The (Horse) Fly s Eye

Transcription

1 The (Horse) Fly s Eye A Search for Highly Energetic Radio Pulses from Extragalactic Sources using the Allen Telescope Array Crab Pulsar: Giant Pulse Detection (sigma = 15.75) Frequency (MHz) Uncalibrated Power 1465 CASPER Workshop II Griffin Foster, Andrew Siemion, Peter McMahon 8 fitted dm = Time (ms) Geoff Bower, Jim Cordes, Griffin Foster, Joeri van Leeuwen, William Mallard, Peter McMahon, Andrew Siemion, Mark Wagner, Dan Werthimer

2 Motivation Exciting Results From Lorimer et al. Lorimer, et. al., A Bright Millisecond Radio Burst of Extragalatic Origin. Science, 318, 7. Announced September 7 Single pulse at L-band 3 Jy, saturated digitizer in one beam Located 3 from the SMC DM = 375 pc/cm 3 implies D ~ 1Gpc Parkes Multibeam Pointing During Lorimer Detection Frequency vs. Time Waterfall for Lorimer Detection

3 The Fly s Eye: The Search for HERPES Highly Energetic Radio Pulses for Extragalactic Sources Goal: Build a set of 44 independent spectrometers for the ATA and use them to search for HERPES over a large portion of the sky. Timeline November 19, 7 - Dan Werthimer and Geoff Bower have lunch to discuss a suggestion made by Jim Cordes to perform a transient search using the ATA. November, 7 - Dan Werthimer tasks a group of mostly undergraduate students at the Center for Astronomy Signal Processing and Electronics Research (CASPER) to begin building a transient instrument. December 22, 7 - Fly s Eye Team installs Fly s Eye at ATA. February, March 8 - Conducted 5 hours of weekend observations. April - August Data analysis underway

4 The Allen Telescope Array Located in Hat Creek, CA (~5 hours from Berkeley) 42 x 6 m dishes,.5-11 GHz usable band 4 independent tunable IFs First light was less then a year ago, October 7 (Galaxy M31) Text Advantages Newly commissioned, so there is plenty of observing time available Each antenna has a wide beam, ~2 FWHM (at L band) With independent, tunable IFs commensal observing is very easy

5 ibob Spectrometer Using standard CASPER libraries and the pocket correlator design the 4 input spectrometer used in Fly s Eye was completed in 3 weeks.

6 The Instrument 44 independent spectrometers - constructed using a system of eleven ibob/iadc quad spectrometers Built using open-source CASPER hardware and software libraries in about one month. Sky Coverage: beams - square degrees Spectrometer Specifications (each): 8 MHz bandwidth, at 143 MHz 128 spectral channels.625 ms readout Fly s Eye Rack at ATA Distributions: Spatial, DM, Power, Pulse Width

7 Instrument Diagnostics PSR B Detection PSR B (36 beams summed, 15 minutes folded) PSR B detected in 41/44 signal paths. PSR B Detections in all 44 Beams (15 minutes folded)

8 Control and Monitoring After initial instrument setup at the ATA all observations were done remotely

9 Observation Diagnostics ibob-g47 ibob-g44 ibob-g ibob-g48 ibob-g45 ibob-g ibob-g46 ibob-g43 AA BB CC DD All observations were done via scripts on a control machine and recorded to a data server (via gulp) For every hour of observations 58 minutes were used in the horseshoe pattern, 1 minutes was used for recording diagnostic data to assure data quality, note the 21 cm line. Packet loss statistics and spectra from diagnostic runs were uploaded to a web server 5 3 ibob-g ibob-g ibob-g

10 Observations Drift scan Fly s Eye observations were conducted in campaign mode on weekends between February and April 8. Initial plan was for fly s eye sky patch observing, eventually transformed to horseshoe constant declination strip. Total observing time thus far is approximately 48 hours. Both North and South pointing observations were performed, primarily North due to kinder RFI environment. Total dataset is approximately 17 terabytes. Fly s Eye Beam Pointing Diagram

11 Instrument Diagnostics Giant Pulses From PSR B The Crab pulsar has be well studied, and is know to produce giant pulses. The pulsar was observed for one hour, during which close to a dozen detectable pulses were detected (in summed data), the brightest of which was distinguishable in approximately half of single dish observations. Crab Pulsar: Giant Pulse Detection (sigma = 15.75) Frequency (MHz) fitted dm = Uncalibrated Power Time (ms) Giant Pulse from PSR B (single beam) Giant Pulses from PSR B (35 beams)

12 Data Analysis Raw gulp Datafile Error Analysis / Dropped Packet Correction IF Seperation (filterbank) IF IF 1 IF 2... IF 43 IF SUM RFI Rejection freq. and time domains De-dispersion DM (dedisperse) Equalization / Normalization De-dispersion DM 1 (dedisperse) De-dispersion DM 2 (dedisperse)... De-dispersion DM (dedisperse) DM search range: 5- pc/cm 3 (with decimation, non-integer spacing) Computing grids used: 5 (Berkeley Wireless Research Center, UC Berkeley EECS, DOE NERSC) Frequency Collapse (dedisperse) Compute RMS / Threshold (seek) Decimate (seek) Log Candidate Pulses Disk Storage (MySQL Database) Total cores (peak): ~ (Itanium64, Xeon, Sparc, Opteron) Fly s Eye Data Analysis Pipeline Total throughput (peak): ~ Mbits/second

13 The RFI Challenge Types of RFI: Wideband, short time period Narrowband, continuous(radio, airplanes, satellites) Narrowband, short time period(radar, more airplanes) Special cases(broken satellites?) Solutions: Pre data analysis: normalization and equalization of spectra, remove abnormal spectra(wideband RFI) time collapse and compute variance, similar to kurtosis methods(narrowband, short time) Post data analysis: flag times with a wide range on DM events Special cases for low(<5 DM) and high(>18 DM) events Current Challenges: RADAR Narrowband, short time period RFI Equalization/Normalization methods

14 The RFI Challenge: Military Radar ARSR-4 Air Route Surveillance Radar Radar transmits on ~12s timescales With a low duty cycle the radar is hard to remove using variance RFI removal In northern California there are 2 nearby sites (Rainbow Ridge, Mill Valley) The stations operate out of phase with each other and at different frequencies false high DM detections Brute force solution: ignore these channels

15 Interesting RFI: Broken Satellites From our first pass script we found ~8 interesting events, one event in particular looked like something we had never seen before, it was too narrow band to be wide band RFI and it was centered around 14 MHz. (Aliens?!?!?!) Then we zoomed in on the data... After zooming in on the event we determined it was not terrestrial RFI or Aliens, but based on the angular speed of the object it is probably a low orbit satellite(transmitting around 14 MHz!).

16 Producing Results MySQL Server Post RFI Rejection Corrected Plots Raw Plots Decimated Waterfalls Web Interface Request Plots After data analysis the data is stored to a MySQL server where our graphing pipeline retrieves data. Using Python we do RFI flagging and generate a series of plots which can be viewed from a web interface The next step of the web interface is to include user interaction to tag plots with i.e. interesting/bad rfi/ nothing... To view an interesting portion of the data we retrieve the raw data from the tape backup, and generate a large waterfall plot of all the data for a time period (the large files sizes of our data causes this step to take ~ minutes to generate). Full Waterfalls User Tag Events

17 Example Preliminary Results 48 hours of observations, 44 spectrometers, minute sets, 9 plot types over a million plots! Result Browser Example Time vs. Sigma Plot Example Time vs. DM Plot

18 New and Future Analysis We have the plot, now we need the eyes! Using our first pass scripts we can find the interesting events in the best data but there is much more to be looked at. Reprocess! - Improve RFI rejection and implement new pulse search algorithms (underway) Improve SNR in some of the data by summing polarizations Include a measurement of the kurtosis to remove intermittent narrowband RFI Try different equalization techniques Improve speed of analysis

19 Assessment of Significance Region allowed by existing experiments Parameter space explored by Fly s Eye I (Spring 8) Parameter Space for Radio Transient Detection Parameter space potentially explored by a one-year commensal Fly s Eye Experiment

20 Future Experiments: Localizing HERPES If we can find them, we should try and localize them. We can easily convert a 4 input ibob Spectrometer to a 4 input Pocket Correlator Fly s Eye 44 input fast readout spectrometer becomes... Fly s Eye 11 x 4 input fast readout correlator Angular Localization of Transient Radio Bursts