seismic processing?

Oil and natural gas exploration -- geology and geophysics
Post Reply
liath Fahad
Silver Member
Silver Member
Posts: 23
Joined: Tue Mar 29, 2016 1:34 pm

seismic processing?

Post by liath Fahad »

Hi, I used the corner filter(band pass) with this frequency values (8,10,45,50). I get ringing effect on the data. my questions are:
1. what is the name of this effect and the cause of it?
2. should be removed and how?
3.on what basis the frequency corner values should be chosen?
4. what are the visual features of seismic data make us decided the processing steps.
5. when I change the frequency values(10,15,45,50) the effect reduced as in pic 3.
6. what is the cause of resolution difference of the two shots as in pic3.
Attachments
Pic 3
Pic 3
Pic 2
Pic 2
Pic 1
Pic 1

GuyM
VIP Member
VIP Member
Posts: 625
Joined: Sat Mar 24, 2012 11:35 pm

Re: seismic processing?

Post by GuyM »

TLDR:

(a) you are seeing an impulse response function from the filter because you have not muted the direct and refracted arrivals and your poor filter design on the high cut has a lot of ringing

(b) You need to test the filter band pass range to attack the kind of noise you want to remove. Make filter test panels eg
5-10-15-20, 10-15-20-25, 15-20-25-30, 20-25-30-35 towards the higher end use 10Hz steps 35-35-45-55 etc.

(c) plot the differences between filtered and unfiltered panels to see what is being removed from the data

(d) cutting high frequencies will lower resolution

(e) you should only remove noise, not reflected signal

====================================================================================================

The "ringing effect" you are seeing is a result of the filter operator and the trace amplitudes.

if you apply a filter to a "spike" (a single sample long amplitude of one) you will see the filter operator - this is called the "impulse response function"
If you have some very strong amplitudes with some very weak ones underneath or around them then with *any* processing operation (FK, migration, filters, deconvolution) you will see that impulse response function.

A filter has a "ringing operator" when the slopes are too small for the filter to be stable. This is called "Gibbs Phenomena" - you need to increase the slope length to make it more stable. In this case the 50HZ-55HZ on the high cut is too small; try 50-60Hz or maybe 55Hz-70Hz to stabilize.

The other part of the solution is to manage your amplitudes better; the general approaches are:

- editing spikes or very noise traces prior to any processing through peak, RMS, signal and noise graphical analysis
- amplitude recovery via Spherical divergence and linear gains (usually in dB/second)
- removing the strong direct and refraction arrivals with a refraction mute
- using surface-consistent amplitude scaling and balancing
- using mild "rolling gate balance' - so balance gates 0-1000ms, 500-1000ms and so on
-using a removable AGC around a process; so you save the AGC scalars then remove them



Your other questions come down to "why am I applying this filter?"

When we are cleaning up seismic, the basic method we use is to throw away unwanted energy.

- the least sophisticated way to do this is throw it away - a mute, trace edit, or editing part of a trace
- the next worst way is to look just at the frequency content of a trace to see if that splits signal from noise

A bandpass filter is just a frequency domain mute.

So - FIRST you need to TEST the filter.

- run a rolling series of filters along the data
- run band pass (0-5-10-15, 5-10-15-20, 10-15-20-25 etc etc) filters
- look where you have signal, and where you have noise
- use this to run expanded filter tests around your desired low and high cuts

So - choose the high cut and low cut accordingly.

You are looking for reflections (to keep) and the noise being removed is either coherent (eg ground roll) or is just random.

As for filter slopes - the first and last point - you need to read up on your theory. Sharp slopes produce unstable filters that have a lot of ringiong.

There's some stuff on filter design and ringing here:
http://seismicreflections.globeclaritas ... ample.html

Here's a discussion of noise edits and mutes, its for marine but the same stuff applies:

http://seismicreflections.globeclaritas ... tudes.html

liath Fahad
Silver Member
Silver Member
Posts: 23
Joined: Tue Mar 29, 2016 1:34 pm

Re: seismic processing?

Post by liath Fahad »

thank you, GuyM after I read your note I check the data. The data was not muted in a proper way. The first break picking was bad so I repeat the picking and do QC in a new method that I'm not sure if it is works?? I plotted the traces of one shot without the effect of time delay by overlapping them as in pic 1. I get what I think the amplitude spectrum of the shot then picked the first break arrival as in pic 2 and 3. I muted above the red line in pic 2 and 3 in pic 4 I plot the amplitude spectrum of the muted part and in pic5 the residual part.
Attachments
pic 5
pic 5
pic 4
pic 4
pic 3
pic 3
pic 2
pic 2
pic 1
pic 1

GuyM
VIP Member
VIP Member
Posts: 625
Joined: Sat Mar 24, 2012 11:35 pm

Re: seismic processing?

Post by GuyM »

Sorry - don't quite understand what you are aiming at there.

Picking the first breaks - for geometry QC and refraction analysis - has nothing to do with muting the data prior to any signal processing to avoid ringing effects.

We'd usually :

- apply geometry to the data
- test gain recovery, usually a combination of Spherical divergence and linear dB/second gains
==> use amplitude decay curves and panel comparisons for a more quantitative assessment of what is right
- run NOISEQC (shallow and deep peak and amplitudes, signal: noise ratio) and graphically display these
==> trace edits from these displays
- pick first breaks and run refraction statics (and QC geometry at the same time)
- apply combined elevation and near surface corrections
- pick (or design based on Linear-Move-Out) a mute for the direct and refracted arrivals

If you think about what we are doing with the combination of elevation and refraction statics, we are essentially stripping gout the weathering layer, and then pouring in a "magical" liquid rock of constant velocity, that replaces the weathered layer and fills in the bumps/dips in the surface topography/elevation.

This means that the direct/refracted arrivals will be much more linear after these corrections, and hence much easier to mute accurately.

Once we have that first part of the structural solution we can move on to the signal processing side of things - addressing ground roll, deconvolution, removing random noise - before going on to do the "structural" bits (veloicties and residual statics, then DMO/PreSTM and so on)

At least - that's how we tend to do things.

You don;t want to filter or mute the data before doing your first break picking, unless the data is terrible and noisy.
Your data looks pretty good!

Guy

GuyM
VIP Member
VIP Member
Posts: 625
Joined: Sat Mar 24, 2012 11:35 pm

Re: seismic processing?

Post by GuyM »

Oh -

In terms of first break QCs I tend to apply the first breaks as a shift with an additional +100ms shift applied.
(that's a push button in the s/ware for us)

That should give you flat "tramlines" and you can see how consistent the picking is.

Toggle between displaying the data based on the shot location (Shot peg or shot station) and the receiver location (receiver peg or station) as an additional QC.

Here's the data autopicked:
pix1.jpg
Here's the data flattened - you can see immediately where the autopicks are not quite right
pix1.jpg
In terms of muting - remember that in this case it was probably the steep high-cut slope of the filter causing an issue, as opposed to just the mute.

<EDIT - the picture embedding is a bit screwy, but you can see the flattened picks picture I hope!>
Attachments
pix2.jpg

Post Reply
  • Similar Topics
    Replies
    Views
    Last post