(07.03.2013 Neal Jackson)

One of the early MSSS HBA datasets, H242+27, was taken using six of the international stations (all except Nancay and Effelsberg). By chance there is a bright PKS compact source (J160913+264129) very close to the centre of this field, making it a good source to test the existing procedures and parsets developed by Adam, Tobias and Olaf, all together. The dataset consisted of L87276 (BEAM_3) on the target and L87277 (BEAM_0) on the calibrator.

Two further scripts have been written on top of the existing software to tie all the analysis together: locapi and lofipi. These were tested on the H242+27 dataset, and are briefly described below.

locapi performs the following steps in an automated fashion (i.e. you start it and come back the next day for the output/error message):

* Make all the necessary BBS and NDPPP parset files

* Find all subbands of the calibrator and average them (by default, to 4ch/4s)

* Run BBS on the calibrator data for the core stations

* Apply these solutions to the target

* Shift, average, and phase up the core to a given source position

* Make IFs with a defined number of subbands per IF

* Concatenate the IFs and convert to circular polarization using mscorpol

In addition, there is an option to look for all NVSS sources within a given radius of a central position and do the last three steps separately for all of those sources, producing thereby n fits files for n sources. Instructions are given in the script and involve specifying a few obvious things (directory locations of data, prefixes of data files, locations of clusterdesc files/mscorpol scripts) and a few less obvious (RA and dec of the phase-up calibrator, and the sbegin parameter which controls the operation of the script).

The only known bug is the combination of subbands into IFs; this will fail if the frequency ranges are non-contiguous. In the H242+27 data, combining 10 subbands per IF appears to work, but I don't know if this is universally good.

Because Parseltongue is not installed on the cep1 cluster, the FITS files need to be moved somewhere else before running lofipi. This script:

* Reads in all files with a specified prefix and ending in '.fits' into AIPS

* Attempts to fringe-fit using the international stations, and the superterp as a reference antenna

* Produces two outputs; a table with the scatter of the fringe-fit (effectively telling you whether the source is detected or not) and a series of figures that look like this:

http://www.jb.man.ac.uk/~njj/PP_TCAL.png

This is a set of plots, one for each telescope and polarization, of delay (colour scale) on axes of time (x) and IF(y). The above is the calibrator; a source which doesn't work is shown below

http://www.jb.man.ac.uk/~njj/PP_TSRC2.png

A summary is also printed of the scatter in these plots, which effectively tells you which sources are detected (and which antennas are probably not working, DE604 in this case).

In the case of the H242+27 field, it looks as though the calibrator (the PKS source) has coherent phase solutions, plus one other, B2 1605+26. However, using the solutions and attempting to image, two further sources are detected (see table). Using solint=10, a few further sources become visible, suggesting that the ionosphere was relatively well behaved during these observations. The less good news is that taking the solint=1 solutions on the PKS sources, copying these to the other sources and attempting to image them is not successful in any of the other sources (including those in which calibrating on the sources themselves is successful). This should preferably be investigated/replicated by others in case of errors on my part, but it does seem to agree with the conclusion reached in the busy week; the temporal stability of the ionosphere is relatively good but the spatial correlation of phases is not.

(Note that the integration time smearing probably restricts the effective field to about half a degree, or a little more). Key in the detection column: x = visible in image with solint=1 in FRING; o=visible with solint=10; ?=possibly visible with solint=10 if you stare very hard at the dirty image.