Hi all,
Please, I have a cmp gather that is aliased with the aliasing not showing up in shotreceiver gather.
Could anyone advice on the best way to remove this in the cmp domain. I am thinking of running fk muting in the taup domain on the fk spectrum of the cmp gather in question, thus muting out the aliased part of the fk spectrum in taup.
Reason for this line of thought is owing to the localization of the fk spectrum of taup gather in the positive part of the fk spectrum but with the aliased cmp data the raparound shows up in the negative part of the fk spectrum.
Thanks for any advice on this.
Regards
CMP aliasing and removal
Re: CMP aliasing and removal
Sorry  missed this one, here's a belated reply.
Generally speaking you'll find that you have more CDPs than you have shots. That's a result of the "natural" CDP spacing being half the group interval.
So  its not uncommon to have (say) 25m between shots, and 25m between receivers, so that you wind up with 2 CDPs per shot.
Now in a shot record that means you have a 25m spacing between each of the traces, but in the CDP domain you will end up with twice that, or 50 metres, because each CDP has half the traces of the shot gather. You also have an uneven distribution of offsets.
One CDP will have all of the odd channels (1,3,5,7,9...) and the next will have all of the even channels (2,4,6,8,10...)
So  that gives you two problems; firstly you might have dipping energy that is not aliased in the shot domain, but is in the CDP domain. And secondly, if you want to do any transform based velocity discrimination work your CDPs have different offset ranges. That tends to get computationally expensive.
There's a few strategies that people have used for this.
The first is "super gathers", where you "mix" two or more CDPs together. This works okay when you have 2 CDPs per shot,m but higher order supergathering can give some off effects in the TauP, Radon or FK domains; where you have high noise levels you will see an imprint of the transform "bounds" and this can create a "blocky" appearance on a VD (colour) display when you stack the data, and that knocks onto the migration and so on.
The second thing is to interpolate shot points. This became very common with flipflop 3D shooting; you essentially infill a missing shot in bwteen the existing ones. As the shot point spacing is often dictated by record length (a ship goes about 25 metres in 810 seconds) this can be helpful. This means you can go from a 25m shot point to a 12.5m Sp spacing, and hence have the same number of traces (and offset distribution) in a Shot record or a CDP gather. It's very effective.
The third thing is to spatially resample the shotrecord to give a higher (or indeed lower) fold, to get that spacing the same. While a lot of data is recorded with 12.5m or even 6.25m groups, this is really oversampling spatially; a quick calculation on Fresnel zones and CDP spacing (as well as vessel movement) indicates that for most purposes the implied 6.25 or 3.125m CDP spacing is not very useful; you can't get that resolution from the data.
What a 12.5m or 6.25m group interval does mean is that the direct and other watervelocity linear arrivals (or indeed steep backscatter) will not be alaised at high frequencies (over 60=80Hz at 1500m/s); this means that you can use this high spatial resolution wavefield to remove aliased noise, then resample spatially in the same way you would temporally.
So, assuming you had 50m shot spacing, and 640 channels at 12.5m groups
 the native data has 8 CDPs for every shot, which will be hard work and aliased
 work with shot records at 12.5m group interval or even 6.25m group interval in the TauP domain to remove linear, water velocity noise
 spatially resample the shot records to 25m groups (320 channels)  so you have 4 CDPs for each shot
 interpolate the shot records to give 12.5m spacing (two interpolations)
The result is 320 fold CDPs and 320 fold shots, with no spatial aliasing.
Of course that's just one strategy. I've used this on 1970s data with 100m group internals and 24 channels to build up 72fold CDPs; that allowed us to run Radon and TauP work, and shifted the understanding of the basin significantly.
Hopefully what you have is nowhere as extreme as this, but the ability to remove steepdip high frequency noise and still preserve detail, especially on legacy data  is extremely useful.
Good luck!
Generally speaking you'll find that you have more CDPs than you have shots. That's a result of the "natural" CDP spacing being half the group interval.
So  its not uncommon to have (say) 25m between shots, and 25m between receivers, so that you wind up with 2 CDPs per shot.
Now in a shot record that means you have a 25m spacing between each of the traces, but in the CDP domain you will end up with twice that, or 50 metres, because each CDP has half the traces of the shot gather. You also have an uneven distribution of offsets.
One CDP will have all of the odd channels (1,3,5,7,9...) and the next will have all of the even channels (2,4,6,8,10...)
So  that gives you two problems; firstly you might have dipping energy that is not aliased in the shot domain, but is in the CDP domain. And secondly, if you want to do any transform based velocity discrimination work your CDPs have different offset ranges. That tends to get computationally expensive.
There's a few strategies that people have used for this.
The first is "super gathers", where you "mix" two or more CDPs together. This works okay when you have 2 CDPs per shot,m but higher order supergathering can give some off effects in the TauP, Radon or FK domains; where you have high noise levels you will see an imprint of the transform "bounds" and this can create a "blocky" appearance on a VD (colour) display when you stack the data, and that knocks onto the migration and so on.
The second thing is to interpolate shot points. This became very common with flipflop 3D shooting; you essentially infill a missing shot in bwteen the existing ones. As the shot point spacing is often dictated by record length (a ship goes about 25 metres in 810 seconds) this can be helpful. This means you can go from a 25m shot point to a 12.5m Sp spacing, and hence have the same number of traces (and offset distribution) in a Shot record or a CDP gather. It's very effective.
The third thing is to spatially resample the shotrecord to give a higher (or indeed lower) fold, to get that spacing the same. While a lot of data is recorded with 12.5m or even 6.25m groups, this is really oversampling spatially; a quick calculation on Fresnel zones and CDP spacing (as well as vessel movement) indicates that for most purposes the implied 6.25 or 3.125m CDP spacing is not very useful; you can't get that resolution from the data.
What a 12.5m or 6.25m group interval does mean is that the direct and other watervelocity linear arrivals (or indeed steep backscatter) will not be alaised at high frequencies (over 60=80Hz at 1500m/s); this means that you can use this high spatial resolution wavefield to remove aliased noise, then resample spatially in the same way you would temporally.
So, assuming you had 50m shot spacing, and 640 channels at 12.5m groups
 the native data has 8 CDPs for every shot, which will be hard work and aliased
 work with shot records at 12.5m group interval or even 6.25m group interval in the TauP domain to remove linear, water velocity noise
 spatially resample the shot records to 25m groups (320 channels)  so you have 4 CDPs for each shot
 interpolate the shot records to give 12.5m spacing (two interpolations)
The result is 320 fold CDPs and 320 fold shots, with no spatial aliasing.
Of course that's just one strategy. I've used this on 1970s data with 100m group internals and 24 channels to build up 72fold CDPs; that allowed us to run Radon and TauP work, and shifted the understanding of the basin significantly.
Hopefully what you have is nowhere as extreme as this, but the ability to remove steepdip high frequency noise and still preserve detail, especially on legacy data  is extremely useful.
Good luck!
Re: CMP aliasing and removal
Hi GuyM,
Thanks for your so many feedback.
The data I am processing has 25m shot interval, 12.5m group spacing, 240 traces and is spatially aliased in the cmp domain after muting in taup domain. I have applied shot interpolation to take it to 12.5m shot interval, 25m group spacing,12.5m cdp interval and 120traces but yet the spatial aliasing is not resolved completely. I still see reflections that are wavy.
Please, is there any other way to resolve this? I actually observed some improvement in the migrated section using the shot interpolated data, although i am reworking my velocities having used cascaded migration in my flow.
Regards,
Nestorkid
Thanks for your so many feedback.
The data I am processing has 25m shot interval, 12.5m group spacing, 240 traces and is spatially aliased in the cmp domain after muting in taup domain. I have applied shot interpolation to take it to 12.5m shot interval, 25m group spacing,12.5m cdp interval and 120traces but yet the spatial aliasing is not resolved completely. I still see reflections that are wavy.
Please, is there any other way to resolve this? I actually observed some improvement in the migrated section using the shot interpolated data, although i am reworking my velocities having used cascaded migration in my flow.
Regards,
Nestorkid
Re: CMP aliasing and removal
So can you work me through the sequence in order?
My recommendation would be:
 antialias filter and resample if appropriate
 amplitude recovery (spherical divergence and linear gain)
 swell noise attenuation, bad trace edits etc
 interpolate to 480 channels, 6.25m trace space
 TauP domain muting to remove aliased linear noise
 deconvolution in the TauP domain (long gap, typically 48ms, operator perhaps 400ms, but test)
 inverse Taup transform to 12.5m group interval spacing
 Kfilter (0.5 nyquist) and trace drop to 25m group interval
 interpolate shots to 12.5m spacing
So now you have 120 fold shots and 120 fold CDPs, which at 12.5m apart
That should address your aliasing.
On a stack, the spatial aliasing is like a diagonal "checkerboard" crosshatching pattern; you can check for spatial alaising on a shot or CDp by looking at the FK spectrum and see if it is wrapping round.
Still  worth baring in mind that not all problems can be easily solved; a 240 x 12.5m streamer is on 3000m; that's pretty short by modern standards for one thing; might be hard to get noise to stack out, especially in deep water.
My recommendation would be:
 antialias filter and resample if appropriate
 amplitude recovery (spherical divergence and linear gain)
 swell noise attenuation, bad trace edits etc
 interpolate to 480 channels, 6.25m trace space
 TauP domain muting to remove aliased linear noise
 deconvolution in the TauP domain (long gap, typically 48ms, operator perhaps 400ms, but test)
 inverse Taup transform to 12.5m group interval spacing
 Kfilter (0.5 nyquist) and trace drop to 25m group interval
 interpolate shots to 12.5m spacing
So now you have 120 fold shots and 120 fold CDPs, which at 12.5m apart
That should address your aliasing.
On a stack, the spatial aliasing is like a diagonal "checkerboard" crosshatching pattern; you can check for spatial alaising on a shot or CDp by looking at the FK spectrum and see if it is wrapping round.
Still  worth baring in mind that not all problems can be easily solved; a 240 x 12.5m streamer is on 3000m; that's pretty short by modern standards for one thing; might be hard to get noise to stack out, especially in deep water.

 Similar Topics
 Replies
 Views
 Last post

 9 Replies
 4344 Views

Last post by GuyM