A Short Course in Seismic

Reflection Profiling
Theory, Wave Propagation in Layered Media,
Data Acquisition, Processing, Interpretation

Larry Braile
Professor
Earth, Atmospheric,
and Planetary Sciences
Purdue University VI. E. Vibroseis
June, 2016
braile@purdue.edu
http://web.ics.purdue/~braile

Science for Planet Earth

 A Short Course in Seismic Reflection Profiling
Theory, Wave Propagation in Layered Media, Data Acquisition,
Processing, Interpretation
Prof. L. W. Braile
Table of Contents

I. Introduction
II. Theory
III. Rock Properties
IV. Some Definitions
V. Seismic Data Acquisition
VI. Seismic Data Processing
Understanding Vibroseis Correlation
VII. Reflection Profiling Interpretation – General Principles
VIII. A Brief Look at Interpretation
IX. A Few Useful References

Science for Planet Earth

VI. Seismic Data Processing
E. Vibroseis Correlation – An Example of
Digital Signal Processing
 Vibroseis Correlation – An Example
of Digital Signal Processing
Larry Braile
Department of Earth, Atmospheric,
and Planetary Sciences
Purdue University
West Lafayette, IN 47907
braile@purdue.edu
http://web.ics.purdue.edu/~braile
URL for this file:
http://web.ics.purdue.edu/~braile/sage/ShortCourseNotes.6.A.Vibroseis.pdf

URL for PPT:

http://web.ics.purdue.edu/~braile/sage/ShortCourseNotes.6.A.Vibroseis.ppt
Understanding Vibroseis Correlation –
Significance and Audience
An example of digital signal processing and linear
systems representation of seismograms.
Vibroseis is an important and commonly used method
for seismic reflection data acquisition.
Most useful at the advanced undergraduate and
graduate level (most effective for students with a
basic understanding of convolution and Fourier
transforms).
Understanding Vibroseis Correlation –
Fundamental Concepts
Illustrate time and frequency domain content of signals.
Example of time and frequency domain representation and
signal processing.
Illustration of amplitude and phase representation of the Fourier
transform.
Applications of the convolution theorem.
Demonstration of the meaning of the phase spectrum.
Explains why the arrival time is at the center of the wavelet for
Vibroseis seismograms.
Explains how cross-correlation with the sweep collapses the
sweep reflections into wavelets.
Illustrate windowing, smoothing and effect of truncation
concepts.
 A Sample Vibroseis Sweep

The sample Vibroseis sweep (above) has relatively low

frequencies and a limited frequency range so that the signals can
be easily viewed in the following illustrations. Typical exploration
seismology Vibroseis sweeps are in the range of ~8 to 80 Hz.
The sample Vibroseis sweep has sinusoidal character and begins
with a frequency of 1 Hz and ends with a frequency of 5 Hz.
Effect of Windowing

The input sweep is commonly “windowed” (tapered at the ends as shown

above) to reduce mechanical problems associated with coupling of the
Vibroseis “pad” to the ground at the beginning and end of the sweep input,
and to reduce truncation effects (Gibbs phenomena) that produce
sidelobes and a longer duration wavelet.
Vibroseis Schematic – Figure 1 (next slide and small image below) is a
schematic illustration of how Vibroseis works. The data were calculated and
plotted with a MATLAB code. The first trace (red) is the sample sweep after
applying a window to smoothly taper the sweep at the beginning and end.
The second trace (blue) is an Earth vertical incidence reflection response
from three interfaces. The resulting response has three spikes (impulses)
representing reflection coefficients. The spikes are located at the two-way
travel times and are scaled in amplitude by the reflection coefficient. Note
that spikes 1 and 3 are positive
amplitude impulses and spike 2
is a negative amplitude impulse.
The next three traces (black)
are the reflection responses to
the downward traveling source
sweep. They are aligned in
time by the reflection travel
times and are scaled by the
spike amplitude (reflection
coefficient). Note that the
response for reflection 2 is
phase reversed.
Vibroseis Schematic (continued) – The sixth trace is the recorded
seismogram (green, uncorrelated) and can be produced by summing the
three reflection responses (black) or, equivalently, by convolving the source
sweep (red) with the Earth response (blue). When Vibroseis data are
collected and recorded from the geophones, the data initially look similar to
the uncorrelated seismogram shown in green. Note that the trace is long (in
this case 16 s) compared to the maximum two-way travel time of the target
Earth model (in this case 7 seconds), and is very “ringy” (sinusoidal
character). The difference
between the total recording
time and the sweep
length is the “listen time.”
The “trick” to Vibroseis is to
cross-correlate the sweep with
the uncorrelated seismogram.
This process collapses the
sweeps into wavelets and
reduces the length of the
seismogram. The resulting
seismogram is shown in the
bottom trace (red).
Figure 1. Vibroseis Schematic (explanation on previous slides)

After Lindseth, 1968

Figure 1. Vibroseis Schematic (explanation on previous slides)

With Sweep
=

After Lindseth, 1968

 Figure 2. Vibroseis Schematic – Cross-Correlation
Illustration of the cross-correlation process – position of one time lag of the pilot
sweep (red) is shown below (on top of the green trace; all time lags are
calculated as the pilot sweep moves from left to right over the uncorrelated
seismogram; also see video and animation in PowerPoint referenced below).
At this position, the cross-correlation coefficient (the sum of the product of the
two signals) is a maximum (the product of two positive parts of the signals is
positive, and the product of two negative parts of the signals is also positive)
and is plotted on the bottom red trace (the seismogram, s(t)) at the lag position
shown by the thin vertical line. Note that the wavelet that is produced is
symmetric and is centered on the two-way travel time for the reflections (1.5,
3.5 and 7 s). The wavelets contain the sweep frequencies (1 to 5 Hz energy).

After Lindseth, 1968

 Figure 1. Vibroseis Schematic – Cross Correlation with pilot
Correlation coefficient is calculated at each
time sample () to form seismogram

Coefficient plotted at
location of thin black
line ()

Crosscorrelation: shift, multiply and sum; repeat

After Lindseth, 1968

Figure 1. Vibroseis Schematic – Cross Correlation, 1st Reflection

Coefficient plotted at
location of thin black line

Arrival Time

Maximum positive correlation at this position

After Lindseth, 1968

Figure 1. Vibroseis Schematic – Cross Correlation, 2nd Reflection

Coefficient plotted at
location of thin black line

Arrival Time

Maximum negative correlation at this position

After Lindseth, 1968

Figure 1. Vibroseis Schematic – Cross Correlation, 3rd Reflection

Coefficient plotted at
location of thin black line Arrival Time

Maximum positive correlation at this position

After Lindseth, 1968

Time and frequency domain representation of Vibroseis
data acquisition and correlation processing: In the next two slides
(small view of the first slide shown below), the vibroseis process is illustrated in
both the time and frequency domain. The three columns illustrate the time
domain (red), frequency domain amplitude spectrum (blue), and frequency
domain phase spectrum (green) for each step (rows) of the process.
Figure 3a (1st row) shows the pilot sweep in time and the amplitude and phase
spectra. Note that the amplitude spectrum shows that the sweep contains
energy from about 1 to 5 Hz and the “edges” of the spectrum are smooth due
to the taper applied to the
sweep. The slope of the phase
spectrum is the delay of the
frequencies showing that the low
frequencies are delayed a small
amount and the higher
frequencies are increasingly
delayed.
Time and frequency domain representation of Vibroseis
data acquisition and correlation processing (continued): Figure
3b (2nd row) shows the time and frequency plots for the Earth response
(reflections). Note that the phase spectrum has constant slope as the spikes
contain all frequencies and each spike is delayed in time.
Figure 3c (3rd row) is the result of convolution (in the time domain) of the sweep
and the Earth response which produces the uncorrelated seismogram shown
in the lower left. The equivalent process in the frequency domain is
multiplication (the convolution theorem), which, for amplitude and phase
spectra means multiplying the
amplitude spectra and adding
the phase spectra. Note that the
phase spectrum of the uncor-
related seismogram (lower right)
is curved but not equal to the
phase spectrum of the sweep.
Time and frequency domain representation of Vibroseis
data acquisition and correlation processing (continued): Figure
3d (1st row) shows the time and frequency plots for the Earth uncorrelated
seismogram, identical to Figure 3c. To process the uncorrelated seismogram,
we cross-correlate with the sweep (shown in Figure 3e, 2nd row). Note that in
the frequency domain, cross-correlation is the same as multiplying with the
conjugate of the spectra (a corollary of the convolution theorem) which means
multiplying the amplitude spectra and adding the negative of one of the phase
spectrum to the other phase spectrum. In this process, the phase spectrum of
the sweep (and associated
delays in the time domain)
disappears and we are left
with the phase spectrum of
the Earth response. The
result is shown in Figure 3f
(3rd row).
Time and frequency domain signals:
sw(t)  SW(f) Vibroseis Sweep
e(t)  E(f) Earth Response (spike series)
u(t)  U(f) Uncorrelated Seismogram (sweeps)
s(t)  S(f) Correlated Seismogram (wavelets)

Operator Symbols:
  Multiplication
  Convolution
  Correlation
 Has Fourier Transform
 || Equals (above and below)
The following plots (sub-figures) are related both
horizontally and vertically.
Figure 3a,b,c. Uncorrelated Seismogram – Time and Frequency



  +



|| || ||


Figure 3d,e,f. Correlated Seismogram – Time and Frequency



  + (conjugate)



|| || ||


Theory and Derivation

Illustrated by Figure 1
 Schematically illustrated
in the time domain in
Figure 2

Illustrated by Figure 3a,b,c

 )

Illustrated by Figure 3d,e,f

 Illustrated by Figure 3f

The following two slides are Figure 3a,b,c and

Figure 3d,e,f in expanded view.
Figure 3a,b,c. Uncorrelated Seismogram – Time and Frequency



  +



|| || ||


Figure 3d,e,f. Correlated Seismogram – Time and Frequency



  + (conjugate)



|| || ||


The Vibroseis Method: The Vibroseis method has a number of
advantages related to the truck mounted source and long sweep (many
seconds). Substantial energy can be put into the ground with the long
sweep, multiple sources (trucks), and repeated sweeps (stacked). The
Vibroseis sweep can be designed and controlled, and it is a mobile and
repeatable source. Because the energy is distributed over time,
environmental effects are reduced. Cross-correlation is a powerful signal
recognition process so Vibroseis can be used in relatively noisy
environments. Unlike explosive sources, if a source location (often called a
VP) needs to be recorded a second or third time, it is relatively easy to repeat
the VP. Disadvantages are: the Vibroseis method may require substantial
equipment and cost for large offset surveys and sometimes produces large
surface waves that are difficult to remove from the record section.
Vibroseis PowerPoint Presentation

Vibroseis courtesy of ION Geophysical

A PowerPoint presentation including some of the figures shown
here, an animation of the Vibroseis correlation process, and
videos of a Vibroseis truck in action, is available at:
http://web.ics.purdue.edu/~braile/sage/SAGE2008Vibroseis.ppt
Download the .ppt file (3.4MB) and the five videos (.wmv files)
and place them in the same folder on your computer.
http://web.ics.purdue.edu/~braile/sage/vid00001.wmv (18.9MB)
http://web.ics.purdue.edu/~braile/sage/vid00014.wmv (40.2MB)
http://web.ics.purdue.edu/~braile/sage/vid00011.wmv (17.7MB)
http://web.ics.purdue.edu/~braile/sage/vid00013.wmv (14.3MB)
http://web.ics.purdue.edu/~braile/sage/vid00009.wmv (19.3MB)
Example of Vibroseis
correlation. First two
traces on the left –
Sweep (6-60 Hz);
numbers to right are
time in seconds. Next
three record sections
– Uncorrelated
seismograms; record
length (sweep length
plus “listen time”) = 15
s. Last record on right
– Correlated
seismograms (5 s);
note prominent
seismic signals
including hyperbolic
curved reflections.

From Yilmaz, 2001.

References
Anstey, Nigel A., Introduction to Vibroseis, 1961.

Anstey, Nigel A., Vibroseis, Prentice Hall, 1991.

Lindseth, Roy O., Digital Processing of Geophysical Data

— A Review, originally published 1968, SEG Course Notes,
1982, 282 pages, Catalog #251A, available at:
http://library.seg.org/doi/book/10.1190/1.9781560802310
http://library.seg.org/doi/abs/10.1190/1.9781560802310.ch7 (Vibroseis).

Yilmaz, O., Seismic Data Processing, SEG, Investigations

in Geophysics, No. 2, Tulsa, 526 p., 1987.

Yilmaz, O., Seismic Data Analysis (vol. 1 & 2), Soc. Explor. Geophys., Tulsa,
Volume I, p. 1-1000; Volume II, p. 1001-2027, 2001.
(http://library.seg.org/doi/abs/10.1190/1.9781560801580.fm).
References
From Lindseth, 1968: The Matlab code and Figure 1 (above ;
and related Figures) were inspired by the schematic
illustration from Lindseth (below).

http://wiki.seg.org/index.php/Dictionary:Fig_V-12
From Lindseth, 1968:

“This uncomplicated model responds well to relatively simple treatment. Crosscorrelation has been
mentioned as a good method to extract a known signal from noise. The conventional method used
by data processors to extract the earth signal from the recorded signal is to crosscorrelate the
sweep with the recorded signal. The process is done before applying NMO. Crosscorrelation of the
sweep with the record has an effect upon the phase exactly opposite to that of convolution,
cancelling completely the phase shift which occurred in the original convolution. In simple terms,
the original product of the field operation is:
Ae ( earth ) • As ( sweep ); Øe + Øs (Amplitude and Phase)
Crosscorrelation of the recorded signal with the same input sweep signal produces:
( Ae • As ) • As ; ( Øe + Øs) - Øs (Amplitude and Phase)
Consolidating and cancelling out terms, the result of the crosscorrelation becomes simply:
Ae • As 2 ; Øe (Amplitude and Phase)
All the influence of the phase of the sweep upon the phase of the field record has been eliminated,
leaving the individual reflection components of the processed record with the phase of the original
earth signal. Each recorded reflection coefficient should now be recognizable as the best
approximation of a spike possible with the range of frequencies contained in the sweep signal. This
is exactly the effect which we wish to obtain with wavelet deconvolution and which is so difficult to
achieve on signals from conventional seismic signal generators. The amplitude of the output is
weighted by the product of the amplitude spectrum of the sweep, but, since the sweep spectrum is
completely white, the squared spectrum is also white, which merely multiplies the earth spectrum
by a constant, which is not serious.”