the subsequent side-branch VSPEAKER adds all
the volume ‡ow rate that ‡ows into the subsequent segments. We change guesses and initial
values accordingly. With these changes, but before running, the guess–target list becomes
GUESS 0b 0c 0e 1h 4e
name BEGIN:Freq BEGIN:TBeg BEGIN:Ph(p) VSPEA:AplVo HX:HeatI
units Hz K deg V W
value 504.22 300.80 100.00 20.000 -11.126
TARGET 2a 4f 10a 10b 10c
name RPN:Targe HX:Est-T HARDE:R(1/z HARDE:I(1/z HARDE:Htot
units K W
value 0.0000 300.00 0.0000 0.0000 0.0000
result 0.0000 300.00 -7.8635E-15 5.8006E-14 -1.7764E-15
67
Figure 5.2: Ho‡er refrigerator results. Lines are DeltaEC results; points are from experimental
data presented in Ho‡er’s thesis. Squares, p1 = 0:015pm . Circles, p1 = 0:03pm .
68
and when we run it the guess–target list becomes
GUESS 0b 0c 0e 1h 4e
name BEGIN:Freq BEGIN:TBeg BEGIN:Ph(p) VSPEA:AplVo HX:HeatI
units Hz K deg V W
value 505.59 302.60 152.52 23.385 -36.043
TARGET 2a 4f 10a 10b 10c
name RPN:Targe HX:Est-T HARDE:R(1/z HARDE:I(1/z HARDE:Htot
units K W
value 0.0000 300.00 0.0000 0.0000 0.0000
result 1.7536E-11 300.00 2.5373E-12 -2.5260E-11 -4.2633E-12
Comparing the values of the guesses here with those before we inserted the VSPEAKER
shows two big di¤erences: the phase of the pressure and the heat rejected at the ambient
heat exchanger.
The change in the phase of the pressure is essentially trivial: The VSPEAKER segment sets
the overall time phase of everything in the model, because the phase of its voltage is …xed
at zero. Previously, the phase of U1 at the top of the model was arbitrarily set to zero by
the BEGIN segment. Now, the VSPEAKER controls the overall time phase of the model, and
the nonzero imaginary part of the speaker’s mechanical impedance causes a large phase shift
between V1 and U1 . This simply shifts all p1 and U1 phasors by about 30 : A further 180
phase shift enters because of the sign convention for positive voltage built into the segment.
The increase in the heat rejected at the ambient heat exchanger from 11.1 W to 36.0
W is meaningful. The extra heat comes from the VSPEAKER, and illustrates how DeltaEC
handles energy for **SPEAKERs and **DUCERs: They are treated as thermally insulated, so
all of the electrical energy going into them shows up either as acoustic power or as heat. This
transducer is not remarkably e¢ cient at converting electrical power into acoustic power; most
69
of its electrical power becomes heat, which is rejected at the ambient heat exchanger in this
model. The extra load on the ambient heat exchanger increases the temperature di¤erence
in that heat exchanger a little, which leads to the small increase in gas temperature seen in
0c, which in turn leads to the small increase in resonance frequency seen in 0b.
The speaker can also be included in the model “in series,” using a VESPEAKER segment,
with yet another recon…guration of the BEGIN and associated guesses and targets to re‡ect
new circumstances. As shown below, we include the cavity above the speaker, which appears
from Ref. [1] to have a volume of approximately a liter. Above that volume, the BEGIN starts
with U1 = 0 and with a guessed p1 : We arbitrarily set the time phase of the entire model
by setting the phase of p1 in that volume to zero in the BEGIN. We note that Ho‡er had
cooling-water tubes attached to the driver housing, to try to remove as much of the driver’s
electrical and other dissipation as possible. To crudely model the e¤ect of these tubes, we
insert an RPN below the driver, which we craft so it removes 75% of the driver’s heat, to
illustrate the use of RPN for such purposes. With these changes, and after running, the
guess–target list becomes
GUESS 0b 0c 0d 2h 2i 7e
name BEGIN:Freq BEGIN:TBeg BEGIN: |p| VESPE: |V| VESPE:Ph(V) HX:HeatI
units Hz K Pa V deg W
value 504.56 301.24 284.59 23.271 116.88 -17.243
TARGET 3a 4a 7f 13a 13b 13c
name RPN:Targe RPN:Targe HX:Est-T HARDE:R(1/z HARDE:I(1/z HARDE:Htot
units K W
value 0.0000 3.0000E+04 300.00 0.0000 0.0000 0.0000
result -5.5422E-13 3.0000E+04 300.00 1.7565E-14 1.8398E-13 2.4869E-14
70
RPN (a)= % of driver heat removed by water tubes on driver box
0.7500 a G or T (t) 14.243 A Watt
18.425 B Watt
Htot Edot - inp * sto Htot rcl - =H2k
!--------------------------------- 6 ---------------------------------
DUCT ambient temperature duct
1.1340E-03 a Area m^2 2.9736E+04 A |p| Pa
0.1190 b Perim m -90.401 B Ph(p) deg
4.2600E-02 c Length m 2.8084E-03 C |U| m^3/s
-169.43 D Ph(U) deg
14.243 E Htot W
ideal Solid type 7.9452 F Edot W
+ 7 HX ambient temperature heat exchanger
+ 8 STKSLAB Stack
+ 9 HX Cold heat exchanger
+ 10 DUCT Cold Duct
+ 11 CONE
+ 12 COMPLIANCE end bulb
+ 13 HARDEND target this to seal the end
+ 14 RPN gross cooling power is net power plus tail dissipation
+ 15 RPN temperature ratio, Tc/T0
+ 16 RPN COP / Carnot’s COP
Notice how the DeltaEC RPN operator =H2k has been used to override DeltaEC’s
normal continuity of H_ between segments, to simulate the removal of 18.425 W of heat by
the water-cooled tubing on Ho‡er’s driver housing.
The standing-wave components of the pressure and volume-‡ow-rate waves in this re-
frigerator model are shown in Fig. 5.3— it is essentially a quarter-wave device. In DeltaEC
models of thermoacoustic engines and refrigerators, it is often also worthwhile to examine
graphs of H_ and E_ and to think about their consistency with the laws of thermodynamics,
e.g., to reveal possible fuzzy thinking about how BEGIN and **END are set up in the model.
In the present case, shown in Fig. 5.4, the acoustic power delivered by the driver drops most
rapidly in the stack and more slowly in the bottom part of the resonator, and the constancy
of H_ tot in each insulated segment (i.e., everywhere but in heat exchangers) is evident. The
positive signs of these power ‡ows indicate ‡ows in the positive-x direction, i.e., from the
BEGIN toward the HARDEND in the model.
Another instructive graph displays gas temperature Tm and solid temperature Tsolid as
functions of x; as illustrated for the present example in Fig. 5.5. In heat exchangers, the
di¤erence Tsolid Tm is due to the time-averaged heat ‡owing from solid to gas. In other
segments, where there is no time-averaged heat ‡owing from solid to gas, these two temper-
atures are equal.
The …nal example in the previous section used six targets and six guesses— the most com-
plicated guess–target summary table we have yet encountered in the Users Guide. In this
section, we discuss how we chose its members, and explain some general guidelines for guess–
target choices.
Recall the purpose of guesses and targets. DeltaEC is numerically integrating equations
71
Figure 5.3: Standing-wave parts of the p1 and U1 wave in Ho‡er’s refrigerator.
Figure 5.4: Acoustic power E_ (dashed blue) and total power H_ (solid red) as a function of x in the
refrigerator model. The BEGIN, COMPLIANCE, and VSPEAKER are all crowded into x = 0: The drop
in H_ marks the ambient heat exchanger and the smaller rise in H_ marks the cold heat exchanger.
72
Figure 5.5: DeltaEC results for gas temperature Tm and solid temperature Tsolid as functions
of x in the uppermost part of Ho‡er’s refrigerator. These temperatures di¤er only in the heat
exchanger.
dp1
= Fmomentum (local and global parameters and variables); (5.10)
dx
dU1
= Fcontinuity (local and global parameters and variables); (5.11)
dx
dTm
= Fx-energy (local and global parameters and variables); (5.12)
dx
dH_ tot
= q_ in heat exchangers; = 0 elsewhere. (5.13)
dx
from the BEGIN segment to the end of the model. Some of the results of that integration
are actually known a priori by the user, and some of the variables needed by DeltaEC
to perform that integration are not known a priori by the user. These exceptional vari-
ables are speci…ed by the user as targets and guesses, respectively, and DeltaEC uses a
shooting-method algorithm, attempting to meet the targets by …nding suitable values for
the guesses. Initial values for guesses are usually left over from a previous run of DeltaEC,
but sometimes they are provided by the user.
For the refrigerator example of the previous section, we chose targets and guesses by
considering what integration results were known by us a priori and what input variables
were not known by us a priori. We looked …rst near the end of the model for possible
targets. We de…nitely needed the three HARDEND variables Re[1=zn ], Im[1=zn ], and H_ tot in
the target list, all targeted to zero, because there is no hole in the end of the apparatus that
would allow nonzero U1 or H_ tot to escape— these are integration results that are known to
us a priori. Experimentally, Ho‡er maintained the hot heat exchanger at 300 Kelvin; but
DeltaEC computes that as a result of each integration pass, so it also had to be a target.
73
We looked …rst at the BEGIN segment for candidate guesses. The beginning gas temperature
was not known a priori by us: DeltaEC had to adjust the beginning Tm to arrive at the
…rst HX Tsolid correctly, so 0c had to be a guess. Next, we did not know the frequency
needed to maintain resonance a priori, so it too had to be a guess. But how was resonance
determined experimentally? By comparing the phases of p1 and U1 at the driver. These
phases are computed by DeltaEC as results. Hence, we added their di¤erence = 0 as a …fth
target. While we were thinking about the driver, we realized that we wanted to imagine that
jp1 j just below the driver was …xed, because Ho‡er adjusted his driver voltage to achieve
a desired amplitude. Thus jp1 j at that location— a calculated result in DeltaEC— had to
be a target, and the driver voltage had to be a guess. The heat rejected in the …rst heat
exchanger had to be guessed because we didn’t know it or control it experimentally, yet it is
required by DeltaEC for each numerical integration. By now we had six targets and four
guesses, so we needed two more guesses. One guess must be jp1 j at the beginning, which
is in the cavity above the loudspeaker, where we had no a priori knowledge of pressure.
One more guess was needed. The overall time phase of a DeltaEC model is meaningless;
somewhere there is always one time-phase variable that can be chosen arbitrarily. However,
here we still had two phases set arbitrarily: the phase of p1 at the BEGIN and the phase of
V1 in the driver. One of these had to be used as a guess. We chose to use the phase of V1 as
that guess.
We chose a point of view where we wanted to learn the cold solid temperature 9H for a
given cooling power 9e. Had we wanted to have DeltaEC tell us the cooling power at a
known cold temperature, we would have used an additional guess–target pair, targeting TC
and guessing Q_ C .
Although some situations that we model in DeltaEC are not this confusing, choosing the
guesses and targets is never easy for a complicated thermoacoustic system. To choose them
wisely, two things are essential: careful thought about the particular thermoacoustic system
itself, and an appreciation of what variables are results of a single DeltaEC integration
from BEGIN to **END and what variables are needed by DeltaEC for each such integration.
We introduce a few general guidelines for this thought process here.
To understand which variables are candidate targets, you must know which variables
are computed as results by DeltaEC during each pass. To understand which variables are
candidate guesses, you must know which variables are needed by DeltaEC for each pass
of its integration. To aid this understanding, look at Eqs. (5.10)–(5.13), where computed
results are the variables on the left-hand sides, and variables needed are on the right-hand
sides. Print these equations (or the more complete versions at the beginning of Chapter 10)
and post them near your workstation, so you can look at them for inspiration when you
are trying to …gure out guesses and targets for a new or modi…ed model. Also look at the
columnar structure of the <.out> …le, where variables needed by DeltaEC for each pass of
its integration are in the left column (as are potential targets) and variables that DeltaEC
computes as results during each pass are in the right column. For a perfect understanding,
there is no good substitute for studying the summaries of the computation algorithms for
each segment, as discussed brie‡y in the tutorial chapters of the Users Guide, and more fully
in Chapter 10.
To understand what variables should be regarded as known a priori by the user, it is
74
Table 5.1: Choosing guesses and targets.
sometimes helpful to think about what variables are (or could be, in principle) experimen-
tally controlled, in contrast to what variables are experimentally observed. These must be
compared with the variables that DeltaEC computes as results during each integration
pass and those that DeltaEC needs as inputs during each integration pass through the
system. Note that our de…nition of an experimental result is more general than usual. In the
Ho‡er refrigerator case, we considered the drive voltage an experimental result because it
is determined experimentally by the condition that the pressure amplitude have the desired
value. The viewpoint expressed in Table 5.2 is appropriate for comparison of DeltaEC
and experimental data. In this case, geometrical parameters are simply …xed. Targets are
experimentally …xed or controlled variables that are results of a single pass of numerical
integration, chosen from among Tm , p1 , and U1 (everywhere but in BEGIN); current magni-
tudes and phases in VDUCERs and voltage magnitudes and phases in IDUCERs; etc. Guesses
may be unknown parameters chosen from among f , the magnitude and phase of U1 -BEGIN
and p1 -BEGIN, Tm -BEGIN, heats at heat exchangers, the magnitude and phase of voltage at
VEDUCERs, etc.
When designing hardware instead of analyzing it, a di¤erent viewpoint may be adopted.
In this case, many geometrical parameters are not yet …xed, but desired operating tem-
peratures, powers, frequency, etc. have been chosen. Often, several geometrical parameters
are included as guesses, and more temperatures and other numerical results are included
as targets. Hence, another useful way to think about guesses and targets is represented by
Table 5.3.
As we did in our description of the choices of guesses and targets for the refrigerator
example above, it is useful to look for targets at the end or middle of the model, then
75
Table 5.2: Interpreting input and result variables: experimental viewpoint.
Experimentally
Experimental result
controlled variable
76
look for guesses in the beginning or middle of the model. Sometimes it is easy to mentally
associate each target with a corresponding guess, as we did for some of them in the discussion
above. A few targets, such as the HARDEND targets in the example above, may be utterly
obvious and necessary, depending only on the model but not on the user’s point of view.
However, many targets depend on the user’s point of view. Does the user want to consider
an electrically driven refrigerator to be driven at …xed frequency, or to be kept on resonance
by active control? Is the user interested in cooling power at a given cold temperature, or
in the cold temperature that is reached for a given cooling load? Thus, it is essential to
think carefully about what we want to think of as “known” and what we want to think of
as “unknown”for every model and for every use of that model.
It is also important to remember that the overall time phase of a DeltaEC model is
arbitrary, to …nd the one and only one location in the model where that time phase is set
(usually in the BEGIN, but sometimes in a transducer or an RPN target), and to make sure
there are no other con‡icting time-phase input variables (multiple side-branch transducers
often being problematic).
77
9.2000E-05 b Volume m^3 18.320 B Ph(p) deg
1.2280E-03 C |U| m^3/s
136.39 D Ph(U) deg
etc. ...
78
If DeltaEC seems to be missing targets by only a little, reduce the convergence tolerance
or adjust one of the other shooting-method tuneup parameters, which are accessible
through the Edit pulldown menu as Options, described in Section 8.3.2.
Consider whether the number of guesses and targets could be reduced by beginning at the
**END and ending at the BEGINning. If so, “‡ip”the model as described in Section 5.7.
Consider other ways to reduce the number of guesses and targets. For example use
VSPEAKER instead of VESPEAKER if possible. Or avoid a transducer by using a nonzero
U1 in BEGIN.
Change the guess-list membership, if you can think of an appropriate alternative to one or
more existing guesses.
In a model with a STK** segment, it is common for a heat upstream of that segment to
be a guess, corresponding to a temperature target downstream of that segment. With
regenerators, pulse tubes, and thermal bu¤er tubes, small changes in the guessed heat
tend to cause large changes in the downstream temperature while the shooting method
is exploring the landscape. If long DUCTs or other temperature-sensitive segments are
downstream of the temperature-target segment, any targets downstream of the long
segments can be wildly a¤ected by the shooting method’s exploratory changes in the
heat guess. Shooting-method stability can be improved in such cases by putting an
RPN segment that …xes the temperature (e.g., “300 =Tm”) immediately upstream of
the long segment(s).
If convergence seems strangely problematical, also consider whether some guesses might
essentially have large …xed o¤sets. The shooting method begins by calculating the
sensitivities of the results to changes in the guesses by changing each guess by a …xed
percentage (described in Section 8.3.2). If a part-per-million change in one guess causes
part-per-million changes in targeted results, but a part-per-million change in a second
guess causes an order of magnitude change in one or more targeted results, the shooting
method will fail. In such cases, it helps to use an RPN input parameter as small guess,
add a large o¤set to it in the RPN’s line b to make a number that’s the size of the
original guess, and sameas the original guess to the RPN result. This will ensure that
ppm changes in all guesses cause ppm changes in all targeted results.
79
5.4.3. Modifying an existing model
Purely acoustic models with only a few guesses and targets, like the bottle in Chapter
2, are usually robust: From a state with all targets met, the user can often change any
variable by a factor of two or so, and DeltaEC’s shooting method will once again hit the
targets. However, DeltaEC’s shooting method gets lost more easily for thermoacoustic
models with more than about …ve targets and …ve guesses, especially with STKDUCT and
STKCONE segments. Some strategies for successful routine modi…cations of complex models,
starting from a state in which targets are met, include:
If an overly bold change has gotten DeltaEC lost, use Restore all guessez, described
in Section 5.7, to recover the previous set of guesses. Then use the incremental plot
feature to change a variable by small steps. In practice, this usually means: Observe
that a big change led to failure to converge; Restore guessez and un-do the big change
manually to return to the secure starting condition; Try a smaller change (such as
1/10 of the original change); If the smaller change still leads to failure to converge,
then restore and undo again, and try an even smaller change; When a small-enough
change has been found, use that change as the step in the incremental plot feature
to let DeltaEC accomplish the desired big change in 10, 100, or more small steps.
(If DeltaEC gets part way through these small steps and then loses its ability to hit
targets, use the plot rewind feature, described in Section 3.2.3, to recover a meaningful
operating point that is at least part way toward the desired change.)
Insert a new segment gradually, by …rst inserting it with near-zero volume and near-zero
impedance, then gradually increasing these with the incremental plot feature. Often
this entails inserting a new segment with near-zero length and gradually increasing the
length. Similarly, the heat of a new heat exchanger can be increased gradually from
zero; the impedance of a new side-branch impedance can be reduced gradually from a
very large value; the impedance of a series impedance can be increased gradually from
zero.
Kill an undesired segment gradually by reducing its length with the incremental plot feature
before Killing it.
Save your model frequently, so you can recover from the unexpected.
The shooting method within DeltaEC can sometimes get stuck around an inappropriate
local minimum. Try manually changing one of the guesses slightly and see if DeltaEC
will loose its fondness for this particular sticky point.
Always check results carefully for reasonableness, particularly when modeling complicated
systems or using any of DeltaEC’s more elaborate features. While DeltaEC is a
useful tool, it is far from foolproof— for example, the shooting method can easily end
up generating devices that are several wavelengths long or wander into the completely
unrealistic realm of negative values for guessed frequency or length.
80
Figure 5.6: Five-inch-diameter standing-wave engine.
81
5.0000E-04 d radius m -85.386 D Ph(U) deg
5.0000E-05 e Lplate m 2210.0 E Htot W
170.63 F Edot W
557.92 G TBeg K
stainless Solid type 306.39 H TEnd K
!--------------------------------- 5 ---------------------------------
HX Ambient HX
1.2670E-02 a Area m^2 6.3081E+04 A |p| Pa
0.4860 b GasA/A 3.2510 B Ph(p) deg
5.0800E-02 c Length m 0.1676 C |U| m^3/s
4.0600E-04 d y0 m -85.829 D Ph(U) deg
-2125.1 e HeatIn W G 84.898 E Htot W
303.00 f Est-T K = 5H? 84.898 F Edot W
306.39 G GasT K
ideal Solid type 303.00 H SolidT K
!--------------------------------- 6 ---------------------------------
ANCHOR subsequent duct is water jacketed. dHdot/dx = dEdot/dx
!--------------------------------- 7 ---------------------------------
DUCT Ambient Duct
sameas 5a a Area m^2 6.9652E+04 A |p| Pa
0.3990 b Perim m -176.59 B Ph(p) deg
3.6500 c Length m 1.8500E-05 C |U| m^3/s
-176.59 D Ph(U) deg
0.6443 E Htot W
ideal Solid type 0.6443 F Edot W
!--------------------------------- 8 ---------------------------------
SURFACE Ambient End
sameas 5a a Area m^2 6.9652E+04 A |p| Pa
-176.59 B Ph(p) deg
1.7385E-16 C |U| m^3/s
95.222 D Ph(U) deg
1.9100E-13 E Htot W
ideal Solid type 1.9100E-13 F Edot W
!--------------------------------- 9 ---------------------------------
HARDEND target this to seal the end
0.0000 a R(1/z) = 9G? 6.9652E+04 A |p| Pa
0.0000 b I(1/z) = 9H? -176.59 B Ph(p) deg
1.7385E-16 C |U| m^3/s
95.222 D Ph(U) deg
1.9100E-13 E Htot W
1.9100E-13 F Edot W
1.3878E-17 G R(1/z)
-4.3970E-16 H I(1/z)
!--------------------------------- 10 ---------------------------------
RPN ensure all ambient-duct dissipation goes to water jacket
0.0000 a G or T = 10A? -5.6843E-14 A zero
5E 5F -
and some <.sp> graphs show the wave, power, and temperature pro…les in Fig. 5.7.
Targets 9a and 9b, the real and imaginary parts of 1=zn ; are of course necessary to impose
U1 = 0 at the closed ambient end of the resonator. Our choices of other targets and guesses
were made to re‡ect the perspective of Ref. [23], in which the as-built hardware was driven
by a …xed heat input (ordinary input 3e) and the ambient heat exchanger was held at …xed
temperature (target 5f). Frequency 0b is a guess, because we want DeltaEC to tell us the
frequency at which it thinks the system should spontaneously oscillate. Initial temperature
0c and initial pressure amplitude 0d are also guesses, because we want DeltaEC to tell us
the hot temperature and the pressure amplitude at which it thinks the system should oscillate
for the given hot heat input. The …nal target and guess, 10a and 5e, allow DeltaEC to tell
us the heat rejected at the ambient temperature, under the constraint that E_ = H_ tot at the
interface between the ambient heat exchanger and the long ambient resonator duct. We judge
this to be a reasonable energy-closure condition for the model because the resonator duct was
82
Figure 5.7: Graphs of the wave (left), total power and acoustic power (top right), and temperature
(bottom right) as functions of x in <5inch.out>. Changing temperature and constant H_ tot show
the location of the stack. The heat exchanger locations are indicated by the steep changes in H_ tot :
83
water-cooled in this apparatus, so the heat arising from E_ dissipated in it might be expected
to be carried away locally by water. (This is of course not strictly true; boundary-layer
thermoacoustic processes redistribute energy along x; as was dramatically demonstrated in
Ref. [24] long ago.) Consistent with this energy-closure condition, the ANCHOR segment just
before the …nal DUCT overrides the default behavior of the subsequent segments (thermally
insulated, except in heat exchangers), ensuring that H_ 2,k = E_ thereafter.
Other choices could be made for the guesses and targets. For instance, the cold-duct
length could be substituted for the frequency as one of the guesses, if we were designing
an apparatus to operate at a desired frequency. The hot-heat-exchanger temperature 3f
could be added to the target list and, simultaneously, the hot-heat-exchanger heat 3e could
be added to the guess list, if we were designing an apparatus to operate at a desired hot
temperature. For now, however, the guesses and targets will be left as they are, since they
re‡ect the point of view adopted in Ref. [23].
To plot some results for this 5-inch engine case, we started DeltaEC with this …le again
and prepared parameters so the plot summary became
Dependent Variables (outputs):
PLOTS 0A 0B 0C 0D 3H 8A
name BEGIN:Freq BEGIN:TBeg BEGIN:|p| BEGIN:HeatI HX:Solid HARDE:|p|
units Hz K Pa W K Pa
Independent Variables (inputs):
Outer loop: 3e HX:HeatI Beg= 9.50E+02 End= 50. Step= -33.
84
Figure 5.8: Standing-wave engine results. Lines are DeltaEC results; points are from experimental
data in Ref. [23].
Figure 5.9: More thermoacoustic segments: stacks, regenerators, and heat exchangers.
85
STKCIRC A stack or regenerator with circular pores. We also use this to model hexagonal
honeycomb stacks.
STKDUCT A “stack” with lateral dimensions much larger than , computed in boundary-
layer approximation. (Used for pulse tubes and thermal bu¤er tubes.)
STKCONE A tapered “stack” with lateral dimensions much larger than , computed in
boundary-layer approximation. (Used for pulse tubes and thermal bu¤er tubes.)
STKPOWERLW A porous medium with friction factor and heat-transfer coe¢ cient speci…ed as
power-law functions of Reynolds number.
TX Tube-array heat exchanger, with the thermoacoustic working gas inside the tubes.
SX Stacked-screen heat exchangers, valid only for greater than hydraulic radius.
PX Power-law heat exchangers, with friction factor and heat-transfer coe¢ cient characterized
by power laws in Reynolds number.
VXT1 and VXT2 Heat exchangers with …xed solid temperature, in which gas temperature
can vary with x:
VXQ1 and VXQ2 Heat exchangers with …xed heat per unit length, in which gas temperature
can vary with x:
ANCHOR This segment overrides what would otherwise be default thermal-insulation mode
in subsequent DUCTs, CONEs, COMPLIANCEs, IMPEDANCEs, SURFACEs, etc., so DeltaEC
treats those segments as if they are immersed in a thermal bath at T = Tm :
86
Restore all guessez. Before beginning iterations during a run, DeltaEC saves copies of
the guess values. It also saves copies of a few output parameters if it caught the user
inadvertently or boldly using a sameas or RPN pointer to an output parameter in a
segment downstream of that sameas or RPN, despite DeltaEC’s “Forward Reference
Condition”warning about the unstable shooting-method behavior that can result from
using such downstream results. (See Section 5.4 for an explanation of this danger.)
Whenever an unsuccessful run overwrites the guesses (leaving you and DeltaEC lost),
you can use this option to restore all of these variables to their earlier values. Simply
Restore guessez, then modify one or more variables yourself (e.g., returning anything
you recently changed to earlier, more successful values), and try again.
Comment lines. If the heading on the …rst line of a segment does not provide enough
space for the user’s notes, additional comment lines can be added below it. Examples
are shown at the beginning of the next section, Section 5.8.1.
Split segment. This option, available by right clicking on the top line of a segment or in
the Edit pulldown menu, automates the process of splitting any segment with length
(e.g., a DUCT, CONE, or STK**) into two segments, each having half of the original length,
correcting the sameas and RPN references everywhere in the model, and correcting the
guess, target, and plot lists. (To partition the lengths unequally while keeping the
correct total length, you can use master–slave linking (Section 2.7.2) after splitting
a segment in equal halves: slave the …rst length to the second, then modify the …rst
length, then clear the length link to allow the two pieces to be modi…ed independently.)
AutoLayout. Sometimes auxiliary windows such as plots and the highlight display become
lost o¤ the edge of the computer’s screen. (For example, undocking a laptop from
a larger screen can cause this.) The AutoLayout option, available in the Display
pulldown menu, brings all of DeltaEC’s auxiliary displays close to DeltaEC itself.
(If this fails for a plot, close it, delete the associated <.ssv> or <.isv> from your
model’s folder, and try again.)
Flip a block of segments. (Implemented in version 6.4.) The number of guesses and tar-
gets can sometimes be reduced by beginning the integration of a model from what
you had previously considered the “end.”Ori…ce pulse-tube refrigerators (described in
Section 5.8) are a particularly good example because we naturally think of them as
beginning at the driver, but they “end” with an accurately known impedance in the
COMPLIANCE tank, while the driver impedance is usually less well known; BEGINning
in the tank can often reduce the number of guesses and targets. Similarly, shooting-
method stability can sometimes be improved by putting regenerators upstream of
thermal bu¤er tubes, instead of the other way around. The “Flip Selected Segs”oper-
ation automates switching back and forth between these two directions of integration,
simplifying an otherwise very tedious and error-prone process. Highlight a block of
segments, such as an entire model or the segments between a TBRANCH and its SOFTEND,
and right-click to access the “Flip Selected Segs” feature. Flipping reverses the order
of the physical segments in the block, and interchanges parameters such as AreaI and
AreaF within unsymmetrical segments such as CONEs. DeltaEC tries to relocate any
87
RPN segments intelligently in the block, but sometimes gives up and puts them at the
end of the block. “Flip Selected Segs” takes care of sameas pointers, RPN-segment
pointers, and plot references, and preserves the guess and target lists. Always proof-
read very carefully after ‡ipping; in complicated models there is always something
that DeltaE handles incorrectly, and intelligent human intervention will be needed
to …x things like TBRANCH impedances feeding into ‡ipped segment blocks and targets
getting moved upstream of the corresponding guesses.
88
Figure 5.10: A Stirling cryocooler.
The physical segments consist of a …rst heat exchanger, at 300 K, the regenerator, and a
second heat exchanger at 80 K. All three are stacked screens. The other segments— BEGIN
and the RPNs— de…ne the end conditions.
The hydraulic radius rh and volumetric porosity for the screens can be calculated from
the screen manufacturer’s specs, using expressions from Ref. [27]:
md
' 1 ; (5.14)
4
d
rh = ; (5.15)
41
where d is wire diameter and m is mesh number (i.e., number of wires per unit length). The
second equation results from the conventional de…nition of hydraulic radius: Gas volume
89
Figure 5.11: Total power (solid black) and acoustic power (dashed blue) vs position in the Stirling
cryocooler example. The total power H_ tot is constant and small in the regenerator, and changes in
the heat exchangers.
divided by gas–solid interface area. This regenerator is a little over 1 cm in diameter and is
5 cm long. The heat exchangers are the same diameter but only 1 mm long.
DeltaEC estimates the temperature di¤erence between the helium gas and the copper
screen wires in the heat exchangers, but it has no provision for estimating the tempera-
ture di¤erence between the screen wires and the “housing” in which they are mounted (a
temperature di¤erence caused by the …nite thermal conductance of the screen wires them-
selves). This is not a serious concern for small machines, but should be checked by hand on
a case-by-case basis.
Line e in the regenerator segment, “ksFrac,” is the fudge factor by which longitudinal
conduction through the regenerator is adjusted due to the spatially intermittent thermal
contact between adjacent screens and due to the conduction of the pressure-vessel wall.
Following Ref. [28], we often set ksFrac somewhere between 0.1 and 0.3, the lower end of
the range being appropriate if the conduction of the pressure-vessel wall is negligible.
Our point of view with respect to boundary conditions in this example is most easily
displayed by running DeltaEC on this …le and examining the guess–target summary,
GUESS 0c 0d 0e 1e 3e
name BEGIN:TBeg BEGIN: |p| BEGIN:Ph(p) SX:HeatI SX:HeatI
units K Pa deg W W
value 300.14 2.8459E+05 -43.119 -35.833 4.9896
TARGET 1f 3f 4a 5a 6a
name SX:Est-T SX:Est-T RPN:Targe RPN:Targe RPN:Targe
units K K
value 300.00 80.000 0.0000 6.2000E-05 -52.000
result 300.00 80.000 -2.2675E-12 6.2000E-05 -52.000
and the BEGIN segment. We are considering the complex volume ‡ow rates at the two ends
to be given, as if we have in mind an “alpha” Stirling machine, with the motions of the
two pistons determining the volumes of the compression and expansion spaces. The volume
‡ow rate at the hot end is set by lines 0f and 0g in the BEGIN segment. The 0 phase of
90
line 0g essentially determines the zero of phase for the entire system. The volume ‡ow rate
3:65 10 4 m3 /s of line 0f (together with the frequency set in line 0b) implies a volumetric
stroke of 2.1 cm3 peak-to-peak at the hot end. The RPNs at the cold end ensure that
DeltaEC’s shooting method arrives there with the desired cold piston stroke and phase.
To arrive at these two targets, DeltaEC adjusts two guesses: the pressure amplitude and
phase in the BEGIN segment (and hence throughout the cooler). We also insist that the solid
temperatures in the two heat exchangers be 300 K and 80 K; DeltaEC achieves these two
targets by adjusting two more guesses: the heat extracted at the hot heat exchanger, and
the temperature in the BEGIN segment. We also insist that the acoustic power and total
power leaving the cold heat exchanger are equal, implying that the cold piston is thermally
insulated. (Acoustic power and total power are also equal at the ambient piston, because
the BEGIN segment assigns H_ 2;k = E_ by default.)
DeltaEC predicts that, under these circumstances, the cooler will reject 36 W at the
hot heat exchanger and will have a cooling power of 5 W. This cooling power accounts for
heat conduction and enthalpy ‡ow through the regenerator, but does not account for any
heat load imposed by frictional irreversibilities in the cold piston, nor any heat load imposed
by the regenerator case conduction except to the extent that it is included in ksFrac.
We now make or suggest a few simple modi…cations to this …le to illustrate additional
features of DeltaEC.
To discover what temperature the cooler would maintain with a heat load of 10 W
instead of 5 W, we remove 3f— the cold heat exchanger temperature— from the target list,
and we remove 3e— the cooling power— from the guess list, and modify it to 10 W. Running
DeltaEC shows that the shooting method is lost with such a dramatic change, so we Restore
guessez, modify 3e back to 5 W, and use an incremental plot of 3e from 5 W to 10 W, step
0.2 W, to reach the desired heat load. Under these circumstances the cold temperature will
be 232 K.
Using 3e as an independent plot variable running from 10 W to 2 W with steps of, say,
0.2 W, and using 3H (cold solid temperature) as a dependent plot variable will generate a
table and plot of cold temperature (and other defaults) vs heat load.
Insertion of two *ESPEAKER segments, before the aftercooler and after the cold heat
exchanger, would model use of linear motors and pistons there.
In the next subsection, we will use TBRANCH and UNION to change this model from an
“alpha” Stirling machine to a “beta” or “gamma” machine, with one power piston on the
hot end and a displacer piston in parallel with the heat exchange elements.
91
Figure 5.12: “Gamma”-style Stirling machine.
of the energy ‡ow H_ tot going down the TBRANCH is guessed, and the temperature at the
SOFTEND of the TBRANCH is targeted to equal the temperature at the UNION. Models with
TBRANCHs and UNIONs tend to have many guesses and targets, since every SOFTEND and
UNION contributes two or three targets (and a few more targets are almost always needed
for temperatures, heats, etc.). Keep in mind that UNION targets di¤er from the usual RPN,
**END and heat-exchanger-temperature targets because their actual values are not speci…ed
by the user— they vary from one integration to the next, depending on what is happening
at the associated SOFTEND segment, as the shooting method varies the guesses.
As an example of use of TBRANCH and UNION in a thermoacoustic setting, we return to the
Stirling cryocooler example above, and convert it to a “displacer”-style [26] Stirling machine,
with a compressor piston at the hot end and a driven displacer piston connecting the hot
and cold ends. In the previous example, P U power (i.e., E_ in DeltaEC vocabulary) ‡owed
in at the BEGIN and out at the end; no explicit segments indicated the pistons there. Here,
with a displacer piston, the cold-end P U power is returned automatically to the hot end,
reducing the hot-end P U power requirement.
The apparatus layout is illustrated in Fig. 5.12. The corresponding display of segment
headings is
0 BEGIN:G P Initialize things
1 COMPLIANCE: space around power-piston motor
2 IESPEAKER: the power piston
3 COMPLIANCE: a little space in front of the power piston
4 TBRANCH:G branch to displacer
5 IESPEAKER: a spring-mounted, driven moving mass
6 RPN:G reassign Tm (Displacer is "axially insulating.")
7 COMPLIANCE: a little volume for the connection
8 SOFTEND: reconnect at UNION
9 SX :GT aftercooler
10 STKSCREEN: regenerator
11 SX :GT cold heat exch
12 UNION: T displacer cold end
13 HARDE: T close the end
14 RPN : T simulate insulation in the displacer piston
and the guess–target summary is the largest we have yet encountered in the Users Guide:
GUESS 0c 0d 0e 4a 4b 4c 6a 9e 11e
name BEGIN:TBeg BEGIN: |p| BEGIN:Ph(p) TBRAN:Re(Zb TBRAN:Im(Zb TBRAN:HtotB RPN:Guess SX:HeatI SX:HeatI
units K Pa deg Pa-s/m^3 Pa-s/m^3 W W W
value 300.22 6373.3 114.43 -3.0131E+09 -3.2836E+09 -6.7651 79.962 -45.267 4.5794
92
TARGET 9f 11f 12b 12c 12d 13a 13b 13c 14a
name SX:Est-T SX:Est-T UNION:|p|SO UNION:Ph(p) UNION:Tm HARDE:R(1/z HARDE:I(1/z HARDE:Htot RPN:Targe
units K K Pa deg K W
value 300.00 80.000 2.5375E+05 -44.127 79.962 0.0000 0.0000 0.0000 0.0000
result 300.00 80.000 2.5375E+05 -44.127 79.962 8.9507E-14 2.8896E-13 -2.499E-11 -4.838E-11
One can think of these guesses and targets as paired up in the following way. The TBeg
guess lets DeltaEC hit the T-ambient target. (These two are so nearly equal, and so
trivially related, that they could easily be omitted.) The two branch-impedance guesses let
DeltaEC reach the two p1 targets at the UNION. The branch-Htot guess lets DeltaEC
hit the RPN target at segment 13, which enforces thermal insulation on the cold end of the
displacer piston. The temperature-reassignment guess, which lets us imagine the displacer
piston to be axially insulating, trivially targets the UNION temperature. The heat removed
at the hot heat exchanger lets DeltaEC reach the target temperature at the cold heat
exchanger. Finally, the two p1 guesses in the BEGIN segment and the heat in the cold heat
exchanger allow DeltaEC to achieve U1 = 0 and H_ tot = 0 at the HARDEND at the end of the
apparatus.
The corresponding <.out> …le is
TITLE Stirling cooler w displacer piston, illustrating TBRANCH--UNION
!->C:\Users\092710\Documents\DeltaEC\PySource\Examples\EnginesAndRefr\gamma2.out
!Created@11:48:32 21-Mar-2016 with DeltaEC version 6.4b2.1!under win32,
using Win 6.1.7601 (Service Pack 1) under Python 2.7.9.
!--------------------------------- 0 ---------------------------------
BEGIN Initialize things
2.0000E+6 a Mean P Pa
55.000 b Freq Hz
300.22 c TBeg K G
6373.3 d |p| Pa G
114.43 e Ph(p) deg G
0.0000 f |U| m^3/s
0.0000 g Ph(U) deg
helium Gas type
!--------------------------------- 1 --------------------------
COMPLIANCE space around power-piston motor
0.1000 a SurfAr m^2 6373.3 A |p| Pa
5.0000E-4 b Volume m^3 114.43 B Ph(p) deg
3.3041E-4 C |U| m^3/s
23.551 D Ph(U) deg
0.0000 E Htot W
ideal Solid type -1.6157E-2 F Edot W
!--------------------------------- 2 --------------------------
IESPEAKER the power piston
2.0000E-4 a Area m^2 2.9239E+5 A |p| Pa
1.0000 b R ohms -39.277 B Ph(p) deg
0.0000 c L H 3.3020E-4 C |U| m^3/s
10.000 d BLProd T-m 23.621 D Ph(U) deg
8.3000E-2 e M kg 40.076 E Htot W
1.0000E+4 f K N/m 21.992 F Edot W
0.0000 g Rm N-s/m 40.076 G WorkIn W
6.0000 h |I| A 19.931 H Volts V
140.00 i Ph(I) deg 6.0000 I Amps A
-312.09 J Ph(V/I) deg
2.9812E+5 K |Px| Pa
ideal Solid type -39.82 L Ph(Px) deg
!--------------------------------- 3 ---------------------------------
COMPLIANCE a little space in front of the power piston
1.0000E-3 a SurfAr m^2 2.9239E+5 A |p| Pa
2.0000E-6 b Volume m^3 -39.277 B Ph(p) deg
2.7635E-4 C |U| m^3/s
18.316 D Ph(U) deg
40.076 E Htot W
ideal Solid type 21.652 F Edot W
!--------------------------------- 4 ---------------------------------
TBRANCH branch to displacer
-3.0131E+9 a Re(Zb) Pa-s/m^3 G 2.9239E+5 A |p| Pa
-3.2836E+9 b Im(Zb) Pa-s/m^3 G -39.277 B Ph(p) deg
-6.765 c HtotBr W G 6.5609E-5 C |U| m^3/s
93.263 D Ph(U) deg
-6.765 E HtotBr W
-6.4851 F EdotBr W
28.137 G EdotTr W
!--------------------------------- 5 ---------------------------------
93
IESPEAKER a spring-mounted, driven moving mass
5.0000E-5 a Area m^2 2.5375E+5 A |p| Pa
0.5000 b R ohms -44.127 B Ph(p) deg
0.0000 c L H 6.5763E-5 C |U| m^3/s
1.0000 d BLProd T-m 93.397 D Ph(U) deg
7.0000E-3 e M kg -6.154 E Htot W
1190.0 f K N/m -6.1539 F Edot W
0.0000 g Rm N-s/m 0.6110 G WorkIn W
1.0000 h |I| A 1.6426 H Volts V
-30.00 i Ph(I) deg 1.0000 I Amps A
-41.929 J Ph(V/I) deg
4.4991E+4 K |Px| Pa
ideal Solid type 169.20 L Ph(Px) deg
!--------------------------------- 6 ---------------------------------
RPN Reassign Tm. (Displacer is "axially insulating.")
79.962 a G or T G 79.962 A Kelvin
inp =Tm
!--------------------------------- 7 ---------------------------------
COMPLIANCE a little volume for the connection
1.0000E-6 a SurfAr m^2 2.5375E+5 A |p| Pa
5.0000E-7 b Volume m^3 -44.127 B Ph(p) deg
5.7702E-5 C |U| m^3/s
103.08 D Ph(U) deg
-6.154 E Htot W
ideal Solid type -6.154 F Edot W
!--------------------------------- 8 ---------------------------------
SOFTEND reconnect at UNION
0.0000 a Re(z) 2.5375E+5 A |p| Pa
0.0000 b Im(z) -44.127 B Ph(p) deg
0.0000 c Htot W 5.7702E-5 C |U| m^3/s
103.08 D Ph(U) deg
-6.154 E Htot W
-6.154 F Edot W
-29.176 G Re(z)
-18.80 H Im(z)
79.962 I T K
!--------------------------------- 9 ---------------------------------
SX aftercooler
sameas 10a a Area m^2 2.8988E+5 A |p| Pa
0.6000 b VolPor -39.755 B Ph(p) deg
1.0000E-3 c Length m 2.6449E-4 C |U| m^3/s
sameas 10d d rh m 4.0412 D Ph(U) deg
-45.267 e HeatIn W G 1.5747 E Htot W
300.00 f SolidT K =9H 27.670 F Edot W
300.22 G GasT K
copper Solid type 300.00 H SolidT K
!--------------------------------- 10 ---------------------------------
STKSCREEN regenerator
1.1670E-4 a Area m^2 2.5407E+5 A |p| Pa
0.6860 b VolPor -44.169 B Ph(p) deg
5.0000E-2 c Length m 5.6226E-5 C |U| m^3/s
1.3900E-5 d rh m -74.284 D Ph(U) deg
0.3000 e ksFrac 1.5747 E Htot W
6.1784 F Edot W
300.22 G TBeg K
stainless Solid type 79.962 H TEnd K
!--------------------------------- 11 ---------------------------------
SX cold heat exch
sameas 10a a Area m^2 2.5375E+5 A |p| Pa
0.6000 b VolPor -44.127 B Ph(p) deg
1.0000E-3 c Length m 5.7702E-5 C |U| m^3/s
sameas 10d d rh m -76.923 D Ph(U) deg
4.5794 e HeatIn W G 6.1540 E Htot W
80.000 f SolidT K =11H 6.1540 F Edot W
79.962 G GasT K
copper Solid type 80.000 H SolidT K
!--------------------------------- 12 ---------------------------------
UNION displacer cold end
8 a SegNum 2.5375E+5 A |p| Pa
sameas 8A b |p|Sft Pa =12A -44.127 B Ph(p) deg
sameas 8B c Ph(p)S deg =12B 8.9952E-15 C |U| m^3/s
sameas 8I d TSoft K =12G 99.432 D Ph(U) deg
-5.4001E-13 E Htot W
-9.1812E-10 F Edot W
79.962 G T K
!--------------------------------- 13 ---------------------------------
HARDEND close the end
0.0000 a R(1/z) =13G 2.5375E+5 A |p| Pa
0.0000 b I(1/z) =13H -44.127 B Ph(p) deg
0.0000 c Htot W =13E 8.9952E-15 C |U| m^3/s
99.432 D Ph(U) deg
-5.4001E-13 E Htot W
-9.1812E-10 F Edot W
-1.5481E-12 G R(1/z)
1.1431E-12 H I(1/z)
!--------------------------------- 14 ---------------------------------
94
RPN simulate insulation in the displacer piston
0.0000 a =A? 1.0729E-12 A ChngeMe
7E 7F -
! The restart information below was generated by a previous run
! and will be used by DeltaEC the next time it opens this file.
guessz 0c 0d 0e 4a 4b 4c 6a 9e 11e
xprecn -3.8026E-3 7.5295E-3 -4.8956E-5 -2.2194E+4 1.4047E+4 -3.6186E-5 -1.5964E-4 4.3401E-4 -3.0683E-5
targs 9f 11f 12b 12c 12d 13a 13b 13c 14a
We are using IESPEAKER segments for both the displacer and the driver, because linear
motors and loudspeakers share the same physical transduction mechanism.
The user will soon discover that this is a surprisingly robust model, considering the large
number of guesses and targets: The model tolerates steps in many independent variables of
several percent without getting lost. The user might next generate cooling-power curves by
using the cold-temperature target as an independent plot variable and the cooling power as
dependent plot variable; or the user might explore the frequency dependence of the cooler,
by using frequency as an independent plot variable.
95
1.1381E+5 A |p| Pa
-14.781 B Ph(p) deg
1.2179E-3 C |U| m^3/s
-30.009 D Ph(U) deg
63.991 E Htot W
66.871 F Edot W
149.91 G TBeg K
149.89 H TEnd K
!--------------------------------- 5 ---------------------------------
STKDUCT Pulse tube
5.6870E-5 a Area m^2 Mstr 1.0284E+5 A |p| Pa
2.6740E-2 b Perim m 5a -43.845 B Ph(p) deg
0.2000 c Length m 1.3033E-3 C |U| m^3/s
1.0000E-5 d WallA m^2 -51.003 D Ph(U) deg
63.991 E Htot W
66.491 F Edot W
149.89 G TBeg K
stainless Solid type 300.23 H TEnd K
!--------------------------------- 6 ---------------------------------
JOIN pulse-tube end effects
1.0284E+5 A |p| Pa
-43.845 B Ph(p) deg
1.2882E-3 C |U| m^3/s
-51.003 D Ph(U) deg
63.991 E Htot W
65.724 F Edot W
300.23 G TBeg K
300.20 H TEnd K
!--------------------------------- 7 ---------------------------------
SX Hot heat exchanger
sameas 5a a Area m^2 2.2011E+4 A |p| Pa
0.6900 b VolPor -19.722 B Ph(p) deg
5.0000E-3 c Length m 1.2889E-3 C |U| m^3/s
6.4500E-5 d rh m -51.331 D Ph(U) deg
-63.991 e HeatIn W G 2.8422E-13 E Htot W
300.00 f SolidT K =7H 12.081 F Edot W
300.20 G GasT K
copper Solid type 300.00 H SolidT K
!--------------------------------- 8 ---------------------------------
IMPEDANCE The orifice
1.0000E+7 a Re(Zs) Pa-s/m^3 1.2938E+4 A |p| Pa
0.0000 b Im(Zs) Pa-s/m^3 11.755 B Ph(p) deg
1.2889E-3 C |U| m^3/s
-51.331 D Ph(U) deg
2.8422E-13 E Htot W
3.7742 F Edot W
!--------------------------------- 9 ---------------------------------
DUCT inertance
1.0000E-5 a Area m^2 Mstr 2.2780E+4 A |p| Pa
1.1210E-2 b Perim m 9a -141.21 B Ph(p) deg
3.0000E-2 c Length m 1.2882E-3 C |U| m^3/s
5.0000E-4 d Srough -51.345 D Ph(U) deg
2.8422E-13 E Htot W
ideal Solid type 3.3272E-2 F Edot W
!--------------------------------- 10 ---------------------------------
COMPLIANCE Reservoir volume
1.2680E-2 a SurfAr m^2 10b 2.2780E+4 A |p| Pa
1.5000E-4 b Volume m^3 Mstr -141.21 B Ph(p) deg
1.6700E-12 C |U| m^3/s
2.0161 D Ph(U) deg
2.8422E-13 E Htot W
ideal Solid type -1.5237E-8 F Edot W
!--------------------------------- 11 ---------------------------------
HARDEND The end
0.0000 a R(1/z) =11G 2.2780E+4 A |p| Pa
0.0000 b I(1/z) =11H -141.21 B Ph(p) deg
0.0000 c Htot W =11E 1.6700E-12 C |U| m^3/s
2.0161 D Ph(U) deg
2.8422E-13 E Htot W
-1.5237E-8 F Edot W
-2.8802E-8 G R(1/z)
2.1522E-8 H I(1/z)
96
Figure 5.13: An ori…ce pulse-tube refrigerator (OPTR). The ori…ce can be purely resistive, as
shown here, but adding inertance in series with the resistive ori…ce improves performance.
DUCT, and COMPLIANCE, respectively. Our use of zero for the imaginary part of the IMPEDANCE
re‡ects our intention that this ori…ce should truly be resistive, with its pressure drop in phase
with mass ‡ux.
For purposes of illustration here, we will regard the geometry of the apparatus as given,
and will explore its performance for a given, …xed input jp1 j. The guess–target summary
indicates our point of view:
GUESS 0c 0f 0g 1e 3e 7e
name BEGIN:TBeg BEGIN: |U| BEGIN:Ph(U) SX:HeatI SX:HeatI SX:HeatI
units K m^3/s deg W W W
value 300.10 6.9646E-03 52.754 -455.16 13.312 -63.991
TARGET 1f 3f 7f 11a 11b 11c
name SX:Est-T SX:Est-T SX:Est-T HARDE:R(1/z HARDE:I(1/z HARDE:Htot
units K K K W
value 300.00 150.00 300.00 0.0000 0.0000 0.0000
result 300.00 150.00 300.00 3.6105E-09 -2.7741E-09 -2.8422E-14
Three of the six targets …x the hot and cold temperatures at 300 K and 150 K. We leave the
amplitude of the oscillatory pressure at the BEGINning at 8% of mean pressure, and leave
the frequency …xed at 300 Hz. Hence, we are asking: What is the cooling power at 150
K, and how much input power, volume ‡ow rate, etc. are required, for …xed frequency and
pressure amplitude, and for this geometry? The result, given in the …le listings above: 13.3
W of cooling power, requiring over 500 W of input power from the compressor.
As with most OPTR models we have worked with in DeltaEC, this one is not par-
ticularly “robust.” An abrupt change of a typical variable by 20% or 30% will likely cause
DeltaEC to get hopelessly lost, unable to adjust its guesses to hit the targets. Much of the
shooting-method “fragility”of OPTR models (and other models with a STKDUCT or STKCONE)
in DeltaEC is due to the fact that small changes in variables near the BEGINning, such as
p1 ; U1 ; and the heat removed at a heat exchanger, have a large e¤ect on temperatures at
the end of the pulse tube. Some of the fragility is due to the fact that OPTR models typi-
cally have a large number of guesses and targets. When you encounter a fragile DeltaEC
model, try to reduce the number of guesses and targets as much as possible (particularly in
initial design explorations when you are changing things dramatically) and, once you have
a convergent model, make only small changes in variables. Try negative values of the Nor-
malization Mode,1 accessible in the Edit, Options pulldown menu; we often …nd that Mode
1
As described in Section 8.3.2, the value of Mode is stored in the <.ini> …le, and it is often convenient
to rename a model’s <.ini> …le to be <default.ini> so all models in a given folder can automatically
bene…t from a good value of Mode.
97
Figure 5.14: Plotting <optr.sp> using Re[p1 ] as the horizontal axis and Im[p1 ] as the vertical axis
yields a pressure-phasor plot. The SegEndMarks option tags the curve with numbers indicating
the ends of the segments. Cloning the plot and switching p1 to U1 yields a volume-‡ow-rate phasor
plot. These can also be created with Phasor Mode in the plotter’s Options pulldown menu.
98
= 4 works much better than DeltaEC’s default Mode = 1. Try reducing DeltaEC’s
convergence tolerance (Section 8.3.2) if you have to use more than 5 guesses and 5 targets.
Remember to subdivide a large change in a variable into a series of small changes, using an
incremental plot. A fast computer and frequent hard-disk saving of satisfactory converged
models will minimize frustration. See Section 5.4 for additional tips.
Examination of the pulse tube segment STKDUCT above shows a possible problem: The
pulse tube …gure of merit H_ tot =E;
_ which can theoretically be as high as 1.000 but is usually
more like 0.8, may be unrealistically high: H_ tot =E_ ' 64.0 W=66.5 W' 0:96. DeltaEC
knows nothing about jet- or streaming-driven convection, and pulse-tube experimentalists
are only beginning to learn how to reliably avoid such convection. (For a discussion of
streaming-driven convection, see Refs. [29] or [30], and [31] and [13].) To force DeltaEC
to accommodate a reduced pulse-tube …gure of merit, the user can introduce an RPN and an
additional guess–target pair. The RPN instruction line can compute the ratio of H_ tot to E_
in the pulse tube, and the target value in line a can be set to something like 0.8. The user
can simulate the thermal loading of streaming-driven convection, etc. by letting DeltaEC
guess an unphysically large value for the cross section of the pulse tube solid wall (line 4d),
which then conducts signi…cant heat from hot to cold, allowing DeltaEC to meet its target
of 0.8. We will not do so here.
Our choices for nearly all input variables in this model were random, not guided by
good design sense or any real cryocooler. In particular, our choice of Re[Zs ] = 1 107 for
the ori…ce impedance above was random. To …nd a better ori…ce setting, use Re[Zs ] as an
independent plot variable, letting it range from 1 107 to 1 108 . The cooling power peaks
at 13.6 W for Re[Zs ] = 2:3 107 : This is not much improvement. For guidance on what to
try changing next to improve the performance, we consider exergy in the next section.
99
as expected.) DeltaEC’s ambient temperature T0 is set to 300 K by default in the BEGIN
segment; this default can be overridden by using =Tzero in a subsequent RPN segment.
The usefulness of the concept of exergy arises from its relationship to irreversible entropy
generation, expressed by the Guoy-Stodola theorem [32],
X_ = T0 S_ gen : (5.17)
In this simpli…ed expression of the theorem, X_ represents the di¤erence in exergy ‡ow
between two locations in an apparatus between which exergy cannot ‡ow in or out laterally,
and the theorem says that this di¤erence is the product of T0 and the rate of irreversible
entropy generation between those two locations. In a perfectly e¢ cient apparatus, there is
no irreversible entropy generation, so X_ = 0 (except where X_ ‡ows in or out, e.g., at a
heat exchanger or a **DUCER or **SPEAKER). Thus, a graph of X_ vs x can help show the
user what parts of an apparatus are most responsible for its ine¢ ciency.
For example, consider the OPTR described in Section 5.8.3. The <.sp> plot of E; _ H_ tot ;
_
and X is shown in Fig. 5.15. Most of the exergy is lost in the regenerator, and the rate of
exergy loss is steepest near the ambient end of the regenerator. This suggests that viscous
losses in the regenerator may be severe. Indeed, the <.out> …le shows a large pressure
drop in the regenerator. The user might try changing the dimensions of the regenerator
to improve the performance. Figure 5.15 also shows signi…cant drops in X_ in the ambient
and cold heat exchangers. There, the analysis of the situation is a little more complicated,
because the heats ‡owing into a heat exchanger contribute exergy according to [13]
X_ = Q_ (1 T0 =Tm ) ; (5.18)
where T0 is ambient temperature and Tm is the local temperature. At the ambient heat
exchanger, Tm = T0 ; so the drop in exergy there is entirely lossy. At the cold heat exchanger,
Tm 6= T0 ; but the loss in thermoacoustic exergy due to Q_ is only 15 W, so most of the drop
in X_ in Fig. 5.15 at the cold heat exchanger is lossy. Thus, the heat exchangers are also
contributing signi…cantly to the poor performance in this OPTR. At this point, the user
might suspect that the poor performance of this OPTR arises from trying to force too much
‡ow through pores that are too tight or an area that is too small.
Subsequent chapters describe additional steady-‡ow and mixture-separation features in
DeltaEC. These phenomena also contribute to X; _ and the terms that they add to Eq. (5.16)
are given in Eq. (8.9).
100
Figure 5.15: Graph of total power (black solid line), acoustic power (blue dashed line), and exergy
‡ow (green dot-dashed line) in the OPTR example. The locations where H_ tot changes are the three
heat exchangers.
101
6. Steady flow, too
The superposition of steady ‡ow and oscillating ‡ow creates many interesting and important
e¤ects that can be modeled with DeltaEC. A nonzero steady ‡ow profoundly a¤ects Tm (x)
in stacks and regenerators, creates Doppler shifts [33] and increases the turbulent dissipation
of acoustic power in DUCTs, and causes a small x dependence in the time-averaged pressure.
In systems such as Reid’s refrigerator (shown below in Fig. 6.1), externally forced steady
‡ow provides the ability to deliver cold gas from a thermoacoustic refrigerator without
a cold heat exchanger. In more conceptually subtle situations, the acoustic oscillations
themselves cause the steady ‡ow, which is then called “streaming” by acousticians [34]. In
some circumstances, such as the thermoacoustic-Stirling hybrid engine shown in Fig. 1.7 or
6.4, we try to understand the streaming well enough to eliminate it. In other circumstances,
such as the self-circulating heat exchanger shown in Fig. 1.9, we try to make streaming as
large as possible without undue dissipation of acoustic power.
is the power that obeys the …rst law of thermodynamics for open control volumes. It is the
sum of the second-order time-averaged thermoacoustic enthalpy ‡ux, the ordinary thermal
conduction in the x direction, and H_ N_ = N_ mwm :
In DeltaEC, mean enthalpy is referenced to zero at temperature T0 , which is set to 300
Kelvin by default in the …rst BEGIN segment of a model. (This default can be overridden
1
DeltaEC’s moles are gram-moles: 6:022 1023 particles (e.g., helium’s molar mass is 0.004002 kg/mol).
We might have chosen to use mass ‡ow or volume ‡ow instead of molar ‡ow as DeltaEC’s steady-‡ow
parameter. We rejected volume ‡ow because it changes with x when Tm depends on x; which seemed
unnecessarily confusing. We chose molar ‡ow instead of mass ‡ow for convenience in the context of mixture
separation, which is introduced in Chapter 7.
102
by using the =Tzero operation in an RPN segment.) For an ideal gas, Eq. (6.1) can then be
regarded as
H_ N_ = N_ mcp (Tm T0 ) : (6.3)
Nonzero N_ also contributes a term
to exergy ‡ow X: _ Like the enthalpy per unit mass wm ; the entropy per unit mass sm is
2
referenced to zero at T0 :
DeltaEC also introduces a new integration variable, the second-order time-averaged
head-loss pressure p2;0;HL ; to keep track of the small but important e¤ects of steady ‡ow
on the time-averaged pressure. The second-order head-loss pressure p2;0;HL is initialized
at zero in BEGIN and evolves with x in most other segments. When N_ and p2;0;HL are
involved in a DeltaEC calculation, the numerical-integration picture that is summarized
in Eqs. (5.6)–(5.9) for thermoacoustics with N_ = 0; namely
dp1
= Fmomentum (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
dU1
= Fcontinuity (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
dTm
= Fx-energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.),
dx
dH_ tot
= Flateral energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; geometry, gas properties, etc.).
dx
must be replaced by this slightly more complicated picture:
dp1
= Fmomentum (p1 ; U1 ; Tm ; H_ tot ; pm ; !; N_ ; geometry, gas properties, etc.),
dx
(6.5)
dU1
= Fcontinuity (p1 ; U1 ; Tm ; H_ tot ; pm ; !; N_ ; geometry, gas properties, etc.),
dx
(6.6)
dTm
= Fx-energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; N_ ; geometry, gas properties, etc.),
dx
(6.7)
dH_ tot
= Flateral energy (p1 ; U1 ; Tm ; H_ tot ; pm ; !; N_ ; geometry, gas properties, etc.),
dx
(6.8)
dp2;0;HL
= Fhead loss (p1 ; U1 ; Tm ; H_ tot ; pm ; !; N_ ; geometry, gas properties, etc.).
dx
(6.9)
2
Entropy per unit mass is referenced to zero at T = T0 and p = pm : Since all parts of a thermoacoustic
apparatus share the same pm ; this unconventional reference pressure does not a¤ect any entropy or exergy
di¤ erences [e.g., Eq. (5.17)] within a DeltaEC model.
103
Please note that dp2;0;HL =dx is usually nonzero even when N_ = 0, because dp2;0;HL =dx
arises from Z
_ 1
m U2;0 ' N m e1 ] dAgas ;
Re [ 1 u (6.10)
2
where m is the molar mass, as explained in more detail near Eq. (8.15).
Our con…dence in the accuracy of the DeltaEC implementation of these equations is
lower than in the equations described in Chapters 2 and 5. We are very con…dent of the e¤ect
of nonzero N_ on Fx-energy and Flateral energy under all circumstances we have encountered, and
con…dent of the e¤ect of nonzero N_ on all the equations when the steady and oscillating ‡ows
are both laminar, but for turbulent ‡ow our understanding of Fhead loss and of the e¤ects of
nonzero N_ on Fmomentum and Fcontinuity are still evolving, and are mostly based on the quasi-
steady approximation, namely that the behavior at any instant of time t is the same as it
would be for steady ‡ow with volume ‡ow rate U = N_ m= m +Re[U1 ei!t ]; an approximation
with obvious shortcomings in acoustics where inertial e¤ects of gas acceleration are usually
as important as anything else. The complicated details of the current steady-‡ow DeltaEC
implementation are given in Chapter 10.
One other variable is sometimes useful in the context of nonzero N_ : The (total) second-
order, time-averaged pressure p2;0;tot corresponds to pressure changes as measured by side-
wall pressure transducers [35], because p2;0;tot is given by
where the reversible part of the second-order time-averaged pressure [36] is given by
jp1 j2 mjU1 j2
p2;0;rev (x) = : (6.12)
4 m a2 4A2gas
The variable p20tot is available in RPN segments. Note that it depends on area, which may
not have a sensible value after segments without area, such as BEGIN, TBRANCH, UNION, or
IMPEDANCE.
User-de…ned gases speci…ed by <.tpf> …les cannot be used with N_ enabled, because Eq.
(6.1) requires m, which is not available in <.tpf> format. To circumvent this shortcoming
when a user-de…ned gas must be used with nonzero N_ , use a <.tpm> …le and set the mole
fraction of one component to zero.
In the rest of this chapter, we illustrate DeltaEC’s steady-‡ow features with three
examples. In the …rst example, nonzero N_ is externally driven and p2;0;HL is of little interest.
In the second example, N_ = 0 is desired, but must be accomplished through careful balancing
of small p2;0;HL e¤ects. In the third example, nonzero N_ is driven by large p2;0;HL e¤ects.
104
Figure 6.1: Standing-wave refrigerator with nonzero steady ‡ow through the stacks, studied by
Reid [37, 38, 39]. Steady ‡ow was injected at the upper pressure node at ambient temperature and
‡owed out at the lower pressure node at a cold temperature.
working gas was injected into the resonator at the upper pressure node. This ‡ow was cooled
as it passed through the stacks, so cold gas ‡owed out of the resonator at the lower pressure
node. There was no concern that acoustic power might ‡ow out into the steady-‡ow piping
because the steady-‡ow connections were made at pressure nodes.
To model this apparatus in DeltaEC, we will consider only one half of the apparatus,
because the other half is identical (though shifted in time phase by 180 ): Focusing only on
the thermal aspects, we begin the model just below the speakers in the right half, and end
it where the right half joins the left half, at the exit port at the bottom. The BEGIN and
the guess–target summary re‡ect the following perspective: The refrigerator is driven at a
known frequency and with a known pressure amplitude near the drivers; a known steady
‡ow of gas at a known temperature goes down through the apparatus; the ambient heat
exchanger solid is held at a known temperature by cooling water. We want to learn the
temperature of the exiting steady ‡ow of cold gas, among other things.
GUESS 0f 0g 3a 5e
name BEGIN:|U| BEGIN:Ph(U) RPN:G or T HX:HeatI
units m^3/s deg W
value 5.5034E-02 8.7817 6.3995E-02 -135.8
TARGET 5f 15a 15b 15c
name HX:Est-T SOFTE:Re(z) SOFTE:Im(z) SOFTE:Htot
units K W
value 300.00 0.0000 0.0000 -66.713
result 300.00 -5.3646E-17 3.9259E-16 -66.713
TITLE segments 1-7 < Bob 9/97; segs 8-14 < Hiller ntbk, pg one, 10/97
!->reid4.out
!--------------------------------- 0 ---------------------------------
BEGIN Initialize in main duct where drivers are attached
3.2388E+05 a Mean P Pa
92.000 b Freq Hz
303.00 c T-beg K
105
6500.0 d |p| Pa
90.000 e Ph(p) deg
5.5034E-02 f |U| m^3/s G
8.7817 g Ph(U) deg G
0.6700 i Ndot mol/s
0.9200 j nL
HeAr Gas type
!--------------------------------- 1 ---------------------------------
DUCT Pre-stack Duct
1.8430E-2 a Area m^2 6234.3 A |p| Pa
0.4813 b Perim m 89.692 B Ph(p) deg
0.1494 c Length m 7.3716E-2 C |U| m^3/s
6.4552 D Ph(U) deg
69.087 E Htot W
ideal Solid type 27.060 F Edot W
!--------------------------------- 2 ---------------------------------
RPN display energies
0.0000 a G or T 27.307 A W H2k
69.087 B W Htot
Htot H2k
!--------------------------------- 3 ---------------------------------
RPN Fix Tm. Inp = frac of heat that should go to mean flow
6.3995E-2 a G or T G 302.38 A K
35.997 B W h2k
H2k inp 5e * - =H2k ; inp 5e * Ndot / m / cp / Tm + =Tm
!--------------------------------- 4 ---------------------------------
RPN verify energy conservation
0.0000 a G or T 35.997 A W H2k
69.087 B W Htot
Htot H2k
!--------------------------------- 5 ---------------------------------
HX Ambient Heat exchanger
1.7211E-2 a Area m^2 6169.2 A |p| Pa
0.5227 b GasA/A 89.798 B Ph(p) deg
1.2700E-2 c Length m 7.4628E-2 C |U| m^3/s
6.3500E-4 d y0 m 6.2352 D Ph(U) deg
-135.8 e HeatIn W G -66.713 E Htot W
300.00 f SolidT K =5H 25.809 F Edot W
302.38 G GasT K
300.00 H SolidT K
copper Solid type -99.802 I H2k W
!--------------------------------- 6 ---------------------------------
STKRECT rectangular-pore stack
1.8824E-2 a Area m^2 5547.8 A |p| Pa
0.7050 b GasA/A 93.314 B Ph(p) deg
0.1524 c Length m 9.1544E-2 C |U| m^3/s
4.0640E-4 d aa m 3.8385 D Ph(U) deg
1.1811E-4 e Lplate m -66.713 E Htot W
6.3500E-3 f bb m 2.3232 F Edot W
302.38 G TBeg K
295.21 H TEnd K
kapton Solid type 8.7823E-12 I H2k W
+ 7 DUCT From cold hx to bolt flange, + a half inch
+ 8 DUCT Duct, beginning 1/2 in below big bolt flange
+ 9 DUCT Elbow, 6 inch diam, 5 inch radius of curvature on centerline
+ 10 DUCT straight part of big black cone
+ 11 CONE The long black plastic cone
+ 12 DUCT little straight section of black cone
+ 13 DUCT half the white "tee", up to the symmetry midpoint
!--------------------------------- 14 ---------------------------------
RPN Expect Htot = Edot + Ndot*m*enth
0.0000 a Target (t) -66.713 A RPNval
Edot Ndot m * enth * +
!--------------------------------- 15 ---------------------------------
SOFTEND pressure node
0.0000 a Re(z) = 12G? 4.5144E-12 A |p| Pa
0.0000 b Im(z) = 12H? -74.05 B Ph(p) deg
sameas 14A c Htot W = 12E? 0.1597 C |U| m^3/s
3.5289 D Ph(U) deg
-66.713 E Htot W
-4.8796E-14 F Edot W
-5.3646E-17 G Re(z)
3.9259E-16 H Im(z)
295.21 I T K
-23.967 J p20hl Pa
!--------------------------------- 16 ---------------------------------
RPN U mean here (liter/sec)
999.00 a Target (t) 5.0776 A RPNval
Ndot m * rho / 1000 *
Output 3H tells us that the refrigerator cools the gas to 295 Kelvin. The <.sp> plots
shown in Figs. 6.2 and 6.3 show details of the temperature pro…le in the stack and the
heat exchanger upstream of it, and how the energy ‡ow is independent of x in the stack
106
Figure 6.2: Temperatures vs x in Reid’s refrigerator. Red line, gas temperature Tm (x): Black line,
solid temperature Tsolid (x):
107
even as energy shifts from N_ mcp Tm to H_ 2;k : The temperature pro…le is so severely curved—
essentially horizontal near the ambient end— that an increase in ‡ow above the present value
of 0.67 mol/s (segment 0), which is about 5 liters per second (segment 13), would probably be
undesirable. Note that the solid temperature equals the gas temperature everywhere except
in the heat exchanger, where the solid is cooler than the gas so it can remove ambient heat
from the gas.
When N_ 6= 0; the energy boundary conditions at the BEGINning and END of a model often
require careful thought. The most commonly desired boundary condition is
On its face, this condition sets total power equal to the sum of acoustic power and steady-‡ow
enthalpy. Comparison of this expression with Eqs. (6.1) and (6.2) here and with Eq. (5.29) in
Ref. [13] shows what this condition deliberately omits: axial thermal conduction in the gas
and solid, and boundary-layer, bucket-brigade thermoacoustic entropy transport in the gas.
Thus, this boundary condition is appropriate whenever a long, large-diameter, unremarkable
DUCT or other segment starts or ends a model with N_ 6= 0. This condition is encountered so
frequently that it is taken to be the default in BEGIN, which can be overridden by accessing
the Default Htot / Other Htot choice in the Optional Parameters in BEGIN. The Default
Htot choice sets the initial value of H_ tot according to Eq. (6.13). This condition must also
sometimes be enforced in subsequent SOFTENDs, where an RPN that expresses Eq. (6.13) can
be used, as illustrated above in segments 14 and 15.
Another new issue is addressed in segment 3 now that N_ 6= 0. This variable, FQ_ N_ ; tells
DeltaEC what fraction of the heat exchanger’s heat Q_ should be allocated to changing
energy in the steady ‡ow through the heat exchanger; the remainder is allocated to H_ 2;k :
When N_ 6= 0, we often use FQ_ N_ as a guess, to let DeltaEC meet temperature and energy-
‡ow targets.
Output J in segment 12 gives p2;0;HL = 24 Pa. Since p2;0;HL is initialized to zero in
BEGIN, this shows that only 24 Pa of steady pressure is needed to push the 5 liters per
second through the stack and heat exchanger. To learn how much of that falls across the
heat exchanger and how much falls across the stack, we could insert an RPN segment between
the heat exchanger and the stack, with p20HL in line b, or display p2;0;HL in a <.sp> plot.
6.3. TASHE
The thermoacoustic-Stirling hybrid engine (TASHE), shown in Fig. 6.4 and described in de-
tail in Refs. [5, 40], will be used to illustrate the use of DeltaEC to design a toroidal ther-
moacoustic apparatus without Gedeon streaming [41]. Without intelligent design, Gedeon
streaming around the loop of the TASHE can convect an enormous amount of heat away
from the hot heat exchanger.
We created this DeltaEC model starting from Backhaus’s 1998 DeltaE model (Ap-
pendix B4 of Ref. [13]; also www.lanl.gov/thermoacoustics/Tashe.out). We removed some
inessential segments to reduce the total number, in an e¤ort to save room in the Users
Guide. We modi…ed some segment types to bring the model up to date in DeltaEC version
108
Figure 6.4: Illustrations of the TASHE. (a) Scale drawing of the toroidal part of the TASHE,
containing the heat exchangers. (b) Overall scale drawing, with the toroidal part at the left end.
(c) DeltaEC’s Schematic view of the TASHE, with some phasor diagrams inserted to illustrate
the wave.
109
6. For selecting guesses and targets, we have adopted the following perspective: At what
frequency, amplitude, and hot temperature will this engine run, for a given Q_ H applied
at the hot heat exchanger? Ambient temperature and gas mean pressure are given, and
hardware dimensions are …xed. In anticipation of adding more targets and guesses later, we
make a special e¤ort to minimize their numbers here. Thus, we do not use gas temperature
0c as a guess and ambient heat-exchanger solid temperature 12f as a target; we simply set
0c manually to a value that brings 12H near 25 C. We also avoid guessing heat-exchanger
heat 20e and targeting HARDEND H_ tot = 0 by using an RPN calculation at segment 19 to
anticipate what the heat-exchanger heat must be based on global …rst-law considerations.
With those simpli…cations, we still have 31 segments, 5 targets, and 5 guesses. This …le is
in the <ExamplesnSteadyFlow> folder, as <tashe1.out>.
!Created@20:43:11 01-Dec-2017 with DeltaEC version 6.4b2.7!under win32
!--------------------------------- 0 ---------------------------------
BEGIN the setup
3.1030E+6 a Mean P Pa
85.815 b Freq Hz G
325.00 c TBeg K
3.1675E+5 d |p| Pa G
0.0000 e Ph(p) deg
0.0000 f |U| m^3/s
0.0000 g Ph(U) deg
helium Gas type
!--------------------------------- 1 ---------------------------------
TBRANCH Split up the flow
-1.8366E+7 a Re(Zb) Pa-s/m^3 G 3.1675E+5 A |p| Pa
1.4082E+7 b Im(Zb) Pa-s/m^3 G 0.0000 B Ph(p) deg
-1263.0 c HtotBr W 1.3686E-2 C |U| m^3/s
-142.52 D Ph(U) deg
-1263.0 E HtotBr W
-1720.1 F EdotBr W
1720.1 G EdotTr W
+ 2 DUCT 180 bend plus brass connecting flange (pg 27 book 4)
+ 3 CONE 4" to 3" Concentric reducer (pg 36 book 4)
+ 4 DUCT 3" FB Duct - Length given in concept.skf
+ 5 CONE 3.5" to 3" Long radius reducing elbow (pg 36 book 4)
+ 6 MINOR minor loss, small end of elbow
+ 7 DUCT FB connector/part of tee (Pg 55 book 4 concept.skf)
!--------------------------------- 8 ---------------------------------
SOFTEND End of feedback branch
0.0000 a Re(z) 2.6323E+5 A |p| Pa
0.0000 b Im(z) 1.1006 B Ph(p) deg
0.0000 c Htot W 0.20533 C |U| m^3/s
-92.847 D Ph(U) deg
-1263.0 E Htot W
-1860.7 F Edot W
-0.11237 G Re(z)
1.6282 H Im(z)
325.00 I T K
+ 9 CONE gas diode plate
!--------------------------------- 10 ---------------------------------
MINOR minor loss here
sameas 9d a Area m^2 3.0830E+5 A |p| Pa
0.8000 b K+ -1.9149 B Ph(p) deg
7.0000E-2 c K- 1.3473E-2 C |U| m^3/s
36.516 D Ph(U) deg
1263.0 E Htot W
1626.8 F Edot W
+ 11 DUCT jetting space
!--------------------------------- 12 ---------------------------------
TX Main room temp water HX (pg 90 book 3)
6.6580E-3 a Area m^2 3.0843E+5 A |p| Pa
0.2275 b GasA/A -2.003 B Ph(p) deg
2.0400E-2 c Length m 1.0922E-2 C |U| m^3/s
1.2700E-3 d radius m 14.689 D Ph(U) deg
-1606.0 e HeatIn W G -342.95 E Htot W
300.00 f SolidT K 1613.3 F Edot W
325.00 G GasT K
stainless Solid type 292.62 H SolidT K
+ 13 DUCT Regen cold end dead space due to ribs (pg 91 book 3)
!--------------------------------- 14 ---------------------------------
STKSCREEN Regenerator (pg 92 book 3) (Ks frac est:pg 20 book 4)
6.2070E-3 a Area m^2 2.6856E+5 A |p| Pa
0.7190 b VolPor 1.5666 B Ph(p) deg
7.3000E-2 c Length m 3.2898E-2 C |U| m^3/s
4.2200E-5 d rh m -38.064 D Ph(U) deg
110
0.3000 e ksFrac -342.95 E Htot W
3402.2 F Edot W
325.00 G TBeg K
stainless Solid type 857.45 H TEnd K
+ 15 DUCT All regen hot end dead space (pg 92 book 3)(area is avg)
!--------------------------------- 16 ---------------------------------
HX HHX (pg 93 book 4) heat xfer area used/not acoustic area
5.6970E-3 a Area m^2 2.6848E+5 A |p| Pa
0.9867 b GasA/A 1.5517 B Ph(p) deg
6.3500E-3 c Length m 3.4276E-2 C |U| m^3/s
7.9400E-4 d y0 m -41.297 D Ph(U) deg
3700.0 e HeatIn W 3357.0 E Htot W
0.0000 f SolidT K 3373.4 F Edot W
857.45 G GasT K
stainless Solid type 950.13 H SolidT K
+ 17 JOIN thermal buffer tube end effects
+ 18 DUCT hhx dead space (pg 94 book 3) stainless used for Qdot
+ 19 STKDUCT Straight section of pulse tube (pg 101 bk 4)
+ 20 JOIN thermal buffer tube end effects
!--------------------------------- 21 ---------------------------------
RPN how much heat must be rejected here? Resonator is insulated
0.0000 a G or T -2094.0 A W
16e 12e + ~
!--------------------------------- 22 ---------------------------------
TX Small water Xger
6.6580E-3 a Area m^2 2.6475E+5 A |p| Pa
0.2690 b GasA/A 1.2086 B Ph(p) deg
1.0160E-2 c Length m 7.4090E-2 C |U| m^3/s
2.2860E-3 d radius m -69.45 D Ph(U) deg
sameas 21A e HeatIn W 1263.0 E Htot W
290.00 f SolidT K 3248.2 F Edot W
325.00 G GasT K
ideal Solid type 195.94 H SolidT K
+ 23 DUCT PT connector (see pg 55 book 4 and concept.skf)
+ 24 UNION Rejoin
+ 25 DUCT Section of 3.5" tee
+ 26 CONE 3.5" to 4" cone to adapt to resonator (pg 93 book 4)
+ 27 DUCT Initial section of resonator
+ 28 BRANCH RC dissipator
+ 29 DUCT Continuation of resonator
+ 30 CONE 7 degree cone - 10.02" final diameter
+ 31 DUCT 10" duct P8
+ 32 SURFACE end cap of resonator
!--------------------------------- 33 ---------------------------------
HARDEND end of duct
0.0000 a R(1/z) =33G 9.4878E+4 A |p| Pa
0.0000 b I(1/z) =33H -178.63 B Ph(p) deg
0.0000 c Htot W 4.3937E-12 C |U| m^3/s
-88.08 D Ph(U) deg
0.0000 E Htot W
-1.9968E-9 F Edot W
-4.2520E-14 G R(1/z)
4.4380E-12 H I(1/z)
GUESS 0b 0d 1a 1b 12e
name BEGIN:Freq BEGIN: |p| TBRAN:Re(Zb TBRAN:Im(Zb TX:HeatI
units Hz Pa Pa-s/m^3 Pa-s/m^3 W
value 85.815 3.1675E+05 -1.8366E+07 1.4802E+07 -1606.0
TARGET 24b 24c 24d 33a 33b
name UNION:|p|SO UNION:Ph(p) UNION:T_SOF HARDE:R(1/z HARDE:I(1/z
units Pa deg K
value 2.6323E+05 1.1006 325.00 0.0000 0.0000
result 2.6323E+05 1.1006 325.00 -1.1441E-16 1.0380E-15
With 3700 W applied to the hot heat exchanger, the apparatus resonates at 85.8 Hz
with an amplitude of 317 kPa at BEGIN. The hot gas temperature is 857 K and the hot solid
temperature is 950 K.
One strange-looking number appears above: Segment 1, TBRANCH, input c, 1263 W,
HtotBr. DeltaEC arrives at the TBRANCH with some value for H_ tot : (It is zero in the present
example, but not always.) H_ tot,br tells DeltaEC how much H_ tot to send into the branch;
the balance goes into the trunk. Neither we nor DeltaEC may know the right value for this
number in the present model, because the contribution of Rayleigh- and jet-driven streaming
to heat convection in the part of the torus that connects the main ambient heat exchanger
(segment 12) and the secondary ambient heat exchanger (segment 22) is unknown. Such
streaming helps determine how much of the heat that comes from dissipation of E_ in the
load and resonator (assumed insulated here) shows up in each of those heat exchangers.
111
Fortunately, the performance of the apparatus is insensitive to H_ tot,br : If we change its value
from 1263 W to 1763; the resonance frequency, pressure amplitude, and hot temperature
change by less than 0.1%, even though the heat rejected at the main ambient heat exchanger
increases by 500 W and that in the secondary ambient heat exchanger decreases by the same
amount. In the absence of any evidence to the contrary, it is sometimes reasonable to set
and this is available as a check-box option in the TBRANCH segment. Another reasonable
option here would be to add H_ tot,br to the guess list, using it to allow DeltaEC to maintain
equal solid temperatures in the two ambient-temperature heat exchangers. Since we are
trying to minimize the number of guesses and targets right now, we choose the …rst option.
Now we let DeltaEC begin to consider Gedeon streaming, by checking the “Enable
Ndot” box in BEGIN. This enables nonzero N_ and the calculation and display of p2;0;HL
throughout the model. For now, we leave N_ = 0 and examine the UNION at segment 24:
!--------------------------------- 24 -----------------------------
UNION Rejoin
8.0000 a SegNum 2.6331E+05 A |p| Pa
2.6331E+05 b |p|Sft Pa = 22A? 1.0991 B Ph(p) deg
1.0991 c Ph(p)S deg = 22B? 0.2837 C |U| m^3/s
325.00 d TSoft K = 22G? -86.775 D Ph(U) deg
-0.6445 e p20HLS Pa (t) 2.2737E-13 E Htot W
1385.9 F Edot W
325.00 G T K
-4038.2 H p20HL Pa
“Allowing steady ‡ow”has turned on the calculation of p2;0;HL . Parameter e here shows
that p2;0;HL = 0:6 Pa at SOFTEND segment 8, which is the end of the TBRANCH and is
supposedly reconnected to the trunk here at segment 22. However, parameter H shows that
p2;0;HL = 4038 Pa here in the trunk.
There is a mismatch of 4 kPa in second-order time-averaged pressure in this UNION
segment. We have already targeted p1 and Tm in the UNION, to satisfy acoustic and ther-
modynamic boundary conditions; now we must add p2;0;HL in the UNION to the target list.
Our corresponding guess will be N_ br in segment 1: We ask DeltaEC how much Gedeon
streaming will ‡ow around the torus, and what e¤ects that will have on operation of the
apparatus.
We make a small increase in N_ br and watch segment 22, to make sure p2;0;HL changes in
the right direction, to make sure we understand what we’re doing. We note that DeltaEC
gets lost if we change N_ br by 0.1 mol/s, so we make smaller changes of 0.01 mol/s until
p2;0;HL in segments 8 and 22 are nearly matched. Then we add 1d and 22e to the guess and
target lists, respectively.
Making those changes and running, we notice that the state plot of Tm (x) is not smooth
in the thermal bu¤er tube. We increase the number of steps in the numerical integration
of STKDUCT (and the other integrated segments), by increasing Nint. The irregularities in
Tm (x) go away above Nint = 30. Running with Nint = 40, the guess–target summary
becomes
GUESS 0b 0d 1a 1b 1d 12e
name BEGIN:Freq BEGIN: |p| TBRAN:Re(Zb TBRAN:Im(Zb TBRAN:Ndotb TX:HeatI
units Hz Pa Pa-s/m^3 Pa-s/m^3 mol/s W
value 86.157 1.9974E+05 -2.0031E+07 1.3845E+07 0.28991 -3950.0
TARGET 22b 22c 22d 22e 31a 31b
112
Figure 6.5: State plot of the TASHE example, with steady-‡ow calculations enabled but Gedeon
streaming not properly eliminated. Solid black curve is Tsolid ; dashed red curve is Tm ; and dot-
dashed green curve is p2;0;HL : The locations of the three heat exchangers are evident as short
plateaus in Tsolid :
and the interesting characteristics of the engine become f = 86 Hz, jp1 j = 200 kPa, Tm;H =
743 K, and Tsolid,H = 834 K. This estimate of performance is probably much more realistic
than the values of 317 kPa, 857 K, and 950 K we found earlier, because now we have included
the Gedeon streaming, which is carrying away a lot of heat from the hot heat exchanger
without contributing to any thermoacoustic power generation.
To decide what to do about the Gedeon streaming, we can either look at the sign of N_ br
or examine temperatures and p2;0;HL in the state plot shown in Fig. 6.5. The regenerator is
running far on the hot side of a linear temperature pro…le, and N_ is streaming through it
from hot to cold. Apparently the gas diode is too tight, overcompensating for the tendency
_ No other segments contribute signi…cantly
of the regenerator to stream in the direction of E:
to p2;0;HL : Examination of the gas-diode segments,
!--------------------------------- 9 ------------------------------
CONE gas diode plate
5.8880E-04 a AreaI m^2 2.0040E+05 A |p| Pa
0.4210 b PerimI m -0.5402 B Ph(p) deg
2.5399E-02 c Length m 8.0839E-03 C |U| m^3/s
1.1400E-04 d AreaF m^2 33.617 D Ph(U) deg
0.1850 e PerimF m 1263.0 E Htot W
5.0000E-04 f Srough 670.28 F Edot W
ideal Solid type
!--------------------------------- 10 ------------------------------
MINOR minor loss here
sameas 9d a Area m^2 1.9709E+05 A |p| Pa
0.8000 b K+ -1.1947 B Ph(p) deg
7.0000E-02 c K- 8.0839E-03 C |U| m^3/s
33.617 D Ph(U) deg
1263.0 E Htot W
113
Figure 6.6: (a) The TASHE of the previous section, with a traditional hot heat exchanger between
the regenerator and the thermal bu¤er tube. (b) The same TASHE, with the hot heat exchanger
replaced by a mixing chamber, self-circulating loop, and ‡ow straightener.
654.07 F Edot W
shows that they are qualitatively in order, with the minor-loss coe¢ cient larger in the +x
direction than in the x direction, as necessary to …ght Gedeon streaming. We surmise that
the area at the small end of the gas diode, 1:14 cm2 ; may be too small. We tentatively increase
that area about 1% (leaving its perimeter constant, since this was an adjustable slit in the
experimental hardware), and we are pleased to see that DeltaEC converges on a solution
with reduced streaming and higher amplitude. We use the incremental plot to continue
increasing that area, watching N_ br and stopping the sequence when it gets close to zero. We
then clear N_ br from the guess list and set it equal to zero, and add the gas-diode’s small-end
area to the guess list. A run then yields <tashe5.out> in the <ExamplesnSteadyFlow>
folder, which shows that an area of 2.47 cm2 stops Gedeon streaming, and with this area the
apparatus runs with f = 86 Hz, jp1 j = 300 kPa, Tm;H = 731 K, and Tsolid,H = 822 K— better
than before. A quick glance at the state plot shows near-linear temperature pro…les in the
stack and thermal bu¤er tube, even with Nint back down at its default value of 10.
114
scribed at the end of the previous section, down to only a few segments in the vicinity of
the hot heat exchanger:
!--------------------------------- 0 ---------------------------------
BEGIN the setup
3.1030E+06 a Mean P Pa
85.895 b Freq Hz
325.00 c TBeg K
2.9837E+05 d |p| Pa
-0.9353 e Ph(p) deg
1.3300E-02 f |U| m^3/s
13.797 g Ph(U) deg
-316.07 h Htot W
0.0000 i Ndot mol/s
helium Gas type
+ 1 STKSCREEN Regenerator (pg 92 book 3) (Ks frac est:pg 20 book 4)
+ 2 DUCT All regen hot end dead space (pg 92 book 3)(area is avg)
+ 3 HX HHX (pg 93 book 4) heat xfer area used/not acoustic area
+ 4 DUCT hhx dead space (pg 94 book 3) stainless used for Qdot
+ 5 JOIN thermal buffer tube end effects
+ 6 STKDUCT Straight section of pulse tube (pg 101 bk 4)
We set the initial conditions in BEGIN to match the conditions of the segment just up-
stream of the STKSCREEN in <tashe5.out>, so the integration proceeds through the rest of
the model as it did in <tashe5.out>, with no need for guesses and targets.
Next, we install a hot self-circulating loop instead of the hot heat exchanger. In place
of the HX, we insert, as a side-branch loop, in sequence: a quarter wavelength of duct, a gas
diode, an adapter cone, a VXQ1 heat exchanger, another adapter, and a quarter-wavelength
return duct. We also insert a ‡ow straightener above the thermal bu¤er tube. We keep
guesses and targets at a minimum: those that are necessary at the UNION and TBRANCH, and
one additional guess–target pair, guessing the input heat and targeting the thermal bu¤er
tube’s …nal temperature.
!->C:\Users\092710\Documents\DeltaEC\PySource\Examples\SteadyFlow - withJoin\hotloop3.out
!Created@08:04:05 02-Dec-2017 with DeltaEC version 6.4b2.7!under win32,
using Win 6.1.7601 (Service Pack 1) under Python 2.7.9.
!--------------------------------- 0 ---------------------------------
BEGIN the setup
3.1030E+6 a Mean P Pa
85.895 b Freq Hz
325.00 c TBeg K
2.9837E+5 d |p| Pa
-0.9353 e Ph(p) deg
1.3300E-2 f |U| m^3/s
13.797 g Ph(U) deg
-316.07 h Htot W
0.0000 i Ndot mol/s
helium Gas type
!--------------------------------- 1 ---------------------------------
STKSCREEN Regenerator (pg 92 book 3) (Ks frac est:pg 20 book 4)
6.2070E-3 a Area m^2 2.5369E+5 A |p| Pa
0.7190 b VolPor 1.9218 B Ph(p) deg
7.3000E-2 c Length m 3.2324E-2 C |U| m^3/s
4.2200E-5 d rh m -30.836 D Ph(U) deg
0.3000 e ksFrac -316.07 E Htot W
3448.0 F Edot W
325.00 G TBeg K
731.79 H TEnd K
stainless Solid type -316.07 I H2k W
!--------------------------------- 2 ---------------------------------
DUCT top half of mixing chamber
6.2000E-3 a Area m^2 Mstr 2.5365E+5 A |p| Pa
0.2791 b Perim m 2a 1.9108 B Ph(p) deg
1.0000E-2 c Length m 3.3238E-2 C |U| m^3/s
-33.235 D Ph(U) deg
-316.07 E Htot W
ideal Solid type 3447.0 F Edot W
!--------------------------------- 3 ---------------------------------
TBRANCH branch into hot loop
5.8567E+7 a Re(Zb) Pa-s/m^3 G 2.5365E+5 A |p| Pa
3.9622E+7 b Im(Zb) Pa-s/m^3 G 1.9108 B Ph(p) deg
0.7468 d NdotBr mol/s G 3.5872E-3 C |U| m^3/s
-32.169 D Ph(U) deg
115
7079.6 E HtotBr W
376.82 F EdotBr W
3070.2 G EdotTr W
!--------------------------------- 4 ---------------------------------
DUCT quarter-wave duct from engine to diode
7.0000E-4 a Area m^2 Mstr 1.0477E+4 A |p| Pa
9.3748E-2 b Perim m 4a -163.77 B Ph(p) deg
4.6000 c Length m 5.5089E-2 C |U| m^3/s
5.0000E-4 d Srough -87.716 D Ph(U) deg
7079.6 E Htot W
ideal Solid type 69.557 F Edot W
!--------------------------------- 5 ---------------------------------
RPN calculate Borda-Carnot minor loss here
0.0000 a G or T 0.18367 A ChngeMe
1 6a 4a / - sqrd
!--------------------------------- 6 ---------------------------------
MINOR asymmetric flow at gas diode
4.0000E-4 a Area m^2 1.0190E+4 A |p| Pa
5.0000E-2 b K+ -173.93 B Ph(p) deg
sameas 5A c K- 5.5089E-2 C |U| m^3/s
-87.716 D Ph(U) deg
7079.6 E Htot W
18.515 F Edot W
!--------------------------------- 7 ---------------------------------
CONE expansion to plenum
sameas 6a a AreaI m^2 Mstr 2.0026E+4 A |p| Pa
7.0899E-2 b PerimI m 7a -176.65 B Ph(p) deg
0.4000 c Length m 5.0311E-2 C |U| m^3/s
sameas 9a d AreaF m^2 Mstr -87.796 D Ph(U) deg
0.4342 e PerimF m 7d 7079.6 E Htot W
5.0000E-4 f Srough 10.101 F Edot W
ideal Solid type
!--------------------------------- 8 ---------------------------------
RPN watch that angle
5.0000 a G or T 16.125 A ChngeMe
7e 7b - pi / 7c / atan
!--------------------------------- 9 ---------------------------------
VXQ1 plenty of room for heat transfer
1.5000E-2 a Area m^2 1.7465E+4 A |p| Pa
0.4000 b GasA/A -176.04 B Ph(p) deg
0.1000 c SolA/A 5.7564E-2 C |U| m^3/s
1.0000E-2 d rh m 92.120 D Ph(U) deg
1.0000E-2 e LenTS1 m 9990.3 E Htot W
6.0000 f Len P1 m -16.133 F Edot W
1.0000E-2 g LenTS2 m 734.64 G TSolBeg P1 K
2910.7 h HeatP1 W G 922.27 H TSolEnd P1 K
1.0000 i FracQN 376.82 I H2k W
stainless Solid type
!--------------------------------- 10 ---------------------------------
CONE adapt between heat exchanger and quarter-wave tube
sameas 7d a AreaI m^2 1.0775E+4 A |p| Pa
sameas 7e b PerimI m -174.31 B Ph(p) deg
sameas 7c c Length m 6.1978E-2 C |U| m^3/s
sameas 11a d AreaF m^2 92.238 D Ph(U) deg
sameas 11b e PerimF m 9990.3 E Htot W
5.0000E-4 f Srough -20.081 F Edot W
ideal Solid type
!--------------------------------- 11 ---------------------------------
DUCT quarter wave connection tube
sameas 4a a Area m^2 2.5365E+5 A |p| Pa
sameas 4b b Perim m 1.9108 B Ph(p) deg
5.0000 c Length m 6.1093E-3 C |U| m^3/s
sameas 4d d Srough 124.84 D Ph(U) deg
9990.3 E Htot W
ideal Solid type -421.19 F Edot W
!--------------------------------- 12 ---------------------------------
SOFTEND end of the branch; this will reconnect to trunk at UNION
0.0000 a Re(z) 2.5365E+5 A |p| Pa
0.0000 b Im(z) 1.9108 B Ph(p) deg
0.0000 c Htot W 6.1093E-3 C |U| m^3/s
124.84 D Ph(U) deg
9990.3 E Htot W
-421.19 F Edot W
-5.4502 G Re(z)
-8.4153 H Im(z)
919.30 I T K
1623.7 J p20HL Pa
!--------------------------------- 13 ---------------------------------
UNION reconnect the "final" end of the loop here
12 a SegNum 2.5365E+5 A |p| Pa
sameas 12A b |p|Sft Pa =13A 1.9108 B Ph(p) deg
sameas 12B c Ph(p)S deg =13B 2.4086E-2 C |U| m^3/s
sameas 12I d TSoft K -27.96 D Ph(U) deg
sameas 12J e p20HLS Pa =13H 2594.7 E Htot W
2649.0 F Edot W
116
731.79 G T K
1623.7 H p20HL Pa
!--------------------------------- 14 ---------------------------------
DUCT bottom half of mixing chamber
sameas 2a a Area m^2 2.5363E+5 A |p| Pa
sameas 2b b Perim m 1.9024 B Ph(p) deg
sameas 2c c Length m 2.4942E-2 C |U| m^3/s
-31.255 D Ph(U) deg
2594.7 E Htot W
ideal Solid type 2648.0 F Edot W
!--------------------------------- 15 ---------------------------------
SX flow straightener (no heat input)
sameas 14a a Area m^2 2.5345E+5 A |p| Pa
0.6500 b VolPor 1.9185 B Ph(p) deg
5.0000E-3 c Length m 2.5100E-2 C |U| m^3/s
5.0000E-4 d rh m -32.614 D Ph(U) deg
0.0000 e HeatIn W 2594.7 E Htot W
0.0000 f SolidT K 2620.4 F Edot W
731.79 G GasT K
731.79 H SolidT K
ideal Solid type 2594.7 I H2k W
!--------------------------------- 16 ---------------------------------
JOIN thermal buffer tube end effects
2.5345E+5 A |p| Pa
1.9185 B Ph(p) deg
2.4527E-2 C |U| m^3/s
-32.614 D Ph(U) deg
2594.7 E Htot W
2560.5 F Edot W
731.79 G TBeg K
731.28 H TEnd K
!--------------------------------- 17 ---------------------------------
STKDUCT Straight section of pulse tube (pg 101 bk 4)
7.0000E-3 a Area m^2 2.5115E+5 A |p| Pa
0.2963 b Perim m 1.6538 B Ph(p) deg
0.2400 c Length m 6.1562E-2 C |U| m^3/s
1.1600E-3 d WallA m^2 -69.113 D Ph(U) deg
2594.7 E Htot W
2546.7 F Edot W
731.28 G TBeg K
300.00 H TEnd K
stainless Solid type 2594.7 I H2k W
!--------------------------------- 18 ---------------------------------
RPN temperature target
300.00 a =A? 300.00 A K
Tm
GUESS 3a 3b 3d 9h
name TBRAN:Re(Zb TBRAN:Im(Zb TBRAN:Ndotb VXQ1:HeatP
units Pa-s/m^3 Pa-s/m^3 mol/s W
value 5.8567E+07 3.9622E+07 0.7468 2910.7
TARGET 13b 13c 13e 18a
name UNION:|p|SO UNION:Ph(p) UNION:p20HL RPN:Targe
units Pa deg Pa
value 2.5365E+05 1.9108 1623.7 300.00
result 2.5365E+05 1.9108 1623.7 300.00
Several features that have not been previously encountered or discussed in the Users
Guide are displayed in this …le. Details about each of these features are given in Chapter
10.
Segment MINOR is employed here with unequal minor-loss coe¢ cients, as appropriate for
an asymmetrical geometry and as necessary to create a pumping e¤ect to move N_ ' 0:8
mole/sec around the loop [42]. Figure 6.7 shows the state plot of p2;0;HL around the loop.
The rise near x = 4 m is the abrupt e¤ect of the gas diode’s MINOR segment. The gradual
decreases elsewhere are the pressure gradients in the DUCTs (and, much smaller, in the VXQ1),
which the gas-diode pump is overcoming. The gradient dp2;0;HL =dx varies along each DUCT,
despite the constancy of the area and ‡ow rate, because of the nonlinear interaction between
U1 and N_ ; which increases dp2;0;HL =dx where jU1 j is largest. The nature of that nonlinear
interaction is controlled in part by the value of the roughness parameter in DUCT segments
4 and 11. As described in detail in Section 10.1.1, the interaction between turbulent steady
‡ow and turbulent oscillating ‡ow is also governed by which of the Optimistic, Average,
or Pessimistic boxes are checked in a DUCT. The p20HL target in the UNION ensures that
117
Figure 6.7: Second-order head-loss pressure vs x in the heat-exchange loop. The abrupt rise marks
the location of the gas diode.
p2;0;HL (x = 16:5 m) = p2;0;HL (x = 0): Recall that the zero of p2;0;HL is set arbitrarily in the
BEGIN segment, so the vertical o¤set of this graph is meaningless for the present discussion;
it only indicates p2;0;HL across the regenerator when N_ = 0 in the regenerator.
Segment 9, VXQ1, is one of the family of “variable” heat exchanger, VX, segments. The
other types of heat exchanger segments— HX, TX, SX, and PX— are assumed to be short, to
have x-independent solid temperatures, and to have constant dTm =dx: The VX** segments
are numerically integrated with Nint steps, so they can be long. VXT1 and VXT2 are intended
to model long heat exchangers with shell-side heat transfer such that the solid surfaces
are held essentially isothermal. The gas temperature Tm (x) and heat transfer per unit
length q(x)
_ are obtained numerically in the VXT** segments. VXQ1 and VXQ2 are intended to
model long heat exchangers with shell-side heat transfer such that the heat input per unit
length is independent of x: The gas temperature Tm (x) and and solid temperature Tsolid (x)
are obtained numerically in the VXQ** segments. VXQ2 and VXT2 allow for two-part heat
exchangers, such as heat exchangers with two passes on the shell side. Additional lengths
for tubesheets, where no heat transfer occurs, are built into both ends of the VX** segments.
For more details, see Section 10.7.3. In the present example, Fig. 6.8 shows temperatures
around the loop. Outside of the heat exchanger, gas and solid temperatures are equal, as
usual. Inside the heat exchanger, Tm increases linearly with x because this is an ideal gas
‡owing though a VXQ1 segment, with cp and q_ independent of x: The di¤erence Tsolid Tm
peaks in the center of the VX, where jU1 j is the smallest and hence turbulent heat transfer
is the least e¤ective at transferring heat from the solid to the gas.
The logistical segments in <hotloop3.out> also have a few peculiar features. In the
BEGIN, H_ tot is set to the value that it had just upstream of the regenerator in <tashe5.out>,
and N_ is enabled and set to zero, to turn on p2;0;HL and steady-‡ow calculations in the rest
of the model. N_ is, of course, zero or negligible in the regenerator, since <tashe5.out>
properly suppressed Gedeon streaming with its gas diode. However, N_ br in the TBRANCH is
large. DeltaEC adjusts this guess to hit the p2;0;HL target at the UNION. We don’t know
exactly what H_ tot,br should be, because we don’t know exactly how much heat might be
convected into the loop by N_ here, but the measurements reported in Ref. [8] indicate that
any such convected heat is much smaller than what is convected out of such a loop at the
118
Figure 6.8: Temperatures around the hot loop. Solid black line, solid temperature. Dashed red
line, gas temperature. The velocity-dependent gas–solid heat-transfer coe¢ cient in VXQ1 is evident
in the di¤erence between the two curves. (Nint = 20 for this plot.)
other end, so we simply set H_ tot,br = E_ br in the TBRANCH. At the UNION, we do not use the
temperature target, because this would force the gas temperature at both ends of the loop
to be equal, which would contradict the entire purpose of the circulating loop. Instead, we
essentially let the 919-K gas coming out of the SOFTEND of the loop pour into the UNION
where Tm = 732 K. We rely on the energy-conserving feature of DeltaEC’s UNION to add
the heat picked up in the loop into the H_ tot leaving the UNION.
Next, as shown in Fig. 6.9, we combine <hotloop3.out> and <tashe5.out>, to make
a complete DeltaEC model of the TASHE, the hot self-circulating heat-exchange loop, and
the resonator. We run <tashe5.out>, Kill its traditional hot heat exchanger, and then
copy and paste the appropriate segments from <hotloop3.out>. We …x a few things that
DeltaEC did not do correctly. A run fails to converge, so we Restore guessez and then clear
all guesses and targets, intending to introduce them one pair at a time. With no guesses
and targets, a run shows mostly sensible results, except at the HARDEND at the end of the
model, where energy is ‡owing out but acoustic power is ‡owing in, in the x direction.
We introduce guesses and targets beginning with the four each that were involved in the
hot loop, since they are closest to being on-target already. Then we add three of the four
UNION targets that were used in <tashe5.out> (skipping the temperature target for the
moment), and three corresponding guesses, two Zbr in the segment-1 TBRANCH and the third
being the slot width in the gas diode in segment 9. At this point, we have 7 targets and 7
guesses, and each change we make is requiring small, gentle steps. We remove the temporary
RPN target we had been using in <hotloop3.out> to keep the temperature near ambient
at the end of <hotloop3.out>, adding instead the segment-35 UNION temperature target.
We add frequency as a guess, and Im(1=zn ) at the HARDEND as a target. Pause to think. In
<tashe5.out>, our perspective was to give the engine a known hot heat input, and observe
performance. Here, we are currently guessing the hot heat input, because that is what we
did in <hotloop3.out>. Clear that guess, and try guessing the main ambient heat instead.
With that change, DeltaEC still tolerates changes to variables reasonably well, given the
fact that we have 8 targets and 8 guesses. In <tashe5.out>, we injected 3700 W of hot
119
Figure 6.9: Sketch of the <combo.out> model. The TASHE feedback loop hangs to the left of the
trunk, and the hot self-circulating heat exchanger loop hangs to the right.
120
heat; let’s increase the heat in our hot loop to the same value here, for a fair comparison
to the earlier model. We begin incremental plotting to higher hot heat, in 5-W steps which
are small enough that DeltaEC stays converged at each step. Along the way, we notice
that the E_ ‡owing in at the as-yet-incompletely-targeted HARDEND is steadily decreasing,
approaching zero at about 3500 W. We stop there, and add the HARDEND’s Re[1=zn ] to the
target list and the BEGIN’s jp1 j to the guess list, and resume our increase of Q_ H up to 3700.
The resulting …le, <combo2.out>, appears below. Whereas <tashe5.out> ran at
an amplitude (in segment 0) of 300 kPa, <combo2.out> runs at 311 kPa. However,
<tashe5.out> required only Tsolid,H = 822 K; <combo2.out> requires solid temperatures
from 1197 K to 1286 K.
This example may not be practical, but it does demonstrate the creation of a DeltaEC
model for a topologically complicated apparatus with 9 targets and 9 guesses.
GUESS 0b 0d 1a 1b 9d 12e
name BEGIN:Freq BEGIN:|p| TBRAN:Re(Zb TBRAN:Im(Zb CONE:AreaF TX:HeatI
units Hz Pa Pa-s/m^3 Pa-s/m^3 m^2 W
value 89.111 3.1114E+05 -1.4792E+07 1.4654E+07 2.0008E-04 -2153.4
GUESS 16a 16b 16d
name TBRAN:Re(Zb TBRAN:Im(Zb TBRAN:Ndotb
units Pa-s/m^3 Pa-s/m^3 mol/s
value 1.9797E+06 9.5090E+06 2.0228
TARGET 26b 26c 26e 35b 35c 35d
name UNION:|p|SO UNION:Ph(p) UNION:p20hl UNION:|p|SO UNION:Ph(p) UNION:T_SOF
units Pa deg Pa Pa deg K
value 2.5486E+05 1.6159 -0.72692 2.5372E+05 1.1616 325.00
result 2.5486E+05 1.6159 -0.72692 2.5372E+05 1.1616 325.00
TARGET 35e 44a 44b
name UNION:p20hl HARDE:R(1/z HARDE:I(1/z
units Pa
value -0.48008 0.0000 0.0000
result -0.48008 -1.3244E-14 3.5147E-12
!->C:\Users\092710\Documents\DeltaEC\PySource\Examples\SteadyFlow - withJoin\combo2.out
!Created@08:42:24 02-Dec-2017 with DeltaEC version 6.4b2.7!under win32,
using Win 6.1.7601 (Service Pack 1) under Python 2.7.9.
!--------------------------------- 0 ---------------------------------
BEGIN the setup
3.1030E+6 a Mean P Pa
89.111 b Freq Hz G
325.00 c TBeg K
3.1114E+5 d |p| Pa G
0.0000 e Ph(p) deg
0.0000 f |U| m^3/s
0.0000 g Ph(U) deg
0.0000 i Ndot mol/s
helium Gas type
!--------------------------------- 1 ---------------------------------
TBRANCH Split up the flow
-1.4792E+7 a Re(Zb) Pa-s/m^3 G 3.1114E+5 A |p| Pa
1.4654E+7 b Im(Zb) Pa-s/m^3 G 0.0000 B Ph(p) deg
0.0000 d NdotBr mol/s 1.4944E-2 C |U| m^3/s
-135.27 D Ph(U) deg
-1651.6 E HtotBr W
-1651.6 F EdotBr W
1651.6 G EdotTr W
+ 2 DUCT 180 bend plus brass connecting flange (pg 27 book 4)
+ 3 CONE 4" to 3" Concentric reducer (pg 36 book 4)
+ 4 DUCT 3" FB Duct - Length given in concept.skf
+ 5 CONE 3.5" to 3" Long radius reducing elbow (pg 36 book 4)
+ 6 MINOR minor loss, small end of elbow
+ 7 DUCT FB connector/part of tee (Pg 55 book 4 concept.skf)
!--------------------------------- 8 ---------------------------------
SOFTEND End of feedback branch
0.0000 a Re(z) 2.5372E+5 A |p| Pa
0.0000 b Im(z) 1.1616 B Ph(p) deg
0.0000 c Htot W 0.21036 C |U| m^3/s
-92.693 D Ph(U) deg
-1651.6 E Htot W
-1794.1 F Edot W
-0.10324 G Re(z)
1.5321 H Im(z)
325.00 I T K
-0.48008 J p20HL Pa
!--------------------------------- 9 ---------------------------------
CONE gas diode plate
121
5.8880E-4 a AreaI m^2 3.1270E+5 A |p| Pa
0.4210 b PerimI m -0.42704 B Ph(p) deg
2.5399E-2 c Length m 1.4669E-2 C |U| m^3/s
2.0008E-4 d AreaF m^2 G 43.826 D Ph(U) deg
0.1850 e PerimF m 1651.6 E Htot W
5.0000E-4 f Srough 1642.7 F Edot W
ideal Solid type
!--------------------------------- 10 ---------------------------------
MINOR minor loss here
sameas 9d a Area m^2 3.0945E+5 A |p| Pa
0.8000 b K+ -1.0163 B Ph(p) deg
7.0000E-2 c K- 1.4669E-2 C |U| m^3/s
43.826 D Ph(U) deg
1651.6 E Htot W
1609.3 F Edot W
!--------------------------------- 11 ---------------------------------
DUCT jetting space
6.8250E-3 a Area m^2 3.0951E+5 A |p| Pa
0.2930 b Perim m -1.0302 B Ph(p) deg
1.9000E-2 c Length m 1.1993E-2 C |U| m^3/s
5.0000E-4 d Srough 28.944 D Ph(U) deg
1651.6 E Htot W
ideal Solid type 1607.8 F Edot W
!--------------------------------- 12 ---------------------------------
TX Main room temp water HX (pg 90 book 3)
6.6580E-3 a Area m^2 3.0968E+5 A |p| Pa
0.2275 b GasA/A -1.1073 B Ph(p) deg
2.0400E-2 c Length m 1.1403E-2 C |U| m^3/s
1.2700E-3 d radius m 24.265 D Ph(U) deg
-2153.4 e HeatIn W G -501.81 E Htot W
300.00 f SolidT K 1595.4 F Edot W
325.00 G GasT K
282.39 H SolidT K
stainless Solid type -501.81 I H2k W
!--------------------------------- 13 ---------------------------------
DUCT Regen cold end dead space due to ribs (pg 91 book 3)
4.9700E-3 a Area m^2 3.0968E+5 A |p| Pa
0.7400 b Perim m -1.1104 B Ph(p) deg
3.1750E-3 c Length m 1.1182E-2 C |U| m^3/s
21.805 D Ph(U) deg
-501.81 E Htot W
stainless Solid type 1594.7 F Edot W
!--------------------------------- 14 ---------------------------------
STKSCREEN Regenerator (pg 92 book 3) (Ks frac est:pg 20 book 4)
6.2070E-3 a Area m^2 2.5488E+5 A |p| Pa
0.7190 b VolPor 1.6250 B Ph(p) deg
7.3000E-2 c Length m 4.0190E-2 C |U| m^3/s
4.2200E-5 d rh m -25.452 D Ph(U) deg
0.3000 e ksFrac -501.81 E Htot W
4560.5 F Edot W
325.00 G TBeg K
1192.4 H TEnd K
stainless Solid type -501.81 I H2k W
!--------------------------------- 15 ---------------------------------
DUCT top half of mixing chamber
6.2000E-3 a Area m^2 Mstr 2.5486E+5 A |p| Pa
0.2791 b Perim m 15a 1.6159 B Ph(p) deg
1.0000E-2 c Length m 4.0993E-2 C |U| m^3/s
6.0000E-4 d Srough -27.604 D Ph(U) deg
-501.81 E Htot W
stainless Solid type 4559.0 F Edot W
!--------------------------------- 16 ---------------------------------
TBRANCH branch into hot loop
1.9797E+6 a Re(Zb) Pa-s/m^3 G 2.5486E+5 A |p| Pa
9.5090E+6 b Im(Zb) Pa-s/m^3 G 1.6159 B Ph(p) deg
606.94 c HtotBr W 2.6239E-2 C |U| m^3/s
2.0228 d NdotBr mol/s G -76.623 D Ph(U) deg
606.94 E HtotBr W
681.51 F EdotBr W
3877.4 G EdotTr W
+ 17 DUCT quarter-wave duct from engine to diode
+ 18 RPN calculate Borda-Carnot minor loss here
+ 19 MINOR asymmetric flow at gas diode
+ 20 CONE expansion to plenum
+ 21 RPN watch that angle
!--------------------------------- 22 ---------------------------------
VXQ1 plenty of room for heat transfer
1.5000E-2 a Area m^2 2.6634E+4 A |p| Pa
0.4000 b GasA/A -174.71 B Ph(p) deg
0.1000 c SolA/A 6.9480E-2 C |U| m^3/s
1.0000E-2 d rh m 92.426 D Ph(U) deg
1.0000E-2 e LenTS1 m 4306.9 E Htot W
6.0000 f Len P1 m -46.285 F Edot W
1.0000E-2 g LenTS2 m 1197.0 G TSolBeg P1 K
3700.0 h HeatP1 W 1285.5 H TSolEnd P1 K
122
1.0000 i FracQN -3.6915E+4 I H2k W
stainless Solid type
+ 23 CONE adapt between heat exchanger and quarter-wave tube
+ 24 DUCT quarter wave connection tube
!--------------------------------- 25 ---------------------------------
SOFTEND end of the branch; this will reconnect to trunk at UNION
0.0000 a Re(z) 2.5486E+5 A |p| Pa
0.0000 b Im(z) 1.6159 B Ph(p) deg
0.0000 c Htot W 2.3017E-2 C |U| m^3/s
105.05 D Ph(U) deg
4306.9 E Htot W
-681.45 F Edot W
-0.73312 G Re(z)
-3.0692 H Im(z)
1280.4 I T K
-0.72692 J p20HL Pa
!--------------------------------- 26 ---------------------------------
UNION reconnect the "final" end of the loop here
25 a SegNum 2.5486E+5 A |p| Pa
sameas 25A b |p|Sft Pa =26A 1.6159 B Ph(p) deg
sameas 25B c Ph(p)S deg =26B 3.7639E-2 C |U| m^3/s
sameas 25I d TSoft K 49.830 D Ph(U) deg
sameas 25J e p20HLS Pa =26H 3198.2 E Htot W
3196.0 F Edot W
1192.4 G T K
-0.72692 H p20HL Pa
!--------------------------------- 27 ---------------------------------
DUCT bottom half of mixing chamber
sameas 15a a Area m^2 2.5489E+5 A |p| Pa
sameas 15b b Perim m 1.6094 B Ph(p) deg
sameas 15c c Length m 3.6364E-2 C |U| m^3/s
48.035 D Ph(U) deg
3198.2 E Htot W
ideal Solid type 3194.4 F Edot W
!--------------------------------- 28 ---------------------------------
SX flow straightener (no heat input)
sameas 27a a Area m^2 2.5474E+5 A |p| Pa
0.6500 b VolPor 1.5619 B Ph(p) deg
5.0000E-3 c Length m 3.5627E-2 C |U| m^3/s
5.0000E-4 d rh m 47.337 D Ph(U) deg
0.0000 e HeatIn W 3198.2 E Htot W
0.0000 f SolidT K 3165.1 F Edot W
1192.4 G GasT K
1192.4 H SolidT K
ideal Solid type 3198.2 I H2k W
!--------------------------------- 29 ---------------------------------
JOIN thermal buffer tube end effects
2.5474E+5 A |p| Pa
1.5619 B Ph(p) deg
3.4935E-2 C |U| m^3/s
47.337 D Ph(U) deg
3198.2 E Htot W
3103.5 F Edot W
1192.4 G TBeg K
1200.5 H TEnd K
!--------------------------------- 30 ---------------------------------
STKDUCT Straight section of pulse tube (pg 101 bk 4)
7.0000E-3 a Area m^2 Mstr 2.5461E+5 A |p| Pa
0.2963 b Perim m 30a 1.3516 B Ph(p) deg
0.2400 c Length m 3.2156E-2 C |U| m^3/s
1.1600E-3 d WallA m^2 -40.217 D Ph(U) deg
3198.2 E Htot W
3062.7 F Edot W
1200.5 G TBeg K
322.59 H TEnd K
stainless Solid type 3198.2 I H2k W
!--------------------------------- 31 ---------------------------------
JOIN thermal buffer tube end effects
2.5461E+5 A |p| Pa
1.3516 B Ph(p) deg
3.1485E-2 C |U| m^3/s
-40.217 D Ph(U) deg
3198.2 E Htot W
2998.9 F Edot W
322.59 G TBeg K
325.00 H TEnd K
!--------------------------------- 32 ---------------------------------
RPN how much heat must be rejected here? Resonator is insulated
0.0000 a G or T -1546.6 A ChngeMe
22h 12e + ~
!--------------------------------- 33 ---------------------------------
TX Small water Xger
6.6580E-3 a Area m^2 2.5427E+5 A |p| Pa
0.2690 b GasA/A 1.2740 B Ph(p) deg
1.0160E-2 c Length m 3.1819E-2 C |U| m^3/s
2.2860E-3 d radius m -40.949 D Ph(U) deg
123
sameas 32A e HeatIn W 1651.6 E Htot W
290.00 f SolidT K 2995.7 F Edot W
325.00 G GasT K
231.46 H SolidT K
ideal Solid type 1651.6 I H2k W
!--------------------------------- 34 ---------------------------------
DUCT PT connector (see pg 55 book 4 and concept.skf)
6.2070E-3 a Area m^2 2.5372E+5 A |p| Pa
0.2790 b Perim m 1.1616 B Ph(p) deg
5.1000E-2 c Length m 3.8215E-2 C |U| m^3/s
6.0000E-4 d Srough -50.713 D Ph(U) deg
1651.6 E Htot W
stainless Solid type 2993.1 F Edot W
!--------------------------------- 35 ---------------------------------
UNION Rejoin
8 a SegNum 2.5372E+5 A |p| Pa
sameas 8A b |p|Sft Pa =35A 1.1616 B Ph(p) deg
sameas 8B c Ph(p)S deg =35B 0.24013 C |U| m^3/s
sameas 8I d TSoft K =35G -86.583 D Ph(U) deg
sameas 8J e p20HLS Pa =35H 2.3874E-11 E Htot W
1198.9 F Edot W
325.00 G T K
-0.48008 H p20HL Pa
+ 36 DUCT Section of 3.5" tee
+ 37 CONE 3.5" to 4" cone to adapt to resonator (pg 93 book 4)
+ 38 DUCT Initial section of resonator
+ 39 BRANCH RC dissipator
+ 40 DUCT Continuation of resonator
+ 41 CONE 7 degree cone - 10.02" final diameter
+ 42 DUCT 10" duct P8
+ 43 SURFACE end cap of resonator
!--------------------------------- 44 ---------------------------------
HARDEND end of duct
0.0000 a R(1/z) =44G 8.5902E+4 A |p| Pa
0.0000 b I(1/z) =44H -178.64 B Ph(p) deg
0.0000 c Htot W 3.5950E-13 C |U| m^3/s
0.0000 d Ndot mol/s 91.562 D Ph(U) deg
2.3874E-11 E Htot W
5.5388E-11 F Edot W
1.4388E-15 G R(1/z)
-4.0109E-13 H I(1/z)
0.0000 I Ndot mol/s
124
7. Mixture separation
A sound wave in a binary gas mixture in a channel causes time-averaged molar ‡uxes of
the two components of the mixture in opposite directions, and can create a signi…cant mole-
fraction gradient along the channel. This occurs because oscillating lateral thermal di¤usion
causes the light and heavy components of the mixture to take turns enriching the central
region of the channel, where the oscillating motion is the greatest; the time phasing is best
when the oscillating enrichment and oscillating velocity are in phase, causing the velocity
oscillations to carry light-enriched gas in one direction and heavy-enriched gas in the other
direction [43, 44, 45]. The light component usually ‡ows in the direction of E:_
The thermoacoustic mixture-separation mole ‡ux and mole-fraction gradient are in some
ways analogous to the time-averaged enthalpy ‡ux and temperature gradient that a sound
wave creates in a stack in a standing-wave thermoacoustic refrigerator.
DeltaEC can model this e¤ect accurately [10].
nL + nH = 1 (7.1)
and
N_ L + N_ H = N_ : (7.2)
In a simple “bulk” ‡ow of the gas mixture (i.e., ‡ow that does not involve mass di¤usion,
mole-fraction gradients, or thermoacoustic mixture separation), it is also true that
N_ L = nL N_ : (7.3)
N_ H = nH N_ (7.4)
is also true in a bulk ‡ow, as expected from the symmetry of the equations. Thus, bulk
‡ow in a binary mixture can be uniquely described by assigning values to just two variables,
which can be chosen as N_ and nL :
Nontrivial ‡ows arise in binary mixtures whenever mass di¤usion or thermoacoustic
mixture separation occurs. For example, for N_ = 0 and no thermoacoustics, a nonzero N_ L
125
and an equal and opposite N_ H arise from ordinary di¤usion in the x direction whenever a
mole-fraction gradient dnL =dx dnH =dx exists; Eq. (7.3) is obviously not satis…ed. Thus,
DeltaEC must routinely handle nontrivial mixture ‡ows, which must be speci…ed by three
independent variables. DeltaEC takes those three variables to be N_ ; N_ L ; and nL :1
For dTm =dx = 0, DeltaEC allows the mole fraction to evolve with x in segment types
MIXBL (boundary-layer approximation), MIXSLAB (between parallel plates), and MIXCIRC
(circular channel), according to the theory described in Ref. [10] and references therein.
Nonzero dnL =dx is supported for three of DeltaEC’s gas mixtures: He-Ar, He-Xe, and
Ne-Xe. The user can create an external …le to specify the properties of other gas mixtures,
as described in Section 12.15.2. The equations being integrated in these MIX** segments are
dp1
= Fmomentum (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, etc.), (7.5)
dx
dU1
= Fcontinuity (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, etc.), (7.6)
dx
dTm
= 0, (7.7)
dx
dnL
= Fmix sep (p1 ; U1 ; Tm ; N_ ; N_ L ; pm ; !; geometry, gas properties, etc.). (7.8)
dx
The similar segment types MIXTBL, MIXTSLAB, and MIXTCIRC allow dTm =dx 6= 0; so the
equations being integrated in these segments are
dp1
= Fmomentum (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, etc.), (7.9)
dx
dU1
= Fcontinuity (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, etc.), (7.10)
dx
dTm
= Fenergy (p1 ; U1 ; Tm ; pm ; !; geometry, gas properties, etc.), (7.11)
dx
dnL
= Fmix sep (p1 ; U1 ; Tm ; N_ ; N_ L ; pm ; !; geometry, gas properties, etc.). (7.12)
dx
In DeltaEC’s BEGIN segment, we specify N_ and nL : If the “bulk NLdot”box is checked,
DeltaEC calculates N_ L using Eq. (7.3); if the bulk NLdot box is not checked, the user must
specify N_ L (perhaps as a guess).
Continuity of nL between segments is enforced. N_ and N_ L split up at TBRANCHs, add at
UNIONs, and are independent of x in all physical segments.
126
the middle or end of a mixture separator whose vigorous thermoacoustic processes maintain
a locally di¤erent nL ; in this case, nL can be regarded as discontinuous [10], and Eq. (7.3)
would not apply if the local thermoacoustic channel’s nL were used, even while the feedstock
‡ow itself is bulk [i.e., the feedstock mole fraction and feedstock ‡ows satisfy Eq. (7.3)]. In
other circumstances, one location of the mixture separator might be …rmly connected to an
essentially in…nite bath of feedstock, so nL would be maintained by net molar or partial-
molar ‡ows to or from the bath that might not satisfy Eq. (7.3). If a **END represents a
location where the acoustics stops and an enriched product ‡ows out of the system, Eq. (7.3)
must apply to the out‡ow.
Here are several possible situations, with discussions of how to initialize variables in BEGIN
and how to choose guesses and targets for each of them in DeltaEC. We only discuss the
mixture-speci…c guesses and targets here; realistic mixture-separation models typically also
include guesses and targets related to p1 and U1 :
Sealed system
The steady-state conditions of Refs. [43, 44] represent sealed systems with N_ = 0 every-
where. Those experiments had no steady in‡ow or out‡ow, but were constrained by the
mole fractions present in the original charge of gas mixture, so getting the right answer
from DeltaEC depends on constraining DeltaEC’s solution with an integral of nL over
the volume of the apparatus to match the total number of moles of light component that
was present in the original charge. Getting the correct answer to that integral would be
the target corresponding to the guessed value of nL in the BEGIN. In practice, we might
approximate such an integral in DeltaEC with a simple, volumetrically weighted average
of values at a small number of locations. Each such value could be captured with an RPN
segment. The BEGIN should have N_ = 0 (and, by default, N_ L = 0); because nothing ‡ows
in or out there.
127
in the feedstock, and the feedstock mixes abruptly into a locally di¤erent mole fraction at
the feedstock-end of the apparatus.
To model that situation with DeltaEC, the easiest approach is to put the BEGIN at
the exit end, because the exit ‡ow is bulk so we have fewer unknowns to guess there.
Suppose we know the incoming N_ and nL . We use these values in the BEGIN (with a minus
sign for the molar ‡ux), check the “bulk ‡ow” box to set N_ L appropriately, and integrate
straightforwardly to the other end of the channel, where we learn what the mole fraction
in the channel is where the feedstock mixes into it. This approach is the easiest because it
involves only two segments— BEGIN and MIXBL— and it involves no targets or guesses.
A seemingly more logical and readable approach to this situation is to begin at the
feedstock end, but then we need more DeltaEC baggage, to work around the fact that we
know less information a priori at that end:
0. BEGIN with the known p1 ; U1 ; feedstock N_ ; and feedstock nL . Since the feedstock
capillary ‡ow is bulk, check the “bulk” box in the BEGIN segment to establish the correct
N_ L .
1. RPN with 1a a guessed mole fraction at the feedstock end of the separation channel
and 1b having inp =nL .
2. MIXCIRC.
3. RPN expressing the bulk condition, Eq. (7.3), for the exit ‡ow, and target it. This
target is associated with the guess in segment 1. For example, put 0. in 3a and “nL Ndot
* NLdot - ”in 3b.
128
expression 0 = nL N_ N_ L by a large number, or divide it by N_ , to give DeltaEC a weighty
enough target.) This target corresponds to the guessed value of nL in segment 0.
Firm connection to feedstock bath in the middle of a long tube, products ex-
tracted at both ends
This time, let’s assume that we know the two exiting molar ‡ow rates and the mole fraction
of the feedstock bath in the middle. The feedstock bath must supply N_ and N_ L in the
center, but not necessarily in bulk proportions.
In hardware, how might we make a …rm connection to a …xed mole fraction in the middle
of a tube? We could have two capillaries there, and pump a huge mole ‡ux into one of them
and a nearly equal mole ‡ux at a slightly di¤erent mole fraction out of the other one. The
net ‡ow— the sum of these two— is not a bulk ‡ow.
0. BEGIN at one end of the channel with a guessed nL and the known N_ : Check the “bulk
‡ow”box to get the known N_ L ; because this is a location of product out‡ow.
1. MIXBL for the …rst part of the channel.
2. RPN Target the feedstock bath’s mole fraction here, with the target value in 2a and
nL in 2b.
3. RPN Guess the incremental N_ in 3a. Add the incremental N_ and N_ L to the separation
tube in 3b using inp Ndot + =Ndot inp 2a * NLdot + =NLdot .
4. MIXBL for the second part of the channel.
5. RPN expressing the bulk condition, Eq. (7.3), and target it. For example, put 0. in 5a
and nL Ndot * NLdot - Ndot / in 5b.
We have 2 targets (segments 2, 5) and 2 guesses (segments 0, 3).
129
Figure 7.1: Apparatus for helium–argon separation used in Ref. [10] and in this example.
130
BEGIN in front of top driver
7.9907E+04 a Mean P Pa
200.00 b Freq Hz
298.15 c TBeg K
3882.6 d |p| Pa G
0.0000 e Ph(p) deg
9.6980E-05 f |U| m^3/s G
-0.9501 g Ph(U) deg G
-2.6300E-08 h Htot W
-3.1800E-07 i Ndot mol/s
0.3406 j nL G
HeAr Gas type
!--------------------------------- 1 ---------------------------------
MIXCIRC tube #1 (bottom)
8.5630E-06 a Area m^2 3000.0 A |p| Pa
1.0000 b GasA/A -101.23 B Ph(p) deg
0.5000 c Length m 8.1411E-05 C |U| m^3/s
1.6510E-03 d radius m -101.23 D Ph(U) deg
-2.6300E-08 E Htot W
0.1221 F Edot W
0.3406 G nL Beg
0.4661 H nL End
!--------------------------------- 2 ---------------------------------
RPN pressure amplitude in center
3000.0 a G or T = 2A? 3000.0 A |p1|
p1 mag
!--------------------------------- 3 ---------------------------------
RPN set Re[z]= rho a
0.0000 a G or T = 3A? 3.6028E-16 A null
rho a * p1 U1 / 1a * real - rho a * /
!--------------------------------- 4 ---------------------------------
RPN set Im[z] = 0
0.0000 a G or T = 4A? 1.8585E-16 A Im z
p1 U1 / 1a * imag rho a * /
!--------------------------------- 5 ---------------------------------
RPN feedstock concentration
0.5000 a G or T (t) 0.0000 A nL Feed
0
!--------------------------------- 6 ---------------------------------
RPN [a]=injection (micromol/sec)
1.5900 a G or T (t) 6.8668E-07 A ChngeMe
1.2720E-06 B ChngeMe
6a 1e-6 * Ndot + =Ndot 6a 1e-6 * 5a * NLdot + =NLdot
!--------------------------------- 7 ---------------------------------
MIXCIRC tube #2 (bottom)
sameas 1a a Area m^2 2350.2 A |p| Pa
sameas 1b b GasA/A 163.69 B Ph(p) deg
sameas 1c c Length m 6.7249E-05 C |U| m^3/s
sameas 1d d radius m 163.84 D Ph(U) deg
-6.1169E-05 E Htot W
7.9026E-02 F Edot W
0.4661 G nL Beg
0.5398 H nL End
!--------------------------------- 8 ---------------------------------
RPN exit bulk condition
0.0000 a G or T = 8A? -3.3295E-16 A null
nL Ndot * NLdot - Ndot /
The <.sp> …le from that run was imported into a spreadsheet to convert nL to nH = 1 nL
for plotting in Ref. [10]. Figure 7.2 shows the result, compared with measurements of the
heavy mole fraction nH that were made at 7 locations along the separation tube.
Next, an incremental plot was made, with feedstock N_ as the independent variable and
with an RPN segment maintaining the molar ‡ow rate in the BEGIN segment equal to 20%
of the feedstock ‡ow. The result is shown in Fig. 7.3.
131
Figure 7.2: Heavy mole fraction nH vs position in tube for one particular operating condition that
is described in the text. The points are measurements from Ref. [10] and the curve is the result of
DeltaEC’s calculation as described in the text.
Figure 7.3: Heavy mole fractions nH at x = 0 and x = 1 m, at various feedstock molar ‡ows,
with one product ‡ow equal to 20% of the feedstock ‡ow and the other product ‡ow equal to 80%
of the feedstock ‡ow. The points are measurements from Ref. [10] and the curves are DeltaEC
calculations as described in the text.
132
Part II: Reference
133
8. General principles
134
The total power ‡ow H_ tot is often independent of x; but there are important exceptions.
In heat exchangers, H_ tot changes by the amount of heat added (or removed), and in transduc-
ers it changes by the amount of power added (or removed). If the ANCHOR logistical segment
(described in Section 11.3.1) is used upstream, H_ tot also changes in every component except
STK**s, because ANCHOR means that acoustic power dissipation is removed locally as heat in
each such component.
Usually the initial values of at least some of these integration variables are unknown at
the BEGIN, so DeltaEC uses a shooting method to …nd initial values that are consistent
with targeted results that are identi…ed by the user elsewhere in the model. The shooting
method is described in the next section.
After the shooting method has met its targets, DeltaEC knows the complete solution
p1 (x); U1 (x); Tm (x); Tsolid (x); H_ tot (x); p2;0;HL (x); and nL (x): Three additional useful func-
tions of x are calculated straightforwardly from that solution: the acoustic power E, _ the
_
exergy ‡ow X, and the total second-order time-averaged pressure p2;0;tot :
_ 1 h ei
E(x) = Re p1 U1 ; (8.8)
2
_ T0 _ T0
X(x) = E+ 1 H_ 2;k + N_ (mwm T0 msm )
Tm Tm
nL 1 nL
+N_ L Runiv T0 ln + N_ N_ L Runiv T0 ln ; (8.9)
nL;0 1 nL;0
p2;0;tot (x) = p2;0;rev + p2;0;HL (8.10)
2 2
jp1 j mjU1 j
= + p2;0;HL : (8.11)
4 m a2 4A2gas
The expression for acoustic power is discussed in Ref. [13] and should be familiar to most
acousticians.
The expression for X_ is an amalgam of the basic thermoacoustic exergy ‡ow derived in
Ref. [13], a steady-‡ow component derived in Ref. [32], and the N_ 6= 0 extension of the ther-
moacoustic mixture-separation expression derived in Ref. [44]. The reference temperature
T0 ; which is the ambient temperature at which heat has no value, is set to 300 K by default in
BEGIN segments and can be changed to a di¤erent value by using the operation =Tzero in an
RPN segment soon after the BEGIN. Steady pressure makes no contribution to the steady-‡ow
component because we assume that the reference pressure for exergy is pm : The second line
of Eq. (8.9) is relevant only for mixtures. The zero-exergy mixture mole fraction, nL;0 ; is the
mole fraction for which the mixture has no value, typically the feedstock’s mole fraction. It
is set to 0.500 by default in BEGIN segments, and can be changed to another value by using
=nLzro in an RPN segment soon after the BEGIN.
The reversible part of the second-order time-averaged pressure, p2;0;rev ; is derived in Ref.
[36], and measurements of p2;0;tot are described in Ref. [35].
These ten functions of x, as well as many others (e.g., the mean density m , the vibration
force F1 described in Section 11.4.1) are available to the user in tabular and graphical forms
via the <.sp> …le, at the end of each segment in the <.out> …le, and/or at any location
via RPN segments.
135
8.1.1. Additional details for N_ and p2;0;HL
DeltaEC’s input parameter Ndot, denoted in the Users Guide by N_ , represents the true
time-averaged molar ‡ow, which is the steady-‡ow quantity under the most direct experi-
mental control and the quantity that carries steady enthalpy ‡ux
H_ N_ = N_ mwm ; (8.12)
where m is the molar mass and wm is the mean enthalpy per unit mass. The total power
is the power that obeys the …rst law of thermodynamics for open control volumes. The total
power H_ tot is the sum of the second-order time-averaged thermoacoustic enthalpy ‡ux, the
ordinary thermal conduction in the x direction, and H_ N_ = N_ mwm :
In DeltaEC, mean enthalpy is referenced to zero at temperature T0 , which is set to 300
Kelvin by default in the …rst BEGIN segment of a model. (This default can be overridden
by using the =Tzero operator in an RPN segment.) For an ideal gas, Eq. (8.12) can then be
regarded as
H_ N_ = N_ mcp (Tm T0 ) : (8.14)
To second order, Z
_ 1
N m ' Re [ 1 u e1 ] dAgas + m U2;0 ; (8.15)
2
where m is the molar mass. However, the second-order time-averaged momentum equation
shows that second-order time-averaged head-loss gradients are generated by U2;0 ; with no
contribution from 1 u e1 : De…ning _ 2
m U2;0 =m; substituting 1 = ( =a ) p1 m T1 in
1
R N2;0
Eq. (8.15), and solving for 2 Re [T1 u e1 ] dAgas in the standard thermoacoustic expression for
H_ 2;k ; it is easy to show that Eq. (8.15) is equivalent to
dTm 1
N_ 2;0 = N_ + H_ 2;k + (Agas k + Asolid ks ) + _
E; (8.16)
mcp dx ma2 mcp
which is the molar ‡ow that is used in most of DeltaEC’s p2;0;HL calculations.1 Note that
N_ 2;0 is often nonzero when N_ = 0; which means that interesting second-order, time-averaged
head losses arise in many circumstances having N_ = 0, such as in the regenerators of TASHEs
and pulse-tube refrigerators. Click the Enable Ndot button in the BEGIN segment’s Optional
parameters to “turn on”the display of p20HL even when N_ = 0.
136
of the thermoacoustic apparatus, then DeltaEC’s numerical integration is straightforward,
proceeding from the initial end to the …nal end. But usually one or more boundary conditions
are instead speci…ed at the …nal end or somewhere in the middle. In such circumstances
DeltaEC uses a shooting method [16], by guessing any unknowns at the BEGIN, integrating
to the **END, comparing the results with the target boundary conditions imposed at that
end and/or elsewhere, and adjusting its guesses until the targets are met. Up to eighteen
guesses and eighteen targets are allowed. One of DeltaEC’s most powerful features is
that the guesses and targets are not limited to the conventional choices, which would be
the initial values of integration variables at the BEGIN as guesses and calculated values of
integration variables at the **END as targets. Any results can be used for targets, and
any upstream variables that have e¤ects on downstream target variables can be used as
guesses. This enables DeltaEC to calculate a resonance frequency, a geometrical dimension,
a temperature, or even the concentration in a binary gas mixture to satisfy desired boundary
conditions.
DeltaEC treats its collection of n guesses and n targets as a system of n nonlinear
equations,
where gi is the ith guess, ti is the ith target value, and ri is the ith calculated result, which
we hope can be made equal to the ith target ti by suitable choices of the n guesses gi : Thus,
DeltaEC must seek values of the n guesses gi that make Fi = 0 for all i: Several meth-
ods for …nding the simultaneous zeroes of complicated, simultaneous functions exist. The
routine incorporated in DeltaEC is called DNSQ, and it is part of the SLATEC Common
Mathematical Library, which is freely available through the internet software repository at
http://www.netlib.org. The algorithm used is a modi…cation of the Powell hybrid method
[46, 47] (a.k.a. the Powell dogleg method). The ordinary DeltaEC user can just as well
imagine that DNSQ is the Newton–Raphson method, which is easy to understand and is
described in Ref. [16] and many tutorial websites. The description below is based on this
oversimpli…cation.2
A Newton–Raphson algorithm X begins with initial “guessed” values of the guesses, and
2
calculates the n values of Fi . If (Fi ) is not close enough to zero, the algorithm estimates
2
the n local slopes
137
The algorithm forms a “better”set of guesses based on the assumption that these functions
Fi are linear, and repeats the process, until it judges the results to be close enough to the
targets, according to
qX qX
2
(wi Fi ) Tolerance (wi ti )2 , (8.23)
where the weighting factors wi passed from DeltaEC to DNSQ are given by
1
wi = if the target is a phase angle, (8.24)
360
1
= if jti j 10 4 and the target is not a phase angle, (8.25)
jti j
= 104 if jti j 10 4 and the target is not a phase angle (8.26)
138
8.3.1. Integration options
Runge-Kutta steps. Default: 10. Recommended range: 1 100: This is an integer Nint
that determines the number of integration steps used to span each STK**, DUCT, CONE,
MIX**, and VX** segment. It does not a¤ect other segment types. It also determines
the resolution in state plots: Nint +1 points per integrable segment. Larger values will
cause a slower, more accurate computation. Larger values are bene…cial for long VX**
segments and when large nonzero N_ ‡ows through STK** segments. Small values will
increase speed at the price of integration accuracy and possible convergence problems.
Minimum Temperature. Default: 10 K. This is a temperature ‡oor to prevent DeltaEC’s
shooting method from exploring unphysical temperatures such as negative tempera-
tures. Brave users with special needs at lower temperatures (and with their own, user-
de…ned thermophysical properties …les!) can set this ‡oor to a lower value. (Some of
DeltaEC’s internal gases use a higher temperature ‡oor. See Chapter 12.)
139
EpsFcn. Default: 10 10 . Recommended range: > 5 10 15 . To estimate @Fi =@gj using
p p
Eq. (8.22), DeltaEC sets gj = gj EpsFcn: (But if gj = 0; then gj = EpsFcn:)
A large value of EpsFcn can sometimes speed iterations when initial values of guesses
are very inaccurate. Too small a value can cause the solver to lose its way completely,
as roundo¤ error in the computer ruins the accuracy with which @Fi =@gj is calculated.
The optimal value is computer-dependent and model-dependent. To guard against
too-small values, EpsFcn has a built-in ‡oor of 2 10 16 ; twice the double-precision
roundo¤ bound (see Wikipedia, “Machine epsilon”).
Step Bound Factor. Default: 100. Recommended range: 0.01–100. This parameter con-
trols the boldness with which DeltaEC …rst tries to adjust the gi , after the @Fi =@gj
have been calculated for the …rst time. Some di¢ cult cases can bene…t from a smaller
Step Bound Factor.
Warnings were generated while parsing... This may indicate a problem with the <.out>
…le. This can occur when a version-5 model is opened in version 6 for the …rst time,
or when the savvy user has made a minor mistake when editing a <.out> …le with a
text editor instead of using DeltaEC’s built-in parameter- and segment-modi…cation
methods. Examine the <.err> …le for speci…c information. When translating a model
from version 5 to version 6, DeltaEC gives warnings whenever it is unsure of itself,
even though it usually makes the right assumptions about the translation.
Success, Partial convergence, or FAILED. The Run Monitor at the bottom of the
main display gives one of these three messages after every run, on a panel of green,
yellow, or red. “Success” indicates that targets were met with the accuracy implied
140
by Convergence Tolerance (or no targets and guesses exist at all). “FAILED” and
Iteration is complete, but some results may not be near their targets in-
dicate that targets have de…nitely not been met to DeltaEC’s satisfaction. If the error
occurs on a model that has displayed good convergence under other conditions, this
may simply be one pathological point. The user may be able to jump-start it by
manually (somewhat intelligently) changing the value of one or two members of the
guess list to put the solver on the right track. Otherwise, consider the tactics de-
scribed in Section 5.4. “Partial convergence” indicates that targets are almost met
to DeltaEC’s satisfaction, and in most cases the user can also be satis…ed. Usually
“Partial convergence” can quickly be changed to “Success” simply by running again.
If the frequent appearance of “Partial success” becomes a nuisance, try changing the
Tolerance and/or Normalization Mode, described in Section 8.3.2.
The plotter indicates such suspicious and failed convergence by plotting the point in red.
Below is a key to interpreting the “info”codes:
info= Signi…cance
1 Partial convergence: The solver considers the iteration successful,
but the residual error is suspiciously large.
2 The solver was making progress, but the maximum number of iterations
has been reached. Another run might make further progress.
4 FAILED: The solver was unable to converge.
Sometimes the user may want to delete the prepended asterisk and number. For example,
if the code is “info=2,”and the remaining error that the solver was working on was already
quite small, it might be quite valid to use this plot point instead of discarding it.
DeltaEC assumes that p1 depends only on x; that u1 depends only on x except within
the viscous boundary layer, and that Tm depends only on x: Thus, DeltaEC cannot
accurately model ‡ow in short-radius elbows, ‡ow and pressure at abrupt transitions
between DUCTs of di¤erent areas (as shown in Fig. 8.1), or inertial or thermal e¤ects in
thermal bu¤er tubes and pulse tubes with signi…cant Rayleigh streaming. (For exam-
ple, in STKDUCT, Tm (x) and Tsolid (x) are equal for all x; which is a very bad approxi-
mation if signi…cant Gedeon streaming, Rayleigh streaming, or jet-driven streaming
141
Figure 8.1: (a) Naive and (b) more-realistic sketches of streamlines where two large-diameter DUCTs
join. DeltaEC performs inherently one-dimensional calculations, so it naively imagines that the
streamlines are as shown in (a). In reality, some high-velocity ‡ow extends into the large-diameter
duct as shown in (b), adding inertance that DeltaEC does not automatically account for, and
resistive minor losses also occur at high Reynolds numbers. Even (b) may not be su¢ ciently
realistic, because the ‡ow at such a transition is seldom left–right symmetrical. These e¤ects can
and usually should be modeled with RPN calculations and MINOR and/or IMPEDANCE segments.
is present.) We use MINOR segments, RPN segments, IMPEDANCE segments, and o¤-line
hand calculations to try to account for 2- and 3-dimensional e¤ects like these.
The derivations of most thermoacoustic equations rely on the assumption that the gas dis-
placement amplitude jU1 j =!A is much smaller than all other relevant dimensions in
the x direction. This assumption is often violated in thermoacoustic components, espe-
cially heat exchangers, whose length in the x direction is seldom larger than 2 jU1 j =!A:
The turbulence algorithm in DUCT, CONE, and VX** is based on the quasi-steady approxi-
mation, namely that the pressure gradient at any instant of time in time-dependent
‡ow is the same as that which would be present for steady ‡ow at the same rate. This
is known to be naive [48, 49, 50, 51].
Similarly, the turbulence algorithm in STKSCREEN and STKPOWERLW is based on the quasi-
steady approximation, which is a poor approximation when gas displacement ampli-
tude is smaller than, or comparable to, the screen pore size. Thus, this aspect of
DeltaEC works poorly at low amplitudes.
Harmonics are often present in p(t) and U (t); but DeltaEC works exclusively with the
fundamentals p1 and U1 : This limitation usually has little impact on the accuracy of
power calculations [13], but experiments being compared to DeltaEC calculations
must use instruments that measure the fundamental components of the wave, not rms
or peak values.
142
Be alert for situations where DeltaEC cannot be expected to understand the user’s in-
tentions. DeltaEC does not warn the user of all such possible inconsistencies and
ambiguities. For example, DeltaEC can’t guess why a user would allow nonzero N_ to
‡ow through an enclosed **SPEAKER or **DUCER, so no change in p2;0;HL is calculated.
The velocity term in Eq. (6.12) for p2;0;tot depends on area, so don’t be surprised if you
disagree with its calculation at an IMPEDANCE segment, since IMPEDANCEs have no area.
Allowing signi…cant nonzero H_ 2;k (when H_ 2;k 6= E)
_ to ‡ow through an IMPEDANCE or a
tight-gapped HX or TX is physically implausible, and leads to errors in the calculation
of N_ 2;0 via Eq. (8.16) and thus to erroneous calculation of p2;0;HL .
VXQ1 and VXQ2 rely on assumptions that may be invalid in some circumstances. See Section
10.7.3 for details.
The shooting method requires that target results change smoothly as guesses are varied,
so the near-linearity assumed in Eqs. (8.22) can be a good-enough approximation for
small variations in the guesses. So, be wary of building step-function or absolute-value
calculations into RPN segments if downstream target results depend on them. For
example, if a model is operating near j 1 j = D at the location of an ori…ce plate in a
DUCT, abruptly switching on a minor loss as shown here may be very problematic if
the model is operating close enough to the transition that DeltaEC’s guess variations
carry it back and forth between j 1 j > D and j 1 j < D .
5 RPN inp = orifice diameter (m); result = area (m2)
1.0000E-2 a G or T 7.8540E-5 A m2
inp 2 / sqrd pi *
6 RPN Turn on minor loss only if |xi_1| > D. Result is 0 or 1.
0.0000 a G or T 1.0000 A K+, K-
U1 mag 5A / w / 5a - sto abs rcl / 1 + 2 /
7 MINOR small orifice plate in big duct
sameas 5A a Area m^2
sameas 6A b K+
sameas 7b c K-
A more subtle abrupt change in results can arise from DeltaEC’s treatment of phase
angles. When displaying its internal complex numbers as magnitude and phase,
DeltaEC selects the phase to be 180 < 180 . Targeting phase angles near
0 , 90 , or 90 is unproblematic. However, when trying to target a phase to be 180 ,
DeltaEC’s shooting method explores guess values that yield results both slightly
above and slightly below 180 . The results that are slightly above 180 will be seen
by the shooting method as slightly above 180 (e.g., 181 becomes 179 ), and the
shooting method will have no chance to hit this target. The simplest way to avoid this
problem is to shift the time phase of the entire model by a large, convenient amount,
such as 180 or 90 . (See Section 2.6.2.) Another simple work-around is to target
sin = 0 instead of = 180 , although this technique can sometimes let the shooting
method “jump” from = 180 to = 0 . More awkwardly, to convert a standard
DeltaEC phase angle to a function that changes smoothly as it passes through 180 ,
use the following trick, which yields ph(U) if ph(U) > 0, and adds 360 to ph(U) if
ph(U) < 0, thus shifting the troublesome discontinuity from 180 to 0 .
7 RPN target phase(U1) here to be 180
180.00
U1 arg U1 arg abs - U1 arg / 180 * U1 arg +
143
9. Good design habits
We pay attention to the interfaces between DeltaEC-modeled heat exchangers and their
environment. Even though DeltaEC is not set up to model anything outside the “skin”
of the solid that is in contact with the thermoacoustic gas in heat exchangers, many e¤ects
beyond that boundary are important, and should be modeled, either using RPN segments
within DeltaEC or using calculations elsewhere (e.g., spreadsheets). For example, in a
water-cooled heat exchanger, non-negligible temperature di¤erences usually exist across the
water boundary layer (due to the …nite thermal conductivity of the water) and between the
inlet and outlet streams (due to the …nite speci…c heat and ‡ow rate of the water). Our
sponsors usually care about the water inlet temperature, not DeltaEC’s heat-exchanger
internal temperature. Similar considerations apply in a drilled metal-block heat exchanger
144
that relies on conductivity of metal to transfer heat between the thermoacoustic gas in the
drilled channels and an external heat source or sink at the perimeter of the block.
145
We also account for unavoidable heat ‡ow through insulation layers wrapped around
components, with RPNs or separate spreadsheets.
In systems with toroidal topology (e.g., TASHE [40, 5]), we suppress Gedeon streaming,
or account for it with nonzero N_ .
9.1.4. Geometry
As a design evolves towards its …nal state, we add more and more geometrical detail to the
DeltaEC model. Even including the short gaps between regenerators and heat exchangers,
or accounting for weld shrinkage that shortens such gaps, is not too much attention to
detail.1 When shop drawings are being prepared, we modify the DeltaEC model to re‡ect
machine-shop realities (e.g., inch-based dimensions; available drill sizes). While assembling
parts, we often check measured dimensions against the DeltaEC model one more time.
9.2.1. Turbulence
We habitually use a value of 5 10 4 for the relative surface roughness in DUCTs and CONEs,
even when it is greater than the actual roughness, because this gives results that …t a lot of
1
If a heat exchanger’s GasA/A is small and its pores are large compared with those of the adjacent STK**,
a gap is absolutely necessary to prevent U1 from getting crowded into a reduced area in the beginning of
the STK**. DeltaEC’s naive one-dimensional calculations do not show the user the need for such gaps.
146
N_ = 0 data over the past 15 years. When steady ‡ow is superimposed, we use the “average”
turbulence setting, which is consistent with halfway between the two lines in Fig. 9 in Ref.
[42].
We use MINOR segments generously wherever appropriate, e.g., corner turning, TBRANCHs
and UNIONs, and area changes (except very gently tapered cones), with Idelchik’s book [52]
for estimates of the K coe¢ cients. We even check the need for MINOR segments at transitions
between heat exchangers and open ducts, where abrupt changes in GasA/A occur. In toroidal
TASHEs [40, 5], we might include almost a dozen MINOR segments.
147
described in Ref. [29] while avoiding cone angles that might create ‡ow separation, and even
then we usually deduct 10% of E_ at the cold heat exchanger from DeltaEC’s raw prediction
of cooling power in pulse-tube refrigerators, and add 10% of E_ at the hot heat exchanger to
the heater power in engines having thermal bu¤er tubes. (This correction also a¤ects the
ambient heat exchangers at the other ends of the tubes.) It is not unreasonable to encode
the math described in Ref. [29] in RPN segments (it takes more than one!) to keep an eye on
this issue in a model with a pulse tube or thermal bu¤er tube.
Large-aspect-ratio refrigerator regenerators can su¤er from regenerator internal stream-
ing, as described in Ref. [53], which must be accounted for, even if only approximately.
(See also “Adiabatic–isothermal mixing”and “Heat leaks— black-body radiation”above.)
148
Figure 9.1: The vertical axis shows the experimentally observed extra TH (blue) or Q_ H (pink), in
excess of DeltaEC’s calculations, divided by DeltaEC’s values, for standing-wave engines that
we have used. The horizontal axis is at the ambient end of the stack, divided by rh : All data
are scaled to jp1 j =pm = 0:10 at the hot end of the stack, using jp1 j3 scaling for extra Q_ H and jp1 j2
scaling for extra TH : Data from three honeycomb (STKCIRC) stacks are shown: “simil honeycomb”
is Ref. [25], “cascade honey” is Ref. [19], and “Tex honeycomb” is Ref. [54]. Unpublished data
from two spiral (STKSLAB) stacks are shown: “Cryenco 100 gpd” is from the engine shown in Fig.
4 of Ref. [55], corroborated by measurements on a half-scale model at Los Alamos, and with the
range of data for heater power indicating low acoustic load (more excess heat) vs. heavy acoustic
load (less excess heat), and “Tek spiral” is from the engine described in Ref. [54] but with the
honeycomb stacks replaced by spiral stacks as shown in Fig. 1.11 in Ref. [13].
149
10. Physical segments
All of DeltaEC’s physically meaningful segment types are listed in this chapter, organized
by function. Logistical and management segments appear in the next chapter. An alpha-
betical index of all segment types is in the Segment Index at the end of the Users Guide.
Input variables:
Area (m2 ) Cross-sectional area A available to the gas, based on inside dimensions of the
DUCT.
Laminarnturbulent Whenever the laminar box is checked, the calculations of dp1 =dx and
dp2;0;HL =dx in the DUCT are based on noninteracting ‡ow, using laminar ‡ow for dp1 =dx
and either laminar or turbulent ‡ow for dp2;0;HL =dx, depending on the steady-‡ow
Reynolds number. The laminar condition is identi…ed in the <.out> …le as a blank
line where input d would otherwise be.
150
Srough This input variable appears if the turbulent box is checked in the Optional Pa-
rameters dialog. Srough is "; the surface roughness inside the DUCT, relative to the
diameter. This is called “relative roughness”in many ‡uid-mechanics books, for steady
‡ow. However, for oscillating ‡ow, we usually regard Srough as a …tting parameter,
with a typical value of " = 5 10 4 : Larger values yield higher turbulent dissipation
_
of E:
Optimistic/Normal/Pessimistic These choices appear if the turbulence box is checked
in the Optional Parameters dialog and if the “Steady ‡ow”box is checked in the BEGIN
segment. These choices control the nonlinear interaction between oscillating ‡ow and
steady ‡ow in the DUCT. “Optimistic” and “Pessimistic” refer to the two calculation
schemes described in Ref. [42], and “Normal” is the average of the two. For the
optimistic case, the turbulent steady and turbulent oscillating ‡ows are assumed to
be independent, simply superimposed. For the pessimistic case, the turbulent steady
‡ow and turbulent oscillating ‡ow are assumed to interact with each other and to
increase each other’s resistive e¤ects according to the quasi-steady approximation. In
the <.out> …le, Optimistic/Normal/Pessimistic is recorded with SrOptm, Srough, or
SrPess to indicate the name of ".
Solid The solid material on the inside surface of the DUCT. (This is used only in the calcu-
lation of s ; which is seldom signi…cant in DUCTs.)
Potential targets:
None.
Master–slave links:
The perimeter of a DUCT can be slaved to its area to keep the cross-sectional shape the same
when area is varied.
The length of a DUCT can be slaved to that of another segment to keep the total length
constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
151
which are equivalent to
1 a2 d2 p1
1+ f p1 + (1 f ) = 0: (10.3)
1+ s !2 dx2
i! m
p1;out = p1;in cos k x U1;in sin k x;
(1 f )kA
i(1 f )kA
U1;out = U1;in cos k x p1;in sin k x; (10.4)
! m
In DUCTs, we de…ne the e¤ective radius to be r0 = 2A= : (For circular cross sections, this
is indeed the radius of the circle. For other cross sections, it is twice the hydraulic radius.)
In narrow DUCTs, for r0 = j < 25, f and f are calculated using complex Bessel functions
For intermediate values, linear interpolation is used to make a smooth match between the
two regimes. While the narrow-duct solution assumes a circular cross section, the shape of
the duct is irrelevant in the boundary-layer approximation. A square duct with dimensions
p
much larger than the penetration depth can be modeled simply by choosing = 4 A, for
example.
When turbulent, p1 and U1 are numerically integrated according to Eqs. (10.1) and
(10.2) above, with f and f modi…ed as described below to account for the turbulence.
DeltaEC’s turbulence algorithm is described in detail in Chapter 7.2 of Ref. [13] and is
based on the quasi-steady approximation: that turbulent oscillatory ‡ow is described by
the Moody friction factor at each instant of time during the oscillatory ‡ow. (See any
engineering ‡uid mechanics textbook, such as Ref. [56], to review the Moody friction factor
as a function of Reynolds number and relative roughness of the pipe wall.) This assumption
must be excellent in the low-frequency limit, in which r0 = ! 0: This limit is approached in
the inertances of the lowest-power pulse-tube refrigerators. We have learned by experience
with a few large systems that the assumption is reasonable for large r0 = ; which is of interest
in the resonators of standing-wave thermoacoustic systems. For experimental validation of
the assumption for intermediate r0 = ; see Ref. [15].
152
To account for turbulence, DeltaEC increases the resistive component of the pressure
gradient, and hence the viscous power dissipation, by
2
NR [fM (1 9 =32) NR dfM =dNR ]
m = : (10.8)
24 rh2 Re [i= (1 f )]
DeltaEC evaluates the Moody friction factor fM and its derivative with respect to Reynolds
number, dfM =dNR ; as functions of Reynolds number and " using the iterative expression
1 18:574
p = 1:7385 2 log10 2" + p ; (10.9)
fM NR fM
which is a remarkably good approximation [56, 57]. (In our own work at Los Alamos, we
have set roughness equal to 5 10 4 for many years, as representative of most turbulent
acoustic losses in ducts, even when the actual surface roughness over pipe diameter is smaller
than this value.) DeltaEC decreases the inertial pressure gradient by
2
1 =r0
m0 = (10.10)
1 =m r0
to correct approximately for the steeper velocity gradient at the wall, which increases the
e¤ective area open to gas that contributes to inertial e¤ects. The factors m and m0 are
applied using
1 1
Im = m Im ; (10.11)
1 f 1 f , laminar
1 1
Re = m0 Re ; (10.12)
1 f 1 f , laminar
1
1 1
f = 1 Re + i Im (10.13)
1 f 1 f
153
Calculation of p1 (x) and U1 (x) with N_ =
6 0:
and "2;0 = mN_ 2;0 = m jU1 j. The derivations of these expressions can be found in the appendix
of Ref. [42].
154
Calculation of p2;0;HL (x) with N_ =
6 0:
How the ‡ow is treated depends on the roughness parameter in the DUCT. There are three
cases: optimistic, pessimistic, and “normal,”which is the average of the other two.
For the optimistic case, the turbulent steady and turbulent oscillating ‡ows are assumed
to be independent, simply superimposed. The steady-‡ow pressure drop is calculated using
!2
dp2;0;HL 1 mN_ 2;0 fMoody,steady
= sign(N_ 2;0 ) m : (10.22)
dx 2 mA 4rh
The steady-‡ow Moody friction factor fMoody,steady is calculated with a Reynolds number
based on the steady velocity:
mjN_ 2;0 j4rh
NR;2;0 = : (10.23)
A
For the pessimistic case, the turbulent steady ‡ow and turbulent oscillating ‡ow are
assumed to interact with each other and to increase each other’s resistive e¤ects, as described
in the appendix of Ref. [42] ignoring any dependence on p1 and in Chapter 14 including the
lowest-order dependence on p1 : From Chapter 14, the steady-‡ow pressure drop is calculated
using
dp2;0 m2 N_ 2;0
2
1 jp1 j cos
= sign(N_ 2;0 ) 1+ 2
+ fM;max
dx 8 m r h A2 2"2;0 pm "2;0
2"22;0 2 j"2;0 j + 1 dfM
NR;max ; j"2;0 j 1; (10.24)
2"22;0 (1 + j"2;0 j) dNR
2 nh q
m jU1 j 2 1
= 1 + 2" 2;0 sin " 2;0 + 3" 2;0 1 "22;0
8 r h A2
2 jp1 j 2 + "22;0 q
+ cos "2;0 sin 1 "2;0 + 1 "22;0 fM;max
pm 3
3=2 + "22;0 q dfM
+ "2;0 3"2;0 1 "22;0 1 + 2"22;0 sin 1 "2;0 NR;max ;
1 + j"2;0 j dNR
j"2;0 j 1; (10.25)
_
Calculation of Tm (x) and H:
Tm is not a¤ected by DUCTs.
The exiting energy ‡ow is left as H_ 2;k;out = H_ 2;k;in in insulated mode, described in Section
11.1.2. This essentially assumes that the side walls are thermally insulated, so the heat
generated by acoustic power dissipation …nds its way to a heat exchanger somewhere else.
If several DUCTs (and/or CONEs) are strung together in insulated mode, the power dissipated
in all of them must show up at one or two nearby heat exchangers.
155
The exiting energy ‡ow is computed as H_ 2;k;out = E_ out in “anchored” mode, described
in Section 11.1.2. This essentially assumes that the duct wall is thermally anchored, e.g.,
by a water bath, so heat generated by acoustic-power dissipation is carried away locally.
Thermoacoustic heat transport in the boundary layer along the perimeter, which in fact
contributes a small di¤erence between H_ 2;k and E_ in ducts, is neglected in “anchored”
mode.
10.1.2. CONE
General description:
Use CONE to …nd the evolution of p1 and U1 with x in tapered channels of any cross-sectional
shape (e.g., square, circular) by giving suitable areas and perimeters. The perimeter is
linearly interpolated between its initial and …nal values, and the area is quadratically inter-
polated between its initial and …nal values. Use this segment only when Tm is independent
of x: (See STKCONE in Section 10.6.1, to allow nonzero dTm =dx.) The calculation is identical
to that for a DUCT if the initial and …nal areas are equal and the initial and …nal perimeters
are equal. As discussed below, the laminar calculation is accurate for circular cones of any
diameter, and for large cones (i.e., with r0 ; ) of any cross-sectional shape.
In a CONE, DeltaEC’s turbulence algorithm is controlled through optional parameter f,
the relative roughness ", whose de…nition for steady ‡ow can be found in ‡uid-mechanics
textbooks [56]: roughness height divided by pipe diameter. We have found that " = 0:0005
works well in high-amplitude acoustics, even if the geometrical roughness is smaller than
this value. Omitting this parameter forces the oscillatory calculation to be laminar at all
Reynolds numbers.
At high enough amplitudes, we believe that a steeply tapered CONE experiences separation
during the half of the cycle when the ‡ow goes from small toward large radius, just like the
separation at high Reynolds number in a too-steeply-tapered di¤user. DeltaEC does not
warn of this condition. To account for such separation, the user could include minor loss,
described in Section 10.2.4.
Input variables:
AreaI (m2 ) Cross-sectional area AI available to the gas at the initial end of the CONE, based
on inside dimensions.
PerimI (m) Perimeter I of the initial inside cross section described above.
AreaF (m2 ) Cross-sectional area AF available to the gas at the …nal end of the CONE, based
on inside dimensions.
PerimF (m) Perimeter F of the …nal inside cross section described above.
156
Laminarnturbulent Whenever the laminar box is checked, the calculations of dp1 =dx and
dp2;0;HL =dx in the CONE are based on noninteracting, laminar ‡ow. The laminar condi-
tion is identi…ed in the <.out> …le as a blank line where parameter f would otherwise
be.
Srough This input parameter appears if the turbulent box is checked in the Optional Pa-
rameters dialog. Srough is "; the surface roughness inside the CONE, relative to the di-
ameter. This is called “relative roughness”in many ‡uid-mechanics books, for steady
‡ow. However, for oscillating ‡ow, we often regard Srough as a …tting parameter, with
a typical value of 5 10 4 : Larger values yield higher turbulent dissipation of E:_
Solid The solid material on the inside surface of the CONE. (This is used only in the calcu-
lation of s ; which is seldom signi…cant in CONEs.)
Potential targets:
None.
Master–slave links:
The initial perimeter of a CONE can be slaved to the initial area, to keep the cross-sectional
shape the same when area is changed.
The …nal perimeter of a CONE can be slaved to the …nal area to keep the cross-sectional
shape the same when area is changed.
Both perimeters of a CONE can be slaved to their respective areas to keep the cross-
sectional shapes the same when either or both areas are changed.
The length and both perimeters of a CONE can be slaved to its two areas, to keep the
cross-sectional shapes of the ends and the wall taper angle constant when either or both
areas are changed.
The length of a CONE can be slaved to that of another segment to keep the total length
constant.
(For an introduction to master–slave links, and a complete listing of possible links for all
segments, see Section 2.7.2.)
157
Format and examples:
Examples of the use of CONE appear throughout Chapters 2, 5, and 6.
158
Calculation of p1 (x); U1 (x); and p2;0;HL (x) with N_ =
6 0:
In CONE, these calculations are exactly the same as given above for DUCT, except that area
A is a function of x in the equations for CONE.
_
Calculation of Tm and H:
Tm is not a¤ected by CONE. (But see also STKCONE in Section 10.6.1, which allows nonzero
dTm =dx:)
The exiting energy ‡ow is computed as H_ 2;k;out = H_ 2;k;in in insulated mode, described in
Section 11.1.2. This essentially assumes that the side walls are thermally insulated, so the
heat generated by acoustic power dissipation is deposited in an adjacent heat exchanger. If
several CONEs (and/or DUCTs) are strung together in insulated mode, the power dissipated in
all of them should show up in a nearby heat exchanger.
The exiting energy ‡ow is computed as H_ 2;k;out = E_ out in anchored mode, described in
Section 11.1.2. This mode essentially assumes that the cone wall is thermally anchored,
e.g., by a water bath, so heat generated by acoustic-power dissipation is carried away exter-
nally. Thermoacoustic heat transport along the perimeter, which in fact contributes a small
di¤erence between H_ 2;k and E_ in CONEs, is neglected.
Input variables:
Area (m2 ) The surface area S inside the COMPLIANCE.
Solid The solid material on the inside surface of the COMPLIANCE. (This is used only in
the calculation of s ; which seldom di¤ers signi…cantly from its value for a solid with
in…nite conductivity and speci…c heat in COMPLIANCEs.)
Potential targets:
None.
Master–slave links:
The surface area of a compliance can be slaved to its volume, so the shape remains the same
when the volume is changed.
159
(For an introduction to master–slave links, and a complete listing of link options for all
segments, see Section 2.7.2.)
10.2.2. SURFACE
General description:
A SURFACE is a surface area with thermal-hysteresis dissipation. It always absorbs acoustic
power. It has no volume. It is often used at the end of a DUCT, before a HARDEND. (See
Fig. 10.1 for two other typical uses for SURFACE.)
Input variables:
Area (m2 ) The surface area S exposed to oscillating pressure in the SURFACE segment, not
accounted for in other segments nearby.
Solid The solid material on the exposed surface. (This is used only in the calculation of s ;
which seldom di¤ers signi…cantly from its value for a solid with in…nite conductivity
and speci…c heat on SURFACEs.)
160
Potential targets; Master–slave links:
None.
Examples of SURFACE also occur throughout Chapter 2 and Sections 5.2 and 5.5.
Calculations:
A SURFACE does not a¤ect temperature or pressure amplitude. Volume ‡ow changes accord-
ing to
!p1 1
U1;out = U1;in 2
S : (10.35)
ma 1 + s 2
(This is the same calculation as would occur in a COMPLIANCE with zero volume.) The
in‡uence of the solid on thermal hysteresis is calculated in boundary-layer approximation:
1=2
k m cp
s = : (10.36)
ks s c s
The exiting energy ‡ow is computed as H_ 2;k,out = H_ 2;k,in in insulated mode. This essen-
tially assumes that the component is thermally insulated, so the heat generated by acoustic
power dissipation must somehow be deposited elsewhere, either upstream or downstream,
such as in an adjacent heat exchanger (or ‡ow out through a BEGIN or **END segment).
The exiting energy ‡ow is computed as H_ 2;k,out = E_ out in anchored mode. Thus, in a
series of such anchored segments, heat generated by acoustic power dissipation is carried
away locally.
SURFACE behaves the same whether N_ = 0 or not. Since Tm is unchanged in SURFACE,
HN_ ,out = H_ N_ ,in .
_
10.2.3. IMPEDANCE
General description:
An IMPEDANCE is a lumped series complex impedance.
Input variables:
Re(Z) (Pa s/m3 ) The real part of the series acoustic impedance Z. If the impedance has
resistance, this number is positive.
Im(Z) (Pa s/m3 ) The imaginary part of the series acoustic impedance Z. If the impedance
has inertance L; this number is equal to !L:
161
Potential targets:
None.
Master–slave links:
In an IMPEDANCE, Im[Z] can be slaved to Re[Z] so jZj remains constant when Re[Z] is
changed. Note, however, that this becomes impossible if Re[Z] is changed to a value that is
too large.
Alternatively, Im[Z] can be slaved to Re[Z] so phase(Z) remains constant when Re[Z] is
changed.
(For an introduction to master–slave links and a complete listing of linking options for
all segments, see Section 2.7.2.)
Linked to one or more RPN calculations, IMPEDANCE gives the user the freedom to cre-
ate nonlinear impedances. The following example shows an impedance with its resistance
proportional to the square of the ‡ow and its inertance proportional to the ‡ow:
!-------------2
RPN real part of impedance
12.0
U1 mag sqrd inp *
!-------------3
RPN imaginary part of impedance
3.0E-4
U1 mag inp *
!-------------4
IMPEDANCE
sameas 2A
sameas 3A
Calculations:
At an IMPEDANCE, volume ‡ow rate and temperature are unchanged. Pressure changes
according to
p1;out = p1;in ZU1 : (10.37)
The exiting energy ‡ow is computed as H_ 2;k;out = H_ 2;k;in in insulated mode, described in
Section 11.1.2. This assumes that the component is thermally insulated, so the heat gener-
ated by acoustic power dissipation must somehow be deposited elsewhere, either upstream
or downstream, such as in a nearby heat exchanger (or ‡ow out through a BEGIN or **END
segment).
The exiting energy ‡ow is computed as H_ 2;k;out = E_ out for anchored mode, described in
Section 11.1.2. This essentially assumes that the component is thermally anchored, such as
by a water bath, so heat generated by acoustic power dissipation is carried away locally.
162
If N_ 6= 0 in IMPEDANCE, the steady ‡ow and the oscillating ‡ow are assumed to be
independent, simply superimposed, as if the ‡ows are laminar. Thus, the di¤erence in p1
across an IMPEDANCE is una¤ected by the presence of the steady ‡ow, and the change in
p2;0;HL is given by
p2;0;HL;out p2;0;HL;in = Re[Z]N_ 2;0 m= m : (10.38)
Since an IMPEDANCE has no de…ned area, p2;0;rev is calculated based on the area of the most
recent upstream segment having area.
10.2.4. MINOR
General description:
Segment type MINOR accounts conveniently for minor-loss e¤ects. The basic equation for a
minor loss for steady ‡ow [56, 52] is simply
1
p= K u2 ; (10.39)
2
where K is the minor-loss coe¢ cient and u is the velocity. DeltaEC’s treatment of minor
loss in MINOR is based on this equation.
Equations (10.40)–(10.47) were derived using the quasi-steady approximation, in which
it is assumed that Eq. (10.39) is obeyed at each instant of time as u varies in time, with K
independent of time except for its dependence on the direction of ‡ow. This approximation
is very good when j 1 j rh : For j 1 j rh or j 1 j rh ; actual pressure drops may di¤er
signi…cantly from values calculated in MINOR.
Since DeltaEC uses U1 to characterize ‡ow, MINOR has an area for converting U1 to
velocity. Because DeltaEC treats oscillating ‡ow, MINOR has two inputs for Ks: K+ for the
minor-loss coe¢ cient in the +x direction, and K for the minor-loss coe¢ cient in the x
direction. The coe¢ cient in the N_ = 0 equations below di¤ers from the 1=2 in Eq. (10.39)
because of the sinusoidal time dependence of the velocity.
As shown in the second example below, the gas diodes in Ref. [42] can be modeled as a
MINOR and a CONE in series.
We often obtain K+ and K from Ref. [52], even though that reference is for steady ‡ow.
This approximation should be fairly accurate when the Reynolds number is large and j 1 j is
much greater than any dimensions of the component.
Input variables:
Area (m2 ) The area A used in the equations below. This is the area on which the minor-
loss coe¢ cients are based. Conventionally, this is usually the smallest area for area-
changing geometries such as di¤users, but it is sometimes the nominal pipe area for
valves and …ttings.
163
Figure 10.1: Examples of the use of MINOR. (a) Ori…ce plate in a channel. (b) Conical gas diode.
Potential targets:
None.
Master–slave links:
None. (Use sameas if you want to keep K+ and K equal to each other.)
164
This is an example of a conical gas diode, such as is described in Ref. [42] and illustrated
in Fig. 10.1(b).
!--------------------------------- 4 ---------------------------------
DUCT quarter-wave duct from engine to diode
7.0000E-04 a Area m^2 S= -2 7705.2 A |p| Pa
9.3748E-02 b Perim m Fn( 4a) -161.41 B Ph(p) deg
4.6000 c Length m 5.4080E-02 C |U| m^3/s
9.0500E-02 d Srough -87.804 D Ph(U) deg
354.15 E Htot W
ideal Solid type 58.812 F Edot W
!--------------------------------- 5 ---------------------------------
RPN surface area of adapter plate
0.0000 a G or T (t) 3.0000E-04 A ChngeMe
4a 8a -
!--------------------------------- 6 ---------------------------------
SURFACE surface area of adapter plate
sameas 5A a Area m^2 7705.2 A |p| Pa
-161.41 B Ph(p) deg
5.4080E-02 C |U| m^3/s
-87.804 D Ph(U) deg
354.15 E Htot W
ideal Solid type 58.812 F Edot W
!--------------------------------- 7 ---------------------------------
RPN calculate Borda-Carnot minor loss here
0.0000 a G or T (t) 0.1837 A ChngeMe
1 8a 4a / - sqrd
!--------------------------------- 8 ---------------------------------
MINOR asymmetric flow at gas diode
4.0000E-04 a Area m^2 7402.7 A |p| Pa
5.0000E-02 b K+ -174.70 B Ph(p) deg
sameas 7A c K- 5.4080E-02 C |U| m^3/s
-87.804 D Ph(U) deg
354.15 E Htot W
10.829 F Edot W
!--------------------------------- 9 ---------------------------------
CONE expansion to plenum
sameas 8a a AreaI m^2 S= -9 1.7471E+04 A |p| Pa
7.0899E-02 b PerimI m Fn( 9b) -177.49 B Ph(p) deg
0.4000 c Length m 4.9949E-02 C |U| m^3/s
1.5000E-02 d AreaF m^2 S= -9 -87.809 D Ph(U) deg
0.4342 e PerimF m Fn( 9d) 354.15 E Htot W
5.0000E-04 f Srough 2.4219 F Edot W
ideal Solid type
2 mjU1 j
p1,out p1,in = (K + K+ ) U1 ; (10.40)
3 A2
m jU1 j2
p2;0;HL;out p2;0;HL;in = (K K+ ) . (10.41)
8A2
165
Pressure calculations for N_ =
6 0:
For N_ =
6 0; changes in p1 across segment type MINOR are given by
m jU1 j
p1,out p1,in = "K U1 for " 1; (10.42)
A2
"2 p 3 1 K K+ 3 "
= 1+ 1 "2 + " sin " (K + K+ )
2 2 K + K+ 4
2 m jU1 j
U1 for j"j 1; (10.43)
3 A2
jU1 j
= "K+ m 2 U1 for " 1; (10.44)
A
m2 N_ 2;0
2
1 jp1 j cos
p2;0;HL;out p2;0;HL;in = K 1+ 2
+ for "2;0 1; (10.45)
2 m A2 2"2;0 pm "2;0
m jU1 j2 jp1 j
= (K K+ ) 1 + 2"22;0 + 2"2;0 cos
8A2 pm
2 jp1 j
(K+ + K ) 1 + 2"22;0 + 2"2;0
cos sin 1 "2;0
pm
4 jp1 j q
2
+ 3"2;0 + 1 + "2;0 =2 cos 1 "22;0
3 pm
for j"2;0 j 1; (10.46)
2 _2
m N2;0 1 jp1 j cos
= K+ 2
1+ 2 + for "2;0 1; (10.47)
2 mA 2"2;0 pm "2;0
where "2;0 = mN_ 2;0 = m jU1 j and is the phase angle between p1 and U1 . Derivations of these
equations neglecting the terms in p1 are given in Ref. [42]; and derivations including the p1
terms appear in Chapter 14.
Other calculations
166
Figure 10.2: Enclosed *EDUCER or *ESPEAKER.
167
L (henry) The electrical inductance L of the coil when it is prevented from moving.
K (N/m) The spring constant K of the transducer. This should include zero-current mag-
netic forces (present in moving-magnet, stationary-iron transducers) and forces from
mechanical springs, but not the gas-pressure forces on the front and back of area A.
jVj or jIj (volts or amps) In VESPEAKER, the amplitude of the oscillating voltage, jV1 j : In
IESPEAKER, the amplitude of the oscillating current, jI1 j :
ph(V) or ph(I) (degrees) In VESPEAKER, the phase of the oscillating voltage V1 : Similarly,
in IESPEAKER, the phase of the oscillating current I1 :
Solid The solid material on the leading and trailing surfaces A: (This is used only in the
calculation of s ; which seldom di¤ers signi…cantly from its value for a solid with in…nite
conductivity and in…nite speci…c heat in *SPEAKERs.)
Im(Ze) (ohm) The imaginary part of the transducer’s electrical impedance Ze when it is
prevented from moving.
Re(T1) (V s/m3 ) The real part of the complex transduction coe¢ cient in the equations
below.
Im(T1) (V s/m3 ) The imaginary part of the complex transduction coe¢ cient in the
equations below.
0
Re(T2) (Pa/A) The real part of the complex transduction coe¢ cient in the equations
below.
0
Im(T2) (Pa/A) The imaginary part of the complex transduction coe¢ cient in the equa-
tions below.
Re(Zm) (Pa s/m3 ) The real part of the mechanical impedance Zm of the transducer.
Im(Zm) (Pa s/m3 ) The imaginary part of the mechanical impedance Zm of the transducer.
168
jVj or jIj (volts or amps) In VEDUCER, the amplitude of the oscillating voltage V1 : In IEDUCER,
the amplitude of the oscillating current I1 :
ph(V) or ph(I) (degrees) In VEDUCER, the phase of the oscillating voltage V1 : In IEDUCER,
the phase of the oscillating current I1 :
Potential targets:
None.
Master–slave links:
0
None. The use of sameas is sometimes convenient in *EDUCERs, to keep = :
IEDUCER: same as VEDUCER, except that current appears in lines i and j instead of voltage.
VESPEAKER
6.000E-04 a Area m^2
6.00 b R ohms
.000 c L H
8.00 d BLProd T-m
5.000E-03 e M kg
1.000E+04 f K N/m
.000 g Rm N-s/m
62.0 h Vin V
-37.2 i Ph(Vin) deg
kapton Solid type
IESPEAKER: same as VESPEAKER, except that current appears in line h and i instead of
voltage.
Other examples of series *ESPEAKERs are in Sections 5.2 and 5.8.2.
Calculations:
An Enclosed transducer IEDUCER, VEDUCER, IESPEAKER, VESPEAKER is an object attached in
series with other segments, as shown in Fig. 10.2. Volume ‡ow rate U1 is nearly unchanged
(except for surface thermal-hysteresis losses— see below), but pressure is changed by the
force exerted by the transducer, obeying the complex canonical equations
V1 = Ze I1 U1 ; (10.48)
0
p1;out p1;in = I1 Zm U1 : (10.49)
If current I1 is speci…ed, then use IESPEAKER or IEDUCER, and …nd p1;out = p1;in +
0
I1 Zm U1 and V1 = Ze I1 U1 .
169
If voltage V1 is speci…ed, then use VESPEAKER or VEDUCER, and …nd I1 = (V1 + U1 )=Ze
and p1;out = p1;in + 0 I1 Zm U1 .
If a …xed electrical load impedance Zext is attached to the transducer, it can be modeled
using an IMPEDANCE segment instead of a *ESPEAKER or *EDUCER segment, with Zimp =
0
Zm =(Ze + Zext ) ; or use one of these Enclosed transducer segments with one or
more RPN targets to maintain the impedance, as shown in the example in Section 4.7.
For *EDUCERs, the complex electrical impedance Ze = Re[Ze ] + iIm[Ze ] and the complex
mechanical impedance Zm = Re[Zm ] + iIm[Zm ] are given explicitly by the user.
0
In the case of *ESPEAKERs, Ze = Re + i!L ; = = Bl=A ; Zm = Rm =A2 +
2
i(!M K=!)=A . Thermal surface losses are computed for area A using the same approach
as for a SURFACE, but Enclosed speakers have area A exposed to p1;in and area A exposed
to p1;out , because typically both sides of the speaker experience oscillatory pressure. This
thermal hysteresis on those two surfaces manifests itself as a small change in volume ‡ow
rate, according to
! 1
U1;out = U1;in 2
A (p1;in + p1;out ) ; (10.50)
ma 1 + s 2
and U1 in Eqs. (10.48) and (10.49) is actually the value after the …rst thermal-hysteresis
loss:
! 1
U1 = U1;in 2
Ap1;in : (10.51)
ma 1 + s 2
The Px outputs for these segments give the complex pressure di¤erence across the trans-
ducer.
None of these segments a¤ects Tm .
The exiting energy ‡ow is computed as H_ 2;k;out = H_ 2;k;in + W _ elec in insulated mode,
described in Section 11.1.2. This assumes that the component is thermally insulated, so the
electrical energy added at the transducer must ‡ow out of the segment, either upstream or
downstream. The exiting energy ‡ow is computed as H_ 2;k;out = E_ out for thermally anchored
mode, described in Section 11.1.2.
The WorkIn output variable displays the electrical power Re[V1 I~1 ]/2 that must be sup-
plied to the transducer. If negative, this is electrical power produced by the transducer.
The Ph(V/I) variable is especially useful for constructing a target related to the electrical
load impedance attached to a *SPEAKER that is used as an alternator; for example, Ph(V/I) =
180 means that the load impedance is purely resistive. See Fig. 4.1a for a more complicated
example with both resistance and capacitance.
It is hard for us to anticipate what a user might intend if nonzero N_ were to ‡ow through
a series transducer, so these segments behave the same whether N_ = 0 or not.
170
Figure 10.3: BRANCH, OPNBRANCH, or PISTBRANCH.
Im(Z) (Pa s/m3 ) The imaginary part of the branch impedance Zbr :
HtotBr (W) How much power H_ ‡ows out of the trunk into the branch.
Im(Z)/k (Pa s/m2 ) The imaginary part of the branch impedance Zbr ; divided by the wave
number k = !=a:
171
HtotBr (W) How much power H_ tot ‡ows out of the trunk into the branch.
Potential targets:
None.
Master–slave links:
The imaginary part of a BRANCH impedance can be slaved to the real part, to hold the
magnitude constant or to hold the phase constant when the real part is changed.
No master–slave options are available for OPNBRANCH and PISTBRANCH.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
OPNBRANCH
0.05 Pa-s/m Re(Z)/k^2
0.20 Pa-s/m2 Im(Z)/k
12.5 W HtotBr
Calculations:
A BRANCH is a side branch with complex impedance Zbr . Pressure is unchanged in the trunk,
but volume ‡ow rate in the trunk changes according to U1;out = U1;in p1 =Zbr .
For an OPNBRANCH, the numbers in lines a and b are multiplied by (!=a)2 and !=a;
respectively, to obtain the impedance Zbr . These dependences on frequency [12] are correct
in the low-frequency limit, !r=a 1. As for BRANCH, pressure is unchanged in the trunk,
but volume ‡ow rate in the trunk changes according to U1;out = U1;in p1 =Zbr .
Similarly, for a ‡anged piston of radius r where the wave number is k = !=a locally, the
PISTBRANCH radiation impedance is given by [59]
0 8 p 1
>
< 4= 8= sin(2kr 3 =4)
aB 2J1 (2kr) 2kr
+ (2kr)3=2
if 2kr > 2:68; C
Zbr = m @1 +i A (10.52)
A 2kr >
: (4= )2kr 1 (2kr)2
3 15
otherwise.
Note the implicit assumption that the outside and inside gases have the same density and
sound speed.
172
Figure 10.4: The side-branched transducers: IDUCER, ISPEAKER, VDUCER, or VSPEAKER.
Thus, the usual method for modeling a ‡anged piston or open tube radiating into 2
solid angle is DUCT, PISTBRANCH, HARDEND, with the HARDEND’s impedance targets targeted
to zero to enforce U1;out = 0, ensuring that U1;in = U1;br = p1 =Zbr at the PISTBRANCH.
Output EdotBr gives the acoustic power E_ br ‡owing out of the branch.
None of these segments a¤ects Tm . In BRANCH and OPNBRANCH, the trunk exiting energy
‡ow is computed as H_ 2;k;out = H_ 2;k;in H_ br in insulated mode, described in Section 11.1.2.
The trunk exiting energy ‡ow is computed as H_ 2;k;out = H_ 2;k;in E_ br in PISTBRANCH in
insulated mode. In thermally anchored mode, described in Section 11.1.2, H_ 2;k;out = E_ out in
all three of these segments.
The behavior of these segments is independent of N_ .
173
steady ‡ow passing by and what fraction is added to the thermoacoustic wave. In a VX*
heat-exchanger segment, FQ_ N_ ; (FracQN) determines these fractions. With a side-branch
transducer, an RPN segment can be used, with user-speci…ed assignment of these fractions
using =Tm and =H2k.)
See also Chapter 14 for a discussion of the saturation of magnetic ‡ux in *SPEAKERs.
See also the series transducers described in Section 10.3.1 above.
174
0
Re(T2) (Pa/A) The real part of the complex transduction coe¢ cient in the equations
below.
0
Im(T2) (Pa/A) The imaginary part of the complex transduction coe¢ cient in the equa-
tions below.
Re(Zm) (Pa s/m3 ) The real part of the mechanical impedance Zm of the transducer.
Im(Zm) (Pa s/m3 ) The imaginary part of the mechanical impedance Zm of the transducer.
jVj or jIj (volts or amps) In VDUCER, the amplitude of the oscillating voltage V1 ; which is
taken to be real. In IDUCER, the amplitude of the oscillating current I1 ; which is taken
to be real.
Potential targets:
None.
Master–slave links:
0
None. The use of sameas is sometimes convenient in *DUCERs, to keep = :
VSPEAKER
6.000E-04 a Area m^2
6.00 b R ohms
.000 c L H
8.00 d BLProd T-m
5.000E-03 e M kg
.000 f K N/m
.000 g Rm N-s/m
-22.5 h AplVol V
ideal Solid type
Calculations:
A side-branch transducer IDUCER, VDUCER, ISPEAKER, VSPEAKER is an object attached as
shown Fig. 10.4 like a branch impedance, but obeying the complex equations V1 = Ze I1 + Ux ,
p1 = 0 I1 + Zm Ux . Pressure is unchanged, but volume ‡ow rate changes according to
U1;out = U1;in Ux . The volume ‡ow rate of the transducer, Ux ; is displayed in the output
column.
If current I1 is speci…ed (note! its phase must be real), then use ISPEAKER or IDUCER,
0
and …nd Ux = (p1 I1 )=Zm and V1 = Ze I1 + Ux .
175
If voltage V1 is speci…ed (note! its phase must be real), then use VSPEAKER or VDUCER,
0
and …nd I1 = (Zm V1 p1 )=(Ze Zm ) and Ux = (V1 Ze I1 )= .
If a …xed electrical load impedance Zext is attached to the transducer, it can be modeled
using a BRANCH segment instead of a *SPEAKER or *DUCER segment, with Zbr = p1 =Ux =
0
Zm =(Ze + Zext ) .
0
In the case of speakers, Ze = Re + i!L ; = = Bl=S ; Zm = Rm =S 2 + i(!M
K=!)=S 2 . Side-branch speakers are assumed to have area S exposed to the oscillating
pressure. Thermal-hysteresis surface losses are computed for that area S using the same
formula as for a SURFACE, Eqs. (10.35) and (10.36). As described in Section 10.2.2, thermal-
hysteresis losses manifest themselves as a small change in volume ‡ow rate.
Note that IDUCER and ISPEAKER will crash if Zm is zero, so it is best to use VDUCER or
VSPEAKER for mechanically ideal or lossless resonant transducers.
None of these segments a¤ects Tm .
The trunk’s exiting energy ‡ow is computed as H_ 2;k;out = H_ 2;k;in + W _ elec in insulated
mode, described in Section 11.1.2. This assumes that the component is thermally insulated,
so the electrical energy added at the transducer must ‡ow out of the segment along the
trunk, either upstream or downstream. The trunk’s exiting energy ‡ow is computed as
H_ 2;k;out = E_ out for thermally anchored mode, described in Section 11.1.2.
The WorkIn output variable displays the electrical power Re[V1 I~1 ]/2 that must be sup-
plied to the transducer. If negative, this is electrical power produced by the transducer.
The Ph(V/I) variable is especially useful for constructing a target related to the electrical
load impedance attached to a *SPEAKER that is used as an alternator; for example, Ph(V/I)
= 180 means that the load impedance is purely resistive. See Fig. 4.1 for a more complicated
example with both resistance and capacitance.
These segments behave the same whether N_ = 0 or not.
176
Figure 10.5: Common smooth-stack geometries, with the x axis perpendicular to the page. (a)
STKSLAB. (b) STKRECT. (c) STKCIRC.
GasA/A The areal porosity = Agas =A = y0 = (y0 + l) ; where Agas is the thermoacoustic
area. The solid fraction 1 is used for the calculation of axial thermal conduction
by the solid.
y0 (m) Half the plate spacing is y0 : In other words, 2y0 is the gap available to the gas— not
the center-to-center spacing of the plates. See Fig. 10.5(a).
Lplate (m) Half the thickness of the solid plate is l: Thus, 2l is the plate thickness. See
Fig. 10.5(a). This variable is used in the calculation of s to account for any imper-
177
fection in the imposition of the isothermal boundary condition imposed on the gas by
the solid.
GasA/A The areal porosity = Agas =A: The solid fraction 1 is used for the cal-
culation of axial thermal conduction by the solid. Often, the user intends =
2
ab= (ab + al + bl + l ) ; this can be enforced with a master–slave link (see below).
a (m) Half of pore width. Thus, 2a is the small dimension of the rectangular pore available
to the gas. See Fig. 10.5(b).
Lplate (m) Half the thickness of the solid plate: Thus, 2l is the plate thickness. This variable
is used for the calculation of s to account for any imperfection in the imposition of
the isothermal boundary condition imposed on the gas by the solid. See Fig. 10.5(b).
b (m) Half of pore breadth 2b: Thus, 2a 2b is the cross-sectional area available to the gas
in a single pore. See Fig. 10.5(b).
GasA/A The areal porosity = Agas =A: The solid fraction 1 is used for the calculation
of axial thermal conduction by the solid.
radius (m) The radius r0 of the circular pore. Thus, r02 is the cross-sectional area available
to the gas in a single pore. (When we use STKCIRC to model hexagonal p “honeycomb”
p
pores, we choose r0 so the pore has the correct cross-sectional area: r0 = 4 3= 2 b '
0:525038b; where b is the ‡at-to-‡at distance across the hexagonal pore.)
Lplate (m) Half the thickness of the solid sheet between pores: Thus, 2l is the plate thick-
ness. This variable is used in the calculation of s to account for any imperfection in
the imposition of the isothermal boundary condition imposed on the gas by the solid.
178
Input variables, STKPIN:
Area (m2 ) The total cross-sectional area A = Agas + Asolid ; but usually not intended to
include the cross-sectional area of the pressure-vessel wall.
2y0 (m) 2y0 is the center-to-center, nearest-neighbor pin-to-pin distance in the hexagonal
lattice of pins. See Ref. [61].
Potential targets:
None.
Master–slave links:
The porosity (= GasA/A) of some of these segments can be slaved to other parameters:
In STKSLAB, the porosity can be slaved to other parameters through the equation
y0
=c ; (10.53)
y0 + l
y0 + l
c= : (10.54)
y0
If c = 1; this simply means that all of the cross section is full of gas and plate, with no
struts, support members, or pressure-vessel wall. Otherwise, c 6= 1 indicates that a fraction
of the area is occupied by such structure; the link keeps that fraction a constant.
In STKRECT, the porosity can be slaved to other parameters through the equation
ab
= ; (10.55)
(a + l)(b + l)
link maintains a constant number of holes, but with the porosity clipped at a maximum
value of 1.
The length of any of these segments can be slaved to the length of another segment to
keep the total length constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
179
Format and examples:
STKSLAB parallel-plate stack
sameas 1 Area
0.724 GasA/A
7.85e-2 m Length
1.8e-4 m y0 (half the plate spacing)
4.0e-5 m Lplate (half the plate thickness)
kapton Solid
Another example of STKSLAB is in Section 5.2. STKCIRC is used in Section 5.5, and
STKRECT is used in Section 6.2.
dp1 i! m
= U1 ; (10.56)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1+ p1 + U1 ; (10.57)
dx ma 1+ s (1 f )(1 )(1 + s ) dx
subject to the condition that the total energy ‡ow H_ 2;k is independent of x, which imposes
the following condition on Tm (x):
h i
_ 2;k 1 Re p1 U~1 1 Tm (f f~ )
dTm H 2 (1+ s )(1+ )(1 f~ )
= 2
h i : (10.58)
dx m cp jU1 j ~ (f f~ )(1+ s f =f )
2 Im f + Agas k Asolid ks
(1+ s )(1+ )
2!Ag a s (1 )j1 f j
For STKSLAB,
tanh[(1 + i)y0 = j ]
fj = ; for j = or ; (10.59)
(1 + i)y0 = j
1=2
k m cp tanh[(1 + i)y0 = ]
s = : (10.60)
ks s c s tanh[(1 + i)`= s ]
180
For STKRECT,
64 X 1
fj = 1 4
; for j = or ; (10.61)
m;n m2 n2 Y mn ( j )
odd
1=2
k m cp f (1 + i)ab= (a + b)
s = ; (10.62)
ks s c s tanh[(1 + i)`= s ]
2 2
where Ymn ( ) = 1 i (b2 m2 + a2 n2 ): (10.63)
8a2 b2
For STKCIRC,
For STKPIN,
j 2ri Y1 [(i 1)ro = j ]J1 [(i 1)ri = j ] J1 [(i 1)ro = j ]Y1 [(i 1)ri = j ]
fj =
(i 1) ro2 ri2 Y1 [(i 1)ro = j ]J0 [(i 1)ri = j ] J1 [(i 1)ro = j ]Y0 [(i 1)ri = j ]
for j = or ; (10.66)
p
k m cp
1=2
J0 ( i!= s ri ) p r2 ri2
s = p f i!= o ; (10.67)
ks s c s J1 ( i!= s ri ) 2ri
p
where = k= m cp ; s = ks = s cs ; ri is pin radius, and ro2 = 3=2 (2yo )2 with 2yo being
the center-to-center distance between nearest-neighbor pins in the hexagonal array of pins.
To turn on steady-‡ow calculations, check the appropriate box in the Optional Parameters
dialog in the BEGIN segment. In the a¤ected STK** segments, the dependence of Tm on x
can be changed signi…cantly by the nonzero N_ :
The display of information is slightly changed. H2k is the “acoustic”-plus-longitudinal-
conduction part of the total energy ‡ux— not including the steady-‡ow part of the total
energy ‡ux, N_ mwm ; where wm is the enthalpy per unit mass. Htot is the sum of H2k and
N_ mwm :
Nonzero N_ changes the temperature pro…le in smooth stacks according to Refs. [38, 39,
13]: The temperature gradient in stacks is computed using constancy of total energy ‡ux
181
H_ tot . In other words, we solve this equation for dTm =dx :
2 0 13
1 Tm f fe
H_ tot = e1 @1
Re 4p1 U A5
2 (1 + s )(1 + ) 1 f e
" #
e )(1 + s f =f )
m cp dTm (f f
+ jU1 j Im fe +
2
2Agas !(1 ) j1 f j2 dx (1 + s )(1 + )
dTm
(Agas k + Asolid ks ) + N_ mwm (10.68)
dx
and DeltaEC uses H_ tot = constant to integrate with respect to x through a stack. The only
N_ 6= 0 feature in this equation is the …nal term, N_ mwm ; where wm is the mean enthalpy
per unit mass. The mean enthalpy is referenced to zero at the environment temperature T0 ,
which is 300 K by default and can be set to another value via =Tzero in an RPN segment. As
usual, the value of H_ tot is determined by conditions in segments other than the stack, such
as the value of N_ set in the BEGIN segment and the heat ‡ows in adjacent heat exchangers.
The stack energy equation, Eq. (10.68), is reasonable only if N_ is essentially of second
order, and hence only if N_ is formally negligible in the …rst-order momentum and energy
equations [38, 39, 13]. Hence, we have not implemented any Doppler e¤ect on dp1 and dU1
in stacks.
In the smooth stacks, dp2;0;HL =dx is calculated with laminar expressions based on N_ 2;0 :
182
General description of tortuous regenerators:
Use STKSCREEN for stacked-screen regenerators (see Ref. [62]). Use STKPOWERLW for any
other regenerator for which friction factor and heat-transfer coe¢ cients follow power laws
in Reynolds number. The algorithms for these segments assume reasonably good internal
thermal contact, so do not trust these segments for rh k and certainly not for rh > k :
DeltaEC has no capability to model the physics of oscillating ‡ow through tortuous porous
media except in the small-pore, nearly isothermal limit.
VolPor The volumetric porosity : This is the ratio of the volume of gas in the regenerator
to the total volume of the regenerator. For plain square-weave screen with m wires
per unit length and wire diameter dwire ; we use
mdwire
'1 ; (10.75)
4
as recommended by Organ [27], or we weigh the regenerator and use its A, x, and
the known density of its wire material to calculate . [See also Eq. (5.14).]
rh (m) The hydraulic radius rh of the regenerator. This is the ratio of the gas volume to
the gas–solid contact area. For plain square-weave screen with wire diameter dwire ; we
use
rh ' dwire ; (10.76)
4 (1 )
as recommended by Organ [27].
ksFrac If the solid material in the regenerator were arranged in smooth columns par-
allel with the x axis, as in the smooth stacks described above in Section 10.5.1,
then the contribution of ordinary thermal conduction in the solid to H_ 2;k would be
(1 ) Aks dTm =dx: However, in STKSCREEN the solid conduction is reduced by the
random point contacts between wires. ksFrac is the reduction factor. Reference [28]
recommends a value of 0.1. We often use a larger value, to include the thermal con-
ductance of the pressure-vessel wall around the regenerator. [See Eq. (9.1).]
183
VolPor The volumetric porosity : This is the ratio of the volume of gas in the regenerator
to the total volume of the regenerator.
Length (m) The length x of the regenerator.
rh (m) The hydraulic radius rh of the regenerator. This is the ratio of the gas volume to
the gas–solid contact area.
ksFrac If the solid material in the regenerator were arranged in smooth columns par-
allel with the x axis, as in the smooth stacks described above in Section 10.5.1,
then the contribution of ordinary thermal conduction in the solid to H_ 2;k would be
(1 ) Aks dTm =dx: However, in porous media such as balls the solid conduction is
reduced by the random point contacts in the x direction between pieces of the solid.
ksFrac is the reduction factor. Reference [28] recommends a value of 0.1 for woven
screens.
f con The friction-factor parameter fcon in the equations below.
f exp The friction-factor parameter fexp in the equations below.
h con The heat-transfer parameter hcon in the equations below.
h exp The heat-transfer parameter hexp in the equations below.
Solid The solid with which the regenerator is made.
Potential targets:
None.
Master–slave links:
The length of either of these segments can be slaved to the length of another segment to
keep the sum of the two lengths constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
184
Other examples of the use of STKSCREEN are in Sections 5.8 and 6.3.
d hu1 i i! dTm
= 2
p1 + hu1 i +
dx ma dx
2i p
Tm s + (gc + e gv ) h 1 dTm s + (gc gv ) h
i! 2i
p1 2i
hu1 i ;
m cp 1 + s + (gc + e
Tg ) i! dx 1 +
v h s + (gc + e T gv ) h
(10.78)
( )
dTm s + + e2i p gv )
h (gc g 2H_ 2;k
= Re Tm +1 Tm p1 hu 1i
dx 1 + s + h (gc + e2i T gv ) A
m cp + h (gc gv )
s g 1
= Im 2i
hu1 i hu 1 i + 2 ks;e¤ +k ;
! 1 + s + h (gc + e T gv )
(10.79)
using
185
these integrals were performed once, o¤-line, to obtain simple functional …ts that DeltaEC
uses during computation.
STKPOWERLW segments are calculated in the same manner as STKSCREEN, with a few
exceptions. The steady-‡ow friction factor and heat transfer coe¢ cients are parameterized
by
fe x p
f = fcon NR ; (10.87)
2=3 he x p
NSt = hcon NR ; (10.88)
where NSt is the Stanton number and the Reynolds number NR is de…ned in the usual way
as
4U rh
NR = : (10.89)
A
[Note: Equation (10.87) is Fanning friction factor, the friction factor used by Kays and
London [63], so instantaneously dp=dx = (f =rh ) 21 u2 :] The pressure equation is replaced
by
dp1 1 f
= i! m hu1 i If 2 fcon NR;1 e x p hu1 i ; (10.90)
dx 8rh
where Z
2
If = sin3 fe x p (z)dz: (10.91)
0
In the volume-‡ow-rate and mean-temperature equations, these parameters are rede…ned for
the power-law stack:
Z =2
he x p 1 2
gc = NR;1 coshe x p 1 (z)dz; (10.92)
0
Z =2
he x p 1 2
gv = NR;1 cos 2z coshe x p 1 (z)dz; (10.93)
0
b( ) = hcon : (10.94)
The trigonometric integrals in Eqs. (10.92) and (10.93) are not evaluated by DeltaEC;
these integrals were performed once, o¤-line, to obtain simple functional …ts that DeltaEC
uses during computation.
m N_ 2;0 4rh
NR;2;0 = (10.98)
A
and "2;0 = mN_ 2;0 = m jU1 j : See Section 8.1.1 for explanation of the use of N_ 2;0 instead of N_
here.
In STKPOWERLW, the expression for head-loss gradient is the same as the pessimistic case in
turbulent DUCTs, but with fM and dfM =dNR given by the STKPOWERLW expression, Eq. (10.87),
and with jp1 j = pm replaced by jp1 j =pm to model isothermal oscillations.
187
The one-dimensional nature of DeltaEC calculations can be particularly unrealistic in
STKDUCT and STKCONE, because DeltaEC assumes that Tm (x) and Tsolid (x) are equal at all
x; despite the fact that these are likely to be very di¤erent if any type of streaming occurs.
PerimI (m) Perimeter I of the initial inside cross section described above.
AreaF (m2 ) Cross-sectional area Agas,F available to the gas at the …nal end of the STKCONE,
based on inside dimensions.
PerimF (m) Perimeter F of the …nal inside cross section described above.
f_wall For fwall < 1, wall thickness = fwall initial perimeter, and is independent of
x. For fwall > 1, wall thickness = local perimeter=fwall, so local wall cross-sectional
area = (local perimeter)2 =fwall.
Potential targets:
None.
Master–slave links:
In STKDUCT, the perimeter can be slaved to the area to maintain the cross-sectional shape
when the area is changed.
In STKDUCT, the perimeter and the area can also be slaved to the length, to maintain
both constant volume and constant cross-sectional shape when the length is changed.
188
In STKCONE, the initial perimeter can be slaved to the initial area, the …nal perimeter
can be slaved to the …nal area, or both perimeters can be slaved to both areas, so shapes
are maintained when areas are changed.
The length and both perimeters of a STKCONE can be slaved to its two areas, to keep the
cross-sectional shapes of the ends and the wall taper angle constant when either or both
areas are changed.
The length of either of these segments can be slaved to the length of another segment to
keep the sum of the two lengths constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
Calculations for N_ = 0:
In STKDUCT and STKCONE, pressure propagates according to Rott’s wave equation, written in
the form
dp1 i! m
= U1 ; (10.99)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1+ p1 + U1 ; (10.100)
dx ma 1+ s (1 f )(1 )(1 + s ) dx
subject to the condition that energy ‡ow H_ 2;k is independent of x, which imposes the fol-
lowing condition on Tm (x):
h i
_ 1 ~ Tm (f f~ )
dTm H2;k 2 Re p1 U1 1 (1+ )(1+ )(1 f~ )
s
= 2
h ~
i : (10.101)
dx m cp jU1 j ~
2 Im f +
(f f )(1+ s f =f )
Agas k Asolid ks
2!Ag a s (1 )j1 f j (1+ s )(1+ )
189
Calculation with N_ =
6 0:
STKDUCT treats nonzero N_ in the same way as STKCIRC. STKCONE is similar, with the addi-
tional variation of A with x: Remember to increase Nint if Tm (x) curves signi…cantly.
10.7.1. HX and TX
General description:
For a general description of all DeltaEC heat exchangers, see the initial paragraphs of the
heat-exchanger documentation, Section 10.7.
In HX the thermoacoustic working gas is between parallel plates; in TX it is inside cylin-
drical tubes. Heat Q_ is positive when it ‡ows from the solid into the thermoacoustic gas,
adding to H_ tot : The temperature di¤erence between the solid temperature Tsolid and the gas
temperature Tm is proportional to Q: _ The proportionality constant is not well validated,
either theoretically or experimentally; we …nd that it usually seems to be accurate within a
factor of 2.
HX and TX are assumed to be so short that the numerical integration from one end to
the other is performed with a single fourth-order Runge-Kutta step. (Most other integrated
segments, i.e., DUCT, CONE, MIX**, STK**, and VX**, are integrated with Nint fourth-order
Runge-Kutta steps.) See VX** in Section 10.7.3 below for longer tubular heat exchangers.
y0 (m) The plate spacing is 2y0 : In other words, 2y0 is the gap available to the gas— not
the center-to-center spacing of the plates.
HeatIn The heat Q_ added to the thermoacoustic gas in the heat exchanger. Positive Q_
indicates heat ‡ow from solid to gas.
190
Solid The solid material on the inside surface of the HX. This is used only in the calculation of
s ; which seldom di¤ers signi…cantly from its value for a solid with in…nite conductivity
and speci…c heat.
Potential targets:
TSolid If used as a target, DeltaEC’s shooting method compares Tsolid to this value.
Master–slave links:
In TX, the porosity can be slaved to other parameters via the equation = min(cr02 =A; 1);
with c calculated when the link is …rst enabled via c = A=r02 : This link maintains a constant
number of holes, up to a maximum porosity of 1.
The length of HX or TX can be slaved to the length of another segment to keep the sum
of the two lengths constant.
(For an introduction to master–slave links, and a complete list of linking options for all
segments, see Section 2.7.2.)
191
Calculations:
In HX and TX, wave propagation is calculated using
dp1 i! m
= U1 ; (10.104)
dx (1 f )Agas
dU1 i!Agas ( 1)f
= 2
1+ p1 : (10.105)
dx ma 1+ s
HX uses parallel plate geometry in computing f , f , and s:
tanh[(1 + i)y0 = j ]
fj = ; for j = or ; (10.106)
(1 + i)y0 = j
1=2
k m cp
s = : (10.107)
ks s c s
Similarly, TX uses cylindrical geometry in computing f , f , and s: For r0 = j < 25, f and
f are calculated using complex Bessel functions
2J1 [(i 1)r0 = j ]
fj = for j = or : (10.108)
(i 1)(r0 = j )J0 [(i 1)r0 = j ]
For intermediate values, linear interpolation is used to make a smooth match between the
two regimes. In both cases, s is calculated using
1=2
k m cp
s = : (10.110)
ks s c s
In TX, the radius r0 is that of one circular pore, so for a heat exchanger comprised of
N circular pores, the total cross-sectional area available to the working gas is N r02 =
(Area)(GasA=A).
HX and TX have no turbulence algorithms.
For N_ = 0; temperature and power change according to
dTm
= 0; (10.111)
dx
dH_ tot dH_ 2;k Q_
= = ; (10.112)
dx dx x
Q_
Tsolid = Tm + ; (10.113)
h Agas xe¤=rh
where x is the heat exchanger length and Q_ is the heat added to the heat exchanger.
Positive heat Q_ (parameter e) ‡ows into the apparatus. The heat ‡ow Q_ is an input for
192
each pass of DeltaEC, and it is often used as a guess for the shooting method described
in Section 2.5. The heat-transfer coe¢ cient h is estimated with
k
h= ; (10.114)
ye¤
and
with hydraulic radius rh equal to y0 for HX and equal to r0 =2 for TX. These expressions may
be quite inaccurate, but we believe they are more useful than nothing. A little experimental
evidence for Tsolid Tm is presented in Ref. [23]. A user who mistrusts these expressions
can use the gas temperature (available through RPN, and as an output in the adjacent stack
segment) instead of the solid temperature for plotting or targeting, and/or use RPNs to
construct custom expressions instead of Eqs. (10.113)–(10.116).
For N_ 6= 0; the wave and energy calculations are unchanged, so all of the heat goes into
H_ 2;k and none into the steady ‡ow. This seems …ne for weak steady ‡ows, such as weak
Gedeon streaming when a TASHE is operated slightly away from the conditions at which
its gas diode enforces N_ = 0: For strong steady ‡ow, the user can override the built-in
calculations with an RPN such as
13 RPN Fix Tm. Inp = frac of heat that should go to mean flow
0.25
H2k inp 12e * - =H2k ; inp 12e * Ndot / m / cp / Tm + =Tm
following the heat exchanger, where 12e is the heat in the upstream heat exchanger and 13a
is the steady-‡ow heat fraction. That fraction is often guessed, to give DeltaEC enough
‡exibility to meet nearby temperature and energy boundary conditions. Alternatively, the
1
fraction can also be set equal to a simple expression like 1 + m jU1 j =jN_ jm in an RPN,
supplying an experimentally unsupported but plausible estimate of how power divides itself
between the steady ‡ow and the acoustic ‡ow.
For N_ 6= 0; the gradients in p2;0;HL are calculated using Eq. (10.70) for HX and Eq. (10.69)
for TX.
10.7.2. SX and PX
General description:
For a general description of all DeltaEC heat exchangers, see the initial paragraphs of the
heat-exchanger documentation, Section 10.7.
In SX, the thermoacoustic working gas ‡ows through a pile of woven screens. The PX
segment allows the user to specify heat-exchanger parameters for which the steady-‡ow
friction factor and heat transfer coe¢ cient are known power laws in Reynolds number.
The SX calculations assume fairly good thermal contact between gas and solid, so these
calculations are valid only for hydraulic radius smaller than thermal and viscous penetration
depths. There is no warning if this condition is not met.
193
Heat Q_ is positive when it ‡ows from the solid into the thermoacoustic gas, adding to
H_ tot : In SX, typically the screen material has high thermal conductivity and the area is small,
because typically the heat must ‡ow radially through the screen wires, from the pressure
vessel around the screens to the thermoacoustic gas.
The temperature di¤erence between the solid temperature Tsolid and the gas tempera-
ture Tm is proportional to Q: _ The proportionality constant is not well supported, either
theoretically or experimentally; we …nd that it seems to be accurate within a factor of 2.
SX and PX are assumed to be so short that the numerical integration from one end to
the other is performed with a single fourth-order Runge-Kutta step. (Most other integrated
segments, i.e., DUCT, CONE, MIX**, STK**, and VX**, are integrated with Nint fourth-order
Runge-Kutta steps.)
VolPor The volumetric porosity : This is the ratio of the volume of gas in the SX to the
total volume of the heat exchanger. For plain square-weave screen with m wires per
unit length and wire diameter dwire ; we sometimes use
mdwire
'1 ; (10.117)
4
as recommended by Organ [27], or we weigh the pile of screens, measure its overall
dimensions A and x; and use the known density of its wire material to calculate .
rh (m) The hydraulic radius rh of the SX. This is the ratio of the gas volume to the gas–solid
contact area. For plain square-weave screen with wire diameter dwire ; we sometimes
use
rh ' dwire ; (10.118)
4 (1 )
as recommended by Organ [27].
HeatIn The heat Q_ added to the thermoacoustic gas in the heat exchanger. Positive Q_
indicates heat ‡ow from solid to gas.
Solid The solid material of the wires. This is used only in the calculation of s ; which
seldom di¤ers signi…cantly from its value for a solid with in…nite conductivity and
speci…c heat in SXs.
VolPor The volumetric porosity : This is the ratio of the volume of gas in the heat
exchanger to the total volume of the heat exchanger.
194
Length (m) The length x of the heat exchanger.
rh (m) The hydraulic radius rh of the heat exchanger. This is the ratio of the gas volume
to the gas–solid contact area.
HeatIn The heat Q_ added to the thermoacoustic gas in the heat exchanger. Positive Q_
indicates heat ‡ow from solid to gas.
Solid The solid with which the PX is made. This is used only in the calculation of s ; which
seldom di¤ers signi…cantly from its value for a solid with in…nite conductivity and
speci…c heat in heat exchangers.
Potential targets:
TSolid If used as a target, DeltaEC’s shooting method compares Tsolid to this value.
Master–slave links:
The length of either of these segments can be slaved to the length of another segment to
keep the sum of the two lengths constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
195
Calculations:
In SX, dp1 =dx and d hu1 i =dx are calculated in the same way as in STKSCREEN, as described
in [62]:
dp1 (1 )2 c1 ( ) c2 ( )NR;1
= i! m 1+ hu1 i + hu1 i ; (10.119)
dx 2(2 1) rh2 8 3
using
where NSt = NN u = NR is the Stanton number, and the Reynolds number NR is de…ned in
the usual way as
4U rh m 4 hui rh m
NR = = : (10.130)
A
Note: this is Fanning friction factor, the friction factor used by Kays and London, so instan-
taneously
dp f 1
= hui2 = 2 f NR hui : (10.131)
dx rh 2 8rh
196
The complex pressure gradient is computed using
dp1
= i! m hu1 i If fcon jNR;1 j1 fe x p
hu1 i (10.132)
dx 8rh2
where Z
2
If = sin3 fe x p
!t d(!t): (10.133)
0
Thermal-relaxation e¤ects due to oscillating pressure are computed using the same equations
for d hu1 i =dx as for screen heat exchangers but with
Z =2
he x p 1 2
gc = jNR;1 j coshe x p 1
!t d(!t); (10.134)
0
Z =2
he x p 1 2
gv = jNR;1 j cos 2!t coshe x p 1
!t d(!t); (10.135)
0
b( ) = hcon : (10.136)
The solid temperature is computed relative to the gas mean temperature using
Q_ rh2
T = Ih ; (10.137)
k hcon jNR;1 j1 he x p Axe¤
where xe¤ = minf2 jhu1 ij =!; xg ; NR;1 is the Reynolds number amplitude (based on the
amplitude of the velocity), and
Z =2 Z =2
4 he x p +1 hexp 4
Ih = cos !t d(!t) = coshe x p 1
!t d(!t): (10.138)
0 hexp + 1 0
(The second form of Ih ; obtained from the …rst via integration by parts, expresses Ih in
terms of gc above.) To maintain DeltaEC’s high speed, the trigonometric integrals are not
evaluated by DeltaEC; we use simple functional …ts to these integrals in DeltaEC.
The head-loss pressure gradient dp2;0 =dx in SX is calculated in the same way as for
STKSCREENs. The nonlinear nature of the screen friction factor f = c1 ( )=NR + c2 ( ) also
causes the mean ‡ow to a¤ect dp1 =dx in SX, according to
dp1 U1
= [c1 ( ) + 2c2 ( )NR;2;0 ] ; j"2;0 j 1; (10.139)
dx 8rh2 A
U1 c1 ( ) c2 ( )NR;1 "22;0 q 3
= + 1 + 1 "22;0 + "2;0 sin 1 "2;0 ;
rh2 A 8 3 2 2
j"2;0 j 1; (10.140)
jU1 j 4rh m
NR;1 = : (10.141)
A
197
[Note NR;2;0 in Eq. (10.139) but NR;1 in Eq. (10.140) — not a typographical error.] We did
not incorporate a mean ‡ow e¤ect on dp1 =dx in STKSCREEN because the energy equation
restricts the magnitude of N_ in stacks. No such restriction exists in heat exchangers.
In PX, dp2;0 =dx is calculated just as in STKPOWERLW. The mean ‡ow a¤ects dp1 =dx in PX
according to
dp1
= i! m hu1 i
dx
m jhu1 ij hu1 i (3 2 j"2;0 j) (1 2 j"2;0 j) dfM
j"2;0 j fM;max + NR;max ; j"2;0 j 1;
4rh 8(1 + j"2;0 j) dNR
(10.142)
jhu1 ij hu1 i
m "22;0 q 3
= i! m hu1 i 1+ 1 "22;0 + "2;0 sin 1 "2;0 fM;max
3 rh 2 2
"22;0 q 3 2
9 1 + 4"2;0 dfM
1+ 1 "22;0 + "2;0 sin 1 "2;0 NR;max ; j"2;0 j 1:
2 2 32 1 + j"2;0 j dNR
(10.143)
which is a hybrid of Eq. (10.132) here and Eqs. (A13) and (A14) in Ref. [42].
For N_ = 0; temperature and power change according to
dTm
= 0; (10.144)
dx
dH_ tot Q_
= ; (10.145)
dx x
where x is the heat exchanger length and Q_ is the heat added to the heat exchanger.
Positive heat Q_ (parameter “e”) ‡ows into the apparatus. The heat ‡ow Q_ is an input for
each pass of DeltaEC, and it is often used as a guess for the shooting method described
in Section 2.5.
For N_ 6= 0; the wave and energy calculations are unchanged, so all of the heat goes into
_
H2;k and none into the steady ‡ow. This seems …ne for weak steady ‡ows, such as weak
Gedeon streaming when a TASHE is operated slightly away from the conditions at which
its gas diode enforces N_ = 0: For strong steady ‡ow, the user can override the built-in
calculations with an RPN such as
13 RPN Fix Tm. Inp = frac of heat that should go to mean flow
0.25
H2k inp 12e * - =H2k ; inp 12e * Ndot / m / cp / Tm + =Tm
following the heat exchanger, where 12e is the heat in the upstream heat exchanger and 13a
is the steady-‡ow heat fraction. That fraction is often guessed, to give DeltaEC enough
‡exibility to meet nearby temperature and energy boundary conditions. Alternatively, the
1
fraction can also be set equal to a simple expression like 1 + m jU1 j =jN_ jm in an RPN,
supplying an experimentally unsupported but plausible estimate of how power divides itself
between the steady ‡ow and the acoustic ‡ow.
198
In SXs, the solid temperature is computed relative to gas mean temperature using
Q_ rh2 (gc gv )
T = (10.146)
k b( ) Axe¤
where xe¤ = minf2 jh 1 ij, xg. In PXs, the solid temperature is computed relative to the
gas mean temperature using
Q_ rh2
T = Ih 1 he x p
; (10.147)
k hcon NR;1 Axe¤
where xe¤ = minf2 jh 1 ij ; xg ; NR;1 is the Reynolds number based on the amplitude of the
velocity, and Ih is de…ned above. Then
These expressions may be very inaccurate, but we believe they are more useful than nothing.
The user who mistrusts them can use the gas temperature (available as an output in the
adjacent stack segment, and through RPN) instead of the solid temperature for plotting or
targeting, and/or use RPNs to construct customized expressions for Eqs. (10.146)–(10.149).
199
Figure 10.6: Example results from VXT1, STKSCREEN, VXQ1 segments in series in a Stirling engine.
The VXT1 extends from x = 0 to x = 0:05 m, with tubesheets 0.005 m long at each end. The
STKSCREEN extends from x = 0:05 m to x = 0:10 m. The VXQ2 extends from x = 0:10 m to
x = 0:24 m, with tubesheets 0.02 m long on each end. Temperatures in upper panel. The gas
temperature (solid line) varies with x in all three segments. The solid temperature (dashed line)
does not vary with x in the VXT1 segment. Lower panel shows H_ tot (solid line) and E_ (dashed line).
200
Figure 10.7: Two perpendicular cross sections through a VXT2, sandwiched between a duct above
and a regenerator below. Thermoacoustic oscillations in the vertical direction pass through the
insides of 33 tubes, of which 7 are exposed in the lower part of this …gure. Cooling water passes
through the “shell” in two passes: rightward across the lower half, and then leftward across the
upper half.
the heat-transfer coe¢ cients on the shell side and on the thermoacoustic side. If the heat
transfer is much better on the shell side, use VXT*. If the heat transfer is much better on
the thermoacoustic side, use VXQ*.
The fourth character in the segment name, 1 or 2, allows the segment to be formed of
1 or 2 separate heat-transfer pieces along x (in addition to the two “tubesheets” at either
end). Thus, “2”is intended for heat exchangers with two “passes”on the shell side. Figure
10.7 shows some of the geometry parameters for a VXT2 (or VXQ2).
To understand the philosophy behind the VX** calculations, consider Fig. 10.8, which
shows a length dx of this segment. In contrast with the STK** segments, where the mean
temperature of the solid is assumed to be the same as the mean temperature of the gas,
here those time-mean temperatures can be di¤erent, so large time-averaged heat ‡ows can
be driven from one to the other. Let Tm (x) be the mean temperature of the gas and Tsolid (x)
be the mean temperature of the solid. Both may be functions of x: Arti…cially, assume that
Tm in the gas is independent of the coordinate perpendicular to x: This is equivalent to
201
Figure 10.8: Schematic of a short section of a VX** segment. The solid lines show the edges
of the solid and the gas. The dashed box is the short control volume, dx long, used to write the
conservation laws for this situation. The space above the solid represents the external heat-transfer
gas such as water or combustion gases, which transfer heat at rate dQ_ to the control volume.
assuming that the heat-transfer bottleneck for the time-averaged heat transfer occurs in a
thin layer adjacent to the surface.
GasA/A The fraction of the total area that is available to gas, Agas =A:
SolA/A The fraction of the total area occupied by solid that contributes to solid thermal
conduction in the x direction, Asolid =A:
rh (m) The hydraulic radius rh of the gas channels. Hydraulic radius is de…ned as the
ratio of the gas volume to the gas–solid contact area. For the present case of an x-
independent channel cross section, rh = Agas = where is the inside perimeter of the
channel.
LenTS1 (m) The length of the …rst tubesheet in the heat exchanger. This region experi-
ences viscous and thermal-hysteresis loss, but not time-averaged heat transfer.
Len P1 (m) The length of the heat-transfer region in VXQ1; the length of the …rst of two
heat-transfer regions in VXQ2.
Len P2 (m) The length of the second heat-transfer region in VXQ2. (Not present in VXQ1.)
LenTS2 (m) The length of the second tubesheet in the heat exchanger. This region expe-
riences viscous and thermal-hysteresis loss, but not time-averaged heat transfer.
202
HeatP1 The heat added to the thermoacoustic gas in VXQ1. The heat added to the ther-
moacoustic gas in the …rst heat-transfer region in VXQ2. Positive heat indicates heat
‡ow from solid to gas.
HeatP2 The heat added to the thermoacoustic gas in the second heat-transfer region in
VXQ2. (Not present in VXQ1.) Positive heat indicates heat ‡ow from solid to gas.
FracQN If “Enable Ndot”is checked in an upstream BEGIN, this variable, FQ_ N_ ; determines
what fraction of heat is added to H_ N_ = N_ mwm : (The remainder is added to H_ 2;k :) In
VXQ2, the same value of FQ_ N_ is applied to both passes.
GasA/A The fraction of the total area that is available to gas, Agas =A:
SolA/A The fraction of the total area occupied by solid that contributes to solid thermal
conduction in the x direction, Asolid =A:
rh (m) The hydraulic radius rh of the gas channels. Hydraulic radius is de…ned as the
ratio of the gas volume to the gas–solid contact area. For the present case of an x-
independent channel cross section, rh = Agas = where is the inside perimeter of the
channel.
LenTS1 (m) The length of the …rst tubesheet in the heat exchanger. This region experi-
ences viscous and thermal-hysteresis loss, but not time-averaged heat transfer.
Len P1 (m) The length of the heat-transfer region in VXT1; the length of the …rst of two
heat-transfer regions in VXT2.
Len P2 (m) The length of the second heat-transfer region in VXT2. (Not present in VXT1.)
LenTS2 (m) The length of the second tubesheet in the heat exchanger. This region expe-
riences viscous and thermal-hysteresis loss, but not time-averaged heat transfer.
TSolP1 The solid temperature in the heat-transfer region in VXT1. The solid temperature
in the …rst heat-transfer region in VXT2.
TSolP2 The solid temperature in the second heat-transfer region in VXT2. (Not present in
VXT1.)
FracQN If “Enable Ndot”is chekced in an upstream BEGIN, this variable, FQ_ N_ ; determines
what fraction of heat is added to H_ N_ = N_ mwm : (The remainder is added to H_ 2;k :) In
VXT2, the same value of FQ_ N_ is applied to both passes.
203
Potential targets:
There are no built-in potential targets in VXQ1 and VXQ2. Use RPN targets as needed.
Master–slave links:
None.
Calculations for N_ = 0:
The …rst-order continuity and momentum equations are taken to be unchanged by the time-
averaged heat transfer. Hence, p1 and U1 evolve according to
dp1 i! m
= U1 ; (10.150)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1 + p1 + U1 ;
dx ma 1+ s (1 f )(1 )(1 + s ) dx
(10.151)
in all four VX** segments, just the same as in stacks. However, the energy equation used
in stacks does not su¢ ce here: H_ 2;k;gas and H_ solid should be considered separately instead of
only considering their sum H_ 2;k ; and the gas mean temperature Tm and solid temperature
Tsolid might sometimes evolve separately with x. So, here there are four equations related to
204
energy ‡ux and energy conservation:
" !#
1 Tm (f ~
f )
H_ 2;k = Re p1 U~1 1
2 (1 + s )(1 + )(1 f~ )
" #
c jU j2
dT (f ~
f )(1 + f =f )
m p 1 m s
+ 2 Im f~ +
2!Agas (1 ) j1 f j dx (1 + s )(1 + )
dTm
Agas k + H_ solid ; (10.152)
dx
dTsolid
H_ solid = Asolid ks ; (10.153)
dx
dH_ 2;k
q(x)
_ = ; (10.154)
dx
dH_ solid Agas k
q(x)
_ = + NN u (Tsolid Tm ) ; (10.155)
dx 4rh2
where q_ is the heat per unit length externally delivered to the heat exchanger (if negative,
it’s the heat removed).
The …rst of these four equations describes the total energy transport in the x direction,
and the second describes the energy transport (via heat conduction) in the x direction in
the solid.
The third equation expresses conservation of energy for the dashed-line control volume
in Fig. 10.8, showing that the heat input dQ_ = q(x)
_ dx equals the di¤erence (dH_ 2;k =dx) dx
between energy transport in and out along x.
The fourth equation similarly expresses conservation of energy for the solid separately
within the control volume.
Equations (10.152)–(10.155) are used to obtain the evolution of the temperatures Tm (x)
and Tsolid (x); and the energy ‡ows H_ 2 (x); H_ solid (x); and q(x);
_ as functions of x; simultane-
ously with the use of Eqs. (10.150) and (10.151) to obtain p1 (x) and U1 (x): Some details are
di¤erent in VXT* segments and VXQ* segments:
In the VXT* segments, Tsolid is independent of x and is a user’s input. Then H_ solid = 0;
Eq. (10.152) is solved for dTm =dx to yield the evolution of Tm (x); Eqs. (10.154) and (10.155)
together yield dH_ 2;k =dx; and Eq. (10.154) yields q(x)
_ = dH_ 2;k =dx: The heat transfer per
unit length q(x)
_ is integrated with respect to x to obtain the total heat transfer rate Q:_ (In
VXT2, the two “passes”can have di¤erent values of Tsolid :) Thus, Eqs. (10.150)–(10.155) are
regarded as …rst-order di¤erential equations for dp1 =dx; dU1 =dx; dTm =dx; and dH_ 2;k =dx (and
205
q(x)
_ is obtained trivially from dH_ 2;k =dx) :
h i
Tm (f f~ )
dTm H_ 2;k 1
2
Re p1 U~1 1 (1+ s )(1+ )(1 f~ )
= 2
h i ; (10.156)
dx f~ )(1+ s f =f )
m cp jU1 j
2!Ag a s (1 )j1 f j2
Im f~ + (f (1+ s )(1+ )
Agas k
dp1 i! m
= U1 ; (10.157)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1+ p1 + U1 ;
dx ma 1+ s (1 f )(1 )(1 + s ) dx
dTm
with given by Eq. (10.156), (10.158)
dx
dH_ 2;k Agas k
= 3:7 mk (Tsolid Tm ) : (10.159)
dx 4rh2
The factor “3.7”in Eq. (10.159) is the laminar, isothermal-wall Nusselt number for a circular
channel (Ref. [63], Table 6-1), so except for the presence of mk this is just the simplest
laminar, circular heat-transfer equation. The multiplier mk accounts for the enhancement
of the Nusselt number due to turbulence, as is described below.
Starting conditions are straightforward for VXT* segments: The starting p1 ; U1; Tm ; and
H_ 2;k are obtained directly from the end of the previous segment, and Tsolid is given by the
user (or as a guess, of course).
In the VXQ* segments, the heat exchange per unit length q_ is independent of x and is set
by the user, as the segment’s heat parameter divided by the appropriate length parameter.
In VXQ2, the two “passes”can, of course, have di¤erent values of heat and length, and hence
di¤erent values of q. _ To make progress, it is necessary to make a couple of assumptions
to simplify the starting equations. Solving Eqs. (10.152)–(10.155) for dTm =dx; dTsolid =dx;
dH_ 2;k =dx; and dH_ solid =dx and integrating them would require initial values for Tsolid and
H_ solid ; in addition to the initial values of Tm and H_ 2;k that are passed into the VXQ* segment
from the end of the previous segment. To avoid having to specify or guess the initial
values of Tsolid and H_ solid , we assume that most of q_ goes into the thermoacoustic gas in
cases of interest to us, so the …rst term on the right side of Eq. (10.155) can be neglected
there. This assumption probably leads only to a slight overestimation of the temperature
di¤erence jTsolid Tm j in cases of interest to us. Second, we assume that dTsolid =dx =
dTm =dx: Examination of Eq. (10.155) (with the aforementioned …rst assumption, and with
q_ independent of x here for a VXQ* segment) shows that this assumption is equivalent to
d(kNN u )=dx = 0; which seems reasonable enough (until someone …gures out a better way to
206
do all this simply enough for DeltaEC). Then the starting equations can be rearranged as
h i
_ 2;k 1 Re p1 U~1 1 Tm (f f~ )
dTm H 2 (1+ s )(1+ )(1 f~ )
= 2
h ~
i ;
dx m cp jU1 j
Im ~
f + (f f )(1+ s f =f )
A k A k
2!Ag a s (1 2
)j1 f j (1+ s )(1+ ) gas solid s
(10.160)
dp1 i! m
= U1 ; (10.161)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1 + p1 + U1 ;
dx ma 1+ s (1 f )(1 )(1 + s ) dx
dTm
with given by Eq. (10.160), (10.162)
dx
dH_ 2;k
= q;
_ (10.163)
dx
4rh2
Tsolid = Tm + q;
_ (10.164)
4:4 mk Agas k
The factor “4.4” in Eq. (10.164) is the laminar, constant-heat-‡ux-density Nusselt number
for a circular channel (Ref. [63], Table 6-1), so except for the presence of mk this is just
the simplest laminar, circular heat-transfer equation. The multiplier mk accounts for the
enhancement of the Nusselt number due to turbulence, as is described below.
Because of the assumptions in the previous paragraph, the starting conditions for the
numerical integration are unambiguous: the four integrated variables Tm ; p1 ; U1 ; and H_ 2;k
all start with the values left over from the previous segment of the DeltaEC model.
Tubesheets are treated exactly like stacks, with H_ 2;k constant and
h i
_ 1 ~ Tm (f f~ )
dTm H2;k 2 Re p1 U1 1 (1+ )(1+ )(1 f~ )
s
= 2
h ~
i ;
dx c jU
m p 1 j ~
2 Im f +
(f f )(1+ s f =f )
Agas k Asolid ks
2!Ag a s (1 )j1 f j (1+ s )(1+ )
(10.165)
dp1 i! m
= U1 ; (10.166)
dx (1 f )Agas
dU1 i!Agas ( 1)f (f f ) dTm
= 2
1+ p1 + U1 ;
dx ma 1+ s (1 f )(1 )(1 + s ) dx
dTm
with given by Eq. (10.165); (10.167)
dx
but with Asolid = Atotal Agas .
The important viscous and thermal functions f and f needed in Eqs. (10.150)–(10.152)
and their variants are calculated as described here. For large displacement amplitudes, the
e¤ective length of the tube is its geometrical length (the total geometrical length, i.e., the
sum of tubesheet and heat-transfer-region lengths). For small displacement amplitudes, the
e¤ective length is in…nite. In between, this interpolation is used:
4 xgeom xgeom xgeom
xe¤ = when < j 1j < : (10.168)
10 j 1 j = xgeom 1 10 2
207
Figure 10.9: E¤ective length as a function of displacement amplitude j 1 j : For N_ disabled, only
the horizontal axis is relevant. For N_ enabled, this plot displays Eqs. (10.185)–(10.187).
(The choices of this functional form, and of 1/10 and 1/2 as transitions, are arbitrary and
without justi…cation, but these details have little e¤ect on the …nal results.) With this e¤ec-
tive length, shown along the x axis in Fig. 10.9, the top half of Fig. 7-1 in Kays and London
is used to estimate the e¤ect of turbulence on f and f in the tube. (That …gure shows
friction factor as a function of Reynolds number, for a few values of x=D:) DeltaEC’s
internal approximations to the Kays and London data are shown here in Fig. 10.10. Power
laws
F = g10 BNR (10.169)
are …t to Kays and London’s high-NR straight lines, and similar power laws with o¤sets are
…t to their low-NR straight lines, with interpolations for 2450 < NR < 10; 000 using a cubic
polynomial matched to the inner ends of the two straight lines in both slope and value.
(Here F is the Moody friction factor, four times the Fanning friction factor that Kays and
London call f:) Then, using the peak Reynolds number, a multiplier
2
NR F
m = max 1; 2
(10.170)
24 rh Re [i= (1 f )]
gives the estimate of how turbulence (including entrance e¤ects) a¤ects what would otherwise
be a laminar boundary layer. This expression is easily derived from the discussion about
turbulence in Section 10.1.1.
Bessel-function (or, for large R; boundary-layer) expressions are used for “starting val-
ues” of f and f : Then m as described above is used to generate modi…ed f and f
for the momentum and continuity equations above, exactly as described in Section 10.1.1:
DeltaEC increases the resistive component of the pressure gradient, and hence the viscous
power dissipation, by m : It decreases the inertial pressure gradient by
2
1 =R
m0 = (10.171)
1 =m R
208
Figure 10.10: DeltaEC’s …ts for Kays and London’s Fig. 7-1. The upper set of curves is the
Fanning friction factor [Fanning = Moody /4 = F=4 in Eqs. (10.169) and (10.170)], and the lower
set is for the heat-transfer coe¢ cient. Within each set, the di¤erent curves are for di¤erent L=D :
in…nity, 100, 50, 25, 15, and 1 from bottom to top.
209
to correct approximately for the steeper velocity gradient at the wall, which increases the
e¤ective area open to gas contributing to inertial e¤ects. It applies the same turbulence
correction to f as to f ; with no justi…cation other than the idea that turbulent eddies
carry heat and momentum similarly. For both f ’s,
1 1
Im = m Im ; (10.172)
1 fj 1 fj, laminar
1 1
Re = m0 Re ; (10.173)
1 fj 1 fj, laminar
1
1 1
fj = 1 Re + i Im (10.174)
1 fj 1 fj
[Note that, perhaps somewhat inconsistently, m enhances heat transfer in f in Eqs. (10.151)
and (10.152) while mk is used in the time-averaged heat-transfer Eqs. (10.159) and (10.164).]
The value of mk needed in Eq. (10.155) is obtained as follows. Power laws
BNR
= A10 (10.175)
are …t to the low-NR and high-NR straight lines in the lower half of Kays and London
Fig. 7-1, with interpolations for 2450 < NR < 10; 000 using a cubic polynomial matched
to the inner ends of the two straight lines in both slope and value. (That …gure shows
heat-transfer factor as a function of Reynolds number, for a few values of x=D: Here
is the heat-transfer factor that Ref. [63] calls NSt 2=3 :) Figure 10.10 shows these …ts. Since
time-averaged heat transfer is desired, these equations are evaluated using the time-averaged
Reynolds number (2= ) jNR;1 j. Finally, a multiplier
mk = (10.176)
laminar ( xe¤ = 1)
gives the turbulence correction factor in the otherwise laminar Eqs. (10.159) and (10.164).
In the presence of nonzero steady ‡ow N_ ; the starting equations are modi…ed and augmented
as follows:
210
" !#
1 Tm (f f~ )
H_ tot = Re p1 U~1 1
2 (1 + s )(1 + )(1 f~ )
" #
c
m p jU 1 j2
dTm (f ~
f )(1 + s f =f )
+ 2 Im f~ +
2!Agas (1 ) j1 f j dx (1 + s )(1 + )
dTm
Agas k + H_ solid + N_ mwm ; (10.179)
dx
dTsolid
H_ solid = Asolid ks ; (10.180)
dx
dH_ tot N_ mcp dTm
q(x)
_ = = in a pass, but = 0 in a tubesheet, (10.181)
dx FQ_ N_ dx
Agas k
q(x)
_ = NN u (Tsolid Tm ) ; (10.182)
4rh2
!2 v !2
u
dp2;0;HL 1 m N_ F u1 jU j
= sign(N_ 2;0 ) m
2;0 t + m 1
(10.183)
dx 2 A
m gas 4r h 2 _
mN2;0
h i 1
N_ 2;0 = N_ + H_ tot N_ mwm + E_ 2 : (10.184)
mcp ma2 mcp
Important: Equation (10.184) omits the (Agas k + Asolid ks ) dTm =dx term that is present in
Eq. (8.16) because we expect it to be negligible most of the time and because DeltaEC
needs N_ 2;0 before it knows dTm =dx in each pass of its numerical
q integration. In Eqs. (10.177)–
(10.183), DeltaEC evaluates m and F on the basis of jNR;1 j2 + NR;2;0 2
; and it evaluates
q
m on the basis of 4 jNR;1 j2 = 2 + NR;2;02
: The square-root factor at the end of the equa-
tion for dp2;0;HL =dx is a rough empirical match to steady-‡ow DUCT results; this factor helps
counteract the fact that high jU1 j biases the friction factor F toward values that are unreal-
istically low for use in this equation. The friction and heat-transfer factors are based on an
e¤ective tube length xe¤ plotted in Fig. 10.9 and given by
2 2
j 1j NR;2;0 1
xe¤ = 1 when + ; (10.185)
xgeom 1000 100
2 2
4 xgeom 1 j 1j NR;2;0 1
= r when + ;
j 1j
2
NR;2;0
2 100 xgeom 1000 4
10 xg e o m
+ 1000
1
(10.186)
2 2
j 1j NR;2;0 1
= xgeom when + : (10.187)
xgeom 1000 4
211
10.8. Adiabatic–isothermal interface loss
10.8.1. JOIN
General description:
The JOIN segment accounts for “small” discontinuities in thermoacoustic variables at the
interface between a heat exchanger or other isothermal segment and an unmixed, thermally
strati…ed, adiabatic region such as a pulse tube. The discontinuity Tm in mean temperatures
is …rst order in the acoustic amplitude, the discontinuity U1 in volume ‡ow rate is second
order, and there is no discontinuity in pressure. See Refs. [64, 65] for Tm ; Ref. [66] for U1 ;
and Ref. [13] and references therein for both Tm and U1 .
When it encounters a JOIN segment, DeltaEC looks both upstream and downstream
(ignoring merely logistical segments such as RPN) to …gure out whether the “heat exchanger”
is upstream and the “pulse tube” is downstream, or vice versa. Both A and dTm =dx are
obtained from the “pulse tube”or other open duct, with dTm =dx = 0 for ordinary DUCTs and
CONEs, and dTm =dx nonzero for STKDUCTs or STKCONEs. (If the relevant duct is a STKDUCT or
STKCONE downstream of the JOIN segment, DeltaEC actually integrates ahead momentarily
to evaluate dTm =dx:)
The JOIN feature does not yet do anything di¤erent for nonzero N_ .
212
1.3511E+05 A |p| Pa
-17.530 B Ph(p) deg
2.7854E-04 C |U| m^3/s
166.83 D Ph(U) deg
-13.711 E Hdot W
-18.763 F Edot W
57.932 G TBeg K
89.997 H TEnd K
!--------------------------------- 13 ---------------------------------
RPN ratio of warm end displacement to p.t. length
6.0000 a G or T (t) 9.8838E-02 A ChngeMe
8C 2 / pi / 0b / 10a / 10c /
!--------------------------------- 14 ---------------------------------
SX cold heat exchanger adj rh
1.1675E-04 a Area m^2 1.5473E+05 A |p| Pa
0.6470 b VolPor -16.785 B Ph(p) deg
5.7150E-03 c Length m 2.7985E-04 C |U| m^3/s
1.2000E-05 d rh m 169.37 D Ph(U) deg
7.1278 e HeatIn W G -6.5828 E Hdot W
Possible targets -21.525 F Edot W
7.1278 G Heat W
copper Solid type 90.000 H SolidT K
Calculations:
The discontinuity in temperature at JOIN is:
Tm jU1 jin dTm
Tm;out Tm;in = jp1 j sin F; (10.188)
m cp !Agas dx
where is the angle by which p1 leads U1 ; dTm =dx is evaluated on the thermally strati…ed,
adiabatic side of the JOIN, and the factor F is given by
v
u
u 2 (Agas k dTm =dx)2
F =ut 2: (10.189)
2 _
(Agas k dTm =dx) + H2;k E _
The factor F is a very crude attempt to account for the two-dimensional nature of the
JOIN problem within DeltaEC’s inherently one-dimensional character. The derivation of
the Tm discontinuity joining condition in the references cited above neglects the thermal
conductivity of the wall of the tube and the boundary-layer shuttle entropy ‡ux. If these
are zero while dTm =dx 6= 0; then F = 1 and Eq. (10.188) represents the joining condition
derived in the references cited above. On the other hand, if the solid wall of the duct or the
boundary layer is e¤ectively a thermal short circuit compared to bulk gas conductivity, then
F = 0 and no temperature discontinuity appears at the JOIN segment. We arbitrarily chose
a simple smooth function that passes through these two limiting cases. (This function was
changed between DeltaEC version 6.3 and version 6.4, to cure a rare situation that had
nonsensically yielded F 1:)
The discontinuity in jU1 j at JOIN is
8 ( 1)
jU1 jout = jU1 jin 2
jp1 j jU1 jin cos
3 ma
16 1 _
= jU1 jin 2
E: (10.190)
3 ma
213
10.9. Mixture separation
10.9.1. MIXBL, MIXSLAB, and MIXCIRC
General description:
The MIXBL, MIXSLAB, and MIXCIRC segments calculate dp1 =dx; dU1 =dx; and dnL =dx for
sound propagation in gas mixtures, with or without steady ‡ow, according to the theory
described in Ref. [10], which builds on the theory described in Refs. [44, 67]. MIXBL performs
the calculations in boundary-layer approximation, MIXSLAB in parallel-plate geometry of
arbitrary gap 2y0 , and MIXCIRC in circular pores of arbitrary radius R. The solid is assumed
to be ideal, so it imposes a perfect isothermal boundary condition on the gas.
GasA/A The areal porosity = Agas =A: (We often …x this at 1.000 and set Area equal to
the gas area.)
y0 (m) Half the plate spacing is y0 : Thus, 2y0 is the gap available to the gas.
GasA/A The areal porosity = Agas =A: (We often …x this at 1.000 and set Area equal to
the gas area.)
Potential targets:
None.
214
Master–slave links:
The length of any of these segments can be slaved to the length of another segment to keep
the sum of the two lengths constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
Calculations:
The evolution of p1 with x in MIXBL, MIXSLAB, and MIXCIRC is given by the same integration
as in DeltaEC’s other ducts,
dp1 i! m
= U1 ; (10.191)
dx (1 f )Agas
because Ref. [68] showed that the …rst-order thermoacoustic momentum equation is un-
changed by mixture-separation e¤ects. The evolution of U1 with x is given by Eq. (8) in
Ref. [68], but they didn’t write the result explicitly [10]. The explicit result, in DeltaEC’s
notation, is
dU1 i!Agas
= 2
f1 + ( 1) [Bf + Cf D + (1 B C) fD ]g p1 ; (10.192)
dx ma
where B and C are given in Ref. [44], and the other variables carry their usual DeltaEC
meanings or are de…ned below.
215
The evolution of the light fraction nL with x is given by solving Eq. (7) in Ref. [10] for
dnH =dx, and integrating numerically for …xed N_ L ; which is equivalent to …xed N_ H and …xed
N_ = N_ H + N_ L . Solving Eq. (7) for dnH =dx yields
dnH N_ H nH N_ 4rh
1 kT
Ru n iv Tm
jp1 j jU1 j (Ftrav cos + Fstand sin )
= 2 ; (10.193)
dx m jU1 j
Fgrad m
Agas D12
4rh mav g !Agas mav g
and then DeltaEC uses dnL =dx = dnH =dx to obtain the desired result.
p In addition to thepusual de…nitions of the viscous and thermal penetration depths =
2 =! m and = 2k=! m cp ; two new penetration depths are de…ned for mixtures:
q
1
2
D = 2
1 + (1 + ") =L + [1 + (1 + ") =L]2 4=L ; (10.194)
2
q
1
2
D = 2
1 + (1 + ") =L [1 + (1 + ") =L]2 4=L ; (10.195)
2
where
1 kT2
"= ; (10.196)
nL (1 nL )
kT is the thermal di¤usion ratio, L = k= m cp D12 , and D12 is the binary mass-di¤usion
coe¢ cient.
In MIXBL, DeltaEC uses the boundary-layer results of Ref. [44]. For f ; f D ; and fD ;
the boundary layer result is
j
fj = (1 i) ; (10.197)
2rh
p p p
L L( D= + D = )
Ftrav = p ; (10.198)
1 + L [(1 + )(1 + L) + " ]
p p p
L+ L ( D= + D = )
Fstand = p ; (10.199)
1 + L [(1 + )(1 + L) + " ]
p p p p h p i
2 2
L (1 ) 1+ L +" L 1 + ( 1) L + " L D
+ D
Fgrad = p :
1 + L [(1 + )(1 + L) + " ] [(1 )(1 L) " ]=
(10.200)
216
In MIXSLAB and MIXCIRC, DeltaEC uses the following expressions:
2 3
2rh 4 G 5;
Ftrav = Re (10.201)
1 f~
2 3
2rh 4 G 5;
Fstand = Im (10.202)
1 f ~
2rh 1
Fgrad =
j1 f j2 (1 ) (1 L) "
2 2 2 2
(1 ) D D (1 + ) LQ ~
Im f 2 2 f D 2 2 fD + f +S + "G ;
S + D + D M
(10.203)
where rh is the hydraulic radius and
2 2
S = 2 1 fD 2 1 f D; (10.204)
D D
2 2
D D
Q = 2 ; (10.205)
These expressions appear in Ref. [10] in the context of a circular tube.2 In MIXCIRC the
three f ’s are given by
2J1 [(i 1)R= j ]
fj = ; (10.208)
(i 1)(R= j )J0 [(i 1)R= j ]
where R is the circle radius, rh = R=2; and j is ; D; or D : In MIXSLAB, the three f ’s are
given by
tanh[(1 + i)y0 = j ]
fj = ; (10.209)
(1 + i)y0 = j
where the plate separation is 2y0 = 2rh and j is ; D; or D :
The head-loss pressure gradient dp2;0;HL =dx is calculated on the basis of laminar ‡ow.
See Eqs. (10.69) and (10.70).
217
the theory described in Ref. [69], which builds on the theory described in Refs. [44, 67, 10].
MIXTBL performs the calculations in boundary-layer approximation, MIXTSLAB in parallel-
plate geometry of arbitrary gap 2y0 , and MIXTCIRC in circular pores of arbitrary radius
R.
The solid is assumed to be laterally ideal, so it imposes a perfect isothermal boundary
condition on the oscillating temperature of the gas. This assumption would correspond to
s = 0 in a STK** segment. However, the solid’s …nite thermal conductivity in the x direction
is included in the calculation of dTm =dx in MIXT** segments, in the same way as in STK**
segments.
Potential targets:
None.
218
Master–slave links:
The length of any of these segments can be slaved to the length of another segment to keep
the sum of the two lengths constant.
(For an introduction to master–slave links, and a complete listing of linking options for
all segments, see Section 2.7.2.)
Calculations:
dp1 i! m
= U1 ; (10.211)
dx (1 f )Agas
dnH
=
dx
jU1 j2 FrT
N_ H nH N_ ( 1)kT
4rh Ru n iv Tm
jp1 j jU1 j (Ftrav cos + Fstand sin ) + m kT dTm
m Tm dx
Agas D12 4rh j1 f j2 !Ag a s
;
Fg ra d
m
m 4rh
jU1 j2 !Ag a s
Agas D12
(10.212)
219
where
hT1 i 1 p1 (1 fD ) 1 dTm
= [1 Bf Cf D (1 B C) fD ] U1 ; (10.214)
Tm pm i!Agas (1 f ) Tm dx
2 2
1 m L kT 1 p1
hnH;1 i = B 1 f +C 1 2 f D + (1 B C) 1 2 fD
mavg " D D pm
2
1 dnH m L kT 1 1 dTm
U1 + 1 2 fD f U1 ; (10.215)
i!Agas dx mavg " i!Agas (1 f ) D Tm dx
and
ie i p dTm =dx
+ 1+ p ; (10.217)
1 f L D D rTcrit
1 jp1 j !Agas
rTcrit = Tm ; (10.218)
pm jU1 j
1 jp1 j !Agas
rncrit = kT : (10.219)
pm jU1 j
In
p addition to the usual
p de…nitions of the viscous and thermal penetration depths =
2 =! m and = 2k=! m cp ; two new penetration depths are de…ned for mixtures:
q
1
2
D = 2
1 + (1 + ") =L + [1 + (1 + ") =L]2 4=L ; (10.220)
2
q
1
2
D = 2
1 + (1 + ") =L [1 + (1 + ") =L]2 4=L ; (10.221)
2
where
1 kT2
"= ; (10.222)
nL (1 nL )
kT is the thermal di¤usion ratio, L = k= m cp D12 , and D12 is the binary mass-di¤usion
coe¢ cient.
For MIXTBL,
j
fj = (1 i) ; (10.223)
2rh
where j is ; ; D; or D and rh is the hydraulic radius, and the four F ’s are given by
220
p p p
L L( D= + D = )
Ftrav = p ; (10.224)
1 + L [(1 + )(1 + L) + " ]
p p p
L+ L ( D= + D = )
Fstand = p ; (10.225)
1 + L [(1 + )(1 + L) + " ]
p p p p h p i
2
L (1 ) 1+ L +" L 1 + ( 2 1) L + " L ( D= + D = )
Fgrad = p ;
1 + L [(1 + )(1 + L) + " ] [(1 )(1 L) " ]=
(10.226)
FrT =
p p p 3 p hp i
L 1+ L 1+ 2
L + 2L + " p1 + L (1 L " ) L(1 + ) D
+ D
p :
1 + L [(1 + )(1 + L) + " ] [(1 )(1 L) " ]=
(10.227)
2rh 1
FrT =
j1 f j2 (1 ) (1 L) "
2 2 2 2
f D D (1 + ) LQ ~ 1 L "
Im 2 2 f D 2 2 fD + f +S + G ;
S + D + D M
(10.231)
221
where rh is the hydraulic radius and
2 2
S = 2 1 fD 2 1 f D; (10.232)
D D
2 2
D D
Q = 2 ; (10.233)
These expressions appear in Refs. [10, 69] in the context of a circular tube.
In MIXTCIRC the three f ’s are given by
tanh[(1 + i)y0 = j ]
fj = ; (10.237)
(1 + i)y0 = j
222
11. Logistical segments
11.1.2. BEGIN
General description:
After the TITLE segment, BEGIN is counted as the zeroth segment of the …le. It is used
to initialize “global” variables that are shared among subsequent segments (e.g., frequency
and mean pressure) and the integration variables required by each pass of DeltaEC to get
started (e.g., real and imaginary parts of p1 and U1 ).
BEGIN segments can be used anywhere in a DeltaEC …le, to set any or all of DeltaEC’s
integration variables or global variables to new values. This can be useful for packing two
or more related DeltaEC models into a single …le or for changing from one gas to another
in the middle of a model. (Do not use such a BEGIN within a teebranch; instead, use RPN
223
“change”assignments to set or reset variables in a teebranch.) When more than one BEGIN
is used in a model, each BEGIN starts a new leg. (Legs are described in Section 11.2.1.)
Starting with DeltaEC version 6.0, BEGIN causes subsequent DUCTs, CONEs, SURFACEs,
and IMPEDANCEs to set H_ tot;out = H_ tot;in ; so the user should imagine such segments to be
thermally insulated. ANCHOR, described in Section 11.3.1, can be used to override this default,
and the user should imagine segments downstream of ANCHOR to be attached to a heat
sink at the local temperature Tm : (ANCHOR is useful for very low power systems and simple
systems where total energy ‡ow is irrelevant, and of course for systems with DUCTs and other
components that actually are well-attached to a heat sink.)
Input variables:
Mean P (Pa) The mean pressure of the gas, pm :
Freq (Hz) The frequency f of the sinusoidal oscillations in subsequent segments. Most
equations in the Users Guide use ! = 2 f:
Default Htot / Other Htot (W) With this choice in the Optional Parameters dialog, the
user can let DeltaEC assign the default value to the total energy ‡ow H_ tot or can
assign a numerical value or let it be a guess.
Enable Ndot / Disable Ndot This check box in the Optional Parameters dialog controls
some calculations and results in many subsequent segments.
Ndot (mol/s) If “Enable Ndot” is checked, the user can specify or guess the steady molar
‡ow rate N_ :
nL If the gas type is a gas mixture, this parameter initializes the light-component mole
fraction nL :
Bulk NLdot / Other NLdot (mol/s) If the gas type is a gas mixture, this check box in
the Optional Parameters dialog controls whether DeltaEC assigns the default, bulk
value to the light component’s steady molar ‡ow rate N_ L or lets the user assign a
di¤erent value.
224
Potential targets; Master–slave links:
None.
BEGIN This example is the most complicated possible format, as saved in an out file.
1.0e6 Pa a Mean P
500. Hz b Freq
300. K c TBeg
3.0e4 Pa d |p|
0.00 deg e Ph(p)
0.00 m3/s f |U|
0.00 deg g Ph(U)
2000. h Htot
1.00E-8 i Ndot
0.800 j nL
1.00E-9 k NLdot
HeAr Gas type
TITLE Argon cover on top of sodium-potassium column, illustrating two BEGINs in one model
!->2beg.out
!Created@10:20:31 29-Mar-07 with DeltaE Vers. 6.0g7 for the IBM/PC-Compatible
!--------------------------------- 0 ---------------------------------
BEGIN Initialize in the argon
1.0000E+05 a Mean P Pa 3.9676E-05 A |U| G( 0f) P
100.00 b Freq Hz 87.799 B Ph(U) G( 0g) P
400.00 c TBeg K
1000.0 d |p| Pa
0.0000 e Ph(p) deg
3.9676E-05 f |U| m^3/s G
87.799 g Ph(U) deg G
0.0000 h Htot W
0.0000 j nL
HeAr Gas type
!--------------------------------- 1 ---------------------------------
DUCT the argon on top
1.0000E-04 a Area m^2 S= -2 1015.7 A |p| Pa
3.5449E-02 b Perim m Fn( 1a) -7.9065E-02 B Ph(p) deg
0.1000 c Length m 1.9924E-08 C |U| m^3/s
89.527 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 6.9504E-08 F Edot W
!--------------------------------- 2 ---------------------------------
BEGIN transition from argon to Na-K
sameas 0a a Mean P Pa
sameas 0b b Freq Hz
sameas 0c c TBeg K
sameas 1A d |p| Pa
sameas 1B e Ph(p) deg
sameas 1C f |U| m^3/s
sameas 1D g Ph(U) deg
sameas 1E h Htot W
NaK-78 Gas type
!--------------------------------- 3 ---------------------------------
DUCT The sodium-potassium column
sameas 1a a Area m^2 1069.7 A |p| Pa
sameas 1b b Perim m -0.1210 B Ph(p) deg
1.0000 c Length m 3.7365E-13 C |U| m^3/s
-0.1156 D Ph(U) deg
0.0000 E Htot W
ideal Solid type 1.9984E-10 F Edot W
!--------------------------------- 4 ---------------------------------
SURFACE Bottom end
sameas 1a a Area m^2 1069.7 A |p| Pa
-0.1210 B Ph(p) deg
3.5753E-17 C |U| m^3/s
83.367 D Ph(U) deg
225
0.0000 E Htot W
ideal Solid type 2.1686E-15 F Edot W
!--------------------------------- 5 ---------------------------------
HARDEND terminate in the liquid metal
0.0000 a R(1/z) = 5G? 1069.7 A |p| Pa
0.0000 b I(1/z) = 5H? -0.1210 B Ph(p) deg
3.5753E-17 C |U| m^3/s
83.367 D Ph(U) deg
0.0000 E Htot W
2.1686E-15 F Edot W
6.3787E-11 G R(1/z)
5.5883E-10 H I(1/z)
Calculations:
which implies that H_ 2;k = E: _ Otherwise, Htot is an ordinary input parameter, which can be
set by the user or guessed.
If the “Disable Ndot”box is checked in the Optional Parameters dialog or the Ndot input
parameter is missing from the <.out> …le, DeltaEC assigns N_ = 0, skips all subsequent
p2;0;HL calculations, and ignores any possible e¤ect of nonzero N_ 2;0 on dp1 =dx. If the “Enable
Ndot” box is checked, p2;0;HL is initialized to zero and its dependence on x is calculated in
all subsequent segments. This also allows input parameter N_ to appear in BEGIN, and allows
other input and output parameters that are relevant to steady ‡ow to appear in subsequent
segments. When this box is checked, nonnegligible dp2;0;HL =dx and nonlinear e¤ects on
dp1 =dx can occur in subsequent segments even if N_ ' 0; because of the di¤erence between
N_ and N_ 2;0 described in Eq. (8.16).
If the gas type is a binary mixture, the initial mole fraction of the light component, nL ;
appears as an input parameter, which can only change in MIX** segments or be reassigned
using =nL in an RPN segment downstream.
Other, less commonly used variables are invisibly initialized in BEGIN:
vol = 0, described in Section 11.4.1.
F1 = 0, described in Section 11.4.1.
Tzero = 300 Kelvin, described above Eq. (6.3) in the context of H_ and below Eq. (8.9)
in the context of X. _
nLzero = 0:5, described below Eq. (8.9) in the context of X. _
N_ L = nL N_ ; if the gas is a binary gas mixture and the “Bulk NLdot”box is checked.
Tsolid = Tm :
226
11.1.3. HARDEND and SOFTEND
General description:
Often, the …nal segment (except possibly RPN segments) will be either HARDEND or SOFTEND.
These contain three or four potential targets.
HARDEND targets are appropriate when the user wants the complex volume ‡ow rate
somewhere to be zero. This is the usual case at the end of a closed thermoacoustic system.
HARDEND is also often used at the end of a closed, sealed TBRANCH.
SOFTEND targets are appropriate when the user wants the complex pressure amplitude
somewhere to be zero. This is often useful for symmetrical systems, where SOFTEND indicates
that the remainder of the apparatus is a mirror image of what is in the <.out> …le and forces
a complex-pressure node. Also use the SOFTEND segment, but without using its targets, to
signal the end of a TBRANCHed sequence of segments that will be reattached elsewhere at a
UNION, or anywhere that a leg must end with circumstances that can best be de…ned using
custom-made RPN targets.
Either **END segment can also be targeted to enforce a non-trivial impedance at the end
of a branch or a model. An example of SOFTEND to model the radiation impedance from an
open-ended tube is shown on the next page.
The third default target in these segments, H_ tot , can be targeted to zero to represent
thermal insulation at a HARDEND or the energy aspect of a mirror-image symmetry location
in SOFTEND.
In both **ENDs, the complex speci…c impedances are made dimensionless by dividing by
m a evaluated at the local temperature.
Disabling the targets in **ENDs is often useful in debugging a new model that doesn’t
readily converge. Set these targets nonzero to model a known, nonzero end impedance— or
use BRANCH (followed by a zero-targeted HARDEND) to work with Z instead of z= a:
Input variables:
None.
I(1/z) If targeted, this number is compared with DeltaEC’s calculation of the imaginary
part of the inverse of the normalized speci…c impedance, Im[1=zn ], at the HARDEND.
Htot (W) Total energy ‡ow H_ tot . This potential target appears in insulated mode. If
targeted, this number is compared with DeltaEC’s calculation of H_ tot at the HARDEND.
Ndot (mol/s) Steady molar ‡ow rate N_ . This potential target appears when “Enable
Ndot” is checked in an upstream BEGIN. If targeted, this number is compared with
DeltaEC’s calculation of N_ at the HARDEND. Useful only in rare circumstances.
227
Potential targets, SOFTEND:
R(z) If targeted, this number is compared with DeltaEC’s calculation of the real part of
the normalized speci…c impedance, Re[zn ], at the SOFTEND.
I(z) If targeted, this number is compared with DeltaEC’s calculation of the imaginary
part of the normalized speci…c impedance, Im[zn ], at the SOFTEND.
Htot (W) Total energy ‡ow H_ tot . This potential target appears in insulated mode. If
targeted, this number is compared with DeltaEC’s calculation of H_ tot at the SOFTEND.
Master–slave links:
None.
!--------------------------------- 97 ----------------------------
HARDEND looks like this if steady-flow calculations are turned off
0.000 R(1/z)
0.000 I(1/z)
0.000 Htot
!--------------------------------- 40 ----------------------------
DUCT final resonator tube, open to the room
8.0000E-4 a Area m^2 Mstr 1023.8 A |p| Pa
0.10027 b Perim m 40a -0.95235 B Ph(p) deg
3.1783E-2 c Length m G 2.2366E-2 C |U| m^3/s
5.0000E-4 d Srough -87.544 D Ph(U) deg
0.68065 E Htot W
ideal Solid type 0.68065 F Edot W
!--------------------------------- 41 ----------------------------
RPN calc radius from area
0.0000 a G or T 0.1461 A kr
1.5958E-2 B m rad
40a pi / sqrt ; # w * a /
!--------------------------------- 42 ----------------------------
RPN radiation impedance from open tube / (rho a / A)
! Levine and Schwinger, Phys Rev 73, 383-406 (1948)
! Z = rho a / A times z ;
! z = (kr)^2 / 4 + j 0.6133 kr
0.0000 a G or T 5.3360E-3 A Re(z)
8.9601E-2 B Im(z)
0.6133 41A * ; 41A sqrd 4 /
!--------------------------------- 43 ----------------------------
SOFTEND radiate sound into 4 pi solid angle in the room
! Note: z calc here uses closest upstream area (40a) internally AND via RPN 41A
sameas 42A a Re(z) =43G 1023.8 A |p| Pa
sameas 42B b Im(z) =43H -0.95235 B Ph(p) deg
0.0000 c Htot W 2.2366E-2 C |U| m^3/s
-87.544 D Ph(U) deg
0.68065 E Htot W
0.68065 F Edot W
5.3360E-3 G Re(z)
8.9601E-2 H Im(z)
293.00 I T K
Calculations:
The normalized impedance in HARDEND and SOFTEND is calculated according to
Ap1
zn = ; (11.2)
m aU1
6
with A from the most recent segment having an area (or A = 10 m2 if no such segment
exists), and with m and a evaluated at the local Tm .
228
11.2. Structured branches and unions
Figure 11.1 shows how TBRANCH and UNION can be used to model complicated branched and
multiply-connected thermoacoustic equipment.
Figure 11.1: (a) Example of an 18-segment DeltaEC model of a linear array of thermoacoustic
components. (b) Example of a model of a multiply-branched apparatus. (c) Example of a model
of a multiply-connected apparatus. Line a of UNION 17 must point to SOFTEND 13, line a of UNION
31 must point to SOFTEND 20, and line a of UNION 32 must point to SOFTEND 26.
11.2.1. TBRANCH
General description:
Use TBRANCH for branched systems that are too complicated for BRANCH or OPNBRANCH.
When it encounters a TBRANCH, DeltaEC treats subsequent segments as the sequential
members of a side branch (which we call a “teebranch”in the Users Guide), until it reaches
a HARDEND or SOFTEND. Then it “returns to the trunk,”treating the rest of the segments as
trunk members. The HARDEND termination of a teebranch is appropriate for a closed-end leg.
The SOFTEND termination of a teebranch is appropriate for later reconnection to the trunk
at a UNION in the trunk.
To manage the plotter display of results from models with one or more TBRANCHes, the
<.sp> …le’s initial column, “leg number,” is incremented each time a TBRANCH or BEGIN is
encountered, and decremented when a SOFTEND or HARDEND is encountered. The user can
display plots of one or more selected legs, suppressing results from uninteresting legs.
229
The complex impedance in the TBRANCH segment tells DeltaEC how to split up the
volume ‡ow rate between the teebranch and the trunk. Usually this complex impedance is
used as two guesses to hit two downstream targets, often either complex 1=zn in a HARDEND
at the end of a dead-ended teebranch or complex p1 at a subsequent UNION for a multiply-
connected teebranch. Unless ANCHOR appears upstream, HtotBr tells DeltaEC how to split
up the total power H_ tot between the branch and the trunk. This is often used as a guess
to hit a downstream temperature target, such as at a UNION. The parameter NdotBr tells
DeltaEC how to split any nonzero time-averaged molar ‡ow N_ between the branch and the
trunk. Similarly, NLdotB speci…es the portion of the time-averaged mole ‡ux of the lighter
component of a mixture, N_ L ; that goes into the branch.
For duct networks in which temperature is constant and p1 and U1 are the only integration
variables of interest, use ANCHOR after the BEGIN segment, so TBRANCH will not worry about
how to split H_ tot and UNION will not think of temperature as a potential default target.
Nested teebranches are allowed, as shown in Figs. 11.1(b) and (c). However, never use a
BEGIN segment within a teebranch to re-initialize variables.
The incoming values of the RPN variables F1 and vol are retained for the trunk, and
these variables are initialized to zero for the teebranch.
Input variables:
Re(Zb) (Pa s/m3 ) The real part of the branch impedance Zbr :
Im(Zb) (Pa s/m3 ) The imaginary part of the branch impedance Zbr :
HtotBr (W) This optional input variable, H_ br ; can be used in insulated mode. It speci…es
how much total power ‡ows into the teebranch. The default value is equal to E_ br +
N_ br mwm ; which is simply E_ br when there is no steady ‡ow.
NdotBr (mol/s) This input variable, N_ br ; appears when “Enable Ndot” is checked in an
upstream BEGIN. It speci…es how much steady ‡ow goes into the teebranch.
NLdotB (mol/s) This input variable, N_ L;br ; appears when “Enable Ndot”is checked in an
upstream BEGIN and when a gas mixture and MIX** segments are in use. It speci…es
how much steady ‡ow of the light component of the mixture goes into the teebranch.
The default value represents bulk ‡ow into the branch, N_ L;br = nL N_ br .
Potential targets:
None.
Master–slave links:
The imaginary part of the branching impedance can be linked to the real part in two possible
ways: to preserve either the magnitude or the phase while the real part is changed.
(For an introduction to master–slave links, and a complete listing of these link options
for all segments, see Section 2.7.2.)
230
Format and examples:
TBRANCH the fork in the road, in the default "insulated" mode
4.412E+07 a Re(Zb) Pa-s/m^3 G
-3.528E+06 b Im(Zb) Pa-s/m^3 G
0.49 c HtotBr G
TBRANCH the fork in the road, if "ANCHOR" segment appears somewhere upstream
4.412E+07 a Re(Zb) Pa-s/m^3 G
-3.528E+06 b Im(Zb) Pa-s/m^3 G
Calculations:
The TBRANCH segment leaves Tm ; p1 ; and p20HL unchanged. Other variables change according
to
F1;br = 0; (11.11)
F1;trunk = F1;in ; (11.12)
volbr = 0; (11.13)
voltrunk = volin : (11.14)
TBRANCH has an extra output, EdotTr, to display the acoustic power ‡owing past the branch
in the trunk.
11.2.2. UNION
General description:
A UNION is used in concert with a TBRANCH and a SOFTEND somewhere upstream, to create
loops of segments as illustrated in Fig. 11.1(c). In this arrangement, when it encounters
a TBRANCH, DeltaEC treats subsequent segments as the sequential members of a branch
until it reaches a SOFTEND; then it “returns to the trunk,”treating the rest of the segments
as trunk members. Subsequently, a UNION segment in the trunk tells DeltaEC where to
connect the teebranch’s SOFTEND back to the trunk.
When reconnected with a UNION, the teebranch’s SOFTEND potential impedance targets
should not be used; instead, enable the UNION’s targets to ensure that p1 (complex) at the
UNION is equal to p1 (complex) at the SOFTEND of the teebranch. Often, guessing the tee-
branch impedance determines how the (complex) volume ‡ow rate splits up at the TBRANCH
and allows the complex p1 target at the UNION to be met.
231
In all but the simplest acoustic duct network, additional potential targets can be used at
a UNION to properly tie the teebranch’s SOFTEND to the trunk at the UNION with continuity
of relevant variables. These include Tm ; p2;0;HL ; and nL :
UNION targets are unusual, because their target values are dynamically rewritten by
DeltaEC during each iteration, so they always re‡ect the current result at the referenced
SOFTEND. (When the user enables a UNION’s target, the target input parameters, e.g., magni-
tude and phase of pressure, can be initialized to any values, because DeltaEC will overwrite
them during each integration pass with the current magnitude and phase of pressure at the
referenced SOFTEND via automatic sameas links from the UNION to the SOFTEND.)
Oscillating volume ‡ow rates are added at the UNION, as are time-averaged energy ‡ows
and mole ‡uxes. The RPN variables F1 and vol are also added.
When TBRANCH and UNION are used for duct networks, where temperature is constant
and hence p1 and U1 are the only variables of interest, an ANCHOR segment upstream of
the TBRANCH will eliminate confusion that might arise from the display of uninteresting,
irrelevant variables.
Input variables:
SegNum The segment number of a SOFTEND, upstream of the UNION, which terminates a
teebranch and which should be connected to the trunk here. See Fig. 11.1(c).
Potential targets:
jpjSft (Pa) If used as a target, DeltaEC’s shooting method compares its calculation of
jp1 j here in the trunk with its calculation of jp1 j at the SOFTEND identi…ed by SegNum.
DeltaEC updates the value of this variable continuously.
Ph(p)S (deg) If used as a target, DeltaEC’s shooting method compares its calculation of
the phase of p1 here in the trunk with its calculation of the phase of p1 at the SOFTEND
identi…ed by SegNum. DeltaEC updates the value of this variable continuously.
TSoft (K) This potential target appears in insulated mode. If used as a target, DeltaEC’s
shooting method compares its calculation of Tm here in the trunk with its calculation of
Tm at the SOFTEND identi…ed by SegNum. DeltaEC updates the value of this variable
continuously.
p20HLS (Pa) This potential target appears when “Enable Ndot”is checked in an upstream
BEGIN. If used as a target, DeltaEC’s shooting method compares its calculation of
p2;0;HL here in the trunk with its calculation of p2;0;HL at the SOFTEND identi…ed by
SegNum. DeltaEC updates the value of this variable continuously.
nLSoft This potential target appears when a gas mixture is in use and any MIX** segment
appears in the model. If used as a target, DeltaEC’s shooting method compares its
calculation of nL here in the trunk with its calculation of nL at the SOFTEND identi…ed
by SegNum. DeltaEC updates the value of this variable continuously.
232
Master–slave links:
None.
!--------------------------------- 13 ---------------------------------
UNION reconnect the "final" end of the loop, segment 8, here
8.000 a SegNum 2.5420E+05 A |p| Pa
2.5420E+05 b |p|Sft Pa = 13A? 1.9444 B Ph(p) deg
1.9444 c Ph(p)S deg = 13B? 3.0515E-02 C |U| m^3/s
781.66 d TSoft K (t) -44.624 D Ph(U) deg
1603.6 e p20HLS Pa = 13H? 2674.7 E Htot W
2666.4 F Edot W
702.66 G T K
1603.6 H p20HL Pa
UNION connect two paths together, matching complex pressure without worrying about any other variables
4.000 a segment number of SOFTEND of the TBRANCH somewhere upstream
3.e4 b |p|Sft Pa
1.234 c ph(p)S deg
Calculations:
Calculations at UNION are simple arithmetic:
where the subscript “in” refers to the closest trunk segment upstream of the UNION (i.e.,
almost always simply the segment immediately preceding the UNION) and the subscript
“end”refers to the SOFTEND that is identi…ed by SegNum in the …rst input parameter of the
UNION, i.e., the SOFTEND that terminates the teebranch that is being joined to the trunk at
the UNION. Note that Eq. (11.18) implies H_ 2;k;out = H_ 2;k;in + H_ 2;k;end only if Tm is successfully
targeted.
In thermally anchored mode, Eq. (11.18) is replaced by H_ tot;out = E: _
11.3. Insulation
11.3.1. ANCHOR and INSULATE
General description:
By default, DeltaEC assumes that the only places total power can ‡ow into or out of a
model are
233
heat exchangers, where heat can be added to the total power (or, if heat is negative,
subtracted from the total power),
BEGIN, unless H_ tot is explicitly set equal to zero (or is set equal to E_ + mwm N_ which in
turn happens to equal zero),
Other segments— stacks and regenerators, ducts and cones, and compliances, surfaces, and
impedances— are assumed to have insulation wrapped around their side-wall boundaries by
default, so whatever total power H_ tot ‡ows into them from the previous segment must ‡ow
out of them to the subsequent segment.
However, sometimes it is useful to change the behavior of ducts, cones, compliances,
surfaces, and impedances so the total power ‡owing out of them is equal to the acoustic
power out of them. In a series of such segments, acoustic power dissipation in each segment
“disappears” from the model, as it might if the segment were thermally anchored to a
heat sink at the same temperature as the local gas temperature Tm : This behavior can be
started by an ANCHOR segment before the …rst of a series of segments that are supposed to be
thermally anchored. Thereafter, such segments obey H_ 2;k;out = E_ out . The complementary
INSULATE segment, or a BEGIN segment, can be used to return the computation to the
default, thermally insulated mode in which H_ 2;k;out = H_ 2;k;in . ANCHOR is most useful for very
low power systems and simple systems where total energy ‡ow is irrelevant, and of course
for components that actually are immersed in a heat sink.
Stacks and regenerators remain insulated when ANCHOR is in e¤ect.
Some users who are interested only in acoustics, not thermoacoustics, habitually put
ANCHOR immediately after BEGIN, to reduce irrelevant clutter.
Use the default, insulated mode for ducts, compliances, or impedances that come be-
tween stacks and heat exchangers, and for modeling insulated acoustic networks attached to
thermoacoustic devices.
234
Table 11.1: List of RPN number-entry methods. Each number is piled on top of the stack.
11.4.1. RPN
General description:
Results of the RPN segment are computed by interpreting the formula (line “b”) in Reverse
Polish Notation, as described in [17, 18] and especially in all of Chapter 4. The formula
is a sequence of numbers, addresses, internal DeltaEC variables, and algebraic operators.
Variables such as Htot can be typed as lower case; DeltaEC will automatically convert to
its preferred case. The formula in an RPN should not be longer than 80 characters. As with
RPN calculators, when operators do not consume all numbers on the “stack,”more than one
output is generated— a feature that can be exploited to display multiple results. The stack,
whose members may be complex numbers, grows downward, from A–J. Valid operators and
inputs are summarized in Tables 11.1–11.6. To display an RPN formula with parentheses,
click on the output parameter, and then click on List Linkages.
235
Table 11.2: RPN variable-entry methods. Each number is piled on top of the stack.
237
Table 11.4: Ordinary RPN binary operations and functions. Each operation or function removes
two numbers from the top of the stack, combines them, and puts the result on top of the stack. In
this table, “x”refers to the number on top of the stack, and “y”refers to the number second from
the top of the stack.
Table 11.5: RPN actions that have no e¤ect on the RPN stack, but have an e¤ect elsewhere.
238
Table 11.6: RPN actions that a¤ect the top two numbers on the RPN stack.
during the course of its other integrations. The derivation in Chapter 14 shows that this
integral gives the total force amplitude exerted by the gas on the pressure vessel. The result
is accessible to the user at any location by putting “F1”in line b of an RPN segment, yielding
the complex number F1 and making it available for further RPN calculations. (As usual,
lower-case typing is accepted.)
DeltaEC initializes F1 to zero at BEGIN and at the beginning of the teebranch of every
TBRANCH. The trunk’s F1 and the referenced-SOFTEND’s F1 are added in UNION. Any other
segment adds its own F1 contribution [according to Eq. (11.23)] to whatever total F1 has
accumulated before it.
These defaults can be overridden via RPN segments that include the change command
“=F1”. The defaults must often be overridden for thermoacoustic systems with elbows or side
239
branches sticking out at di¤erent angles, to resolve the components of the force-amplitude
vector in di¤erent directions and to take care of minus signs in toroidal acoustic paths.
For a 90 miter elbow, split the elbow into two segments. Displaying and then rezeroing
the force variable between those segments (with “F1” and “=F1”, respectively) is a useful
approach. For a smoothly radiused 90 elbow, strictly speaking we ought to do an integral
of a vector velocity around the corner to obtain a vector force. This is beyond DeltaEC’s
current capability. However to the extent that U1 does not vary from one end of such
an elbow to the other, that integral yields simply 2= times DeltaEC’s full-elbow DUCT
integral for the vertical direction and 2= times DeltaEC’s full-elbow DUCT integral for the
horizontal direction. The user who requires F1 in such an elbow can use RPN calculations
to implement the factor of 2= ; as shown in the second example below.
R
For IESPEAKER and VESPEAKER segments, DeltaEC evaluates m U1 dx in Eq. (11.23)
as simply the speaker’s moving mass times its velocity phasor.
Example 1
To …nd the oscillating velocity of a freely suspended, 14-kg pressure vessel enclosing a
half-wavelength resonance driven by a noncompliant transducer at the far end, use this series
of segments.
!--------------------------------- 0 ---------------------------------
BEGIN no transducer at this end
2.000E+06 a Mean P Pa
173.53 b Freq Hz
300.00 c TBeg K
1.0000E+05 d |p| Pa
0.0000 e Ph(p) deg
0.0000 f |U| m/s
0.0000 g Ph(U) deg
!--------------------------------- 1 ---------------------------------
SURFACE first end
sameas 2a a Area m^2
!--------------------------------- 2 ---------------------------------
DUCT
1.0000E-02 a Area m^2
0.3545 b Perim m
1.0000 c Length m
!--------------------------------- 3 ---------------------------------
SURFACE noncompliant transducer at this end
sameas 2a a Area m^2
!--------------------------------- 4 ---------------------------------
HARDEND
0.0000 a R(1/z) (t)
0.0000 b I(1/z) = 4H?
!--------------------------------- 5 ---------------------------------
RPN (a)=mass of pressure vessel (kg). [A]= velocity phasor of pressure vessel (m/s).
14.000 a G or T (t) (-8.8680E-05, 0.1309 ) m/s
F1 5a / i / w /
!--------------------------------- 6 ---------------------------------
RPN viscous drag force phasor (newtons)
0.0000 a G or T (t) ( 1.3536 , -1.3536 ) N
F1 p1 3a * 0d 1a * - -
guessz 0b
targs 4b
We arbitrarily de…ned zero phase to be that of the pressure at segments 0 and 1. Segment
5 shows that the phase of the velocity is almost 90 , which is consistent with the phases of
the pressure on the two ends of the half-wavelength resonator at segments 1 and 3. Segment
6 subtracts the normal end-cap forces from the total force to display what’s left over, which
must be the integrated viscous shear force on the side walls.
Example 2
To …nd oscillating forces and torques on the half-wavelength refrigerator shown in Fig. 11.2,
we use a DeltaEC model like this:
240
Figure 11.2: Example 2. Left, a sketch of a half-wavelength, two-speaker, standing-wave refriger-
ator. Right, force diagram, in which the directions of the arrows indicate the “positive” direction
of calculated forces. In both parts of the …gure, “c.g.” indicates the center of gravity.
Most of the input and output have been eliminated from the listing for clarity, leaving
only the relevant RPN segments fully displayed. Segment 13 captures F1 for the vertical
components, segment 15 captures F1 for the elbow using the approximation that U1 is
uniform in the elbow, and segment 19 captures F1 for the horizontal part of the resonator.
Then segment 20 adds up the previous results to yield the net horizontal and vertical F1 s
for this half of the full refrigerator.
Considering the symmetry of the full refrigerator as shown in the …gure (only half of
which is modeled above with DeltaEC ) to get the minus signs and factors of two correct,
we see that the total horizontal force exerted on the pressure vessel is 2 [20B], the total
241
vertical force is zero, and the total counter-clockwise torque around the center of gravity
“c.g.” is 2 ([20A]LH + [20B]LV ) where LH and LV are distances between the center of
gravity and the axes of the parts of the refrigerator as shown in the …gure. Both the total
horizontal force and the torque have time phases of about 90 , relative to the time phase
of the velocity at the mirror-image plane, which is set to 0 with segment 18.
The vol operator in segment 16 causes the cumulative volume to be displayed: The total
volume of gas in the DUCTs, HXs, STKSLAB, and CONE. (The other segments in this example
have no gas volume.)
11.4.2. VOLUME
General description:
Result A of a VOLUME segment is the sum of the volumes in all duct, cone, stack, compliance,
and heat exchanger segments beginning with BegAddr and ending with EndAddr (parameter
letters are inconsequential). Porosity is ignored in calculating this volume, so the result
is the volume of the gas plus the volume of internal …ns, plates, etc. in heat exchangers
and stacks. (In contrast, the vol variable, accessible in RPN segments, gives the cumulative
242
gas volume, not including internal …ns and plates.) This segment is intended to give an
indication of the overall size of an apparatus for doing trade-o¤ analysis during design.
Input variables:
Potential targets:
Line “a”is the target at which VOLUME output line “A”can be aimed.
Master–slave links:
None.
11.4.3. CONSTANTS
General description:
The CONSTANTS segment allows the user to input a dozen real numbers in one segment.
This is useful when a subsequent RPN segment depends on a large number of nonstandard
parameters that the user must control or that DeltaEC must guess. (Without CONSTANTS,
a dozen RPN segments would be required.) This segment performs no calculations and has
no potential targets.
243
11.5. Interfaces to external programs and …les
11.5.1. BLKDATA
General description:
BLKDATA is a specialized segment designed for users who have a quantity of tabular data that
they want DeltaEC to follow. The data may represent experimental results, the results
from other computational methods, or simply an irregular parameter space to be explored
that is not well served by DeltaEC’s incremental-plot capability (e.g., irregular step size
or parametric plotting). The data …le is in text format, delimited with spaces or tabs, and
it may contain up to 14 columns and any number of lines, all with the same number of
entries. The …rst one or two lines can optionally contain descriptive strings that are read
into the parameters’description and unit …elds, respectively. Only one BLKDATA segment per
model is allowed. The data …le must have the su¢ x <.blk> and its root name is supplied
in the heading of the BLKDATA segment. When the …le is opened, the number of lines and
columns are counted, and initial column headings are read in, if present. The number of
input parameters for the segment is equal to the number of columns in the <.blk> …le.
During a run, each numeric line of the <.blk> …le is read sequentially, and DeltaEC runs
the model once for each line. Each line of the …le is inserted into the input parameters of
the BLKDATA segment, one line per data point. The use of sameas provides a direct means
to tie the values of DeltaEC input parameters to the tabulated data.
Enforcing values that normally appear as DeltaEC outputs is a little more subtle. This
requires an RPN target to be sameas’ed to an input parameter in the BLKDATA segment, so
the RPN’s calculated output value can be forced to hit the targeted value. This target must
be used, of course, and an appropriately related input parameter must be added to the guess
list. Segment 5 in the example below illustrates such a target.
Input variables:
The heading …eld, which is to the right of “BLKDATA,” must contain the root name of the
<.blk> …le. For example, if the user’s data is in <mydata.blk>, the BLKDATA segment
looks like the example below.
244
then DeltaEC will immediately transform the BLKDATA segment to this:
!--------------------------------- 1 ---------------------------------
BLKDATA mydata
511.40 a Freq Hz
6653.0 b |p|@0 Pa
3.1000E-02 c SkrAmp A
and will read subsequent lines of <mydata.blk> into place in subsequent iterations of the
run. If the …le <mydata.blk> is a log of experimental data, we can ask DeltaEC to
emulate the experimental conditions by inserting appropriate sameas statements, as in the
following fragment that matches the experimental applied frequencies, pressure amplitudes,
and speaker currents:
!--------------------------------- 0 ---------------------------------
BEGIN the setup
7.8000E+04 a Mean P Pa
sameas 1a b Freq Hz
300.00 c TBeg K
sameas 1b d |p|@0 Pa
0.0000 e Ph(p)0 deg
0.0000 f |U|@0 m^3/s
0.0000 g Ph(U)0 deg
air Gas type
ideal Solid type
!--------------------------------- 1 ---------------------------------
BLKDATA mydata
511.40 a Freq Hz
6653.0 b |p|@0 Pa
3.1000E-02 c SkrAmp A
+ 2 SURFACE cavity back surface
+ 3 DUCT back duct
!--------------------------------- 4 ---------------------------------
VESPEAKER ...using an IESPEAKER would be simpler,
! but would not illustrate BLKDATA plus target.
1.5000E-04 a Area m^2 68.286 A |p| Hz
4.6000 b R ohms -91.445 B Ph(p) deg
5.5000E-05 c L H 5.2965E-06 C |U| m^3/s
3.3000 d BLProd T-m -90.455 D Ph(U) deg
4.0000E-04 e M kg 1.8081E-04 E Hdot W
8560.0 f K N/m 1.8081E-04 F Edot W
0.8000 g Rm N-s/m 1.8597E-03 G EdotIn W P
0.1945 h |V| V 0.1945 H Volts V
243.00 i Ph(V) deg 2.2615E-02 I Amps V
32.263 J Ph(V/I) deg
sameas 0 Gas type 121.99 K |Px| Pa
ideal Solid type -145.97 L Ph(Px) deg
!--------------------------------- 5 ---------------------------------
RPN use to match speaker current output to measured value
sameas 1c a = 5A? 2.2615E-02 A
4I
11.5.2. SYSEXEC
General description:
The SYSEXEC segment allows DeltaEC to send data to an external program via the com-
puter’s …le-handling system, execute that external program, and bring any results of that
external program back into DeltaEC. The use of sameas allows any input or output num-
bers in DeltaEC to be brought into the SYSEXEC segment’s input parameters so they can
be sent to the external program, and subsequently allows any DeltaEC input parameters
or RPN tokens to be set equal to results of the external program.
The …rst action taken by SYSEXEC is to write all of its 14 real input parameters to a text
…le called <ToSys.dat> in single-column format. SYSEXEC then pauses DeltaEC execution
and runs an external program using the entire text string speci…ed in the heading of the
SYSEXEC segment. The external program might or might not make use of the numbers in
<ToSys.dat>. If the external program’s results should be brought back in to DeltaEC, the
245
external program must put those results in a …le named <FromSys.dat> in the same ASCII,
single-column format. (If more than 14 values are required, the leftovers might be read in
through a BLKDATA segment or a second SYSEXEC segment.) When the external program has
…nished running, control returns to DeltaEC, which then reads up to 14 parameters from
the …le <FromSys.dat> if it exists, putting those numbers in the output list of the SYSEXEC
segment, and resumes its numerical integrations.
When the SYSEXEC segment is encountered by DeltaEC during a run, DeltaEC creates
the …le <ToSys.dat> containing the parameters a through n:
3.0000
4.0000
6.0000
7.0000
2.0000
87.000
2.0000
1.0000
10.000
13.000
5.0000
4.0000
11.000
9.0000
and passes the segment’s heading …eld, sort tosys.dat > fromsys.dat, to the operating
system as a command string. Next, the operating system sends the newly created …le
<tosys.dat> to the sort function. The sort command sorts the numbers into ascending
order. The results of the sort are piped into the <fromsys.dat> …le using the ‘greater than’
character, i.e., the ‘>’character, in the command string. Finally, the SYSEXEC segment reads
the <fromsys.dat> …le and puts the results in the output column of the segment, so the
segment ends up looking like this:
! Sort input parameters into output parameters (Windows version)
!-----------------------------------1----------------------------------
RPN a silly example
87.0 a 13.000 A
5 8 +
!--------------------------------- 2 ---------------------------------
SYSEXEC sort tosys.dat > fromsys.dat
3.0000 a In 1 1.0000 A Out 1
246
4.0000 b In 2 2.0000 B Out 2
6.0000 c In 3 2.0000 C Out 3
7.0000 d In 4 3.0000 D Out 4
2.0000 e In 5 4.0000 E Out 5
sameas 1a f In 6 4.0000 F Out 6
2.0000 g In 7 5.0000 G Out 7
1.0000 h In 8 6.0000 H Out 8
10.000 i In 9 7.0000 I Out 9
sameas 1A j In 10 9.0000 J Out 10
5.0000 k In 11 10.000 K Out 11
4.0000 l In 12 11.000 L Out 12
11.000 m In 13 13.000 M Out 13
9.0000 n In 14 87.000 N Out 14
Then an RPN or sameas in a subsequent segment that points to output 2N will access the
largest number in the original list, which was 87.000 in this example.
As an entertaining exercise, put a SYSEXEC segment at the end of a model, sameas some
of the SYSEXEC inputs to all of the guesses in the model, sameas other SYSEXEC inputs to
all of the targeted results in the model, and write a little batch …le or python script (if
python exists on your computer) to append the numbers in <ToSys.dat> as a single line
in another …le. Put the name of that batch …le or python script in the SYSEXEC’s header.
Ignore <FromSys.dat>. Running the model will then produce a nice table of the values of
all the guesses and targeted results during each pass of the shooting method, showing the
details of what DeltaEC’s shooting method is actually doing.
The SYSEXEC segment can also be used to “insert” a user-created segment in a model.
Use <ToSys.dat> to send pm and f (both from segment 0) and integration variables such
as p1 , U1 , and Tm (from the preceding segment) to the operating system, run your own
program to change the values of integration variables such as p1 , U1 , and Tm , return your
new values of p1 , U1 , and Tm to DeltaEC in <FromSys.dat>, and …nally use a subsequent
RPN segment with tokens such as =Tm to pass your custom-calculated values to the next
segment in DeltaEC.
247
12. Gases (and liquids)
248
k0he=0.0025672*ta**0.716
amuhe=0.412e-6*ta**0.68014
k0ar=(1.39e-4*ta**0.852-1.5e-8*(ta-300.)*(ta-300.))*(1.+2.e-8*pm)
amuar=(1.77e-7*ta**0.852-25.e-12*(ta-300.)*(ta-300.))*(1.+2.e-8*pm)
k0=x1*k0ar+x2*k0he-(k0ar+k0he)*x1*x2**1.5
mu=x1*amuar+x2*amuhe+0.2*(amuar+amuhe)*x1*x2
diff12= 7.49d0*(1.d0 - 0.05d0*x2)*(ta/295.d0)**1.66d0/pm
ktherm= 0.38d0 * x2**1.2d0 * x1**0.8d0
249
For humid air, DeltaEC uses the ideal-gas equation of state for v and a slightly non-
linear temperature dependence for w (due to the T dependence of cp for water vapor in
the ASHRAE tables). Also in accordance with ASHRAE recommendations, we set and k
equal to their dry-air values.
In wet air, in accordance with Hiller’s measurements [73], the oscillating thermodynamics
in DeltaEC is that of humid air at saturation, while the mean ‡ow thermodynamics in-
cludes the enthalpy of the liquid when N_ 6= 0. In other words, if there is no mean ‡ow, the cal-
culation proceeds exactly as for an ideal-gas mixture and ignores the condensate, but if mean
‡ow is nonzero DeltaEC performs stack integrations with ideal-gas-mixture properties in
the momentum and continuity equations but with the heat of condensation/evaporation
(plus the small enthalpy of the condensate itself) included in the energy equation as it is
integrated to …nd dTm =dx: In all cases, we ignore the dynamic e¤ects of oscillating di¤usion
of the water vapor through the air, which have been described by Raspet [74] and Slaton
[75].
Although the enthalpy of wet air calculated in DeltaEC’s gas-properties algorithm
includes the latent heat of freezing and melting as the condensate passes through 0 C, we
have not yet incorporated this latent heat, of the liquid-to-solid phase transition, into the
numerical integrations in STK** segments, so integrating through Tm = 0 C in wet air is of
dubious value. (Anyway, we have not thought of any good reason to use wet air in DeltaEC
below 0 C, because we believe the stack would simply plug up with ice.)
The saturated vapor pressure is accurate from 100 C to 370 C; other properties are
reasonably accurate from 50 C to 150 C.
Ideal gas approximation for equation of state and speci…c heat. Transport from Ref. [70].
k0=0.0003609*ta**0.7512
mu=0.3577e-6*ta**0.6885
Ideal gas approximation for equation of state and speci…c heat. Transport from Ref. [70].
k0=0.002627*ta**0.744
mu=0.19361e-6*ta**0.6723
Ideal gas approximation for equation of state and speci…c heat. Transport from Ref. [70].
k0=0.002795*ta**0.686
mu=0.2726e-6*ta**0.6721
250
12.10. Carbon dioxide (CO2)
These equations provide a 1% match to the webbook.nist.gov/chemistry/‡uid/ data from
300 K to 1100 K at 1 bar, 5 bar, and 10 bar.
DATA cpco2/ 560.0d0, 1.13d0, -0.00045d0, 4.0d-9, 1.5d0, 100.0d0 /
DATA aco2/ -0.015d0, 7.0d0, 1.0d-5, 400000.0d0 /
DATA rhoco2/ 4.3d-9, 1.7, 70.0 /
DATA kco2/ 2.8d-5, 1.1395d0, 4.0d-4, 4.0d-10, -2.2d-8, 625.0d0 /
DATA muco2/ 1.4187d-7, 0.8216d0, 1.5416d-7, 0.8094d0, -8.0d-12, 650.0d0, 8.0d-7 /
oldgam=1.2857d0
massavg=0.04401d0
r=runiv/massavg
cp=cpco2(1)+cpco2(2)*ta+cpco2(3)*ta**2+cpco2(4)*pm**cpco2(5)*ta*dexp(-ta/cpco2(6))
enth=cpco2(1)*(ta-tzero)+cpco2(2)/2.*(ta**2-tzero**2)+cpco2(3)/3.*(ta**3-tzero**3)
& -cpco2(4)*pm**cpco2(5)*cpco2(6)
& *(dexp(-ta/cpco2(6))*(1+ta/cpco2(6))-dexp(-tzero/cpco2(6))*(1+tzero/cpco2(6)))
entro=cpco2(1)*dlog(ta/tzero)+cpco2(2)*(ta-tzero)
& +cpco2(3)/2.*(ta**2-tzero**2)-cpco2(4)*pm**cpco2(5)*cpco2(6)
& *(dexp(-ta/cpco2(6))-dexp(-tzero/cpco2(6)))
a=dsqrt(oldgam*r*ta)+aco2(1)*(1+dexp(-pm/aco2(4)))*ta+aco2(2)-aco2(3)*pm
rho=pm/r/ta+rhoco2(1)*pm**rhoco2(2)*dexp(-ta/rhoco2(3))
beta=(pm/r/ta**2+rhoco2(1)/rhoco2(3)*pm**rhoco2(2)*dexp(-ta/rhoco2(3)))/rho
gamma=ta*beta**2 * a**2/cp + 1.0d0
k0=kco2(1)*ta**kco2(2)+kco2(3)+kco2(4)*pm+ kco2(5)*(ta-kco2(6))**2
mu=muco2(1)*ta**muco2(2)+(pm/1.0d6-1.0d0)*(muco2(3)*ta**muco2(4)
& -muco2(1)*ta**muco2(2))+muco2(5)*(ta-muco2(6))**2+muco2(7)
251
goto 900
real*8 function gasprop(ta,a,b,c,d,e,f)
real*8 z,ta,a,b,c,d,e,f
z=(ta-1400)/200
gasprop=a+z*(b+ z*(c + z*(d +z*(e+f*z))))
return
end
252
bp=bp+ta*beta**2*(ct1+2.*ct2*t1)/rho/cp/cp
cpp=-ta*(beta**2+bt)/rho
c So far, everything is evaluated at p=0.
a=a+ap*pm
rho=rho+rp*pm
beta=beta+bp*pm
cp=cp+cpp*pm
gamma=1.+ta*beta**2*a**2/cp
mu=e2*rho**(1./3.)*exp (e1*rho/ta)
To employ a user-de…ned gas, select its …lename as the gas type in a BEGIN segment.
253
User-de…ned gases speci…ed by <.tpf> …les cannot be used with nonzero N_ , because Eq.
(6.1) requires m, which is not available in <.tpf> format. To circumvent this shortcoming
when a user-de…ned gas must be used with nonzero N_ , use a <.tpm> …le as described in
the next Section, and set the mole fraction of the unused component to zero in the BEGIN
segment of the model.
254
Equation (12.8) is a compromise between simplicity and ‡exibility. It is intended for use
with most coe¢ cients set to zero in a variety of simple expressions. Finally, the numbers in
lines 6 through 9 are used in these equations:
As an example of the use of these coe¢ cients, consider the example below for neon, a mixture
of two isotopes.
! user-defined mixture: neon, a mixture of two isotopes, 20 and 22.
! Natural abundance is nL = 0.91
! mL(kg/mole) mH(kg/mole) gammaL gammaH:
0.01999 0.02199 1.6667 1.6667
! k pure L (W/m-K):
0. 0. 0. 0. 0. 0.00122 0.65
! k pure H (W/m-K):
0. 0. 0. 0. 0. 0.00116 0.65
! mu pure L (kg/m-s):
0. 0. 0. 0. 0. 7.41E-7 0.66
! mu pure H (kg/m-s):
0. 0. 0. 0. 0. 7.70E-7 0.66
! k mixture (W/m-K):
0.
! mu mixture (kg/m-s):
0.
! D12 (m2/s):
0.53E-4 0. 1.72
! kT:
0.0267 1.0 1.0
To employ a user-de…ned mixture, select its …lename as the gas type in a BEGIN segment.
255
13. Solids
256
Prior to version 3.5b2, DeltaE’s stainless steel properties were very inaccurate at cryo-
genic temperatures.
257
To request a user-de…ned solid, simply use the root …le name as the solid type. The
<.tpf> …le should be in the same directory or folder as the model …le. If the name matches
any pre-de…ned solid name, the (constant) user-de…ned properties will replace DeltaEC’s
internal calculations.
User-de…ned gases and user-de…ned solids share the <.tpf> …le extension. DeltaEC
can tell the di¤erence between a solid and a ‡uid via the number of non-comment lines in
the …le.
258
Back Matter
259
14. Derivations
Here we present our derivations of a few equations that are used in DeltaEC but are not
derived in the archival thermoacoustics literature.
The right-hand side of this equation is what DeltaEC’s RPN force operator, F1, calculates;
this operator is described in Section 11.4.1. If F1 = 0; the vessel does not vibrate. Otherwise,
if the vessel is freely suspended and we want to learn its acceleration amplitude, we simply
divide by its mass M ; if we want its velocity amplitude, we divide by i!M ; its displacement
amplitude, we divide by ! 2 M:
Equation (14.1) includes both the normal force exerted by gas pressure on the vessel and
the shear force exerted by viscosity and gas velocity on the vessel. Newton’s laws say that
this must be true, but it can also be veri…ed directly, as follows. Calculus shows that
@ @v @
( v) = +v : (14.2)
@t @t @t
Switching to “sum over repeated indices”notation (as in Ref. [36]), the continuity equation
is
@ @
= ( vk ) (14.3)
@t @xk
and the momentum equation is
@vi @vi @p @ 0
= vk + ik ; (14.4)
@t @xk @xi @xk
where 0ik is the viscous stress tensor. Substituting Eqs. (14.3) and (14.4) into (14.2) and
doing a little algebra to combine terms yields
@ @ 0
( vi ) = ( vi vk + p ik ik ) : (14.5)
@t @xk
261
Now integrate over the whole volume of gas, and use Green’s formula to convert the volume
integral of the divergence on the right side into a surface integral:
Z I
@ 0
( vi ) dV = ( vi vk + p ik ik ) dSk ; (14.6)
@t
where dV is a di¤erential element of the volume and dSk is a vector normal to an element
of surface with area dS: The velocity is zero on the unmoving surface, so
Z I
@ 0
( vi ) dV = (p ik ik ) dSk : (14.7)
@t
The right side is the total force exerted on the gas by the solid vessel bounding the gas. It
is the sum of the pressure forces and the viscous forces. The x component of the left side is
Eq. (14.1), which is what DeltaEC’s F1 operator calculates.
1
pminor = K(t) [ m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 ) jju1 j cos !t + u2;0 j (14.9)
2
Z2
1
= K(t) [ m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 ) jju1 j cos !t + u2;0 j d(!t):
4
0
(14.10)
262
In the interval 0 < !t < 2 ; the argument of the absolute value changes sign at
!t = Arccos( u2;0 = ju1 j) 0 and !t = 2 Arccos( u2;0 = ju1 j) = 2 0 if ju2;0 =u1 j < 1;
at these zero crossings, the value of K changes from K+ to K and back again. Thus
0 1
Z0 2Z 0
1 @ A [ m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 )2 d(!t)
pminor = K+ K
4
0 0
(14.11)
if u2;0 = ju1 j < 1: Evaluating this (e.g., with the symbolic-mathematics software Mathemat-
ica) yields
m ju1 j2 j 1j
pminor = (K K+ ) 1 + 2"22;0 + 2"2;0 cos
4 2 m
j 1j
(K+ + K ) 1 + 2"22;0 + 2"2;0 cos sin 1
"2;0
m
4 j j q
+ 3"2;0 + 1 + "22;0 =2 1 cos 1 "22;0 (14.12)
3 m
m ju1 j2 j j
pminor = K+ 1 + 2"22;0 + 2"2;0 1 cos for "2;0 1; (14.13)
4 m
m ju1 j2 j 1j
= K 1 + 2"22;0 + 2"2;0 cos for "2;0 1: (14.14)
4 m
263
The time average of the …rst term on the right-hand side, to second order, is
c1
m u2;0 : (14.18)
8rh2
Taking the time average of the second term involves the integral
Iscreen = [ m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 ) jju1 j cos !t + u2;0 j (14.19)
Z2
1
= [ m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 ) jju1 j cos !t + u2;0 j d(!t):
2
0
(14.20)
In the interval 0 < !t < 2 ; the argument of the absolute value near the end of the equation
changes sign at !t = Arccos( u2;0 = ju1 j) 0 and !t = 2 Arccos( u2;0 = ju1 j) = 2 0
if ju2;0 =u1 j < 1; so we can write
0 1
Z0 2Z 0
1 @ A[
Iscreen = m + j 1 j cos (!t + )] (ju1 j cos !t + u2;0 )2 d(!t): (14.21)
2
0 0
ju1 j2 nh q i
m
Iscreen = 1+ 2"22;0 1
"2;0 + 3"2;0 1
sin "22;0
j 1j 2 + "22;0 q
+2 cos "2;0 sin 1 "2;0 + 1 "22;0 : (14.22)
m 3
dp2;0;HL c1 c2 jp1 j
= m u2;0 m ju1 j2 1 + 2"22;0 + 2"2;0 cos sin 1
"2;0
dx 8rh2 2 rh pm
2 jp1 j q
+ 3"2;0 + 2 + "22;0 cos 1 "22;0 (14.23)
3 pm
for j"2;0 j 1:
In SX, j"2;0 j 1 may sometimes be of interest. In that case,
dp2;0;HL c1 c2 jp1 j
= m u2;0 m ju1 j2 sign ("2;0 ) 1 + 2"22;0 + 2"2;0 cos : (14.24)
dx 8rh2 4rh pm
264
We follow the outline in the appendix of Ref. [42], but we change the time dependence
from sine to cosine, to be consistent with other results in this section, and we keep some
dependence on 1 by writing Eq. (A6) in Ref. [42] as
dp m + j 1 j cos(!t + )
= fM (t) u(t) ju(t)j ; (14.25)
dx 2D
where u = U=S in Ref. [42]. Using Eq. (A4) in Ref. [42] for fM yields
0 1
2 Z0 2Z 0
dp2;0;HL m ju1 j 1 @ A fM;max + dfM NR;max jcos !t + "2;0 j
= 1
dx 2D 2 dNR 1 + j"2;0 j
0 0
j 1j
1+ cos (!t + ) (cos !t + "2;0 )2 d(!t); (14.26)
m
where "2;0 = u2;0 = ju1 j and 0 = Arccos( u2;0 = ju1 j) as for the STKSCREEN derivation above.
We neglect ju1 j2 NR;max j 1 j because it is fourth order. Then
0 1
2 Z0 2Z 0
dp2;0;HL m ju1 j fM;max @ A 1 + j 1 j cos (!t + ) (cos !t + "2;0 )2 d(!t)
=
dx 2D 2 m
0 0
0 1
2 Z 0 2Z 0
m ju1 j dfM NR;max @ A jcos !t + "2;0 j 1 (cos !t + "2;0 )2 d(!t):
2D dNR 2 1 + j"2;0 j
0 0
(14.27)
The integral in the …rst line was evaluated above, in Eqs. (14.21)–(14.22). The integral in
the second line was evaluated to obtain the bottom line in Eq. (A10) in Ref. [42]. Pulling
all the pieces together, and using 1 = m = p1 = pm for adiabatic oscillations, yields
dp2;0 ju1 j2 nh q
m 2 1
= 1 + 2"2;0 sin "2;0 + 3"2;0 1 "22;0 fM;max
dx 2 D
2 jp1 j 2 + "22;0 q
+ cos "2;0 sin 1 "2;0 + 1 "22;0
pm 3
3=2 + "22;0 q dfM
+ "2;0 3"2;0 1 "22;0 1 + 2"22;0 sin 1 "2;0 NR;max ;
1 + j"2;0 j dNR
j"2;0 j 1; (14.28)
2
dp2;0 m u2;0 1 jp1 j cos
= sign(u2;0 ) 1+ 2
+ fM;max
dx 2D 2"2;0 pm "2;0
2"22;0 2 j"2;0 j + 1 dfM
NR;max ; j"2;0 j 1: (14.29)
2"22;0 (1 + j"2;0 j) dNR
265
These results appear in Section 10.1.1 for DUCT and 10.1.2 for CONE.
STKPOWERLW and PX
The result for dp2;0;HL =dx in STKPOWERLW and PX is the same as for DUCT and CONE above,
except that we replace pm with pm to model isothermal oscillations.
Using the law of cosines for triangles, we obtain the magnitude of the pole-tip ‡ux
q
j pt;1 j = [L jI1 j]2 + [(Bl) jU1 j =!A]2 2 [L jI1 j] [(Bl) jU1 j =!A] sin (14.34)
266
93.911 i Ph(I) deg G 59.466 I Amps A
70.164 J Ph(V/I)deg
4.3376E+05 K |Px| Pa
174.36 L Ph(Px) deg
ideal Solid type -1273.9 M HeatIn W
!------------------------------------ 16 ---------------
RPN A = IL flux [weber]; B = bEMF flux [weber]; C = alpha [degree]
0.0000 a Target (t) 1.9148 A weber
0.5801 B weber
150.93 C deg
15D 15i - 15d 15C * 2 / pi / 0b / 15a / 15c 15h *
!------------------------------------ 17 ---------------
RPN peak pole-tip flux. [weber = tesla.m^2]
0.0000 a Target (t) 1.7099 A weber
16A sqrd 16B sqrd + 2 16A * 16B * 16C sin * - sqrt
Dividing by pole-tip e¤ective area would turn 17A into the magnetic …eld, in tesla.
Pulsed combustion
Suppose fuel is injected into an air stream in a duct, and combustion occurs. If the injection
and combustion are steady, then conservation of matter applied to a small region around
the combustion zone implies that
N_ out = N_ in + N_ ; (14.35)
where N_ in and N_ out are the molar ‡ow rates of the incoming fresh air and the outgoing com-
bustion products, respectively, and N_ is the additional molar ‡ow rate due to the injection
of the fuel and the change in mole numbers accompanying the combustion. Similarly,
Q_
Tout = Tin + ; (14.36)
mcp N_ out
where Tin and Tout are the temperatures of the incoming and outgoing streams, respectively,
Q_ is the rate at which the combustion releases heat, and m and cp are the average molar
mass and the heat capacity per unit mass of the outgoing stream.
In a thermoacoustic system, the combustion might be time dependent, because of the
pressure dependence of the chemical reaction rate and/or a time dependence in the fuel
injection. Suppose that the region of interest in the duct is small enough that the pressure
is spatially uniform at every instant of time. Let the instantaneous heating rate be q(t);
_ so
the average heating rate must be
2Z =!
!
Q_ = q(t)
_ dt: (14.37)
2
0
Similarly, let the instantaneous rate of increase of moles in the region of interest, due to fuel
injection and combustion, be n(t);
_ whose average must be
2Z =!
!
N_ = n(t)
_ dt: (14.38)
2
0
267
Assuming ideal gases, the local rate of change of temperature due to q(t)
_ is
q(t)
_ 1 q(t)
_
T_ (t) = = ; (14.39)
V m cp V pm =Tm
where V is the volume in which the combustion occurs, and the rate of change of that volume
due to this heating is
Similarly, n(t)
_ contributes a time-dependent volume change
V_ n (t) = m n(t)=
_ : (14.42)
V_ q (t) + V_ n (t)
(t) = ; (14.43)
( _ pm
1) Q=
Z2
1
1 = (!t)ei!t d(!t): (14.44)
0
Depending on the circumstances, the phase of the complex variable might be related to
the local phase of either p1 or U1 :
Thus, the change in DeltaEC’s complex volumetric-‡ow-rate amplitude across the re-
gion of interest is
1 Q_
U1 = 1; (14.45)
pm
and the local increment in acoustic power at the fundamental frequency is
1 Q_ 1 h f i
E_ = Re p1 1 : (14.46)
pm 2
For air, ( 1) = = 2=7; and we might hope for jp1 j =pm 0:2: If j 1 j 1; then the
_
e¢ ciency of conversion of Q to acoustic power could be roughly 3%. This is small, but not
negligible, compared with normal thermoacoustic-engine e¢ ciencies.
(This derivation, new in the 6.3 Users Guide, might bene…t from critical analysis by
users! Feedback will be appreciated.)
268
To implement this in DeltaEC via RPN segments, the user will have to know Q_ and
N_ ; make an estimate of the time dependence for q(t) _ and n(t)
_ (including the time phase
relative to p1 or U1 ), and do the math (or eyeball it graphically) to obtain a numerical value
for 1 : Then put all those values into RPN segments, and use Eqs. (14.35), (14.36), and
(14.45) above to impose local changes in DeltaEC’s integration variables N_ ; Tm ; and U1 :
Here is one way to implement it:
!--------------------------------- 0 ---------------------------------
BEGIN at a pressure node, where fresh air is easily injected.
1.0000E+05 a Mean P Pa
60.000 b Freq Hz
900.00 c TBeg K
0.0000 d |p| Pa
0.0000 e Ph(p) deg
2.0000 f |U| m^3/s
0.0000 g Ph(U) deg
2.0000E-02 i Ndot mol/s
air Gas type
!--------------------------------- 1 ---------------------------------
DUCT pipe from pressure node to somewhere more interesting
3.0000E-02 a Area m^2 Mstr 1.5548E+04 A |p| Pa
0.6140 b Perim m 1a -90.098 B Ph(p) deg
2.5000 c Length m 2.0572E-02 C |U| m^3/s
5.0000E-04 d Srough 119.96 D Ph(U) deg
349.21 E Htot W
ideal Solid type -138.43 F Edot W
!--------------------------------- 2 ---------------------------------
RPN (a) is time-average combustion power (Watt)
! Verify (A) > 1, to ensure enough fresh air. The fuel is methane,
! which has 50 MJ/kg LHV and 0.016 kg/mole. We need 2 moles of O2,
! and so 10 moles of air, per mole of CH4.
1000.0 a G or T 1.6000 A Verify
Ndot inp 50e6 / 0.016 / / 10 /
!--------------------------------- 3 ---------------------------------
BEGIN Change from air to natural-gas combustion products.
sameas 0a a Mean P Pa
sameas 0b b Freq Hz
sameas 0c c TBeg K
sameas 1A d |p| Pa
sameas 1B e Ph(p) deg
sameas 1C f |U| m^3/s
sameas 1D g Ph(U) deg
sameas 0i i Ndot mol/s
NGcbProd Gas type
!--------------------------------- 4 ---------------------------------
RPN (a) is methane’s ratio of DeltaNdot to Qdot (mole/Joule)
! This is the value for methane, assuming H2O stays vapor:
! 3 moles combustion products / moleCH4.
! This segment increments Ndot appropriately.
3.7500E-06 a G or T 2.3750E-02 A mol/s
3.7500E-03 B mol/s
2a inp * # Ndot + =Ndot
!--------------------------------- 5 ---------------------------------
RPN guess Tm increment
1105.8 a G or T G 2005.8 A K
Tm inp + =Tm
!--------------------------------- 6 ---------------------------------
RPN target increment in Htot
sameas 2a a G or T =6A 1000.0 A W
Htot 1E -
!--------------------------------- 7 ---------------------------------
RPN (a) is |Lambda1|, assuming sinusoidal combustion and
! an extra 10% for Vdot sub n. Set phase of Lambda1 same as that of p1.
! (A) is the new U1. (B) is the increment in U1 due to pulsed combustion.
1.1000 a G or T (-1.0278E-02, 1.4681E-02) A m3/s
(-5.3554E-06, -3.1429E-03) B m3/s
gamma 1 - gamma / 2a * pm / p1 arg inp cmplx * # U1 + =U1
!--------------------------------- 8 ---------------------------------
RPN (A) is the acoustic power created by pulsed combustion
0.0000 a G or T 24.433 A W
Edot 1F -
...
...
269
15. Symbols
Variables not found here are intermediate factors, constants, or functions de…ned in the text
close to where they are used.
English letters
A area, m2
a sound speed, m/s
a half of rectangular pore’s width, m
B magnetic …eld, tesla
B …rst virial coe¢ cient, m3 /mol
b a constant or function
b half of rectangular pore’s breadth, m
C compliance, m3 /Pa
C a constant or function
c a constant or function
c heat capacity per unit mass, J/kg K
cp isobaric heat capacity per unit mass, J/kg K
COP coe¢ cient of performance, Q= _ W_ or Q=_ E_
2
D12 binary mass di¤usion coe¢ cient, m /s
D diameter, m
d diameter, m
d di¤erential
E_ acoustic power, W
e 2.71828...
F force, N
F a factor, fraction, friction factor, constant, or function
F a function
f frequency, Hz
f spatially averaged di¤usion function
f friction factor
fM Moody friction factor
H_ rate at which energy ‡ows, W
h heat transfer coe¢ cient per unit area, W/m2 K
270
I electric
p current, amp
i 1
J Bessel function
K spring constant, N
K minor-loss coe¢ cient
ksFrac fudge factor by which (1 )ks is adjusted in STKSCREEN
k thermal conductivity, W/m K
kT thermal di¤usion ratio
k real or complex wave number, m 1
L = k= m cp D12
L electric inductance, henry
L length, m
l length, m
l half of plate thickness, m
M mass, kg
m molar mass, kg/mol (e.g., 0.004002 kg/mol for helium)
m screen mesh number, wires/m
m multiplier for turbulent e¤ects
m0 turbulence multiplier for inertial pressure gradient
NN u Nusselt number
NR Reynolds number
NSt Stanton number, = NN u = NR
N_ molar ‡ow rate, mol/s (One mole = 6:022 1023 molecules)
n mole fraction
p pressure, Pa
Q heat, J
Q_ rate at which heat is transferred (thermal power), W
q_ rate at which heat is transferred per unit length, W/m
R acoustic resistance, Pa s/m3
R electric resistance, ohm
R mechanical resistance, kg/s
R radius, m
Runiv universal gas constant, J/mol K
r radius, m
r radial coordinate, m
S entropy, J/K
S surface area, m2
S_ rate of entropy generation, J/K s
s entropy per unit mass, J/kg K
T temperature, K
t time, s
U volume ‡ow rate, m3 /s
u x component of velocity, m/s
V volume, m3
V voltage, volt
271
v molar volume, m3 /mol
v vector velocity, m/s
W work, J
W_ rate at which work is done (mechanical or electrical power), W
w enthalpy per unit mass, J/kg
X exergy, J
X_ rate at which exergy ‡ows, W
x coordinate along sound-propagation direction, m
Y Neumann function
y coordinate perpendicular to x direction, m
Z acoustic impedance, Pa s/m3
Z electric impedance, ohm
z coordinate perpendicular to x and y, m
z speci…c acoustic impedance, kg/m2 s = Pa s/m
zn = z= a
Greek letters
= (1= )(@ =@T )p ; thermal expansion coe¢ cient, K 1
ratio of isobaric to isochoric speci…c heats
big di¤erence
penetration depth, m
small di¤erence
ik the unit tensor, equal to unity for i = k and zero for i 6= k
" surface roughness
" in mixtures, function de…ned by Eq. (10.196)
"20 = mN_ 20 = m jU1 j
" = mN_ = m jU1 j
s correction factor for …nite solid heat capacity
e¢ ciency, W_ =Q_ H or E=
_ Q_ H
phase angle
thermal di¤usivity, m2 /s
dynamic viscosity, kg/m s
kinematic viscosity, m2 /s
displacement of gas in x direction, m
perimeter, m
3.14159...
density, kg/m3
Prandtl number
ik nine-component viscous stress tensor, Pa
transduction coe¢ cient, N/amp = volt/(m/s)
!t
volumetric porosity
! = 2 f; radian frequency, s 1
272
Subscripts
273
thermal
viscous
0 “environment”or “ambient”or independent of time
1 …rst order, usually a complex amplitude
2 second order
2; k sum of second-order enthalpy and ordinary conduction
+ positive-x direction
negative-x direction
univ universal (i.e., Runiv ' 8:314 J/mol K is the molar gas constant)
Special symbols
Im[ ] Imaginary part of
Re[ ] Real part of,
hi spatial average perpendicular to x
jj magnitude of complex number
overbar time average
overdot time derivative, time rate
tilde complex conjugate
274
Bibliography
[1] T. J. Ho‡er. Thermoacoustic refrigerator design and performance. PhD thesis, Physics
department, University of California, San Diego, 1986.
[2] T. J. Ho‡er. Concepts for thermoacoustic refrigeration and a practical device. In Chair-
man Paul Lindquist, editor, Proceedings of the 5th International Cryocoolers Confer-
ence, pages 93–101, Wright-Patterson AFB, OH, August 1988. Wright-Patterson Air
Force Base.
[3] S. L. Garrett, D. K. Perkins, and A. Gopinath. Thermoacoustic refrigerator heat ex-
changers: Design, analysis, and fabrication. In G. F. Hewitt, editor, Heat Transfer
1994: Proceedings of the 10th International Heat Transfer Conference. Vol. 4, pages
375–380, Rugby, England, 1994. Institution of Chemical Engineers.
[4] G. W. Swift. Thermoacoustic engines. J. Acoust. Soc. Am., 84:1145–1180, 1988.
[5] S. Backhaus and G. W. Swift. A thermoacoustic-Stirling heat engine: Detailed study.
J. Acoust. Soc. Am., 107:3148–3166, 2000.
[6] R. Radebaugh. A review of pulse tube refrigeration. Adv. Cryogenic Eng., 35:1191–1205,
1990.
[7] R. Radebaugh. Development of the pulse tube refrigerator as an e¢ cient and re-
liable cryocooler. Proc. Institute of Refrigeration (London), pages 11–29, 2000.
http://cryogenics.nist.gov/Papers/Institute_of_Refrig.pdf, last viewed March, 2016.
[8] S. Backhaus, G. W. Swift, and R. S. Reid. High-temperature self-circulating thermo-
acoustic heat exchanger. Appl. Phys. Lett., 87:014102, 2005.
[9] G. W. Swift and S. Backhaus. Gas diodes for thermoacoustic self-circulating heat ex-
changers. In CP838, Innovations in Nonlinear Acoustics: 17th International Symposium
on Nonlinear Acoustics, edited by A. A. Atchley, V. W. Sparrow, and R. M. Keolian,
pages 399–406, Melville, New York, 2001. American Institute of Physics.
[10] G. W. Swift and D. A. Geller. Continuous thermoacoustic mixture separation. J.
Acoust. Soc. Am., 120:2648–2657, 2006. Erratum J. Acoust. Soc. Am. 124:2421, 2008.
[11] S. L. Garrett. Understanding Acoustics: An Experimentalist’s View of Acoustics and
Vibration. ASA Press, 2017.
[12] L. E. Kinsler, A. R. Frey, A. B. Coppens, and J. V. Sanders. Fundamentals of Acoustics.
John Wiley and Sons, 4th edition, 1999.
275
[13] G. W. Swift. Thermoacoustics: A Unifying Perspective for Some Engines and Refriger-
ators. First Edition: Acoustical Society of America Publications, Sewickley PA, 2002.
Second Edition: ASA Press and Springer, 2017.
[14] A. Migliori and G. W. Swift. Liquid sodium thermoacoustic engine. Appl. Phys. Lett.,
53:355–357, 1988.
[15] M. Iguchi, M. Ohmi, and K. Maegawa. Analysis of free oscillating ‡ow in a U-shaped
tube. Bull. JSME, 25:1398–1405, 1982.
[16] W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling. Numerical Recipes
in FORTRAN 77: The Art of Scienti…c Computing, Second Edition. Cambridge Uni-
versity Press, 1992.
[17] Reverse Polish notation. In Wikipedia. Retrieved on March 26, 2016, from
https://en.wikipedia.org/wiki/Reverse_Polish_notation.
[18] B. Brown, 2001. Post…x Notation Mini-Lecture. Retrieved on 3/26/2016 from
http://bbrown.kennesaw.edu/web_lectures/post…x/.
[19] D. L. Gardner and G. W. Swift. A cascade thermoacoustic engine. J. Acoust. Soc. Am.,
114:1905–1919, 2003.
[20] G. W. Swift. Encyclopedia of Applied Physics, volume 21, chapter Thermoacoustic
engines and refrigerators, pages 245–264. Wiley, for American Institute of Physics,
1997.
[21] N. Rott. Damped and thermally driven acoustic oscillations in wide and narrow tubes.
Z. Angew. Math. Phys., 20:230–243, 1969.
[22] N. Rott. Thermally driven acoustic oscillations, part III: Second-order heat ‡ux.
Z. Angew. Math. Phys., 26:43–49, 1975.
[23] G. W. Swift. Analysis and performance of a large thermoacoustic engine. J. Acoust.
Soc. Am., 92:1551–1563, 1992.
[24] P. Merkli and H. Thomann. Thermoacoustic e¤ects in a resonance tube. J. Fluid Mech.,
70:161–177, 1975.
[25] J. R. Olson and G. W. Swift. Similitude in thermoacoustics. J. Acoust. Soc. Am.,
95:1405–1412, 1994.
[26] I. Urieli and D. M. Berchowitz. Stirling Cycle Engine Analysis. Adam Hilger, Bristol
UK, 1984.
[27] A. J. Organ. Thermodynamics and Gas Dynamics of the Stirling Cycle Machine. Cam-
bridge University Press, 1992.
[28] M. A. Lewis, T. Kuriyama, F. Kuriyama, and R. Radebaugh. Measurement of heat
conduction through stacked screens. Adv. Cryogenic Eng., 43:1611–1618, 1998.
276
[29] J. R. Olson and G. W. Swift. Acoustic streaming in pulse tube refrigerators: Tapered
pulse tubes. Cryogenics, 37:769–776, 1997.
[30] J. R. Olson and G. W. Swift. Suppression of acoustic streaming in tapered pulse tubes.
In R. G. Ross, Jr., editor, Cryocoolers 10, pages 307–313. Plenum, New York, 1999.
[32] A. Bejan. Advanced Engineering Thermodynamics. Wiley, New York, 2nd edition, 1997.
[37] R. S. Reid, W. C. Ward, and G. W. Swift. Cyclic thermodynamics with open ‡ow.
Phys. Rev. Lett., 80:4617–4620, 1998.
[39] R. S. Reid. Open cycle thermoacoustics. PhD thesis, Georgia Institute of Technology,
School of Mechanical Engineering, 1999.
[41] David Gedeon. DC gas ‡ows in Stirling and pulse-tube cryocoolers. In R. G. Ross, Jr.,
editor, Cryocoolers 9, pages 385–392. Plenum, New York, 1997.
277
[46] M. J. D. Powell. A hybrid method for nonlinear equations. In P. Rabinowitz, edi-
tor, Numerical Methods for Nonlinear Algebraic Equations, pages 87–114. Gordon and
Breach, London, 1970.
[50] M. Ohmi and M. Iguchi. Critical Reynolds number in an oscillating pipe ‡ow. Bull.
JSME, 25:165–172, 1982.
[51] M. Ohmi, M. Iguchi, and I. Urahata. Flow patterns and frictional losses in an oscillating
pipe ‡ow. Bull. JSME, 25:536–543, 1982.
[52] I. E. Idelchik. Handbook of Hydraulic Resistance. Begell House, New York, 3rd edition,
1994.
[54] J. R. Olson and G. W. Swift. A loaded thermoacoustic engine. J. Acoust. Soc. Am.,
98:2690–2693, 1995.
278
[61] G. W. Swift and R. M. Keolian. Thermoacoustics in pin-array stacks. J. Acoust. Soc.
Am., 94:941–943, 1993.
[63] W. M. Kays and A. L. London. Compact Heat Exchangers. McGraw-Hill, New York,
1964.
[64] P. J. Storch, R. Radebaugh, and J. E. Zimmerman. Analytical model for the refrigera-
tion power of the ori…ce pulse tube refrigerator. Technical note 1343, National Institute
of Standards and Technology, 1990.
[65] N. T. Weiland and B. T. Zinn. Open cycle traveling wave thermoacoustics: Mean
temperature di¤erence at the regenerator interface. J. Acoust. Soc. Am., 114:2791–
2798, 2003.
[66] J. L. Smith and M. Romm. Thermodynamic loss at component interfaces in Stirling cy-
cles. In Proceedings of the 27th Intersociety Energy Conversion Engineering Conference,
pages 5.529–5.532. Society of Automotive Engineers, 1992.
[67] G. W. Swift and P. S. Spoor. Thermal di¤usion and mixture separation in the acoustic
boundary layer. J. Acoust. Soc. Am., 106:1794–1800, 1999. Errata J. Acoust. Soc. Am.
107:2299, 2000; 109:1261, 2001.
[69] D. A. Geller and G. W. Swift. Thermoacoustic mixture separation with an axial tem-
perature gradient. J. Acoust. Soc. Am., 125:2937–2945, 2009.
[74] Richard Raspet, William V. Slaton, Craig J. Hickey, and Robert A. Hiller. Theory
of inert gas-condensing vapor thermoacoustics: propagation equation. J. Acoust. Soc.
Am., 112:1414–1422, 2002.
279
[75] William V. Slaton, Richard Raspet, Craig J. Hickey, and Robert A. Hiller. Theory
of inert gas-condensiong vapor thermoacoustics: Transport equations. J. Acoust. Soc.
Am., 112:1423–1430, 2002.
[77] O. J. Foust, editor. Sodium-NaK Engineering Handbook. Gordon and Breach, New
York, 1972.
280
Segment index
282
MIXBL STKDUCT
reference summary, 214 examples, 95
tutorial discussion, 125 reference summary, 187
MIXCIRC STKPIN
examples, 129 examples, 180
reference summary, 214 reference summary, 177
tutorial discussion, 125 STKPOWERLW
MIXSLAB reference summary, 183
reference summary, 214 STKRECT
tutorial discussion, 125 examples, 104, 180
MIXTBL reference summary, 177
reference summary, 217 tutorial discussion
MIXTCIRC with steady ‡ow, 104
reference summary, 217 STKSCREEN
MIXTSLAB examples, 88, 108
reference summary, 217 reference summary, 183
tutorial discussion
OPNBRANCH ignoring steady ‡ow, 88
examples, 172 with steady ‡ow, 108
reference summary, 171 STKSLAB
PISTBRANCH examples, 63, 180
examples, 172 reference summary, 177
radiation impedance from ‡anged tube, SURFACE
172 examples, 63, 81
reference summary, 171 reference summary, 160
PX SX
reference summary, 193 examples, 88
reference summary, 193
RPN SYSEXEC
examples, 56, 88 examples, 246
reference summary, 235 reference summary, 245
tutorial discussion, 50
TBRANCH
SOFTEND examples, 56, 91, 108, 118
as open-tube radiation impedance, 228 reference summary, 229
examples, 104, 108, 228 tutorial discussion
reference summary, 227 ignoring steady ‡ow, 35
tutorial discussion with steady ‡ow, 108
ignoring steady ‡ow, 35 TITLE
with steady ‡ow, 104, 108 reference summary, 223
STKCIRC TX
examples, 81, 180 reference summary, 190
reference summary, 177
STKCONE UNION
reference summary, 187 examples, 56, 91, 108, 118
283
reference summary, 231
tutorial discussion
ignoring steady ‡ow, 35
with steady ‡ow, 108
VDUCER
examples, 175
reference summary, 173
tutorial discussion, 26
VEDUCER
examples, 169
reference summary, 167
tutorial discussion, 26
VESPEAKER
as alternator, see IESPEAKER
examples, 70, 169
magnetic ‡ux saturation, 266
never use with R = !L = 0, 167
reference summary, 167
VOLUME
reference summary, 242
VSPEAKER
examples, 67, 175
magnetic ‡ux saturation, 266
never use with R = !L = 0, 167
reference summary, 173
tutorial discussion, 26
VXQ1
examples, 114
reference summary, 199
tutorial discussion
with steady ‡ow, 114
VXQ2
reference summary, 199
VXT1
reference summary, 199
VXT2
reference summary, 199
284
Subject index
286
temperature ‡oor, 139 Limitations, 141, 145, 146
user-de…ned, 28, 253, 254 Linking multiple models, 58, 235
Graphs, see State plots, Incremental plots, List linkages, 53
Phasor plots, Schematic view Looped models, examples, 36, 56, 92, 108,
Guesses and targets 115
convergence di¢ culty, 78, 80
examples, 19, 71, 90, 92, 119 Magnetic ‡ux saturation, 266
general guidance, 19, 71, 77, 136 Master–slave links, 30, 52
in RPN, 54, 58 Messages, error and warning, 140
maximum number, 136 Mixture separation, typical end conditions,
meeting targets, 78, 80 126
mixture separation, 126 Mixture-separation equation, 126, 134
non-standard guesses, 58 Mode, see Normalization mode
non-standard targets, 54, 239 Momentum equation, 10, 61, 134
phase target near 180 , 143 Multi-model plots, 58
restoring, 86 Multiple models, see Linking multiple mod-
rewinding incremental plots, 48 els
shooting-method parameters, 139 Nint, 139
troubleshooting, 78, 80 Nonlinear e¤ects, 6
with nonzero N_ , 108, 111 Normalization mode, 139
Harmonic content, 141 Notation, 7, 270
Head-loss equation, 103, 134 energy, 136
Heat exchange, external T , 144 Notes, user-speci…ed, 87
Highlights display, 34 Numerical integration method, 136
Hydraulic radius of screens, 89, 147, 183 Numerical options
integration steps, 139
Incremental plots, 44 shooting method, 139
cloning, 41 temperature ‡oor, 139
examples, 67, 84
multiple models, 58 Open tube, radiation impedance, 171, 228
rewinding, 48 Options, numerical, 139
Inherent limitations, 141 Overview of DeltaEC, 1, 10, 61, 134
Installation, 7
Phases near 180 , 143
Integration variables, 134
Phasor plots, 33, 42, 97
acoustics only, 11
Plots, see State plots, Incremental plots,
thermoacoustics, 63
Phasor plots, Schematic view
with mixture separation, 126
Power notation, 136
with steady ‡ow, 103
Pressure, second-order, 103
ksFrac examples, see Looped models
including vessel-wall conduction, 145 inherent limitations, 142
recommended value, 145 subtleties in calculation, 136
Pulsed combustion, 267
Legs, 42, 229
License, 8 Radiation impedance
287
open end of ‡anged tube, 172 in DUCT, 29, 142, 146, 150, 152
open end of un‡anged tube, 228 in MINOR, 147, 163
Rayleigh streaming, 147 in PX, 196
Reporting bugs, 7, 140 in STKPOWERLW, 142, 187
Restoring guesses, 86 in STKSCREEN, 142, 187
Rewinding incremental plots, 48 in SX, 196
Run monitor, 15 in VX**, 142, 204
288