Help, spatial frequency content analysis results

[fvnktion]

I have been working on a project analyzing spatial frequency content of different laser scanned blanket wrinkle configurations. I do not understand the results that i am getting. I am getting a sinc2 function with amplitude shifts in the fundamental frequencies as i take the fft ( power spectral density) of discrete cross sections in the blanket, see link for plots. The only difference, it seems, between different cross section locations in the blanket is the amplitude of different fundamental frequencies of the sinc2 seen in my matlab PSD calculations.

-All analysis has been done using matlabs fft function, multiplying by the conjugate and scaling the frequency axis. I have used 1e3 pt. FFT, 10e3 pt. FFT for maximum clarity of content.

-Initial laser scan data was run through an interpolating smoothing function and a mesh was created using matlab. I then created a smoothed point cloud file and took discrete 2d cross sections for fft (PSD) calculation.

-To validify my matlab algorithm, I have added noise to a modulated sine wave and the same fft fucntion seems to pull out the correct spectral content, no matter the amount of noise that i add.

Does this sinc2 looking result seem realistic to any of you?? Do you have any interpretations of why i might be getting this?? It would make sense if the majority the initial data resembled a trianlgle function, but it does not seem to, to me. Thank you for your responses.

Link to my resulting plots and input plots:

http://www.geocities.com/fvnktion/psdStuff.htm

Michael

 

[logbook]

I tried the link but couldn't see the pictures. The page has rectangular place holders but the images wouldn't display :-(

 

[logbook]

My mistake. Looks like the pictures are really big and my slow dialup link just didn't display anything.

 

[logbook]

How big is the fundamental? It looks like there is a large spike off the top of the graph. What you are describing as a sinc^2 function looks like side-lobes on the FFT.

Google for "spectral leakage".

Using the von Hann window function reduces the spectral leakage and your side-lobes are lower. If you simulate with a non-cohereent sinewave you will see the same thing; use a sine wave that does not have an integer number of cycles across the FFT sample interval.

Also note that FFTs are best done at powers of two length so use 1024 points not 1000.