LITTLE THINGS AIPS CALIBRATION Recipe
Designed to deal with a mixed array VLA-EVLA
=====================================================================
Suggestion: STARTING with D ARRAY and then working through C array and B array is recommended.
D array data will allow you to have a look at the whole field (the strong continuum sources
that might give problems will be obvious here), is more stable, and will give you the best feel
for the data. The steps in this recipe apply to all configurations.
0. Basic stuff
B array configuration
From observing log (available online at
http://www.vla.nrao.edu/cgi-bin/oplogs.cgi):
* note antennas out for EVLA tests
* note antennas that have moved recently
* note antennas with any problems during the observing session
1. FILLM
1a. Request archive data:
1b. FILLM ---> *.CH 0.1, *.LINE.1
default FILLM
datain 'ARCHIVE:DDO43/Archive/AH927_Q080122.xp
band 'l';vlaobs 'AH927';
doall -1 ; qual 1; $ restricts FILLM to the galaxy+calibrators of interest;
nfiles 0; ncount 1; $ read one file;
outna 'ddo43-b1';
outdisk 1; outseq 0;
douvcomp= -1; $ allow channel/IF-dependent weights;
doweight 10; $ use memo 108 weights (i.e., put weights in 1/Jy^2);
bparm= -1,-1; $ avoid opacity & gain corrections;
cparm 0;
cparm(4)=25.1; $ One needs to explicitly flag shadowed antennas
$ since the ModComps were retired on 27 June 2007.`
$ This requires FILLM 31DEC08 after 18Nov08 MNJ
cparm(7)= 0; $ Assigns new FREQID if frequency changes by more than
$ the max. Doppler shift between sources 180 degrees
$ apart. Setting this to -1 forces all data to have same
$ FREQID.
doconcat=-1; $ Change this to DOCONCAT=1 to add data to an existing file.
***Note: FILLM's channel 0 will ONLY be used for initial flagging.
2. TASAV ---> LINSAV.1
***We TASAV right away, because VLANT changes the AN table.
default TASAV
outna 'DDO43-1BeTa;
outcla 'LINSAV';
outdi 2; $ Ideally set this to a different disk from indisk.
getn *.LINE
3. UVCOP ---> LINCOP.1
***Discard the first and last channels:
1st and last 10 channels if 127 channels total
1st and last 20 channels if 255 channels total
***We discard these channels because (1) they're pretty much useless;
(2) they seem to confuse BPASS (which takes the solution from
channel N as the initial guess for channel N+1); (3) their noise
characteristics are quite different from the rest of the channels, which
can be confusing e.g. in clipping and imaging.
default UVCOP
outcla 'LINCOP';
bchan 11; echan 127-10; $ Use bchan 21; echan 255-20; if 255 channels total
uvcopprm 0
uvcopprm(4) 1 $ report progress
getn *.LINE
***From now on we operate on LINCOP data unless otherwise specified.
4. LISTR/SCAN ---> *.listr
***We do this before VLANT because we need to know FREQIDs for VLANT.
default LISTR
optype 'SCAN';
docrt -1;
outpr 'LT1:DD043-B1.listr;
getn *.LINCOP
---> *.listr
5. VLANT ---> AN/1, CL/2
***Note that VLANT can be run only for data observed from 1992 onwards.
Data earlier than 1991 produce the following error message:
VLANT1: Task VLANT (release of 31DEC07) begins
VLANT1: ANT DATA UNAVAILABLE FOR YEAR 1991 DATA START WITH 1992
VLANT1: Purports to die of UNNATURAL causes
For earlier data we skip baseline corrections entirely, and hope for the
best. If there are clear and systematic phase gradients with time,
consult the baseline corrections at
http://www.vla.nrao.edu/astro/archive/baselines/
and apply corrections via CLCOR (which is basically what VLANT does).
***AN table: For now, this recipe resets to the original AN table just
before running VLANT the last time. This means that only one run
of VLANT updates the AN entries, leading to a correct AN table as
needed for (e.g.) UVFIX.
***We maintain the convention that CL/1 is the original CL table and
CL/2 has all the corrections that should have been applied on-line but weren't
-- mainly, antenna position corrections.
5a. default VLANT
getn *.LINCOP
5b. If VLANT does not create a new CL table (no antennas moved), or
the data were taken before 1992:
default TACOP
getn *.LINCOP
getona *.LINCOP
inext 'CL' ; inver 1 ; ncount 1 ; outver 2
***The goal here is to keep the recipe uniform for all data.
6. PRTAN AN/1
default PRTAN
docrt 132
getn *.LINCOP
Location Of VLA Antennas
N36 ( 7)
N32 (26)*
N28 (27)
N24 (25)*
N20 ( 2)
N16 (13)*
N12 (18)*
N8 ( 9)
N4 (12)
*(19) W4 E4 (15)
( 6) W8 E8 (10)
*( 1) W10 ( )
*(24) W12 E12 (20)
( 8) W16 E16 (14)*
( ) E20 (21)*
*(16) W24 E24 (28)
*(17) W28 E28 ( 3)
( ) E32 (11)*
(22) W36 E36 (23)*
VLA:_OUT ( 4)
VLA:_OUT ( 5)
VPT:_OUT (29)
* => EVLA ANTENNA
***To choose the reference antenna the following algorithm should be followed:
1. should be present throughout the run
2. should be on an "inner" pad, but NOT N1/E1/W1 (to avoid shadowing)
3. NOT on the master pad (since those are always weird)
4. NOT an EVLA antenna during this transition
5. try to avoid the north arm in the smaller configurations (to avoid shadowing)
6. NOT listed in any interesting way in the log file (to avoid problems with the
reference antenna)
7. preferably consistent with other recent runs
8. should be a fairly stable antenna (but can't tell until TVFLG/CALIB)
===> Refant: 15 (E4)
7. Calibrators
7a. ***Check out the calibrators in the on-line calibrator manual:
http://www.vla.nrao.edu/astro/calib/manual/index.shtml
***Max baseline at 21 cm in B array is 54.3 klambda
C array is 16.2 klambda
D array is 4.9 klambda
primary (flux/bandpass) calibrators:
0137+331= 3C48
0542+498= 3C147
***You can ignore the uv-ranges for these, since there are now
models for the most important ones.
secondary (gain) calibrator: 0702+445
0702+445 J2000 T 07h02m53.6790s 44d31'11.940"
0659+445 B1950 T 06h59m16.4860s 44d35'35.850"
-----------------------------------------------------
BAND A B C D FLUX(Jy) UVMIN(kL) UVMAX(kL)
=====================================================
20cm L X P P P 2.40 30 visplot
===> secondary cal --- UVMax restriction of UVMax=30 klambda
7b. SETJY ---> SU/1
***Set aparm(2) to corespond to date of observation.
If date <1990 aparm(2)=3
If 1992<date<1998 aparm(2)=2
If date >1998 aparm(2)=0
***We need to enter a flux density for each primary (flux) calibrator.
If we have only one FREQID, this is easy:
default SETJY
sources '0137+331','0542+498','' $ primary (flux) calibrator(s)
optype 'CALC'; freqid=1; $ First FREQID
aparm 0,0; $ data taken after 1998
getn *.LINCOP;
===> / Flux calculated using known spectrum
> SETJY1: BIF = 1 EIF = 1 /Range of IFs
> SETJY1: '0137+331 ' IF = 1 FLUX =15.8922 (Jy calcd)
> SETJY1: '0542+498 ' IF = 1 FLUX =21.9624 (Jy calcd)
> SETJY1: / Using (1999.2) VLA or Reynolds (1934-638) coefficients
***If we have more than one FREQID, we assume that the frequency offsets used are small,
so that the flux densities of the flux calibrators are nearly the same for each FREQID.
Typical frequency offsets are of order the bandwidth, +/-3 MHz; at 1420 MHz,
for a worst-case spectral index of -1, this leads to an error of
2*3 MHz/1420 MHz= 0.4% -- not worth worrying about. So, we use any FREQID which covers
all primary flux calibrators. If flux calibrator A is observed only with FREQID 1,
while flux calibrator B is observed only with FREQID 2, we have to run SETJY twice:
default SETJY
sources '0137+331','' $ primary (flux) calibrator A
optype 'CALC'; freqid=1; $ FREQID for A
aparm 0,0; $ data taken after 1998
getn *.LINCOP
default SETJY
sources '0542+498','' $ primary (flux) calibrator B
optype 'CALC'; freqid=2; $ FREQID for B
aparm 0,0; $ data taken after 1998
getn *.LINCOP;
7c. CALRD
***Read in models of flux density calibrators:
***Note: These models are in J2000 coordinates. If your data are in B1950,
change the model images to B1950 with EPOSWTCH. We will later use
UVFIX to fix the uv-data.
default CALRD
object '3c48';band 'L';
default CALRD
object '3c147';band 'L';
8. PRTUV
***We use this to find the integration times on calibrators & sources.
default PRTUV
cparm 0; cparm(9)=103; $ Pick a baseline -- here, baseline 1-3
docrt 132;
getn *.LINCOP
---> calib: 10s
source: 10s
9. UVFLG ---> FG/1
***We toss the EVLA-EVLA baselines, to avoid dealing with aliasing. There should still
be plenty of VLA-EVLA baselines to allow antenna-based solutions for the EVLA, but
keep an eye out for oddities (e.g., in BPASS)!
***NOTE: Do NOT run UVFLG if EVLA=0 (try PRINT EVLA to check) -- otherwise you'll delete
ALL of your data.
default UVFLG
outfgver 1;opcode 'flag';reason 'EVLA';
antenna=EVLA;baseline=EVLA;
getn *.CH0 $ Note that we use FILLM's CH0 for initial flags -- we'll
$ TACOP later.
10. TVFLG ---> FG/1
10a.
***We *only* use the original CH 0 from FILLM for initial flagging. Here we flag
calibrators only, to remove any gross, obvious problems.
- Check the first scan carefully -- often the system isn't "organized" on
this first scan.
- look for hot pixels and hiccups.
- We are NOT quacking, because (1) FILLM's NX table isn't correct;
(2) QUACK flags data from the beginning-of-scan, whereas we want
to flag data from antennas-on-source.
default TVFLG
calcode '*'; $ calibrators only
docat -1; $ avoid saving temporary files
dohist -1; $ avoid creation of history entries
Freqid 1; $ must step through all FREQIDs!
docalib -1;
flagver 1;outfgver 1; $ keep all flags in FG/1
dparm 0;
dparm(3) 1 $ show baselines twice, to treat all antennas identically --
$ this displays, for example, baseline 27-1 as well as 1-27
dparm(6)=10 $ time resolution: should be set to the calibrators' integration
$ time, in seconds
getn *.CH0 $ note this is the ONLY time we use FILLM's Channel 0!
***Within TVFLG:
- Set useful defaults:
SMOOTH=1 to avoid averaging date before displays
SCAN= 20 to use a long time for median filters (AMP/PH DIFF)
FLAG ALL CHANNELS
FLAG STOKES FULL (usually -- sometimes you'll want NORR or NOLL)
***If individual Stokes need to be flagged, make sure you set the
STOKES FLAG to correspond to the polarization that is displayed on
the TV.
SWITCH SOURCE FLAG to ONE-SOURCE to avoid inadvertantly flagging
your galaxy (though sometimes you'll want to of course)
- Be sure to inspect BOTH polarizations!
***We suggest the following steps:
- Set the above defaults.
- Flag first integration in every scan (manual QUACK) -- this should be
the first integration AFTER most antennas are on-source, which is why
we can't use the usual QUACK.
- Inspect the following:
AMPLITUDE to check for missing records or antennas
AMP DIFF to check for variable gains
PHS DIFF to check for variable atmosphere/gains
- If your data set is in D array (and, if your source is southern, then also check for this
problem in the C array), keep an open eye for solar interference. It will be obvious
if in TVFLG you choose a SORT BY BASELINE display, showing you how the short baselines
behave, the ones affected by solar interference. If solar interference is affecting your data
then in CALIB you should use a UVRANGE.
***Note: occasionally, flagging using UVFLG can be more straightforward
(e.g., deleting an antenna).
10b. TABED FG/1
***Here we TABED the CH 0 flags to LINCOP (with FREQID= -1). After this we're done with
FILLM's channel 0.
default TABED
opty 'repl';
inext 'fg';
inver 1 ; outver 1;
bcount 1;ecount 0 ;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.CH 0
getona *.LINCOP
---> LINCOP FG/1
11. BPASS
***This is a first-order BPASS leading to a new Channel 0.
The goal is to avoid closure errors in Channel 0 calibration due
to huge delays (phase slopes) on VLA-EVLA baselines.
We divide each visibility by the vector average of the inner
3/4 of the band (i.e., an on-the-fly channel 0). Thus we remove
source structure (though getting the weights wrong) and take care
of the amplitude scale.
***There is a split here between the easy case (one FREQID for all sources)
and the Galactic HI case (multiple FREQIDs, usually different for the
bandpass calibrator and the galaxy (and phase calibrator)). Check
LISTR/SCAN to see which you're doing.
11a. BPASS: one FREQID for all sources ---> BP/1
default BPASS
calsour '0137+331','0542+498','' $ Select bandpass calibrators
docal 1 ; gainuse 2; $ apply VLANT changes. Probably irrelevant.
flagver 1; $ apply initial flags
refant 15; $ Change this to your refant
Qual -1;
solint 0; $ one solution per scan
minamper 7 ; minphser 7; $ report closures > 7%/7d
smooth 0; $ no smoothing
soltype '' ; weightit 0; $ L1, L1R, etc. seem less stable
bpassprm 0;
bpassprm(5) 0; $ derive "channel 0" on a record-by-record basis --
$ more biased than averaging first, but avoids
$ some subtle pitfalls (see EXPLAIN file)
bpassprm(2) 1; $ some closure info is printed
bpassprm(6) 2; $ print avg. closure errors > 2%
bpassprm(7) 2; $ print avg. closure errors > 2d
ichansel 0; $ derive channel 0 from inner 3/4 of the band
freqid 1; $ here we have only one FREQID
getn *.LINCOP
---> BP/1
11b. BPASS: Multiple FREQIDs ---> BP/1,2,3
***The overall plan here is as follows (assuming FREQIDs 1 and 2 refer to
the offset [bandpass calibrator] frequencies, and FREQID 3 refers to that
of the galaxy & phase calibrator):
(1) run BPASS once for FREQID=1 (-> BP/1) and once for FREQID=2
(-> BP/2).
(2) check both BP tables with POSSM. They should look virtually
identical.
(3) If they do appear virtually identical, we concatenate them:
(a) write out both tables [TBOUT],
(b) concatenate the two [vi/emacs],
(c) read them back in [TBIN] as BP/3,
(4) If they do NOT appear identical, there is something wrong.
The case we've come across involves the use of a front-end
filter combined with the use of unexpected LOs, so that one
FREQID was observed through the edge of the filter. The
resulting bandpasses show a strong slope in the amplitude gains
for most VLA antennas. So far we've seen this only for central
frequencies around 1423 MHz. In this case, simply copy the "good"
BP table to BP/3 using TACOP.
(5) modify BP/3 to refer to FREQID=3 [TABED].
***It is a VERY good idea to use POSSM carefully throughout to be sure
you're doing what you think you're doing.
11b1. BPASS FREQID 1 ---> BP/1
default BPASS
calsour '0137+331','0542+498','' $ Select bandpass calibrators
docal 1 ; gainuse 2; $ apply VLANT changes. Probably irrelevant.
flagver 1; $ apply initial flags
refant 15; $ Change this to your refant
Qual -1;
solint 0; $ one solution per scan
minamper 7 ; minphser 7; $ report closures > 7%/7d
smooth 0; $ no smoothing
soltype '' ; weightit 0; $ L1, L1R, etc. seem _less_ stable -- weird
bpassprm 0;
bpassprm(5) 0; $ derive "channel 0" on a record-by-record basis --
$ more biased than averaging first, but avoids
$ some subtle pitfalls (see EXPLAIN file)
bpassprm(2) 1; $ some closure info is printed
bpassprm(6) 2; $ print avg. closure errors > 2%
bpassprm(7) 2; $ print avg. closure errors > 2d
ichansel 0; $ derive channel 0 from inner 3/4 of the band
freqid 1; $ 1st offset FREQID
getn *.LINCOP
---> BP/1
11b2. BPASS FREQID 2 ---> BP/2
***Same as 11b1, now on FREQID 2
default BPASS
calsour '0137+331','0542+498','' $ Select bandpass calibrators
docal 1 ; gainuse 2; $ apply VLANT changes. Probably irrelevant.
flagver 1; $ apply initial flags
refant 15; $ Change this to your refant
Qual -1;
solint 0; $ one solution per scan
minamper 7; minphser 7; $ report closures > 7%/7d
smooth 0; $ no smoothing
soltype '' ; weightit 0; $ L1, L1R, etc. seem _less_ stable -- weird
bpassprm 0;
bpassprm(5) 0; $ derive "channel 0" on a record-by-record basis --
$ more biased than averaging first, but avoids
$ some subtle pitfalls (see EXPLAIN file)
bpassprm(2) 1; $ some closure info is printed
bpassprm(6) 2; $ print avg. closure errors > 2%
bpassprm(7) 2; $ print avg. closure errors > 2d
ichansel 0; $ derive channel 0 from inner 3/4 of the band
freqid 2; $ 2nd offset FREQID
getn *.LINCOP
---> BP/2
11b3. POSSM to compare BP/1 and BP/2
default POSSM $ to check BPASS results
flagver 1;
aparm 0, 1, 0.7, 1.3, -180, 180, 0, 2, 0, 0; $ Plot BP, with amp/ph ranges
source '0137+331','0542+498','1331+305','' $ POSSM doesn't work with
$ source '' for some reason!
solint -1; $ Separate plots for each scan
nplots 9; $ 9 plots per page
bparm 0;
dotv 1;
freqid 1; bpver 1 ; grch 1;
freqid 2; bpver 2 ; grch 2;
getn *.LINCOP
tvinit
****If worried: try POSSM on secondary calibrator, applying this BPASS
(if BPASS stable, secondary should look flat).
11b4. *If* BP/1 and BP/2 appear virtually identical, concantenate them
to form BP/3:
11b4a. Write the two tables to disk:
default TBOUT
docrt 500;
getn *.LINCOP
inext 'bp' ; inver 1;
outfile 'LT1:bp1.out;
inext 'bp' ; inver 2;
outfile 'LT1:bp2.out;
11b4b. Outside AIPS: concatenate the two tables
cd $LT1
cat bp1.out bp2.out > bp3.out
emacs bp3.out
- change NAXIS2 to be equal to the *sum* of NAXIS2 in the two tables
- delete from first ***END*PASS*** through ***BEGIN*PASS***, inclusive
11b4c. Read the new concatenated table in as BP/3:
default TBIN
infile 'LT1:bp3.out;
getona *.LINCOP
---> reads in BP/3
11b5. *If* BP/1 and BP/2 are NOT virtually identical, pick the one that
best matches the behavior of the phase calibrator (possibly by doing a
quickie BPASS on the phase calibrator and comparing the results), and
copy that BP table to BP/3:
default TACOP
inext 'bp';
inver 2; $ Set this to the "good" BP table
getn *.LINCOP
getona *.LINCOP
---> BP/3
***You should IGNORE the "bad" FREQID for all subsequent processing.
11b6. TABED BP/3 to set FREQID=-1 (so we can use the same BP table for
everyone) --> BP/4
default TABED
opty 'repl';
inext 'bp';
inver 3 ; outver 4;
aparm 0;
aparm(1) 8; $ Changing column 8 = FREQID
keyval= -1,0; $ ...which we change to FREQID= -1
getn *.LINCOP
---> BP/4
11c. POSSM to check BP table
11c1. Plot BP table itself
default POSSM $ to check BPASS results
flagver 1;
aparm 0, 1, 0.7, 1.3, -180, 180, 0, 2, 0, 0; $ Plot BP, with amp/ph ranges
source '0137+331','0542+498','1331+305','' $ POSSM doesn't work with
$ source '' for some reason!
solint -1; $ Separate plots for each scan
nplots 9; $ 9 plots per page
bparm 0;
dotv 1;
freqid 1; bpver 1; $ for single-FREQID data sets
freqid 3; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
tvinit
11c2. Apply BP table to 2ndary calibrator & plot individual baselines
default POSSM
flagver 1;
aparm 0; $ Plot data
solint -1; $ Separate plots for each scan
nplots 9; $ 9 plots per page
aparm 0;
aparm(1) 1; $ vector average
source='0702+445','' $ Secondary (phase) calibrator
docal 1 ; gainuse 2 ; doband 1; $ average all BP entries
dotv 1;
freqid 1; bpver 1; $ for single-FREQID data sets
freqid 3; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
tvinit
11c3. Apply BP table to 2ndary calibrator & vector average all data
default POSSM
flagver 1;
aparm 0; $ Plot data
solint 0; $ average all time
nplots 0; $ average all baselines
aparm 0;
aparm(1) 1; $ vector average
source='0702+445','' $ Secondary (phase) calibrator
docal 1 ; gainuse 2 ; doband 1; $ average all BP entries
dotv 1;
freqid 1; bpver 1; $ for single-FREQID data sets
freqid 3; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
tvinit
***This plot should be flat in both amp. and phase as a function of
frequency, with no slope. If some channels are off, note which ones
those are and keep an eye out for interference or other bad data.
If there are large errors, consider running BPASS on the secondary
calibrator and using that to correct the galaxy. Note that this
will be somewhat painful since AIPS does not allow incremental BP
tables -- unlike SN or CL tables.
12. AVSPC ---> NEWCH0.1 (2,3)
***AVSPC must be run once for each FREQID (unfortunately FREQID=-1
purports to work, but creates an empty data set). This entails
some nasty bookkeeping for data sets with multiple FREQIDs. Here
we assume that we have one or three FREQIDs. If BPASS checks above show
that one FREQID is useless, you should ignore that one entirely in this
and all subsequent processing.
***We will use these NEWCH0 files for (1) further flagging, and
(2) time-dependent gain calibration.
12a. FREQID=1 --> NEWCH0.1
default AVSPC
docalib -1;gainuse 0; flagver -1; $ do NOT apply flags
doband 1;
freqid 1; bpver 1; $ for single-FREQID data sets
freqid 1; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
outname inna ; outcl 'NEWCH0';
---> NEWCH0.1
12b. FREQID=2 (if multiple-FREQID data set) ---> NEWCH0.2
***Skip FREQID=2 if the corresponding BP table looked irrelevent (see 11b5).
default AVSPC
docalib -1; gainuse 0; flagver -1; $ do NOT apply flags
doband 1
freqid 2; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
outname inna ; outcl 'NEWCH0' ; outse= freqid;
---> NEWCH0.2
12c. FREQID=3 (if multiple-FREQID data set) --> NEWCH0.3
***Skip FREQID=3 if the corresponding BP table looked irrelevant (see 11b5).
default AVSPC
docalib -1;gainuse 0; flagver -1; $ do NOT apply flags
doband 1 ;
freqid 3; bpver 4; $ for multiple-FREQID data sets
getn *.LINCOP
outname inna ; outcl 'NEWCH0' ; outse= freqid;
---> NEWCH0.3
12d. LISTR/SCAN
***It is a VERY good idea to run LISTR/SCAN on each of the NEWCH0
data sets at this point, to be sure each has the data you expect.
13. TABED LINCOP FG/1 -> NEWCH0 FG/1
***We use TABED to set FREQID to -1 (apply to all FREQIDs) in the FG
table, since AVSPC will change all FREQIDs to 1 in the NEWCH0 data sets.
13a. NEWCH0.1
default TABED
opty 'repl';
inext 'fg';
inver 1 ; outver 1;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.LINCOP
getona *.NEWCH0.1
---> NEWCH0.1, FG/1
13b. If multiple FREQIDs: NEWCH0.2
default TABED
opty 'repl';
inext 'fg';
inver 1 ; outver 1;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.LINCOP
getona *.NEWCH0.2
---> NEWCH0.2, FG/1
13c. If multiple FREQIDs: NEWCH0.3
default TABED
opty 'repl';
inext 'fg';
inver 1 ; outver 1;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.LINCOP
getona *.NEWCH0.3
---> NEWCH0.3, FG/1
14. CALIB -> NEWCH0.1(,2,3) SN/1
***If we only have one FREQID, all CALIBs are run on the same NEWCH0.1.
If we have multiple FREQIDs, CALIBs for a given source must be run
for all NEWCH0.1,2,3 in which that source appears (probably easiest to check
with LISTR/SCAN).
***Note: if 'SN' table must be destroyed: task 'extdest'; inext 'sn'; invers 0.
***If solar interference is affecting your data than in CALIB you should use a UVRANGE.
Use UVPLT to find the range affected.
After calibration, if UVPLT still shows signs of solar interference, it means that
not enough short baselines were discarded; therefore the calibration has to be redone
and UVRANGE to be reset. Note that the Sun might rise or set, especially during a long
B-array run, in which case you might wish to split the calibration by timerange in
a set affected and a set without solar interference.
14a. Primary (flux density) calibrators --> SN/1
default CALIB
calsour '0137+331','' $ flux density calibrator #1
freqid -1;
docal 1 ; gainuse 2;
flagver 1;
refant 15; $ Change this to your refant
solint 0;
aparm 4,0,0,0,0,2; $ min 4 antennas; print closures
soltype 'L1'; solmode 'A&P'; weightit 1; $ true L1 minimization
solcon 0;
minamper 10; minphser 10; $ complain if >10%/10d off
cparm 0,0,10,10,1; $ complain if avg > 10%/10d off
snver 1;
get2n 3C147_L.MODEL.1
getn *.NEWCH0 $ ***NOTE: must run this for all NEWCH0.1,2,3 in which
$ this calibrator appears!
---> SN/1
default CALIB
calsour '0542+498','' $ flux density calibrator #2
nmap 1 ; ncomp 1e6,0 ; inver 1 ; cmethod 'DFT';
freqid -1;
docal 1 ; gainuse 2;
flagver 1;
refant 15; $ Change this to your refant
solint 0;
aparm 4,0,0,0,0,2; $ min 4 antennas; print closures
soltype 'L1'; solmode 'A&P'; weightit 1; $ true L1 minimization
solcon 0;
minamper 10; minphser 10; $ complain if >10%/10d off
cparm 0,0,10,10,1; $ complain if avg > 10%/10d off
snver 1;
get2n 3C147_L.MODEL.1
getn *.NEWCH0 $ ***NOTE: must run this for all NEWCH0.1,2,3 in which
$ this calibrator appears!
---> SN/1
14b. Secondary (phase) calibrator --> SN/1
***Check uv restrictions for secondary calibrators carefully.
For 0702+445: no restrictions, so uvra= 0,0.
default CALIB
calsour '0702+445','' $ phase calibrator
uvrange 0,0;
wtuv 0.0; $ may have to set wtuv 0.01 if solutions are
$ crazy and uvrange is not 0,0
freqid -1;
docal 1 ; gainuse 2;
flagver 1;
refant 15; $ Change this to your refant
solint 0;
aparm 4,0,0,0,0,2; $ min 4 antennas; print closures
soltype 'L1'; solmode 'A&P'; weightit 1; $ true L1 minimization
solcon 0;
minamper 10; minphser 10; $ complain if >10%/10d off
cparm 0,0,10,10,1; $ complain if avg > 10%/10d off
snver 1;
getn *.NEWCH0 $ ***NOTE: must run this for all NEWCH0.1,2,3 in which
$ this calibrator appears!
15a. TABED all SN tables to NEWCH0.1 --> SN/2,3
***This step is only required if we have more than one FREQID (NEWCH0).
We copy everything to NEWCH0.1 to match the FREQID=1 case.
***Skip this step if you only have one FREQID!
default TABED
opty 'repl';
inext 'sn';
inver 1 ; outver 0;
aparm 0;
aparm(1) 6; $ Changing column 3 = FREQID
keyval= -1,0; $ ...which we change to FREQID= -1
getn *.NEWCH0.2
getona *.NEWCH0.1
---> NEWCH0.1, SN/2
default TABED
opty 'repl';
inext 'sn';
inver 1 ; outver 0;
aparm 0;
aparm(1) 6; $ Changing column 3 = FREQID
keyval= -1,0; $ ...which we change to FREQID= -1
getn *.NEWCH0.3
getona *.NEWCH0.1
---> NEWCH0.1, SN/3
15b. GETJY SN/1-3, SU/1
***Find flux density of secondary calibrator, and set SN table amplitude
gains to reflect a common flux density scale.
default GETJY
sources '0702+445','' $ Secondary (phase) calibrators)
calsour '0137+331','0542+498','' $ Primary (flux) calibrators
freqid -1;
snver 0; $ Use all SN tables
getn *.NEWCH0.1
Task GETJY (release of 31DEC07) begins
> GETJY1: Source:Qual CALCODE IF Flux (Jy)
> GETJY1: 0702+445 : 1 T 1 2.45351 +/- 0.00696
15c. CLCAL/MERG to merge all SN tables -> SN/4
***This step is only required if we have more than one FREQID (NEWCH0).
Skip this step if you only have one FREQID!
default CLCAL
opcode 'MERG'; $ Merge SN tables, for ease of plotting etc.
refant 15; $ Change this to your refant
getn *.NEWCH0.1
---> SN/4 $ creates a merged SN table -- doesn't change CL tables at all.
Must now copy this new SN table to all other NEWCH0:
default TACOP
inext 'SN';
getn *.NEWCH0.1
getona *.NEWCH0.2
getona *.NEWCH0.3
16. SN table checks
16a. SNPLT last SN table
16a1. SNPLT phase:
default SNPLT
inext 'sn';inver 0;
pixrange 0;
opcode 'alsi';do3col 1;dotv 1;
nplots 9;
factor 2; symbol 5;
xinc 1; optype 'phas';
getn *.NEWCH0.1
===> Note any phase jumps (on the phase calibrator) for future flagging.
The EVLA antennas, even after applying VLANT, still show quite
a bit of phase drift. This is OK so long as a linear interpolation
between the phases looks like it will work.
16a2. SNPLT amplitude:
default SNPLT
inext 'sn';inver 0;
pixrange 0;
opcode 'alsi';do3col 1;dotv 1;
nplots 9;
factor 2; symbol 5;
xinc 1; optype 'amp';
getn *.NEWCH0.1
===> Note whether the amp. is roughly constant for a given
antenna/pol'n/IF.
***We have seen a couple of cases where the first phase cal scan has a
significantly different amplitude gain for the EVLA antennas.
The reason is not clear but the raw data do show this effect, so
CALIB is doing the right thing.
16b. LISTR/GAIN print SN table
default LISTR
optype 'gain'; Inext 'sn'; inver 1;
freqid -1;
dparm 5,0; $ Amp & phase
factor 0; docrt 132;
outprint '';
antennas 0; $ may have to list missing antennas explicitly, to avoid
$ column overrun. To list missing antennas, use the form
$ ANTENNAS -3,2,0 to have antennas 3 and 2 left out.
getn *.NEWCH0.1
---> check for phase jumps and other inconsistencies.
***We have seen a couple cases where the first phase cal scan has a
significantly different amplitude gain for the EVLA antennas.
The reason is not clear but the raw data do show this effect, so
CALIB is doing the right thing.
17. UVFLG ---> NEWCH0.1 FG/1
***If the SN table shows a phase jump on the phase calibrator, you should
flag the data between the two phase cal scans which show the jump
(since those cannot be calibrated).
***Note that we flag NEWCH0.1, *regardless* of whether the galaxy appears
in this data set. This is because later on (step 20/TVFLG) we flag the
NEWCH0 data in order from inseq 1 through inseq freqid_max, copying the
FG table from one file to the next. UVFLG itself doesn't care whether
the flags you enter actually do anything -- it just adds entries to the
FG table, which are then applied (or ignored if irrelevant) by other
tasks.
***This step may of course be skipped if there are no obvious phase jumps.
This example deals with a phase jump on antenna 18.
default UVFLG
antenna 18,0; $ the antenna which "jumped"
timer 0 6 33 0 0 6 59 0; $ the source scan between the offending ph.cal scans
opcode 'FLAG';
reason 'phase jump';
outfgver 1;
getn *.NEWCH0.1
---> NEWCH0.1 FG/1
18. CLCAL NEWCH0 ---> CL/3
***For multi-FREQID data this becomes rather complicated, since
we need a new CL table for every NEWCH0 file, to allow detailed checks
and second-order flagging.
***Note that there is no need to work around any phase jumps, since the
intervening data are flagged (see step 17 above [UVFLG]).
18a. CLCAL for the primary calibrators ---> CL/3
default CLCAL
sour= '0137+331','0542+498','' $ Primary (flux) calibrators
calsour= sour;
interpol 'SELF';
gainver 2 ; gainuse 3;
refant 15 $ Change this to your refant
dobtween -1; $ Don't interpolate entries for different sources
snver 1; $ if single FREQID
snver 4; $ if multiple FREQIDs
getn *.NEWCH0.1 $ do this for all NEWCH0 with primary (flux)
$ calibrator data
18b. CLCAL for the phase calibrator and galaxy ---> CL/3
default CLCAL
sour= '0702+445','DDO43','' $ Secondary (phase) calibrator + galaxy
calsour= '0702+445','' $ Secondary (phase) calibrator
interpol 'SIMP';
gainver 2 ; gainuse 3;
refant 15 $ Change this to your refant
dobtween -1; $ Don't interpolate entries for different sources
cutoff 120; $ Don't extrapolate/interpolate beyond 120 minutes
snver 1; $ if single FREQID
snver 4; $ if multiple FREQIDs
***If your data set used +/- frequency switching for the phase
calibrator (our observations did not, but some archival data may),
you should use BPARM with SAMPTYPE='BOX' to select a smoothing time
which covers both frequency settings. LISTR/SCAN on LINCOP will
help you choose this; normally something like 12 minutes should be OK.
bparm 12/60 ; samptype='BOX';
getn *.NEWCH0.1 $ do this for all NEWCH0 with secondary (phase)
$ calibrator or galaxy data
***At this point we have a new CL table for all NEWCH0 files.
19a. ANBPL
***We use ANBPL to check the data weights. Data with very high weights
(factor 5-10 or more above normal) should be flagged with UVFLG.
default ANBPL
docalib 1;gainuse 3;
flagver 1;
bparm 2,17,0; $ Plot antenna-based weight vs. time
nplots 9; dotv 1;
docrt 132; $ Print as well as plotting weights -- useful for
$ finding exact times of bad weights
$ Note you can also use outprint to send to a file.
opcode 'alsi'; $ Plot all IFs together
do3col 1; $ ...using different colors
getn *.NEWCH0.1 $ Must do this separately for every NEWCH0 file
19b. UVFLG to eliminate very high weights
***Should UVFLG on NEWCH0.1, even if bad Weights are seen in
NEWCH0.2 or NEWCH0.3 -- we'll be copying FG/1 from NEWCH0.1 to
NEWCH0.2 for subsequent second-order flagging.
20. TVFLG FG/1
20a. TVFLG on calibrators: NEWCH0.1
The calibrators should now have constant amplitude and zero
phase...apart from source structure for the primary (flux) calibrators,
and any uvrange for the secondary (phase) calibrators. If you see
huge problems on the calibrators, you may have to re-run CALIB etc.
default TVFLG
calcode '*'; $ calibrators only
docat -1; $ avoid saving temporary files
dohist -1; $ avoid creation of history entries
docalib 1 ; gainuse 3; $ apply the new CL table
flagver 1;outfgver 1; $ keep all flags in FG/1
dparm 0;
dparm(3) 1; $ show baselines twice, to treat all antennas identically --
$ so, this displays baseline 27-1 as well as 1-27.
dparm(6)=10; $ time resolution: should be set to the calibrators'
$ integration time, in seconds
getn *.NEWCH0.1
***Within TVFLG:
- Set useful defaults:
SMOOTH=1 to avoid averaging date before displays
SCAN= 20 to use a long time for median filters (AMP/PH DIFF)
FLAG ALL CHANNELS
FLAG STOKES FULL (usually -- sometimes you'll want NORR or NOLL)
***If individual Stokes need to be flagged, make sure you set the
STOKES FLAG to correspond to the polarization that is displayed on
the TV
SWITCH SOURCE FLAG to ONE-SOURCE to avoid inadvertantly flagging
your galaxy (though sometimes you'll want to of course)
- Be sure to inspect BOTH polarizations!
***We suggest the following steps:
- Set the above defaults
- Check AMPLITUDE, AMP DIFF, PHS DIFF. Be wary of known source
structure and uv-range limits!!
***Note: occasionally, flagging using UVFLG can be more straightforward
(e.g., deleting an antenna).
20b. If we have multiple NEWCH0s (FREQIDs):
(1) Copy the FG table:
default TABED
opty 'repl';
inext 'FG' ; inver 1 ; outver 2;
bcount 1;ecount 0;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.NEWCH0.1
getona *.NEWCH0.2
Re-run TVFLG with same inputs as above, except:
flagver 2 ; outfgver 2;
getn *.NEWCH0.2
(2) Assuming there are three FREQIDs, we do this yet again:
Copy the FG table:
default TABED
opty 'repl';
inext 'FG' ; inver 2 ;outver 2;
bcount 1;ecount 0;
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...to FREQID= -1
getn *.NEWCH0.2
getona *.NEWCH0.3
Re-run TVFLG with same inputs as above, except:
flagver 2 ; outfgver 2;
getn *.NEWCH0.3
21. Calibration/flagging checks: calibrators
21a. UVPLT
***Check the amp & phase vs. uv-distance for all calibrators.
Amplitude should match the results of SETJY/GETJY.
If there are obvious outliers which are not expected due to source
structure, go back and flag those (and possibly re-run CALIB etc.).
default UVPLT
calco '*' ;
docal 1 ; gainuse 3;
flagver 1; $ set this to the latest FG version -- may be >1 if
$ there are multiple NEWCH0s (FREQIDs).
dotv 1;
do3col 1;
getn *.NEWCH0.1 $ do this for each NEWCH0 file
bparm 0 $ amp. vs. uv-distance
bparm 0,2 $ phase vs. uv-distance
21b. IMAGR
default IMAGR
sources '0702+445','' $ calibrator to image
docalib 1; gainuse 3; $ apply latest calibration
flagver 1; $ apply latest flags -- set this to the
$ highest-numbered FG table
outname 'junkcal'; $ some obviously cruddy name
cellsize 1; $ for B configuration
cellsize 3.5; $ for C configuration
cellsize 10; $ for D configuration
imsize 1024; $for B array
imsize 512; $for C and D array
niter 1000;
nbox 1 ; clbox -1,5,512,513; $ calibrator should be in the center
minpa 121;
uvwtfn 'NA'; robust 0.5;
dotv 1;
getn *.NEWCH0.xx $ whichever file has the calibrator you're imaging
---> Shouldn't see obvious calibration errors or striping. CLEANed
flux density should roughly match SETJY/GETJY.
22. TVFLG on the galaxy: NEWCH0.xx
***This is our first run of flagging on the galaxy.
- Check the first scan carefully -- often the system isn't "organized"
on this first scan
- Look for hot pixels and hiccups.
- We are NOT quacking, because (1) FILLM's NX table isn't correct;
(2) QUACK flags data from the beginning-of-scan, whereas we want
to flag data from antennas-on-source.
22a. If there's only one NEWCH0 (FREQID):
default TVFLG
calcode '-CAL'; $ non-calibrators only
docat -1; $ avoid saving temporary files
dohist -1; $ avoid creation of history entries
docalib 1 ; gainuse 3; $ apply the new CL table
flagver 1;outfgver 1; $ keep all flags in FG/1
dparm 0;
dparm(3) 1; $ show baselines twice, to treat all antennas identically --
$ this displays baseline 27-1 as well as 1-27
dparm(6)=30; $ time resolution: should be set to the sources'
$ integration time, in seconds
getn *.NEWCH0.1
***Within TVFLG:
- Set useful defaults:
SMOOTH=1 to avoid averaging date before displays
SCAN= 20 to use a long time for median filters (AMP/PH DIFF)
FLAG ALL CHANNELS
FLAG STOKES FULL (usually -- sometimes you'll want NORR or NOLL)
***If individual Stokes need to be flagged, make sure you set the
STOKES FLAG to correspond to the polarization that is displayed on
the TV
SWITCH SOURCE FLAG to ONE-SOURCE to avoid inadvertantly flagging
your galaxy (though sometimes you'll want to of course)
- Be sure to inspect BOTH polarizations!
***We suggest the following steps:
- Set the above defaults
- Flag first integration in every scan (manual QUACK) -- this should be
the first integration AFTER most antennas are on-source, which is why
we can't use the usual QUACK.
- Check AMPLITUDE, AMP DIFF, PHS DIFF. Be wary of known source
structure and uv-range limits!!
***Note: occasionally, flagging using UVFLG can be more straightforward
(e.g., deleting an antenna).
22b. If there's more than one NEWCH0 (FREQID):
***We assume here that the galaxy is in NEWCH0.3. If not, you should
first TACOP FG/2 from NEWCH0.3 to the file with the galaxy in it.
This will now be FG/3.
default TVFLG
calcode '-CAL'; $ non-calibrators only
docat -1; $ avoid saving temporary files
dohist -1; $ avoid creation of history entries
docalib 1 ; gainuse 3; $ apply the new CL table
flagver 2;outfgver 2; $ or flagver 3 ; outfgver 3 if you had to TACOP
$ the FG table from another file
dparm 0;
dparm(3) 1; $ show baselines twice, to treat all antennas identically --
$ this displays baseline 27-1 as well as 1-27
dparm(6)=30; $ time resolution: should be set to the sources'
$ integration time, in seconds
getn *.NEWCH0.xx $ This should be the file with the galaxy data.
***Within TVFLG:
- Set useful defaults:
SMOOTH=1 to avoid averaging date before displays
SCAN= 20 to use a long time for median filters (AMP/PH DIFF)
FLAG ALL CHANNELS
FLAG STOKES FULL (usually -- sometimes you'll want NORR or NOLL)
***If individual Stokes need to be flagged, make sure you set the
STOKES FLAG to correspond to the polarization that is displayed on
the TV
SWITCH SOURCE FLAG to ONE-SOURCE to avoid inadvertantly flagging
your galaxy (though sometimes you'll want to of course)
- Be sure to inspect BOTH polarizations!
***We suggest the following steps:
- Set the above defaults
- Flag first integration in every scan (manual QUACK) -- this should be
the first integration AFTER most antennas are on-source, which is why
we can't use the usual QUACK.
- Check AMPLITUDE (for missing data), AMP DIFF . PHS DIFF may
occasionally be useful, but it's likely to be mostly random,
unless you have a strong continuum source near your galaxy.
***Note: occasionally, flagging using UVFLG can be more straightforward
(e.g., deleting an antenna).
23. Calibration/flagging checks: sources
23a. UVPLT
***Check the amp vs. uv-distance for the galaxy.
If there are obvious outliers which are not expected due to source
structure or RFI (i.e., not mostly on short spacings), go back and
flag those. Note any obvious short-spacing horrors, which may
be due to solar or terrestrial RFI.
default UVPLT
docal 1 ; gainuse 3;
flagver 1; $ set this to the latest FG version -- may be >1 if
$ there are multiple NEWCH0s (FREQIDs).
dotv 1;
do3col 1;
source 'DDO43',''
getn *.NEWCH0.1 $ whichever file holds the galaxy
bparm 0 $ amp. vs. uv-distance
23b. IMAGR
default IMAGR
sources 'DDO43','' $ the galaxy
docalib 1; gainuse 3; $ apply latest calibration
flagver 1; $ apply latest flags -- set this to the
$ highest-numbered FG table
outname 'junk' $ some obviously cruddy name
cellsize 1; $ for B configuration
cellsize 3.5; $ for C configuration
cellsize 10; $ for D configuration
imsize 1024; $ for B configuration
imsize 512; $ for C and D configurations
niter 1000;
nbox 0;
minpa 121;
uvwtfn 'na'; robust 0.5;
dotv 1;
getn *.NEWCH0.xx $ whichever file has the source you're imaging
---> Shouldn't see obvious calibration errors or striping. Note
that this "channel 0" includes HI emission, so you may see some
odd effects (e.g., very woofly noise in B configuration) -- don't
panic!
### If you find a strong continuum source rippling your map even in a 512x512 imsize
D array configuration channel zero image, you may have to use self-cal.
24. TASAV -> CH0SAV.1,2,3
default TASAV
outna 'DDO43-1MidTa
outcla 'ch0sav';
outdi 2; $ Ideally set this to a different disk from indisk,
$ in case of disk crashes
getn *.NEWCH0 $ loop over NEWCH0 files (= FREQIDs)
outse inseq
25. TABED SN, FG tables to LINCOP
***Use TABED to ensure FREQID=-1 for all tables.
25a. NEWCH0.xx FG/yy -> LINCOP FG/2
default TABED
opty 'repl';
inext 'fg';
aparm 0;
aparm(1) 3; $ Changing column 3 = FREQID
keyval= -1,0; $ ...which we change to FREQID= -1
inver 1; $ if single FREQID
inver 3; $ if multiple FREQIDs: set this to max. flag table number
outver 2;
getn *.NEWCH0.1 $ if single FREQID
getn *.NEWCH0.3 $ if multiple FREQIDs: set this to file you flagged
$ on most recently (usually the file with the
$ galaxy in it)
getona *.LINCOP
---> LINCOP FG/2
25b. NEWCH0.xx SN/yy -> LINCOP SN/1
default TABED
opty 'repl';
inext 'sn';
inver 1; $ if single FREQID
inver 4; $ if multiple FREQIDs
outver 0;
aparm 0;
aparm(1) 6; $ Changing column 3 = FREQID
keyval= -1,0; $ ...which we change to FREQID= -1
getn *.NEWCH0 $ if multiple FREQIDs: all should have same merged SN
$ table so you can use whichever NEWCH0 file you want
getona *.LINCOP
---> LINCOP SN/1
26. CLCAL LINCOP SN/1 ---> CL/3
***We do CLCAL directly on LINCOP rather than copying, to avoid
(even more) confusion in the multiple-FREQID case.
***Note that there is no need to work around any phase jumps, since the
intervening data are flagged (see step 17 above [UVFLG]).
26a. CLCAL for the primary calibrators ---> CL/3
default CLCAL
sour= '0137+331','0542+498','' $ Primary (flux) calibrators
calsour= sour;
interpol 'SELF';
gainver 2 ; gainuse 3;
refant 15; $ Change this to your refant
dobtween -1; $ Don't interpolate entries for different sources
snver 1 ;
freqid= 1 $ You must run CLCAL once for each FREQID with
$ the relevant calibators present
getn *.LINCOP
26b. CLCAL for the phase calibrator and galaxy ---> CL/3
default CLCAL
sour= '0702+445','DDO43','' $ Secondary (phase) calibrator + galaxy
calsour= '0702+445','' $ Secondary (phase) calibrator
interpol 'SIMP';
gainver 2 ; gainuse 3;
refant 15; $ Change this to your refant
dobtween -1; $ Don't interpolate entries for different sources
cutoff 120; $ Don't extrapolate/interpolate beyond 120 minutes
snver 1 ;
***If your data set used +/- frequency switching for the phase
calibrator (our observations did not, but some archival data may),
you should use BPARM with SAMPTYPE='BOX' to select a smoothing time
which covers both frequency settings. LISTR/SCAN on LINCOP will
help you choose this; normally something like 12 minutes should be OK.
bparm 12/60 ; samptype='BOX';
freqid= 1; $ You must run CLCAL once for each FREQID with
$ the phase calibrator or galaxy present.
getn *.LINCOP
27. Calibration/flagging checks: calibrators
27a. WIPER
***Check the amp & phase vs. uv-distance for all calibrators.
Amplitude should match the results of SETJY/GETJY, and phase should be
zero, apart from known structure (reflected in source model for gain
calibrators and uvrange for phase calibrators).
If there are obvious, unexpected outliers, go back and flag those
(and possibly re-run various tasks).
default WIPER
calcode '*'
docal 1 ; gainuse 3;
doband 1; bpver 1; $ for single-FREQID data sets
doband 3; bpver 4; $ for multiple-FREQID data sets
freqid 1; $ set this to match the calibrator
flagver 2; $ should be the latest FG table
smooth 7, 117; $ boxcar average over all channels -- use
$ smooth 7, 235 if you started with 255 channels
dotv 1;
do3col 1;
bparm 0;
getn *.LINCOP
bparm(2) 1; $ amp. vs. uv-distance
bparm(2) 2; $ phase vs. uv-distance
27b. POSSM
***Check vector average of all data for each calibrator.
Amplitude should match the results of SETJY/GETJY; phase should be
flat, and consistent with zero (corresponding to a point source
at the origin)...apart from known source structure and possibly HI
absorption.
default POSSM
calcode '*';
docal 1 ; gainuse 3;
doband 1; bpver 1; $ for single-FREQID data sets
doband 3; bpver 4; $ for multiple-FREQID data sets
freqid 1; $ set this to match the calibrator
flagver 2; $ should be the latest FG table
aparm 0; $ Plot data
solint 0; $ average all time
nplots 0; $ average all baselines
aparm 0;
aparm(1) 1; $ vector average
source='0702+445','' $ Secondary (phase) calibrator
uvrange= 0,0; $ should be set to eliminate known source structure,
$ as in CALIB
dotv 1;
tvinit;
getn *.LINCOP
27c. IMAGR
default IMAGR
docal 1 ; gainuse 3;
doband 1; bpver 1; $ for single-FREQID data sets
doband 3; bpver 4; $ for multiple-FREQID data sets
freqid 1; $ set this to match the calibrator
flagver 2; $ should be the latest FG table
outname 'test'
cellsize 1; $ for B configuration
cellsize 3.5; $ for C configuration
cellsize 10; $ for D configuration
imsize 1024; $ for B array
imsize 512; $ for C and D array
niter 200; $ reasonable for a point source
nbox 1 ; clbox -1,5,512,513; $ calibrator should be in the center
minpa 121;
uvwtfn 'na'; robust 0.5;
dotv -1; $ so you can go eat lunch
getn *.LINCOP
***You shouldn't see obvious calibration errors or striping. CLEANed
flux density should roughly match SETJY/GETJY. If you do have
bad stuff, UVLSF will likely take care of it, so don't get too worked up.
28. Calibration/flagging checks: sources
28a.1 WIPER
***Check the amp vs. uv-distance for the galaxy.
If there are obvious outliers which are not expected due to source
structure or RFI (i.e., not mostly on short spacings), go back and
flag those. Note any obvious short-spacing horrors, which may
be due to solar or terrestrial RFI.
default WIPER
docal 1 ; gainuse 3;
doband 1; bpver 1; $ for single-FREQID data sets
doband 3; bpver 4; $ for multiple-FREQID data sets
freqid 1; $ set this to match the galaxy
flagver 2; $ should be the latest FG table
smooth 0; $ do not smooth: smoothing will decrease your noise and
$ consequently show you a lower flux level than the level
$ that your data has in fact
dotv 1;
do3col 1;
sources 'DDO43',''
bparm 0;
getn *.LINCOP
bparm(2) 1; $ amp. vs. uv-distance
bparm(2) 2; $ phase vs. uv-distance
Use bparm(3)=1;bparm(6)=0;bparm(7)=20 if you want to force a range to the Y axis of the WIPER plot.
***Be careful with smoothing when doing WIPER on the source, it should NOT be done at this step.
Smoothing will decrease your noise and consequently show you a lower flux level than the level
that your data has in fact.
***It is very probable that when doing a WIPER on all the line data, some channels with junk will
ruin your WIPER display and you will not be able to really make out what the clipping level
should be. If the junk consists of random hot pixels then just run another WIPER in which you
force the y axis to a 0-20 range (bparm(3)=1;bparm(6)=0;bparm(7)=20). If the junk comes
in a structured manner than further investigations are needed to identify which baselines,
in which channels might be missbehaving. Use WIPER only to identify the problematic baselines
and channels, and use TVFLG or UVFLG to remove them form the data.
28a.2
Up to which level most of the values (leaving aside the very hot pixels) comfortably fit in?
----> 6Jy
***This is the value that you will be using in CLIP when combining your data!!!
28b. IMAGR
default IMAGR
docal 1 ; gainuse 3;
doband 1; bpver 1; $ for single-FREQID data sets
doband 3; bpver 4; $ for multiple-FREQID data sets
freqid 1; $ set this to match the calibrator
flagver 2; $ should be the latest FG table
outname 'test';
cellsize 1; $ for B configuration
cellsize 3.5; $ for C configuration
cellsize 10; $ for D configuration
imsize 1024; $ for B configuration
imsize 512; $ for C configuration
imsize 256; $ for D configuration
niter 1000; $ a light clean, just to see what we've got. niter=0
$ would be ok too, esp. if you IMLIN afterwards.
nbox 0 ; clbox 0;
minpa 121;
uvwtfn 'NA'; robust 0.5;
dotv -1; $ so you can go eat lunch
getn *.LINCOP
***You shouldn't see obvious calibration errors or striping. You
should however see of order 100 mJy of continuum sources in the field,
as well as some indication of your galaxy. The latter may be quite
confusing for B configuration, which resolves out most of the
structure. Don't fret until you've combined all the array
configurations.
***If you find a strong continuum source rippling your map even after trying a 512x512
imsize in D array configuration cube, you may need to self-cal.
28c. Noise Estimations:
***The easiest way to calculate the expected sensitivity is to say
7
S(mJy) = ----------------------------
sqrt{ N (N-1) delta_nu t}
where
7 is a constant that depends on the system temperature of the receivers(a quite conservative value)
N = number of antennas in the array
delta_nu = channel resolution in MHz
t = integration time in hours.
You need to use for N the number of Antennas which have on average been giving good data during
the run. The total time should be the actual time spent on source. No need to be super precise.
One simply wants to get a ballpark figure which is good to 10-20%.
Note that for data which are Hanning smoothed (archive data) the channel spacing is equal to the
resolution delta_nu. Without Hanning smoothing, the resolution is ~1.4 x delta_nu. Note that
after offline Hanning smoothing, if you preserve all channels, the resolution becomes 2 x delta_nu;
it reverts to a new "double the old delta_nu" if you delete every other channel.
***Finally, if you want to know the noise in a single visibility, use N=2 and t = 1/360, and
probably multiply by sqrt{2} because a visibility is a single polarisation as well.
----> The Expected noise level is: X
----> The rms Noise level in a line free channel is: xx
29. TASAV -> EndTaB.LINSAV.1
default TASAV
outna 'DDO43-EndTa
outcla 'linsav';
outdi 2; $ Ideally set this to a different disk from indisk,
$ in case of disk crashes
getn *.LINCOP
30. FITTP
31. Celebrate your victory with an appropriate beverage!
================================================================================
================================================================================
The National Radio Astronomy Observatory is a facility of the National Science Foundation
operated under cooperative agreement by Associated Universities, Inc.
