Expansions for infinite or finite plane circular time-reversal

mirrors and acoustic curtains for wave-field-synthesis

Tim Mellow
Mellow Acoustics Ltd., Farnham, Surrey GU9 9QH, United Kingdom

€inena)
€rkka
Leo Ka
Nokia Research Center, 00180, Helsinki, Finland
(Received 22 July 2013; revised 30 December 2013; accepted 21 January 2014)
An acoustic curtain is an array of microphones used for recording sound which is subsequently
reproduced through an array of loudspeakers in which each loudspeaker reproduces the signal from
its corresponding microphone. Here the sound originates from a point source on the axis of symme-
try of the circular array. The Kirchhoff-Helmholtz integral for a plane circular curtain is solved ana-
lytically as fast-converging expansions, assuming an ideal continuous array, to speed up
computations and provide insight. By reversing the time sequence of the recording (or reversing the
direction of propagation of the incident wave so that the point source becomes an “ideal” point
sink), the curtain becomes a time reversal mirror and the analytical solution for this is given simulta-
neously. In the case of an infinite planar array, it is demonstrated that either a monopole or dipole
curtain will reproduce the diverging sound field of the point source on the far side. However,
although the real part of the sound field of the infinite time-reversal mirror is reproduced, the imagi-
nary part is an approximation due to the missing singularity. It is shown that the approximation may
be improved by using the appropriate combination of monopole and dipole sources in the mirror.
C 2014 Acoustical Society of America. [http://dx.doi.org/10.1121/1.4864459]
V

PACS number(s): 43.60.Tj, 43.20.Bi [HCS] Pages: 1256–1277

I. INTRODUCTION The second problem to be studied is that of the time re-

In this paper, we investigate two practical problems
which may be evaluated using the Kirchhoff-Helmholtz
boundary integral. The first is the “acoustic curtain”1 in
which a planar array of microphones is used during a record-
ing and the output of each microphone is heard through a
corresponding loudspeaker in an array during playback. This
is known as “wave field synthesis.”2,3 In practice an array of
microphones is usually made just large enough to make a
reasonable approximation to an infinite array and the spacing
between each is made small enough to avoid spatial aliasing.
It should be noted that the sound field on the opposite side of
the curtain from the sources is reproduced on both sides of
the loudspeaker array if they are omnidirectional monopoles.
The same also applies to dipoles except that the two sides
have opposite phase.
Although it is not physically possible to reproduce the
sound field on the same side of the acoustic curtain as the
sources, this field may be calculated using a technique
known as near-field acoustic holography.4,5 The Green’s
function within the Kirchhoff-Helmholtz boundary integral
is itself in integral form so that the resulting expression may
be considered as a double Fourier transform. The first trans-
form produces a spatial spectrum of the field within the
array, which is propagated in the spatial frequency domain
to the (parallel) observation plane where the second or
inverse transform takes place.
a)
Author to whom correspondence should be addressed. Electronic mail:
leo.m.karkkainen@nokia.com
versal mirror. Here the recording from the microphone array
is played back through the loudspeaker array in reverse.
However, if the recording is that of a simple periodic point
source, we find that the wave from the loudspeaker array
simply converges toward the position of the original source
and then diverges again on the other side without reproduc-
ing the singularity of the original.6 This is because the imagi-
nary part of the pressure changes sign as the converging
wave switches to a diverging one. It is therefore zero at the
position of the original point source instead of infinity,
although the finite real part is faithfully reproduced. In order
to illustrate this, let us consider a diverging spherical wave
such as that produced by a point source
eikr sin kr þ i cos kr
p~d ðrÞ ¼ ikqcU~ d ¼ kqcU~ d ; (1)
4pr 4pr
where r is the radial ordinate, U~ d is the volume velocity of
the source at r ¼ 0, k ¼ x/c is the wave number, x is the
angular frequency, c is the wave propagation speed, and q is
the density of the medium. Then the expression for a con-
verging wave, such as that produced by an “ideal” point
sink, is
eikr sin kr þ i cos kr
p~c ðrÞ ¼ ikqcU~ c ¼ kqcU~ c ; (2)
4pr 4pr
where U~ c is the volume velocity at r ¼ 0. Notice that in both
cases the real part of the pressure is finite at the origin r ¼ 0
whereas the imaginary part is singular. Imagine now that a
converging spherical wave from a spherical time-reversal

1256 J. Acoust. Soc. Am. 135 (3), March 2014 0001-4966/2014/135(3)/1256/22/$30.00 C 2014 Acoustical Society of America
V
mirror passes through the origin and comes back out again
as a diverging one (without any reflections from the inner
surface of the mirror). There is neither a source nor a sink at
the origin so that, for reasons of continuity, we now have
U~ d ¼ U~ c ¼ U~ which yields
ikr
e sin kr þ i cos kr
p~d ðrÞ ¼ ikqcU~ ¼ kqcU~ ; (3)
4pr 4pr
e ikr
sin kr  i cos kr
p~c ðrÞ ¼ ikqcU~ ¼ kqcU~ ; (4)
4pr 4pr
and a resultant time-reversal field of
 
eikr eikr
p~tr ðrÞ ¼ p~d ðrÞ þ p~c ðrÞ ¼ ikqcU~ 
4pr 4pr
sin kr
¼ kqcU~ d ; (5)
2pr
which is continuous at r ¼ 0. Hence the real parts of p~d ðrÞ
and p~c ðrÞ have the same signs but signs of the imaginary
parts are opposite. Also, the resultant pressure gradient at the
origin @ p~ðrÞ=@rjr¼0 is zero, so that as the converging waves
pass through the origin and back out again as diverging ones,
it is equivalent to their being reflected from a rigid termina-
tion at the origin.
It has been said that when time-reversing the waves pro-
duced by dropping a pebble in a pond, a pebble must rise out
of the water at the end of the sequence7 in order to truly rep-
licate the original field. Hence one could place an active
point sink at the location of the original source,6 which is in
effect a point source with reversed polarity. We see that
combining such a source from Eq. (3) with the time-reversal
field of Eq. (5) gives the ideal point sink field of Eq. (4). In
other words, p~c ðrÞ ¼ p~tr ðrÞ  p~d ðrÞ.
Alternatively one could use a passive sink in the form of
a sphere, which would need to have a surface impedance
equal to that of free space and a diameter somewhat larger
than the wavelength. This is described in greater detail in
Appendix C.
Returning now to the problem of the planar time-
reversal mirror, using a sink would require a priori knowl-
edge of the original source position. To overcome this, Conti
et al. propose “near-field time-reversal,”8 which calculates
the time-reversal field in a similar manner to near-field
acoustic holography. Furthermore, Rosny and Fink demon-
strate the advantage of using a purely dipole mirror in near-
field time reversal in order to achieve sub-wavelength
resolution.9
However, in this paper we show that a reasonable
approximation (albeit limited by the classical half-
wavelength diffraction limit) may be made outside the im-
mediate vicinity of the source without using sinks or near-
field techniques (Fourier transforms) in order to represent a
practical physical situation. It is interesting to note that while
the amplitude of the imaginary part of the pressure may be
zero at the location of the original point source, it can
become quite large on either side at high frequencies, albeit
of opposite polarity, like a dipole. Hence it is possible to
J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014
T. Mellow and L. Ka
observe a “caustic” or “focusing” effect in the time-reversal
field. Discretization effects are not considered here as the
mirror is continuous, but this does give us the theoretical
performance limit of a discretized system and most impor-
tantly yields analytical solutions to the Kirchhoff-Helmholtz
boundary integral by means of suitable expansions. For
instance, the expansion for the pressure-gradient (used in the
monopole integral) draws on a previous result for a resilient
disk in free space.10 The fast converging expansions devel-
oped here are much more amenable to computation than
evaluating the original integrals numerically (especially
those for the infinite mirror and curtain) and have been used
by the authors to generate computer animations for further
insight.
Also, complicating factors such as room reflections are
not considered as the purpose of this paper is to study the in-
herent limitations of the problem with relatively simple and
easily computable expressions. Hence, the curtain and mirror
are evaluated in free space, although a model of an infinite
wall is included automatically in the monopole case through
the boundary condition of zero velocity in the plane in which
it is located.
The Kirchhoff-Helmholtz boundary integral is a mathe-
matical expression of the Huygens-Fresnel principle, which
states that if each point on a wave-front is replaced by a
point source, the outward traveling wave will be repro-
duced.11,12 The fact that this also produces an inward trav-
eling wave need not matter if the surface on which the
point sources are located fully envelopes the original source
and we are only interested in the “external” field. This is
also the case if the point sources cover an infinite plane
because the opposite side of a plane to the original source is
still fully isolated from it. In an infinite array, the sources
may be monopoles in which case it is necessary to evaluate
the velocity at each point. Alternatively, they may be
dipoles in which case the pressure at each point must be
evaluated.
If field strength at an arbitrary location r due to a point
source at r0 is expressed by the Green’s function g(r|r0) and
the pressure at r0 is p(r0), the Kirchhoff-Helmholtz boundary
integral may be written as
ðð
@ @
p~ðrÞ ¼ gðrjr0 Þ p~ðr0 Þ  p~ðr0 Þ gðrjr0 ÞdS0 ;
@n0 @n0
(6)
where the first term (or monopole part) is the integral of the
product of the inward pointed normal gradient of the bound-
ary values of p(r0) and g(r|r0) over the surface and the sec-
ond term (or dipole part) is the integral of the product of the
boundary values of p(r0) and the inward pointed normal gra-
dient of g(r|r0) over the surface. In this form, we have a
somewhat asymmetrical radiator because there will be a
degree of cancellation (depending on the surface pressure
and pressure-gradient values) between the monopole and
dipole terms on one side of the surface, but in this paper we
shall consider each term separately and how best to combine
them.
€ rkka
€inen: Equations for virtual acoustic sources 1257
 In the case of a finite circular planar source, the in the plots is at w ¼ 0 and z ¼ 1. In the equations, polar coor-
Kirchhoff-Helmholtz integral is still valid over the surface of dinates are also used where w ¼ r sin h and z ¼ r cos h.
the source provided that the boundary conditions in the plane
beyond its perimeter are met. That is, we would have zero
normal pressure-gradient (or velocity) if it were a monopole II. FINITE PLANE CIRCULAR ACOUSTIC CURTAIN
or zero pressure if it were a dipole. OR TIME-REVERSAL MIRROR USING MONOPOLE
While the theory of the Kirchhoff-Helmholtz inte- (PRESSURE-GRADIENT) SOURCE ARRAY
gral11,12 predicts that it is sufficient to use either monopole
A. Normal pressure-gradient or velocity in the
pressure-gradient (velocity) sources or dipole pressure sour- xy-plane (z 5 0) due to original point source
ces to create an infinite planar acoustic curtain, the choice of
combination to give the best results in a time-reversal mirror Before solving the monopole Kirchhoff-Helmholtz
is not predicted. However, in this investigation it is shown integral, we need to derive an expansion for the pressure-
that a pure dipole time-reversal mirror yields best results gradient in the plane of the curtain/mirror due to the orig-
between the plane of the source and the mirror, whereas a inal point source. This expansion will then be inserted
combination of a double-strength monopole and anti-phase into the monopole part of the Kirchhoff-Helmholtz inte-
dipole time-reversal mirror works best elsewhere. The mirror gral. The point source shown in Fig. 1(a) produces the
is created by reversing the direction of propagation of the pressure field
incident wave from the original source, which is achieved by
changing the signs in the exponents representing the p~S ðw; /; zÞ ¼ ikqcU~ S gðw; /; zjr0 ; h0 Þ; (7)
pressure-gradient and pressure terms in the cases of the
monopole and dipole integrals, respectively (i.e., changing where k ¼ x/c is the wave number, q is the density of air, c
eikr to eikr). The point source is a particularly useful is the wave propagation speed, and U~ S is the volume velocity
example because it can form the basic building block of any of the source. The Green’s function gðw; /; zjr0 ; h0 Þ is given
source distribution. in cylindrical-spherical coordinates by
Section II of this paper deals with finite and infinite cur-
tains and mirrors using the monopole part of the Kirchhoff- e7ikr1
Helmholtz integral while Sec. III deals with those using the gðw; /; zjr0 ; h0 Þ ¼ ; (8)
4pr1
dipole part. In both cases, Subsection A derives the fields in
the plane of the curtain/mirror, which are the pressure- where w, /, and z are the radial, azimuthal, and axial cy-
gradient in Sec. II and the pressure in Sec. III. Subsection B lindrical ordinates, respectively, r0 and h0 are the radial
then presents expressions for the pressure field where the dis- and inclinational spherical coordinates, respectively, and
tance from the origin to the observation point is greater than r1 is the distance from the source to the observation point
the radius of the mirror/curtain (r > a). Subsection C where
presents those for where the distance from the origin to the
observation point is less than the radius of the mirror/curtain
(r < a). Full derivations are given in Appendix A. In order to
avoid duplicating equations, it should be noted that the expo-
nents in this paper can have either negative signs, represent-
ing diverging waves, or positive ones, representing
converging waves. In such cases, the sign will be in the form
7 where the upper sign represents the acoustic curtain solu-
tion and the lower sign represents the time-reversal mirror
one. Similarly, the spherical Hankel function will appear in
the form hð2n1Þ
n , where the “2” denotes the acoustic curtain
solution and the “1” represents the time-reversal mirror one.
Some examples of calculated fields are presented using these
solutions and discussed. In particular, the differences
between the monopole and dipole time-reversed fields and
the original point source are highlighted, with some sugges-
tions for optimum combinations.
In all of the three-dimensional pressure-field plots, the
circular curtain/mirror is located in the xy-plane with its cen-
ter at the origin. Due to symmetry, the magnitudes of fields
are the same on both sides of the xy-plane. Therefore, for
convenience, they are plotted on the positive-z side only in
coordinates w/d
normalized axially-symmetric cylindricalpffiffiffiffiffiffiffiffiffiffiffiffiffiffi FIG. 1. Geometry of (a) field due to original point source and (b) field repro-
duced by time-reversal mirror or acoustic curtain. Although field reproduced
and z/d, where the radial ordinate is w ¼ x2 þ y2 , and d is by the latter originates from the negative-z side of the curtain, the same field
the axial distance between the original point source and the is reproduced on both sides of the x-y plane so it is convenient to calculate it
origin. The normalized position of the original point source on the positive-z side.

1258 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
 r12 ¼ r02 þ w2 þ z2  2r0 ðw cos / sin h0 þ z cos h0 Þ: (9) @
g0 ðw; /jr0 ; h0 Þ ¼ gðw; /; zjr0 ; h0 Þjz¼0
@z
We eventually let r0 ¼ d and h0 ¼ 0. Note that the sign of the 1 @
exponent can be positive or negative. A negative sign denotes ¼ gðw; /jr0 ; h0 Þ; (11)
w cos / @h0
an outgoing wave (or source) and a positive sign denotes an
incoming one (or sink). The normal pressure-gradient in the where
z ¼ 0 plane is given by

@ e7ikr
p~ ðw; /; zÞjz¼0 ¼ ikqcU~ S g0 ðw; /; zjr0 ; h0 Þ; (10) gðw; /jr0 ; h0 Þ ¼ ; (12)
@z S 4pr

where g0 is the Green’s function normal gradient. By reci- and r 2 ¼ r02 þ w2  2r0 w cos / sin h0 . The Green’s function
procity, this can be obtained from a result previously derived can then be expanded using Gegenbauer’s addition
as follows10 theorem13

8
>
>
> ik X1
>
> ð2q þ 1Þhqð2n1Þ ðkr0 Þjq ðkwÞPq ðcos / sin h0 Þ; r0 > w
< 4p q¼0
gðw; /jr0 ; h0 Þ ¼ (13)
>
>
> ik X1
>
> ð2q þ 1Þjq ðkr0 Þhqð2n1Þ ðkwÞPq ðcos / sin h0 Þ; w > r0 ;
: 4p q¼0

ð2Þ
where hq is the spherical Hankel function for an outgoing
wave defined by @ k2 qcU~ S X
1
p~S ðw; zÞjz¼0;0wd ¼  ðq þ 1Þ
@z 2pw q¼0
hð2Þ
p ðkrÞ ¼ jp ðkrÞ  iyp ðkrÞ: (14)
 ð4q þ 3Þj2qþ1 ðkdÞ
ð1Þ ð2n1Þ
hq is the spherical Hankel function for an incoming wave  h2qþ1 ðkwÞP2qþ2 ð0Þ: (20)
defined by
The particle velocity in the z ¼ 0 plane is then given by
hð1Þ
p ðkrÞ ¼ jp ðkrÞ þ iyp ðkrÞ; (15)
1 @
u~0 ðwÞ ¼ p~ ðw; zÞjz¼0 : (21)
and jp, yp are the spherical Bessel functions defined by ikqc @z S
rffiffiffiffiffiffiffi
p The point source will be replaced with a planar array of sour-
jp ðkrÞ ¼ Jpþ1=2 ðkrÞ; (16)
2kr ces at z ¼ 0 with this velocity distribution.
rffiffiffiffiffiffiffi
p
yp ðkrÞ ¼ Ypþ1=2 ðkrÞ: (17) B. Pressure field where the distance from the origin to
2kr
the observation point is greater than the radius of the
After gathering Eqs. (10)–(13), differentiating with respect mirror or curtain (r > a)
to h0 and setting h0 ¼ 0 and r0 ¼ d, we have We now replace the point source with a circular monop-
@ k2 qcU~ S X
1 ole array of radius a in the x-y plane with its center at the
p~S ðw; zÞjz¼0 ¼  ðq þ 1Þ origin, as shown in Fig. 1(b), using the monopole Kirchhoff-
@z 4pw q¼0
Helmholtz boundary integral. Then the reconstructed pres-
 ð2q þ 1Þhqð2n1Þ ðkdÞjq ðkwÞPqþ1 ð0Þ: sure field p~M ðr; hÞ is given by integrating in w0 and /0 over
(18) the surface of the circular mirror/curtain

However, only the odd terms are non-zero, so that ð 2p ð a

@
p~M ðr; hÞ ¼ 2 p~S ðw0 ; z0 Þjz0 ¼0
@ k2 qcU~ S X
1
0 0 @z0
p~S ðw; zÞjz¼0;w>d ¼  ðq þ 1Þð4q þ 3Þ
@z 2pw q¼0  gðr; hjw0 ; /0 ; z0 Þjz0 ¼0 w0 dw0 d/0 ; (22)
ð2n1Þ
 h2qþ1 ðkdÞj2qþ1 ðkwÞP2qþ2 ð0Þ;
where g(r, hjw0 , /0, z0) is the Green’s function in spherical-
(19) cylindrical coordinates defined by

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1259
 gðr; hjw0 ; /0 ; z0 Þ ¼ eikr1 =ð4pr1 Þ; (23) If a < d, then the outer solution (r > a) is given by

k2 qcU~ S X
P X
where r and h are the radial and inclinational spherical coor- Q
ð2Þ
dinates, respectively, w0, /0, and z0 are the radial, azimuthal, p~M ðr; hÞjr>a;d>a ¼ 7i ð1Þpþq gðMÞ
p;q h2p ðkrÞ
4p2 p¼0 q¼0
and axial cylindrical ordinates, respectively, and r1 is the dis-
ð2n1Þ ðM1Þ
tance from the source to the observation point where  h2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ; (25)

r12 ¼ r2 þ w20 þ z20  2rðw0 cos /0 sin h þ z0 cos hÞ: (24)

ðMÞ ðM1Þ
where gp;q and fp;q ðkaÞ are given by Eqs. (A10) and (A11),
In Appendix A, Eq. (22) is solved to give the following respectively.
solutions: If a  d, then the outer solution (r > a) is given by

8 9
< ikr  =
~
kqcU S e 0
e ikr1
2k P X
X Q
ð2Þ
p~M ðr; hÞjr>a;0da ¼i þ 7 ð1Þpþq  gðMÞ ðM2Þ
p;q h2p ðkrÞj2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ ; (26)
8p : r0 r1 p p¼0 q¼0 ;

ðM2Þ
where fp;q ðkaÞ is given by Eq. (A15), r02 ¼ r 2 þ d2 þ 2rdjcos hj and r12 ¼ r2 þ d 2  2rdjcos hj.

C. Pressure field where the distance from the origin to the observation point is less than the radius of the mirror
or curtain (r < a)
If a < d, then the inner solution (0  r  a) is given by

8 9
 
kqcU~ S < e7ikr0 e7ikr1 P X
2k X =
Q
ð2n1Þ
p~M ðr; hÞj0ra;d>a ¼i  7 ð1Þpþq  gðMÞ
p;q j2p ðkrÞh2qþ1 ðkdÞf ðM3Þ
p;q ðkaÞP 2p ðcos hÞ ; (27)
8p : r0 r1 p p¼0 q¼0 ;

ðMÞ ðM3Þ
where gp;q and fp;q ðkaÞ are given by Eqs. (A10) and (A24), respectively.
If a  d, then the inner solution (0  r  a) is given by

8 9
kqcU~ S <eikr0 þe7ikr0 eikr1 e7ikr1 2k XP X =
Q
pþq ðMÞ ðM4Þ
p~M ðr;hÞj0ra;0da ¼ i þ 7 ð1Þ gp;q j2p ðkrÞj2qþ1 ðkdÞfp;q ðkaÞP2p ðcoshÞ ;
8p : r0 r1 p p¼0 q¼0 ;

(28)

ðM4Þ
where fp;q ðkaÞ is given by Eq. (A34), r02 ¼ r 2 þ d2 þ 2rdjcos hj and r12 ¼ r2 þ d 2  2rdjcos hj.

III. INFINITE PLANAR ACOUSTIC CURTAIN OR TIME-REVERSAL MIRROR USING MONOPOLE

(PRESSURE-GRADIENT) SOURCE ARRAY
ðM4Þ ðM4Þ
If a ¼ 1, then fp;q ðkaÞja!1 ¼ 2ð1Þpþq in the time-reversal mirror solution and fp;q ðkaÞja!1 ¼ 0 in the acoustic cur-
tain solution. Hence Eq. (28) reduces to

kqcU~ S
p~M ðr; hÞja¼1 ¼ i
84p
0 1
>
> ikr0 ikr0 ikr1 ikr1 P X
X Q
>
> e þ e e  e 2k
<@
>
2r0
þ
2r1
þ gðMÞ j2p ðkrÞj2qþ1 ðkdÞP2p ðcos hÞA ; time-reversal mirror
p p¼0 q¼0 p;q

>
>
> eikr0
>
>
: ; acoustic curtain:
r0
(29)

1260 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
IV. FINITE PLANE CIRCULAR ACOUSTIC CURTAIN OR TIME-REVERSAL MIRROR USING DIPOLE (PRESSURE)
SOURCE ARRAY

A. Pressure field in plane of curtain or mirror due to original point source

Before solving the dipole Kirchhoff-Helmholtz integral, we need to derive an expansion for the pressure in the plane of
the curtain/mirror due to the original point source. This expansion will then be inserted into the dipole part of the Kirchhoff-
Helmholtz integral. The point source shown in Fig. 1(a) produces the pressure field given by Eq. (7). If we set r0 ¼ d and
h0 ¼ 0 in Eq. (13), before inserting it in Eq. (7) and note that only the even terms are non-zero [using the identities of Eqs.
(A5) and (A6)], we can write the following expansions for the pressure field due to the point source:

k2 qcU~ S X
1
ð1Þp ð4q þ 1ÞCðp þ 1=2Þ ð2n1Þ
p~S ðw; 0Þjd>w ¼  h2q ðkdÞj2q ðkwÞ; (30)
4p3=2 q¼0 p!

k2 qcU~ S X
1
ð1Þp ð4q þ 1ÞCðp þ 1=2Þ ð2n1Þ
p~S ðw; 0Þjw>d ¼  j2q ðkdÞh2q ðkwÞ: (31)
4p3=2 q¼0 p!

B. Pressure field where the distance from the origin to the observation point is greater than the radius of the
mirror or curtain
We now replace the point source with a circular dipole array of radius a in the x-y plane with its center at the origin, as
shown in Fig. 1(b), using the dipole Kirchhoff-Helmholtz boundary integral. Then the reconstructed pressure field p~D ðr; hÞ is
given by integrating in w0 and /0 over the surface of the circular mirror/curtain

ðp ða
@
p~D ðr; hÞ ¼ 2 p~S ðw0 ; 0Þ gðr; hjw0 ; /0 ; z0 Þjz¼0þ w0 dw0 d/0 ; (32)
p 0 @z0

where g(r, hjw0 ,/0,z0) is the Green’s function in cylindrical-spherical coordinates defined by

gðr; hjw0 ; /0 ; z0 Þ ¼ eikr1 =ð4pr1 Þ; (33)

where

r12 ¼ r2 þ w20 þ z20  2rðw0 cos /0 sin h þ z0 cos hÞ: (34)

In Appendix A, Eq. (32) is solved to give the following solutions:

If a < d, then the outer solution (r > a) is given by

k2 qcU~ S X
P X Q
ð2Þ ð2n1Þ
p~D ðr; hÞjr>a;d>a ¼ i ð1Þpþq gðDÞ ðD1Þ
p;q h2pþ1 ðkrÞ  h2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ; (35)
8p2 p¼0 q¼0

ðDÞ ðD1Þ
where gp;q and fp;q ðkaÞ are given by Eqs. (A42) and (A43), respectively.
If a  d, then the outer solution (r > a) is given by
8
 ikr 
kqcU~ S < 2k XP XQ
e 0 eikr1
p~D ðr; hÞjr>a;0da ¼i 6sgnðcos hÞ   ð1Þpþq
8p : r0 r1 p p¼0 q¼0
9
=
ð2Þ
gðDÞ h
p;q 2pþ1 ðkrÞj 2q ðkdÞf ðD2Þ
p;q ðkaÞP 2pþ1 ðcos hÞ ; (36)
;

ðD2Þ
where fp;q ðkaÞ is given by Eq. (A46), r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1261
C. Pressure field where the distance from the origin to the observation point is less than the radius of the mirror
or curtain
If a < d, then the inner solution (0  r  a) is given by
8
 7ikr 
kqcU~ S < 2k XP XQ
e 0 e7ikr1
p~D ðr; hÞj0ra;d>a ¼i 6sgnðcos hÞ þ  ð1Þpþq
8p : r0 r1 p p¼0 q¼0
9
=
ð2n1Þ
 gðDÞ j
p;q 2pþ1 ðkrÞh2q ðkdÞf ðD3Þ
p;q ðkaÞP2pþ1 ðcos hÞ ; (37)
;

ðDÞ ðD3Þ
where gp;q and fp;q ðkaÞ are given by Eqs. (A42) and (A54), respectively.
If a  d, then the inner solution (0  r  a) is given by

  ikr 
kqcU~ S e 0 þ e7ikr0 eikr1  e7ikr1
p~D ðr; hÞj0ra;0da ¼ i 6sgnðcos hÞ 
8p r0 r1
)
2k XP X Q
pþq ðD4Þ
 ð1Þ gp;q j2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ ; (38)
p p¼0 q¼0

ðD4Þ
where fp;q ðkaÞ is given by Eq. (A64), r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

V. INFINITE PLANAR ACOUSTIC CURTAIN OR TIME-REVERSAL MIRROR USING DIPOLE (PRESSURE) SOURCE
ARRAY
ðD4Þ ðD4Þ
If a ¼ 1, then fp;q ðkaÞja!1 ¼ 2ð1Þpþq in the time-reversal mirror solution and fp;q ðkaÞja!1 ¼ 0 in the acoustic cur-
tain solution. Hence Eq. (38) reduces to

kqcU~ S
p~D ðr; hÞja¼1 ¼ i
84p
  ikr 
>
> e 0 þ eikr0 eikr1  eikr1
>
> sgnðcos hÞ 
>
> 2r0 2r1
>
>
>
< P X
X Q 
2k ðDÞ
  g j2pþ1 ðkrÞj2q ðkdÞP2pþ1 ðcos hÞ ; time-reversal mirror (39)
>
> p p¼0 q¼0 p;q
>
>
>
>
>
> eikr0
>
: sgnðcos hÞ ; acoustic curtain:
r0

FIG. 2. Real part of reconstructed pressure field of finite monopole acoustic FIG. 3. Imaginary part of reconstructed pressure field of finite monopole
curtain at kd ¼ 20, where a ¼ d and PM ¼ 4pd~ p M =ðkqcU~ S Þ from Eq. (28). acoustic curtain at kd ¼ 20, where a ¼ d and PM ¼ 4pd~ p M =ðkqcU~ S Þ from
Plot is in cylindrical coordinates (w,z), where w ¼ r sin h and z ¼ r cos h and Eq. (28). Plot is in cylindrical coordinates (w,z), where w ¼ r sin h and
with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd. z ¼ r cos h and with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.

1262 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources

FIG. 4. Real part of reconstructed pressure field of finite dipole acoustic cur-
tain at kd ¼ 20, where a ¼ d and PD ¼ 4pd~ p D =ðkqcU~ S Þ from Eq. (38). Plot
is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h and with
the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.
VI. DISCUSSION
A. Acoustic curtain
Not surprisingly, the solutions given in Eqs. (29) and
(39), for infinite monopole and dipole acoustic curtains,
respectively, are simply the same as that for a point source
on the negative-z side of the plane when observed on the
positive side, although the field is symmetrical either side of
the xy-plane (except that the polarity is reversed in the case
of the planar dipole source).
Interestingly, Eqs. (28) and (38) for the finite circular
monopole and dipole planes, respectively, both have three
terms. The first and second terms relate to point sources on
the negative and positive-z sides of the xy-plane, respec-
tively. However, when we evaluate the field on the positive-
z side of an acoustic curtain, we find that their combination
eliminates the point source on the positive side of the xy-
plane leaving just the one on the negative side. However, the
waves from that source can only be “seen” on the positive
side because as soon as we evaluate the field on the negative
side they appear to radiate from a source on the positive side
onto the negative side. In other words, we have symmetry in
the case of a monopole curtain or anti-symmetry in the case
of a dipole one. The third term, or “diffraction” term, has a
complicated far-field directivity pattern at kd ¼ 20 with 60
FIG. 5. Imaginary part of reconstructed pressure field of finite dipole acous-
tic curtain at kd ¼ 20, where a ¼ d and PD ¼ 4pd~p D =ðkqcU~ S Þ from Eq. (38).
Plot is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h and
with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.
J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014
T. Mellow and L. Ka
FIG. 6. Real part of pressure field of original point source (or real part of
reconstructed pressure field of infinite monopole or dipole time-reversal mir-
ror) at kd ¼ 20, where PS ¼ 4pd~ p S =ðkqcU~ S Þ from Eq. (7), PM ¼
4pd~p M =ðkqcU~ S Þ from Eq. (29), and PD ¼ 4pd~ p D =ðkqcU~ S Þ from Eq. (39).
Plot is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h.
Also, r0 ¼ d and h0 ¼ 0.
side lobes between 0 and 90 off axis, none of which peak
at more than 23 dB below the central lobe or “Airy disk.”
The 6 dB width of the central lobe is just 60.7 .
The real and imaginary fields due to a finite monopole
acoustic curtain are shown in Figs. 2 and 3, respectively,
while the real and imaginary fields for a finite dipole one are
shown in Figs. 4 and 5, respectively. Admittedly, the recon-
structed fields are not optimum because the infinite plane has
been simply truncated at a ¼ d. Hence, there are discontinu-
ities in the velocity (monopole) or pressure (dipole) bound-
ary conditions equivalent to a rectangular window function
and a better solution would be to include an apodization
function.14 However, these plots give some indication of
what might be produced by a practical setup.
B. Time-reversal mirror
The time reversed solution for the infinite plane is a lit-
tle more complicated because the first two terms of Eqs. (28)
and (38) represent the imaginary part of the field of a point
FIG. 7. Imaginary part of reconstructed pressure field of infinite monopole
time-reversal mirror at kd ¼ 20, where PM ¼ 4pd~ p M =ðkqcU~ S Þ from Eq. (29).
Plot is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h and
with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd. See Fig. 6 for real part.
€ rkka
€inen: Equations for virtual acoustic sources 1263
FIG. 8. Imaginary part of reconstructed pressure field of infinite dipole time-
reversal mirror at kd ¼ 20, where PD ¼ 4pd~ p D =ðkqcU~ S Þ from Eq. (39). Plot
is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h and with
the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd. See Fig. 6 for real part.
FIG. 9. Imaginary part of pressure field of original point source at kd ¼ 20,
where PS ¼ 4pd~ p S =ðkqcU~ S Þ from Eq. (7). Plot is in cylindrical coordinates
(w,z) where w ¼ r sin h and z ¼ r cos h. Also, r0 ¼ d and h0 ¼ 0.
FIG. 10. Error function due to the difference between the imaginary monop-
ole and dipole pressure fields of Figs. 7 and 8, where PM ¼ 4pd~ p M =ðkqcU~ S Þ
and PD ¼ 4pd~ p D =ðkqcU~ S Þ. Plot is in cylindrical coordinates (w,z) where
w ¼ r sin h and z ¼ r cos h and with the expansion limits set to
P ¼ Q ¼ 4ka ¼ 4kd.
1264 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014
FIG. 11. Combined imaginary part of reconstructed pressure field of infinite
time-reversal mirror as follows: For z/d > 1, plot shows double-strength
monopole field less single-strength dipole field. For z/d < 1, plot shows
single-strength dipole field only, where PM ¼ 4pd~ p M =ðkqcU~ S Þ and
PD ¼ 4pd~ p D =ðkqcU~ S Þ. Plot is in cylindrical coordinates (w,z), where
w ¼ r sin h and z ¼ r cos h and with the expansion limits set to
P ¼ Q ¼ 4ka ¼ 4kd.
FIG. 12. Difference between imaginary reconstructed pressure fields shown
in Fig. 9 and Fig. 11, where PS ¼ 4pd~ p S =ðkqcU~ S Þ, PM ¼ 4pd~
p M =ðkqcU~ S Þ,
and PD ¼ 4pd~ p D =ðkqcU~ S Þ. Plot is in cylindrical coordinates (w,z), where
w ¼ r sin h and z ¼ r cos h and with the expansion limits set to
P ¼ Q ¼ 4ka ¼ 4kd.
FIG. 13. Real part of reconstructed pressure field of finite monopole time-
reversal mirror at kd ¼ 20, where a ¼ d and PM ¼ 4pd~ p M =ðkqcU~ S Þ from Eq.
(28). Plot is in cylindrical coordinates (w,z), where w ¼ r sin h and z ¼ r cos h
and with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.
€rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
FIG. 14. Imaginary part of reconstructed pressure field of finite monopole FIG. 16. Imaginary part of reconstructed pressure field of finite dipole time-
time-reversal mirror at kd ¼ 20, where a ¼ d and PM ¼ 4pd~ p M =ðkqcU~ S Þ reversal mirror at kd ¼ 20, where a ¼ d and PD ¼ 4pd~ p D =ðkqcU~ S Þ from Eq.
from Eq. (28). Plot is in cylindrical coordinates (w,z), where w ¼ r sin h and (38). Plot is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h
z ¼ r cos h and with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd. and with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.

source on the negative-z side and the real part (see Fig. 6) on decays with increasing z. This error is distributed unequally
the positive-z side, respectively. Also, the third term does between the monopole and dipole fields as follows:
not vanish as the radius approached infinity but simplifies as 
=ð~p M Þ þ p~E ; z>d
shown in Eqs. (29) and (39), and thus provides an additional pSÞ 
Jð~ (41)
=ð~ p M Þ þ p~E ; z < d;
imaginary term. However, the monopole and dipole time-
reversed imaginary pressure fields shown in Figs. 7 and 8, 
=ð~p D Þ þ 2~pE; z > d
respectively, differ somewhat from that of the simple point pSÞ 
Jð~ (42)
=ð~ p D Þ; z < d;
source shown in Fig. 9. To begin with, the singularity at the
center of the original point source is absent. This is due to except at (z, w) ¼ (d, 0) or in the immediate vicinity, due to
the fact that the waves from the planar source travel into the the loss of the singularity. It is perhaps not surprising that
focal point and then out again the other side. Hence, the the dipole field in the vicinity of the plane has the least error
phase changes in the plane of the focal point where the imag- because the dipole Kirchhoff-Helmholtz integral actually
inary pressure is positive on one side and negative on the contains the pressure distribution of the original point source
other. Therefore it is substantially canceled in the plane of within the plane. Therefore, the error in the plane should be
the focal point where the positive and negative imaginary zero. The imaginary pressure field of the original point
pressures meet, as can be seen in Figs. 7 and 8. Another no- source can be approximated using the following combination
ticeable difference is that these time-reversed imaginary of monopole and dipole planar sources:
pressure fields have ripples, due to interference patterns, 
which are absent in the original point source. When taking =ð2~ p M  p~D Þ; z > d
pSÞ 
Jð~ (43)
the difference between them, as shown in Fig. 10, an error =ð~ p D Þ; z<d
function becomes apparent, which is given by
as shown in Fig. 11, where the interference patterns have
J0 ðkwÞ largely disappeared. Unfortunately one cannot have simulta-
p M  p~D Þ ¼ p~E ¼ kqcU~ S
Jð~ : (40)
2pðz þ dÞ neously a good reconstruction of =ð~ p S Þ for z > d and z < d

Hence it represents an incoming cylindrical wave which

passes through the z axis and travels back out again. It also

FIG. 15. Real part of reconstructed pressure field of finite dipole time-
reversal mirror at kd ¼ 20, where a ¼ d and PD ¼ 4pd~ p D =ðkqcU~ S Þ from Eq.
(38). Plot is in cylindrical coordinates (w,z) where w ¼ r sin h and z ¼ r cos h
and with the expansion limits set to P ¼ Q ¼ 4ka ¼ 4kd.
FIG. 17. Error magnitude and phase between reconstructed pressure field of
finite monopole time-reversal mirror and original point source versus angle
off-axis h1 at a distance of r1 ¼ 2d from the position of the source,
where a ¼ 2d, r02 ¼ r12 þ 4d2 þ 4r1d cos h1, r2 ¼ r12 þ d2 þ 2r1d cos h1, and
cos h ¼ (r2 þ d2  r12)/(2rd). Plot of 20 log10(|PM/PS|). As would be expected
from a point source, PS is independent of h1.

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1265

FIG. 18. Error magnitude and phase between reconstructed pressure field of
finite dipole time-reversal mirror and original point source versus off-axis
angle h1 at a distance of r1 ¼ 2d from the position of the source,
where a ¼ 2d, r02 ¼ r12 þ 4d2 þ 4r1d cos h1, r2 ¼ r12 þ d2 þ 2r1d cos h1, and
cos h ¼ (r2 þ d2  r12)/(2rd). Plot of 20 log10(|PM/PS|). As would be expected
from a point source, PS is independent h1.
because the field superpositions are different. Again, this
breaks down at z ¼ d and w ¼ 0, and in the immediate vicin-
ity, due to the loss of the singularity. The difference between
the imaginary pressure field of Fig. 11 and the original point
source is shown in Fig. 12. Apart from the missing singular-
ity at the center, there are also errors in the plane of the point
source and either side of it due to the phase reversal.
Otherwise, the reconstruction appears to be fairly accurate.
Furthermore, the above relationships appear to hold at any
frequency.
The real and imaginary fields due to a finite monopole
time reversal mirror are shown in Figs. 13 and 14, respec-
tively, while the real and imaginary fields for a finite dipole
one are shown in Figs. 15 and 16, respectively. The same
comments made above, with respect to the acoustic curtain,
about the lack of a smooth windowing function apply here
also. For illustration, the errors of the monopole and dipole
time-reversal mirrors are plotted against the off-axis angle in
Figs. 17 and 18, respectively. At kd ¼ 20, the error is gener-
ally less than 63 dB between an angle of 6 and 56 .
Beyond this angle, the error increases because we are in the
region where the imaginary part of the pressure changes
sign. Also, there is a significant error on axis (at 0 ) due to
the finite nature of the radiator, which produces on-axis
peaks and nulls in the near field pressure. A similar phenom-
enon is seen with circular pistons,15 but this could no doubt
be mitigated through the application of suitable apodization.
It is worth noting that the figures shown in this paper
can be demonstrated more clearly using animation by
plotting
<ð~
p Þ cos t  =ð~
p Þ sin t: (44)
VII. CONCLUSIONS
The Kirchhoff-Helmholtz boundary integral has been
applied to the case of a point source, where the surface of
integration is planar and transparent. As expected, when the
plane of integration is infinite, and therefore fully bounds the
1266 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014
source, either the monopole or the dipole part (or both) can
be used to recreate a sound field outside the boundary that
results from sources inside it, provided that the normal ve-
locity or pressure values, respectively, are known at the
boundary. Using time reversal, the real part of the field
inside the boundary may also be recreated, although the
imaginary part is an approximation due to the differing
directions of propagation on either side of the source. The
resulting phase change in the plane of the source removes
the singularity present in the original.
However, it has been shown that the accuracy of the
approximation may be improved by using the right combina-
tion of the monopole and dipole parts of the boundary inte-
gral: A mirror comprising dipole pressure sources gives the
best results in the region between the mirror and the location
of the original point source, whereas a combination of
double-strength monopole velocity sources and anti-phase
dipole ones works best on the side of the source furthest
from the mirror. Because these superpositions are different,
it is not possible to obtain the best possible results every-
where simultaneously, so the choice of transducers will
depend on the application.
For a dipole array, the electrostatic loudspeaker has the
advantage of being an almost pure pressure source with very
little mechanical impedance. Also, a single transducer can
act as an array simply by partitioning the electrodes. Indeed,
such a loudspeaker has already been used successfully as a
form of an acoustic curtain for several decades by employing
an analog delay line to feed electrodes in the form of concen-
tric rings in order to replicate the phase and amplitude rela-
tionships of a notional incident wave and thus locate a
virtual point source behind the loudspeaker.16 Because of the
lack of suitable high-voltage low-power amplifiers, such
loudspeakers are still generally driven from stepping-up
transformers with the delay line relying on high-voltage pas-
sive components. Hence further development of suitable
semiconductors is needed in order for the processing to be
carried out digitally.
A potential application for the time-reversal principle is
in audio-visual entertainment whereby it could be possible to
make sounds appear to come from locations in front of the
loudspeakers. In order to do this, the phase relationship of
the notional incident wave would need to be reversed such
that the increasing phase lag with distance from the center of
the mirror would become an increasing phase advance. In
other words, it would be the reverse of the acoustic-curtain
loudspeaker just described, but with a suitable matrix array
and digital processing the position of the source could be
adjusted electronically.
An important application is in ultrasound where it would
be useful to be able to predict the position of the focal point
and adjust it electronically. Piezoelectric transducers with
oscillating thickness are commonly used because of their
near ideal piston movement and high efficiency at very high
frequencies.
As increasing processing power is making these kinds
of systems more feasible, the search for alternative and
affordable kinds of transducers that can fulfil the require-
ments of multi-loudspeaker systems is an active area of
€rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
research with renewed interest in thermo-acoustics17 and
electro-active polymers (artificial muscles).18
In this paper, expansions have been derived in order to
speed up the computations and these have been applied to fi-
nite time-reversal mirrors and acoustic curtains as used for
wave field synthesis. The increase in speed is especially
apparent in the case of an infinite mirror or curtain. For
example, the plots of Fig. 7 and Fig. 8 took less than one-
quarter of the time taken using numerical integration. The
expansions given in Secs. II–V agree with numerical calcula-
tion of the original integrals (see Appendix B). Hence they
are rigorous solutions, not approximations.
APPENDIX A: DERIVATIONS OF EXPANSIONS USED
TO CALCULATE FIELDS
1. Finite plane circular acoustic curtain or time-
reversal mirror using monopole (pressure-gradient)
source array
a. Pressure field where the distance from the origin to
the observation point is greater than the radius of the
mirror or curtain
The Green’s function of Eq. (23) can be expanded in the
same fashion as Eq. (13)

8
>
> ik X1
>
>
< 4p ð2p þ 1Þhð2Þ
p ðkrÞjp ðkw0 ÞPp ðcos /0 sin hÞ; r > w0
p¼0
gðr; hjw0 ; /0 ; z0 Þjz0 ¼0 ¼ gðr; hjw0 ; /0 Þ ¼ (A1)
>
> ik X1
>
>
: 4p ð2p þ 1Þjp ðkrÞhð2Þ
p ðkw0 ÞPp ðcos /0 sin hÞ; w0 > r:
p¼0

Then inserting Eqs. (20) and (A1) in Eq. (22) enables the two integrals in Eq. (22) to be separated as follows:
ða ð 2p
ik3 qcU~ S X
1 X
1
p~M ðr; hÞjr>a;d>a ¼ ð2p þ 1Þðq þ 1Þð4q þ 3Þ  jp ðkw0 Þj2qþ1 ðkw0 Þdw0 Pp ðcos /0 sin hÞd/0
4p2 p¼0 q¼0 0 0
ð2n1Þ
 hð2Þ
p ðkrÞh2qþ1 ðkdÞP2qþ2 ð0Þ: (A2)

The Legendre function Pp can be expanded using the following addition theorem13 (after setting one of the three angles in the
original formula to p/2)

X
1
Pp ðcos /0 sin hÞ ¼ Pp ð0ÞPp ðcos hÞ þ 2 ð1Þq Pq q
p ð0ÞPp ðcos hÞcos q/0 ; (A3)
q¼1

which leads to the identity

ð 2p X1
sin 2pq
Pp ðcos /0 sin hÞd/0 ¼ 2pPp ð0ÞPp ðcos hÞ þ 2 ð1Þq Pq q
p ð0ÞPp ðcos hÞ
0 q¼1
q
¼ 2pPp ð0ÞPp ðcos hÞ; for integral q: (A4)

It is also noted that13

pffiffiffi
p ð1Þp Cðp þ 1=2Þ
P2p ð0Þ ¼ ¼ pffiffiffi (A5)
p!Cð1=2  pÞ pp!

and

P2pþ1 ð0Þ ¼ 0; (A6)

so that, after substituting Eqs. (A4)–(A6) in Eq. (A2) and excluding the odd terms, only the radial integral remains as follows:

   
k3 qcU~ S X
1 X 1
ð1Þpþq ð4p þ 1Þð4q þ 3Þ 1 3
p~M ðr; hÞjr>a;d>a ¼ i C p þ C q þ
2p2 p¼0 q¼0 p!q! 2 2
ða
ð2Þ ð2n1Þ
 h2p ðkrÞh2qþ1 ðkdÞP2p ðcos hÞ j2p ðkw0 Þj2qþ1 ðkw0 Þdw0 : (A7)
0

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1267
A general solution to the integral over w0 can be written as19

ð n
k2 w20 ð2Þ ð2Þ
hð2Þ ð2Þ
m ðkw0 Þhn ðkw0 Þdw0 ¼  hð2Þ ð2Þ
m ðkw0 Þhn1 ðkw0 Þ  hm1 ðkw0 Þhn ðkw0 Þðn  mÞ
kðm þ n þ 1Þðn  mÞ
o
 ðhð2Þ ð2Þ
m ðkw0 Þhn ðkw0 ÞÞ=ðkw0 Þ ; (A8)

where hð2Þ
m ðkw0 Þ is defined in Eq. (15) so that in the case of Eq. (A7) we ignore the ym(kw0) terms, leaving just the jm(kw0)
terms. Hence, after truncating the summation limits to P and Q, the final expression for the pressure field becomes

k2 qcU~ S X
P X
Q
ð2Þ ð2n1Þ
p~M ðr; hÞjr>a;d>a ¼ 7i ð1Þpþq gðMÞ ðM1Þ
p;q h2p ðkrÞ  h2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ; (A9)
4p2 p¼0 q¼0

where

ð4p þ 1Þð4q þ 3ÞCðp þ 1=2ÞCðq þ 3=2Þ

gðMÞ
p;q ¼ (A10)
p!q!ðp þ q þ 1Þð2q  2p þ 1Þ

and

ðM1Þ  

fp;q ðkaÞ ¼ k2 a2 j2p ðkaÞj2q ðkaÞ  j2p1 ðkaÞj2qþ1 ðkaÞð2q  2p þ 1Þ j2p ðkaÞj2qþ1 ðkaÞ =ðkaÞ : (A11)

The solution for d  a is a little more complicated because the spherical Bessel and Hankel functions have to be exchanged in
part of the integral in order to achieve convergence
   
k3 qcU~ S X1 X1
ð1Þpþq ð4p þ 1Þð4q þ 3Þ 1 3
p~M ðr; hÞjr>a;0da ¼i C pþ C qþ
2p2 p¼0 q¼0 p!q! 2 2
 ðd
ð2n1Þ
 h2qþ1 ðkdÞ j2p ðkw0 Þj2qþ1 ðkw0 Þdw0
0
ða 
ð2n1Þ ð2Þ
þ j2qþ1 ðkdÞ j2p ðkw0 Þh2qþ1 ðkw0 Þdw0 h2p ðkrÞP2p ðcos hÞ; (A12)
d

so that, after applying the integral solution of Eq. (A8), together with the following Wronskian:20

 
ð2n1Þ ð2n1Þ
k2 d 2 j2q ðkdÞh2qþ1 ðkdÞ  j2qþ1 ðkdÞh2q ðkdÞ ¼ 7i; (A13)

we obtain

k2 qcU~ S X
P X
Q
ð2Þ
p~M ðr; hÞjr>a;0da ¼ 7i ð1Þpþq gðMÞ ðM2Þ
p;q ð6ij2p ðkdÞ þ j2qþ1 ðkdÞfp;q ðkaÞÞh2p ðkrÞP2p ðcos hÞ; (A14)
4p2 p¼0 q¼0

where
n o
ðM2Þ ð2n1Þ ð2n1Þ ð2n1Þ
fp;q ðkaÞ ¼ k2 a2 j2p ðkaÞh2q ðkaÞ  j2p1 ðkaÞh2qþ1 ðkaÞð2q  2p þ 1Þðj2p ðkaÞh2qþ1 ðkaÞÞ=ðkaÞ : (A15)

However, this is not the simplest form. Let us now use Eq. (A10) to recast Eq. (A14) into the following form:

1268 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
  X X Q 
kqcU~ S P
ð2Þ ð1Þpþq ð4q þ 3ÞCðp þ 1=2ÞCðq þ 3=2Þ
p~M ðr; hÞjr>a;0da ¼ i 2ik ð4p þ 1Þj2p ðkdÞh2p ðkrÞP2p ðcos hÞ 
8p p¼0 q¼0
pp!q!ðp þ q þ 1Þð2q  2p þ 1Þ

2k XP X Q
ð2Þ
7 ð1Þpþq gðMÞ h
p;q 2p ðkrÞj 2qþ1 ðkdÞf ðM2Þ
p;q ðkaÞP2p ðcos hÞ ; (A16)
p p¼0 q¼0

which is simplified by applying the following identity:

X
1
ð1Þpþq ð4q þ 3ÞCðp þ 1=2ÞCðq þ 3=2Þ
1¼ (A17)
q¼0
pp!q!ðp þ q þ 1Þð2q  2p þ 1Þ

to the term in braces, which can be proved by exchanging p and q in Eq. (A26) and letting h ¼ 0 so that
P2q(cos h) ¼ P2p þ 1(cos h) ¼ 1.13 Hence

 XP
kqcU~ S ð2Þ
p~M ðr; hÞjr>a;0da ¼i 2ik ð4p þ 1Þj2p ðkdÞh2p ðkrÞP2p ðcos hÞ
8p p¼0

P X
2k X
Q
pþq ðMÞ ð2Þ ðM2Þ
7 ð1Þ gp;q h2p ðkrÞj2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ : (A18)
p p¼0 q¼0

The single expansion in p only contains odd terms. Noting that Pp(x) ¼ (1)p Pp(x), we can write the following expansion
containing both odd and even terms:
 X1
kqcU~ S  
p~M ðr; hÞjr>a;0da ¼ i ik ð2p þ 1Þjp ðkdÞhð2Þ p ðkrÞ Pp ðcos hÞ þ Pp ðcos hÞ
8p p¼0
1 X
X 1 
2k ð2Þ
7 ð1Þpþq gðMÞ h
p;q 2p ðkrÞj2qþ1 ðkdÞf ðM2Þ
p;q ðkaÞP 2p ðcos hÞ : (A19)
p p¼0 q¼0

Using the identity of Eq. (13) and truncating the summation limits to P and Q, this simplifies further to
8 9
 
kqcU~ S < eikr0 eikr1 P X
2k X =
Q
pþq ð2Þ
p~M ðr; hÞjr>a;0da ¼i þ 7 ð1Þ  gðMÞ ðM2Þ
p;q h2p ðkrÞj2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ ;
8p : r0 r1 p p¼0 q¼0 ;

(A20)

where r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

2. Pressure field where the distance from the origin to the observation point is less than the radius of the mirror
or curtain
In order to achieve convergence in the region r  a, we simply exchange the spherical Bessel and Hankel functions in Eq.
(A7) as follows:
   
k3 qcU~ S X
1 X1
ð1Þpþq ð4p þ 1Þð4q þ 3Þ 1 3
p~M ðr; hÞj0ra;d>a ¼i C pþ C qþ
2p2 p¼0 q¼0 p!q! 2 2
 ðr ða 
ð2Þ ð2Þ ð2n1Þ
 h2p ðkrÞ j2p ðkw0 Þj2qþ1 ðkw0 Þdw0 þ j2p ðkrÞ h2p ðkw0 Þj2qþ1 ðkw0 Þdw0  h2qþ1 ðkdÞP2p ðcos hÞ;
0 r
(A21)

so that, after applying the integral solution of Eq. (A8), together with the following Wronskian:

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1269
  
ð2Þ ð2Þ
k2 r2 j2p ðkrÞh2p1 ðkrÞ  j2p1 ðkrÞh2p ðkrÞ ¼ i (A22)

we obtain

k2 qcU~ S X
1 X
1
ð2n1Þ
p~M ðr; hÞj0ra;d>a ¼ 7i ð1Þpþq gðMÞ ðM3Þ
p;q ðij2qþ1 ðkrÞ þ j2p ðkrÞfp;q ðkaÞÞh2qþ1 ðkdÞP2p ðcos hÞ; (A23)
4p2 p¼0 q¼0

where
n o
ðM3Þ ð2Þ ð2Þ ð2Þ
fp;q ðkaÞ ¼ k2 a2 h2p ðkaÞj2q ðkaÞ  h2p1 ðkaÞj2qþ1 ðkaÞð2q  2p þ 1Þðh2p ðkaÞj2qþ1 ðkaÞÞ=ðkaÞ : (A24)

However, this is not the simplest form. Let us now use Eq. (A10) to recast Eq. (A23) into the following form:

 X 1
kqcU~ S ð2n1Þ
p~M ðr; hÞj0ra;d>a ¼ i 62ik ð4q þ 3Þj2qþ1 ðkrÞh2qþ1 ðkdÞ
8p q¼0
X 1 pþq 
ð1Þ ð4p þ 1ÞCðp þ 1=2ÞCðq þ 3=2Þ
 P2p ðcos hÞ
p¼0
pp!q!ðp þ q þ 1Þð2q  2p þ 1Þ
1 X
2k X 1
ð2n1Þ
7 ð1Þpþq gðMÞ ðM3Þ
p;q j2p ðkrÞh2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞÞ; (A25)
p p¼0 q¼0

which is simplified by applying the following identity:13

X
1
ð1Þpþq ð4p þ 1ÞCðp þ 1=2ÞCðq þ 3=2Þ p
P2qþ1 ðcos hÞ ¼ P2p ðcos hÞ; 0h (A26)
p¼0
pp!q!ðp þ q þ 1Þð2q  2p þ 1Þ 2

to the term in braces. It is derived from

X
1
P2qþ1 ðcos hÞ ¼ Ap;q P2p ðcos hÞ; (A27)
p¼0

where
ð p=2
P2qþ1 ðcos hÞP2p ðcos hÞsin hdh
Ap;q ¼ ð0 p=2 : (A28)
P2p ðcos hÞP2p ðcos hÞsin hdh
0

Hence

 X1
kqcU~ S ð2n1Þ
p~M ðr; hÞj0ra;d>a ¼ i 62ik ð4q þ 3Þj2qþ1 ðkrÞh2qþ1 ðkdÞP2qþ1 ðcos hÞ
8p q¼0

1 X
2k X 1
ð2n1Þ
7 ð1Þpþq gðMÞ j
p;q 2p ðkrÞh 2qþ1 ðkdÞf ðM3Þ
p;q ðkaÞP2p ðcos hÞ : (A29)
p p¼0 q¼0

The single expansion in q only contains odd terms. Noting that Pq(x) ¼ (1)q Pq(x), we can write the following expan-
sion containing both odd and even terms:

1270 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
  X1
kqcU~ S  
p~M ðr; hÞj0ra;d>a ¼ i 7ik ð2q þ 1Þjq ðkrÞhð2n1Þ
q ðkdÞ Pq ðcos hÞ  Pq ðcos hÞ
8p q¼0
X1 X 1 
2k pþq ðMÞ ð2n1Þ ðM3Þ
7 ð1Þ gp;q j2p ðkrÞh2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ : (A30)
p p¼0 q¼0

Using the identity of Eq. (13) and truncating the summation limits to P and Q, this simplifies further to
8 9
 
kqcU~ S < e7ikr0 e7ikr1 P X
2k X =
Q
ð2n1Þ
p~M ðr; hÞj0ra;d>a ¼i  7 ð1Þpþq  gðMÞ
p;q j2p ðkrÞh2qþ1 ðkdÞf ðM3Þ
p;q ðkaÞP 2p ðcos hÞ ;
8p : r0 r1 p p¼0 q¼0 ;

(A31)

where r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

The solution for d  a is a little more complicated because the spherical Bessel and Hankel functions have to be
exchanged in part of the integral in order to achieve convergence. First, for r  d, we have
   
k3 qcU~ S X
1 X1
ð1Þpþq ð4p þ 1Þð4q þ 3Þ 1 3
p~M ðr; hÞj0rd;0da ¼i C pþ C qþ
2p2 p¼0 q¼0 p!q! 2 2
 ðr ðd
ð2Þ ð2n1Þ ð2n1Þ ð2Þ
 h2p ðkrÞh2qþ1 ðkdÞ j2p ðkw0 Þj2qþ1 ðkw0 Þdw0 þ j2p ðkrÞh2qþ1 ðkdÞ h2p ðkw0 Þj2qþ1 ðkw0 Þdw0
0 r
ða 
ð2Þ ð2n1Þ
þ j2p ðkrÞj2qþ1 ðkdÞ h2p ðkw0 Þh2qþ1 ðkw0 Þdw0  P2p ðcos hÞ; (A32)
d

so that, after applying the integral solution of Eq. (A8), together with the Wronskians of Eqs. (A13) and (A22), and truncating
the summation limits to P and Q, the final expression for the pressure field becomes

k2 qcU~ S X
P X Q
ð2Þ
p~M ðr; hÞj0rd;0da ¼ 7i ð1Þpþq gðMÞ
p;q ð6ij2p ðkrÞh2p ðkdÞ
4p2 p¼0 q¼0
ð2n1Þ ðM4Þ
ij2qþ1 ðkrÞh2qþ1 ðkdÞ þ j2p ðkrÞj2qþ1 ðkdÞfp;q ÞP2p ðcos hÞ; (A33)

where
n o
ðM4Þ ð2Þ ð2n1Þ ð2Þ ð2n1Þ ð2Þ ð2n1Þ
fp;q ðkaÞ ¼ k2 a2 h2p ðkaÞh2q ðkaÞ  h2p1 ðkaÞh2qþ1 ðkaÞð2q  2p þ 1Þðh2p ðkaÞh2qþ1 ðkaÞÞ=ðkaÞ : (A34)

Second, for r > d, we have

   
k3 qcU~ S X
1 X1
ð1Þpþq ð4p þ 1Þð4q þ 3Þ 1 3
p~M ðr; hÞjd<ra;0da ¼i C pþ C qþ
2p2 p¼0 q¼0 p!q! 2 2
 ðd ðr
ð2Þ ð2n1Þ ð2Þ ð2n1Þ
 h2p ðkrÞh2qþ1 ðkdÞ j2p ðkw0 Þj2qþ1 ðkw0 Þdw0 þ h2p ðkrÞj2qþ1 ðkdÞ j2p ðkw0 Þh2qþ1 ðkw0 Þdw0
0 d
ða 
ð2Þ ð2n1Þ
þ j2p ðkrÞj2qþ1 ðkdÞ h2p ðkw0 Þh2qþ1 ðkw0 Þdw0  P2p ðcos hÞ; (A35)
r

so that, after applying the integral solution of Eq. (A8), together with the Wronskians of Eqs. (A13) and (A22), and truncating
the summation limits to P and Q, the final expression for the pressure field becomes

k2 qcU~ S X
P X
Q
ð2Þ
p~M ðr; hÞjdra;0da ¼ 7i ð1Þpþq gðMÞ
p;q ð6ij2p ðkdÞh2p ðkrÞ
4p2 p¼0 q¼0
ð2n1Þ ðM4Þ
ij2qþ1 ðkdÞh2qþ1 ðkrÞ þ j2p ðkrÞj2qþ1 ðkdÞfp;q ðkaÞÞP2p ðcos hÞ; (A36)

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1271
 ðM4Þ
where fp;q ðkaÞ is given by Eq. (A34). Again, using similar arguments to those above, this can be simplified to

kqcU~ S e7ikr0 e7ikr1 eikr0 eikr1 2k X P X Q
p M ðr; hÞj0ra;0da ¼ i
~  þ þ 7 ð1Þpþq gðMÞ
p;q j2p ðkrÞ
8p r0 r1 r0 r1 p p¼0 q¼0

ðM4Þ
 j2qþ1 ðkdÞfp;q ðkaÞP2p ðcos hÞ : (A37)

2. Finite plane circular acoustic curtain or time-reversal mirror using dipole (pressure) source array
1. Pressure field where the distance from the origin to the observation point is greater than the radius of the
mirror or curtain
It was shown in a previous paper10 that, in the case of a uniform driving pressure p~0 in the region 0  w0  a, the follow-
ing solution may be given to Eq. (32):
ða
p X
ik~ 1
Cðp þ 3=2Þ ð2Þ
p~D ðr; hÞ ¼ pffiffiffi0 ð1Þp ð4p þ 3Þ  j2pþ1 ðkw0 Þdw0 h2pþ1 ðkrÞP2pþ1 ðcos hÞ; (A38)
p p¼0 p! 0

where p~þ ðw0 Þ ¼ ~

p  ðw0 Þ ¼ p~0 =2. We now replace p~0 with the non-uniform pressure distribution p~S ðw0 ; 0Þ so that
ða
ik X 1
Cðp þ 3=2Þ ð2Þ
p~D ðr; hÞjr>a;d>a ¼ pffiffiffi ð1Þp ð4p þ 3Þ  p~S ðw0 Þjd>w0 j2pþ1 ðkw0 Þdw0  h2pþ1 ðkrÞP2pþ1 ðcos hÞ; (A39)
p p¼0 p! 0

which after inserting Eq. (30) can be written

k3 qcU~ S X
1 X1
ð1Þpþq ð4p þ 3Þð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ
p~D ðr; hÞjr>a;d>a ¼ i
4p2 p¼0 q¼0 p!q!
ða
ð2n1Þ ð2Þ
 j2q ðkw0 Þj2pþ1 ðkw0 Þdw0 h2q ðkdÞh2pþ1 ðkrÞP2pþ1 ðcos hÞ; (A40)
0

so that, after applying the integral solution of Eq. (A8) and truncating the summation limits to P and Q, the final expression
for the pressure field becomes

k2 qcU~ S X
P X
Q
ð2Þ ð2n1Þ
p~D ðr; hÞjr>a;d>a ¼ i ð1Þpþq gðDÞ ðD1Þ
p;q h2pþ1 ðkrÞ  h2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ; (A41)
4p2 p¼0 q¼0

where

ð4p þ 3Þð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ

gðDÞ
p;q ¼ (A42)
p!q!ðp þ q þ 1Þð2p  2q þ 1Þ

and
ðD1Þ  

fp;q ðkaÞ ¼ k2 a2 j2p ðkaÞj2q ðkaÞ  j2pþ1 ðkaÞj2q1 ðkaÞð2p  2q þ 1Þ j2pþ1 ðkaÞj2q ðkaÞ =ðkaÞ : (A43)

The solution for d  a is a little more complicated because the spherical Bessel and Hankel functions have to be exchanged in
part of the integral in order to achieve convergence

k3 qcU~ S X
1 X1
ð1Þpþq ð4p þ 3Þð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ
p~D ðr; hÞjr>a;0da ¼ i 2
2p p¼0 q¼0 p!q!
 ðd ða 
ð2n1Þ ð2n1Þ ð2Þ
 h2q ðkdÞ j2q ðkw0 Þj2pþ1 ðkw0 Þdw0 þj2q ðkdÞ h2q ðkw0 Þj2pþ1 ðkw0 Þdw0 h2pþ1 ðkrÞP2pþ1 ðcos hÞ;
0 d
(A44)

1272 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
so that, after applying the integral solution of Eq. (A8), together with the Wronskian of Eq. (A13), we obtain

k2 qcU~ S X
1 X
1  
pþq ðDÞ ðD2Þ ð2Þ
p~D ðr; hÞjr>a;0da ¼ i ð1Þ g p;q 7ij2pþ1 ðkdÞþj 2q ðkdÞf p;q ðkaÞ h2pþ1 ðkrÞP2pþ1 ðcos hÞ; (A45)
4p2 p¼0 q¼0

where
( )
 
ðD2Þ ð2n1Þ ð2n1Þ ð2n1Þ
fp;q ðkaÞ ¼ k2 a2 j2p ðkaÞh2q ðkaÞ  j2pþ1 ðkaÞh2q1 ðkaÞð2p  2q þ 1Þ j2pþ1 ðkaÞh2q ðkaÞ ðkaÞ : (A46)

However, this is not the simplest form. Let us now use Eq. (A42) to recast Eq. (A45) into the following form:
 X1
kqcU~ S ð2Þ
p~D ðr; hÞjr>a;0da ¼ i 72ik ð4p þ 3Þj2pþ1 ðkdÞh2pþ1 ðkrÞP2pþ1 ðcos hÞ
8p p¼0
(1 )
X ð1Þpþq ð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ

q¼0
pp!q!ðp þ q þ 1Þð2p  2q þ 1Þ

1 X
2k X 1
pþq ðDÞ ð2Þ ðD2Þ
þ ð1Þ gp;q h2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ ; (A47)
p p¼0 q¼0

which is simplified by applying the following identity:

X
1
ð1Þpþq ð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ
1¼ (A48)
q¼0
pp!q!ðp þ q þ 1Þð2p  2q þ 1Þ

to the term in braces, which can be proved by exchanging p and q in Eq. (A26) and letting h ¼ 0 so that
P2qþ1(cos h) ¼ P2p(cos h) ¼ 1.13 Hence
 X1
kqcU~ S ð2Þ
p~D ðr; hÞjr>a;0da ¼ i 72ik ð4p þ 3Þj2pþ1 ðkdÞh2pþ1 ðkrÞP2pþ1 ðcos hÞ
8p p¼0

1 X
2k X 1
pþq ðDÞ ð2Þ ðD2Þ
þ ð1Þ gp;q h2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ : (A49)
p p¼0 q¼0

The single expansion in p only contains odd terms. Noting that Pp(x) ¼ (1)p Pp(x), we can write the following expansion
containing both odd and even terms:
 X1
kqcU~ S
p~D ðr; hÞjr>a;0da ¼i 6ik ð2p þ 1Þjp ðkdÞhð2Þ
p ðkrÞðPp ðcos hÞ þ Pp ðcos hÞÞ
8p p¼0

2k X1 X 1
ð2Þ
 ð1Þpþq gðDÞ h
p;q 2pþ1 ðkrÞj 2q ðkdÞf ðD2Þ
p;q ðkaÞP 2pþ1 ðcos hÞ : (A50)
p p¼0 q¼0

Using the identity of Eq. (13) and truncating the summation limits to P and Q, this simplifies further to
  ikr 
2k X P XQ
kqcU~ S e 0 eikr1
p~D ðr; hÞjr>a;0da ¼ i 6sgnðcos hÞ   ð1Þpþq
8p r0 r1 p p¼0 q¼0

ð2Þ
gðDÞ h
p;q 2pþ1 ðkrÞj2q ðkdÞf ðD2Þ
p;q ðkaÞP2pþ1 ðcos hÞ ; (A51)

where r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1273
2. Pressure field where the distance from the origin to the observation point is less than the radius of the mirror
or curtain
In order to achieve convergence in the region r  a, we simply exchange the spherical Bessel and Hankel functions in
Eq. (A40) as follows:

k3 qcU~ S X
1 X 1
ð1Þpþq ð4p þ 3Þð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ
p~D ðr; hÞj0ra;d>a ¼ i
2p2 p¼0 q¼0 p!q!
 ðr ða 
ð2Þ ð2Þ ð2n1Þ
 h2pþ1 ðkrÞ j2pþ1 ðkw0 Þj2q ðkw0 Þdw0 þj2pþ1 ðkrÞ h2pþ1 ðkw0 Þj2q ðkw0 Þdw0 h2q ðkdÞP2pþ1 ðcos hÞ;
0 r
(A52)

so that, after applying the integral solution of Eq. (A8), together with the Wronskian of Eq. (A22), we obtain

k2 qcU~ S X
1 X
1  
pþq ðDÞ ðD3Þ ð2n1Þ
p~D ðr; hÞj0ra;d>a ¼ i ð1Þ g p;q ij 2q ðkrÞþj2pþ1 ðkrÞf p;q ðkaÞ h2q ðkdÞP2pþ1 ðcos hÞ; (A53)
4p2 p¼0 q¼0

where
( )
 
ðD3Þ 2 2 ð2Þ ð2Þ ð2Þ
fp;q ðkaÞ ¼k a h2p ðkaÞj2q ðkaÞ  h2pþ1 ðkaÞj2q1 ðkaÞð2p  2q þ 1Þ h2pþ1 ðkaÞj2q ðkaÞ ðkaÞ : (A54)

However, this is not the simplest form. Let us now use Eq. (A42) to recast Eq. (A53) into the following form:

kqcU~ S X1
ð2n1Þ
p~D ðr; hÞj0ra;d>a ¼ i 2ik ð4q þ 1Þj2q ðkrÞh2q ðkdÞ
8p q¼0
(1 )
X ð1Þ ð4p þ 3ÞCðp þ 3=2ÞCðq þ 1=2Þ
pþq
 P2pþ1 ðcos hÞ
p¼0
pp!q!ðp þ q þ 1Þð2p  2q þ 1Þ
!
1 X
2k X 1
pþq ðDÞ ð2n1Þ ðD3Þ
 ð1Þ gp;q j2pþ1 ðkrÞh2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ ; (A55)
p p¼0 q¼0

which is simplified by applying the following identity:13

X
1
ð1Þpþq ð4p þ 3ÞCðp þ 3=2ÞCðq þ 1=2Þ
sgnðcoshÞP2q ðcos hÞ ¼ P2pþ1 ðcos hÞ (A56)
p¼0
pp!q!ðp þ q þ 1Þð2p  2q þ 1Þ

to the term in braces. It is derived from

X
1
sgnðcos hÞP2q ðcos hÞ ¼ Ap;q P2pþ1 ðcos hÞ; (A57)
p¼0

where

ð p=2
P2q ðcos hÞP2pþ1 ðcos hÞsin hdh
0
Ap;q ¼ ð p=2 : (A58)
P2pþ1 ðcos hÞP2pþ1 ðcos hÞsin hdh
0

Hence

1274 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
  X1
kqcU~ S ð2n1Þ
p~D ðr; hÞj0ra;d>a ¼ i 2ik ð4q þ 1Þj2q ðkrÞh2q ðkdÞP2q ðcos hÞ
8p q¼0

1 X
2k X 1
pþq ðDÞ ð2n1Þ ðD3Þ
 ð1Þ gp;q j2pþ1 ðkrÞh2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ : (A59)
p p¼0 q¼0

The single expansion in q only contains odd terms. Noting that Pq(x) ¼ (1)q Pq(x), we can write the following expansion
containing both odd and even terms:

 X1
kqcU~ S  
p~D ðr; hÞj0ra;d>a ¼ i ik ð2q þ 1Þjq ðkrÞhqð2n1Þ ðkdÞ Pq ðcos hÞ  Pq ðcos hÞ
8p q¼0
X1 X 1 
2k pþq ðDÞ ð2n1Þ ðD3Þ
 ð1Þ gp;q j2pþ1 ðkrÞh2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ : (A60)
p p¼0 q¼0

Using the identity of Eq. (13) and truncating the summation limits to P and Q, this simplifies further to

  7ikr 
2k X P XQ
kqcU~ S e 0 e7ikr1
p~D ðr; hÞj0ra;d>a ¼ i 6sgnðcos hÞ þ  ð1Þpþq
8p r0 r1 p p¼0 q¼0

ð2n1Þ
 gðDÞ j
p;q 2pþ1 ðkrÞh2q ðkdÞf ðD3Þ
p;q ðkaÞP 2pþ1 ðcos hÞ ; (A61)

where r02 ¼ r2 þ d 2 þ 2rdjcos hj and r12 ¼ r 2 þ d2  2rdjcos hj.

The solution for d  a is a little more complicated because the spherical Bessel and Hankel functions have to be
exchanged in part of the integral in order to achieve convergence. First, for r  d, we have

k3 qcU~ S X
1 X
1
ð1Þpþq ð4q þ 1Þð4p þ 3ÞCðp þ 3=2ÞCðq þ 1=2Þ
p~D ðr; hÞj0rd;0da ¼ 1
4p2 p¼0 q¼0 p!q!
 ðr ðd
ð2Þ ð2n1Þ ð2n1Þ ð2Þ
 h2pþ1 ðkrÞh2q ðkdÞ j2q ðkw0 Þj2pþ1 ðkw0 Þdw0 þ j2pþ1 ðkrÞh2q ðkdÞ h2pþ1 ðkw0 Þj2q ðkw0 Þdw0
0 r
ða 
ð2Þ ð2n1Þ
þ j2pþ1 ðkrÞj2q ðkdÞ h2pþ1 ðkw0 Þh2q ðkw0 Þdw0 P2pþ1 ðcos hÞ; (A62)
d

so that, after applying the integral solution of Eq. (A8), together with the Wronskians of Eqs. (A13) and (A22), and truncating
the summation limits to P and Q, the final expression for the pressure field becomes

k2 qcU~ S X
P X Q
p~D ðr; hÞj0rd;0da ¼ i ð1Þpþq gðDÞ
p;q
8p2 p¼0 q¼0
ð2n1Þ ð2Þ ðD4Þ
 ðij2q ðkrÞh2q ðkdÞ7ij2pþ1 ðkrÞh2pþ1 ðkdÞþj2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞÞP2pþ1 ðcos hÞ; (A63)

where
( )
 
ðD4Þ ð2Þ ð2n1Þ ð2Þ ð2n1Þ ð2Þ ð2n1Þ
fp;q ðkaÞ ¼ k2 a2 h2p ðkaÞh2q ðkaÞ  h2pþ1 ðkaÞh2q1 ðkaÞð2p  2q þ 1Þ h2pþ1 ðkaÞh2q ðkaÞ ðkaÞ : (A64)

Second, for r > d, we have

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1275
 k3 qcU~ S X
1 X 1
ð1Þpþq ð4p þ 3Þð4q þ 1ÞCðp þ 3=2ÞCðq þ 1=2Þ
p~D ðr; hÞjd<ra;0da ¼ i
4p3 p¼0 q¼0 p!q!
 ðd ðr
ð2Þ ð2n1Þ ð2Þ ð2n1Þ
 h2pþ1 ðkrÞh2q ðkdÞ j2pþ1 ðkw0 Þj2q ðkw0 Þdw0 þ h2pþ1 ðkrÞj2q ðkdÞ j2pþ1 ðkw0 Þh2q ðkw0 Þdw0
0 d
ða 
ð2Þ ð2n1Þ
þ j2pþ1 ðkrÞj2q ðkdÞ h2pþ1 ðkw0 Þh2q ðkw0 Þdw0 P2pþ1 ðcos hÞ; (A65)
r

so that, after applying the integral solution of Eq. (A8), together with the Wronskians of Eqs. (A13) and (A22), and truncating
the summation limits to P and Q, the final expression for the pressure field becomes

k2 qcU~ S X
P X Q
p~D ðr; hÞjdra;0da ¼ i ð1Þpþq gðDÞ
p;q
8p2 p¼0 q¼0
 
ð2n1Þ ð2Þ ðD4Þ
 ij2q ðkdÞh2q ðkrÞ7ij2pþ1 ðkdÞh2pþ1 ðkrÞ þ j2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞ P2pþ1 ðcos hÞ; (A66)

ðD4Þ
where fp;q ðkaÞ is given by Eq. (A64). Again, using similar arguments to those above, this can be simplified to
  7ikr 
kqcU~ S e 0 e7ikr1 eikr0 eikr1
p~D ðr; hÞj0ra;0da ¼i 6sgnðcos hÞ þ þ 
8p r0 r1 r0 r1
XP X Q 
2k pþq ðD4Þ
 ð1Þ gp;q j2pþ1 ðkrÞj2q ðkdÞfp;q ðkaÞP2pþ1 ðcos hÞ : (A67)
p p¼0 q¼0

APPENDIX B: ALTERNATE FORMULAS FOR CALCULATING FIELDS USING NUMERICAL INTEGRATION

1. Finite plane circular acoustic curtain or time-reversal mirror using monopole (pressure-gradient) source array
This appendix presents the numerical integrals, in cylindrical coordinates, which the authors used to validate the analyti-
cal solutions in the form of expansions
ð 2p ð a
@
p~M ðw; zÞ ¼ 2 p~ðw0 ; z0 Þjz0 ¼0 gðw; /; zjw0 ; /0 ; z0 Þju¼z0 ¼0 wdwd/0 ; (B1)
0 0 @z0

where
! 7ikpffiffiffiffiffiffiffiffiffiffi
w20 þd 2
@ 2 ~ 1 e
p~ðw0 ; z0 Þjz0 ¼0 ¼ 6k dqcU S 16 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ; (B2)
@z0 ik w20 þ d2 4pðw20 þ d2 Þ

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2
2 2
eik w þw0 2ww0 cosðuu0 Þþðzz0 Þ
gðw; /; zjw0 ; /0 ; z0 Þ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (B3)
4p w2 þ w20  2ww0 cosð/  /0 Þ þ ðz  z0 Þ2

2. Finite plane circular acoustic curtain or time-reversal mirror using dipole (pressure) source array

ð 2p ð a
@
p~D ðw; zÞ ¼ 2 p~ðw0 ; z0 Þjz0 ¼0 gðw; /; zjw0 ; /0 ; z0 Þj/¼z0 ¼0 wdwd/0 ; (B4)
0 0 @z 0

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2ffi
7ik 2
w0 þðdz0 Þ
e
p~ðw0 ; z0 Þ ¼ 7ikqcU~ S qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (B5)
4p w20 þ ðd  z0 Þ2

1276 J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 €rkka
T. Mellow and L. Ka €inen: Equations for virtual acoustic sources
 ! pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 2
@ 1 eik w þw0 2ww0 cos/0 þz
gðw; /; zjw0 ; /0 ; z0 Þj/¼z0 ¼0 ¼ ikz 1 þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi  :
@z0 2
ik w þ w0  2ww0 cos /0 þ z2
2 4pðw2 þ w20  2ww0 cos /0 þ z2 Þ
(B6)

APPENDIX C: PASSIVE ACOUSTIC SINK would only be substantially resistive for smaller wavelengths
and the resistance would be less than one-half of the flow re-
The solution to the following homogeneous Helmholtz
sistance of the material.
spherical wave equation
  1
@2 2 @ http://www.holophony.net/. This website contains some illuminating ani-
2
þ  þ k p~ðrÞ ¼ 0 (C1) mations. (Last viewed on January 28, 2014).
@r2 r @r 2
A. J. Berkhout, “A holographic approach to acoustic control,” J. Audio
Eng. Soc. 36(12), 977–995 (1988).
3
is given by A. J. Berkhout, D. de Vries, and P. Vogel, “Acoustic control by wave field
synthesis,” J. Acoust. Soc. Am. 93(5), 2764–2778 (1993).
4
eikr eikr E. G. Williams, J. D. Maynard, and E. Skudrzyk, “Sound source reconstruc-
p~ðrÞ ¼ p~þ þ p~ ; (C2) tions using a microphone array,” J. Acoust. Soc. Am. 68(1), 340–344 (1980).
r r 5
E. G. Williams, Fourier Acoustics (Academic Press, London, 1999), pp.
33–34.
6
where p~þ is the strength of the diverging wave and p~ is that J. de Rosny and M. Fink, “Overcoming the diffraction limit in wave
of the converging one. The velocity in the direction toward physics using a time-reversal mirror and a novel acoustic sink,” Phys.
Rev. Lett. 89, 124301 (2002).
the origin is given by 7
B. E. Anderson, M. Griffa, C. Larmer, A. J. Ultrich, and P. A. Johnson
“Timing reversal,” Acoustics Today 4(1), 5–15 (2008).
1 8
S. G. Conti, P. Roux, and W. A. Kuperman, “Near-field time–reversal
u~ðrÞ ¼ p~ðrÞ
ikqc amplification,” J. Acoust. Soc. Am. 121(6), 3602–3606 (2007).
  ikr   ikr 9
J. Rosny and M. Fink, “Focusing properties of near-field time reversal,”
p~þ i e p~ i e Phys. Rev. A 76(6), 065801 (2007).
¼ 1 þ þ1 : (C3) 10
qc kr r qc kr r T. J. Mellow, “On the sound field of a resilient disk in free space,”
J. Acoust. Soc. Am. 123(4), 1880–1891 (2008).
11
P. M. Morse and K. U. Ingard, Theoretical Acoustics (McGraw-Hill, New
If we have a sphere of radius R with an anechoic surface im- York, 1968), pp. 319–321.
pedance of qc, we can write the boundary condition 12
L. L. Beranek and T. J. Mellow, Acoustics: Sound Fields and Transducers
(Academic Press, Oxford, 2012), pp. 536–541.
p~ðRÞ 13
I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
¼ qc (C4) Products, 6th ed., edited by A. Jeffrey (Academic, New York, 2000), p.
u~ðRÞ
930, Eq. (8.533); p. 963, Eq. (8.794.1); p. 959, Eq. (8.756.1); p. 887, Eq.
(8.334.2); p. 782, Eq. (7.112.1); p. 968, Eq. (8.828.1).
so that 14
J. F. Kelly and R. J. McGough, “An annular superposition integral for axi-
symmetric radiators,” J. Acoust. Soc. Am. 121(2), 759–765 (2007).
e2ikR
15
H. Backhaus and F. Trendelenburg, “Uber € die Richtwirkung von
p~þ  p~ : (C5) Kolbenmembranen (On the directivity of piston membranes),” Ztschr. f.
1 þ 2ikR  techn. Phys. 7, 630–635 (1927).
16
P. J. Walker, “New developments in electrostatic loudspeakers,” J. Audio
Hence, when kR is very large or the wavelength is very small Eng. Soc. 28(11), 795–799 (1980).
17
in comparison to the radius of the sphere, p~þ (which is the T. Sugimoto and Y. Nakajima, “Acoustic characteristics of a flexible
sound generator based on thermoacoustic effect,” J. Acoust. Soc. Am.
strength of the reflected wave) becomes vanishingly small
133, 3231 (2013).
and we are left with the converging wave without any reflec- 18
T. Sugimoto, A. Ando, K. Ono, Y. Morita, K. Hosoda, D. Ishii, and K.
tions. However, this only holds for small wavelengths. Nakamura, “A lightweight push-pull acoustic transducer composed of a
Another problem for large wavelengths is that if such a pair of dielectric elastomer films,” J. Acoust. Soc. Am. 134(5),
EL432–EL437 (2013).
sphere were fabricated from a piece of porous absorbent ma- 19
Symbolic computation by MathematicaV.
R

terial, the air within it would act as compliance and the resis- 20
M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions
tive part of the impedance would be small. In other words, it (Dover, New York, 1964), p. 390, Eq. (9.1.16).

J. Acoust. Soc. Am., Vol. 135, No. 3, March 2014 € rkka

T. Mellow and L. Ka €inen: Equations for virtual acoustic sources 1277