PSN-L Email List Message

Subject: FFT Bin magnitude
From: "Randall Pratt" rpratt@.............
Date: Sun, 8 Jul 2012 13:19:25 -0500


Hi,

 

I was looking for aliasing in my system and have questions.  My method is to
compare 2 fft results from different sample rates of the same data.  I used
240 sps and an fft of 16384 length.  Then I used every 16th sample of the
same dataset and put these points into a 1024 fft.  Therefore the bin width
of both fft computations is .0146hz and the first 512 bins of both will
cover the freq range 0 to 7.5hz.  Dividing the magnitudes bin by bin should
then show points where a frequency in the slower sampled data exists out of
proportion to the faster sample rate and indicate an alias.  The scheme
appears to work quite well and I have a program working that will read a
..psn file and list results. The next step would be to compute and list the
possible offending frequencies.

 

Now the question is this.  Why if I use the same data, the same bin size and
1/16th of the points in 1/16 of the total number of bins covering the same
frequencies do the fft bin magnitudes differ by approximately a factor of
16?  Shouldn't the same amplitude per freq bin exist in both computations to
give a valid power computation?  

 

Randy

 












Hi,

 

I was looking for aliasing in my system and have = questions.  My method is to compare 2 fft results from different sample rates of the = same data.  I used 240 sps and an fft of 16384 length.  Then I used = every 16th sample of the same dataset and put these points into a = 1024 fft.  Therefore the bin width of both fft computations is .0146hz and the = first 512 bins of both will cover the freq range 0 to 7.5hz.  Dividing the magnitudes bin by bin should then show points where a frequency in the = slower sampled data exists out of proportion to the faster sample rate and = indicate an alias.  The scheme appears to work quite well and I have a program = working that will read a .psn file and list results. The next step would be to = compute and list the possible offending = frequencies.

 

Now the question is this.  Why if I use the same = data, the same bin size and 1/16th of the points in 1/16 of the = total number of bins covering the same frequencies do the fft bin magnitudes = differ by approximately a factor of 16?  Shouldn’t the same = amplitude per freq bin exist in both computations to give a valid power = computation? 

 

Randy

 


[ Top ] [ Back ] [ Home Page ]