Fourier Transform Applications

Fourier Transform:
Applications
Seismograms
Eigenmodes of the Earth
Time derivatives of
seismograms
The pseudo-spectral
method for acoustic wave
propagation
1
Fourier: Applications Modern Seismology Data processing and inversion
Fourier: Space and Time
Space
Space Time
Time
xx space
spacevariable
variable tt Time
Timevariable
variable
LL spatial
spatialwavelength
wavelength TT period
period
k=2/
k=2/ spatial
spatialwavenumber
wavenumber ff frequency
frequency
F(k)
F(k) wavenumber
wavenumberspectrum
spectrum =2f angular
=2f angularfrequency
frequency
Fourier
Fourierintegrals
integrals
With
Withthe thecomplex
complexrepresentation
representationof
ofsinusoidal
sinusoidalfunctions
functionseeikx(or
ikx
(or
(e
(eiwt))the
iwt
theFourier
Fouriertransformation
transformationcan
canbe
bewritten
writtenas:
as:

1

ikx
f ( x) = F ( k ) e dx
2

1
F (k ) =
2

f ( x)e ikx dx
2
The Fourier Transform
discrete vs. continuous
Whatever continuous
Whateverwe wedo
doon
onthe
the
computer
computerwith
withdata
datawill
will
be
bebased
basedononthe
thediscrete
discrete 1

ikx
f ( x) = F ( k ) e dx
Fourier
Fouriertransform
transform 2

1
F (k ) =
2 f ( x)e ikx dx
discrete

N 1
1
Fk =
N

j =0
f j e 2 ikj / N , k = 0,1,..., N 1
N 1
f k = F j e 2 ikj / N , k = 0 ,1,..., N 1
j =0
3
The Fast Fourier Transform
... the latter approach became interesting with the introduction of the
Fast Fourier Transform (FFT). Whats so fast about it ?
The FFT originates from a paper by Cooley and Tukey (1965, Math.
Comp. vol 19 297-301) which revolutionised all fields where Fourier
transforms where essential to progress.
The discrete Fourier Transform can be written as

N 1
1
u k =
N
j
u
j =0
e 2 ikj / N
, k = 0 ,1,..., N 1
N 1
uk = u e
j =0
j
2 ikj / N
, k = 0 ,1,..., N 1
4
... this can be written as matrix-vector products ...

for example the inverse transform yields ...
1 1 1 1 K 1 u 0 u 0
1
2 3 K N 1 u1 u1
1 2 4 6 K 2 N 2 u 2 u 2
=
M M M M M
M M M M M
( N 1) 2
1 L
N 1
L L u N 1 u N 1
.. where ...
= e 2i / N
5
... the FAST bit is recognising that the full matrix - vector multiplication
can be written as a few sparse matrix - vector multiplications
(for details see for example Bracewell, the Fourier Transform and its
applications, MacGraw-Hill) with the effect that:
Number of multiplications
full matrix FFT
N2 2Nlog2N
this has enormous implications for large scale problems.

Note: the factorisation becomes particularly simple and effective
when N is a highly composite number (power of 2).
6
Number of multiplications
Problem full matrix FFT Ratio full/FFT
1D (nx=512) 2.6x105 9.2x103 28.4

1D (nx=2096) 94.98
1D (nx=8384) 312.6
.. the right column can be regarded as the speedup of an algorithm

when the FFT is used instead of the full system.
7
Spectral synthesis
The
Thered
redtrace
traceisisthe
thesum
sumof
ofall
allblue
bluetraces!
traces!
8
Phase and amplitude spectrum
The spectrum consists of two real-valued functions of angular
frequency, the amplitude spectrum mod (F()) and the phase
spectrum ()
i ( )
F ( ) = F ( ) e
In many cases the amplitude spectrum is the most important

part to be considered. However there are cases where the
phase spectrum plays an important role (-> resonance,
seismometer)
9
remember
z* = a ib = r (cos i sin )
i
= r cos ri sin( ) = r
z = zz* = (a + ib)(a ib) = r
2 2
10
The spectrum
Amplitude
Amplitudespectrum
spectrum Phase
Phasespectrum
spectrum
space
Fourierspace
Fourier space
Physicalspace
Physical
11
(FFT)
>>
Most >>help
helpfftfft
Mostprocessing
processingtools
tools
(e.g.
(e.g.octave,
octave,Matlab,
Matlab, FFT
FFTDiscrete
DiscreteFourier
Fouriertransform.
transform.
FFT(X)
Mathematica, FFT(X)isisthe
thediscrete
discreteFourier
Fouriertransform
transform(DFT)
(DFT)ofofvector
vectorX.X. For
For
Mathematica, matrices, the FFT operation is applied to each column.
matrices, the FFT operation is applied to each column. For N-D For N-D
Fortran,
Fortran,etc)
etc)have
have arrays,
arrays,the
theFFT
FFToperation
operationoperates
operateson onthe
thefirst
firstnon-singleton
non-singleton
intrinsic functions
intrinsic functions
dimension.
dimension.
for
forFFTs
FFTs FFT(X,N)
FFT(X,N)isisthe
theN-point
N-pointFFT,
FFT,padded
paddedwith
withzeros
zerosififXXhas
hasless
less
than N points and truncated if it has more.
than N points and truncated if it has more.
FFT(X,[],DIM)
FFT(X,[],DIM)ororFFT(X,N,DIM)
FFT(X,N,DIM)applies
appliesthe
theFFT
FFToperation
operationacross
acrossthe
the
dimension DIM.
dimension DIM.
For
Forlength
lengthNNinput
inputvector
vectorx,x,the
theDFT
DFTisisaalength
lengthNNvector
vectorX,X,
with
withelements
elements
Matlab FFT NN
X(k)
X(k) = sum
= sum x(n)*exp(-j*2*pi*(k-1)*(n-1)/N),
x(n)*exp(-j*2*pi*(k-1)*(n-1)/N),11<= <=kk<=
<=N. N.
n=1
n=1
The
Theinverse
inverseDFT
DFT(computed
(computedby byIFFT)
IFFT)isisgiven
givenbyby
NN
x(n)
x(n)==(1/N)
(1/N)sum
sum X(k)*exp(
X(k)*exp(j*2*pi*(k-1)*(n-1)/N),
j*2*pi*(k-1)*(n-1)/N),11<=
<=nn<= <=N.
N.
k=1
k=1
See
Seealso
alsoIFFT,
IFFT,FFT2,
FFT2,IFFT2,
IFFT2,FFTSHIFT.
FFTSHIFT.
12
Frequencies in seismograms
13
Amplitude spectrum
Eigenfrequencies
14
Sound of an instrument
a - 440Hz
15
Instrument Earth
26.-29.12.2004 (FFB )
0S2 Earths gravest tone

T=3233.5s =53.9min
Theoretical eigenfrequencies
16
Fourier Spectra: Main Cases
random signals
Random
Randomsignals
signalsmay
maycontain
containall
allfrequencies.
frequencies.AAspectrum
spectrumwith
with
constant
constantcontribution
contributionof
ofall
allfrequencies
frequenciesisiscalled
calledaawhite
whitespectrum
spectrum
17
Gaussian signals
The
Thespectrum
spectrumof
ofaaGaussian
Gaussianfunction
functionwill
willitself
itselfbe
beaaGaussian
Gaussian
function.
function.How
Howdoes
doesthe
thespectrum
spectrumchange,
change,ififIImake
makethe
theGaussian
Gaussian
narrower and narrower?
narrower and narrower?
18
Transient waveform
AAtransient
transientwave
waveform
formisisaawave
waveform
formlimited
limitedinintime
time(or
(orspace)
space)inin
comparison
comparisonwith
withaaharmonic
harmonicwave
waveform
formthat
thatisisinfinite
infinite
19
Puls-width and Frequency Bandwidth
time (space) spectrum
Widening frequency band

Narrowing physical signal
20
Spectral analysis: an Example
24 hour ground motion, do you see any signal?

21
Seismo-Weather
Running spectrum of the same data

22
Some properties of FT
FT is linear Shifting corresponds to
signals can be treated as the sum changing the phase (shift
of several signals, the transform theorem)
will be the sum of their transforms
f (t a ) e ia F ( )
FT of a real signals F ( a ) e ia f (t )
has symmetry properties
Derivative
F ( ) = F * ( )
the negative frequencies can be dn
obtained from symmetry f (t ) (i ) n F ( )
properties
dt
23
Fourier Derivatives
.. let us recall the definition of the derivative using Fourier integrals ...

x f ( x) = x F (k )e dk
ikx

= ikF (k )e ikx dk

... we could either ...
1) perform this calculation in the space domain by convolution
2) actually transform the function f(x) in the k-domain and back
24
Acoustic Wave Equation - Fourier Method
let us take the acoustic wave equation with variable density
1 1
2
p =
x p
c
2 t x

the left hand side will be expressed with our

standard centered finite-difference approach
1
1
[p (t + dt ) 2 p ( t ) + p ( t dt ) ] = x x p
c 2 dt 2

... leading to the extrapolation scheme ...
25
Acoustic Wave Equation - Fourier Method
1
p ( t + dt ) = c 2 dt 2 x x p + 2 p ( t ) p ( t dt )

where the space derivatives will be calculated using the Fourier Method.
The highlighted term will be calculated as follows:
1
P jn FFT P n ik P n FFT x P jn
multiply by 1/
n n
1 1 1 1 1 n
x P FFT x P ik x P FFT x x Pj
n

j

... then extrapolate ...
26
... and the first derivative using FFTs ...
function df=sder1d(f,dx)
% SDER1D(f,dx) spectral derivative of vector
nx=max(size(f));
% initialize k
kmax=pi/dx;
dk=kmax/(nx/2);
for i=1:nx/2, k(i)=(i)*dk; k(nx/2+i)=-kmax+(i)*dk; end
k=sqrt(-1)*k;
% FFT and IFFT

ff=fft(f); ff=k.*ff; df=real(ifft(ff));
.. simple and elegant ...
27
Fourier Method - Comparison with FD - Table
Difference (%) between numerical and analytical solution

as a function of propagating frequency
160
140
120
Simulation time
100
3 point 5.4s
80 5 point 7.8s
Fourier 33.0s
60
40
20
0
5 Hz 10 Hz 20 Hz
28
Numerical solutions and Greens Functions
3 point operator 5 point operator Fourier Method

10 10 10
Impulse response (analytical) concolved with source

Impulse response (numerical convolved with source
9 9 9
8 8 8
7 7 7
Frequency increases
6 6 6
5 5 5
4 4 4
3 3 3
2 2 2
1 1 1
0 0 0
500 1000 1500 500 1000 1500 500 100
29

Fourier Transform Applications

Uploaded by

Copyright:

Available Formats

Fourier Transform Applications

Uploaded by

Document Information

Original Description:

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

Fourier Transform Applications

Uploaded by

Copyright:

Available Formats

Fourier Transform:

The discrete Fourier Transform can be written as

... this can be written as matrix-vector products ...

full matrix FFT

this has enormous implications for large scale problems.

Problem full matrix FFT Ratio full/FFT

1D (nx=512) 2.6x105 9.2x103 28.4

.. the right column can be regarded as the speedup of an algorithm

In many cases the amplitude spectrum is the most important

0S2 Earths gravest tone

Widening frequency band

24 hour ground motion, do you see any signal?

Running spectrum of the same data

... we could either ...

1) perform this calculation in the space domain by convolution

2) actually transform the function f(x) in the k-domain and back

let us take the acoustic wave equation with variable density

the left hand side will be expressed with our

... leading to the extrapolation scheme ...

% FFT and IFFT

.. simple and elegant ...

Difference (%) between numerical and analytical solution

3 point operator 5 point operator Fourier Method

Impulse response (analytical) concolved with source

You might also like