Observations and data analysis

All data analysis was performed using the MIRIAD package. The observations were done at 1408 and 1708 MHz, with a bandwidth of 128MHz using the usual ATCA setup, although only the 1408 MHz data is used here. All data was collected from the ATCA, with the timing positions presented by Camilo et. al.[2] from the Parkes 64m telescope being used to identify the pulsars that we remove later. The cluster was imaged in a number of different configurations (table 1), over many days.

Table: Observing configurations. Field centre 0:24:06.00,-72:03:00.00 (J2000)

Observation date	99-Feb-12 -- 99-Dec-31	99-Jan-25	00-May-08	00-Dec-23	00-Dec-25	00-Dec-27
Configuration	many: 1.5A, 1.5B, 6A, 6C, 6D	750B	375	750C	750C	750C
Image type	Large array for high res image	5 antennae for extended sources, 6th for removal of bright pulsars
620mmComments	635mmMany days combined to make a very high res image with very high S/N of $28 \mu Jy$	630mmBecause of maintenance, 3rd antenna missing for most of day, and many faults during observing $\Rightarrow$ only 4 antennae left. Data unused currently				615mmLoss of several hours because of error and thunderstorm

The 1.5 and 6km array data, that are not sensitive to the extended component of the unresolved pulsars, were mosaiced together to form a sensitive high resolution image of the core[3]. The 375 and 750m arrays were combined to form a low resolution image of the core and was well matched with the spatial distribution of the known pulsars. An attempt was made to verify, using UVGEN, that a source with characteristic size of the pulsar distribution half width of 70''[2] would not be attenuated appreciably by the $\sim 60$'' beam size of the 375/750 data. This work is incomplete. A self-calibration of the low resolution image using itself as a model would be extremely difficult, since there are only 5 antennae that are used to form the small baselines. The individual small array images were self calibrated (for amplitude and phase) using GPSCAL, with the model being McConnell and Ables' (2000) 6km mosaic. This might have the side-effect of decreasing the amplitude of anything that does not appear in the 6km mosaic - which would include the extended flux we are searching for. Subtracting the point sources in the high resolution image should leave us with the extended component we are looking for in the low resolution image (as well as some sources that vary between the observations). Because McConnell and Ables' (2000) 6km mosaic has modelled the point sources well, we can subtract them from the UV data, removing both the sources and their sidelobes from the image. We do this by forming the UV data for the smaller arrays individually, and subtracting the clean components of McConnell and Ables' (2000) mosaic formed by using the 1.5 and 6km arrays, using UVMODEL. Because this will remove the pulsars that appear in McConnell and Ables' (2000) high resolution mosaic (ie, those ones that are bright enough for a few of the observations), we mask the clean image first, using MATHS, to exclude the pulsars (and anything else within 130'' of the core), and assume that anything that appears within this mask and is similar in all the days observations is a background source, so remove those manually from the low resolution data. Each of the pulsars scintillate, and some are quite dramatic - if one happens to be bright on any one day, then we may falsely conclude that we have found the extended flux, when the extra flux really belongs to a single pulsar, and not 200 of them. Hence, we form an image using just the antenna 6 baselines (hence the UV coverage is poor, with only $\sim 6$km baselines) for each day, and IMFIT to each of the sources that are both above $3 \sigma$, and that correspond to one of the already known and located pulsars. The sources were both modelled as point sources, and Gaussian (since they may vary within the day's observation, due to scintillation), but the difference between the two was minimal, so the point source method was chosen for simplicity. We assume that the pulsars we do not know about never get bright enough to skew our extended flux measurements significantly. Since the 6km baseline image is hard to calibrate because of the lack of good UV coverage, we assume that the two background sources near the core are reasonably constant, and scale the pulsars by an amount that scales the brightest of these `calibrators' to the level obtained in the McConnell and Ables' (2000) 1.5/6km mosaic. The scaled fluxes from the pulsars are then subtracted from the UV data from each individual day, again with UVMODEL, and a combined image is again formed. This image should now only contain the extended component due to the unresolved pulsars in the centre, with the main sources of noise being low level confusion.