Correlation of mm - VLBI Observations Roger Cappallo ~ Mike Titus MIT Haystack Observatory 2010.1.25
Talk Flow Practical experiences at the correlator Preparation Hardware Correlation Post-processing Software Correlation & Future Plans
Recent mm-vlbi Experiments processed at Haystack April 2007 & 2009 SgrA* - CARMA/SMTO/JCMT at 230 GHz - size measurement (and by implication position!) of emitting region March 2008 SgrA* - VLBA at 86 GHz - parallax measurement
Pre-experiment recording checks sually do fringe test of all stations via evlbi (ftp downloads) Must gather and correct a priori information prior to attempting this! Results fed back to the stations and any needed fixes applied, repeat until ready
Experimentʼs Passage through the Correlator gather logs and schedules inspect incoming media (disks) - put into library compile logs and schedule in correlator (vex) format pick scans to find fringes; select and run pre-pass scans construct processing & fourfit control files schedule, then perform production processing analyze results/evaluate stations -- clean-up as necessary export results release disks
mm setups atypical Schedules often have incorrect a priori info, which needs to be discovered and corrected prior to processing: - source positions - station positions - frequency sequences - LO offsets - record mode (e.g. channel layout)
Collect and insert station info in corr. control files: Need clock information; offset from GPS, drift rate and any jumps Gather station logs, summarize and check station comments -- used to edit schedule Note which IFs are which bands, configure as SB or LSB as appropriate
Pre-pass and production phase pre-passes are a sampling of scans throughout schedule to make sure setup is correct and stable (e.g. clock model) production run -- after pre-passes check OK production can take up to 4X observe time, due to 1 Gb/s playback limit of Mk4 (2 IFs recorded at 2 Gb/s played back at ½ speed)
Mark 4 correlator characteristics 16 stations 16 channels / station 1 gb/s per station (actual playback rate) 2 gb/s record rate processed 1/2 speed fringe rate window width typically 4 Hz
Demanding nature of mm VLBI high frequencies high fringe rates - e.g. 260 KHz on SMTO-JCMT baseline upper edge of rotator envelope - different phase increment mode - errors uncovered in digital generator of acceleration model
After correlation... Output needs to be examined even more carefully than usual: - Make sure clock/rate OK and LO stable - Was accumulate period adequate (residual fringe rates within window)? - Any scans missed or need re-correlation? - Are manual phase cals needed? Segment in the HOPS post-processing package over coherence time Possibly import into AIPS for further analysis
Wideband Postprocessing per pass limit: 16 ch x 32 MHz/ch = 512 MHz multiple passes glued together via fourmer post-processing software fourfit modified to do fringe fitting over 64 ch coherently combining separate IFʼs can be challenging...
Mk4 Fringe Plot 0854+201.uponwn, 096-0740, JC JCMT - CARMA, fgroup, pol LL Fringe quality 9 Amp. and Phase vs. time for each freq., 14 segs, 4 APs / seg (2.00 sec / seg.), time ticks 2 sec SNR 12.0 PFD 3.5e-26 Intg.time 27.868 Amp 0.929 Phase 61.3 Sbdelay (us) -0.000592 Mbdelay (us) 0.000023 Fr. rate (Hz) -0.048494 Ref freq (MHz) 229361.0000 AP (sec) 0.500 Exp. Exper # Yr:day Start Stop FRT mm096 3271 2009:096 074002.00 074030.00 074230.00 Corr. date: 2009:112:002014 Fourfit date: 2010:025:130903 Position (J2000) 08h54m48.8749s +20 06 30.640" Validity L Parity Bias Level L L J (S 10) C (S 7) J C J C
Mk4 Fringe Plot 0854+201.uponwn, 096-0740, JC JCMT - CARMA, fgroup, pol LL Fringe quality 9 Amp. and Phase vs. time for each freq., 14 segs, 4 APs / seg (2.00 sec / seg.), time ticks 2 sec SNR 12.4 PFD 2.3e-28 Intg.time 27.868 Amp 0.960 Phase -105.6 Sbdelay (us) -0.000235 Mbdelay (us) -0.000134 Fr. rate (Hz) -0.048135 Ref freq (MHz) 229361.0000 AP (sec) 0.500 Exp. Exper # Yr:day Start Stop FRT mm096 3271 2009:096 074002.00 074030.00 074230.00 Corr. date: 2009:112:002014 Fourfit date: 2010:025:131000 Position (J2000) 08h54m48.8749s +20 06 30.640" Validity L Parity Bias L J (S 10) C (S 7) J C
Mk4 Fringe Plot 0854+201.uponwn, 096-0740, JC JCMT - CARMA, fgroup, pol LL Fringe quality 9 Amp. and Phase vs. time for each freq., 7 segs, 9 APs / seg (4.50 sec / seg.), time ticks 5 sec SNR 15.2 PFD 4.2e-45 Intg.time 27.868 Amp 0.837 Phase -46.6 Sbdelay (us) -0.008473 Mbdelay (us) -0.000464 Fr. rate (Hz) -0.048602 Ref freq (MHz) 229361.0000 AP (sec) 0.500 Exp. Exper # Yr:day Start Stop FRT mm096 3271 2009:096 074002.00 074030.00 074230.00 Corr. date: 2009:112:002014 Fourfit date: 2010:025:130716 Position (J2000) 08h54m48.8749s +20 06 30.640" Validity L Parity Bias L J (S 10) C (S 7) J C
Mk4 Fringe Plot 0854+201.uponwn, 096-0740, JC JCMT - CARMA, fgroup, pol LL Fringe quality 7 Amp. and Phase vs. time for each freq., 13 segs, 49 APs / seg (24.50 sec / seg.), time ticks 30 sec SNR PFD 20.7 0.0e+00 297.725 0.498 0.3 Intg.time Amp Phase Sbdelay (us) 0.003674 Mbdelay (us) 0.000002 Fr. rate (Hz) -0.050530 Ref freq (MHz) 229361.0000 AP (sec) 0.500 Exp. Exper # Yr:day Start Stop FRT mm096 3271 2009:096 074002.00 074500.00 074230.00 Corr. date: 2009:112:002014 Fourfit date: 2010:025:131238 Position (J2000) 08h54m48.8749s +20 06 30.640" Validity L Parity Bias Level L L J (S 10) C (S 7) J C J C
Mk4 Fringe Plot 0854+201.uponwn, 096-0740, JC JCMT - CARMA, fgroup, pol LL Fringe quality 0 Amp. and Phase vs. time for each freq., 13 segs, 49 APs / seg (24.50 sec / seg.), time ticks 30 sec SNR PFD 1.8 1.0e+00 297.725 0.042 157.2 Intg.time Amp Phase Sbdelay (us) 0.000477 Mbdelay (us) -0.000367 Fr. rate (Hz) -0.045118 Ref freq (MHz) 229361.0000 AP (sec) 0.500 Exp. Exper # Yr:day Start Stop FRT mm096 3271 2009:096 074002.00 074500.00 074230.00 Corr. date: 2009:112:002014 Fourfit date: 2010:025:133952 Position (J2000) 08h54m48.8749s +20 06 30.640" Validity L Parity Bias Level L J (S 10) C (S 7) J C J C
Future Plans Software correlation using DiFX - mm VLBI well-suited for software correlation, due to small N, linear run-time dependence on BW
Computers outpacing VLBI data rates Deller & Brisken 2009
Deller & Brisken 2009
Future Plans Software correlation using DiFX - mm VLBI well-suited for software correlation, due to small N, linear run-time dependence on BW Currently working on interface from DiFX into Mk4 / HOPS post-processing suite Considering a complete integration of DiFX into Mk4 processing system