GEPH316 Field Data: Frequency filtering data

GEPH316 Seismic Data Processing

Field Data Workshop 2000

11 Frequency filtering data


Previous: Frequency filter design	Up: Contents	Next: Statics

Frequency filtering

In this exercise we explore the effect of frequency filtering on the real seismic data. If you feel uncomfortable with the ideas of Fourier analysis, it is useful to experiment with various forms of filters and discover by experiment what filtering does to the actual seismic data.

Low-pass filtering

For this section, the shot at 05m provides a suitable set of test traces. A copy of the section is included in the handout. Separate out the seismic shot s05 and inspect the character of the noise in the data traces.

suwind < line > s05 key=sx min=05 max=05
suxwigb < s05  title="s05" label1="Time (s)" label2="Trace numbers" &

Which traces are most noisy and what is the character of the noise?
Are there some unwanted parts of the real seismic arrivals that have a different frequency character from the desired seismic arrivals?
What frequencies are present in the useful data and what frequencies are present in the noise? Produce the spectra to check your conclusions.

suspecfx < s05 | suxwigb title="s05 spectra" label1="Frequency (Hz)" label2="Trace numbers" &

Most of the sharp noise seems to occur above about 200Hz, so apply the following low-pass filter to try to cut out the high frequency noise. The point half-way down the filter taper slope will be 205 Hz.

sufilter < s05 > s05.low1 f=190,200,210,220 amps=1.,1.,0.,0.
suxwigb < s05.low1  title="s05 low-pass filtered at 205 Hz" label1="Time (s)" label2="Trace numbers" &

You should be able to see the low frequency seismic arrivals more clearly on the low-pass filtered traces. Write a short description of the improvement in the appearance and clarity of the traces.
It is often difficult to visualise the full character of the noise that has been removed. It is best to apply the reverse of the filter to the data, which should just leave the noise, in order to see exactly what has been removed. In this case apply the following high-pass filter to see all the high frequency noise.

sufilter < s05 > s05.high1 f=190,200,210,220 amps=0.,0.,1.,1.
suxwigb < s05.high1 title="s05 high-pass filtered at 205 Hz"  label1="Time (s)" label2="Trace numbers" &

On the high-pass filtered traces you should be able to see the high frequency noise that has been removed.
Note that a small proportion of the refracted wave energy has been removed as well.
Why has the refracted wave energy only been removed from the first few traces?
- Consider the implications of the variation of frequency absorption with distance. The high frequency content of the direct waves will rapidly be attenuated with distance.
Look at the strong high frequency arrival at about 0.100s on trace 11. It is the strongest of a series of pulses stretching back towards time zero at zero distance (the axis of the box).
Measure the first arrival time of this pulse on trace 12, which is 36m from the shot point. What is the velocity of this arrival? What is this seismic event?
- The velocity should be roughly 36m/0.105s = 343ms^-1.
- This is the velocity of sound in air, which is 330ms^-1, so this is likely to be the air-wave generated by the sound of the sledge-hammer hitting the aluminium plate.
- Such an air wave is quite common in hammer seismics and it has a frequency of about 300Hz. It can be removed from seismic traces using a high-cut filter with a cut-off frequency of about 200Hz.
Inspect the high-frequency noise spectra to check your conclusions.

suspecfx < s05.high1 | suxwigb title="s05 spectra high-pass filtered at 200 Hz" label1="Frequency (Hz)" label2="Trace numbers" &

See whether the air wave can be observed more clearly on the traces using a high-cut filter with a cut-off frequency of about 300Hz.

sufilter < s05 > s05.air f=270,290,310,330 amps=0.,0.,1.,1.
suxwigb < s05.air title="s05 air wave  high-pass filtered at 300 Hz" label1="Time (s)" label2="Trace numbers" &

Use the following lines if you want a series of print-outs of the filtered traces.

supswigp < s05.low1  title="s05 low-pass filtered at 200 Hz" label1="Time (s)" label2="Trace numbers" | lp -d mono
supswigp < s05.high1 title="s05 high-pass filtered at 200 Hz"  label1="Time (s)" label2="Trace numbers" | lp -d mono
supswigp < s05.air   title="s05 air wave high-pass filtered at 300 Hz" label1="Time (s)" label2="Trace numbers" | lp -d mono

Clear away the unwanted xwigb graphical windows.

zap xwigb

High-pass filtering

Produce and examine the complete s05 spectrum.

suspecfx < s05 | suxwigb title="s05 spectra" label1="Frequency (Hz)" label2="Trace numbers" &

What frequency are the surface waves? What frequency limit could be used to cut off the lower frequency surface waves but not damage the higher frequency direct wave content?
- You may find that the frequency limit will vary from trace to trace.
- Probably a value around 60Hz is the best average discrimination frequency.
Design a high-pass filter to kill the surface waves, similiar to the one below which has a cut-off at 55 Hz half-way down the filter taper. Apply it to the s05 shot and inspect the resulting wiggle trace.

sufilter < s05 > s05.high2 f=40,50,60,70 amps=0.,0.,1.,1.
suxwigb < s05.high2  title="s05 high-pass filtered at 55 Hz"  label1="Time (s)" label2="Trace numbers" &

Have you reduced the later surface waves?
Have you created unwanted effects, particularly on traces 10 to 12? What might have caused these unwanted effects?
In order to investigate the problems, design the complementary low-pass filter to show the surface waves that have been removed. Apply it to the s05 shot and inspect the resulting wiggle trace.

sufilter < s05 > s05.low2 f=40,50,60,70 amps=1.,1.,0.,0.
suxwigb < s05.low2  title="s05 low-pass filtered at 55 Hz"  label1="Time (s)" label2="Trace numbers" &

Is it now possible to decide what might have caused these unwanted effects?
- Because the surface waves stray off the end of each of the later traces, the frequency domain filtering tends to affect the start of the traces as well.
Carefully inspect this picture, which is predominantly of surface waves. The later parts of the traces appear to be very smooth oscillations.
Are there any 'unsmooth' oscillations or additional peaks at the front ends of the surface waves? What may be causing these effects?
- Some traces appear to have extended or double troughs.
- Some traces appear to have leading extended or double peaks.
- This position on the traces is suitable for a refracted or reflected arrival. Hence it may be dangerous to remove all these surface and other waves by using a low-cut filter.
- On large scale seismic sections, this should be less of a problem because of the scale and frequency band of the reflections.

Use the following lines if you want a series of print-outs of the filtered traces.

supswigp s05.high2 title="s05 high-pass filtered at 45 Hz"  label1="Time (s)" label2="Trace numbers" | lp -d mono
supswigp s05.low2  title="s05 low-pass filtered at 45 Hz"  label1="Time (s)" label2="Trace numbers" | lp -d mono

Clear away the unwanted xwigb graphical windows.

zap xwigb

Filtering the complete line

What filtering is it safe to apply to all the traces in the line in order to remove high frequency noise?
Apply the following band-pass filter, with cut-offs below 60Hz and above 300Hz half-way down the taper, to the complete seismic line to reduce unwanted high frequency noise.
Inspect the output section.


subfilt < line.cdp > line.filter  fstoplo=50 fpasslo=70 fpasshi=240 fstophi=260 verbose=1
suxwigb < line.filter  title="Line band-pass filtered betwen 60 and 250 Hz"  label1="Time (s)" label2="Trace numbers" &

Remove the old windows and unwanted files using:

zap xwigb
rm s05.*
rm line.cdp
ls


Previous: Frequency filter design	Up: Contents	Next: Statics

Go to <--->