TIME and FREQUENCY

Time

To s(t )  Ao cos( ot   o )
Ao
2
o 
To
1
fo 
To
o

o f = frequency [Hz]
 = angular frequency [radian per second]

Note s(t )  Ao cos( ot   o )  Ao cos  o cos  ot  Ao sin  o sin  ot

Frequency

A sine wave is fully described by three quantities:

Ao , fo ,  o

Amplitude Phase

Ao

o

fo fo

1
Amplitude A and phase  define a phase vector (or phasor) Ae ,
j

i.e., a complex number a  jb where

a  jb  Aej  Acos  jAsin

Euler’s Formula

a  A cos  b  A sin 
b
A  a2  b2   arctg  
a
Note a and b are the amplitudes of cosine and sine components

The spectrum of a signal s t  can be described by a complex function

S f  :

s t  Fourier Transform S f 

PERIODIC SIGNALS


s (t )  ao   n an cos n o t  bn sin n o t
1

2
s (t ) is periodic with period To 
o

A periodic signal is a linear combination of cosine and sine waves

with frequencies that are integer multiples of the fundamental frequency
(Fourier series)

2
 Since:
To To
2
2 1
 s(t )dt  a T o o
ao 
To  s(t )dt
To
T 
 o 2
2
To To
2
2 2
 s(t ) cos  tdt
a1To a1 
 s(t ) cos  o tdt 
2 To To
o
T
 o 
2 2

and so on
To To
2 2
2 2
an 
To  s(t ) cos n tdt
T
o bn  
To  s(t ) sin n tdt
To
o
 o 
2 2

 bn 
Defining: An  an2  bn2  n  arctg  
 an 
we can also write:

s (t )  Ao   n An cos[n o t   n ]
1

Amplitude Phase
1
2
Ao
A3
A2
A1
3

f0 2 f0 3 f0 f0 2 f0 3 f0

3
 NON-PERIODIC SIGNALS

Let us start with the Fourier series of a periodic signal and let us
extend To to 
To
2
1
s (t ) 
To  s(t )dt 
To

2

 To T0


2 0  2 2

n cos n 0t  s (t ) cos n 0tdt  sin n 0t  s (t ) sin n 0tdt 
1 2  T T 
 0  0
 2 2 

If To  0 d
n 0 
 


Thus, assuming that  s(t )dt  


(Note: not a necessary condition)

1  
  

2 0   
s (t )  cos  t  2 s (t ) cos  tdt  sin  t  2 s (t ) sin  tdt d
  

4
that is

 a( ) cos t  b( ) sin  td

1
s (t ) 
2 0

where 
a( )  2  s (t ) cos tdt



b( )  2  s (t ) sin tdt


i.e., a generic signal s(t) is the sum of an infinite number of sine and
cosine waves

Defining:
A( )  a 2 ( )  b 2 ( )

 b( ) 
 ( )  arctg  
 a ( ) 

 A( ) cost   ( )d

1
s (t ) 
2 0

and since d  2df


s (t )   A( f ) cos2ft   ( f )df
0

5
 Amplitude  Phase
A


f f


S ( f )  A( f )e j ( f ) Unilateral spectrum of s(t)

In exponential form:


1  A ( ) j  t   (  )  A ( )  j  t   (  )  
s (t ) 
2 0  2 e 
2
e d  


1  A ( ) j ( ) j t A ( )  j ( )  j t 

2  
0
2
e e 
2
e e d  


1  A ( ) j (  ) j t 

A (  )  j (   ) j t
0

2  0 2 
  e e d  e e d 

2 

6
i.e.,  
1
 S ( ) e
j t
s (t )  d
2 

A( ) j ( )  0
e
2

where
S ( ) 
A( )  j (  )  0
e
2

S( f ) Bilateral spectrum of s(t)
Note  1
S( f )  S( f )
2
 
S ( f )  S * ( f ) it occurs when s(t) is real

Moving from  to f :
 

 S ( f )e
j 2ft
s (t )  df Inverse Fourier Transform



S ( )  a ( )  jb( ) 
1 1
with S ( ) 
2 2
 

 s ( t ) cos  t  j sin  t dt   s (t ) e

 j t
 dt
 
thus
 

 s(t )e
 j 2ft
S( f )  dt (Direct) Fourier Transform


Note A dual sign convention can be also used for the exponential terms, i.e., positive
argument for the Direct FT and negative argument for the Inverse FT.

7
 PROPERTIES
Linearity

If s1(t) S1(f) and s2(t) S2(f) then

as1 (t )  bs2 (t )  aS1 ( f )  bS 2 ( f )

Delay
If s (t )  S ( f ) then s (t  t0 )  S ( f )e  j 2ft0
Indeed:
  

 s(t  t )e  s( )e  s( )e

 jt  j (  t 0 )  jt 0  j
0 dt  d  e d
  

Viceversa: S ( f  f 0 )  s (t )e j 2f 0t

Symmetries
 
If s (t ) is real  S ( f )  S * ( f )
Indeed:
   
S ( )   s (t )e  jt
dt   s (t ) cos tdt  j  s (t ) sin tdt
 
    
  
 

S r ( ) S i ( )

Even symmetry because

Odd symmetry because
cos(t )  cos(t )
sin(t )   sin(t )

A( )  A(  )  even
 ( )   (  )  odd

8
 
If s(t) is real and even  S ( f ) is real and even

Indeed, the imaginary part is null because s (t ) sin t

is an odd function


If s(t) is real and odd  S ( f ) is imaginary and odd

Indeed, the real part is null because s (t ) cos t

is an odd function

Note These symmetries also hold in dual direction: from f to t

Value in the origin

 
S ( 0)   s(t )dt


 
and viceversa s ( 0)   S ( f )df


9
 EXAMPLES

Dirac delta function

 

 Aimp(t )e
 j 2ft
s (t )  Aimp (t )  S ( f )  dt  A


s S

A A

t f

Reciprocally, a constant function has a Fourier transform that

is a Dirac delta function in f=0.

Sine wave

A0 j 0 j ot A0  j 0  j 0t
s (t )  A0 cos( 0t   0 )  e e  e e
2 2


A A 
S ( )    0 e j 0 e j ot  0 e  j 0 e  j 0t e  jt dt 
  
2 2
 
A0 j 0 A
 e  e  j (  0 ) t dt  0 e  j 0  e  j (  0 ) t dt 
2 
2 

A0 j 0 A
 e imp (   0 )  0 e  j 0 imp (   0 )
2 2

10
 A

If  0  0  s (t )  A0 cos  0t

(real and even)  0 0


  
 0  0
A

If  0    s (t )  A0 sin  0t
2 
(real and odd)
 0 0


 0 0 

Rectangular pulse
T0 T
s (t )  A0  t  0
A0 2 2

T0 T
s (t )  0 t e t 0
2 2

T0 T0 (real and even function)


2 2 T0

  2

  dt 
 jt  j t
S ( )  s (t ) e dt  A0 e
 T0

T 2
 0
2
A0 T0
 A0  T0
cos tdt 

2 sin 
2

2

11
  sin fT0
S ( f )  A0T0
fT0

A0T0

1 2
T0 T0

Examples A(f) ( f )
s(t )

1)

A(f) ( f )
s(t )

2)

A(f) ( f )
s(t )

3)

12
 LINEAR SYSTEMS
(TIME-INVARIANT)
X 1 (t ) Y1 (t )

X 2 (t ) Y2 (t )

aX 1 (t )  bX 2 (t ) aY1 (t )  bY2 (t )
System output can be fully described in the frequency domain
through the Transfer function

aX 1 (t   )  bX 1 (t ) aY1 (t   )  bY1 (t )
System output can be fully described in the time domain
through the Impulse response

IMPULSE RESPONSE

s(t )
t

 
s (t )   k s (k )r (t  k )

where
r (t )
1

t
0 

13
 
s (t )   k s (k ) p (t  k )


where p(t)
1 unit area


0  t

If  approaches zero

  d 

s (k )  s ( ) s (t )   s( )imp(t   )d
 p (t )  imp (t ) 


If
imp(t) h(t)= impulse response

then
s(t) g(t)

with 
g (t )   s ( )h(t   )d
 
 
convolution integral

h(t) is an exhaustive description of the linear system

14
Example
h(t ) s(t )
h(t )

g (t )  s (t )  h(t )

convolution symbol

 
Note
 s( )h(t   )d   h( )s(t   )d
 

TRANSFER FUNCTION

s (t ) g (t )
 
S( f ) G( f )

   

 g (t )e dt   e  s( )h(t   )ddt 

 j 2ft  j 2ft
G( f ) 
  

 

  s( )e
 j 2f
 h(t   )e  j 2f (t  ) ddt 
  

15
  

 s( )e  h(t   )e
 j 2f  j 2f ( t  )
 d (t   )d 
 

H(f)= transfer function

 
G( f )  S ( f ) H ( f )

H(f) is an exhaustive description of the linear system

EXAMPLES
S( f ) s (t ) A
A

f0

H( f )
0.5

f0

G( f ) g (t ) A
2

A
2

f0

The output of a linear system when the input is a sine wave

at frequency f 0 still is a sine wave at frequency f 0

16
 signal
S( f ) s (t )
noise

H( f )
Low-pass filter

f0

G( f ) g (t )

Note

G ( f ) is a product of complex functions:

G( f )  S ( f ) H ( f )

GA ( f )  S A ( f ) H A ( f ) G ( f )  S ( f )  H  ( f )
product of amplitude spectra Sum of phase spectra

17
SUMMARIZING

h(t )
s (t ) g (t )
H( f )

g (t )  s(t )  h(t )
The signal is considered as a temporal pulse train.
The output is the sum of the responses generated by each pulse.
 

 S ( f ) H ( f )e
j 2ft
g (t )  df

The signal is considered as a sum of sine waves at different frequencies.
The output is the sum of the sine waves generated by each input frequency.

Note Calculating g(t) in frequency domain can be faster than in time domain!

EXAMPLES OF TRANSFER FUNCTIONS


Delay
s (t )  S ( f )

s (t   )  S ( f )e  j 2f

H ( f )  e  j 2f

 2f

Note There is neither amplitude or phase distortion

18
Phase shift

e  j 0 f  0
H( f ) 
e j 0 f  0

0
cos(1t ) cos(1t   0 )  cos 1 (t  )

1
1

cos( 2t ) 0
cos( 2t   0 )  cos  2 (t  )

 2
2

cos(1t )  cos( 2t ) cos 1 (t   1 )  cos  2 (t   2 )

phase distortion

19
Derivative 
s (t )  S ( f )
s ' (t )  ?
 

 s (t )e
'  j 2ft

dt  s (t )e  j 2ft  
   j 2f  s(t )e
 j 2ft
dt
 

If s(t) is null at   and   s ' (t )  j 2f S ( f )
This is also true for a sine wave and thus for any periodic signal.

j
H ( f )  j 2f  2fe 2


2f 2 Note Both amplitude
and phase distortions are

 generated by a derivative
2

Integration 
s (t )  S ( f )
t 
g (t )   s ( )d  G ( f )  ?
0
 
s (t )  g ' (t )  S ( f )  j 2f G ( f )

 S( f )
G( f ) 
j 2f
1 j 1  j 2
H( f )    e
j 2f 2f 2f

Note Both amplitude and phase

distortions are generated by integration

20
 Mechanical seismograph
Case
x0

Spring (k) Damper

Paper roll
x1
Mass (m)

Pen

h=0.05
h = damping factor h=0.5
h=0.7
h=1
h=0.05
1 k h=0.5
f0  h=0.7
2 m h=1

Resonance
frequency

FREQUENCY RESPONSE

ma  F  mx1  mg  k ( x1  x0 )   ( x1  x0 )

Analysis in the frequency domain (using phasors)

  0  mg  k ( x1  x0 )
  0  mx1   k ( x1  x0 )   ( x1  x0 )
mx1  x1  kx1  kx0  x0

x0  jx0
k  j
since

x1  jx1 x1  x0
k  m  j
2

x1   2 x1

21
 The signal on the paper roll is: xs  x1  x0

 k  j 
xs    1 x0
 k  m  2
 j 
m
 m 2 2
H ( )   k
k  m 2  j 1  m  2  j 
k k
Defining:

2
k
0  resonant angular frequency
 02
m
H ( ) 
d
 2 
2m 0
damping factor
1  2  j 2d
0 0

Geophone Leaf spring

Damping
resistor

coil coil

N S N
magnet

22
 Amplitude frequency response (www.geospace.it)

Geophone groups

Surface

Reflector

The use of a geophone group gives a directional response to the group,

designed to enhance near-vertical upward travelling reflected waves and to
minimize any horizontally travelling coherent noise (ground roll)

23
 single group
x

Vertical direction:    

0 0


t


Horizontal direction:

2 t

3t x
t 
v

Impulse response

Vertical direction:
*
4

t[s]


1
Horizontal direction:
*
t [s]
t
Transfer function
H( f )
4
Vertical direction:
f [Hz]

Horizontal direction:

t 3 t t 3 t
 jw  jw jw jw
H ( w)  e 2
e 2
e 2
e 2

t 3
 2 cos w  2 cos w t 1 f [Hz]
1
2 2 4t
x t
Other directions: as for horizontal, with 0  t 
v

24
 Group of N geophones

Horizontal direction:

t 3 N 1
H ( w)  2 cos w  2 cos w t  .....  2 cos w t
2 2 2

t t  t 3 N 1 
sin w  H ( w)  2 sin w cos w  cos w t  .....  cos w t 
2 2  2 2 2 

t t t
sin w  H ( w)  sin wt  sin 2 wt  sin wt  .....  sin Nw  sin( N  2) w
2 2 2

1 sin( Nft )
H ( f ) norm 
N sin(ft )

Group of N geophones
x
)sin( Nf
1 sin( Nft ) 1 V
H ( f ) norm  
N sin(ft ) N sin(f x )
V
Example

Assumptions: V1  300m / s ; 10 < seismic band < 140 Hz

1
 10 NT  0.1
NT
1 1
T
 150 T 
150

hence N  15
x  V1T  2m

1 1
NT T

25
 Ghost negative reflection coefficient (-1)

Hydrophones t0
P

airgun

primary

ghost
sea floor

travel path difference  2P

2P
primary-ghost delay 
V

Ghost
Transfer function: H( f ) 1 e j 2f
Actually, ghost occurs both at source and receiver position, so the comprehensive transfer function is:

H( f )  (1 e j 2f )2
2
H( f )  1 e j 2f  1 cos2f  j sin2f  2  2cos2f
2

H( f )

0 fo 2fo 3fo f
1
where f o 


26
 Ringing
-1 -1 -1
t0


primary

z
1st multiple

2nd multiple

Sea floor
R R R

2z
 where v  1500 m/s
V

Ringing
sea floor primary

layer primary

sea floor 1st multiple

layer 1st multiple

time

sea floor 2nd multiple

layer 2nd multiple

27
Ringing
-1 -1 -1 Air
Model for sea floor multiples

z Water

R R R Solid
Transfer function:

 

H(w)  1  R  e jw  R2  e j2w  ....    R  e jw
n

n0
2z
where  with v  1500 m/s
v

Ringing
1
since R < 1 the series converges H (w) 
1  R  e jw
Actually, ringing also occurs when the reflected wave is back to the surface.
Thus, the comprehensive transfer function which includes both descending and
ascending multiples is:
1
H (w) 
1  R  e   jw 2
Norm.

for example,
with R =.5 and z=30m

28
 NON-LINEAR SYSTEMS

x1 (t ) y1 (t )

x2 (t ) y2 (t )

ax1 (t )  bx2 (t ) y3 (t )  ay1 (t )  by2 (t )

Example

x1 (t ) ( )2 2
x1 (t )

2
x2 (t ) x2 (t )
”
a 2 x1 (t )  b 2 x2 (t )  2abx1 (t ) x2 (t )
2 2
ax1 (t )  bx2 (t )
”

Non-linear systems generate armonic distortion.

A2 A2
A cos  ot A2 cos 2  o t   cos 2 ot
( )2 2 2

A2
A 2

fo 2 fo
Non-linear systems cannot be described through impulse response
or transfer function.

29
 A/D – Analog to Digital Conversion

Advantages of digital signals:

* digital signals can be processed
* digital signals are more robust to noise than analog signals

Can we convert an analog signal into a digital signal without

losing any information ?

Tc
clock

Tc  sampling interval

The question can be rephrased as:

How short should we make Tc to preserve all the information ?

30
 SPECTRUM of the SAMPLED SIGNAL
s(t ) s (t ) periodic A( f ) 
S( f ) impulsive

T0 T0 f0 f0

s (t ) A( f ) 
s (t ) impulsive S( f ) periodic

fc
Tc

The spectrum of a digital signal is periodic

The operation of sampling, in time domain, consists of a product

s(t )

Tc

31
 pulse train 
frequency comb
s(t ) S( f )
1 1
Tc

1
Tc 2Tc fc 2 fc con fc 
Tc

Indeed s(t ) is periodic and its Fourier series coefficients are:

Tc Tc
2 2
1 1 1 1
a0   imp(t )dt  T a1   imp(t ) cos  tdt  T c
…. an  1
Tc Tc c Tc Tc c Tc
 
2 2
1 per t  0
Tc
2
1
b1  
Tc T
 imp(t ) sin  tdt  0
c
…. bn  0
 c
2
. 0 per t  0

The operation of sampling, in frequency domain, consists of a convolution


S( f )

f max * =
fc 2 fc

Sampling an analog
signal makes its
spectrum periodic
f max

 fc fc 2 fc

32
Sampling (or Nyquist) theorem

In order to preserve all the information, a band-limited analog signal

must be sampled at a rate:

f c  2 f max

Assuming a sampling frequency f c , the Nyquist frequency is the

highest frequency that the analog signal can contain in order to be
fully reconstructed from the sampled version.

fc
f Nyquist 
2

ALIASING

sampling

f max

f Nyq f
c

reconstruction

f Nyq

33
Example

Aliasing of a 200Hz sine wave sampled at a rate of 250Hz.

A 50Hz sine wave shares the same samples.
f Nyq f c

ANTI-ALIAS FILTER

anti-alias filter

f max f Nyq
sampling

reconstruction

f Nyq

f Nyq f
c

34
 DISCRETE FOURIER TRANSFORM
N is even

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

period
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

DISCRETE FOURIER TRANSFORM

N is even

Nyquist frequency
sampling
frequency
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

DC component

zero
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

zero

35
 DISCRETE FOURIER TRANSFORM
N is odd

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

period
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

DISCRETE FOURIER TRANSFORM

N is odd

Nyquist frequency sampling

frequency
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

DC component
Nyquist frequency
zero
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

36
 AUTOCORRELATION


R (t )   s( ) s(  t )d


Normalized autocorrelation:


 s( )s(  t )d

RN (t )  


s ( )d
2



Properties

At t = 0
R (0)  signal energy
R N ( 0)  1
Range
 1  RN (t )  1

Simmetry

R (t )  R (t )

Thus, the following is also a valid definition of autocorrelation:


R (t )   s( )s(  t )d


37
 Spectrum
R (t )  s (t )  s (t )

s (t )  S ( f )
  j 2ft  j 2f   
s (t )   s(t )e

dt   s( )e

d  S (  f )  S ( f )
2
  
S A ( f )S A ( f )  S ( f )
R (t )  S ( f ) S ( f )
S ( f )  S ( f )  0

Note. Signals that share the amplitude spectrum also share the
autocorrelation function regardless their phase spectra.

Matched filter
If the output of a linear system excited by s(t) consists of the convolution of
s(t) with s(-t), the system is said “matched filter” of s(t).
 
s (t ) S (f) R(t )

Example: white noise

38
Example: sine wave

39
Example: sine wave + noise

noise

sine wave
+
noise

autocorrelation

40
 CROSS-CORRELATION

C (t ) 

 s( ) g (  t )d  s(t )  g (t )

Also defined as: C (t )   s( ) g (  t )d (but the function is flipped)


Normalized cross-correlation:


 s( ) g (  t )d
C N (t )  
1/ 2
 2 

 s ( ) d    g 2 ( )d 
   

Range
 1  C N (t )  1

Spectrum

s (t )  g (t )  s (t )  g (t )


s (t )  S ( f )

 
g (t )  G ( f )

   S A ( f )  GA ( f )
s (t )  g (t )  S ( f ) G ( f )
S ( f )  G ( f )

41
Examples: impulsive wavelets + noise

Zero-phase wavelet

Examples: sweep overlap

Zero-phase wavelet

42
Zero-phase wavelets

43
 Vibroseis
non-impulsive source
vibration mode “sweep” (e.g., 7-70 Hz)
sweep length 7-20 sec
record time two-way traveltime to the
target + sweep length

filter in matched filter

t2
controlling signal v(t )  A(t ) sin( at  b )
2
instantaneous frequency wi (t )  a  bt

In order to increase the energy

input into the ground, a number
of vibrators can be used
simultaneously in a group:

0.5
sweep

-0.5

-1
0 100 200 300 400 500 600 700 800 900 1000
ms
10
Ampl. Spectr. [dB]

-10

-20

-30
0 50 100 150 200 250
Hz
400
Klauder wavelet

200

-200
-500 -400 -300 -200 -100 0 100 200 300 400 500
ms

“sweep” signal, Amplitude Spectrum and “sweep” autocorrelation (Klauder wavelet)

44
 Klauder wavelet 10-20Hz
400

200

-200

-400
-500 -400 -300 -200 -100 0 100 200 300 400 500
Klauder wavelet 10-40Hz
400

200

-200
-500 -400 -300 -200 -100 0 100 200 300 400 500
Klauder wavelet 10-80Hz
300

200

100

-100
-500 -400 -300 -200 -100 0 100 200 300 400 500
ms

Sweep autocorrelation according to different bandwidths

Seismic record before cross-correlation Seismic record after cross-correlation

45
 2D FOURIER TRANSFORM

F (k , )   f ( x , t ) e  j ( kx   t ) dxdt
 

e  f ( x , t )e
 jkx  j t
 dtdx
 
 

e 
 j t
 f ( x , t ) e  jkx dxdt
 

A 2D transform consists of 2 sequential 1D transforms

2D INVERSE TRANSFORM

1
 F ( k ,  ) e
j ( kx   t )
f ( x, t )  dkd 
4 2

TIME DOMAIN FREQUENCY DOMAIN

46
 TIME DOMAIN FREQUENCY DOMAIN

ENVELOPE EXTRACTION
The envelope of a signal is the outline of its amplitude variations.

The envelope is extracted as the module of the complex signal, also

called analytic signal.
analytic signal
env(t )  sa (t ) where sa (t )  s(t )  jsˆ(t )
Hilbert transform

47
Hilbert transform
The Hilbert transform is the convolution of s(t) with the Hilbert
transformer 1/t.

s( )

1 1
sˆ(t )  s(t ) *  d
t  t 

In the frequency domain:

Sˆ ( w)  S ( w) H ( w) where H ( w)   j sgn(w)
  
j
 j  e 2  w  0
H ( w)   j sgn(w)   

 j 
j e w0 
2

Hilbert transform
The Hilbert transformer is a linear operator.
Impulse response:
1
h (t ) 
t
Hyperbolic function

Transfer function:

H ( w)   j sgn(w)
Phase shift

48
Analytic signal

In the time domain:

sa (t )  s(t )  jsˆ(t )

In the frequency domain:

S a ( f )  S ( f )  jSˆ ( f )

f 0 Sa ( f )  S ( f )  j * (  j ) * S ( f )  2S ( f )

f 0 Sa ( f )  S ( f )  j * j * S ( f )  0

Analytic signal

S( f )

Sa ( f )

49
Examples
Examples of envelope extraction

50