- 24th December 2005, 06:38 #1

- Join Date
- May 2004
- Location
- oriental
- Posts
- 494
- Helped
- 5 / 5
- Points
- 4,947
- Level
- 16

## matlab spectrum

Hi all,

I am a Matlab newbie :)

How to get frequency spectrum of a vector data? (something like [3 4 2 10 ...])

I heard FFT is the time-frequency fransform. And I tried fft() function in Matlab.

But it return a complex number. Is it stand for both frequency and phase?

Any suggestions will be appreciated!

Best regards,

Davy

- 24th December 2005, 06:38

- 24th December 2005, 09:42 #2

- Join Date
- Sep 2002
- Posts
- 302
- Helped
- 22 / 22
- Points
- 9,373
- Level
- 23

## matlab frequency spectrum

Hi

Yes, the fft results contain both amplitude and phase information.

To plot only the amplitude try abs( fft( ... ) );

Another useful MATLAB function for psd plotting is**pwelch**.

Regards

3 members found this post helpful.

- 24th December 2005, 10:57 #3

- Join Date
- May 2005
- Posts
- 288
- Helped
- 4 / 4
- Points
- 3,090
- Level
- 13

## frequency spectrum matlab

Hello Circuit_Seller;

In fft() you get fft. But if you want to plot it, you need frequency, that is you must write:

plot( f , fft )

How you can get frequency vector?

1 members found this post helpful.

- 24th December 2005, 11:18 #4

- Join Date
- Sep 2002
- Posts
- 302
- Helped
- 22 / 22
- Points
- 9,373
- Level
- 23

## spectrum matlab

Hi

If N is the FFT length and Fs is sampling frequency then

f = 0 : Fs / N : ( Fs - 1 ) / N;

Regards

2 members found this post helpful.

- 24th December 2005, 12:25 #5

- Join Date
- May 2005
- Posts
- 288
- Helped
- 4 / 4
- Points
- 3,090
- Level
- 13

## matlab frequency plot

Circuit_seller wrote:

If N is the FFT length and Fs is sampling frequency then

f = 0 : Fs / N : ( Fs - 1 ) / N;

Fs / N > ( Fs - 1 ) / N

Can you explain more? How did you take this result?

- 24th December 2005, 18:51 #6

- Join Date
- Aug 2005
- Location
- Salvador-BA-Brazil
- Posts
- 226
- Helped
- 28 / 28
- Points
- 2,906
- Level
- 12

## plot spectrum matlab

Lets try to help:

In order to plot a simple estimation of the spectrum using fft you must pay attention to some details.

Try to increase resolution using zero padding.

Try to avoid spectral leakage, it is easy when you know the spetrum and are just confirming it using fft.

use the command fftshift it takes the center of the signal to origin. eg Yn=abs(fftshift(fft(y,n)))/N where N is the number of points of the signal and n >N is the number of points of the DFT ( n-N is zero padding).

In order to plot the frequency you do:

Having a DFT with n points the vector of frequency is f=[-n/2: (n-1)/2]*Fs/n using this and the fftshift the values in the plot are directly given in Hertz.

I hope have helped

2 members found this post helpful.

- 25th December 2005, 07:44 #7

- Join Date
- Sep 2002
- Posts
- 302
- Helped
- 22 / 22
- Points
- 9,373
- Level
- 23

## frequency plot matlab

Hi

Sorry,

f = 0 : Fs / N : Fs - 1 / N;

- 25th December 2005, 07:44

- 6th January 2006, 18:33 #8

- Join Date
- Jan 2006
- Posts
- 287
- Helped
- 19 / 19
- Points
- 3,089
- Level
- 13

## matlab plot spectrum

type in matlab command window, the following one by one

help plot

help stem

help polar

furthermore, suppose that Z=complex matrix, then

real(Z) , GIVES U ONLY REAL VALVUES,,,

complex values can nicely be plot in polar graph, by plotting:

abs(Z) against angle(Z)...

i think it will help u..

- 17th January 2006, 15:21 #9

- Join Date
- May 2005
- Posts
- 288
- Helped
- 4 / 4
- Points
- 3,090
- Level
- 13

## plot frequency spectrum matlab

Hello;

Circuit_Seller wrote this relation:

f = 0 : Fs / N : Fs - 1 / N

- 18th January 2006, 09:45 #10

- Join Date
- Sep 2002
- Posts
- 302
- Helped
- 22 / 22
- Points
- 9,373
- Level
- 23

## matlab fft spectrum

Hi

When you perform a FFT with length N, [0 Fs] is divided to N frequency bins.

So to produce frequency bins i wrote that expression.

Regards

- 18th January 2006, 09:45

- 10th July 2006, 17:39 #11

- Join Date
- Jul 2004
- Posts
- 53
- Helped
- 7 / 7
- Points
- 2,213
- Level
- 10

## fourier spectrum matlab

I am uploading 2 PDF files on Fourier Transform / Spectrum calculation and display in Matlab. I used it for the lab of DSP.

You can also use the below function for displaying the frequency spectrum in MATLAB. As "N" the number of points for fft increases the spectrum becomes more accurate to the theoratical values.

Here 'signal' is the data vector for which u want to see the frequency spectrum. 't' is the time vector for which the signal exists , 'ts' is the sampling time , 'N' is the number of fourier transform points.

function[freq,mag]=fouriertransform(signal,t,ts,N);

S=fft(signal,N);

CS=[S(N/2+1:N) S(1:N/2)];

freq=[-N/2+1:N/2]/(N*ts);

mag=abs(CS);

figure;

plot(freq,mag);

6 members found this post helpful.

- 10th July 2006, 18:49 #12

- Join Date
- Jul 2006
- Posts
- 117
- Helped
- 8 / 8
- Points
- 2,111
- Level
- 10

## fft spectrum matlab

Hi

Well fft gives a complex function.

To plot absolute value and d phase type following

plot(abs(fft))

plot(angle(fft))

u can also refer to help files

- 12th July 2006, 15:01 #13

- Join Date
- Aug 2005
- Posts
- 15
- Helped
- 6 / 6
- Points
- 1,592
- Level
- 9

## frequency spectrum in matlab

Perhaps that tutorial about FFT can help you

3 members found this post helpful.

- 12th September 2006, 10:11 #14

- Join Date
- May 2005
- Posts
- 27
- Helped
- 0 / 0
- Points
- 1,696
- Level
- 9

## phase spectrum matlab

This is my understanding about DFT. DFT builds the relationship between the samples of signal x(t) and the samples of its FT X(f).

time-domain sampling interval is T, the sample freq is FS=1/T. and freq domain sample interval is f0, so time domain truncation duration is 1/f0=T0, and both sampling number in the t-domain and f-domain are N=T0/T.

so the t-domain data is x(T*n) ,n=[0:N0-1]. data point in t-domain is 0:T:T*(N0-1).

and x(t)'s discrete fourier tranform is fft(T*x(T*n)). and the freq point is 0:f0:FS.

Added after 1 minutes:

This is my understanding about DFT. DFT builds the relationship between the samples of signal x(t) and the samples of its FT X(f).

time-domain sampling interval is T, the sample freq is FS=1/T. and freq domain sample interval is f0, so time domain truncation duration is 1/f0=T0, and both sampling number in the t-domain and f-domain are N=T0/T.

so the t-domain data is x(T*n) ,n=[0:N0-1]. data point in t-domain is 0:T:T*(N0-1).

and x(t)'s discrete fourier tranform is fft(T*x(T*n)). and the freq point is 0:f0:FS.

- 20th August 2007, 11:08 #15

- Join Date
- Apr 2007
- Posts
- 94
- Helped
- 6 / 6
- Points
- 1,342
- Level
- 8

## matlab spectrum plot

a=abs(fft

b=angle(fft)

plot(a)

plot(angle(b))

- 20th August 2007, 14:04 #16

- Join Date
- Jun 2007
- Posts
- 2
- Helped
- 0 / 0
- Points
- 984
- Level
- 7

## matlab fourier spectrum

in MatLab help is good example for fft,

roughly output spectrum in dB is:

20log10(abs(fft(sin(2*pi*fc*tt))))

in this case the amlitude is 2 peak-to-peak that corresponds 10 dBm@50 Ohm for sine signal with amplitude 2Vp-p. MatLab shows power 0dB@fc

Can somebody say how take into account this 50 Ohm output impedance to get proper power?

- 30th September 2007, 11:20 #17
## matlab plot frequency

if you want to plot the frequency and phase response of the vector you have, you can use the function

**freqz**....

- 10th October 2007, 12:07 #18

- Join Date
- Aug 2007
- Posts
- 157
- Helped
- 8 / 8
- Points
- 1,772
- Level
- 9

## spectral plot matlab

select fft box from DSp tool box.

- 31st October 2007, 16:38 #19

- Join Date
- Aug 2005
- Posts
- 39
- Helped
- 5 / 5
- Points
- 1,462
- Level
- 8

## matlab plot frequency spectrum

let s be your signal;

N = NFFT

Fs = sampling freq

y = fft(s,N);

p = abs(y);

f=Fs/N.*(0:(N/2)-1);

plot(f,p(1:N/2));

will give you the power spectrum

1 members found this post helpful.

- 3rd November 2007, 18:27 #20

- Join Date
- Nov 2007
- Posts
- 2
- Helped
- 0 / 0
- Points
- 917
- Level
- 6

## spectrum in matlab

Originally Posted by**davyzhu**

the solution of this problem is that first find the fft then try the command freqz()

that have the arguments fft result,1,1024.