Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.atnf.csiro.au/computing/software/miriad/userguide/node94.html
Дата изменения: Unknown
Дата индексирования: Mon Apr 11 01:51:16 2016
Кодировка:

Preparing your Data

Before you calibrate, you must prepare your data. This consists of loading, flagging, perhaps converting to ``channel-0'' datasets, and then splitting.

1. ATCA data will be initially in RPFITS format. Miriad's task to read RPFITS files is described in Chapter 8. Alternatively, the data may be in FITS format (e.g. VLA data). Again see Chapter 8 for information on loading data in this format. Generally you will want to load all of an observation into a single Miriad file (i.e. use atlod to read multiple RPFITS files into a single output if necessary).

2. You should now flag your data, and possibly convert to channel-0 (using uvaver). Tasks for these operations are described in Chapter 10. For CABB data, you should not convert to channel-0.

   Things to look out for while flagging are:

   • Shadowing: when in a short array, it is possible for one antenna to be partially blocked by another antenna. To flag based on this, use select=shadow(22.5) in your uvflag command.
   • Seeing monitor measurements: the seeing monitor is an independent interferometer which constantly measures the atmospheric phase stability to a geostationary satellite. The baseline length is 220m, and the satellite elevation is 60 degrees. The measurement is expressed in microns of excess path, and is contained in the dataset variable smonrms. For high frequency observations you may wish to discard data whose excess path length rms is greater than some threshold, and this can easily be achieved with the seeing parameter that the select keyword accepts.
   • Tracking accuracy: the tracking accuracy of the ATCA antenna is usually adequate at low frequencies, but at higher frequencies (where the primary beam is smaller), tracking errors become more deleterious. These errors may occur during windy weather (gusts can blow the antennas off the directed tracking position). In addition, poor tuning of the antenna drive systems can sometimes cause poor tracking, particularly after a source change.

     Note that tracking errors should be distinguished from pointing errors. Tracking errors are the inability of the antenna drive system to follow the requested path. Pointing errors correspond to the difference between the position that the astronomer requests and the actual position. Tracking errors are caused by flaws in the drive servo system, whereas pointing errors are a sum of tracking errors and errors in the antenna pointing model.

     Since October 2003, tracking errors have been saved in the Miriad dataset. The uv variables corresponding to the rms and maximum tracking error in a cycle are axisrms and axismax.

     Both the maximum and rms give two tracking error values for each antenna, nominally corresponding to the tracking error in the azimuth and elevation axes. However the ATCA on-line system produces a single composite value, which is replicated for the two axes in the dataset.

     When the tracking error is less than $2'$, the ATCA on-line system believes that this error is sufficiently small, and thus considers the antenna to be tracking. However this tolerance is not generally good enough for 3-mm observations. To flag based on this criterion, use the pointing parameter that the select keyword accepts.

3. As Miriad datasets can contain only a single set of calibration tables, it is rather poor at handling the calibration of datasets containing multiple sources and multiple frequency bands. For calibration purposes, it is best to work on datasets containing a single source and single frequency band. So, it is best to break the multi-source, multi-band dataset into a collection of single-source, single-band datasets. The best task to do this is uvsplit. Task uvsplit generates the names of the output datasets itself, forming these from the source name and the central frequency (in MHz) of the data. It is rather unforgiving if you already have files with one of the names that it wants to use. Make sure your directory is free of files that may usurp uvsplit's name choice. Task uvsplit allows you to perform extra selection if you wish, which may be convenient if you only want to deal with part of your observation at a time.

   UVSPLIT
   vis=multi.uvxy	The input dataset.
   select	Extra selection.

   For CABB data, or large spectral line data-sets, if disk space is low, it may be useful not to split off the program source. Rather you can split off the calibrators, determine the calibration tables from them, copy the calibration back to the multi-source file, and then image directly from the multi-source file. This way you avoid making a second copy of your program source data. For example, to avoid the source ``vela'' from being split off, use

   UVSPLIT
   vis=multi.uv	The input dataset.
   select=-source(vela)	Do not select vela data.

   There are a few things to look out for when using uvsplit. If you don't care about the files already extant in the current directory, you may use options=clobber to make uvsplit overwrite any files it wants to create.

   If the parent file contains more than one frequency band at the same frequency, then uvsplit will output two datasets in the normal way, but these will have .1 and .2 etc. appended to the end of their filenames, indicating which frequency band the data has come from.

   Further to that, sometimes it can be handy to operate on only one IF as an individual dataset, and this can be achieved by using options=nosource. This creates a dataset for each IF in the parent file, but will not split out the individual sources. A similar thing can be done to keep all IFs in each source file (options=nofreq).

4. Handling CABB's large fractional bandwidths can sometimes cause problems, and it may be preferable to split the band up into smaller bandwidth chunks. This can be done using the maxwidth parameter. This parameter should be set to the largest bandwidth allowed in a single dataset, in GHz. For example, to split the 2 GHz CABB continuum bands up into 128 MHz chunks, use

   UVSPLIT
   vis=multi.uv	The input dataset.
   maxwidth=0.128	The maximum bandwidth for each output dataset, in GHz.

   Splitting the data into smaller bandwidth chunks in order to more accurately calibrate it has been dubbed "divide and conquer".