Signal filtering and FFT

Hello,

I have a real time vibration data for 3.0599 seconds sampled at 9600 Hz.

I need to compute the FFT of the signal and find the peaks.

My frequency of interest is 0 to 600 Hz. I used low pass filter to filter the signal,

Fs = 9600; 
Fn = Fs/2; 
Wp = 600/Fn; 
Ws = 601/Fn; 
Rp = 10; 
Rs = 50; 
[n,Ws] = cheb2ord(Wp,Ws,Rp,Rs); 
[c,b,a] = cheby2(n,Rs,Ws); 
[soslp,glp] = zp2sos(c,b,a); 
figure;
freqz(soslp, 2^16, Fs)
fAW1 = filtfilt(soslp, glp, AW1);
figure;
plot(f1,fAW1)

But I am not satisfied with the FFT results that I am getting from my filtered signal.

Is the procedure I am adopting correct? If not, what am I supposed to do to improve the result?

Thanks in advance.

[Star Strider]

‘Is the procedure I am adopting correct?’

It’s impossible to determine that with the information you posted. Your code appears to be correct. If you use the freqz function with your filter, you can see if it is performing as you want it to.

Note that you apparently have a significant amount of broadband noise, and no frequency-selective filter can eliminate that. You would have to use wavelet denoising, and there is no certainty that even it would do everything you want.

[vanrapa]

Thanks for your reply.

I am looking into wavelet denoising for my signal now. Do you know of any procedure to choose the appropriate wavelet type and level pertaining to my signal?

[Star Strider]

My pleasure.

I would begin with the wdenoise (link) function, and go from there. That might be all you need.