What does SETI@home do?

1
What does SETI_at_home do?

• Mini-project for CSE 260
• Qian Peng
• 15 November 2001
• Ref. http//setiathome.ssl.berkeley.edu/

2
Data Source

• Sky survey conducted by National Astronomy and
  Ionospheric Centers 305 meter radio telescope in
  Arecibo, Puerto Rico.
• Covers 28 of sky, most sky will be observed two
  or three times during the course of the project.
• Recorded on 35GB DLT tape. Complete survey
  requires 1100 types, total of 39 TB data.
• Tapes are mailed to Berkeley.

3
What to Look for

• Extraterrestrial intelligence signals are of
  unknown bandwidth and time scale
• Wide band (unlikely), narrow band (most likely)
• Both continuous and pulsed signals are possible
• Need to reject local radio frequency interference
• Local source tends to be constant while sky
  drifts past the telescope (12 seconds for a
  target to cross the focus)
• Look for 12-second Gaussian signal
• Need to correct Doppler shift (frequency change)
  caused by relative motion

4
Distributing Data

• Embarrassingly parallel
• Frequency bands are independent of one another
• An observation of one portion of the sky is
  independent of an observation of another part
• Work unit of 107 sec. 10 kHz data sent to client
• Survey covers 2.5 MHz data, centered at 1420 MHz
• Broken into 256 blocks, each 9766 Hz wide
• Need to record at 20 kbps for signal up to 10 kHz
• 20kpbs x 107 s. additional info -gt 340 KB each
  unit

5
2.4 to 3.8 Tflops for Each Work Unit

• Baseline smoothing to reject broadband noise
• For Doppler shift rates from -50Hz/s to 50Hz/s,
  de-chirp data, for bandwidth resolutions from
  0.075 to1220 Hz in 2X steps
• FFT to generate time-ordered power spectra
• Search for short duration signals above a
  threshold
• For each frequency
• c2 fit to Gaussian (telescope beam profile)
• Search for triplets (three evenly spaced signals)
• Search for pulses (lots of equally spaced weak
  signals)

6
SETI_at_home Server Side

• 4 splitter workstations to divide 2.5MHz data
  into 256 sub-bands using FFT inverse FFT, and
  form work units. Temporary storage.
• 3 Sun Enterprise 450 servers
• User database (work units completed, membership)
• Science database
• time, sky coordinates, frequency for each work
  unit
• Parameters of potential signals detected by
  clients
• Work unit storage (handle distribution, store
  results)

7
What Happens to Qualified Signals

• Each work unit sent to multiple clients,
  cross-check results
• Most signals rejected as RFI or test signals
  based on a large database of known RFI sources
• Search remaining data for multiple detection
• Follow up with dedicated observation
• Have another team cross-check (different
  telescope, etc), then interferometry measurements.

8
E.T.I. ?!