Today, I finished up the IDL
assignment for the tutorials this week. The assignment was to
create
a program
that would take a function that was a combination of two sinusoidal
functions with varying frequencies and amplitudes with some added
random background
noise and to use Fourier transforms to produce the power spectrum and
identify the two frequencies. The amplitudes were 5 and 2 (chosen
arbitrarily by me) with the random noise varying between about -20 and
+20, Even with the noise four times larger than the first
amplitude,
the FFT (Fast Fourier Transform) found its frequency with at least a
99% accuracy. However, typically (though not in this image),
because the noise was about 10 times as large as the second amplitude,
it only found its frequency at a 90% confidence level. The 90%
level also returned some false positives, which were returned
at the same frequencies every time the program was run, which is
actually kind of
funny and makes me question just how random IDL's random number
generator is. This also proves just how useful the Fourier
transforms and power spectra are for finding
periodicity in seemingly cluttered and useless signals.
 |
Contact
Information
Christopher Daly
Solar REU Program
Physics Department, EPS 264
Bozeman, MT 59715
email: cjd37@njit.edu
telephone: (201)-956-1198