Three Years of SETI@home A Status Report

Download Report

Transcript Three Years of SETI@home A Status Report

The SETI@home,
SERENDIP, SEVENDIP,
Astropulse, and SPOCK
SETI Programs
‘Dan Werthimer, Dave Anderson, Jeff Cobb,
Paul Demorest, Eric Korpela, Cecile Kim, Geoff Marcy
University of California, Berkeley
http://seti.berkeley.edu/
NOT FUNDED
NOT FUNDED
NOT FUNDED
Porno in space:
FUNDED!
Drake Equation
• N=R fs fp ne fl fi fc L
• N = number of communicating
civilizations in our galaxy
Planet Detection
First Radio SETI
• Nikola Tesla (1899)
– Announces “coherent signals from Mars”
• Guglielmo Marconi (1920)
– Strange signals from ET
• Frank Drake (1960)
– Project Ozma
– one channel, 1420-1420.4 MHz
Signal Types
1. Artifact
(radio, radar, ~TV, ????)
2. Deliberate (easy to decode, pictures, language
lessons)
First civilization we contact is likely to be a
billion years ahead of us.
Targeted Search Strategy:
Project Phoenix
- Seti Institute
Sky Survey Strategy:
Serendip, SETI@home
-
UC Berkeley
Beta
- Harvard
Southern Serendip
- Australia
Meta II
- Argentina
Seti Italia
- Medicina Obser.
Quick History of Berkeley SETI
• Radio SETI
– SERENDIP
Search for
Extraterrestrial Radio
Emissions from
Nearby Developed
Intelligent Populations
1.00E+10
1.00E+09
1.00E+08
1.00E+07
1.00E+06
1.00E+05
1.00E+04
1.00E+03
1.00E+02
• SERENDIP I-III (1979-1997)
1.00E+01
• SERENDIP IV (1997-)
1.00E+00
• SERENDIP V (2004-)
1979
1984
1989
1994
1999
2004
Channels
Bandwidth [kHz]
Resolution [/kHz]
Beams
Sensitivity ([-25] W/m2)
Drake FOM [Hz*Sr*m3/W1.5]
The Berkeley Radio SETI Family Tree
SERENDIP
SERENDIP II
OSU
SETI Italia
SERENDIP III
SERENDIP IV
SETI@home
Data Recorder
Southern
SERENDIP
SETI@home
Clients
SETHI@Berkeley
HI Survey
AstroPulse
Pulse Survey
SERENDIP V
SETI@home II
Data Recorder
SETI@home II
Clients
SETI Programs at the
University of California
NAME
TIME SCALE
SEARCH TYPE
SERENDIP
1 second
Radio sky survey
SETI@home
ms to second Radio sky survey
ASTROPULSE us to ms
Radio sky survey
SEVENDIP
ns
Visible targetted
SPOCK
1000 seconds Visible targetted
SERENDIP IV
Photos Courtesy NAIC Arecibo Observatory, a facility of the NSF
•
168M channels
•
•
100 MHz Band centered on 1420 MHz •
Carriage House 1 line feed
Operating since 1997
Why SETI@home?
• Coherent Doppler drift correction
– Narrower Channel Width->Higher Sensitivity
• Variable bandwidth/time resolution
• Search for multiple signal types
– Gaussian beam fitting
– Search for repeating pulses
Problem: Requires TFLOP/s processing power.
Solution: Distributed Computing
The SETI@home Client
SETI@home Statistics
TOTAL
RATE
4,324,355 participants
(in 226 countries)
2,000 per day
1,168,254 years
computer time
1,200 years per day
1021 floating point
operations
55 Tera-flops
Structure of SETI@home
Master
Science
Database
Candidate Identification
Result Verification
Online
Science
Database
The Internet
Tapes from
Arecibo
Data splitters
Work Unit Storage
Data Server
Volunteer
Statistics
Database
Web Server
3.8 Million Volunteers
The Input and Output
• 1 Work-Unit=9.8 kHz x 220 samples (107 sec.)
– 256 Workunits across 2.5 MHz band centered on
1420.0 MHz.
– Workunits overlap in time by ~25 sec.
– Each workunit sent to multiple computers for result
verification
– Typically 4 TFLOP/workunit.
• Output=Typically ~5 potential signals.
Spikes
• Power distribution in
the Fourier
transformed data is
exponential if no RFI.
• SPIKE: Any bin in the
spectrum above 22X
the mean power
(7.8x10-25 W/m2)
Gaussians
• Weighted 2 fit to
beam profile (vs
time).
• Gaussian must exceed
a power and 2
threshold
• Score inversely
proportional to
probability of arising
due to noise
• Sensitivity 8.4x10-25
W/m2
Triplets
• Three evenly spaced
spikes above 7.75X
the mean power.
(5.3X10-25 W/m2)
Pulses
• Modified Fast folding
algorithm w/ dynamic
threshold
• Logarithmically
spaced periods from
3ms to 35s
• Sensitivity as low as
10-26 J/m2
Candidate Identification
• Candidate: A signal or group of signals
– Within a positional window (~1 beamwidth typ.)
– Within a frequency window (variable)
– Above a score or power threshold (variable)
– With time separation » typical transient RFI timescale
• Score:
– Relative ranking of a candidate’s probability of arising
due to random noise.
– Should be independent of signal type
– Can also include probability of coincidence /w
celestial objects
Gaussian Candidates
AstroPulse
• Sky survey
– Covers decs 0 to 30
– ~3 years of data recorded so far.
• Good time resolution
– Sensitive to 0.4 µs radio pulses at 21 cm
• DM range
– -100 to +100 pc/cm3
• Sensitivity
– 10-18 W/m2 peak (Coherent de-dispersion)
Pulsed vs. CW
Concentrating power into short bursts can be more
efficient than a “constantly on” transmitter.
Pulsed signals can be easier to see above
background noise.
Dispersion
… eventually becoming very weak.
However, we can correct for dispersion ...
AstroPulse
• Only ~1.5 searches for single pulses
on µs timescale before (O’Sullivan, Phinney)
• Pulsar searches: ms time scales, folded
• SETI@home: 0.8 ms single pulses.
• With interesting astrophysics as well
as SETI applications.
– Evaporating primordial black holes?
– Pulsars, Other astrophysical exotica?
Computation
… but it takes a lot of CPU
time!
To search DMs up to 100
pc/cm3 in real time, we need
about 500 GigaFLOPs.
(This would take ~1000 years
of your PC working full time)
Conclusion: We need more
computers!
BOINC
• Berkeley Open Infrastructure
for Network Computing
– General-purpose distributed
computing framework.
– Open source.
– Will make distributed computing
accessible to those who need it.
(Starting from scratch is hard!)
AstroPulse/BOINC
• AstroPulse will be the first to use BOINC.
• It is a good “beta-test” application:
– Simple data analysis/reduction.
– “Only” needs a few thousand computers.
– Other projects which plan to use BOINC:
– SETI@home II
– Global climate modeling/prediction (Oxford)
AstroPulse Testing
Sample batch of data
run through shows
expected noise
characteristics, and little
else …
… so (hopefully) little
RFI contamination for
this type of signal.
HI Column Density
OPTICAL SETI
• OPTICAL PULSE SEARCH
– Pulsed laser power output continues to grow.
– Petawatt pulses achieved at Livermore Labs.
(Mjoule in 1nS)
– can detect at earth technology at 1Kpc
– little background noise, even from bright
stars in whole visible band
OSETI Detector
• 3-Photomultiplier fast
coincidence detector
– Sensitive to 1ns
pulses
• Low background
– False alarm rate: 1
per 300 hours (10-6
Hz)
– Double false alarm
rate: 1 per 600 years!
• Good sensitivity
– 10-8 W/m2 peak
– 10-19 W/m2 average
Optical SETI
• Uses Leuschner
Observatory (UCB)
– Automated 0.8m
telescope
• Targeted Search
– Nearby F,G,K,M stars
– ~2,000 stars
observed so far
– Soon to include
galaxies
Amy Reines and Geoff Marcy
10-meter Keck
Telescope
Survey: 650 F8 – M5 V, IV
Hipparcos
V < 8.5
B-V > 0.55 (F8V)
Sep > 2 arcsec
Age > 2 Gyr
Doppler Instruments
• Echelle Spectrometer
• Resolution: 60,000
• Iodine Abs. Cell.
– Superimpose I2 lines
– Wavelength Calib.
Piggyback ALFA Sky Survey
• SETI Instruments
– Dedicated spectrometer (SERENDIP V)
• 300 MHz bandwidth, 2 pols, 7 beams
• 5 * 109 channels, 0.8 Hz resolution
– SETI@home II data recorder
• 10 MHz, 1 pol, 7 beams
• Steps across 300 MHz band
Piggyback ALFA Sky Survey
• Improved sensitivity
– Tsys, integration time
• Uniform sky sampling
– galactic plane concentration
• Multibeam RFI rejection
• Larger Bandwidth
Our Generous Sponsors
•
The Planetary Society
•
The SETI Institute
•
The University of California
•
Informix
•
Sun Microsystems
•
EDT
•
Friends of SETI@home
•
Netscreen
•
Network Appliance
•
Intel
•
Fujifilm
•
O’Reilly & Associates
•
IBM
•
SpaceSounds
•
Quantum
•
Dillon Engineering
•
HP
•
NAIC, Arecibo Observatory
•
Xilinx
•
~4 million volunteers
Maybe, someday, the U.S. Government
•SETI HAIKU
Seti.berkeley.edu