Download presentation
Presentation is loading. Please wait.
Published byCody Atkins Modified over 9 years ago
1
Coherent Dedispersion Pulsar Timing Machines Matthew Bailes + Swinburne, Caltech, ATNF, CASPER
2
Magnetic Field - Period Diagram Crab pulsar Pulsars go this way “Noisy” “Smooth”
3
Timing Procedure: “Fold” data at apparent Period (up to ~million pulses) Use pulsar ephemeris, site position Calculate shift Calculate Site Arrival Time (SAT) Correct for Clocks (UT, UT1, etc) Transform to Solar System Barycentre (SSB) Barycentric arrival time (BAT/TOA) Subtract pulsar model Fit for pulsar parameters
4
Pulsar Timing Count the Pulses!
5
PSR J1600-3053 Form a Profile Cross Correlate Get time of arrival
6
Time of Arrival (TOA) error width Signal to noise ratio Gaininformation PSR flux Temperature
7
Pulse Dispersion FFT x Filter -1 FFT -1 No Dispersion!
8
Pulsar (in)Coherent Dedispersion Uses computers/FPGAs and digitizers to remove dispersion almost “perfectly” Rules: Don’t distort profiles Have high dynamic range Attain Polarisation Purity Remove Dispersion changes Don’t change any equipment Tension!
9
Radioastronomy in Software No hard definition Pulsar “backend” written in software, not hardware Requires: Digitizer IO system to computers Software Advantages: Bandwidth follows Moore’s law
10
Radio Astronomy in Software Finite amount of information: ~B N tel N p (voltages/second) ~ giga-samples/sec To make spectra for a Single dish: ~N p 5 log 2 N FFT B PKS 512 MHz, 1024 channels, 2 polns 2x50x512x10 6 ~50 GFlops! Mpt FFTs ~500 Gflops Gpt FFTs ~ Teraflops
11
Example: Chart Recorder (power monitor):
12
RFI detector (not real code - contains bug!)
13
S2TCI Model (van Straten + Wiedtfeld?) York A/D Processors 8 Video Tapes 8 Video Tapes CPU 2 x 16 MHz 128 Mb s -1 Mark I Switch
14
Issues Error rate ~10 -5 Required two custom devices (@70K each) 1000s of video tapes required Limited BW Playback < real time Months until result
15
CPSR1 Model (Anderson/van Straten) Caltech A/D Processors COTS Disk COTS DLT Tapes COTS 2 x 20 MHz 160 Mb s -1 SUN COTS DLT Tapes COTS Switch COTS
16
CPSR1 Issues Error rate ~10 -5 -> Zero 1000s of DLT tapes required (5K/day) No custom playback unit required Still Limited BW Playback < real time Weeks until result
17
Processors Custom A/D RAM Dell CPSR2 Model (Bailes/Ord/ van Straten/Hotan) 2 x 2 x 64 MHz 1024 Mb s -1 Real time +5 mins Caltech A/D RAM DellSwitch
18
CPSR2 Real time RFI excision (< hour to implement) 5 minute delay until profiles cf weeks-months Much greater sensitivity (~20 MSPs) 2bit device (worse residuals on 0437!) Running since Aug 2002-now Gb/s VLBI recorder CGSR unreliable (COTS too cheap)
19
Best Results?? Typically few to ~100 ns RMS 0437-4715 van Straten et al. (130 ns) 3 years 1713+0747 Nice et al. (200 ns) (Arecibo) Verbiest et al. (2009) 200 ns (Parkes) 0437-4715 Verbiest et al. (2008) 200ns 9.9 years 1909-3744 Verbiest et al. (2009) 166 ns ~6 years
20
Pulsar Profiles with Zero Smearing PSR J0437-4715 timing Joris Verbiest PhD 5.75 ms
21
Verbiest et al. (2008) Incoherent S2 CPSR2 CPSR1
22
Pulsar Parameters
23
PSR J0437-4715 Changing inclination angle di/dt (10,000 sigma!) Annual “orbital” parallax i=137.58(6) deg! GR “test” van Straten et al. (2001) Mass of a white dwarf Mc=0.25(1) dP b /dt= 200 sigma 100,000 sigma 5,000,000 sigma
24
PSR J0437-4715 (Deller et al. 2008B) Parallax works! D=156.3(1.3) pc D Pb =156 (2) pc (G/G) < 2x10 -12 yr -1.
25
0437 Results (Deller et al 2008) No anomalous accelerations Three different distance measures accurate to 1-2% Parallax works! d G /d t < 2x10 -12 G No unseen “Jupiters” within 226 AU No black holes/neutron stars approaching No extreme gravitational wave background
26
PSR J1909-3744 40 s wide!!
28
Shapiro Delay
29
Astrophysics: e = 1.3(2) x 10 -7 a = 569,397.318 km (a-b) = 5 M 2 = 0.212(4) Mo D = 1260 pc
31
Giant Pulses (Knight et al. 2006) Unresolved on us timescales From young or millisecond pulsars Power-law distribution of energies MWA Bhat et al. (2007) < 2 s wide!
32
DFB3 DFB3 (Manchester/Hampson/Brown) APSR (Bailes/van Straten/Jameson) CABB A/D 2 x 16 x 64 MHz 8096 Mb s -1 8 Gb s -1 PFB PFB -1 Switch Processors APSR
33
APSR Issues/Features 10 seconds until profile 1 GHz BW @ 2 bits coherent dedispersion 512 MHz BW @ 4 bits coherent dedispersion 256 GHz BW @ 8 bits coherent dedispersion Adaptive RFI cancellation mode (Kesteven & Manchester) Coherent filterbank & fold for >= 20 pulsars at once! Single pulse mode, RT calibration Very robust, jetisons to disk if required 64 MHz/Server real time ! BUT… PFB “staining”
36
DFB3 + lightening
37
APSR + lightening
38
DM = 120, 3 ms MSP with APSR
40
Key Software PSRDADA Open source ring buffers and resends DSPSR Open source coherent dedisperser Links to ring buffer or disk files Writes psrchive files Nehalem 128 MHz, real time coherent dedispersion! PSRCHIVE Open source pulsar profile manipulator
41
A/D RAM CPU Processors CASPSR Model (van Straten/Jameson) 2 x 400 MHz (8 bit) No PFB stains RAM 12,800 Mb s -1 12.8 Gb s -1 Switch CPU
42
CASPSR/TerryBOB Issues/Features FFTs become enormous Expect ~1 minute delay until profiles seen NO PFB issues!!! On edge for Clovertown @ 2 servers Very extendable N servers if required Moore’s law on our side
45
Future: ROACHPSR? 800(1024) MHz BW x 2 x 8 bits 24 port 10 Gb Switch 4-6 Data streamers (Nehalems with 48 GB RAM) 8-12 Processors (Supermicros with 2xGPUs) 2-3 x CASPSR RFI excision/removal 13xCASPSR beam spectrometer? 200 MHz out of 400 MHz, 2 bits
46
Parkes 13 beam x 400 MHz Pulsars Spectroscopy 10cm spectroscopy 1024 MHz, 8192 channels 13cm spectroscopy 512 MHz 10/50cm timing 3 GHZ @ 1024 MHz 700 MHz @ 64 MHz VLBI
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.