Page 262 - Acquisition and Processing of Marine Seismic Data

5.5 BAND-PASS FILTER 253

the velocities of direct wave and seabed components of a specified frequency band by
reflection are approximately between 1480 applying a suitable filter operator. The fre-
and 1520 m/s, although higher and lower quency band that a frequency filter keeps is
velocity values can be obtained in areas of termed the pass-band, defined by low (f 1 ) and
inclined seafloor. Slightly higher velocity high (f 2 ) frequency cut-off values. Frequency fil-
values are expected for subsurface reflection ters are typically classified according to the
hyperbolas, since the wave velocity characteristics of their pass-band regions.
generally increases with depth. Several different filter designs exist to prop-
iii. Checking the database: Some specific erly modify the amplitude spectrum of the input
information in the database such as fold data:
map, stacking chart, and direction of the
• Low-pass filters (Fig. 5.13A) remove the
survey line constructed by plotting x and y
frequencies higher than a specified cut-off
coordinate pairs for each CDP or shot gather
frequency value. Low-frequency cut-off (f 1 )
can be analyzed (Fig. 5.11) to ensure that
is zero.
the database has correct entries after
• High-pass filters (Fig. 5.13B) remove the
geometry loading.
frequency band lower than a specified cut-off
iv. Analyzing the brute stack: The brute stack of
frequency value (f 1 ). High-frequency cut-off
the seismic line can be prepared to check
(f 2 ) is the Nyquist frequency.
whether the data after loading the geometry
• Band-pass filters (Fig. 5.13C) keep the
seems normal. Fig. 5.12 shows the result of a
specified frequency band between a low- and
wrong channel order definition. In this a high-frequency cut-off value (f 1 and f 2 ,
example, the processor incorrectly supposes respectively).
that the channel numbers decrease along • Notch filters (Fig. 5.13D) remove only one
the offset.
specified frequency value or a very narrow
frequency band typically used to filter out
specific monofrequency noise components,
such as 50- or 60-Hz powerline interference.
5.5 BAND-PASS FILTER
Not all of the amplitudes from different fre-
Frequency filtering is the process of directly quency components of a seismic trace may be
modifying the amplitude spectrum of the seis- useful, because the seismic record contains both
mic data. The main motivation for frequency fil- signal and noise components with different fre-
tering is that the signal and noise amplitudes quency bands. In general, raw marine seismic
appear at different frequency components in data contains both high-frequency random
the seismic data. If noise frequencies are close noise and low-frequency swell noise compo-
to the signal frequencies, then the reflection sig- nents embedded in the data (Fig. 5.14). There-
nals may also be removed along with the noise fore, it is generally required to remove both
by frequency filtering. high- and low-frequency amplitudes, and such
As expressed in Section 4.5, seismic traces are a filtering is generally achieved by applying a
composed of the summation of several sinusoids band-pass filter designed to keep the amplitude
with different phase, amplitude and frequency components within a frequency band while
characteristics. We can decompose the traces removing the amplitudes outside this band of
into their different sinusoidal components of dif- the input amplitude spectrum. The pass-band
ferent frequencies using spectral analysis, and of a band-pass filter is schematically defined
then we can remove the undesired amplitude as a rectangular window in the frequency