Version: V1.0 -2.0
Two dimensional filtering techniques use the spatial characteristics of the input seismic data to supress unwanted noise and enhance the continuity of desired events. TP2DF contains :-
These two algorithms are fundamentally different in that FX Decon adapts its filters automatically to the data, whereas XT Filter has fixed characteristics determined by initial parameters.
FX Deconvolution is a filter for enhancing the continuity of linearly dipping events. Normally used on post-stack data, it preserves the most dominant dipping energy within a 2-D window, while suppressing random noise and very short or non-linearly dipping events (i.e those having curvature). Go to FX deconvolution setup The input section is divided up into a series of rectangular, overlapping windows in space and time. A new F-X filter is designed and applied individually for each space - time window and the results blended across the window overlap zones.
At every frequency F, a symmetric convolutional prediction filter is designed which, given a set of 2n+1 complex amplitude vs frequency spectra from the input window, will predict the nth trace's spectral component from the n traces on either side. The filter is designed using the least squares method, similar in priciple to that used in classical predictive deconvolution, but with the following important differences.
If a window of seismic data contains events predominantly having the same dip, or linear dips, the traces are linearly predictable in space. Any energy that is not predictable, in the across trace direction, may be assumed to be noise. In the frequency (F) domain, the characteristic of a single noise free dipping reflector is that, at any frequency, the power spectrum remains constant from one trace to the next, whereas the phase spectrum changes linearly, i.e. it is predictable in the across trace (X) direction.
XT-Filter is a simple but effective time variant 2-dimensional dip filter for reducing the effect of noise in cases where the predominant dip of the noise is different from that of the data (go to XT filter setup).Such noise can take several forms, e.g.
Noise in these all categories can often be separated to some degree from signal on the basis of dip. For example, an isolated noise spike on a single trace will have much of its energy at high frequencies and wave numbers and will be attenuated relative to continuous events by a filter which rejects high dips.
Unlike a conventional F-K filter, the XT filter is phase insensitive and cannot distinguish between positive and negative dips. However, it does not suffer from the edge effects, wrap-around and spatial aliasing phenomena which can be a problem with F-K filters.
Remember that if events in the input data are spatially aliased, then they cannot normally be suppressed with a conventional X-T or F-K filter.
Spatial aliasing exists where high frequencies are combined with steep dips, such that at a frequencyfHz, the maximum apparent apparent dip dipmax exceeds
dipmax = 500 /f msec per trace
For example, if your data contains frequencies of 50Hz, dips greater than 10 msec/trace will be spatially aliased.
FXDECON F-X Deconvolution
XTFILT X-T Filter
X-T Filter time Variant Dip for XTFILT
Process ID
(type string) TracePrep
standard process ID
Process title
(type string) A brief
description of an instance of the process
Filter type
(type option, WIENER )
In this version of the program, only one type of filter is allowed, a
symmetrical (zero phase) Wiener prediction operator.
White noise %
(type single, limits
0.0 to 100.0)
This is an optional constant which can be added to the zero
lag value of the autocorrelation function used in calculating the filter
coefficients. No white noise will allow the filter to be maximally adapted to
the data. 100% white noise will degrade the filter to be an unweighted mix of
the traces. Use this parameter to control the amount of mixing. 50% white noise
will give a very mixed appearance to the data.
Trace window, traces
(type long
integer, limits 10 to 2000)
Owing to spatial and temporal variations in dip,
it is necessary to limit the window width over which FXD operators are
calculated. For more laterally uniform sections, a larger number should be
used. This number of traces will be used in calculating each filter. The filter
is a two sided, symmetrical filter - the number of output points convolved with
the full length of the filter is NtrW-NFilt.
Trace blend, traces
(type long
integer, limits 0 to 2000)
Adjacent trace windows will be linearly blended
across a zone this number of traces wide.
Filter width, traces
(type long
integer, limits 2 to 50)
The number of coefficients in the prediction
filter. Each trace will be predicted from this number of adjacent traces, half
on each side.Increasing the filter width improves the efficiency of the filter,
but more input traces are involved in the calculation of each output trace,
possibly causing a more mixed appearance.
Time window, msec
(type single,
limits 100.0 to 6000.0)
The section is divided vertically into time windows
of this length. Choose a time window so that dips will not change significantly
over the length of the window. If the window is too long, events in the shallow
part of the section may be contaminated by noise deeper down.
Time blend zone, msec
(type single,
limits 0.0 to 1000.0)
The time windows are overlapped and blended top and
bottom by this number of milliseconds, to give the output section a more
uniform appearance. Note that the time window parameter above includes the
blend zones. Maximum blend zone length is half the time window.
Low cut frequency, Hz
(type single,
limits 0.0 to 1000.0)
Minimum frequency used in designing filters and
generating output. Frequencies below this are ignored. Set it to the lowest
frequency expected in the data.
High cut frequency, Hz
(type
single, limits 4.0 to 1000.0)
Maximum frequency used in designing filters
and generating output. Frequencies above this are ignored. Set it to the
highest frequency expected in the data.
Addback from input %
(type single,
limits 0.0 to 100.0)
Percentage of the input which is added back into the
output, to produce amore "natural" appearance.
RMS o/p amplitude
(type single,
limits 0.0 to 10000.0)
On output, traces will be RMS balanced to this
value.
Process ID
(type string) TracePrep
standard process ID
Process title
(type string) A brief
description of an instance of the process
Min Freq Hz :
(type single, limits 0
to 250)
This is the lower frequency limit in Hz for the input passband. Set
this to a value slightly below the estimated lowest frequency present in the
input.
Max Freq Hz :
(type single, limits 0
to 250)
This is the upper frequency limit in Hz for the input passband. Set
this to a value slightly above the estimated highest frequency present in the
input.
The XT Filter will limit the bandwidth of the output to the passband
determined by Min Frequency and Max Frequency
No. of Bands :
(type long integer,
limits 1 to 20)
The range between the Min and Max frequencies will be split
up into this number of bands, on a logarithmic frequency scale, so that the
bands are of equal width in octaves. This parameter determines the accuracy
with which the linear dependence of spatial cutoff versus frequency is
implemented. The filters used for the band splitting are 24dB per octave 2nd
order Butterworth type.
Because of the smoothing caused by the finite
slopes of the filter cutoffs, increasing the number of bands will only improve
the performance of the filter up to a point, after which the only difference
will be an increase in run time.
You should aim for a band width of around
half an octave - in practice this means 8 bands for a 5 to 80Hz (4 octave)
passband, or as few as 4 bands for a 10 to 40Hz (2 octave) passband.
Band Overlap :
(type single, limits
0.5 to 1.5)
This parameter determines the degree of band overlap, required
to give a smooth frequency passband to the output data. A value of between 1.0
and 1.2 is recommended. You can safely leave this parameter set to 1.0 in
nearly all cases.
O/p RMS Ampl. :
(type single, limits
0.0 to 10000)
The RMS amplitude for the output traces, after addback. If you
set this value to zero, no RMS balance will be applied.
On this page, enter the minimum and maximum dips to be passed by the XT filter, as a function of time. You can enter up to five time control points for the filter. Parameters are linearly interpolated between control points. At times less than that for the first control point, the parameters are set equal to their values at the control point. Below the last control point, the parameters are set equal to their values at the control point.
If all the control point values are zero, the control point will be ignored.
Time
(type single, limits 0.0
to 10000)
The time from which this control point applies
Minimum Dip
(type single,
limits 0.0 to 100.0)
The minimum dip to accept, in msec per trace spacing,
as measured from the input section. Normally set to zero. If you set this to a
value other than zero, horizontal events and dips up to this value will be
strongly attenuated.
Maximum Dip
(type single,
limits 0.0 to 100.0)
The maximum dip to accept. Set to a value close to the
maximum dips that you want to keep on your section, usually between about 2 and
10 msec per trace.
Addback
(type single, limits
0.0 to 100.0)
This allows a percentage of the input data to be mixed back
into the output traces. Between 10 and 80 percent will restore a more "natural"
look to the more "mixed" output that may otherwise result..
FX window too small The combination of filter width and blend zone was wider than the chosen trace window
X-T filter times must increase The values entered in the X-T time variant dip page must increase monotonically, from top to bottom
No XT filter times were specifiedAll the time on the X-T time variant dip page were zero.
X-T filter forward pass complete A message to show that the forward pass of the X-T filter is complete
X-T filter reverse pass complete A message to show that the reverse pass of the X-T filter is complete