MATLAB Central Newsreader - time and frequency domain

time and frequency domain

beda meda — Sat, 10 May 2008 09:59:05 -0400

Hi,
I need to do this: I want to draw Hamming window in time and
frequency domain. In time domain it is simple, I just write
hamming(256) and plot. But I can't find function to view the
same in frequency domain. You can see what I want at:
http://www.fit.vutbr.cz/~cernocky/speech/opora/zre_opora.pdf
page 38, chapter 5.3.2, pictures on the right.
Thanks for help.

Re: time and frequency domain

Lars — Sat, 10 May 2008 10:38:03 -0400

I don't have the chzech-toolbox installed. Sorry!

Re: time and frequency domain

jay vaughan — Sat, 10 May 2008 16:34:02 -0400

"Lars " <mustermann.klaus.TO.REMOVE@gmx.de> wrote in message
<g03tua$bpi$1@fred.mathworks.com>...
> I don't have the chzech-toolbox installed. Sorry!

Try this.

f1 = zeros(1,512);
f1(128:383) = hamming(256);

f2 = zeros(1,512);
f2(128:383) = 1;

fft_f1 = fftshift(fft(f1,2048));
fft_f2 = fftshift(fft(f2,2048));

subplot(2,2,3); plot(f1);
subplot(2,2,1); plot(f2);
subplot(2,2,4); plot(abs(fft_f1)); xlim([896 1151])
subplot(2,2,2); plot(abs(fft_f2)); xlim([896 1151])

J

Re: time and frequency domain

beda meda — Sat, 10 May 2008 19:29:05 -0400

"jay vaughan" <jvaughan5.nospam@gmail.com> wrote in message
<g04ipq$nbi$1@fred.mathworks.com>...
> "Lars " <mustermann.klaus.TO.REMOVE@gmx.de> wrote in message
> <g03tua$bpi$1@fred.mathworks.com>...
> > I don't have the chzech-toolbox installed. Sorry!
>
> Try this.
>
> f1 = zeros(1,512);
> f1(128:383) = hamming(256);
>
> f2 = zeros(1,512);
> f2(128:383) = 1;
>
> fft_f1 = fftshift(fft(f1,2048));
> fft_f2 = fftshift(fft(f2,2048));
>
> subplot(2,2,3); plot(f1);
> subplot(2,2,1); plot(f2);
> subplot(2,2,4); plot(abs(fft_f1)); xlim([896 1151])
> subplot(2,2,2); plot(abs(fft_f2)); xlim([896 1151])
>
> J

Thank you SO MUCH!

Re: time and frequency domain

Chen Sagiv — Sat, 10 May 2008 19:58:02 -0400

Hi,

It is not important when you plot the abs of the FT, but
shouldn't you use instead:

> f1 = zeros(1,512);
> f1(128:383) = hamming(256);
>
> f2 = zeros(1,512);
> f2(128:383) = 1;
>
> fft_f1 = fftshift(fft(ifftshift(f1),2048));
> fft_f2 = fftshift(fft(ifftshift(f2),2048));
>
> subplot(2,2,3); plot(f1);
> subplot(2,2,1); plot(f2);
> subplot(2,2,4); plot(abs(fft_f1)); xlim([896 1151])
> subplot(2,2,2); plot(abs(fft_f2)); xlim([896 1151])

Best,

Chen
"jay vaughan" <jvaughan5.nospam@gmail.com> wrote in message
<g04ipq$nbi$1@fred.mathworks.com>...
> "Lars " <mustermann.klaus.TO.REMOVE@gmx.de> wrote in
message
> <g03tua$bpi$1@fred.mathworks.com>...
> > I don't have the chzech-toolbox installed. Sorry!
>
> Try this.
>
> f1 = zeros(1,512);
> f1(128:383) = hamming(256);
>
> f2 = zeros(1,512);
> f2(128:383) = 1;
>
> fft_f1 = fftshift(fft(f1,2048));
> fft_f2 = fftshift(fft(f2,2048));
>
> subplot(2,2,3); plot(f1);
> subplot(2,2,1); plot(f2);
> subplot(2,2,4); plot(abs(fft_f1)); xlim([896 1151])
> subplot(2,2,2); plot(abs(fft_f2)); xlim([896 1151])
>
> J

Re: time and frequency domain

Greg Heath — Sat, 10 May 2008 22:47:24 -0400

On May 10, 3:58=A0pm, "Chen Sagiv" <chensagiv...@gmail.com> wrote:
> Hi,
>
> It is not important when you plot the abs of the FT, but
> shouldn't you use instead:
>
> > f1 =3D zeros(1,512);
> > f1(128:383) =3D hamming(256);
>
> > f2 =3D zeros(1,512);
> > f2(128:383) =3D 1;
>
> > fft_f1 =3D fftshift(fft(ifftshift(f1),2048));
> > fft_f2 =3D fftshift(fft(ifftshift(f2),2048));

Absolutely not!

Hope this helps.

Greg

Re: time and frequency domain

Chen Sagiv — Sun, 11 May 2008 11:00:22 -0400

Hi Greg,

No, it doesn't help. Can you explain your very short
response ?

Thanks,

Chen

Greg Heath <heath@alumni.brown.edu> wrote in message
<3c59d7c3-618b-44c2-816a-
b58be9319b4d@24g2000hsh.googlegroups.com>...
> On May 10, 3:58=A0pm, "Chen Sagiv"
<chensagiv...@gmail.com> wrote:
> > Hi,
> >
> > It is not important when you plot the abs of the FT, but
> > shouldn't you use instead:
> >
> > > f1 =3D zeros(1,512);
> > > f1(128:383) =3D hamming(256);
> >
> > > f2 =3D zeros(1,512);
> > > f2(128:383) =3D 1;
> >
> > > fft_f1 =3D fftshift(fft(ifftshift(f1),2048));
> > > fft_f2 =3D fftshift(fft(ifftshift(f2),2048));
>
> Absolutely not!
>
> Hope this helps.
>
> Greg

Re: time and frequency domain

Greg Heath — Sun, 11 May 2008 12:30:42 -0400

Corrected for the unforgivable sin of TOP-POSTING!

On May 11, 7:00=A0am, "Chen Sagiv" <chensagiv...@gmail.com> wrote:
> Greg Heath <he...@alumni.brown.edu> wrote in message
>
> <3c59d7c3-618b-44c2-816a-
> b58be9319...@24g2000hsh.googlegroups.com>...
>
> > On May 10, 3:58=3DA0pm, "Chen Sagiv"
> <chensagiv...@gmail.com> wrote:
> > > Hi,
>
> > > It is not important when you plot the abs of the FT, but
> > > shouldn't you use instead:
>
> > > > f1 =3D3D zeros(1,512);
> > > > f1(128:383) =3D3D hamming(256);
>
> > > > f2 =3D3D zeros(1,512);
> > > > f2(128:383) =3D3D 1;
>
> > > > fft_f1 =3D3D fftshift(fft(ifftshift(f1),2048));
> > > > fft_f2 =3D3D fftshift(fft(ifftshift(f2),2048));
>
> > Absolutely not!
>
> > Hope this helps.

> Hi Greg,
>
> No, it doesn't help. Can you explain your very short
> response ?

There is no point to using ifftshift.

When I ran your code the spectra exhibted hashy
artifacts.

Did you cut what you wrote, paste it into a command
line, and compare the plots with the originals??

The results are obviously incorrect.

Hope this helps.

Greg

Re: time and frequency domain

Chen Sagiv — Sun, 11 May 2008 13:26:03 -0400

Dear Greg,

I refer you to a previous discussion at:

http://www.mathworks.com/matlabcentral/newsreader/view_threa
d/168642#430044

Chen

Greg Heath <heath@alumni.brown.edu> wrote in message
<97e5fc3c-ddbc-4c0c-abd9-
c52e9a3ea38c@s50g2000hsb.googlegroups.com>...
> Corrected for the unforgivable sin of TOP-POSTING!
>
> On May 11, 7:00=A0am, "Chen Sagiv"
<chensagiv...@gmail.com> wrote:
> > Greg Heath <he...@alumni.brown.edu> wrote in message
> >
> > <3c59d7c3-618b-44c2-816a-
> > b58be9319...@24g2000hsh.googlegroups.com>...
> >
> > > On May 10, 3:58=3DA0pm, "Chen Sagiv"
> > <chensagiv...@gmail.com> wrote:
> > > > Hi,
> >
> > > > It is not important when you plot the abs of the
FT, but
> > > > shouldn't you use instead:
> >
> > > > > f1 =3D3D zeros(1,512);
> > > > > f1(128:383) =3D3D hamming(256);
> >
> > > > > f2 =3D3D zeros(1,512);
> > > > > f2(128:383) =3D3D 1;
> >
> > > > > fft_f1 =3D3D fftshift(fft(ifftshift(f1),2048));
> > > > > fft_f2 =3D3D fftshift(fft(ifftshift(f2),2048));
> >
> > > Absolutely not!
> >
> > > Hope this helps.
>
> > Hi Greg,
> >
> > No, it doesn't help. Can you explain your very short
> > response ?
>
> There is no point to using ifftshift.
>
> When I ran your code the spectra exhibted hashy
> artifacts.
>
> Did you cut what you wrote, paste it into a command
> line, and compare the plots with the originals??
>
> The results are obviously incorrect.
>
> Hope this helps.
>
> Greg

Re: time and frequency domain

Chen Sagiv — Sun, 11 May 2008 13:30:22 -0400

Greg Heath <heath@alumni.brown.edu> wrote in message
<97e5fc3c-ddbc-4c0c-abd9-
c52e9a3ea38c@s50g2000hsb.googlegroups.com>...
> Corrected for the unforgivable sin of TOP-POSTING!
>
> On May 11, 7:00=A0am, "Chen Sagiv"
<chensagiv...@gmail.com> wrote:
> > Greg Heath <he...@alumni.brown.edu> wrote in message
> >
> > <3c59d7c3-618b-44c2-816a-
> > b58be9319...@24g2000hsh.googlegroups.com>...
> >
> > > On May 10, 3:58=3DA0pm, "Chen Sagiv"
> > <chensagiv...@gmail.com> wrote:
> > > > Hi,
> >
> > > > It is not important when you plot the abs of the
FT, but
> > > > shouldn't you use instead:
> >
> > > > > f1 =3D3D zeros(1,512);
> > > > > f1(128:383) =3D3D hamming(256);
> >
> > > > > f2 =3D3D zeros(1,512);
> > > > > f2(128:383) =3D3D 1;
> >
> > > > > fft_f1 =3D3D fftshift(fft(ifftshift(f1),2048));
> > > > > fft_f2 =3D3D fftshift(fft(ifftshift(f2),2048));
> >
> > > Absolutely not!
> >
> > > Hope this helps.
>
> > Hi Greg,
> >
> > No, it doesn't help. Can you explain your very short
> > response ?
>
> There is no point to using ifftshift.
>
> When I ran your code the spectra exhibted hashy
> artifacts.
>
> Did you cut what you wrote, paste it into a command
> line, and compare the plots with the originals??
>
> The results are obviously incorrect.
>
> Hope this helps.
>
> Greg

Hi again,

As a new member to the newsgroup, I just realized what you
meant by top-posting. I do hope this will be the worst sin
in my life, but I will try to avoid it from now on.

Best,

Chen

Re: time and frequency domain

Greg Heath — Sun, 11 May 2008 20:47:07 -0400

On May 11, 9:26=A0am, "Chen Sagiv" <chensagiv...@gmail.com> wrote:
> Greg Heath <he...@alumni.brown.edu> wrote in message
>
> <97e5fc3c-ddbc-4c0c-abd9-
> c52e9a3ea...@s50g2000hsb.googlegroups.com>...
>
> > Corrected for the unforgivable sin of TOP-POSTING!
>
> > On May 11, 7:00=3DA0am, "Chen Sagiv"
> <chensagiv...@gmail.com> wrote:
> > > Greg Heath <he...@alumni.brown.edu> wrote in message
>
> > > <3c59d7c3-618b-44c2-816a-
> > > b58be9319...@24g2000hsh.googlegroups.com>...
>
> > > > On May 10, 3:58=3D3DA0pm, "Chen Sagiv"
> > > <chensagiv...@gmail.com> wrote:
> > > > > Hi,
>
> > > > > It is not important when you plot the abs of the
> FT, but
> > > > > shouldn't you use instead:
>
> > > > > > f1 =3D3D3D zeros(1,512);
> > > > > > f1(128:383) =3D3D3D hamming(256);
>
> > > > > > f2 =3D3D3D zeros(1,512);
> > > > > > f2(128:383) =3D3D3D 1;
>
> > > > > > fft_f1 =3D3D3D fftshift(fft(ifftshift(f1),2048));
> > > > > > fft_f2 =3D3D3D fftshift(fft(ifftshift(f2),2048));
>
> > > > Absolutely not!
>
> > > > Hope this helps.
>
> > > Hi Greg,
>
> > > No, it doesn't help. Can you explain your very short
> > > response ?
>
> > There is no point to using ifftshift.
>
> > When I ran your code the spectra exhibted hashy
> > artifacts.
>
> > Did you cut what you wrote, =A0paste it into a command
> > =A0line, and compare the plots with the originals??
>
> > The results are obviously incorrect.

> Dear Greg,
>
> I refer you to a previous discussion at:
>
> http://www.mathworks.com/matlabcentral/newsreader/view_threa
> d/168642#430044

Let me repeat

When I ran your code the spectra exhibted hashy
artifacts. Did you actually run it??

I read the reference thread. What is your point?

Greg

P.S. You may want to take a look at

fftshift([0 1 2 3 4 5 6 7])
ifftshift([0 1 2 3 4 5 6 7])
fftshift([0 1 2 3 4 5 6 7 8])
ifftshift([0 1 2 3 4 5 6 7 8])

Re: time and frequency domain

Chen Sagiv — Mon, 12 May 2008 08:13:03 -0400

Greg Heath <heath@alumni.brown.edu> wrote in message
<0d25ec05-ce98-47fa-868e-
d33bb49dfec8@f36g2000hsa.googlegroups.com>...
> On May 11, 9:26=A0am, "Chen Sagiv"
<chensagiv...@gmail.com> wrote:
> > Greg Heath <he...@alumni.brown.edu> wrote in message
> >
> > <97e5fc3c-ddbc-4c0c-abd9-
> > c52e9a3ea...@s50g2000hsb.googlegroups.com>...
> >
> > > Corrected for the unforgivable sin of TOP-POSTING!
> >
> > > On May 11, 7:00=3DA0am, "Chen Sagiv"
> > <chensagiv...@gmail.com> wrote:
> > > > Greg Heath <he...@alumni.brown.edu> wrote in message
> >
> > > > <3c59d7c3-618b-44c2-816a-
> > > > b58be9319...@24g2000hsh.googlegroups.com>...
> >
> > > > > On May 10, 3:58=3D3DA0pm, "Chen Sagiv"
> > > > <chensagiv...@gmail.com> wrote:
> > > > > > Hi,
> >
> > > > > > It is not important when you plot the abs of the
> > FT, but
> > > > > > shouldn't you use instead:
> >
> > > > > > > f1 =3D3D3D zeros(1,512);
> > > > > > > f1(128:383) =3D3D3D hamming(256);
> >
> > > > > > > f2 =3D3D3D zeros(1,512);
> > > > > > > f2(128:383) =3D3D3D 1;
> >
> > > > > > > fft_f1 =3D3D3D fftshift(fft(ifftshift
(f1),2048));
> > > > > > > fft_f2 =3D3D3D fftshift(fft(ifftshift
(f2),2048));
> >
> > > > > Absolutely not!
> >
> > > > > Hope this helps.
> >
> > > > Hi Greg,
> >
> > > > No, it doesn't help. Can you explain your very short
> > > > response ?
> >
> > > There is no point to using ifftshift.
> >
> > > When I ran your code the spectra exhibted hashy
> > > artifacts.
> >
> > > Did you cut what you wrote, =A0paste it into a command
> > > =A0line, and compare the plots with the originals??
> >
> > > The results are obviously incorrect.
>
> > Dear Greg,
> >
> > I refer you to a previous discussion at:
> >
> >
http://www.mathworks.com/matlabcentral/newsreader/view_threa
> > d/168642#430044
>
> Let me repeat
>
> When I ran your code the spectra exhibted hashy
> artifacts. Did you actually run it??
>
> I read the reference thread. What is your point?
>
> Greg
>
> P.S. You may want to take a look at
>
> fftshift([0 1 2 3 4 5 6 7])
> ifftshift([0 1 2 3 4 5 6 7])
> fftshift([0 1 2 3 4 5 6 7 8])
> ifftshift([0 1 2 3 4 5 6 7 8])

Hi Greg,

I ran the code, and indeed it looks nasty.
My point is however very simple.

If you take the code suggested in the beginning of this all
discussion and display the imaginary rather the abs value
you will get a phase, which you should be aware of if you
account for the real and imaginary values of the FT
separately. This is my point !

Try out:

f1 = zeros(1,512);
f1(128:383) = hamming(256);

f2 = zeros(1,512);
f2(128:383) = 1;

fft_f1 = fftshift(fft(f1,2048));
fft_f2 = fftshift(fft(f2,2048));

subplot(2,2,3); plot(f1);
subplot(2,2,1); plot(f2);
subplot(2,2,4); plot(imag(fft_f1)); xlim([896 1151])
subplot(2,2,2); plot(imag(fft_f2)); xlim([896 1151])

Best,

Chen

Re: time and frequency domain

Greg Heath — Tue, 13 May 2008 00:09:18 -0400

On May 12, 4:13 am, "Chen Sagiv" <chensagiv...@gmail.com> wrote:
> -----SNIP
> If you take the code suggested in the beginning of this all
> discussion and display the imaginary rather the abs value
> you will get a phase, which you should be aware of if you
> account for the real and imaginary values of the FT
> separately. This is my point !

There are two impotant points:

1. The difference between the fft of x1(t) and
   x2(t) = x(t-t0) is a complex multiplier linear
   phase shift exp(-2*pi*f*t0). Therefore, if
   X1 = fft(x1) is real, X2 = fft(x2) will be
   complex.
2. The fft of a zero padded version of x will also
   be complex.

close all, clear all, clc, k=0

x0 = ones(1,16);
x1 = [ones(1,16) zeros(1,16)];
x2 = [zeros(1,8) ones(1,16) zeros(1,8)];
x3 = [zeros(1,16) ones(1,16)];

k=k+1,figure(k)

subplot(2,2,1); plot(x0); axis([-5 35 -0.1 1.1])
subplot(2,2,2); plot(x1); axis([-5 35 -0.1 1.1])
subplot(2,2,3); plot(x2); axis([-5 35 -0.1 1.1])
subplot(2,2,4); plot(x3); axis([-5 35 -0.1 1.1])

X0 = fftshift(fft(x0));
X1 = fftshift(fft(x1));
X2 = fftshift(fft(x2));
X3 = fftshift(fft(x3));

k=k+1,figure(k)
subplot(2,2,1); plot(imag(X0)); axis([-5 35 -11 11])
subplot(2,2,2); plot(imag(X1)); axis([-5 35 -11 11])
subplot(2,2,3); plot(imag(X2)); axis([-5 35 -11 11])
subplot(2,2,4); plot(imag(X3)); axis([-5 35 -11 11])

Hope this helps,

Greg

Re: time and frequency domain

Chen Sagiv — Wed, 14 May 2008 06:27:02 -0400

Greg Heath <heath@alumni.brown.edu> wrote in message
<08b54431-1628-4ae0-9506-
1ed9fe6469c3@f36g2000hsa.googlegroups.com>...
> On May 12, 4:13 am, "Chen Sagiv" <chensagiv...@gmail.com>
wrote:
> > -----SNIP
> > If you take the code suggested in the beginning of this
all
> > discussion and display the imaginary rather the abs
value
> > you will get a phase, which you should be aware of if
you
> > account for the real and imaginary values of the FT
> > separately. This is my point !
>
> There are two impotant points:
>
> 1. The difference between the fft of x1(t) and
> x2(t) = x(t-t0) is a complex multiplier linear
> phase shift exp(-2*pi*f*t0). Therefore, if
> X1 = fft(x1) is real, X2 = fft(x2) will be
> complex.
> 2. The fft of a zero padded version of x will also
> be complex.
>
> close all, clear all, clc, k=0
>
> x0 = ones(1,16);
> x1 = [ones(1,16) zeros(1,16)];
> x2 = [zeros(1,8) ones(1,16) zeros(1,8)];
> x3 = [zeros(1,16) ones(1,16)];
>
> k=k+1,figure(k)
>
> subplot(2,2,1); plot(x0); axis([-5 35 -0.1 1.1])
> subplot(2,2,2); plot(x1); axis([-5 35 -0.1 1.1])
> subplot(2,2,3); plot(x2); axis([-5 35 -0.1 1.1])
> subplot(2,2,4); plot(x3); axis([-5 35 -0.1 1.1])
>
> X0 = fftshift(fft(x0));
> X1 = fftshift(fft(x1));
> X2 = fftshift(fft(x2));
> X3 = fftshift(fft(x3));
>
> k=k+1,figure(k)
> subplot(2,2,1); plot(imag(X0)); axis([-5 35 -11 11])
> subplot(2,2,2); plot(imag(X1)); axis([-5 35 -11 11])
> subplot(2,2,3); plot(imag(X2)); axis([-5 35 -11 11])
> subplot(2,2,4); plot(imag(X3)); axis([-5 35 -11 11])
>
> Hope this helps,
>
> Greg
>

Hi Greg,

Thanks for the example.

Chen