The Multibeam correlator is a cut-down version of the correlator developed for the Parkes HI multibeam survey. It is a 2-bit digital autocorrelation spectrometer that takes its input signals from the LBA DAS (Data Acquisition System). The Parkes correlator had 16 boards, one for each beam and three for cross correlation interference detection, while the Hobart correlator has 4 board. Each board produces a maximum of 2048 spectral channels and the maximum bandwidth for the correlator is 64 MHz. For bandwidths of 16 MHz and less the correlator boards can be combined to produce more spectral channels across the band. The DAS takes a maximum of two input signals (IFs) and so at bandwidth of 32 and 64 MHz the best configurations have 2048 channels across the band, while bandwidth of 16 MHz and less 4096 channels per IF can be obtained.
Setting up for a Multibeam correlator observation
- Select the required receiver on the PC, make sure that the quadrature hybrid is set appropriately (normally out)
- Move the required receiver to the focus position.
- Set the frequency and level on the agilent to the appropriate settings for the line to be observed. On hobart, at the command line, type,
oscillator_gui.sh, to get the agilent settings window.(NOTE: Updated for new system)
- Connect the IF signals from the receiver into the DAS
- Take the signals from A4/A5 into the 30 dB amplifiers and then into the rotary attenuators (all in rack 3).
- Take the outputs from the rotary attenuators to the RF inputs on the frequency translator (rack 6)
- Take the output from the Rohde and Schwarz SMY (use this as it can be controlled remotely by survey_ng), into a splitter (rack 5) and then into the the LO inputs on the frequency translator. Make sure that the SMY is set to the appropriate frequency and a level of 10 dBm.
- Take the IF outputs from the frequency translator and connect them into the 96/16 MHz bandpass filters at the top of rack 6.
- Take the output of the filters and connect them to the connectors marked 3A/3B (these signals will now appear on connectors 6A/6B in rack 3).
- Take the output from 6A/6B in rack 3 into the inputs of the programmable attenuators in rack 2
- Take the output from the programmable attenuators and connect them to the square-law detector inputs in rack 3.
- Take the output from the square-law detectors in rack 3 and connect them to the connectors marked 1A/1B.
- Check that connectors 3A/3B in rack 1 are connected into VFC #0 and 1 (these are hardly ever touched and so should be OK).
- Select the appropriate DAS profile (e.g. if you want a bandwidth of 4 MHz you need COR4000.PRO). LOAD the profile and remember to do SETUP too! (The notes on using the new DAS are here
- Adjust the rotary attenuators so that the levels on the DAS display are in the centre of the range.
- Adjust the levels on the programmable attenuators so that the power meter on the square law detect reads between −5 and −4 dBm.
- Make sure that the CAL (noise diode) can be controlled by sam26m. To do this, type the command
calu -m sam26m -r rakbus in an xterm on smerd (logged in as observer).
- The next step is to start the correlator software on newsmerd. To do so, type
./cor_ng from the
/home/observer directory. This will start up a number of processes in xterms, including a number of minimized ones. If the script fails to produce the
SPD windows on your desktop, (with
CONFIG minimized), then it’s likely there is an existing instance of the correlator software running which needs killing. Use
ps -ef | grep CORDAT and so on to get rid of these (after making sure that no-one has a timed schedule waiting!).
- Note that synccc and dummsy MUST be running for the correlator to be able to send data to newsmerd. So you should see open windows with messages from each of these programs. If not, search for zombie processes with those names and kill them. Then restart. You can restart just dummsy from a window on newsmerd, if it has died.
- In the SPD window, set the graphics device to /xs to view the spectra. There are no GUIs for the correlator when running on newsmerd. Communication with the correlator is accomplished either by scripted commands (using
survey_ng) or by
dscom (a command line interface).
- To test that the correlator is working correctly, it is useful to test it with
dscom before starting a schedule. In a terminal on newsmerd, type
dscom. Then issue the following commands into the prompt. [Not recommended for beginners, see KYA307 notes below.]
stop (Wait until the correlator responds by saying it has stopped)
fc (To close any file the correlator may have been writing to)
conf (To test the configuration of the correlator. To proceed, you need to check the output in the CONFIG window and press return. If the configuration shown does not match what you want, then you’ll need to set the configuration by
conf das_xxyy_16_4096 (replacing
das_xxyy_16_4096 with the desired configuration).
go (To start the correlator)
Check the output spectra in SPD (with
sel 1 2). If the spectra look like a bandpass, then all is well. Stop the correlator by entering
stop in DSCOM, followed by
- If the output spectrum is all zeros, then it’s likely that the DAS is not connected to the correlator. This is commonly seen after LBA observations as the VSIB recorder uses the correlator outputs and may not have been disconnected afterwards. To fix this, you need to go to the rear of the DAS module and reconnect the 1-inch ribbon cables marked CORR 1 and CORR 2 to the DAS’s output ports, unplugging the VSIB cables if necessary.
- If the output is a very distorted bandpass, there’s likely either a mismatch between the DAS and correlator profiles, or a configuration error with the correlator. First, check that the DAS profile is correct, that it has been set up correctly and that the levels are ok. If they are ok, try re-issuing the
conf command in DSCOM, then restarting the correlator with
go. If this still produces a distorted output, try re-loading a different DAS and correlator configuration and then returning to your desired configuration. The recipe below should do this but remember that pauses are required between each
stop command, and that each
conf command requires you to press return in the CONFIG window before proceeding.
(set up the COR8000.PRO DAS profile)
(set up the intended DAS profile - COR4000.PRO in this example)
If it still fails, it is probably necessary to reboot the correlator. Best done by pressing the white button on the correlator computer. It’s normally concealed behind a panel, which is directly above the “Multibeam Correlator Rack Power Supply” which is at floor level in rack 5.
After the correlator is working properly and you have quitted from
dscom, you can start a schedule with following commands as observer on newsmerd
/usr/local/pkg/correlator/survey_ng -a sys26m -i 0 -n 2 -r rakbus -s sam26m -f hart6ghz_n188.out (Where hart6ghz_n188.out is the name of your schedule)
After the schedule finishes (or if you kill it with a Ctrl-C), survey_ng unfortunately kills the dummsy program. This means that the correlator will not function, until that process is restarted. The easiest thing to do is to just
cd ; ./dummsy in an xterm on newsmerd. NOTE: You can’t just
./cor_ng again, as this will not work while any surviving processes of the previous run are alive. If you can’t get the correlator to start with
./cor_ng find and close any terminals called
dummsy before restarting the correlator.
ps -ef | grep synccc and the like will help.
Creating a schedule
To schedule a spectral line observation with the multibeam correlator you need to use the
corsched.pl perl script that lives in the
/home/observer/corsched directory on newsmerd. The input files to
corsched.pl have to have lines in the following format :
srcname hh:mm:ss.s -dd:mm:ss coord_mode velocity int_time obs_mode cor_mode file
For example the line
mon_r2 06:05:21.6 −06:22:27 b1950 0.0 600 poff monr2 auto
means that a 600 second (10 minute) observation is to be made of the source Mon R2, whose B1950 coordinates are 06:05:21.6, −06:22:27. The
poff observing mode means that a 10 minute integration will be made at an offset (reference) position, followed by a 10 minute integration at the onsource position. You are prompted for the offset for the reference integration when the
corsched.pl script is run. The
KYA307 students 2012 please read the notes below:
MOST IMPORTANT: If you change any cables or wires, please put them back where they were before you leave. You should not have to change any wires, assuming that you don’t use the S/X receiver. It is fine to do the exercises using only the cooled, multi band receiver.
Your data goes into a file with a name like c120881126.rpf in directory /home/observer/dat. The file name has c for correlator data, 12 for 2012,
088 for the day of the year (March 28) 1126 for the UT start time, and file type rpf for radio-physics fits format. Files in .rpf format can be read by
ASAP analysis programs.
— For observations using the cooled, multi band receiver, you need to set the agilent oscillator for the first LO. Set it by opening a window on newsmerd, then :
ssh -X oper@hobart
Now enter the Agilent frequency as:
oscillator_control.sh Agilent 14.3 ghz
you can also enter the sml frequency as
oscillator_control.sh SML01 637 mhz
you can check the frequency and level with a line like
Now exit from hobart. Do not change the level (dBm) of the oscillator, it should stay at 16.0 dBm.
Note that 14.3 is a good frequency for observing near 8.5 GHz sky frequency, since 14.3 - 5.2 = 9.3. So a line at 8.585 GHZ will be mixed with 14.3 GHz to come out at 5.715 GHz, then mixed again with 5.2 GHz to come out at 615 MHz. Then you can set the SML oscillator at 637 MHz and send that LO to the frequency translator to get 1152 MHz as the band centre corresponding to the original sky freq of 8.585 GHz. Of course you will use a different SML frequency if the line freq is different. And you can change the Agilent if the line is more than a few hundred MHz from 8.5 GHz. Try to keep the SML (third LO) frequency in the range 600 to 850 MHz, and keep the Agilent (1st LO) high enough above the line so that 5.2 MHz comes out well below the line after the first mix, i.e. 5.715 > 5.2 in the example above.
A good source to observe as a calibration and system check is 30 Dor, a bright HII region in the LMC. Its position is
05:38:42.3 −69:06:03. (J2000)
It should be bright in the continuum, so you can do some scans on it using bruce to make sure you’re seeing it. Since it’s so far south, it is circumpolar for us (latitude 42.9 - ( 90–69 ) = 21.9 deg is the elevation at inferior transit (at LST=17:38), 63.9 is the elev at superior transit (at LST=05:38). 30 Dor should show some bright recom lines in the range 8 - 9 GHz (see Peck et al. 1997, Astron. J. 486, 329).
Simon and I recommend that you should not follow the instructions above to run the program DSCOM. This program has a habit of locking up the pipe between the correlator and newsmerd. You can check that the correlator is working by running a small, dummy file as the input to survey_ng.
I have made one that should start the correlator and run two short, dummy scans. The output should appear on the graphics window, under the control of the spd window. Start the test program with:
/usr/local/pkg/correlator/survey_ng -a sys26m -c newsmerd -i 0 -n 2 -r rakbus -s sam26m -f KYA307-test.out
(NOTE: This should be executed in
observer@newsmerd:~/corsched Which means you will need to change to this directory using
cd /home/observer/corsched if you are not already there.)
You should see messages in the dummsy and synccc windows, including the file name created for the output.
After about 30 seconds the correlator will dump a scan to newsmerd. It will not plot automatically. To see the most recent spectrum, go to the spd window and set the graphics device to xs by typing /xs asuming this has not already been done, then type any command, e.g.
sel 1 2
which just selects for plotting spectra 1 and 2. Then spd will continuously plot the spectra as they come in. You can automatically average the spectra for each scan by typing
If you want to save the average at the end of the scan, type
before the next scan starts. You can then subtract the saved scan from the current scan with
SPD commands can be found here:
When survey_ng finishes, or if you kill it, the correlator communication program dummsy will die. Start it again in the window where it was running, by typing
on the command line. If the window closes, you can open a fresh window with
then in this window type dummsy.
If you have trouble with the tasks started by ./cor_ng, you can control-c in their windows to kill them. When they are all gone, you can try starting cor_ng again. Sometimes the processes disappear, but they are still running. To kill zombie processes like synccc, CONFIG, XFER, CORDAT, and dummsy, first find out the process id by doing
ps -ef | grep NAME
Where NAME is the name of the process, synccc, CONFIG, XFER, CORDAT or dummsy.
then kill the process with a command like
kill −9 nnnnn
where nnnnn is the PID (process identification) number shown by the ps -ef command. Be careful what you kill !