Overview of Talk ALMA Phasing Project (APP) Hardware & So?ware Some Commissioning Results Some DiFX Details Road Forward 11/16/15 EHT CorrelaLon - G. Crew 2
ALMA Phasing Project (APP), 1 Diverse Team (led by Shep Doeleman, PI) MIT Haystack Observatory (prime) NRAO (correlator hardware, &c) MPIfR (so?ware and analysis) ASIAA (correlalon hardware and simulator) NAOJ (oplcal fiber link system) U. Concepcion (observalonal support) Onsala (polarizalon conversion) Harvard CfA (firmware & reviews) DiFX Community support 11/16/15 EHT CorrelaLon - G. Crew 3
ALMA Phasing Project (APP), 2 History and some key Milestones 4- year project awarded Aug. 2011 PDF in Nov. 2012, ALMA board approval followed CDR in May 2013 Hardware/so?ware development & delivery Hardware Accepted Nov./Dec. 2014 First VLBI Light Jan 2015 (ALMA APEX band 6) True VLBI Mar 2015 (ALMA PICO band 6) VLBA/EHT Commissioning July/Aug 2015 (band 3,6,7) So?ware/Commissioning review Oct 2015 Cycle 4 (Call for Proposals Mar 2016) starts Oct 2016 Limited extension to support Cycle4/Cycle5. 11/16/15 EHT CorrelaLon - G. Crew 4
ALMA Phasing System (APS) The (in)famous cartoon Antenna Front End Samplers, WVR Optical Fiber Tx Central Electronics Central Variable Reference Photonic LO 10 MHz Maser H Maser 5 MHz CRG: 125 MHz, &c MASER_V_GPS_COUNTER CLOA GPS Correlator: Station Optical Fiber Rx/Demux Eng. Port 1 PPS Corr GPS ALMA Control System Connects to all sub systems Alma Scheduling Blocks Tunable Filterbank Commands Phasing Tunable Filterbank Card (TFB) Form 32 x 62.5 MHz slices Phase shift each slice Station Card Geometric Delay Correlator: Baseline Correlator Interface Card (CIC) Route data to corr cards VLBI sum input Correlator Card (CC) Cross Correlation Form VLBI Sum VLBI sum antenna mask Scaling of VLBI Sum Delay to align VLBI sum Long Term Accum/Final Add Correlator: Control Correlator Hardware Protocols TFB Commanding VLBI Phasing Interface Card 2 Cards per Quadrant VDIF Formatting 10 GbE Packetization Synchronization PPS/Clock Monitor Phased Sum Data Each Quadrant: 2Pol x 32ch x 62.5MHz 2 bits/sample Data Rate: 16 Gb/s 128 LVDS Pairs CDP Computer Cluster Baseline Visibilities Generate Channel Averages Fast WVR Phase Corrector φ Phase Corrections (Fast) 8 x 10 GbE AOS (high site) OSF (low site) 8 x 10 GbE Fiber MUX Fiber DEMUX VLBI 16 Gbps Recorders OSF (low site) TelCal Computer Slow Phase Solver and Monitor Uses source model parameters to calculate phase corrections Monitors phasing efficiency by correlating sum with reference φ Phase Corrections (Slow) VLBI Observing Mode Scheduling Block Control of TFB: Spectral Windows / Channel Averages Control of CIC: VLBI sum on CAI 63 Control of CC: VLBI antenna sum mask 2 bit Quantization of VLBI sum Control of Phasing Interace Cards: Synchronizatioed commanding Packet Time setting Monitor GPS Maser Drift Control of VLBI Recorders VLBI Scheduler Control of Source Model Monitor Sends Parameters to TelCal Computer OLD things NEW things DONE! 11/16/15 EHT CorrelaLon - G. Crew 5
APS Scan Sequencing VLBI Scan Record On... Correlator Scan Sequence... TelCal Solving... TFB Commands... WVR Data... Time Unphased Phased Phased Phased 11/16/15 EHT CorrelaLon - G. Crew 6
Some Commissioning Results January 12-13 2015 first light VLBI March 30 2015 first interconlnental fringe ALMA- PICO March September A variety of phasing experiments (which turned up some intereslng issues with the ALMA correlator) July/August 2015 3 sessions in 3 bands in 3 days Unfortunately it was a bad week for the ALMA correlator (very unstable for non- APS reasons) 11/16/15 EHT CorrelaLon - G. Crew 7
Aug 27 th phasing (all fixed! 25 ant.) Baselines from sum and reference to comparison antennas Un- Phased Phased 11/16/15 EHT CorrelaLon - G. Crew 8
Band 3 Amplitude (ALMA w/vlba) 11/16/15 EHT CorrelaLon - G. Crew 9
Band 3 Phasing (ALMA w/vlba) 11/16/15 EHT CorrelaLon - G. Crew 10
Phasing Efficiency Experiments Phasing Efficiency (Mar-Sep 2015) Average(<Comp. Sum> / <Comp. Ref>) 7 6 5 4 3 2 Sqrt(N) APP0150 Mar30 Apr24 May07 May29 Jun20 Jun19 Aug27 Aug25 Sep14 Mar30(band6) Jul31(band6) Aug01(band7) Aug01(band3) 1 0 0 10 20 30 40 50 Phased Antennas (N) 11/16/15 EHT CorrelaLon - G. Crew 11
ALMA- PICO March 30 dual pol, HP Mk4/DiFX fourfit 3.10 rev 1009 Amp. and Phase vs. time for each freq., 7 segs, 70 APs / seg (22.40 sec / seg.), time ticks 30 sec 3c273.yizlev, 089-0404, HP Hx - Px, fgroup B, pol LL Fringe quality 5 Error code G SNR 41.3 Int time 135.983 Amp 0.724 Phase 100.4 PFD 0.0e+00 Delays (us) SBD 0.004884 MBD 0.008442 Fringe rate (Hz) -0.063007 Ion TEC 0.00 Ref freq (MHz) 226166.1969 AP (sec) 0.320 Exp. Exper # Yr:day Start Stop FRT Mar30l 1120 2015:089 040504.00 040720.00 040612.00 Corr/FF/build 2015:243:151932 2015:243:152457 2015:240:083820 RA & Dec (J2000) 12h29m 6.6997s +2 03 8.598" U Validity 11/16/15 EHT CorrelaLon - G. Crew 12
ALMA- PICO March 30 dual pol, IP Mk4/DiFX fourfit 3.10 rev 1009 Amp. and Phase vs. time for each freq., 7 segs, 70 APs / seg (22.40 sec / seg.), time ticks 30 sec 3c273.yizlev, 089-0404, IP Iy - Px, fgroup B, pol RL Fringe quality 5 Error code G SNR 35.3 Int time 135.983 Amp 0.619 Phase -150.3 PFD 0.0e+00 Delays (us) SBD 0.013016 MBD 0.008225 Fringe rate (Hz) -0.062985 Ion TEC 0.00 Ref freq (MHz) 226166.1969 AP (sec) 0.320 Exp. Exper # Yr:day Start Stop FRT Mar30l 1120 2015:089 040504.00 040720.00 040612.00 Corr/FF/build 2015:243:151932 2015:243:152257 2015:240:083820 RA & Dec (J2000) 12h29m 6.6997s +2 03 8.598" U Validity 11/16/15 EHT CorrelaLon - G. Crew 13
ALMA- PICO March 30 dual pol, HQ Mk4/DiFX fourfit 3.10 rev 1009 Amp. and Phase vs. time for each freq., 7 segs, 70 APs / seg (22.40 sec / seg.), time ticks 30 sec 3c273.yizlev, 089-0404, HQ Hx - Qy, fgroup B, pol LR Fringe quality 5 Error code G SNR 31.4 Int time 135.983 Amp 0.550 Phase -41.5 PFD 0.0e+00 Delays (us) SBD 0.000152 MBD 0.003640 Fringe rate (Hz) -0.062798 Ion TEC 0.00 Ref freq (MHz) 226166.1969 AP (sec) 0.320 Exp. Exper # Yr:day Start Stop FRT Mar30l 1120 2015:089 040504.00 040720.00 040612.00 Corr/FF/build 2015:243:151932 2015:243:152315 2015:240:083820 RA & Dec (J2000) 12h29m 6.6997s +2 03 8.598" U Validity 11/16/15 EHT CorrelaLon - G. Crew 14
ALMA- PICO March 30 dual pol, IQ Mk4/DiFX fourfit 3.10 rev 1009 Amp. and Phase vs. time for each freq., 7 segs, 70 APs / seg (22.40 sec / seg.), time ticks 30 sec 3c273.yizlev, 089-0404, IQ Iy - Qy, fgroup B, pol RR Fringe quality 4 Error code G SNR 49.4 Int time 135.983 Amp 0.867 Phase -124.5 PFD 0.0e+00 Delays (us) SBD 0.008408 MBD 0.003420 Fringe rate (Hz) -0.063063 Ion TEC 0.00 Ref freq (MHz) 226166.1969 AP (sec) 0.320 Exp. Exper # Yr:day Start Stop FRT Mar30l 1120 2015:089 040504.00 040720.00 040612.00 Corr/FF/build 2015:243:151932 2015:243:152332 2015:240:083820 RA & Dec (J2000) 12h29m 6.6997s +2 03 8.598" U Validity 11/16/15 EHT CorrelaLon - G. Crew 15
Some DIFX Details Hardware upgrades at Haystack and Bonn Zoom banding issues PolConvert Other issues with the process today Analysis issues 11/16/15 EHT CorrelaLon - G. Crew 16
Hardware Upgrade at Haystack Upgrade strategy: Recent upgrade to support APP commissioning/ EHT March 2015 campaign is mostly done (slll have bandwidth issues with Mark6 recorders) Significant upgrades for March 2017/2018 based on lessons learned from March 2015 produclon Target is ~8 site playback units with 64 Gbps processed with 4 x 16 Gbps passes 11/16/15 EHT CorrelaLon - G. Crew 17
Some DiFX Details: Zoom Banding ALMA Tunable Filter Bands (Recorded Channels) Overlap! ALMA Sampling is 125 MSps (not 2 n MSps) 1/16 of channel must be discarded HOPS requires 2 n channels 11/16/15 EHT CorrelaLon - G. Crew 19
Some DIFX details, Polconvert Linear Circular ALMA is X/Y, not R/L (quarter- wave plates for 66 antennas was not an allowed oplon) DiFX doesn t need to know it s X/Y not R/L polconvert tool developed to clean up the DiFX mess (whatever we put in v2d and vex for DiFX to run) Works in AIPS only today (and there is no cleanup) CalibraLon data (D terms) needs to be taken 11/16/15 EHT CorrelaLon - G. Crew 20
Road Forward (ALMA) APP deliverables for Cycle 4 Finish programmalc materials, documentalon Help ALMA work out VLBI support procedures Finish Commissioning: Polconvert requires a (good) 4- hour session Finish analysis of July/Aug sessions Validate execulon via Scheduling Block APP enhancements for Cycle 5 (and beyond) Improved handling of fixed delays Source Models Phasing oplmizalon 11/16/15 EHT CorrelaLon - G. Crew 21
Road Forward (SubBands) Channels/SubBands/FFT/DFT/zoom- mode/resampling R1DBE: 15 x 32- MHz channels (64 MSps) per IF/pol ALMA: 32 x 62.5- MHz channels (TFB - - 125 MSps) Overlapped? Not Overlapped? R2DBE(now): 1 x 2048- MHz channel (4x1024 MSps) IniLal development was aimed for 32/64 v 62.5 MHz R2DBE(future): 2 n - MHz channels? (if needed) R2DBE(future): 2 n 5 m - MHz channels? (if needed) DBBC3 (in development) will be even more capable Channelizes the way we originally envisioned SWARM correlator (EHT) is 143*N MSps ALMA flexibility/1- channel R2DBE allow other oplons 11/16/15 EHT CorrelaLon - G. Crew 22
More on ALMA Sub- Band Tuning ALMA sub- band tuning Tunable filter banks tune in 2.0 GHz/2 16 steps (30.517578125 khz) Per- sub- band LO offsevng is required in DiFX in order to compensate for the frequency mismatch DiFX support with zoom mode by A. Deller based on work with ASIAA enoise simulator Small LO offset was tested with SPT equipment (R2DBE) and the Wf/Gg geodelc R1DBE (v1.4) Not using this yet (ALMA Correlator performance) If we do, Large LO offsets are needed 11/16/15 EHT CorrelaLon - G. Crew 23
Post CorrelaLon Work Ahead Changes to difx2mark4 and fourfit remain Possible difx2mark4 enhancements to bewer accommodate DFT vs FFT and zoom channels (e.g. round up # channels to nearest power of 2) Modify fourfit (and other HOPS tools) to use DFT rather than FFT processing (i.e. relax assumplons of 2 n ) and modify internal data structures for the significantly larger data sets required by EHT and ALMA. 11/16/15 EHT CorrelaLon - G. Crew 24