SETI Receivers: How Should We Scan?

SETI League Technical Manual -- Systems

SETI Receivers: How Should We Scan?

Several of our members have asked (and doubtless others have wondered) about the time dimension of sweeping the SETI water hole. The underlying reason for recruiting as many as 5,000 amateur SETI observers worldwide is that it makes it easy to divide up the whole sky. But as more and more SETI League members gear up for our Project Argus sky survey, the question of how to divide up the frequency spectrum begs our attention. Let us assume for the sake of argument that we're attempting to scan the entire Water Hole (1420 to 1660 MHz), at a nominal resolution of 10 Hz. Let's further assume we're using the most popular amateur SETI receiver to date, the Icom model 7000 series scanning receiver. If we can establish a 12.5 kHz IF and audio bandwidth (this might require some receiver modifications), that receiver can be programmed to scan in 12.5 kHz steps. We know that a SoundBlaster type sound card in a PC, with appropriate software, can easily analyze a 12.5 kHz chunk of audio spectrum. Dividing the 240 MHz wide water hole by 12.5 kHz, we can see that our receiver needs to scan 19,200 channels.

OK, but where does the 10 Hz resolution come in? Our DSP software is performing, let us say, a 1024 point Fast Fourier Transform (FFT). That means it divides the audio band into 1024 individual bins. The width of each of these bins is: (12.5 kHz) / (1024) = 12.2 Hz, pretty close to our resolution goal. So our actual number of 12.2 Hz-wide bins surveyed is: (19,200 channels) x (1024 bins / channel) = just under 20 million!

Now there is the problem of how long it should take to sweep thru all 19,200 channels. That is, what is the minimum dwell time required at each frequency step, in order to perform the required digital signal processing? Unfortunately, there's no simple answer. Signal processing time is a function of the computer used (CPU, clock speed, memory, hard disk access time, etc) as well as the DSP software chosen, and bin resolution selected. For PCs (as opposed to high-end dedicated DSP boxes) minimum processing time is likely to be in the seconds. But it doesn't necessarily pay to go with the minimum time. Since sensitivity improves by the square root of integration time, it makes sense to dwell in each channel for minutes.

A practical maximum dwell time per channel is the longest time a celestial object remains within the beamwidth of the antenna. If the antenna is fixed (operated in drift-scan mode) this value is likely to be ten minutes or so for a typical TVRO dish. Which means we really don't want to scan at all, but rather sweep out a 360 degree swath of sky at one channel one day, the next channel the following day. For 19,200 channels, each 12.5 kHz wide, this means one complete sweep takes 19,200 days. This approach makes the search of the entire water hole take 56.2 years. At this rate, I doubt that many of our members will live long enough to make it all the way from the hydrogen to the hydroxyl line! But by the time we've gotten much past hydrogen, we can expect new technology (affordable dedicated DSP boxes, for example) and new search strategies to emerge. Hardware cost keeps coming down, and computer power seems to be doubling every year. If we dare extrapolate this trend, it suggests that within a decade we'll have seen a 1,000-fold improvement in DSP technology. Such improvement would let us cover the entire water hole, for a given swath of sky, in just a matter of days.

So I guess the bottom line is, let's start out crude, and become more elegant with the passing years. But let's start now; we wouldn't want to miss The Call.

email the
Webmaster

Top of Page