| Author |
 Topic  |
|
|
thuang
USA
Posts |
 Posted - 08/03/2005 : 11:24:03 PM
|
hi, i just started learning origin (version 7.5) for the first time two days ago, and i need to do a whole sequence of modifications to fourier transformed data sets. i've been become familiar with origin's features but there's just some things i can't really figure out by myself so i'm looking for help in this forum. so i have some questions here, and sry if this is a really long post. and uh...one more thing is that i've never heard of fourier and his work until last week and i don't really understand it, but i i think i can get around that if i can smooth out these problems i have with understanding origin. so here goes:
anyway, i have data sets from 2 epr absorbance spectra, one from a doubly spin labeled spectrum, and one from a singly spin labeled function. i integrate these two to get the normal obsorbance spectra, and i figured out that much with origin analysis > calculus > integrate. next i FFT both integrated spectra and i get a fft data set and a graph for each spectrum and i ran into my first problem.
well, the next step is dividing one transformed spectrum into the other, getting a new transformed data set, and i'm not sure how to do that. i tried creating a new worksheet where the x variable was the frequency that is the same for both transformed spectra, and then i created a column for dividing the real parts of both spectra and another column to divide the imaginary parts of the spectra. but i'm not sure if this is how it's done.
1)so my first question is how exactly do i divide one transformed function into the other, to make a new function that is still a fourier transformed one. i'm guessing that this all has to do with math operations with the fft data columns but i'm not sure.
after that i need to fit the real part of this new transformed function to a sum of two gaussians with an offset, and then subtract the offset. i read other posts where people have used multipeak fit, replicas, and user defined functions but i'm not sure which one is best for me. at the same time i think i need to keep the fit like a transformed function, because the next step is to reverse fourier transform the gaussian fit.
2) so the 2nd question is what is the best way to get a fit of the real part of the new transformed function as a sum of 2 gaussians and an offset? i get the impression that user defined is the most flexible way.
3)but then i don't know how to keep the function in a way that allows it to be reverse transformed. i think one of the posts and the manual said i may have to keep the imaginary part. but then how does that work if the fit is based on only the real part? and would the fit give me a new set of data? if it does then it would be simple to subtract the offset from the y column.
after i have this untransformed fit, i need to multiply it by a constant and then integrate it from infinity to -infinity.
4) so would the final function i have after the reverse transform be in the form of a data set? then i can probably just set a new column as the final function times the constant and then integrate that?
5)and then i have one last question about the fft function: in the manual it warned that normalize and shift need to be unchecked in the fft window, or else the resulting functions would have problems when doing a reverse transform. i don't really know what this means, but i did try fft forward with and without shift and normalize and the results were different. but when i fft backwards with or without these 2, the real and imaginary were the same but then the Phi and the power were different. and i have no idea what those mean anyway. what should i do about shift and normalize? do you think it will affect my final calculations?
whew, well that was a really long post. i hope i'm not taking too much of anyone's time with this. actually, i was thinking, origin seems to do EVERYTHING by performing operations on just data sets. or maybe i just know enough yet. wouldn't it be easier if all these steps could be done on functions in the form of equations? or maybe not?
ok, that aaalllll the questions i have. so far. any help would be greatly appreciated, thanks. sry for being such a noob |
|
|
thuang
USA
Posts |
Posted - 08/04/2005 : 4:42:01 PM
|
| i guess my post really was too long. bump? |
 |
|
|
Mike Buess
USA
3037 Posts |
Posted - 08/04/2005 : 6:50:18 PM
|
Yes, a lot of material. Perhaps better to break it up into separate posts. Anyway, here's a start...
1) Presumably you have two FFT worksheets - FFT1 and FFT2 - and you want to divide the Real column of one by the real column of the other. This will do that part...
FFT1_Real /= FFT2_Real;
Alternatively, plot FFT1_Real and use Analysis > Simple Math.
2) Plot FFT1_Real (now divided) and select Analysis > Fit Multi-peaks > Gaussian. If that works fine, but you might need to resort to the Advanced Fitting Tool. (Definitely a separate topic.)
3) Not sure if you need the imaginary since I don't really understand what you're trying to obtain. If you do need it then handle like real...
FFT1_Imag /= FFT2_Imag;
4) Yes, you'll get another dataset.
Mike Buess
Origin WebRing Member |
|
|
|
Leo_Li
China
Posts |
Posted - 08/05/2005 : 08:35:20 AM
|
5) The "Shift" and "Normalize" option for FFT doesn't affect the FFT process at all, but they determine how the transformed data will be presented. To know more, click '?' on the upper-right corner of the dialog to bring up the help window.
For inverse FFT (iFFT), it assume that your Real and Img data are neither shifted nor normalized (on the other hand, iFFT has no way to know how the input data are shifted or normalized), so if you want to perform a symmetric operation, "Shift" and "Normalize" shall be cleared.
Leo
OriginLab
|
|
|
| |
Topic |
|
|
|
manipulating fourier transformed data sets