A postscript version of this document can be obtained here
and the WSRT
For reducing WSRT data with Miriad, please note the following:
Be sure that the WSRT MeasurementSet has been converted to a UVFits file in the correct way. This is done with the Aips++ task
ms2uvfits ms=<infile> fitsfile=<outfile> multisource=T writesyscal=T combinespw=T
This ensures that the proper calibration information is written in the UVFits file. If you have requested to get the data in UVFIts format, the conversion has been done for you.
Read the UVFits file with an adapted version of the fits reader (ivcFits instead of fits). This ensures a correct interpretation of the (WSRT dialect of the ) UVFits header, as well as that some extra calibration information is written into the Miriad file. Only by using ivcFits you will be able to do a correct reduction of the data.
The UVFits files written by ms2uvfits have the Y-coordinate axis of the telescope positions swapped (because classic Aips requires this). This means that this axis runs the other way from what is used for ATCA. This implies that flagging on shadowing will flag the wrong dish!!!! For WSRT data one should use an adapted version of uvflag that is available from the download link.
The ms2uvfits converter does not write the velocity information in the UVFits file. If you have done a spectral-line measurement, you have to give this info by hand. Use the keyword velocity=velsys, refval, refchan of the task ivcFits. The default is refchan = 1. For a standard WSRT line observation, the reference channel is nref = N/2=1 , where N is the number of channels you used.You also have to put the restfrequency of the line in the Miriad uvfile by hand (using the task puthd, the restfrequency of the H I line is 1.420405752 GHz)
Standard also the auto-correlation spectra are written into the data file. These are a problem for the Miriad task selfcal since that task wants input files that contain cross-correlation data only. In order to be able to run selfcal, you have to flag the auto-correlations and copy the uvfile with uvcat with options=unflagged.
Unlike standard practise for ATCA, one has to apply the Tsys correction to the data manually. It may be wise to apply these corrections to the data, in particular at 21 cm. This can be done with the task attsys, available in the latest Miriad version from Epping, or from our web pages. Note however, that the estimates of the Tsys values are always accurate and can be noisy. One should inspect the Tsys using varplt to check whether it does more damage than good. If you do not apply the correction, one has to recover the gain changes with amplitude selfcalibration.
We have modified the model information of some of the standard calibrators as it is used by mfcal. This in order to cater for the lower frequency bands used by the WSRT. However, using these new model spectra requires recompiling the module calstoke.for from the Miriad library, apart from re-linking mfcal. We intend to make a WSRT version of mfcal and make that available through these pages.
The new model spectra implemented are for 3C 48 and 3C 147, and are taken from the VLA pages (exact ref soon to follow).
Note that CTD93, a calibrator often used at the WSRT, is not known to Miriad. The flux of CTD 93 at 21 cm is about 4.83 Jy.
Data from the DZB correlator suffer very badly from Gibb's ringing in the spectra and this can cause big problems in your data. You may have observed with on-line Hanning or Hamming smoothing to suppress this. If you have not done so, you can use the program uvfil that applies a Hanning smoothing to the visibility spectra. The effect of uvfil is the same as the on-line smoothing of the WSRT (exactly the same thing happens) and the existence of uvfil could even be a reason for you never to use on-line smoothing (if Hanning is OK for your data. Why throw away information that you can also throw away later?). This is different from ATCA where there are subtle differences between on-line an off-line smoothing. Other filters (like Hamming or Tukey) will be implemented in this task in the near future.
Due to the large volume of WSRT data sets, it can happen that tasks like invert or uvmodelrun into problems with scratch files larger than 2 GB. Adapted versions of these programs (execs only) that do not have this problem are available from the download link . They are called invertMore and uvmodelMore.
The numbering of the antennae in Miriad is 1-based, while the standard WSRT numbering is 0-based. So, e.g. ant(1) is RT0. This is a source of confusion. ant(11) is RTA, ant(12) is RTB, ant(13) is RTC and ant(14) is RTD.
An adapted version of uvplt is available. This new version allows to plot data from 91 baselines instead of the maximum 36 baselines of the previous version. The latest version of Miriad from Epping should have this modified uvplt, it is also available from our download link.
We are working on a seperate version of Miriad that is optimised for the WSRT. In the foreseeable future this will be released as a complete package.
For question & remarks, contact Tom Oosterloo
The source of the adapted software and binaries compiled under Suse 8.3 can be copied from our download link.
Date: 2004/09/22 10:17:17