Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.atnf.csiro.au/vlbi/dokuwiki/lib/exe/fetch.php/difx/difxwsbonn.pdf Äàòà èçìåíåíèÿ: Thu Sep 25 05:47:08 2008 Äàòà èíäåêñèðîâàíèÿ: Tue Apr 12 07:08:31 2016 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: mdi

Status of the Bonn Correlator

MPIfR ­ BKG correlator

Walter Alef Helge Rottmann


Bonn correlator centre
· MK IV hardware correlator (XF)
­ ­ ­ ­ 15 Up 16 32 up ­ Da Mark 5s: 8 x 5A, 4 x 5B, (1 x 5B+, 2 x 5C) to 1 Gbit/s data rate frequency channels lags (default continuum) to 1024 (tested) ta export:

· To AIPS using MK4IN · To geodetic database (CALC/SOLVE etc.)
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


Bonn correlator centre
· MK IV operations
­ GMVA observations at 83 GHz
· 13 stations · 512 Mbit/s · 10 days per year

­ MPI-based observations
· e.g. observations above 100 GHz

­ EVN observations with special requirements ­ Geodetic VLBI (~50%)
· Handled by partner BKG and Uni Bonn
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


DiFX Software Correlator
· Standard DiFX implementation
­ Input from Mark 5A/B or 20 TB Raid ­ 1 Gb via Internet possible + 1 Gb from Eb ­ Cluster with 176 compute cores in 22 nodes
· 20 Gbit Infiniband interconnect · 1 Gb Ethernet

present status

· Manpower:
­ open postdoc position with geodesists ­ H. Rottmann & W. Alef: very limited time available at present for work on DiFX
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


DiFX plans for 2008/2009
· Realise production system similar to MK IV / VLBA · Verify geodetic quality of correlation
­ Implement phase-cal extraction ­ File output to MK IV format database

· Phase-out MK IV correlator · Add >~ 4 Mark 5C · Convert Mark 5A/B to Mark 5C
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


DiFX plans for 2008/2009
· Extend cluster to ~ 500 compute cores (November 2008)
­ More than ~10 stations @ 1 Gbit (1:1 in time)

· Direct rack cooling · rebuild correlator room · Manpower:
­ geodetic postdoc (hopefully soon) ­ H. Rottmann starting 03/09 full time on DiFX ­ J. Anderson for Lofar application
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


Upgrade plans for hardware
· ~1 6 Ma rk 5 C · Internet connectivity > 1 Gbit/s
­ ­ ­ ­ Depending on In 2012 10 Gb Big disk buffer Asynchronous correlation demand it/s ? : hundreds of TB data transfer + Mark 5C mixed

9 Sept 2008

DiFX WS @ Max Planck Institute for Radioastronomy


Geodetic correlation: needs
· massive subnetting · typically only 1 frequency setup
­ except for cases with LO offset

· one start time, different stop times · output 1 file per baseline/scan + station file + file root (vex excerpt) incl. Model translation program needed · "version control" of output files
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


MK IV correlator VEX files
· observe vex file + clocks + eop + pcal mode · log vex file: station -> scan name > 8pack > file name (possibly byte position) · correlator vex file: possible correlator setups · experiment vex file: setup parameters for each experiment (int. time, lags, polarisations) · vex file with setups for phase-cal extraction
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


geodetic correlation
· · · · · · · · scan-based parallel sub-net correlation n x integration time fits in 1 minute scans of 20 s +, (soon even shorter) operator start script with scans mount requests for Mark 5 modules status display of playbacks and correlator export with data selection to CALC/SOLVE via DBEDIT
DiFX WS @ Max Planck Institute for Radioastronomy

9 Sept 2008


MK IV Output files
· Each scan creates a subdirectory with scan name · Each correlation of scan creates root file with version as file extension (date encoded to base 26) · each station has file with state counts and phase-cal · each baseline has file with raw data · each run of fringe fitter creates file for baseline and sub-frequency (S/X LL RR ...)
9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


geodetic verification: step 1
· · · ·

·

30 s of data of 4 stations, 1 lower sideband The data was exported to AIPS comparison was done in MIRIAD The total phase as a function of frequency (delay) and the total phase over time were calculated The differences between the results of DiFX and the MK IV correlator were calculated
DiFX WS @ Max Planck Institute for Radioastronomy

9 Sept 2008


geodetic verification






Results: the mean phase difference is well within 1 of zero, and the r² values indicate that far less than 1% of the variation in the data can be explained by linear regression. Thus, no significant differences exist between the phase outputs as a function of time, between the MK IV and DiFX correlators.
DiFX WS @ Max Planck Institute for Radioastronomy

9 Sept 2008


9 Sept 2008

DiFX WS @ Max Planck Institute for Radioastronomy


9 Sept 2008

DiFX WS @ Max Planck Institute for Radioastronomy


Next step of comparison
1 h 3 station INT3 (Wettzell, Ny ålesund, Tsukuba) Problems encountered so far: "Illegal VLBA format" of translated Ts data. Non-standard track assignments of Ts Mark IV formatted Mark 5 data starting on integer second before scan starts. manual phase-cals.

9 Sept 2008 DiFX WS @ Max Planck Institute for Radioastronomy


Next step of comparison






DiFX data export and fringe fitting via AIPS to data-base. Problems with importing to AIPS encountered due to "unusual" (=geodetic) frequency setup. Under investigation. Bonn and Leipzig analysis centres will perform CALC/SOLVE comparison. Results so far: looks good on first inspection
DiFX WS @ Max Planck Institute for Radioastronomy

9 Sept 2008


full geodetic verification
· phase-cal · 24 h full geodetic run ~ 8 stations · documentation of algorithm Later: · output to MK IV format

9 Sept 2008

DiFX WS @ Max Planck Institute for Radioastronomy