The development of the DFT: Difference between revisions

From Class Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
Line 1: Line 1:
If we have a signal, such as following:
If we have a signal, such as following:

[[Image:DFT1.jpg]]
[[Image:DFT1.jpg|400px]]

How do we put it into computer?
How do we put it into computer?
We can use A/D converter and a low pass filter to sample the signal with wanted instead of from t = -\infty to \infty:
We can use A/D converter and a low pass filter to sample the signal with wanted instead of from t = -\infty to \infty:

[[Image:DFT2.jpg]]
[[Image:DFT2.jpg|1000px]]

But then we have to make it periodic, so we convolve it with impulse function with NT apart to have impulse function in both time and frequency domain:
But then we have to make it periodic, so we convolve it with impulse function with NT apart to have impulse function in both time and frequency domain:

[[Image:DFT3.jpg]]
[[Image:DFT3.jpg|1000px]]
From the equations of final signal in both time and frequency domain, we can see that in the computer we have x(n) and in the frequency domain the areas of the impulse functions is:

<math>\sum_{m=-\infty}^\infty \frac{1} {NT}\sum_{n=0}^{N-1}X(nT)\, e^{-j2\pi\frac{m}{NT}nT} \delta\left(f-\frac{n}{NT}\right)</math>
From the equations of final signal in both time and frequency domain, we can see that in the computer we have x(n) and in the frequency domain:
Which is the definition of the DFT(x(n)).

<math>\sum_{n=0}^{N-1}x(nT)\, e^{-j2\pi\,fnT}\,\cdot\,\frac{1} {NT}\sum_{m=-\infty}^\infty\delta\left(f-\frac{n}{NT}\right)=\sum_{m=-\infty}^\infty \frac{1} {NT}\sum_{n=0}^{N-1}x(nT)\, e^{-j2\pi\frac{m}{NT}nT} \delta\left(f-\frac{n}{NT}\right)</math>

Then the areas of the impulse functions is:

<math>\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}</math>

Which is the definition of the <math>DFT(x(n))</math>.

==Property of the DFT==

*DFT is periodic.
Proof:

<math>X(m+N)=\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{n(m+N)}{N}}=\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}e^{-j2\pi\frac{nN}{N}}=\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}=X(m)</math>

==Inverse DFT==

Let's try to get back x(l) if we have X(m)

<math>X(m)=\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}</math>

Let's try do some trick to this DFT, let's sum it up and with exponents tag along.

<math>\sum_{m=0}^{N-1}X(m)\, e^{j2\pi\frac{lm}{N}} = \sum_{m=0}^{N-1}\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}e^{j2\pi\frac{lm}{N}}=\sum_{n=0}^{N-1}x(n)\sum_{m=0}^{N-1}e^{j2\pi\frac{(l-n)m}{N}}</math>

Note: <math>\sum_{m=0}^{N-1}e^{j2\pi\frac{(l-n)m}{N}}</math> will be N if <math>l=n</math>.

Let <math>e^{j2\pi\frac{(l-n)m}{N}}=r^m</math> then <math>r=e^{j2\pi\frac{(l-n)}{N}}</math>

So we know the sum of <math> r^m=S=1+r+r^2+r^3+...+r^{N-1}</math>

Therefore, we can divide r at both side of equation and get <math>\frac{S-1}{r}=S-r^{N-1}, S(\frac{1}{r}-1)=\frac{1}{r}-r^{N-1}, S=\frac{1-r^N}{1-r}</math>

Think about this, if <math>l\ne\, n, r=e^{j2\pi\frac{(l-n)}{N}}, S =\frac{1-e^{j2\pi(l-n)}}{1-e^{j2\pi\frac{(l-n)}{N}}}=0</math>, since <math>e^{j2\pi(l-n)}=1</math> for any integer of l and n.

So <math>\sum_{m=0}^{N-1}X(m)\, e^{j2\pi\frac{lm}{N}}=\sum_{n=0}^{N-1}x(n)N\delta_{n,l}=Nx(l)</math>

<math>x(l) = \frac{1}{N}\sum_{m=0}^{N-1}X(m)\, e^{j2\pi\frac{lm}{N}}\equiv\, IDFT(X(n)) </math>

Key points to note:

* By doing the DFT, we make the signal periodic in both time domain and frequency domain.

<math> X(m+N)=X(m),\,\, x(n+N)=x(n) </math>

* <math> X(m) \,\,for\,\,\, m>\frac{N}{2}</math> corresponds to <math> X(m-N)</math>

==Approximation of Fourier integral==

We can kind of see the DFT as an approximation to the Fourier integral.

<math>X(m)=\sum_{n=0}^{N-1}x(n)\, e^{-j2\pi\frac{nm}{N}}=\frac{1}{T}\sum_{n=-\frac{N}{2}}^{N-1-\frac{N}{2}}x(n)\, e^{-j2\pi\, nT\frac{m}{NT}}T\cong\frac{1}{T}\int_{-\frac{NT}{2}}^{\frac{NT}{2}}\,x(t)e^{-j2\pi\,tf}\,dt</math>

Revision as of 02:09, 20 November 2008

If we have a signal, such as following:

DFT1.jpg

How do we put it into computer? We can use A/D converter and a low pass filter to sample the signal with wanted instead of from t = -\infty to \infty:

DFT2.jpg

But then we have to make it periodic, so we convolve it with impulse function with NT apart to have impulse function in both time and frequency domain:

DFT3.jpg

From the equations of final signal in both time and frequency domain, we can see that in the computer we have x(n) and in the frequency domain:

Then the areas of the impulse functions is:

Which is the definition of the .

Property of the DFT

  • DFT is periodic.

Proof:

Inverse DFT

Let's try to get back x(l) if we have X(m)

Let's try do some trick to this DFT, let's sum it up and with exponents tag along.

Note: will be N if .

Let then

So we know the sum of

Therefore, we can divide r at both side of equation and get

Think about this, if , since for any integer of l and n.

So

Key points to note:

  • By doing the DFT, we make the signal periodic in both time domain and frequency domain.

  • corresponds to

Approximation of Fourier integral

We can kind of see the DFT as an approximation to the Fourier integral.