a pod is at ion

Upload: beatriceportiniari

Post on 29-May-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 A Pod is at Ion

    1/13

    Project on Non-linear Apodisation

    Techniques

    for

    Sidelobe Suppression

    Submitted by

    Satam Choudhury

    Electronics and Communication Engg.

    National Institute of Technology, Silchar

  • 8/9/2019 A Pod is at Ion

    2/13

    Acknowledgement

    I would like to express my heartfelt gratitude for my supervisor, Dr. Amit Kumar

    Mishra, Dept. of Electronics and Communications Engg., Indian Institute of Technology,

    Guwahati, for guiding me through my summer project on Non-linear Apodisation

    Techniques for Sidelobe Suppression. My special thanks to Mr. Rajiv Panigrahi, Dept. of

    Electronics and Communications Engg., Indian Institute of Technology, Guwahati ,for

    constantly clarifying my doubts and also guiding my project work.

    I am beholden to the Ms. Jharana Rabha of System Simulation Laboratory for

    providing me with the necessary facilities for accomplishing the project.

  • 8/9/2019 A Pod is at Ion

    3/13

    We have considered a sinusoidal signal comprising of 3 components, 2, 2.5 and 5 kHz

    and the signal is sampled at a rate 1.5 times the Nyquist rate. The spectrum analysis is done

    using the following techniques:

    Dual Apodisation

    fig.1

    fig.2

  • 8/9/2019 A Pod is at Ion

    4/13

    The two versions of the image are the IPRs shown in Fig 1. and Fig 2. (Hanning). The

    resultant dual-apodized IPR, shown below has the narrow mainlobe of the sinc IPR and the

    small sidelobes of the Hanning IPR. The combination of a uniformly weighted image with a

    normalized version of the same image computed with any weighting function chosen to

    produce low sidelobes is the basic principle behind Dual apodisation. The combination

    technique uses a min function applied to the detected image on a pixel-by-pixel basis.

    Applied to complicated images, DA results in an effective weighting function that can differ

    from sample to sample in the image. Thus it results in a weighting function and an IPR that

    are both nonlinear and spatially varying.

  • 8/9/2019 A Pod is at Ion

    5/13

    Complex Dual Apodisation

    To make use of the complex values we operate on the real (I or in-phase) and

    imaginary (Q or quadrature) components of the undetected images separately. If the value

    of a component has a sign that is different for the sinc IPR than for the Hanning IPR, then

    there must be some weighting function intermediate between uniform weighting and

    Hanning weighting for which the value of the component would be zero.

    CDA provides significantly lower sidelobes than DA for the one-dimensional case (1-D).

    SVA

    We shall now discuss spatially variant apodization (SVA), which allows each pixel in an

    image to receive its own frequency domain aperture amplitude weighting function from a

    continuum of possible weighting functions. SVA effectively eliminates finite-aperture induced

  • 8/9/2019 A Pod is at Ion

    6/13

    sidelobes from uniformly weighted SAR data while retaining nearly all of the good mainlobe

    resolution and clutter texture of the unweighted SAR image.

    (simulated at 2-times Nyquist frequency)

  • 8/9/2019 A Pod is at Ion

    7/13

    (simulated at 1.5 times Nyquist frequency)

    We observe that for non-integer multiples of Nyquist sampling frequency, the

    resolution is not accurate and the sidelobe suppression is not complete.

    There is another important limitation of Spatially Variant Apodisation. At higher

    smapling rates, the mainlobe resolution is not good and the sidelobes are not suppressed as

    well.

  • 8/9/2019 A Pod is at Ion

    8/13

  • 8/9/2019 A Pod is at Ion

    9/13

    However the proposed algorithm intends to address this problem of sidelobe

    suppression at non-integer multiples of Nyquist sampling frequency.

    1. At first, the zero crossing points are determined. For every three consecutive zerocrossing points, there is a maxima( appositive maxima) and a minima(a negative

    minima). These maxima are determined for a uniform weighting function. The same

    process is carried out for two or more windowed-weighted functions and the points of

    maxima are determined. The maxima points are compared and those points which

    occur in the all the weighted functions are retained while other maxima( the

    sidelobes) are ignored.

    2. A particular case may occur when two closely placed frequency components overlapeach other. This may be illustrated as follows:

    The following signal is composed of 6 frequency components: 2, 2.9, 3.5,3.8,4, 5 kHz

  • 8/9/2019 A Pod is at Ion

    10/13

    In this case we observe that there is overlap of two closely spaced frequencycomponents. It is obvious that in between zero-crossing points there is more than 1 maxima.

    So we need to determine the total positive minima in that interval and subdivide the interval

    into [zero-crossing point 1, 1st

    +ve minima], (1st

    +ve minima, 2nd

    +ve minima](nth +ve

    minima, zero crossing point 2]. The same process is repeated for other weighted window

    functions and the maxima are compared. This process can help us in considerably resolving

    closely spaced frequency components.

    Program for Uniform Weighting, Hanning weighting and Dual apodisation

    clc;

    clear all;

    close all;

    N= 256;

    l = 64;%length of seq

    t=sqrt(-1);

    n = 0:1:l-1;

    eps=power(10,-10);

    %frequency components

    F1 = 5000;%in Hz

    F2 = 2000;

    F3= 2500;

    Fs = 3*F1; %sampling freq

    f1 = F1/Fs;

    f2= F2/Fs;

    f3= F3/Fs;

    %given signal

    x = sin(2*pi*f1*n)+sin(2*pi*f2*n)+sin(2*pi*f3*n);

    f= (-Fs/2):(Fs/N):(Fs/2-(Fs/N));%range

    x1= [x zeros(1,N-l)];%zero padding

    y1= fftshift(fft(x1));

  • 8/9/2019 A Pod is at Ion

    11/13

    m1= abs(y1); figure(1);

    plot(f,m1);grid on;title('Uniform weighting'); ylabel('Amplitude');

    %windoww= (hann(l))';

    x2= x.*w;

    x3= [x2 zeros(1,N-l)];

    %fft of x1 and x2

    y3= fftshift(fft(x3));

    m3= abs(y3);

    figure(2);plot(f,m3);grid on;title('Hanning Weighting'); xlabel('frequency in

    Hz');ylabel('Amplitude');

    %DA

    for k= 1:N

    mag1(k)=min(m1(k),m3(k));end

    figure(3);plot(f,mag1);grid on;title('DA'); xlabel('frequency in

    Hz');ylabel('Amplitude');

    Program for Complex Dual Apodisation

    clc;

    clear all;

    close all;

    N= 256;

    l = 64;%length of seq

    t=sqrt(-1);

    n = 0:1:l-1;

    eps=power(10,-10);

    %frequency components

    F1 = 5000;%in Hz

    F2 = 2000;

    F3= 2500;

    Fs = 3*F1; %sampling freq

    f1 = F1/Fs;

    f2= F2/Fs;f3= F3/Fs;

    %given signal

    x = sin(2*pi*f1*n)+sin(2*pi*f2*n)+sin(2*pi*f3*n);

    f= (-Fs/2):(Fs/N):(Fs/2-(Fs/N));%range

    x1= [x zeros(1,N-l)];

    y1= fftshift(fft(x1));

    m1= abs(y1);

    w= (hann(l))';

  • 8/9/2019 A Pod is at Ion

    12/13

    x2= x.*w;

    x2= [x2 zeros(1,N-l)];

    y2= fftshift(fft(x2));

    m2= abs(y2);

    I1=real(y1);

    I2=real(y2);

    Q1=imag(y1);

    Q2=imag(y2);

    I=zeros(1,N);

    Q=zeros(1,N);

    %I-Q indepenent comparison

    for k = 1:N

    if(sign(I1(k))~=sign(I2(k)))

    I(k)=0;

    else

    I(k)=min(abs(I1(k)),abs(I2(k)));end

    if(sign(Q1(k))~=sign(Q2(k)))

    Q(k)=0;

    else

    Q(k)=min(abs(Q1(k)),abs(Q2(k)));

    end

    end

    mag=I+t*Q;

    figure(3);plot(f,abs(mag),'r');grid on;title('CDA'); xlabel('frequency in Hz');

    Program for Spatially Variant Apodisation

    clc;clear all

    close allN= 256;l = 64; %length of seqt=sqrt(-1);n = 0:1:l-1;

    %frequency componentsF1 = 2000;F2 = 5000;

    F3= 3000;

    Fs = 4*F2; %sampling freqf1 = F1/Fs;f2 = F2/Fs;f3= F3/Fs;

    xn = sin(2*pi*f1*n)+sin(2*pi*f2*n);

    f= (-Fs/2):(Fs/N):(Fs/2-(Fs/N));%range

  • 8/9/2019 A Pod is at Ion

    13/13

    x1= [xn zeros(1,N-l)];%zero paddingy= fftshift(fft(x1));

    p=sign(angle(y));

    y1=p.*y;I1=real(y);Q1=imag(y1);n1=length(y1);I=zeros(1,N);Q=zeros(1,N);w1= zeros(1,N);w2=zeros(1,N);

    for k=3:N-2w1(k)=-I1(k)/(I1(k-2)+I1(k+2)+eps);

    if(w1(k)