ocw.nctu.edu.twocw.nctu.edu.tw/course/engineeringmathematicsi/ch04.pdf · mingsian r. bai 1 sin 2...

66
Fourier Series and Fourier Transforms 機械工程學系 白明憲 教授

Upload: vutuyen

Post on 23-May-2018

217 views

Category:

Documents


1 download

TRANSCRIPT

  • Fourier Series and Fourier Transforms

  • MingsianMingsian R. R. BaiBai22

    0 1 2 cos particular solution of the form cos sin

    How about an arbitrary periodic input? Fourier series

    * ( ) is periodic ( 2 ) ( ) , : integer, 2 is called the e op riod

    a y a y a y k tY A t B t

    f t f t Pn f tn P

    + + = = +

    + =f ( )

    * : periodic discrete freq.

    f t

  • MingsianMingsian R. R. BaiBai33

    0

    1

    1

    1

    1( ) ( cos sin ) , 2

    even periodic odd per

    2 : ;22 , 1, 2, : superharmonic

    fundamental frequency

    History of Fouri

    2

    ierodic

    Fourier Series

    n nn

    n

    n t n tf t a a bP P

    Pn n n

    P

    =

    =

    = = =

    = + +

    K s

    Oscillates increasingly faster as .' & ' are called the Euler coefficients.n n

    na s b s

  • MingsianMingsian R. R. BaiBai44

    2 2

    2

    2

    Orthogonality ( ) cf. linear al

    1. cos 0 , sin

    gebra

    0 , 0

    2.

    cos cos , , 0

    sin sin , , 0

    s

    co

    d P d P

    d d

    d P

    mnd

    d P

    mnd

    n t n tdt dt nP P

    m t n t dt P m nP P

    m t n t dt P m nP P

    m t

    + +

    +

    +

    = =

    =

    =

    Note :

    2sin 0

    d P

    d

    n t dtP P

    +=

    ( ) ( )2

    cos cos 1 cos cos2

    cos 2 1cos2

    = + +

    +=

    (cos,sin are orthogonal functions, see Sturm-Liuille)

  • MingsianMingsian R. R. BaiBai55

    0

    2 20

    To find ,

    c( ) os2

    d P d Pnd d

    aa n tdtf d

    Pt t a

    + += + sinn

    n tdt bP+

    1

    20

    2 2

    00 = 22

    (DC bias, me1 ( an val )) ued

    d P d

    P

    P

    d dn

    d

    dt

    a P

    a f t dtP

    Pa

    + +

    =

    +

    =

    =

    DC0

    2a

    ( )f t

  • MingsianMingsian R. R. BaiBai66

    2

    0

    To find ,

    1co) c( s os2

    n

    d P

    d

    a

    n t n tdt aP P

    f t +

    =2 2

    1

    cos cos

    sin cos

    d P d P

    md dm

    m

    m t n tdt a dtP P

    m t n tbP P

    + +

    =

    + +

    2 2

    0 0

    2

    2cos

    1 ( & can b1 ( )cos e combined here )2

    d P

    n d

    d P d P

    n nd d

    nn ta

    n tdt a dt a PP

    a a af t dtP P

    + +

    +

    = =

    =

    Q

    2

    0

    To find ,

    1 sin sin2

    ( )

    n

    d P

    d

    b

    n t n tdt aP

    fP

    t +

    =2

    cos sind P

    md

    m t n tdt aP P +

    +2

    1

    2 2 2

    2

    sin sin sin

    1 ) ( sin

    d P

    dm

    d P d P

    m n nd d

    d P

    n d

    m t n t n tb dt b dt b P

    n tb f t dtP P

    P P P

    +

    +

    +

    =

    +

    + = =

    =

  • MingsianMingsian R. R. BaiBai77

    Thm 1 If ( ) is a bounded periodic function which, in any one period, has at most a finite number of local maxima and minima and a finite n

    Fourier conve

    umber of po

    rg

    in

    ence

    ts of

    the

    orem

    discont

    f t

    01

    inuity, then

    Pf: see p.81 ,

    (

    )

    Greenberg, Foundations of

    ( ) 1 cos sin2 2

    Applied Mathematics.

    n nn

    f t f t n t n ta a bP P +

    =

    + = + +

    interchangibility of

    Dirichlet cond. (sufficient cond.)

    Q

    converges to the mean

  • MingsianMingsian R. R. BaiBai8

    ( )1

    0

    1

    0

    0

    1 ( ) cos sin (synthesi

    2 ( ) cos , (analysis equations)

    * Alternative f

    2 ( )

    orm:

    sin

    s equation)

    2 , 2

    2T

    n

    n n n

    n

    n

    n

    n

    n

    T

    n

    a f t t dtT

    b f t t

    nn

    dt

    f t a a t b t

    T

    n TT P

    =

    = = = =

    =

    = + +

    =

    , 0,1, 2,

    *Plot Time-domain: ( ) Frequency-domain: ,

    *Parameters : sin sin , (rad/sec) 2 (cycle/sec, CPS, Hz)

    FSn n

    P n

    f t a b

    t t f

    =

    =

    K

  • MingsianMingsian R. R. BaiBai9

    Ex. Half-wave rectifierFind the Fourier series of a 2 -periodic function

    0 0 ( )

    sin 0t

    f tt t

    =

    t

    ( )tf

    0 2 3

    2

    0

    2 20

    1 00

    Sol: 2 / 2 Let

    1 1 ( ) cos sin cos

    1 1 cos(1 ) cos(1 ) 1 cos 1 ( 1) , 12 1 1 (1 ) (1 )

    1 cos 2 1: sin cos 04

    n

    n

    Pd

    n ta f t dt t nt dt

    n t n t n nn n n n

    tn a t t dt

    = ==

    = =

    + + + = + = = +

    = = = =

  • MingsianMingsian R. R. BaiBai10

    ( ) ( )

    0

    0

    21 0

    0

    1 1( )sin sin sin

    1 1 sin(1 ) sin(1 ) = 0, 12 1 1

    1 1 sin 2 11: sin (or By l'

    sin

    Hos2

    s

    4 2

    in 1 cos cos2

    nn tb f t dt t nt dt

    n t n t nn n

    t tn b t dt

    = =

    + = +

    = = = =

    = +

    Q

    pital rule)

  • MingsianMingsian R. R. BaiBai11

    1 sin 2 cos 2 cos 4 cos6 cos8Thus, ( )2 3 15 35 63

    * The "smoother' the function, the fewer terms the F.S. takes to converge. Low-frequency components prevail.

    Thm 4 term-by-term integra ti

    t t t t tf t

    = + + + + +

    L

    The integral of any periodic function ( ) which satisfies the Dirichlet condition can be found by term-by-term integration of the F.S. of the fu

    on

    term-by-term differentia

    n

    t

    ction.

    Thm 5 If ( ) i

    ion

    f t

    f t s a periodic function which satisfies the Dirichlet condition, and is everywhere , and if ( ) also satisfies the Dirichlet condition, then ( ) can be found by term-by-te

    continrm dif

    uoferentia

    ust

    f tf t

    ion of the F.S. of ( ).

    Ex. Square wave is a counter-example.f t

  • MingsianMingsian R. R. BaiBai12

    Half-Range Expansions0

    1

    1( ) cos sin2

    == + + n n

    n

    n t n tf t a a bP P

    even odd

    0

    0

    ( ) is an function, i.e., ( ) ( ) Let

    1 1 1 ( )cos ( )cos ( )cos

    ev

    ( )

    e

    n

    P Pn P P

    f t f t f td P

    n t n t n ta f t dt f t dt f t dtP P P P P P

    t t

    ==

    = = +

    0

    0

    0 0 0

    1 ( ) 1 ( )cos ( ) ( )cos

    1 1 2 ( )cos ( )cos ( )cos

    P

    P

    P P P

    n t n tf t d t f t dtP P P P

    n t n t n tf t dt f t dt f t dtP P P P P P

    = +

    += + =

  • MingsianMingsian R. R. BaiBai13

    0

    01

    1( ) is an function, ( ) ( ) ( ) cos2

    ,

    1 ( )sin 0 (Fourier cosine series)

    If ( ) is an f

    even

    even odd

    odd

    2 ( )c

    unction, )

    os

    (

    =

    = = +

    =

    =

    =

    nn

    n

    Pn

    P

    P

    n ta

    n tf t f t f t f t a aP

    n tb f t dtP P

    f t dtP P

    f t f t1

    0

    ( ) ( ) sin

    1 ( )cos 0,

    2

    odd even

    (Fourier( ) sine ser esi is )n

    =

    =

    =

    =

    n

    Pn

    n

    Pn P

    n tb f t dtP

    n tf t f t bP

    n ta f t dtP P

    P

  • MingsianMingsian R. R. BaiBai14

    : consider ( ) only in (0, p] for solving PDEs

    ( ), 0 : arbitrary func( ), 0

    tion( )

    ( 2 ) ( ), periodic extension

    Half-range expansion

    <

  • MingsianMingsian R. R. BaiBai15

    Alternative forms of Fourier Series

    2 2 2 2

    0

    1

    2 20

    1

    cos sin

    cos sin

    ( )2

    Single trigonometri

    c form:

    2

    n n

    n n

    n n n n

    n

    n nn

    n t n ta bp p

    a bn t n tp pa b a b

    af t

    a a b

    =

    =

    +

    + + +

    = +

    = + +

    n

    nnb

    na

    2 2= +n n nA a b

    ( )2 2 100

    0 01 1

    0 01 1

    cos cos sin sin

    sin cos cos sin

    Let , , tan /2 2

    ( ) cos

    ( ) sin

    n n n

    n n n

    n n n n n n n

    n nn n

    n nn n

    n t n t n tp p p

    n t n t n tp p p

    aA A a b b a

    f t A A A A

    f t A A A A

    = =

    = =

    +

    + +

    = = + = =

    = + = +

    = + = +

  • MingsianMingsian R. R. BaiBai16

    0

    ( ) (synthesis equation)

    1 ( ) (analysis equation)

    n

    n

    i tn

    n

    T i tn

    f t C e

    C f t e dtT

    =

    =

    =

    1

    1

    1

    (rad/sec) 2 (Hz)

    2 (period),2 ,

    2Fundamental frequency:

    Note that ( , ) Heisenberg uncertainty principle

    n

    f

    T p

    n n nT p

    TT

    =

    =

    = = =

    =

    Complex form of FS

  • MingsianMingsian R. R. BaiBai17

    0

    1

    0

    1

    0

    1

    00

    cos sin

    2 2

    2 2

    ( )2

    2

    2

    Defi

    Complex FS de

    ne

    rivation:

    , ,2 2

    n n

    in t p in t p in t p in t p

    n n

    in t p in t pn n n n

    n

    n

    n

    n nn

    n t n ta bp p

    e e e ea bi

    a ib a ibe e

    af t

    a

    a

    a a ibC C

    =

    =

    =

    +

    +

    +

    + +

    = +

    = +

    = +

    = =

    20 0

    2 2

    2cos

    2

    ( ) ( ),

    1 1 ( ) (1)2 2

    1 1 1( )cos ( )sin2 2

    1 1 ( ) sin2 2

    n

    n nn

    in t p i tn n

    d p

    d

    d p d pn nn d d

    n

    d p

    d

    n

    n t i dtp

    a ibC

    f t C e C e

    C a f t dtp

    a ib n t n tC f t dt i f t dtp p p p

    n tf tp p

    =

    +

    + +

    =

    +

    +=

    = =

    = =

    = =

    = =

    2( ) (2)

    d p in t pd

    f t e dtp

    +

  • MingsianMingsian R. R. BaiBai18

    2

    2 2

    1Similarly, ( ) (3)21 1( ) ( )

    2(1),(2),(3): ( ) , ,0,1, ,

    n

    d p in t pn d

    d p d pin t p i tn d d

    C f t e dtp

    C f t e dt f t e dtp T

    n

    +

    + +

    =

    = =

    =

    K K

    1(1 ) 11

    11

    1

    2 2

    Ex. ( ) 1 1 in one period ( 1)

    1 1Sol: 2 2 (1 ) 2(1 )

    ( 1) ( 1) (1 )sinh(1) (1 2 1

    t

    in t in inin t pt

    n

    n nin

    f t e t p

    e e e e eC e e dtp in in

    e e in ein n

    +

    = < =

    = = = + +

    = = = + +

    ( )1)

    ( )

    n

    in tn

    nf t C e

    ==

  • MingsianMingsian R. R. BaiBai19

    0

    ( ) (synthesis equation)

    1 ( ) (analysis equation)

    n

    n

    i tn

    n

    T i tn

    f t C e

    C f t e dtT

    =

    =

    =

    ( ) FS nf t C

    t

    ( )f t

    FS

    nC

    1n =0 1 21(time domain)continuous,periodic

    (freq. domain) Spectrum( ) discrete, aperiodic

    In Digital Signal Processing (DSP), the following Fourier Series pair is used to represent a continuous and periodic signal.

  • MingsianMingsian R. R. BaiBai20

    ApplicationEx. Steady-state response due to a periodic input

    2 0 0.02 25 ( ) . ( )

    2 0 ( 2 ) ( ),

    t ty y y r t st r t

    t tr t r t p

    + < < + + = =

    + <

    =

    =

    +

    1( ) { ( ) ( )}ih t FT E H =

    t

    1

    ( )h t

  • MingsianMingsian R. R. BaiBai57

    { } { }{ }

    { }

    1 1

    * Three ways of descrbing a :: ( ) ( ) ( Impulse response function

    Transfer function Frequency re

    LTI)

    : ( ) ( ): ( ) ( ) ( )sponse fun

    sys

    ction

    tem

    s j

    h t LT H s FT HH s LT h t

    H FT h t H s

    =

    = =

    =

    = =

    FRF is only meaningful for stable LTI systems.

    ( )h t

    ( )H s ( )H

    LT FT

    s j=

    system

  • MingsianMingsian R. R. BaiBai58

    Sampling theoremEx. fundamental theorem of DSPAssume fundamental theorem of DSP

    ( ) ( ) ( ), where ( ) ( )

    being an

    Nyquist-Shannon

    implu , se train ( ) is the Dirac delta function a

    s cn

    x t x t s t s t t nT

    t

    =

    = =

    nd is the sampling period.Find the Fourier transform ( ) in terms of ( ).s c

    TX X

  • MingsianMingsian R. R. BaiBai59

    1 2 ( ) 2 ( ) ( property)1 2( ) ( ) ( )

    (impulse train to impluse tra

    modulation

    in)1Knowing ( ) ( ) ( ) ( ) ( )

    2

    s

    s

    jk tFT FTs

    jk t FTs

    k k

    FTs c c

    e k

    s t e S kT T

    x t x t s t X S

    = =

    = =

    =

    ( theorem)convoluti

    1 1( )

    n

    ( )2

    o

    ) (2s c

    X X S

    = = 2( )cX ( )s

    kk

    T

    =

    Sol: Since ( ) is a function, it has the following representation (Dirichlet conditions violated):

    ( ) ( ) , where (sampling frequency)

    1 ( )

    periodic FS

    2sjk tk

    n k

    j

    s

    kk

    T

    s t

    s t t nT a e

    a t eT

    = =

    =

    == = /2

    /2

    1 1( )s sT

    t jk t

    kTdt s t e

    T T

    =

    = =

  • MingsianMingsian R. R. BaiBai60

    ( ) ( ) ( ) [ ( )]( ) (freq-shifted impuse response)

    1 1( ) ( ) ( ) ( )

    ( ) ( 1( ) ( ) )) (

    c s c s

    c s

    s c s c sk k

    FTs sc c s

    k

    X k X k dX k

    X

    x t s t X kT

    X k X kT T

    x t X

    =

    = =

    =

    =

    = =

    = =

    Q

    Oversampled, no aliasing: 2 > s N Undersampled, aliasing: 2 < s N

  • 61 MingsianMingsian R. R. BaiBai

    Demo of Aliasing: Alice in the Wonderland

    20 0

    0

    Linear chirp/sweep sine:[ ] cos(0.5 ), * / /nsec

    Instantaneous frequency

    Time duration: 0 ~ sec (sec) Frequency sweeping range: 0 ~ (rad/sample)

    sx n n r f

    nn

    nr

    = =

    = =

    %MATLAB programfs=10000;nsec=4; %time durationr=2; %aliasing factorn=0:1:nsec*fs;w0=r*pi/fs/nsec;x=cos(0.5*w0.*n.*n);

    No aliasing:

    Aliased:

    InterpretationFor the digital discrete-time signal, the sampling frequency must be greater than two times the maximum frequency of the analog continuous-time signal.

    At least two samples within one cycle of a sinusoid must be sampled.

  • MingsianMingsian R. R. BaiBai62

    [ ]

    Ex. Amplitude-modulated signals bandpass (narrowband) signals, Why modulation? 15 ( ) ( ) cos ( )cx t a t t t

    = +

    Q ( km at 20 kHz 15 mm at 20 GHz)

    amplitude

    carrier frequency

    phase

    t

    ( )x t( ): baseband info-bearing signal

    (audio,speech,video )a t

    K

    carrier

    ( )a t

    [ ]cos ( )ct t +

    ( )x t

  • MingsianMingsian R. R. BaiBai63

    *

    ( ) BasebandDefine complex amplitude (phasor)

    ( ) ( ) ( ) ( ) ( )cos ( ) Re ( ) ( )

    ( ) ( ) Re ( )

    ( ) ( )

    RF ba

    =

    nd

    2

    c

    c

    c c

    j t

    j tc

    j t j t j t

    j t j t

    u t a t ex t a t t t u t e

    X j x t e dt u t e e dt

    u t e u t e

    =

    = + =

    = =

    +

    ( )

    ( ) ( )

    ( )

    -

    ( ) *

    *

    *

    1 = ( ) ( )2

    1 1 = ( ) ( )2 21 1 = ( )2 2

    cc

    c c

    j t

    j tj t

    j t j t

    c c

    e dt

    u t e u t e dt

    u t e dt u t e dt

    U j j U j j

    +

    +

    +

    +

  • MingsianMingsian R. R. BaiBai64

    ( )X j12

    c c

    (RF band)

    ( )U j1

    (Base band)

  • MingsianMingsian R. R. BaiBai65

    HW (Wylie)

    Sec.9.1: 18, 30Sec.9.3: 4(a), (b), 11Sec.9.4: 18Sec.9.5: 13

    Due date: 1/12/2009

  • MingsianMingsian R. R. BaiBai66

    Final Exam

    : CD, 1/12/2009

    :

    Laplace transformFourier tranaform

    Fourier Series and Fourier Transforms 2 3 4 5 6 7 8 9 10 11Half-Range Expansions 13 14Alternative forms of Fourier Series 16 17 18 19Application 21 22 23 24WylieFourier Integral and Fourier Transform 27 28 29Applications of Fourier Transform (FT) Vibration diagnostics 32 Time Versus Frequency Diagram 34 35 36 37 38 39 40 41 42 43Summary 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64HW (Wylie) Final Exam