How do I eliminate large peaks using an FFT function?

Hi,

I have a small data set and wanted to try to use a FFT function to smooth out some of the larger peaks in the data curve. The data is of an increasing temperature ramp (about 900 - 1000C) and there's a linear trend. I realize this may not be the best method to smooth my data, but I wanted to try to do this with a FFT and see what comes out.

The data is from 1 second measurents so Fs = 1. My code is below. As it is right now, I get a choppier plot after I use the "fix" function. I think removes the small fluctuations and leaves me with the more defined signals (kind of the opposite of what i want).

if true
  % 
A = importdata('P:\TAT_1_FFT_test.dat');
Fs = 1;
dt = 1;
L = size(A,1);
t = 1:dt:L;
NFFT = nextpow2(L);
n=L/2;
ti = 1:1:2^NFFT;
Y = fft(A,2^NFFT);
amp_spec = abs(Y)/n;
subplot(2,1,1)
plot(t,A);  grid on
xlabel('Time (s)')
ylabel('Temperature (C)')
subplot(2,1,2)
freq= (0:(2^NFFT - 1))/(2*n*dt);
plot(freq, amp_spec(1:(2^NFFT)); grid on
xlabel('Frequency (Hz)')
ylabel('Amplitude')
fY = fix(Y/100)*100;
ifY = ifft(fY);
cy = real(ifY);
figure
subplot(2,1,1)
plot(t,A);  grid on
xlabel('Time (s)')
ylabel('Temperature (C)')
subplot(2,1,2)
plot(ti(1:L),cy(1:L)); grid on
axis
xlabel('Time (s)');
ylabel('Temperature (C)');  
end

I'm wondering if I need a high passs filter. I've been trying to look into them, but I'm not sure how to customize them and what I should look at to customizq them.

Any help is appreciated!

Thanks,

Daniel

Answer 1

[Image Analyst]

No, a high pass filter will let high frequency signals pass (survive). Spikes and peaks are high frequency signals and a high pass filter won't affect them much. You can to blur them to smooth them, so you want a low pass filter. Zero out the high frequencies and you should see your signal look smoother because the high frequencies (spikes, peaks) will be filtered out. Please post a screenshot of your signal.