AIPS HELP file for TDSTEP3 in 31DEC24
As of Mon May 6 14:37:03 2024
TDSTEP3: Third step in time-dependent imaging
INPUTS
INNAME Input UV data (name)
INCLASS Input UV data (class)
INSEQ Input UV data (seq. #)
INDISK Input UV data disk drive #
TD_TIMES List of time limits for the
"short" time intervals (days)
PRTLEV Debug level displays:
1 INPUTS for each task
2 IMHEAD for each task
OUTNAME Output image name (name)
OUTDISK Output image disk drive #
******** IMAGR parameters **************
NCHAV Number of chan. to average.
CHINC Channel incr. between maps.
CELLSIZE 1.E-12 (X,Y) size of grid in asec
IMSIZE 0.0 8192. Minimum image size
NFIELD 0. 4096. Number of fields (max 4096)
DO3DIMAG -1.0 1. > 0 => use different tangent
points for each field
FLDSIZE -1. 8192. Clean size of each field.
RASHIFT RA shift per field (asec)
DECSHIFT DEC shift per field (asec)
UVTAPER 0. (U,V) Gaussian taper
units are kilo-lambda
UVRANGE 0. Min & max baseline (klambda)
GUARD -1.0 0.9 x,y guard band fractional
radius
ROTATE Rotate image CCW from N by
ROTATE degrees
ZEROSP 0-spacing fluxes and weights
SEE HELP!!
UVWTFN UV dist. weight function
UVSIZE 0. Array size for doing uniform
weights. 0 -> actual field
size.
ROBUST Robustness power: -5 -> pure
uniform weights, 5 => natural
UVBOX 0. 128. Additional rows and columns
used in weighting.
UVBXFN Box function type when UVBOX
> 0. 0 -> 1 round pill box.
XTYPE 0. 10. Conv. function type in x
default spheroidal
YTYPE 0. 10. Conv. function type in y
default spheroidal
XPARM Conv. function parms for x
YPARM Conv. function parms for y
NITER * 0.0 Maximum # of Clean components
BOXFILE Input file of field params
and Clean boxes; ' ' => use
FLDSIZE, RASHIFT, DECSHIFT,
NBOXES, CLBOX only.
OBOXFILE * Output file for final Clean
boxes
GAIN * 0.0 2.0 Clean loop gain
FLUX * Minimum Clean component (Jy)
MINPATCH * 0.0 Min. BEAM half-width in AP.
BMAJ * -999.9 FWHM(asec) major axis Clean
* restoring beam.
BMIN * -999.9 FWHM(asec) minor axis Clean
* restoring beam.
BPA * -360.0 360.0 Clean beam position angle
OVERLAP -1.0 1 => restore components to
overlapped fields, >=2=>
expect overlaps in Cleaning
ONEBEAM -1.0 1.0 > 0 use only 1 dirty beam
per scale in multi-facet
Cleans
OVRSWTCH -0.9 0.9 Not 0 => switch from OVERLAP
>= 2 to OVERLAP 1 - see HELP
FACTOR * -5.0 5.0 Speedup factor see HELP
CMETHOD * Modeling method:
* 'DFT','GRID',' '
IMAGRPRM Task enrichment parameters
(1) Antenna diameter (m)
(2) Source Spectral index
(3) Frequency scaling factor
(4) > 0 -> SDI Clean factor
(5) >0 => scale residuals
(6) Half-width in x of box
(7) Half-width in y of box
(8) Filter components whose
neighborhood is weaker than
IMAGRPRM(8) Jy. 0 -> don't
(9) Radius in pixels for the
IMAGRPRM(8) test.
(10) multiplier of image size
to get beam size: 0 => 2;
2, 1, 0.5 0.25 supported
(11-16) Multi-scale controls
(17) spectral index radius
0 -> no correction
(18) Limit grids (see help)
(19) Dynamic range limit
(20) Retry factor (see help)
IMAGRPRM ? Task enrichment parameters
? (1) Antenna diameter (m)
? (4) > 0 -> SDI Clean factor
? (5) >0 => scale residuals
? (6) Half-width in x of box
? (7) Half-width in y of box
? (8) Filter components whose
? neighborhood is weaker than
? IMAGRPRM(8) Jy. Can TELL
? only if non-zero on GO.
? 0 -> no filtering.
? (9) Radius in pixels for the
? IMAGRPRM(8) test.
? (11-15) Multi-scale controls
? (18) Limit grids (see help)
? (19) Dynamic range limit
? (20) Retry factor (see help)
IM2PARM Yet more parameters:
(1) Auto boxes: allowed #
(2) : island level
(3) : peak required
(4) : limit wrt max
(5) : extend boxes
(6) : edge skip
(7) reset boxes for next chan
(8) TV timeout interval: init
(9) timeout after 1st: in sec
(11) baseline-dependent avg
max time in sec
(12) field size 0 -> infinite
(13) Number channels averaged
IM2PARM ? Yet more parameters:
? (1) Auto boxes: allowed #
? (2) : island level
? (3) : peak required
? (4) : limit wrt max
? (5) : extend boxes
? (6) : edge skip
? (9) timeout after 1st: in sec
NGAUSS 0.0 10.0 Number of scales to use
WGAUSS 0.0 Scales in arc sec >= 0
FGAUSS 0.0 Minimum flux for each resol.
MAXPIXEL * 0.0 500000.0 Maximum pixels searched in
* each major cycle.
IN3NAME Spectral index image name
IN3CLASS Spectral index image class
IN3SEQ Spectral index image sequence
number
IN3DISK Spectral index image disk
IN4NAME Spectral curvature name
IN4CLASS Spectral curvature class
IN4SEQ Spectral curvature sequence
number
IN4DISK Spectral curvature disk
FQTOL Frequency tolerance in kHz
(primary beam & spec index)
DOTV * -1.0 4096.0 Display residuals on TV ?
Start with field = DOTV
BADDISK -1.0 1000.0 Disks to avoid for scratch.
HELP SECTION
TDSTEP3
Type: Procedure
Use: See HELP TDEPEND for a full discussion. TDSTEP3 taks as input
a fully calibrated and edited single source file. It then
loops over time intervals doing:
a. SPLIT to make 'ISPLIT' containing all data for time range(i)
b. IMAGR of NFIELDS+1 where NFIELDS is the number of facets
required to image the full area and facet NFIELDS+1 is
centered on the time-variable target source. BOXFILE must
include all NFIELDS+1 facets and should do an UNClean box on
the target source wherever it occurs in facets numbered <=
NFIELDS and should do UNClean boxes surrounding the source
in facet NFIELDS+1. The latter is really only required if
you do auto-boxing. Note that you will be doing deeper
imaging at a later stage.
c. UVSUB subtracts facet NFIELDS+1 from ISPLIT and appends the
data in a file with class 'APPEND'
d. It then deletes ISPLIT and all images and beams.
The output of this step is a UV data set from which the
time-variable target source has been removed (as best one can
at this early point). It has name OUTNAME, class 'APPEND',
sequence 1, and disk OUTDISK.
Adverbs:
INNAME.....Input UV data file (name). Standard defaults.
INCLASS....Input UV data file (class). Standard defaults.
INSEQ......Input UV data file (seq. #). 0 => highest.
INDISK.....Input UV data file disk drive #. 0 => any.
SOURCES....Source name - specify at most 1.
TD_TIMES...A list of break times for imaging intervals in which the
target source is not likely to change. Time interval I
is TD_TIMES(I) to TD_TIMES(I+1) in days.
PRTLEV.....For debug purposes: 1 => INPUTS for each task are
printed. 2 => IMHEADER for each input file is also
printed. 3 => also INPUTS for each zap, 4 also IMHEADER
for each zap.
OUTNAME....Output image name (name). Standard defaults.
Used for data set called class APPEND only.
OUTDISK....The disk drive # of almost everything. 0 => highest with
space but you shoukd not depend on this.
************ Adverbs strictly for IMAGR ************************
NCHAV......NCHAV is the number of channels to be averaged together
in in the gridding process. 0 => 1. If this value is less
than the total number of channels, then a multi-channel
image will result. Note that the last output channel may
include data from fewer than NCHAV input channels unless
ECHAN, BCHAN, CHINC, and NCHAV are very carefully
chosen.
The term "averaged" applies to the output image; each
channel is kept separate in the uv data used in the
imaging so that it may be gridded and model subtracted
at the correct frequency. IM2PARM(13) overrides this and
actually averages a number of channels together on the
fly before gridding. Be aware that the values of NCHAV,
CHINC, and IM2PARM(13) interact and should be consistent.
CHINC......Number of input channels to skip between images. 0 => 1
The i'th output channel includes input channels
BCHAN + (i-1)*CHINC
through
MIN (ECHAN, BCHAN + (i-1)*CHINC + NCHAV - 1).
See also IM2PARM(13) for considerations.
CELLSIZE...(X,Y) pixel separation in asec.
IMSIZE.....(X,Y) The minimum desired size of the fields regardless of
the FLDSIZE for component search.
NFIELD.....The number of fields to map in the antenna beam. Up to
4096 are allowed. Note that only 64 fields may be
described in adverbs, but 4096 are allowed. If you want
to set Clean boxes in advance for more than the first
field, or wish to specify RASHIFT, DECSHIFT, FLDSIZE, or
BCOMP for fields > 64, you must use the BOXFILE option.
If the multi-scale option is used, the actual number of
fields imaged will be NFIELD*(Number of scales) and that
product is limited to 4096.
TDSTEP3 SPECIAL USAGE: facet NFIELD+1 containing the
target source must also be specified and will be imaged.
Mark the target with an UNClean box in the facet(s) which
would normally contain it. In facet NFIELD+1, Clean only
the target source - if auto-boxing you will have to put
UNClean boxes around the target to keep those areas from
being Cleaned in the special facet and, of course, put a
Clean box on the target in the special facet. See
BOXFILE for info on UNClean boxes.
DO3DIMAG...> 0 => make the images by shifting the tangent point to the
field center. This is more accurate for significant shifts
than simply moving the w term to the field center which is
what happens with DO3DIM <= 0. It turns out that this
option costs only about 1 percent of the cpu time when it is not
needed and may make Cleans go very much faster when it is
needed. Beginning April 2009, DO3D false will do a more
accurate implementation called by some "faceting in the
uv plane". That mode should be reasonably accurate and
the resultings should FLATN more easily.
FLDSIZE....(X,Y) field size in pixels for the component search during
Cleaning; one per field. Should be in the range 32X32 to
8192X8192. Output image size will be increased to the next
highest power of two (or IMSIZE if that is greater), but
only the region specified will be searched for
components. Default is IMSIZE-10. Set FLDSIZE(1,i) and
FLDSIZE(2,i) = -1, if you want there to be NO clean box
initially in field i. This isn't necessary if you are
using auto-boxing (IM2PARM(1) > 0). TV options may be
used to delete, change and create Clean boxes
interactively. The BOXFILE option and the NVSS WWW
server may help in entering these values;see below.
RASHIFT....RA shift of the phase center of each field from the tangent
point of the uv data in asec. Map center = tangent point +
shift. If X>0 shifts map center to east. NOTE: RASHIFT is
a shift in RA scaled by cos (Dec_0) as
Ra_new(i) = RA_0 + RASHIFT(i) / cos (Dec_0)
where _0 => the tangent point in the uv data. This is a
change for 15OCT99 from shifts in -SIN projection (which
do not work for -NCP data and large angles). If the UV
data have been rotated then RASHIFT and DECSHIFT refer to X
and Y in the new coordinate system.
The BOXFILE option and the NVSS WWW server may help in
entering these values;see below.
DECSHIFT...Declination shift of map center from tangent point of each
field in asec. Map center = tangent point + shift. If Y>0
shifts map center to north.
The BOXFILE option and the NVSS WWW server may help in
entering these values;see below.
UVTAPER....(U,V) Gaussian taper (kilo-lambda) at 30 percent level
UVRANGE....(Minimum,Maximum) baseline (kilo-lambda) in map.
GUARD......Fraction of the x and y radius for which uv samples are not
allowed. < 0 => just enough to avoid mathematical errors
in the convolution.
0 => 0.3 * SQRT(taper weight at 0.3 from edge).
ROTATE.....Rotation angle to be applied in degrees.
ZEROSP.....(1)= zero spacing Stokes I flux density. Zero spacing flux
is placed at the center of FIELD 1.
(2)= zero spacing Stokes Q flux density
(3)= zero spacing Stokes U flux density
(4)= zero spacing Stokes V flux density
(5)= weight for zero spacing flux.
Both ZEROSP(1) and ZEROSP(5) must be > 0 to apply this
option. The zero-spacing data sample is appended to the
end of the input data set and participates in any uniform
weighting, gridding, etc. in the same way any other sample
does. ******* NOTE THAT THIS IS NOT THE SAME AS OTHER
IMAGING TASKS USAGE OF ZEROSP. ****************
UVWTFN.....Weighting function of (u-v) plane in 2 character code.
If the 1st character is N use "natural" weighting (the
weights attached to the data with no variation due to
local density). Otherwise, use "uniform" weighting
in which the weights are scaled by the local density of
weights under control of adverbs UVSIZE, UVBOX, UVBXFN,
and ROBUST.
The second character (and also the first) controls any
alteration of the weights to be done before they are
used in the natural or uniform weighting:
2nd character = S => take square root of weight_in
2nd character = V => take fourth root of weight_in
2nd character = O => use 1.0
UVWTFN = 'CS' => take 1 / square root of weight_in
UVWTFN = 'CV' => take 1 / fourth root of weight_in
UVWTFN = 'C?' => take 1 / weight_in where ? is any
character except S, V, O
UVSIZE.....Size of the array used to count samples for uniform
weighting. Does not have to be a power of two and can be
smaller than or bigger than the image size. The default is
the size of the first output image.
ROBUST.....Briggs' "robustness" parameter. "Uniform" weights are
tempered by a constant being added to the local density of
weights. ROBUST = -4 is nearly pure uniform weighting,
ROBUST = +4 is nearly pure natural weighting. Use of this
option requires a second array in the "AP" memory and may
therefore force the data to be sorted. The option is
turned off if ROBUST < -7 and uniform weighting is turned
off is ROBUST > 7. See HELP ROBUST - the AIPS ROBUST
differs numerically from that of Briggs.
UVBOX......(U,V) box size for weighting. This is the support radius
over which a sample is counted. I.e., the sample or its
weight is counted over an area 2*UVBOX+1 cells on each side
in the UV plane, where the UV cell size is (after
correcting units) given by 1 / (UVSIZE(i) * CELLSIZE(i)).
UVBXFN.....If UVBOX > 0, UVBXFN controls how the samples are counted
as a function of u and v (UVBXFN < 0) or of radius (UVBXFN
> 0). In the latter case, the function is 0 for radius >
UVBOX. Functions are pill box, linear, exponential, and
Gaussian for ABS(UVBXFN) = 1-4, resp. 0 -> 1. See HELP
UVBXFN.
XTYPE......Convolution function type in X-direction
1=Pill-box, 2=exponential, 3=Sinc, 4=Exp*Sinc,
5=Spheroidal, 6=exp*BESSJ1(x)/x. <= 0 or > 5 -> 5.
YTYPE......Convolution function type in Y-direction
XPARM......Array containing parameters for XTYPE.
See HELP UVnTYPE when n=convolution type.
YPARM......Array containing parameters for YTYPE.
NITER......Clean iteration limit. 0 => no Cleaning.
BOXFILE....Input text file used to simplify the specification of large
numbers of fields and/or large numbers of Clean boxes.
Leading and trailing blanks from all lines in the text
file are discarded, so "column 1" below means the first
non-blank column in the card.
This option is used to specify field parameters for
fields 1 through NFIELD which are then copied to the
fields used for additional scales (if any). To specify a
field's parameters, put the letter F or f in column 1
followed by the field number, the X and Y FLDSIZE values,
the RASHIFT amd the DECSHIFT for the field (separated by
blanks). Any field specified in this way overrides the
corresponding parameters given in the adverbs. Thus,
F 2 450 450 -25.5 6.7
specifies that field 2 is to have a FLDSIZE of 450x450 with
an RASHIFT of -25.6 and a DECSHIFT of 6.7 arcsec. If this
is the only F card in the file, then fields 1 and 3 through
NFIELD are set by the adverb values. As an alternative, a
field may also be specified with a "coordinates" card
having a C or c in column one. After the C, give the field
number, the X and Y FLDSIZE values and the center Right
Ascension (HH MM SS.S) and Declination (signDD MM SS.S)
separated by blanks. Thus
C 2 450 450 11 34 45.67 -00 14 23.1
specifies that field 2 is to have a FLDSIZE of 450x450 with
a center RA of 173.6902917 degrees and a center Declination
of -0.23975 degrees. All 9 numbers must be given; the sign
is optional for positive declinations and is given only on
the degrees term.
To set a BCOMP include put the letter B or b in column 1
followed by the field number and the value of BCOMP to
be used. Thus, to include no components from field 98 and
some from 99 include the lines:
B 98 0
B 99 243
Fields 1 through NFIELD*(Number of scales) may be
specified.
To set Clean boxes, specify one box per line, as field
blc-x blc-y trc-x trc-y (5 integers) e.g.
1 200 205 220 222
1 230 232 240 241
2 100 100 130 121
...
or circular "boxes" as
field -1 radius center-x center-y (5 ints) e.g.
001 -1 10 210 214
001 -1 5 235 237
....
Column 1 must contain a numeric character (part of the
field number); otherwise the line is treated as some
other sort of line. Fields with no boxes specified --
and auto-boxing turned off (IM2PARM(1) = 0) --- default
to the size specified by IMSIZE and FLDSIZE (see above
and including FLDSIZEs read from this file). This option
overrides NBOXES and CLBOX if any boxes for field one
appear in the file. Otherwise, those adverbs are used
for field 1. E.g. BOXFILE 'FITS:BOXES'
Fields 1 through NFIELD*(Number of scales) may be
specified. If you do not give boxes for a field >
NFIELD, then the boxes for the corresponding field at
full resolution (0 scale) are copied to those at the
lower resolutions (larger scales).
If BOXFILE = ' ', NBOXES and CLBOX apply unchanged as do
the FLDSIZE, RASHIFT, and DECSHIFT adverbs.
The NVSS WWW server may help in preparing these values;see
below.
The number of Clean boxes per field is limited to
min [ 4096, (64*4096)/(NFIELD*NGAUSS) ]
To specify that a field has no Clean boxes, specify the
BLC and TRC as four zeros.
To mark regions which should never be Cleaned, include
lines beginning with U or u. Following that character
with at least one blank, enter the facet number and four
values as for Clean boxes (i.e either rectangular or
circular). The UNClean boxes may be changed with the TV
when DOTV is true. Thus, for example
U 2 100 100 130 121
u 001 -1 5 235 237
will protect one area in each of facets 1 and 2 from
Cleaning. This option is primarily used to isolate a
source to one special facet and to keep it from being
Cleaned in another facet. Modeling routines may then use
the special facet - or all facets except the special one
- to isolate the source. Task CCEDT is also used for
this purpose when UNClean boxes have not been used.
When combining more than one spectral channel or IF, you
may wish to alter their relative weights in a temporary
fashion. The BOXFILE option allows this with W cards:
W in column 1, then a weight, then a channel number (0 ->
all), and last an optional IF number (absent -> 0, 0 ->
all). For example:
W 0.1 1
W 0.5 2
W 0.8 2 3
W 0.1 63
Assigns weight 1 to all channels 3-62, weight 0.1 to all
channels 1 and 63, weight 0.5 to channel 2 except for 0.8
in channel 2, IF 3. These "weights" multiply the weights
already assigned to the data.
When imaging with multiple facets, especially with
multiple scales, you may wish to have some of the facets
ignored in the Cleaning. To do this, give "I" cards with
a facet number or a range of facet numbers as
I 18
I 23 36
to image facets 18 and 23 through 26 but never consider
them for Cleaning. Note, when giving a range, the first
number must be lower than the second number.
The BOXFILE option is essential when NFIELD > 64.
OBOXFILE...Output text file to record the Clean boxes used. If
BOXFILE is also used and OBOXFILE points at a new file,
then IMAGR starts by copying all of BOXFILE to OBOXFILE.
Then, each time a TV REBOX or TVBOX is selected the file
is rewritten (as the TV interaction ends) with all of the
Clean boxes currently in force for all fields. The lines
in the file containing other kinds of information are
retained throughout. Thus, one can set OBOXFILE=BOXFILE
or for safety make a new OBOXFILE but then use that as
input the next time.
GAIN.......The Clean loop gain. 0 => 0.10.
FLUX.......Stop Clean when abs(resid. image max) < FLUX in Jy.
If FLUX < 0 then Clean stops after the first negative
Clean component (the actual value of FLUX is then
irrelevant). Note that on restarts and in OVERLAP < 2
mode, when the residual levels are known for all fields,
the Clean is stopped if all fields are below 1.05 *
FLUX. When each facet is Cleaned, the Clean proceeds
until the next component is less than 1.0 * FLUX however.
This "slop" is to prevent expensive major cycles being
initiated to deal with weak bumps that appear after the
model has been subtracted and the fields re-imaged.
MINPATCH...Minimum half width of the portion of the beam which is used
in the AP minor Clean. (init 51) Use 51 for deep Cleans of
extended sources. Use a large value if the beam has big
side-lobes.
BMAJ.......The FWHM (asec) major axis of the restoring beam. If 0;
value obtained from fitting to the beam. If <0; output
will contain the residual image. This is for the
point-source resolution. It will be corrected to the
other scales (if any).
BMIN.......The FWHM (asec) minor axis of the restoring beam.
BPA........The position angle in the unrotated image of BMAJ.
OVERLAP....If <= 0, components found in one field are restored only
to that field even if they occur at a coordinate found in
one or more other fields. If >0, components from each
field are restored to all fields that they overlap. For
OVERLAP<2, all fields are Cleaned in each major cycle.
For OVERLAP >= 2, one field is Cleaned and its components
subtracted from the uv data before the next field is
imaged and Cleaned. This prevents the slow convergence
which happens when Clean boxes in two fields actually
cover the same emission. It also reduces the affects of
sidelobe bumps of strong sources appearing to be sources
in the weaker fields. If OVERLAP>0, the output CC files
will have been merged. If OVERLAP>=2, you can control
which field is Cleaned next, but only if the task runs
interactively (set DOWAIT=TRUE before GO IMAGR). OVERLAP
>= 2 must be used if you use the multi-scale option
below. OVERLAP>=2 is believed to be a superior Clean
method, but it may be slower than OVERLAP=1. See a
suggestion below. Set OVERLAP = N (where N > 2) to force
all images to be recomputed with filtering every N
Cleans. You may also force recomputing from the TV menu.
ONEBEAM....> 0 => do only one beam pattern per scale for either
value of DO3DIMAG. Note that the facet beams are
different, but it has been argued that they are not
enough different to matter with uv-plane based Cleans.
OVRSWTCH...Parameter to allow switching between OVERLAP >= 2 mode
and OVERLAP = 1 mode when the peak residual is less than
abs(OVRSWTCH) * Initial_Peak where Initial_Peak is the
peak found at the beginning of the Clean. If OVRSWTCH
> 0, ONEBEAM is not changed in the switch. If OVRSWTCH
< 0, ONEBEAM is made true in the switch. This implements
the suggestion below but avoids the tricky parts of a
restart. Note that switching is not allowed for
multi-scale imaging (NGAUSS > 1). In OVERLAP 1 mode the
Clean boxes are examined for overlap between facets
frequently and the smaller of the overlapped boxes is
removed from the list.
------------------------
SUGGESTION: Use OVERLAP=2 and ONEBEAM=FALSE at the start
of major multi-facet Cleans and run them until the high
dynamic range signals have been Cleaned. Then restart
with those Clean components using OVERLAP=1 and
ONEBEAM=TRUE for the weaker components in a more
efficient Clean. NOTE that you must be careful to Clean
each source region in only one facet if you use OVERLAP=1
mode. Multi-scale Clean is not available in OVERLAP < 2.
------------------------
FACTOR.....FACTOR>0 causes deeper Clean in each major cycle, speeding
Clean, maybe "eating" extended structure. OVERLAP=2
mode may need speeding with FACTOR and/or larger
MAXPIXEL.
FACTOR=0 => the normal Clark Clean. FACTOR=-0.3 is good for
deep Cleans of extended structure.
CMETHOD....This determines the method used to compute the model
visibility values.
'DFT' uses the direct Fourier transform, this method is the
most accurate.
'GRID' does a gridded-FFT interpolation model computation.
' ' allows the program to use the fastest method,
except that DFT will be used on images <= 128 for
accuracy reasons.
IMAGRPRM...Correction control parameters (SEE EXPLAIN IMAGR):
(1) If > 0 then make frequency dependent primary beam
corrections assuming an antenna diameter of IMAGRPRM(1)
meters. Can change with TELL which only makes sense if
you are going to repeat the subtraction with a filtered
set of components (see IMAGRPRM(8)). Note that VLA and
ATCA arrays (TELESCOPE header parameter) use the
default primary beam parameters defined elsewhere in
AIPS, while other antennas actually use IMAGRPRM(1) as
the diameter of a "standard" telescope. See FQTOL
below also.
(2) Visibility amplitudes will be corrected to the average
frequency assuming a spectral index of IMAGRPRM(2).
Note: the typical optically thin synchrotron spectral
index is about -0.7.
(3) If > 0, then the u,v and w terms are scaled by
IMAGRPRM(3) before imaging.
(4) If > 0, then SDI Clean will be used when the fraction
of residual pixels in the Clean boxes stronger than
half the maximum residual exceeds IMAGRPRM(4). <= 0 ->
never use or allow SDI Clean. Can change with TELL.
If SDI Clean is enabled, the output CC files will have
been merged.
(5) If > 0 then scaling of residuals is requested and
(6) Half-width in x of box to determine the dirty beam area
(default = 5)
(7) Half-width in y of box to determine the dirty beam area
(default = 5) Can change (5)-(7) with TELL.
(8) If non-zero, select only those Clean components having
> ABS(IMAGRPRM(8)) Jy within a radius of IMAGRPRM(9)
cells of the component. If IMAGRPRM(8) < 0, the abs
value of the flux near the component is used. This is
an optional filter to remove weak isolated components
which can cause a significant bias. Can change with
TELL but only if it was non-zero to begin with. A copy
of the input data has to be made for this option and it
is only made if IMAGRPRM(8) is non-zero. If this
option is selected, the output CC files will have been
merged. Note that IMAGRPRM(8) should always be <= 0
for images of Q, U, and V Stokes parameters since
negative brightnesses are valid. Filtering is done on
restarts, when requested from the TV, on certain
Cleaning failures, on normal completion (after which
the task may resume Cleaning depending on IMAGRPRM(9)
until the completion points such as NITER and FLUX are
reached a second time) and on final exit. If ALLOKAY
>= 2, the filter is not applied on the restart. If the
filtering option was selected at the start (IMAGRPRM(8)
non zero), it may be turned off by setting IMAGRPRM(8)
exactly 0 and running TELL. To delete all negative
regions, set IMAGRPRM(8) to a tiny positive number.
(9) The abs(IMAGRPRM(9)) is the radius in cells for the
area in which fluxes are computed. If IMAGRPRM(9) < 0,
the Clean will restart following the "final" filtering
on the assumption that enough changes are made by the
filter that more Cleaning will be needed.
abs (IMAGRPRM(9)) < 1.1 => 3.1. Can change with TELL.
(10) = multiplier of max image size to set beam size.
Values of 2, 1, 0.5, and 0.25 are allowed. 0 => 2.
Smaller beam images are a bit faster, but less accurate
in the early Clean cycles. The largest beam image used
is 2048 on a side except when IMAGRPRM(1) > 0.75. When
IMAGRPRM(10) is 1, the limit is 4096 and when
IMAGRPRM(10) > 1.5, the beam is twice the image size
limited by 32768.
Multi-scale experimental controls are based on
BeamRatio = (field beam area) / (min beam area)
(11) Multi-scale experimental control: select which
field to Clean using peak fluxes (in Jy/beam) weighted
by 1 / (BeamRatio)**IMAGRPRM(11). This is important.
(12) Multi-scale experimental control: decrement the
value of IMAGRPRM(11) used above by IMAGRPRM(12) each
time an non-point resolution field is Cleaned until it
is 0.
(13) Multi-scale experimental control: use gain =
GAIN * [ {1/BeamRatio} ** IMAGRPRM(13) ]
(14,15) Multi-scale experimental control: use
factor = FACTOR * (1 - IMAGRPRM(14) * [ 1 -
1.0/({BeamRatio} ** IMAGRPRM(15)) ]
(16) Multi-scale experimental control: use maxpixel
= MAXPIXEL + IMAGRPRM(16) * (BeamRatio)
Multi-scale experimental control limits:
0 <= IMAGRPRM(11) <= 1.0 not changed by TELL
0 <= IMAGRPRM(12) <= 0.1 changed by TELL
0 <= IMAGRPRM(13) <= 1.0 changed by TELL
0 <= IMAGRPRM(14) <= 1.0 changed by TELL
0 <= IMAGRPRM(15) <= 1.0 changed by TELL
0 <= IMAGRPRM(16) changed by TELL
(17) 1 => use a spectral-index image represented in
IN3NAME, IN3CLASS, IN3SEQ, IN3DISK below to correct the
Clean component model for each channel. IN4NAME et al
will also be used as a curvature image iff IN3NAME are
specified.
IMAGRPRM(17)-0.5 is used as a radius in pixels over
which the spectral index image is averaged. When it is
small (0 < IMAGRPRM(17) <~ 1), the spectral index is
interpolated rather than averaged. See FQTOL below as
well. When doing spectral index, the primary beam
correction (IMAGRPRM(1)) costs very little extra.
(18) In OVERLAP>=2 mode, when imaging multiple fields,
IMAGR grids and FFTs multiple fields in an attempt to
determine the next one to Clean. Multiple fields are
done to reduce I/O in this search which may otherwise
have to re-read the work file several times to find the
next field to Clean. The limit on the number of fields
done depends on the maximum size of the AP, the size of
the images, etc - trying to guess when I/O will be
expensive in time. Sometimes, IMAGR will make more
images than are needed at a subsequent excess cost. To
limit the number of fields imaged at any one try, set
IMAGRPRM(18) to the maximum number you want to allow.
The task will now reduce the maximum number when the
multiple fields all have similar maxima - i.e. after
the wide dynamic range early cleans are done.
(19) In OVERLAP>= 2 mode, when Cleaning a field with a
small bright source, it is possible to Clean too
deeply. Then weak lumps due to sidelobes of strong
sources in other fields are treated as sources in the
present field. By the time the present field is
Cleaned again, these errors become very apparent. This
parameter is used to limit the weakest source Cleaned
in this major cycle to IMAGRPRM(19) times the strongest
source in this cycle. The default is the sum of the
maximum sidelobe outside a radius of 5 pixels and the
maximum sidelobe outside a radius of MINPATCH pixels.
IMAGRPRM(19) is also used to limit the depth of a SDI
Clean major cycle. SDI never goes deeper than 0.33 of
the peak, but even that may be too much.
(20) In OVERLAP >= 2 mode, the objective function of the
selected field after it is re-imaged is compared to the
objective function of the second best field (without
re-imaging). If the second best now appears better
than the selected field by a factor greater than
IMAGRPRM(20), then the task will try another field.
0 = 1.005. (Values < 1 are converted to 1/IMAGRPRM(20)
and, finally, values > 5 => 1.005.)
IM2PARM....Even more IMAGR parameters:
Auto-Clean boxing (can be changed by TELL):
(1) IMAGR can create Clean boxes automatically. In
OVERLAP 2 mode it will do this only in the facet
about to be Cleaned. In OVERLAP < 2, it looks at
every facet at each major cycle. It will find no
more than the strongest IM2PARM(1) boxes each time
it looks. <= 0 => don't do. Limit 50.
(2) The auto-boxing starts by finding islands of
emission > IM2PARM(2) * rms in the residual image.
This defines the size of the box if it is
accepted. (0 -> 3.0)
(3) A box can only be accepted if its peak brightness
is > IM2PARM(3) * rms in the residual. A box is
also accepted only if the peak in it is not
already in a Clean box.
< IM2PARM(2) -> IM2PARM(2) + 2.0
(4) A box is also only accepted if its peak brightness
is > IM2PARM(4) * maximum residual in the whole
image. < 0.01 OR > 0.9 -> 0.1
(5) The box determined by the island may be extended
outward in all directions by IM2PARM(5) pixels.
< -1 => 1. Note that -1 means compressed by 1 in
radius or in all directions for rectangles.
(6) The residual image is examined only in an ellipse
(circle if IMSIZE(1) = IMSIZE(2)) of radius in X
of IMSIZE(1)/2 - IM2PARM(6) and in Y of
IMSIZE(2)/2 - IM2PARM(6). <= 0 -> 5
(7) When imaging an output cube, should channel N+1
begin with the boxes of channel N or only with
those set up by BOXFILE, CLBOX, etc.?
> 0 - begin with initial boxes in each channel
< 0 - pass boxes along to next channel
= 0 => +1 when auto-boxing, -1 when not doing
auto-boxing
TV timeout controls:
(8) The first TV display resumes Cleaning after
IM2PARM(8) seconds. 0 -> 600
(9) After the first, the TV display resumes Cleaning
after IM2PARM(9) seconds. 0 -> 30.
Baseline-depndent and frequency averaging:
(11) The maximum elapsed time over which averaging of
data may be done in seconds. 0 -> infinite
(12) The desired field of view radius in arc minutes
which is not to be distorted by time averaging in
a baseline-dependent fashion. <= 0 -> infinite
or no averaging. The field of view is the region
in which averaging is not to reduce the amplitude
by more than 1 percent on any baseline. No data
separated by more than 268.5 wavelengths divided
by IM2PARM(12) are averaged together. It might
be wise to make this parameter larger than the
field of view about which you care.
(13) Average IM2PARM(13) channels together in the
on-the-fly averaging when IM2PARM(12) > 0. Note
that you can get channel averaging without time
averaging by setting IM2PARM(11) to a small
enough (but > 0) value while setting IM2pARM(12)
to an appropriate value. If IM2PARM(13) is 1,
IMAGR will compute the number of channels based
on the maximum possible baseline and the value of
IM2PARM(12). This parameter must be <= NCHAVG
and both NCHAVG and CHINC should be integer
multiples of IM2PARM(13).
Future expansion:
(10)
(14) - (40)
NGAUSS.....Number of scales to use. 0 -> WGAUSS=0 and NGAUSS =
1. The total number of scales is NGAUSS and the
total number of fields imaged will be NFIELD*NGAUSS.
Clean boxes specified for fields 1-NFIELD will be copied
to the corresponding fields NFIELD+1 to NFIELD*NGAUSS
unless (using BOXFILE) you have specified them already.
RASHIFT, DECSHIFT, FLDSIZE are specified only for fields
1-NFIELD. See multi-scale Clean discussion in the
Explain section.
WGAUSS.....The FWHM of the circular Gaussian source models to be
used. If a point source is to be used (which is
recommended), then WGAUSS(1) should be 0 and WGAUSs(2)
through WGAUSS(NGAUSS) > 0. IMAGR will go through the
widths to insure that, if a WGAUSS of 0 is used, it is
the first one. All resolutions now have components from
all resolutions restored to them scaled appropriately
with the widths of the fatter of the two resolutions.
FGAUSS.....The minimum flux in Jy/beam for each scale in the
same order as WGAUSS. Must be >= FLUX to have much
effect.
MAXPIXEL...The maximum number of pixels that are searched for
components inside the ``AP'' in each major cycle. <= 0
=> 20000. This number affects the cpu usage significantly.
Too many causes the task to search over many points it will
never use. Too few causes the task to do many more small
major cycles, also at great expense. Use this with great
caution, but big wins are possible using larger sizes on
very large Cleans. OVERLAP=2 mode may need speeding with
FACTOR > 0 and/or larger MAXPIXEL.
IN3NAME....Image name of spectral index image; no default.
IN3CLASS...Image class of spectral index image; no default.
IN3SEQ.....Image sequence of spectral index image; 0 -> highest.
IN3DISK....Disk of spectral image image; 0 -> any.
IN4NAME....Image name of spectral index curvature image; no
default. Curvature images should be base 10 rather than
base e - they differ by a factor of 2.3. Also the
reference frequency for them is 1 GHz. These are changes
done 2010-07-13.
IN4CLASS...Image class of spectral index curvature image; no
default.
IN4SEQ.....Image sequence of spectral index curvature image;
0 -> highest.
IN4DISK....Disk of spectral curvature image image; 0 -> any.
FQTOL......Frequency tolerance in kHz. Spectral channels with FQTOL
are handled together (use the same average CC model) when
applying the primary beam and spectral index
corrections. Default is to do each channel separately
which can take a long time.
DOTV.......Display residuals on TV channel 1. > 0.5 => display field
number DOTV initially. Can be changed interactively and by
TELL. When using this option, you may interact with the
residual images, selecting which field is examined in what
window, resetting the Clean boxes, and stop the Cleaning of
the current channel. IMAGR uses DOTV in the form of the
nearest integer; set it only to integer values.
BADDISK....This array contains the numbers of disks on which it is
desired that scratch files not be located. BADDISK has no
effect on input and output maps.
EXPLAIN SECTION
See EXPLAIN IMAGR for more information on the
imaging parameters.