Professional Documents
Culture Documents
A4.1 Definitions
Vrms
Vstk
input time
Tn
Tdn
Td
T0
Tg
The following pages derive kinematic equations for DMO-PSI in the time domain.
PSI is a migration-like process that deals exclusively with spatial parameters (x,
h) at constant time layers. This process is independent of velocities and is
computed efficiently in the Fourier transform domain (Kx, Kh).
DMO-PSI is virtually velocity independent. The derived point of tangency (Tg, k) is
defined for a specified velocity to enable comparison with EOM.
References
ID
111
322
373
452
544
619
620
658
690
697
Authors
Forel D., and Gardner,
Gardner, G. H. F.,
Ottolini, R. A.
Ottolini, R. A., and
Bancroft, J. C.,
Fowler, P. J.
Bancroft, J. C.,
Canning, A., and
Fowler, P. J.
Fowler, P. J.
Date
May. 1988
May. 1986
Nov. 1982
Mar. 1984
Jun. 1998
Nov. 1997
Nov. 1997
Mar. 1996
Jun. 1998
Sep. 1998
Title
A Three-Dimensional Perspective on TwoDip moveout and prestack imaging
Migration of reflection seismic data in
The migration of common midpoint slant
A comparison of prestack migration
A comparative overview of prestack time
A kinematic comparison of DMO-PSI and
A two-pass approximation to 3-D prestack
A comparative overview of prestack time
A comparative overview of dip moveout
Journal or Publisher
Geophysics
Exp. Abs., Offshore
Dissertation,
Geophysics
Exp. Abs. EAGE
Exp. Abs. SEG 1997
Exp. Abs. SEG 1997
Geophysics
Exp. Abs. EAGE
Exp. Abs. SEG
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Vol
Pages
53 604-610
75-84
49 237-249
1-18
1571-1574
1575-1578
61 409-421
1-17
1744-1747
4h 2
T =T 2 .
V
2
n
(A4.1)
The time Td (after DMO) may be found using Tc on the unit circle, i.e.,
Td
Tn
h 2 b2
2 h 2 b2
=
, where Tc =
, giving Td = Tn
.
Tc 2h
V
h
V
(A4.2)
h 2 b 2 2 4h 2
T 2 .
T =
2
V
h
2
n
(A4.3)
h2 b2
Tg = T
,
h
4k 2
T =T 2 ,
V
2
k
2
g
(A4.4)
h 2 b 2 2 4k 2 h 2
T 2 2
.
T =
2
V h b 2
h
2
k
(A4.5)
Equating both methods of DMO, i.e. equations (A4.3) and (A4.5), yields the
definition of the new offset k:
k 2 = h2 b2 .
Appendix 4 Page 2
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
(A4.6)
(h)
(h)
(k)
2h/V
Tc
2h/V
Td
Tn
Tc
DMO
Special
offset
for NMO
Td
Tn
NMO
Tg
Tg
GDMO
a)
b)
Figure A4.1 DMO ellipses for defining a) conventional NMO and DMO, and b)
GDMO.
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 3
t2
4x2
+
= 1,
Tn2 T 2V 2
(A4.7)
NMO removal
4h 2
T =T + 2 ,
V
(A4.8)
Tg2 = Td2 +
4k 2
,
V2
(A4.9)
2
n
4( x 0 b )
,
V2
2
Td2 = T02 +
4( x b )
+ t 2 = Td2 ,
2
V
(A4.10)
hg = h
Tg
T
(A4.11)
(A4.12)
2t dt
8x
+ 2 2 = 0,
2
Tn dx T V
(A4.13)
4 x0Tn2
dt
=
.
dx
T0T 2V 2
(A4.14)
8( x b )
dt
+ 2t
= 0,
2
V
dx
(A4.15)
4( x 0 b )
dt
=
.
dx
T0V 2
(A4.16)
Appendix 4 Page 4
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
x0
CMP
h
TV/2
CMPh
T0
Td
Tn
Ri
R0
NMO
Tg
Tdn
Dip-NMO
Figure A4.2
Equating the two slopes from equations (A4.14) and (A4.16),
4 x0Tn2
4( x 0 b )
,
=
2 2
T0T V
T0V 2
(A4.17)
T2
b = x0 1 n2 ,
T
(A4.18)
4h 2
T2 2
V
b = x0 1
2
T
(A4.19)
which becomes
b=
4 x0 h 2
.
T 2V 2
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 5
(A4.20)
4 x02
T = T 1 2 2
T V
2
0
2
n
(A4.21)
Solving the NMO equation (A4.8) for Tn, and substituting into (A4.21) we get
2 4 h 2
4 x02
T = T 2 1 2 2
V T V
2
0
(A4.22)
which becomes
T02 = T 2
4
V2
2
4 x 2h 2
x0 + h 2 20 2
T V
(A4.23)
Note that the bracketed term is the equivalent offset he that will be derived in
section A4.7.
The above derivation for he is based on the prestack migration ellipse that
assumes a constant velocity. A more general derivation that only assumes RMS
velocities at the scatterpoint is presented in A4.7.
Appendix 4 Page 6
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
4( x 0 b )
4k 2
,
+
V2
V2
2
Tg2 = T02 +
(A4.24)
4
V2
2
4x 2h 2
x0 + h 2 20 2 ( x0 b )2 k 2 ,
T V
(A4.25)
4 x 2h 2
2 x0 b 20 2 .
T V
(A4.26)
which simplifies to
Tg2 = T 2
4
V2
4
V2
8 x02 h 2 4 x02 h 2
2 2 2 2 .
T V
T V
(A4.27)
or
16 x 2 h 2
Tg2 = T 2 1 4 0 4 ,
T V
(A4.28)
b2
Note also that Tg2 = T 2 1 2 .
h
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 7
(A4.29)
A4.6 Define the PSI offset hg-PSI for the point of tangency Tg
After GDMO, PSI is applied to the prestack volume. Energy at the point of
tangency Tg will be rotated to the prestack migration gather at an offset hg-PSI.
The offset hg-PSI is the radial distance from the scatterpoint location
(x = xs, h = 0) to the GDMOd location (x = b, h = k), i.e.,
hg2 PSI = ( x0 b ) + k 2 ,
2
(A4.30)
8 x02 h 2
= x +h 2 2 .
TV
2
0
Appendix 4 Page 8
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
(A4.31)
(A4.32)
V 2
2
T 2 ( x h ) 2
+ 0 +
V 2
2
T0 2 he2 2
= 2 + 2 ,
2 V
(A4.33)
rearranging, we get
T0V 2
2
+ (x + h )
T0V 2
2
=2
+ he
2
T0V 2
2
2
+ (x h ) ,
2
(A4.34)
T0V
2
+ (x + h )
1
1
2
2
2
2
T V
T V
2
= 2 0 + he2 0 + ( x h ) ,
2
2
T0V 2
T0V 2
T0V
2
2
2
+ ( x + h ) =4
+ he +
+ (x h )
2
2
2
T V
4 0 + he2
2
T0V
2
+ (x h )
(A4.35)
(A4.36)
2
T0V
T V
T V
2
2
2
2
2
+ x + 2 xh + h =4 0 + 4he + 0 + x 2 xh + h
2
2
2
1
T0V 2
2 T0V 2
2
2
4
+ he
+ (x h )
2
2
T V 2
T V 2
2 T V 2
2
2
4 xh = 4 0 + he2 4 0 + he2 0 + ( x h ) ,
2
2
rearranging gives
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 9
(A4.37)
(A4.38)
T0V 2
2
+ (x h )
T V 2
2
= 0 + he2
2
T0V
xh
2
+ he
(A4.39)
T0V
2
+ (x h )
T0V 2
2
x h
T0V
2
2
2
h
+
+ he +
e
2
2
2
T0V
2
T0V
+ he
2
2
2
x 2 2 xh + h 2 =+ he2 +
he2 = x 2 + h 2
( xh )2
2
T0V
2
+ he
2
( xh )2
T0V 2
2
+ he
xh
2
+ he
, (A4.40)
2
2 xh ,
(A4.41)
(A4.42)
where the denominator term is simplified using equation (A4.33) to get the total
two way time T giving
4x 2h 2
h = x +h 2 2 .
T V
2
e
Appendix 4 Page 10
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
(A4.43)
(1 + x )
= 1+
x x 2 x3
+ ... .
2 8 16
(A4.44)
in the DSR equation and in the hyperbolic form of the equivalent offset.
Starting with the exact traveltimes of equation (A4.33) we have
T 2 ( x + h ) 2
T = 0 +
V 2
2
T 2 ( x h ) 2
+ 0 +
V 2
2
T 2 h 2 2
= 2 0 + e2 .
2 V
2
T02V 2
2
T0 4 ( x h )
+ 1 +
2
T02V 2
T0
4he2 2
= 2 1 + 2 2 .
2 T0 V
or
4 ( x + h )2
1 +
T02V 2
4 ( x h )2
+ 1 +
T02V 2
4h 2 2
= 2 1 + 2 e 2 .
T0 V
Using only the first two terms of the power series expansion we get
2 ( x + h )2 2 ( x h )2
2he2
.
+
+
=
+
1
2
1
1 +
2 2
T02V 2
T02V 2
T0 V
Solving for he,
2 ( x + h) + 2 ( x h)
4he2
,
=
2 2
T0 V
T02V 2
2
he2 = x 2 + h 2 .
Note that this form has no cross term and is an asymptotic solution for the exact
solution.
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 11
Approximate Solution #2
This following solution takes the dip-dependent moveout DD-MO correction and
Post-stack migration equations and then uses a simple geometry substitution for
the cosine term. The geometry is defined in Figure 4A.2, where we know that this
method will stack a dipping event, but will smear energy along the dip.
Note that Tdn is the moveout corrected data that lies on the zero-offset hyperbola.
2
2
Substituting Tdn = T0 +
gives
T =T +
2
2
0
4 x2
V2
2
2
into T = Tdn +
4 ( x 2 + h 2 cos 2 )
V2
=T +
2
0
4h 2 cos 2
V2
4 ( x 2 + h 2 h 2 sin 2 )
V2
(11.1 and 2)
2
4h e
=T +
.
V
2
0
Recall the geometry of zero-offset Kirchhoff migration from Figure 4.8 in the
Migration course notes as;
xmax
sin =
2x
.
VTn
4 x2h2
h = x +h 2 2 .
Tn V
2
e
Appendix 4 Page 12
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Approximate Solution #3
This solution will approximate the square-roots in the DSR with the first three
terms of the power series, and then equate them with the hyperbolic form of the
equivalent offset moveout equation.
Starting again with the DSR equation (7.6, or 11.7),
T 2 ( x + h ) 2
T = 0 +
V 2
2
T 2 ( x h ) 2 2
+ 0 +
,
V 2
2
(A4.34)
we will express the square-roots in the above power series form and then only
consider the first three terms. We now start with
1
1
2
2
4 ( x h )2 2
4 ( x + h)
+ 1 +
,
1 +
T02V 2
T02V 2
(A4.34)
2
4
2
4
4 ( x + h ) 16 ( x + h ) 4 ( x h ) 16 ( x h )
... + 1 +
... ,
1 +
2T02V 2
8T04V 4
2T02V 2
8T04V 4
(A4.34)
2 ( x + h ) 2 + ( x h ) 2 2 ( x + h ) 4 + ( x h ) 4
... ,
2 +
2 2
4 4
T0 V
T0 V
(A4.34)
T
T= 0
2
Expanding,
T
T= 0
2
T
T= 0
2
2
2
4
4
( x + h ) + ( x h ) ( x + h ) + ( x h )
T = T0 1 +
... ,
T02V 2
T04V 4
(A4.34)
(
T 2 = T02 1 +
+
... , (A4.34)
T02V 2
T04V 4
T04V 4
T 2 = T02 +
4 ( x2 + h2 )
V2
4 ( x2 + h2 )
T02V 4
4
4
2 ( x + h ) + ( x h )
... ,
2 4
T0 V
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 13
(A4.34)
4
4
2
x 2 + h 2 ) ( x + h ) + ( x h )
(
4 2
2
T = T + 2 ( x + h ) +
... ,
2T02V 2
V
T02V 2
2
0
(A4.34)
T 2 = T02 +
he2 = x 2 + h 2
(x
+
+ h2 )
T02V 2
4 2
he ,
V2
( x + h ) 4 + ( x h ) 4
... ,
2T02V 2
2
2
2
2 ( x 2 + h 2 ) ( x 2 + h 2 + 2 xh ) + ( x 2 + h 2 2 xh )
... ,
he2 = x 2 + h 2 +
2T02V 2
2
2
2
2 ( x 2 + h 2 ) ( x 2 + h 2 ) + 2 xh + ( x 2 + h 2 ) 2 xh
... ,
he2 = x 2 + h 2 +
2 2
2T0 V
(A4.34)
(A4.34)
(A4.34)
(A4.34)
2
2
2
2 ( x 2 + h 2 ) ( x 2 + h 2 ) + 2 xh ( x 2 + h 2 ) + 4 x 2 h 2 + ( x 2 + h 2 ) 2 xh ( x 2 + h 2 ) + 4 x 2 h 2
...
he2 = x 2 + h 2 +
2T02V 2
,
(A4.34)
he2 = x 2 + h 2 +
8 x 2 h 2
... ,
2T02V 2
(A4.34)
4 x2h2
... ,
T02V 2
(A4.34)
he2 = x 2 + h 2
Note this solution has a T0 in the cross term, not T as in the exact solution.
Appendix 4 Page 14
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Approximate Solutions #4
We will now start with the exact solution for he, i.e. equation (A4.43)
he2 = x 2 + h 2
4x 2h 2
.
T 2V 2
Note that the velocity is defined at T0, not at the time of the input sample T. This
expressed in the following form of the equation
he2 = x 2 + h 2
4 x2h2
.
T 2VT20
One possibility is to convert the time T into T0 with the hyperbolic form of the
equivalent offset moveout correction to give
he2 = x 2 + h 2
4 x 2 h2
.
T02VT20 + 4he2
We now have an implicit equation for he, but since the cross term is quite small,
this equation is quite good for getting an iterative solution of the form,
2
e , n +1
4 x2 h2
= x +h 2 2
.
T0 V + 4he2, n
2
We can also replace the equivalent offset in the cross term with the approximate
solution #1, i.e.,
he2 = x 2 + h 2
4 x2 h2
.
T02V 2 + 4 ( x 2 + h 2 )
These forms are all OK, (just OK), as he is defined at T. We really want to define
the velocity (which is still at T0) but with a function of (x, h, T), the input sample
location.
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/
Appendix 4 Page 15
Comments
The above equations may imply that the equivalent offset is computed each
time one input sample is summed into all the prestack migration gathers.
That is not the case.
An efficient operation copies one input trace into all surrounding prestack
migration gathers. Since one input trace may cover a number of offset bins
in the same prestack migration gather, we only need to know the starting
bin number, and the times at which the data moves to a new offset bin.
At this point, the only way I know how to do this practically is to create a
table for T, he, and bin number as a function of T0. I then find the times of
the bin transitions, which define the size of the loops that sum the data into
the corresponding offset bins. The cosine of the dip can be included in
this table to provide a dip limiting filter at the time the prestack migration
gathers are formed.
If the velocity function is spatially independent, then one set of tables can
be generated for all displacements and offset.
When the velocity varies with depth, one input sample can occupy a
number of different moveout corrected times. Similarly, one input sample
may have a number of different equivalent offsets in the same prestack
migration gather. This data mapping can be defined by starting at T0 = 0
and then defining T and he for the defined sample on the input trace.
Building a table also helps to define these areas and aids in selecting the
desirable data.
Appendix 4 Page 16
Downloaded 10 Nov 2011 to 198.3.68.20. Redistribution subject to SEG license or copyright; Terms of Use: http://segdl.org/