Spectral analysis of signal and subsets.

Hello,

I had a question regarding the spectral output I get when looking at a signal. Suppose I have a signal u (~12000 data points), and I split it into two parts u1 and u2,after running it through a filter to remove the component within a certain frequency range.

u = u1 +u2;

Now, I run the individual signals u, u1 and u2, through a pwelch command after padding it to nearest power of two. I use the default values of pwelch (8 windows, 50% overlap). fs=13.88 Hz, the sampling rate.

[pxx_u,Fu]=pwelch(u,[],[],[],fs); [pxx_u1,Fu1]=pwelch(u,[],[],[],fs); [pxx_u2,Fu2]=pwelch(u,[],[],[],fs);

When I do this,and compare all the spectral outputs, I see a strange thing happening at lower frequencies. The energy in the total signal(u) in a few locations is slightly lower than the component u1. How can this be possible? Is there an explanation to this, or is something wrong with my spectral calculations? Any inputs will be appreciated. I have attached the plot of the spectra from the 3 signals below.

Thanks, Nathan

Answer 1

[Nathan P]

Looks like the link didn't show up:

Here is the image link,

http://imgur.com/DKq4X7P&9XgVV8Y#0

Thanks, Nathan