DF For AVS Vs Hydrophone - Cray

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

JULY 01 2001

Directivity factors for linear arrays of velocity sensors 


Benjamin A. Cray; Albert H. Nuttall

J Acoust Soc Am 110, 324–331 (2001)


https://doi.org/10.1121/1.1373706

CrossMark

 
View Export
Online Citation

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


Directivity factors for linear arrays of velocity sensors
Benjamin A. Craya) and Albert H. Nuttall
Naval Undersea Warfare Center Division, 1176 Howell Street, Newport, Rhode Island 02841

共Received 19 November 1999; revised 11 January 2001; accepted 23 March 2001兲


Some of the features unique to beamforming a linear array of acoustic velocity sensors, which are
not present with scalar-sensing elements 共such as conventional pressure sensors兲, are described in
this paper. Four types of sensors are considered here: a uniaxial motion sensor, which measures
acoustic particle velocity along a single axis; a biaxial motion sensor measuring velocity in two
orthogonal directions; a triaxial motion sensor that measures all three orthogonal components of the
velocity vector; and a sensor, denoted as an acoustic vector sensor, that measures acoustic pressure
as well as the complete velocity vector. Comparisons are made of the directivity index for each type
of sensor and for linear arrays of sensors. It is shown that uniaxial velocity sensors can have a
maximum directivity factor three times greater than an omnidirectional pressure sensor, a gain in
directivity index of 4.8 dB. Not surprisingly, this directivity gain is highly dependent on signal
arrival direction. Indeed, a uniaxial velocity sensor’s directivity can be less than that of an
omnidirectional pressure, indicative of a loss in signal level. These comparisons further indicate that
a single vector sensor can provide 6 dB of directivity gain, four times the directivity of a pressure
sensor. Line arrays of directional sensors can have a directivity index approximately 5 dB greater
than that of an identical line array of pressure sensors for approximately all azimuthal array

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


steerings. 关DOI: 10.1121/1.1373706兴
PACS numbers: 43.60.Qv, 43.30.Wi 关JCB兴

I. INTRODUCTION by D’Spain by quantifying, and comparing, the directivity


gains of conventional pressure-sensing arrays to those of lin-
Acoustic vector sensors are defined here to be sensors
ear vector-sensing arrays. The directivity index (N di) of an
that measure the amplitude and phase of acoustic pressure
and acoustic particle motion in a given direction at a collo- array is defined as a decibel measure of the improvement in
cated point. These sensors measure both a scalar 共pressure兲 the signal-to-noise ratio 共SNR兲 that a beamformed array pro-
and vector quantity, i.e., acoustic particle motion 共accelera- vides in an ideal isotropic noise field with a perfectly corre-
tion, velocity, or displacement兲. In this work only acoustic lated plane-wave signal, relative to an omnidirectional array
particle velocity will be considered. A single-axis, or uniaxis, element located in an ideal reflection-free field, or simply a
acoustic motion sensor measures one component of the free-field environment.
acoustic field vector, biaxial sensors measure motion in two Nehorai and Paldi7 developed an analytical model to
orthogonal directions, while a triaxial sensor measures all compare the direction-of-arrival 共DOA兲 estimation perfor-
three orthogonal components. These sensors can be config- mance of an array of vector sensors to that of an array that
ured into an array of elements, each element of which may measures acoustic pressure only. The paper treated both
be delayed, weighted, and summed, as commonly done with single and multiple acoustic sources in an ideal free-field
standard pressure hydrophones. environment. Cramer–Rao lower bounds 共CRLB兲 on DOA
A group of researchers1–13 have recently documented accuracy for the single-source single-vector sensor as well as
some aspects of beamforming arrays of vector sensors. multiple-source multivector sensors were given. The CRLB
D’Spain et al.1–3 constructed and deployed a 16-element ver- provides a universal measure of achievable unbiased estima-
tical line array that measured the three components of acous- tion accuracy and is commonly used by signal processors to
tic particle velocity as well as acoustic pressure. The low- compare the merits of different high-resolution DOA meth-
frequency array DIFAR, as in DIrectional low-Frequency ods. Hawkes and Nehorai8,9 examined the influence of a
Analysis and Recording, consisted of pressure hydrophones plane boundary near a vector-sensing array had on DOA es-
with three orthogonally mounted geophones. The output timations, and, as done previously in the paper by Nehorai,
from a geophone sensor is directly proportional to acoustic expressed for the CRLB for various stand-off distances from
particle velocity. At-sea data collected from the array were ideal pressure-release and rigid boundaries.
beamformed conventionally 共delayed and summed with uni- Pesotskii and Smaryshev10 considered the directional
form amplitude weighting兲 and adaptively, using a minimum gain of point, line, and planar arrays consisting of monopole
variance Capon approach on all four quantities. Both single- pressure sensors, and single-axis dipole sensors. Four types
element and whole array beamforming results were pre- of array configurations were examined—a perfectly absorb-
sented. ing baffle, a compliant baffle, a rigid baffle, and a free-field
This paper extends the theoretical framework presented environment. A comparison was then made for each type of
sensor on each type of baffle. The study concluded that the
a兲
Author to whom correspondence should be addressed. combination of measuring acoustic pressure with a single

324 J. Acoust. Soc. Am. 110 (1), July 2001 0001-4966/2001/110(1)/324/8/$18.00


monic time-dependence exp(i ␻ t) of the acoustic pressure
and particle velocities will be suppressed henceforth.
The gradient of acoustic pressure and the particle veloci-
ties are related via Euler’s linearized momentum equation.11
For harmonic waves,

⳵ v共 3 兲
ⵜp 共 r兲 ⫽⫺ ␳ 0
⳵t
⫽⫺i ␻␳ 0 共 v x 共 r兲 nx ⫹ v y 共 r兲 ny ⫹ v z 共 r兲 nz 兲 , 共6兲
where the density of the fluid medium is denoted ␳ 0 .
Pressure-gradient hydrophones measure the quantity ⵜ p
by measuring the phase difference between two spatially
separated hydrophones. A well-defined dipole 共or a cardioid
acoustic response兲 can be obtained with the paired hydro-
FIG. 1. Array coordinates and geometry for an equispaced line array of
multiaxis velocity sensors along the x-axis. phones. The technique introduces finite difference approxi-
mation errors due to the geometry of the paired hydrophones
and not to the inherent sensitivity of either hydrophone.
component of the acoustic velocity increases the direction Each orthogonal component measured by an ideal veloc-
gain four-, three-, and two-fold for a point, line, and planar ity sensor has an amplitude response proportional to the co-

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


array in the free field, respectively. This is in agreement with sine of the conical angle ⌰ between the plane-wave inci-
a subset of the results presented here. dence arrival direction and each component normal 共the
component’s maximum response axis兲. Thus

V j 共 ␪ , ␾ 兲 ⫽V cos共 ⌰ j 兲 , j⫽x,y,z, 共7兲


II. SINGLE SENSOR RESPONSE
where direction cosines
In Cartesian 共Fig. 1兲 coordinates, single-frequency, time-
harmonic acoustic plane waves, which have wavefronts cos共 ⌰ x 兲 ⫽cos共 ␪ 兲 sin共 ␾ 兲 , 共8兲
propagating in the direction of acoustic wave vector ⫺k, can
be characterized by acoustic pressure or by acoustic particle cos共 ⌰ y 兲 ⫽sin共 ␪ 兲 sin共 ␾ 兲 , 共9兲
velocity. Thus
cos共 ⌰ z 兲 ⫽cos共 ␾ 兲 , 共10兲
p 共 r,t 兲 ⫽Re兵 P 共 ␪ , ␾ 兲 exp关 i 共 k•r⫹ ␻ t⫹ ␾ p 兲兴 其 , 共1兲
and V is the amplitude of the incident plane-wave acoustic
v x 共 r,t 兲 ⫽Re兵 V x 共 ␪ , ␾ 兲 exp关 i 共 k•r⫹ ␻ t⫹ ␾ x 兲 其 , 共2兲
velocity. The pressure is omnidirectional, that is, P( ␪ , ␾ )
v y 共 r,t 兲 ⫽Re兵 V y 共 ␪ , ␾ 兲 exp关 i 共 k•r⫹ ␻ t⫹ ␾ y 兲 其 , 共3兲 ⫽ P.
Euler’s momentum equation relates the gradient of the
and
pressure field to acoustic particle velocity. For single fre-
v z 共 r,t 兲 ⫽Re兵 V z 共 ␪ , ␾ 兲 exp关 i 共 k•r⫹ ␻ t⫹ ␾ z 兲 其 , 共4兲 quency harmonic plane waves, this relationship reduces14 to
the well-known definition for a fluid’s characteristic imped-
where 储 k储 (⫽ ␻ /c) is the acoustic wave number, ␻ is the
ance 兩 p(r,t) 兩 / 兩 v (r,t) 兩 ⫽ ␳ c. Thus one may define an equiva-
circular frequency, c is the sound speed, r⫽ 关 x,y,z 兴 is the
lent particle velocity amplitude, or normalized pressure, that
position vector, P( ␪ , ␾ ) is the amplitude of the sound wave
is measured by the pressure sensor
pressure, and V x ( ␪ , ␾ ), V y ( ␪ , ␾ ), V z ( ␪ , ␾ ) are the ampli-
tudes of the component velocities in the x, y, and z direc- v p 共 r,t 兲 ⫽V exp关 i 共 k•r⫹ ␻ t 兲兴 , 共11兲
tions, respectively. The amplitudes are real-valued and de-
pendent on azimuthal angle ␪ and polar angle ␾. The where V⫽⫺ P/ ␳ c.
complete acoustic velocity field may be written as Equation 共5兲 represents the velocity vector, whereas Eq.
共1兲, or equivalently, Eq. 共11兲, is the scalar pressure field. To
v共 3 兲 共 r,t 兲 ⫽ v x 共 r,t 兲 nx ⫹ v y 共 r,t 兲 ny ⫹ v z 共 r,t 兲 nz . 共5兲 combine the scalar and vector quantities, a weighting vector
The vectors 兵 nx ,ny ,nz 其 are mutually perpendicular unit vec- 共which may have complex-valued components兲 can be intro-
tors aligned as shown in Fig. 1. The elements of a uniaxis duced: w⫽w x nx ⫹w y ny ⫹w z nz .
vector-sensing array measure only one of the velocity vector
components, such as v x , v y , or v z , while the multiaxis ar-
rays measure two or more of these orthogonal vectors. A III. BEAMFORMING AN ARRAY OF VELOCITY
superscript may be used to denote the dimension of the mo- SENSORS
tion sensor.
For acoustic planewave propagation, the components of A possible approach to beamforming an array of triaxial
the collocated motion sensor are assumed to be in-phase, velocity sensors would be to delay, weight, and sum the
with the component phases set to zero. For brevity, the har- velocity components from each sensor separately. Thus

J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors 325
v共兺3 兲 共 ␪ , ␾ 兲 ⫽V cos共 ⌰ x 兲 冕 N⫺1

n⫽0
w xn exp关 i 共 k⫺ks 兲 •rn 兴 nx
where the amplitude shading coefficients w xn , w yn , and w zn
are arbitrary. The wave vector k corresponds to the acoustic
plane wave arriving from any given direction, ( ␪ , ␾ ), and ks
N⫺1
is the wave vector which corresponds to the chosen steered,
⫹V cos共 ⌰ y 兲 兺
n⫽0
w yn exp关 i 共 k⫺ks 兲 •rn 兴 ny or look direction ( ␪ s , ␾ s ) of the array.
One may then be tempted to simply take the norm 共or
N⫺1 length of the weighted sum of all the vector components兲 of
⫹V cos共 ⌰ z 兲 兺
n⫽0
w zn exp关 i 共 k⫺ks 兲 •rn 兴 nz , 兺 in Eq. 共12兲 as a means to obtain the beam response of the
v(3)
summed vector components. That is, the norm may be writ-
共12兲 ten as

␤ 共 3 兲共 ␪ , ␾ 兲 ⫽ 冑冉 兺 N⫺1

n⫽0
w xn v xn 冊 冉兺
2


N⫺1

n⫽0
w yn v yn 冊 冉兺
2


N⫺1

n⫽0
w zn v zn 冊 2

, 共13兲

where v xn ⫽V x ( ␪ , ␾ )exp兵 i(k⫺ks )•rn 其 , etc. This approach may vary at each element array location. The resulting scalar

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


results, however, in taking the square-root of the sum of the field may then be delayed and summed in a conventional
squared vector velocity components; a nonlinear processing manner. Thus in general,
technique.

冏兺 冏
Another approach to array beamforming would be to N⫺1 2
steer each velocity sensor toward the source, and then obtain B 共3兲
共 ␪,␾ 兲⫽ w共n3 兲 •v共n3 兲 e i 共 k⫺ks 兲 •rn , 共15兲
the squared magnitude of the sum of the sensors. That is, n⫽0

form the inner product of the summed velocities in Eq. 共12兲,


v(3)
兺 , with a steered triaxial velocity sensor of unit amplitude and for the line array,
and then take the magnitude-squared of the product. Thus for

冏兺
the line array in Fig. 1, (rn ⫽ 关 x n ,0.,0.兴 ), this approach yields N⫺1

B 共L3 兲 共 ␪ , ␾ 兲 ⫽ 共 w xn V x ⫹w yn V y ⫹w zn V z 兲
B 共 3 兲 共 ␪ , ␾ 兲 ⫽ 储 v共兺3 兲 共 ␪ , ␾ 兲 •v共 3 兲 共 ␪ s , ␾ s ,0兲 储 2 n⫽0

⫽g 共 3 兲 共 ␪ s , ␾ s , ␪ , ␾ 兲 2
冉 冊冏
2

⫻exp i x ␣共 ␪,␾ 兲 共16兲

冏兺 冊冏
.

冉 c n
N⫺1 2

⫻ w n exp i x n ␣ 共 ␪ , ␾ 兲 , 共14兲
n⫽0 c
Hence, the result here is a linearly weighted combination of
where all three measured velocity components. That is, this ap-
proach yields the square of the sum of the weighted velocity
g 共 3 兲 共 ␪ s , ␾ s , ␪ , ␾ 兲 ⫽V 共 sin ␾ sin ␾ s cos共 ␪ ⫺ ␪ s 兲 components. A special case of the general weighting vector,
w(3)
n , would be constant direction cosine weights.
⫹cos ␾ cos ␾ s 兲 ,
The most general linear processing would be to augment
Eq. 共15兲 with a measurement of acoustic pressure. Processing
and
all 4N measured quantities, i.e., pressure and three velocity
␣ 共 ␪ , ␾ 兲 ⫽cos共 ␪ 兲 sin共 ␾ 兲 ⫺cos共 ␪ s 兲 sin共 ␾ s 兲 . components at each element location, yields

The quantity g (3) reaches a minimum of V when ( ␪ , ␾ )


⫽( ␪ s , ␾ s ) and ( ␪ , ␾ )⫽( ␪ s ⫹ ␲ ,⫺ ␾ s ). Hence, without a cor-
responding measurement of pressure, the triaxial velocity
B 共 pv兲
共 ␪,␾ 兲⫽ 冏兺
N⫺1

n⫽0
共 w共n3 兲 v共n3 兲 ⫹w pn v pn 兲 e i 共 k⫺ks 兲 •rn 冏 2

. 共17兲

sensor will have directional ambiguity.


This second approach to beamforming is limited in its There are now 4N degrees of freedom to manipulate, in or-
capability. The restriction of weighting values to be essen- der to maximize some measure of performance, such as the
tially the direction cosines and constant from element to el- array output SNR. These 4N degrees of freedom are the N
ement disallows emphasis of some element outputs over oth- pressure weights 兵 w pn 其 and the N velocity weights for each
ers and it imposes the same relative weighting to the three orthogonal direction, namely, 兵 w xn 其 , 兵 w yn 其 , 兵 w zn 其 for 1⭐n
velocity components. ⭐N.
A more general linear processing technique would form For reference, the power response for a line array of
the inner product of the velocity at each element location unit-amplitude scalar pressure sensors located along the
with an arbitrary element weighting vector, w(3) n , which now x-axis is

326 J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors
B 共p兲
共 ␪,␾ 兲⫽ 冏兺
N⫺1

n⫽0
w pn


exp i x n ␣ 共 ␪ , ␾ 兲
c 冊冏
2

. 共18兲
A similar calculation for a uniaxial sensor that measures
the v x component of particle velocity gives

IV. ANALYTICAL EVALUATION OF DIRECTIVITY


I⫽ 冕 冕
0
2␲ ␲

0
共 w x0 cos共 ␪ 兲 sin共 ␾ 兲兲 2 sin共 ␾ 兲 d ␾ d ␪

Directivity is a fundamental measure of array perfor- 4␲ 2


mance; the directivity factor, DF, of an array can be ex- ⫽ w . 共24兲
3 x0
pressed as15
Upon substitution in Eq. 共19兲,
4␲B共 ␪s ,␾s兲
DF⫽ . 共19兲 DF 共 x 兲 ⫽3 cos2 共 ␪ s 兲 sin2 共 ␾ s 兲 . 共25兲
兰 20 ␲ 兰 0␲ B 共 ␪ , ␾ 兲 sin共 ␾ 兲 d ␾ d ␪
Likewise, for the velocity components v y and v z ,
The directivity index is 10 log(DF).
Calculation of the DF typically requires the numerical DF 共 y 兲 ⫽3 sin2共 ␪ s 兲 sin2共 ␾ s 兲 , 共26兲
evaluation of the above double integral. However, for certain
and
array geometries and frequencies 共e.g., for an equally spaced
line array of pressure sensors兲, the integrals may be analyti- DF 共 z 兲 ⫽3 cos2共 ␾ s 兲 . 共27兲
cally evaluated exactly.
Equations 共25兲, 共26兲, and 共27兲 indicate that a single velocity
The maximum array response at steering angles ( ␪ s , ␾ s )
sensor may have a maximum directivity factor three times
for a given sensor type can be obtained from Eqs. 共15兲, 共17兲,
greater than an omnidirectional pressure sensor, that is, a
and 共18兲. Hence, for unit-amplitude pressure, triaxial, and

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


gain in directivity index of 4.8 dB. Furthermore, directivity
vector sensors, the maximum linear array responses may be
factors of velocity sensors are highly dependent on azimuth
written as
and elevation steering angles ( ␪ s , ␾ s ). At certain steering

B 共p兲
共 ␪s ,␾s兲⫽ 冏兺 冏
N⫺1

n⫽0
w pn ,
2
angles, the directivity factor of a uniaxial velocity sensor can
be less than that of an omnidirectional pressure sensor.
The most general type of single array element would be

冏兺N⫺1 共20兲
a vector sensor. The maximum response of such an element
共3兲
B 共 ␪s ,␾s兲⫽ a 共 ␪ s , ␾ s 兲 w xn ⫹b 共 ␪ s , ␾ s 兲 w yn is given as
n⫽0

冏 2 B 共 p v 兲 共 ␪ s , ␾ s 兲 ⫽ 共 w p0 ⫹a 共 ␪ s , ␾ s 兲 w x0 ⫹b 共 ␪ s , ␾ s 兲 w y0
⫹c 共 ␾ s 兲 w zn , ⫹c 共 ␾ s 兲 w z0 兲 2 , 共28兲

and and the two-fold integral becomes

B 共 pv兲
共 ␪s ,␾s兲⫽ 冏兺
N⫺1

w pn ⫹a 共 ␪ s , ␾ s 兲 w xn ⫹b 共 ␪ s , ␾ s 兲 w yn
I⫽ 冕 冕
0
2␲ ␲

0
B 共 p v 兲 共 ␪ , ␾ 兲 sin共 ␾ 兲 d ␾ d ␪
n⫽0

冏 2 4␲
⫽ 共 3w 2p0 ⫹w x0
2
⫹w 2y0 ⫹w z0
2
兲. 共29兲
⫹c 共 ␾ s 兲 w zn , 共21兲 3
Hence, the directivity factor of a single element that mea-
where a( ␪ , ␾ )⫽cos( ␪ )sin( ␾ ), b( ␪ , ␾ )⫽sin( ␪ )sin( ␾ ), and sures all three components of particle velocity and acoustic
c( ␾ )⫽cos( ␾ ) for all ␪,␾. These maximum array responses pressure is
provide the numerator 关 B( ␪ s , ␾ s ) 兴 for DF in Eq. 共19兲.
The effort in obtaining expressions for the directivity B pv共 ␪ s , ␾ s 兲
factors is then in analytically evaluating the double integral DF 共 p v 兲 ⫽ . 共30兲
共 w 2p0 兲 ⫹ 31 共 w x0
2
⫹w 2y0 ⫹w z0
2

given in Eq. 共19兲, to be noted by I.
The goal now becomes that of determining the weights
A. Directivity factors for a single element, N Ä1
that maximize the above ratio. Without loss in generality,
It is straightforward to derive the directivity factor of a one may let w p0 ⫽1 and solve the following equations simul-
single omnidirectional pressure sensor. Namely, taneously:

I⫽ 冕 冕 0
2␲

0

兩 w p0 兩 2 sin共 ␾ 兲 d ␾ d ␪ ⫽4 ␲ 共 w p0 兲 2 . 共22兲 ⳵ DF p v
⳵ w x0

⳵ DF p v
⳵ w y0

⳵ DF p v
⳵ w z0
⫽0. 共31兲

Hence, upon substituting Eqs. 共22兲 and 共20兲 into Eq. 共19兲,
This leads to two sets of optimal weights, one corre-
4 ␲ 共 w 2p0 兲 sponding to a minimum DF p v and the other, given below,
DF 共 p 兲 ⫽ ⫽1. 共23兲 for maximum directivity, i.e.,
4 ␲ 共 w 2p0 兲
w x0 ⫽3a 共 ␪ s , ␾ s 兲 , 共32兲
As expected, the directivity index is zero and the single pres-
sure sensor is omnidirectional. w y0 ⫽3b 共 ␪ s , ␾ s 兲 , 共33兲

J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors 327
w z0 ⫽3c 共 ␾ s 兲 . 共34兲
Substituting the optimal real weights in Eqs. 共32兲, 共33兲,
and 共34兲 into Eq. 共30兲 yields
DF p v ⫽1⫹3 共 a 共 ␪ s , ␾ s 兲 2 ⫹b 共 ␪ s , ␾ s 兲 2 ⫹c 共 ␾ s 兲 2 兲⫽4, 共35兲
since a( ␪ , ␾ ) ⫹b( ␪ , ␾ ) ⫹c( ␾ ) ⫽1 for all ␪,␾.
2 2 2

Therefore, the maximum directivity for a single sensor


that measures all velocities as well as acoustic pressure is
DI⫽10 log(4)⫽6 dB, and this holds for any steering angle.
This gain is equivalent to the gain obtained from a four-
element line array of pressure sensors 共at the frequency for
which there is half-wavelength separation between ele-
ments兲. The four-element line array would be longer than the
one-element vector-sensing array, and would retain an am- FIG. 2. Maximum directivity for an N-element equispaced line array at
biguous back lobe response. Of course, one could configure broadside ( ␪ s ⫽90°) utilizing optimal real weights.
the four pressure elements, not as a linear array, but as a
pressure-gradient sensing tetrahedron. This, then, would
For large N and for u s ⫽cos( ␾ s ) not close to ⫾1, one may
have the equivalent properties of the single-vector sensor.
argue, as in the method of stationary phase, that for smooth
weightings, the sum of the exponential terms in Eq. 共39兲

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


B. Directivity factors for an N-element line array essentially samples the remainder of the integrand at u
⫽u s . Therefore,
To simplify the derivation here, the line array elements


will now be taken to be equispaced along the z-axis at spac- N⫺1 N⫺1
1
ing d z ⫽␭/2. This array orientation removes the azimuthal I⬵ ␲ 共 1⫺u s2 兲 兺 兺
⫺1 n⫽0 m⫽0
w yn w ym e i 共 n⫺m 兲 ␲ 共 u⫺u s 兲 du
angle dependence 共␪兲 from the exponential term e i(k̄⫺k̄s )•r̄0
关given in Eq. 共17兲兴 and simplifies the subsequent integration N⫺1

for directivity. The integral given by Eq. 共22兲 for pressure ⫽2 ␲ sin 共 ␾ s 兲
2

n⫽0
w 2yn . 共40兲
sensors may be integrated directly over azimuth angle ␪, and
assuming real-valued weights, we obtain The DF then becomes, substituting Eq. 共40兲 into Eq.
共19兲,
兺 兺 w pn w pm 冕0 exp关 i ␲ 共 n⫺m 兲共 cos共 ␾ 兲
N⫺1 N⫺1

I⫽2 ␲ 2 sin2共 ␾ s 兲共 兺 n⫽0
N⫺1
w yn 兲 2
n⫽0 m⫽0
DF y ⫽ N⫺1
. 共41兲
兺 n⫽0 共 w yn 兲 2
⫺cos共 ␾ s 兲兲兴 sin共 ␾ 兲 d ␾ . 共36兲
This agrees with the numerical calculations presented in Fig.
Completing the ␾-integration yields an exact expression for
2, which will be described in more detail in Sec. V, and
the DF of a line array of pressure sensors, i.e.,
shows that, as the number N of array elements increase, and
N⫺1
共 兺 n⫽0 w pn 兲 2 for ␾ s ⫽90°, the directivity of a uniaxial velocity-sensing
DF p ⫽ N⫺1 2
. 共37兲 array approaches a gain that is 3 dB greater than that of a
兺 n⫽0 w pn
pressure-sensing array. Approximations for DF x , DF z , and
This ratio is a maximum for uniform weights and results16 in even DF p v can be obtained in a similar manner.
an optimal directivity of DI⫽10 log(N).
Similar analytical expressions for the DF of vector-
sensing arrays are more difficult to obtain and require one to V. DERIVATION OF ARRAY GAIN
approximate the DF by assuming that there are a large num- Array gain provides an SNR metric for an array’s per-
ber of array elements N. For example, consider an array of formance in various types of noise fields, directivity is lim-
uniaxial velocity sensors, again aligned on the z-axis, that ited to ideal isotropic noise. The formulation here will allow
measure the v y component of particle velocity. Then for the analysis of general noise fields, provided the array’s

I⫽ 冕 冕
2␲

0 0

sin2共 ␪ 兲 sin2共 ␾ 兲 冏兺
N⫺1

n⫽0
w yn exp关 in ␲ 共 cos共 ␾ 兲
covariance, denoted as R, is known or otherwise measured.
Results, however, are presented for the specific case of
spherically isotropic noise. To simplify the derivation of ar-

冏 2

⫺cos共 ␾ s 兲兲兴 sin共 ␾ 兲 d ␾ d ␪ . 共38兲


ray gain, column matrices are introduced. Hence, the array’s
beam response is recast as
N⫺1
Letting u⫽cos( ␾ ) and u s ⫽cos( ␾ s ) and integrating over ␪ ␤ 共 p v 兲 共 ␪ , ␾ 兲 ⫽V 兺 e n* 共 ␪ s , ␾ s 兲 e n 共 ␪ , ␾ 兲
yields n⫽0

I⫽ ␲ 冕 1

⫺1
共 1⫺u 兲 2
冏兺
N⫺1

n⫽0
w yn e 冏
in ␲ 共 u⫺u s 兲
2

du. 共39兲
⫻ 关 w pn ⫹w xn sin␾ cos ␪
⫹w yn sin ␾ sin ␪ ⫹w zn cos ␾ 兴 , 共42兲

328 J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors
FIG. 3. Maximum directivity for an N-element equispaced line array at FIG. 4. Directivity versus azimuthal steering angle ( ␪ s ) for a 10-element
endfire ( ␪ s ⫽0°) utilizing optimal real weights. equispaced line array at elevation steering angle ␪ s ⫽90°.

where e n ( ␪ , ␾ )⫽exp关 ik(x n sin( ␾ )cos( ␪ )⫹y n sin( ␪ ) sin( ␪ ) weights in Eq. 共43兲 are real, Eq. 共49兲 can be written as
⫹z n cos( ␾ )) 兴 , and the 4N weights 兵 w pn 其 , 兵 w xn 其 , 兵 w yn 其 , B 共0p v 兲 ⫽W T R r W, 共50兲
兵 w zn 其 , are defined to be real but otherwise arbitrary.
where R r is the real part of the covariance matrix R.

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


Two 4N⫻1 real column matrices, matrices of weights
and direction cosines, are defined as We define the array gain as the quotient of the output
SNR to the input SNR. There follows
W⫽ 关 w p0 . . . w pN⫺1 w x0 . . . w xN⫺1 w y0 . . . w yN⫺1
关 W T D 共 ␪ s , ␾ s 兲兴 2
⫻w z0 . . . w zN⫺1 兴 , T
共43兲 AG⫽ , 共51兲
W TR rW
D 共 ␪ , ␾ 兲 ⫽ 关 1 . . . 1 cos共 ⌰ x 兲 . . . cos共 ⌰ y 兲 . . . cos共 ⌰ z 兲 . . . 兴 . where we have normalized the input noise pressure power at
Then, when the plane-wave signal arrival angle coincides unity.
with the steering angle, namely, ( ␪ , ␾ )⫽( ␪ s , ␾ s ) the array The 4N optimum real weights that maximize this array
signal output waveform is expressible as gain are
Y s ⫽VW T D 共 ␪ s , ␾ s 兲 . 共44兲 W 0 ⫽ ␣ R r⫺1 D 共 ␪ s , ␾ s 兲 , ␣ arbitrary, 共52兲
The array signal output power is then while the corresponding optimum array gain is
B 共s0p v 兲 ⫽ 兩 Y s 兩 2 ⫽ 兩 V兩 2 关 W T D 共 ␪ s , ␾ s 兲兴 2 . 共45兲 AG 0 ⫽max AG⫽D T 共 ␪ s , ␾ s 兲 R r⫺1 D 共 ␪ s , ␪ s 兲 . 共53兲
At the same time, the array output noise waveform can This relation was used in the following section to obtain the
be expressed as numerical values for maximum directivity 共optimum array
gain in an ideal isotropic noise field兲 in Figs. 2–6.
Y 0 ⫽W T V 0 , 共46兲

冋册
with the 4N⫻1 column matrix VI. DIRECTIVITY CALCULATIONS
␤共p兲 Figure 2 compares the maximum directivity index of an
␤共x兲 N-element pressure sensing line array to velocity-sensing
V 0⫽ , 共47兲 line arrays. The elements of the arrays are equispaced at
␤共y兲
half-wavelength spacing and are placed along the x-axis in
␤共z兲
the free field. The arrays have been steered to broadside
and the four N⫻1 column matrices ( ␪ s ⫽ ␾ s ⫽90°) in Fig. 2 and endfire ( ␪ s ⫽0°, ␾ s ⫽90°) in
␤ 共 p 兲 ⬅ 关 e 0* 共 ␪ s , ␾ s 兲v p0 . . . e N⫺1
* 共 ␪ s , ␾ s 兲v pN⫺1 兴 T , Fig. 3.
For a plane-wave signal arriving at broadside, that is,
␤ 共 x 兲 ⬅ 关 e 0* 共 ␪ s , ␾ s 兲v x0 . . . e N⫺1
* 共 ␪ s , ␾ s 兲v xN⫺1 兴 T , along the y-axis, there would be no particle velocity in the x-
or z-directions. Hence, for this array steering, there would be
␤ 共 y 兲 ⬅ 关 e 0* 共 ␪ s , ␾ s 兲v y0 . . . e N⫺1
* 共 ␪ s , ␾ s 兲v yN⫺1 兴 T , 共48兲 no additional directivity gains from measuring the velocities
␤ 共 z 兲 ⬅ 关 e 0* 共 ␪ s , ␾ s 兲v z0 . . . e N⫺1
* 共 ␪ s , ␾ s 兲v zN⫺1 兴 T . v x or v z . Similarly, for plane waves arriving at endfire, only
the x-component of acoustic particle velocity is nonzero.
This leads to the array output average noise power in the Since the directivity calculations are based on a half-
form wavelength equispaced line, an N⫽10-element pressure-
sensing array has an acoustic aperture 5␭ 共a physical aperture
B 共0p v 兲 ⫽ 兩 Y 0 兩 2 ⫽W T V 0 V H T
0 W⫽W RW, 共49兲
of 4.5␭兲 for a directivity of 10 dB. Notice, in Fig. 2, that the
where R is a Hermitian 4N⫻4N covariance matrix of the array that measures both pressure and velocity provides 10
totality of 4N received noise waveforms. Because all the dB of directivity for an acoustic aperture of only 1.5␭, and

J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors 329
comparison to the endfire beamwidth. That is, with increas-
ing aperture, the directivity is determined entirely by the ar-
ray beamwidth and is minimally influenced by the individual
element response.
The figures also show the directivity of an array that
measures both acoustic pressure and the y-component of par-
ticle velocity. Including the pressure increases the directivity
for both array steerings. Notice that, for a single element
(N⫽1) and for either array steerings, the uniaxial velocity
sensor provides a 4.8-dB gain over the omnidirectional
pressure-sensing element. Measurement of both pressure and
velocity yields a 6-dB gain.
In Fig. 4, a comparison is made between the variation of
directivity with azimuth for a pressure-sensing and vector-
FIG. 5. Directivity versus azimuthal steering angle ( ␪ s ) for a 10-element sensing line array at a fixed frequency ( f ⫽c/2d x ). A 10-
equispaced line array at elevation steering angle ␪ s ⫽75°. element line array, again aligned along the x-axis and in the
free field, is assumed and the elevation angle is fixed at
the total number of channels, namely, 6, is still less than the broadside. In Fig. 5, the elevation angle is fixed at 75°,
number of channels of the pressure-sensing array. whereas in Fig. 6, the elevation angle is 60°. In the free field,
The difference between the directivity gains shown in the optimal directivity of an array of omnidirectional pres-

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


these figures is due to the orientation of the velocity sensors sure sensors is proportional to the number of array elements,
( v x and v y ) with respect to the incident plane-wave signal. N, for all array steerings. Hence, the constant directivity in-
In Fig. 2, the difference in directivity between pressure sens- dex of 10 dB for the pressure-sensing array is observed. No-
ing and velocity sensing ( v y -component only兲 remains es- tice in Fig. 4 that the array of uniaxial velocity sensors, de-
sentially constant with increasing array aperture. This is be- noted v y , provides 3-dB gain over pressure sensors for most
cause the azimuthal 共that is, in the ␪-plane shown in Fig. 1兲 azimuthal steering angles. This gain is due to the differences
beamwidth narrows, with increasing aperture, identically for in the cosine and omnidirectional element response patterns.
the pressure and velocity sensors. Hence, the gain shown is However, at oblique elevation angles, this gain diminishes in
due to a constant difference between elevation angle 共the proportion to the cosine of the elevation angle.
␾-plane in Fig. 1兲 beamwidths. The uniaxial velocity sensor Biaxial and triaxial sensors further improve directivity,
has a cosine element response, whereas the pressure sensor is particularly for incident angles far from broadside where the
assumed omnidirectional. In elevation, the directivity of the gains can be very large. For example, in Fig. 5, the directiv-
velocity sensor is twice that of the pressure sensor. Hence, ity index of the uniaxial array is 5 dB at 77° azimuth. On the
there is a constant difference of 3 dB. other hand, the triaxial array directivity is 14 dB. Notice that
For endfire steering, the difference between the directiv- the maximum gain is achieved when the pressure and all
ity index of the uniaxial ( v x ) velocity-sensing array and the three components of particle velocity are measured and that
pressure-sensing array diminishes with increasing array ap- this optimal gain is essentially constant for all array steer-
erture. Again, the endfire beamwidth narrows identically for ings. However, measuring each component of acoustic ve-
both the pressure- and velocity-sensing arrays with increas- locity can be difficult. Additional self-noise mechanisms ex-
ing aperture. However, since the arrays are steered to endfire, ist with velocity sensors.
the effect of the broad angular difference between the cosine
and omnidirectional element response becomes negligible in
VII. CONCLUSIONS

Presented in this study are formulations for the beam


response and directivity of arrays of uniaxial, biaxial, triaxial
velocity sensors, and vector sensors; sensors that measure
pressure and all components of acoustic velocity at a collo-
cated point. To simplify the development, each orthogonal
component of a multiaxis velocity sensor was assumed to be
collocated and the orientation of the components was un-
varying from sensor to sensor. However, the amplitude shad-
ing coefficients for each component of particle velocity was
allowed to be arbitrary.
Figures 4, 5, and 6 indicate that a 10-element line array
of vector sensors, with elements separated by one-half an
acoustic wavelength, has a directivity index approximately 5
dB greater than that of an identical line array of pressure
FIG. 6. Directivity versus azimuthal steering angle ( ␪ s ) for a 10-element sensors for approximately all azimuthal array steerings. This
equispaced line array at elevation steering angle ␪ s ⫽60°. implies that vector-sensing arrays, albeit with a larger num-

330 J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors
4
ber of output channels, may be physically smaller than K. T. Wong and D. Zoltowski, ‘‘Closed-form underwater acoustic
pressure-sensing arrays, without a loss in directivity. direction-finding with arbitrarily spaced vector hydrophones at unknown
By definition, the uniaxial velocity sensor is more direc- locations,’’ IEEE J. Ocean Eng. 22, 566–575 共1997兲.
5
J. C. Nickles, G. L. Edmonds, R. A. Harris, F. H. Fisher, W. S. Hodgkiss,
tional than an omnidirectional pressure-sensing element, and
J. Giles, and G. L. D’Spain, ‘‘A Vertical Array of Directional Acoustic
it is this directionality that provides the constant 3-dB in- Sensors,’’ IEEE Oceans ’92 Conference Proceedings, pp. 340–345, 1992.
crease in the array directivity index shown in Fig. 2. This 6
P. Gerstoft and J. T. Goh, ‘‘Performance evaluation of horizontal and
gain then, for broadside array steering only, is due simply to vertical vector sensor arrays in shallow water environments,’’ Proceedings
the assumed differences in the pressure and uniaxial element 16th International Conference on Acoustics, pp. 1643–1644, 1998.
7
responses. A. Nehorai and E. Paldi, ‘‘Acoustic vector-sensor array processing,’’
IEEE Trans. Signal Process. 42, 2481–2491 共1994兲.
It should be noted that pressure sensors on or near 8
M. Hawkes and A. Nehorai, ‘‘Surface-Mounted Acoustic Vector-Sensor
acoustic baffles often have an element response pattern ap- Array Processing,’’ Proc. IEEE International Conference on Acoustics,
proximated well by a cosine raised to a fractional power near Speech and Signal Processing (ICASSP96), Atlanta, GA, May 1996.
9
unity. For these configurations, the directivity index for a M. Hawkes and A. Nehorai, ‘‘Acoustic vector-sensor beamforming and
linear pressure-sensing array and any type of linear velocity- capon direction estimation,’’ IEEE Trans. Signal Process. 46, 2291–2304
sensing array steered to broadside, would be approximately 共1998兲.
10
A. V. Pesotskii and M. D. Smaryshev, ‘‘Comparative evaluation of the
equivalent. However, vector sensors may be able to provide
efficiency of receiving arrays consisting of combination monopole-dipole
additional gains 共on an ideal anechoic baffle兲, as suggested in receivers in a free field and near a plane barrier,’’ Sov. Phys. Acoust. 35,
Figs. 3, 4, and 5 for other array steerings. Pressure-sensing 289–291 共1989兲.
directivity remains constant, whereas the multiaxis velocity 11
L. Camp, Underwater Acoustics 共Wiley-Interscience, New York, 1970兲.
12
sensor gains vary with array steering and these gains are due G. L. D’Spain, F. H. Fisher, W. S. Hodgkiss, J. C. Nickles, G. L. Ed-

Downloaded from http://pubs.aip.org/asa/jasa/article-pdf/110/1/324/10681684/324_1_online.pdf


to measuring the additional vector components of acoustic monds, and R. A. Harriss, ‘‘The Vertical DIFAR Array,’’ Proceedings of
the Deployable Surveillance Workshop, April 1992.
particle velocity. 13
G. L. D’Spain, W. S. Hodgkiss, and G. L. Edmonds, ‘‘Energetics of the
deep ocean’s infrasonic sound field,’’ J. Acoust. Soc. Am. 89, 1134–1158
1
G. L. D’Spain, W. S. Hodgkiss, G. L. Edmonds, J. C. Nickles, F. H. 共1990兲.
Fisher, and R. A. Harriss, ‘‘Initial Analysis of the Data from the Vertical 14
M. C. Junger and D. Feit, Sound, Structures, and Their Interaction 共MIT
DIFAR Array,’’ Proceedings of Mastering the Oceans Through Technol- Press, Cambridge, MA, 1986兲.
ogy, 共OCEANS 92兲, Newport, Rhode Island, October 26–29, 1992. 15
2 A. H. Nuttall and B. A. Cray, ‘‘Efficient Calculation of Directivity Indices
V. A. Shchurov, ‘‘Modern State and Prospects for Use of Underwater
for Certain Three-Dimensional Arrays,’’ NUWC-NPT Technical Report
Acoustic Intensity Measurements,’’ Scientific Information Services Pre-
print, Pacific Oceanological Institute, Far Eastern Branch of Russian 11,129, Naval Undersea Warfare Center Detachment, New London, CT,
Academy of Sciences, 1998. 26 July 1996.
16
3
G. L. D’Spain, W. S. Hodgkiss, and G. L. Edmonds, ‘‘The simultaneous A. H. Nuttall and B. A. Cray, ‘‘Approximations to Directivity for Linear,
measurement of infrasonic acoustic particle velocity and acoustic pressure Planar, and Volumetric Apertures and Arrays,’’ NUWC-NPT Technical
in the ocean by freely drifting swallow floats,’’ IEEE J. Ocean Eng. 16, Report 10,798, Naval Undersea Warfare Center Division, Newport, RI, 25
195–207 共1991兲. July 1997.

J. Acoust. Soc. Am., Vol. 110, No. 1, July 2001 B. A. Cray and A. H. Nuttall: Directivity factors 331

You might also like