How to see the frequencies present in a measured signal.

Hello,

I measured a pulse signal sampled with a frequency of 64Hz and want to see the fre-quencies responsible for the signal. Can some explain how to plot the power spectrum using FFT function to see the fre-quecy components responsible for the pulse.The pulse data of 10secs duration is at-tached below.

[dpb]

There's a specific example at the end of

doc fft

if memory serves...

[Gova ReDDy]

But I don't understand how to use that fft to my signal.

Can someone please explain me hwo to do it.

[Joseph Cheng]

the documentation for fft has examples of how to do it. http://www.mathworks.com/help/matlab/ref/fft.html

The first example can be adapted with the information you gave above.

[Gova ReDDy]

Yes,I tried like this and resultant was

a1=data(1,1:640);
 plot(a1)
 a1_fft = abs(fft(a1));
a1_fft = a1_fft(1:length(a1)/2);
plot(a1_fft);

But from the output pulse signal(posted in the above) it can be understood that it only has 1Hz frequency component but the FFT output shows 1Hz components in larger and also other frequencies of 11 ,21,25,31Hz.

Can someone explain why the other frequency components are also in larger magnitude.

Answer 1

[Image Analyst]

What make you think a square wave has only one frequency? A sine wave does , but a square wave is made up of an infinite number of sine waves . Hence you get power at multiple frequencies. The FFT of a single square pulse is a sinc function, but when you convolve with a comb function to get a square wave train, you'll get tons of sinc's all adding together. Here are some links with basic info on FFT theory:

Steve's Blog

http://www.cv.nrao.edu/course/astr534/FourierTransforms.html

David Young's File Exchange

[Joseph Cheng]

Also by just looking at the sample plot you can see not all the square waves are of same duty cycle.

[Joseph Cheng]

Also you may want to plot the x axis as Fs*time

[Gova ReDDy]

Yes,I want to have the x-axis in terms of time. Can I know how to do that for the above data provided.

[Image Analyst]

Who's to say it's not time? It's just plotted in elements, so each element could correspond to a sliver of time. You have 10 seconds covering 640 elements, so that means each unit on the x axis is 1/64 second. Just say x = 1:(1/64):10 and plot(x,a1) that way. The x axis for the second plot is in terms of spatial frequency, not time.

[Gova ReDDy]

a1=load('a1.mat');
plot(a1.a1);
a2_fft = abs(fft(a2));
a2_fft = a2_fft(1:length(a2)/2);
for x=1:1:length(a2_fft)
t(x)=x/64;
end
plot(t, a2_fft);
xlabel('Frequency(Hz)');
ylabel('Magintude(dB)');

I used like this and got he result as shown here

Now, I only see the first peak frequency as 0.015 and second as 0.17Hz but the origianl signal should contain 1Hz as first peak freuqnecy.

I don't what is wrong with my implemenataion.Can somoeone please help me out in plotting FFT of 'a1' signal to show the Single-Sided Amplitude Spectrum of a1 over the scale of 10Hz. Thanks

[Gova ReDDy]

Hello, can somone explain what is wrong with my implementation of FFT to show the frequencis upto 10Hz.

[dpb]

What was your sampling frequency and length? The relationship is

df=1/T

where T is n*dt and n is number of samples and dt the sample rate. From that you can set the frequency axis appropriately for your sample case.

Again, I commend the example at

doc fft

to your attention -- it has exactly the right steps illustrated for a sample case; simply plug your numbers in in the appropriate places. The only difference is they chose to use the power of 2 length FFT, but the same is true if you just use the signal length.

[Gova ReDDy]

I am confused.Can you please help me out to plot this.