User Tools

Site Tools


correlator:espresso

Espresso

Espresso is a suite of scripts designed to take the drudgery out of file-based correlation. The idea is to automate as much of the processing as possible. While providing a rapid way to set up and run correlation jobs, the scripts assume the user is 'intelligent' (that's another way of saying the scripts are dumb) - check their output. All the scripts will give some help if initiated with the –help switch. Please read those help messages before using. A typical Espresso correlation would look like this:

disk_report.py > ~/disk.json

creates a summary of the disks in $DIFX_MACHINES used for storing baseband data.

Transfer the data to the correlator, e.g.:

ssh rey075@cass-01-per.it.csiro.au
globus-url-copy <options> -r file://<path>/v255z/ATCA/ sshftp://user@magnus-data/<path>/v255z-At/

Following the naming convention for the directories on the correlator (<expname>-<Telescope>) is important.

This transfer can be automated using the make_copyfiles.py and run_copyfiles.sh scripts on cass-01-per:

ssh rey075@cass-01-per.it.csiro.au
scp cormac@magnus:disk.json ~          #be sure to get a newly generated version of disk.json
cd ~/transfers/v255z
dtsum v255z     # alias for du -cs -BG > du.txt
make_copyfiles.py ~/disk.json ./du.txt
run_copyfiles.sh copy_v255z_*.sh

make_copyfiles.py will identify destination nodes with sufficient space and create a bunch of shell scripts with the relevant transfer commands. The run_copyfiles.sh command can then be used to initiate these transfers under screen. Use screen -r to view the running transfers.

Update the disk report once the data have been transferred.

disk_report.py > ~/disk.json

Get the .key file from ATNF and run through our local version of SCHED (this is important to get the best station positions for Mopra and ATCA). This is no longer generally needed unless the vex file needs modification.

cd $CORR_HOME/2015/September/v255z/sched
getatnf v255/v255z/v255z.key             # alias for wget ftp://ftp.atnf.csiro.au/pub/people/vlbi/v255/v255z/v255z.key
sched < v255z.key
cp v255z.vex ../
cd ../

Currently “prestart = 0” can be inserted into .key file to avoid scan start times being 5s different from vex file used for observations (SCHED version issues). Also may need certain setup files from ATNF if these are not available locally. But should always get station positions from database at Curtin as this ($STADB) should be up-to-date with the latest station positions.

You may also have to edit the TRACKS section in the vex file to add the following line:
S2_recording_mode = <samprate>x<nchan>-<nbits>;
e.g. for 32 Msample/sec recording, 4 channels, 2 bits, this would be:
S2_recording_mode = 32×4-2;

If an ATCA pad other than the default W104 was used, then update the station position. Similarly update ASKAP position if necessary. (If Tid used a station other than what was specified in the vex file, it is probably best to re-run SCHED. However you could use updatepos.py to swap between the two 34m antennas, DSS34 and DSS45, as these have the same slewing parameters etc.)

updatepos.py ATCA AT_W102 v255z.vex

Create an input (.datafiles) file for lbafilecheck.py:

disk_exper.py v255z ~/disk.json

Create filelists, machines, threads and run file with lbafilecheck.py:

lbafilecheck.py v255z.datafiles

Note: mk5scans.py can add time ranges to Mark5 file lists after creation, but this is deprecated and not usually necessary.

Edit a .v2d file as required for your experiment. (Note: getEOP.py to download the latest EOPs.)

getEOP.py <date> >> v255z.v2d

where <date> is the start date of the experiment in MJD or VEX format.

Run vex2difx and check that number of jobs, etc. look reasonable and .v2d and .vex files parse correctly.

vex2difx v255z.v2d

Start a clock search job (output will go $CORR_DATA/<expname>/clocks). Choose appropriate mjdStart and mjdStop, scan or source selection in the .v2d file.

espresso.py -c v255z_1

Set up the pipeline directories if they don't exist yet; can use the bash alias: mkpipe v255z
Run the LBA AIPS pipeline to get clock estimates (either use mkpipe or copy a template .inp file from ~corrtest/progs/Pypeline/lba_template.inp and edit appropriately).

cd $CORR_DATA/v255z/clocks
difx2fits v255z_1 V255Z.CLK.FITS
LBA.py $PIPE/v255z/in/v255z.inp

The fitted clock residuals will be in $PIPE/v255z/out/v255z.clocks. Or you can see the raw data graphically in $PIPE/v255z/out/v255z_FRING_DELAY.pdf and v255z_FRING_RATE.pdf.

If you need to inspect the FRING output, e.g. to determine the time of a clock jump, you can use the Pypeline script:

clocks.py -p "1000 V252AZ UVDATA 1 1"

where 1000 is the AIPS user number, and “V252AZ UVDATA 1 1” is the AIPS name, class, disk and sequence number.

Update the clocks in the .v2d file:

cd $CORR_HOME/2015/September/v255z
updateclock.py -e 2012y067d04h00m00s -s 'AT,HO,MP' -o '0.5,-.01,0.034' -r '5.1,-3.3,1.2' -f 6642 v255z.v2d

Units for the clock offset (-o) are microsec, the clock rate (-r) mHz, and the observing frequency (-f) MHz.

If this experiment requires multiple correlation passes, then make copies of the .v2d file for each pass, e.g.: 'v255z-line.v2d' and 'v255z-cal.v2d'. The naming convention '<expname>-<passid>.v2d' must be followed.

Run the production correlation job

espresso.py -a v255z-cal

Go make yourself an espresso. You may have time for more than one.

Occasionally jobs will fail to complete. chklogs is a bash alias to check which jobs did not appear to complete as expected, based on the messages at the end of the *.difxlog files. Specify the output data directory, or by default chklogs uses the current working directory.

Run the pipeline on the complete dataset, and transfer the pipeline data and summary to the wiki:

cd $CORR_DATA/v255z
difx2fits -d V255Z.FITS
LBA.py $PIPE/v255z/v255z.inp
cd $PIPE/v255z/out
lba_feedback.py v255z-cal.wikilog v255z-line.wikilog > v255z.txt
archpipe v255z

Notes:

  • Generally may want to set options/output file name in difx2fits (run with no options for help). Especially important for generating output from correlation with multiple phase centres (option –phasecenter <p>).
  • lba_feedback.py can now take multiple .wikilogs, e.g. for different correlator/pipeline passes. Naming convention is {expt}-{pass}.

Archive the correlator output directory using the archivec.py script to access the Pawsey data store.

<code> archivec.py $CORR_DATA/v255z/ /projects/VLBI/Archive/LBA/v255 </code>

See the archive page for notes on distribution of output data to PIs.

On cortex, link the .FITS files from the archive to the ftp site so the PI can download them: <code> ssh cormac@cortex.ivec.org ln /pbstore/groupfs/astrotmp/as03/VLBI/Archive/Curtin/v255/v255q/*.FITS /pbstore/groupfs/astrotmp/astronomy/anonymous/v255q </code> Note: cortex.ivec.org has a special “feature” that makes it necessary to make a soft link in one's home area to the /pbstore partition for transfers to work: <code> ssh cormac@cortex.ivec.org ln -s /pbstore ~/pbstore </code>

correlator/espresso.txt · Last modified: 2016/05/24 12:47 by cormac