1
Nonlinear and dissipative Nearfield Acoustical Holography
algorithms based on Westervelt Wave Equation
Yaying Niua)
Yong-Joe Kimb)
Acoustics and Signal Processing Laboratory, Department of Mechanical Engineering, Texas A&M
University, 3123 TAMU, College Station, TX 77843-3123, USA
When a conventional, linear, lossless Nearfield Acoustical Holography (NAH) procedure is
applied to reconstruct the three-dimensional (3-D) sound fields that are radiated from a
high-level noise source and include significant nonlinear components, it can result in
significant reconstruction errors. Here, a nonlinear, dissipative, planar NAH procedure is
introduced that can be used to identify nonlinear noise characteristics in the 3-D nearfield
of the high-level noise source from two-dimensional sound pressure data measured on a
hologram surface. The proposed NAH procedure is derived by applying perturbation and
renormalization methods to the nonlinear, dissipative Westervelt Wave Equation. In order
to validate the proposed procedure, the nonlinear and dissipative sound pressure field
radiated from a high-level pulsating sphere at a single frequency is calculated from the
spherical Burgers Equation. An improved SONAH procedure is applied to reconstruct the
source sound pressure field that is input to the proposed nonlinear projection
procedure. Within 2.5 m nearfield reconstruction distance from the pulsating sphere, the
nonlinear sound pressure field reconstructed by applying the proposed NAH procedure
matches well with the directly-calculated field at the maximum reconstruction error of 0.5
dB.
1 INTRODUCTION
The sound fields radiated from high-level noise sources include significant nonlinear
components that can lead to shock wave propagations. Thus, it is critical to identify noise
propagation characteristics in the nearfield of the sources in order to design optimal noise control
schemes to suppress the nonlinear noise components before they turn into shock waves.
However, the conventional, linear, lossless NAH procedures1-9
cannot be used to visualize the
highly nonlinear and dissipative sound fields correctly. Therefore, it is here proposed to develop
a novel, nonlinear, dissipative, NAH procedure in an open 3-D space to consider the nonlinear
and thermoviscous effects. Combined with a linear NAH procedure, the proposed NAH
procedure can be used to nonlinearly project the sound pressure data measured on a 2-D planar
a)
2
measurement surface close to the noise sources into a 3-D space to visualize both linear and
nonlinear acoustic fields.
Hamilton and Blackstock10
summarized nonlinear acoustics research work done by many
researchers11-17
. In particular, perturbation and renormalization methods are used by Ginsberg13-
16 and Nayfeh
17 to investigate nonlinear acoustic wave propagations in planar, cylindrical, and
spherical cases. According to Ginsberg’s investigation on 2-D planar and 3-D cylindrical
nonlinear wave propagations based on renormalization procedures13-16
, the particle velocity
along a wave propagating direction can advance or retard wavefronts, while the transverse
particle velocity can bend the rays of the nonlinear waves. For a spherical wave, however, the
transverse particle velocity decays in the factor of 1/r compared to the radial velocity17
. Thus,
the bending of the rays is negligible and the rays remain straight in this case.
Among the nonlinear and thermoviscous acoustic equations for perfect gases10
, the
Westervelt Wave Equation10,11
(WWE) is selected to derive the proposed nonlinear and
dissipative NAH procedure. The WWE is solved by using perturbation and modified
renormalization methods to give linear and nonlinear NAH reconstructed sound pressure fields,
which include the nonlinearity-induced steepening effects of wavefronts but do not include the
transverse particle velocity-induced bending of rays. Thus, the proposed nonlinear NAH
procedure can be applied to nonlinear wave propagation problems only with straight rays: e.g.,
spherical and 1-D planar wave cases. In addition, local nonlinearity caused by source surface
displacements is not included in the WWE. That is, the WWE is appropriate for sound waves
only with dominant cumulative nonlinearity away from a noise source surface.
The proposed nonlinear and dissipative NAH procedure is valid for monofrequency source
cases. Thus, the first-order, linear sound pressure at a fundamental frequency is used to calculate
the second-order, nonlinear sound pressure at twice of the fundamental frequency. In multi-
frequency source cases, however, the Fenlon’s solution10,12
indicates that each frequency
component consists of the harmonics, summations, or differences of other frequency components.
Thus, the fundamental frequency component can include nonlinear components and cannot be
directly applied for the calculation of its second-order nonlinear component.
A nonlinear and dissipative pulsating sphere simulation is performed to validate the
proposed NAH procedure applicable to spherical sound fields. The generalized Burgers
Equation in the spherical coordinates10
is solved by using the aforementioned perturbation and
renormalization procedures to obtain the nonlinear and dissipative sound pressure field radiated
from the high-level pulsating sphere. The calculated hologram data is backward projected to the
source surface, and then the reconstructed source data is input to the proposed, nonlinear,
forward, NAH projection procedure. The sound pressure field reconstructed by using the
proposed NAH procedure is compared with the directly-calculated sound pressure field.
2 THEORY
2.1 Perturbation Procedure for Westervelt Wave Equation
The nonlinear and dissipative Westervelt Wave Equation10,12
(WWE) is represented as
2 3 2 22
2 2 4 3 4 2
0 0 0 0
1 p b p pp
c t c t c t
, (1)
where p is the acoustic pressure, 0 is the ambient fluid medium density, c0 is the speed of sound,
β = (γ + 1)/2 is the nonlinearity coefficient for an ideal gas, and b is the sound diffusivity that
3
includes viscous and thermal conduction effects10
. The two terms on the right-hand-side (RHS)
of Eq. (1) thus represent thermoviscous and cumulative nonlinear effects, respectively.
By using the perturbation method13-17
, the acoustic pressure, p in the Cartesian coordinates,
(x, y, z) can be expanded as
2
1 2, , , , , , , , ,p x y z t O O p x y z t p x y z t , (2)
where ε is the small perturbation parameter (e.g., acoustic Mach number, ε = u0/c0 where u0 is the
particle velocity magnitude) and pn is the n-th order acoustic pressure (n = 1,2,3,). The sound
diffusivity, b is a small number in the order of ε in air10
. By substituting p in Eq. (2) into Eq. (1)
and neglecting the O(ε3) and higher-order small terms, the first- and second-order equations can
be associated with the O(ε) and O(ε2), respectively: i.e.,
22 1
1 2 2
0
10
pp
c t
, (3a)
2 3 2 22 2 1 1
2 2 2 4 3 4 2
0 0 0 0
1 p p pbp
c t c t c t
. (3b)
Eq. (3a) is a linear, lossless, homogeneous wave equation and Eq. (3b) is a nonlinear, dissipative,
inhomogeneous wave equation. The left-hand-side (LHS) of Eq. (3b) is a linear wave equation
with the unknown of the second-order pressure, p2, while the RHS consists of the dissipative and
nonlinear source terms that are composed of the first-order acoustic pressure, p1. Thus, once the
first-order sound pressure solution is obtained from Eq. (3a), the second-order equation becomes
a linear, inhomogeneous partial differential equation of which solution can be obtained
analytically for certain cases.
2.2 Nonlinear, Dissipative, Planar NAH Projection
In practical measurements of the sound pressure field generated by a monofrequency sound
source, the measured sound pressure data at the fundamental frequency contains the only first-
order, linear sound pressure components. The linear NAH algorithms1-9
can be thus applied to
the backward projection from the hologram surface to the sound source surface at the
fundamental frequency. Here, an improved Statistically Optimal Nearfield Acoustical
Holography (SONAH) algorithm9 is applied to backward project the hologram sound pressure
field at z = zh to the source surface at z = z0. As described in Ref. 9, this improved SONAH
procedure allows accurately reconstructing a source sound pressure field even with the hologram
data measured by using a microphone array with a small measurement aperture. The linearly
reconstructed sound pressure field on the source plane is then input to the nonlinear and
dissipative NAH forward projection procedure.
In the Cartesian coordinates, (x,y,z), the acoustic pressure solution of Eq. (3a) at the
fundamental frequency of ω can be obtained from the conventional, linear, planar NAH
projection1-3
in a discretized form as
0
11
1 1 0
0 0
1, , , Re , , ,
yx
y zmnx
NNin k y ik z zim k x i t
xm yn
m nx y
p x y z t P k k z e e e eN N
, (4)
where z0 denotes the sound source surface location, z is a reconstruction surface location (z ≥ z0),
Nx and Ny are the number of measurement points along the x- and y-directions, respectively, and
4
(kxm,kyn) = (mkx,nky) are the discrete wave numbers. In Eq. (4), the wave number spectrum,
P1(kxm,kyn,z0,ω) = P1mn is obtained by applying the spatial Fast Fourier Transform (FFT) to the
source sound pressure data, p1(x,y,z0,ω) that is obtained by applying the linear, backward
SONAH procedure to the measured sound pressure data on the hologram plane at z = zh. The
projection relation between the reconstruction and source surfaces is
0 0
1 1 0 1, , , , , , zmn zmnik z z ik z z
mnP m n z P m n z e P e
. (5)
In Eq. (5), the z-direction wave number, kzmn is the function of ω, kxm, and kyn: i.e.,
2 2 2 2( ) ( )zmn x yk k m k n k , (6)
where k = ω/c0 is the acoustic wave number. By substituting Eq. (4) into Eq. (3b) and writing
the second-order acoustic pressure solution as p2 = Re(p21e-it
+ p22e-i2t
), Eq. (3b) can be
decomposed into two inhomogeneous Helmholtz equations: i.e.,
0
1132 2
21 140 00
, ,yx
y zmnx
NNin k y ik z zim k x
mn
m nx y
ibk p x y z e e P e
N N c
, (7)
( )( ) 0
2 2
22
1 11 12
1 1( )( )2 2 40 0 0 00 0
4 , ,
2
y yx x
zmn z l m q nyx
N NN Ni k k z ziq k yil k x
mn l m q n
l q m nx y
k p x y z
e e P P eN N c
. (8)
In Eqs. (7) and (8), p21 is associated with the “dissipative” sound pressure component at the
frequency of ω, while p22 corresponds to the cumulative “nonlinear” sound pressure component
at the frequency of 2ω. In Eq. (8), the high wave number components (i.e., l ≥ Nx and q ≥ Ny) are
set to zero since they are the subsonic components decaying out exponentially during the forward
projections along the z-direction. Similarly, the P1 and kz terms with the subscripts of (l-m)(q-n)
have only non-zero values when the subscripts, (l-m) and (q-n) are within [0, Nx-1] and [0, Ny-1],
respectively.
The solution of p21(x, y, z) in Eq. (7) can be written as
11
21 21
0 0
, ,yx
yx
NNin k yim k x
mn
m n
p x y z P z e e
. (9a)
By substituting Eq. (9a) into Eq. (7), the wave number spectrum, P21mn can be represented as
0
3( )0 1
21 4
0
( )
2zmnik z zmn
mn
x y zmn
ib z z PP z e
iN N c k
. (9b)
In Eq. (9), homogeneous solutions are neglected to consider the only particular solution since the
transient responses associated with the homogeneous solutions decay out quickly. Reflective
waves propagating in the negative z-direction are also neglected in Eq. (9) since no reflection
occurs at the infinite boundary of the open 3-D space. Eq. (9) also satisfies the assumption that
there is negligible “dissipation” at the source surface: i.e., P21mn(z0) = 0. The dissipative sound
pressure spectrum, P21mn(z) increases as z increases in the supersonic region, i.e., (kxm)2 + (kyn)
2 <
k2. Therefore, a renormalization procedure is required to remove this secular term to obtain a
uniformly valid solution.
The solution of p22(x, y, z) in Eq. (8) can be written as
5
11
22 22
0 0
, ,yx
yx
NNiq k yil k x
lq
l q
p x y z P z e e
, (10a)
where the nonlinear wave number spectrum, P22lq is represented as
2 0
( )( ) 0 2 0
( )20 2
1( /2)( /2)2 2 4
0 0 2
22 ( )1 ( )12
1 1( )( ) 2 22 2 40 00 0 2 ( )( )
2,when 2 and 2
2
2,
zlq
zmn z l m q ny zlqx
ik z z
l q
x y zlq
lq i k k z zN ik z zN
mn l m q n
m nx y zlq zmn z l m q n
z z eP l m q n
N N c ik
P ze e
P PN N c k k k
otherwise
, (10b)
where
222 2
2 4zlq x yk k l k q k . (10c)
Similar to Eq. (9), homogeneous solutions are neglected in Eq. (10). Eq. (10) also satisfies the
assumption that there is neither a reflective wave generated at the infinite boundary of the open
3-D space nor a nonlinear component at the source surface, i.e., P22lq(z0) = 0. As in the
dissipative sound pressure spectrum shown in Eq. (9b), the nonlinear sound spectrum, P22lq(z)
includes the secular term that increases as z increases in the supersonic region.
2.3 Renormalization
In order to obtain the uniformly valid solution of Eq. (3) at a large z value, the secular terms
with the growing factor of (z – z0) in Eqs. (9b) and (10b) are eliminated by using a
renormalization procedure. From Eqs. (4), (9), and (10), the total sound pressure is written in the
complex form as
2
1 21 22, , , , , , , , ,i t i t i tp x y z t p x y z e p x y z e p x y z e . (11)
In order to eliminate the secular terms with (z – z0) in Eqs. (9b) and (10b), a strained coordinate,
ζ is introduced as
1 , , , ,z z x y z t , (12)
where ε is the small perturbation parameter (i.e., acoustic Mach number) and the function, z1 is
determined to eliminate all of the second-order small secular terms. When compared to the three
strained coordinates defined in the three directions of a 3-D space in Refs. 15 and 16, the only
one strained coordinate is used in Eq. (12) to renormalize the secular terms in the full, 3-D space.
By substituting Eq. (11) into Eqs. (4), (9), and (10) and retaining up to the second-order terms, a
condition for removing the secular terms is obtained: i.e., the strain coordinate, ζ can be found by
setting the absolute value of the following renormalization function, f to zero for a given set of
(x,y,z,t),
6
0
0
2 0
11
1
0 0
31
04
0
22( )1( /2)( /2) 2
04
0 0 2
1, , , ,
2
2 , when an
2
yx
y zlqx
zlq
zlq
NNiq k y ik zil k x i t
lq zlq
l qx y
ik zlq i t
zlq
ik zl q i t
x y zlq
f x y z t e e z P ik e eN N
Pibz z e e
c ik
Pz z e e l
N N c ik
d are evenq
. (13)
The uniformly valid total sound pressure is then represented as
0
( )( ) 0 2 0
11
1
0 0
( )1 ( )122
1 1( )( ) 2 240 00 0 2 ( )( )
1, , ,
2
yx
y zlqx
zmn z l m q ny zlqx
NNiq k y ik zil k x i t
lq
l qx y
i k k zN ik zNi t
mn l m q n
m nx y zlq zmn z l m q n
p x y z t e e P e eN N
e eP P e
N N c k k k
, (14)
where the subscripts, l and q satisfy l ≠ 2m or q ≠ 2n. The nonlinear and dissipative sound
pressure components are calculated by subtracting the linear NAH forward projected sound
pressure from the total sound pressure in Eq. (14).
2.4 Nonlinear and Dissipative Pulsating Sphere Simulation
In this section, a pulsating sphere, with the radius of rs, that radiates a nonlinear and
dissipative sound pressure field is considered to validate the proposed nonlinear, dissipative,
planar NAH procedure. Note that an acoustic monopole cannot be used to simulate the nonlinear
and dissipative sound field since its infinitely small size compared to the smallest wave length
makes it impossible to effectively radiate nonlinear sound fields10
. Therefore, the radius, rs is
here set to be in the same order of magnitude with the wave length, λ.
The nonlinear and dissipative sound pressure field generated from the pulsating sphere is
thus obtained from the generalized Burgers Equation10
in the spherical coordinates, (r, θ, φ) with
the assumption of the θ- and φ-direction symmetry conditions: i.e.,
2
3 2 3
0 0 02
p p b p p p
r r c c
, (15)
where τ = t-(r-rs)/c0 is the retarded time and r is the radial location (r ≥ rs) of a receiver. Similar
to the WWE, the spherical Burgers Equation also describes cumulative nonlinear effects.
According to the aforementioned perturbation and renormalization procedures, the sound
pressure, p in Eq. (15) has the uniformly valid solution: i.e.,
1,i
s
ep r Pr
, (16)
where α is the strained coordinate determined by the following two equations, i.e.,
1 , ,r r r , (17)
7
1 1 22, , ( ) ln
ii
s s
s
e rg r Pr r C r r C e
r
. (18)
In the renormalization function Eq. (18), P1 is the acoustic pressure amplitude on the sphere
surface at r = rs and C1 and C2 are the constants defined, respectively, as
2
1 13
02s
bC Pr
c , (19a)
2 2
2 13
0 02sC i P r
c
. (19b)
The linear sound pressure data on the hologram surface (z = zh) at the fundamental frequency, ω
can be obtained from Eq. (16) by replacing α with r = rhn where rhn represents the radial location
of the n-th receiver on the hologram surface. This hologram data is linearly backward projected
to the source surface at z = z0 and the reconstructed source sound pressure data is then used for
the nonlinear, forward NAH projections in the 3-D space. The linear and nonlinear sound
pressure fields directly-calculated from Eq. (16) in the 3-D reconstruction space are compared
with the reconstructed 3-D sound pressure fields to validate the projection performance of the
proposed NAH procedure.
3 PULSATING SPHERE SIMULATION RESULTS
3.1 Simulation Setup
Figure 1 shows the pulsating sphere simulation setup in a free field. The center of the
pulsating sphere with the radius of rs = 0.25 m is placed at (x,y,z) = (0.75,0.75,-0.25) m. The
sound pressure amplitude of the pulsating sphere on its undisturbed surface at r = 0.25 m is 6 kPa
and the excitation frequency is 1 kHz. A 31×31 acoustic pressure transducer array with the
sampling intervals of ∆x = ∆y = 0.05 m is placed at z = 0.05 m to obtain hologram sound
pressure data. Thus, the hologram height is 0.05 m when the plane on the pulsating sphere
surface at z = 0 m is defined as the source surface.
3.2 Directly-Calculated Sound Pressure Fields
The total (i.e., linear, nonlinear, and dissipative) sound pressure field is directly calculated
from Eq. (16), after α is determined by numerically searching for the roots of |g(r,α,τ)| = 0 in Eq.
(18). The linear, lossless component that is obtained by replacing α with r in Eq. (16) is shown
in Fig. 2(a) at the fundamental frequency of 1 kHz. The linear sound pressure amplitude in Fig.
2(a) decreases proportionally to 1/r. The cumulative nonlinear sound pressure field is obtained
by subtracting the linear and dissipative sound pressure from the total sound pressure (see Fig.
2(b)). The nonlinear sound pressure field in Fig. 2(b) has the minimum sound pressure level
(SPL) at (x,y,z) = (0.75,0.75,0) m (i.e., on the sphere surface at r = 0.25 m) where the nonlinear
SPL is approximately 115 dB. The nonlinear SPL is approximately 10 dB less than the linear
SPL at z = 3 m. Although not shown here, the dissipative sound pressure field that can be
obtained by considering only the dissipation source term in Eq. (15) and following the identical
perturbation and renormalization procedures in Sections 2.1 – 2.3 is negligible with low SPLs
when compared to the linear and nonlinear sound pressure fields.
8
3.3 Linearly Backward Projected Sound Pressure Fields
The sound pressure data calculated on the hologram surface at the fundamental frequency of
1 kHz is linearly backward projected onto the source surface by using the improved SONAH
procedure9. The linearly backward projected sound pressure field by using the improved
SONAH procedure is well matched with the directly-calculated sound pressure field as shown in
Figs. 3(a) – 3(c). Although the reconstruction error is relatively large at approximately 0.6 dB
along the measurement aperture edges, a spatial window can be used to suppress this
reconstruction error in the nonlinear forward NAH projections. The reconstructed sound
pressure data on the source surface at z = 0 m is input to the nonlinear NAH projection procedure.
3.4 Linearly and Nonlinearly Forward Projected Sound Pressure Fields
The sound pressures along the four reconstruction lines in Figs. 2(a) and 2(b) are
reconstructed by using the proposed nonlinear NAH procedure. The reconstruction lines are
selected in the “nearfield” of the pulsating sphere since the spatial-FFT-based, linear, forward
NAH procedure is limited to “nearfield” reconstructions due to the farfield reconstruction error
caused by spatial-FFT-periodicity-induced ghost images1-3
.
Figures 4(a) – 4(d) show the directly-calculated and NAH-projected SPLs along
reconstruction lines 1 – 4. Here, the NAH-projected “linear” SPLs are obtained by applying the
conventional, linear NAH procedure to the source data as in Eq. (4) and the NAH-projected,
“nonlinear” SPLs are calculated by subtracting the NAH-projected “linear” sound pressure
components from the “total” components obtained from the renormalization procedure as in Eq.
(14). The NAH-projected, linear SPLs match well with the directly-calculated, linear SPLs
along the all four reconstruction lines. The NAH-projected, nonlinear SPLs also match well with
the directly-calculated, nonlinear SPLs along the reconstruction lines 3 and 4. Along off-
centered reconstruction lines 1 and 2, the discrepancies between the directly-calculated and
NAH-projected nonlinear components in the region close to the source surface at z = 0 m are
caused by the assumption of no nonlinear component at the planar source surface, although the
pulsating sphere simulation results in nonlinear components at the source surface (see Fig. 2(b)).
Figures 5(a) – 5(d) show the reconstruction errors that are defined as the SPL differences
between the directly-calculated and NAH-projected SPLs in Fig. 4. Along off-centered
reconstruction lines 1 and 2, Figs. 5(a) and 5(b) show that the nonlinear NAH reconstruction
errors decrease quickly around the source surface as z increases. Along reconstruction line 3 at
the center of the measurement aperture, the nonlinear reconstruction error is within 0.2 dB, while
the linear reconstruction error increases oscillatory as z increases. The latter indicates that the
spatial-FFT-periodicity-induced ghost image effects and the measurement aperture truncation
errors become more significant as z increases further from the pulsating sphere1-3
. Along
reconstruction line 4, Fig. 5(d) shows that both the linear and nonlinear NAH reconstruction
errors are approximately 0.1 – 0.3 dB around the center and become larger as the reconstruction
location is further away from the center to the measurement aperture edges where the aperture
edge truncation errors and the ghost image effects are more significant than those at the center.
Note that Eq. (10) can be directly applied for the nonlinear NAH projections without
renormalizations although the renormalization procedure is necessary for long distance farfield
projections. Figures 6(a) and 6(b) show the effects of the renormalization along reconstruction
line 3 extended up to z = 2.5 m (i.e., z 7.3 where is the wave length at 1 kHz). The same
line and marker legends in Figs. 4 and 5 are reused in Fig. 6 except no renormalization case. In
Fig. 6(a), as z increases farther from the source surface at z = 0 m, the nonlinear component
becomes relatively significant compared to the linear component: i.e., the SPL difference
9
between the linear and nonlinear components becomes small as z increases. In Fig. 6(b), the
linear reconstruction error increases oscillatory up to 1.2 dB at z = 2.5 m, due to the spatial-FFT-
periodicity-induced ghost image effects and the measurement aperture truncation errors. The
nonlinear reconstruction error of the no renormalization case in Fig. 6(b) has almost identical
values with that of the renormalization case represented by the solid line with triangle markers.
The maximum nonlinear reconstruction error is approximately 0.5 dB at z = 2.5 m.
For the no renormalization procedure case, the exact and reconstructed nonlinear sound
pressure fields along with the reconstruction errors on x-y plane at z = 1.25 m and x-z plane at y =
0.75 m are shown in Fig. 7. Since both linear and nonlinear NAH reconstructions have
significant edge truncation errors, the x- and y-coordinates are chopped to the range of [0.4, 1.1]
m in Fig. 7. The reconstructed sound pressure fields agree well with the directly-calculated ones:
see Figs 7(a), 7(b), 7(d), and 7(e). Similar to the results shown in Figs. 5(a) to 5(c), the nonlinear
NAH reconstruction errors on the x-z plane decrease as z increases (see Fig. 7(f)).
4 CONCLUSIONS
In this article, the novel, nonlinear, dissipative, planar NAH algorithm is introduced that is
based on the first- and second-order perturbation solutions of the Westervelt Wave Equation
(WWE) and the modified renormalization procedure. The proposed NAH procedure is
applicable to sound fields with dominant cumulative nonlinearity and straight wave rays such as
spherical waves and 1-D plane waves. The WWE is decomposed into the first-order,
homogeneous and second-order, inhomogeneous wave equations. The conventional, spatial-
FFT-based, linear, planar NAH technique is applied to solve the first-order wave equation. The
resulting linear sound pressure solution is input to the second-order wave equation as the
inhomogeneous source terms to find the second-order, nonlinear sound pressure solutions. The
secular terms in the second-order solutions are removed by using the modified renormalization
approach to obtain the uniformly valid solution.
By applying the aforementioned perturbation and renormalization procedures to the
spherical Burgers Equation, the sound pressure field radiated from the high-level pulsating
sphere is calculated to validate the proposed NAH algorithm. In the nearfield of the pulsating
sphere up to z = 1.2 m 3.5, the SPLs reconstructed by using the proposed NAH procedure
match well with the directly-calculated SPLs at the maximum reconstruction error of
approximately 0.3 dB for the linear NAH-projections and 0.2 dB for the nonlinear NAH-
projections.
As z value increases farther from the sound source up to z = 2.5 m 7.3, the spatial-FFT-
based linear reconstruction solution suffers from the more significant spatial-FFT-periodicity-
induced ghost image effects and the measurement aperture truncation errors, which increase the
linear reconstruction error up to 1.2 dB. Then, this linear reconstruction error degrades the
performance of the proposed nonlinear NAH procedure with the maximum reconstruction error
of 0.5 dB.
In the proposed nonlinear NAH projections, the secular terms have negligible effects on the
uniformly valid solutions. Thus, it can be concluded that the proposed nonlinear and dissipative
NAH procedure can be applied to reconstruct the sound pressure fields successfully in the
nearfield of the pulsating sphere regardless of the renormalization procedure.
5 REFERENCES
1. J. D. Maynard, E. G. Williams, and Y. Lee, “Nearfield acoustic holography: I. Theory of
generalized holography and the development of NAH,” J. Acoust. Soc. Am. 78, 1395 – 1413
(1985).
10
2. W. A. Veronesi and J. D. Maynard, “Nearfield acoustic holography (NAH): II. Holographic
reconstruction algorithms and computer implementation,” J. Acoust. Soc. Am. 81, 1307 –
1322 (1987).
3. E. G. Williams, Fourier Acoustics: Sound Radiation and Nearfield Acoustical Holography,
15 – 114 (Academic, San Diego, CA, 1999).
4. J. Hald, “STSF – a unique technique for scan-based Nearfield Acoustical Holography
without restriction on coherence,” Brüel & Kjær Technical Review, No. 1, 1 – 50 (1989).
5. Y.-J. Kim, J. S. Bolton, and H.-S. Kwon, “Partial sound field decomposition in
multireference near-field acoustical holography by using optimally located virtual
references,” J. Acoust. Soc. Am. 115, 1641 – 1652 (2004).
6. H.-S. Kwon, Y.-J. Kim, and J. S. Bolton, “Compensation for source nonstationarity in
multireference, scan-based near-field acoustical holography,” J. Acoust. Soc. Am. 113, 360 –
368 (2003).
7. J. Hald, “Patch near-field acoustical holography using a new statistically optimal method,”
Brüel & Kjær Technical Review, No. 1, 40 – 50 (2005).
8. H.-S. Kwon, Y. Niu, and Y.-J. Kim, “Planar nearfield acoustical holography in moving fluid
medium at subsonic and uniform velocity,” J. Acoust. Soc. Am. 128, 1823 1832 (2010).
9. Y.-J. Kim and Y. Niu, “Improved Statistically Optimal Nearfield Acoustical Holography in
subsonically moving fluid medium,” J. Sound Vib. 331, 3945 – 3960 (2012).
10. M. F. Hamilton and D. T. Blackstock, Nonlinear Acoustics, 41 – 308 (Academic, San Diego,
CA, 1998).
11. P. J. Westervelt, “Parametric acoustic array,” J. Acoust. Soc. Am. 35, 535 537 (1963).
12. F. H. Fenlon, “Derivation of the multiple frequency Bessel-Fubini series via Fourier analysis
of the preshock time waveform,” J. Acoust. Soc. Am. 53, 1752 – 1754 (1973).
13. J. H. Ginsberg, “Multi-dimensional non-linear acoustic wave propagation, part II: The non-
linear interaction of an acoustic fluid and plate under harmonic excitation,” J. Sound Vib. 40,
359 – 379 (1975).
14. J. H. Ginsberg, “A re-examination of the non-linear interaction between an acoustic fluid and
a flat plate undergoing harmonic excitation,” J. Sound Vib. 60, 449 – 458 (1978).
15. J. H. Ginsberg, “Propagation of nonlinear acoustic waves induced by a vibrating cylinder. I.
the two-dimensional case,” J. Acoust. Soc. Am. 64, 1671 – 1678 (1978).
16. J. H. Ginsberg, “Propagation of nonlinear acoustic waves induced by a vibrating cylinder. II.
the three-dimensional case,” J. Acoust. Soc. Am. 64, 1679 – 1687 (1978).
17. S. G. Kelly and A. H. Nayfeh, “Non-linear propagation of directional spherical waves,” J.
Sound Vib. 72, 25 – 37 (1980).
6 FIGURES
Fig. 1 – Sketch of nonlinear, dissipative pulsating sphere simulation setup.
11
(a)
Reconstruction line 1:
(x,y) = (0.45,0.75) m
2: (x,y) = (0.6,0.75) m
3: (x,y) = (0.75,0.75) m
4: (y,z) = (0.75,1) m
(b)
Reconstruction line 1
2
3 4
Fig. 2 – 3-D sound pressure fields directly-calculated from pulsating sphere on x-y plane at z =
0 m and x-z plane at y = 0.75 m when excited at 1 kHz: (a) Linear, lossless sound pressure field,
and (b) Corresponding nonlinear sound pressure field at 2 kHz.
(c)(b)(a)
Fig. 3 – Linear sound pressure fields on x-y plane at z = 0 m: (a) Directly-calculated sound
pressure field, (b) SONAH-reconstructed sound pressure field, and (c) Reconstruction error (i.e.,
dB differences between reconstructed and directly-calculated sound pressure levels (SPLs)).
(a) (b)
(c) (d)
Fig. 4 – Directly-calculated and NAH-projected SPLs along reconstruction lines 1 – 4: (a)
Reconstruction line 1, (b) Reconstruction line 2, (c) Reconstruction line 3, and (d)
Reconstruction line 4.
12
(a) (b)
(c) (d)
Fig. 5 – NAH reconstruction errors along reconstruction lines 1 – 4: (a) Reconstruction line 1,
(b) Reconstruction line 2, (c) Reconstruction line 3, and (d) Reconstruction line 4. (a)
(b)
Fig. 6 – Reconstruction results with and without renormalization along reconstruction line 3
extended up to z = 2.5 m (The line and marker legends are same as Figs. 4 and 5 except no
renormalization case): (a) Directly-calculated and NAH-projected linear and nonlinear SPLs,
and (b) Reconstruction errors. (a) (b) (c)
(d) (e) (f)
Fig. 7 – Directly-calculated and NAH-projected nonlinear sound pressure fields on x-y plane at
z = 1.25 m and x-z plane at y = 0.75 m: (a) Directly-calculated results on the x-y plane, (b)
Reconstructed results on the x-y plane, (c) Reconstruction errors on the x-y plane, (d) Directly-
calculated results on the x-z plane, (e) Reconstructed results on the x-z plane, and (f)
Reconstruction errors on the x-z plane.