| T O P I C R E V I E W |
| alaudo |
Posted - 02/16/2005 : 5:47:55 PM
Origin Version (Select Help-->About Origin): 6.1
Operating System: Win XP Prof
Hallo, all!
I've discovered another problem (at least a problem for me). At the moment I am testing FFT-transformation (see next post) in Origin 6.1 by producing simple functions as sums of several (up to 5) sinus functions with different periods and trying to distinguish single components in the calculated spectrum.
Whenever I plot a new function I can indicate how many points it must have. Changing this parameter, to me, should make the function finer or sparser -- the more points we have, the finer the function is. Nonetheless, changing this parameter leads to strange distortions of the functions, as if I have changed the period or added another component. Did anyone already have this problem?
Alexander.
|
| 4 L A T E S T R E P L I E S (Newest First) |
| alaudo |
Posted - 02/21/2005 : 11:05:51 AM
I played around with different values and FFT-transformations and found out, that there is no problem. Actually there is a problem, but it is only the problem of undersampling.
So, I withdraw my question. Thanks! |
| alaudo |
Posted - 02/21/2005 : 10:44:49 AM
Hi,
here is the picture I produced. Increasing # of points drastically changes the function .
What's the problem? What I expect is just a sinusoid with the freq 10 Hz. Instead of that I get strange wave-form functions.
Imagine you are right and it is the problem of insufficient sampling rate. In the first case we have 100 points for the signal bw 0 and 10, meaning 10 points per unit. I would expect the points to lie randomly or sequentially bw -1 and 1, but not like here - one period in 10 units.
Maybe I am wrong?
I am producing the picture for FFT transformation of these functions.
Edited by - alaudo on 02/21/2005 10:45:55 AM |
| greg |
Posted - 02/17/2005 : 4:52:12 PM
You should also be aware that Origin's FFT will truncate or pad a dataset with zeroes to achieve 2^N points.
http://www.originlab.com/www/support/resultstech.aspx?ID=570&language=English
|
| Mike Buess |
Posted - 02/17/2005 : 02:45:54 AM
Hi Alexander,quote:
Whenever I plot a new function I can indicate how many points it must have. Changing this parameter, to me, should make the function finer or sparser -- the more points we have, the finer the function is.
The effect on the spectrum depends on how you increased the number of points in the time domain. Say you started with ten periods of a pure sine wave sampled at 2 points per period. Doubling the number of points by adding ten more periods sampled at the same rate will indeed double the resolution by producing twice as many points in the same frequency range. Doubling the points by sampling 4 points per period will double the spectral width without affecting the resolution.
A pure sinusoid sampled at a rate less than two points per period will produce a peak at a frequency that is the difference between the actual frequency and some multiple of the sampling frequency. (This is called aliasing.) The frequency will appear to change if you are undersampling your curve and then changing the sample rate.
Mike Buess
Origin WebRing Member
Edited by - Mike Buess on 02/17/2005 02:48:41 AM |
