Scribd Logo
Upload

Welcome to Scribd!

  • 72 views

    FFT Scaling

    Uploaded by

    thanh910
    The document discusses Matlab's FFT and IFFT functions and how to properly handle the scaling when using them. It provides the definitions that Matlab uses for the FFT and IFFT. It then works through an example of taking the Fourier transform of a Gaussian pulse, transforming it back, and multiplying it by its intensity to verify that the scaling is correct. The example shows propagating solitons using the nonlinear Schrodinger equation solved in the frequency domain.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    FFT Scaling

    Uploaded by

    thanh910
    72 views5 pages
    The document discusses Matlab's FFT and IFFT functions and how to properly handle the scaling when using them. It provides the definitions that Matlab uses for the FFT and IFFT. It then works through an example of taking the Fourier transform of a Gaussian pulse, transforming it back, and multiplying it by its intensity to verify that the scaling is correct. The example shows propagating solitons using the nonlinear Schrodinger equation solved in the frequency domain.

    Original Description:

    FFT

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses Matlab's FFT and IFFT functions and how to properly handle the scaling when using them. It provides the definitions that Matlab uses for the FFT and IFFT. It then works through an example of taking the Fourier transform of a Gaussian pulse, transforming it back, and multiplying it by its intensity to verify that the scaling is correct. The example shows propagating solitons using the nonlinear Schrodinger equation solved in the frequency domain.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    72 views5 pages

    FFT Scaling

    Uploaded by

    thanh910
    The document discusses Matlab's FFT and IFFT functions and how to properly handle the scaling when using them. It provides the definitions that Matlab uses for the FFT and IFFT. It then works through an example of taking the Fourier transform of a Gaussian pulse, transforming it back, and multiplying it by its intensity to verify that the scaling is correct. The example shows propagating solitons using the nonlinear Schrodinger equation solved in the frequency domain.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 5

    Figuring out Matlabs FFT/IFFT and their scaling

    The FFT/IFFT in Matlab always irritate me since I screw up the scaling and then get confused about why my program
    doesnt work correctly. Inevitably I always trace the root of the problem back to the fft scaling. Part of the problem
    resides in the fact that although taking a simple Fourier transform and then its inverse is simplistic and will always
    give the correct results there are these issues. If the quantity being transformed is located in the center of the window
    (as ours always is) fftshifting is always required. The fftshifting issue is simple, however, things get tricky since basic
    NLSE code will work if one ignores the fft scaling so when one modies the code and then it doesnt work you are left
    scratching your head: It worked before! Why not now? The reason is that the scaling in the fourier domain must
    be correct if you are going to do autocorrelation or solve the Generalized NLSE for example anything where your are
    multiplying multiple things in the Fourier domain.
    Lets start out by reviewing the fft/ifft denitions in matlab and then work out an example to verify our approach..
    Matlab Denitions
    Matlab denes its fft using the following:

    X(k) =
    N

    n=1
    x(n)e
    j2(k1)(n1)/N
    , 1 <= k <= N (1)
    This corresponds to the analytic result:

    X(w) =

    x(t)e
    iwt
    dt (2)
    We see that this actually corresponds to our inverse FT although it does not contain the 1/2 scaling
    Matlab denes its ifft as:
    x(n) =
    1
    N
    N

    k=1

    X(k)e
    j2(k1)(n1)/N
    , 1 <= n <= N (3)
    This of course corresponds to our Fourier transform:
    x(t) =

    X(w)e
    iwt
    dw (4)
    Lets look at the spectral domain of ||
    2

    We start out by dening some time vector T and a width T0 we then dene a Gaussian eld so that we can check
    our code:
    (t) = A0e
    (T.
    2
    )/(2T0
    2
    )
    (5)
    To get this eld in the frequency domain we will use the ifft as the Fourier Transform!
    () = fftshift(ifft[fftshift[(t)]]) scale (6)
    For the Gaussian we are looking at:
    () =

    2 T0 exp

    2
    T0
    2
    2

    (7)
    where scale is dened as scale = dTN
    1
    As noted above simply transforming and then back transforming always works but to verify that our code is correct
    we check against the nonlinear term in the NLSE as noted above
    old() = (); (8)
    old(t) = fftshift(fft[fftshift[old()]])/scale; (9)
    I = abs(old(t)).
    2
    ; (10)
    So we did just go in a circle: We dened our eld in the time domain then transformed to the frequency domain. We
    then transformed back Hence we are just left with what we started with: (but we are doing this to check our code one
    step at a time)
    (t) = A0e
    T
    2
    /(2T0
    2
    )
    (11)
    Now dene the nonlinear term:
    NL(t) = old(t). I; (12)
    Analytically this just gives us:
    NL(t) = A0
    3
    e
    3T
    2
    /(2T0
    2
    )
    (13)
    in the time domain
    Now to get this in the frequency domain we just perform:
    () = fftshift(ifft[fftshift[NL(t)]]) scale; (14)
    Which analytically corresponds to:
    () = A0
    3

    2
    3
    T0e
    w
    2
    T0
    2
    /6
    (15)
    NLSE using RKF
    The NLSE may be solved in the frequency domain simply by integrating:


    S(, z)
    z
    =
    i
    2
    2

    2

    S(, z) + iF

    S(t, z)|
    2

    S(t, z)

    (16)
    Using the Runge-Kutta-Fehlberg Method of fth order:
    2
    Looking specically at higher order soliton propagation we get:
    2
    1
    0
    1
    2
    0
    0.2
    0.4
    0.6
    0.8
    1
    0
    2
    4
    6
    8
    10
    Time (T/T0)
    RKF Soliton Propagation
    Distance (Z/Z0)
    I
    n
    t
    e
    n
    s
    i
    t
    y
    Figure 1: Propagation of a third order soliton over one soliton period using RKF.
    Well this is easy to get with the split step so lets look at the next two:
    And nally some serious computation to get a nice view of N = 5..
    Refrences for this kind of thing: P. L. Francois, Nonlinear propagation of ultrashort pulses in optical bers: total
    eld formulation in the frequency domain JOSAB 276 (1991).
    3
    2.5
    2
    1.5
    1
    0.5
    0
    0.5
    1
    1.5
    2
    2.5
    0
    0.2
    0.4
    0.6
    0.8
    1
    0
    2
    4
    6
    8
    Time (T/T0)
    RKF Soliton Propagation
    Distance (Z/Z0)
    I
    n
    t
    e
    n
    s
    i
    t
    y
    Figure 2: Propagation of a fourth order soliton over one soliton period using RKF.
    4
    2
    1.5
    1
    0.5
    0
    0.5
    1
    1.5
    2
    0
    0.2
    0.4
    0.6
    0.8
    1
    0
    5
    10
    Time (T/T0)
    RKF Soliton Propagation
    Distance (Z/Z0)
    I
    n
    t
    e
    n
    s
    i
    t
    y
    Figure 3: Propagation of a fth order soliton over one soliton period using RKF.
    5

    You might also like