| Author |
 Topic  |
|
|
hjbb
Taiwan
4 Posts |
 Posted - 03/27/2011 : 01:54:05 AM
|
Origin Ver. and Service Release (Select Help-->About Origin): 8Pro
Operating System:Win7
I am doing the FFT of one sinusoidal damped signal. And I expect there will be "sidelobes" at frequency spectrum because with retangular window used.
However, the result is not the same as I expect.
Why?? |
|
|
Penn
China
644 Posts |
Posted - 03/29/2011 : 04:00:39 AM
|
Hi,
Would you please send your data to us? And then we can look further into it. You can follow the instruction in this page to send us a file.
Penn |
 |
|
|
hjbb
Taiwan
4 Posts |
Posted - 03/29/2011 : 06:30:44 AM
|
Dear Penn:
Thanks for your kindly reply.
But I just use the signal like exp(-t/t0)*cos(2*pi*0.01*t) to do the FFT.
And I expect there will be sidelobes in spectrum because finite data and retangular window used.
Maybe I misunderstanding the signal processing about that.
And thanks once again for your opinion.:)
|
|
|
|
Penn
China
644 Posts |
Posted - 03/29/2011 : 11:18:29 PM
|
Hi,
The side lobes are decided by your data. Maybe you can try our built-in data to see the side lobes. Just import the data "Signal with Discrete Frequencies.dat" located in "<Origin Installation Directory>\Samples\Signal Processing\" folder. Then perform FFT on this data. You can obviously see the side lobes.
For the signal like exp(-t/t0)*cos(2*pi*0.01*t), I have a try too. In column A, I fill row number from 1 to 1000. In column B, the first 300 rows are calculated by the expression above, where t0=200, and the left rows are padded with zero. Then perform FFT on column B, you can also see side lobes (maybe need to zoom in the graph).
Penn |
Edited by - Penn on 03/29/2011 11:18:55 PM |
|
|
|
hjbb
Taiwan
4 Posts |
Posted - 03/31/2011 : 11:47:51 PM
|
Hi Penn,
Thank you for your suggestion. But I think there are sidelobes in FFT spectrum because you "pad the zero".
Without padding the zero, there should be also sidelobes because it is finite data.
Because the data is not infinite, so it means (in the view of signal processing) there is an retangular window multiply your infinite data.
And the FFT result is the convolution of each other and that's why the sidelobes exist.
So without padding zeros , there must be sidelobes.
I am wondering is that any implicit process in the Origin to remove this pheonomena?
Sorry for lengthy discription.
Thanks in advance. :) |
|
|
|
Penn
China
644 Posts |
Posted - 04/01/2011 : 02:53:39 AM
|
Hi,
I have compared the results with several other softwares by using the same data generated from your signal expression, all the results are the same, but cannot see the side lobes. So, I think the result is correct. By the way, you can see our algorithm on how to perform FFT in this page.
Penn |
|
|
|
hjbb
Taiwan
4 Posts |
Posted - 04/01/2011 : 02:59:12 AM
|
Thank you very much.:)
Maybe I misunderstand something.
|
|
|
| |
Topic |
|
FFT window and sidelobes