How to find the frequency from a non-uniform time data series while doing FFT?

Hi...

I have a discrete meteorological data set of age versus O isotope data. I would like to find the frequency-amplitude spectrum using fft in matlab. I am not able to determine the corresponding frequency values to plot the spectrum. It will be of great help if someone can provide a clue.

Thanks

FFTs really don't like data sampled irregularly in X. You may be able to do a huge FFT with regular X, but many of the corresponding Ys set to zero. Either that or crank out an old style DFT. I had a project where my X's were logarithmic. I never found a shortcut and had to go the DFT route.

thank you very much...can I perform the DFT directly in matlab....I am new to matlab as well as in the FFT analysis...please guide me

I am afraid I know very little of Matlab. The project I did the DFT work for was about 15 years ago using Fortran. There should be some Matlab wizards out there that could help you.

thank you sir

Hi Jeet_rio,

Assume you have N samples at regular intervals T.
If you obtain their DFT, the separation in frequency between bins is 1/(NT).
For example: suppose you have N=256 samples taken at intervals of T=10 seconds. The total duration of the register is NT=2560 seconds.
The interval between bins is the inverse of that, i.e. 1/2560 Hz.
So, the first value of the DTF (or FFT) corresponds to frequency 0 (mean value).
The following value correspons to the component of frequency of (1/2560)Hz, the folowing to (2/2560)Hz, and so on.
Note that FFT is just an efficient way for carry on the calculation of DFT.
Regards

Z



P.D.: Sorry. I missed the part of "non-uniform", thiking that the issue was about the relationsgip between time axis scale and frequency axis scale.

I'm decent with Matlab, but by no means an expert. Here are some articles I turned up on the web regarding non-uniform time discrete fourier transforms, and some interpolation techniques (guy gets data from tide guages, might be similar enough concept to use his implementation to fill in the gaps on your dataset).

NUFFT PPT slides

**broken link removed**

thank you...thanks a lot...I will try and let you know