Appendix A. Hints for Reducing High Frequency VLA Data in AIPS
Applying the standard centimeter wavelength recipe in AIPS to reduce
high frequency data (22 or 43 GHz) taken with the VLA may work in the
smaller arrays (C and D) and occasionally even in the larger (A and B)
array configurations, but is most often just a source of frustration.
Nevertheless, VLA data taken at these high frequencies in the largest
array configurations can be calibrated in almost all cases with only a
few minor adjustments to the centimeter wavelength recipe.
The reasons for more complicated calibration lies in part in the
achieved high resolution, resolving the standard flux density
calibrators, in particular 3C48. However, most of the problems are
caused by the atmosphere, where the troposphere introduces rapid phase
fluctuations between the antenna elements of the interferometer. Both
effects scale with baseline length expressed in units of wavelength,
but the latter also heavily depends on the current weather; phases are
sometimes observed to wind on time scales of less than a minute. This
causes decorrelation during your calibrator and target source scans,
and requires you to determine phase-only calibration, before the flux
density (i.e. gain) calibration should be attempted.
Here such an approach for reducing high frequency VLA in AIPS is
described to help to overcome the most common problems. It is assumed
that the reader has some experience with reducing data in AIPS, and is
familiar with the 'standard recipe', tools to examine the data,
self-cal, and if appropriate, spectral line and polarization
calibration issues. If not, it is recommended to read the AIPS
cookbook first (in particular chapter 4).
High frequency calibration already starts when loading the data,
requiring specific parameters in FILLM to be set (actually, one
ideally would want to use these FILLM inputs for all frequencies..).
Run FILLM with:
- doweight 1 To apply Tsys weights for each individual IF and polarization
- douvcomp -1 Do not compress (compression discards individual IF weights)
- cparm(8) 0.05 A short time for your CL-table entries (in min); 0.05min = 3s
- bparm 20, 1 To apply opacity and gain curve corrections (bparm 0 default)
This creates a CL-table that can be interpolated over very short
intervals, hopefully short enough to cover the atmospheric phase
fluctuations accurately. The default is 5 minutes, good for centimeter
wavelengths, but way too long for proper interpolation of your high
frequency phases. Also, now you have 'nominal sensitivity' weights for
individual IF/Pol entries, which reflect sensitivity differences
between the receivers, IFs, etc. To retain this 'nominal sensitivity'
weighting you are required to set docalib=2 (and a non-negative value
for gainuse) in all the calibration tasks during the remainder of the
data calibration!
After loading your data, check your CL-table entries, e.g. LISTR with
optype 'gain', PRTAB with dohms 1, or SNPLT on a short (few minutes)
time range with optype 'amp'; make sure the entries are at the
interval you expect (much less than a minute) and that the opacity and
gain curve corrections have been applied (gains a few percent
deviating from a value of one). Inspect your continuum or 'channel 0'
data (gains, system temperatures), and flag bad data. For example, you
also may wish to flag antenna 1, which is known to have bad optics at
43 GHz, and the antennas without a 43 GHz receiver (currently in March
2002, antennas 7, 9 and 15, but for earlier observations you may want
to check the receiver status page, or your observation log - these
antennas may have been left present in your data when you first do a
pointing scan in X-band). Standard tools for data inspection and
flagging are described in chapter 4 of the AIPS cookbook (LISTR,
UVPRT, VPLOT, SNPLT, UVFLG, TVFLG, EDITR, and many more). Make sure
that at least your calibrators are 'clean'. Run VPLOT on your
calibrators with a reference antenna close to the center of the array
(determined by using PRTAN) to get an indication how rapidly your
phases fluctuate; use antenna [reference antenna] 0, solint 0 and
bparm 0 2 0 (phase only).
Note that fast-switching, when used, will have changed the source
names you used in making the observe schedule file. Your sources will
have been renamed to their J2000 positions, making it less easy to
recognize the calibrator and target scans when you run LISTR (optype
'scan').
Run SETJY on your absolute flux density calibrator: 3C286 = J1331+305
= B1328+307, or 3C48 = J0137+331 = B0134+329. And maybe it is a good
idea to make a copy your correct CL-table number one (actually all
tables) with TASAV before continuing, so in case of accidents, you
have CL-table one with the opacity/gain corrections applied (INDXR
does not have this option when re-creating the CL-table yet, but it
may have it in the future).
Run CALIB, at this stage to correct for phase only, with a small
solution interval (depending on your signal to noise, e.g. 20 seconds)
on all your calibrator sources. You might want to obtain a model for
3C286 or 3C48 from
www.aoc.nrao.edu/~smyers/calibration and run CALIB
on these sources with the model separately. For 3C286, a point source
approximation probably also will work - welcome to the realm of
'trial-and-error'.
Inputs to CALIB:
- calsour 'cal1', 'cal2', etc To define your calibrators; ALL but 3C48
- docalib 2 To apply nominal sensitivities, ESSENTIAL!
- gainuse 0 Apply latest CL-table (is one/first here)
- refant [reference antenna] Pick a well behaved one in the array center
- solint 20/60 For 20 seconds - may have to try some values
- solmode 'p' Do phase calibration only at this stage!
- snver 1 Collect solutions in SN-table one
And if you have 3C48 as absolute flux density calibrator, or when you
wish to use a model for 3C286, you should re-run CALIB with the
previous/above values used and:
- calsour '3C48'' or '3C286'' Or any other name you have for the source
- in2disk [disk with model]
- get2name [model catalog #] To specify the model to be used
- invers 0 Use the models latest CC-version (i.e. one)
- ncomp 0 Use all the CC-components of the model
This may work, but there is no guarantee. Some tricks to apply, in no
particular order, in your data set or CALIB to obtain a larger
relative portion of good versus bad solutions would be:
- flag some more bad data points on your calibrator sources
- discard antennas with uncertain baseline positions (see observing log file)
- apply baseline corrections (info.aoc.nrao.edu/vla/html/baselines.shtml)
- do, or do not, use a model for 3C286 versus a point source approximation
- choose a different reference antenna (the one you have might be misbehaving)
- decrease the uvrange To weigh nearby antennas more in the solution
- use soltype 'L1' To be less sensitive to outlying points
- increase or decrease solint Increase for weak, decrease for strong sources
- decrease aparm(7) (default 5) To include more noisy but valid (?) solutions
- decrease aparm(1) (default 6) To require less antennas for a good solution
- recreate 'CH 0' from 'LINE' To get up to 25% more bandwidth on calibrators
Note that at 43 GHz in A-array the unprojected uv-distance between the
outer two antennas is already 0.5 Mega-wavelengths, and the outer 6
antennas - the default for aparm(1) - require good solutions out to 2
Mega-wavelengths for CALIB to accept the solution for your outermost
antenna. Hence, it is a good idea to have aparm(1) set to e.g. four
(or three if you're cautious!).
Check the resulting SN-table number one with LISTR (optype 'gain',
dparm 1 0) or SNPLT (inext 'sn', optype 'phas'), and judge whether you
have enough solutions and whether you believe the phases shown are
likely to reflect the variation caused by the troposphere. If not,
fiddle around with your data and/or parameters in CALIB as suggested
above and try again. In case the majority of solutions are fine, you
may want to edit spurious points in your SN-table with e.g. SNEDT,
EDITA, SNCOR, or CLCOR.
Once you are satisfied with the phases in your SN-table, you would
want to apply phase corrections to minimize decorrelation in your
calibrator scans before you determine the absolute flux density
scale. To insert the corrections, run CLCAL with:
- sources '' Correct phases for all sources
- calsour 'cal1', 'cal2', etc Include ALL your calibrators
- interpol '2pt' ('simp' will average phases over a scan)
- snver 1; gainver 1;gainuse 2 To apply SN#1 to CL#1, creating CL#2
- refant [reference antenna] Select the same antenna as used in CALIB
In less straight forward observations your data may not be suited to
run CLCAL only once, e.g. when you are switching frequencies - if in
doubt, consult chapter 4 of the AIPS cookbook. It is however very
simple to run CLCAL multiple times. Inspect your new CL-table two for
unexpected dubious inter- and extrapolations (LISTR with optype
'gain', dparm 1 0, or SNPLT with inext 'cl', optype 'phas') and
backtrack possible problems.
If you have flagged more data, delete your SN table two (as it does
not overwrite bad solutions for flagged data). Now re-run CALIB with
the corrected phases to obtain the flux density scale: (note that you
preferably want to use models for 3C286 or 3C48)
- calsour 'cal1', 'cal2', etc To identify your calibrators; not 3C48, 3C286
- docalib 2 To apply nominal sensitivities, ESSENTIAL!
- gainuse 0 Apply latest CL-table (is two/second here)
- refant [reference antenna] Pick the same antenna as used before
- solint 0 Scan lengths; phase variations are applied
- solmode 'a&p' Do full calibration to get the flux densities
- snver 2 Collect solutions in a new SN-table (two)
And run CALIB again for the absolute flux density calibrator 3C286, or
3C48, using a model and the previous/above values used:
- calsour '3C48'' or '3C286'' Or any other name you have for the source
- uvrange 0 To use full uv range without restrictions
- in2disk [disk with model]
- get2name [model catalog #] To specify the model to be used
- invers 0 Use the models latest CC-version (i.e. one)
- ncomp 0 Use all the CC-components of the model
The same tricks may be used as for phase-only to get a more good over
bad ratio in the solutions. Check your SN-table two thoroughly; the
phases 'must' be zero or very close to zero (therefore interpol '2pt'
is preferred over interpol 'simp' in CLCAL), and you want to make sure
the gains of your reference antenna do not scatter too much for
individual sources (as the flux density scale is not yet fixed, the
average gain per source will vary - after GETJY in the next step the
gains per antenna over all sources should be similar). If you can
identify misbehaving antennas, flag them and re-run CALIB as many
times as needed to re-create SN-table two. Cautious users will start
from the beginning.
Run GETJY to obtain the secondary calibrator flux densities:
- sources 'cal1', 'cal2', etc Do not enter the source used in SETJY above
- calsour '3C48'' or '3C286'' Or any other name you have used in SETJY
- snver 2 Point to the flux density/gain solution table
If you used a model for the flux density calibrator, your flux scale
will actually be tied to the flux density of the calibrator , and NOT
to the value entered in the SU-table by SETJY. However, the SU-table
entry should be non-zero for GETJY to work, hence the SETJY step at
the beginning.
Carefully note the flux densities reported by GETJY and do not blindly
trust these values. LISTR or SNPLT may point out problematic antenna
solutions which may indicate you to flag some more data and start
over. Hopefully you have some prior flux densities for your secondary
calibrators (from
aips2.nrao.edu/vla/calflux.html), or even better,
you observed one or more of the sources regularly monitored by NRAO
staff (see
www.aoc.nrao.edu/~smyers/calibration). You
want to check the values, because it sometimes happens that the flux
densities deviate considerably from the expected values and make no
sense. This could be the case if the pointing solutions that were
determined prior to your primary calibrator scan are inappropriate for
this particular primary calibrator scan or, for example, if your
secondary calibrators are in a different part of the sky, and the
cloud cover on your single primary calibrator scan differs from the
cloud cover in the opposite part of the sky (or even worse, a
combination of these). In some cases you may be forced to approximate
the flux density scale by entering a (recent) flux density for one of
your secondary calibrators, ignoring the primary calibrator scan and
accepting an introduced flux density uncertainty. If you decide you
have to restart, do not forget to reset the flux densities of all your
calibrators with SETJY (and optype 'rejy'), before entering a zerosp
for a new flux density calibrator source (also with SETJY). Re-iterate
until you are happy with the flux density scale.
The final flux density calibration table is obtained by running CLCAL
again:
- sources '' Calibrate flux densities for all sources
- calsour 'cal1', 'cal2', etc Include calibrators to use for your targets
- interpol '2pt', or 'simp' (no real difference for per-scan solutions)
- snver 2; gainver 2;gainuse 3 To apply SN#2 to CL#2, creating CL#3
- refant [reference antenna] Select the same antenna as used in CALIB
From here you are almost set to continue as usual according to the
'standard recipe', i.e. polarization and bandpass calibration if
appropriate, and splitting into single source data sets. However,
remember to set docalib to two in all these tasks as long as you are
working on the multi-source data set and haven't applied initial
phase, flux density (including polarization, bandpass) and 'nominal
sensitivity' calibration with SPLIT. After SPLIT, the individual
weights will have been entered in the data, properly scaled by the
latest CL-table you've made. Using your single source calibrated data,
use docalib false in your subsequent imaging and analysis tasks.
If you anticipated checking your fast-switching calibration by
including a 'check source' (an at least moderately strong source
observed a few times (about once an hour) with the same fast-switching
parameters at about the same distance from your fast-switching source
as your target source, but not necessarily in the same direction), you
can now asses a snapshot of your calibration by imaging this
source. If the fast-switching has worked perfectly, your check source
has the expected morphology, expected flux density, and the expected
position. The position error should indicate how good the astrometry
on your target source is. If you did not include a check source, all
but the astrometry and spatial dependence of the calibration can be
inferred from your fast-switching source by imaging an uncalibrated
scan (use a modified SN/CL table by skipping the calibration on this
scan) with the calibration derived from the two neighboring scans.
This section was contributed by:
Lorant Sjouwerman - VLBA Scientific Services - lsjouwerman@aoc.nrao.edu
Go to: 
Contents
Previous