I while back, you guys helped me get my FFT working, along with computing the power spectrum. I have been busy with all sorts of things, but I finally got around to finishing up the core functionality. I can display the power spectrum of any of the data files (with pretty colors!) , zoom in, and check for very simple signals.
The zooming is slightly buggy with color normalization, but I can fix that later.
For signal detection, it just looks for simple clock cycles (peaks at regular intervals). I know this is incredibly simplified for signal detection, so do you guys have any suggestions for more detection techniques (preferrably ones that are simple to code)?
Thanks Jason. Looks interesting. I will run it by the scientists at the SETI Institute for their comments.
I like the color scheme on analysis3! some easy to read titles would also be useful to the viewer. name of target, date, time, center frequency, frequency resolution of the smallest channel at every zoom level.
Thanks for the replies guys.
As for showing more information, I'm working on it. I just wanted to get the core application working before making it pretty.
Thanks for your contributed images of your analysis. Looking for simple sinusoids is the basis of many SETI observations.
Usually, we are most interested in signals that "drift" with time, in an earth-based reference frame. The reason is that the Earth rotates, and we don't ordinarily expect that ET knows the location of our array on the earth. Anyway, why would they choose the ATA over other observatories? Hence we expect to see effects of acceleration caused by earth rotation (see http://setiquest.org/forum/topic/baudline-analysis-voyager#comment-1843 for a recent posting), which over short periods of time causes the signal to slant across the waterfall with a slope in the range of 0.01 Hz/s to as much as 10 Hz/s, depending on the observation frequency, position on earth, position of source in sky, acceleration at source, etc.
Signals that stand vertically in the waterfall are not accelerating at all w.r.t. the array on earth. These kinds of signals are usually labeled radio frequency interference and discarded without further inspection. If you can examine your plots for signals that are slanted lines, or possibly, not even straight lines, these are the most interesting signals.
Some wobbly signals look like a random walk in frequency (http://en.wikipedia.org/wiki/Random_walk -- scoll down and you'll see multidimensional random walks that look like some of our stranger signals). We interpret these as evidence 1) most likely resulting from an unstable oscillator (clock) on earth, or 2) signals that have been interfered with during travel to earth by the interstellar medium (ISM).
The ISM and interplanetary medium (IPM) both cause a loss of coherence, leading to a broadening and speckle-like response in the waterfall. Broadening up to 0.01 Hz is expected just due to the interplanetary medium (solar wind), and are even worse if we observe close to the sun. We can't predict the effects of the ISM very well. The SonATA detector is set to keep signals broadened by no more than 10 Hz. Symaski62 has bird-dogged some interesting signals in setiQuest waterfalls (see http://setiquest.org/forum/topic/tau-ceti-1420-mhz#comment-1842).
Take a look at the results of your analysis and see how they compare to published waterfalls, would be one suggestion. Also, do you see things other people don't? Especially, do you see signals that are not simple straight lines (e.g. the face of Elvis -- no, just kidding!).
I have been trying make something to highlight weirdness in the noise but I really have no clue what I am doing. Basically I just generate a bunch of pictures and see if anything catches my eye on the bottom that I can't see on the top.
Here is an example with obvious signals and two things that barely catch my eye for whatever reason.
it will help everyone if you label images with information about what the source is - datafile, target, frequency etc. etc. finding a random waterfall image among the hundreds of thousands (or more) in our datasets is a difficult chore.
it is from the lagrange-4-3991 data set. I was just using that picture as an example of what I was doing. It isn't anything that anyone should really look into. More of a here is what I am doing tell me why its stupid kind of thing.