Sunteți pe pagina 1din 13

GEOPHYSICS, VOL. 66, NO. 4 (JULY-AUGUST 2001); P. 11951207, 21 FIGS.

Seismic array design by spatial convolution

Albin K. Kerekes

energy that propagates elsewhere, distribute the required total


ABSTRACT charge into smaller unit charges to retain the high-frequency
bandwidth while maintaining the required pressure eld for
Deriving the response of an array is one thing, design-
penetration, and attenuate ground-roll prone spatial frequen-
ing an array to match a desired response is quite another.
cies before propagation takes place.
The rst is easy, the second is not. Given a selected pass
The fundamental function of source arrays lies in transmit-
and reject requirement in the spatial frequency domain,
ting/radiating the seismic energy much the same way as an-
an array can be obtained that best matches such require-
tennas transmit radio signals. The source environment is of
ment within the limits of available hardware.
relatively high energy and transmission/radiation character-
Optimum spatial arrays for 2-D and 3-D/4-D seismic
istics depend upon the constructive and destructive interfer-
surveys can be designed using the technique of spatial
ence within the immediate vicinity of the source. In contrast
convolution. Such a technique relies upon uniform ar-
to the source, the receiver environment is of relatively low
rays of differing shapes and sizes as building blocks.
energy, where the emphasis is in the selective reception of
These building blocks are convolved in space because
weak signals often buried in source-generated and ambient
their selected responses matching notch points against
noise. For such a difference, the assumption of reciprocity be-
side lobes to achieve a desired end result in the spatial
tween source and receivers, and the consequent convolution
frequency domain. The nal array design can be made
of respective array responses can be deceiving. Receiver ar-
optimum for a given set of requirements, such as signal
rays are used mainly to attenuate spatial frequencies that are
preservation within the passband, attenuation within the
alias prone from both source-generated and ambient sources,
reject band, and azimuthal distribution for 3-D/4-D seis-
attenuate ambient random noise by the multiplicity of ele-
mic surveys. For any given design, solely the number and
ments, and secure overall coupling by the multiplicity of
the spacing of the elements limit the optimization.
elements.
A rule of thumb has been observed which shows that
Depending upon the priorities of the use, the design of arrays
the required number of elements in a 2-D array for
may vary. Since each seismic prospect has distinct and unique
3-D/4-D seismic is equal to the square of the number
eld conditions and exploration objectives, the array designs
of elements in a 1-D equivalent array for 2-D seismic. It
for both source and receiver must address such uniqueness.
is also observed that for a given number of elements, nar-
For these reasons, my paper focuses solely on the technique
row azimuth designs can offer greater attenuation than
of array design, avoiding any recommendation for a particular
wide azimuth designs.
set of source and receiver arrays for general eld use.
The optimum in array attributes can be dened as that ar-
INTRODUCTION
ray conguration which provides the best possible performance
addressing the particular requirements of the seismic prospect
Although there are many advocates of the point source, and the exploration objectives while falling within the limita-
point receiver approach to seismic eld acquisition, the fact tions set by the contractual hardware and economics. A more
remains that weve always collected better data with arrays. general optimum can be dened in terms of the pass and re-
This was true in 2-D surveys, and, in view of the so called ject bands, which are derived from the exploration objectives.
acquisition footprints in the otherwise uniformly and sym- We look for little or no attenuation in the passband, while we
metrically sampled 3-D and 4-D surveys, it seems to be true seek maximum attenuation in the rejectband. These two bands
today. Arrays have distinctly different functions at the source are generally next to each other, often overlapping. Ground
and at the receiver. roll and alias-prone interference is generally of low spatial fre-
Source arrays may be used to direct and/or beam-steer the quency. In directional characteristics, the main lobe provides
seismic energy towards the seismic targets while attenuating the passband, whereas side lobes are the rejectband we need

Manuscript received by the Editor February 22, 1999; revised manuscript received September 29, 2000.

Liberty Seismic Consultants Inc., P.O. Box 1118, Bandera, Texas 78003. E-mail: albin@pomarosa.com.
c 2001 Society of Exploration Geophysicists. All rights reserved.
1195
1196 Kerekes

to attenuate. In this sense, we can also say that optimum in ar- and simple characteristics. They are composed by a number
ray performance is that which provides maximum attenuation of elements (N ) spaced at a uniform in line spacing (d). The
next to the main lobe. Uniform (linearly spaced), circular, or response can be expressed by the following:
podded (closely grouped) arrays are popular not because they  
 sin Ndk 
are optimum but, rather, because they are easily designed R(k) 
= 20 log  , (1)
and implemented. In a geophysical sense, the easiest in design N sin dk 
and implementation is seldom, if ever, the optimum. where R(k) is the response in decibels, N is the number of ele-
Intricate but optimum arrays can be designed by a relatively ments, d is the spacing between elements, and k is the spatial
simple technique using the spatial convolution of uniform ar- frequency in wavenumbers. Such a response in decibels in the
rays. The process starts with the responses of the basic convolu- domain of spatial frequency (k) is equally simple. The num-
tional elements being manipulated to yield a desired response ber of lobes from the main lobe at k = 0 to the rst-alias lobe
in the spatial frequency domain. Once such a response is ob- at k = 1/d is equal to the number of elements in the array.
tained, the actual array is dened by the spatial convolution The notch positions are at k = 1/Nd, 2/Nd . . . N /Nd. Side-lobe
of those basic convolutional elements. In fact, I start with the attenuation will be symmetrical between the main lobe and
desired pass and reject response and obtain the dimensions of the rst alias lobe, and its maximum in decibels is given by
the array that will yield such a desired response. 20 log(N ). Figure 2 shows the response of a uniform array of
The paper begins with a brief summary of the spatial con- seven elements spaced at d interval.
volution used for constructing array responses. Next, 1-D and We can convolve a uniform array by itself, for example,
2-D array designs are discussed in detail with illustrations to (N 1, d1) (N 2, d2), where N 1 = N 2 and d1 = d2. The resul-
their pertinent responses in the spatial frequency domain. tant array will have increased attenuation with lobes and notch
positions unchanged. Figure 3 shows an example of a three-
THE PROCESS OF SPATIAL CONVOLUTION
element array being repeatedly convolved with itself. The top
The spatial convolution of two arrays can be best described is the three-element array with the response on the left and the
as follows. Upon each element of one array, we position a com- layout on the right. The middle right illustrates the convolution
plete version of the other array. The resultant array thus will of this three-element array with itself, i.e., for each element of
have a total number of elements that is the product of the num- the one, we position a complete array of the same, resulting in a
ber of elements in the two arrays being convolved. Figure 1 nine-element array with resulting unit-weights of 1, 2, 3, 2, 1.
illustrates the spatial process for both 1-D and 2-D arrays. The The lateral component on the drawing is only to facilitate the
multiplication of the amplitude spectra and the summation of illustration of the otherwise 1-D array design. The middle left
the phase spectra give the response of such a resultant array. If
symmetrical arrays are desired and used for which the spatial
phase response is zero, the design involves solely the multipli-
cation of the amplitude spectra (summing the decibels). The
method was rst described in a part of an oral presentation
at the 49th Annual SEG Meeting (Kerekes, 1979) which re-
ceived Geosources Technical Paper Award from the late Harry
Mayne for its originality.

CONVOLUTION APPROACH TO ARRAY DESIGN

In this section, 1-D arrays are used to illustrate the basic


idea behind the convolution approach to array design. We start
with a close look on uniform arrays. These arrays have distinct

FIG. 2. Uniform array characteristics and response shown for


a seven-element array.

FIG. 3. Example of convolution for N (the number of elements)


FIG. 1. The process of spatial convolution. and d (the spacing of elements) being kept constant.
Seismic Array Design by Spatial Convolution 1197

shows the corresponding response with notch positions iden- The design technique described in this paper involves the
tical to the top case but with increased attenuation. Further changing of elements and spacings in order to induce attenua-
convolution of the kind results in the case illustrated at the tion at any desired spatial frequency band by the interaction of
bottom of Figure 3. pertinent notches and lobes. We must reduce the spacing of el-
We can convolve two uniform arrays of different number of ements on any component array if we wish to move its notches
elements but the same spacing, i.e., (N 1, d1) (N 2, d2) where towards the higher spatial frequencies. With some practice and
N 1 = N 2 but d1 = d2, or the same number of elements but the use of a commercially available interactive computer pro-
different spacing, i.e., (N 1, d1) (N 2, d2) where N 1 = N 2 but gram, such design technique is relatively easy to master. Pass
d1 = d2. The resultant array will have increased attenuation and reject bands can be manipulated to yield attenuation that
with lobes and notches according to N 1 and N 2, or d1 and d2, is proportional to the particular signals we wish to attenuate.
respectively. Figure 4 shows an example of these types of arrays. Complete optimization, as outlined in the introduction of this
We can also convolve two uniform arrays that have a paper, is therefore possible for a given number of elements and
different number of elements and a different spacing, i.e., spacing limitations.
(N 1, d1) (N 2, d2) where N 1 = N 2 and d1 = d2. The resultant Figure 6 illustrates the process for a 24-element array as the
array will have variable attenuation with side lobes and notches result of three successive convolutions. The upper portion is
according to the pertinent N 1, N 2, d1, and d2. Figure 5 shows the result of the following convolutions:
an example of these types of arrays.
In summary, uniform arrays can be convolved in three dis- A1(N=2,d=5) A2(N=3,d=10) A4(N=2,d=10)
tinct manners, each of which yields predictably differing results. A6(N=2,d=12.5) = A7 (2)
These are (1) number of elements and spacing being constant,
(2) number of elements being different with spacing being con- The responses in color to the left on Figure 6 are the individual
stant, or number of elements being constant with spacing being component arrays involved in the convolution. The response
different, or (3) number of elements being different and spac- in black is the resultant array response, which is the sum of all
ing being different. the component responses in decibels. Notches of component
arrays will produce notches in the resultant array response. By
altering the position of notches in any of the component arrays
(changing the spacing), the attenuation and shape of the resul-
tant array changes in a rather predictable manner. On the right
are the layouts corresponding to the spatial convolutions of the
component arrays. The process is indicated by the markings on
the right edge from A1 through A7. A1 is the two-element ar-
ray of 5 m spacing and A2 is the three-element array of 10 m
spacing. A3 is the result of A1 convolved in space with A2 (i.e.,
for each element of A2 we have a complete array of A1). This
result of A3 is then convolved with the array A4, which is a
two-element array of 10 m spacing. The convolution yields the
next result marked A5. This result is then convolved with the
array A6, which is a two-element array of 12.5 m spacing, to
yield the nal result marked A7. The lower portion of Figure 6
shows the change introduced by altering the interval from 10
to 12 m on the three element (blue) array. The previously opti-
mum resultant array which has maximum side lobe attenuation
FIG. 4. Example of convolution for N (the number of elements) near the main lobe, has changed to one that has maximum side
being variable and d (the spacing of elements) being kept
constant. lobe attenuation away from the main lobe.
Figure 7 shows two subsequent alterations to the original in
Figure 6. The upper portion shows a change of interval from
10 to 11 m for the two-element (pink) array. The resultant
array has a narrow rejectband with the second side lobe being
maximally attenuated. The lower portion shows changing the
interval from 10 to 11.5 m on the three-element (blue) array
and from 10 to 11 m on the two-element (pink) array. The result
of these changes produced a wider reject band with maximum
attenuation in the middle.
Once the design is nal and optimum for our purpose, these
arrays designed in the wavenumber domain can be constructed
in the spatial domain, i.e., in the eld. The process involves
the spatial convolution of the component arrays as shown to
the right of each corresponding response. The uppermost on
Figure 6 results in a eld array with a combination of weight
and distance tapering (variable weight, variable spacing). The
FIG. 5. Example of convolution for both N (the number of subsequent examples on Figures 6 and 7 result in distance ta-
elements) and d (the spacing of elements) being variable. pered arrays (variable spacing).
1198 Kerekes

FIG. 6. Examples of design techniques. The responses of the components are shown on the left with the corresponding geometric
convolution processes on the right. The box insert shows the particular arrays being convolved with the corresponding color on the
graphs. Highlight circling shows changes between the upper and the lower examples.

FIG. 7. Examples of design techniques. The responses of the components are shown on the left with the corresponding geometric
convolution processes on the right. The box insert shows the particular arrays being convolved with the corresponding color on the
graphs. Highlight circling shows changes between the upper example in Figure 6 and these examples.
Seismic Array Design by Spatial Convolution 1199

Looking at these results it is fairly clear that if we start with the number of elements per array would remain the same. By
an array in the eld, followed by deriving its response to match increasing the dimension of our interest, we must consider in-
a given pass and reject scenario in the wavenumber domain, creasing the number of elements in the arrays we intend to
we would require many trials before a near satisfactory result use. A simple expectation is the power of two. If 12 geophones
could be obtained. The procedure in this paper is an inverse per receiver were adequate in our 2-D survey, the 3-D might
process to the above. It starts in the response domain by ad- require as much as 12 12 = 144 geophones.
dressing the pass and reject scenario, then moves to the eld Similarly to the 1-D array design, we use uniform arrays
to construct the array that corresponds exactly to the desired with a minimum number of elements as building blocks for
response. Such arrays are likely to be complex, requiring a the design of 2-D arrays. One of the simplest of these building
marked chain for the eld layout. Based on my three decades blocks is a square of four elements. Figure 8 shows the response
of eld experience and contrary to common intuition, such in the domain of azimuth, spatial frequency, and amplitude in
complex arrays do not require more time or greater eld effort. decibels (color). The graph also shows a slice response taken
Once the layout crew is equipped and trained, it will achieve at 90 azimuth. These responses can be derived separately for
comparable daily production rates to that which uses a simple each azimuth by taking the 1-D response for that azimuth using
uniform design. the following equations.
In summary, we can design 1-D arrays for 2-D seismic sur- For odd number of elements, we have
 
veys using uniform arrays of minimal number of elements as  m 
building blocks for the convolutional process. We add their g + 2 gi cos(2i dk) 
 0
responses in decibels to match a desired response of pass and  
= 20 log  
i=1
R(dk) 
reject characteristics in the spatial frequency domain. We can
  m

achieve this by manipulating the notch positions of these com-  g0 + gi 
ponent arrays (varying the spacing of elements) to interfere  i=1

with the other component arrays in order to affect attenuation  
N1
of resulting side lobes as needed. Once the desired optimum m= , (3)
in pass and reject characteristics is achieved in the spatial fre- 2
quency domain and knowing the attributes of all the compo- and for even number of elements, we have
 m 
nent arrays, we proceed to construct the total array in space   
2 gi cos{(2i 1)dk} 
by the spatial convolution of these component arrays. Such a 
 i=1 
= 20 log  
process can be extended into the domain of 2-D arrays for 3-D
R(dk) 
and 4-D seismic surveys.  m

 gi 
 i=1

2-D ARRAY DESIGN
 
N
As seismic acquisition moves from two dimensions to three, m= , (4)
from 1-D to 2-D arrays, we can not reasonably expect that 2

FIG. 8. Array response of a square of four elements. The response is in decibels as indicated by the color in the domains of azimuth
angle and spatial frequency. A single slice along the wavenumber axis at 90 is shown in the upper part of the gure.
1200 Kerekes

where R(dk) is the response as the function of the dimension- notches will fall on lobes, and we will obtain lobe attenuation.
less (dk) domain, g0 is the center weight, gi are the weights of This type of convolution process will yield uniform square ar-
elements from center outward, N is the total number of ele- rays with increased number of elements. Figure 11 shows this
ments in the array, d is the spacing of elements, and k is the type of successive convolution. Such uniform array combined
spatial frequency in wavenumbers. with the stacking process produces a response similar to the
A convenient display of results is a 3-D display where the popular stack array with anti-aliasing reliance on the stack
x-axis is the wavenumber, the y-axis is the azimuth, and the (Anstey, 1986). A more effective anti-aliasing can be achieved
z-axis is the response amplitude in color. We will examine re- by the use of tapered arrays (Kerekes, 1997).
sponses with respect to the main lobe, which is the passband, We can convolve the square of four elements with another
and the side lobes, which usually represent the rejectband. square of four, which has spacing that results in optimum side
These uniform arrays can be of simple geometry like a square lobe attenuation close to the main lobe. The resultant array
of four elements, a parallelogram, a hexagon, and so on. Fig- will be a 4 4 = 16 elements array. Naturally this process can
ure 9 shows some of these examples with the corresponding be continued with another convolution to yield 4 4 4 = 64
responses. These fundamental building blocks may or may not elements, and so on. Figure 12 shows examples of convolution
be symmetrical along all azimuths. If they are not, the convolu- where we varied the spacing to achieve such optimization. This
tion must take spatial phase response into consideration. The kind of optimization consists in placing the maximum attenu-
pentagon and the triangle are examples of arrays with nonsym- ation, which is the reject zone, next to the main lobe, which is
metrical characteristics along azimuths. Initially, we should use the pass zone.
symmetrical and uniform arrays, such as the square, the rectan- Figure 13 compares the above-mentioned 64-element opti-
gle, the cross, and the hexagon. Since they are symmetrical, the mum array with the uniform equivalent. The actual attenua-
spatial phase response for these will be zero for all azimuths. tion achieved by this 64-element, wide-azimuth, 2-D array is
Their amplitude responses can be derived and shown for all roughly equivalent to the attenuation of an 8-element, single-
azimuths in a 3-D plot of spatial frequency, amplitude, and azimuth, 1-D array. It seems that if we wish to retain the
azimuth. same rejectband attenuation, going from 2-D seismic surveys
We will start with the square of four elements. We can con- to 3-D/4-D seismic surveys, from single azimuth to wide az-
volve this uniform array by itself. The resultant array will be imuths, we should increase the number of elements approxi-
a 4 4 = 16 elements array. Similarly to the 1-D case, further mately by a power of two. A lesser number of elements for
convolution of this kind will increase attenuation on the lobes 3-D/4-D seismic surveys will likely compromise attenuation,
and around the notches, but will leave the lobe and notch posi- azimuthal distribution or both.
tions unchanged. This type of array will be centrally weighted We can also convolve squares with rectangles, triangles,
with uniform spacing of elements. Figure 10 shows responses pentagons, and hexagons of different spacing. Such con-
of this type of convolution. volution offers increased versatility for the total number of
We can convolve the square of four elements with another elements, but the process of optimization becomes increasingly
square of four, which has spacing that is half of the rst one. difcult and azimuthal distribution will likely become nar-
The resultant array will be a 4 4 = 16 elements array. Some row. Lets look at two options of convolution with the hexagon.

FIG. 9. Samples of some basic 2-D uniform arrays with 90 slice-responses.


Seismic Array Design by Spatial Convolution 1201

FIG. 10. Successive spatial convolution of identical square arrays. The number and position of the side lobes
remain constant while the attenuation of the reject zone increases.

FIG. 11. Successive spatial convolution of square arrays where the spacing of the second array is half of the rst
array. The results yield uniformly spaced arrays along the 90 slice.
1202 Kerekes

FIG. 12. Spatial convolution of square arrays where the spacing have been arranged such that side lobe attenuation
is progressively maximized near the main lobe. Based upon the assumption that most harmful interference is
located near the passband represented by the main lobe, this process is considered optimum for any particular
case.

FIG. 13. Comparison of the uniform and optimized arrays. The signicant difference is that the uniform array
reaches maximum attenuation of side lobes away from the main lobe, whereas the optimized array attenuates
side lobes in the immediate vicinity of the main lobe.
Seismic Array Design by Spatial Convolution 1203

Figure 14 shows the combined results. The middle image shows left). Figure 18 shows an example of a four-circle design with
the convolution of the rst with a second unit that has half the the 1-D equivalent of an eight-element array (upper left). In all
spacing of the rst. If we convolve the hexagon with an identi- of these cases in Figures 1618, the 1-D equivalent is designed
cal duplicate, we get the response shown by the bottom image. to yield side-lobe attenuation near the main lobe.
I would value the middle image more but, given certain objec- Contrast is shown on Figure 19 depicting the response of a
tives, the bottom image might serve well. Everything depends single-circle and an optimized four-circle array, both limited
upon our requirements of passband, rejectband, attenuation, to a total of 24 elements per receiver. The single-circle array
and azimuthal uniformity. has good azimuthal uniformity and poor attenuation. The four-
circle array has poor azimuthal uniformity but good attenua-
WIDE-AZIMUTH DESIGNS
tion. Increasing the number of elements per array will likely
Wide-azimuth arrays can be designed by a circular distri- increase attenuation, azimuthal uniformity, or both.
bution of elements. The rather popular single radius array NARROW-AZIMUTH DESIGNS
offers excellent azimuthal uniformity but very poor attenua-
tion. Figure 15 shows the response of a single circle together Narrow-azimuth arrays may be adequate and even desirable
with the 1-D equivalent of a two-element array. Given a con- for certain 3-D/4-D seismic prospects where the emphasis is in
stant number of elements per array, we may increase the num- the strike direction. Midpoint consistency for stacking requires
ber of circles of differing radii to obtain increasingly bet- nondipping reectors, more so if reection quality is poor. In
ter attenuation. Such increase, however, will cost azimuthal such cases, the 3-D seismic survey may be oriented predom-
uniformity. inantly in the strike direction for both azimuthal and offset
The spatial convolution technique can be very useful in the distribution. Arrays, therefore, can be oriented in the same di-
design of multiple circles with distinctly different radii to ac- rection, permitting the design for narrow azimuths. Having a
commodate an ideal and optimum balance of attenuation and constant number of elements per array, we can increase atten-
azimuthal uniformity. Such design strategy can secure optimum uation by limiting the response to a certain range of azimuths.
results within the usually narrow hardware limitation of ele- There are a great many ways that narrow-azimuth arrays can
ments per receiver. The radii are derived using the principles be designed for a given set of requirements. One of the sim-
outlined in the 1-D array designs to seek the best t to a given plest designs involve the use of crossarrays. This is similar to
pass and reject requirement. Figure 16 shows an example of a the previously discussed multiple and concentric circles except
two-circle design with the 1-D equivalent of a four-element ar- that we take out the azimuths that are not required.
ray (upper right). Figure 17 shows an example of a three-circle Figure 20 shows a cross array of six-arms, four elements per
design with the 1-D equivalent of a six-element array (upper- arm, together with the corresponding response. This would be

FIG. 14. Spatial convolution of hexagon arrays. The middle response shows the case where the second arrays
spacing have been halved to achieve juxtaposing peak-to-trough. The bottom response shows the convolution of
identical arrays. According to the requirements of the exploration objectives, either could be optimum.
1204 Kerekes

FIG. 15. Response of a one-circle array. The response shown in the upper left is a 1-D reference equivalent. The middle left is the
90 slice.

FIG. 16. Response of a two-circle array. The response shown in the upper left is a 1-D reference equivalent. The middle left is the
90 slice.
Seismic Array Design by Spatial Convolution 1205

similar to a four-circle array with about 67% of the azimuths imuths of 330 30 and 150 210 . Figure 21 shows a narrower
taken out. The six arms spread out to cover 60 , 30 of az- cross array of six-arm, four elements per arm, together with the
imuths, around the principal y-axis. Attenuation is centered corresponding response. Again, this would be similar to a four-
around the 0 , whereas the worst is at 90 azimuth. The array is circle array with about 83% of the azimuths taken out. The six
designed for good performance within 30 from its principal arms spread out to cover 30 15 of azimuths around the prin-
axis, and attenuation is optimized near the main lobe for az- cipal y-axis. Attenuation is centered around the 0 , whereas the

FIG. 17. Response of a three-circle array. The response shown in the upper left is a 1-D reference equivalent. The middle left is the
90 slice.

FIG. 18. Response of a four-circle array. The response shown in the upper left is a 1-D reference equivalent. The middle left is the
90 slice.
1206 Kerekes

FIG. 19. Comparison of the one- and four-circle arrays. As we move from wide-azimuth to narrow-azimuth
distribution, the response tends to loose azimuthal uniformity. The four-circle design in this case also shows
increased side-lobe attenuation near the main lobe.

FIG. 20. Cross array of 30 . The layout is shown on the left, and the upper right is a 0 slice along the orientation of the cross array.
Seismic Array Design by Spatial Convolution 1207

FIG. 21. Cross array of 15 . The layout is shown on the left, and the upper right is a 0 slice along the orientation of the cross array.

worst is at 90 azimuth. The array is designed for good perfor- often overlapping receiver array will always be more efcient.
mance within 15 from its principal axis, and attenuation is Proper array design technique for 3-D/4-D seismic surveys is
optimized near the main lobe for azimuths of 345 15 and not only simple and available, but should be an essential part
165 195 . of every 3-D/4-D seismic survey worldwide.
Comparing the 3-D responses, it seems that the narrower the
azimuths, the greater the attenuation around the principal axis CONCLUSIONS
for a given number of elements per array. Such is evidenced
by the noticeably poorer response around the 90 azimuth The design technique using spatial convolution of uniform
for the narrower design. Elements being constant, we can and symmetrical arrays offers a relatively simple way to match
increase attenuation by limiting the response to a certain problems with solutions. The predictability of simple uniform
range of azimuths. Such an array optimization process to array responses is being used to facilitate such design tech-
address a given set of conditions is not only cost-efcient, but nique. The method can be used for both 2-D and 3-D/4-D seis-
also paramount to best resolve the exploration objectives. mic surveys with standard, state-of-the-art seismic equipment,
and it is limited only by the available number of elements and
spacings of the hardware. The examples shown here are in-
SPATIAL SAMPLING AND ARRAYS
tended only for the illustration of the method and are prob-
The spatial sampling provided by the source and receiver ably too large to be cost-efcient today in most cases. These
intervals is an integral part of array design, setting the require- examples are not in any way recommended for any particu-
ments for anti-aliasing characteristics. Such sampling is usually lar seismic survey. However, the technique described here can
derived from the exploration objectives, the seismic velocities, make a signicant contribution in reducing the so-called acqui-
far-offsets, maximum dips to be imaged and other specics. Be- sition footprints within the ever-expanding limits of available
cause we generally undersample in space due to economic lim- hardware. As the seismic industry moves towards a better geo-
itations, arrays are used to minimize the side effects of such phone and more geophones per receiver station, the technique
undersampling, particularly the effects of aliasing. Spatially described here can, and hopefully will, help the eld geophysi-
aliased seismic energy, mostly from surface waves (ground roll cist to design increasingly better arrays.
and the like) is especially harmful to the integrity of 3-D/4-D
seismic data. A large bin size combined with a small circle ar- REFERENCES
ray, or often just a small number of bunched geophones per
Anstey, N. A., 1986, Whatever happened to ground roll: The Leading
group, leaves the door open for the spatially aliased and other Edge, 5, no. 12, 1921.
harmful interference. Interpolation might help smoothing the Kerekes, A. K., 1979, Seismic array designA practical approach: Pre-
sented at the 49th Ann. Internat. Mtg., Soc. Expl. Geophys.
footprints and it is relatively inexpensive in terms of cost. 1997, The stack array and the spatial alias: The Leading Edge,
In terms of signal enhancement however, a properly designed, 16, 345348.

S-ar putea să vă placă și