Searches of the Harvard SETI Group

ARECIBO, 1978

Our first high resolution SETI began in 1978 with a search, at the Arecibo 1000-foot dish, of 200 interesting candidate objects. That project used off-line long Fourier transforms (64K-points) to search 1 kHz instantaneous bandwidth (IBW) segments centered on the 21 cm line of neutral hydrogen, with a resolution bandwidth (RBW) of 0.015 Hz (the highest resolution and sensitivity ever achieved in SETI). Here is a block diagram showing how this search was configured. Real-time compensation for Doppler shifts relative to the heliocenter was used, effectively producing a chirped receiver (thus affording excellent immunity to non-chirped radio interference). The tiny bandwidth of the search (1 kHz at 1.4 GHz: a part in 10^6) required a transmitting civilization to precompensate their carrier beacon for their motion relative to our sun: this is a rather restrictive scenario, although it is a task an advanced civilization could accomplish, if they so desired. This search was described in Science, 201, 733-735 (1978).

"SUITCASE SETI," 1981-82

With support from NASA and The Planetary Society, Paul Horowitz spent a year as an NRC postdoctoral fellow at Ames Research Center (1981-82), where he and colleagues from Stanford University and NASA built a high-resolution hardware spectrometer that could handle in real time the kind of signal processing that was used in the earlier Arecibo search. Specifically, this "Suitcase SETI" hardware implemented the FFT in firmware, achieving 64K-channel spectrum analysis (0.03 Hz RBW, 2 kHz IBW) simultaneously in each of two polarizations, along with simple baseline/peak searching and archiving. It included a phase-continuous programmable local oscillator for real-time compensation of site Doppler. Here is the block diagram of suitcase seti. Suitcase SETI travelled to Arecibo in March 1982, where it searched 250 candidates (stellar and other), mostly at the second harmonic of HI, at 2.84 GHz. Once again, RFI rejection was impressive; once again, no confirmed signals. However, as a test of the hardware we looked at the maser source W49(OH), producing a spectrum of such detail that, if plotted at 200 dpi, would stretch across the 1000-foot dish. This search was described (with others) in Icarus, 67, 525-539 (1986).

SENTINEL, 1983-85

With sponsorship by The Planetary Society, and with the permission of NASA, we reconfigured Suitcase SETI as a dedicated search at the Harvard/Smithsonian 84-foot steerable Cassegrain radiotelescope at Harvard, Massachusetts. This search, known as "Sentinel," was the first dedicated high-resolution SETI, covering the northern sky in a transit mode at the 21 cm line. Unlike the earlier targetted searches at Arecibo, we chose an all-sky transit search because the larger beam size (30 arc minutes, compared with 3 arc minutes at Arecibo) corresponds to a full search of the visible sky in about 200 days. Once again, the system was sensitive only to precompensated carrier transmissions to our heliocentric frame. As previously, we found good RFI rejection, but no confirmed signal sources. Sentinel is described in the Icarus article referenced above.

META, 1985-94

Sentinel and its predecessors achieved high resolution at the expense of total frequency coverage (2 kHz for Suitcase SETI and Sentinel, 1 kHz for the earlier off-line search), which required a transmitting civilization to target our star specifically, in order to permit Doppler precompensation to the heliocentric frame. Furthermore, the long required integration time (30 seconds, for a B-tau product of unity) prevented immediate reobservations of interesting candidates, being comparable to the source transit time of 2 minutes. What was needed was a spectrometer of much greater bandwidth, in order to cover contiguous bands centered on magic frequencies as seen in magic inertial rest frames; good choices for the latter are i) the galactic barycenter, ii) the local standard of rest, iii) the heliocenter, and iv) the cosmic microwave background (CMB) rest frame. The uncertainties in our knowledge of these frames was of order 30 km/s, corresponding to +/-150 kHz of Doppler uncertainty at the 21 cm line. We thus embarked on a project to build an 8-million channel spectrometer, to achieve 400 kHz IBW at 0.05 Hz RBW. This was META (Megachannel ExtraTerrestrial Assay), funded by The Planetary Society (through a gift from Steven Spielberg). META was a dedicated all-northern-sky transit search, with successive spectra alternating among the rest frames listed above; these are typically frequency offsets of order +/-1 MHz. As with its predecessors, META used an agile local oscillator to compensate for Doppler chirp caused by site acceleration, which provides a characteristic changing Doppler signature for narrowband signals of extraterrestrial origin. This requires a frequency chirp of order -0.1 Hz/s. During the 20-second integrations the doppler chirp amounts to some 50 frequency channels, thus nicely discriminating against terrestrial radio interference. The dedicated search covered most of the northern sky (-30 degrees to +60 degrees declination) with the Harvard/Smithsonian 26 m equatorial radiotelescope operated in meridian transit mode. Each potential source passed through the antenna beam pattern in approximately 2 minutes, during which the three reference frames are covered once in each antenna polarization. META's hardware, designed in 1983, consisted of GaAsFET low-noise frontends in each polarization, image-reject downconverters with programmable phase-continuous 2nd LO, 7-bit quadrature digitizers, a 144-point channelizing DFT feeding an array of 144 68000-based 64K-point FFTs, and a central workstation of modest performance. Its block diagram shows the signal processing, including the special-purpose array of 144 68000-based processors. META was the first megachannel SETI, and ran for a decade before being replaced by BETA in 1995. In an analysis of 5 years of data, during which 60 trillion channels were searched, we found 37 candidate events exceeding the average detection threshold of 1.7e-23 W/m^2, none of which has been detected upon repeated reobservations. In spite of lack of a confirmed signal, META permits one to set some interesting limits on the prevalence of advanced civilizations that transmit in ways that the search would have detected. For a technical summary, see Astrophysical Journal, 415, 218 (1993); a non-technical version appears in The Planetary Report, 13, 5 (Sept/Oct 1993).

BETA, 1995-

Given the results of META and its predecessors, and the fact that SETI elsewhere has similarly found occasional candidates that have the right characteristics but do not repeat in observations made much later (a characteristic that led to a meeting on "Intermittency in SETI" at the SETI Institute in January 1994), we felt that the next search system should incorporate means for i) rapid and automatic reobservation of candidate events, ii) better discrimination of interference, through a simultaneous 3-beam configuration, and iii) coverage of the full 1.4-1.7 GHz "waterhole" band of frequencies. Thus was born "BETA" (Billion channel ExtraTerrestrial Assay), the current search of the Harvard SETI group, which was switched on in October, 1995. BETA took four years to design and build, with support from The Planetary Society, NASA (grant NAGW-2872), the Bosack/Kruger Foundation, and the Shulsky Foundation. It uses the 26-meter dish with dual (east-west) feedhorns (and a third low-gain terrestrial discone) to feed a 240 million channel Fourier spectrometer (80 million channels of 0.5 Hz resolution and 40 MHz instantaneous bandwidth for each feed) whose outputs feed an array of programmable "feature recognizers." The latter sift through 250 MByte/s of spectral data, seeking distinctive spectral features that transit from the east to the west horn without appearing in the low-gain terrestrial antenna. BETA's contemporary hardware consists of HEMT low-noise frontends, an array of 63 quadrature mixer/digitizers with GPS phase-locked local oscillators, and an array of 63 4-million-channel complex FFT boards feeding a flexible state-machine based feature recognizer/correlator array resident in a set of Pentium motherboards; the latter communicate with a UNIX workstation via thin-wire Ethernet. Here is a functional block diagram of BETA. BETA searches the 1.4-1.7 GHz waterhole as 8 hops of 40 MHz, each hop taking 2 seconds (16 seconds for a full cycle through the waterhole); thus each potential source is visited 8 times at each frequency hop, in each sky beam. A good candidate (seen first in east, then west, never terrestrial) triggers the antenna to leapfrog a few beamwidths to the west, inviting the source to perform an encore. If that ever happens, the antenna will break off its survey and go into sidereal tracking mode, repeatedly moving on and off the candidate source, archiving all integrated spectra.