Methods of Seismic Data Processing PDF

Methods of Seismic Data Processing
Gary F. Margrave
Geophysics 557/657
Course Lecture Notes, Winter 2006
The Department of Geology and Geophysics

The University of Calgary
Geophysics 557/657
Course Lecture Notes
420 Pages
Winter 2005
by
G.F. Margrave, Associate Professor, P.Geoph.
The CREWES Project
Department of Geology and Geophysics
The University of Calgary
Calgary, Alberta, T2N-1N4
403-220-4604
gary@geo.ucalgary.ca
Table of Contents
Section Title Page Number

Chapter 1: Synthetic Seismograms 30 pages
The Big Picture 1-2
Elastic Waves 1-7
Well Logs 1-9
Gardner's Rule 1-11
The Wave Equation 1-14
Traveling Waveforms 1-17
Normal Incidence Reflection Coefficients 1-19
Synthetic Seismogram Algorithms 1-23
Synthetic Seismogram Examples 1-28
P-S Synthetic Seismogram Construction 1-30
Chapter 2: Signal Processing Concepts 76 pages

Convolution 2-2
Convolution by Replacement 2-5
Convolution as a Weighted Sum 2-6
Matrix Multiplication by Rows 2-7
Matrix Multiplication by Columns 2-8
Convolution as a Matrix Operation 2-9
Fourier Transforms and Convolution 2-13
Fourier Analysis and Synthesis 2-19
Fourier Analysis Example 2-21
Fourier Transform Pairs 2-23
The Dirac Delta Function 2-25
The Convolution Theorem 2-27
Sampling 2-29
The Discrete Fourier Transform 2-33
The Fast Fourier Transform 2-37
Filtering 2-38
The Z Transform 2-39
Crosscorrelation 2-44
Autocorrelations 2-46
Spectral Estimation 2-48
Wavelength Components 2-53
Apparent Velocity (or phase velocity) 2-56
The 2-D F-K Transform 2-58
F-K Transform Pairs 2-62
-p Transforms 2-63
Properties and uses of the -p Transform 2-68
Inverse -p Transforms 2-71
Least Squares -p and f-k Transforms 2-74
Chapter 3: Amplitude Effects 32 pages

Seismic Wave Attenuation 3-2
True Amplitude Processing 3-8
Automatic Gain Correction (AGC) 3-9
Trace Equalization (TE) or Trace Balancing 3-13
Constant Q Effects 3-14
Minimum Phase Intuitively 3-18
Minimum Phase and the Hilbert Transform 3-21
Minimum Phase and Velocity Dispersion 3-25
Array Theory 3-27
Chapter 4: The Convolutional Model and Deconvolution 61 pages

Bandlimited Reflectivity 4-2
The Convolutional Model 4-4
Frequency Domain Spiking Deconvolution 4-12
Finding a Wavelet's Inverse 4-20
Wiener Spiking Deconvolution 4-23
Prediction and Prediction Error Filters 4-28
Gapped Predictive Deconvolution 4-32
Burg (Maximum Entropy) Deconvolution 4-36
The Minimum Phase Equivalent Wavelet 4-39
Vibroseis Deconvolution 4-41
Deconvolution Pitfalls 4-47
Reflectivity Color 4-55
Q Example 4-58
Chapter 5: Surface Consistent Methods 29 pages

Seismic Line Coordinates 5-2
A Surface Consistent Convolutional Model 5-5
Surface Consistent Methods 5-9
Statics and Datums 5-12
Statics with Uphole Times 5-17
Surface Consistent Residual Statics 5-19
Refraction Statics 5-25
Chapter 6: Velocity Definitions and Simple Raytracing 26 pages

Velocity in Theory and Practice 6-2
Instantaneous Velocity 6-3
Vertical Traveltime 6-4
Vins as a Function of Vertical Traveltime 6-6
Average Velocity 6-8
Mean Velocity 6-10
RMS Velocity 6-11
Interval Velocity 6-13
Snell's Law 6-18
Raytracing in a v(z) Medium 6-20
Measurement of the Ray Parameter 6-24
Raypaths when v = vo + cz 6-25
Chapter 7: Normal Moveout and Stack 38 pages

Normal Moveout 7-2
Stacking Velocity 7-5
Normal Moveout and Reflector Dip 7-6
NMO for a V(z) Medium 7-10
Dix Equation Moveout 7-13
Normal Moveout Removal 7-15
Extension of NMO and Dip to V(z) 7-17
NMO for Multiple Reflections 7-22
CMP Stacking 7-27
Post Stack Considerations 7-30
ZOS: A Model for the CMP Stack 7-34
Fresnel Zones 7-36
Chapter 8: Migration Concepts 52 pages
Raytrace Migration of Normal Incidence Seismograms 8-2
Time and Depth Migrations, A First Look 8-5
Elementary Constant Velocity Migration 8-6
Huygen's Principle and Point Diffractors 8-9
The Exploding Reflector Model 8-14
F-K Migration, Geometric Approach 8-20
F-K Migration, Mathematics 8-25
F-K Wavefield Extrapolation 8-27
Recursive F-K Wavefield Extrapolation for v = v(z) 8-31
The Extrapolation Operator 8-33
Vertical Time-Depth Conversions 8-36
Time and Depth Migration in Depth 8-37
Kirchhoff Migration 8-40
Finite Difference Concepts 8-43
Finite Difference Migration 8-46
Chapter 9: The Third Dimension 32 pages

Impulse Responses 9-2
Wave Propagation 9-6
Fresnel Zones 9-7
Wavelength Components 9-10
Apparent Velocity (or phase velocity) 9-13
The F-K Transform 9-15
F-K Transform Pairs 9-19
F-L transform Computation 9-20
3-D Migration by Double 2-D 9-24
Exploitable Symmetries 9-27
Mapping Strategies 9-29
Time migration of traveltime maps 9-31
Chapter 10: Seismic Resolution Limits 35 pages

Resolution Concepts 10-2
Linear v(z) resolution theoru for zero offset seismic data 10-18
Chapter 11: Study Guide 9 pages

Geophysics 557 Final Exam Study Guide 11-2
Exam Sampler 11-7
Lecture Notes
Geophysics 557
Chapter 1
Synthetic Sei sm ogram s
Methods of Seismic Data Processing 1 -1

The Big Picture
T h e s i m p l es t m o d e l o f s ei s m i c d at a i s t h at o f a w av e le t
c on v o lv e d wi t h re f l ec t i v it y . T h e p i c tu re i s s i m p l e a n d
a p pea li n g . A c om p a c t p uls e o f so u n d i s s e n t d ow n in t o
t h e e ar t h a n d s ca l ed c op i e s o f i t a re re f l ec t e d f r om t h e
m a j or f or m a ti o n b ou n da ri e s .
T he s e e c ho e s ar e r e c o r de d o ve r t he e xt en t o f t he
s e i s m i c ex pe r i m e nt an d a na l yz e d . S i nc e e a c h e c ho i s a
s c a l e d c o py of t h e s o ur c e w av e f o rm , s i mp l e c o m pa r i s o n
m ak e s i t i s e a s y t o de d uc e t he re l a t i v e s t r e ng th o f t he
di f f e r e nt r e f l e c t i n g ho r i z on s . T he e s t i m at e d s e t o f
r e f l e c t i o n c oe ff i c i e nt s i s c a l l e d t h e r ef l e c t i vi ty f unc t i o n
o f t he e a r t h b e ne a t h t he s ur ve y .
I t s a n i c e c o n c ep t bu t is i t v a li d ? H o w c a n i t b e
d e f en ded f ro m ba si c p h ys i c al p r i n c i p le s ? W h at
a s s u m p t i on s ( t h er e a r e a l w ay s a s su mp t io n s i n p h ys i cs )
a r e re q u i r ed ? W h e n a r e t h ey j u s t i fi e d a n d w h en a re t h e y
n ot ?
1-2 Synthetic Seismograms

The Big Picture
O n c e w e s ta r t t o t hi n k ab o u t th e i de a , w e c a n i mm e d i at e l y
c o m e up w i th a l o t o f q ue sti o ns suc h as:
• How can we procede if we don't know the source waveform?
• What if several echos are very closely spaced?
• How can we tell where the echo came from?
• I s n't th e r e at te n ua ti o n o f se i s m i c e ne r g y a nd do e s n ' t t hi s
c h an g e th e s o ur c e w a vef o r m?
• What is convolution anyway? (And why should I care?)
• What about multiple bounce echos? Don't they confuse things?
• If things are so simple, how come seismic processing is so
complicated? Maybe those processors are just fooling us ...
• How can I decide how much source energy I need?
• What are the limits of the detail that can be resolved?
• What are the tradeoffs with Vibroseis and dynamite?
• What is reflectivity anyway? (And why should I care?)
• What's this band-limited stuff?
• W h y c a n' t I j u st t ru st t h e s eis m ic p ro c es so r t o take c a re
o f t he se m es sy d e ta i ls ?
I 'm su r e th a t y o u ca n th in k o f m o r e q u e st io n s a s w e ll . A l l
o f t he se q ue s ti on s h a ve t h e ir r e l e va n ce a n d I h o p e to
a d dr ess m an y of t h e m in th is co u r se . A t t he e n d , y o u
sh o u ld h av e a g o o d u nd e r s ta n di ng o f t h e st r e n g th s a nd
w e a k n e ss e s o f th e c o nv o lu ti on a l m o de l a n d t hi s s ho u ld
h e lp y ou fo r m a h e a lt h y , sce p t ica l v ie w o f f in al s e is mi c
im a g e s .

The Big Picture
S e is m i c d at a p r oc e ss i n g i s t y p ic a ll y d i v i d e d i n t o m a n y
st e p s t h ou g h t h e r ea li t y i s t h at t h e s e is m i c re f l ec t i on
p ro c es s d oe s n ot c le an ly se p ar at e i n t o d i s cr e t e
p ac ka ge s . W e h a v e a so u rc e w h i c h s e n d s o u t a
co m p l i c at ed , l ar ge l y u n k n ow n wa v ef or m w h i ch e x p an d s,
a t t en ua t es , r ef l ec t s , t ra n s m it s , c h an g e s m od e s , a n d
ge n e ra l l y s c at t er s a bo u t w h i l e a s et o f re c ei v e rs
p la c id ly r e c or d s w h at e v er co m e s t h e ir wa y. A n d
ge n e ra l l y w h at h i t s t h e re c or d er s is f ar mor e
co m p l i c at ed t h an t h e s i m p l e d i re c t e c h os t h at w e w an t :
Receivers
Surface wave
All kinds of waves
sweep across the
receivers
P-wave
reflection S-wave
reflection
G o d wo u l d n ot p r oc e ss s ei s m i c d at a t h e wa y we d o . ( I' v e
r ec e iv e d a r ev e l at i on o n t h at p oi n t . ; - } ) I n s t ea d , H e
w ou l d b ac k t h e wa v e s d ow n i n t o t h e e ar t h u n d o in g a l l
p h y si c al e f f ec t s a t t h e p oi n t w h e re t h e y o c c u rr e d . W e
a r e p r ev e n t ed f ro m d o in g t h i s l ar ge l y b e c au s e o f
i gn o ra n c e o f t h e s u b s u rf ac e st r u c t u re . T h at i s, i n o rd er
t o u nd o t h e p h ys i c al e ff e c t s o f wa v e p r op a ga t io n , we
r eq u i r e k n ow l ed g e o f t h e s u b s u rf a ce p r op e r ti e s th a t
c on t r ol t h o se e f f ec t s . U n fo rt u na t el y , th o s e a re t h e v e ry
p r op e r ti e s w h ic h we h o p e t o d i sc o v er w i t h t h e s ei s m i c
e x p er i m e n t in th e f i rs t p l ac e . P ro bl e m s o f t h i s so rt a re
c om mo n in g eo p h ys i c s a n d a r e c al l ed " i n v e rs e p r ob l em s" .
The Big Picture
S o , f ac e d w i t h t h e n e e d t o f i n d a s ol u t i on i n s p i t e o f
a l m o st t o t al ig n or an c e , we su b d i v i d e,
c om p ar t m en t a li z e, a s s u m e , a n d a p p r ox i m at e u n t i l w e
r ea ch a r es t at e m e n t o f t h e p r ob l em wh i c h i s s o v as t l y
s i m p l if i e d t h a t w e c an a c t u al l y so l v e it . A n ex a m p l e o f
such a t r e m en do u s o v e r si m p l i f i ca t io n is th e
" c o n v ol u t i on al m o d el " o f th e s e is m i c t ra c e w h i ch is o f
c e n t ra l im p or t an c e t o d e c on v o lu t i on t h e or y.
C on ti n u i n g w it h sw e ep i n g g en er al i ti e s , we c an gr ou p
m o st p h y si c al l y ba se d s ei s m i c p r oc e s se s i n t o o n e o f t wo
g ro u p s : im a gi n g p r oc e ss e s a n d d ec o n v ol u t i on p r oc e s se s .
I m a gi n g p ro c es s e s a tt e m p t t o d et e rm in e t h e co rr e c t
s p at i al p os i t io n o f t h e e c h os a n d a r e t yp i f i ed b y n m o
r em o v al , c m p s t ac ki n g , a n d m i gr at i on . D e c on v ol u t i on
p r oc e ss e s a t t e m p t t o r e m ov e t h e i l l u m i n at i n g w av ef o rm
a n d m ax i m i ze t h e r e so lu ti o n o f t h e s e i sm ic i m ag e .
E x am p l e s a r e g ai n r ec ov e r y, s t at i s t i ca l d e c on v o lu ti o n ,
i n v er s e Q f i l t er i n g, a n d wa v e l et p r oc e ss i n g .
Deconvolution
techniques

The Big Picture
I n o r d e r t o u n der s t an d t h e i m p l i c at i on s o f o u r si m pli f i ed
t h e or ie s , it is i m po rt a n t t o u n d e rs t an d a s m u c h a s
p o ss i bl e a bo u t t h e m or e r ea li s t ic p h ys i c s t h at w e a r e
a p pro x im a t i n g. T h e r ef or e , i n a d d it i on t o st u dy i n g
m a th em a t ic a l s i m p l i fi c at i on s s u c h a s t h e c on v ol u t i on a l
m o d el , w e wi l l n ot h es i t at e t o ex a m i n e o f t h e m o s t
i m p o rt an t p h ys i c al m e c h an i s m s i n v ol v ed i n s ei s m i c w av e
p r op a ga t io n .
deconvolution methods imaging methods
the convolutional model one-way scalar waves
primaries, multiples, etc
elastic wave theory
anelastic wave theory
physics of continuous media

Elastic Waves
T h e si m ple s t e la st i c m a t er i al r eq u ir e s 2 f u n d a m en t a l
c on st a n t s t o d e s c ri b e t h e re la t io n be t we e n st r e ss a n d
s t ra in k n ow n a s H oo ke ' s l aw :
σ ii = λΔ + 2μεii, i=x,y,z Δ = εxx +εyy+εzz

(Sherrif and Geldart,
σ ij = μεij , i=x,y,z, i≠j Exploration Seismology, 1981)
H e re σ i j d e n ot e s t h e c om p o n e n t s o f t h e s t r es s t e n so r
a n d e i j t h e c o m p on e n t s o f t h e s tr ai n t en so r. λ a n d μ a r e
c al l ed t h e L a m e co n s t an t s a n d μ is a ls o o f t en kn o wn a s
t h e s h ea r m o d u l u s. μ i s z er o f or a f lu id . O t h er c o n st a n t s
a r e o f t e n a l s o re f er e n c ed s u c h a s Yo u n g' s m o d u l u s , E ,
P oi s s on ' s r at io , σ , a n d t h e bu l k m od u lu s, k . T h e s e
c on s t an t s a r e a l l r e la t ed i n v ari o u s w ay s a n d a n y t wo
s u ff i c e t o d e sc r i be t h e el as t i c m a t er ia l .
μ 3λ+2μ λ 3λ+2μ
σ = k =
E = 2 λ+μ
λ+μ 3
The description of elastic wave in such a medium, requires the

application of Newton's second law (f=ma). This leads to the
incorporation of the density, ρ, as a necessary constant in the role
of "mass" in Newton's second law. Thus, analysis of elastic waves
in the most simple elastic solid (homogeneous and isotropic),
requires three parameters: any two of: λ, μ, E, σ, and k, plus the
density, ρ.

Elastic Waves
It is well established in theory1,2,3 that a homogeneous, isotropic

elastic solid supports two distinct types of body waves:
compressional and shear. Compressional or P waves are
characterized by particle motion parallel to the direction of wave
propagation. Shear or S waves have particle motion transverse to
the direction of wave propagation. P and S waves have velocities
of propagation given by:
λ+2μ μ
α = β =
ρ ρ
We may choose to regard α and β as fundamental constants

(together with ρ). Some relationships are:
α2–2β2 α 21 –σ
λ = ρ α2–2β 2 μ = ρβ 2
σ = =
2 α –β
2 2
β 1 – 2σ
3.5
2.5
1.5
0.2 0.25 0.3 0.35 0.4 0.45
Poisson's ratio
1: Waters, Reflection Seismology, 1987

2: Sherrif and Geldart, Exploration Seismology, 1982
3: Aki and Richards, Quantitave Seismology Theory and Methods,
1980,
Well Logs
Well logging is a technology designed to make geophysical

measurements in a bore hole. Well logs are the most common way
to get information about the elastic parameters of rocks which are
needed for making synthetic seismograms. Three very common
logs, which are of interest to us, are
SON ... P-wave interval transit time

SSON ... S-wave interval transit time
RHOB ... density
The interval transit time logs are usually provided in units of

microseconds/lu (lu= meters or feet). Thus, the P and S wave
velocities are found as:
6 6
10 10
α = β =
son sson
Units for density logs can vary. Be careful to work with consistent
units.
Digital well logs are usually packaged in ascii flat files in either GMA
or LAS format. The LAS format is more modern and flexible and is
to be preferred.

Well Logs
Here are some example logs from 8-8, an oil well in the Blackfoot
field
100/08-08- 100/08-08-
1400 023-23 W4 1400 023-23 W4
mannville mannville
1450 1450
1500 coal_1 1500 coal_1

coal_2 coal_2
coal_3 coal_3
1550 glauc_ch_top 1550 glauc_ch_top
glauc_1 glauc_1
glauc_ss_top glauc_ss_top
glauc_base glauc_base
1600 1600
miss miss
base base
1650 1650
3 3 2 2 1 3 2 2 2 2 2 1
5 0 5 0 5 0 8 6 4 2 0 8
0 0 0 0 0 0 0 0 0 0 0 0
Units of log SON Units of log RHOB
Faster More dense
Why do these logs appear to have a negative correlation?

Gardner's Rule
W e l l l og s a re o f t en i n ad e q u at e , i n c om p l et e , o r m i s si n g .
O n e c om mo n e x am p le o f t h i s c om e s f ro m t h e f ac t t h a t
s on i c lo gs ( SO N ) a re ru n m u c h m or e f r eq u en t l y t h an
d e n s it y l og s. T h u s we a re o f t e n f ac e d w it h t h e n ee d t o
c re at e a s e is m o gr am w it h o u t d e n s i t y i n f or m a ti o n .
G a rd ner et a l. ( 1 ) , f ol l ow ed th e r ea so n ab l e a p p r oa c h o f
s ee ki n g a n em pir i c al r el at i o n sh i p b et w ee n P - w av e
v e lo c it y a n d d en si t y. B e lo w i s a c r os sp lo t o f a a n d r fo r
B l ac k fo ot 8 - 8 w h i ch i n d i c at es a re as on a bl e c or r el at i on
e x is t s :
3000
2800
2600
2400
2200
2000
1800
2000 3000 4000 5000 6000 7000
P-wave velocity
1 Ga rd ner , G .H. F ., G a rd ne r , L . W . , an d G re gor y, A . R. , 19 74, Fo rm a tio n

v e lo cit y an d d en s it y - t h e d iag n o st i c b a si s f or s tra tig r ap hic t r a ps ,
G eo p hy s ics , 39 , 77 0-7 80
Gardner's Rule
Gardner et al. sought and found a relationship of the form:
ρ = a αm
T h e c on s t an t s a a n d m c an be d et e rm in e d f ro m fi t t i n g a
s t ra ig h t l in e t o a n p l ot o f lo g(ρ ) v e rs u s l og ( α ) . B e l ow
a r e t h e r es u l t s o f s ev e ra l s u c h f i t s t o B l ac kf o ot 8- 8 .
3200
m=.46
3000
m=.30
2800
m=.25
2600
2400
2200
2000
1800
2000 3000 4000 5000 6000 7000
P-wave velocity
G a rd ner et a l . d et e rm in e d a n d r e co m m e n d e d m = . 2 5 a s a
r ea so n ab le v al u e . H ow e v er , a s we c an s e e, t h e d at a
s u p p o rt q u i t e a ra n ge o f a lt e rn a t i v es . ( Th e v al u e o f α i s
l ar ge l y d e p e n d e n t o n t h e u n it s u s e d a n d i s n ot q u o t ed
h e re . ) T h u s , th e c ar ef u l a p pli c at i on o f G ar d n er ' s r u l e
r eq u i r es a bi t o f a n a l ys is .

Gardner's Rule
Here are the three pseudo density logs from the three fits on the
previous page.
1650 1650 1650
1600 1600 1600
1550 1550 1550
1500 1500 1500
1450 1450 1450
m=.46 m=.30 m=.25

1400 1400 1400
2000 2500 3000 2000 2500 3000 2000 2500 3000
Density Density Density
Actual density log from Blackfoot 8-8
Result from a Gardner type regression against P-wave
velocity

The Wave Equation
T h e g re at s u c c es s o f p h y si c s in e x p l ai n i n g o u r w or l d a n d
f u e li n g t h e gr ow t h o f t ec h n o lo gy i s ba se d f u n d am en t a ll y
u p o n d i f f er en t i al e q u a t io n s a n d m or e sp ec i f ic a ll y p ar t ia l
d i f f er en t i al e q u a t io n s . P D E ' s a r e t h e m a t h em a t i ca l
s t at e m en t o f t h e a p p l ic at i o n o f b as i c p h y si c al l a ws t o
c om ple x s ys t e m s . F or e x am p l e , a c on s i d e ra ti o n o f a
c on s t an t d e n s i t y f lu id l e ad s t o t h e 's c al ar w av e
e q u at i on ' wh i c h i s c en tr al t o m os t ge op hy s ic a l i m ag i n g
a l g or it h ms. T h e S W E i s a d i r ec t c o n s eq uen c e o f
N e w to n ' s s e co n d l aw a n d H oo ke 's l aw a s a p p l ie d t o t h e
f l u id .
∂2 Ψ ∂2Ψ ∂2 Ψ 1 ∂2 Ψ
2
+ 2
+ 2
– 2 2
= f x,y,z,t
∂x ∂y ∂z v x,y,z ∂t
H e re Y i s t h e p r e ss u r e , v i s t h e v el o ci t y o f w av e
p r op a ga t io n , a n d f ( x ,y ,z , t) r e p re s en t s a n y p os s ib l e
s ou r c es .
T h o u g h i t is h ar d l y o b v i ou s , t h e s o lu t i o n s t o t h i s
e q u at i on a re t ra v el i n g w av e s . A gr e at d e al o f in te r es t i n g
p h y si c al ef f e ct s ca n b e s t u d i ed w it h t h e S W E i n c lu din g :
• propagation of primaries and multiples

• reflection and transmission at interfaces
• head waves and surface waves
• ray theory, Snell's law
• characterization of sources
• arrays of sources and receivers

The Wave Equation
T h e r e is a p ow e rf u l m e t h od o f s ol u t i on o f P D E 's t h a t i s
o f c o n s id e r ab le r el e v an c e e x p lo ra t io n s e is m o lo gy . T h i s i s
t h e m e th o d o f s o lu t i o n by G r ee n ' s fu n ct i o n s. W e w il l n o t
d e v el o p i t h e re b u t s i m p l y st a t e th e im p or t an t r es u l t s .
T h e e ss e n c e o f t h e t h eo ry is t o d e v e lo p a s ol u t i on t o
t h e P D E o f i n t e re s t f or a " p o in t s ou rc e " a n d t h e n t o
s h ow h ow the re s p on s e to a r b it r ar y s ou r c e
c on f i g u ra ti o n s ca n be c on st r u c t ed fr om t h e e l em en t ar y
s ol u t i on . T h e S W E , w h en s p e ci al i ze d f o r t h e G r ee n ' s
f u n c t i on p ro bl e m l o oks li k e:
∂2 G ∂2G ∂2G 1 ∂ 2G
2
+ 2
+ 2
– 2
= δ x–xo,y–yo ,z–zo,t–t o
∂x ∂y ∂z v x,y,z ∂t
2
T h e t e rm o n t h e ri g h t o f t h e eq u al s ig n i s a D i ra c d el t a
f u n c t i on a n d r ep re s en t s a m at h e m a ti c al i m p u l s e a t a
s i n gl e p o i n t i n s p ac e , ( xo ,y o , zo ) , a n d a t a n in st a n t o f
t i m e , t o . T h e s o lu t i on to t h e G re e n ' s f u n c t i on p ro bl e m ,
G ( x, y ,z ,t ) , i s k n ow n e x ac t l y f or c o n st a n t v e lo c it y a n d
a p pro x im a t el y fo r a n u m b e r o f m or e c o m p l i ca t ed
s i tu a t i on s . G c on t a in s a l l p h y si c al e f fe c t s d u e t o t h e
i m p u l s iv e s ou r c e a n d i s p r op e rl y c al le d a n " i m p u l s e
r es p o n se " .
T o o b t ai n t h e r es p o n s e t o g en e ra l s ou r ce c o n fi g u ra t io n s ,
w e im a g in e t h e so u r ce to b e c om p os e d o f a s et o f s c al e d
i m p u l s es . T h en co n s t ru c t t h e G re en ' s f u n c t i on s f or a ll o f
t h e se i m p uls e s a n d si m p l y s u p e ri m p o se t h e s e G r ee n ' s
f u n c t i on s . T h i s i s a n e x am p l e o f t h e m a t h em a t i ca l
p r oc e ss o f " co n v ol u t i on ". W e w il l l ea rn m o re a b ou t
c on v o lu ti o n l at e r i n t h i s c ou r s e. F o r n ow , it i s e n ou g h t o
v i su a l iz e i t a s a ge n e ra l s u p e r p os i t io n o f s c al e d " i m p u l s e
r es p o n se s " .
The Wave Equation
T he r e s ul t w e ha ve j us t o b t a i ne d i s s o i m po r t a nt t h at w e
r e s t a t e i t i n d i f f e r e nt t er m s :
T h e w a v ef ie ld d u e to a so u r ce ha vi n g e xt en d e d sp a t ia l
a n d te m p o ra l f o rm c a n b e co n s id e re d t o be t h e
c o n v o lu t io n o f t h e e a rt h 's i m p u ls e res p o n se w it h t h e
e xt en d e d so ur ce . T h is re su l t h o l d s fo r a n y l i ne a r w a v e
e qu a t i o n a n d e x te n d s t o el a st i c, a n is o t ro p i c a n d
a t t en u a t in g m ed ia .
T h e t wo c om p o n en ts o f t h i s r e su lt , t h e e ar t h ' s i m p u l s e
r es p o n se , I r , a n d t h e s ou r c e wa v ef or m , w s , a re b ot h
a b s t ra ct e n t i t ie s th a ta r e d i f f ic u l t t o q u a n t if y . I r i s
g en e r al ly ve r y c om p l i c at e d a n d c on t ai n s a l l p h ys i ca l
e ff e c t s. w s i s a c om p l e t e ch a ra c t er iz at i on o f t h e s ou r c e
w av e fi e ld a n d c an be co n s id er ed a s t h e sp ec i f ic a t io n o f
t h e wa v ef i e ld a t a ll p o i n t s o n a s u rf a ce s u r ro u n d i n g t h e
s ou r c e.
Impulse response
Response to 3 sources
Traveling Waveforms
T he s im plest m a thematic al w a ve equat io n is th e scala r w a ve

e q ua t io n. I n a coustic m e di a o r s im p le e lastic media ,
compression a l w a ves a r e g o ve rne d b y it . In 1- D , th e s c al ar
w a ve equ a tio n is :
∂2 ψ 1 ∂2 ψ
2
= 2 2 (1)
∂z v ∂t
Where ψ represents the propagating wave. We now show that

ψ = f t±z/v (f is an arbitrary function)
is a solution to (1).
∂f 1 ′ ∂2 f 1 ′′
= ± f , = 2f
∂z v ∂z
2
v
∂f
2
′ ∂ f ′′
= f , = f
∂t ∂t
2
Substitution of the second partials of f into (1) results in an

immediate identity. Thus f is a solution to (1) with the form of f
being arbitrary except that it must be twice differentiable.

Traveling Waveforms
As an example of a waveform, consider the Ricker wavelet defined

by:
2 2
w τ = 1–2 πfdomτ exp – πfdomτ τ->
-0.05 0 0.05
Shown for fdom=30Hz
Note that the Ricker wavelet is centered where its argument equals
zero. Thus w(t+z/v) represents a wavelet centered at t+z/v = 0 or
z = -vt. So we conclude:
w t+z/v = Wavef o rm trave li ng z=-vt

i n th e - z d ir ec ti o n
Wa vef o r m tr a vel i ng
w t–z/v =
i n t he +z d i r e c ti o n z=vt
Similarly, cos(ω (t-z/v)), cos(k(z - vt)), and cos(ω t-kz) all

represent cosine waves traveling in the +z direction.
1
1.01 sec
0.5 1.0 sec
0
-0.5
-1
400 450 500 550 600
z-> (meters)
cos 2π30 t–z/1000 Plotted versus z for t=1.0 and 1.01 (sec)
Normal Incidence Reflection Coefficients
(Adapted from E.S. Krebes, Course Notes in Theoretical Seismology)
Incident Reflected
Consider a v e r ti c a l l y displacement displacement
tr a ve li ng com pr es s i o na l Z f t–z/α1 g t+z/α1
w a ve incident o n a
h or i z o nt a l i nterface. I n
α1,ρ 1
o r de r t o d e s c r ib e the
r e f l e cti o n a n d α2,ρ2
tr a n s mi ss i o n t ha t o c c u r ,
i t can b e s ho w n tha t t w o
c o n di t i o ns m us t b e h t–z/α2
s a ti sf i e d : Transmitted
displacement
continuity of displacement: f + g = h (1)
continuity of normal pressure: ??? (2)
To develop a form for the second equation, we use Hookes

law which says stress is proportional to strain.
stress = (applied force)/area

strain = (change in length)/length

Consider an infinitesimal elastic element

whose ends undergo displacement u1 and u2: u1
dz
Δl u2–u1 ∂u
Strain = = ≈
l dz ∂z
u2
Now, invoking Hooke's law:
Force ∂u
stress = pressure = = k
area ∂z
Where k is a constant formed from the material constants. To

determine k, we can use dimensional analysis:
2
l
force units mass sec2 mass l
pressure = = 2
= 3
(length units)2 l l sec
So, k looks like: k = ρα Thus the pressure continuity equation is:

2
∂f ∂g ∂h
ρ 1α21 + ρ1α21 = ρ2 α22 (evaluated at the interface)
∂z ∂z ∂z
∂f –1 ′ ∂ g 1 ′ ∂h –1 ′
But since = f, = g, = h
∂z α1 ∂z α1 ∂z α2
ρ 1α1f′ – ρ1α1g′ = ρ2α2h′
Which can be immediately integrated to give:
ρ 1α1f – ρ1α1g = ρ2α2h (2)

Assume that an interface occurs at z=0, then if the boundary

conditions are applied there, the two equations determining normal
incidence reflection and transmission are:
f + g = h (1)
I 1f – I 1g = I 2h (2)
Where impedance = Ik = ρkαk , k= 1,2
and where f,g, and h are understood to be evaluated at z=0.
Multiplying (1) by I2, and subtracting it from (2) leads to:
I1–I 2
g = f = –Rf
I1+I 2
Similarly, we can obtain:
2I 1
h = f = Tf
I1 +I2
The quantities R and T are known as the normal incidence reflection

and transmission coefficients:
I2 –I1 2I 1
R = , T =
I1 +I2 I1+I2
I 2–I1+2I1
Note that: R+T = = 1
I1+I 2

I2–I 1 2I 1
R = , T =
I1+I 2 I1+I 2
R and T are often written in terms of the contrast and average of

impedance across the layer:
1
I = I1+I 2 , ΔI = I 2–I 1
2
I 1 = I–.5ΔI , I 2 = I+.5ΔI
Straight forward algebra then gives:
1 d ln I
ΔI
R = ≈ Δz
2I 2 dz
I–.5ΔI
T = 1–R =
I
R can be written in terms of ρ and α as:
Δ ρα ρΔα+αΔρ 1 Δα Δρ
R = = = +
2ρα 2ρα 2 α ρ
Note that the definition of R is such that an impedance increase

gives a positive RC but that the reflected pulse is flipped in
polarity.

Simple "Primaries Only" Impulse Response.
Layered Earth, Normal Incidence, Acoustic
Model Impulse
Response
k=0
V1,R 1 *R1
t=Δt
k=1 *R1
V2,R 2
1 –R * * 1–R1 *R2
1
t=2Δt
2
k=2 *1– R1 *R2
V3,R 3
1– R * 1– R * * 1– R1 1– R2 *R3
1 2
t=3Δt 2 2
k=3 *1– R1 1– R2 *R3
Vj,R j j–1 j–1
1– Rk * * 1–Rk *Rj
k = 1
k = 1
j–1
t=jΔt
2
k=j * 1– Rk *Rj
k=1
Model layers have a Δ Zj

constant traveltime Δt =2
"thickness": Vj
Vn,R n
k=n-1
n–1 n–1
1– Rk * * 1–Rk *Rn
k = 1 k=1
n–1
t=nΔt 2
k=n * 1– Rk *Rn
k=1

Computation of a 1-D Synthetic Seismic Impulse Response
(Including All Multiples)
t=Δt t=2Δt t=3Δt t=4Δt t=5Δt t=6Δt t=7Δt t=8Δt
R0
t
R1
z
R2
R3
E ar th model i s bui lt of
l ay er s of e qua l tr a ve lti me
" thic kness" Δ t
R4
R5 Completed node
Current node
R6
Note: All Raypaths are
actually vertical. They are
R7 shown slanted for illustrative
purposes.
R8
R9
At the designated point, 6D4 and 6U5 The c omp le te s e is mogr am i s
are known and we wish to compute obta ined by r ecur s ive ca lcula tion
6U4 and 6D5: be ginning i n the up pe r le f t. All
nod es on a ny up war d tr ave ling r a y
6U4 = R4*6D4 + (1+R4)*6U5 ar e comp le tel y ca lc ul ate d bef ore
pr oc ed ing to the nex t d e pth.
6D5 = (1-R4)*6D4 -R4*6U5
Adapted from: Reflection Seismolgy, K.H. Waters, 1981

J.H. Wiley
From Impulse Response to Source Waveform
Response
Impulse Response Source Waveform Response
t=Δt *R 1 t=Δt *R 1
2
t=2Δt * 1–R1 *R2 t=2Δt * 1–R1 *R2
2
2 2
t=3Δt * 1–R1 1–R2 *R3 t=3Δt 2
* 1–R1 1–R2 *R3
2
j–1 j–1
2
t=jΔt * 1–Rk *Rj t=jΔt *
2
1–Rk *Rj
k = 1 k = 1
n–1 n–1
2
t=nΔt * 1–Rk *Rn t=nΔt *
2
1–Rk *Rn
k = 1 k = 1
T h e " p r i m ar ie s o n l y"
T o o b t ai n t h e s o u rc e
impulse re s p on s e
wa v ef o rm re s p on s e
c on s i s ts of a time
f ro m the i m pu ls e
s er i es o f s c al ed a n d
re s p on s e , s im ply
d e la ye d im pu ls e s
re p l ac e ea c h s p i ke o f
t h e i m pu ls e re s p on s e
by t h e p r od uct o f t h e
s p i ke and s o u rc e
wa v ef o rm . T h i s i s t h e
m at h e m a ti c al p ro c es s
o f co n v ol u t i on
Impulse Responses and Seismograms
F o r a li n e ar ea rt h , i t c an be s h ow n t h at i f w e a re g i v en
t h e w av e fo rm s i gn a t u re o f a n on - im p u l si v e so u rc e a n d
t h e i m p u l se re s p on s e o f a n ea rt h m od el , t h e n :
s t = Ir t •ws t
where:
Ir t is the earth impulse response
ws t is the source waveform
st is the earth response to the source waveform

T h e g e n er al p ro of o f t h i s r es u l t c o m es f ro m " G re e n 's
fu n ct i o n a n a ly s is " a n d is t r u e f or a n y l i n e ar wa v e
eq ua t i on ( e l as t ic , s c al ar , et c ) G en er al l y I r c on t a in s a l l
p h ys i c al ef f ec t s t h e t h e or y i s c ap ab l e o f p r od u c in g , a n d
u su a l ly t h a t i s m o re t h a n w e w an t .
T he m o s t c o m m o n u s e o f 1 - D s ei s m o gr a m s i s i n t h e
i n te rp re t a t io n o f p ro c es s ed s e i sm i c s e c t io n s. I n t h is
ca s e m o st of th e p h y si ca l e f f ec ts ( m u lt ip l es ,
t r a n s m is s io n l o ss es , a tt en u a ti o n ) h a v e b e e n r em o v ed i n
t h e p r o ce s si n g. T h e re f o re , c o m m o n p ra c ti ce r e p l a ce s
I r ( t ) w it h r ( t ) w h er e :
n o rm a l i n c i d en ce re f le c t i on co ef f i c ie n t s
rt =
p o s it i o n ed i n 2 - w ay v e rt i c al t r av e l t im e
Thus: s t = r t •ws t
s( t ) g i v en b y t h i s re s u lt i s th e m o st c om m on 1-D
se i s m og ra m c o m p u t e d i n ex p l o ra ti o n g eo p h y si c s .
1 - D S y n th e ti c S ei s mog ram Su m ma r y
• A c o m p l et e s ol u t i on , g en er at i n g a l l m u l t ip le s a n d
t ra n sm is s i on ef f e c t s, c an b e c on s t r u c te d . S o m e
m et h o d s a ls o i n c l u d e a t t e n u at i on .
• A ss u m pti o n s: ra y th eo ry , 1 - D , n o rm a l i n c id en c e
• G e o p h ys i c al w el l l og s , p r ov i d i n g P - w av e v e l oc i t ie s
a n d d e n si t i es , a re u se d . T h e y a re u su a ll y r es am p l ed t o
a v a ri ab l e d e p t h l a ye ri n g w it h e q u al D t s te p s .
• M e t h od i s i n h er e n t ly a lg or i t h m i c . N o a n a l yt i c c l os ed
fo rm s ol u t i on a v ai l ab l e.
• I n p r ac t ic e , m ult i p l es a n d t ra n s m is s i on lo s se s a re n o t
u s u al ly in cl u d e d . R e f le c t i on c oe f fi c i en ts i n t im e a r e
si m ply c on v ol v ed w it h a so u rc e r e sp o n s e.

Example of Synthetic Seismogram Creation by
Convolution of Reflectivity and Wavelet.
Time Domain View
Wavelet
Synthetic Seismogram
Reflection Coeficients
0 0.2 0.4 0.6 0.8 1 1.2

Time (secs)

Example of Synthetic Seismogram Creation by
Convolution of Reflectivity and Wavelet.
Frequency Domain View
-10
-20
Reflectivity
-30
-40 Wavelet
-50
-60
-70
-80
Synthetic Seismogram
-90
-100
0 50 100 150 200 250
Frequency (Hz)

P-S Synthetic Seismogram Construction
The SYNTH Algorithm
Define Layered Model
Vp, Vs, and

R e sa m pl e d
density logs
l o gs
Loop over layers: k=1 to nlayers

Iterative Snell's law raytracing
1 ) Ra yt r ac e
I nc i de nc e
P P OR P S
A n gles
Free
2) Zoeppritz PP AND surface
RCs PS P
S S
Primary reflections
3) Map RCs to to,
Input
apply wavelet.
wavelet
Re s pons e of
Next layer l ay e r k
+
Accumulated
gather after k-1
layers
= Accumulated
gather aft er k
la yer s

Lecture Notes
Geophysics 557
Chapter 2
Sign al P rocessi ng

Convolution
C on v o lu ti o n i s t h e m a t h e m at i c al p r oc e s s o f " s h i ft i n g ,
s ca li n g , a n d su mm i n g" a w av e fo rm t o p r od u c e a n o u t pu t
by s u p er p o si t i on . G e n e ra ll y, t w o i n p u t s i gn a ls a r e
re q u i r ed , sa y r a n d w , w it h w b ei n g t h e wa v ef or m a n d r
a s e ri e s o f s c al in g c oe f f ic i e n t s. F or ex a m p l e , l e t r= [1 0
0 - .5 .5 0 - 1 ] a n d le t w = [ - . 5 1 - . 5 ] , t h en t h e
c on v ol u t i on o f r a n d w i s:
O u t p u t sa m p l e n u m b e r
j 0 1 2 3 4 5 6 7 8
k
r0w0 r0w1 r0w2 = r0*w
0 -.5 1 -.5 0 0 0 0 0 0
r1w0 r1w1 r1w2 = r1*w +

1 0 0 0 0 0 0 0 0 0
r2w0 r2w1 r2w2 = r2*w +

2 0 0 0 0 0 0 0 0 0
r3w0 r3w1 r3w2 = r3*w +

3 0 0 0 .25 -.5 .25 0 0 0
+ r4w0 r4w1 r4w2 = r4*w
4 0 0 0 0 -.25 .5 -.25 0 0
+ r5w0 r5w1 r5w2 = r5*w
5 0 0 0 0 0 0 0 0 0
+ r6w0 r6w1 r6w2 = r6*w
6 0 0 0 0 0 0 .5 -1 .5
s = r•w
-.5 1 -.5 .25 -.75 .75 .25 -1 .5

2-2 Signal Processing Concepts
Convolution
I n t h e p r ev i ou s s l id e, we d e s cr i be d a t a bl u l ar m e th o d
f or c om p u t i n g t h e c on v o lu t i on o f r a n d w t o y i el d s.
T h i s c a n b e w ri t t en m a th em a t ic a ll y a s fo ll o ws :
s = r• w
sj = Σ rk wj–k
k
T o s ee t h a t t h i s s u m m a t io n e x p r es s i on is eq u iv a le n t t o
t h e t ab u la r m e t h od , co n s id er t h e e x am p l e o f j =4 :
s4 = r0 w4–0+r 1w4–1+r2w4–2+r3w4–3+r4 w4–4+r 5w4–5+r6w4–6

s4 = r0 w4+r1w3 +r 2w2+r3 w1+r4w0 +r5w–1+r6w–2
N ot e t h at t h e le n g t h o f s i s t h e c om b i n ed l en g t h s o f
r a n d w le s s 1 :
length s = length r +length w –1
T h u s , m a t h em a t i ca ll y , e v er yt i m e a c o n v ol u t i on i s
p er f or m ed t h e r es u l t i n c re as e s i n l e n gt h . T h i s c re at e s a
b i t o f a h e ad e r ( bo okk e ep i n g ) p r ob l em in se i s m i c d at a
p ro c es s in g a n d i s n ot u s u al l y a l lo we d . T h a t i s, i f a
s e i sm ic t r ac e is c on v ol v ed wi t h a f i l t er o p e r at or , t h e
r e su lt i s t r u n c at e d a t t h e s a m e l e n gt h a s t h e se i sm ic
t r ac e .

Convolution
We have seen that the convolution of discretely sampled

vectors is written:
sj = Σr w
k
k j–k
The analagous result for continuous functions is:

∞
st = r τ w t–τ dτ
–∞
We now show that the order of convolution is immaterial.

Let:
τ′=t–τ, dτ′=–dτ, τ=t–τ′

–∞
Then: st = – r t–τ ′ w τ′ dτ ′
∞
And: st = r t–τ′ w τ′ dτ′

–∞
So: s = r•w = w•r
We also note that convolution is linear in the sense that:
a+b •c = a•c + b•c

Convolution by Replacement
( e m p ha s is o n in pu t s am ple s )
Consider the discrete convolution of a three point boxcar, b, with
an eleven point time series, r.
0.1
0.05 r 1
0 • 0.5 b
-0.05 0
0 2 4
-0.10 2 4 6 8 10 12
E ac h i n p u t sa m p l e i s c on s i d e re d se p a ra te l y. T h e
b ox c a r i s m u lt i p l i ed by t h e i n p u t sa m p l e re s u lt i n g i n a
s c al e d b ox c ar . T h e s c al e d b ox c a r c o n t ri b u t es t o
o u t p u t sa m p l e lo c at i on s b eg i n n i n g a t the
p os i t io n o f t h e i n p u t sa m p l e . T h us t h e
b ox c ar i s sc a le d b y ea c h sa m p l e o f r
0.1
a n d re p l ic a t ed a t t h e lo c at i on o f
0.05 th e r s am ple . E ac h o u t p u t
sa m p l e r ec e iv e s m u lt i p l e
0
co n t ri b u t i on s w h i c h a r e s u m m e d .
-0.05 I n p u t s am ple s 1 ,2 a n d 6 a r e
sh o wn e x p l ic i t l y c o n t ri b u t in g.
-0.1
0 2 4 6 8 10 12
0.1 0.1
0 =0
-0.1 -0.1
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14

Convolution as a Weighted Sum
( e m p h as i s o n o u t p u t s am p l es )
Consider the discrete convolution of a three point boxcar, b, with

an eleven point time series, r.
0.1
0.05 r 1
0 • 0.5 b
-0.05 0
0 2 4
-0.10 2 4 6 8 10 12
T o co m p u t e a n o u t pu t 0.
s am p l e , p o si t i on t h e 1
b ox c ar o v e r s om e r
s am p l e s , m u l t ip ly t h e r 0
s am p l e s b y t h e bo x c ar
w ei g h t s, a n d s u m . T h e -0.05
c om pu t at i on o f o u t pu t -0.1
0 2 4 6 8 10 12
s am p l e s 1 a n d 7 i s
i ll u s t r at ed . T h i s i s a
p r oc e ss o f s m o ot h i n g
0.1
o r a v er ag i n g t h e i n p u t .
-0.1
0 2 4 6 8 10 12 14

Matrix Multiplication
by Rows
Consider the a 4x4 matrix equation such as:
a 11 a12 a13 a 14 b1 c1
a 21 a22 a23 a 24 b2 c2
= eqn 1
a 31 a32 a33 a 34 b3 c3
a 41 a42 a43 a 44 b4 c4
This is equivalent to the following system of equations:
c1 = a11b 1 + a12b 2 + a 13b 3 + a14b 4

c2 = a21b 1 + a22b 2 + a 23b 3 + a24b 4
eqns 2a-2d
c3 = a31b 1 + a32b 2 + a 33b 3 + a34b 4
c4 = a41b 1 + a42b 2 + a 43b 3 + a44b 4
T h us t h e el e m en t s o f t h e v ec t o r C a r e c om p u t e d b y
t ak in g e ac h ro w o f A , m u l t i p l yi n g i t by t h e v e ct o r B , a n d
s u m m i n g t h e r e su l t s . T h is p ro c es s i s f am il i ar t o m o s t
s t u d e n t s o f li n e ar a l ge br a a s " m at r i x m u l t i p l i ca t io n b y
r ow s" . I t c a n be wr i t t en s ym bo li c al l y a s t w o n e s t ed
c om pu t at i on lo op s :
c=zeros(1,4);
for irow=1:4
for jcol=1:4
c(irow)=c(irow) + a(irow,jcol)*b(jcol);
end
end

Matrix Multiplication
by Columns
M at r i x m u l t i p li c at i on " b y c ol u mn s" i s l e s s we l l k n ow n
t h an th e c o rr es p o n d i n g p ro c es s " b y r ow s" bu t i t
p r ov i d e s a u s ef u l in tu it i v e i n s i gh t t o co n v ol u t i on .
E x am i n at i on o f e q u at i on s 2 a - 2 d s h ow s t h at t h e c ol u m n s
o f A h a v e b ee n m u l t ip li e d by a si n g l e c or re s p on din g
e le m e n t o f B . T h u s w e ca n e x p r es s t h e m a t ri x
m u l t i p l ic a t io n a s a su m o f c o lu m n v ec t or s , e ac h o n e
b ei n g a s c al ed v e rs i on o f a c o lu m n o f A .
a 11 a 12 a 13 a 14 c1
a 21 a 22 a 23 a 24 c2
b1 + b2 + b3 + b4 =
a 31 a 32 a 33 a 34 c3
a 41 a 42 a 43 a 44 c4
W ri t t en a s c om p u t at i on l o op s , t h i s a m o u n t s t o r e v er s in g
t h e o rd er o f t h e l oo p s i n t h e m u l t ip l i c at i on s " by r ow s "
c=zeros(1,4);
for jcol=1:4
for irow=1:4
c(irow)=c(irow) + a(irow,jcol)*b(jcol);
end
end

Convolution as a Matrix Operation
C on si d e r t h e c on v o lu t i on o f a re f l ec t i v it y se q u e n c e , r ,
w it h a w av e le t , w , to y i e l d a se i s m i c t ra c e , s . T h i s i s
u s u al l y wr i tt e n a s t h e c o n v ol u t i on in t e g ra l:
s(t) = w(t – τ)r(τ)dτ

–∞
W h e n we h a v e d i s c re t e, fi n i t e l en gt h a p p r ox i m a t io n s t o
t h e se q u a n t i ti e s , th e c o n v ol u t i on i s u s u al l y w ri t t e n a s a
s u m m a ti o n . I f r j i s t h e re f le c t i vi t y se r ie s wi t h j = 0 , 1 ,. . . n ,
an d wk is t h e p o ss i b ly n o n - c au s al wa v el e t w it h k =-
m . . .0 . . . m , t h e n :
k–m
s k = Δt Σ
j = k+m
wk–jr j
U s u al l y, in t h e se e xp re s si o n s, t h e Δ t t e rm i s d ro p p e d o r
s et t o u n i t y. I t i s u se f u l t o w r it e o u t a fe w t e rm s o f t h i s
s u m m a ti o n :
s0 = + w0r0 + w–1r1 + w–2r 2 +

s1 = + w1r0 + w0 r1 + w–1r 2 +
T h e s am e o p er at i on c an b e a c h i e v ed b y m at r ix
m u l t i p l ic a ti o n wh er e th e w av e l et , w , is l oa d ed i n t o a
s p e ci a l m at r i x c al l ed a T o ep l i t z o r c on v ol u t io n m a t ri x .

I t i s a s i m p l e ex e rc i s e o f m at r ix m u l ti p l i c at i on b y r ow s t o
c h e c k th a t t h e f o ll ow i n g m at r i x e q u at i on c om pu t es t h e
c on v ol u t io n o f w w i th r
w0 w–1 w–2 w–3 r0 s0

w1 w0 w–1 w–2 r1 s1
=
w2 w1 w0 w–1 r2 s2
w3 w2 w1
w0 rm sn
N ot e t h e s ym m e t ry o f t h e W m a t ri x w h i c h h as t h e
w av e le t s am ple s c on s t an t a l on g t h e d i ag on al s . A n o t h er
w ay t o v i e w W i s t h at e ac h c o lu mn c on t a in s t h e w av e le t
w i th t h e z er o t i m e s am ple a l ig n e d o n t h e m ai n d i ag on a l.
N ow , im a gi n e d o i n g th e m a t r ix m u l ti p l i c at i on by c ol u m n s
i n s t ea d o f r ow s a n d we ge t t h e m o s t i n t u i t iv e v i e w o f
c on v ol u t io n " by r ep la c em e n t " .
w0 w–1 s0
w1 w0 s1
r0 + r1 + =
w2 w1 s2
w3 w2
sn

A s a n e x am p l e o f c o n v ol u t i on by m at r ix m u l t i p l ic a ti o n ,
h e re i s a n i l l u st r at i on o f t h e c on v ol u t i on o f a r e fl e c t iv i t y
s er i es a n d a m i n i m u m p h as e wa v le t t o y i e ld a 1 -D
s ei s m o gr am .
A s a s ec o n d ex am p l e, h er e i s th e c on v o lu t i on o f a
r ef l ec t i v i t y s er i es a n d a z er o p h as e wa vl e t t o y ie l d a
z e r o p h as e s ei s m og r am .

T h es e e x am p l es o f co n v ol u t i on b y m ar t i x m ult i p l i c at i on
s h ow e x p l ic i t l y w h at i s m e an t w h e n we s ay t h at
c on v ol u t io n i s a s t at i on a ry p r oc e ss . I n t u i t i v el y , th is
p h r as e m e an s t h at t h e o p e ra t io n d o es n ot ch a n g e wi t h
t i m e i n s om e se n s e . P re c i se l y, i t m ea n s t h at t h e
w av e f or m s i n t h e co l u m n s o f t h e c o n v ol u t io n m at r i x a re
a ll id en t i c al . T h at is , th e wa v el e t w h i ch i s s c al e d a n d
u s e d t o re p l ac e e ac h r ef l e ct i v i t y s p i k e d o e s n ot c h a n ge
w i th t i m e . A s w e s h al l s e e, m a n y p h y si c al p ro ce s s es
v i ol at e th is a s su m p t i on a n d i t i s q u i t e p o s si b l e t o
g en er al i ze the co n v ol u t i on o p er at i on to m od el
n o n s ta t io n ar y p r oc es s e s.
W h en t h e a ss u m p t i o n o f s t at i on a ri t y i s m a d e i n t h e
c on te x t o f s t at is t i c al d ec o n v ol u t i on t h e or y, it m e an s
p r e ci s e ly t h e s a m e t h i n g. W e a s s u m e t h a t t h e t i m e s e ri e s
w e m ea su re d ( t h e s ei s m i c t ra c e ) i s re l at e d t o t h a t wh i c h
w e w an t ( th e r ef l e ct i v i t y) b y a s ta t io n ar y c o n v ol u t i on
o p er at i on . G i v en t h at , we e x p e c t t h at a st a ti o n ar y
i n v e rs e o p er at o r wi l l s u f fi c e t o r ec o ve r t h e r e fl e c t iv i t y.

F ou rie r T ran s fo r ms a n d C o n v ol u t io n
C on si d e r t h e co n v ol u t i on i n t eg r al for c on t i n u o u s
f u n c t i on s :
iωu
Now, let g be a complex sinusoidal function: gu = e
∞
iω t–τ iωt
Then: ht = fτe dτ = e F ω (1)
–∞
∞
–iωτ
where Fω = fτe dτ (2)
–∞
T h i s r em a rk ab l e r es u l t s h ow s t h at i f we c on v o lv e A N Y
f u n c t i on , f , wi t h a co m p l e x s in u s o id , t h e r es u l t i s t h e
s am e c om p l e x si n u s o id m u lt i p l i ed by a c o m p l ex
c oe f f ic i e n t . T h i s c o m p l ex c o ef f i ci e n t , F ( w ) , i s c om pu t ed
f ro m f( t ) a n d i s k n ow n a s t h e F o u ri e r T r an s f or m o f f ( t ) .
T h o s e w h o h av e st u die d m a t h em a t i c al p h ys i c s wi l l
r ec og n i ze t h at t h i s m e an s t h at t h e c om ple x s in u so i d s
a r e e ig e n f u n c t io n s o f t h e c on v o lu ti o n o p er at o r a n d t h e
F o u ri e r T r an s f or m p ro v id es t h e ei g en v a lu e s .

Fourier Transforms and Convolution
H e re we s e e t h e r e su l t o f c o n v ol v i n g 1 0 , 3 0 , a n d 7 0 H z
c om ple x s in uso id s w it h a 3 0 Hz R i c ke r w av e le t . I n ea ch
c as e , o n l y t h e r ea l p ar t s o f t h e c om ple x s i n u s oi d s a r e
p l ot t e d . W e se e t h a tt h e 1 0 H z s i n u s oi d i s d im in i s h e d b y
7 3 % , t h e 7 0 H z by 9 3 % , a n d t h e 30 Hz is u n at t e n u at e d .
( T h e d i st o rt i on s i n t h e s i n u s oi d s a re a rt i f ac t s o f t h e
d i s p l ay n ot t h e c on v o lu t i o n a l g or i th m . )
10 Hz. Maximum amplitude = .27
1 1
0 0
-1 -1
0.3 0.4 0.5 0.6 0.7 0.8 0.3 0.4 0.5 0.6 0.7 0.8
30 Hz. Maximum amplitude = 1.0

1 1
0 30Hz Ricker 0
-1 -1
0.3 0.4 0.5 0.6 0.7 0.8 0.3 0.4 0.5 0.6 0.7 0.8
Convolve
70 Hz Maximum amplitude =.064

1 1
0 0
-1 -1
0.3 0.4 0.5 0.6 0.7 0.8 0.3 0.4 0.5 0.6 0.7 0.8
IN OUT
H e r e i s t h e a c t u al F o u ri e r a m p l it u de s p e ct r u m o f t h e
3 0 H z R ic k er w av e l et .
10
-10
-20
-30
-40
-50
-60
0 20 40 60 80 100
Frequency (Hz)
Since "decibels down" are computed by
dbdown = 20*log10(F(ω)/Fmax)
we can use the results from the previous figure to compute:
dbdown(10Hz) = 20*log10(.27) = -11.4 decibels

dbdown(30Hz) = 20*log10(1.0) = 0 decibels
dbdown(70Hz) = 20*log10(.064) = -23.9 decibels
S o F ( w ) , t h e F ou r ie r T ra n s fo rm o f a fu n ct i o n f ( t ) , i s a
q u ic k wa y o f c om pu t in g t h e re l at i v e a t t e n u at i on o f
d if f e re n t s i n u s oi d s w h e n t h e y a r e c on v o lv e d wi t h f ( t ) .

A co n v ol u t i on ca n a f f e ct n o t o n ly t h e a m p l i t u d e o f a
s i n u so i d bu t i t s p h as e a s we l l . T h e R i ck e r w av e le t i s
kn o wn a s a z er o p h a se f u n c t i on w h ic h m e an s t h a t i t d o es
n ot h av e a p h as e ef f ec t . L e t u s r ep ea t t h e a n al ys i s bu t
t h i s t im e w i t h a f u n c t i on w h i c h h as a kn o wn p h as e
e ff e c t . F or t h i s p u r p os e , w e c o n s id e r a R i c ke r w av l e t
w it h a 9 0o p h as e s h i ft .
0.15 0.15
0.1 0.1
0.05
0.05
0
0
-0.05
-0.05
-0.1
-0.1 -0.15
-0.1 -0.05 0 0.05 0.1 -0.1 -0.05 0 0.05 0.1
30 Hz. Ricker zero phase 30 Hz Ricker 90o phase
N o t e t h at z e ro p h a se wa ve f or m s a re a l wa y s s ym m et r i c
w h il e 9 0o p h as e r e su lt s i n a n a n t i s ym m e t ri c w av e f or m .
W e m i gh t e x p ec t t h e 9 0o R i ck e r t o h av e t h e s am e ef f e c t
o n th e a m p l it u de o f s i n u s oi d s b u t so m e a n d d i t i on a l
e ff e c t a s we l l. T o s ee , w e re p e at th e a n al y si s o f p as s i n g
c om ple x si n u s oi d s th ro u gh i t .

H e r e we re p e at t h e r e su lt o f c o n v ol v i n g 1 0 , 3 0 , a n d 7 0
H z c o m p l ex s in uso id s wi t h a 30 Hz R i c ke r wa v el e t b u t t h i s
t i m e t h e R i c ke r h a s 9 0o p h as e . T h e a m p li t u d e
a tt e n u a ti o n o f t h e s i n u s oi d s i s t h e s am e a s be f or e b u t
n o w t h e re i s a n a d d i t i on al 9 0o p h as e l ag . (When
c om p ar i n g t h i s f i gu r e w i t h - 2- o f t h i s s e ri e s , n o te t h at
t h e re h as b e en a n x - ax i s s c al e c h an g e o n a l l p lo t s. )
R esult with 90o Ricker
10 Hz. Result with 0o Ricker Maximum amplitude = .27
1 1
0 0
-1 -1
0.45 0.5 0.55 0.6 0.65 0.45 0.5 0.55 0.6 0.65
Maximum amplitude = 1.0

30 Hz. 1
1
30Hz Ricker
0 0
90 o
-1 -1
0.45 0.5 0.55 0.6 0.65 Convolve 0.45 0.5 0.55 0.6 0.65
70 Hz Maximum amplitude =.064

1
-1
0.45 0.5 0.55 0.6 0.65 0.45 0.5 0.55 0.6 0.65
IN OUT
H e r e i s a c om p le t e d e s c ri p t i on o f t h e 9 0o , 30 H z . R i c ke r
i n t h e t i m e d o m ai n a n d a m p l it u de a n d p ha s e s p ec t r u m i n
t h e F o u ri e r d om a in . W e h av e s ee n t h at th e F ou r ie r
d o m ai n p r ov i d e s a c on v e n i en t d e s cr i p t i on o f t h e e ff e c t
o f c on v o lv i n g t h e wa v e l et wi t h c om p l e x s i n u s oi d s .
0.1
0.05
Time Domain 0
-0.05
-0.1
-0.15 -0.1 -0.05 0 0.05 0.1 0.15
Time
-20
Fourier Domain
Amplitude -40
Spectrum
-60
0 20 40 60 80 100
Frequency
100
Fourier Domain 0
Phase Spectrum
-100
0 20 40 60 80 100
Frequency
&OURIE R ! NALYS IS AND 3 Y NT HES IS
4 H E G R EA T U T IL I T Y O F T H E & OU R I ER T RA N S FO RM C O M E S F R OM
I T S A B I LI T Y T O D E CO M P O SE A N Y F U N C T IO N I N T O A S E T O F
C OM P L EX S I N U S OI D S ) N T H E C O N T IN U O U S C AS E T HE
F R EQ U E N C I ES O F T H E S I N U S OI D S R A N GE F R OM d T O d A N D
H A V E A M P L IT U DE S A N D P H AS ES W H I C H A R E C OM P U T ED F R OM
T H E F OR W AR D & OU R IE R T R AN S F OR M
d
nI WT
(W HT E DT
nd
4 H IS E Q U AT I ON C OM P UT ES T H E C OM P L E X C O EF I C I EN T S ( W
OF T HE C OM PLE X S I N U SO I D S WHIC H WH E N S UMMED
I N T E GR AT E D W IL L Y IE L D H T 5 S U AL L Y ( W I S D EC O M P OS E D
I N T O T WO S E P AR AT E R EA L F U N C T IO N S

AMPLIT UDE S PEC TRUM !W (W 2E ( W )M ( W
n
)M ( W
PHAS E S PE CT RUM F W TAN
2E ( W
4 HE I N V ER S E & O U RI E R T RA N S FO RM E X P R ES S ES T HE
C ON S T R U C T IO N O F H T A S A S U P E R P OS I T IO N O F C O M P L EX
S I N U S OI D S
d
I WT
HT (W E DW
P nd
) F W E W I S H T O U S E C YC L IC A L F RE Q U E N C Y F I N S T E AD O F A N GU L A R
F R EQ U E N C Y W W P F T H E & O U RI E R T RA N SF O RM P AI R I S
d
n PIFT
(F HT E DT
nd
d
PIFT
HT (F E DF
nd
-ETHODS OF 3EISMIC $ATA 0ROCESSING

Fourier Analysis and Synthesis
A s a n e x am p l e c on s i d er t h e G a u s s i an fu nct i on :
2
–α 2 t
ht = e
U s in g s t an d a rd t e c h n i q u es o f i n t e gr al c al c u lu s , t h e
F o u ri e r t ra n s fo rm o f t h e G a u ss i an c an be s h o wn t o be :
π – ω /4α 2
Hω = α e
half width = 1/α

half width = 2α
h(t) H(ω)
N o t e t h at t h e h a l f wi d t h s , a s re p re s e n t ed by t h e i r 1 / e
p o in t s a re in v e r se l y p ro p or t i on a l. I n f ac t :
ΔtΔω = α 2α = 2
–1
T h i s i s a n e x am p l e o f a g en e r al p r op e rt y wh i c h s ay s t h a t
t h e " wi d t h " o f a ti m e d om a i n fu n c ti o n is i n v er s el y
p r op o rt i on a l t o it s wi d t h i n f re q u e n c y. I t ca n be s h ow n ,
g iv e n a s u i t ab l e m e as u re o f w i d t h , t h at :
( w id th i n t i m e ) ( w id th i n f r eq u en c y ) > = a co n s t an t
B r ac e we l l ( 1 9 7 8 , T h e F o u ri e r T ra n sf o rm and its
A p p l ic a ti o n s) s h ow s t h e c o n s ta n t t o b e 1 /2 a n d t h at t h e
e q u al i t y h ol d s f or th e G a u s s i an .

Fourier Analysis Example
0.08
0.06
0.04
H er e i s a m i n i m u m p h as e
0.02
w av e l et c on s t ru ct e d wi t h
0
a . 0 0 1 s e c s am ple r at e
-0.02
a n d a 3 0 H z d om in a n t
-0.04
f r e q u en cy .
-0.06
-0.08
0 0.05 0.1 0.15 0.2
time (sec)
T h e i s t h e " a m p l it u de
1
0.8 s p e ct r u m o f t h e w av e l et
0.6 d i s p l ay e d wi t h a l i n ea r
0.4 v e rt i c al s c al e . N o t e t h at
0.2 t h e f r e q u en c y a x i s s t op s
0
0 100 200 300 400 500 at 500 Hz wh i c h is
frequency (Hz)
1 /( 2 * . 0 0 1 se c ) .
0
-20
H e re the amplitude
-40
sp ec t r u m i s d i s p l ay e d w i t h
a d ec i b el v e r t ic al sc a le :
-60
db =
-80
0 100 200 300 400 500 2 0 *l o g1 0 ( A ( f ) /A m ax )
frequency (Hz)
3
2 T h i s i s t he ph as e s p e c t r um .
1 N o t e t ha t t h e ve r t i c a l s c a l e
0
-1 i s i n r ad i an s .
-2
-3
0 100 200 300 400 500
frequency (Hz)
A t t h i s po i nt , F o ur i e r an a l ys i s m a y l o o k l i k e an e xe r c i s e i n
g r a ph m a ki n g ; ho w e ve r , i t s ut il i t y w i l l b e c o m e c l e ar o n t h e
n e xt pa g e .

Fourier Analysis Example
A
B
Sum of components Sum of components

80 80
60 60
40 40
20 20
Cumulative sum of Fourier

Individual Fourier components
components
0 0 0.05 0.1 0.15 0. 0 0 0.05 0.1 0.15 0.2
2
time (sec) time (sec)
H e re we s ee t wo e q u i v al e n t wa y s o f v i e wi n g t h e F o u r ie r
t ra n s fo rm in fo rm a t i on o n t h e p r e v io u s p ag e. I n A , t h e
i n d i v id u al F ou r ie r c om po n e n t s a r e s h ow n f r om 1 0 t o 7 0
H z , p ro p er l y s c al e d f or t h ei r a m p l i t u d e a n d p h as e . T h e
s u m o f a l l 1 3 c om p on e n t s y i e ld s th e w av e l et a t t h e t op
w h ic h i s q u i t e si m i l ar t o t h e t r u e w av e l et s h o wn o n t h e
p r ev i o u s p a g e. A d d i n g i n t h e r em a i n in g f r eq u e n c y
c om po n e n t s ( 0 - > 1 0 H z a n d 7 0 - > 5 0 0 H z ) wi l l re c on s t r u c t
t h e wa ve l e t e x ac t l y. T h e f i gu r e o n t h e ri g h t c on t ai n s t h e
s am e in fo rm a t i on ex c e p t t h at ea c h tr ac e is t h e s u m o f
t h e f re q u e n c y c om p o n en ts be t we e n i t s f re q u e n c y a n d 1 0
H z . T h i s g i v es a g oo d i ll u s t ra t io n o f h o w t h e wa v el e t
t ak es fo rm a s it s sp ec t r u m i s s u m med .

Fourier Transform Pairs
The table below is reproduced from:
Brigham, E.O., 1974, The Fast Fourier Transform, Prentice Hall
N o t e : I t i s a r em a rk ab le f ac t t h at n o s i gn a l c a n h a v e
fi n i t e l en g t h ( i . e . c om p a c t s u p p or t ) i n bo t h t h e t i m e
a n d fr e q u e n c y d o m a in s .

Fourier Transform Pairs
The table below is reproduced from:

Brigham, E.O., 1974, The Fast Fourier Transform, Prentice Hall

The Dirac Delta Function
T h e D ir ac d el t a fu nct i on wa s i n v en t e d b y P . A. M . D i ra c t o
h a n d l e p ro bl e m s i n t h e d e v el op m en t o f q u a n t u m
m e c h an i c s . S i n c e t h e n , i ts u n i q u e a b il i t y t o r e p re s en t a
" un it s p ik e " i n t h e c on t i n u ou s f u n c ti o n d om a i n . I t ca n b e
d e f i n ed a s t h e l im it i n g fo rm o f a s h ar p ly p e ake d f u n c t i on
w h os m a x i m u m p ro c ee d s t o i n f i n it y a s it s wi d t h sh r i n ks
t o z e r o i n s u c h a w ay t h at it s a re a re m a in s u n i t y.
8
A se ri e s o f b ox c ar s w i t h
7
b4
u nit a r ea c on v e rg e s i n
6 t h e l im it t o t h e d e lt a
f u n c t i on :
5
4 b∞ = δ t
b3
3
b2 It can be thought of as:
2
b1
1 0, t≠0
δt =
0
-0.5 -0.25 0 0.25 0.5 ∞, t=0
T h e m os t i m p o rt a n t p r op e rt y o f t h e d e l t a f u n c t i on i s i t s
be h av i or u nd er i n t e gr at i on . I f f ( t ) is a n y f u n c t i on , t h e n :
b f t0 , if a<t0<b
f t δ t–t0 dt =
a 0, otherwise
This is known as the sifting property of the delta function.

The Dirac Delta Function
Consider the Fourier transform of the delta function:
∞
–i ωt –iωt0
δ t–t 0 e dt = e
–∞
T h us i t h a s a c on s t an t , u n i t a m p l i t u d e s p ec t r u m ( a ls o
kn o wn a s a " wh i t e " s p e c t ru m) a n d l i n e ar p h a se .
Consider the action of the delta function under convolution:

∞
δ t–t 0 f τ–t dt = f τ–t0
–∞
T h u s t h e d e lt a f u n c t i on sh i f t s f ( t ) to p l a ce i t s o ri g in a t
t h e l o ca t io n wh e r e t h e a rg u m e n t o f t h e d el t a f u n c t i on
v an i s h e s. T h i s i s c a ll e d a " st a ti c sh if t " i n s e is m i c d a t a
p r oc e ss i n g . S i n c e c on v ol u t io n c an b e d o n e in th e
F o u ri e r d o m ai n b y m u l t i p l ic a t io n o f t ra n sf o rm s , we c an
c on c l u d e t h a t a s t at i c s h i ft c an be d o n e b y:
fτ FFT Fω Mult IFFT f τ–t o
–iωt o
e
T h at i s , a s t at i c sh i f t i s e q u i v al en t t o a l i n e ar p h as e
sh i f t . F i n a ll y, if we i n v e rs e F o u ri e r t ra n sf o rm t h e
eq u a t io n a t t h e to p o f t h e p ag e, w e en d u p w i t h a
d ef i n i t io n o f t h e d e lt a f u n c t io n i n t e rm s o f i t s F o u ri e r
co m p o n en t s :
1 ∞ ∞
δ (τ − t 0 ) =
2π
∫ −∞
e
iω (τ − t0 )
= ∫ −∞
e
2 π i f ( τ −t0 )
T h us t h e d e l ta f u n c t i on h as u n i t a m p li t u d e sp e c t ru m
a n d a p h as e s p e c t ru m t h at is li n e ar i n f r eq u en c y a n d
w it h s lo p e p r op o rt i on a l t o t h e t i m e s h i ft .

The Convolution Theorem
C o n s id er t h e c o n t in u ou s c o n v ol u t i on o f f a n d g:
∞
ht = f τ g t–τ dτ (1)
–∞
W e c an r e p re s en t f a n d g i n t er m s o f t h e i r sp ec t r a
a s: ∞ ∞
1 i ωτ 1 iϖ t–τ
fτ = F ω e dω an d g t–τ = Gϖe dϖ
2π –∞ 2π –∞
S u bs t i t u t in g t h es e i n t o ( 1 ) :
∞
∞ ∞
1 1
iωτ iϖ t– τ
ht = F ω e dω Gϖe dϖ dτ
2π –∞ 2π –∞
–∞
∞
Interchanging ∞
1 1 i ω– ϖ τ iωt
FωGϖ dτ e dωdϖ
the order of
integration ht = e
2π 2π –∞
–∞
∞
1 i ωt
The term in [ ] is the
ht = F ω G ϖ δ ω–ϖ e dωdϖ
Dirac delta function.
2π –∞
∞
The d el ta functi on 1 iωt
col lap ses one of the ht = F ω G ω e dω
f r eq uency inte gra ls 2π –∞
H er e w e h a ve h ( t ) re p r es e n t ed a s t h e i n v e r se F o u r i er
t r an s f or m o f " s om e t h i n g " . B y i n fe r en c e , t h a t so m e th in g
m u st be t h e F o u ri e r t ra n sf or m o f h . T h u s :
Hω = FωGω

The Convolution Theorem
T h e r es u l t we h av e j u s t d e ri v e d is o n e o f t h e m o st
f u n d am en t a l a n d i m p or t an t in a l l o f si g n al p ro ce s s in g . I t
t e ll s u s t h a t we c an c on v ol v e t w o s i gn a ls b y m u l t i p ly i n g
t h e ir s p ec t r a a n d i n v e rs e F o u r i er t ra n sf o rm i n g t h e
r es u l t . T h e re as on t h at t h is is i m p or t an t is t h at t h e r e i s
a n e x t r em e l y f as t a l g or it h m fo r p er f or m i n g t h e d i g it a l
F o u ri e r t r an s f or m c al l ed t h e f a st F ou r ie r t ra n sf o rm
( FF T) . Us i n g t h e F FT a c on v ol u t i on c an be d o n e b y:
f(t) g(t)
FFT FFT
F(ω) G(ω)
Multiply
H(ω)
IFFT
h(t)
N o t e t h at m u l t ip l y in g co m p l e x s p e c tr a i s:
iφF ω iφG ω
H ω = F ω G ω = AF ω e AG ω e
i φ F ω + φG ω
= AF ω AG ω e
T h a t i s w e c an v i ew i t a s m u l t i p l yi n g t h e a m p l it u de
s p e ct r a a n d a d din g t h e p h as e s p e c tr a.

Sampling
T h e a n a ly t ic a n al y si s o f c on t i n u ou s s ig n al s i s m o s t u s ef u l
f or g ai n in g a c o n c ep t u a l u nd er s t an d i n g o f si g n al
p r oc e s si n g . I n a c t u al p ra ct i c e ; h ow e v er , t h e v as t
m a j or i t y o f w or k is d o n e wi t h d i s c re t ly s am p l e d
f u n c t i on s . T h e p r oc e ss o f s a m p l in g a co n t i n u ou s f u n c t i on
i n t i m e ca n be v i ew e d a s a m ult i p l i c at io n b y a s am p l in g
c om b.
T im e Do m a in Fr e q u e nc y D o ma in
Continuous
Gaussian
Continuous
spectrum
Gaussian
Times Convolved with
F o ur i er
1/Δt t r a ns fo r m
Sampling
of
Comb
s am pl i ng
c om b
Comb spacing = Δt
Equals Equals
Sampled Gaussian
Gaussian Spectrum
and aliases
1/Δt

Sampling
S o w e h av e s e en th a t s am p l i n g i n t h e t i m e d om a i n
c a u se s t h e re p l i ca t io n o f th e c on t i n u ou s s p ec t r u m i n
t h e f r eq uen c y d om a i n . T h e s p a ci n g b et w e en t h es e
s p e c t ra l a l i as es i s 1 / Δ t a n d i t i s c u s t om a ry t o re s t ri c t
o u r a t t en t i on t o t h e p ri m ar y fr e q u en cy ba n d l i ei n g
b e t we en - 1 /( 2 Δ t) a n d 1 /( 2 Δ t) . T h e f r eq uen c y F n =
1 / ( 2 Δ t ) i s c al l e d t h e N yq u i s t fr e q u e n c y a n d i s t h e
l i m i t in g f re q u e n c y o f t h e s am p l ed d at a .
Fnyquist = 1/(2Δt)
Spectrum of
sampled
data
showing
aliasing.
F -2Fn -Fn 0 Fn 2Fn
S p e ct r u m o f
s am p l ed
d a t a wi t h
m in i m a l
a li as i n g .
F -2Fn -Fn 0 Fn 2Fn

Primary
frequency band

Sampling
T h e u n a li a se d s am p l i n g o f a n y c on t i n u ou s s i gn a l r eq uir e s
t h a t t h e s i gn a l h av e i t s p o we r r es t r ic t e d t o a f re q u e n c y
b an d: - f m ax < f < fm a x . S u c h s i g n al s a r e s ai d t o b e
b an d li m i t e d . A ba n d l im it e d s i gn a l c a n be d i gi t al l y
s am p l ed , wi t h ou t a l i as in g , w i t h a s am p l e s i ze o f
Δ t = 1 / ( 2 f m ax ) . I t i s a f u n d a m en ta l th eo re m ( T h e
S am p l i n g T h eo re m , P ap o u l i s, S i gn a l A n a ly s is , p 1 4 1 ,
1 9 8 4 ) t h at s u c h a b an dli m i t e d , c on t i n u o u s , s i gn a l c an
b e e x ac t l y re c ov e r ed f ro m i ts d i gi t al s am p l e s b y a
p ro c e s s kn o w n a s " s in c f u n c t io n i n t e rp o l at i on " .
Ti me D o ma in F r e q ue n cy Do m a in
Spectrum
S ampled band limited
of sampled,
function
unaliased,
continuous
function
Interpolation
site C on volve d
w i t h a si n c
f un c ti on Mu l t ip l ie d
by a
b oxca r
Recovers the original

continuous function
Re cove rs the
s p ectr um of
the conti nuous
fu nction

Sampling
I n o r d er t o m i n i m i ze a l i as i n g, r aw a n a l og s ei s m i c d at a i s
p as s e d t h r ou g h a n a n al og a n t i al i as f i l t er p r i or t o
d i g it i za t io n . A t y p ic a l a n t i al i as fi l t er h as a n a m p l i t u d e
s p e ct r u m w h i c h be g i n s t o r ol l o f f a t 5 0 % t o 6 0 % o f
f n yq uis t a n d re ac h e s v er y la rg e a t t en u at i on ( > 6 0 d b ) a t
f n yq uis t .
0
H e re is t he
s p ec tr u m o f a n -20
ant ia l ia s f il te r
-40
f o r us e pr io r to
sampli n g a t -60
. 0 04 s ec .
-80
-100
-120
0 2 4 6 8 10 120 140
0 0 0 0 0
Frequency (Hz)
R u l e o f t h u m b : S a m p le y o u r d at a s u c h t h at t h e
e x p e c t ed s ig n a l f re q u e n c ie s a re l e ss t h an h al f f n yqu ist .
s am ple
r at e N y q u i st
.008 s 6 2 .5 H z
C om mo n s a m p l in g
.004 s 125 H z
ra t es and their
N y q u i st f re q u e n c ie s .002 s 250 H z
.001 s 500 H z
A l i as i n g i s a l s o a p o s s ib i li t y w h e n re s am p l i n g se i s m i c
d at a . I f t h e n e w sa m p l e i n t er v al i s m o re c oa rs e t h an t h e
o l d , th en a n a n t i al i as f i lt e r s h ou l d be a p p l i ed .
The Discrete Fourier Transform
T h e gr ea t u t il it y o f th e co nt i nu ous F o u ri er tr a n sf o rm t o
de c o m p o se f un c t io n s in t o f u n d a m en t a l co m p l ex s in u s o id s
ca n be a p p li ed d ir ec tl y to d i sc re t ely s a m p l ed ti m e
do m a in fu n c ti o n s. Co n si d er a fu n c ti o n h( t ) w h ic h i s z e ro
ev ery w h er e ex c ep t a t N t im e s d ef in e d b y t= k Δ t , k = 0 ,1 , 2
.. . N - 1 , w h e re i t t a k e s t he va lu es h k . T h is fu n c ti o n ca n
be w ri tt e n w i th t h e d ira c d e lt a f u n ct i o n a s :
N–1
ht = Σ h kδ t–kΔt
k = 0
If we now take the Fourier transform of h(t) we have:

∞
N–1 N–1 ∞
Hω = Σ h kδ t–kΔt e
–i ωt
dt = Σ h k
–∞
δ t–kΔt e
–i ωt
dt
k = 0 k = 0
–∞
N–1
Hω = Σ h ke
–i ωkΔt
k = 0
He r e w e ha v e a n a na l y tic e x pr e ssi o n fo r t he F o ur i e r
tr a n sfo r m o f th e h k sa m p le s w h ic h is d e f in e d f o r a l l ω .
We ha v e a lr e a d y se e n th at t he p h e n o me no n o f a li a sin g
lim it s th e us a b le fr e qu e n cy b a n d t o -π /Δ t - > + π / Δ t .
Fu r t he rmo r e, l in e a r al g e b r a te lls u s th a t N f r e q ue nc ie s in
th is b a n d sh o ul d s uf fic e to d e t e r m in e th e N h k . S o w e a r e
le a d t o c o ns id e r sa m pl in g th e fr e q u e nc y d o m a in a t ω ν =
2πν/(NΔ t) , ν = 0 ,1, 2 .. . N - 1.
N–1
Hυ = Σ h ke
–i2πυk/N
k = 0

D i sc r e te e x p on e n t i al s h av e a w e ll k n ow n o r t h og on a li t y
p r op e r ty s u c h t h at :
U s in g t h i s , i t i s n o t d if f i c u l t t o sh o w t h a t t h e h k s am p l e s
c an be re c ov e re d fr om t h e H ν by :
N–1
1
hk = Σ H υe
i2πυk/N
Inverse DFT
N υ = 1
T h i s r e su lt t og et h e r w it h :
N–1
Hυ = Σ
k = 0
h ke
–i2πυk/N
Forward DFT
f or m t h e d i sc r et e F o u ri e r t r an s f or m p a ir . T h e y a re t h e
d i r ec t a n al og t o th e c o n t in u ou s F o u r ie r t ra n sf o rm
r el at i on s . L i ke t h e F T, t h e DF T i s c om p l e t e i n t h at t h e h k
a r e e x ac t l y r e c ov e ra b l e f ro m t h e i r s p ec t r u m , t h e H ν .

H e r e i s a p i ct o ri al r e p re s en t a t io n o f t h e d e v el o p m e n t o f
t h e D F T fr om t h e c on ti n u o u s c as e :
T im e Do m ai n F r e q ue n cy Do m ai n
Sampled band limited Spectrum is Spectrum

function N samples long periodic of sampled,
with unaliased,
Δt period continuous
2πN/Δt function.
-fnyq fnyq
Co nv o l ve d wi th th e
tr a nsf or m o f t h e 1/Δt
samp li ng c o mb
1/Δf Times a
sampling Δf
comb
Principle DFT The

s am pl ed
band
s pe ct ru m
Principle
band
T ID FT -fnyq fnyq
The sampled time series
becomes periodic
with period T=NΔt fnyq = 1/(2Δt) T = 1/Δf ΔfΔt = 1/N

T h e s am p l i n g o f t h e s p ec t r u m o f a d i s cr e t e t i m e s e ri e s
c au s e s t h at s e ri e s to b ec o m e p e ri od i c w i t h p e ri od T =
N Δ t. T h is h as s i gn a l p r oc e s si n g c on s e q u e n ce s t h at a r e
a p pa r en t wh en w e c on s i d er a p p l y in g a f il t e r wi t h t h e
D FT a n d t h e c o rr es p o n d i n g c on v o lu ti o n .
Time series Principle Period

showing time
domain aliases
Filter operator:
T h e c on v o l u t io n o p e r at i on t h at d up li c at e s m u l t ip li c at i on
w it h t h e D FT i s c al l ed c i rc u l ar c o n v ol u t i on . N ot e t h at t h e
f i lt e r o p er at or p la c ed o n t h e l as t s am p l e o f t h e p r in ci p l e
p e ri o d a p p e ar s t o " wr ap a r ou nd " a n d a f f e c t t h e fi r s t
s am p l e . T o a v o id t h i s p ro bl e m , i t i s c om m o n t o p ad t h e
t i m e s er i es wi t h a l e n gt h o f z e ro s c h os e n w it h t h e l e n gt h
o f t h e f il t e r o p e r at or i n m i n d .
Principle Period
Zero
Pad

The Fast Fourier Transform
T h e f as t F o u ri e r t r an s f orm ( FF T ) is n ot h i n g m or e t h a n a
c l ev e r w ay o f c al c u l at i n g t h e D FT wh i c h ge t s i m p r es s iv e
p e r fo rm a n c e r es u l t s . T h e c on v ol u t i on o f a n N l en g t h
o p er at o r i n t h e t im e d o m ai n r e q u i re s o n t h e o r d er o f N 2
f l oa ti n g p oi n t o p e r at i on s . T h e s am e c om pu t at i on i n t h e
f r eq uen c y d o m ai n wi t h t h e F F T r e q u i re s r ou g h l y N * lo g( N )
o p er at i on s . H o w ev e r, we m ust be c ar e f u l wi t h th is
s t at e m e n t b ec au s e , ge n e ra ll y , t h e tw o N ' s a r e n ot t h e
s am e. T h i s i s be c au s e t h e F F T a l go ri t h m re q u i r es t h at
t h e ti m e s er i es le n g t h b e a " m ag i c n u mbe r" w h i c h is
u s u al l y a p ow e r o f 2. ( Al s o t h e t w o ti m e se r ie s b e in g
c on v ol v ed m u s t be t h e s am e l e n gt h . ) T h i s is a c h ie v e d b y
a tt a c h in g a z e r o p ad t o t h e t im e s er i es . T h u s if N i s t h e
l e n gt h o f t h e t im e d o m ai n o p e ra t or a n d i f N 2 i s t h e f i rs t
p o we r o f 2 g re at e r t h a n N , t h e n we m u s t co m p a re N 2 t o
N 2 lo g( N 2 ) . ( O f t e n e v en t h i s i s n o t e n ou gh be c au s e t h e
z er o p ad m u s t b e l o n g e n ou g h t o a v o i d o p e ra t or w r ap
a ro u n d . ) T h e b ot t om l in e is t h at s h o rt o p e ra t or s ( l e ss
t h a t ~ 6 4 p oi n t s ) a re o f t e n a p p l ie d f a st e r wi t h
c on v ol u t io n w h i l e lo n g o p er at o rs a re M U C H f as t e r wi t h
F FT ' s . T h e d ia gr am b el ow i s a d ap t e d f r om H at t o n e t a l.
a n d s h o ws t h e ba s i c t r ad e of f .
Time domain
Convolution
compute time
FFT
Operator Length
Filtering
W e h av e se e n t h at c on v o lu t i on w it h a wa v e f or m
su r p r es s e s a n d p o s s ib l y p ha s e sh i f t s s om e f re q u e n c i es
re l at i v e t o o t h e rs . T h is f i l t er i n g a c t io n i s o f t en ex p l o it e d
t o e n h an c e s i gn a l a n d su r p r es s n o is e . H e re w e se e a
co m p a ri s on o f f i v e d i ff e re n t z er o p h as e f i l te r s i n b ot h
t h e t im e a n d f re q u e n c y d om a i n s. T h e i n v e rs e r el at i o n sh i p
be t we e n t e m p or al w i d t h a n d f re q u e n c y ba n d wi d t h is
re ad i l y a p p a re n t .
Wavelet 1
Five Wavelet 2
Generic Wavelet3
Wavelets
Wavelet 4
Wavelet 5
-0.15 -0.1 -0.05 0 0.05 0.1 0.15
-20
-40
Their Wavelet 2 Wavelet3 Wavelet 4 Wavelet 5
-60
Fourier Wavelet 1
-80
Am pl itu de
-100
Spec tra
-120
0 50 100 150 200

Frequency (Hz)

The Z Transform
T h e p er i od i c i ty o r c i rc u l ar i ty i n h e re n t i n bo t h t i m e a n d
f r eq uen c y is n i c el y c ap t u r ed b y a p o we rf u l m e t h od o l og y
k n ow n a s t h e Z tr an s f or m . C on si d e r th e t i m e s er i es , [ 1 -
. 5 - .3 0 . 1 0 ] , w h er e i t is a s su m ed t o s t ar t a t t = 0 a n d
i n c r em e n t by ⎯ t . W e re p r es e n t th is s er ie s in t h e Z
d o m ai n by a p o ly n om i a l i n z :
0 1 2 3 4
H z = 1z –.5z –.3z +0z +.1z
1 2 4
= 1–.5z –.3z +.1z
S o w e s e e t hat t h e e x p onent o f z gi v e s t h e s am p l e n um b e r
a n d he n c e determines t h e s a m pl e t i me ( nΔt ). N ot e a l s o
t he f o l lowing:
• Negative times correspond to negative exponents of z
• M u l t i pl i c a t i on b y z n d e l ay s t h e t i m e s e r i e s by n
s a mp l es i f n i s p os i t i ve a nd a dv a n c es i t b y n s am pl e s
f o r n e g at i v e n .
T h e g re at u t il i t y o f t h e Z t r an s f or m li e s i n it s a b il i t y t o
r ep re s en t d i sc r et e co n v ol u t i on a n d t h e D FT a s o p e r at i on s
w i th p ol y n om i al s . I t i s n o t d i f fi c u l t t o sh o w t h at t h e
c on v ol u t io n o f t wo t i m e s e ri e s , f a n d g, c an b e re al i ze d
b y s i m p l y m u l t i p l yi n g t h e i r Z t r an s f or m s a n d r ea d in g o f f
t h e r e su l t . ( S ee W at e rs ( p 1 3 3 ) f or a p r oo f. )
1 2 1 2
F z = f0 +f1z +f2z + ... G z = g0 +g 1z +g 2z + ...
Hz = FzGz =
1 2
f0 g0+ f0 g1+g0f1 z + f0g 2+f1g 1+g0f2 z + ...
h = f• g

The Z Transform
T he f a ct t ha t c o nv o lu t io n i s d o ne b y m u lt i p li c at io n o f Z
t r a n s fo r m s i s r e m in is c e nt o f t he F o u r ie r t r a n s fo r m . I n
f a ct , i f w e l e t z = e - iω Δ t t h e n t he Z t r a ns f or m b e c om e s :
N–1 N–1
Gz = Σ gkz
k
Gω = Σ gke
–iωkΔ t
k = 0 k = 0
A s wi t h t h e D FT , i f w e n o w c on s i d e r o n l y d i s c re t e
f re q u e n c i es ω ν = 2 πν/( N Δ t ) , ν = 0 , 1, 2 . .. N - 1 , t h en w e
s ee n t h at t h e Z t r an s f or m , w i t h z = e - iω Δt , is p r ec i s el y
t h e D FT .
N–1
Gν = Σ g ke
–i2πυk/N
k = 0
T he Z transfo r m is more general t ha n th e DF T s in ce z c an b e

a ny c o mp le x number. I n f a ct t he D F T amount s to evalua t ing
th e Z tranfo r m a t N d iscrete l o ca tio n s around th e un it cir c le
in th e comple x z p lane.
imag(z)
Complex z plane ω2
ω1
ωo
real(z)
ων
ωN-1
ων+1
The Z Transform
C o n s i d e r t h e el e m e n t al c ou pl et F( z) = 1 -a z. N ow i f w e
c o n v ol v e F( z ) wi t h an o t h e r ar b i t ra ry t im e s e ri e s g ( z ) ,
t h e n w e r ep re s e n t t his as : H ( z) = F( z ) G ( z) . Su pp os e t h at
o n l y F( z) an d H ( z) ar e k n o wn t o u s an d w e w i sh t o
r e c ov e r G ( z) . In t h e z t r a n s fo r m d om ai n we c an s im p l y:
Hz
Hz = FzGz ∴ Gz =
Fz
So we define the inverse of any time series as:
–1 1
F z =
Fz
For F(z) = 1 -az, this gives:
–1 1 2 3
F z = = 1+az+ az + az +
1–az
T h i s s er i e s , c a l le d t he g e om et r i c s er i e s , i s k n ow n t o
c o n v e r ge a b so l u t e ly p ro v i d e d t hat | az | < 1 . S i n c e w e ar e
e s p e c i al l y i n t e re s t e d in t h i s re s u l t ev a l u at e d on t h e u n i t
c i r c l e ( | z| = 1 ) t h e n w e n e e d | a| < 1 . It i s c ust o m a ry t o
t a l k a b ou t t he lo c at i o n of t he "ze r o" of t h i s c o u p l e t
d ef i n e d b y:
1
1–az 0 = 0 ⇒ z0 =
a
I f |a | < 1 , t h e n w e se e t h at z o m u st li e o u t s i d e t h e u n i t
c i r c le in o r d e r f or th e i n v er s e t o c on v e rg e . S u ch a n
i n v e r se is sa i d t o b e s t ab l e ( p h y ic a ll y r ea li za bl e ) . N ot e
a l so t h a t F ( z ) i t s el f i s t r i v ia l ly s t ab l e.

The Z Transform
A n y c a u s a l, st a b le t i m e se ri es wi th a ca u sa l , s t a bl e in v e rs e
i s s a i d t o b e m i n im um p h ase . T h u s o u r e le m en t a l c o u p le t ,
1 - a z , is m i n im u m p h a s e wh en e v er |a |< 1 . A n y m o r e c o m p l ex
t im e se ri es c a n a lw a y s be fa ct o r ed i n to a se t o f
el e m en t a r y c o u p l et s .
N–1
Gz = Σ k
g kz = z–z 0 z–z1 z–zN–1
k = 0
W e s a y t h a t G ( z ) i s m i ni m u m p h a s e i f a ll i t s e le m e n ta l
c o u pl e ts a r e m i ni m u m p ha s e . T h a t i s e q ui v a le n t t o
s a y in g t h a t a l l o f t h e r o o ts o f t he p o ly n o mi al G (z ) m us t
l i e o u ts i d e t h e u n it c i r cl e i n t h e c o m pl e x z p la ne . I f a ll
r o o ts l i e i n s i d e t he u ni t c i r c le , G (z ) i s s a id t o b e
m ax im um p ha s e a nd o th e r w is e i t i s m ix e d p h as e .
imag(z) •
z=z0
Complex z plane
real(z)
•
• z=z1
z=zN-1
A minimum ph as e t ime series ha s al l i ts zeros outside the

unit circle.

The Z Transform
T h e z er o s o f F ( z ) c o r re sp on d t o p o l es f o r F -1 ( z ) . T h u s, fo r
t h e c a se o f a t i m e se ri es w h o s e Z tr a n s fo rm h as a
d e n om i n a t or, we s ee th at t h e s t a bi li t y c o n d it i o n re q u ir es
t h a t a l l p o l es a l s o l i e o u ts i d e t h e u n it ci rc l e. T h e m o s t
g en e ra l t im e s er ie s c a n b e w ri t te n a s a Z tr a n s fo rm w it h
b o t h n u m e ra t o r a n d d en o m i n a t o r s uc h a s:
Az z–α 0 z–α 1
Hz = =
Bz z–β 0 z–β 1
W e s a y t h e c o r r e s p o nd in g t im e s e r ie s i s m i n im um p h as e
i f a ll α i a n d a ll β i l i e o ut s i d e t h e u n it c i r c le . T h e
f ol lo w in g t he o r e m f o ll ow s i m me d ia te l y:
T h e r es ul t an t o f t he s e quent i al c onv ol ut i on o f a n y
nu m ber o f m i n i mu m phas e time series i s al s o
min im um ph a s e.
Conversely:
I f a ny ti m e s e r i e s i n a se qu e nce of co nvo l u ti o n s i s
n ot mi n i m um ph a s e , th e n th e r e s ulta nt is no t
m i nimu m p ha s e .
T h o ug h th e se s ta te me nt s se e m ir o n c la d , k e e p in m in d
t he u n sta t e d a s su mp t io n t ha t al l th e se tim e se r ie s h a ve
t he s am e sa m p le r a te . T hu s th e r e s a mp li ng o f a ti me
s e r ie s is a n o p e r a ti o n w h ic h l ie s o ut sid e t he sco p e o f
t he s e th e o r ems .

Crosscorrelation
G i v e n t w o si g n al s , r a n d s , t h e c r os s c or re l at i on p ro v id es
a n u m e ri c al ch a ra c t er i za t i on o f th ei r s i m i la ri t y.
T h e c a lc u l at i on o f s ⊗r
Z e ro l ag :
s0 s1 s2 s3 s4 s5 •••
Multiply aligned samples and sum:
r0 r1 r2 r3 r4 r5 •••
c0 = s0 r0+s1 r1+s2 r2+
F i r s t p o si t i v e l ag :
s0 s1 s2 s3 s4 s5 •••
r0 r1 r2 r3 r4 r5 •••
c1 = s0 r1+s1 r2+s2 r3+
F i r st n e ga ti v e l ag :
s0 s1 s2 s3 s4 s5 •••
r0 r1 r2 r3 r4 r5 •••
c–1= s 1r0 +s 2r1+s 3r2+

C r o ss c o r r el a ti o n
T h e ge n e ra l f or m f o r t h e c ro s sc o rr el at i on o f s a n d r c an
b e w ri t t en :
cj = Σs r
k
k k+j
O r, fo r c on t i n u o u s s ig n al s :
cτ = s t r t+τ dt
–∞
P r op e rt i e s o f cr os s c or re l at i on s :
• I f ei t h e r s o r r is a n i n f i n it e l en g t h ra n d om s i gn a l ,
t h e n c j = 0 f or a l l j .
• T h e m ax i m u m o f c d e fi n e s t h e " l ag " a t wh ic h s a n d
r a r e m o st s im il ar wh e n a l ig n e d .
• A c r os s c or re l at i on c an be c om p u t e d by ti m e
re v e rs i n g s a n d c o n v ol v in g. C a n y ou p r ov e t h i s ?
• Th e a u t o co rr e la t io n is a sp ec i al c as e of
c ro ss c o rr el at i on w h en r =s .

Autocorrelations
T h e a u to c o r r e la t io n, φ , of a sig n a l, s, is a
ch a r a ct e r iz a t io n o f i ts se l f sim i la r it y . It c a n b e c o mp u te d
a s f o llo w s:
Z e ro l ag :
The signal s s0 s1 s2 s3 s4 s5 •••
A copy of s s0 s1 s2 s3 s4 s5 •••
φ0 = so +s 1+s2+s3 +
2 2 2 2
F i r st p os i t i v e la g:
s0 s1 s2 s3 s4 s5 •••
s0 s1 s2 s3 s4 s5 •••
φ1 = so s1+s1 s2+ s2s 3+
S e c on d p o si t i v e l ag :
s0 s1 s2 s3 s4 s5 •••
s0 s1 s2 s3 s4 s5 •••
φ2 = so s2+s1 s3+ s2s 4+

Autocorrelations
The general form for the autocorrelation of s can be written:

length (s)
φj = Σ
k = 0
sk sk+j
Properties of the autocorrelation:
• φ o >= φ j for all j. The zero lag is always largest.
• I f s is a n in f i n it e le n gt h r a n d o m se q u en c e , th e n
φ o g iv e s th e s u m o f sq u a r es o f th e s eq u e n ce
a n d a l l o t h er φ j a r e z er o .
• The Fourier transform of the autocorrelation gives

the power spectrum (squared amplitude spectrum)
of the signal, s.
• The autocorrelation has no phase information.
The autocorrelation is often normalized such that φ o=1:
length (s)
Σ sks k+j
φj =
k = 0
length (s)
Σ
k = 0
sk
2

Spectral Estimation
T h e ge n e ra l p r ob l em o f e st i m a t in g a m p l i t u d e o r p o we r
s p e ct r a o f a n u n k n ow n s i gn a l em be d d e d i n n oi s e o r o t h e r
u n w an t e d s i gn a ls i s c al l ed sp ec t r al es t i m at i on . I t a r is e s
i n m a n y c on te x t s i n s ei s m i c d a t a p ro c es s i n g b u t m o s t
n ot a bl y i n d ec on v ol u t io n t h e or y. T w o sa m p l e p r ob le m s :
• G iv e n a sm a l l n u m b er o f l ag s o f a p os s i bl y i n f in i t e
a u t o c orr e la t io n , e s ti m a t e t h e p o we r s p ec t r u m o f t h e
u n d e rl y in g p h ys i c al p r oc e s s.
• G i v e n a s m al l p o rt i on o f a p os s ib l y in f i n i t e t i m e
s e ri e s , es t i m at e t h e a m p l it u de sp ec t r u m o f t h e
u n d er l yi n g p h y s ic al p r oc es s .
T h e s e t wo p ro bl e m s a r i se re p e at ed ly , a n d i n a v ar i et y o f
c on t e x t s , i n s ei s m i c d at a p ro ce s s in g t h e or y. H ow e v er ,
t h e y a re e s se n t i al l y s i m i l ar d i f f er in g o n l y in t h e n at u r e
o f t h e i n p u t : e. g a n a u t o c or re l at i on o r a g e n er al t i m e
s er i es .
W e s h al l c on s i d e r t w o a p p r oa c h e s : t h e wi n d o we d D FT ,
a n d t h e m a x im um e n t ro p y s p ec t r u m ( B u rg s p e c t ru m) .
C on si d e r t h e co n s t ru c t i on o f a n el e m e n t ar y s e is m o gr am
b y co n v ol u t i on :
Wavelet
Seismogram
Reflectivity
0 0.2 0.4 0.6 0.8 1

Spectral Estimation
I f we c om p u t e t h e a u t o c or re la t io n s of t h es e t h r ee
f u n c t i on s , w e o b t ai n :
Autocorrelation of wavelet
Autocorrelation of seismogram
Autocorrelation of reflectivity
-0.2 -0.1 0 0.1 0.2

W e s ee fr om t h i s re s u l t t h a t t h e a u t oc o rr el at i o n o f t h e
s e is m o gr am i s q u i t e s i m i la r t o t h a t o f t h e w av e l et . T h u s
i t i s re as on a bl e t o a sk i f we c a n e s t im a t e t h e w av e le t
p o we r s p e ct r u m fr om the c e n t ra l l ag s of the
a u t oc or re l at i on o f th e se i s m og ra m . F u rt h e rm o r e, w e w i ll
d o t h i s wi t h ou t u s in g a n y d i r ec t kn o wl e d ge o f t h e
w av e le t . S o, w e wi l l t ak e th e sa m p l es fr om - . 1 t o . 1 o f
t h e s e is m o gr am a u t oc o rr el a ti o n a n d c om p ute t h e i r p ow er
s p e c t ru m . I f w e s im p ly t r u n c at e t h e a u t oc or re l at i on we
o bt ai n t h e r es u l t s h ow n be l ow :
Estimate with boxcar window
Exact result
Frequency (Hz)
0 20 40 60 80 100
Spectral Estimation
T h e p r e c ed i n g s p e c t ra l es t i m at e i s n o t b ad b u t c an b e
im p ro v ed by t ap e ri n g t h e s am p l es n e ar t h e e d g e o f t h e
ch o s en wi n d o w i n st e ad o f s i m p l y t ru n ca t in g. T h e m e t h od
o f t ap e r in g i s re f er r ed t o a s " w i n d ow i n g " a n d a n u m b e r
o f s p e c i al w in d ow s h av e b e en d e v is e d .
boxcar
hanning mwindow
bartlett
boxcar
mwindow
Hanning
Bartlett
Exact result
H er e a re t h e
r e su l ta n t e st im at e s
f ro m a p p ly in g t he
v a ri o u s w in d ow
p ri or t o e s t i ma t in g
t h e p ow er w it h t h e
D F T. A l l w in d ow s
d o a r ea s on ab l e
j ob t ho ug h t h e
e d ge s e em s t o b e
w i t h B a rt le t t a nd
H an n in g.
F requency (Hz)
0 20 40 60 80 100
Spectral Estimation
The DFT is a polynomial in z containing no denominator terms.
N–1
Gz = Σ g kz
k k
z = e
–i2πυk/N
k = 0
C on s e q u e n t l y, t h e D FT s p e ct r al es t i m at e c o n t ai n s o n l y
z e ro s ( n o p o l es ) in t h e z p l an e a n d is so m e t im es c a ll e d
a n a l l- z e ro s es t i m at e . A n a l t er n at i v e es t i m at e w as
d e v el o p ed by J .P . Burg ( s ee C l aer b ou t , 1976,
F u n da m e n t al s o f G e op hy s ic a l D a t a P r oc es s i n g) wh i c h
s ee ks t o p r od u c e a s p ec t r al m od e l u si n g a Z t r an s f or m
w it h o n ly d en o m i n at o r t e rm s . T h i s m at h e m at i c al
d e v el o p m e n t o f t h e B u r g sp ec t r u m , a l so c a ll e d t h e
m ax i m um en t r op y sp ec t r al e s t im a t e or a ll - p o l es
e st i m a te , i s be yo n d t h e sc o p e o f t h i s p r es e n t at i on .
N e v e rt h e l es s , t h i s i n t u i ti v e c on c e p t o f t h e B u rg
t e ch niq ue h e l p s u s u n d er s t an d i t s ba si c b eh a v i or . A s a n
a l l - p o le s e s ti m a t e, it i s v er y e ff e c t iv e a t m o d e li n g
s p e ct r a wh ic h h av e is o la t ed s p i ke s b u t le s s s o f or
s m oo t h s p ec t r a. F u r t h e rm o re , B u r g d ev e l op e d t h e
m e t h od u s i n g p r ed ic t i on o p er at or s t o p re d i c t t h e t i m e
s er i es o u t s id e o f t h e t r u n c at i on ra n ge s o t h at t h e
c on c e p t o f a wi n d o w d o es n o t a p p l y t o t h e B u rg
s p e ct r u m . T h e F o ur ie r s p ec t r u m p la c e s
z e r os c l o se t o t h e u ni t c i rc l e
a n d s o c a n m o de l a p hy si c a l
p r oc e s s w i t h a s m oo th
s p ec t r u m h a vi n g n ot c h es .
Frequency
T h e B u rg s p ec t ru m p la c e s p o le s c l o se t o t he
u ni t c i rc l e a nd s o c a n m od el s pi k es i n a n
u nd er l yi n g p hy s ic a l p ro ce s s
Spectral Estimation
A s m ig h t b e ex p e c t ed f ro m t h e p r ec e d i n g d is c u s s io n , t h e
B u r g s p e ct r u m d oe s n o t d o a g oo d j o b i n t h i s c as e :
E xact result
Burg (maximum entropy) spectrum (l=30)
0 20 40 60 80 100 120
H o we v er , th is d oe s n o t m ea n th a t t h e B u r g s p e ct r u m i s
w it h o u t m e ri t . H a t t on e t a l. ( p ag es 3 6 - 3 8 ) gi v e a n
e x ce l l en t a n a l ys is s h ow i n g t h e su per i or it y o f th e B u r g
t e ch n iq u e o v e r t h e D F T i n t h e c a se o f t h e re s ol u t io n o f
t wo c l os e ly s p a ce sp e c t ra l p e ak s. F u r th er m or e , a s w e
s h al l s e e, t h e B u r g t e c h n iq u e l ea d s t o a v e ry e f f ec t i v e
d e c on v o lu t i on m e t h od .

Wavelength Components
C o n sid er a se rie s o f p la n a r w a v e fro nt s p r op a g a ti ng a s

s ho wn be lo w.
λx
θ X
λz λ
θ
Wave propagation direction
Z
T he di st an c e between w a ve f r o nt s , m e a s ur e d
pe r p e nd i c ul a r t o t h e m, i s de fi ne d as t he w a ve l e n g t h, λ .
We c an a l s o s pe a k o f t h e w av e l e ng t h "c om po ne n t s " i n
t h e va r i ou s c o o r di n at e di r e c t ion s . F or ex am pl e , t he
ho r i z on t a l w a ve l e ng th, λ , i s t he d i s t a nc e b e t w e e n
x
w av e f r on t s me a s u re d i n t h e x c oo r d i na t e di r e c t ion . Th us :
λ λ
λx = and λz =
sin θ cos θ

W e s e e t h at t h e c o mp o ne n ts o f w a ve l e ng t h a r e n e v e r
l e ss t h an t h e w a ve le n g th i t s e lf . I n f a ct , f o r a v e r ti ca ll y
t r a v e li n g w a v e , λ x i s i nf i ni t e . T h e c o mp o ne n t s a d d a s
i n ve r se s qu a r e s :
1 1 1
= +
λ2 λ2x λ2z
I t is o f t e n c o nv en i e nt t o d e a l wi t h v e c t o r q ua nt i t i e s s o
w e d e f i ne t he wa v e nu mb e r , k , a n d i t s c om po ne nt s a s t h e
i nv er s e o f t h e w a v e l en g t h a n d i t s c om po ne nt s .
k = λ k x = λx k z = λz
–1 –1 –1
2 2 2
k = kx + kz
k is th e m a g ni tu de o f a ve c to r , k, w h ic h p o in ts in th e
d ir e c tio n o f w a v e p r o pa g a t io n an d w h os e co m p o ne n t s a r e
th e i nv e r se w a ve l e ng th s.
I n 3- D , w e ha v e p la na r w a v e fr o n ts i ns te a d o f lin e a r b u t a
sim p le e xt e ns io n o f th is r e su lt st ill h o ld s:
2 2 2 2 The dispersion relation

kx + ky + kz = k for scalar waves.
Where:
–1
k y = λy

This geometric relation between components of the wavenumber

vector is fundamental to the study of wave propagation. It can be
considered as the Fourier domain equivalent of the scalar wave
equation. A fundamental result from theory is that the extrapola-
tion of surface recorded data into the subsurface (z direction)
requires knowledge of kz. On the surface, we can measure kx, ky,
and f, and since fλ=v, this allows kz to be calculated form the dis-
persion relation:
Since kz must be a real number (in order to be interpreted as an

inverse wavelength) we see from this equation another fundamen-
tal result. Not all values of (kx,ky,f) can be considered as wavelike.
In fact, we must have
in order for a triplet of (kx,ky,f) to be a propagating wave.

Apparent Velocity
T h e w a ve le n gt h c o m p o n en t s a n d t h e co r re sp o n d i ng
w a ve n u m ber s a re cl o se ly r ela te d t o t he w a v e v el o ci t y
a n d it s co m p o n en t s w h ic h a r e ca l le d a p p a r en t v el o c it ie s.
R e ca ll in g th e ba s ic r el a ti o n , λ f = v , w e se e t h a t t he
a d d i ti o n f o rm u l a f o r w a v el en gt h c o m p o n en t s:
1 1 1
= +
λ2 λ2x λ2z
leads directly to:
1 1 1
2
= 2
+ 2
v vx vz
where vx = fλx and vz = fλz
If we use wavenumber components, we have:
f f f
v = vx = vz =
k kx kz
N oth i ng p h y si ca l a c tu a l ly p r o p a ga t es a t a n y o f t he
a p p a r en t v e lo c it i es . R a t h er, th e y a re s im p l y rel a t ed t o
t h e a rbi t ra r y c h o ic e o f c o o rd i n a t e d i re ct i o n s a n d c a n be
v is u a l iz e d a s t h e w a ve le n gt h a lo n g a co o rd i na te d ir ec ti o n
d i vi d ed by t h e ti m e b et w e en w ave cr es ts ( i .e. t h e p e ri o d
o f t h e w a v es .)

Apparent Velocity
x Receivers on surface
z
θ λx
A se ri es o f p la n e
w a v ec res t s a p p r o a c h a λz
h o r iz ont a l a n d a ve rt ic a l
r ec o rd i n g a r ra y . E a ch a rra y
s ee s th e a p p a re nt
w a v ele n gt h a l o n g t h e
s u rf a ce o n w h ic h it i s
d e p lo yed a n d c a n m e a su re
t h e a p p a ren t v el o c it y o f
t h e w a v ef ro n t s a lo n g t h a t
s u rf a ce . T h e a n gle θ i s
c a l le d th e “ em er gen c e
a n g le ”. T h e w a v el en gt h
c o m p o n e nt s a re :
λ λ
λx = and λz =
sin θ cos θ
And the apparent velocities are:
similarly

Th e 2 - D F - K Tra nsf o rm
The f-k transform is a fundamental tool which essentially allows the

direct computation of wavenumber components and frequency for
a multidimensional wavefield. In 2-D, it can be written:
and in 3-D
The inverse transforms are mathematically similar:
2-D
3-D
Th e s e i n v e rs e t r an s f or m s have the p h y si c al
i n t er p r et a t io n of p re s en t i n g a wa v e f i el d as a
s u p e rp o s it i on o f in div i d u a l F ou r ie r co m p o n en t s o r " p l an e
w av e s" .

Th e 2 - D F - K T ransform
A f ew f - k t r an s f or m s a re k n ow n a n al yt i c al l y. P er h ap s
t h e m o s t i n p o rt a n t i s t h e t r an s f or m o f a si n g l e l in ea r
e v en t . U s i n g th e m a t h e m at i c s o f D i ra c d el t a fu nct i o n s,
a se i sm ic wa v e f i el d c o n s is t i n g o f a n i so l at e d l in ea r
e v en t c an be w ri t t e n :
kx
• Horizontal events in (x,t) are vertical in (kx,f) and vice-versa.

• All events in (x,t) with then same apparent velocity, vx, are col-
lected into a single linear event in (kx,f). The different events are
distinguished by their phase spectra but have differing phase spec-
tra.
T h e 2 -D F - K T ransfor m
If w e conside r al l possibl e linear events characterized b y
vx=v/sin(θ) , then we have:
x kx
θ =0o
θ =15o
vt θ =30o f/v
θ =50o
θ =90o θ =90o
θ =50o
θ =30o
θ =0o θ =15o
The previously encountered fact that f/v > kx is reexpressed in this

analysis though the fact that a portion of (kx,f) space is not popu-
lated. As a general rule of thumb, we see that large kx values are
only wavelike at high frequencies. This fact will turn out to be fun-
damental in describing the ability of seismic images to resolve
small features. Small features require large kx values which, in turn,
require a large temporal frequency bandwidth.
In proceding from analytic to discrete f-k transforms, it turns out

that the implementations of the Fourier transform integrals are
approximate but the forward and backward DFKT are exact invers-
es of each other. This fact is a great convenience in data process-
ing and is not generally true of other transforms such as the Radon
transform.

Th e 2- D F - K Tr ansf o rm
Wh e n w e pr o c e d e fr om the c o n ti n uo us F - K t r an s f o r m t o th e
di s cr e te , a s i t ua ti o n di r e ctl y a na l o g o us t o t he 1 -D c a s e
occ ur s. T ha t i s , the a c t of s p at i a l s a m pl ing ind uc es a
periodicity i n th e (ω , k ) do ma i n . Un l i k e te m po r a l aliasing,
s pa tial aliasing i s a l w a ys pr e s e n t.
-knyq knyq
ω = ωmax
ω ω ω
Principle
Band
He r e w e s ee o n e ev e n t s h o wi n g s p a t i al al i as i n g an d
an o t h e r wh ic h d o es n ot . G i v e n a s p at i a l s am pl e ra t e of Δx
an d an ap p ar e n t v e l oc i t y v a t h e n a l l t e m p o ra l f re q uen c i e s
h i g h er t h an :
va
f = 2 π ωcrit = va k nyquist =
2 Δx
crit
wi l l b e s p a t i al l y al i as e d . Fo r e x c el l e n t i l l u s t r at i on s o f
sp at i al a l ia s i n g se e Ha t t on et al . p p 4 3 - 4 5 an d Y i lm az
p p 6 2 -6 9
FK Transform Pairs
Space-time domain FK amplitude spectra
A si n gl e fl at
ev ent.
Wa vel et i s
30 Hz
( domi nan t)
and m in i mum
phas e
S ix even ts with
e me rgen ce
a ngle s: 0, 10,
30, 50, 70, &
90 degr ees .
V elo city i s 20 00
m /se c.
S ix e v en ts w it h
e m e r g en c e
a n gl es : 0 , - 10 ,
- 30 , - 50 , - 7 0 , &
- 90 d e g re e s .
V el oc i ty i s
2 0 0 0 m / se c .
A single
diffraction
hyperbola.
Veolocity is
2000 m/sec
Many
diffraction
hyperbolae.
Veolocity is
2000 m/sec

τ-p Transforms
T h e τ - p t r an s f or m, a l s o kn ow n a s t he R ad on t r a ns f or m
o r s l an t s t a c k , i s a v e r y u s ef u l d at a p r oc es s i n g t o o l d ue
t o i t s a bi l i ty t o d e c om po s e a s e is m i c m a t r i x i nt o e v e n t s
o f c o ns t a nt h o r i zo nt a l s l o w ne s s , p . I t ' s c l o se r e l at i o n t o
t he f -k t ra ns f o r m i s c a pt u r e d i n t he " pr o je c t i o n s l i c e
t he o re m " w hi c h s ho w s t ha t t h e τ - p t r a ns f o r m m a y b e
c om pu t e d f ro m a n f - k t r a ns f o rm w hi ch h as b e en
i nt e r po l at ed t o " p ol a r " c o o rd i na t e s . ( S ee D e an s , S . R . , 198 3,
Th e R ad on Tr an s fo r m a nd S om e o f I ts A p pl ica ti on s , J o hn W ile y an d S o n s ) .
C o ns i de r t he e x pr e s s i on f or a f o rw a r d f -k t r an s f o r m:
φ(k x,f ) = Ψ(x,t)e

2π i (k xx – f t)
dx dt (1)
–∞
W e h a v e se en h ow t h is e xp r es si o n t ra n s fo r m s li ne a r
ev en t s in ( x ,t ) i n to li n ea r ev en t s in ( kx ,f ) :
sin θ sin θ ω
t = x kx =
v v
θ =0o
x kx
θ =15o
vt θ =30o ω /v
θ =50o
θ =90o θ =90o
θ =50o
o
o θ =15o θ =30
θ =0
(x,t) space (ω,kx) space

τ-p Transforms
N ote t h a t s in(θ) /v ( ho ri z o n t a l s lo wn es s ) c a n a l s o b e
wr it t e n a s d t /d x o r t h e ra y p a ra m et er p . T h u s , ra d i a l l in e s
i n t h e f- k t ra n s f o rm a re l i n es o f c o n s t a n t p . T h i s c a n b e
ex am in e d f u rt h er b y a s u bs t it u t io n o f va ri a bl es in th e f - k
i n te gr a l ( 1 ) :
∞
kx
φ(p,f ) = Ψ(x,t)e
2π i f (p x – t)
dx dt where p= (2)
–∞ f
H e re p h a s b e en ex pli c i t ly i n t r od u c e d a s t h e ra ti o o f k x
a n d f a n d h e n c e i s c on st a n t a l on g ra d i al l i n es in ( kx , f)
s p ac e . S o φ ( p , f) c an b e re g ard ed a s a " p ol ar c oo rd in a t e"
r ep r e se n t at i on o f φ ( kx , f ) . N o w, co n s id er t h e m e an i n g o f
e q u at i on 2 f o r c on s t an t p b y p e rf o rm i n g t h e t
i n t eg r at i on fi r st :
∞
φ(p,f ) = ψ(x,f)e
2π i f p x
dx (3)
–∞
where
∞
ψ(x,f) = Ψ(x,t) e
–2π i f t
dt (4)
–∞
Then, compute the inverse Fourier transform (f->t) of (3)

∞
2π i f τ
ϕ(p,τ) = φ(p, f ) e df (5)
–∞
Now, substitute (3) into (5):

∞
∞
2π i f τ
ϕ(p,τ) = ψ(x,f )e
2π i f p x
dx e df
–∞
–∞

τ-p Transforms
Interchange the order of integration:

∞
∞
2π i f (p x + τ)
ϕ(p,τ) = ψ(x, f )e df dx
–∞
–∞
The inner integral gives Ψ(x, px+τ ), so:
ϕ(p,τ) = Ψ(x, px + τ) dx (6)

–∞
Equat io n (6 ) i s th e c o nv e nt io na l e qu a tio n fo r t he τ- p
transform (c o mpare w i th Y il ma z (Seis m ic Da t a Processing ,
1 987 ) equat io n 7 .5) . S e veral th in g s ca n b e learned from th is
d e velo pment:
• T h e τ- p t ra ns f o rm c a n b e c o m p u t e d f ro m t h e f - k
( F o u ri e r) t ra ns f o rm by a c o o r d in a t e ch an ge fr o m ( f ,k x )
to ( f, p ) f o l lo w ed b y a n i n v er se F o u r ie r t r a n sf o r m f ro m
f- >τ . T h i s a m o u n t s to c h a n gi n g t o p o la r c o ord i n a te s in
th e F ou ri er d o m ain .
• T h e τ - p t r an s f or m m ay e q u i v al e n t ly b e c o m p u t e d b y
e q u at i on ( 6 ) wh i c h is a p ro c es s k n ow n a s " sl an t
s t ac ki n g " x
F o r f ix e d ( p ,τ ) , eq u a t io n 6
r ep r ese n ts a su m m a t io n
t h ro ug h the f un c t io n t
Ψ ( x ,t ) along a l in e a r
t ra j ec to r y . H en c e it i s
c a l le d sl a n t st a c k i ng .
Ψ(x,t)
τ-p Transforms
• S i n c e t h e a n al yt i c τ- p t r an s f or m i s c o m p u t ab l e f r om
the 2 - D F ou r ie r t r an s f or m ( a n d v ic e - v e rs a) t h e
i n f om a ti o n c o n t en t i s t h e s am e in e it h e r d o m ai n . T h e
f ac t t h a t t h e 2 - D F o u r i er t r an s f or m i s c o m p l et e m ea n s
t h at t h e a n a ly t i c τ - p t ra n s fo rm is a ls o. W e w i ll s e e t h a t
t h i s i s n o t t ru e fo r t h e d i g it a l τ- p t ra n sf o rm .
-kxnyquist kx kxnyquist
H e re w e s e e an illustration o f th e r ep r e sentation of ( f , kx )
space i n b ot h re ct an gu l a r an d p o lar co ordinates . T h e ra dial
l in e s a r e li n e s o f c onstant p an d a r e a ll s h ow n t o t er m ina t e
( w h er e p o ss i bl e ) a t t h e s a m e c onstant f . T o c ompute t h e
d i s cr et e τ- p t ransfor m transfo rm , spectra l v alu e s a r e
i n t er p olated f r om t h e r ectangular ( f ,k x ) g r id t o re g u larl y
sampled f l ocation s on e a c h ra dia l li n e :

τ-p Transforms
0 p
A f t e r i nt er p ola t i on o n t o ra di a l li nes , ( f ,k x ) s p ac e b ec om es
( f, p ) s p ac e . A n inverse Fourier t ransform f r om f t o τ
complet e s t h e journey t o (τ, p) s pa c e .
C l o s e i ns pe c ti on o f t he f i g ur e o n t h e p r e v i ou s p a ge
s h ow s w hy t he d i s c r et e τ - p t ra ns f o r m h a s d i f f i c u l t y e v e n
t h ou gh t h e a na l yt i c τ - p t r a ns f o rm i s c o mp l et e. I t i s
i m po s s i bl e t o p i c k a s e t o f d i s c re t e p v a l ue s w hi c h c ov er
t h e ( f , kx ) g r i d u n i f o r ml y . E i t he r t h e y a r e t o o f a r a p ar t a t
t h e g r i d e d ge s o r t h e y a r e t o o c r o w de d n e ar t he c e nt er .
I n e i t he r c a s e, it c an b e s h ow n t ha t t he r e i s a l w a ys
" i nf o rm at i on l os s " i n g o i ng t o t h e d i s c r e t e ( τ , p ) s p ac e
a nd b a c k a g ai n . P u t a no t he r w ay , m e r e l y t ra ns f o r mi ng
d at a t o ( τ , p) s p ac e a n d b ac k ( w i t ho ut a ny τ - p
p ro c e s s i ng ) w i l l a l w a ys a l t e r t h e d at a i n s om e w a y. T hu s
t h e d i s c r e t e τ - p t ra ns f o r m i s n o t c o mp l e t e i n t h e s a me
s e ns e t ha t t he d i s c r e t e ( f , kx ) t ra ns f o r m i s .

Properties and uses of the τ-p Transform
T h e m o st o b v io u s p r o p e r t y o f a τ-p t r a ns fo r m i s t h at i t
m a ps a lin e a r e v e n t in ( x, t) to a p o in t in (τ- p ).
x p
το po
το
τ
Δx
t
Δt
Δt –1
= vapp = p o
Δx
Less obviously, hyperbolae map to ellipses:
x p
το το
τ
t

T hu s w e c a n e xp e c t t ha t b an dl i m i t e d p r op a g at i n g b o d y
w a ve s i n a c on s t a nt v e l oc i t y e ar t h w i l l ma p t o a c om pa c t
r e g i o n of (f , p) s p ac e de f ine d b y p max = 1 / v. In t h e f - k
t r a ns for m , t hi s c o r r e s po n ds t o a t r i a ng u l ar r e g i o n.
p kx
f=fmax
f f
T hu s w e e x pe c t t ha t a p pa r e nt v e l o c i t y f i l t e r i ng c a n be
d on e in e i t h er d o ma i n by e s s e nt i a l l y m ut i ng
( s ur pr e s s i ng ) t h a t p o r t i on o f t he d o ma i n c o r r es p on di ng
t o t he u n d e s i ra b l e v el o c i t i e s .
A l ia si n g a f fe c t s t h e (τ,p ) t r a n sf o r m m u ch a s it d o e s t he
( f, k ) t r a n sfo rm . I f t h e (τ,p ) t r a n sf o r m i s co n st r u ct e d by
s la n t st a ck i ng i n ( x, t) o r ( x ,f ) th e n it i s no t di r e c tl y
a f fe c ted b y ho r iz o n ta l a l ia si ng. Bu t t he c ho i ce o f Δp a n d
t he n um be r o f p v a lu e s is a d iff ic ul t o ne a nd l e a d s di r e c tl y
t o p a li a sin g .
A rule of thumb for Δp is
1 Δk x
Δp = ≈
fmax (xmax – xmin) fmax
T uner, G., 1 99 0, Al i asin g i n t he tau- p t ransform and the remov al of spati al al iased
c oherent noi se: Geophy sic s, 5 5 , 1 49 6- 1 50 3

Other uses of the τ-p Transform:
• S in c e a s la n t s ta ck is l es s a ff ec t ed by s p a t ia l a l ia si ng
t h a n a n f - k t ra n s fo r m , i t ca n b e u s ed t o i n te rp o l a te t o
f i ne r t ra c e sp a c in g s a n d " u n a li a s " d a ta . U se d in t h is
f a s h io n it i s o f te n c a ll ed a " s m a rt i n te rp o l a t o r" . (Y ilmaz ,
O. , 1987, S eism ic Dat a P roces sing , p435. )
• I t c an b e s ho w n th a t m u lt ip le s a r e n o t p e r i o di c o n a n
of fse t tr a c e i n th e (x ,t ) d o ma i n b u t a r e in t he (τ, p) .
(T rei tel et al. , 1982, Pl ane- wav e dec ompo si ti on of sei sm ogram s, Geophysi c s,
47, 1375- 1401) T h is m e a ns th a t p r e d ict iv e de c o nv o lu ti o n
fo r mu lt ip le su r p r e ss io n o f te n w o r k s b e t te r o n (τ ,p )
g a th e r s.
• M i g r at i on c an a ls o b e d on e in t h e (τ, p ) d o m ai n . (Die bol d,
J.B. , a nd Stoffa , P.L., 198 1, The tr av el time e qua ti ons, ta u-p ma ppi ng, a nd
inve r sion of common midpoi nt da ta , Geophysi cs, 46 , 23 8-2 54 )

Inverse τ-p Transforms
T h e p r o c es s o f re co n st r uc t i o n o f t h e s ei sm i c d a t a i n ( x , t )
s p a c e gi v en it s τ- p t ra n s fo rm is ca l l ed a n in v er se τ- p
t ra n s f o rm . T h er e a r e a n u m be r o f wa ys t o d o th i s p r o c es s
t h o u gh we sh a l l d i s cu s s o n l y t wo : F o u ri er m et h o d s a nd
f il t er ed ba c k p ro j e c ti o n .
T he F o u ri er m et h o d i s o bv io u s f ro m t h e d is c u ss io n o f
t h e f o rw a rd τ - p t r a n s fo rm . T he m a jo r s t ep i s t h e
r e c o n st ru ct io n o f t h e 2 - D ( f, kx ) t ra n sf o rm w h i ch
r e q u ir e s a n i nt er po l a t io n o nt o a r ec t a n gu la r g ri d f r o m a
p o l a r o n e . T h is w i ll o bv io u s ly h a v e n u m er i ca l d if fi cu lt ie s
t h o u gh t he y a re c o n t ro ll a bl e. F o ll o w in g the
i n te rp o la t io n , an i n v er s e 2- D F o ur ie r t ra n sf o rm
c o m p le te s t h e p ro c es s.
F il t e r e d b a ck p r oj e ct io n a vo id s t h e ( f, k x ) d o ma in a n d
r e c o ns t r uc ts t h e i m a g e d ir e ct ly w it h a c o nv o lu ti o na l
f i lt e r f o ll o w e d b y a n i n ve r s e s la n t s ta ck . C o n s id e r t he
e x pr es s io n f o r t h e i n v e r s e 2 -D F o u r ie r t r a n s fo r m :
∞
–2π i (k xx – f t)
Ψ(x,t) = φ(k x,f )e dk x df (1)
–∞
N ow , letting k x = f p and converting t h e wavenumbe r integral

i n to a p integral g ives :
∞
Ψ(x,t) = f φ(p,f )e
–2π i f p x 2π i f t
e dp df (2)
–∞

T h e t er m i n b ra c ke ts c a n be c o n si d e re d t o b e th e p r o d u c t
o f t w o f u n ct i o n s o f f . H e n ce , it m u st b e a c o n v o l u t io n in
t im e :
∞
Ψ(x,t) = α(t)•β(p,x,t) dp (3)

–∞
where • denotes a convolution over time and

∞
α(t) =
2π i f t
fe df (4)
–∞
∞
2 π i f (t –p x)
β(p,x,t) = φ(p,f )e df = ϕ(p,t –px) (5)
–∞
N o te t ha t ϕ (p ,τ) i s th e fo r w a r d s la n t st ac k . S ub sti tu tio n

o f ( 5) in t o (3 ) r e su lts in :
Ψ(x,t) = α(t)• ϕ(p,t – px) dp (6)

–∞
E q u at io n ( 6) e x pr e sse s fi lte r e d b a ck pr o j e c tio n f r o m th e

(τ, p) d o ma in t o th e (x ,t ) d o m a in . E a ch p o int in (x ,t ) i s
co n st r u cte d b y i nt e g r a ti ng a lo n g a lin e a r t r a j e ct o r y in
(τ, p) , j u st lik e t h e fo r w a r d s la nt st a ck . U nl ik e t he fo r w a r d
o p e r a ti o n, th e in te g r a ti on i s fo l lo w e d b y a c o nv o lu tio n
w h ich is a fo r m o f a f ilt e r .

A n ot her w ay t o d o t h e i n v e r s e t r an s fo r m i s s u g ge s t e d b y
e q u a t i on 2 . Ra t h e r t h an c on v o l v e i n t h e t i me d om ai n w i t h
a fi l t e r o p e r at or w e c a n d o t h e r e c on s t r u c t i on i n t he f
d o m a i n . Ta ki n g a fo r wa rd Fo u ri e r t r an s f or m ( t -> f ) o f ( 2 )
g i v e s:
∞
–2π i f p x
F Ψ(x,t) = ψ(x,f) = f φ(p,f )e dp (7)
–∞
T o us e (7 ) f o r t he i n ve r s i o n, w e f i r s t t r an s f o rm ϕ (p,τ) t o
φ( p, f ) . T he n, f o r e a c h x, w e m ul t i p l y φ(p , f ) b y a p
d e pe n de nt p ha s e s hi f t an d i nt e gr a t e o ve r p a nd t he r e s ul t
i s s c a l e d b y f . A f t e r c o ns t r u c t i n g ψ (x, f ), an i n ve r s e
F o ur i e r t r an s f o r m f r o m f - > t c o mp l e t e s t h e p r oc e s s .

Least Squares τ-p and f-k Transforms
W e h av e s e e n t hat a c o n v e n i e n t m et hod of i m p l e m e n t i n g
f or w ar d a n d i n v e r s e τ-p t ra n s f or m s i s i n t he f r e q u e n c y
d o m ai n : ∞
φ(p,f ) = ψ(x,f)e
2π i f p x
dx (1)
–∞
ψ(x,f) = f φ(p,f )e
–2π i f p x
dp (2)
–∞
H e r e (1 ) is th e fo r w a r d t r a ns fo r m f r o m ( x,f ) to (p ,f ) a nd
( 2) is t he in ve rse t r a ns fo r m . V ir tu a ll y a ny in te g rat io n c an
b e i mp le me n t e d a s a n e q ui va le nt ma t r ix op e r a t io n fo r
d isc r e te da t a . C or r e sp o n di ng t o (1 ) a nd ( 2) w e h a ve :
φj (f) = Σ Rj kψ k(f) Rj k = exp(2π i f p j x k) (3)

k
ψ k (f) = fΣ R k j φj (f)
* *
Rk j = exp(– 2π i f p j x k) (4)
j
These can be written:

φ= Rψ (5)
ψ = fR φ
*T
(6)
φ1 (f) ψ 1(f) R1 1 R1 2 R1 3
φ2 (f) ψ 2(f) R2 1 R2 2 R23
φ= ψ= R= etc
φ3 (f) ψ 3(f) R3 1 R3 2 R3 3

R at h e r t ha n c o mp ut e t h e f or w a rd t r an s f or m d i r e c t l y , t he
l e a s t s qu ar e s t e c hn i qu e u s e s e q ua t i on ( 6 ) t o p os e a n
i n v e rs e p r o b l e m f or t he τ - p s pe c t r um .
ψ = fR φ
*T
(6)
R f ψ = RR φ
–1 *T
–1
φ = RR Rf ψ
*T –1
(7)
E qu a ti o n (7 ) is t he st a nd a r d le a st sq u a r e s e sti m at e o f
th e τ- p s pe ctr um . I t is u su a lly su p e r io r in th e se n se t ha t
th e (x ,t ) d o ma in d at a ca n b e r eco n str u c te d fr o m it w it h
fe w e r a r t ifa c ts . T hi s fo r m ul a tio n a ssu m e s th a t t he
nu m b e r o f p t r a ce s e x ti ma t e d w il l b e n o la r ge r t ha n t he
nu m b e r o f x t r a ce s. E ve n w he n th e d a ta is p e r fe c tl y
r e g u la r in x a n d th e nu m b e r o f p a nd x tr a c e s ar e t he
sa me , th e le as t sq u a r e s m e t ho d is u su a lly s up e r io r
b e ca u se th e τ-p t r a n sfo r m i s i nc o mp le t e . T hi s m e a n s t ha t
th e fo r w a r d a n d r e v e r se τ- p pr o c e ss e s le a ve a r t ifa c ts in
th e d a ta . T he le a st sq u ar e s a p p r o a ch m in im iz e s su ch
ar tif a cts .

T h e m o re i n co m p le te a n d i n co n si st e n t a tr a n s fo rm p a i r
a r e, t h e m o r e t h e le a s t s q u a re a pp ro a c h be co m es u s ef u l .
T h i s m ea n s i t is e sp e ci a l ly p re f err ed fo r s la nt s t a c ks a l o n g
p a r a bo li c a n d h y p er bo li c t ra j ec t o ri es w h ic h a re i n c o m p l et e
ev e n i n t h e a n a l y t ic s en s e .
A n o t h e r e x a m pl e o f a n i n c om p l e t e t ra n s f or m i s t h e
d i s c r e t e F ou r i e r t r an s f or m f or i rr e g u l ar l y s am p l e d d a t a .
I t c a n a l s o b e p o s e d a s a n i n v e r s e p r o bl e m :
ψ=F ϕ
*T
–1
ϕ = FF Fψ
*T
where
F1 1 F1 2 F1 3
F F F
F = 2 1 2 2 23 Fm n = exp(2π i k x mxn )
F3 1 F3 2 F3 3
S e e M a r f ur t, e t a l. , ( 199 6, P i tf al ls of u si ng co n v e nt io n a l R a do n
t r a nsf o r ms on p o o rl y s a mpl ed d at a: G e oph y sic s, 6 1, 1 46 7- 148 2) f or a
m o r e c o mp le t e d i s cu s s io n .

Lecture Notes
Geophysics 557
Chapter 3
Amplitude Effects

Seismic Wave Attenuation
As a seismic wave propagates through the earth, it suffers
attenuation (amplitude decay) for a number of reasons:
Attenuation Mechanism #1: Geometric Spreading (or Spherical
Divergence)
A s s ei s m ic e n er gy p ro p a ga t e s
a w a y f ro m a s o u r ce ( o r f o c a l
p o i n t ) th e co n se rv a t i o n o f to t a l
en e rg y re q u ir es t ha t t h e e n er gy Surface A2
f o u n d o n t h e wa v e fr o n t s u rf a c e
A 1 a t s o m e t im e t 1 e q u a l th at
o n su r fa c e A 2 a t so m e ti m e t 2 .
E t 2 = ε t 2 A 2 = E t 1 = ε t1 A 1 Surface A1
w h e re ε i s t h e e n e rg y p e r u n i t
ar e a. Si n c e t h e d i s p l ac e m e n t
w av e am p l i t u de, u, is
p r op or t i on a l t o t h e s q u a r e
r oo t o f ε , w e d e d uce :
u2 A1 R1 t1 u0
= = = or
ut =
u1 A2 R 2 t 2 Rt
T h e p r o p er i n t erp r et a ti o n o f th i s r esu l t is t h a t th e w a v e
a m p l it u d e d ec a y s a s 1 / R w h er e R i s t h e r a d iu s o f
c u rv a tu r e o f t h e w a v e fr o n t. I n t h e c a se o f a co ns t a n t
v el o ci t y m e d iu m , R i s si m p ly th e d is t a n ce t ra v el le d ;
h o w e v er, i n a l a y er ed m e d iu m , R c a n b e s h o w n t o b e
p ro p ort io n a l to ( V2 r ms / V 1 ) t w h e re V 1 i s t h e v elo ci t y o f
t h e f irs t la yer. ( N ewm an, Geo phys ics, 1971, p 481-488, Hu br al, P ., and
Krey, T., In te rv al V elocit ies f rom Seism ic Refle ct ion Time Meas ure men ts , 1980,
So ciety o f Ex plor atio n G eoph ysicist s)
3-2 Amplitude Effects

Th u s we deduce t h a t t h e e f fe c t s of spheri c a l diverg en c e c a n
b e a p prox i m ate l y c ompensated f or by applyi n g a " gain
c or re c ti o n " g iv e n b y:
2
G t spreading = G0Vrms t t (Compare with Hatton et al., page 56)
Attenuation Mechanism #2: Absorption (or inelastic attenuation)
I n a p e rf e ct l y el as t i c m ed iu m, t h e t ot a l en e rg y o f t h e
p r op a ga t in g w av e fi e l d re m a in s a c on s t an t . H o we v er , t h e
e ar th is n o t p er f ec t l y e l as t ic a n d p r op ag at i n g s ei s m i c
w av e s g ra d u al l y d i e o u t o v e r t i m e . T h e p ri m ar y
m e c h an i s m f or t h i s i s th e c o n t in uo u s c on v e r si on o f a
s m al l p o r ti o n o f t h e s e is m i c en e r gy t o h e at d u e t o
i rr e v er s ib l e a n el as t i c b e h av i or o f ro c ks . I t is c u s t om ar y
t o t al k a b ou t t h e p a ra m et e r Q wh i c h c h ar ac t er i ze s t h i s
e n er g y l o ss :
energy
Q = per frequency cycle
energy loss
V ar i o u s at t e nu at i on t h eo r i e s e x i s t w i t h t h e s i mp l e s t b e i n g
t h e "c o n s t an t Q " t h e or y o f K j ar t a ns son 1 a nd o t he r s . M os t
e mp i r i c a l e v i de n c e i s c o ns i s te nt w i t h t h e a s s um pt i on t ha t
Q i s i n de p e nd e nt of f r e qu e nc y at l e a s t o ve r t he s e i s m i c
b a ndw i d t h . T he c on s t a nt Q t h e o ri es a l l p re d i c t an
a mp l i t ud e l os s g i v e n b y :
1 Kjar tan ss o n, E, 1 979, C o ns tan t Q-Wave P ro pa ga tion an d A tten u atio n, JGR , V 84,
p4 737-474 8
Thus, the constant Q theory refers to a Q which is independent

of frequency but predicts an attenuation which is a first order
exponential in both time and frequency.
f t
Constant Q Exponential Decay Surface

N o t e t h a t Q = ∞ i s a p e rf e c t ly e la st i c m a te r ia l wh i l e Q =
0 i s p e rf e c t ly a b s or p t iv e . A h i gh l y a b so rp t iv e r oc k h as a
Q o f 2 0 - 5 0 w h i l e v e ry co m p e t en t l i m e st o n es a n d
d o lo m i t es c an h av e Q o f 2 0 0 o r m o re . T h e c om m on fi r s t
o r d e r c or re c t io n f o r Q e ff e c t s i s t o a p p l y a s i m p l e ,
f re q u e n c y i n d e p en d en t , e xp o n e n t i al g ai n c or re c t i on . I f
w e wr i t e:
– πft/Q –α t
e ≈ e ; α=πfdom/Q = the attenuation constant
Typica l ly , th e a t te n uation constant i s ex p res se d i n db/se c

which w oul d be
Assuming f do m of 2 0 a n d a Q o f 1 0 0 l e ad s to a "t y pi c al "

v a l u e o f 1 2 d b /s e c .

Attenuation Mechanism #3: Transmission losses

I n o u r e x a m in a ti o n o f t he t h eo r y o f t h e 1 - D s y n th et ic
s ei sm o gr a m w e s a w t h a t t h er e i s a c o n t in u o us
a m p l it ud e d ec a y d u e t o t ra n sm i ss io n l o ss es . I f f a c t, w e
f o u nd t h a t t h e e a rt h ' s i m p ul se r e sp o n se r e su lt ed i n t h e
r ec o r d in g o f t h e n 't h r e f le c t io n c o e ff ic ie n t a t t h e
s u rf a c e m u lt i pl ie d b y a t ra n sm i ss io n l o ss t e rm :
nth reflection coefficient
recorded at surface = Rn (Transmission losses)
n–1
2
where transmission losses
= k = 1
1–Rk
T hi s e f f e c t i s h i g h l y d ep e nd e n t u p o n l o c a l g eo l o gy a nd
i s d i f f i c ul t t o e s t i m at e w i t h a n y p r e c i s i on . I t i s
c u s t om ar y t o i g no re it a n d " ho pe " t ha t i t i s e i t h er s m al l
o r i nc l ud e d i n t he " db / s ec " c o r re c tio ns a l r e ad y
d i s c us s e d.
Attenuation Mechanism #4: Mode Conversion
A s wa v es p ro p ag at e i n a n e la st i c m ed i u m, t h ey a re
c on s t an t l y be i n g c o n v er t e d f ro m P t o S a n d th e r ev e r se
a t ev e r y i m ped a n c e c on t r as t . T h e s e m od e c on v e rs i on s
o c c u r bo th u po n re f le c t i on a n d t r an s m i s si on a n d a re
d e s cr i be d b y t h e f am o u s Zo ep pri t z e q u at i on s ( s e e A ki
a n d R i c h ar d s , 1 98 0 , o r t h e C RE W E S Zo ep pri t z e x p l or er
a t w ww . cr e we s .o rg ) . I f , a s i n c on v e n t i on al s ei s m i c , o n l y
t h e v e r ti c al c om p o n e n t o f g ro u n d m ot i on i s r ec o rd e d ,
t h e n i t i s r ar el y p o ss i b le t o a d d re s s t h i s e ff e c t . T h e
s ol u t i on i s t o r e co rd a l l t h r ee c om p o n en ts o f g ro u n d
m o t io n a n d p r oc e ss t h e d at a a s e la st i c wa v es . T h i s i s
t h e s u b j ec t o f l e ad i n g e d g e re s e arc h a r ou n d t h e w or ld .

Attenuation Mechanism #5: Scattering
R a n do m s c a tt e r in g o ff s m a ll i r r e g ul ar i ti e s c au s e s t he
d is p e r s a l o f s e is m ic w a v e fi e l ds a n d a n a pp a r e nt l o s s o f
e n e r g y . I f a f u ll 3 -D w a ve fi e ld h a s b e e n r e co r d e d t h e n
s u c h s ca t te r e r s c an b e i m a g e d b y m ig r a ti o n b ut t h e l o s t
w a v e fi e ld e n e r g y i s n ot r e s t or ed .
Attenuation Mechanism #6: Refractions and critical angles
S n el l' s l a w go v er n 's t h e a n gl es o f ref le ct i o n a n d

re fr a ct i o n w h en a w a v e in t era c t s w it h a n i m p ed a n c e
c o n tr a st .
θ θ
sin θ sin φ
=
v1 v2
φ
I n t h e n o r m al c as e wh er e v 2 > v 1 , t h e re ex is t s a "c r i t i c al
a n g le " o f i n c i d e n c e b e yo n d wh ic h n o t r a n s m i s si o n o c c u r s .
v1
sin θ crit =
v2

En e rgy in c id en t a t o r be y o n d t h e cr it i ca l a n gl e i s t h ro w n
ba c k to th e su rf a c e a s p o s t- c ri t ic a l r ef le ct i o n s a n d
re fr a ct i o n s. I t is n o t a v a i la b le t o i ll um i n a t e d ee pe r
re fl ec t o rs . T h i s i s e sp e ci a ll y n o t ic a b le in t h e n ea r su r fa c e
w h e re th e v el o c it y c o n t ra s t a t t h e b a se o f t h e
w e a t he ri n g ca n a p p ro a ch 1 / 2 o r l es s. S in c e t h e a rc si n o f
1 / 2 i s 3 0 d eg ree s, t h is m ea n s o n l y a n a rr o w co ne o f
en e rgy p en et r a te s to th e su b su rf a c e.
Surface source
Base of weathering Transmitted cone

True Amplitude Processing
W e h a v e s ee n t h a t a s im p l e m o d el f o r a 1 - D s e is m o g ra m
p re d i ct s th at t h e se is m i c d a t a c o n s i st s o f b a n d - l im i t ed
re f le ct i o n c o e f ic i en t s:
s t = w t •r t
W he r e w is a se is mi c w av e le t, r is th e e ar th 's r e fle cti vi ty
e x pr e sse d a s a fu n ct io n o f 2- w a y ve rtic a l t r a ve lti me , s i s
th e se i sm o g r a m , an d • d e n ot e s c on v o lut io n . S in ce w
g e n e r a ll y co n ta in s sig n if ica n t e ne rg y o nl y o ve r so m e
ch a r a ct e r is tic fr equ e n cy b a n d w id th , i f w e vie w th e
co n v ol ut io n a s a m ul tip li ca ti o n in th e fr e q u e nc y do m a in ,
w e se e th a t s i s i nd e e d a b a n dl im ite d ve rsio n o f r . (I f w i s
n o t z e r o p h as e th e n th e r e is a p ha s e sh ift a s w e l l.)
T h er e a r e m a n y r ea l ea r th w a v e p ro pa ga t io n e ff ec t s
w h i ch ca us e t he r a w se is m ic d a t a t o d e vi a t e co n s id e ra b ly
f ro m t h is m o d el . T r u e a m p l it u d e p ro c e ss in g i s a " ho ly
gr a il " o f t he s ei sm i c d a t a p ro c es si n g w o r ld a n d re fe rs t o
a p ro ce ss in g seq u e n ce w h ic h , w h en c o m p le te , y i el d s d a t a
w h i ch is a cc u ra t el y re p res en t a bl e a s b a n d li m it ed
re fl ec t io n c o e ff ic ie n ts .
W h il e n o t y et s tr ic t l y p o ss ib l e, m a n y d a ta p ro ce ss in g
f l o w s c o m e q u it e c l o se t o b e i ng t ru e a m p li tu d e
p ro c es si n g. G e n er a l ly , t ho u g h n o t e x cl u si v el y , t hi s
m ea n s the a vo i d a n ce of s t a ti st ic a l a m p li tu d e
c o rr e c t io n s like AGC in fa v o r of d et er mi n is ti c
c o rr e c t io n s l i k e s p h er i ca l d i ve rg e nc e a nd e x p o n en ti a l
ga i n .
I t i s n ot u n c om m on t o f i n d m od e rn p r oc e s se d s ei s m i c
d at a wh i c h i s r ou g h ly p r op o rt i on a l t o w el l l og d e ri v ed
r ef l ec t i on c oe f fi c i en ts o v e r l im it e d ti m e z o n e s .

Automatic Gain Correction (AGC)
A u t om a t ic g ai n c or re c t io n m e t h od s a t t e m p t t o p e rf o rm
n e c e ss ar y a m pli t u d e a d j u s t m e n t s t o se i sm ic d at a
b as e d p u r el y o n s t at i st i c s o f t h e o b se r v ed a m p l i t u d e
d e c ay . T h e y s h ou l d be c o n t ra st e d wi t h d et e rm in i s t i c
m et h o d s w h ic h u s e a p h ys i c al m o d e l o f o n e o r m or e
d e c ay p ro c es s es t o d et e rm in e c or re c t i on f ac t or s .
G en e r al ly , A G C m e t h od s a r e si m ple a n d e f fi c i en t bu t
t e n d t o p ro d u c e u n p h y si c al a m p l i t u d e d i s t or t io n s . T h e y
a re u s ef u l i n s i t u at i on s w h er e p h ys i c al l y m e an i n g f u l
a m p l i tu d es a re l es s i m p o rt a n t t h an " we l l ba la n c ed
t r ac e s.
T h er e a r e m an y A G C a l g or it h ms in c om m o n u s e . A
s i m p l e , ef f e c ti v e m et h o d i n v ol v e s t h e d e f in it i on o f a
t e m p o ra l w in do w s i ze a n d t h e m e as u re m e n t o f t h e
t r ac e rm s a m p l it u de o v er t h at w i n d ow . T h e wi n d o w i s
t h e n i n c re m e n t e d a n d th e m ea su r m e n t re p e at e d . T h e
r e su l t i s a se t o f rm s a m p l i t u d e m ea s u re m e n t s a t
d i s c re t e t i m es wh i c h d ef i n es a n ' am p l it u d e m o d e l ' o f
t h e t ra c e. T h i s m od el i s t h e n l i n ea rl y i n t e rp o l at e d t o
t h e t r ac e sa m p l e r at e a n d t h e A GC ' d t r ac e i s c o m p u t e d
b y d i v i d in g t h e o r i gi n al tr ac e b y t h e a m p l it u d e m od e l .
r c s w it h t h eo r et i c a l
a mp l it u d e d e c ay
r e f l ec t i o n c o ef fi c i e nt s

i n t er p o l a t ed r m s a m p l i tu d e m o d e l
d i s c re t e r m s m e a s u r e s
A b ov e i s t h e co n s t ru c t i on o f a n r m s a m p l i t u d e m o d e l
f ro m m ea su r e s ev e r y . 1 s e c on d s a n d t h e n i n t e rp o l at e d .
B e l ow i s t h e a p p l ic a ti o n o f t h at m od e l t o t h e t r ac e .
A G C ' d r e s u l t. T r a c e d i vi de d a m p l i tu de m o de l .
r m s a m p l it ud e m o d e l
T r a c e s h ow i ng a m p l it ud e d e c ay

B e l ow i s a c om p ar i so n o f a d e t er m i n i s t ic a m p l i t u d e
r es t or at i on a n d s ev e ra l d if f e re n t A G C p ro d u c t s. N ot e
t h at t h e re l at i v e e v e n t 's t an do u t ' ( t h e a m p l i t u d e r at i o
b et w ee n a n y t wo e v e n t s) i s be s t p re s er v ed b y
d e t er m i n i st i c m e t h od s a n d s ec o n d ar il y by lo n g A G C
o p er at or s .
AGC .4 sec
operator.
AGC .1 sec
operator.
AGC .0 2 5 sec
operator.
D e t er m in i st ic g a i n
S y nt he t i c w it h t - 1
a m p l i tu d e d e c a y
S y nt h e t ic w i th n o
a m p l i t ud e l os s e s
T h e t wo m os t c om m o n m i st a ke s w i th A GC a r e t o u se i t
e x cl u s i v l y fo r a l l g ai n a d j u s t m e n t s o r t o a v oi d i t e n t ir e ly .
I n t h e f i r st c as e , A GC s h ou l d be u s ed w i th c au t io n if t h e
i n t en ded i n t e rp r e t at i on m e t h od p l ac e s e m p ha s is o n
r el i ab le a m p l it u d e i n f or m at i on . I n th e s e c on d c as e , A G C
o f t e n l e ad s t o su per i or r es i d u al st a t ic s a n d v el o ci t y
a n a ly s es si d e f l ow s e v en w h en n e v er u s e d i n a m ai n fl ow .

A co m p a ri s on o f A G C o p e ra t or l e n gt h s o n a r ea l , r aw ,
s e is m i c tr ac e s h o ws h o w t h e c h o ic e o f o p er at or l e n g t h
c a n d r as t ic a ll y a f f ec t t h e e v en t c h ar ac t er . A l s o
a p p ar e n t is t h e e f fe c t kn o wn a s a n A G C sh a d ow z o n e .
T h i s o c c u r s wh en a p ac ka ge o f e n e rg y ( i n t h e c as e t h e
f r is t br ea ks ) h as m u ch h i g h er a m p l i tu d e t h an a d j a c en t
e v e n t s. T h e a d j a ce n t e n e rg y t e n d s t o h av e a
s u r p re s s ed a m p l i t u d e o v er r ou g h ly t h e l e n g th o f t h e
A GC o p e r at or .
A G C " s h a d ow " z on e s
2. 0 se c A GC
1 .0 se c A GC
.5 se c A GC
. 2 5 se c A GC
R a w t ra ce
A m aj o r c on ce r n wh e n u s in g a n A GC i s th a t se r io u s
d i s t or t i on s i n t h e e m be d d e d wa v e l et ca n o cc u r i f t h e
A GC o p e ra t or l e n gt h i s s h or t er t h a n t h e s ou rc e
w av e f or m . T h i s c a n re s u lt i n a s t ro n g d eg r ad at i on o f
t h e p er f or m an c e o f d e co n v ol u t i on a l go ri t h m s . T h i s wi l l
b e co m e m o r e c le ar a f t er t h e re ad e r h a s s t u d i ed
d e c on v o l u t io n i n t h e n e x t c h ap t e r .
Trace Equalization (TE) or Trace Balancing
T r ac e s fr om r aw f i el d r ec o rd s c an o ft e n h av e w il d l y
v ar yi n g t o ta l ( r m s ) p o we r l ev e l s. T h e re a re m a n y
p o ss i bl e c au s es i n c l u d i n g: s h ot s t re n g t h v ari a ti o n ,
g eo p h on e c ou p li n g v ar i t io n , n ea r s u r f ac e ge ol og y
c h an g e s, so u rc e - r ec e i v er o ff s e t , a n d m o re . ..
E v en i n c a se s wh e r e d et e rm i n i s t i c g ai n i s p r ef e rr e d ,
s om e s or t o f t r ac e b al an c i n g sh o u l d s t il l b e p er f or m e d .
O t h e rw is e , h ig h r m s p ow er t ra c es ( w h i ch a re o f t e n t h e
n oi s i es t tr ac e s ) , w il l d o m i n at e i n st a c ki n g a n d c r os s-
c or re l at i on s .
A s im p l e m e th o d is ca l le d t ra c e e q u a li za t io n , o r T E , a n d
i s u s u a ll y s y n on o m ou s wi t h t r ac e b al an c i n g . T E i s a v er y
s i m p l e p r oc e s s i n w h i ch a l l t r ac e s a re a d ju st e d t o h a v e
t h e s am e rm s p ow er l e v el a c c or d i n g t o:
o u t pu t t r ac e = i n p u t t ra c e/ ( rm s p ow e r o f i n p u t t ra c e)
A c om mo n v ar ia n t o f T E i s t o c om p u t e t h e r m s p o we r
o v e r a p ar t i c u la r t im e z o n e i n st e ad o f t h e en t i r e t ra c e. I f
t h e t i m e z o n e v ar i es in w id t h , t h en c ar e m u s t b e t ak en
t o n or m al i ze t h e rm s p ow er m ea su r e s f or t h i s e f fe c t .
C au t i on sh o u l d a l w ay s b e ex e r ci s e d wh en i n t e rp re t in g
se i s m i c p l ot s wh e r e a t ra ce e q u al i za t io n o r A G C h as
be e n a p p l ie d a s a n o p t i on i n p l ot t i n g . W h i l e t h is m ay b e
a c on v en i e n c e i t m e an s t h a t t h e d a t a d i s p l ay m ay n ot
t ru l y re p r es e n t t h e d a ta a s s t o re d o n d i s k o r t ap e . F o r
ex a m p l e, d at a th a t i s w il d l y u n b al an c e d f ro m t r ac e - t o-
t ra ce m a y a p p e ar t o h a v e g oo d a m p l it u de v a ri at i on ,
le ad in g to e r ro n eo u s p r oc e ss i n g d e c i si o n s.
Constant Q Effects
S t r i c tl y s p e ak i n g, c on s t an t Q t h e or y r ef e rs t o a Q w h i ch
i s i n d e p en den t o f f r eq uen c y b u t m a y s t i l l d e p en d o n
t i m e . F o r s i m p l i ci t l y, we wi l l a ls o a ss u me time
i n d e p en d en c e . N ot e t h at t h e a t t en u a t io n c an be w ri t t en
as :
exp –πft/Q = exp –πfx/(vQ) = exp –πx/(λQ)
= exp –πn λ/Q
w h e r e w e h a ve us e d λ f= v a n d n λ = x/λ i s t he n um b e r o f
w a v e le n g th s th a t fit i n th e di sta n ce tr av e le d. T h us , a s a
w a v e fo r m p r o p ag a te s, it is co n ta n tl y b e in g at te nu a te d
w it h th e hig he r fr equ e n cie s b e in g a tt e n ua t e d fa s te r . I f
W (f ) is t he sp e c tr u m o f o u r s o ur c e w av e f or m, w ( τ), t he n
a ft e r p r o pa g a t in g a tim e t , t he a m p lit ud e sp e c tr u m o f
th e p r o p a g a tin g w a v e fo r m h a s b e c o m e :
Wp f = W f exp –πft/Q
If w e a ss um e Q=50, a n d a s pecific s h a pe f o r |W (f )| , t h e n w e
c a n c ompu te th e ampl it u de s p ec trum o f t h e pr o pagating
w aveform at any t im e:
0
-50 Wf
.5 sec
-100
1.0 sec
-150
1.5 sec
-200
2.0 sec
-2500
50 100 150 200 250
frequency (Hz)
Constant Q Effects
T hu s w e se e t ha t s e i s mi c d at a m u s t a c t ua l l y c o nt a i n a
w av el e t w i t h c o nt i nu ou s l y d e c r e a s i ng b a nd wi d t h. T h i s
m e a ns t h e d a t a s i gn al s p e c t r um i s a c t ua ll y a f un c t i o n o f
t i m e a n d i s s a i d t o b e n on s t at i o na r y ( o r, e q ui v a l e nt l y ,
t i m e- v a ri a nt ) . D e pe nd i ng u p on t he v a l ue t a k en t o
c ha r ac ter i z e t h e b ac k g r ou n d n o i s e , we o b t ai n t he s e
s pe c i f i c m ax i m um s i g n al f r e qu e n c y e s t i m at e s ( b a s e d o n
t he p r e c e di n g g r a ph ) :
time .5 sec 1.0 sec
noise 1.5 sec 2.0 sec
100 db down 180 Hz 120 Hz 80 Hz 70 Hz Table showing

130 Hz
predicted signal
75 db down 80 Hz 70 Hz 55 Hz
band for Q=50
50 db down 80Hz 60 Hz 45 Hz 40 Hz
25 db down 45 Hz 35 Hz 30 Hz 25 Hz
Here is a repeat of the Q analysis for Q=100.

0
-50
Wf
.5 sec
1.0 sec
-100
1.5 sec
-150
2.0 sec
-200
50 100 150 200 250
frequency (Hz)
time .5 sec 1.0 sec
noise 1.5 sec 2.0 sec
Ta ble sh o wi ng
100 db down +200 Hz 185 Hz 140 Hz 120 Hz
p re d ic t ed
75 db down 180 Hz 130 Hz 105 Hz 80 Hz sign al ban d
fo r Q= 1 0 0

Constant Q Effects
Thus fa r we have di scu ssed th e e ffect s o f a ttenuatio n o n th e
ampl it ud e spectrum o f t he propagati ng waveform but th e
p hase effects are a ls o dramatic.
Consider a 1-d earth with constant Q properties:

x
A 1-d attenuating earth:

in
out
{output spectrum} =
{input spectrum}*exp(-pi*f*x/(v*Q))
• An input impulse suffers attenuation at all non-zero frequencies

• The amount of attenuation is proportional to x/v = t
• T h e a t te nuation is n e ce ssaril y couple d w i th m in i mu m
phase d i sp er si o n ( Futte rman, W.I ., 19 62, Dispe rsi ve Body W av es, JGR, 67 ,
52 79 -5 29 1)
I f th e e a r t h b eha v e s lin e a r l y, th e n w e ca n s til l a r g u e th a t

su p e r p o sit io n h o ld s. T h us t h e i mp u lse r esp o ns e o f an
e a r t h w it h r e f le c tiv it y { r } is t he s up e r p o sit io n of a se t o f
d e la y e d a nd p r o g r e s siv ly m o r e a t te n ua t e d w a ve f o r m s:
in
A three reflector
earth
out
T hr e e s u pe r i mp o s e d w av e f or ms s h ow i ng i nc r e a s i ng
a t t e nu a t i o n wi th i nc r e as i n g t i m e .

Constant Q Effects
Matrix Model of the Linear Attenuating Earth
Stationary Impulse response of

N on-stationary Q impulse
Earth a constant Q Earth
response matrix
Response
T he c on s t r uc ti on o f a n on s t a t i o na r y mu l t i p l e f r e e
s y nt h e t i c s e i smo g r am i s s h ow n f o r a c on s t a nt Q e ar t h
ha vi n g 3 r e f l e c t o r s . Th e ma t r i x m ul t i p l i c a t i o n s h ow n he r e
i s p e rf or m i ng a c o nvo l ut i on as d e s c r i b e d o n pa g e 2 -1 1.
T he c o nvo l ut i on ha s b e e n m ad e n on s t a t i o na r y b y
c h an g i ng t h e w a ve l e t i n e ac h c o l um n o f t h e c on vo l ut i on
m at r i x .

Minimum Phase, Intuitively
I nfinitely m a n y w avelets c a n be c ons t ruc t e d whic h hav e th e

s ame a m pl i t ude s pectrum b y making different assumptions
a bo u t phas e .
-40
-80
0 100 200
Frequency (Hz)
-0.1 0 0.1 0.2

s ec o n d s
H o we v er , o n l y a f e w o f th es e h av e a n y p r ac t i ca l u s e. T h e
m i n i m u m p ha s e w av e l et i s d i st i n g u i sh e d f r om a ll o t h e r s
b y be i n g t h e m o st f ro n t- lo ad e d o f t h e " c au s al " wa v e l et s .

in Linear, causal,
out
attenuating
earth
t=0
M in i m u m p h a se w a v el et s a r is e n a t u ra l l y i n t h e ea rt h . O nl y
t h e a ss u m p t io n s o f ca u s a l it y a n d l in e a ri t y a r e n ee d ed t o
s h o w th a t a tt e n u a t io n in th e e a rt h i n a m in i m u m p h a s e
p ro ce s s. ( F utte r ma n, 1 962, JG R vol 73 , p 3 917- 393 5)
T h e a m p l i t u d e s p e ct r u m a l on e i s s u f f ic i e n t t o d e t er m i n e
u n i q u e l y t h e m i n i m u m p h as e w av e le t . T h e p h as e
s p e ct r u m , φ ( f ) , m ay b e c om pu t ed a s :
φ f = H ln A f
where H denotes the Hilbert Transform.

I t is a co m m o n m ist a ke t o th in k t ha t " m in im um ph a se "

r e f e r s to a p ar tic ul ar p h a se sp e ct r u m w h ic h, i f pr e se r v e d ,
m a in ta in s a d a ta se t 's "m in im u m p h a se n e ss ". W e ha v e j us t
se en t h at t his is no t th e c a se . In st e a d , m in im um p h a se
r e f e r s t o a p a r ti cu la r m at h e ma t ica l r e la ti o ns hi p e x ist in g
b e t w e e n th e a mp li tu de a n d p h a se sp e ct r a so th a t
k n o w le d g e of e i th e r o n e i s s uf fic ie n t to c o mp u te th e
o th e r .
W h e n a s ei sm ic d a t a s et i s sa id t o be m in i m u m p h a s e , w e
ge n er a l ly m ea n th at th e e m be d d ed w a v el et h as t h i s
p ro p er ty a n d n o t t h e tr a c es th e m s el ve s . C er t a in l y t h e
ea rt h 's r ef l ec t iv i ty f u n ct i o n i s n o t m i n i m u m p h a s e.
T r u e o r F a l s e : If a da t as e t i s min i m um ph a s e a l re a dy , t hen a
z er o ph a s e filter w i l l preserve m in i mu m p h as e b e c aus e i t
do e s not change t he p h as e i n an y wa y.
T rue o r F a l se : I f t he a m p li tu d e s p e ct r u m o f a m i ni m um
p h as e d a ta set is c h a nge d , th e n th e p ha s e sp e c tr u m mu s t
a ls o ch a n g e t o p r e s e r v e t h e m in im u m p ha s e r ela t io n sh ip .
T ru e o r F a l s e : I t h a s b ee n p ro v en be yo n d d ou b t t h a t
s ei s m i c d a t a f r om i m p u l s iv e so u rc e s i s m i n i m u m p h a se .
T r u e o r F a l s e : Al l phys ic a l processes ar e m i n i mu m phas e .
True or False: A minimum phase process can have zero amplitude

over part of its spectrum.
T r u e o r F als e : A ba n d l i m it e d p r o c es s c a n n e ve r t ru l y b e
m i n im u m p h a se .

Minimum Phase and the Hilbert Transform
T h e c o n c ep t o f m i n im u m i s i nt i m a te ly l in ked w i th t h a t o f
c a u sa l it y . F or o u r p u rp o s es , a ca u s a l t im e ser ie s i s o n e
w h i ch v a n i sh e s f o r t< 0 . T h e i n ve st ig a t io n o f c a u s a l
f un c t io n s is f a c il it a t ed b y t h e f o l lo win g F o u ri er tr a n sf o rm
p a i r:
h(t)
1 1 1.0
ht = + sgn t
2 2
Graph of the step function h(t)
Thus h(t) is the unit causal function also called the step function.
T he F o ur i e r tr a nsf o r m o f the s t e p f u nc t i o n i s : ( See P apo ulis , A. ,

1984, Sig nal A na lysis, McG raw-Hill for a pr oo f. )
i
H ω = πδ ω –
ω
If f(t) is any causal function, then:
ft = ftht
Taking the Fourier transform of both sides of this result gives:
1 i
Fr ω + iFi ω = Fr ω + iFi ω • πδ ω –
2π ω
Equating real and imaginary parts gives:
1 1 1
Fr ω = Fr ω + Fi ω •
2 2π ω
1 1 1
Fi ω = Fi ω – Fr ω • ω
2 2π
T hus , t he s pectrum of a c ausal f unction h as its real a nd
i maginary p arts linked b y t he relations :
1 1 1 1
Fr ω = π Fi ω • and Fi ω = – π Fr ω •
ω ω
If we write out the convolution integrals, we obtain:
∞ ∞
1 Fi ϖ –1 Fr ϖ
Fr ω = π dϖ Fi ω = dϖ
ω–ϖ π ω–ϖ
–∞ –∞
T h e s e i n t e gr al s a r e c al l e d H i lb e rt t r an s f or m s a n d we s ay
t h at t h e re al a n d im a g in a ry p a rt s o f a c au s al si g n al f o rm
a H i l b er t t ra n sf o rm p a ir . I n o u r c as e, w e a c t u a ll y w an t t o
r el at e t h e a m p l it u d e a n d p ha s e o f a c au s al s i gn a l t o o n e
a n o t h er , n o t t h e r ea l a n d i m ag i n ar y p ar t s. H ow e v er ,
r ec al l i n g t h at :
iφ ω
Fω = Aωe ⇒ ln F ω = ln A ω + iφ ω
The answe r s eems immediate that the p has e an d lo g
amplitude spectrum ar e Hilber t t ransfor m p airs. How ever; w e
must as k:
- Under what circumstances c an w e t ak e the log
spectru m?
- Does the l og spectrum still correspon d t o a c ausa l time
domai n function?
T h e a n sw e r to th e f ir st q u e st io n is th a t w e c a n ta k e th e
lo g so l o ng a s A (ω ) ≠ 0 . T his is e q u iva l e nt t o sa y in g th a t
th e t im e se r ie s f (t ) mu st ha v e a st a b le i nv e r se .

For the second question, consider the z transform of f(t):
∞
Fz = Σ fkz
k
k = 0
W e k n ow th a t, s i nce f i s causal, t he n F ( z) c o nta i n s no

neg at i ve po w e r s of z . N o w con s i d e r t he w e l l k n ow n s er i e s
e xp a ns ion f o r the log a r i t hm :
2 3
1–u 1–u
ln u = 1–u – + +
2 3
w h ich is v al id i n th e r e g i o n 0 < | u | < 2 . S in ce F (z )
co n ta i ns o nly p o si tiv e p o w e r s o f z a nd ln ( u) c on t ai ns o nl y
p o sit iv e p o w e r s o f u , th e n ln (F ( z )) c o nt a in s o n ly p o si tiv e
p o w e r s o f z a n d th e r efo r e i s t he tr an sf o r m o f a ca u sa l
ti m e se r i e s. T h e r efo r e w e c o nc lu d e :
∞
∞
–1 ln A ϖ 1 φϖ
φω = dϖ ln A ω = π dϖ
π ω–ϖ ω–ϖ
–∞
–∞
For our purposes, we state the following important result:
F or a c au s a l , s t a b l e f unc t i o n w i th a c au s a l , s t a b l e
i nv e r s e , t h e p ha s e a nd l og am pl i t ud e s p e c t r um f or m a
H i l b e rt t ra ns f or m p a i r . I n pa r t i c u l ar , t h e p ha s e ma y b e
c o mp ut e d a s t he Hi lb er t t r a ns for m o f t he l og o f t he
a mp l i t ud e s p e c t r um . S uc h a f u nc t i o n i s s a i d t o b e
mi ni mu m p ha s e .

W e h a ve d emonstrated th a t a ca u sal, st a bl e time s eri es w i th

a c a us a l, st a bl e in v er se i s comple te ly d et er mi n ed by e it h er
it s amp l itude o r phase sp ec t ru m . G i ve n one, t he o th e r ca n
be computed . I n p a rt ic u la r, t he p ha s e s p ect ru m i s compu t ed
as:
φ ω = H ln A ω
w h e r e H d e n o te s t he H il b e rt t r a n s fo r m . W e c a ll e d s uc h a
w a ve f or m m i ni m u m p h a s e . T he r e a s o n f o r t h is n a m e
c o m e s f r o m a t h e o r e m ( R o b in s o n , E. A . a nd Tr ei te l, S. , 1 98 0,
G eo p h ys i ca l S ig n a l A na l ys is , P r en t i ce -H al l) w hi ch s h o w s t ha t, f o r a ll
c a u s a l w a ve le t s w it h t h e s a m e a mp li t ud e s p e ct r um , t h e
m in i mu m p ha s e w av e le t a r r iv e s t he s o o ne st w i th t h e
m o s t e n e r g y . M a th e m at ic a ll y, t hi s i s s t at e d b y p r o vi n g
t h a t t h e p ar t ia l e n e r g ie s : p
Ep = Σ 2
fk
k = 0
a r e l a r g er f o r t h e m i ni m um p ha s e w a ve l e t t ha n f o r a ny
o t he r w a ve l e t f or al l p . T hi s pr o of i s e q ui va l e nt t o s a yi n g
t h at t he p ha s e de l a y o f t h e mi n i mu m p ha s e w av e l e t i s t h e
s m al l e s t po s s i b l e de l a y al l o w e d b y c a us a l i t y , f o r e ac h
f r e qu e nc y .
R e ca l li n g t h a t t h e H i l b er t t r an s f or m i s j u st a c on v ol u t i on
wi t h 1 /ω , i t f ol l ow s th a t t h e m i n i m u m p h as e sp e c t u m
fo r a n y p ar t ic u l ar f re q u e n c y is i n f l u en c e d b y th e
a m p l i t u d e sp ec t r u m a t a ll f re q u e n c i es . P u t a n o t h e r w ay ,
a ch a n g e t o t h e a m p l it u de s p e ct r u m a t a p ar t i cu la r
fr e q u en cy wi l l c h an g e th e m i n i m u m p h as e sp ec t r u m a t
a l l f re q u e n c i es .

Minimum Phase and Velocity Dispersion
W e ha v e sh o w n h o w to c al cu la te t he ph a se o f a m in im um
p ha s e w a v e le t g ive n it s a m p lit ud e sp e c tr u m a n d ha v e
a lso in di ca te d th a t t he co n st an t Q at te n u a tio n m o d e l i s
m in im um p h a se . T hu s, if w e r e p r e se n t a si ng le
p r o p ag at in g co m p le x sin u so id a s:
–i2πf(t–z/v)
w f,t,z = A f e
W e c an i n fe r t h e v e lo c it y o f t h i s wa v e b y f ol l ow in g t h e
m o t io n o f a p oi n t o f c on st a n t p h as e. W i t h c om ple t e
g en e r al it y , w e c an f ol l ow t h e p o in t o f z e r o p h as e b y
e q u at i n g t h e p h as e t o z er o a n d s o lv i n g fo r z / t . T h u s w e
d e d u c e i t s v el o ci t y t o be z /t = v . I f th e sa m e w av e
p r op a ga t es t h r ou g h a co n s t an t Q m e d i u m , t h e n w e h av e :
–πft/Q + iH(–πft/Q)
wQ f,t,z = w f,t,z e
–πft/Q –i2πf(t–z/v+φQ (f))
or wQ f,t,z = A f e e
t f
where φQ f ≈ ln (Kjartansson, 1979)
πQ f0
Then solving for the velocity:
t f 1 f
t–z/v+ ln = 0 ⇒ z
t = v f ≈ v 1+ ln
πQ f0 πQ f0

Minimum Phase and Velocity Dispersion
Thu s w e s e e t h a t , i n an at tenu at i ng m e di um , velocity
becomes f requency de p ende n t , a phenomenon know n as
di s per s i on. T h e v e l oc i t y d is p e r si o n p r ed ic t e d b y t h is
t h eo ry i s st r on g e st fo r lo w Q v al u es . T h e f i g u re b el o w
p l ot s v el oc i t y v e rs u s f re q u e n c y f or d i f fe r en t t h r e e
d i ff e re n t Q ' s. N ot e t h e n e ar ly c on s t an t b eh a v i or f o r Q
o f 2 0 0 a n d t h e s t ro n g v ar ia t io n f o r Q o f 1 0 .
T h e w or d "d i s pe r s io n " a ri s e s because a p u ls e w i l l t e n d t o

s pr e ad out (d i s pe rs e ) as i t s var i ous f requencies pr o pag a t e
a t different v e l ocities. It i s t h i s di s p er s ion w h ic h l e a ds t o
t h e c h a r acteristic p u l s e s hap e o f a m in imu m phas e wavelet.
T h e pu l s e s h a pe i s s t r on g l y influenced b y t h e n e a r s urf a c e
b e c au s e i t h as dr a m atic a lly l o we r Q valu es .
0.1
0.05
-0.05
-0.1
0 0.05 0.1 0.15 0.2
Array Theory
T h e u se o f a rr a y s o f so ur c es a n d re ce i ve rs i s c o m m o n p l a c e
i n ex p l o r a t io n se is m o l o g y . T h e es s en t ia l d e t a i ls o f t h ei r
use are s tr a i gh t fo rw a rd c o n s eq u e n ce s of l in e a r
s u p er p o s it i o n a n d si gn al p ro c es s in g . C o n si d e r:
Thr ee single
f re q ue ncy
sour ce s a t hal f -
wa vle ngth
spa ci ngs
v = 2000 m/sec
f = 30 Hz
l = 2000/30 =
67 m
The sum ma tio n o f t he th ree sou rce

λ/2 wav efields . No te t he d ram atic
at te nu atio n o f sp ecific rayp at hs.
Tho ug h n ot v ery a ppar ent , t he
cent ra l por tio n o f t he wa vef ield
also ha s les s cu rv at ure.
Tw o m o r e so u rc es
at in t er med iat e
lo ca tio ns . T he
s ou r ce s pa cin g
d ecr ea se s to a
q ua rt e r- wav e len g t h
Th e su mm at ion of all f ive

λ/2 wav ef ields in creas es t he
r eje ctio n o f s te eper
r aypat h s.

Array Theory
He r e is the summati on of
λ/2 fiv e wav efi e lds fr om the
pr ev ious page .
We increase
the array
length with
two more
sources.
λ/2 No w t h e arr ay rej ect s

mor e ra ypat hs . Lo ng er
arra ys rej ect mo re, unt il
an in fin ite len g th pas ses
only v er tical ra ypat hs .
We
in creas e
t he arra y
len gt h
ag ain with
+ t wo m ore
s ou rces.
Thi s is a ve r y l ong ar r a y
wi th 9 e le me nts. The
λ/2 e ffe c ts ar e quite
dr ama tic . We note the
occ ura nc e of seve r a l
str ong r e je cti on
notche s.

Array Theory
T he respo ns e of a n array c an b e a nalyze d b y considering the
i dealized r espons e of a seque nce o f uni t spikes.
1/dx
9
8
7
notches a t n/L
dx 6
wher e n=1, 2,.. .
5
4
3
L 2
1
L= 9*dx = 9*16.7 = 150.3 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
Array in space domain wavenumber (m ) -1
Fourier transform
o f th e array.
S in ce th e a r r a y i s p u r e ly a f un ct io n o f x , its r e sp o n se i s
p ur ely a fu n cti o n o f k x . T ha t is, it w ill b e i nd e p e n de nt o f
k z o r f . H o w e ve r ; in o r d e r t o u se t he ar r a y r e sp o n se
ch a r t, w e n e e d a w a y to e st im a te k x fo r a n e ve nt o f
in te r e st . We c a n d o t hi s b y pi ck in g a h o r iz o n ta l su r f ac e
o f in t e r e st an d me as ur ing t h e ho r i zo n ta l a p p a r e n t
w a v e le n g th a lo n g it :
S i n c e w e u s u a l l y d on ' t h a v e a
m on o c h ro m at i c w a v e f ie ld , t h en λ
w e u s u a l ly m e a s u re a p p ar e n t λx
h or iz o n t a l v e l oc it y and
c om p u t e k x f r om :
kx 1 sin θ
= =
f va v
Thu s, we must pick a f requency of interest t o perform the
analysis
Array Theory Wavelet: 30 Hz,
Minimum phase
He r e w e se e a n
a rra y s i mul a ti on
fo r a br oa dba nd
w a ve fro nt w i th
the wav el e t
sh ow n at the
ri g ht. T he sa me
a rra ys as
s i m u l a t e d
pre vi ou sl y f or a λ/2
30 Hz si ng le
f r e qu e nc y
so urce a re
sh ow n.
λ/2
λ/2
λ/2

Array Theory
H e re a re b r oa d -b a nd s n a ps h ot s o f t h e s im ul at io n o f a n i mp ul s i ve
s o ur c e a n d f o u r d i f fe re nt a r r ay s. T he s ma l l b o x a t t he t o p o f
e ac h c o l um n g i ve s t h e p h y si ca l s i ze o f t h e a rr ay . I m ag es a r e
p l ot t e d w i th a s li gh t v e r ti ca l e xa gg e ra t io n a nd e ac h w av e f ro n t i s
a ct u a ll y c i rc u l a r. E ac h s o u r ce s o n f i gu r at i on i s s ho w n f ul l- b an d
a nd b ro k e n i n to f i ve d i ff e re nt s ub - b an ds . T he a rr a y s a l w ay s
a ff ec t h ig h f r e qu e nc ie s m or e s t ro n g ly a n d t he l on g e r a r ra y s
p r o du c e a n u n d i s to r t e d w a ve f ro m o n l y f or n e a rl y v e rt i ca l
t r a ve l pa t hs . T he f u ll - ba nd i ma g es a r e t h e s am e a s t ho s e o n t he
p r e vi ou s p ag e.
Array Theory
A m a j or ef f ec t o f a q u i si ti o n a rr a y s is th a t t h ey r es u lt in
a v a r ia b le ( n o n st a t io n a r y ) em b ed d ed w a v el et . F o r a gi v en
re fl ec t o r, t h e w a v el et w i ll v a r y w i t h o f fs et . F o r a g iv en
t ra c e, th e w a v el et w il l v a ry w it h ti m e. T h is h a s
s ign i fi c a n t im p li c a ti o n s fo r d e co n v o l u ti o n t h eo r y w h i ch
a s su m es a st a t io n a r y w a v e le t.

Lecture Notes
Geophysics 557
Chapter 4
The C onvolutional Model and
Deconvolution

Bandlimited Reflectivity
T h e u l t im a t e g oa l o f s e is m i c d at a p ro ce s s in g i s t o
d e t er m i n e t h e ea rt h ' s re f le c t i v it y a t a f u n c t i on o f
p o si t i on b e n ea t h t h e su r v e y. S i n c e s ei s m i c so u rc e s d o
n ot ge n e ra te u s e f u l p ow e r a t a l l f re q u e n c i es , i t i s
g en e r al ly a c c ep t e d t h at a n y r ef l ec t i v i t y es t i m at e m u s t
b e " b an d l i m i te d ". T h i s m e an s th a t t h e b e st p os s ib l e
r es u l t f r om f u ll y p r oc e ss e d se i sm ic d at a i s t h at i t
r ep r e se n t s b an d l i m i t ed re f l ec t i v i ty . W e c an t h i n k o f t h i s
r es u l t a s be i n g t h e t ru e ( br oa d ba n d ) r e fl e c t iv i t y
c on v o lv e d wi t h a z er o p h as e w av e l et .
E v en t h i s m od e s t g oa l i s ra re l y f u l ly re al i ze d .
D e co n v ol u t i on i s o n e o f o u r m a j or t oo ls f o r a c h ie v i n g
t h i s e n d . S h o r tc o m i n gs i n o u r t h e or y a n d a l go ri t h m s a n d
l ac k o f k n ow le d g e t o g u i d e t h e m u s u al l y m e an s t h a t o u r
f i n al e s t im a t e wi l l h a v e so m e u n des i re d p h as e r ot at i on o r
a n i n co rr e c t a m pli t u d e s p e c t ru m . T h e f i gu r es be l ow a n d
o n t h e n e x t p a ge il l u s t ra t e th es e c on c e p t s .
Embedded wavelet
Nonwhite (20 Hz dominant) and

minimum phase reflection coeffi-
cients
Bandlimited (10-70Hz) and 60°

phase rotated Reflection coeffi-
cients
Bandlimited (10-70Hz)
Reflection coefficients
4-2 The Convolutional Model and Deconvolution

Bandlimited Reflectivity
T h e c on s e q u e n c e o f a li m i t e d f re q u e n c y b an d i s lo s s o f
r es ol u t i on . T h a t is we c an n o t d is t i n g u is h c l os e ly sp a c ed
r ef l ec t i v i t y s p i ke s. A n u n kn o wn p ha s e r ot at i on m ake s i t
d i f f ic u l t t o d e t e rm i n e t h e p re c i se l o ca t io n o f a
r ef l ec t i v i t y sp ik e o r it s a m p l i t u d e . N o t e t h at t h e
p r es e n c e o f a p h as e r ot a te d w av e le t c an n o t b e d et e c t ed
w it h t h e p h a se s p e c t ru m o f t h e t r ac e a l on e .
Amplitude Spectra
cients

cients
Phase Spectra

cients

cients

The Convolutional Model
T h e m a j o ri ty o f th e th e o ry o f th e d ec o n v o lu t io n o f
s eis m ic d ata is ba s ed o n a ser ie s o f si m p li fy i n g
a s su m p t io n s c o n ce rn in g t he n a t u re o f th a t d a t a . T h es e
a s su m p t io n s a r e u su a l ly en c a p su l a te d a n d ref er en c ed a s
" T h e C o n v o lu t i o n a l M o d el " . W e h a v e a l rea d y s ee n t h a t , in
a l in e a r 1 - D ea rt h, w e c a n w r it e th e t he c o n st ru c t io n o f a
s y n th e ti c sei s m o gra m a s a co nv o l u ti o n o f a s o u rc e
w a v e fo r m a n d a n i m p u ls e r es p o n se :
where:
s t = Ir t •ws t
Ir t is the earth impulse response
ws t is the source waveform
s t is the earth response to the source waveform
T h e a ss u m pti o n o f l i n ea ri t y s i m p l y m ea n s t h at a li n e ar
c om b i n at i on o f so lu ti o n s t o t h e g ov e r n in g 1 - D w av e
e q u at i on i s a l s o a s ol u t i on . W h i le th is i s a n i m p o rt a n t
r es u l t f ro m p h y s ic s , f or t h e p u r p os e o f p r ov i d i n g a b as e
f or d ec o n v ol u t i on t h e ory , i t i s p r ac t i c al ly u se l es s . T h e
p r ob l em i s t h at A L L o f t h e p h ys i c s a n d g eo lo gy o f t h e
p r ob l em is c o n t ai n ed i n t h e i m p u l s e re s p on s e . T h at is , i f
w e c o n si d e r a n a tt e n u a ti n g e ar th , w i t h m u l t i p l es a n d
t ra n s m i ss i on l o ss e s , t h e n a ll o f t h e s e ef f ec t s a r e
c on t a in e d in t h e im p u l se r e sp o n s e. I n f ac t , t h e
c on v o lu ti o n al r es u l t a b ov e , is v a li d in 2- D o r 3 - D a n d
t h e re f or e t h e i m p u l s e r es p o n se c an a ls o c o n t ai n s u c h
e ff e c t s a s e l as t ic m od e co n v e rs i on s a n d sp h er i c al
d i v e rg en c e i n a d d i t i on t o th o s e a lr e ad y m en t i on ed . S o ,
a l t h ou gh t h i s re s u l t c an be p ro v en f ro m a v e ry g en e r al
t h e or y, it i s t o o ge n e ra l t o b e o f u s e t o u s. I n s t ea d , w e
m u s t m ak e a n u m b e r o f si m pli f yi n g a s su mp t io n s t o
f ra m e t h e c on te x t o f d e c on v o l u t io n t h e or y.
S h er if f a nd G eldart (Exploration Seismology, 1995, Cambridge
University Press) p re se nt th e convolutional mod el by
d ecomposi n g t h e e a rt h' s impuls e r esponse a s :
Ir t = ns t • p t • e t
where
re presents n e a r surf a c e e f f ect s be nea t h b o t h
ns t t h e source a n d receiver
r epre se n t s a ll e ff e c ts n o t o th e r wise mo d ele d
pt su c h as m u lti p le s , a b s o r p ti o n , mode
c o nv e r s io n s, e t c.
i s t h e "i mp u l s e re s p o n s e" ( t h e i r t e rm ) o f t h e
et t ar g et r ef l e c t or s . " t h i s i s t h e s ig n a l t h a t
s ei s m i c re f l e c t io n wo r k is i n t e n d e d t o f i n d ".
T hi s t e r m in o lo g y i l l u s tr a t e s s o m e o f t h e t y p ic a l
c o n fu s io n s u r r o un di ng t h e c o nv ol ut io na l m o d e l. C on s id e r
t h e ir d e fi n it i o n o f e ( t ) . I f i t i s t r u ly t h e i m p u ls e
r e s p o ns e o f t h e t ar g e t r e fl e ct o r s t h e n i t c o n ta in s a ll
m ul t ip le s , a b s o r pt io n, m o d e c on ve r s io ns a s w e l l a s
p r im ar i e s f r o m t ha t z o ne . T h is m e a ns i t i s N OT t h e
s ig n a l w e w i s h t o u n co ve r a n d t h us t he i r d e f in i ti o n i s
s e l f- c o nt r a d ic t o r y . A ls o , p ( t) i s s up p os e d t o b e a
c o n vo lu ti o n a l o pe r a t or w h ic h m o de l s a d i v e r s e r a n g e o f
e ff e ct s w i th o u t a ny j u s t if i c at io n t h a t t h is i s e ve n
p os s ib le . I n f a ct , m o s t o f t h e m e nt io ne d e f fe ct s a r e
n on s t at io na r y ( s e e 2 - 12 f o r a d e fi n it i on ) a n d t h e r e fo r e
c a n no t b e m o d e le d a s a c o nv o lu t io n. T hi s i s t h e
p r e s e n ta ti o n i n a n e x ce ll e n t, h ig h ly r eg ar d e d r e fe r e n ce
w o r k s o i t i s u nd e r s ta n da b le t h a t t h e r e i s a g r e at d e a l
o f c o nf u s io n s u r r o un di ng t he c o nv o lu t io na l m o d e l i n t h e
i n du s t r y.
W e n o w m o di f y t h e m o de l o f S he r i f f a nd G e l da r t w i t h
t he i n t e nt o f p re s e r v i ng it s s pi r i t b ut m a ki n g i t l o gi ca l l y
c o ns i s t e nt . F i rs t w e c o m bi n e t h e s o ur c e w a v e f o rm a nd
t he n e a r s u rf a c e e f f e c t s i n t o a n e qu i va l e nt w a v e l et :
we t = ws t • ns t
N e x t we d is c ar d p ( t ) a s c on t ai n i n g n o n s t at i on ar y e ff e c t s
w h ic h a re b ey on d t h e s c op e o f t h e m o d el a n d a l l ow e( t )
t o be a n i m pu ls e r e sp o n s e i n a l i m i te d s en s e o f t h e
t ar ge t r e f le c t or s:
s t = we t • e t + noise t
He r e we h a v e al s o in t ro d u c e d a d d i t i v e , s t at i on ar y, w h i t e
n o i se . T h e ea r t h ' s im p u l s e r es p o n s e i s f u rt her a s su m e d t o
b e:
e t = m t •r t
where:
rt = the earth's primary reflection series
mt = the subset of the earth's multiple reflection

response which can be modeled as a stationary process.
Thus we can write:
s t = we t • m t • r t + noise t

N ote t h a t t h e m u lt i p le t e rm c a n be e q ua ll y wel l a s so ci a t e d
wi t h t h e wa v e le t i ns t ea d o f t h e re fl e ct i vi t y so t ha t we c a n
wr it e :
s t = wm t • r t + noise t
wm t = we t • m t
T hi s r e s ul t i s a g o od s ta r t in g p o in t f o r d e c on vo lu ti on
t h e o r y s in ce i t p r e s e n ts t he s e is m i c t r a ce a s t h e
c o n vo lu ti o n o f a w a ve le t w it h t h e e a r th 's r e fl e c t i v it y . I t
i s e m ph a s iz e d t ha t o ur g o a l i s d e d uc in g t h e e a r th 's
r e fl e ct i vi t y a nd N OT i t s i m p ul s e r e sp o n s e . T he t w o a r e
v e r y d if f e r e n t.
W e r e ma r ke d t h a t t h e n o i s e i s m o de l e d a s b e i ng
" s ta t i o na r y" a nd " w hi t e " i n n a t ur e . S t a t i o na ry i n t h i s
c o nt e x t m e a ns t ha t t h e b a s i c f e a t ur e s o f t he s pe c t r um
d o n ot c ha ng e w i t h t i me . T ha t i s , i f we ex t r ac t e d
s pe c t r a f r om s ma l l w i nd ow s r a ng i ng u p a n d d ow n n ( t )
w e w o ul d f i nd e s s e nt i a l l y t he s am e s p ec t r al s ha p e .
G au s s i a n o r u n i f or m l y d i s tri b ut ed n o i s e c a n b e s ho w n t o
h a v e t h i s p r o pe r t y . T h e c on v o l ut i o n o f t w o s t a t i on ar y
s i g na l s i s a l s o s t at i o na r y. A n e x a mp l e o f a n o ns t a t i on ar y
s i g na l i s t h e i m pu l s e r e s p on s e f r om a c o ns t a nt Q e ar t h .
A s we h a v e s e e n, t he s p ec t ra l r e s po ns e c ha ng e s
s y s t e ma t i c al l y wi th t i me .
A w h i t e s p e c t ru m is o n e t h a t h as c on s t an t p o we r a t a l l
f re q u e n c i es ( e . g . " w h i t e n o is e " ) . A n i n fi n i t e l e n gt h
s i gn a l o f G a u ss i an o r u n if o rm l y d i st r i bu t e d n o is e c an b e
s h ow n t o h av e a w h it e s p e c t ru m . F i n it e l e n gt h n oi s e
s eq u en c e s h av e a p p r ox i m at e l y w h i t e sp e c t ra w h en
s m oo t h ed w it h a sh o r t o p er at or .

T ho u g h n o t s t r ic t l y p a r t o f t he c o n vo lu ti o n a l m o d e l, a
f u r th e r a s s u mt io n i s o ft e n m a d e ( i n t he c o nt e xt o f
d e co nv o lu t io n t h e o r y ) t h at t he r e fl e c t i v it y , r ( t) , i s a
w h it e a n d s ta ti o na r y t i me s e r ie s . I t c an b e e as i l y
d e mo n s tr a t e d u s in g s o n ic lo g s t ha t real e a r th
r e fl e ct i vi t y d oe s n ot h a ve a w hi te s p e c tr u m b ut i n s te a d
s h o w s c o n s id e r a b le s p e c tr a l c ol or e vi d e n ce d b y a
p r o no un ce d r o ll o ff i n p o w e r a t t h e l o w f r e q u e nc ie s .
H e re w e s ee a n e xa m p le o f a re a l r ef l ec t iv i t y ( in t im e ,
c o m p u t ed fr o m a s o n ic lo g a ss u m in g c o n s t a n t d en s it y) a n d
i ts F o u ri er s p ec t ru m :
0.4
0
0.2 -10
-20
0
-30
-0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 50 100 150
Time (sec) Frequency (Hz)
Re al r cs com pute d fro m a s oni c S pe ctr um o f the re a l rcs . Not e
lo g a t cons ta nt de ns it y the 20db rol l of f f rom 10 0 to 0
Hz .
C o nt r a s t t hi s w i t h a c o mp ut e r g e ne r at e d r a nd o m
r ef le c ti v i t y d e s ig n e d wi th a w hi t e s pe c t r um :
0.2 0
0.1 -10
0
-20
-0.1
-30
-0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 0 50 100 150
Time (sec) Frequency (Hz)
C omp u t e r ge n era t ed p se u d o S pe c tr u m o f t h e p s eu d o
r an d om rcs. r a nd om r cs . N o te t he
e s s en ti a ll y f la t ( wh it e)
s pe ct ru m.
I n o u r ba s ic c o n v o l ut i o n a l m o d e l, w e a ss u m ed th e e ff ec t s
o f m u lt i p les c o u ld be tr ea t ed a s a co nv o l u ti o n o f t h e
s o u rc e w a ve fo rm w it h a " m u l ti p le o p er a t o r" :
wm t = w t • m t
I n o u r d e v e l op me nt o f t h e 1 - D s e i s mo g r am , w e e x a mi ne d
a n a l g o ri t h m w hi c h i s c ap ab l e o f g en e ra t i ng a l l p o s s i b l e
m u l t i pl e s . C o u ld t h i s o p e r at i o n h av e be e n p e r f or m e d a s
a c o nv o l ut i o n? T he ge n e ra l a ns w e r t o t hi s q u e s t i o n i s
" n o" b e c a us e t he m ul t i p l e t r a i n gr o w s i n l e ng t h a s t i m e
i nc r e a s es a n d i s t h us f u nd a me n t al l y n on -s t a t i on ar y .
H o w e v e r , c e r t a i n c l as s e s o f m u l ti pl e s c a n b e m od el e d b y
a c o nv o l ut i on i n c l ud i n g s u rf a c e gh os ts a nd w a t e r b o t t o m
m u l t i pl e s . I n g e ne r al , i f w e r es tri c t o u r a t t e nt i on t o t h e
p o r t i o n o f a n i m pu l s e r e s po ns e l a t e r i n t i me t ha n a
m a jo r m ul t i p l e g e ne r at o r , t he n t he m u l t i pl e c o n t ri b u t i on
f r om t h a t g e ne r at i ng i n t e rf a c e c a n b e m od e le d a s t h e
c o nv o l ut i o n o f a m u l t i pl e o pe r at o r w it h t he s o ur c e
w av ef o r m. H o w ev er , a s a no t he r c a v e a t , e v e n wa t e r
b ot t om m u l ti pl e s o n f a r o f f s e t t r a c e s s h ow n o n- pe r i od i c
s pa c i ng a n d s o v i o l a t e o ur m o de l .
Summary of assumptions:
• Ea rt h's i m p u l s e r e s p on se c on s i s t s o f a r e f le c t i v i t y s e ri e s
p o s si b l y c o n v ol v e d w it h a m u l t i p l e o p e ra t or . It is al s o
s t at i o n ar y.
• T h e e ff ec t o f t h e source wavefo rm m a y be modeled a s a
simple s ta t ionary convolu ti o n wit h t he e a rt h' s impuls e
re sponse .
• Any noise is additive, white, and stationary.
• Optionally, Earth's reflectivity series is white and stationary.
Here we i llust rate the steps involve d i n the const ructio n o f a
multiple-fr ee synthetic s ei smic t race u sing a pseudo random
re fl ectivity:
0.2
0.1 Pseudo
random
0 reflectivity
-0.1
-0.2
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (sec)
0.1 0.1
M in i m u m
Minimum
p h as e
0 phase
w a v el e t t o 0
wavelet
s ca l e
enlarged
-0.1 -0.1
0 0.2 0 0.05 0.1 0.15 0.2
Time (sec) Time (sec)
0.02
0.01
Noise free
0
seismogram
-0.01
-0.02
0 0.2 0.4 0.6 Time
0.8(sec) 1 1.2 1.4 1.6 1.8
Amplitude Spectra
0
-20
Wavelet Reflectivity
-40
-60 S eismogram
-80
-100
20 40 60 80 100 120 140
Frequency (Hz)
0.02
Noise free
0.01
s e i s m o g ra m
0 with noise
superimposed
-0.01
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (sec)
0.02
0.01 Noisy
seismogram
0
-0.01
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Time (sec)
Amplitude Spectra
-20
Reflectivity
-40
-60 Noisy seismogram

Wavelet
-80
20 40 60 80 100 120 140
Frequency (Hz)
De f i n i ti o n : “ Th e em be d d e d wa v el e t ”. A s y o u n o w k n ow ,
th er e a re m an y “ wa v el e t s” i n e x p lo ra t io n s ei s m ol o gy .
T h e p h ra s e e m b ed ded w av e l et r e fe r s t o a w av e l e t
d e r i ve d b y f i tt i ng an y s e i s mi c t r a c e t o t h e c o nv o l ut i o n a l
m o de l . T ha t i s , t h e e m b ed de d w a v e l e t i s w h a te v e r s i g n a l
m us t b e c on v o l ve d w i t h t h e r e f l e ct i vi t y t o g i ve t h e t r a c e
u nd e r c o n s i de r a t i o n. E v e n w he n t h e c o n vol u t ion a l m o d e l
i s a p o o r f i t t o t h e d a t a , a n e mb e dd e d w a v el e t c an s t i l l
b e e s t i ma t e d i n t he l e a s t -s qu a r e s e ns e .
Frequency Domain Spiking Deconvolution
Perhaps the ea siest d e c o nv o lu t io n t e ch n iq u e to
c o nc e p tu a li z e is th e fr eq ue n c y d o m a in m eth o d . I t i s
su g ge st e d b y t he sp e c tr a we e x a m in e d in o u r d isc u ssi o n o f
t he co n v o lu ti o na l m o d e l:
Reflectivity
-10
-20
Wavelet
Noisy seismogram
-30
-40
-50
N oise free seismogram
-60
-70
20 40 60 80 100 120 140

Frequency (Hz)
H e r e w e s e e t he b as i c i d e a t ha t u n d e r l i es a l l
d e c o nv o l ut i on c on c e pt s : T he a mp l i t ud e s pe c t r al s ha p e
o f t h e s e i s m i c t ra c e ( s e i s mo g ra m i n t h i s c a s e ) i s
e s s e nt i a l l y s i mi l a r t o t ha t o f t h e u nk no wn w a v e l e t . G i v e n
t hi s , a l l t ha t r e m ai ns i s t o d e du c e t he w av e l e t ' s p h a s e
a nd t h en w e c a n d e s i g n a n i n v e r s e f o r i t. W e o b s e r v e
t ha t t h e n o i s y s e i m og r a m i nt ro du c e s a f u r t he r
c o mp l i c at i o n i n t ha t w e m us t r e s t r i c t o u r a t t e n t i on t o
t he s i g na l f r e q ue nc y ba nd .
N o t e t h a t w e a re re l yi n g o n t h e r ef l e c ti v i t y t o h a v e a
w h it e s p e c t ru m so t h at w e c an a t t ri b u t e a ll s p e c t ra l
" c h ar ac t e r" t o t h e w av e le t .
I f w e c a n c o m p u te t he a m p li t u de s pe ct r um o f t h e
w a v e le t b y s m o o t h in g th e a m p li tu d e sp e ct ru m o f t h e
s eis m ic tr a c e, t h en w e ca n i nv o ke th e m i ni m u m p h a s e
a s su m p t io n to c o m p le te ly s p ec if y t h e u n k n o w n w a v el et .
H er e i s t h e h e lp f il e f ro m t h e M a t la b ro u t in e , d ec o n f ,
w h i ch do es fr eq u en c y d o m ain d e co n v o l u ti o n :
% deconf algorithm
% [trout,specinv]=deconf(trin,trdsign,n,stab,phase) • Compute the power
% [trout,specinv]=deconf(trin,trdsign,n,stab)
spectrum of the design trace.
% [trout,specinv]=deconf(trin,trdsign,n)
% Add in th e stab power.
% DECONF performs a frequency domain deconvolution of the • Con volve the power
% input trace spectrum with a boxcar
% smoother to estimate the
% trin= input trace to be deconvolved
wavelet power spectrum.
% trdsign= input trace to be used for operator design
% n= number of points in frequency domain boxcar smoother • Compute the wavelet phase
% stab= stabilization factor expressed as a fraction of the spectrum with the Hilbert
% zero lag of the autocorrelation. This is equivalent to being transform.
% a fraction of the mean power. • Compute the spectrum of
% ********* default= .0001 **********
the input trace.
% phase= 0 ... zero phase whitening is performed
% 1 ... minimum phase deconvolution is performed • Divide the input trace
% ************** default= 1 *************** spectrum by the estimated
% wavelet spectru m.
% trout= output trace which is the deconvolution of trin • Inverse FFT to give
% specinv= output inverse operator spectrum. The time domain
deconvolved trace.
% operator can be recovered by real(ifft(fftshift(specinv)))
W e n ot e t h a t t h e d e c on v ol u t i on o p e r at or c an be
d es i gn e d o n o n e t r ac e a n d a p p l i ed t o a n o t h e r. T h i s i s t o
s i m u l at e t h e p ra c ti c e o f d e s ig n i n g th e o p er at o r o n a
s e gm en t o f t h e t r ac e t o a v o id l e t t in g su ch t h i n g s a s
s u r f ac e wa v es i n f l u en c e t h e d e s ig n . T h e o t h er
s i g n if i c an t p a ra m et e rs a r e : t h e l en g t h o f a b ox c ar
s m o ot h e r, a s t ab i li z at io n f ac t or , a n d a f l ag fo r z er o o r
m in i m um p h a se . I n o r d e r t o s p e c if y n , w e re c al l t h at t h e
f r eq u en c y s am p l e s i ze o f th e D F T s p e c t ru m i s Δ f = 1 / T
w h e re T i s t h e t ra ce l en gt h i n s e co n d s . T h u s , a
s m o ot h e r o f l e n gt h F s mo ot h ( i n H er t z) wi l l h av e a n u m b er
o f p o in ts gi v e n b y: F
n smooth = smooth
= TFsmooth
Δf
T he s t a b il i z a ti o n f a ct o r i s d e s ig n e d t o p r e ve n t t h e
o pe r at o r d e s ig n f r o m b e in g u n du ly i n fl u e nc e d b y n oi s e
a nd t o a vo id t h e p o s si b il i t y t h a t a d iv i s io n b y z e r o
m ig ht o cc ur w he n t he s pe c tr u m i s i nv e r te d . I t c a n b e
t h o ug h t o f a s w hi t e n oi s e a d de d t o t h e s p e c tr u m w it h a
c e r t a in p o w e r l e v e l. T ha t p o w e r l e v e l i s :
stab power (db below mean power) = 10*log10(stab)
So the default stab o f .0001 is a d b level o f 10*(-4) o r 4 0

dbdown f ro m mea n pow er .
If we c hoo s e a f r e quenc y s moot h er o f 1 0 Hz -> 10*1.6 = 16

points, and de f aul t th e s t a b f ac t or , t hen, th e
deconvolution of t h e n oi s e f r ee s e i smog r a m giv es :
0.1
0.08
0.06 exact rcs
0.04
deconf
estimate
0.02
noise free
0
seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

In the frequency domain our result looks like:
0 Exact rcs
-20
-40
-60 deconf estimate

N oise free seismogram
-80
-100
0 50 100 150 200 250
Frequency (Hz)
W e c a n s e e t h at t h e e st i m at e is q u i t e go od . W e c an b e
m o re p r e ci s e a b ou t h ow g oo d i t i s b y u si n g t h e M a t la b
f u n c t io n m x c or r wh i c h c om p a re s t w o t i m e s er i es a n d
re t u r n s t h e m ax i m u m o f t h ei r c ro s s co rr e la t io n a n d t h e
l ag a t wh ic h it o c c u r s. T h e r e su l t s i n :
m ax c or re l at i on = . 3 9 a t l ag o f . 1 s am ple s
If w e n o w r u n th e s am e pr o c e ss w it h t he sa m e
pa r a m e t e r s o n t he n o isy se i sm o g r a m w e o b t a in q ui te a
di ffe r e n t r e s ul t a s s ho w n o n th e n e xt p ag e .

0.18 Frequency Domain Spiking Deconvolution
0.16
0.14
0.12 exact rcs

0.1
deconf stab =.5
0.08
0.06 deconf stab = .01

0.04
deconf, stab=.0001
0.02
0 noisy seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
0.12
High cut filtered: 70-80 Hz rolloff
0.1
deconf stab =.5
0.08
0.06 deconf stab = .01
0.04
deconf stab=.0001
0.02
0 exact rcs
-0.02
-0.04 Filtered results Max corr Lag
Here are some deconf stab=.0001 0.0603 1.3000

-0.06
0
results 0.2
from 0.4 0.6deconf
0.8 stab1 = .01
1.2 1.4 1.6 1.8 1.6000
0.0728
maxcorr: deconf stab =.5 0.1510 3.0000

W e c a n d e v e lo p a s i m p l e m a t h em a t i ca l m o d el fo r
d e c on v o lu t i on i n t h e f r eq u en c y d o m ai n . F ir s t , t h e
c on v o lu ti o n al m o d e l f or a m u l t i p l e- fr e e se i s m i c t ra c e i s
(1)
w h er e r i s r ef l e ct i v i t y, w i s t h e wa v e l et , a n d n i s a d d i t iv e
n oi s e . I n t h e f re q u e n c y d om a in , t h i s be c om es
(2)
G e n e ra ll y , t h er e w il l b e a r an g e o f fr e q u en ci e s , c al l ed t h e
s i gn a l ba n d , o v er w h i c h t h e R ( f ) W ( f ) t e r m d o m i n at e s
o v e r N ( f ) . D en o t in g t h e b ou n d s o f t h is f re q u e n c y b an d
b y f m i n a n d f m a x, w e c an h av e t h e a p p r ox i m at i on
( 3)
w h er e t h e v er t i c al b ars ( e. g . |S ( f) | ) d e n ot e a b s ol u t e
v al u e s o r a m pli t u d e s p e ct r a. N o t e th a t b y u s i n g
a m p li t u d e s p ec t r a, we a r e d i s c ard in g t h e p o s si b i li t y o f
e st i m a t in g t h e wa ve l e t p h a se d ir e ct l y f ro m t h e d at a.
T h e n e x t st e p , s p ec t r al s m oo th in g i s d if f i c u l t t o fu ll y
j u s t i fy m at h e m a ti c al l y. D e n ot i n g a sm o o t h e d s p e c t ru m
b y a n o v er b ar , t h e " w h i t e re f le c t i v it y " a ss u mp t io n m ea n s
t h at
R (f ) ≈ 1 (4)
W e t h en a r g u e t h at s m oo t h i n g |S ( f ) | y i e l d s a n e s t im a t e
o f t h e a m p l it u d e s p e c t ru m o f t h e em b e d d e d wa v el e t .
T h o u g h we k n ow t h i s i s n ot p r e ci s e ly t ru e , i t i s
a p pro x im a t el y s o i n m a n y u s ef u l s i t u at i on s .

T h us we h av e t h e e s t im a t e
W (f ) est = S(f ) ≈ W (f ) (5)
T h e a m pli t u d e sp ec t r u m o f th e d e c on v ol u t i on o p er at o r i s
j u s t t h e in v e r se o f t h i s
(6)
G e n e ra ll y , th is s p e c t ra l d i v is i on is p r ob l em a t i c i f t h er e
a r e f r eq u e n c i es wh er e t h e e s t im a t ed w av e l et ' s s p e c t ru m
i s v er y sm a l l. W h er e i t is s m a ll u su a ll y m e an s t h at t h er e
w as n ot m u c h ra d i at e d s ou r c e p o w er a n d s o n o is e i s
l i ke ly d o m i n an t . S i n c e t h e s e s m al l v a l u es a r e i n v er t e d ,
t h e y be c om e v er y i m p o rt an t i n D( f) . G i v e n t h e s e
c on s i d e ra ti o n s, i t i s c u s t om a r y t o a d d a s m a ll c on s t an t
t o t h e e s t im a t ed wa v el e t' s a m p l it u d e s p e c t ru m p ri or t o
i n v er s i on . T h en
(7)
w h er e : ( 8)
T h e c o n s ta n t μ i s c al l e d t h e " w h it e n o i se f ac t or " o r
" s t ab i li t y fa c t or" a n d is a sm a l l p os i t iv e n u m b e r u s u a ll y
b et w ee n .0 1 a n d . 0 00 0 0 1 .
L as t l y, w e m ust e s t im a t e th e p h a s e s p e c t ru m o f D ( f ) .
U n d e r th e m in i m u m p h as e a s s u m p t i on a n d u s i n g H t o
d e n ot e t h e H i lb e rt tr an s f or m , w e h a v e
(9)

w h er e w e u s e |D ( f ) | a s gi v e n b y e q u at i on ( 7 ) . N ot e t h a t
t h e s t ab i li t y f ac t or a l so g u a rd s a g ai n s t t ak i n g t h e
l og ar it h m o f z er o i n e q u at i on ( 9 ) . S o , we n o w h av e t h e
a m p li t u d e a n d p h as e s p ec t r u m o f t h e d e c on v ol u t i on
o p er at or a n d w e a r e re ad y t o a p ply i t t o t h e s ei s m i c
t r ac e. A g ai n i n t h e f re q u e n c y d om a in , t h i s i s
(10)
I f we s u bs t i t u t e e q u a t io n ( 3 ) in to eq u at i on ( 1 0 ) w e c an
o b t ai n a n ex pre s si on fo r th e e m b ed d ed wa v el e t
r em a in in g a f t er d e c on v o l u t io n
(11)
N e g le c t i n g t h e n oi s e t e rm , we es t i m at e t h e e m b ed ded
w av e le t a s
(12)
As s u m in g a b an d p a ss fi l t er i s a p p l ie d fo l lo wi n g
d e c on v o lu t i on , w e c an r eg ar d W D ( f ) a s ef f e ct i v e ly z e ro
o u ts i d e t h is b an d w i d t h . E q u a t io n ( 1 2 ) c an b e f u r t h er
w ri t t en a s
(13)
I n t h e l as t s t e p , t h e a p p ro x im a t e u n i t y f ol lo ws o n l y i f t h e
a s s u m p t i on s o f s t at i on ar y w av e le t , w h i t e r ef l e ct i v i t y,
a n d m in i m um p h as e a re a p p r ox i m a t el y v al i d . I f th e f i rs t
t w o f ai l th en w e e x p e c t a n on - wh i t e a m p l i t u d e sp ec t r u m
f or W D ( f ) a n d i f t h e l as t f ai ls t h e n we ex p e c t a re s id u al
p h a se sp e c t ru m .
Finding a Wavelet's Inverse
If w symbolizes a wavelet and x is its unknown inverse, then the
two are related by:
w• x = 1
Here, the • d enotes convolution and 1 i s a unit vector. In
matri x n ot ation, this i s written:
w0 0 0 0 0 x0
1
w1 w0 0 0 0 x1
0
w2 w1 w0 0 0 x2 =
0
w3 w2 w1 0 x3
w0
Here we have assumed that both w and x are causal. In general

such an inverse will require infinitly many terms to produce and
exact result so we will look for an approximate finite length inverse.
If n is the length of the inverse and m is the length of the wavelet,
then the above matrix equation is:
n
Vector D is the
n X desired output
which, in this
m W = m D
case is a spike
at zero lag.

T h u s w e h a v e c h o s en n <m p r o v id i n g m o r e eq u ati o n s t h a n
u n kn o w n s a n d a r e in a p osi t io n t o s ee k a l ea s t s q u a re s
s o lu t i o n . T h e c la s s ic le a s t s q u a re s a p p ro a c h i s ( se e H att o n
et a l. , 1 9 86 p 3 1 ):
T T
W WX = W D The normal equations
–1
T T
X = WW WD The estimated X
w0 w1 w2 w3 w0 0 0 0 0 x0 w0 w1 w2 w3 1
0 w0 w1 w2 w1 w0 0 0 0 x1 0 w0 w1 w2 0
0 0 w0 w1 w2 w1 w0 0 0 x2 = 0 0 w0 w1 0
0 0 0 w3 w2 w1 0 x3 0 0 0
w0 w0 w0
M u l t i p li c at i on by W T d oe s a cr os s c or re l at i on be c au s e i t
c an b e e as i l y s ee n t o b e c o n v ol u t i on w it h t h e t i m e
r ev e rs e d w av e l et . T h i s c an be se e n t o b e:
φ0 φ1 φ2 φ3 x0
w0
φ1 φ0 φ1 φ2 x1
0
φ2 φ1 φ0 φ1 x2 =
0
φ3 φ2 φ1 x3
φ0
Where φj is the jth lag of the autocorrelation of w.

We have seen that the process of finding the m-length causal

inverse, x, of a causal wavelet, w, reduces to solving the m by m
linear system:
φ0 φ1 φ2 φ3 x0
w0
φ1 φ0 φ1 φ2 x1
0 Where φj is the jth lag of
φ2 φ1 φ0 φ1 x2 = the autocorrelation of w.
0
φ3 φ2 φ1 x3
φ0
This remarkable result says that we don't need to know the

wavelet itself, just m lags of its autocorrelation. And, if we are
content to be off by an arbitrary scale factor, then we can replace
wo by 1. How this is possible is a consequence of the following
facts:
• A causal, stable wavelet with a causal, stable inverse IS minimum

phase. (Karl, J.H., An Introduction to Digital Signal Processing,
Academic Press, 1989, see pages 35-37)
• Th e F o ur i e r tr a ns f o r m o f t he a u tocorrelati o n i s t he p o w e r
s p e c tr u m of th e w a ve l e t (Wiener-K h i n tc h i ne Theorem) . T hu s
th e ph a s e i nfor m a ti o n is no t p r e s e nt i n t he a ut ocor r e lat i o n.
T hu s , th e problem o f e s t i ma ting th e inverse t o a mini m um

p ha s e wavelet i s r e d uce d to o n e o f e s ti m a ti n g th e
a ut ocor r e lat i o n o f th e un k no w n w av e l e t. Mo s t te c h niqu e s
d o s o imperfectly.

Wiener Spiking Deconvolution
T he o r i g i na l d e c o nv o l ut i on t e c h ni qu e, a n d s t i l l t h e
w or k ho r s e o f t h e m e t ho d ol o g y i s a t i m e d o ma i n m e t h od
r e f e rr e d t o a s W i e ne r d e c on v ol u t i o n. I t re s t s o n t h e
t i m e d o ma i n c o mp u t at i o n o f t h e i nv e r s e o f a m i n i mu m
p h as e f i l t e r g i v e n i t s a ut o c or r e l at i on . B e l o w we s e e t h e
a ut o c or r e l at io ns o f t he s yn t he t i c t r ac e w hi c h w e h a v e
b e e n ex a m i ni ng :
6
5
autocorrelation of wavelet
4
autocorrelation of
3
noisy seismogram
2 a u t oc o r r el at io n
o f n o is e f r ee
s e i sm o gr am
1
autocorrelation of
0
synthetic reflectivity
-1
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
T hu s we are r e m in d e d of the f ac t th a t th e
a ut o co r r e la ti o n o f t he s e is m o g r a m i s v e r y s i mi l ar t o t h e
a ut o co r r e la ti o n o f t he w a ve l e t. T hi s i s a c o n s e q ue n ce o f
o ur a s su m pt io n t ha t t he r ef l e ct iv i ty i s a r a nd o m , w hi t e
s e q ue n c e a n d c an b e d e mo n s tr a t e d m a th e ma ti ca ll y a s
f o ll ow s:

Recall the expression for the convolutional model:
s t = wm t •r t + n t
w hi c h e x pr e s s e s t h e s e i s mi c t r a c e , s , a s a c o n v ol u t i on
b e t w e en a wa v e l e t w i t h a p os s i b l e m u l t i pl e t r a i n, w m ,
a nd a re f l e c t i c i ty , r , p l us a d di t i v e ra nd om n o i s e , n . S i n c e
a n a u t oc o r r e l at i o n i s f o rm e d b y t i m e re v e r s i ng t h e t r a c e
a nd c o nv o l v i ng i t w i t h i t s e l f , w e h a v e :
A s t = s t • s –t = wm t •r t + n t • wm –t •r –t +n –t
= wm t •r t •wm –t •r –t + wm t •r t •n –t +
n t •wm –t • r –t + n t •n –t
S in ce th e o r d e r o f co n vo lu ti o n i s u ni mp o r t a nt , t he f ir s t
te r m in th is e x p r e ss io n c a n b e se e n t o b e th e c o nv o lu tio n
o f th e a ut o co r r ela ti o ns o f w m a nd r . T h e se c o nd a n d th ir d
te r m s b o th inv o lv e th e cr o ss co r r e l at io n s b e t w e e n t w o
r a n do m se q u e n ce s, r an d n, a n d h e n ce a r e z e ro w h il e th e
la st te r m is t he au to c o r r a lt io n o f n . T hu s
A s t = A w t •A r t + A n t
Since r and n are both random sequences by assumption, their
autocorrelations are delta functions and we obtain:
A s t = A w t + pnδ t
w h e r e p n is t he m e a n n o ise p o w e r . S o w e se e th a t th e
a u to co r re la ti o n o f se i sm o g r a m a n d w a v e le t s ho u ld b e
e q u al e xce pt f o r th e p o ss ib i lit y o f a sli g ht i nc r e a se in th e
z e r o la g po w e r.

T he p r oo f t ha t t he a ut o c or r e l at i on o f t he w av l e t c a n b e
o bt a i ne d f ro m t h at o f t he s e i s mo g r am r e l i e s o n
s t a t i s t i c a l p ro pe r t i e s t ha t c an n e v e r b e ex a c t l y s at i s f i e d
i n p r a c t i c e . T h e re f o r e t w o p r o bl e m s a r i s e : w e m u s t
c ho o s e h o w m an y l a g s o f t h e a u t o c o r r el a t i o n t o a l l o w
i nt o t he s o l ut i o n, a n d we m u s t g ua r d a g ai n s t t h e
p o s s i b i l i t y t ha t t he s pe c tru m o f t he t r un c at e d
a ut o c or r e l at io n m i g ht c o nt a i n z e r o s . T he f i r s t i s
" s o l v e d " b y m ak i ng t he n u m be r o f l ag s a u s e r
p a r am e t e r, w hi l e t h e s e c o nd r e qu i r e s t h e a dd i ti on o f a
" s ta b " f ac t or t o t he z e r o l ag o f t he a ut o c o rr e l a t i on .
T hu s t he n o r m al e qu a t i o ns w hi ch m us t b e s o lv e d f o r t h e
w av el e t i nv er s e a r e m o di f i e d t o :
φ0+λ φ1 φ2 φ3 x0
1
φ1 φ0+λ φ1 φ2 x1
0
φ2 φ1 φ0+λ φ1 x2 =
0
φ3 φ2 φ1 x3
φ0 +λ
W h e r e φ is t h e a u tocorrelati o n o f th e s e is m ic t r a c e , λ i s
t he s ta b f a ct o r , a n d x is t h e un k n o w n in ver se o p e ra to r . In
c o mp a r i ng t hi s a lg o r i th m with fr e q u e n cy d o m a in d e c o n, i t
is n o te d t h a t t h e y a r e nea r ly t h e F o u r ier e q u iv a le n t s o f o n e
a n o th e r . W i nd o w i n g t h e a u t o co r re la t io n i n W iener d eco n i s
e q u iv a le n t to s m o o th in g t h e p o w e r sp e c tr u m in f r e q u e n cy
d e c o n. T he n u m b e r o f la g s i n t h e a u tocorrelati o n a n d th e
n um be r o f p o in ts in t h e fr equ e n c y d o m a in sm o o t her a r e
in v e r s e ly re la t e d . R ea so n in g v e r y lo o sely , w e h a ve:
1 1 T nsamps
nlagsΔt ≈ ⇒ nlags ≈ =
nsmoothΔf nsmooth Δt n smooth
S i nc e o ur s y nt h e t i c ha s a 2 mi l s a mp l e ra t e , i t h as r o ug h l y
8 0 0 s am pl e s , s o w e e x pe c t t he 1 6 po i n t s mo o t he r w e
us e d t o b e s i mi lar t o 8 0 0/1 6 = 5 0 l ag s o f t h e
a ut o c o r r e l at i on . He r e a r e i s t he h e l p f i l e f r o m t he M a t l a b
r o ut i n e d e c o nw :
% Algo rith m:
% [trout,x]=deconw(trin,trdsign,n,stab) • Co mpu te th e a uto co rre latio n o f
% [trout,x]=deconw(trin,trdsign,n) th e inp u t seismic trace .
%
• Wind ow th e au to correla tion
% routine performs a Weiner style deconvolution of the
% input trace
(bo xca r) to o nly n la gs
% • Se t up th e n o rm al equ a tio n s fo r
% trin= input trace to be deconvolved th e wie ne r in verse, a dd th e sta b
% trdsign= input trace to be used for operator design fac tor to th e d ia gon al, a n d solve
% n= number of autocorrelogram lags to use (and length of
• Co nvo lve th e in ve rse o pe ra to r
% inverse operator
% stab= stabilization factor expressed as a fraction of the
w ith th e se ism ic tra ce.
% zero lag of the autocorrelation.
% ********* default= .0001 **********
%
% trout= output trace which is the deconvolution of trin
% x= output inverse operator used to deconvolve trin
U s i ng e s s e nt i a l l y c o m pa r a b l e p ar a me t e r s to the
f r e qu e n c y d o ma i n e x a mp l e , w e o b t ai n f o r t he n o i s e f r e e
c a s e:
0.1
0.08
0.06 Exact rcs
0.04
deconw estimate
0.02
0 Noise free
seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
F r o m m a xc o r r , w e o b ta in a m a xi m um c r o ss c o r r e la ti on
c o e f fi c i e nt b e tw e e n t h e e s ti ma te d r cs a n d t h e e xa ct
o ne s o f . 3 9 a t a l a g o f . 2 s e c o nd s . V e r y c l os e t o t h e
r e s u lt f r o m d e co n f.
In the frequency domain these results look like: Exact rcs
0
-20
-40
-60 deconw estimate
Noise free seismogram

-80
-100
0 50 100 150 200 250
Frequency (Hz)
H e r e a re so m e sa m p le d e c o ns o f th e n o is y t r a c e w hi ch
h a ve a lr ead y be e n f ilt e r ed b a ck t o 7 0H z .
High cut filtered: 70-80 Hz rolloff
0.14
0.12 exact rcs
0.1
deconw stab =.5
0.08
0.06 deconw stab = .01
0.04
deconw stab=.0001
0.02
0 noisy seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Filtered results Max corr Lag
Here are some deconw stab=.0001 0.1262 1.4000
results from
deconw stab = .01 0.1414 1.6000
maxcorr:
deconw stab =.5 0.1802 3.0000

Prediction and Prediction Error Filters
A f or wa r d p re d i c t io n f i lt e r i s a l i n ea r c on v ol u t i on a l
o p er at or wh i c h i s d e s i gn e d t o p r ed i c t t h e n e x t el e m en t
i n a s eq u e n c e g iv e n t h e v a l u es p r e c ed i n g i t . W e c an
w ri t e t h is p ro ce s s u s in g t h e m a t ri x f or m o f c on v ol u t i on
as :
w0 0 0 0 x0 w1
w1 w0 0 0 x1 w2
w2 w1 w0 0 x2 = w3 Eqn 1
0
wm w0 xN wm+1
M u lt iplyin g b o th s id es o f t h is by t he transpose o f t h e
Toepl it z W m a tr ix a n d f o rming t h e normal equations a s w e
d id f o r in v ers e f il te ri ng gi v es:
φ0 φ1 φ2 φN x0 φ1
φ1 φ0 φ1 x1 φ2
φ2 φ1 φ0 x2 = φ3 Eqn 2
φN φ0 xN φN+1
H e r e , i n c on t r as t t o t he n o rm al e qu at i o ns f o r i n v e rs e
f i l t e r i ng , we h a v e t he s i g na l a ut o c o rr e l a t i on a pp e ar i ng
o n b ot h s i d e s o f t he e qu a t i o n. T h e s o l ut i o n t o t h e s e
e qu at i o ns g i v e s a p r e di c tio n f i l t e r x , wh i c h, i n p r a c t i c e ,
i s u s e d t o p r e di c t v a l ue s " o f f t h e e n ds " o f t he s e qu e nc e
o n wh i c h i t w as d e s i g ne d. W e m i g ht s us pe c t t h at s i nc e
t he r e i s n o p ha s e i nf o r ma t i on go i ng i nt o t he p r e di c t i o n
f i l t e r d e s i g n t ha t t he f i l t e r w i l l be m in i mu m p h a s e a n d
t ha t i s i nd e e d t h e c a s e .
W e n ow w is h t o d ra w a p ar al l el b et w ee n p r e d ic t i on
f i lt e ri n g a n d t h e d e s ig n o f i n v er s e fi l t e rs . I t t u r n s o u t
t h at t h e r e la t io n s h ip i s n ot wi t h p r ed i c t i on f il t e rs bu t
w it h a c l os el y re l at ed f il t e r, t h e p r ed i c t i on e r ro r f i l t er .
T o d e r iv e t h i s , n ot e t h a t e q u at i on 1 c an be wr it t e n a s
t h e f ol l ow i n g ex pre s si on w it h z t ra n sf o rm s :
–1
w z x z = z w z – w0
Now, we can reformulate this into:
–1 –1
w z x z – z w z = –z w0
N o t e t h a t t h e l e ft h a n d si d e is e s se n ti a l ly t h e d if f ere n c e
be t wee n t h e p re d ic t ed va lu e s, w ( z ) x ( z ) , a n d t h ei r a c tu al
v a lu e s, z - 1w( z ) . H e nc e it i s t er m ed t h e p re d ic t io n e rr o r .
M a n ip u l a t i n g f u rt h e r:
–1 –1
w z z –x z = z w0
multiply by z w z 1–zx z = w0 eqn 3
T h e o p e ra t o r , 1 -z x (z ), i s c a l l e d a p r e di c ti on e rr o r f i l t e r
o f u n i t l ag b e c a us e i t a s s e rt s t ha t w e c a n o pe r at e o n
w (z ) t o y i el d w 0 f o l l ow e d b y a s eq ue nc e o f z e r os . T h at
i s , w e c a n' t p os s i b l y p r e di c t t he f i r s t v a lu e i n a
s e qu e n c e, s o t he e rr o r i n t ha t p r e di c t i o n m u s t a l w a ys b e
1 0 0 % , h o w e v e r , w e a s s e r t t hr o ug h e qu at i o n 3 , t ha t a l l
o t h e r v a l ue s c an b e p r ed i c t e d w i t ho ut e rr o r . O f c o ur s e
t hi s wo n' t be p o s s i bl e i n g e ne r al a n d w e w i l l o b t ai n a
l e as t s qu a r e s s o l ut i o n w hi ch m i ni mi z e s t h e p r e di c ti on
e rr o r .

F o r m al ly , e q u a tio n 3 is i de n t ica l, to w i th in a sca le fa ct o r ,
o f th e z t r a ns fo r m e xp r e s sio n fo r th e de s ig n o f an
in ve r se f ilt e r fo r w . T ha t is, w - 1 m u st s a tis fy :
–1
wzw z = 1
If we write the matrix expression for equation 3, we have:
w0 0 0 0 1 w0
w1 w0 0 0 –x 0 0
w2 w1 w0 0 –x 1 = 0
0
wm w0 –x N 0
Forming the normal equations as before leads to:

2
φ0 φ1 φ2 φN 1 w0
φ1 φ0 φ1 –x 0 0
φ2 φ1 φ0 –x 1 = 0 eqn 4
φN φ0 –x N 0
A s ex p e ct e d , eq u ati o n 4 is ne a r ly i d en t i ca l t o t h e n o rm a l
eq ua ti o n s f o r a W i en e r in v e rs e fi l te r. T h u s we m ake t wo
c o n c lu s io n s:
• P re d ic t io n fi lt e rs a n d p re d i ct i o n er ro r f il t ers a r e
m i n im u m p h a se .
• S p i ki n g ( W i en er ) d e co n v o lu t i o n i s i d en t ic a l t o u n it
l a g p re d i ct i o n e rr o r fi l te ri n g.
• T h u s d e c o n v o lu t i o n r em ove s t h e p re d ic t a bl e p a rt o f
t h e tr a c e.
H a vi ng d e si g ne d a pr edi ct io n fi lt e r to pr e di ct o n e sa m p le
a h e a d, it is a s im pl e m a tte r to d e s ig n o ne to p r e d ic t α
sa m p le s a h e a d b y m o di fy in g e q u at io n 1 t o g iv e :
w0 0 0 0 x0 wα
w1 w0 0 0 x1 wα+1
w2 w1 w0 0 x2 = wα+2 eqn 5
0
wm w0 xN wα+m
P ro c ed i n g a s be fo r e, w e fo r m t h e no rm a l e q ua t i o n s
eq u i va le nt t o e q u a ti o n 2 :
φ0 φ1 φ2 φN x0 φα
φ1 φ0 φ1 x1 φα+1
φ2 φ1 φ0 x2 = φα+2 eqn 6
φN φ0 xN φα+N
T he s ol u t i o n t o e q ua t i o n 6 g i ve s a N + 1 l on g p r e di c ti on
o pe r a t o r w hi c h a t t e m pt s t o pr e d i c t α s a mp l e s a he a d. It i s
c a l l e d a g a p pe d p r e di c t ion o p e ra t o r an d pl a y s an e s s e n t i a l
r o l e i n t he s ur pr e s s ion o f m ul t i p l e s w h i c h f i t t h e
c o nv ol u t i o na l mo de l .

Gapped Predictive Deconvolution
W e h av e s ee n t h a t W i en e r sp ik i n g d ec on vo l u t io n i s
e q u i v al en t t o u nit l ag p re d i c t i on er r or f i l t er i n g. A s i m i l ar
t e ch n iq u e is t o u s e p r ed i c t i on f i l t er s o f so m e l ag o t h e r
t h an z e ro t o c om p u t e t h e p r e d i ct a bl e p ar t o f a s ei s m i c
t ra c e a n d s u bt r ac t i t f ro m t h e o r ig i n al t ra c e. T h u s, i f t h e
l ag u s e d is 1 , w e s h ou l d ge t t h e s am e r e su l t a s W i en e r
s p i ki n g d ec o n v ol u t i on . T h is t e c h n iq u e i s m o st u s e f u l in
a t t e n u at i n g m u l t ip le s t h at f i t t h e c on v ol u t i on a l m od e l .
S u c h a m u l t i p l e is t h e w at e r b ot t o m m u l t i p l e wh i c h c an
b e si m ula t ed (Backus, M.M., Geophysics, vol 24, p233-261,
1959) b y c on v o l v in g o u r se i s m og ra m wi t h t h e i m p u l s e
r es p o n se o f a n o c ea n . 2 s e c on d s d e e p ( 2 - wa y t i m e ) a n d
a n o c e an bo t t om rc o f . 4 ( h u ge ) .
0.1
Noise free
0.08
seismogram with
water bottom
0.06
multiple
0.04 Water bottom

impulse response.
0.02 Scaled by .1
0 Noise free
seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
I t is d if fic u lt to se e t he e f fe c ts o n t h e s e is m o g r a m b u t if
y o u lo o k cl o se l y a t .4 se c o nd s be h in d a m a j o r r e f le c to r,
t hen y o u sh o u ld s e e a r eve rse p o l a r it y im a g e o f it
su p e r i m po s e d o n th e se i sm o g r a m .

On the autocorrelations, we see a significant new side lobe has
developed at a lag of .4 seconds.
4
A ut o c o r r ela ti o n
3
of wa v ele t
2 A u t o c o r re lat i on of
n o i se fr e e sy nt h et ic
1 p l us wa te r bo t to m
m u lt ip le
0 A u t o c o r re lat i on of
n o is e fr e e sy nt h et ic
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
B a s e d o n t he s e d is p la ys w e a r e l e a d t o s e le ct a
p r ed i c ti o n g a p o f 1 80 s a m pl e s ( . 3 6 s e co n ds ) a nd a n
o pe r a t or l e n g th o f 5 0 s a m pl e s ( a s i n s p ik in g d e c o n) .
H e r e i s t h e h e lp f il e f r o m t h e M a t la b f u n ct io n d e co n pr :
% [trout,x]= deconpr(trin,trdsign,nop,nlag,stab)
% [trout,x]= deconpr(trin,trdsign,nop,nlag)
%
Alg o rit h m:
% DECONPR performs Wiener predictive deconvolution by calling
• De sig n a ga ppe d,
% PREDICT to design a prediction filter, nop long with lag nlag
% and stab factor, using trdsign. The predicted part of trin, min im um ph as e pre dict io n
trinhat, filt er ( w it h st ab f act o r)
% then formed by convolving the prediction operator with trin, fr om t he au toco rr ela tio n of
% and trout is computed by delaying trinhat by nlag samples and tr ds ig n.
% subtracting it from trin. The prediction operator is returned • Co n vo lv e t h e p re dict io n
% in x. op era to r w ith t ri n t o f or m
% th e pre dict ab le pa rt .
% trin= input trace to be deconvolved
• Su bt ract t h e p re dica tb le
% trdsign= input trace used to design the prediction operator
par t of t rin f ro m tr in t o
% nop= number of points in the prediction operator
% nlag= prediction lag distance in samples fo rm t he ou tpu t t ra ce.
% stab= stabilazation factor expressed as a fraction of the zero
% lag of the autocorrelation.
% ************ default= .0001 ***********
%
% trout= deconvolved output trace
% x= prediction operator
%
% See also: Peacock and Treitel, Geophysics vol 34, 1968
% and the description of PREDICT

S o , ru n n in g d ec o n p r o n th e n o i se f r ee s y n t h et i c w it h
m u l t i p l es u s in g t h e g ap a n d o p er at o r l en g t h m en ti o n ed
a n d a d e f au l t s t ab f ac t o r gi v e s:
0.1
Estimate of
multiples
0.08
0.06 True multiple

free seismogram
0.04
Estimate of
multiple free
0.02
seismogram
0 Seismogram with
multiples
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
I f w e e x a m in e t he a ut o co r r ela tio n s, w e se e th a t th e
p e r io d ic ity in th e a ut o co r r e l a tio n s a t la g o f .4 se co n d s
h a s b e e n su r p r e ss e d .
4
True
3 autocorrelation
with out multiples
2 Es t imat e o f
au to c o rr ela tio n
1 w ith o ut
mu lt ipl es
0 A utocorrelation
with multiples
-1
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Max Coeff lag
Re su lts fr om us ing max corr t o com par e
With multiples 0.9358 -0.1000
with t he nois e f ree, m ult iple free
se ismogram After deconpr 0.9736 -0.1000

H e re w e c om pa r e t h e re s u l t s f r o m f o l l ow i n g ou r pr e v i ou s
de con pr b y a de c o nw ( n = 5 0 ) wi t h a s i ng l e de c o nw w i t h
n= 1 8 0 + 50 = 2 3 0.
0.16
0.14
0.12 exact rcs
0.1
deconw (n 230)
0.08
de conp r (la g 1 80, n
0.06 50) followe d by
de conw ( n 50 )
0.04 Noi se fre e
se ismogram w ith
0.02
mult ipl es
0 No ise f ree
s eis mo gra m
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Autocorrelations
7
6 exact rcs
5
deconw (n 230)
4
d econp r (la g 1 80 ,
3 n 5 0) fol lowed by
d econw (n 5 0)
2 N oise fre e
se i smog ram wi th
1 mul ti pl e s
No is e f re e
0
s eis m ogr am
-1
-0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Max Coeff lag
Re sults from using maxcor r to compare
the t wo de cons on thi s page w ith the deconpr + deconw 0.3451 -0.3000
e xac t r c s long deconw 0.3892 -0.2000

Burg (Maximum Entropy) Deconvolution
B ur g d e c on v o lut io n i s clo se ly r e la te d to W ie n e r sp ik in g
d e co n vo l ut io n in th a t it i s a c co m pl ish e d u sin g p r e d ic tio n
e r r o r f ilt e r s o f un it la g . T he t e ch n iq ue w a s d e si g ne d b y
J. P. Bu r g ( C la e rbout, J. F ., 19 7 6, F u ndamentals of
G eophysi ca l Da t a P roces si ng, Mc G raw-Hill) in r e sp o ns e to
d o ub t s a b o u t t he W i e ne r te c hn iq u e o f w in d o w in g th e
a u to co r re la ti o n. He r e a so n e d t h at w in d o w in g th e
a u to co r re la ti o n ca u se d th e no rma l e q u a tio n s to d e si g n a
p r e d ict io n filt e r a s th o u g h t he d a ta ha d th e pr o p e r t y th a t
it s a ut o co r r ela tio n va n ish e d a ft e r n la g s . I t ce rta in ly
se ems r e a so n ab le to e x p e ct a b ette r sp e ct r a l e st im a tio n
fr o m a n a lg o r it h m th a t e xp e c ts t he a ut o co r r ela tio n to
co n ti nu e in s o me r e a s o na b l e w a y . B u r g f o un d a te ch niq u e
w h ich d e si g ne d a p r e d ic tio n e r r o r f ilt e r d ir e c tl y fr o m th e
d a ta r a th e r t h an f ir st fo r m in g t he au to c o r r e l at io n a nd
w in d o w in g it. H is t e c hn iq ue w i ll n o t b e de ve l o pe d h e r e
b u t w e w ill qu o te th e fo l lo w in g p r o p e r ti e s:
• The Burg pr e diction e r r or f i lter m in imiz e s t he sum
s quar e d er r o r f r om f o r wa r d and ba c k wa r d pr ediction.
• T h e s o- c al l ed B u r g s p e c t r u m i s c o m p u t e d as t h e in v er s e
of t h e s p e c t ru m o f t h e p r e d i c t i on e r r or f i l t er . ( T h i s i s n o t
e x p l i c i t ly d on e i n B u rg d e c on v o l u t i on . )
• The Burg prediction error filter is minimum phase.
• T ho u g h n o t c o m pu te d d ir e c tl y , t h e B ur g t h e o r y c an
b e s h o w n t o b e e q u iv a le n t t o a W ie ne r t he o r y w h ic h ,
i n s te a d o f t r un ca ti n g t h e a u to co r r e la t io n, e x tr a po la t e s
i t i n a w a y w h ic h m a x im iz e s t he r a n do m ne ss ( e n tr o py )
o f t he i m pl ie s s i g na l. ( Ka n a s ew i ch , E.R ., 1 9 81 , T i m e
S e qu e n ce A n a ly s is in G e o p h y si cs ( 3r d E d it i o n ), U ni v ers it y
o f A l be rt a P res s )

Here is the help file from the Matlab function deconb:
% A l go ri t hm :
% [trout,pefilt]=deconb(trin,trdsign,l) • De si g n a un i t l a g
%
% routine performs a Burg scheme deconvolution of the
pr e di c ti on e rr or f i l te r of
% input trace l e ng th l on t rds i g n.
% • Co nv ol ve t he pr e di ct i on
% trin= input trace to be deconvolved e r ro r f i l te r wi t h tr i n to
% trdsign= input trace to be used for operator design f or m the o utp ut t ra ce .
% l= prediction error filter length (and length of
% inverse operator
%
% trout= output trace which is the deconvolution of trin
% pefilt= output inverse operator used to deconvolve trin
N ot e t ha t t h e r e i s no s t a b f a c t o r i nv ol v e d ( t he a l g o r i t hm
i s al w a y s s t a b l e ) a nd t h at w e m us t c ho o s e t he l e n g t h o f a
pr e d i c t i o n e r r o r f i l t e r i n s t e ad o f t he n umb e r o f l a g s on an
a ut o c o r r e l at i on f unc t i o n. Ho w e ve r , as a r ou g h g ue s s , w e
m i g ht c o ns i d e r l t o b e s i m i l ar t o t h e n um b e r o f l a g s .
He r e i s t h e r es u l t f r om d ec o n v o lv in g o u r n oi s e fr e e
s yn t h et i c w i t h l = 5 0 . It ac h i e v e s a m a x i m u m c ro s s
c or r e la t i on o f . 4 3 5 5 at a l ag o f ze r o, c o n s id er a bl y be t t e r
t h a n t he ot her al go r it hm s .
0.1
0.08
0.06 exact rcs
0.04
estimated rcs
from deconb
0.02
0 noise free
synthetic
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
B e lo w i s t h e r es u l t fr o m d ec o n v o l v in g t h e n o is y
s ei sm ogr a m w it h t h r ee d i ff er en t p red ic t io n f il t er le n gt h s .
A l l re su l t s h a v e be en h i gh c ut fi l te re d a t 7 0 H z .
0.16
0.14
0.12 exact rcs
0.1
deconb l=12
0.08
0.06 deconb l=25
0.04
deconb l=50
0.02
noisy
0
seismogram
-0.02
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
max corr lag

R esu lt s fr o m u sin g m axc o r r deconb l=50 0.2227 -1.4000
t o c o mp ar e e ac h
de c o nv o lu t io n w it h t h e deconb l=25 0.2168 -1.4000
exa ct r c s: deconb l=12 0.2093 -1.4000
S o w e s ee t h at , a t l e a s t o n t h is s y nt he t ic , t he B u r g
a lg o r it hm d o e s a n e xc e ll e n t j ob , i s v e r y s t a b le , a n d n o t
v e r y s e ns i ti v e t o t h e c h o ic e o f t he p a r am e t e r l .

The Minimum Phase Equivalent Wavelet
A n y wa v el e t, n o m a t te r w h at i t s a m p l it u d e o r p h as e
s p e ct r u m , c an be s ai d t o h av e a re l at ed w av e l et c al l ed
i t s m i n i m u m p h as e eq u iv a l en t . I f t h e gi v e n w av e le t h a s
a n a m p li t u d e sp ec t r u m wh i c h is p os i t iv e d ef i n i t e, t h en
i t s m in im u m p h a se e q u i v al en t h a s t h e s a m e a m p l i t u d e
s p e ct r u m b u t a p h a s e s p ec t r u m c om p u t e d a s t h e H i l be r t
t ra n s fo rm o f t h e lo ga ri t h m o f t h e a m p l i tu d e s p e c t ru m. I f
w e l e t w( t ) d en o t e t h e a r b it r ar y w av e l et a n d F a n d H b e
t h e f or wa r d F ou r ie r a n d H il b e rt t r an s f or m s re s p ec t i v e ly ,
then:
F [ w ( t ) ] = W ( f) = A ( f ) e x p ( i φ ( f) ) .
I n t h i s e x p r es s i on , W ( f ) is t h e c o m p l ex - v a lu ed F o u r ie r
s p e ct r u m a n d A ( f ) a n d φ ( f ) a re t h e re al - v al u e d a m p l i t u d e
a n d p h as e s p e c t ra . N ow , i f W ( f ) i s p os i t iv e e v er yw h e re ,
t h e n t h e m in i m u m p h a s e eq uiv a le n t w av e l et h as a F ou r ie r
s p e ct r u m gi v e n b y:
W m i n( f ) = A ( f ) e x p ( i φ m i n( f ) ) .
w h er e
φ m in ( f) = H [ l n ( A( f ) ) ] .
I n th e m or e g en e r al c as e , w h en A ( f ) m i gh t h av e a z er o
s om e wh e r e o r w h e n l ar g e p or t i on s o f i t s d o m ai n a r e
d om in a t ed b y n oi s e, it is c u s to m ar y t o c om p ute t h e
m i n i m u m p h as e e q u i v al en t b y:
W m i n( f ) = A μ ( f ) ex p( iφ μ _ m in ( f) ).
φ μ _ m i n( f ) = H [ l n ( Aμ ( f ) ) ] .
A μ ( f ) = A ( f) + μ m a x ( A( f ) )
I n t h e f i n al ex p r e ss i on , μ i s a s m a ll n u m b er , t yp i c al l y
be t w ee n 1 0- 1 a n d 1 0- 6, w h os e e x ac t v a l u e d ep en ds o n
t h e s i gn a l- to - n oi s e r at i o a n d t h e s p e c t ra l s h ap e o f t h e
s ig n al sp ec t r u m .
The Minimum Phase Equivalent Wavelet
I n i t s m o s t g en er al fo rm , t h e m in i m u m p h as e e q u i v al en t
w av e le t i s n o t u n i q u e b ec au s e o f i ts d ep en d e n c e o n t h e
" w h i t e n oi s e f ac t or " μ .
T h e s im u la t io n be l ow s h o ws a [ 1 0 ,2 0 , 6 0 , 7 0] z er o- ph as e
O rm sb y w av e l et a n d t h re e o f i ts m i n i m u m p h as e
e q u i v al en t s . N o t e t h at a l l h av e d i st i n c l y d i f f er e n t p h as e
a s a r e su l t o f t h e i r d i f fe r in g μ v al u e s .
0
-
20 Minimum phase equivalent μ=.01
-
-
-
80
-100
Original zero phase wavelet
-120
-140
0 5 100 15 200 250
0 Frequency0(Hz)
0 Original zero phase wavelet
-0.1 0 0.1 0. 0.3

time (seconds) 2
Vibroseis Deconvolution
E x p l or at i on w it h V i br os e is ® s ou rc e s i s f u n d am en t a ll y
d i f f er en t f ro m t h e u s e o f e x p l os i v e s ou r c es a n d n e e d s
s p e ci a l c on s i d e ra ti o n i n o u r t h eo re t i ca l d e v e lo p m e n t .
I n s t e ad o f a n u nkn ow n i m p u l s i ve s ou r c e w av e f or m ,
v i br os e is a t te m p t s t o c re at e a k n ow n ex t e n d e d s ou r c e
kn o wn a s a sw e ep . A s w ee p i s t y p i c al ly a s ig n al w h i ch
m o v es c o n t in u o u s ly t h r ou g h a s p ec i f i ed fr e q u e n c y b an d
g en e r at i n g o n l y o n e f r eq u en c y i n s t an t an e ou sl y. T y p ic a ll y
s we e p s a r e l i n ea r ( t h e s am e t i m e i s s w ep t a t ea ch
f re q u e n c y) b u t n o n - l in e ar sw e ep s , w h i c h e m p h a s iz e t h e
h i gh f re q u e n c i es , a re a ls o c om mo n . H er e i s a 1 0 - 7 0 H z ,
8 s e c on d , li n e ar s w ee p :
sweep
Time (sec)
T h e c o n v ol u t i on al m od e l s t il l f i t s t h i s so u rc e e q u a ll y we l l
a s t h e im p u l si v e so u r ce . T h a t i s , gi v e n a r ef l ec t i v i t y r ( t ) ,
w e c an si m u l at e th e ea rt h ' s r es p o n s e by c o n v ol v i n g t h e
s we e p w i t h r ( t ) :
reflectivity
sweep
c on vo l ve d
with
r e f l e c ti vi t y
O b v io u s ly , t h i s is a d i f f er en t s or t o f re c or d t h an t h e
i m p u l s iv e s ou r c e a n d i s m u c h m o r e d i f f ic u l t t o in te r p re t e
b ec au se t h e s ou rc e wa v e f or m i s s o e x t en d ed . W e n e ed a
m e t h od o f co l la p s in g t h e s ou r ce t o a c om p a c t p u l se .
T h a t t u r n s o u t t o b e t h e c ro ss c or re la t io n m et h o d .

According to the convolutional model, the vibroseis record is:
M o st o f t h e se t e rm s we re d e f i n ed a l r ea d y i n o u r
d i s c u s si on o f t h e c on v ol u t i on a l m od e l . W e r e p e at t h e
d e f i n it i on s h e re :
the uncorrelated vibroseis record
the vibroseis sweep
near s ur fa c e e ff ec ts a n d v ib rator d i st o rt io n
a convolutional approximati o n t o Q ef fec t s
the subset of all multiples which are convolutional
the desired reflectivity
zero mean, white noise
Now we cross correlate with the sweep and rewrite the model as:
where
is the correlated vibroseis record
is the autocorrelation of the sweep

(Klauder wavelet)
i s t h e ef f e ct i v e “ e art h f i lt e r”
R o u gh ly s p ea ki n g , t h i s s ay s t h a t we c an u s e t h e
c on v o lu ti o n al m o d el fo r co rr e la t ed v ib r os e is d a t a i f w e
r eg ar d t h e s ou r c e w av e f or m a s t h e a u t o c or re l at i on o f
t h e s we e p . T h is is o f t en c al l ed t h e K l au d e r wa v el e t .

T h u s , i n t h e s im p le s t c as e , w e ex pec t a c o rr el at e d
v i br os e is r e co rd t o be t h e s we e p a u t o c or re l at i on
c on v o lv e d wi t h 0.25
t h e re f le c t iv i t y .
0.2
autocorrlation of
0.15 10-70 sweep wv t
(Klauder wavelet)
0.1
0.05
-0.05
-0.1
30 Hz minimum
phase pulse
wm t
-0.15
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25
0.5
0.4 wm t • r t
0.3
0.2 wv t • r t
0.1
0 rt
-0.1
-0.2
0 0.2 0.4 0.6 0.8 1 1.2

D e co n v ol u t i on o f t h e v i br os e i s s yn t h e t ic p r es e n t s a
s p e ci a l p r ob le m si n c e w e c an n o t a ss u m e t h e wa v el e t i s
m i n i m u m p h as e. M o st a p p ro ac h e s to t h i s p ro bl e m i n v ol v e
a t t e m p t i n g t o m o d i f y t h e t h e c or re l at ed v i br os e is r ec o rd
s o th a t i t s e m be d d e d wa v el e t i s m or e n e ar ly m i n i m u m
p h a se . A n i m med i at e p ro bl e m a ri s es be c au s e a m i n i m u m
p h a se wa v ef o rm c an no t be ba n d l im it e d y e t t h e v i b ro se i s
w av e le t i s e x p l i ci t l y ba n d l im it e d . T h is m e an s t h at A L L
m e t h od s wh i c h a t t em pt t o p r ec on d it i o n t h e e m b ed ded
v i br os e is wa v ef o rm m u s t em p lo y a wh it e n o i se o r " st a b"
f ac t or t o e x te n d t h e sp ec t r u m . I t i s u s u a ll y go od
p r oc t i c e t o e n s u re t h at t h i s f ac t or i s t h e sa m e a s t h a t
u s ed l at e r i n t h e d ec on a l g or it h m.
G i v e n t h i s, a n d a s su m in g t h at t h e e m be d d e d w av e le t i s
t h e K la u d er w av e le t , i t i s a s t ra i g h t f or wa rd ex e rc i s e in
s i gn a l p ro ce s s in g t o d es i g n a c o n v er s io n o p er at or w h i ch
c on v e r ts th e Kl au der w av e le t t o i t s m i n i m u m p h as e
e q u i v al en t :
0.25 0 spectrum of minimum

phase equivalent of
0.2
-20 Klauder wavelet
0.15
Klauder
0.1
wavelet -40
0.05
-60
0
-0.05 -80
Spectrum
-0.1
of Klauder
minimum phase equivalent -100
-0.15 wavelet
of Klauder wavelet
-0.2 -120
-0.1 -0.05 0 0.05 0.1 0 50 100 150 200 250
Frequency
(Hz)

A t t h i s p oi n t it i s a p p r op r i at e t o a s k w h y t h e v ib ro s ei s
r e co rd sh o u l d be d e c on v o lv e d a t a ll . A f t e r a ll , t h e s i g n al
i s ge n e ra t ed w i th a wh i t e sp e c t ru m o v e r t h e sw e p t ba n d
a n d i s n o m i n al l y z e r o e ls e wh e r e. T h us t h e z e ro p h as e
v i b ro se i s sy n t h et i c w h ic h c o n si s t s o f Kl a u d er w av el e t
c o n v ol v ed w it h re f l ec t i v i ty i s a lr e ad y o p ti m a l. T h e
a n sw e r, o f c o u rs e , l i es i n t h e o t h e r e ar t h f il t e ri n g
e f f ec t s s u c h a s t h e n e ar s u r fa c e ef f ec t , m u l t i p l es , a n d
a bs or p t i on ( Q ) . T h u s , a m in i m a l v i b ro se i s m o d e l f or
d e c on v o l u t io n t h e or y i s:
sx t = wv t • n s t • r t
U n l i ke u n l i ke t h e i m p uls i v e c a se , t h e g oa l o f v ib r os ei s
d e c on v o l u t io n i s t o re c ov e r r ( t ) o n ly o v e r t h e s we p t
b an d, e v en i n t h e n o is e fr e e c as e . W it h t h i s i n m i n d , we
u s e t h e 3 0 H z m i n im u m p h as e wa v el e t t o re p r es e n t t h e
n e ar s u rf ac e ef f e ct s a n d t h e K la u d e r w av e le t fo r t h e
s ou rc e a n d c r ea t e t h es e s yn t h e t i c s e is m o gr am s :
0.5
30 Hz min ph s an d m in p hs
Klau der wav lets con v olv ed
0.4
wit h r eflect iv ity
30 Hz min phs and Klauder

0.3 wavlets convolved with
reflectivity
30 Hz min phs
0.2
wavlet convolved
with reflectivity
0.1 Klauder convolved with
reflectivity
0 r eflectivity
-0.1
0 0.2 0.4 0.6 0.8 1

T h e r es u l t s o f W ei n e r d ec o n o n t h e s e sy n t h e ti c s i s
b el ow . N o t e t h a t we c om pa r e t o t h e r e fl e c t iv i t y
c on v o lv e d w it h t h e K la u d e r w av e le t a n d n o t t o t h e
r ef l ec t i v i t y it s e lf . A l l o f t h e s e d e c on v ol u t i on s h av e b e en
f i lt e re d ba c k to t h e s w ep t b an d .
0.45
A Decon of 30 Hz min phs and
0.4 min phs Klauder wavlets
convolved with reflectivity
0.35
Deco n o f 30 Hz min phs an d
0.3 B Kla ud er wav let s co nv olv ed
wit h ref lectiv it y
0.25
De con of 30 Hz min
0.2 C p hs wa vle t convolve d
wi th r ef le ctivi ty
0.15
D Decon of Klauder
0.1 convolved with
0.05
reflectivity
0
E Klauder convolved
with reflectivity
-0.05
0 0.2 0.4 0.6 0.8 1
Max corr Lag
W e c an s ee t h at t h e m in im u m p h as i n g o f A 0.3795 0.8000
t h e v i br os e i s re c or d p ro d u c es a b e t te r
0.3712 -2.3000
d e c on b u t it a p p e ars t h e re s u l t h a s a 9 0 B
d e gr e e p h a s e r ot at i on . I n f ac t t h i s is t h e C 0.4978 -0.1000
c as e : -0.4138 -9.1000
0.15 D
-90° rotation of A
0.1
0.05
E
0
-0.05
0 0.2 0.4 0.6 0.8 1
Deconvolution Pitfalls
T h e assumption s behind deconvolution t h e or y h e l p u s t o
u nde rs t a n d i t s b as i s an d , sometimes , t o anticipa t e p r ob lem s
b e f or e t h e y a ri s e . H e re w e w i l l examine s o m e c o m mon
deconvolution "pitfalls".
Mixed-wavetypes in the design gate.
T h e m o st c o mm o n e x a mp le h e r e is th e Surface
waveform
o cc ur an ce o f a su r f ac e w av e o r si mi la r
co h e r e n t n o ise t r a in i n t he d e si g n g a t e .
T h e s im ul a te d su r fa ce w a v e b e g i ns a t . 2 Reflection
waveform
se c o nd s. 0 0.2 0.4 0.6 0.8 1 1.2
0.3
Contaminated
0.25 -10 with surface
wave
0.2
-20
0.15 Contaminated
with surface -30
0.1
wave
-40
0.05
Simple S imple
0 -50
synthetic synthetic
-0.05
-60
-0.1 0 20 40 60 80 100
0 0.2 0.4 0.6 0.8 1 1.2
Frequency
(Hz)
Deconvolution Results
Contaminated
with surface
wave
S imple
synthetic
RCs
0 0.2 0.4 0.6 0.8 1 1.2

Here is a closeup of the ends of the traces so that the
considerable phase distortion can be appreciated:
Deconvolution Results
Contaminated
with surface
wave
Simple
synthetic
RCs
0.7 0.75 0.8 0.85 0.9 0.95 1
S o we s e e t h at t he pr e s e n c e o f t h e s ur fac e wa ve ha s
c a us e d a g r e a t de a l o f ph as e d i s t o r t i o n e v e n qu i t e f a r
f r o m t he on s e t of t h e w a ve . S i n c e t he p ha s e c or r e c ti on s
ap pl i e d b y m i ni m um p ha s e d e c o nvo l u t i o n ar e de d uc e d
f r o m a s mo o t he d r e p re s e n t a t i o n of t h e a mp l i t ud e
s pe ct r um , t he pr e s e n c e of t h e s ur f a c e w av e p e ak i n t h e
am p l i t ud e s pe c tru m c a u s e s e rr o ne o us p ha s e s t o b e
c o mp ut e d .

Filtering before deconvolution.

T h e i s s u e of f i l t e ri n g b e f or e d e c on v ol u t i o n c an b e a
c o m p l e x on e w h i c h t ak e s s om e s u r p r is i n g t wi s t s . It m i g h t
s e e m t h at on e c o u l d d og m a t i c al l y i n s i st t h a t a l l f i l t e ri n g
b e f or e d e c on v ol u t i on m u s t be m i n i m u m p h a s e. How e v e r;
a s w e sh al l s e e, t hat i s o f t en i n c or r ec t . I t g re a t ly h e l p s
t h e d ec i s i on p r oc e s s t o c o n s i d e r w h e t h e r t h e u n f i l t er e d
d at a i s i n t h e " m i n i m u m p has e s t at e " o r n o t .
W e w il l s a y t h a t s ei sm i c d a t a i s in t h e " m i ni m u m p h a se
s ta t e " i f t h ere is a s in g le e m be dd e d w a v el et a nd t h a t
w ave le t i s m i n im u m p h a se . I f d a t a is in t he m in i m u m
p h a s e st a t e, t h en a ny f il te ri n g sh o u l d be m i n im u m p h a se
t o t ry t o p r es erv e t h a t st a t e. I f n o t , z er o p h a s e fi lt er in g
m i gh t a ct u a l ly be pr ef er red i f it c a n be a rg u ed t o m o v e
t h e d a t a t o wa rd s t he m i ni m u m p h a se s ta t e .
T he s u r fa ce w a v e s y nt he t ic w h ic h w e p r e se nt e d e a r li e r
i s n o t i n t he m i ni m um p ha s e s ta t e b e ca u s e i t c o n ta in s
t w o e m b e d de d w a ve le t s : t he m in im u m p h as e r e fl e c ti o n
w a ve le t a n d t he n o n -m in im um p h as e s u r f ac e w a v e fo r m .
T he r ef o r e , m in im um p ha s e f i l te r in g before
d e co nv o lu t io n m ig h t n o t b e a p p r op r ia te . I n f a ct , a z e r o
p ha s e f il t e r d e si g n e d t o k no ck d o w n t h e s u r fa ce w a v e
p e ak s h o ul d m o ve t h e d at a t o w ar d s t he m in i mu m p h as e
s t at e . T h e e xa m pl e o n t he n e x t p ag e s h o w s t h at t hi s i s
th e ca s e .

0.6
0.5 Deconv ol ved surfac e w ave

syn theti c w it h mi nim um
phase f il teri ng to remov e
0.4 the surfac e w ave.
Deconvolved surface wave
synthetic with zero phase
0.3 filtering to remove the
surface wave.
Deconvolved synthetic
0.2 with surface wave
present.
Deconvolved minimum
0.1 phase synthetic
0 RCs
-0.1
0 0.2 0.4 0.6 0.8 1 1.2
Surface wave spectral peak

-10
-20
-30
-40
Uncontaminated spectrum
-50
-60
After filtering out the surface wave

-70
-80
-90
0 20 40 60 80 100
Frequency (Hz)

Design gate considerations.
T y p ic a ll y t h e d e c o nv ol ut i o n o p e r a to r i s d e si g ne d o ve r a
s u b s et o f t h e t r ac e c ho s e n f o r i t s h i g h s ig n al t o n o is e
r a t io . C o n s id e r a ti o n s :
- include the zone of interest

- include large dominant reflectors
- exclude surface waves and below basement
- don't design on noise
- highly non stationary data should avoid very long gates
- o p e r a t o r l e n g th s ho u l d be n o m o r e th a n 1/ 3 to 1/ 2
o f th e ga te le n g t h
0.5
1 80 m il o per ato r
0.4
d es igne d o ver .2 to .4
s eco n ds
60 mil operator designed

0.3 over .2 to .4 seconds

0.2
over .4 to .6 seconds

0.1
over entire trace
0 RCs
-0.1
0 0.2 0.4 0.6 0.8 1 1.2

Filtering after deconvolution.
R e al s e i s mi c d at a a l w ay s c o nt a i ns a g r e at d e a l o f
a pp ar e nt l y r a nd o m n o i s e . W e ' v e a l r e ad y m e nt i o ne d t he
i n ad v i s ab i l i t y o f d e s i g ni ng t he o p e ra t o r o n n oi s y d a t a.
A l s o , i t i s a l m os t a l w a ys n e c e s s ar y t o f i l t e r d a t a b ac k
t o s o me s i g na l b an d a f t e r d e c on v ol u t i o n. M o s t d e c o n
a l g or i t h ms c a nn ot d i s t i n gu i s h s i g n al f ro m n o i s e a n d s o
w hi ten b o t h. T hi s c a n h a v e a d i s a s t r ou s e f f e c t o n s u c h
n oi s e s e ns i t i v e p r o g ra ms a s r e s i du al s t a t i c s .
0
noisy data spectrum
-20
-40
-60 noiseless
data
-80 spectrum
0 50 100 150 200 250
Frequency (Hz)
0.2
noisy
0.1
seismogram
0
noiseless
seismogram
-0.1
0 0.2 0.4 0.6 0.8 1 1.2

0.5
RCs filt er ed b ack

0.4
to 60 Hz.
Deconvoled noisey
0.3
seismogram, filtered
back to 60 Hz.
0.2 Deconvoled noisey

seismogram
Deconvoled
0.1
noiseless
seismogram
0 RCs
-0.1
0 0.2 0.4 0.6 0.8 1 1.2
T h o ug h th e r esu lts o f th e d e c o nv o lu tio n o f n o is y da t a a r e

a lm o st a lw a y s b ette r w h e n f ilt e r e d b a c k , in g e n e r a l, th e
m a tch t o th e R C' s is st ill m u ch w o r se th a n w i th o ut n oi se .

Iterative deconvolution.
I t i s o f te n a s s u m e d t h at d e c on v o lu t i o n i s s om e t h i n g t h a t
n e ed s d oi n g o n c e a n d i s t h e n b e st f or go t te n . T h i s
a t t i t u d e u su a l ly le ad s t o u n der wh i t e n e d d at a w it h
r es i d u al p h as e ro t at i on s . S i n c e t h e a s s u m p t i on s o f
d e c on v o lu t i on a re n e v er p r ec i s el y m e t, i t i s o f t en u s e f u l
t o a p p l y s ev e r al d i ff e re n t d e c on s f or d i f fe r en t p u r p o se s .
F o r e x am ple , we m ay u se p r ed i c t i v e d e c on t o a t t ac k a
m u l t i p l e a n d t h e n s p i k in g d ec on v ol u t io n t o s h ar p en
r es ol u t i on .
M or e i m p or t an t l y, d ec on v ol u t io n a l g or i th m s c an n o t
d i s t in g u i s h b e tw e en s ig n al a n d n oi s e. T h u s w e m u s t t h in k
o f t h e m a s wh i t e n in g t h e s p e ct r u m o f si g n al p l u s n o is e .
I f d e co n v ol u t i on i s t h e n fo l lo we d b y a n y p r oc e s s w h i ch
c an r ej e c t n oi s e w h i le re t ai n i n g s i g n al , t h e re s u l ti n g d at a
w il l h a v e a n o n - w h it e , l ow e r r e so l u t io n , s p e ct r u m . T h e
m o st c om m o n e x am p le o f t h i s i s C M P s t ac ki n g . T h us i t i s
o f t e n n e ce s s ary t o r u n a p o s t - st a c k d e c on v ol u t i on o r
w h it e n i n g s t e p t o e n su re m ax i m u m r e so lu ti o n . I f p o s t-
s t ac k m in i m u m p h as e d ec o n v ol u t i on i s d es i r ed , c ar e
s h ou l d b e t ak en t o en s u r e t h at z er o- p ha s e f i lt e r in g wa s
n ot d on e a f t er t h e p re - s t ac k d e c on v o lu t i on .

Reflectivity Color
Here i s a r e fle c tiv it y series c o mp ut e d fr o m a n A lb er ta well
a nd p lo tt e d versus t im e .
0.06
0.04
0.02
-0.02
-0.04
-0.06
0 0.2 0.4 0.6 0.8 1
Time (sec)
Since standard deconvolution algorithms assume a white

reflectivity spectrum, we are motivated to compute the
spectrum and see if it is white.
White spectrum is flat
0
-10
Anti-alias filter rolloff
-20
-30
Low fr e que ncy de ca y i s ty pi ca l we ll l og
-40 beha vi er a nd indic a tes consi de r abl e
spe ct ra l col or .
-50
-60
0 50 100 150 200 250
Frequency (Hz)
So we see that this spectrum is non-white.

What color is it?

Reflectivity Color
Example of a reflectivity estimate via Weiner deconvolution for a
non-white reflectivity. The traditional whitened estimate is shown
(A) along with a color corrected estimate (B) and the original well
log (C).
0.25
0.2 A
0.15
0.1 B
0.05
0 C
-0.05
-0.1
0 0.2 0.4 0.6 0.8 1
Time (sec)
Reflectivity Color
A: Reflectivity estimate from normal Weiner decon
B: A convolved with a zero phase color restoration operator
C: A convolved with a minimum phase color restoration operator
D: Original well log at 2 ms sample rate
N u m b ers gi ve ( M a x im u m co r re la t io n c o ef f, l a g a t m a x
{ s a m p l es} ) f o r t h e c o rr el a ti o n be tw ee n a n e st im a t e a n d
t h e a n sw e r g iv e n in D
0.35
0.3
A
(.8037,-.2)
0.25
0.2 B
(.8470,-.2)
0.15
0.1
C
(.8816,-1.8)
0.05
0 D
-0.05
-0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Time (seconds)
Q Example
H e r e i s a n e xa mp l e com p ar i n g a s ta tion a r y, mi n i m um ph a s e
s y nt he t i c w i th a s e r i e s o f c o nsta nt Q s y n th e ti c s . E a c h
c o n s ta nt Q s y nth e ti c ha s th e s am e 30 H z , mini m um ph a s e
w a ve l e t c o nv ol v e d w ith i t as t he s t at i o na r y s y nth e ti c do e s .
0.14 wavelet
0.12 Q=25
0.1 Q=50
0.08 Q=100
0.06 Q=150
0.04 Q=200
0.02 stationary
0 RC's
-0.02
0 0.2 0.4 0.6 0.8 1 1.2
T h e e ff e c t o f Q a t t e n u at i on c an b e s e en t o h a v e a t l ea s t
t h r ee c h ar ac t e ri s t ic s : a p r og r es s iv e l os s o f f r eq u e n c y
c on t e n t w i t h i n c r ea s i n g t i m e , a p ro g re ss i v e l os s o f
o v e r al l a m p l i t u d e , a n d a p r og r es s iv e t i m e d e l ay . T h e
c on s t r u c ti o n o f o n e o f t h e se s yn t h e t i cs i s d et a il e d o n
t h e n e x t p a ge .

Q Example
E a c h o f t h e Q sy n t heti c w a s c r e a t e d b y f ir s t co n st r u ct in g
t he " Q m a t r ix " w h ic h a pp l ie s a Q r e s po n s e t o a t im e s e r ies
v ia a g e n e ra liz ed co n v o lu ti o n. H e re t h e p r o c e s s i s d e p ict e d
g r a p h ic a ll y fo r th e Q =2 5 ca se.
input time
RCs in Q= 25
Q Matrix
time seismogram
E a c h c ol u mn o f t h e Q m a t r i x c o nt a i ns t h e Q = 2 5 i m pu l s e
r e s p o ns e f o r t he i np ut t i me o f t he c o l u mn c on vo l ve d w i t h
t he 3 0 Hz m i ni m um pha s e s o ur ce wa ve f o r m . If w e F o ur i e r
t r an s f o rm e a c h c o l um n, w e c a n s e e d i r e c t l y t he Q
a m pl i t u de a nd p ha s e r e s po ns e :
input time input time

Am pl itude s p ec tr um Phase spectrum of Q
o f Q m a tr ix matrix
Q Example
T h e f i rs t s t ep in d ec o n v ol v i n g t he Q s yn th et i c s is t o
d e t er m i n e e x p on e n t i al g ai n c o rr ec t i on s a n d a p p l y t h e m .
T h e re s u lt i s:
0.14
0.12 Q=25 db/sec = 17
0.1 Q=50 db/sec = 11
0.08 Q=100 db/sec = 5
0.06 Q=150 db/sec = 2
0.04 Q=200 db/sec = 0
0.02 stationary
0 RC's
-
0.02 0 0. 0.4 0.6 0. 1 1.2
2 8
T h e s e g a i n f ac tor s w er e de t e r m i ne d e m pi r i c a l l y a s i s
s t an da r d p r ac t i c e . I t a pp e ar s t ha t Q =2 5 m ay b e a b i t
u nd e r g a i ne d .

Q Example
N ex t w e r un W e i ne r d e c o nv o l ut i on w i t h t he s am e
p a r am e t e rs f o r e ac h t r a c e ( 3 0 l a gs a n d . 0 0 0 1 wh i t e
n o i s e ).
0.14
MaxCC Lag
0.12 Q=25 0.1557 -0.5000
0.1 Q=50 -0.2965 -3.6000
0.08 Q=100 -0.3676 -3.0000
0.06 Q=150 -0.3733 -2.8000
0.04 Q=200 -0.3693 -2.6000
0.02 Q=∞ -0.4811 -2.3000
0 R C's
-0.02
0 0.2 0.4 0.6 0.8 1 1.2
I t 's c lea r f ro m t h is e xa m p l e t h a t t h e d ec o n v o l ut i o n re su l t
d eg ra d e s s te a d il y w it h d ec re a si n g Q . Kee p in m i n d t ha t
t h is i s a " be st ca s e" sc en a r io : no n o i se , n o m u lt ip l es ,
m i ni m u m p h a s e so u r ce , a n d w h it e re fl ec ti v it y . A l s o , ev en
t h e Q= 2 5 c a se i s n o t a n u n rea s o n a b le a t t en u a ti o n le ve l
be ca us e t he m a x i m u m t im e i n t h e sy n t h et ic is o n l y 1
s ec o n d . S i n ce t/ Q d et er m in e s t h e a c tu a l a t te n ua t i o n ,
1 / 2 5 is t he s a m e a s 2 / 5 0 o r 3/ 75 .

Geophysics 557 Final Exam Study Guide
What are the expressions for P and S wave velocities in terms of the Lame constants?
How does the Vp/Vs ratio depend on poison's ratio?
How is the normal incidence reflection coefficient (for P waves) related to impedance?
What is impedance?
How are normal incidence reflection and transmission coefficients related?
What is meant by the term "impulse response"?
What physical effects are modeled in construction of a normal incidence seismogram a
discussed in lecture?
Under what conditions can the earth response to a real source be modeled as a
convolution of a source waveform with an impulse response?
What is the major use for 1-D synthetic seismograms?
Why are multiples and transmission losses not typically included in such models?
What information is typically input to the 1-D synthetic seismogram computation?
Describe convolution by replacement.

Describe convolution as a weighted sum.
Use the integral form of convolution to prove that convolution is linear. That is, the
convolution of c with a+b is equal to the convolution of c with a plus the convolution of c
with b.
What happens when any function is convolved with a complex sinusoid?
What is technical meaning of a phrase " a 90 degree wavelet" ? 45 degree? any degree?
What is the definition of the 1-D Fourier transform? The inverse transform?
How is the Fourier phase spectrum defined? The amplitude spectrum?
How are the "time width" and "frequency width" of a function related?
What is a Dirac Delta function? What is its Fourier transform?
When g and f are convolved in time, what happens to their Fourier spectra? What
happens to their amplitude spectra? Their phase spectra?
How is the Nyquist frequency related to the temporal sample rate?
What is the frequency sample rate of a time series of length T?
To avoid aliasing a signal of frequency Fmax, what sample rate must be used?
From a practical viewpoint, the signal frequencies in your data should be less than a
constant, c, times Fnyquist. What is a good value for c?
What sample rate should you use for 80 Hz signal? For 130 Hz? For 40 Hz?
What is an anti aliasing filter? When should it be used?
What are the two main purposes of a zero pad when doing a discrete Fourier transform
(DFT)?
What is circular convolution?
What is the relationship between the DFT and the fast Fourier transform (FFT)?
How is the Z transform related to the DFT?
A filter which has a pole close to the unit circle at some frequency does what to that
frequency? How about a zero close to the unit circle?
Where are the zeros of a minimum phase filter located in the z plane?
What is the definition of minimum phase in terms of causality and stability?
How is an inverse filter defined in the z plane?
What is meant by a stable inverse?
11-2 Study Guide

What is meant by the statement that "This seismic data is minimum phase"? (Note that
the statement is technically always false but it has a practical, definite meaning.)
In order to assert that seismic data is "minimum phase" at some stage of the processing
what conditions must be met?
How is cross correlation defined? What does it mean? What does the cross correlation
lag mean?
What is an autocorrelation? What is the expected autocorrelation of a random sequence?
What are the two central problems of spectral estimation?
What is the role of the "window" in spectral estimation?
What kind of spectrum is well modeled by the Burg spectrum?
What is the 2-D Fourier transform of a linear event with apparent velocity v? Draw a
sketch showing (x,t) space and (f-k) space for a range of different apparent velocities.
What is the most likely apparent velocity and where is it found in (f,k) space?
What is spatial aliasing? For a given apparent velocity and spatial sample rate, what is
the critical frequency at which spatial aliasing begins?
How can convolution be expressed as a matrix operation? Draw a diagram showing the
Toeplitz matrix symmetry.
Describe the six basic modes of seismic attenuation.

Geometric spreading corresponds to what conservation law?
Under geometric spreading, amplitude decreases proportional to what?
How is Q defined?
What is the formula for amplitude loss in a constant Q theory?
What is the formula for transmission losses in a layered medium?
What phenomenon is responsible for trapping large amounts of seismic energy in the
near surface?
What is true amplitude processing?
What do constant Q models predict about the signal bandwidth of seismic data?
The phase effects associated with Q attenuation are known as what?
What assumptions are required to derive these phase effects?
What is unique about a minimum phase wavelet?
What is the most important property of minimum phase wavelets from the viewpoint of
deconvolution theory?
What is meant by velocity dispersion?
What is the convolutional model? Write a mathematical expression for the model as it is
applied in deconvolution theory. Define each term and state the assumptions which
constrain each.
Can all types of multiples be included in the convolutional model? Why or why not?
The convolutional model expects the seismic trace to be stationary, What is meant by
this? Is it a reasonable expectation?
What are the essential steps of frequency domain spiking deconvolution?
How is the seismic wavelet estimated in frequency domain spiking deconvolution?
If a wavelet is known to be minimum phase, then its inverse can be found by solving a
set of matrix equations whose left hand side involves not the wavelet itself but a
statistical measure of it. What is this measure? Explain intuitively how this result is
equivalent to computing a wavelet's phase spectrum from its amplitude spectrum under
the minimum phase assumption.
Methods of Seismic Data Processing 11- 3

Wiener spiking deconvolution assumes that the autocorrelation of the seismogram is
similar to the autocorrelation of the wavelet. Justify this assumption by argument from
the convolutional model.
What is meant by the 'stab' factor or 'white noise' factor in deconvolution?
It is customary in seismic data processing to follow deconvolution by a bandpass filter. Is
this a sensible practice? Either justify it or refute it by argument from deconvolution
theory.
What is the relationship between prediction error filters and spiking deconvolution?
Explain why deconvolution keeps the prediction error and rejects the predictable part of a
seismic trace.
How is gapped predictive deconvolution implemented? What is a typical example of a
type of multiple which it is designed to attenuate? How can the prediction gap be chosen?
How are midpoint and offset defined in terms of source and receiver coordinate?
Illustrate with a diagram.
The near surface is generally assumed to cause effects which are a function of what
coordinates?
The subsurface effects are often assumed to be a function of what coordinates?
How is a static delay defined? What physical effect is often used to justify the
assumption of static delays in the near surface?
What is the definition of source static? Receiver static?
What is meant by the term "datum"? How does its choice effect the statics application?
What datum is most appropriate for pre stack processing?
What are surface consistent methods? Why are they useful? List some examples of
common surface consistent applications?
What is the definition of vertical traveltime?

How can instantaneous velocity be computed as a function of vertical traveltime?
What is a time-depth curve and what is it used for?
How is average velocity defined? In what sense is it an average?
What is the mean velocity and how is it defined?
Define Vrms in terms of instantaneous velocity and as is relates to the mean and average
velocities.
Which is always greater the average or the rms velocity?
What is interval velocity? Define at least two type of interval velocities.
What is the expression for the addition of two interval velocities?
What is the expression for the "Dix" interval velocity calculated from two closely
positioned rms velocity measurements?
Under what conditions can an interval velocity be said to approximate a local wave
propagation velocity?
How can imaginary interval velocities result from a Dix interval velocity calculation?
Derive the traveltime equation for normal moveout. Use a diagram to show the meaning
of all quantities. What is the shape of the traveltime curve?
What is the shape of the wavefront in the nmo experiment as it approaches the receivers?
How must the nmo equation be modified if the reflector is dipping?
What is the stacking velocity for a dipping reflector beneath a constant velocity
overburden?
11-4 Study Guide

How must the nmo equation for a dipping reflector be modified to take the azimuth of the
seismic line into account?
What is the definition of stacking velocity? Explain why stacking velocity is always a
function of offset.
Using a diagram, derive the geometric relation between wavelength components and
wavelength for a periodic planar wavefront.
What are wavenumbers?
What is apparent velocity? What are the mathematical limits (upper and lower) of
apparent velocity?
How does apparent velocity relate to wavelength components and wavenumbers?
What is Snell's law? Snell's law can be considered as the conservation of what quantity?
What is a v(z) medium?
When raytracing in a v(z) medium, what quantity is conserved and how is it defined?
Derive the distance traveled and traveltime integrals for raytracing in a v(z) medium.
How can the ray parameter be measured?
When velocity increases linearly with depth, what shape are the raypaths? The
wavefronts? (Exact equations not necessary.)
For the nmo experiment in a v(z) medium, explain how the result that stacking velocities
may be approximated by rms velocities arises. What assumptions are required? In
practice, when can we expect it to be roughly valid?
The Dix equation moveout can be interpreted as allowing the replacement of the real v(z)
medium by a constant one with properly chosen parameters. Explain this.
Explain why interpolation of trace sample values is needed in nmo removal. What is
moveout stretch? Why does it arise?
What are residual statics? How are they computed? What is their purpose? What
processes should be run on seismic data prior to attempting a residual statics solution?
What is velocity analysis? How is it performed? What processes should be run on
seismic data prior to attempting a velocity analysis?
Do statics and moveout removal commute? That is, do you get the same result regardless
of the order of the processes? If not what is the preferred order?
In the extension of nmo and dip to v(z), what quantity must be measured in addition to
stacking velocity in order to allow the computation of apparent dip and the "dip
correction" of stacking velocities?
What can be said about the staking velocities of multiples? Where will they be found on
a stacking velocity analysis chart?
After stacking, the power of random noise can be expected to be reduced by what factor?
Considered as an "f-k" process, stacking can be said to pass what portion of the offset
wavenumber spectrum?
Are "f-k" filters applied to cmp gathers likely to improve a stack? What if they are
applied to shot or receiver gathers?
What is a zero offset section? How does is serve as a model for a stack?
What is the relation between traveltime gradient measured on a stack and the normal
incidence ray parameter?
What information is needed for the raytrace migration of a normal incidence
seismogram?

What are the algorithmic steps in normal incidence raytrace migration?
Time migration processes are biased towards what class of rays? How are these rays
handled? When is time migration a valid process? When is depth migration a valid
process?
What is the migrator's formula? How can it be used?
Explain post stack migration by replacement of each point with a wavefront. How are the
wavefronts defined?
What is meant by "migration dip"?
Should data from an area where all geologic dips are less than 5 degrees be migrated?
Why? Should a steep dip algorithm or a low dip algorithm be used?
What is Huygen’s principle?
What is the traveltime curve of a point diffractor for post stack migration? For pre stack
migration?
What is a diffraction chart? Explain how diffraction curves can be used to construct a zos
image from a geologic model?
What is the exploding reflector model? Why is it useful?
In the exploding reflector model, what is the mathematical expression for the migrated
section? For the zos image?
Using the exploding reflector model to explain wavefield extrapolation. What is the
mathematical expression for an extrapolated section? What is the relationship between
any extrapolated section and the migrated depth section?
What is the dispersion relation? Use the dispersion relation to derive the mapping which
defines f-k migration. Use a diagram to illustrate the mapping of the f-kx spectrum to the
kx-kz spectrum. What is the meaning of the evanescent boundary? What determines the
maximum kx wavenumber after migration? What determines it before migration? Draw a
flow chart for f-k migration.
What is f-k wavefield extrapolation? Derive the expression for the f-k phase shift
required to shift the datum by ∆z. Explain how recursive f-k phase shifting can be used to
create a v(z) migration algorithm. Draw a flow chart.
What is the geometric shape of the wavefield extrapolation operator? What are the two
distinct components of the operator? How can is be applied in the (x,t) domain?
What is the major distinction between time and depth migration? Can time migration
produce a depth section and depth migration a time section? Explain.
What is Kirchoff migration? What is the shape of the Kirchoff migration operator
(constant velocity) when applied post stack? Pre stack?
Describe a general method to determine the shape of the Kirchoff migration operator.
Your method must be valid for any (x,y,z) location, any velocity, and pre or post stack.
What is pre stack time migration? When is it a valid process? Should it be inferior, the
same, or superior to stacking and post stack time migration?
What is DMO? Describe a flow using DMO that should give similar results to pre stack
time migration. What are the strengths and weaknesses of the DMO approach? Under
what circumstance is DMO->stack->migration exactly the same as pre stack migration?
What kind of velocities should be input to the NMO removal step in a flow involving
DMO? How can these velocities be obtained?
Describe, without equations, the essential steps in CSP migration? How does CSP
analysis effect velocity resolution?
11-6 Study Guide

What is the central (most difficult and most important) problem in the application of
depth migration to the thrust belt? Describe at least one approach to solving this problem.
What is wavelet processing? What are the essential steps in wavelet processing? When
should it be done in a processing flow? When is it necessary? What are two common
methods of wavelet estimation?
What is impedance inversion? When should it be run in a processing sequence? How can
the convolutional model (from deconvolution theory) be used to justify impedance
inversion? What is the major computation involved in impedance inversion? Describe at
least two common problems with impedance inversions that are difficult to solve.
What is the expected behavior of the amplitude spectrum of the radiated waves from a
dynamite sources as a function of charge size?
Explain how Q effects necessarily lead to a time variant (i.e. non-stationary) signal
bandwidth. What is the relationship between spectral width and wavelet width?
What is a "corner frequency"? When can it be observed? What does it mean?
For a constant velocity earth, what are the equations which express the limits of
observable scattering angle due to aperture, record length, and spatial aliasing? Make a
sketch of their basic form.
Staring from the theory of f-k migration, derive an expression for the maximum kx after
migration as a function of frequency, velocity, and scattering angle. Explain the
relevance of this to the problem of resolving small horizontal features? What steps can be
taken in recording or processing to increase horizontal resolution?

Exam sampler. There will be between 30-35 multiple choice questions and 4-
8 short answer questions.
PLEASE ANSWER ALL OF THE FOLLOWING QUESTIONS.
There are a total of 100 marks (points) for the examination. You have about 100
minutes for the exam.
Write all work directly on the examination sheets. If you need more room, you
may attach a work sheet with your name and the question number on it. PLEASE
HAND IN THE EXAMINATION SHEET AND ALL WORK SHEETS WITH
YOUR ANSWERS.
Multiple Choice Questions (2 points each)
INSTRUCTIONS: For each question, there are two best (most correct) responses.
Choosing both correct responses and no incorrect ones is worth two points. One
correct and one incorrect is worth one point, and any other result (including more
than two selections) is worth zero. Write your answers in the space provided below
each question.
1) The 1-D synthetic seismogram, as discussed in lecture:

a) can be made to contain all possible multiples.
b) is useful for modeling AVO and converted waves.
c) applies the source waveform of a band limited source by correlation.
d) is an excellent model of a trace on a stacked and migrated section, provided
that all possible multiples are included in the solution.
e) is based on ray theory and normal incidence reflection and transmission
coefficients.
answer _________
13) Average velocity:

a) characterizes the shape of diffraction curves on a cmp stack.
b) is depth divided by the vertical traveltime to that depth.
c) is a mathematical average over depth.
d) can be measured on an f-k plot.
e) is a mathematical average over traveltime.
answer _________
20) Post-stack F-K migration:

a) easily handles variable velocity.
b) is useful to explain the transformation of the data spectrum under migration.
c) shows that a constant frequency, f, maps to a hyperbola in (kx,kz).
d) is a steep dip algorithm.
11-8 Study Guide

e) works by applying a phase shift to the f-k transform.
answer _________
31) Minimum phase:

a) is the state of all raw seismic data.
b) is the desired state of final migrated sections.
c) means that the phase is the Hilbert transform of the log of the amplitude
spectrum.
d) is the smallest possible phase.
e) is only possible for time series which have inverses.
answer _________

Short answer problems (10 points each)
Please work directly on the examination sheets. Show all work.
1) Deconvolution
Suppose a dynamite dataset has mistakenly had a zero phase bandpass filter
applied before deconvolution. Assuming that the original data was noise free,
with a minimum phase wavelet and a white reflectivity:
a) Write a time domain convolutional expression for a single trace after the bandpass filter was
applied but before deconvolution. Then rewrite this expression in the frequency domain.
b) By working through the steps in frequency domain deconvolution, derive a

frequency domain expression for the embedded wavelet remaining after standard
minimum phase deconvolution. Show each step in the frequency domain
deconvolution process and indicate any smoothing but you may assume that stab
factors (white noise) are not needed.
c) Derive an expression for a correction filter which can be applied after

deconvolution to give the desired result from the deconvolution process.
11-10 Study Guide

Methods of Seismic Data Processing 11-11

Methods of Seismic Data Processing PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Methods of Seismic Data Processing PDF

Uploaded by

Copyright:

Available Formats

Methods of Seismic Data Processing

The Department of Geology and Geophysics

Section Title Page Number

Chapter 2: Signal Processing Concepts 76 pages

Chapter 3: Amplitude Effects 32 pages

Chapter 4: The Convolutional Model and Deconvolution 61 pages

Chapter 5: Surface Consistent Methods 29 pages

Chapter 6: Velocity Definitions and Simple Raytracing 26 pages

Chapter 7: Normal Moveout and Stack 38 pages

Chapter 9: The Third Dimension 32 pages

Chapter 10: Seismic Resolution Limits 35 pages

Chapter 11: Study Guide 9 pages

Methods of Seismic Data Processing 1 -1

1-2 Synthetic Seismograms

Methods of Seismic Data Processing 1 -3

Methods of Seismic Data Processing 1 -5

deconvolution methods imaging methods

the convolutional model one-way scalar waves

primaries, multiples, etc

elastic wave theory

anelastic wave theory

physics of continuous media

1-6 Synthetic Seismograms

σ ii = λΔ + 2μεii, i=x,y,z Δ = εxx +εyy+εzz

The description of elastic wave in such a medium, requires the

Methods of Seismic Data Processing 1 -7

It is well established in theory1,2,3 that a homogeneous, isotropic

We may choose to regard α and β as fundamental constants

1: Waters, Reflection Seismology, 1987

Well logging is a technology designed to make geophysical

SON ... P-wave interval transit time

The interval transit time logs are usually provided in units of

Methods of Seismic Data Processing 1 -9

1500 coal_1 1500 coal_1

1550 glauc_ch_top 1550 glauc_ch_top

Why do these logs appear to have a negative correlation?

1-10 Synthetic Seismograms

1 Ga rd ner , G .H. F ., G a rd ne r , L . W . , an d G re gor y, A . R. , 19 74, Fo rm a tio n

Gardner et al. sought and found a relationship of the form:

1-12 Synthetic Seismograms

1650 1650 1650

1600 1600 1600

1550 1550 1550

1500 1500 1500

1450 1450 1450

m=.46 m=.30 m=.25

Methods of Seismic Data Processing 1 -13

• propagation of primaries and multiples

1-14 Synthetic Seismograms

T he s im plest m a thematic al w a ve equat io n is th e scala r w a ve

Where ψ represents the propagating wave. We now show that

Substitution of the second partials of f into (1) results in an

Methods of Seismic Data Processing 1 -17

As an example of a waveform, consider the Ricker wavelet defined

Shown for fdom=30Hz

w t+z/v = Wavef o rm trave li ng z=-vt

Similarly, cos(ω (t-z/v)), cos(k(z - vt)), and cos(ω t-kz) all

continuity of displacement: f + g = h (1)

continuity of normal pressure: ??? (2)

To develop a form for the second equation, we use Hookes

stress = (applied force)/area

Methods of Seismic Data Processing 1 -19

Consider an infinitesimal elastic element