Private parameters for program NCLEAN
This document contains an overview of the parameter interface of the program NCLEAN. The program also uses a number of public interfaces; references to these are also listed. show that The remainder of the document describes the individual parameters in alphabetical order. This description centers on the Help texts, which have been designed to guide the user to the proper choice at each junction, even if his knowledge of the overall workings of the program is only superficial.
See also:
Prompt: ONE antenna pattern (grp.fld.chn.pol.1.seq)
Expected input: Character *32: 1 to 1 values
Specify ONE antenna pattern
(group.field.channel.polarisation.type(=1).sequence_number)
Prompt: Window area l,m,dl,dm
Expected input: Integer: 1 to 4 values
The CLEAN window is defined as the union of up to 32 rectangular areas (which may arbitrarily overlap). For BEAM clean this is also the window within which source responses will be subtracted.
You are being prompted for these areas one by one until you give a null reply (<CR> only). An area is specified by four numbers: l,m, dl,dm, where
l,m l and m in grid points for the area centre; (l,m)=(0,0) at the map
centre, increasing toward the upper right (i.e. with decreasing RA
and increasing DEC)
dl,dm width and height of the area in grid points
Prompt: Limit in fraction of map maximum
Expected input: Real: 1 value
Specify the level to which to clean in fraction of the initial map maximum.
Prompt: UV-radius range for clipping (m)
Expected input: Real: 1 to 2 values
Specify the (circular) UV-plane radii (in metres) between which you want to clip the data. The default is to clip everywhere.
Prompt: Amplitude range to be discarded
Expected input: Real: 1 to 2 values
Specify amplitude range (in Westerbork Units) of visibility magnitudes that you want to discard.
In the annulus defined by CLIP_AREA, values between the limits you specify will be discarded.parameter CLIP_AREA
NOTE: It would be more natural to define a range within which visibilities are considered valid. As it is, only the lower limit is actually useful, allowing you to define a rejection threshold for interference. To do so, specify your threshold for the lower and 'infinity' for the upper limit, e.g.
<threshold>,100000
Prompt: UV-clean memory size
Expected input: Integer: 1 value
Specify the work memory size in bytes for the UV Clean option, to be allowed in defining the beam patch and to be used in executing the Fourier transforms.
The default shown is normally adequate; a larger value may speed up the execution of major cycles in UV Clean.
NCLEAN will not accept a value in excess of 4000000 (4 MB).
Prompt: Maximum number of components to find
Expected input: Integer: 1 value
Specify the maximum number of components to be cleaned.
Prompt: Reporting interval for components found: | terminal, printer
Expected input: Integer: 1 to 2 values
You may give two numbers <n> here, indicating that every <n>-th component must be reported. The first number applies to your terminal window, the second to the log file. A value of 0 means 'no reporting at all.
Example:
2,0 specifies that every other component will be typed, and none logged.
Prompt: Major cycle depth
Expected input: Real: 1 value
Specify the level relative to the initial map maximum in the CLEAN window to which you want to clean in one major cycle.
Prompt: Data multiplication factor
Expected input: Real: 1 value
Specify the factor by which to multiply the input map-data.
This option is only relevant for DATA clean. The first input-map and all maps created by DATA clean will be multiplied by this factor.
Prompt: Correct antenna pattern for mapping taper: YES/NO
Expected input: Yes/No: 1 value
The gridding convolution in map-making is usually compensated for by multiplying the output map and antenna pattern with a taper function that rises toward the map edges.
If this is the case, answering YES here will instruct the program to account for this effect; this will reduce the aliasing errors in the residual map, allowing you to clean a somewhat larger part of the map. YES will in general produce a result with less aliasing, NO the reverse.
Prompt: Work memory size
Expected input: Integer: 1 value
The memory workspace in bytes needed is 12 times the size of the rectangle enclosing all selected areas in the map plane. NCLEAN normally allocates up to 300 KB for this purpose. To satisfy the present need, you must either accept the value suggested here or specify a smaller set of areas, or both.
NCLEAN will not accept a value in excess of 4000000 (4 MB).
Prompt: Grating factor
Expected input: Real: 1 value
In the minor cycles of UVCOVER, DATA and COMPON clean, the maximum error made in only subtracting part of the antenna pattern is estimated by <number of sources> times <ypical sidelobe level in antenna pattern> times GRATING_FACTOR.
You may lower this latter factor if you are not worried about the effects of far-out sidelobes on the minor-cycle cleaning process. A lower value will allow more source components to be collected in minor cycles before a major cycle must be started.
Prompt: Clean loop-gain factor
Expected input: Real: 1 value
For each component found, the antenna pattern shifted to its position is subtracted from the CLEAN window. To prevent overshoots, it is customary to subtract not the complete component but only a fraction of it. The magnitude of this fraction is defined here.
Prompt: Multiplication factor
Expected input: Real: 1 value
Specify the factor by which to multiply the residual map before restoring.
Normally one should use the default value of 1; a value of 0 serves to make a map of the CLEAN components only.
Prompt: Operation wanted
Expected input: Character *24: 1 value
CLEANing:
BEAM Hogbom clean: Use the map and the beam to clean. This
method is inaccurate away from the map centre due to the
aliasing associated with visibility gridding.
UVCOVER Standard Clark Clean: Find and provisionally subtract sources
in some map areas first (minor cycles), then properly subtract
the sources found (major cycle). The number of minor cycles
between major cycles is determined by the program; the user can
steer this through a few control parameters.
The method suffers from the same limitations as BEAM, but is
faster for cleaning extended sources.
DATA Cotton-Schwab Clean: Find sources as in UVCOVER, but perform
the major cycle on the original SCN-file visibilities, making a
new map from the residuals as input for the next major cycle.
This method rigorously avoids the aliasing limitations of BEAM
and UVCOVER, at the price of being very much slower.
UREST Use a clean component list and a map to restore the clean
components in the map.
For exploring the data before committing more serious work:
HISTO Produce only a histogram of selected areas in the map and/or
antenna pattern
COMPON Execute minor cycles as for UVCOVER, but omit the following
major cycle. The result is a source model and a map in which
these sources are provisionally subtracted in selected areas.
It is the fastest of the four CLEAN variants but produces an
inaccurate residual map.
This option may be used to quickly get a feel for the
minor-cycle control of UVCOVER and DATA Clean, or to make an
initial data model for NCALIB SELFCAL or NMODEL UPDATE.
Other options:
QUIT Terminate NCLEAN.
Prompt: Prussian hat value
Expected input: Real: 1 value
For extended sources a prussian hat (i.e. a nominal additional peak value on the central value of the antenna pattern) may give better clean results. Values of .1 to .4 could be tried.Cornwell
residual map?
Expected input: Yes/No: 1 value
Specify if the residual map must be written.
For BEAM and UVCOVER cleaning, this map is an automatic by-product of the process and the choice is whether or not to write it to the .WMP file. The map will be given the same indices as the input map except for an incremented sequence number.
For UVDAT cleaning, the residual map must be constructed by making a new map from the original visibilities, in which the CLEAN components just found are subtracted. The new map will OVERWRITE the input map.See elsewhere for the rationale
Prompt: restored map (YES/NO)?
Expected input: Yes/No: 1 value
Specify if a restored map must be written.
A restored map consists of the CLEAN components convolved with a hypothetical beam that has no sidelobes, superimposed on the residuals. It is an approximation to what you would have observed with complete contiguous UV coverage (hoth in hour angle and baseline) up to the longest baseline available.
You will be given the option to suppress the residuals (parameter MAP_FACTOR), in which case you get a map of the CLEAN components only.
Prompt: dl, dm arcsec, pa deg
Expected input: Real: 1 to 3 values
Specify the restore beam width:
dl width of beam in arcsec (full-halfwidth)
dm width of beam in arcsec (full-halfwidth)
pa position angle of skewed beam in degrees (anti-clockwise;
0 deg is horizontally to the right (+l direction)
The default beam is a two-dimensional Gaussian truncated at a level of .xx relative to the maxiumum. If an antenna pattern for your map is available, the l,m beam widths are derived from it; else a rule-of-thumb formula is used:
half-width = 12 arcsec * 1400/frequency(MHz)
Prompt: ONE input map: grp.fld.chn.pol.0.seq
Expected input: Character *32: 1 to 1 values
Specify ONE input map (group.field.channel.polar.type(=0).sequence_number)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.