Sie sind auf Seite 1von 12

Geophysical Prospecting, 2001, 49, 675686

Vibroseis deconvolution: a comparison of cross-correlation and


frequency-domain sweep deconvolution

K.F. Brittle, L.R. Lines* and A.K. Dey


University of Calgary, Department of Geology and Geophysics, Earth Science Building 240, 2500 University Drive N.W., Calgary,
Alberta T2N 1N4, Canada

Received January 2001, revision accepted July 2001

ABSTRACT
Ideally, traditional vibroseis processing produces a band-limited zero-phase Klauder
wavelet through cross-correlation of the sweep with the recorded signal. An
alternative wavelet processing method involves deconvolving the sweep from the
recorded vibroseis trace. This deconvolution can be achieved through frequency-
domain division. We compare and contrast the advantages and disadvantages of
sweep deconvolution versus cross-correlation on synthetic and real data.

INTRODUCTION with the use of cross-correlation, where the sweep is cross-


correlated with the trace creating an embedded Klauder
Research on vibroseis deconvolution followed the paper of
wavelet (Yilmaz 1987). Sheriff (1990) defined the Klauder
Hagedoorn (1962) entitled `In pursuit of the errant seismic
wavelet as the autocorrelation of a linear vibroseis sweep.
pulse'. In that paper, Hagedoorn compared pulses from PP
This definition includes non-linear sweeps as they can
and PS reflections for explosive sources. He basically found
be characterized by numerous linear sweeps. Additional
that the PS recordings experienced severe `high-cut' filtering
problems for vibroseis are the mixed-phases associated with
due to the near surface, compared with PP reflections. Hoffe
the combination of the zero-phase Klauder wavelet produced
and Lines (2000, paper presented at the SEG/EAGE summer
by cross-correlation and the minimum-phase effects during
research workshop, Boise, Idaho, USA) showed similar
acquisition. There are changes in amplitude and phase due to
results for multicomponent vibroseis recording in western
linear and non-linear filter effects from the vibrator system as
Canada. As a result of this, we attempt to restore high-
well as `earth filtering' effects on the waves as they travel
frequency information in vibroseis recordings by using the
through the earth. This results in the higher frequencies being
ideas on vibroseis deconvolution presented by Robinson
attenuated more than the low frequencies and the creation of
(1999) and Robinson and Saggaf (2001). (Although Robin-
a mixed-phase wavelet, a result of the convolution of the
son's analysis was in the time domain, whereas our analysis is
minimum-phase earth effects and the Klauder wavelet. If
in the frequency domain, the same principles apply.)
the sweep had a perfectly flat amplitude spectrum, cross-
The vibroseis source is widely used in seismic acquisition,
correlation (which involves multiplication of amplitude
as it is a non-destructive method with a controllable
spectra) would have the same effect as deconvolution.
frequency range. The higher the preserved frequencies from
However, since this is often not the case, sweep deconvolu-
the vibroseis source, the better the resolution of the thin beds
tion could remove the effect of the sweep's amplitude
associated with heavy oil reservoirs and reservoir geophysics.
spectrum if the sweep contained in the data is known.
In vibroseis data, the recorded trace has an embedded sweep.
This paper examines the potential of utilizing frequency-
It is necessary to remove the sweep from the trace to resolve
domain division to remove the sweep from the recorded
the reflectivities of the beds. This is traditionally completed
trace. Several models are examined to determine the
advantages and disadvantages of this method. The sweep
*E-mail: lines@geo.ucalgary.ca shape, earth filtering and the effects of noise are examined to

q 2001 European Association of Geoscientists & Engineers 675


676 K.F. Brittle, L.R. Lines and A.K. Dey

attempt to predict the success of this technique when applied The conversion into the frequency domain allows the
to field data. Field data are also examined and compared with processor to remove the sweep through division,
a synthetic trace generated from sonic and density logs.
X v
Rv: 5
Sv
C R O S S - C O R R E L AT I O N
There are several advantages to this method. In the noise-free
Cross-correlation is the industry standard for removal of an case, where the entire spectrum is covered by the sweep, a
embedded sweep. Vibroseis deconvolution utilizes the con- perfect answer for the reflectivity can be obtained. Another
cept that the autocorrelation of a sweep is equal to a Klauder advantage of this method is that the spectrum is whitened
wavelet. The autocorrelation is a measure of the similarity of with no residual effects of the sweep remaining. In cross-
a waveform with a time-shifted version of itself, with the correlation, the sweep shape creates a filter whose spectrum is
maximum occurring at zero-lag. After cross-correlation, the equal to the square of the sweep's spectrum. The complete
trace will contain only information that is common to both removal of the sweep for the swept frequencies is important
waveforms; therefore the method acts as a filter for unwanted as additional resolution may be obtained in the taper zone of
frequencies. A required filter can be implemented by the sweep. This would create the opportunity during field
designing a specific sweep. The results of cross-correlation acquisition for increased taper lengths in order to achieve
are extremely sensitive to changes in the sweep. Problems that maximum amplitudes, as well as including the use of non-
occur in association with the sweep can be partially linear sweeps with no permanent effects on the seismic data.
attributed to attenuation and improper vibration (Ristow Several steps are required for the removal of the sweep. If
and Jurczyk 1975). the sweep and trace are different lengths they should be
The basic seismic convolutional model for a vibroseis padded with zero values. Following this a Fourier transform
source is is completed for both the trace and the sweep. The trace is
then divided by the sweep in the frequency domain to return
xt rt*st; 1
the trace with no embedded sweep. Finally, an inverse
where x(t) denotes the recorded trace, r(t) denotes the Fourier transform is completed on the new trace to calculate
geological reflectivity, s(t) denotes the sweep and * denotes the time-domain deconvolved trace.
the convolution operator. To remove the sweep, the trace is
cross-correlated (^) with the sweep. The equation for the
correlated sweep is EFFECT OF SWEEP DESIGN

xCC t rt*st ^ st: 2 When the vibroseis technique was first implemented a linear
sweep was the signal of choice. A sweep is an oscillating
This equation can be simplified to
signal with time-variant amplitude and a frequency that
xCC t rt*kt; 3 varies monotonically with time (Goupillaud 1976). A linear
sweep is one that has an instantaneous frequency that is a
since the cross-correlation of two identical sweeps is defined
linear function of time, where the autocorrelation is the
as a Klauder wavelet, k(t). Cross-correlation collapses the
Klauder wavelet. The sweep has tapered ends designed to
sweep to a Klauder wavelet at impedance contrasts and filters
control the background levels of the sweep autocorrelation
the data with the sweep parameters. The filtering of the data
and the side-lobe reverberations (Goupillaud 1976).
may be a problem if the sweep was improperly designed or if
Non-linear sweeps do not have a frequency that is a linear
there were errors during acquisition.
function of time. When change in the frequency is fastest,
there will be depressed amplitudes (Goupillaud 1976). When
F R E Q U E N C Y- D O M A I N S W E E P a non-linear sweep is used, a different wavelet will be
DECONVOLUTION generated by the autocorrelation of the sweep. Four different
sweeps (Fig. 1) were used to examine the effect of the sweep
Frequency-domain sweep deconvolution (FDSD) is based on
shape on the results of cross-correlation and frequency-
the same convolutional model as cross-correlation. However,
domain sweep deconvolution. For cross-correlation, the
it uses the frequency domain to simplify the equation to
corresponding Klauder wavelet marks each impedance
X v RvSv: 4 boundary (Fig. 2). When the sweep is removed with

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
Vibroseis deconvolution 677

Figure 1 Sweeps designed and applied in the synthetic modelling to compare the methods of cross-correlation and FDSD.

cross-correlation, the amplitude spectrum of the reflectivity is The noise after cross-correlation is filtered in the upper and
filtered by the power spectrum of the sweep. For frequency- lower frequencies due to the sweep shape. This helps to
domain sweep deconvolution, in these synthetic examples, remove the effect of high- and low-frequency noise. The
the trace contained data for the entire spectrum, giving results of the synthetic modelling with the addition of noise
an exact answer for the reflectivity (Fig. 3). The results of are shown in Fig. 4. The synthetic models show that the noise
FDSD were filtered with a 10-15-135-150 band-pass filter makes only a small impression on the results of cross-
to remove any of the frequencies that were not part of the correlation. There are minor changes visible in association
sweep. The results of this method are independent of the with the low-amplitude impedance contrasts. However, all of
sweep shape. the noise outside the sweep's frequency range has been
removed, leaving a limited effect on the data.
For frequency-domain sweep deconvolution, the noise is
EFFECT OF NOISE
also additive to the convolutional equation. When the sweep
The signal-to-noise ratio is critical for the acquisition of good is removed in the frequency domain, the noise is divided by
quality data. Much of the noise can be controlled with proper the sweep. This results in
receiver patterns, good base plate to ground coupling and RvSv Nv
amplitude control in the field. For this study, random noise is X FDSD v 1 ; 8
Sv Sv
modelled into the convolutional equation to test the effect of
where the sweep is removed from the first part of the
the noise on the deconvolution methods. The noise is additive
summation but the noise is amplified. The noise can be
to the convolutional equation and is written as
reduced with two methods. The first is to remove the
xt rt*st 1 nt; 6 frequencies that are not included in the sweep with a band-
pass filter. This will remove all the amplified noise. The
where n(t) denotes random noise. The signal-to-noise ratio
second method is to add a small (1%) noise factor to the
for the additive noise is based on the mean power spectrum of
sweep before FDSD, eliminating any division by zero and
the trace.
reducing the amplification of the noise. The results of the
For cross-correlation, both terms in the summation are
frequency-domain sweep deconvolution for the sweep are
convolved with the reversed time sweep. This replaces the
shown in Fig. 5. A 1% noise factor was added to the trace
sweep on the left-hand side of the addition with a Klauder
before the sweep removal. The noise related to the linear
wavelet but leaves the noise cross-correlated with the sweep,
sweep was efficiently removed with both the noise factor and
resulting in
the band-pass filter. The quality of the results is dependent on
xt rt*kt 1 nt ^ st: 7 the amount of band-pass filtering. Any remaining random

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
678 K.F. Brittle, L.R. Lines and A.K. Dey

Figure 3 Comparison of frequency-domain sweep deconvolution for


Figure 2 Comparison of cross-correlation for linear and non-linear linear and non-linear sweeps. The results of the sweep removal by
sweeps. The results of cross-correlation are dependent on the FDSD are independent of the sweep shape.
embedded Klauder wavelet and subsequently the sweep.

Larner (1984) converted the mixed-phase wavelet to zero-


noise in the trace will be further reduced when the data are
phase by constructing a minimum-phase equivalent for the
stacked.
Klauder wavelet. The phase conversion filter is convolved
with the vibroseis record to convert the seismic trace to
EFFECT OF MINIMUM-PHASE GROUND
minimum-phase. Once this is completed, it is possible to use
F I LT E R I N G
minimum-phase spiking deconvolution to remove the mini-
As a source penetrates the ground and is measured by a series mum-phase, creating a zero-phase record. To determine the
of geophones the signal is altered by a filtering effect. The minimum-phase equivalent of a boxcar amplitude spectrum,
filter is more significant for a vibroseis source than for a noise factor is added to the Klauder wavelet. The minimum-
dynamite. Gibson and Larner (1984) and Cambois (2000) phase equivalent is dependent on the level of noise that is
have discussed the effect of earth filtering. All the authors added to the spectrum, controlling the results of this method.
identified problems associated with the earth-filtering effect Cambois (2000) suggested that it is more efficient to alter the
and suggested that it has an embedded minimum-phase low-frequency component of the Klauder wavelet in order to
wavelet. This is supported by the lack of problems for an remove the problems associated with the boxcar spectrum
impulsive dynamite source. Since a vibroseis source creates an than to add the noise factor. This eliminates problems
embedded zero-phase wavelet and the earth filter is mini- associated with adding noise to the frequencies that contain
mum-phase, the result is a mixed-phase wavelet. Gibson and real data.

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
Vibroseis deconvolution 679

Figure 4 Effect of noise on the results of cross-correlation. A 3:1 signal-to-noise ratio was applied to the trace before cross-correlation, with most
of the noise being filtered during the cross-correlation.

For initial modelling a minimum-phase earth filter is the earth filter after the sweep has been removed for the
replicated with a simple minimum-phase wavelet (Fig. 6). swept frequencies. However, problems arise in the prediction
The minimum-phase wavelet was convolved with the trace of the earth filter for the frequencies not covered by the
before the addition of any random noise. This gives a sweep.
convolutional equation of Through modelling, an initial examination of the two
methods was made. The first test is for a linear sweep with
xt rt*et*st 1 nt; 9
the minimum-phase earth-filtering effects. The results of the
where e(t) is the minimum-phase earth filter. With frequency- methods are comparable with both cross-correlation and
domain sweep deconvolution, the spectrum is whitened for FDSD, showing a phase rotation and attenuation of the
the sweep and any remaining spectral shape is due to earth higher-frequency amplitudes for the seismic after the sweep is
filtering. It should be possible to determine the spectrum of removed (Fig. 7). The phase rotation for cross-correlation is

Figure 5 Effect of a 3:1 signal-to-random noise ratio on frequency-domain sweep deconvolution. The method amplifies the noise in the
frequencies not associated with the sweep. The noise amplification can be controlled with a small (0.1%) noise factor during FDSD, with any
remaining noise eliminated with a band-pass filter. The band-pass filter defines the result's amplitude spectrum.

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
680 K.F. Brittle, L.R. Lines and A.K. Dey

due to the combination of the Klauder and minimum-phase cross-correlation with a linear sweep (Fig. 9). For FDSD the
wavelets. For FDSD it is due to the band-limited nature of the result is identical for the Gaussian sweep and the linear
sweep and the minimum-phase wavelet. sweep, with the filtering of the amplitude spectrum being a
To whiten the amplitude spectrum and remove the effects result of the minimum-phase earth filtering.
of the high-frequency attenuation, and to attempt to remove In an attempt to remove the earth-filtering effects,
the effects of the minimum-phase earth filtering, minimum- minimum-phase spiking deconvolution is applied to the
phase spiking deconvolution is applied. In spiking deconvo- data (Fig. 10). The result of the minimum-phase deconvolu-
lution there is an assumption that the embedded wavelet is tion on FDSD for the Gaussian sweep is identical to the linear
minimum-phase. For vibroseis data this is incorrect as cross- sweep, with a slight phase rotation as a result of the band-
correlation produces a Klauder wavelet that is then con- limited nature of a vibroseis source. The minimum-phase
volved with the minimum-phase earth filtering producing a deconvolution process cannot estimate the minimum-phase
mixed-phase wavelet. This incorrect assumption creates a wavelet for data outside the bandwidth of the sweep, limiting
phase rotation in the data. the accuracy of the minimum-phase wavelet removal. The
For a linear sweep, the results of the minimum-phase estimate of the spectrum for the minimum-phase wavelet in
assumption in spiking deconvolution are a small phase the case of cross-correlation fails more significantly. The
rotation. For both cross-correlation and FDSD, the mini- spectrum cannot be estimated accurately for any frequency
mum-phase wavelets spectrum is accurately estimated for the and the phase rotation is dependent on the sweep.
sweep frequencies (10150 Hz). The wavelet is incorrectly
estimated for the frequencies outside these, resulting in the
T H E M I N I M U M - P H A S E E Q U I VA L E N T
small phase rotation (Fig. 8). If the sweep covers a smaller
frequency range then the phase rotation is increased, with Gibson and Larner's (1984) method of removing the
a 1090 Hz sweep having an approximately 408 phase minimum-phase wavelet depends on determining the mini-
rotation. mum-phase equivalent of the Klauder wavelet. The
An advantage of FDSD is the independence of the result associated filter is used to convert the embedded zero-phase
of the sweep shape. The phase rotation in the data after wavelet in the trace after cross-correlation to minimum-
minimum-phase spiking deconvolution is dependent on the phase. This allows an accurate application of minimum-
band-pass filter that is applied to data, limiting the phase spiking deconvolution.
bandwidth of the data. The results for cross-correlation The results of this process are affected by the level of white
depend on the band-limited nature of the sweep and the noise that is added to the data as a stability factor (Fig. 11).
sweep shape. The phase rotation both before and after This has been modelled for both the linear and the Gaussian
spiking deconvolution varies for different taper lengths and sweeps described in the previous sections of this paper. The
sweep shapes. This is revealed by modelling a Gaussian- linear sweep's minimum-phase equivalents vary with the level
shaped sweep with the embedded minimum-phase wavelet. of white noise added to the Klauder wavelets spectrum.
When the sweep is removed with cross-correlation, the data However, the variation is greater for the Gaussian sweep.
are filtered by the Klauder wavelet and the minimum-phase An examination was completed to compare the quality of
wavelet. This creates different results compared with those of the minimum-phase wavelet removal from the modelled data

Figure 6 Minimum-phase wavelet for modelling the minimum-phase filtering effects of the earth.

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
Vibroseis deconvolution 681

Figure 7 The results for the removal of a linear sweep when the trace Figure 8 The results of minimum-phase spiking deconvolution on
is affected by minimum-phase earth filtering. There is a phase both cross-correlation and FDSD when minimum-phase earth
rotation for both cross-correlation and FDSD. processes affect the trace. The results are similar for both FDSD
and cross-correlation for a linear sweep.

in the cases with and without the minimum-phase equivalent


calculation. The minimum-phase equivalents at 10% and The same comparison can be made for a non-linear sweep,
0.1% white noise were determined for both the linear and in this example the Gaussian sweep described in the previous
Gaussian sweeps. The comparison was made between the sections of this paper (Fig. 13). For the Gaussian sweep, the
noise levels required to determine efficiently the minimum- results of FDSD are identical to those for a linear sweep.
phase equivalent of the Klauder wavelet and the filter to Cross-correlation combined with minimum-phase deconvo-
eliminate the mixed-phase wavelet. lution shows a larger phase shift compared with FDSD. The
For a linear sweep (Fig. 12), the results for a simple results when the minimum-phase equivalent Klauder wavelet
minimum-phase deconvolution for both the FDSD and cross- method is implemented show extreme variations in quality
correlated data show only slight phase rotations of ^ 108. If depending on the amount of noise added as a stability factor.
the minimum-phase equivalent for the Klauder wavelet is The low 0.1% stability factor gives a result that is
determined and the associated filter is applied to the trace, the approximately 1808 out of phase, while the 10% stability
results show slight errors. The first is a minor time shift of factor improves the result significantly.
10 ms as a result of the filtering to minimum-phase and the An advantage associated with the FDSD method is the
subsequent minimum-phase deconvolution. As a result of the complete removal of the sweep from data, eliminating
difficulties associated with determining the minimum-phase some of the effects related to the zero-phase Klauder wavelet
equivalent of the Klauder wavelet, there is a significant interacting with minimum-phase earth filtering. There
amount of side-lobe reverberation in the trace. will still be some phase rotation after minimum-phase

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
682 K.F. Brittle, L.R. Lines and A.K. Dey

Figure 9 The results for a Gaussian sweep for both cross-correlation Figure 10 The results for a Gaussian sweep for both cross-correlation
and FDSD when there is an embedded minimum-phase wavelet. The and FDSD when there is an embedded minimum-phase wavelet and
amplitude spectrum for cross-correlation is affected by the Klauder minimum-phase spiking deconvolution is applied. Both traces show a
wavelet and the minimum-phase wavelet, while the FDSD amplitude phase rotation. However, the phase rotation for FDSD is independent
spectrum is independent of the sweep and relies only on the of the sweep shape. The amplitude spectrum for cross-correlation is
minimum-phase wavelet. affected by the Klauder wavelet and the minimum-phase wavelet,
while the FDSD amplitude spectrum is independent of the sweep and
relies only on the minimum-phase wavelet.
deconvolution and the subsequent incorrect estimate of the
minimum-phase wavelet that is related to the limited
bandwidth of the sweep. The examples shown are for a the sweep was 300 ms. The shotpoint interval was 25 m and
10150 Hz sweep and give a phase rotation of 108. A sweep the geophone interval was 5 m.
of 1090 Hz increases the phase rotation to 408. The data were processed with exactly the same flow, with
the exception of the initial sweep removal for both cross-
correlation and FDSD. By retaining identical processing
F I E L D D ATA
flows, a direct comparison between cross-correlation and
The method of FDSD was tested on a data set acquired by frequency-domain sweep deconvolution can be made.
Petro-Canada in 1999. A direct comparison was made Figure 15 shows a comparison between both methods after
between cross-correlation and FDSD by following identical the sweep has been removed and a band-pass filter has been
processing sequences to minimize the effects of the processing applied to remove noise outside the sweep. As discussed in
on the results. The data were acquired with a 1080 Hz the modelling section, the noise outside the sweep's frequen-
sweep at a sample rate of 2 ms and a sweep length of 16 s cies was amplified for FDSD. However, the noise was
(Fig. 14). The taper length at both the beginning and end of substantially limited with the addition of a 1% stability

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
Vibroseis deconvolution 683

Figure 11 Minimum-phase equivalents with 10%, 5%, 1% and 0.1% stability factors for both the linear sweep and the Gaussian sweep. The
minimum-phase equivalent depends on the amount of white noise that is added as a stability factor to the spectrum of the Klauder wavelet. The
wavelets produced with different stability levels vary less for the linear sweep compared with the Gaussian sweep. After Gibson and Larner
(1984).

factor. The rest of the noise was removed with the band-pass at 1.90 s. At this time, cross-correlation shows small doublets
filter. that are not visible on the synthetic; rather they show up as an
Both methods yield similar results before any surface- increase in amplitude but do not become peaks. With an
consistent deconvolution is completed. With the completion increase in phase rotation, these peaks can be recreated.
of the minimum-phase deconvolution process there is a phase Another significant difference is the wavelet shape for the
rotation associated with both methods. The comparative lowest picked wavelet. The wavelet for cross-correlation
phase rotation between the two methods was determined to shows a slightly wider wavelet with a non-symmetrical shape
be 18.58. compared with the synthetic data. This irregularity can be
However, the interest lies in a comparison of both modelled if a 208 phase rotation is applied to the synthetic.
cross-correlation with no minimum-phase conversion and A third comparison can be made at the top pick, where
frequency-domain sweep deconvolution with the real reflec- the result of FDSD is closer to the synthetic compared with
tivity. This can be completed by tying the seismic data to a cross-correlation.
well. In this case, the well is slightly off-line. However, an
excellent tie was completed between the seismic and the
CONCLUSIONS
synthetic. The synthetic was created with a 1080 Hz
Klauder wavelet to replicate the ideal result of the field Both the modelling and the field data show positive results
data. The tie was made at the clastic/carbonate interface, the for the use of frequency-domain sweep deconvolution. The
largest and most significant change in velocity. The results of most important factor is the elimination of the phase shift
this tie are shown in Fig. 16. There is some stretch associated versions of the Klauder wavelet from the data. The
with the seismic, which is expected due to the location of the deconvolution of the Klauder wavelet helps to remove the
well in comparison with the line. filtering of the data that occurs when cross-correlation is
Both the cross-correlation and FDSD methods show slight performed. One advantage of this includes more promising
dissimilarities compared with the synthetic data. The most results in association with amplitude-attenuation-dependent
significant difference between cross-correlation and FDSD is processing such as frequency-change measurements for steam

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
684 K.F. Brittle, L.R. Lines and A.K. Dey

Figure 12 Comparison of cross-correlation, FDSD and effects of the white-noise levels on the minimum-phase Klauder wavelet equivalent for a
linear sweep. The cross-correlation with minimum-phase deconvolution does not include the minimum-phase conversion.

Figure 13 Comparison of cross-correlation, FDSD and effects of the white-noise levels on the minimum-phase Klauder wavelet equivalent for a
Gaussian sweep. The results for FDSD are independent of the sweep shape, but are dependent on the frequency content of the sweep and the data
recovered. The cross-correlation with minimum-phase deconvolution does not include the minimum-phase conversion.

Figure 14 Amplitude spectrum of the sweep used for the seismic acquisition.

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
Vibroseis deconvolution 685

Figure 15 Shot record and associated amplitude spectrum for both cross-correlation and FDSD. The two methods yield similar results. The noise
amplified outside the sweep range was removed with an 8-12-75-85 band-pass filter.

Figure 16 A comparison of the well log synthetic, cross-correlation and FDSD for the clastic/carbonate boundary. The two methods were
compared with the synthetic data from a local well log, and examination indicated that the phase rotation is less significant for FDSD. There is
some stretch between the seismic and synthetic as the well is slightly off-line.

injection in heavy oil reservoirs. A second advantage is frequency-filtering effects. The first is the minimum-phase
associated with the phase as discussed by Gibson and Larner earth attenuation and the second is the band-limited nature of
(1984), Cambois (2000) and Robinson (1999). Through the the data. For the frequencies of the sweep, it is possible to
process of removing the Klauder wavelet in FDSD, it is estimate the earth attenuation. However, for the non-sweep
possible to reduce the phase rotation associated with the frequencies the minimum-phase wavelets spectrum cannot be
mixed-phase vibroseis data, a combination of the earth determined, giving the small phase rotation found in the
attenuation and the zero-phase Klauder wavelet. Once the models. The results of the field data also show that there is
Klauder wavelet is no longer a factor, there are two only a small phase rotation and that the FDSD method

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686
686 K.F. Brittle, L.R. Lines and A.K. Dey

gives better results. The phase rotation significantly increases REFERENCES


for cross-correlation with non-linear sweeps, while the
results of FDSD are identical for all sweeps of identical Cambois G. 2000. Zero-phasing the zero-phase source. The Leading
Edge 19, 7275.
frequencies.
Gibson B. and Larner K. 1984. Predictive deconvolution and the
zero-phase source. Geophysics 49, 379397.
Goupillaud P.L. 1976. Signal design in the `Vibroseis' technique.
ACKNOWLEDGEMENTS
Geophysics 41, 12911304.
The authors thank the sponsors of the Consortium for Hagedoorn J.G. 1962. In pursuit of the errant seismic pulse.
Research in Elastic Wave Exploration Seismology (CREWES) Geophysical Prospecting 10, 148165.
Ristow D. and Jurczyk D. 1975. Vibroseis deconvolution. Geo-
and the Alberta Oil Sands Technology and Research
physical Prospecting 23, 363379.
Authority (AOSTRA) for funding this research. Husky Robinson E.A. 1999. Seismic Inversion and Deconvolution Part B:
Energy Inc. and Petro-Canada are gratefully acknowledged Dual Sensor Technology. Pergamon Press, Inc.
for providing the vibroseis data. We are also grateful to Scott Robinson E.A. and Saggaf M. 2001. Klauder wavelet removal
Hafner of Veritas for proposing the idea of sweep deconvolu- before vibroseis deconvolution. Geophysical Prospecting 49, 335
340.
tion in an unpublished manuscript that he made available to
Sheriff R.E. 1990. Encyclopedic Dictionary of Exploration Geophy-
us. Finally, we note our appreciation of Dr James Martin and sics. Society of Exploration Geophysicists, Tulsa, OK.
an anonymous referee for their constructive remarks on the Yilmaz O. 1987. Seismic Data Processing. Investigations in
manuscript. Geophysics, Society of Exploration Geophysicists, Tulsa, OK.

q 2001 European Association of Geoscientists & Engineers, Geophysical Prospecting, 49, 675686

Das könnte Ihnen auch gefallen