J. Acoust. Soc. Jpn.

(E) 13, 1 (1992)

The mechanism of sound production in organ pipes

and cavity resonators

Samuel A. Elder

Physics Department, United States Naval Academy,

Annapolis, Maryland, 21402 USA

A new unified method of treating the sounding mechanism of organ pipes and cavity
resonators replaces the control volume method which until recently has been the most
widely accepted theory. It is shown that shear layers contain a dividing surface whose
motions control the production of sound in the resonant cavity. Both jet-drive and
force drive contributions are found to be involved in the sounding mechanism. The
present approach shows the drive system to be inherently linear over a range of shear
layer widths, so that sinusoidal oscillation is readily achievable. For voicing situa
tionscalling for spectra rich in harmonics, nonlinear drive is attained by resorting to a
smaller ratio of jet width to lip cutup which, at high enough Reynolds number, permits
the growth of eddies which result in impulsive action at the lip.

Key words:Sound production, Cavity resonator, Organ pipe, Separated shear layer,
Jet instability, Flow tone

PACS number:43. 75. Np, 43. 28. Py, 43. 28. Ra, 43. 50. Nm, 43. 20. Ks

Then in 1937 Brown 1) published results of an in

1. INTRODUCTION
1.1 History of the Development of Organ Pipe
Physics
The physics of organ pipes, a problem of classical
mechanics left unsolved by the sudden onset of mod
ernphysics at the beginning of the twentieth century,
had received some consideration by two scientific
giants of the nineteenth century. H. Helmholtz, in
his "On the Sensations of Tone," expressed the view
that pipe drive is due to fluctuating volume flow
from the jet into the pipe. Rayleigh, in his "Theory
of Sound" disagreed, believing that the oscillation is
sustained, not by a volume flow but by a force or
dipole source caused by the motions of the jet in the
mouth of the pipe. Rayleigh also pointed to the fact
that jet motion is inherently unstable and discovered
how to calculate appropriate propagation constants
for jets and free shear layers of simple structure, an
important component to the understanding of
jet/edge phenomena. He did not attempt a complete
theory of the organ pipe problem however, and no
further progress was made by his contemporaries.
vestigationof edgetone oscillation including photo
graphicvisualizations of a ribbon jet playing on a
sharp edge, with eddies being cast off at the period
of the emitted tone. He showed that the frequency of
an edgetone oscillator is proportional to the speed
of the jet. This led to speculation that organ pipes
are driven by the same mechanism as edgetones, ex
ceptthat for organ pipes, the frequency is held
nearly constant because of the coupling to the pipe
resonator." This view was generally accepted for
a number of years but did not lead to a method of
calculating the jet blowing pressure and sound
amplitude for organ pipes.
The first progress in developing a quantitative
theory of pipe mechanism came in the form of three
papers delivered at the 5th International Congress on
Acoustics in 1965 by scientists from Germany:
Bechert", Cremer 4) and Ising." These were later
followed up with a seminal paper" by Cremer and
Ising in 1968 and a related doctoral dissertation by
Ising 7) in 1969. The "German" method was an
adaptation of Helmholtz's concept. Pipe drive was

11

assumed to take place by flow of jet fluid, Qj, into
the parallel resonant mechanical circuit consisting
of positive mouth reactance and negative pipe re
actancebranches as shown in Fig. 1. The pipe is
then assumed to act as a linear amplifier, producing
an acoustic flow QM=-(Zp/Zs)Qj, where Zp is the
acoustic impedance of the cavity looking into the
mouth and where Zs is the combined series imped
anceof the resonant system. The acoustic flow in
turn, acting on the jet fluid in the plane of the
mouth, produces more jet-drive flow Qj, completing
the feedback loop. Using root locus methods they
were able to account at least qualitatively for many of
the observed properties of organ pipe oscillation.
To get improved agreement with experiment they
found it necessary to assume that entrained jet fluid
added to the drive, resulting in an oscillation ampli
tudeproportional to the square of the jet speed.
A divergent line of reasoning was followed by
Coltman, 8) in a paper which appeared also in 1968.
Following Rayleigh, Coltman reasoned that the jet
interacted with the pipe by generating a fluctuating
force across the mouth asscoiated with its own
a.
Fig. 1 Jet drive model. a. Qj is jet-drive
flow, QM and Qp represent acoustic flow.
b. Parallel-resonant equivalent circuit,
consisting of pipe impedance, Zp, and
mouth impedance, ZM.
Fig. 2 Force drive model:series
circuit. •¢p is acoustic driving
12
J. Acoust. Soc. Jpn.(E) 13, 1 (1992)
decay, a "force-drive" theory as opposed to "jet-
drive." This meant that the reactive elements of the
pipe were driven in series rather then in parallel (see
Fig. 2) and in addition implied that the amplitude of
the oscillation should be proportional to the square
of the jet speed without the need of entrained fluid,
as postulated for the jet-drive approach. However
in later measurements of the timing of the arriving
jet pulse at the lip of the pipe, Coltman 9) found that
the phase relations seemed to support the jet-drive
theory.
In 1973, Elder10) used a control volume approach
to show that both jet-drive and force-drive were con
sistentand suggested that there should be a third,
nonlinear, contribution to the oscillation due to
intermodulation distortion caused by the nonlin
earityof the momentum conservation process. This
was followed in 1976 by a paper due to Fletcher11)
which showed that Elder's nonlinear term was un
necessary,leaving only the contributions due to
force-drive and jet-drive. Fletcher and his associates
have developed a comprehensive feedback theo
ry12,13)which goes further than any to date in ex
plainingthe physics of organ pipes. Nevertheless
there remain some areas where the model appears to
be deficient, such as in the prediction of threshold
b.
blowing pressures, and the amplitude of underblown
regimes.
In 1980 an important new approach was intro
ducedby Yoshikawa and Saneyoshi,14) who sug
gestedthe possibility of direct feedback from the
action of the jet as a vibrating diaphram. This idea
is the starting point for a new concept developed by
Elder15) whose theory represents the latest effort in
the field of organ pipe physics and which will be
described in this paper.
1.2 Recent Studies of Cavity Resonators
A closely related phenomenon is the cavity reso
nator,or "Helmholtz" resonator problem. Wind
blowing past a slot or cavity along the side of a
moving vehicle may cause sound and vibration with
frequency proportional to the speed. If a resonant
volume or plate is present to couple with the oscilla
tion,a strong fixed-frequency tone may be produced
at certain speeds. This sometimes occurs in auto
mobileswith a single window open, manifesting itself
as a very low frequency (infrasonic) rumble or feel
resonant ingof pressure sensation to the occupant of the auto
pressure. mobile.The effect has been studied extensively in
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

of the layer with step-change in velocity profile-the

a. Helmholtz "vortex sheet." Imagine a uniform mass

of fluid moving over an inviscid stationary fluid.

Rayleigh17) showed that the slightest disturbance will

cause the interface between the layers to break into

a wave which will begin to grow exponentially. If

the disturbance has an angular frequency, ƒÖ, the

surface of discontinuity will have a transverse dis

placement:
b.
(1)
where A is the amplitude of the disturbance at x,

t=0, and where Uc, the wave speed, is half the

speed of the upper fluid. Such waves, however,

are an expression of vorticity and will eventually

roll up into a row of discrete eddies if there is space

enough, as demonstrated by Rosenhead18) who

carried out Rayleigh's solution to the second order.

This is shown in Fig. 4.

For the equivalent case of a complete jet consist

Fig. 3 Cavity resonator compared with
ingof a layer of fluid of width, ƒÂ, moving at speed
organ pipe. a. Cavity mouth showing in

flectionpoint locus (dividing streamsur

face)with displacement, ƒÌ. b. Organ pipe

mouth showing jet emerging from flue of

a.
width ƒÂ. Dividing surfaces are along the
inflection point loci of the shear layers on

either side of jet.

b.

recent years on account of its application to military

vehicles.16) Besides its being an undesirable sound,
c.
or "noise," the cavity resonator problem differs
from organ pipe dynamics in the fact that only a
single free shear layer, or "halfjet," is present at the d.

mouth of the cavity in contrast to the organ pipe jet,

which has two shear layers (see Fig. 3). Fig. 4 Vortex sheet instability. a. Stability
wave. b. Beginning of nonlinear asym
1.3 A Unified Approach to Organ Pipes and Cavity metrydevelopment. c. Wave begins to
Resonators roll over. d. Distinct eddy formed
We will consider first the case of the cavity reso (adapted from Rosenhead18)).
nator,showing how the oscillation of the single free
shear layer creates forces and drive flow in the mouth a.

of the cavity. Then the same approach will be ap

pliedto the organ pipe jet, where two shear layers
move in coupled motion in such as manner as to co
operatein the creation of acoustic flow in and out
of the pipe. b.

2. THE SHEAR LAYER AS A

VIBRATING MEMBRANE
2.1 Shear Layer Instability Fig. 5 Sinuous and varicose jet waves. a.
The simplest case of shear layer instability is that Sinuous. b. Varicose.

13
 J. Acoust. Soc. Jpn.(E) 13, 1 (1992)

U, surrounded above and below by masses of sta velocity profiles for the ideal halfjet and "full" jet
tionaryfluid, instability also is present. However and their real-world viscous counterparts. At the
for this case, the two vortex sheets making up the upstream, or trailing, edge of a slot the existing wall
edges of the jet will tend to move in synchronism boundary layer at that point must be continuously
with each other, forming either a "sinuous" pair of converted to a separated shear layer. The stream
waves or a "varicose" pair, in the words of Ray linethat forms at the edge itself must therefore
leigh,19) as shown in Fig. 5. In this case the wave begin as a stagnation point in order to match the
speed depends on the wavelength. Rayleigh's solu zero-velocity boundary layer condition at the wall.
tionfor the sinuous case leads to a wave speed: As the flow along this streamline accelerates up to
the local stream speed, U, the streamline will be
(2a)
for thin jets and comea dividing line for flow in the outside fluid,
which is moving, and flow of the fluid inside the
(2b)
cavity, which initially is not. As the flow pro
in the limit for very thick jets, which is of course to
gressesalong the mouth of a cavity or pipe, viscous
be expected since for a very thick jet we should be
(or turbulent) shear will produce a profile of posi
aware of only one shear layer at a time. The jet tivecurvature below this point due to acceleration of
waves also begin with exponential growth and tend
fluid while above the point the curvature will be
to roll up into a row of eddies, first to one side and negative due to deceleration of fluid. This means
then the other, as may be seen in Brown's photo
that the streamline that begins at the edge is actually
graphs 1)of edgetone jets. the locus of points corresponding to the inflection
point in the velocity profile. The inflection point
2.2 The Inflection Point Streamline:A Flow
locus is therefore the dividing streamsurface which
Divider
separates outside from inside fluid as waves form
Inviscid vortex sheets are a useful idealization, but
along the shear layer.
in the real world the edge of a jet or separated shear

layer always consists of a finite boundary layer of

2.3 Stability Wave Functions for Realistic Shear
thickness, ƒÂ0. In Fig. 6 is a comparison of the
Layers
For realistic halfjets and jets as shown in Fig. 6b
and Fig. 6d, the wave speed and growth rate must
a. b. be computed by numerical integration based on
some assumed velocity profile shape. One must
also consider spatially growing waves rather than
temporally growing waves assumed by Rayleigh.
Some useful tables have been prepared by
Michalke,20) based on a hyperbolic tangent curve
shape, which give complex wave propagation con
stantsas a function of the nondimensional fre
quency.The parameters evaluated by Michalke
for spatially growing disturbances along shear layers
c. d. of infinite extent are:
Cr=nondimensional phase velocity

β=nondimensional frequency

αr=nondimensional real propagation constant

αi=nondimensional imaginary propagation con

stant.

To make use of these for obtaining practical esti

matesfor a real shear layer we need to know the

angular frequency ƒÖ, the free stream velocity, U•‡,

Fig. 6 Velocity profiles:a. Ideal halfjet. b.
Real shear layer. c. Ideal (top-hat) jet. d. and the slope of the velocity profile at the inflection

Real jet. point, (•ÝU/•Ýz)0. The following auxiliary formulas

14
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

are useful: create an acoustic wave across the mouth, either

from the resonance of the cavity or from a dipole
Uc=CrU∞=wave convection speed
source at the downstream edge. The sum of the
d=U∞/(2∂U/∂z)0=length parameter
acoustic particle velocity and the stability wave then
β=ωd/U∞,=nondimensional frequency
add up to zero at x=0. For the case of the cavity
k=2π/λ=αr/d=real propagation constant
resonator, Elder21,22) has shown that the form of the
α=-αi/d=imaginary propagation constant.
total wave disturbance near x=0 is:
The wave function for a separated shear layer of
infinite extent will then take the form: (4)
which is the sum of the acoustic particle displace
(3)
mentand the negative of the stability wave. Here
"k"
3. WAVE FUNCTIONS FOR SHEAR is twice the value which would be obtained
WAVES FORMED IN THE MOUTH from the formulas in Section 2.3 and the exponen
OF A CAVITY tialgain term is missing, while uM is the acoustic
3.1 Kutta Conditions at Leading and Trailing particle velocity amplitude. As explained in Ref.
23), the doubling of the propagation constant occurs
Edges
on account of the fact that the shear layer is being
Except near boundaries fluids behave as though
they were nearly inviscid, and over distances small accelerated by the sound field, causing the wave
speed to be halved relative to that in the "labora
compared to the acoustic wavelength a free shear
layer is approximately incompressible. Therefore tory"system. Note the absence of an imaginary
for some purposes we may deal with the shear layer component to the propagation constant. Expo
by potential flow methods. This was the basis of nentialgain disappears very quickly in the growth of
Rayleigh's analysis of the instability problem. the wave because of the stretching of the shear layer
that is produced by its transverse motion, reducing
However, near either the upstream or downstream
edges, the potential flow solution leads to infinite the strength of the fluctuating vorticity sources. Sta
bilitywaves become neutrally stable once their lateral
velocity. In a real fluid discontinuities in the
velocity cannot occur, nature always finding a way displacement approaches the shear layer width.
At the downstream edge, eddy creation is more
to avoid the catastrophy. For the case of cavity
resonators and organ pipe jets there seem to be two common, there being room beyond the edge for the
ways in which this is achieved: either by (1) the eddy to develop. However, for the case of the
casting off of an eddy, or (2) by forming a stability laminar cavity resonator, it has been shown23) that
wave/acoustic wave pair at the edge. flow reattachment may occur at x=H, allowing the
Both eddies and stability waves are, in fact, mani acoustic wave/stability wave solution to take place
festationsof the vorticity that is resident in the shear in the same manner as at x=0. On the other hand,
layer. Eddies may thought of as the "particle" when the downstream boundary condition is met by
nature of vorticity while stability waves exhibit its casting off an eddy, the acoustic wave is not generat
"wave"
nature. In order for eddies to show up it edat this edge and the total wave function in the
is necessary that there be enough room in the mouth vicinity of x=H becomes the stability wave alone:
region for them to spin. The initial size of the eddy (5)
is of the order of the thickness of the upstream
boundary layer. Therefore, when the boundary 3.3 Organ Pipe Edge Conditions
layer thickness, 80, is small compared to the mouth For the organ pipe jet, there are two upstream
dimension, H, eddies are an option for assuring edges, one for each shear layer. The condition
velocity control at the upstream edge. More typi ξ=0 cannot be met simultaneously by both shear
cally,80 is of the order of H and the vorticity shows layers at x=0 unless an additional boundary condi

up as a stability wave. tionis imposed, namely that •ÝƒÌ/•Ýx=0 at x=0. This

is sometimes called a "full" Kutta condition.

3.2 Cavity Resonator Edge Conditions Creamer and Ising 6) invoked this condition in

In order to insure that the velocity vanish at x=0 their description of organ pipe jet motion, showing

(i.e. at the upstream edge), forces in the shear layer that it leads to a second pair of stability waves

15
 J. Acoust. Soc. Jpn.(E) 13, 1 (1992)

along the jet, moving in the upstream direction. For we must require that the jet-drive flow, Qj, must be
the top-hat profile considered by Cremer and Ising, introduced by means of a "branch" circuit into the
this leads to a jet centerline wave of the form: loop containing the reactive elements, so that the
(6) total acoustic discharge into the cavity, Qp, is given
by:
where
(11)
To establish the legitimacy of this relation, which is
a kind of flow-conservation rule, we compute Qp by
and where C1 and C2 are given in terms of k1, k2, α1, integrating the dividing surface velocity over the
α2. entire surface of the mouth:
For jet with a more realistic profile, numerical

results are available to estimate the complex pro (12)

pagationconstant13) needed to describe the wave
motion.
where two-dimensional symmetry is assumed and
the integration is over the streamwise dimension H
4. THE DIVIDING SURFACE AS A and lateral dimension b of the mouth.
VIBRATING MEMBRANE To simplify the discussion, let us assume an ideal
halfjet with velocity profile given by Fig. 6a. Sub
4.1 Relation between v and ƒÌ
stitutingv from Eq.(10) and neglecting the small
Since the streamsurface through the locus of
flow deficit due to the stagnation point at the edge,
inflection points divides the shear layer into inside
we obtain:
and outside fluid, 24) its wave motions are in some

ways analogous to the motions of the reed in a reed (13)

wind instrument. The transverse velocity of the But we have indicated that ƒÌ(0, t)=0 so this may be

dividing streamsurface contains contributions from written:

the acoustic wave across the mouth and from the (14)
component of the main flow velocity of the shear
The first term on the right is the integral of the
layer or jet as it is deflected either into or out of the
acoustic particle velocity, or the acoustic discharge
mouth. To determine the velocity in relation to
through the mouth, QM, while the second term is of
the wave displacement 6, we use the method of
the form of jet-drive, QJ, which has been used in
Rayleigh 25) and Lamb 26):
previous papers, for the case in which 6 has a finite
Let the surface in question be represented by the
value at the downstream edge. We have shown,
equation
therefore, that the distinction between QP and QM,
(7) and with it the parallel-resonant drive situation,

We then require that occurs whenever ƒÌ(H, t) does not vanish.

Jet-drive is present for organ pipe oscillation, but

(8) not necessarily for cavity resonators. Where it

holds, we must assume that at least a portion of the

where U is the local stream speed and v is the "mem
acoustic response of the system, (QM)par, will be via
brane"velocity, •Ýz/•Ýt. Carrying out the differentia
parallel-resonant drive:
tion,

(9) (15)
This may be written
If there is assumed to be no fluid motion below the
(10) dividing surface, the effect of the motion of the shear
By substituting the appropriate value of ƒÌ we may layer is to cause an amount of fluid Qj to enter the
obtain v as a function of x over the mouth. cavity each cycle, but not to remove any. There is
therefore both a dc and an ac component of QJ.
4.2 The Flow Conservation Rule For a real shear layer, there is some motion of the
If the cavity or pipe is to be driven by a parallel fluid below the dividing surface, so that fluid will
resonantcircuit representation, as shown in Fig. 1, be also ejected from the cavity each cycle, but in

16
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

general this will be of a smaller amount. The

ejected fluid, Qj', does not produce additional

acoustic gain (such as indicated by Eq.(15)), since

it is applied outside the cavity.

4.3 Forces on the Dividing Surface

The lateral displacement of the vorticity con
centratedin the portion of the shear layer between
Fig. 7 Mouth of cavity from above, show
the upstream and downstream edges produces fluc ingcoordinate system.
tuatingforces on the surrounding fluid which result

in a pressure difference •¢p across the mouth in

series with the resonant elements. This produces

an additional acoustic flow in the cavity given by:

(16)

so that the total contribution to acoustic discharge

in the cavity may be written:
(21)
(17)
Now if the fluid upstream and downstream of the
Combining the terms in QJ and making use of fact cavity is assumed to be free of forces in the z-direc
that Zs=ZM+Zp, we may write this in the form: tion,we may set F=0, giving:

(18)

This shows clearly that the acoustic oscillation con

sistsin general of two parts:a jet-drive contribution
that is proportional to QJ, and a force-drive con

tributionthat is not.
To compute the force term from which •¢p is (22)
found, we must take the material derivative of the
dividing surface velocity which is proportional to Noting that the integrands of all the terms on the
the force on the fluid per unit volume fluid due to right side of this equation involve at least one space
shear layer motion: derivative of 6, and therefore operate on the stability
wave alone, it is clear that the term on the left must
(19) contain any reaction forces associated with the
or
purely acoustic field. If we further separate the
integral on the left into two parts, the first con
tainingno x-dependent terms and the latter a func
tionof x, we may then identify the former term as
due to the acoustic pressure difference, ZSQM, that
(20) drives the standing wave in the cavity:
Given a function ƒÌ(x, y, z, t) we may determine the

total force by integrating this expression over the x

and y dimensions of the mouth and vertically, over
the portion of the shear layer containing the sources

of the fluctuating vorticity. The geometry of the cav

itymouth is shown in Fig. 7. Formally we may rep

resentthe total force on the fluid in the region of the

mouth as:
(23)

17
 J. Acoust. Soc. Jpn.(E) 13, 1 (1992)

where SM is the area of the mouth (bH). This

gives, for the "force-drive" acoustic pressure dif
ference:

(24)

For later reference, we shall identify the terms on

the right in Eq.(24) by the labels A, B, C, and D
respectively. It should be remarked that term-B
appears in Eq.(24) only because we must take into
account the fact that the dividing streamline begins
(and sometimes ends) at a stagnation point, and so
the fluid speed U, which is constant over most of
the path, must be considered a variable over a small
"boundary layer" region where acceleration and

deceleration take place. This is an important source

of coupling to the surrounding fluid.

5. APPLICATIONS OF THE THEORY

5.1 Cavity Resonator in Laminar Boundary Layer
We will treat this case first because it represents
the simplest application of the theory. In Fig. 8 is
shown a sequence of dividing streamline configura
tionsobtained by computer-assisted hotwire ex
periment 23)on a cavity acted on by a laminar bound
arylayer. Such experiments have shown that
when a cavity mounted in a flat wall under condi
tionsof laminar flow is driven at resonance, reat
tachmentoccurs at the downstream edge. Con
sequentlythe "acoustic" boundary condition is
applied at both upstream and downstream edges and
the Ei function specified in formula (4) applies uni
formlyalong the mouth.
Before we can substitute this function into Eq.
(24), however, we must specify the z-dependence of
ξI. For Rayleigh's ideal halfjet, ξ is proportional
Fig. 8 Dividing streamline for laminar
to e•}kz, where the plus sign applies below the inflec
cavity resonator (from Ref. 23)). Circles
tionpoint and the minus sign above. 27) As a first
represent data points obtained by com
approximation, let us therefore take the integration puter-assisted hotwire measurements.
over z to be equivalent to multiplying the integrand Solid line is a theoretical fit based on
by a constant of order 2/k. Now we may make
Eq.(4).
use of the ƒÌI function to integrate •¢p. The result

is:

18
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

(29)
Here at last is the correct explanation of the "U2"
dependence of the acoustic amplitude!
To find the amplitude and frequency of the oscilla
(25)
tionat a given speed we make use of Eq.(18), in the
where we have left the pressure terms in the order
form:
A, B, C, and D.

In a feedback oscillator the total phase shift

(30)
around the "loop" must be 360•‹ to produce self

sustainedoperation. Now for all types of jet/edge which may be read as follows:for simultaneous

resonator systems, feedback is applied from the values of wind speed, U, oscillation amplitude, UM,

downstream edge or lip to the upstream edge with a and sounding frequency, ƒÖ, that give the same com

180•‹ shift. Since the feedback is applied via some plexnumber on either side of the equation, we have

form of acoustic radiation, and so travels from the a solution point. This is a form of "root locus"

downstream to upstream edge effectively instanta method and the solution may be performed either

neously(the acoustic wavelength being large compar by computer or by graphical analysis.28) Without

edto H), travel time of information along the jet or actually going through this process, which is rather

shear layer must adjust itself to one-half period of tedious, we can gain some useful insights by ex

the oscillation. This means that the disturbance aminingthe known form of the ZS impedance

along the jet or shear layer must be approximately function 29):

one-half wavelength long between x=0 and x=H,

a fact that has been experimentally

(31)
verified by

Coltman 9) for organ pipes and edgetones with

where L is the depth of the cavity (here assumed to
laminar jets, and by Fletcher 13) for turbulent jets.
be a long tube), k0 is the acoustic propagation con
This can be expressed in terms of the real propaga
stant,c is the speed of sound, AL is the cavity end
tionconstant:
correction, and RL is the linear part of the system
(26) resistance, containing losses due to radiation, viscous
where kL is the value of the propagation constant damping and thermal damping. The nonlinear
in the "laboratory" coordinate system, as distin part of the resistance is due to formation of period
guishedfrom "k" which is measured in the accelerat icallyemitted vortices at the mouth orifice at large
edsystem of the shear layer itself. To illustrate the sound amplitude, 30) while the imaginary terms rep
difference, we rewrite Eq.(4) in the form: resentthe mouth and cavity reactances that form
the resonant system.
(27)
Amplitude of the oscillation is ultimately limited
from which it is evident that the effective value of by the nonlinear term in ZS. So long as the phase
the propagation constant is kL=k/2. This effect
condition for positive feedback is satisfied, the oscil
may be observed in the actual data for dividing
lationwill tend to build up until system losses rep
surface waves along the mouth of a laminar cavity resentedby the real part of ZS are just matched by
resonator in Fig. 8, obtained from Ref. 23), where
a real, amplitude-dependent term on the left side of
it is plain that, in the system moving up and down Eq.(30). The threshold for equality of the two
with the shear layer, the mouth dimension H is one sides of Eq.(30) can occur only when the amplitude
full wavelength long. Another way of looking at
is large enough for the nonlinear term in Zs to be of
it is to trace the timing of the zero-crossing locus in
size comparable to the linear term. Thus by equat
Fig. 8 and compare it with the progress of the wave
ingthese two terms we can get an idea of the actual
maximum, which moves exactly half as fast. The
amplitude at which the oscillation will reach dy
result implies that:
namicequilibrium.
(28) For a pipe closed at the opposite end, the linear
Substituting this in Eq.(25) we discover that, at portion of the impedance at the lowest axial mode
resonance, all but term-B vanishes, obtaining for the can be estimated from the "Q" or quality factor of
laminar cavity resonator the simple drive-pressure: the system at low amplitude from the formula 7):

19
 J. Acoust. Soc. Jpn.(E) 13, 1 (1992)

(32)
where Sp is the cross sectional area of the pipe.
Equating this to the nonlinear term we obtain:

(33)
Since ƒÌ(H)=0 for laminar cavity at resonance, there

is no contribution from QJ and the left side of Eq.

(30) is

which is real and so may be equated to approxi

matelytwice the nonlinear resistive impedance,
giving:
Fig. 9 Dividing streamline for turbulent
cavity resonator (from Ref. 21)). Circles
Thus the speed of the wind needed to excite a cavity represent data points obtained by com
of slot dimensions H and b, quality factor Q and puter-assisted hotwire measurements.
Dashed line is faired to data. Theoretical
resonant frequency f is on the order of: ξI-curve can be fitted near x=0 while ξII

(34) function can be fitted near x=H.

This may be compared with some actual data, taken

from Ref. 23), for which the resonant frequency, f, Therefore the value of kH to use in Eq.(25) should
is 300Hz, and low-amplitude "Q" of the cavity is be
26.3, and other parameters are given by:
(36)
(The integration for turbulent boundary case is the

same as for laminar, despite the fact that the ƒÌII

Substituting in Eq.(34), we obtain a computed value function applies over only part of the path. This is

of U=1,382cm/s which is in the neighborhood of because we need integrate either function only over

the measured value, 1,300cm/s. the integrals that contain functions of x.) Sub

stitutionof kH into this equation indicates that all

5.2 Effect of Turbulent Boundary Layer on Cavity four force terms contribute to the drive term re

Resonance sultingin:

In Fig. 9 is shown a sequence of dividing stream

linesfor data obtained on a cavity resonator mount

(37)
edin turbulent boundary layer flow. 21) Synchronous
Interestingly, in this case, the force contains an
computer averaging was needed to reduce the
expression (shown in curley brackets) that depends
background noise caused by turbulence. We as
on the relative wave speed in the accelerated system,
sumethat, since the turbulence is small scale, the
Uc/U. The expression involving the wave speed
average flow behaves in a quasi-laminar fashion so
has real values only for the range of Uc/U equal to
that it is still reasonable to talk about a dividing
or greater than 0.2929.
streamline for the flow. Experimentally it is found
The theory therefore predicts that cavity re
that the 61 function fits the experimental data best
sonancewill not occur in a turbulent boundary
near x=0 while, in the vicinity of x=H, it is neces
layer unless the wave speed exceeds a critical thresh
saryto use a ƒÌII, function to get a match. This
old.This can be translated into a critical Strouhal
seems to indicate that there are no acoustic sources
number:
in the jet in the vicinity of the downstream edge, and

we may assume that k and kL are essentially in

distinguishablehere. The average value of k over

from which we may derive the Strouhal number for
the path from x=0 to x=H is therefore given ap
the lowest shear-wave mode:
proximatelyby: Strouhal Number=(fH/U)cr=(3/4)(Uc/U)cr
(35) (38)

20
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

This prediction is strikingly confirmed in the data of sinuous, doubling the total drive force. To cal
DeMetz and Farabee31 who observed a distinct culatethe force drive terms, one must determine the
cut-off effect for resonant cavities with lower value for k1H. It seems likely that
Strouhal number!
(40)
From Fig. 9 it is clear than ξ does not vanish at
x=H for this case, so there should be a contribution Over the early portion of its path, the jet centerline
to QJ. However since the peak value ofξarrives travels very fast on account of the action of the
90° out-of-phase with the sound pressure in the counter wave at x=0, so that there is little phase
pipe, QJ does not contribute to the power stroke shift. Over the rest of the path, the acoustic particle
for turbulent cavity resonance. displacement is smaller than the displacement due
to the stability wave, which has large gain in a jet,
5.3 Organ Pipes so that k1 and k, are approximately the same over
As shown in Fig. 3, the organ pipe jet may be the entire path of the wave.
thought of as a pair of shear layers of opposite We will not attempt to integrate Eq.(24) for the
vorticity, separated by the jet width, δ.Although fulljet. On account of the standing wave conditions
on account of their proximity, the two shear layers near x=0 and the fact that k1* is a complex number,
exhibit coupled motion, with propagation constants a simple result such as Eq. (25) is not feasible.
characteristic of the system as a whole, force-drive Some things can be said about the solution however.
and jet-drive may still be thought of as coming from Generally speaking, jet width is small compared to
each layer separately. The lower layer produces the cut-up, H. For a typical diapason pipe, for
forces across the mouth plane according to Eq. (24), example, the ratio of δ/H might be 1/30, while the
and delivers a jet-drive flow ratio ofδ0/H is even smaller. Therefore organ pipe
jets voiced for bright tone must fall into the limiting
(39)
category for which eddies are expected to travel
to the pipe each cycle. On account of the in along the jet. Currently halfjet resonators with
stabilitygrowth of the jet displacement along its δ0/H in this range are being studied at the U.S.

path,ξ(H,t) does not vanish for the organ pipe and Naval Academy (results are still unpublished), and
we expect a contribution from jet-drive so long as it has been found that below a critical value of δ0/H,
there is a component of QJ in phase with the sound the spectrum of the radiated sound suddenly be
pressure in the pipe. In an ideal top-hat jet, there comesvery rich, indicating transition to a new type
is no moving fluid below the dividing surface and of drive. It appears that for shear layers thin
so no return flow is removed from the pipe by the enough to permit the buildup of spinning vortices
lower layer. The upper layer, however, removes in the mouth, a large inertia is added to the motion
an amount-QJ from the jet, that is, the same of the layer, producing impulsive drive at the lip,
amount out-of-phase, and introduces no fluid from with a resulting spectrum not unlike that of a di
the outside. Provided the maximum displacement apasonpipe. The matter is still under study, but
at the lip,ξ(H), is at leastδ/2, all the jet fluid con already shows much promise for increasing the
tributesto QJ. understanding of organ pipe physics, particularly
The drive flow QJ introduced inside the pipe is with regard to the manipulations involved in voic
amplified by the acoustic resonator to give an ing.
acoustic standing wave through the mouth of the
pipe, QM,with total acoustic discharge into the pipe 5.4 Organ Pipe with Wide Jet.
QP as given by Eq. (11). The flow-QJ produces Until now we have tacitly assumed that jet widths
no amplification, but does combine with the acoustic are narrow enough that upper and lower shear layers
flow in the mouth,-QM, to produce the total radiat are able to move in a coupled, sinuous mode and
eddischarge,-QP. Therefore, the total radiation that both shear layers are able to react with the
from either end of an open pipe is the same, as cavity. Clearly, as the width of the jet becomes
shown experimentally by Coltman.32) greater, there must come a point at which one or
The forces from the upper layer act in phase with both of these assumptions is not valid. Assuming
those from the lower layer when the jet motion is that the width of the shear layer lies within the range

21
 J. Acoust. Soc. Jpn.(E) 13, 1 (1992)

for instability for the halfjet, what are the limits on The historical controversy as to whether pipes are
δthat distinguish halfjet from fulljet resonator and driven by force-drive or jet-drive is thus resolved
is there a continuous operational mode for the in in favor of both. Jet-drive is possible whenever the
betweenstate? Clearly there are two ratios that jet arrives at the lip with finite displacement and
need to be specified in order to distinguish single with a component in phase with the sound pressure
shear layer from double shear layer situations:δ/λ, in the pipe.
whereλis the wavelength of the stability wave along The present approach shows the drive system to
the shear layer, determines the complex propagation be inherently linear over a range of shear layer
constant for the top-hat profile, whileδ/H deter- widths, so that sinusoidal oscillation is readily
mines whether one or two shear layers interact with achievable. For voicing situations calling for
the lip. For the former, Rayleigh showed that the spectra rich in harmonics, nonlinear drive is attained
top-hat eigenvalues change smoothly into step by resorting to a smaller ratio of jet width to lip
profileeigenvalues forδ/λ ≫1, so we are assured cutup which, at high enough Reynolds number,
that there is a proper value of wave propagation permits the growth of eddies which result in im
constant,κ+jα, for all widths of the jet. Likewise pulsiveaction at the lip.
it seems obvious that for δ/H≫1, it must be impossi
blefor the full jet to interact with the lip by crossing ACKNOWLEDGMENTS
into the pipe, regardless of whether the sinuous The work herein reported was supported by the
mode is present or not. United States Naval Academy Research Council.
It would be interesting to investigate experi

mentallyas to whether there is an in-between oscil REFERENCES

latorystate in which interaction at the lip is dom
1) G. B. Brown,"The vortex motion causing edge
inatedby lower shear layer dynamics while a
tones," Proc. Phys. Soc. London 49, 493-507,
weak upper shear layer oscillation shows some 508-521 (1937).
coupling of the entire jet with the acoustic field in 2) J. Jeans, Science and Music, 1st paperback ed.
the pipe. For this case, we should expect the prop (Cambridge University Press, Cambridge, 1961),
agationconstant to be somewhere between that for p.131.
halfjet and that for jet, while at resonance the value
3) D. Bechert,"Schwingungen eines akustisch
gesteurten ebenen Freistrahls," 5th Int. Congr.
of kH might be in the neighborhood of 1.25ƒÎ, with
Acoust., Liege, Paper L35 (1965).
the possibility of a critical value of Uc/U according
4) L. Cremer,"Berechnung der selbsterregten
to the way in which type A and D forces are in Schwingungen von Orgelpfeifen," 5th Int. Congr.
volved. Acoust., Liege, Paper L64 (1965).
5) H. Ising, Uber die Klangerzeugung in Orgel
6. CONCLUSIONS pfeifen,"5th Int. Congr. Acoust. Liege, Paper
M55 (1965).
A new unified method of treating the sounding
6) L. Cremer and H. Ising,"Die selbsterregten
mechanism of organ pipes and cavity resonators
Schwingungen von Orgelpfeifen," Acustica 19,
has been developed to replace the control-volume 143-153 (1968).
method which until recently has been the most 7) H. Ising,"Uber die Klangerzeugung in Orfel
widely accepted theory. It has been shown that pfeifen,"Doktor-Ingenieur Thesis, Berlin (1969).
shear layers contain a dividing surface which acts
8) J. M. Coltman,"Sounding mechanism of the flute
and organ pipe," J. Acoust. Soc. Am. 44, 983-992
as a vibrating membrane whose motions control the
(1968).
production of sound in the resonant cavity. This
9) J. M. Coltman,"Jet drive mechanisms in edge
vorticity-containing shear layer, which is the prin tones and organ pipes," J. Acoust. Soc. Am. 60,
cipleagent in the dynamics of jet motion, can be 725-733 (1976).
studied apart from the whole jet. In this way in 10) S. A. Elder,"On the mechanism of sound produc
sightsinto organ pipe physics can be found through
tionin organ pipes," J. Acoust. Soc. Am. 54,1554-
1564 (1973).
the study of the simpler but related phenomenon of
11) N. H. Fletcher,"Jet-drive mechanism in organ
cavity resonators.
pipes," J. Acoust. Soc. Am. 60, 473-475 (1976).
Both jet-drive and force drive contributions are 12) N. H. Fletcher and S. Thwaites,"The physics of
found to be involved in the sounding mechanism. organ pipe," Sci. Am. 248(1), 94-103 (1983).

22
 S. A. ELDER:SOUND PRODUCTION IN ORGAN PIPES

13) See, for example, the tutorial material in the text by 23) S. A. Elder, T. M. Farabee, and F. C. DeMetz,
N. H. Fletcher and T. D. Rossing, The Physics of "Mechanisms of flow -excited cavity tones at low
Musical Instruments (Springer-Verlag, New York, Mach number," J. Acoust. Soc. Am. 72, 532-549
1991), Chap. 16. (1982).
14) S. Yoshikawa and J. Saneyoshi,"Feedback excita 24) We must necessarily restrict our attention to the
tionmechanism in organ pipe," J. Acoust. Soc. Jpn. case for which the vorticity has not rolled up into
(E) 1, 175-191 (1980). an eddy.
15) S. A. Elder,"On the theory of resonant cavity 25) See Ref. 17), p.391.
excitation by an unstable jet," paper submitted to 26) H. Lamb, Hydrodynamics, 6th ed., 1st paperback
J. Acoust. Soc. Am.(1992). ed. (Dover, New York, 1945), Sect. 9.
16) for a comprehensive bibliography, see W. K. Blake, 27) Ideally, in a fully developed velocity profile, the
Aero-Hydroacoustics for Ships, Vol. I (David average inflection point of a free shear layer lies
Taylor Naval Ship Research and Development along the plane z=0. In the simple model, velocity
Center, Carderock, MD, 1984), pp.219-228a. perturbations move about this point while the
17) Lord Rayleigh (J. W. S. Strutt), The Theory of average profile remains stationary, even though the
Sound (Dover, New York, 1945), Vol. 2, Chap. 21. actual velocity profile moves up and down with the
18) L. Rosenhead,"The formation of vortices from a wave as it descends into the cavity along the path.
surface of discontinuity," Proc. R. Soc. London 28 ) See examples of similar analysis in Refs. 6), 21), and
Ser. A134, 170-192 (1931). 23).
19) See Ref. 17), p.402. 29 ) See Refs. 6) or 21).
20) A. Michalke,"On spatially growing disturbances 30) U. Ingard and H. Ising,"Acoustic nonlinearity of
in an inviscid shear layer," J. Fluid Mech. 23, 521- an orifice," J. Acoust. Soc. Am. 42, 6-17 (1967).
544 (1965). 31) F. C. DeMetz and T. M. Farabee,"Laminar and
21) S. A. Elder "Self-excited depth-mode resonance for turbulent shear flow induced cavity resonances,"
a wall-mounted cavity in turbulent flow," J. Acoust. #77-1293 AIAA 4th Aeroacoustics Conf., Atlanta,
Soc. Am. 64, 877-890 (1978). GA (October 3-5, 1977).
22) S. A. Elder,"Forced oscillations of a separated 32) J. M. Coltman,"Sound radiation from the mouth
shear layer with application to cavity flow-tone of an organ pipe," J. Acoust. Soc. Am. 46, 477
effects," J. Acoust. Soc. Am. 67, 774-781 (1980). (1969).

23