Models in Newstar
(contributed by W.N. Brouw 931011, edited by J.P. Hamaker 940411)
Source models in Newstar are in principle meant to belong to a certain observation, although models can be converted from/to belonging to other observations.
Each model has a header. In this header the following information is stored:
the reference right-ascension and declination ( apparent or
epoch ) to which the l,m coordinates of the source components refer. It is assumed that this is also the centre of the primary beam of the individual telescopes.
If the model list was made through the ADD and EDIT options of NMODEL , the type will be local . If it is generated through NCLEAN or the FIND option of NMODEL , the type and other data are taken from the corresponding map.
A local list can be converted to an apparent or epoch type though the CONVERT or EDIT options of NMODEL . In this case the reference data provided (either from a scan header or manually) will be copied into the model header.
An apparent or epoch list can be converted to local in the same way.
An apparent or epoch list can be converted to another apparent or epoch list with the NMODEL EDIT or CONVERT options. For EDIT, the new reference data from the user will replace the existing header data, without changing the actual source components. CONVERT will also convert the existing source and header data to the new coordinates, i.e. the new source list will be for the new RA, DEC, FREQ, INST, using the model components' spectral index, rotation measure, primary beam and sky position.
An apparent or epoch list can be converted to an epoch or apparent list by NMODEL EDIT or CONVERT. For EDIT, the new values will replace the old header data, and only the change in direction to the North Pole ('phi') will be applied to the individual sources. If CONVERT, only a reference scan can be used, and the source components will be shifted, rotated, beamed, spectrally indexed etc. to the new coordinates.
Each source in the source model has the following information:
, in general in Westerbork Units) of the source. This value is always posaitive.
the l and m coordinate on the sky for an E-W interferometer pointed at the reference coordinates in the model header; l and m are thus given as offsets to the header position.
a number to be used to select particular sources only for certain operations. In general only type 0 will be used in the programs (e.g. self-calibration, updates, subtractions), but sometimes type = 1 are used (in a different way, e.g. in
UPDATE they will not be updated, but subtracted from the observations before updating starts). Other types are not used at present. The user can manually set them (NMODEL
EDIT) to e.g. temporarily 'delete' them from his model.
the source intensity has been corrected for the primary beam of the instrument.
Whenever NMODEL CONVERT is used, the above data will be adjusted to reflect the new header data. Note that a beamed source will remain unchanged if the frequency and/or instrument is changed but no spectral index is available.
All of the above information can be changed by the NMODEL HANDLE EDIT and FEDIT options.
The beam mode can be changed (and the intensity changed) by the NMODEL [DE]BEAM options. This operation will only be applied to sources for which it is appropriate.
Models can be combined with the NMODEL xxx option, provided either
The two models will be automatically converted and merged into a single model with one of the header s being used.
Conversion of a model to visibilities always occurs in the context of a particular scan in a SCN file. In this process, the model is automatically converted from the model-header coordinates to those in the scan header.
In this process, beamed source components will be de-beamed before being used; de-beamed components will first be beamed for the header coordinates, then de-beamed for the scan coordinates). In addition non-clean components will be smeared for the appropriate bandwidth and integration time.
The user can suppress the smearing and the beaming operations (NMODEL MODEL_ACTION keyword). He has no control over the other conversions (spectral index, different position etc.).
In addition to the corrections discussed so far, position-dependent instrumental polarisation (somewhat unfortunately often referred to as cross polarisation) should be corrected for in the calculation of model visibilities. It is understood in principle how this must be done, but the implementation has been suspended until proper measured parameyters become available.
The primary beam for the WSRT is currently modeled as a circularly symmetric 
function. Pointing measurements show that this is a good approximation out to the first null. If positions further out were to be included, it would probably be better to model the beam as a J0 (Bessel) function or as an 
.
For the ATCA the beam is given as an inverse polynomial, which is correct to the first side-lobe (except for non-symmetry around the legs).
The NMODEL UPDATE, XUPDATE and SUPDATE options try to match a givel source model the data as well as possible to an observation in a .SCN file.The update process extracts its information from residuals of the observed visibilities after all model components of types 0 and 1 have been subtracted. Updating is a process of fine-tuning the parameters of the model components and is therefore not applicable to clean components.
To make the process not too slow, each source component is fitted separately, special care being taken that components close together do not unduly influence each other's updates. For the process to work it is necessary that observation(s) span as large an HA range as possible; otherwise the exact distance between to closely-spaced components is difficult to determine.
The fitting process is linear, and uses both the real and imaginary parts of the visibilities. Since the equations used are linearized, more than one iteration is in general needed for the process to converge. In the equations no account is taken of secondary effects (e.g. a shift in l,m will lead to different amplitudes due to primary beam and frequency dependencies). It is assumed that these effects will also iterate out. In those cases where this does not happen, it probably would not have happened if they were taken into account; due to dependencies between equations.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.