CGG
CNASE
Coherent noise attenuation and coherent signal enhancement
Release: v1.09
Last updated: October 2012
1. Function
1.1. General
CNASE stands for coherent noise attenuation and signal estimation and is designed for
wide patch 3D shots. CNA (coherent noise attenuation) is used for rejecting coherent noise
trains such as ground roll or guide waves. Alternatively, CSE (coherent signal estimation)
is used to pass signal in NMO-corrected shot gathers. CSE can be used as a random noise
attenuator.
Although CNASE is designed for 3D, it can also be run on 2D shots as well as 3D and 2D
receiver gathers.
geovation User Guide CNASE 1
CGG CNASE
geovation User Guide CNASE 2
CGG CNASE
2. Function Call
2.1. Description
Column Contents
1 *
3-7 CNASE
9-10 First option:
• NA: Coherent noise attenuation
• signal output = data - noise estimate
• noise output = noise estimate
• SE: Coherent signal estimation
• signal output = signal estimate
• noise output = data - signal estimate
12-13 Second option:
• 2: Estimation of coherent signal and noise simultaneously (see
Example 4)
15-16 Input buffer
23-24 Output buffer
31-80 Parameters
2.2. Parameters
2.2.1. Mandatory parameters
Computation parameters
Bn
n = Postprocessing loop number to be executed (integer).
Range: n > 0, where n is a valid loop number in the job
MAXOFFf
f = Maximum offset between shot and station (floating point).
Range: f> 0
Noise velocities (NA, NA 2, SE 2)
NOISEV(a,b,c,d)
(a,b,c,d) = Apparent velocities of the noise reject region in FK space (floating point).
Parameter NOISEV is mandatory for options NA, NA 2, and SE 2. It is ignored for the SE
option. See Section 3, “Recommendations”, p. 8.
Units: distance units / s
Range: 1.0 ≤ a ≤ b ≤ c ≤ d
CORRIDOR(LENGTHc, WIDTHd)
(c,d) = Corridor size to determine the total number of traces. See Section 3,
“Recommendations”, p. 8.
geovation User Guide CNASE 3
CGG CNASE
c = Length of the corridor (floating point). Units: distance units. Range: c > 0
d = Width of the corridor. The WIDTH parameter is only used for 3D shots (floating
point). Units: Distance units.
Range: d > 0. Default: d = LENGTH
FREQ(a,b,c,d)
(a,b,c,d) = Processing frequency range, in Hz (floating point).
These four frequency values define the low-cut, low-pass, high-pass, and high-cut
frequencies for estimating the noise. The low-cut and high-cut frequencies are included
to taper the frequency response in FK space. This parameter describes the range of
noise frequencies for both CNA and CSE.
Range: 0.0 < a ≤ b ≤ c ≤ d
Signal velocities (SA, NA 2, SE 2)
SIGNALV(a,b)
(a,b) = Apparent velocities of the signal region in FK space (floating point).
SIGNALV is mandatory with options SE, NA 2, and SE 2. SIGNALV is ignored with the
NA option.
Units: Distance units / s
Range: 1.0 ≤ a ≤ b
2.2.2. Optional parameters
NTn
n = Maximum number of traces in a group (integer).
Range: n > 0
Default: 500
Important
If NT is not specified and the number of traces in a group exceeds 500, the job
aborts.
Noise velocities (NA, NA 2)
DOWNDIP
If this flag is coded, the program rejects noise with negative dips.
Default: Positive dips are rejected.
See Section 3, “Recommendations”, p. 8 and Example 3.
OVERSMPf
f = Oversampling factor to decrease sampling interval and improve interpolation
(floating point).
Units: number of samples / cycle
Range: f≥ 1.0
Default: 2.0 (Nyquist)
geovation User Guide CNASE 4
CGG CNASE
Parameter CASCADE
This parameter applies strong noise attenuation by dividing the noise velocity range into
smaller overlapping velocity segments. If CASCADE is not coded, by default, PASSBND
and SLOPE are ignored and a single velocity segment is assumed (original velocity). See
Section 3, “Recommendations”, p. 8 and Example 1.
General coding:
CASCADE(PASSBNDp, SLOPEs)
where
PASSBNDp
p = Width of each individual noise velocity segment (floating point).
Units: distance units / s
Range: p > 0
Note
This is a mandatory parameter when CASCADE is coded.
SLOPEs
s = Width of the interior slopes for all segments (floating point).
Units: distance units / s
Range: s > 0
Default: s = average of starting and ending noise velocity slopes
Application window
For further information, refer to Section 3, “Recommendations”, p. 8.
TAPERf
f = Start and end times (in ms) of taper for the application window (floating point).
Range: f ≥ 0
Default: 200
GATE(Oc,Ss,Ee,...)
(o,s,e, ...) = Offset, start, and end times for application window (floating point).
Units:
• o in distance units,
• e, s in milliseconds
Range: o ≥ 0; 0 ≤ s ≤ trace length
Default values: o = 0.0; s = 0.0; e = trace length
Parameter EXTRAP
This parameter defines extrapolation options for multiple GATE parameters. By default, use
linear extrapolation wherever possible.
General coding:
geovation User Guide CNASE 5
CGG CNASE
EXTRAP(LOCATION=w, TYPE=t)
where
LOCATION=w
w = Location where the extrapolation takes place;
Possible values: ALL, NEAR, FAR, or OFF
Default: ALL
TYPE=t
t = Type of extrapolation.
Possible values: LINEAR or CONSTANT
Default: LINEAR
Outlier trace
See Section 3, “Recommendations”, p. 8.
THRESHf
f = Threshold scalar for editing corridor traces (floating point).
Range: f > 0
Default: 10
ZERO
If this flag is coded, output traces with a power greater than THRESH*median are zeroed.
By default, traces are not zeroed (output original trace).
WINDOW(a,b)
(a,b) = Threshold window for finding median power of corridor traces, in ms (floating
point).
Range: 0.0 ≤ a ≤ b
Default: last 1000 ms, or trace length if length ≤ 1000 ms
Simultaneous noise and signal estimation (NA 2, SE 2)
SIGFREQf
f = Minimum signal frequency, in Hertz (floating point).
Range: 0 ≤ f ≤ Nyquist
Default: 5.0
PREWHITf
f = Prewhitening percentage (floating point).
Range: f > 0
Default: 1.0
Other parameters
MUTE
This flag turns on the mute recovery. This flag must be used to apply mute recovery for
preserving zero input samples on output. Normally, CNA is run on shots with no zeros
geovation User Guide CNASE 6
CGG CNASE
on input and therefore MUTE is not needed. It can be used for CSE when the NMO-
corrected input shot gathers have been muted (see Example 2).
Default: no mute recovery
KTAPERf
f = High wavenumber taper as a percentage (floating point). See Section 3,
“Recommendations”, p. 8.
Options concerned: NA, NA 2, SE 2
Range: 0 ≤ f ≤ 100
Default: 5
AGCd
d = Internal AGC window length (floating point).
An AGC scaling is applied within the process to assure consistent amplitudes. The AGC
scaling is stored and removed prior to output so no net AGC scaling occurs. Parameter
AGC specifies the window length used.
Note
It is recommended to use the default value.
Range: d > 0
Default: 500.0
Parameter ALIAS
If coded, this parameter dumps the maximum unaliased frequency (fm). By default, the
program does not dump fm.
General coding
ALIAS(INTERVLa,FWEIGHTb)
where
INTERVLa
a = Station interval (shot order) or shot interval (station order). Floating point.
Range: a ≥ 1.0
FWEIGHTb
b = Weight assigned to FREQ values (floating point). See Section 3,
“Recommendations”, p. 8.
Range: 0.0 ≤ b ≤ 1.0
Default: 1.0
Auxiliary output buffer
OS1=AB
AB = Auxiliary output buffer containing the noise estimate. See Section 3,
“Recommendations”, p. 8.
geovation User Guide CNASE 7
CGG CNASE
3. Recommendations
3.1. Flexible trace header attributes: module-specific
information
3.1.1. Output trace: header attributes updated
• If header attribute TR_MAXI is not equal to the value −1 (its known maximum value), the
new maximum value of the processed trace is transcribed.
3.2. How does CNA/CSE work?
CNA/CSE use the following two steps for each output trace:
1. An offset and azimuth-limited corridor of traces around the output trace is collected to
build a 2D zone for noise estimation. The number of traces in the corridor usually ranges
from five to ten. See CORRIDOR parameters LENGTH and WIDTH.
2. In the FX domain, a modified FK filter is applied to the corridor of traces to estimate the
noise/signal on the output trace.
• CNA estimates the noise directly and subtracts the noise estimate from the input to
obtain the signal.
• CSE estimates the signal directly using NMO-corrected shot gathers.
3.3. FK comparison
Even though CNA/CSE is FK-based, it offers two potential advantages over ordinary FK
filtering:
• Ordinary FK filtering is global in nature and tends to smear large-amplitude noise,
whereas CNA/CSE is local in nature and isolates the effect of noise to a small corridor
of traces. See THRESH.
• Because ordinary FK filtering assumes regular spatial sampling, it is usually not
appropriate for 3D shots. CNA/CSE works in the FX domain and is able to account for the
irregular shot-receiver spacing common in a 3D shot.
3.4. Signal versus noise trade-off
Before we describe the parameters below, the following table (Table 1) should be
understood when using CNA. There is usually a trade-off between attenuating noise and
preserving signal. When noise attenuation is too aggressive, we may inadvertently subtract
some signal as well (left column). Conversely, in attempting to keep more signal, we may
also end up with more noise than desired (right column).
Table 1. Signal versus Noise
Attenuate more noise Keep more signal
(and maybe signal) (and maybe noise)
Do Not Apply GATE Apply GATE
Larger FREQ Range Smaller FREQ Range
geovation User Guide CNASE 8
CGG CNASE
Attenuate more noise Keep more signal
(and maybe signal) (and maybe noise)
Cascaded: PASSBND One Pass: No PASSBND
Use NA - not NA 2 Use NA 2 - not NA
Follow-up With DOWNDIP flag No DOWNDIP flag
3.5. Output and auxiliary buffers
Both CNA and CSE output the noise-attenuated shot to the output buffer OB. The auxiliary
buffer (OS1=AB) is used to store the noise estimate. AB = IB – OB.
For example, if OB is a zero trace (see ZERO parameter), then AB = IB. See also Example 3.
3.6. CORRIDOR parameters LENGTH and WIDTH
The CORRIDOR parameter specifies the corridor size thereby determining the total number
of traces in the corridor. WIDTH describes the width of a corridor centered on the shot-
receiver azimuth. The traces that fall within the corridor are sorted by offset, and those
traces with offsets within LENGTH/2 of the output trace offset are used in the offset-sorted
corridor.
The WIDTH is typically set to equal LENGTH, and because CNA/CSE is a local method, the
corridor should only contain five to ten traces isolated to a single receiver line. The average
number of traces used to construct corridors are printed for each shot. The output trace
is always included in the corridor.
Figure 1 shows an example of WIDTH = LENGTH = 1500 ft from a geometry with 200 foot
station intervals. The corridor of traces for the given shot,receiver pair are highlighted and
are used to build the modified FK filter.
As a rule of thumb:
WIDTH = LENGTH < 2 * average receiver line spacing
Normally, there is a trade-off between attenuating noise and preserving signal, and one
should be aware of the differences between CNA and CSE when using WIDTH and LENGTH.
• For CNA, larger WIDTH and LENGTH values tend to do a better job of preserving signal;
smaller values are usually better at attenuating coherent noise.
• For CSE,smallerWIDTH and LENGTH values tend to do a better job of preserving signal;
larger values are usually better at attenuating noise.
The WIDTH parameter is only used for 3D shots.
Table 2. Signal versus Noise
Attenuate more noise Keep more signal
(and maybe signal) (and maybe noise)
Smaller LENGTH/WIDTH Larger LENGTH/WIDTH
CNA
(3-6 traces) (7-12 traces)
Larger LENGTH/WIDTH smaller LENGTH/WIDTH
CNE
(7-12 traces) (3-6 traces)
geovation User Guide CNASE 9
CGG CNASE
Figure 1. Example of a corridor of traces for a shot,receiver pair
Seven traces fall within the corridor for this output trace.
3.7. Random noise and CSE
In CSE, the data can still benefit from a partial random noise attenuation by not including
all the noise frequencies. This more conservative approach may lead to less smearing.
Remember, the FREQ parameter only describes the range of noise frequencies for both
CNA and CSE; all frequencies in the original data are still output.
If CSE is used to attenuate random noise, be sure to include the full range of frequencies
occupied by the random noise. The shot must also be NMO corrected, and if high-frequency
noise attenuation is desired, statics should also be applied.
3.8. NOISEV for option NA
The CNA apparent NOISEV parameter must have four values to describe the noise reject
region in FK space. The four velocities must be increasing and describe the reject region
shown in Figure 2.
3.9. SIGNALV for option SE
The CSE apparent SIGNALV parameter only has two values to describe the symmetrical
signal pass region in FK space. There is a cosine taper applied between velocities V1 and
V2. See Figure 3.
3.10. Picking velocities
CNA/CSE velocities can be picked from a suitable FK analysis tool that displays velocity
information to see in what regions the noise/signal resides.
A receiver line nearest the shot (where the apparent velocities are valid) can be used to
pick CNA velocities. The center of the reject region in Figure 2 can be chosen to match the
reciprocal of the coherent noise dip according to the following relation:
Velocity = INTERVL / Dip
where Dip is in sec/trace, and INTERVL is the station interval in, for example, ft/trace.
For example, if the noise dips at 0.030 sec/trace and the station interval is 200 feet, then
the corresponding velocity = 200/0.030 = 6667 ft/sec.
geovation User Guide CNASE 10
CGG CNASE
3.11. CASCADE parameters (PASSBND and SLOPE)
A stronger noise attenuation can be achieved with the PASSBND parameter by dividing the
NOISEV range into smaller overlapping velocity segments. CNA is applied to each segment
and the results are cascaded as shown below.
The PASSBND value describes the width of each individual segment and must be specified
to use overlapping segments. Each segment has a low-cut, low-pass, high-pass, and high-
cut value. The passband width is defined as the high-pass low-pass, and is the same for
all segments.
3.11.1. Double Overlap
The double overlap method is used to provide continuous coverage across the NOISEV
passband by making the low-pass value equal to the previous high-pass value. To honor the
original NOISEV passband, CNA may slightly adjust the segment PASSBND value. It honors
the specified beginning and ending NOISEV slopes used in the first and last segments. The
SLOPE value represents the width of the interior slopes for all segments. The SLOPE default
is the average of the starting and ending NOISEV slopes.
The following example uses five segments with identical starting and ending NOISEV slopes
(200). Therefore, all interior slopes also equal 200.
3.12. DOWNDIP for option NA
By default, CNA rejects up-dip or positive dipping events which are defined by increasing
time with increasing offset. Up-dip events lie in the reject region of FK space shown in
Figure 2. To reject down dips or negative dipping events which decrease in time as the
offset increases, the DOWNDIP flag must be used. Down-dip events correspond to negative
wavenumbers and lie in the reject region which is the mirror image of Figure 2.
geovation User Guide CNASE 11
CGG CNASE
Figure 2. Noise reject region in FK space for CNA
The reject region for noise is described by the four velocities. At each frequency,
cosine tapers in the K direction are applied outside the dashed region. See the
KTAPER parameter description. Velocities must be such that: V1<V2<V3<V4
There are also cosine tapers applied in the frequency direction, as
specified by the FREQ parameter, between low-cut and low-pass
frequencies and between high-cut and high-pass frequencies. There
is no noise estimate for frequencies less than the low-cut frequency.
geovation User Guide CNASE 12
CGG CNASE
Figure 3. Signal pass region in FK space for CSE
The pass region in FK space is described by the two velocities, V1 and V2. The mirror
image of V1 and V2 is also shown since the pass region is symmetrical to include
negative and positive dips (velocities). At each frequency, a cosine taper is applied
in the K direction outside the dashed area. There are also cosine tapers applied
in the frequency direction, as specified by the FREQ parameter, between low-
cut and low-pass frequencies, and between high-cut and high-pass frequencies.
There is no signal estimate for frequencies less than the low-cut frequency.
3.13. THRESH, WINDOW, and ZERO
These parameters are designed to detect high-amplitude traces which are obvious
anomalous traces with no signal. CNA/CSE finds the median power of all the traces in the
corridor over the time WINDOW. The first and second values are the start and end times
of the window.
Any traces with a power greater than THRESH*median are excluded from the corridor. This
helps reduce smearing effects caused by high-amplitude noise. If the power of the output
trace is greater than THRESH*median, it is left unfiltered unless the ZERO flag is used, and
then the output trace is zeroed.
3.14. Application window parameters GATE, TAPER,
and EXTRAP
The GATE parameter lists the start (S) and end (E) application times for specified offsets
(O). The noise estimate is found for the whole trace but only the noise found in the gate is
subtracted from the data. Input data outside the GATE times is unchanged. A cosine TAPER
is applied to the noise estimate before and after the start and end GATE value.
geovation User Guide CNASE 13
CGG CNASE
For example, if the following parameters are coded:
GATE(O0, S500, E2000), TAPER200
the noise between 500 and 2000 ms are fully removed. The noise between 300 and 500
ms, and 2000 and 2200 ms is partially removed because of the 200 ms TAPER. Because the
times for only one offset were given, this single application window is applied to all traces.
To supply multiple offsets and application times, repeat the O, S, E sub-parameters. The
start and end times default to 0 ms and the trace length can be omitted if desired. For
example, if trace lengths are 4000 ms and the following values are coded:
GATE(O0,E1000,
O1000, S800, E3000,
O2000, S1200,
O3000 ),
the application times for each offset would be as follows:
Offset Start (ms) End (ms)
0 0 1000
1000 800 3000
2000 1200 4000
3000 0 4000
By default, start and end application times are linearly extrapolated for both offsets less
than the minimum specified offset, and offsets more than the maximum specified offset.
Parameter EXTRAP allows a customized extrapolation via two sub-parameters:
• LOCATION indicates whether extrapolation takes place before the minimum specified
offset (NEAR), beyond the maximum specified offset (FAR), for both locations (ALL), or
for neither location (OFF).
• TYPE specifies whether to use LINEAR or CONSTANT extrapolation.
A valid pair of application times must have an ending time greater than the starting time.
Traces with invalid GATE values are not filtered. This is a desirable feature for multiple
offsets when the goal is to not filter the trace when the gate length becomes so small that
no filtering is intended.
Example
EXTRAP(LOCATION=LINEAR,TYPE=FAR),
GATE(O500, S500, E5000,
O2100, S1400, E5000,
O7000, S3000, E5000)
In this example, the idea is to filter the gather over a noise cone. Traces with offsets greater
than 7000 have their gate times linearly extrapolated. At some offset, the extrapolated
start times may begin to exceed the ending time, 5000. Since these are invalid gates, no
filtering is applied to these traces.
In effect, the application gate pinches out.
3.15. KTAPER (CNA)
The maximum wavenumber, Kmax, is defined by the high-cut FREQ and the V4 velocity in
Figure 2. The KTAPER value is a percentage applied to Kmax to define the roll-off region
where a cosine taper is applied.
The taper in the K direction runs from Kmax− KTAPER* Kmax to Kmax.
geovation User Guide CNASE 14
CGG CNASE
3.16. Spatial aliasing
Coherent noise is spatially aliased at frequencies greater than fa if fa > 1/(2*dip), where the
dip is in sec/trace and fa is in Hz. We recommend attenuating the noise only at frequencies
that are unaliased. Therefore, we need to know what themaximum unaliased frequency is.
Although we cannot change the maximum dip of the data, we can limit the frequency range
by setting the high-pass FREQ value equal to the maximum unaliased frequency, f, such
that:
fa = 1/(2*dip)
We can either measure the dip (sec/trace) or make use of the fact that the dip is related
to the apparent velocity as shown below:
apparent velocity = interval / dip.
CNA estimates the maximum unaliased frequency by using the NOISEV parameter and the
interval.
In general, the interval is defined to be the difference in shot-receiver offsets between two
consecutive traces after ordering the traces in the corridor according to ascending offset.
For 2D, the interval is constant, but for 3D, the interval varies with offset and azimuth. For
3D, when the coherent noise travels at a constant propagation velocity for all offsets and
azimuths, the apparent velocity (interval/dip) is also constant. For example, both interval
and dip decrease at the same rate as the receiver line moves further off the shot. This effect
can be observed on 3D shots when the noise travels at a constant propagation velocity
for all traces. The noise is linear (with larger dips and intervals) when the shot is on the
receiver line, but the noise is hyperbolic (with smaller dips and intervals) when the shot
is off the receiver line.
For each velocity segment, CNA finds the frequency where the data begins to spatially alias.
Combining the above formulas, fa = 1/(2*dip) and velocity = interval/dip we have:
Maximum Unaliased Frequency at V2 = V2 /(2*interval)
where
V2 is the low-pass velocity for the velocity segment (second value). Based on the general
interval definition above, the 2D and 3D intervals are defined as follows:
• 2D: The geometry line item: station interval (for shot gathers) or shot interval (for receiver
gathers).
• 3D: The shot/receiver location that is nearest the center location of all shots/receivers is
used to find the average interval over all corridors.
Since the interval is only a function of the geometry, the maximum unaliased frequency
increases as V2 increases with multiple velocity segments (see Section 3.11, “CASCADE
parameters (PASSBND and SLOPE)”, p. 11). The maximum unaliased frequency is only
a guide and does not need to be honored precisely. This is especially true for 3D because
the average interval is only an estimate that represents an interval that actually changes
with offset and azimuth. Small amounts of aliasing may be tolerable for the smaller velocity
segments.
3.17. Quality control (QC) diagnostics
For each velocity segment, CNA dumps the velocities, the actual noise frequencies that are
used, and the maximum unaliased frequency at V2 to the log file.
geovation User Guide CNASE 15
CGG CNASE
For shot-ordered data, a 2D geometry with a station interval of 15:
FREQ(1,2,35,40), NOISEV(500,700,1700,1900), PASSBND200
the QC dump looks like:
Double Overlap Noise Velocity Noise Frequency Used Maximum
V1 V2 V3 V4 Lcut Lpass Hpass Hcut Unaliased
Frequency at V2
500.0 700.0 900.0 1100.0 1.0 2.0 35.0 40.0 23.3 Hz
700.0 900.0 1100.0 1300.0 1.0 2.0 35.0 40.0 30.0 Hz
900.0 1100.0 1300.0 1500.0 1.0 2.0 35.0 40.0 36.7 Hz
1100.0 1300.0 1500.0 1700.0 1.0 2.0 35.0 40.0 43.3 Hz
1300.0 1500.0 1700.0 1900.0 1.0 2.0 35.0 40.0 50.0 Hz
3.18. ALIAS parameter FWEIGHT
Normally, the QC dump is only informational, and does not alter the input FREQ values.
FWEIGHT values less than one can alter the actual frequencies by honoring the maximum
unaliased frequency at V2 estimate. The weights, FWEIGHT and 1-FWEIGHT, are assigned
to the high-pass FREQ and maximum unaliased frequency at V2, respectively, and the
weighted average is used for the new high-pass frequency. In the above example, if
FWEIGHT is 0.5, then the dump looks like:
Double Overlap Noise Velocity Noise Frequency Used Maximum
V1 V2 V3 V4 Lcut Lpass Hpass Hcut Unaliased
Frequency at V2
500.0 700.0 900.0 1100.0 1.0 2.0 29.2 34.2 23.3 Hz
700.0 900.0 1100.0 1300.0 1.0 2.0 32.5 37.5 30.0 Hz
900.0 1100.0 1300.0 1500.0 1.0 2.0 35.0 40.0 36.7 Hz
1100.0 1300.0 1500.0 1700.0 1.0 2.0 35.0 40.0 43.3 Hz
1300.0 1500.0 1700.0 1900.0 1.0 2.0 35.0 40.0 50.0 Hz
Weighted averaging has decreased the frequencies only for the first two velocity segments.
The averaged frequency values are never allowed to be less than the original FREQ values.
Note that the original slopes are maintained whenever the high-pass frequency changes.
If FWEIGHT is zero, then the maximum unaliased frequency at V2 is completely honored for
the high-pass frequency and the dump looks like:
Double Overlap Noise Velocity Noise Frequency Used Maximum
Unaliased
Frequency at V2
V1 V2 V3 V4 Lcut Lpass Hpass Hcut
500.0 700.0 900.0 1100.0 1.0 2.0 23.3 28.3 23.3 Hz
700.0 900.0 1100.0 1300.0 1.0 2.0 30.0 35.0 30.0 Hz
900.0 1100.0 1300.0 1500.0 1.0 2.0 35.0 40.0 36.7 Hz
1100.0 1300.0 1500.0 1700.0 1.0 2.0 35.0 40.0 43.3 Hz
1300.0 1500.0 1700.0 1900.0 1.0 2.0 35.0 40.0 50.0 Hz
3.19. Output
At each frequency in the FREQ range, CNA/CSE models the data as:
geovation User Guide CNASE 16
CGG CNASE
D = S + N + n
where
S = Coherent signal, N = Coherent noise, and n = random noise.
CNA finds an estimate, N^, for the noise, and CSE finds an estimate, S^, for the signal.
Table 3 summarizes the outputs for CNA/CSE inside and outside the FREQ range. The output
data equals the input data outside the FREQ range. For a complete table including CNA2/
CSE2 see CNA2 and CSE2.
Table 3. CNA/CSE Outputs
Model: D = S + N + n Option Estimates Output Buffer Auxiliary Buffer
Frequencies inside the
NA N^ D – N^ N^
FREQ range
SE S^ S^ D - S^
Frequencies outside
NA/SE 0 D 0
the FREQ range
4. Examples
Example 1. One-pass versus cascaded CNA for multiple noise trains
Testing and quality control is required for optimal parameter selection, particularly for
CORRIDOR WIDTH and LENGTH.
If there are multiple coherent noise trains at different velocities, for efficiency, and better
signal preservation, you may want to try to attenuate the whole range of noise trains in
one pass as in the following example.
***************************************************************************
* DLOOP 1
* RUNEX DD FILE=local:(PROJDIR)/CNASE/cnase1_INnmo,
Y=(SP_NB)
**
* CNASE NA DD FF FREQ(1, 2, 20, 25),MAXOFF10000,B2,NT120,
CORRIDOR(LENGTH200),
NOISEV(4200,4300,8000,8500),
* ENDLP
* DLOOP 2
* WUNEX FF FILE=local:(PROJDIR)/CNASE/+
cnase1_OUT
*******************************************************************************
* PROCS X(YB1)
For more aggressive noise attenuation (see Table 3), you can try a cascaded approach
using the PASSBND parameter.
***************************************************************************
* DLOOP 1
* RUNEX IB FILE=local:(PROJDIR)/CNASE/cnase1_INnmo,
Y=(SP_NB)
**
* CNASE NA IB OB FREQ(1, 2, 20, 25),MAXOFF10000,B2,NT120,
CORRIDOR(LENGTH1000, WIDTH1000),
NOISEV(4200,4300,8000,8500),
CASCADE(PASSBND740, SLOPE200),
* ENDLP
* DLOOP 2
* WUNEX OB FILE=local:(PROJDIR)/CNASE/+
geovation User Guide CNASE 17
CGG CNASE
signaloutput
*******************************************************************************
* PROCS X(YB1)
Example 2. CNA/CSE quality control
The output buffer should be checked for residual output noise and the auxiliary buffer
should be checked for residual signal left in the noise estimate. There is a trade-off between
these two alternatives when there is overlap between signal and noise in FK space. For
example, higher velocities can be used to reject more of the noise in CNA, but may start
to attenuate the signal as well.
The following example applies CSE to a single NMO-corrected muted shot and outputs both
a signal and noise files for quality control. The application times are from 500 to 2000 ms:
***************************************************************************
* DLOOP 1
* RUNEX IB FILE=local:(PROJDIR)/CNASE/nmo_corrected,
Y=(SP_NB)
**
* CNASE SE IB OB FREQ(2, 4, 25, 30),MAXOFF10000,B2,NT120,
CORRIDOR(LENGTH1000, WIDTH1000),
OS1=AB,
MUTE,SIGNALV(16000, 17000),THRESH5,
GATE(O0.0, S500, E2000)
* ENDLP
* DLOOP 2
* WUNEX OB FILE=local:(PROJDIR)/CNASE/signaloutput
* WUNEX AB FILE=local:(PROJDIR)/CNASE/noiseoutput
* ENDLP
***************************************************************************
* PROCS X(YB1)
Example 3. Two-pass CNA for positive and negative dips using DOWNDIP
Normally, coherent noise trains have a positive dip where the arrival times increase with
increasing offset. If, however, there are negative dips present also, a second pass of CNA
with the DOWNDIP parameter can be used as shown in the following job. The positive and
negative dip noise traces can be summed together for QC purposes.
***************************************************************************
* DLOOP 1
* RUNEX IB FILE=local:(PROJDIR)/CNASE/data
**
* CNASE NA IB OB FREQ(1, 2, 40, 50),MAXOFF10000,B2,NT120,
OS1=AB,
CORRIDOR(LENGTH12, WIDTH12),
NOISEV(130, 150, 230, 300),
THRESH5,ZERO,WINDOW(400, 100),
* ENDLP
* DLOOP 2
* WUNEX AB FILE=local:(PROJDIR)/CNASE/+
positivdip_noise
****
* CNASE NA OB OC FREQ(1, 2, 40, 50),MAXOFF10000,B3,
OS1=AC,
CORRIDOR(LENGTH12, WIDTH12),
NOISEV(130,150, 230, 300),
THRESH5,ZERO,WINDOW(400, 100),
DOWNDIP,
* ENDLP
* DLOOP 3
* WUNEX AC FILE=local:(PROJDIR)/CNASE/+
negdip_noise
* ENDLP
*******************************************************************************
geovation User Guide CNASE 18
CGG CNASE
* PROCS X(YB1)
Example 4. SE option plus second option
***************************************************************************
* DLOOP 1
* RUNEX ++ FILE=local:(PROJDIR)/CNASE/nmo_corrected_shots,
Y=(SP_NB)
**
* CNASE SE 2 == 04 FREQ(1, 2, 25, 30),
CORRIDOR(LENGTH100, WIDTH100),
OS1=NN,
SIGNALV(8000, 25000),
NOISEV(250, 300, 4000, 4500), OVERSMP10.0,
MAXOFF10000,
WINDOW(0, 7000),
THRESH5,
B4,NT120,
* ENDLP
* DLOOP 4
* WUNEX 04 FILE=local:(PROJDIR)/CNASE/output_shots
* ENDLP
***************************************************************************
* PROCS X(YB1)
geovation User Guide CNASE 19