Cancelling a Running Pulsar Observation
So you’ve got an urgent thing to do with the telescope and Jim Palfreyman (0407882718) is using it to observe the Vela pulsar, you can’t get hold of him and you need to do your thing.
If your thing is oiling the gears or polishing the focus cabin (or other such tasks that don’t involve the receiving and recording chain), then just use vdesk to move the dish, lock it off with the key, hit the stop button, and do your thing. When finished restore the key and stop button, and observing will continue automatically. Email Jim (firstname.lastname@example.org) to let him know the stoppage times.
If you are doing some observing then it is imperative you follow these instructions and close down the Vela observation properly otherwise your data almost certainly will be stuffed.
- Go to window 1 or window 2 where the action happens. One window will be running and the other may be scheduled to run. If you have a shell prompt then nothing is waiting. You pretty much have to ^C each window that is not at a shell prompt.
- The top left window is where the recording on hovsi happens. Use ^C to stop that.
- The bottom left window is sleeping waiting for observing to finish. Use ^C to stop the sleep. The telescope used to park when doing this. It no longer does. It is vital for your experiment that you stop this.
- Top right there is a window with date/time and two numbers on each line. Use ^C to stop that. This automatically changes attenuator levels. It is vital for your experiment that you stop this.
- Bottom right has a window running onsource. Use ^C to stop this.
- Go to the Field System (window 5) and change the log file using:
log=<your file> (you will probably do this anyway for your experiment, but this is just a reminder.)
- Email Jim (email@example.com) to let him know what you’ve done.
Starting a Vela Pulsar Observation
- Connect to the vnc session on newsmerd remotely by using:
If the server is not running you need to run vncserver -alwaysshared -geometry 1280×720 preferably from the desktop.
- Window 1 or 2 typically has all the main windows required and window 5 has the Field System which is also needed. If the field system is not running type:
fs (as oper@hobart)
- If the cryo temps aren’t showing regularly in the field system (not important for pulsars, but for others using S/X) type:
- Make sure the dish is parked. In the Field System type:
- Select the receiver using the OTTER interface on Window 3, bottom right (run ./OTTER_interface.sh on newsmerd if you cannot find it). Most observations will be in L-band which is receiver 2. Wait for the platform to move into place (this can be checked on the live page). Make sure it is set to Direct and Quad OUT. The noise diodes should also be off.
- If the receiver platform doesn’t move, the switch on the receiver selector is probably set to Manual. If you are working remotely you may facepalm now, followed by a curse to the person who left it that way. If there is no one at the observatory then go start your car.
- Check the L-band temperatures, voltages and currents. Currently the 20K stage is typically around 12K and the 70K around 82K. The voltages should be reasonably close to what’s stated (say 10–15%) and the currents should be about 3,10,15 mA on 1,2,3 respectively for both A and B.
- Set the local oscillators. This will be in window 3 top left, (or run ./oscillator_gui.sh on oper@hobart if you cannot find it). For an observation at 1376 MHz (a good spot currently) the settings are:
Agilent 4.1 GHz/16 dBm
SML01 876 MHz/7 dBm
SML02 876 MHz/7 dBm
SMY01 off (can be set to 688 MHz/7 dBm if you are at the observatory and want to confirm the signals are being received using the spectrum analyser)
- Configure Palfreymans Delight. This will be in window 3 bottom left, (or run ./PalfreymansClient.sh on newsmerd if you cannot find it). You will want:
Channel 1 on Multifeed RCP
Channel 2 on Multifeed LCP
DAS attenuation to roughly 11 and 14. (This will be automatically adjusted later on).
DC Input to External SLD
DC Attenuation to 5 dB.
- These last two setting will not affect pulsar observation but are standard for other observations.
- Point the telescope to Vela (visible when LMST is between 23:07:44 and 18:04:00). In window 1, bottom right (oper@hobart) type (assuming its 2015 day 183):
- If Vela is not up, it will move there and track when it rises.
- Wait for the telescope to say Tracking on the live page. The oscilloscope traces should be showing pulses and The Flashing Blue Light™ should be flickering. Gently adjust the Balance knob on the DC amplifier above the chart recorder for maximum effect. Turn the volume knob up on the speakers to hear the pulsar. Remind the tour that they are hearing pulses that left the pulsar 1000 years ago.
- If you are only setting this up for a tour demo - then you can stop here, just park the dish (source=stow in the field system) when finished.
- Configure the BG3 cable. Behind cabinet 6 make sure the rainbow cable 1 goes into Ch 1 and 2 goes into Ch 2. The small cable with the large connector should go into Recorder Out. The other end of those three cable should go directly to the back right connector on the board of the vsib box.
- Configure the DAS profile. If remote, then use vncviewer das:1 with the password con… otherwise walk over to the DAS.
- Type P for Profile and L for Load and enter PSR64_N.PRO. After getting Waiting for Setup, press S. If you get a communications error then try loading VSOP_HO.PRO, do Setup, and then reload PSR64_N.PRO. If there are still issues, close down the DAS software and re-launch with VSOP_HO.PRO, then PSR64_N.PRO.
- Press Space so you can see the levels. These need to be 128 ± 4 at worst. The attenuation levels entered previously adjust these. Each 0.1 dB roughly moves the level by 1. Again, you only need to be in the ball park at this point, because it is automatically adjusted later.
- Move onto Vela. In Window 1/2, bottom right type:
- Now to observe until Vela sets, in Window 1/2 top left:
./record 2015_183 1376 set
- Or for something shorter:
./record 2015_183 1376 1h
- Note that 19h of observation takes about 4 Tb of disk space.
- We are now recording. If you have set the levels up correctly, the channel figures should be reading 33,33,17,17 or something close.
- In the other hovsi window (Window 1/2, top right) type:
./agc 2015_183 1376 (but use the date and frequency you are recording in!)
- This will pull data from the last fully recorded 10 second data file and check the levels. If incorrect, it will adjust the attenuators on Palfreyman’s Delight automatically. This is repeated every 5 minutes. This is automatically killed when recording is finished.
- In window 1/2 (bottom left) type:
./stow set (as oper@hobart if it’s not there).
- This waits until Vela sets then kills the onsource script and parks the dish when the experiment is over. Again very important so Vela is not tracked into the limits as it sets. Note that ./stow 5h 50m will wait 5h 50m before ending the experiment.
- Window 1, top far-right has a newsmerd window, type:
at -f fire_cal 1200
- or some time during your observation. This fires the noise diodes for polarisation calibration.
- This only needs to be done occasionally or when the setup has changed.
- If after a power failure, /mnt/sirius16 is not mounted then (as root) type:
- Move to the calibration source:
- On newsmerd in jimp, type in without <CR>:
- You now need the year (e.g. 2014), day-of-year (e.g. 176) and observation frequency (e.g. 1376). Type the following command:
./record_fluxcal 3C353_O 2014_176 1376
- Now hit <CR> on fire_cal, followed by record_fluxcal. Wait 90 sec.
- Now we need to temporarily move north of the source. In Window 1/2, bottom right type:
- Then repeat fire_cal and record_fluxcal commands above, but change it to 3C353_N
- Then repeat using 3C353_S
It is now possible to kick off an observation in advance, even if one is running. The receiver needs to be in place, directories created, noise diode test completed, oscillators set up, DAS configured, and levels roughly set.
As an example, Vela is currently set and it is day 184 in 2015.
- In window 1/2 bottom right type:
date; ~/jimp/velars -s -d; date; sleep 10s; ~/jimp/track vel15184
- In window 1/2 top left (hovsi) type:
date; ~/jimp/velars -s -d; date; sleep 10s; ~/jimp/record 2015_184 1376 set
- In window 1/2 top right (hovsi) type:
date; ~/jimp/velars -s -d; date; sleep 10s; ~/jimp/agc 2015_184 1376 19
- In window 1/2 bottom left (hobart) type:
date; ~/jimp/velars -s -d; sleep 10s; date; ~/jimp/stow set
The 10 sec sleep is just to avoid a race condition where for 1 second, vela is up and down at the same time!
- The Field System will sometimes pop up errors on baseband converters, e.g. “?ERROR ch −308 vd total power integrator overflow”. These are benign but the most annoying aspect is they cause the vnc window to pop to the top and become the focus for your keyboard. Not good! Type one or all of:
check=*,-rx (for temperature warnings)
(Note that -vd and -ifd need to be adjusted depending on the error.)
Observing at Ceduna
The receiver at Ceduna is manually loaded by Bev. These instructions are for observing in C band (4.8 GHz).
- Log onto ares and then type:
- If the field system is not running in a window then type fs.
- To set the oscillators type:
- For C band, the Agilent is irrelevant and SML01 and SML02 should be set to 436 MHz at 7 dB for a central frequency of 4816 MHz. Note the band is reversed.
- For L band 1376 MHz, the Agilent is set to 6.2 GHz and SML01 and SML02 should be set to 428 MHz at 7 dB.
- For L band 1668 MHz, the Agilent is set to 6.2 GHz and SML01 and SML02 should be set to 720 MHz at 7 dB.
- If L band has been set up with circulars then 1668 is most likely optimal.
- To start the DAS, run das, load vsop_ho.pro first (or if a comms error) and then load psr64_n.pro
- To set the attenuation levels run:
- Note that O means On and X means Off. Try to get as close to 127 on the DAS as possible.
- To record, using the field system move on target and then:
- And then move to an appropriate xraid to record as per Hobart.
- psrchive software is on genesis if local processing is required. Otherwise:
- After recording, the data can be moved to a portable usb drive. To mount, as root, type something like:
mount /dev/sdd1 /mnt/usb1
- Then to move data (not as root) type:
rsync —stats —ignore-existing —remove-source-files -v -t -r /exports/xraid/Al_1/pulsar /mnt/usb1
- Note that this rsync command will remove files, but leave the directory structure behind. It’s usually a good idea to clean up that structure afterwards. Also, if this rsync command is interrupted (by ^C or otherwise), then if run again it will pick up from where it left off.
Some Useful Plotting Commands
All commands use a 10 second .ar file that live in the RDSI structure.
- To plot a waterfall diagram:
psrplot -p Y -j Fp -D 10/xs blah.ar
- To integrate a whole 10 sec file:
psrplot -p D -j FTp -D 10/xs blah.ar
- To pick out a single pulse (say number 42):
psrplot -p D -j Fp -c subint=42 -D 10/xs blah.ar
- To pick out a single pulse (say number 42) and centre on 0.5:
psrplot -p D -j Fp -c subint=42 -j ‘centre cof’ -D 10/xs blah.ar
- To pick out a single pulse (say number 42), centre on 0.5, and zoom:
psrplot -p D -j Fp -c subint=42 -j ‘centre cof’ -c ‘x:range=(0.45,0.55)’ -D 10/xs blah.ar
- To pick out a single pulse (say number 42), centre on 0.5, zoom, and put x units in milliseconds:
psrplot -p D -j Fp -c subint=42 -j ‘centre cof’ -c ‘x:range=(0.45,0.55)’ -c ‘x:unit=ms’ -D 10/xs blah.ar
Once the data has been collected it needs to be processed. This is done using the hex cluster which is connected to the RDSI storage. But first the data has to be moved there.
Typically this is run from a vncsession onto hex0. If the server is not running type (as pulsar@hex0):
vncserver -alwaysshared -geometry 1280×720
Log onto sirius (aka mk5ce) and type:
rsync —stats —ignore-existing —remove-source-files -v -t -r /exports/sirius_internal16/pulsar/J0835–4510_S/ /mnt/rdsi/pulsar/process/MTP26M/J0835–4510_S; date
Next we need to process. Log onto a hex node (say hexb) and:
mv 2014_123 /mnt/rdsi/pulsar/process/MTP26M/hexb/J0835–4510_S
This moves the directory into a spot for hexb to process. This is not strictly necessary (i.e. all files could be in one spot) but makes things efficient when processing large amounts of data. It does require manual supervision but it is still way faster. Now start processing:
lbapsr_master.csh /mnt/rdsi/pulsar/process/MTP26M/hexb Hobart 8
This can be repeated for as many hex nodes as required.
After a directory is completed:
mv 2014_123 /mnt/rdsi/pulsar/MTP26M/J0835–4510_S
This puts it in with the rest. Now go there:
Note that in all the following commands, do NOT use a trailing / on the directory name! To move all the raw files into their own directory run:
To create .ft files in a timing directory:
To create today.ft (which is an integration of all the pulses for the day):
To create an all.tim file in the timing directory:
To run tempo2:
After deleting bad observations in tempo2, save a new .tim file in tempo2 format as fixed.tim. Save a .par file as today.par.
To create the all.spa file in the spa directory:
Note that all of these commands only run on one core. You can background them and run up to 8 on a single hex machine.
Down the track and it’s time to free up space by deleting those large lba files, type:
This will search the spa output for bright and consecutive pulses and moves those .lba files to a raw.keep directory. Once you’ve verified this command has done that to your satisfaction, copy/paste the rm -rf 2014_123/1376/raw command to delete the raw directory. Please use the copy/paste - this prevents typos and deleting the wrong directory!!
To pull out good timings of all data so far so this in the J0835–4510_S directory:
cat 2014*/1376/timing/*fixed.tim |grep -v C › allsofar_fixed.tim
Then to pull out 5000 random samples:
shuf -n 5000 allsofar_fixed.tim > allsofar_fixed_5000.tim
The use tempo2 to see it all:
tempo2 -gr plk -npsr 1 -nobs 10000 -us -grdev 20/xs -f /imports/rdsi/pulsar/MTP26M/J0835–4510_S/J0835–4510.eph allsofar_fixed_5000
To create daily.tim (which is the arrival times of all the today.ft files):
To use tempo2 to see these daily summaries (way less points to deal with and great for creating an ephemeris):
tempo2 -gr plk -us -grdev 20/xs -f /imports/rdsi/pulsar/MTP26M/J0835–4510_S/J0835–4510.eph daily.tim
The Current Program
Observing of the Vela pulsar (J0835–4510) is currently occurring whenever the 26m telescope is otherwise free. Vela is visible 17 hours a day (LMST 00:00–17:00) and will generate 4 Tb of data if a full day is observed. This is being stored on rdsi. In the past the raw “.lba” files were processed and discarded. We now plan to keep them so further processing can be conducted on interesting events.
This Vela pulsar collection is currently by far the largest single pulse archive in the world. And expanding. There is plenty of scope for further study on the data collected. If this interests you then please contact Jim Palfreyman (firstname.lastname@example.org).
Old notes - this may no longer be accurate
Mt Pleasant is equipped with a dual-pol, 64 MHz bandwidth baseband recorder and a 48 core cluster that can be used together as a pulsar coherent dedispersion system. The telescope signal is routed into the ATNF data acquisition system (DAS) in the same way as for VLBI or correlator observations. The output of the DAS is fed into the eVLBI recorder via the “BG3” cable and the eVLBI software (“vsib_record”) is used to record data to “hovsi” which is connected (via a dedicated ethernet cable) to the 24 Tb host “sirius”.
The data can be processed at any time, during or after the observations. A set of scripts have been written to automate the transfer and processing of baseband data for pulsar observations. When running, these scripts will copy data from the XRaid to the Origin 3400 one file at a time and run the coherent dedispersion and folding software (“dspsr”) on each file, using an ephemeris from the collection kept in $TZPAR. The Origin 3400 originally had 24 processors but one of the blades developed a fault before it was given to the observatory and will not run reliably. It is best to leave this blade turned off (there is a red sticker over the power button) or the whole machine will crash on average once a day. The remaining blades have a total of 20 processors. At best (for low dispersion measures) this allows processing at a rate roughly 1/3 real-time. There is 1GB of memory per processor, which is usually enough but can be insufficient if very high-resolution processing is attempted (single pulses with lots of frequency channels and phase bins for example).
Processed files are transferred across the network to a TPAC-based mass storage device that is mounted as /imports/tpac/pulsar on the radio astronomy server, ares.
When using vsib_record to take pulsar data, please ensure that the DAS is running the profile called PSR64_N.PRO. This profile disables the automatic gain control system, which is essential when observing bright pulsars (otherwise, the AGC tries to remove each incoming pulse and you end up with a large artificial dip on the trailing edge of the profile). This profile operates the DAS in 64 MHz mode, which requires a special cable (the BG3 cable) to interface between the correlator ports on the DAS and the input to the VSIC converter card. This is quite different to the “normal” eVLBI mode of operation where the S2 data output port it used. The vsib_record program should be run with the arguments “-m 2 -w 64″ to ensure correct recording.
Data can be recorded to any of the mass storage devices connected to the eVLBI machine, “hovsi”, in a subdirectory called “pulsar”. For example, to use the first set of XRaid disks, record to “/data/xraid0_0/pulsar”. Within the pulsar directory, you should construct a series of nested directories named after the source, day of year and observing frequency. This directory structure is critical. At the moment, no information regarding the telescope coordinates or system setup can be stored in the baseband data files produced by vsib_record, so the pulsar processing software reads all the information it needs from the names of the directories in which the data are located. You should start with the source name in J2000 format. For the Vela pulsar this would be “J0835–4510″. Inside the source directory you should create one (or more) directories named for the date, in the format “YYYY-DOY”. For example, 2007_183. Inside the date directory, you should create directories named after the observing frequency in MHz. Files should be recorded inside these frequency directories.
Once data is flowing to a storage area on hovsi, you can begin processing it with the Origin. See Aidan for details as the system is still somewhat experimental.
- We now need to the relevant directories on hovsi for calibration and observation. In the Window 1/2 on newsmerd there is a hovsi terminal at top left, type:
- If these directories are not mounted, say after a power failure, then the special interface to sirius may need to be configured (as root):
ifconfig eth0 192.168.1.2 (unlikely to need this)
- You now need the year (e.g. 2014), day-of-year (e.g. 176) and observation frequency (e.g. 1376). Type the following commands:
- For calibration we need to temporarily move off Vela. In the Field System (Window 5) type:
- We now need to run the noise diodes for a calibration. In a newsmerd window type:
calu -m sam26m (and press return when prompted with rakbus.)
samtest -c 5 -d 0.5 -i 0 -n 1 -s 120 -f 30 (followed by sam26m but DON’T press return yet!)
- Now on hovsi (Window 3, top left):
vsib_record -m 2 -w 64 -f 10s -t 90s -o CAL_J0835–4510 (but press Return on samtest first, then this one.)
- Wait until vsib_record has finished. samtest will follow soon after.
- Move back onto Vela. In the Field System (Window 5) type: