f-k filters
TRANSCRIPT
f-k filters
1
Mitch Withers, Res. Assoc. Prof., Univ. of Memphis
2
Consider a 2-d delta function, πΏ π₯, π¦ = &1, π₯ πππ π¦ = 00, π₯ ππ π¦ β 0
3
Consider a 2-d delta function, πΏ π₯, π¦ = &1, π₯ πππ π¦ = 00, π₯ ππ π¦ β 0
/!"
"/!"
"πΏ π₯, π¦ ππ₯ππ¦ = 1 A point source
4
Consider a 2-d delta function, πΏ π₯, π¦ = &1, π₯ πππ π¦ = 00, π₯ ππ π¦ β 0
/!"
"/!"
"πΏ π₯, π¦ ππ₯ππ¦ = 1
Similarly, the 2-d sifting property is, /
!"
"/!"
"πΏ π₯ β π₯#, π¦ β π¦# π π₯, π¦ ππ₯ππ¦ = π π₯#, π¦#
π¦
π₯#π₯
πΏ π₯ β π₯#
A point source
a line source sifts values of π π₯, π¦ at π₯# for all π¦.
5
Consider a 2-d delta function, πΏ π₯, π¦ = &1, π₯ πππ π¦ = 00, π₯ ππ π¦ β 0
/!"
"/!"
"πΏ π₯, π¦ ππ₯ππ¦ = 1
Similarly, the 2-d sifting property is, /
!"
"/!"
"πΏ π₯ β π₯#, π¦ β π¦# π π₯, π¦ ππ₯ππ¦ = π π₯#, π¦#
π¦
π₯#π₯
πΏ π₯ β π₯#
A point source
a line source sifts values of π π₯, π¦ at π₯# for all π¦. /
!"
"πΏ π₯ β π₯# π π₯, π¦ ππ₯ = π π₯#, π¦
π₯#π₯
π¦
π π₯#, π¦
6
The mathematics of the 2-d FT is ambivalent to our choice of units so one choice can be time and space, a seismic reflection line for example.
Ξ¦ π, π = πΉ[π π‘, π₯ ] = /!"
"/!"
"π(π‘, π₯)π!$%&(()*+,)ππ‘ππ₯
7
The mathematics of the 2-d FT is ambivalent to our choice of units so one choice can be time and space, a seismic reflection line for example.
Ξ¦ π, π = πΉ[π π‘, π₯ ] = /!"
"/!"
"π(π‘, π₯)π!$%&(()*+,)ππ‘ππ₯
Where π is the familiar frequency in cycles/second and π is spatial frequency in cycles/length.
π₯
ππ is the wavelength
π = ./
is the spatial frequency
Wavenumber is πβ = 2ππ = %&/
similar to π = 2ππ.
8
The mathematics of the 2-d FT is ambivalent to our choice of units so one choice can be time and space, seismic reflection line for example.
Ξ¦ π, π = πΉ[π π‘, π₯ ] = /!"
"/!"
"π(π‘, π₯)π!$%&(()*+,)ππ‘ππ₯
Where π is the familiar frequency in cycles/second and π is spatial frequency in cycles/length.
π₯
ππ is the wavelength
π = ./
is the spatial frequency
Wavenumber is πβ = 2ππ = %&/
similarly to π = 2ππ.
We use πβ here to distinguish wavenumber from spatial frequency π though you may encounter π for both. Know which it is by context.
Spatial frequency not wavenumber.
9
A few familiar properties in 2-d
πΉ π π₯ β π₯#, π‘ β π‘# = πΊ(π, π)π!$%& +,!*()! Shifting property
10
A few familiar properties in 2-d
πΉ π π₯ β π₯#, π‘ β π‘# = πΊ(π, π)π!$%& +,!*()! Shifting property
πΉ π π₯, π‘ + ππ π₯ β π₯#, π‘ β π‘#= πΊ(π, π) 1 + ππ!$%& +,!*()!
Where π is a real scalar.Replication property (an echo)
11
A few familiar properties in 2-d
πΉ π π₯ β π₯#, π‘ β π‘# = πΊ(π, π)π!$%& +,!*()! Shifting property
πΉ π π₯, π‘ + ππ π₯ β π₯#, π‘ β π‘#= πΊ(π, π) 1 + ππ!$%& +,!*()!
Where π is a real scalar.Replication property (an echo)
πΉ π.(π₯, π‘) β π%(π₯, π‘) = πΊ.(π, π)πΊ%(π, π)
where
Convolution property (think f-k filter)
π. π₯, π‘ β π% π₯, π‘ = /!"
"/!"
"π. π’, π£ π% π₯ β π’, π‘ β π£ ππ’ππ£
12
A few familiar properties in 2-d
πΉ π π₯ β π₯#, π‘ β π‘# = πΊ(π, π)π!$%& +,!*()! Shifting property
πΉ π π₯, π‘ + ππ π₯ β π₯#, π‘ β π‘#= πΊ(π, π) 1 + ππ!$%& +,!*()!
Where π is a real scalar.Replication property (an echo)
πΉ π.(π₯, π‘) β π%(π₯, π‘) = πΊ.(π, π)πΊ%(π, π)
where
Convolution property (think f-k filter)
π. π₯, π‘ β π% π₯, π‘ = /!"
"/!"
"π. π’, π£ π% π₯ β π’, π‘ β π£ ππ’ππ£
πΉ π.(π₯, π‘)π%(π₯, π‘) = πΊ.(π, π) β πΊ%(π, π) Windowing
13
Plane wave fronts
ποΏ½βοΏ½
π, = apparent velocity = 1234 5
π
Incidence angle
14
Plane wave fronts
ποΏ½βοΏ½
π, = apparent velocity = 1234 5
π
Incidence angle
π β 90Β° Horizontally traveling wave (ground roll and surface waves)
π β 0Β° Vertically traveling wave (body waves and reflections)
15
Plane wave fronts
ποΏ½βοΏ½
π, = apparent velocity = 1234 5
π
Incidence angle
π β 90Β° Horizontally traveling wave (ground roll and surface waves)
π β 0Β° Vertically traveling wave (body waves and reflections)
π£ =ππ =
M1 πM1 π
=ππ
16
Plane wave fronts
ποΏ½βοΏ½
π, = apparent velocity = 1234 5
π
Incidence angle
π β 90Β° Horizontally traveling wave (ground roll and surface waves)
π β 0Β° Vertically traveling wave (body waves and reflections)
π£ =ππ =
M1 πM1 π
=ππ
π, =ππ,
βΞπ‘Ξπ₯
!.
=ΞπΞπ
A horizontal seismic line measures π, & π, not π£& π.
17
Plane wave fronts
ποΏ½βοΏ½
π, = apparent velocity = 1234 5
π
Incidence angle
π β 90Β° Horizontally traveling wave (ground roll and surface waves)
π β 0Β° Vertically traveling wave (body waves and reflections)
π£ =ππ =
M1 πM1 π
=ππ
π, =ππ,
βΞπ‘Ξπ₯
!.
=ΞπΞπ
Slope in t-d plot
Slope in f-k plot
A horizontal seismic line measures π, & π, not π£& π.
18
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
19
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
High π,Large π
Low π,Small π
20
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
k
f
RefractionsP-wave
S-wave
Surface waves (ground roll)
Air wave
High π,Large π
Low π,Small π
High π,Large π
Low π,Small π
21
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
k
f
RefractionsP-wave
S-wave
Surface waves (ground roll)
Air wave
High π,Large π
Low π,Small π
High π,Large π
Low π,Small π
We can apply a fan filter to remove ground roll (surface waves) and keep signals of interest based on incidence angle and horizontal velocity.
22
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
k
f
RefractionsP-wave
S-wave
Surface waves (ground roll)
Air wave
High π,Large π
Low π,Small π
High π,Large π
Low π,Small π
We can apply a fan filter to remove ground roll (surface waves) and keep signals of interest based on incidence angle and horizontal velocity.
But be careful, all the rules on windowing apply.
23
d
t
Refractions
P-wave
S-wave
Surface waves
Air wave
k
f
RefractionsP-wave
S-wave
Surface waves (ground roll)
Air wave
High π,Large π
Low π,Small π
High π,Large π
Low π,Small π
We can apply a fan filter to remove ground roll (surface waves) and keep signals of interest based on incidence angle and horizontal velocity.
But be careful, all the rules on windowing apply.
Can still filter specific frequencies (e.g. low pass, high pass). Filters can also be designed to remove updipsignals (right side) or downdip (left side) and many other features.
Many additional possible filters introduce greater opportunity for misinterpreting filter artifacts as structure.
24
Consider a rectangular surface array recording of a horizontal plane wave traveling in the π₯% direction.
π₯.
π₯%
This looks like a DC offset in the π₯.direction and a constant frequency sine wave in the π₯% direction.
Wav
e fro
nts
25
Consider a rectangular surface array recording of a horizontal plane wave traveling in the π₯% direction.
π₯.
π₯%
This looks like a DC offset in the π₯.direction and a constant frequency sine wave in the π₯% direction.
Wav
e fro
nts
π.
π%
βπ#
+π#
The FT of a constant frequency sine wave is a delta function. When traveling in the π₯% direction the impulses lie on the π% axis.
26
π₯.
π₯%
π.
π%
βπ# +π#
Wave fronts
Similarly
27
π₯.
π₯%
π.
π%
βπ# +π#
Wave fronts
Similarly
π₯.
π₯% Wave f
ronts
π.
π%
βπ#
+π#
28
There is a great deal more that can be done with F-K analysis (e.g. reflection seismology) and with arrays (they donβt always have to be rectangular or evenly spaced; array analysis) and it would be easy to spend an entire semester on just one of those topics.