Power spectrum of short signals

  • Thread starter tct212
  • Start date
  • #1
2
0
I am looking for a way to calculate the power spectrum of short signals of variable duration (between 200 and 500 ms). Standard methods seem wildly inaccurate. Are there any tricks for short segments?

Thanks anyone for a response.
 

Answers and Replies

  • #2
Dr Transport
Science Advisor
Gold Member
2,439
582
If you have access to Matlab, you should be able to take the FFt of the signal, square it and get your power spectrum. Do a google search for Power Spectrum and Matlab and you'll get a load of hits.
 
  • #3
2
0
For short signals?

Thank you Dr Transport. Yes I do have access to Matlab and FFT works great for segments longer than 500 ms. When the segments are short, fft does a very bad job. I am looking for some improved algorithm that may be better optimized for short segments. Any insights would be appreciated.
 
  • #4
f95toli
Science Advisor
Gold Member
3,085
577
The length of the segment in ms does not matter. What matters is the length of the segment relative to the longest period of the signal of interest (i.e. the bandwidth of the signal).
You need at sample at least one entirely cycle of each frequency component you are trying to extract in order to get sensible results; my guess would be that you are hitting that "limit" when you go below 500 ms.
 

Related Threads on Power spectrum of short signals

  • Last Post
Replies
2
Views
2K
  • Last Post
Replies
3
Views
12K
Replies
3
Views
4K
Replies
6
Views
1K
  • Last Post
Replies
6
Views
2K
  • Last Post
Replies
3
Views
48K
  • Last Post
Replies
18
Views
6K
  • Last Post
Replies
1
Views
2K
  • Last Post
Replies
10
Views
2K
  • Last Post
Replies
16
Views
7K
Top