A GUIDE TO JODRELL MULTIBEAM POINTED HI OBSERVATIONS

version 1.0 10th June 2000. Created

version 2.0 14th June 2000. html'ed

The Multibeam and its software were primarily designed for scanning. As such, there is a dearth of infomation, software and expertise for pointed observations. This document is written to help you plan, make and reduce pointed observations, and maybe avoid some of the mistakes, frustrations and time-wasting (our own and anyone around kind and foolish enough to help us) we encountered. Please do not hesitate to contact us with any comments or questions. It is intended that this be a 'live' document, with other observers adding material, correcting any mistakes or changes etc. Please update the version number each time you do this.

Observational planning

The normal setup is 64MHz bandwidth, 1024 channels x 2 pols x 4 beams.

There is horrendous broad-band interference between 1397-1391MHz (5000-6100km/s). You will not get any useable data in this region. There is marrow-band interference at 9800km/s and 11100-12000km/s.

We got good baselines at night, even at elevations 10-20deg; we got terrible baselines (10MHz ripple, 100mJy-1 Jy amplitude) in the daytime even when away from the sun.

We got 6mJy noise/channel (after Hanning smoothing once) in 30 mins. But the overall flux calibration may be 20% down, based on UGC3574 (from Lister Staveley-Smith's 1985 Ph.D thesis, which lives in the Library).

Observational setup

We used 64MHz bandwidth centred at 1394.5MHz (Sampler set at 128MHz, LO set at 388.625MHz). This was fine, but includes the bad interference at 1370MHz, and leaves 1420MHz a bit close to the edge of the band for good bandpassing. 1400MHz might be a better choice.

Pointing strategy

Beams are 12' FWHP, 20' apart in a rhombus


     2
    1 3
     4

Normally you will put the source in each of the four beams in turn for some fixed number of 5-second 'scans' or 'cycles', and make some number of these complete circuits. The telescope takes ~30s to move between beams, so the decision is a trade-off between this overhead (and the strain it puts on the telescope), and getting reference beams close enough in time to get good baselines. We tried two strategies:

beams 214412 in that order, = 2 circuits x 3 beams x 60 cycles (~35 mins for 30mins data)
beams 1234123412341234, = 4 circuits x 4 beams x 24 cycles (~45 mins for 32mins data).

We did not find any great improvement in baselines when we went to the shorter dwell-time in each beam. However, in the LIVEDATA reduction software, you cannot look forward or back more than 80 cycles for reference beams. This suggests 40 cycles as a good dwell time in each beam.

Sched files:

You will need a target.sch file for each source. Here is one giving two circuits of all four beams with 40 cycles/beam:


 $ unit     1
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 1
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 fitsname = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     2
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 2 
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     3
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 3
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     4
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 4 
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     5
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 1 
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     6
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 2 
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     7
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 3 
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   24
 raj = 20:13:57.7
 decj = +29:02:05
 track
 $ unit     8
 freq = 1394.5
 bandwidth 64
 config = mb4mod
 refbeam = 4
 enable refbeam
 fcc_p_trk = disabled
 fcc_rot= 0.0
 source = C132511
 cycles =   40
 raj = 20:13:57.7
 decj = +29:02:05
 track
 closef
 stop

To make them en masse, we made a template.sch with AAAAAAA,BBBBBBBBBB,CCCCCCCCC for RA, dec, name and used e.g.


cat template.sch | sed 's/AAAAAAA/NGC1058/' | sed 's/BBBBBBBBBB/02:43:29.7/' 
   | sed 's/CCCCCCCCC/+37:20:30/' > NGC1058_3.sch

You have to ftp these onto ARTHUR in the MULTI$SCH directory, then call them from TKMULTI.

Observing

Set up mbcor and tkmulti as per Notes on running the Multibeam Correlator Software with tkmulti. Note that you log in as multi, but the file owner is mbiers.

Livedata can be persuaded to reduce pointed observations, as follows.

Set up livedata as follows:
Read directory /arthur_data/rpfits/data
Write directory /scratch/oddjob_1/mscals/MYDATA
Auto-queue OFF
Bandpass calibration ON
Monitor output OFF (doesn't work for now, but it would be very nice to have it)
Write data ON
Statistics OFF
Integrations 1
Log Interval 200
Observation mode DWELL
Prescale mode OFF (see below)
Bandpass mode MEDIAN
Velocity frame BARY
Fit order 0
Tukey 25% OFF
Bandpass interval determines how often (in cycles) the bandpass is recalculated. We found 10 was ok
Number of precycles 80 (this was the largest we could get Livedata to accept)
Number of postcycles 80 (")
Check field name OFF
Check central beam only OFF
Check time OFF
Check position ON, min 10, max 300, jump 300 (or anything > 40) (Jump is the maximum allowed difference in positions before data gets discarded.)
Gridder ON, using oddjob
In the Gridder window:
- Parameter set GENERAL
- Autosize ON
- Image size 4' x 4'
- Channels 1-1024
- Projection SIN
- Av type see below
- Smoothing radius 6'
- Clip fraction see below
- Weighting 0
- Polarisation 0 (honest)
- Processor oddjob
  use 50% (DON'T put this up, and reduce it if running more than one gridder/gridzilla)
- Truncated 16-bit OFF
Search path /scratch/oddjob_1/mscals/mydata
File suffix mscal
Output FITS directory /scratch/oddjob_1/mscals/mydata/fits
Output FITS file name livedata.fits

When the baseline is good, medianing is fine, mean with say 5-10% clipping gains you maybe 20% S/N. When the baseline is bad, all this is irrelevant compared with getting as smooth a ripple as possible, on top of which you might hope to see your galaxy. So we turned off the prescaling and used a mean with no clipping, all in the hope of a smooth baseline. But you should experiment.

All directories must exist before livedata or gridder try to write to them!

Then you can select chunks of .mbf files and feed them to livedata; they will come out as mini-cubes of processed fits data.

If you wish to combine different .mscal files to make a single spectrum, run gridzilla separately, and select the files you want with mouse and CTRL-mouse as on a PC.

To look at your spectra

Run kview& from a terminal window.

In kview, select View, then Profile Mode: Line.

Then Set1, and select from your fits directory a .scancounts.fits file.

You will see a block map telling you how many Scans you got at each position.. Use the mouse/drag to zoom into the white region and NOTE THE x,y VALUES. Then use Set1 to select the .fits file for the same source, and you will get a spectrum at last in the Profile Window. Use the mouse/drag to zoom into the spectrum as you wish.

To actually get a 1D .fits spectrum, use the following miriad script with syntax miriad.script fitsname objectname xpix ypix

xpix and ypix are one plus the pixel number from the .scancounts.fits files (.scancounts.fits starts at pix zero)

miriad.script outputs a 1d fits file, called filenames.fits

miriad.script:

fits in=$1 op=xyin out=tmp1
imsub in=tmp1 out=tmp2 region=box'('$3,$4,$3,$4')'
reorder in=tmp2 out=tmp3 mode=312
puthd in=tmp3/object value=$2 type=ascii
fits in=tmp3 out=tmp4 op=xyout
echo $1 | sed 's/.fits/s.fits/' | sed 's/.*p291_//' >tmp5
cat tmp5 | awk '{print "mv tmp4",$0}' >tmp6
echo 'rm -rf tmp*' >> tmp6
chmod 744 tmp6
./tmp6

Then use the very wonderful SLAP (currently in /scratch/aife_1/amsr/scathach_1/amsr/code/slap) to look at your spectra. Note that the infile must be entered at the infile prompt, on a separate line due to lower/upper case handling.

Phew.

Last modified: Wed Jun 14 17:36:07 BST 2000