Downloaded from [Link] on: ago.

28, 2024

Calculation of pressure fields from arbitrarily shaped, apodized, and excited

ultrasound transducers

Jensen, Jørgen Arendt; Svendsen, Niels Bruun

Published in:
I E E E Transactions on Ultrasonics, Ferroelectrics and Frequency Control

Link to article, DOI:

10.1109/58.139123

Publication date:
1992

Document Version
Publisher's PDF, also known as Version of record

Link back to DTU Orbit

Citation (APA):
Jensen, J. A., & Svendsen, N. B. (1992). Calculation of pressure fields from arbitrarily shaped, apodized, and
excited ultrasound transducers. I E E E Transactions on Ultrasonics, Ferroelectrics and Frequency Control,
39(2), 262-267. [Link]

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
 You may not further distribute the material or use it for any profit-making activity or commercial gain
 You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
 262 IEEE TRANSACTIONS ON ULTRASONICS. FERROELECTRICS. AND FREQUENCY CONTROL. VOL. 39. NO. 2, MARCH 1992

Calculation of Pressure Fields from Arbitrarily

Shaped, Apodized, and Excited
Ultrasound Transducers
J ~ r g e nArendt Jensen and Niels Bruun Svendsen

Abstract- A method for the simulation of pulsedpressure The paper proceeds along the following lines. Section I1 for-
fieldsfromarbitrarily shaped, apodized and excitedultrasound mulates the problem and details the underlying theory. Section
transducers is suggested. It relies on the Tupholme-Stepanishen 111 gives various implementation details and Section IV lists
method for calculating pulsed pressure fields, and can also handle
the continuous wave and pulse-echo case. The field is calculated a number of examples for different transducer geometries and
by dividing the surface into small rectangles and then summing apodization functions. It will be shown that the method is fast
theirresponse. A fast calculation isobtainedby using the far- and gives accurate answers.
field approximation. Examples of the accuracy of the approach
and actual calculation times are given.
11. THEORY

I. INTRODUCTION The purpose of thispaper is to deviseafastandaccu-

rate method for calculating the pulsed pressure field emitted

T HE MOST IMPORTANT component in acquiring high

quality imagesfor medicalultrasound scanners is the
[Link] it determines the quality of
from an arbitrarily shaped,apodized, and excitedultrasound
transducer.
It is assumed that the transducer is mounted in an infinite,
thedata acquired, and thus the quality of the images and rigid baffle. Enforcing appropriateboundaryconditions, the
parameters displayed. Considerable effort has thereforebeen emitted field can be found by solving the wave equation for
spend ondesigning transducersandcharacterizing the field the velocitypotential 4) [4], [ 5 ] :
emitted and received [l], [2].
Several methods for calculating the pressure field have been
developedforassisting in the design and characterization of
varioustransducer geometries. Most of the methods can be
traced to the fundamental solutions of Rayleigh, King, and from which the pressure is calculated as:
Schoch, of which a review can be found in [3].
Themostpowerful approachseems to be the method
developed by Tupholme and Stepanishen [4]-[6], which gives
an exact solution for a transducer modeled as a planar piston where p0 is the mean density of the media, CO is the propaga-
vibrating uniformly in an infinite rigid, planar baffle. Analytic tion velocity. and p 1 is the over pressure.
expressions for several transducer types have been found [ 5 ] , Thecoordinate system shown in Fig. 1 is used in the
[7], but closed form solutions can not be found for all types. calculation. The particlevelocitynormal to the transducer
Especially the introduction of exotic geometries orapodization surface is denoted by ,[)(F2 + $3. t ) . The solution to the
of thetransducersurfaceleads to analyticallyunsolvable homogeneous wave equation using Greens function is [S]:
integrals.
In this paper we will develop a simulation approach based
on the Tupholme-Stepanishen approach, which can simulate
transducers with any apodization of the transducer surface and
with any excitation of the transducer. where S denotes the transducer surface. g is the time-dependent
Green’s function and is
ManuscriptreceivedJune 12, 1991; revised October 21, 1991; accepted
October 22, 1991. This work was supported in part by the Danish Technical
Research Council, with grant 1 6 4 2 1 8 , in part by Briiel and K j z r A/S, in part
by Novo’sFoundation, in part by H.C. 0rsted’s Foundation, and in part by
Trane’s Foundation.
J. A. Jensen is with the ElectronicsInstitute,Building 349, Technical
University of Denmark,2800Lyngby,Denmark,andhe is currentlyvisiting where I 1‘ - 2‘ - 3‘ I is the distance from the ’lUface
theDepartment of BiomedicalEngineering,DukeUniversity,Durham,NC to the point where the field is calculated The integral is a
27706. statement of Huygens’ principle, that foraplanar vibrating
N. B. Svendsen is withtheDanishAcousticalInstitute,Akademivej,
Building 356, DK-2800 Lyngby, Denmark. surface each point on the source generates a spherical wave,
IEEE Log Number 9105552. and
resulting
the field is found by integrating these waves at
0885-3010i92$03.00 Q 1992IEEE

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.
JENSEN AND SVENDSEN: CALCULATION OF PRESSURE FIELDS
ULTRASOUND
FROM TRANSDUCERS 263

Transducer The sound pressure for the incident field then is

/

or

Noteheretheseparationbetweentheexcitation and the

transducer geometry. The 7 l e ( t ) includes the electromechanical
Fig. 1. Coordinatesystem for calculatingtheincident field impulse response of the transducer [ 2 ] .
Explicit solutionsfor anumber of transducer geometries
have been found. Analytical expressions for the circular, flat
the point of interest under the assumption of radiation into an transducer can be found in [ 5 ] ,and for the circular, concave
isotropic, homogeneous, nondissipative medium. geometry in [7],[9].
If aslightly curved transducer is used, an additionalterm It must beemphasized that onlytwo approximations are
is introduced as shown in Morse and Feshbach [S]. This term used here. The first is the assumption of a large and slightly
is calledthesecond order diffractionterm in Penttinen and curved transducer, and the second assumption is that of sepa-
Luukkala [7]. It can be shown to vanish for a planar transducer, rability between excitation and transducer geometry. Trans-
and as long as the transducer is only slightly curved and large ducerscan be constructed in which thisis avery good
compared to thewavelength of the ultrasound, the resulting approximation, so that thepressure field calculated by this
expression is a good approximation to the pressure field [7]. method is in good agreement with the measured field.
If it is assumed that surfacevibration accountingfor the The geometric featuresof the transducer are contained in the
excitationfunction and electromechanicalimpulseresponse apodized spatial impulse response h,(T;, Fz, t ) and from this
can be split into a spatial component a(T; +Tfj)and a temporal the field for any excitation function, including the continuous
component U, ( f 2 ) then:
wave case, can be calculated. Further, it has been shown that
thepulse echo field received by the emitting transducercan
becalculated by [ l l ] :

where a(,?) is denoted the spatial source velocity distribution

[ 101. This implies that the vibration amplitude at a certain point
on the surface does not depend on time, so the amplitude of
vibration is not influenced by the shape of the excitation.
where vPe is the pulse-echo electromechanical impulse re-
Thefunction
sponse including the excitation function. So the emitted pulsed
hcL([Link]>t-tz)=
L
n(Fz+F3)g(T;.t I F'2+F3,t2)dzF3 (6)

is called the apodized spatial impulse response and it relates

field, thereceivedfield,andthe continuouswavecase can
be derived from the apodized spatial impulse response as the
electromechanical impulse response usually can be determined
from a simple measurement.
the transducer geometry to the acoustical field. By this function
Thus, the original problem is transformed into calculating
we can write
the apodized spatial impulse response. Closed form solutions
$ ( F 1 , ? 5 > t=
) lJE(t) *ha(r;:F2,tt) have been found for some cases as mentioned previously, but
t (7)
not for all geometries and rarely when apodization is used.
where ~ ~ is ( tthe) piston velocity waveform, and the velocity
potential is written as a convolution in time between this and A. Simulation Method
the apodized spatial impulse response. In analytic calculations the solution is found by evaluating
If the particlevelocity is assumed to beuniform over the whichpart of a sphere with center at the field point that
surface of the transducer, (5) can be reduced to [6]: intersects the transducer surface [9]. Thearea of the strip on the
radiator surface divided by the distance to the field point gives
dj(71.,72.t) = l t w e ( t 2 ) L g ( F 1 , t I F2+?3,tz)d2F3 dt2 (8) the spatial impulse response at that time instance in the case
of uniform vibration. When apodization is used, the different
where the last integral equals the traditional spatial impulse areas on the strip should be suitably weighted.
response. In this simulation method the problem is reversed. A spher-
Note that 11, depends on the difference between 71 and ?2, ical wave is emittedfrom apoint on the aperture and all
thus it is spatiallyvarying. To emphasize this h, is written spherical waves are summed at the field [Link] by
h a ( ? l , f 2 >t ) . theinverse of the distance from theaperturepoint to the

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.
 264 IEEE TRANSACTIONS ON ULTRASONICS, FERROELECTRICS. AND FREQUENCY CONTROL. VOL. 39, NO. 2, MARCH 1992

field point. The apodizedspatialimpulseresponse is then

approximated by

where r, denotes the points on the transducer surface.

Just dividing the surfaceintopoints has the drawback
that quitealargenumber of points mustbe used as the
variance of h, at a time instance depends on the number of
spherical waves received in one timeinterval. The number
of points canbedrasticallyreduced by dividing the surface
into small rectangles and then summing theresponsesfrom
theserectangles. Thisis the approach usedhere. A similar
approach has also been studied by Ocheltree and Frizzel for
the continuous wavecase [12]. Here we are studyingthe pulsed
field case in which the continuous wave casecan be calculated
(b)
as aspecial case.
Fig. 2 . Isodistance curves on thetransducer surface. (a) Near-field.
Dividing the transducer surface into squares introduces an (b) Far-field.
approximation to the true geometry, and the field will deviate
from the true [Link] problem is reduced by using small
Y
squares, where the distance to the field point is large compared
to the size of the squares. Thus, it is appropriate to use a far-
field approximation,whencalculating thecontribution from
eachindividual element. The exactsolution for the impulse
responsefromarectangular piston is derived in [ 6 ] so only
an intuitive explanation for the far-field solution is given here.
As theimpulseresponse at apoint in front of apiston is
proportional to howlarge part of thepiston that contributes Fig. 3. Description of thepistonandits orientation.
to the response at a given time, the problem of deriving the
response is reduced to geometricconsiderationsconcerning
thedistancebetweenthe field pointand the differentparts
of [Link] piston surfacethe
isodistance curveslooks like shown inFig. 2(a), but if the
distance increases the curves tends to straight lines, which is
shown in Fig 2(b). The first is the near-field situation and the
latter is thefar-fieldsituation.
To calculate the far-field response from therectanglea
description of thepistonandthelocation of the field point
as shown in Fig. 3 is needed. The piston is described by its
length and width and to fix the location of the field point, the Fig. 4. Far-field response
piston is placed in a coordinate system in the XY-plane with
the center at theorigin. Thenthe location is defined by the field
Based on the mentioned description of the system we get
point’s position vector, split up into a unit vector (xe,ye,z,)
andadistance, 1.
In generalthe far-field spatial impulse responsehasthe
shape of a trapezoid as shown in Fig. 4, where ti is the time-
of-flight from the nearest corner of the piston to the field point.
Likewise t 2 and t 3 are the time-of-flight from the second and
third nearest corner, and t l is the time-of-flight from the corner where wy and W , are the side lengths of the rectangle.
with the largest distance to the field point. In special cases two The arrival times are then calculated by
or more of the t’s areequal.
The trapezoid shape responsecanbecalculatedbycon-
volving two rectangular pulses. The width of these pulses are
calculated by projecting the lengthandwidth of thepiston
onto the line through the rectangles center and the field point.

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.
JENSEN AND SVENDSEN: CALCULATION OF PRESSURE FIELDS FROM ULTRASOUND TRANSDUCERS 265

tJ = t l + at, + atz. (14)

The amplitude of the trapezoid is afunction of piston area

and distance to the field point. The shape depends on At1 and
At,, but the area of the trapezoid integrated over the whole
timeinterval is alwaysequal to

B. Arbitrarily Shaped Piston

Thisquitesimple result for the far-field response of a
rectangular piston can be used when designing a fast numerical
method for calculating the response of a piston of any shape.
As for thepoint source representation the piston is split up
into small squares (or rectangles), but instead of representing Time Is] XIO-7
the small areas by point sources the far-field responsefrom
the rectangles is used. This makes it possible to use larger and Fig. 5. Simulated (-) and true (- - -) spatial impulse response of con-
therefore fewer squares to describe the piston. The resulting [Link] on they-axis is [Link] to
response is calculated by summing up the responses from t = ;;.63pS.
all thesquares. Apodizing is obtained by multiplyingthe
individualresponses by an apodizingfactor, that couldbe of the rectangles. By this method accurate and reliable results
afunction of e.g. theradius of thetransducer. In asimilar should be assured.
manner a time delay can be added to the response, giving a
different phase for different parts of the transducer, like in a
IV. EXAMPLES
phasedarray.
In this section severalexamples of use of theprogram
C. Far-FieldRegion are shown. Responsesare compared to analytic solutions
and guidelines for choosing the number of elements and the
The size of the rectangles must be chosen so that. the field
resulting computing times are given.
point lies in the far-field region. This is given by [l]:
The first example is for a concave, nonapodized transducer
n

wL with an aperture radius of 8 mm and a focal distance of 150

I>>- (16) mm. A n analytic expression is found for this geometry [9] and
4x
where 1 is thedistance to the field point, W thelargest can thus be compared to simulated responses.
dimension of the rectangle and X the wavelength, which equals The spatial impulse response at a distance of 120 mm from
the surface is shown in Fig. 5 from on the acoustical axis and
c o / f , where f is frequency. I f f is the highest frequency in the
response simulated, the side length should obey the relation out in steps of 1 mm. The transducer was divided into 3177
squares with a side length of 0.25 mm. Calculating the ten lines
W << d w ~ . (17) of the spatial impulse response at a sampling frequency of l 0 0
MHz took 1.5 S on an HP/Apollo 90001425t workstation', and
Examples indicating how close the side length can be chosen is shown as dashed lines in Fig. 5 . We see that the program
to this limit and what accuracy then is obtained, are given in quite accurately tracks the theoretical spatial impulse response
SectionIV. off the acoustical axis. On the axis it is, however, more difficult
to get the exact position and shape of the abrupt changes in
[Link] OF PROGRAM thespatialimpulseresponse due to the employment of the
The prime application of this program is to investigate fields far-field approximation.
from transducers of shapes withunknownanalytic solutions In the next example the concave transducer was apodized
and to studytheinfluence of apodization andphasing of with a Gaussian distribution function defined as
elements. It is very difficult to imaginethe shape of these a(T) = e-a;(f)l
fields, and therefore whether correct results are calculated. In (18)
order to solve this problem, the program has been divided where R is the radius of the aperture and 7' the distance from
into two parts. The first calculates the position and orientation the center. a p was chosen to be 2. The spatial impulse response
of thesmallrectangles describing the transducer, and the is shown in Fig. 6. The characteristicelimination of sharp
second performsthe field calculation. The two partsare edges in the spatial impulse response is seen.
independenttherebyenablingthepossibility of thoroughly To showthat thepulse-echo responsecan be calculated
testing and calibrating the field calculation, which is transducer to good accuracya singleexample is shown in Fig. 7. The
independent. The onlyuncertainty is then theplacement of measured and simulated responses were obtained at a distance
therectangles. This part of the programcan,however,be
interfaced to a CAD program, that can visualize the placement 'This workstation has roughly the calculation speed of a 40-MHz 486 PC.

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.
 266 IEEE
TRANSACTIONS ON ULTRASONICS,
FERROELECTRICS,
FREQUENCY
AND CONTROL,
VOL. 39, NO. 2, MARCH 1992

0.9 1 1,
0.8 -
13 0.7

0.6
-
-

''
2
1
-
0.5

0.4 -

3 0.3-
i
: '\:
B
z 0.2- t

0 1 2 3 4 5 6 7 8 01
-10 -8 -6 -4 -2 0 2 4 6 8
Time [I) xio-l
Lared dislance [mm]

Fig. 6. Simulated spatial

impulse
response of Gaussian apodized
con-
'[Link] on they-axis is [Link] to
t = 7T.S3/tS.

of 120 mm from the transducer surface. Themeasured pressure

field was acquired by moving a needle in steps of 0.2 mm with
an accuracy of 0.006 mm, measuring in a plane containing the
acoustical axis of the transducer. Before the measurement, the
transducer and the needle were aligned so that the needle was
parallel to the acousticalaxis. The datawere sampled at a
frequency of 100 [Link] simulated field was calculated
by measuring vp, as the response from a planar reflector, and
then using (11) to calculate the field. The simulation was done
with a element size of 0.25 mm, and the calculation took 12.1
seconds.
The envelope of the RF-signals is shown as a contour plot
with 6 dB betweenthecontours. The plotsspan 20 mm in
the lateral direction and 4 ps in the axial direction. An other
example at 60 mm from thetransducersurface isshown in
Fig. 8, indicating the good agreement between simulation and
Fig. 7. Measuredandsimulatedpulse-echoresponse for a concave trans-
[Link] details and examples can be found in [Link] at .r =l20 [Link] on the !/-axis [Link]
P11. corresponds to t = 77.S3ps.
An important parameter to be selected is the element size,
whichessentially determines theaccuracy of theresult. The
calculation time on other computers can be determined from:
size shoulddepend on thedistance to thefield, as this
determines how well the far-field approximation is. In Table NCT . Nsquares
T= (20)
I the quantity (w2f/(41co)) is shown against the normalized Lin
mean square error (MSE) defined by
whereNsquareS is the number of elements and L,,, the
Linpack rating. So for the newer HPiApollo 730 workstation
with a Linpack rating of 22 MFlops, it would take 0.11 S to
calculate the response with 0.1-mm rectangles.
where N is the number of samples in the responses.
The figures shownwere calculated for the concave, non- V. CONCLUSION
apodizedtransducermentionedpreviously at 120 mmfrom Amethodfor the calculation of fields from arbitrarily-
the surface and 10 mm off the acoustical axis at a sampling shaped and apodized transducers has been given. A calculation
frequency of 100 MHz. Also shown in the table is the time in the order of a few seconds were obtained by splitting
calculation time and calculation time multiplied by the number the transducer into small squares, and summing their far-field
of Flops obtained in the Linpack benchmark and divided bythe responses.
number of elements (NCT). All experiments were conducted Any excitation of the transducercanbehandled and the
on an HPiApollo 9000/425t with a Linpack ratingof 1.8 Mflop. continuous wave solution found by Fourier transforming the
Using the last number in thetablearough estimate of the calculatedresponses. It was also shown how to obtainthe

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.
 the geometric approximation of usingsquares. We therefore
hope to improve the accuracy and still attain short calculation
time by improving on this approximation in future work.

ACKNOWLEDGMENT
Klaus Bolding Rasmussen is thanked for his help during the
development of the technique and its implementation. Dr. Ole
Trier Andersen is thanked for valuable discussions during the
preparation of the paper.

REFERENCES
G. S. Kino, Acoustic waves, Devices, imaging & una106 signul procesJ-
ing. EnglewoodCliffs, NJ: PrenticeHall,1987.
J . W. [Link],and F. S. Foster,“Ultrasoundtransducersfor
pulse-echo medical imaging,” IEEE Trans. Biomed. Eng., vol. BME-30.
no. 8, pp. 4 5 3 4 8 1 , Aug. 1983
G. R. Harris,“Review of transientfieldtheoryforabaffledplanar
piston,” J . Acoust. Soc. Am., v01 70, pp. 1&20, 1981.
G . E. Tupholme,“Generation of acousticpulses by baffledplane
pistons,” Mathematika vol.16,pp.209-224,1969.
P. R. Stepanishen, “The time-dependent force and radiation impedance
onapiston in arigidinfiniteplanarbaffle,” J . Acousr. Soc. .4m., vol.
49, no. 3, pp. 841-849,1971.
-, “Transient radiation from pistons in a infinite planar baffle.” .1.
Acoust. Soc. Am. vol. 49, pp. 1627-1638, 1971.
[Link], “Theimpulse responseandpressure
nearfield of acurvedultrasonicradiator,” J. P h y ~ .D.,vol. 9, pp.
1547-1557, 1976.
P. M . Morse and H. Feshbach, Methods of Theoreticul Physics, Port I.
-
L
NewYork:McGraw-Hill.1953.
[Link], F. S. Forster and J. Hunt, “Transient fields of concave annular
B arrays,” L’ltrason. Imaging, vol. 3, pp. 37-61, 1981.
F
G. R. Harris,“Transientfield of a baffledplanarpistonhaving an
arbitrary vibration amplitude distribution,” J . Acousr. Soc. Am., vol. 70,
pp, 186-204, 1981.
J. A. Jensen, “A model for the propagation and scattering of ultrasound
in tissue,”J. Acoust. Soc. Am., vol. 89, no. 1, pp. 182-191, 1991.
K. B. Ocheltree and L. A. Frizzel, “Sound field calculation for rectan-
gular sources,” IEEE Trans. Ultrason., Ferroeiec., Freq. Contr., vol. 36.
1 no. 2, pp.242-248,Mar.1989.
- 8 - 6 4 - 2 0 2 4 6 g

[Link] dislancc [mm]

(b)
Fig. 8. Measuredandsimulatedpulse-echoresponseforaconcavetrans-
Jfirgen Arendt Jensen was born in Roskilde, Den-
mark, in [Link] MSc. degree in
[Link] at .r=60 mm. The time on they-axis is [Link]
electricalengineering in 1985from the Technical
correspondsto t = 38.92ps.
University of Denmark. He received the Ph.D. de-
TABLE I gree in [Link]
TABLEOVERACCURACY
AND COMPUTATION
subject of his Ph.D. was digital signal processing of
TIMEFOR DIFFERENT ELEMENT
SIZES
medicalultrasoundimages.
Heisa VisitingScientist atthe Department of
No. of Side &
U MSEa Calculation N@ Biomedical Engineering, Duke University, Durham,
elements length 4f,” % time (S) NC. He currently holds a postdoctoral fellowship at
the Technical University and has published a num-
788 0.5 mm 0.0169 14.7 0.1 233.77
ber of reports and papers on digital signal processing, ultrasound acoustics,
3177 0.25 mm 0.0042 6.4 0.3 173.93 andmedicalultrasoundimaging.
20 046 0.1 mm 6.76.10-4 3.5128.64 1.4
“MSE is mean square error (see text).
‘NCT the normalized calculation time (see text).
Niels Brnun Svendsen was born in Copenhagen,
Denmark, in 1962. He received the BSc. degree in
pulse-echo field, so all characteristics of the field can be found electrical engineering in 1986 from the Engineering
by this simulation method. Academy of Denmark.
The accuracy of the approach is on the order of 3 to 5 per- He has been with the Danish Acoustical Institute
since 1986 as aStaff Scientist, where he has worked
cent compared to the theoretical spatial impulse response (see withcomputeralgorithms in thefields of sound
(19)), when a reasonable number of elements is used ensuring propagation, siren coverage planning, and noise and
short calculation times. During the work with the program, it vibrationanalysis. In 1987 hecarriedoutastudy
on coherent Doppler simulation at TNO Institute of
was found that the main cause for the deviations to the true Applied Physics, Delft, The Netherlands. Currently,
response, was the use of the far-field approximation rather than he is engaged in the del /elopment of a noise analysis system.

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on November 13, 2009 at 09:53 from IEEE Xplore. Restrictions apply.