Many small signals are routinely detected in the collected data files. These signals arise from interfering signal sources, nearby computers, or are noises generated inside the system.
The key to eliminating these signals before unusual signal reports are posted is to check four characteristics - the intensity profile, the frequency drift rate, the bandwidth, and the modulation.
First, check the suspect signal against Known Interferences (like 60 Hz power line noise and its harmonics, for example) for a match in frequency, duration, and modulation. Also check any location or direction associated with a Known Interference. If the signal does not match a known interference, get answers to the following questions:
1. Does the signal intensity profile follow the antenna reception pattern (about 18 minutes is the 3 db beam width). The signal should be down 3 db nine minutes on either side of the center of the signal due to rotation of the antenna by the earth.
a. The signal is probably terrestrial interference if an identical signal is picked up in another data file collected at dramatically different celestial coordinates (Right Ascension and Declination).
b. If an interference signal is generated locally (i.e., not in the posted data file), it will disappear from its location in time if the same file is examined again.
2. Does the signal Doppler shift with at least sidereal motion (about 6 hz/min for our typical antenna orientation, a 90 degree elevation).
a. The signal is probably terrestrial interference if it has a Doppler shift less than the sidereal motion, or no Doppler shift at all.
b. The Doppler shift will probably appear nearly linear on a deep space object over the 18 minute time window. If the shift is strongly nonlinear the signal is probably from a satellite or some other nearby, rapidly moving source.
3. Is the signal narrow in bandwidth?
a. Only artificial processes produce signals with line widths of 2 hz or less.
4. Does the signal have any detectable modulation?
a. An ET signal will probably be either continuous or repeating to increase S/N.
When signals are detected in data files (especially with a simple file screening tool like SETIEasy), volunteers are encouraged to go through the four-point checklist above with tools like CoolEdit, MathCAD, Matlab, and/or Speakeasy before posting a signal report on one of the servers. The Completion post should briefly describe the tests made on the file and the results, so others with different software packages can apply an assortment of tests if necessary as well. However, if you do not have all of the software listed above (or if you use other signal processing software), do not feel inhibited from making the post. The discussion group format is used precisely to permit people to collaborate on achieving a complete result, each one contributing what they can to the analysis.
Dan discusses some of the interference signals routinely observed in the document below, and analyzes them in CoolEdit and MathCAD.
From: Dan Kennedy
Date: 11/4/00
Time: 9:17:11 AM
I am noticing new artifacts in the files. These are relatively steady signals
not unlike those down near zero. Except these occur in the upper end of
the spectrum.
They are all but invisible using even CoolEdit. However, I have found that viewing the high frequency side is more sensitive if the entire graph is selected. That causes the presentation to reverse the color of the selected region. So instead of being dark, it brightens and signal appear dark.
I see them clearly at 3975, 3905, 3854, 3664, 3545, 3485, 3424 and 3304 Hz. A spectral scan across the entire file using CoolEdit gives this graph. It shows minor glitches corresponding to the 3664 and 3785 lines.
Plotting the values on a frequency graph shows that they are regularly spaced in reference to high-order 60 Hz harmonics.
There is an offset from the exact 60 Hz harmonic. But we have noticed before that the 60 hum line is not exactly 60 Hz. In CoolEdit it looks very close to 60 Hz. But looking closely reveals that it is just a bit higher than 60.
Looking more closely, with MathCAD, I can see the 3664 Hz harmonic and 60 Hz fundamental better.
Applying the more exact detected fundamental frequency of 60.07 Hz and plotting the 55th to 66th harmonics makes a better match with most of the artifact frequencies. The above are done with million point FFTs.
This shows some interesting features. It suggests that there may also be a signal at 3364 Hz. After re-examining the CoolEdit display I can see it here.
The missing signals at 3604 and 3725 are also visible when I know where to look. So that completes the set of harmonic artifacts.
That leaves the strange offsets at 3854 and 3975. The one at 3854 is solid. Yet it is 10 Hz off the expected 64th harmonic at 3844 Hz..
The anomalous signal is vertically centered above. Where I would expect a 3844 Hz harmonic, I find a 3854 Hz artifact. A 4 million point MathCAD FFT (4194304) shows the signal well.
I don't see anything significant in these signals except the unexplained offsets of the 64th and 66th harmonics. Plus it is fun to play with. On to the next file.
- Dan
Return to the ASRG SETI Page