J.C. Piquette, S.E. Forsythe - Low-Frequency Echo-Reduction and Insertion-Loss Measurements (2001)

You might also like

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

Low-frequency echo-reduction and insertion-loss measurements

from small passive-material samples under ocean


environmental temperatures and hydrostatic pressures
Jean C. Piquette and Stephen E. Forsythe
Naval Undersea Warfare Center, Division Newport, 1176 Howell Street, Newport, Rhode Island 02840
共Received 26 April 2001; accepted for publication 18 July 2001兲
System L is a horizontal tube designed for acoustical testing of underwater materials and devices,
and is part of the Low Frequency Facility of the Naval Undersea Warfare Center in Newport, Rhode
Island. The tube contains a fill fluid that is composed of a propylene glycol/water mixture. This
system is capable of achieving test temperatures in the range of ⫺3 to 40 deg Centigrade, and
hydrostatic test pressures in the range 40 to 68 950 kPa. A unidirectional traveling wave can be
established within the tube over frequencies of 100 to 1750 Hz. Described here is a technique for
measuring the 共normal-incidence兲 echo reduction and insertion loss of small passive-material
samples that approximately fill the tube diameter of 38 cm. 共Presented also is a waveguide model
that corrects the measurements when the sample fills the tube diameter incompletely.兲 The validity
of the system L measurements was established by comparison with measurements acquired in a
large acoustic pressure-test vessel using a relatively large panel of a candidate material, a subsample
of which was subsequently evaluated in system L. The first step in effecting the comparison was to
least-squares fit the data acquired from the large panel to a causal material model. The material
model was used to extrapolate the panel measurements into the frequency range of system L. The
extrapolations show good agreement with the direct measurements acquired in system L.
关DOI: 10.1121/1.1402115兴
PACS numbers: 43.58.Vb, 43.20.Fn, 43.20.Ye, 43.58.Bh 关SLE兴

I. INTRODUCTION only the echo reduction, not the insertion loss. Such methods
also often involve backing the sample with some other ma-
The normal-incidence echo reduction and insertion loss terial in order to avoid the influence of unwanted reflections
are quantities that are often needed to characterize the per- from the tube end. Hence, the echo reduction that is mea-
formance of passive acoustical materials such as absorbers, sured in this manner is characteristic not only of the sample
decouplers, and acoustic windows. These are often deter- under test, but also of the backing material. As frequency is
mined in a panel test.1 Unfortunately as test frequency is lowered significantly the methods employed in pulse tubes
lowered, panel size must be increased in order to avoid the clearly become inapplicable, especially once the wavelength
disturbing influence of waves originating at the panel edges. in the tube fill fluid exceeds the tube length.
For example, conventional panel measurements, as con- An alternative kind of rigid-walled-tube-based measure-
ducted in the Acoustic Pressure Tank Facility 共APTF兲 of the ment is considered here. System L, one of a group of rigid-
Naval Undersea Warfare Center, Newport, RI 共NUWC兲 are walled tubes available for acoustic testing 共under ocean en-
typically limited to frequencies above 10 kHz for samples of vironmental conditions of temperature and hydrostatic
76⫻76-cm (30⫻30-in.) lateral dimensions. The lower- pressure兲 at NUWC, has the capability to create a unidirec-
frequency limit for panels of this size is reduced to about tional traveling wave in steady state. 共This is accomplished
3400 Hz if the ONION method is used.2 However, frequen- through the use of active terminations of the tube, thus simu-
cies even lower than this are often of interest. Panel tests lating a tube of infinite length.兲 Hence, it is possible to place
generally must be excluded as test frequency is lowered sig- a sample of interest near the tube’s center and to test the
nificantly, owing to the high cost of fabricating large samples sample without needing to back the sample in some special
and the inability of existing test facilities to accommodate way. If desired, the sample can be affixed to a backing plate,
large samples. For example, APTF is limited to panels less or the intrinsic properties of a nonbacked sample can be
than 1.8 m on a side in order that the panel can fit through tested instead.
the facility access port. Even for a panel of this size, tested A description of system L is given in Sec. II. The mea-
using conventional methods, the low-frequency limit for surement concept for using system L to determine low-
echo reduction is about 2 kHz and that for insertion loss is frequency echo reduction and insertion loss is described in
about 3 kHz. The ONION method would only extend the Sec. III. Measurements both with and without a sample
low-frequency cutoff to about 1.5 kHz for both measure- present in the tube are required to effect the present tech-
ments. nique. The required ‘‘without-sample’’ measurements, and
An obvious way to attain lower frequencies without in- their use to adjust nominal hydrophone sensitivities and po-
creasing sample size is to conduct tests in a rigid-walled sitions as well as the speed of sound of the tube fill fluid, are
tube, such as an impedance or pulse tube.3 However, such discussed in Sec. IV. The required ‘‘with-sample’’ measure-
methods, which typically utilize pulsed sound, usually obtain ments are discussed in Sec. V. Also described in Sec. V are

1998 J. Acoust. Soc. Am. 110 (4), October 2001 0001-4966/2001/110(4)/1998/9/$18.00

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
an array of six hydrophones is mounted along the upper in-
terior length of the tube to aid in the production of a travel-
ing wave. The first such hydrophone is located at an offset
distance of 15 cm from the projector array closest to the tube
cover 共shown at the right side of Fig. 1兲. Denoting that hy-
drophone’s position as the zero, or reference, position, the
remaining hydrophones are situated at distances of 32, 64,
104, 136, and 168 cm from the reference hydrophone. The
last such hydrophone is thus also located about 15 cm from
the proximate projector array. By monitoring the amplitudes
and phases of the waves received by each of these six hy-
drophones, more than enough information is acquired to de-
termine the drives that must be applied to each of the two
projector arrays in order to establish a unidirectional travel-
ing wave within system L. 共The U.S. Navy’s first use of
active cancellation methods, performed at what was then the
FIG. 1. Schematic representation of the system L tube. The sample under Underwater Sound Reference Laboratory in Orlando,
test 共not shown兲 is positioned near the tube center, and must approximately Florida, is described in Ref. 4. The current version of the
fill the 38-cm tube diameter. Also not shown is a series of hydrophones hardware and software in system L is a modern computer-
situated along the upper interior length of the tube. The ‘‘first’’ such hydro-
phone is located at an offset distance of about 15 cm from the projector based descendant of the techniques described in this early
array nearest the removable tube cover, shown on the right side of the figure. reference.兲
This hydrophone offset was selected to avoid the near-field region of the
proximate projector array. This first hydrophone is taken to be the ‘‘refer- III. MEASUREMENT CONCEPT
ence’’ hydrophone. The five other hydrophones are nominally located at 32-,
64-, 104-, 136-, and 168-cm distances from the reference hydrophone. A sample is mounted approximately midway between
共These values of the nominal hydrophone positions are adjusted, by amounts
the two projector arrays of the system L tube 共note again Fig.
of less than 1 cm, using a least-squares procedure described in the text.兲
Positioning the sample near the tube center results in three of these six 1兲. The sample is oriented such that a normal to its surface is
hydrophones being situated to either side of the sample. The measurements parallel to, and approximately coincident with, the tube cen-
obtained from one such triad of hydrophones 共in the reflection region terline. 共If the sample does not entirely fill the cross section
shown兲 are used to decompose analytically the components of the standing
wave that appears in this area of the tube when a sample is present. Mea-
of the tube, it is necessary to correct the measurements. A
surements obtained from the triad of hydrophones on the other side of the suitable waveguide model is described in the Appendix.兲 The
sample 共in the transmission region shown兲 are used both to help establish a two projectors in system L are driven in a manner that estab-
unidirectional traveling wave in that region, and to determine the amplitude lishes acoustic waves within the tube that simulate those that
of the through-the-sample transmitted wave.
are present in a standard normal-incidence panel test.1 In
such a test on one side of the panel there appear two waves,
comparison measurements acquired from a relatively large one of which approaches the sample 共and which is called the
test panel, evaluated in APTF. A description of a causal ma- ‘‘incident’’ wave兲, and the other of which recedes from the
terial model used to extrapolate the APTF measurements into sample 共and which is called the ‘‘reflected’’ wave兲. On the
the system-L frequency range is also given. A summary and opposite side of the sample there appears only one wave,
the conclusions are provided in Sec. VI. Finally, the Appen- which recedes from the sample 共and which is called the
dix provides the details of a simple waveguide model that is ‘‘through-the-panel transmitted wave’’兲. The two waves ap-
used to correct the system L measurements for samples that pearing in the reflection region of system L combine to cre-
fail to fill completely the cross section of the tube. ate a standing wave; the single wave appearing in the trans-
mission region is a unidirectional traveling wave.
The reflected wave unquestionably loads the projector
II. DESCRIPTION OF SYSTEM L
array that it impinges upon, thereby modifying its output.
System L 共see Fig. 1兲 is a horizontal tube designed for However, there are three hydrophones situated between that
acoustical testing of underwater materials and devices, and is projector and the sample, and both the amplitudes and phases
part of the Low Frequency Facility of the Naval Undersea of the received signals are recorded at each of these three
Warfare Center in Newport, RI. The tube is of 38 cm inner hydrophones. Hence, there is more than sufficient informa-
diameter, and is 243 cm in length. It contains a fill fluid that tion available to decompose analytically the standing wave
is composed of a propylene glycol 共antifreeze兲/water mix- into its individual, oppositely directed, components. Once
ture. 共The antifreeze is added to the water in a ratio of 1 part the amplitudes of these standing-wave components are
antifreeze to 3 parts water. The resulting mixture has a speed known, they can be used to compute the sample’s pressure
of sound of about 1600 m/s, which is slightly greater than reflection coefficient, and hence the sample’s echo reduction
that of pure water.兲 Pressurization apparatus allows hydro- 共taken here to be ⫺20 times the logarithm to the base 10 of
static test pressures as high as 68 950 kPa, thus simulating an the magnitude of the pressure reflection coefficient兲.
ocean depth of up to about 6.9 km. There are also three hydrophones situated at 共approxi-
Each end of the tube is fitted with a projector array mately兲 known positions in the transmission region. Ampli-
consisting of seven pressure-compensated transducers. Also, tude and phase are also accurately measured at these hydro-

J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement 1999

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
phones. The amplitude of the unidirectional traveling wave is driven to steady state while the voltage driving the other
appearing there can thus be combined with the incident-wave projector is held at zero. Of course, driving the system in this
amplitude determined from the reflection-region measure- manner creates a standing wave within the tube.

冉冊
ments to compute the sample’s pressure transmission coeffi- Consider the following equation that defines the
cient, and hence its insertion loss. 共This is taken here to be

冉冊
system-L transfer matrix:
⫺20 times the logarithm to the base 10 of the magnitude of
a 11 a 12
the pressure transmission coefficient.兲 m1
Pulsed sound would not be an adequate method to a 21 a 22 m2
achieve the measurements of interest. With the sample lo-
cated near the tube center as described, the distance between
the sample and the projector array in the transmission region
a 31
a 41
a 32
a 42
⫻ 冉 冊
d1
d2

m3
m4
. 共1兲

a 51 a 52 m5
is less than 99 cm. Even neglecting turn-on transients of the
source共s兲, and response transients of the sample, this avail- m6
a 61 a 62
able distance would be only one cycle in the fill fluid at
about 1500 Hz. It is clear that sufficient data could not be Here, the a matrix is the 共initially unknown兲 transfer matrix;
accurately acquired to characterize the sample to frequencies the d matrix is the ‘‘drive’’ matrix, i.e., the complex drive
as low as 100 Hz if it were not possible to create a traveling voltages applied to each of the two projectors; and the m
wave. matrix is the matrix of measured complex responses at each
of the six hydrophones in system L. 共It should be understood
IV. ‘‘NO-SAMPLE’’ EXPERIMENTS AND ASSOCIATED
that the elements of the a matrix must be determined not
LEAST-SQUARES ADJUSTMENTS only for the case in which no sample is present, but must also
be determined afresh for each sample and each condition of
A. Background temperature, hydrostatic pressure, and frequency of interest.兲
Even relatively modest errors in sensitivity 共of a few When measurements are acquired as described above,
tenths of a dB兲, or in the ‘‘known’’ hydrophone locations 共of each of the 12 elements of the transfer matrix can be deter-
a few tenths of a cm兲, can produce several dB of error in the mined, six at a time. For example, when one projector is
determinations of the amplitudes of the three wave fields driven 共e.g., the projector that corresponds to element d 1 of
共incident, reflected, transmitted兲 of interest. In order to cir- the d matrix兲, while the other projector is shorted out 共i.e.,
cumvent this problem, least-squares adjustments of the hy- d 2 ⫽0兲, notice that Eq. 共1兲 reduces to the six equations
drophone sensitivities and positions, and also of the speed of a 11d 1 ⫽m 1 ; a 21d 1 ⫽m 2 ; a 31d 1 ⫽m 3 .
sound of the tube fill fluid, are required.
共2兲
The notion behind the least-squares adjustments is the a 41d 1 ⫽m 4 ; a 51d 1 ⫽m 5 ; and a 61d 1 ⫽m 6 .
determination of those hydrophone sensitivities and posi-
Since m 1 through m 6 are the measured hydrophone re-
tions, and that speed of sound of the fill fluid, that best sat-
sponses, and since the applied drive d 1 is obviously also
isfy the requirements for echo reduction and insertion loss
known, it is clear that Eq. 共2兲 can be used to determine the
when no sample is present. Naturally, with no sample present
unknown 共complex兲 elements a 11 through a 61 of the transfer
one ideally should find a null reflected wave 共infinite echo
matrix. In a similar manner, the remaining six unknown ele-
reduction兲 and should find a through-the-sample transmitted
ments of the transfer matrix are determined when the system
wave of amplitude equal to the incident wave 共zero insertion
is driven under the conditions d 1 ⫽0 and d 2 ⫽0. Once all the
loss兲.
elements of the a matrix have been determined by this pro-
Although six hydrophones are present in system L, and
cess, projector drives that could be used to create a unidirec-
although all six are typically used in the creation of a trav-
tional traveling wave in the system L tube can be computed.
eling wave within the tube when it is used in its standard
In principle, a traveling wave can be built within the
mode of transducer calibration, in the present measurement
tube based on the responses of any two 共or more兲 of the six
only three hydrophones are available for this purpose. Once
hydrophones. To do this, a submatrix equation, consisting of
a sample has been placed inside the tube near its center, a
only those rows of Eq. 共1兲 that are associated with the hy-
standing wave will be created in the reflection region. Three
drophones that are to be used to build the traveling wave, is
of the six hydrophones are located in this region, and thus
created. Unlike the situation in which the elements of the
are unavailable for use in building a traveling wave. The
transfer matrix were initially determined, in the present case
‘‘no-sample’’ measurements in the system L tube help to es-
the d elements of the submatrix equation are unknown, but
tablish the fact that a good traveling wave can indeed be built
the a elements are known. Also, the m elements of this sub-
using only these three hydrophones.
matrix equation are ‘‘seeded’’ with the desired hydrophone
responses, which correspond to the expected hydrophone re-
B. Determination of the ‘‘no-sample’’ system transfer sponses to a unidirectional traveling wave. These desired
matrix
values are determined by evaluating the unidirectional
Two no-sample data sets are required. The two no- traveling-wave function exp关ikx兴 at the 共nominal兲 ‘‘known’’
sample data sets involve measurements of the complex volt- hydrophone positions x. 共The implied time dependence
age responses at each of the six hydrophones for each of two exp关⫺i␻t兴 is suppressed.兲 The resulting submatrix equation
drive conditions. In each of these tests, one of the projectors can then be solved for the unknown elements of the d matrix.

2000 J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
Since the submatrix equation has only two unknown quanti- to j⫽5 are sums over hydrophones. 共One of the six hydro-
ties, viz., d 1 and d 2 , but three equations result from the val- phones in the system is taken to be a reference hydrophone,
ues at each of the three available hydrophones in the trans- whose position and calibration are taken to be known ex-
mission region, the problem is solved using linear least- actly.兲 Finally, the quantities W 1 through W 4 are dimension-
squares methods. The problem must be solved at each less weighting factors, which determine the extent to which
measurement frequency and hydrostatic pressure of interest. the associated sums affect the cost function, and whose val-
The best-fit values of d 1 and d 2 obtained at each frequency ues are at our disposal. The following numerical values were
can then be used to evaluate the ‘‘forward’’ problem that found to give good results:
would be observed if these two computed drives were in fact
applied to the projectors in the system L tube. By substituting W 1 ⫽0.2; W 2 ⫽0.1; W 3 ⫽0.1 W 4 ⫽0.1. 共4兲
these computed drives into Eq. 共1兲, together with the now-
This scheme for determining corrections was found to pro-
known a-matrix elements for the no-sample case, the re-
duce reasonable changes to the initial positions, calibrations,
sponses at all six hydrophones can then be computed.
and fill-fluid sound speed. For example, the results of one
typical calculation produced the following set of values:
C. Least-squares adjustments of hydrophone
sensitivities and positions, and of the sound speed xc 1 ⫽0.998 35; xc 2 ⫽1.000 40; xc 3 ⫽1.003 61;
of the fill fluid
xc 4 ⫽0.994 764; xc 5 ⫽1.001 14;
To effect the required least-squares adjustments, a 共5兲
‘‘cost’’ function must be defined. There is considerable lati- ec 1 ⫽0.988 471; ec 2 ⫽0.997 562; ec 3 ⫽1.006 29;
tude available in the definition of this function. It is impor-
tant that the cost function be defined in a way that ensures ec 4 ⫽1.023 04; ec 5 ⫽1.010 18; cc⫽0.999 278.
that the adjustments remain relatively small. The principle of
It can be seen that the correction factors do not differ much
‘‘fitting-to-goal’’ was also adopted in designing this function.
from unity, and are distributed mostly randomly, as would be
One such condition is that the amplitude of the reflected
expected of true measurement errors in the adjusted quanti-
wave appearing in the reflection region of the tube when no
ties. In this way, it is found that the largest absolute change
sample is present should be as small as possible. The second
in hydrophone position is slightly more than 0.7 cm. The
condition is that the amplitude of the wave appearing in the
largest change in hydrophone calibration is found to be
transmission region of the tube when no sample is present
slightly less than 0.2 dB. The change in the speed of sound of
should be as close as possible to the amplitude of the inci-
the hydrophone fill fluid is less than 1.17 m/s.
dent wave. An auxiliary condition imposes the requirement
The results obtained by applying this procedure to the
that the adjustments be small relative to the initial values of
data acquired with no sample in the system-L tube, for one
the quantities being adjusted, since it is not expected that the
typical case, are shown in Figs. 2共a兲 and 共b兲 and Fig. 3. In
initial hydrophone sensitivities and positions, as well as the
Fig. 2共a兲 the effective echo reduction seen at the three hy-
speed of sound of the tube fill fluid, would ever be grossly in
drophones used to build the traveling wave in the through-
error.
the-sample transmitted-wave region is shown. It is seen that
The specific form of the adopted cost function is


typical echo reductions in this region are of the order of 70
n 5
dB, except at obvious tube resonances. These extremely high
f⫽ 兺
i⫽1
兩 R i 兩 2 ⫹ 共 1⫺ 兩 T i 兩 兲 2 ⫹W 1 兺 共 1⫺xc j 兲 2
j⫽1
echo reductions are to be expected, since the traveling-wave
drives have been computed using the transfer function as

⫹W 2 冋兺 5

j⫽1

共 1⫺xc j 兲 ⫹W 3
2 5

兺 共 1⫺ec j 兲 2
j⫽1
determined at these particular hydrophones.
In Fig. 2共b兲 the effective echo reductions computed at
the other triad of hydrophones, that is, the three hydrophones

⫹W 4 共 1⫺cc 兲 2 . 册 共3兲
that were not used to establish the traveling wave, are shown.
The associated pressure reflection coefficients are those used
in the computation of the R i quantities of Eq. 共3兲. The typical
The outermost sum of Eq. 共3兲 is taken over the test frequen- value seen in Fig. 2共b兲 of the associated echo reduction is
cies of interest. 共This sum excludes certain frequencies at about 40 dB. 共When no corrections are applied to the hydro-
which there are obvious resonances in the system.兲 The phone calibrations and positions, and to the sound speed of
quantities 兩 R i 兩 are the amplitudes of the effective pressure the tube fill fluid, the typical echo reductions seen at these
reflection coefficients. The quantities 兩 T i 兩 are the amplitudes hydrophones are about 25 dB. Thus, the small corrections to
of the effective pressure transmission coefficients. The quan- these quantities have a significant impact upon the reduction
tities xc j are hydrophone-position ‘‘correction factors,’’ of measurement error.兲 The typical value of 40 dB for the
taken initially to be unity. When multiplied times the initial echo reduction suggests that a reflected ‘‘error wave’’ of am-
‘‘known’’ hydrophone positions, these factors produce ‘‘cor- plitude about 1% of that of the incident wave is present in
rected’’ hydrophone positions. The quantities ec i are hydro- the reflection region of the tube. If an error wave of this same
phone calibration correction factors. These are also taken amplitude is also present when a sample of 20-dB intrinsic
initially to be unity. The quantity cc is the tube fill-fluid echo reduction is being measured in system L using the
sound-speed correction factor. The sums that run from j⫽1 present method, an error of slightly less than 1 dB in the

J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement 2001

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
sample is present. This is the insertion loss associated with
the T i quantities of Eq. 共3兲. The typical insertion loss seen in
this figure is of the order of magnitude of about 0.1 dB. This
again suggests that a through-the-sample transmitted error
wave of amplitude about 1% of that of the incident wave is
present in the transmission region. The respective insertion-
loss errors, in samples that are characterized by 20- to 30-dB
intrinsic insertion loss, are again in the 1- to 3-dB range,
respectively.

V. ‘‘WITH-SAMPLE’’ EXPERIMENTS
A. Acoustic Pressure Tank Facility „APTF…
measurements
Before acquiring any data with samples in system L, a
relatively large sample was first tested in APTF. This sample
共a decoupler兲 was measured in order to obtain baseline re-
sults that could be used to judge the quality of the system L
measurements. The dimensions of the test panel were ap-
proximately 82⫻92 cm (32 41 in.⫻36 41 in.), and the sample
was of approximately 5-cm 共2-in.兲 thickness. This relatively
large panel size enabled the use of standard measurement
FIG. 2. 共a兲 Effective echo reduction, obtained at the three hydrophones techniques to approximately determine sample response in
located in the transmission region shown in Fig. 1, when no sample is
the frequency interval of interest. Although highly accurate
present in the tube. The very large numerical values are obtained simply
because the LS determination of the projector drives is only a slightly over- measurements at the frequencies of interest in system L
determined problem, with the two required drives being determined from the could not be conducted in APTF, the measurements could be
three hydrophones available in the transmission region using a least-squares extrapolated into the desired frequency regime. This was
process. 共TW⫽traveling wave; ER⫽echo reduction兲. 共b兲 Effective echo re-
possible because the sample of interest consisted of a single-
duction, obtained at the three hydrophones located in the reflection region
shown in Fig. 1, when no sample is present in the tube. It can be seen that layer material; edge diffraction was found not to be too
the typical value obtained for the echo reduction is about 40 dB, ignoring strong; and new causal theoretical expressions for the sound
obvious tube resonances. If no adjustments of the hydrophone sensitivities speed and loss of the material could be employed in the
and positions, as well as the speed of sound of the tube fill fluid were
performed 共as described in the text兲, the typical echo reduction obtained at
extrapolations. We developed these theoretical expressions
these hydrophones for this case is about 25 dB. 共TW⫽traveling wave; using the appropriate Hilbert transform implementation of
ER⫽echo reduction兲. the causality principle.5 The transforms were carried out ana-
lytically. The results of the analytical calculations were also
measurement can be expected. For a sample of 30-dB intrin- verified numerically.
sic echo reduction, the measurement error could be some- The model we used to derive a causal relationship be-
what greater than 3 dB. tween the sound speed and loss of the sample, based on the
In Fig. 3 is shown the effective insertion loss when no use of Kramers–Kronig relations,5 is of the form
␻ ␻
⫽ ⫹ 关 ␣ 共 ␻ 兲兴 , 共6兲
c phase共 ␻ 兲 c ⬁
where ␻ denotes the angular frequency; c phase( ␻ ) and ␣共␻兲
are the phase speed and loss of the sample material expressed
as functions of ␻, respectively; c ⬁ is the phase speed at
infinite frequency; and the 关 兴 notation denotes the Hilbert
transform. The numerical value of c ⬁ is to be determined by
least-squares fitting.
In order that Eq. 共6兲 yield reasonable results without
requiring the introduction of specialized methods of
analysis,6 it is helpful to use only functions for ␣共␻兲 that
remain finite as ␻ →⬁. We chose to introduce a loss function
FIG. 3. Effective insertion loss, obtained in the transmission region shown
in Fig. 1, when no sample is present. This insertion loss is computed by that is approximately linear with frequency, at least over a
combining the results obtained from the triad of hydrophones located in this major segment of the frequency interval of our experiments,
region with those obtained from the triad in the reflection region, required to and to use a filtering function to eliminate the divergence for
determine the incident-wave amplitude. It is seen that the typical insertion large frequencies. The loss function we chose to introduce is
loss obtained is about 0.1 dB, except at obvious tube resonances. If no
adjustments of the hydrophone sensitivities and positions, and the fill-fluid a 1兩 ␻ / ␻ 0兩
sound speed, were performed, the typical insertion loss obtained is about 1 ␣共 ␻ 兲⫽ , 共7兲
dB. 共TW⫽traveling wave; IL⫽insertion loss兲. 1⫹b 共 ␻ / ␻ 0 兲 4

2002 J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
where a 1 is a constant that measures the ‘‘strength’’ of the
loss, and which is to be determined by least-squares fitting.
关Note that the function defined in Eq. 共7兲 is approximately
linear for frequencies below ␻ 0 .兴 The constants b and ␻ 0 are
parameters that describe the filter characteristics, and their
values are at our disposal. The choice of a filter with a
fourth-order pole is somewhat arbitrary, but this ensures rela-
tively rapid convergence of the Hilbert-transform integral.
Fortunately, the Hilbert transform of a( ␻ ) as given by Eq.
共7兲 can be carried out analytically in a straightforward fash-
ion. The result is

关 ␣ 共 ␻ 兲兴 ⫽⫺

a 1 ␻ 0 ␻ ␲ 冑b ␻ 2 ⫹ ␻ 20 ln 冉 冊冊
b␻4
␻ 40
. 共8兲
2 ␲ 共 b ␻ 4 ⫹ ␻ 40 兲

Here ln denotes the natural logarithm.


Panel measurements, as acquired in APTF, directly mea-
sure the amplitudes and phases of the incident, reflected, and
through-the-panel transmitted waves. Unfortunately, it is not
possible to avoid the contaminating influence of the wave
originating at the sample edge for the current sample
(32 14 in.⫻36 41 in.) at the frequencies of interest. Thus, even if
the causal model described above were an exact description
1
of sample performance, exact agreement between theory and FIG. 4. Insertion loss and echo reduction of the 82⫻92-cm (32 4 -in.
1
measurement would not be expected. However, since the ⫻36 4 -in.) panel at a hydrostatic test pressure of 689 kPa 共100 psi兲. Dots are
model involves only two adjustable parameters 共c ⬁ and a 1 兲, data; solid lines are from a least-squares fit of a causal panel model to the
measurements. Two parameters were adjusted in the fits. 共a兲 Insertion loss.
it was hoped that least-squares fitting would help to discrimi-
共b兲 Echo reduction.
nate against edge effects, essentially treating these effects as
a form of 共coherent兲 noise. Since many frequencies are in-
volved in the fitting process but only two constants are being
adjusted, it was felt that the combined effects of the many model parameters were found to be c ⬁ ⫽3.5⫻104 cm/s and
test frequencies involved, covering the wide frequency inter- a 1 ⫽2.1 cm⫺1. At the 500 psi hydrostatic test pressure,
val of 250 to 15 000 Hz, might reduce the influence of the shown in Figs. 5共a兲 and 共b兲, the two fitted model parameters
coherent noise. were found to be c ⬁ ⫽3.9⫻104 cm/s and a 1 ⫽1.1 cm⫺1.
The causal dispersion model of Eqs. 共6兲–共8兲 was incor- These values, and their changes with hydrostatic pressure,
porated into a single-layer reflection/transmission model of are reasonable for the sample under test. At 100 psi, the
the sample, and the resulting model was fitted to the APTF sound speed in the material, which can be computed from the
data acquired from the panel to determine c ⬁ and a 1 at each causal model of Eqs. 共6兲–共8兲 using the fitted constants, is
hydrostatic pressure. In Figs. 4共a兲 and 共b兲 and Figs. 5共a兲 and about 2.7⫻104 cm/s, while the associated loss factor is about
共b兲 are shown the results of the measurements and of the 0.53 cm⫺1. At 500 psi, the computed sound speed in the
fitting process for two of the hydrostatic pressures that were material is about 3.3⫻104 cm/s, while the loss factor is about
tested. Data are shown as dots, while the theoretical results, 0.28 cm⫺1.
based on least-squares fitting of the causal model, are shown B. Sample measurements in system L
as solid lines. The solid lines are extended down to 100 Hz in
order to cover the low-frequency limit to be tested in system An approximately circular subsample of 38-cm 共15-in.兲
L. 共The low-frequency cutoff for the insertion-loss measure- diameter was next cut from one corner of the 32 41 -in.
ments in APTF was 250 Hz, while the low-frequency limit of ⫻36 14 -in. decoupler panel for testing in system L. A corner
the echo-reduction measurements was 2000 Hz. These low- cut was used to permit retesting of the panel in APTF, should
frequency limits are well below those usually applied in that that have proved necessary. The subsample was then
facility.兲 As can be seen, the theory agrees reasonably well mounted in system L, and suitable measurements acquired.
with the measurements over much of the data, although the Recall that the basic concept behind the present mea-
effects of edge diffraction are clearly present, especially in surement method involves driving the two transducers to cre-
Fig. 5共b兲. ate a traveling wave on one side of the sample and a standing
In fitting the APTF measurements, the filter parameters wave on the other. However, there is no need in practice to
of the causal model were set such that ␻ 0 actually produce such waves within the system when a
⫽2 ␲ 40 000 radians/s and b⫽0.5 共dimensionless兲 in all sample is present. The same kinds of standing-wave mea-
cases. At the 100 psi hydrostatic test pressure used in the surements used to determine the system transfer function
measurements presented in Figs. 4共a兲 and 共b兲, the two fitted when no sample is present in the tube can also be used when

J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement 2003

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
1
FIG. 5. Insertion loss and echo reduction of the large panel (32 4 -in.
1
⫻36 4 -in.) at a hydrostatic test pressure of 3450 kPa 共500 psi兲. Dots are
data; solid lines from a least-squares fit of a causal panel model to the
measurements. Two parameters were adjusted in the fits. 共a兲 Insertion loss.
共b兲 Echo reduction.
FIG. 6. Insertion loss and echo reduction, at a hydrostatic test pressure of
689 kPa 共100 psi兲, of an approximately circular, 38-cm diameter, subsample
1 1
a sample is present as well. That is, when a sample is present cut from the 32 4 -in.⫻36 4 -in. panel, as tested using the new procedure in
system L. Dots are data; solid lines are extrapolations of the causal model
a transfer-matrix equation of the form of Eq. 共1兲 still applies fitted to the large-panel measurements shown in Fig. 4. 共a兲 Insertion loss. 共b兲
to the measurements. Of course, the elements of the transfer Echo reduction.
matrix will obviously be different depending upon whether
or not a sample is present, and will obviously be different for
each sample and each particular test circumstance. Moreover, VI. SUMMARY AND CONCLUSIONS
the same technique used to determine the projector drives
that would produce a unidirectional traveling wave in the A technique for measuring the echo reduction and the
transmission region when no sample is present can again be insertion loss of small passive-material samples in NUWC’s
used to determine the drives needed to establish such a wave system L was described. The method allows for the measure-
even when a sample is present. Once the measurements that ments to be acquired over the full range of ocean environ-
determine the transfer matrix elements for the with-sample mental conditions in the frequency range of 100 to 1750 Hz.
case have been acquired, there is no need to actually estab- ‘‘No-sample’’ measurements are acquired in system L and
lish these waves in the tube. The amplitudes of the waves are used to adjust the hydrophone sensitivities and positions.
that would appear in each half of the tube can be determined The speed of sound of the tube fill fluid 共a propylene glycol/
instead from the with-sample transfer matrix, and the ampli- water mixture兲 is also adjusted. With-sample measurements
tudes of these waves can be combined in the usual way to are obtained and are used to determine the elements of the
determine sample echo reduction and insertion loss. transfer matrix of the tube/sample system. 共The corrected
The results obtained when the decoupler subsample was quantities deduced from the no-sample tests are held fixed in
evaluated in this way are shown in Figs. 6共a兲 and 共b兲 and determining the elements of the with-sample transfer ma-
Figs. 7共a兲 and 共b兲. The dots once again are the measured trix.兲 This transfer matrix is then used to determine the echo
values, obtained by the present method, and the solid lines reduction and the insertion loss of the sample under test.
are theoretical values based on extrapolating the APTF mea- A causal material model was developed and applied to
surements. The two adjusted parameters determined at each measurements acquired from a relatively large sample panel
hydrostatic pressure tested in APTF were used in the theo- evaluated in NUWC’s APTF. The model was used to ex-
retical expressions of the causal model to compute the solid- trapolate the APTF measurements into the frequency range
line curves shown in the figures. As can be seen, agreement of system L. The extrapolations showed good agreement
between the measurements and the extrapolations is quite with the direct measurements acquired using the new
good. method.

2004 J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
FIG. 8. Boundary conditions and pertinent acoustic variables for a branch-
ing pipe.

共1兲 A is the relative surface area of the sample 共A⫽1 com-


pletely fills the pipe兲.
共2兲 L, R are the complex pressures of the left- and right-
moving waves. Here, we assume without loss of gener-
ality that an outgoing traveling wave field is established
at the right tube end and that the pressure of the incom-
ing wave at the left is unity.
共3兲 l 1 , l 2 , r 1 , r 2 are the complex pressures of the left- and
right-moving waves on either side of the sample in the
upper 共sample-only兲 branch of the pipe.
共4兲 l 3 , l 4 , r 3 , r 4 are the complex pressures of the left- and
right-moving waves on either side of the fluid sample in
the lower 共fluid-only兲 branch of the pipe.
FIG. 7. Insertion loss and echo reduction, at a hydrostatic test pressure of 共5兲 R s , T s , R s⬘ , T s⬘ are the reflection and transmission co-
3450 kPa 共500 psi兲, of an approximately circular, 38-cm diameter, sub-
1 1 efficients of the sample in the forward and reversed di-
sample cut from the 32 4 -in.⫻36 4 -in. panel, as tested using the new proce-
dure in system L. Dots are data; solid lines are extrapolations of a causal rection. For multilayered samples, R s and R s⬘ are in gen-
model fitted to the large-panel measurements shown in Fig. 5. 共a兲 Insertion eral different, but T s and T s⬘ are equal 共see Rudgers8兲.
loss. 共b兲 Echo reduction. 共6兲 T f is the transmission coefficient of the fluid 共lower
branch兲. This number has magnitude 1.0 but has a phase
The new technique is currently available for evaluation associated with the travel time through a layer of fluid
of customer samples in system L. equal to the thickness of the panel 共required to keep the
definition of transmission coefficient consistent as a
phase shift relative to the incident wave for both fluid
ACKNOWLEDGMENTS and sample兲. The reflection coefficient of the fluid
We gratefully acknowledge the efforts of Alfred sample is 0.
Garceau, who prepared the samples and acquired the data
The relevant continuity conditions are
from system L. We are also grateful to Dr. James J. Dlubac,
Naval Surface Warfare Center, Carderock Division, West Be- 共1兲 l 1 ⫽R s r 1 ⫹T s⬘ l 2 共transmission and reflection through the
thesda, Maryland, for providing material samples and for sample兲,
valuable technical consultation. The work was supported by 共2兲 r 2 ⫽T s r 1 ⫹R s⬘ l 2 共transmission and reflection through the
the Office of Naval Research 共Code 321兲. sample兲,
共3兲 l 4 ⫽T f l 3 共transmission through the sample thickness of
fluid兲,
APPENDIX: AREA CORRECTION FOR SAMPLES
共4兲 r 4 ⫽T f r 3 共transmission through the sample thickness of
THAT DO NOT FILL THE TUBE COMPLETELY
fluid兲,
In our experimental apparatus it is impossible to com- 共5兲 1⫹L⫽l 1 ⫹r 1 共continuity of pressure at the top left
pletely fill the system L tube with a material sample. This fork兲,
results in an experimental error since the effect of the smaller 共6兲 1⫹L⫽l 3 ⫹r 3 共continuity of pressure through the bottom
sample is based on a mix of the sample volume and the fluid left fork兲,
volume that is replacing the rest of the ideal sample’s outer 共7兲 R⫽l 2 ⫹r 2 共continuity of pressure at the top right fork兲,
perimeter. For low frequencies, following the analysis in 共8兲 R⫽l 4 ⫹r 4 共continuity of pressure at the bottom right
Kinsler and Frey7 of the effective impedance of a branching fork兲,
pipe, we can define the relevant acoustic quantities following 共9兲 1⫺L⫽A(r 1 ⫺l 1 )⫹(1⫺A)(r 3 ⫺l 3 ) 共continuity of vol-
Fig. 8: ume velocity at left fork兲,

J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement 2005

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40
TABLE I. Comparison of analytical and CHIEF results.

Frequency Analytical R s CHIEF R s Analytical T s CHIEF T s

500 ⫺0.034 154⫺0.005 671 7i ⫺0.034 106⫺0.006 822 5i 0.961 71⫹0.135 58i 0.961 93⫹0.135 56i
1000 ⫺0.073 975⫺0.027 533i ⫺0.075 313⫺0.028 784i 0.906 18⫹0.261 51i 0.905 07⫹0.261 33i
1500 ⫺0.113 55⫺0.066 766i ⫺0.117 98⫺0.068 186i 0.834 04⫹0.375 63i 0.832 08⫹0.0374 69i
2000 ⫺0.146 77⫺0.122 63i ⫺0.152 98⫺0.123 03i 0.746 61⫹0.475 01i 0.743 96⫹0.471 84i
2500 ⫺0.168 05⫺0.192 29i ⫺0.175 41⫺0.199 75i 0.646 41⫹0.556 52i 0.641 03⫹0.550 5i
3000 ⫺0.173 14⫺0.271 08i ⫺0.181 64⫺0.284 7i 0.537 81⫹0.617 52i 0.527 65⫹0.603 65i

共10兲 R⫽A(r 2 ⫺l 2 )⫹(1⫺A)(r 4 ⫺l 4 ) 共continuity of volume Here, T f is e ikd , the phase correction for traveling through
velocity at right fork兲. the thickness of fluid, d, at the sample edge; A is the area
correction ratio.
共A1兲
The above expressions for transmission and reflection
coefficients were tested using a model of the system L tube
The above system can be solved for R s and T s , eliminating
and a sample with realistic material properties based on our
all but the variables L, R, R s⬘ , T s⬘ , and A. At this point, it can
version of CHIEF 共combined Helmholtz integral equation
be assumed that two sets of measurements were done: first
formulation.兲9,10 The system L tube and sample were simu-
with the sample in its ‘‘normal’’ position, with measured re-
lated, the ends of the tube were driven by imposing unit
flected and transmitted amplitudes L 1 and R 1 and then with
surface velocity on the left and right ends of the tube, suc-
the sample in the ‘‘flipped’’ position 共turned 180 deg along
cessively, and the pressure field was measured at points in
the pipe axis to face the other way兲 with corresponding am-
the tube corresponding to actual hydrophone locations. The
plitudes L 2 and R 2 for the reflected and transmitted waves.
sample was defined as being 2-in. thick 共two 1-in. layers兲
共In fact, the flipped response can be derived by using the two
with the following material properties 共shear was assumed
with-sample measurements to calculate incident, transmitted,
negligible兲:
and reflected waves with the outgoing wave in the opposite
direction, so no more measurements are needed.兲 ␳ 1 ⫽1800 kg/m3, c l1 ⫽2600⫹520i m/s
The above-described reflection and transmission coeffi-
cients can then be written as 共 imaginary part represents loss兲 , c s1 ⫽0,

共1兲 R s ⫽R s (L 1 ,R 1 ,A,T s⬘ ), ␳ 2 ⫽800 kg/m3, c l2 ⫽1000⫹200i m/s, c s2 ⫽0.


共2兲 R s⬘ ⫽R s⬘ (L 2 ,R 2 ,A,T s ), The radius of the sample was taken to be 0.7071 times the
共A2兲
共3兲 T s ⫽T s (L 1 ,R 1 ,A,R s⬘ ), radius of the tube, giving an area ratio of 0.5. The results are
共4兲 T s⬘ ⫽T s⬘ (L 2 ,R 2 ,A,R s ), given in Table I. As can be seen by inspection, the agreement
is good at all frequencies, with deviations increasing with
where the R s ( ), etc., are the functional forms derived from
frequency as the assumptions about the plane wave unifor-
Eq. 共A1兲 above with the L 1 , L 2 , etc. substituted for the L
mity of the field begin to break down.
and R in the original solution.
Solving the above system gives
共 1 兲 R s ⫽⫺ 共 L 1 L 2 兲 ⫹ 共 R 1 ⫺T f 兲共 R 2 ⫺T f 兲 ⫹A 共共 R 1 ⫺T f 兲
R. J. Bobber, Underwater Electroacoustic Measurements 共U. S. Govern-
1

ment Printing Office, Washington, 1970兲, pp. 287–299.


⫻ 共 ⫺R 2 ⫹T f 兲 ⫹L 1 共 1⫹L 2 ⫺T 2f 兲兲 /D, 2
J. C. Piquette, ‘‘Transmission coefficient measurement and improved su-
blayer material property determination for thick underwater acoustic pan-
共 2 兲 R s⬘ ⫽⫺ 共 L1L2 兲 ⫹ 共 R 1 ⫺T f 兲共 R 2 ⫺T f 兲 ⫹A 共共 R 1 ⫺T f 兲 els: A generalization and improvement of the ONION method,’’ J. Acoust.
Soc. Am. 92, 468 – 477 共1992兲.
⫻ 共 ⫺R 2 ⫹T f 兲 ⫹L 2 共 1⫹L 1 ⫺T 2f 兲兲 /D, 3
See Ref. 1, pp. 305–310.
共A3兲 4
L. G. Beatty, ‘‘Acoustic impedance in a rigid-walled cylindrical sound
channel terminated at both ends with active transducers,’’ J. Acoust. Soc.
共 3 兲 T s ⫽A 共共 A⫺1 兲共 1⫹L 1 兲共 1⫹L 2 兲 T f ⫺R 1 共共 A⫺1 兲 R 2 T f Am. 36, 1081–1089 共1964兲.
5
See, for example, C. W. Horton, ‘‘Dispersion relationships in sediments
⫹T 2f 兲 ⫺1 兲 /D, and sea water,’’ J. Acoust. Soc. Am. 55, 547–549 共1974兲.
6
K. R. Waters, M. S. Hughes, J. Mobley, and J. G. Miller, ‘‘On the appli-
共 4 兲 T s⬘ ⫽A 共共 A⫺1 兲共 1⫹L 1 兲共 1⫹L 2 兲 T f cability of Kramers–Kronig relations for untrasonic attenuation obeying a
frequency power law,’’ J. Acoust. Soc. Am. 108, 556 –563 共2000兲.
⫺R 2 共共 A⫺1 兲 R 1 T f ⫹T 2f 兲 ⫺1 兲 /D,
7
L. E. Kinsler, A. R. Frey, A. B. Coppens, and J. V. Sanders, Fundamentals
of Acoustics, 3rd ed. 共Wiley, New York, 1982兲, pp. 235–237.
8
where A. J. Rudgers and C. A. Solvoid, ‘‘Apparatus-independent acoustical-
material characteristics obtained from panel-test measurements,’’ J.
D⫽L1L2⫹A 2 共 1⫹L1⫹L2⫹L1L2⫺R1R2 兲 Acoust. Soc. Am. 76, 926 –934 共1984兲.
9
G. W. Benthien, D. Barach, and D. Gillette, CHIEF User’s Manual 共Naval
⫹ 共 R1⫺T f 兲共 T f ⫺R2 兲 ⫺A 共 L1⫹L2⫹2L1L2 Oceans System Center, 1988兲.
10
H. Schenck, ‘‘Improved integral formulation for acoustic radiation prob-
⫺2R1R2⫹R1T f ⫹R2T f 兲 . lems,’’ J. Acoust. Soc. Am. 44, 41–58 共1968兲.

2006 J. Acoust. Soc. Am., Vol. 110, No. 4, October 2001 J. C. Piquette and S. E. Forsythe: Low-frequency echo and loss measurement

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 160.36.178.25 On: Sun, 21 Dec 2014 13:23:40

You might also like