You are on page 1of 18

GEOPHYSICS, VOL. 42, NO. 7 (DECEMBER 1977); P. 1321-1338, 14 FIGS.

, 4 TABLES

BEAM FORMING WITH VIBRATOR ARRAYS

M. E. ARNOLD*

Seismic field tests conducted near Tulsa, Okla- beaming was exhibited from surface seismometers.
homa, and the Handy area in north-central Texas used Visual inspection and statistical analysis of re-
linear arrays of vibrators to concentrate sound waves flection continuity indicated that it makes no differ-
into a beam, which is directed vertically when all ence in the Handy area whether beam-forming tech-
vibrators operate in-phase or at an angle from the niques are used in the field or are applied later in
vertical when relative time delays are introduced to processing. This result was anticipated, since pre-
each vibrator. vious seismic work in the Handy area indicated that
The soundwave directivity was verified in the Tulsa the random noise level was low enough to minimize
area by recordings from subsurface seismometers, the theoretical advantage of field summing.
and at the same time reflection enhancementby wave

INTRODUCTION CONCEPT OF EXPERIMENTS


The use of source arrays to achieve concentrated The term “beam forming”, as usedhere, describes
and directed beams is not new. Multiple and extended the process of introducing relative time delays be-
antennas have been used for concentration and di- tween vibrators in an array so that the waves from
rection of electromagnetic radiation. Arrays of the vibrators are made to arrive simultaneously at a
sources have been used in underwater sound and distant seismometer group. For instance, seismic ray
sonar applications for similar purposes. Arrays of paths from each of nine sources to a common seis-
seismic sources described in the patent literature mometer group are shown in Figure 1. If the reflection
endeavor to concentrate sound waves into a beam from each of the sourcesis to arrive at seismometer
and to a lesser extent attempt to achieve directivity group 8 in-phase, then the initiation of signals from
of the beam (Bodine, 1956; Poulter, 195I; Swift, sources 1 to 9 must be delayed progressively. The
1955; and Born et al, 1955). Needless to say, the appropriate delay values depend on the depth, dip
achievement of seismic wave beams and, in particu- angle, and medium velocity.
lar, directivity of these beams requires a high level The field configuration of seismometer spreads
of coherence and precise phasing of the sources. for determining the effectiveness of beam forming is
shown in Figure 2. Essentially, subsurfaceseismom-
/ Tulsa test area eters in each vertical plane were positioned along
In an area near Tulsa, seismic field tests were con- the quarter-circle of radius 300 ft about the center
ducted that used linear arrays of vibrators to concen- of the vibrator array and a 12-group surface spread
trate sound waves into a beam. Objectives of these was placed east of the vibrator array.
field testswere to establish (1) the directivity of sound The vibrators were arranged in east-west linear
wave beams for various time delays, and (2) the ef- arrays. One array consisted of five vibrators and the
fect of the beam direction on shallow reflections. other of nine vibrators, both with units spaced at

Manuscript received by the Editor April 20, 1976; revised manuscript received March 30, 1977.
*Amoco Production Co., Tulsa, OK 74102.
0 1977 Society of Exploration Geophysicists. All rights reserved.

1321
1322 Arnold

FIG. I. Vertical section showing raypathsfrom sources I to 9 to a reflector and thence to seismometergroup 8.

25ft intervals. For each array, monofrequency sig- frequency driving signals from 33 to 100.5 Hz at
nals driving each vibrator were in-phase so that the 4.5Hz intervals. Each monofrequcncy signal was
beam was directed vertically. For the nine-vibrator generated ten times, the recordings of which were
array, another experiment had the driving signals summed to obtain intermediate seismograms. The
between adjacent vibrators delayed successively 15 intermediate seismogramsthus obtained were then
1.5 msec easterly and westerly and finally, successive summed to obtain the final seismogram.
delays of 3.0 msec easterly. There were 15 mono- We w’ere concernedwith the reduction and analyses

FIG. 2. Configuration of vertical seismometerspreadsand surface seismometer spread (only two seismometer
groups shown) employed to determine the effectiveness of beam forming using vibrator arrays.
Beam Forming with Vibrator Arrays 1323

of signals recorded by the vertical seismometer Table 1. Boundary distance between near- and far-field
as a function of specified frequency for vibrator
spreads. To reduce the processing required, only
arrays tested in the Tulsa area.
signalsof frequencies 37.5,55.5, 82.5, and 100.5 Hz
were used. Inspection of these signals indicated a Radius
high value of harmonic distortion. For this reason,
the signals were filtered with a frequency response Frequency 9 x 25 Array 5 x 25 Array
which was maximum at the driving signal fre_quenLy
37.5 Hz 375 ft 94 ft
and had a half-power bandwidth of approximately 55.5 Hz 555 ft 139 ft
3 Hz. 82.5 Hz 825 ft 206 ft
100.5 Hz 1005 ft 252 ft
The first-arrival amplitude on the filtered seismo-
grams is considered to be the maximum amplitude
of the wavelet envelope which occurs near the center
of the wavelet at a time approximately equal to the far-field is tabulated in Table I for the 9 x 25 and
duration of the signal, i.e., 222 msec after the onset 5 x 25 arrays at frequencies of 37.5, 55.5, 82.5, and
of the wavelet. The maximum amplitude of each 100.5 Hz (Kraus, 1950). A velocity of 8000 ftisec
wavelet envelope was measured in decibels and was used. Table 1 indicates that, for the vertical seis-
corrected to a reference recording gain level. The mometer spreads, seismometers were far-field only
adjusted amplitudes of signals recorded by vertical for the 5 x 25 array.
and in-line horizontal components of the three- The far-field radiation distribution R, given by
component seismometers were summed vectorially Wolff and Maker (1930) for arra! s with uniformly
to obtain the amplitude and incidence angle of the spaced sourcesof equal power and with signals uni-
transmitted signal at the seismometer position in formly delayed is
the vertical plane.
sin Nz’
R,=
THEORETICAL CONSIDERATIONS N sinz’ ’
The radiation distribution of linear arrays usually where
is determined by measurementsin the “far” field, or
Fraunhofer region, on the assumption that the z’=z+--,
$
incident wave is plane. In the “near” field, or Fresnel 2
region, a spherical or curved wavefront is used. For
md
our experiment, the boundary between near-field and
z=h s’na

I
3 - COMPONENT’
SEISMOMETERS

FIG. 3. Graph of interval velocity to a depth of 300 ft derived from velocities measured in previous tests and
configuration of east-west vertical seismometer spread relative to a 9 x 25 vibrator array.
1324 Arnold

cy= angle between the array normal and the radial are direct paths, whereas least-time raypaths to the
line from array center to seismometer(vertical two shallowest seismometers (seismometers E-5, E6,
angle), N5, and N6 in Figure 2) involve reflection.
N = number of sources= nine- and five-vibrator
ANALYSIS OF SIGNAL AMPLITUDES
arrays,
d = distance between sources= 25 ft, In Figures 4 through 8, normalized signal amph-
A = v/f(v = 8000 ftisec), tudes for the antenna theory, the mathematical model
+ = 2wf7, and and field recordings in the vertical planes are plotted
7 = time delay between sources= 0.0 set, 0.0015 in polar coordinates. Figures 4 through 8 are, re-
set, or 0.003 sec. spectively, for arrays of one vibrator; a 9 X 25
array with a time delay of 0 msec between adjacent
The vertical angle of maximum radiation (y) is
vibrators;a 9 x 25 array with a time delay of 1.5 msec
between adjacent vibrators, both in an easterly and
y = sin-’ -& in a westerly direction; a 9 x 25 at-ray with a time
( >
delay of 3.0 msec between adjacent vibrators in an
Equation (1) was used to calculate the radiation easterly direction; and finally, a 5 x 25 army with a
distribution R,, and this was called the antenna time delay of 0 msec between adjacent vibrators.
theory. Equations (1) and (2) were employed for In Figure 4 the radiation distribution for a single
both the 9 x 25 and 5 x 25 arrays, although these vibrator from antennatheory is circular in both verti-
equations are less applicable to the former. This far- cal planes, i.e., signal amplitudes are equal at a
field approximation may be minor when compared constant distance from the vibrator in all directions.
to other discrepanciesbetween the theoretical model The amplitudes computed from the mathematical
and the physical properties of the subsurface;speci- model are also identical in the east-west and north-
fically, the subsurfaceis not an isotropic and homo- south plots but due to layering, the maximum values
geneous medium. Thus, the radiation distribution (1 .O) occur at +45 degrees and -45 degrees and a
values computed from equation (I) may be expected relative minimum value (0.83) occurs vertically
to only roughly approximate amplitudes measured beneath the vibrator (origin). The observed ampli-
from recorded signals. tudes from the single vibrator should also be identical
In addition, an amplitude comparison was made in the east-west and north-south plots but, unfor-
of signals derived from a mathematical model based tunately, were near the level of ambient noise. The
on geometric optics. This mathematical model uses dissimilarity of the field-recorded 37.5 Hz signal
the measured interval velocities of the near subsur- amplitudes suggeststhat the reliability of these mea-
face. A graph of these velocities is shown in Figure surements is extremely low. The 37.5-Hz signals
3, together with a cross-section showing a nine- could not be extracted from noise for angles from
vibrator array, and seismometers in the east-west -30 degrees to +30 degrees. The 55.5-Hz signals
vertical plane. on the east-west array were recorded improperly
The traveltime and amplitude decay (geometric and are not shown in Figure 4. The 82.5- and
divergence and reflection-transmission losses) for 100.5-Hz signals are similar on the east-west and
each vibrator-seismometer raypath were computed north-south plots; however, shallow lobes (maxima
from the horizontally layered model. The simulated at +75 degrees and -75 degrees) on the east-west
signal recorded by each seismometerthen is the sum plots are absent on the north-south plots. It should
of the sinusoidal wavelets from all vibrators in an be noted that for a single vibrator the theoretical
array, each suchwavelet being appropriately reduced radiation distribution is independentof driving-signal
in amplitude and adjusted in time for traveltime and frequency.
delay time differences. The simulated signal thus In Figure 5 the radiation distributions are shown
derived was filtered and the maximum envelope for a 9 x 25 array with a time delay of 0 msec. The
amplitudes measured in a manner identical to that antenna theory amplitude is maximum at 0 degrees
used for the field-recorded signals. For each com- in the east-west plane and the width of the vertically
bination of vibrator array and time delay, the wavelet directed beam decreases (beam concentration in-
envelope amplitudes were normalized to the highest creases) with increasing frequency. In the north-
value. south plane, the contour from antenna theory has a
The least-time raypath for all but the two shallow- constant value of 1.0. The signal amplitudes com-
est seismometersin the vertical seismometer spreads puted from the mathematical model are similar to
Beam Forming with Vibrator Arrays 1325

East North
-

f = 55.5 cps

z = 55.5 :ps
Field-reccrded signal
mplitudes not measurable

f = 32.5 cpa

f = 100.5 cps

----- Antenna theory


- Mathematical model
---w___ Field recclrdings
FIG. 4. Radiation distribution of a single vibrator with frequencies of 37.5, 55.5, 82.5. and 100.5 Hz

those determined from antenna theory in the east- between +45 to +90 degrees and from -45 to -90
west piers; but amplitude valuesare considerabiy less degrees; however, due to the iayering, contours for
than 1.O, since the normalization value occurs on the the mathematical model are characterized,by a pro-
north-south plot. Contours obtained from the mathe- nounced minimum at 0 degrees. Maximum values
matical model in the north-south plane are nearly for field-recorded amplitudes occur on the north-
coincident with the contours from antenna theory south plots, with the exception of the plots for
1326 Arnold

100.5Hz signals, which is believed to be unreliable north-south plots, the field-recorded amplitudes are
due to noise. The contours for field-recorded ampli- similar to the mathematical model’s for angles from
tudes on the east-west plots indicate that, due to mea- approximately -45 to f45 degrees, with the ex-
surement error, the beam increases slightly in width ception of the polar plot for 100.5Hz signals. Outside
as the driving-signal frequency increases, rather of this range, the contours for field-recorded ampli-
than to concentrate as expected from theory. On the tudes decrease in value so that two distinctive lobes

f = 55.5 cps

f = 82.5 cp~

LEGEKD
----- Antenna theory

- Mathematical model
------- Field recordings
FIG. 5. Radiation distributionfor a 9 X 25 vibrator array with a time delay of 0 msec between adjacent vibrators,
frequencies of 37.5, 55.0, 82.5, and 100.5 Hz.
Beam Forming with Vibrator Arrays 1327

are formed that beam at anglesof approximately +45 a 9 x 25 array with time delays of I .5 msec in an
degrees and -45 degrees, respectively. The width easterly direction and also 1.5 nlsec in a westerly
of these lobes increases as the frequency increases direction. Field-recorded amplitudes for these two
from 37.5 to 82.5 Hz. time delays complement each other in the east-west
In Figure 6 we show the radiation distribution for plane and thereby provide measurementsfrom -90

0’
----- Antenrz. :kecry

------- Field recordings, 1.5 ms west


-.-I'..-. Field re:Drdings, 1.5 ms enst
FIG. 6. Radiation distributionfor a 9 X 25 vibrator array with a time delay of I .5 msec in an easterly and also in a
westerly direction, and frequencies of 37.5, 55.5, 82.5, and 100.5 Hz.
1328 Arnold

Ksrth
1.0 10 1.0 50 0 50 1.0
93' 90' 90. 90.

f = 37.5 irps

LSGEND
----- ,Ar.tenna thegry
- i,!:rthem&tical nodei
------- Field rzcor,jings
FIG. 7. Radiation distribution for a 9 x 25 vibrator array with a time delay of 3.0 msec inan easterly direction,
and frequencies of 37.5, 55.5, 82.5, and 100.5 Hz.

to +90 degrees. In the north-south plane, however, closely in that the main lobe is directed at an angle
field-recorded amplitudes for these two time delays of +30 degrees. Significant departures in patterns
should be nearly identical. Departures between these occur only on side lobes-most notably for 55.5.
two contours are not large and are indicative of the and 100.5-Hz signals. On the east-west 37.5-Hz
reliability of the field-recorded amplitudes. All three plot, the antenna theory and the mathematical model
radiation distributions in the east-west plane agree are similar; however, the maximum value on the
Beam Forming with Vibrator Arrays 1329

main lobe for field-recorded amplitudes is reduced -East


appreciably, since the maximum field signal value
occurs in the north-southplane. Antenna theory con-
tours in the north-south plane again are circular but
vary erratically with increasing frequency. The con-
tours in the north-south plane determined from the
mathematical model are similar in shape at all fre-
quencies. Contours for field-recorded amplitudes on
the north-south plots, in general, exhibit two main
lobes between +30 to +45 degrees and from -30
to -45 degrees.
In Figure 7 we show the radiation distribution for
a 9 x 25 array with a time delay of 3.0 msec in an
easterly direction between adjacent vibrators. In
general, the mathematical model and field-recorded
contours exhibit a pronounced lobe in the east-west
plots. This lobe is between +60 and +75 degrees
for the mathematical model and at +45 degrees for
the field-recorded amplitudes at frequencies of 37.5,
55.5, and 82.5 Hz. At 100.5 Hz, the two lobes are
coincident at +75 degrees. The antenna theory is in
close agreement with the mathematical model at
anglesless than +75 degrees;however, the maximum
value for the antenna theory occurs at +90 degrees.
On the north-south plots the antenna theory contours
again are circular and are varying erratically with
increasing frequency. The shape of the mathematical
model on the north-south plots is similar in shape to
the north-south plots in Figure 6; however, contour
values are appreciably less. For 37.5-, 55.5-, and
to a lesser extent, 100.5-Hz signals, the field-
recorded contours on the north-south plots exhibit
pronounced lobes at +30 and -30 degrees, which
is similar to Figure 6.
In Figure 8, a 5 x 25 array with a time delay of
0 msec between adjacent vibrators on east-west plots
only is shown. Comparison of these plots with the
east-west plots for a 9 x 25 array in Figure 5 demon-
stratesthat shorter linear arrays result in appreciably
less concentration (increased width) of the vertically
directed beam. As for the 9 x 25 array, the vertical
beam defined by antenna~theoryand the mathematical
model tend to decrease in width as the frequency ----- Antenna thi,ory
increases. Also, as in Figure 5, the contour for the
- Mathematic.A model
mathematical model displays a minimum at an angle
_____e_ Field reccrdinns
of 0 degreesat frequenciesof 37.5 and 55.5 Hz; how-
ever, unlike correspondingcontours in Figure 5, this FIG. 8. Radiation distribution for a 5 x 25 array with
minimum persists as the frequency is increased to a time delay of 0 msec between adjacent vibrators,
and frequencies of 37.5, 55.5, 82.5, and 100.5 Hz.
82.5 and 100.5 Hz. The contours for field-recorded
amplitudes, in general indicate, with the exception
of 100.5-Hz signals, that the beam is lessconcentrated
in a vertical direction than is indicated by contours model. Notably, the contours for field-recorded
determined from antennatheory and the mathematical amplitudes for a frequency at 37.5 Hz exhibit two
1330 Arnold

0 ms

I I I I I I
1

2
lo5 ms
east
3
4

3.0 ms
east

1.5 ms
west

FIG. 9. Seismic signal traces from seismometergroups 1 through4 of the surfaceseismometerspread(Figure 2)


recorded in the experimental area. A 9 x 25 vibrator array with time delays between adjacent vibrators of
0 msec, 1.5 msec in theeasterly direction, 3.0 msec in an easterly direction, and 1.5 msec in a westerly direction
were employed.

lobes directed at +45 degrees and -45 degrees, adjacent vibrators, of course, is to direct selectively
respectively. sound wave transmission and thereby enhance the
prominence of reflections from dipping subsurface
REFLECTION ENHANCEMENT IN TULSA
strata. In Figure 9 we show four seismic signal traces,
TEST AREA
respectively, from seismometer groups I through 4
The purpose of applying time delays between of the surface seismometer spread (Figure 2) for
Beam Forming with Vibrator Arrays 1331

each of four time delays between adjacent vibrators. top of Figure 9 a synthetic trace ~primary reflection
These seismic signal traces result from summing only), derived from subsurfacevelocity data, is shown
recordings of monofrequency signals from 33 to for identification of principal reflcctions.
100.5 Hz at 4.5.Hz intervals. Each monofrequency The effectiveness of beam forming may be gauged
signal was generated ten times, the recordings of by observing in Figure 9 the Oolagnh and Woodward
which were summed to obtain intermediate seismo- reflections as the time delay between adjacent vibra-
grams. time delays employed were 0 msec, I .5 msec tors is varied. The Oolagah reflection. at a time of
in an easterly direction 3.0 mbcc in an easterly di- 0.210 set, appearsprominent on aI1 traces for a time
rection, and I .5 msec in a westerly direction. At the delay of 0 msec. For a time delay of 1.5 msec in an

AREA 0 F
VIBRATO R ARRAY

V IBR .ATOR
AR RAY

co LUMNS
6 5 4 3 2 1
0 0 . . . . .
7 . . . . . .
. . . l . .
3 . . . . . .
;: VIBRATOR
4 . . l l . z POINT
3 . . 0 . . .

2 . . . . . .-
100 FIEET
1 . . . . . .F \

1
I I
lOOFEET
N

SCALE
0 2000 FEET
1 1
FIG. IO. Location of the Handy area test site in the north-central Texas and layout of the vibrator array and
seismometer spread.
1332 Arnold

FIG. 11. A previously obtained seismogram section A from dynamite detonationsand a vibrator correlogram
section B exhibit interfering reflections from complicated subsurface structure. The sections are shown at
differing horizontal scales. The diagrams above the sections indicate the positions of source pattern and test
spreads for the signals analyzed in this report.

easterly direction, the Oolagah reflection at seismom- gate beam forming capabilities of linear arrays of
eter group 4 is increased in amplitude relative to this vibrators. The tests were performed at a site de-
reflection at seismometer group 1. For a time delay signatedas the Handy area (Figure lo), located about
of 3.0 msec in an easterly direction and also 1.5 msec 15 miles north of Whitesboro, Texas, where re-
in a westerly direction, the Oolagah reflection is flections from beds with opposing dips are known
barely perceptible. Thus, it is concluded that the to interfere.
Oolagah formation has a west component of dip. The
Woodward reflection at 0.322 set is perceptible on TEST SCENARIO
the traces for a time delay of 0 msec and appearsmost A six-fold section displayed in Figure lla illu-
prominent on traces for a time delay of 1.5 msec in stratesthe seismic events which led to the implemen-
the westerly direction. Thus, it is concluded that the tation of this test program. This sectionwas assembled
Woodward formation has an east component of dip. from profiles recorded in the mid- 1960s using dy-
namite. The test line location, indicated above the
Handy test area section, is at a position of maximum interference
A series of tests was conducted to further investi- between seismic events for recording times greater
Beam Forming with Vibrator Arrays 1333

than I .9 sec. These events may be reflections from where + or - is used when the source is located
beds dipping in opposite directions and/or diffrac- down- or up-dip, respectively.
tions from faults. The correlogram section in Fig- In Table 2, delay time T in msec between two ele-
ure 1lb covers a somewhat shorter length of the ments (vibrator locations) in the in-line direction is
same seismic line. The two sectionsare displayed in tabulated as a function of dip angle in range for
different horizontal scales but do illustrate similarity. reflection at 2.45 set with an average velocity of
The equipment used at the Handy area consistedof 13,000 ft/sec. As noted in Table 2, negative delay
eight Wabco Model Y-6OOB vibrator trucks and an time r implies that moveout time decreaseswith in-
analog recording truck. Each vibrator had a delay- crease in range; consequently, delay of source initia-
line unit to provide the precise phasing of sources tion along the source array should increase from
required to accomplish beam forming. For these tests near-source to far-source.
the vibrators used 14- to 57-Hz signals of 7-set
FIELD SUMMING VERSUS COMPUTER
duration.
SUMMING
A 24-group seismometer spread was located south
of the vibrator array. Each group consisted of 36 The test program objective was to investigate the
HS- 1, 14-Hz seismometersin a 6 x 6 pattern having relative merit of introducing vibrator time delays in
25-ft in-line and 20-ft crossline intervals. the field versus the data processing introduction of
time delays.
CALCULATION OF DELAY time By way of introductionto the problem, let the noise
The appropriatebeam forming delay values depend N at some remote detector be random and not asso-
upon the depth and dip angle of the reflector, the ciated with the signal S. Let S be a reflection from
offset distance, and the medium velocity. For the one vibrator so that the signal-to-noise ratio at the
simple straight-ray case, these parametersare related remote detector is 1.
as follows: Now, assumeindividual seismogramsare recorded
at the remote detector for each of nine vibrators of a
lOOOL(X * TOV sin 0) linear array and, subsequently, 9 common traces are
(3)
‘= V[(TOV)2 + X2 k 2TOVX sin 01u2 ’ added during processingwith reflections in-phase. In
other words, the signal-to-noise ratio improves by
where
the factor fi or 3 (after numerous authors, e.g.,
r = time delay in msec between consecutivevibra- Smith, 1956).
tor locations, Now, consider insteadthat the nine vibrators of the
L = distance in ft between consecutive vibrator linear array are driven so that the reflections are
locations, in-phase at the remote detector. For this condition
X = offset distance in feet (range), of maximum radiation, and assuming that mutual
V = average velocity in ft/sec, coupling between vibrators is negligible, the signal
To = two-way normal incidence traveltime in set is improved nine-fold (Stratton, 1941), while noise
from seismometer group to reflector, remains the same. Thus, S/N = 9 for this case.
0 = dip angle of reflector, and This result indicates that a three-fold (or approxi-

Table 2. Delay time 7 in msecbetween tkvoelements of an in-line source array as a function of dip angle and range for
a reflection time of 2.45 set and average velocity of 13,000 ft/sec. When the source array is located up-dip
from the seismometer spread, the dip angle is positive.

Dip in degrees

Range, ft -10 0 10 20 30 40 50 60

0 +1.4 +0.1 -1.3* -2.6 -3.8 -4.9 -5.9 -6.6


1,000 +1.6 +0.3 -1.0 -2.4 -3.6 -4.8 -5.8 -6.6
2,500 +2.0 +0.7 -0.7 -2.4 -3.3 -4.5 -5.6 -6.5
4,000 +2.3 +1.0 -0.3 -1.7 -3.0 -4.3 -5.4 -6.4
8,000 +3.1 +1.9 +0.7 -0.7 -2.1 -3.4 -4.8 -5.9

*Negative delay time 7 implies that moveout time decreases with increase in range.
1334 Arnold

1 2
FlELb SUM (T=O) DPC SUM (T=O)

Window 1

-
1

2 Window 2

3 Window 3

.d- -
FIG. 12. The signals on correlogram no. I (field sum) were obtained using eight vibrators in array column 1
driven simultaneously. The signals on correlogram no. 2 (DPC-sum) are the nondelayed composite of eight
individually recorded correlograms for each vibrator along array column 1.
Beam Forming with Vibrator Arrays 1335

Table 3. Average correlation coefficient (C,,) and standard error of the mean (a,,,) for three time windows of traces
l-24 of correlograms from vibrator columns l-6 as a function of field or DPC summation.

Window 1 Window 2 Window 7


Place of
summation c av ffM c a” a.u c,, U.M

Field 0.64 0.09 0.64 0.09 0.86 0.04


DPC 0.60 0.09 0.63 0.08 0.83 0.05

mately lo-dB) improvement in signal-to-noise ratio tion coefficient; that is, consistent values of correla-
is anticipated if beam forming is done in the field tion (small standarderror) for reflection signals with
rather than during subsequentprocessing. This im- large values of signal-to-noise ratio and scattered
provement, however, assumesthat random noise is of values of correlation (large standard error) for re-
significant level; and in the Handy area random noise flection signals with small values of signal-to-noise
is not much of a problem. Record quality shown by ratio. The standard errors from these tests are about
seismogram sections in Figure II is good. Conse- equal in all comparisonsof field- and DPC-summed
quently, one anticipates that it should make little correlograms. Average correlations, however. are
difference here whether beam forming is done in the consistently higher for field-summed correlograms
field or subsequentlyat the processing center. than for DPC-summed correlograms, although the
As indicated by the map in Figure 10, six vibrator differences are not statistically significant and
array lines, spaced at lOO-ft intervals, were centered quantitatively support the similarity observed in the
at a point about 900 ft from the reflection spread. visual comparisons.
Each of the vibrators was driven individually and In Figure 13, correlograms from column 3 source
seismograms recorded. Each vibrator moved ap- array are shown. As before, visual comparison of
proximately 10 ft from one side to 10 ft on the other field- and DPC-summed correlograms indicates
side of the vibration point during the recording of that it makes very little difference in the Handy area
15 individual vibrations. Then. in the same in-line whether delay times are introduced in the field or in
array, all vibrators were driven simultaneously and the processing center.
the seismogramsrecorded for delay times of 0, + 1, The correlograms in Figure 13 do, however, illu-
+3, +5, and -3 msec. This procedure was repeated strate important effects of beam forming with these
at lines 2 through 6. specified delays on the character of events on traces
The correlograms displayed in Figure 12 permit l-12. Note, in particular, the effect of beaming on
visual comparisonof correlogramsfrom array column the events in window 2 centered at 2.45 sec. When
1 field-summed and summed in the Digital Pro- delay times of +3 msec are used, the amplitude of
cessing Center (DPC-summed). This similarity of that flat dip reflection is enhanced and the high-
continuity and quality of reflections indicates that in amplitude, right-hand (south) dipping event is nearly
the Handy area it is immaterial whether beam form- absent. When delay times of -3 msec are used, then
ing is done on-line or off-line. the south-dippingreflection has high amplitude, while
A quantitative evaluation of trace-to-trace correla- flat or north-dipping reflections are severely at-
tability over three trace segmentson zero-time-delay tenuated. When delay times of +5 msec are used.
correlogramsfrom all six linear arrays was estimated then “overbeaming” occurs and the amplitudes are
from correlation coefficients. Using calculated very low for all reflections with times more than
maximum values of correlation coefficients, average 1.5 sec.
correlation coefficient C,, and standard error of the A statistical study of trace-to-trace correlation of
means (T,~< were determined and listed in Table 3.’ events on traces 1-l 2 within windows 1, 2, and 3 for
Large values of average correlation coefficients field-summed correlograms for six adjacent arrays
are assumed to imply better continuity of reflection and the five different delay times is summarized in
(perfect correlation is 1.OO). The standard error is Table 4.
expected to vary inversely with the average correla- In window 1, trace-to-trace correlation for delays
of 0, +l, and -3 msec is 0.75, 0.74, and 0.72,
respectively. These values are expected since -1
‘Average standard deviation of the six sets of statistics is
assumed representative of all sets so that a,,, = o(n,)‘l’, msec delay is optimum for maximum reflection within
where n, = 6. this window and beam-power loss is less than 50 per-
1336
Beam Forming with Vibrator Arrays 1337

Tablel~A_veragecnrrelation coefficient (C,) and standard error of the means (au) for three time windows of traces
1-12 of field-summed correlograms from vibrator columns l-6 as a Gnction of d@iagtime 7.

Delay time Window 1 Window 2 Window 3


in msec
(7) C,V U.M C,V CM c a” fl.u

0 0.75 0.10 0.65 0.09 0.97 0.05


+I 0.74 0.08 0.66 0.10 0.91 0.06
+3 0.55 0.07 0.63 0.08 0.73 0.07
+5 0.55 0.07 0.47 0.06 0.66 0.07
-3 0.72 0.09 0.79 0.07 0.95 0.07

cent for delays within 2 msec of optimum. However, The time-shift offset relation of these distributions
when delays are +3 and +5 msec, one anticipates a supports the premise that the reflections enhanced
reduction of amplitude of the south-dipping reflection in the two casesare actually from different reflecting
and, hence, a reduction in correlation coefficient conditions. This interpretation is, of course, sup-
c a>. The 17 to 20 percent reduction of trace-to-trace ported by visual inspection of the seismograms
correlation to 0.55 is clear evidence of the power of themselves.
beam forming in discriminating against coherent In window 3 the gently south dipping multiple re-
events. flections have better than 0.91 correlation for time
In window 2, reflections from flat-to-north dipping delays of 0, fl, or -3 msec, but correlation is re-
and south dipping beds interfere. The average corre- duced to 0.73 with a delay of +3 msec and to 0.66
lation coefficient has a small value of 0.47 for +5 with a delay of +5 msec. This indicates that multiple
msec delay becauseof “overbeaming,” but is nearly reflections can be attenuated by choosing the proper
constant at approximately 0.65 for delays of 0, + 1, beaming direction.
and +3, and is 0.79 for -3 msec delay. For 0 and
+l msec delays, the beam is not precisely focused CONCLUSIONS
for either reflecting condition but broad enough to Field tests indicate that sound waves may be con-
result in about equal reflection from both reflecting centratedinto a beam which, in turn, may be directed
conditions and, consequently, similar values of to enhancethe prominence of reflectionsfrom dipping
average correlation. However, flat-to-north dipping subsurfacestrata by utilizing a linear array of vibra-
reflection is selectively enhanced when +3 msec tors. The directivity of sound wave transmission
delay is used and south dipping reflection is nearly is demonstrated by recordings from subsurface
optimum when -3 msec delay is used. The value seismometers. The enhancement of reflections by
of average correlation coefficient does not, of course, appropriately directing the sound wave beam is dem-
indicate that discimination between the two reflected onstrated by seismograms recorded from surface
events is visibly improved by beam forming. Evi- seismometer spreads.
dence that the correlation coefficients refer, in the The configuration and direction of sound wave
case of window 2, to continuity of different re- beams are depicted by radiation distribution contours
flections is obtained from the statistics by analyzing on vertical polar plots in-line and cross-line to the
distribution of maximum correlation coefficients as vibrator array. The radiation distribution contours
a function of correlation time shift. Such a distri- derived from field-recorded signal amplitudes, in
bution is sh&wn in Figure 14 where the number of general, conform to radiation distributioncontoursfor
occurrencesof maximum correlation coefficient as a signal amplitudes derived from a mathematical
function of different times shifts from zero-time model of near-surface strata in the test area but are
correlation is plotted with dots if obtained from cor- not generally conformable to radiation distribution
relograms with delays of +3 msec and with triangles contours derived from antenna theory, due in part to
if obtained from correlograms with delays of -3 erroneousassumptionsthat (1) the region of investiga-
msec. Visual inspection of the distributions indicates tion is in the far-field, and (2) the region of investiga-
the coefficients from correlogramswith delays of -3 tion is in a homogeneousand isotropic medium.
msec are located principally left of the distribution of Seismograms recorded from a surface spread in
coefficients from correlograms with delays of +3 the Tulsa test area demonstrate that the prominence
msec. Actually, the centers of the distributions are of shallow reflections on signal traces may be con-
3 msec apart. trolled effectively by variation in driving-signal time
FIG. 14. The distribution of maximum correlation coefficients as a function of correlation time shift is shown
by plotting the number of occurrencesof different amountsof time shift each side of the zero correlation. Oc-
currencesfor coefficients from correlogramswith delays of +3 msec are plotted with dots, Mhile occurrences
for coefficients from correlograms with delays of -3 msec are plotted with triangles.

delays between adjacent vibrators in an in-line array, REFERENCES


and also that dip of the correspondingreflectors may
Bodine, A. G., Jr., 19.56, Geophysical transducer: U.S.
be inferred from variation in reflection prominence.
Patent No. 2,735,507.
Visual inspection and statistical analysis of reflec- Born, W. T.. Faust, L. Y., and Wolf, A., 1955, Seismic
tion continuity on correlograms employing beam geophysical exploration: U.S. Patent No. 2,720,933.
Kraus. J. D.. 1950. Antennas: Nelc Y(vl\, McGraw-Hill
forming indicate that it is immaterial in the Handy Book Co.. Inc.
area whether beam-forming techniquesare usedin the Poulter, T. C., 1951, Seismic explorat~~~nusing elevated
field or applied later during processing. This result charge: U.S. Patent No. 2,545,380.
Smith, M. K., 1956, Noise analysis and multiple seismom-
was anticipated. since in the Handy area the random eter theory: Geophysics, v. 21, p. 357.
noise level relative to signal was low enough to Stratton, J. A., 1941, Electromagnetic theory: New York,
McGraw-Hill Book Co., Inc.
minimize the theoretical advantage of field summing
Swift. L. M.. 1955. Refraction shooting: U.S. Patent No.
over process summing of individually recorded 2,713,395.
correlograms, This conclusion assumessourceswith- Wolff, I.. and Malter, L., 1930, Dircctlonal radiation of
sound: J. Acoust. Sot. Am., v. 2, p. 232-233.
in a !ie!d+;ummed array are not mutually coupled.

You might also like