Bandpass and highpass filter

Hi I have this problem. I have this vecktor z that was given to me. I fourier transform it and get Z.

----------------------------------------------

[Z,nu]=tdftfast(z);

plot(nu, abs(Z));

----------------------------------------------

Plot of Z looks like this: http://imageshack.com/a/img909/5763/5SZS0I.jpg

I now want to filer out all 3 peaks. For the first one I use this code:

----------------------------------------------

[b,a] = butter(10,0.1);

freqz(b,a);

dataOut = filter(b,a,z);

[Z1,nu1]=tdftfast(dataOut);

plot(nu, abs(Z1));

x=ifft(Z1);

----------------------------------------------

And get this plot witch is good: http://imageshack.com/a/img537/2999/Y3sfCD.jpg

But for the next two peaks i get some problems. I use this code for the second peak:

----------------------------------------------

[b,a] = butter(10,[0.2 0.3],'bandpass');

freqz(b,a);

dataOut = filter(b,a,z);

[Z2,nu2]=tdftfast(dataOut);

plot(nu, abs(Z2))

z2=ifft(Z2);

----------------------------------------------

But now when i plot i get this and it is all wrong: http://imageshack.com/a/img538/7708/7awCCq.jpg

It is the same thing with peak number 3. What is it that I do wrong? Don't worry about a, b, dataOut and so on since i do not run the both codes at the same time. How do I bandpass peak nr 2 and highpass peak nr 3?

I also wonder how i find the carrier frequency?

Thanks!

[Star Strider]

I would use the MATLAB fft function instead of the one you are using. Follow the documentation for details on how to determine the various frequencies.

Also, you need to have the sampling rate (in seconds) or sampling frequency (in Hz) to determine the frequencies of your component signals. You also need those data to design your filter correctly.

[qwfds dsafsad]

If i use fft function or not doesn't matter. My problem and the question is how i filter out the two other peaks. The fft function has nothing to do that that or am i wrong?

How i find the sampling rate or sampling frequency is one thing i don't know. That is what i meant with "I also wonder how i find the carrier frequency?". Do you have any suggestions?

[Star Strider]

I suggested the MATLAB fft function because the documentation is clear and concise.

If you do not already know the sampling rate or sampling frequency of your signal, there is probably no way to determine it, unless you also know the frequency of at least one peak in your signal, and your signal has very little noise. Then you can infer it, and possibly guess it with sufficient accuracy to have useful results.

[qwfds dsafsad]

I looked at it now and do i really need to know the sampling frequency? By looking at the pictures i attached in my first post i can find the carrier frequency. 0,25 for peak 2 and 0,45 for peak 3. That must be enough?

[Star Strider]

If you are certain that those are the actual frequencies (in Hz), you can then find their indices in the nu-vector and calculate the sampling frequency, and from it the Nyquist frequency. Again, see the documentation for details.

I have no idea what function you are using to calculate your fft, but it might describe in its documentation how it determines the ‘nu’ values and what they mean. You might be able to get the relevant information from it, if it assumes (or infers) a sampling frequency from your ‘z’ data.