These re-measurements have provided positions with a typical accuracy (standard error) of around 0.6 arcsec in each coordinate for the stronger compact sources, flux densities at both 4800 and 8640 MHz, and structural information for the individual source components which is sufficient to outline the gross morphology of the sources.

2. OBSERVATIONS

2.1 Observing method

Observations were made simultaneously centred on two frequencies of 4800 MHz and 8640 MHz. The total correlator bandwidth at each observing frequency was 128 MHz, although only the central 50% of each band was used so as to reduce "bandwidth smearing" of the synthesised images caused by frequency averaging. For each band, two orthogonal linear polarisations were combined and the data averaged on-line so as to provide a total intensity.

Phase and position calibration of the target sources was performed using standard ATNF Compact Array calibrator sources, which were observed once every 30 target sources (see below). We flux calibrated the data using daily observations of the standard source PKSB1934-638, for which we have adopted flux densities of 6.33 Jy (at 4800 MHz) and 2.59 Jy (at 8640 MHz).

Our observing method made brief "cut" observations of the target sources, each lasting about 45 seconds. Every source was observed 3 times: once approximately 4 hours East of transit, once near transit, and once approximately 4 hours West of transit. The profile resulting from each cut observation was reduced and displayed in real time, thus permitting data quality to be continuously monitored. When 3 cuts had been obtained at the three different hour angles, the profiles were reduced and combined to produce positions and fluxes for all source components.

2.2 Scheduling

For each band, we further divided the sources into "packets" of 30 target sources plus one nearby phase-calibrator source. Typically, the calibrator source lay within 10 degrees of the centre of the source group. The calibrator was observed at the start of each packet and this information used to calibrate the positions of the associated target sources in the CASNAP program (see below).

In principle, 4 hours of observing in the East could be followed by 4 hours near transit and then 4 hours in the West, so that three days of continuous observing would complete a declination band. In practice, however, we had to vary the times spent observing in each sky "window" so as not to return to a part of the sky which had already been observed 24 hours earlier.

Approximately 10% of the target objects had to be re-observed in order to obtain satisfactory data. Where this occurred, we tried to re-observe all three cuts for the source on the same day so that any possible variability would not influence the data quality.

2.3 Data Reduction

The principle steps in the CASNAP reduction were:

• 1: We calibrated each of the target fields in a 30-source "packet" for both amplitude and position using the associated calibrator observed at the start of the packet.
• 2: The calibrated visibilities were averaged over the 45s observing time for each source.
• 3: We flagged sources which had anomalously high or low visibilities relative to the median amplitude over all baselines. However, we ignored anomalously high visibilities on the shortest baseline (either 75 or 300m, depending on the array configuration) because of the possibility of extended emission.
• 4: For each snapshot or "cut", we determined both a "dirty source" cut profile and the related "dirty beam" cut profile. The dirty source cut profile was determined by Fourier transforming the observed visibilities. We applied a scaling factor for both the beam and source cut during this process to account for the fore-shortened baseline spacings.
• 5: We deconvolved each source cut using a standard implementation of the Hegbom CLEAN algorithm, and accumulated the CLEAN components. The CLEAN was terminated either when the residuals converged or when the maximum number (500) of CLEAN iterations was reached.
• 6: Based on the known half-power beam width (HPBW) of the array, we tested whether any of the CLEAN components were unrealistically close to each other. If so, we re-combined them into a single component.
• 7: We sorted the CLEAN-ed components for each cut into flux order and retained only the two most significant. This, we justified as follows: at the flux levels at which we were working, the statistics of the source counts showed that there was very little chance of two unrelated sources occurring in the same field. And our three-cut observing method did not permit us to delineate accurately the structure of sources which had more than two components.
• 8: We optimised the position and flux values of each of the "cut-components" by minimising the square of the difference between the modelled and observed visibilities.
• 9: We displayed the CASNAP results on-line in order to monitor the data quality and logged the modelled cut-component values to a database for further processing by the POSNAP program. This completed the CASNAP reduction.

The POSNAP program combined the reduced cut-component data from the individual cuts and automatically determined source component parameters once sufficient cuts had been obtained. The main steps in the POSNAP procedure were:

• 1: The program checked that the cuts contributing to each source were separated in hour angle by at least 10 degrees and were well distributed across the sky. Experience convinced us that at least three cuts were required for adequate source fitting.
• 2: We modelled the positional uncertainty for each cut-component as a normal distribution whose dispersion was the quadrature sum of both a flux-independent and flux-dependent contribution. The component's uncertainty in position orthogonal to the cut was approximated as a uniform distribution of width equal to the processed field size, 120 arcsec.
• 3: We used the positional uncertainties and cut-component widths to model the probability that a cut-component appearing in each of the cuts was produced by a real source component. To do this, all possible cut- component combinations were determined and a probability product computed for each combination.
• 4: We ranked all products and iteratively processed the data to attempt to remove any degeneracy within an East-Transit-West cut-component combination. Such degeneracy would occur, for example, when the hour angle at which a particular cut observation was made was very similar to the relative orientation of two or more source components on the sky. The probability products were recomputed and surviving combinations re- ranked.
• 5: If the highest-ranked probability product allowed a non-degenerate, and significant, estimate of integrated flux, we noted this combination as representing a "real" source component. We then removed the cut- components contributing to the real source component from any other combinations and the residual fluxes were re-assigned. Steps 4 and 5 were repeated until either all the cut-components were "used" or the maximum allowable number of real sources components was reached.
• 6: We computed the best position and position error for each real source component using the spatial location and extent of its associated probability product. The integrated flux for a source component was computed as the noise-weighted average of the non-degenerate cut-component contributions. The maximum and minimum widths for a source component were determined directly from the CASNAP cut profiles.
• 7: We cross-correlated source components found at both of our observing frequencies (4800 and 8640 MHz) based on positional coincidence within a "3-sigma" region of combined uncertainty and computed spectral indexes based on the integrated fluxes where appropriate.
• 8: Finally, we manually inspected the source list and removed a few anomalies produced by the automated reduction procedure by combining multiple components into a single component for a few extended sources. In the great majority of cases, however, we have preferred to list the direct output from the automated process so that the user can make up his or her own mind as to the complexity or otherwise of the sources.

This completed the POSNAP reduction.

The main reason we were led to produce our own reduction software was that the reduction of data for several thousand sources using conventional, manual methods, such as AIPS or MIRIAD, is a very onerous task. Furthermore, conventional methods do not permit data quality to be assessed during observing. The main advantages of our method were that it was much faster and more automated than conventional methods. Its main disadvantage was that it did not combine the data from each cut prior source to fitting. This resulted in a poorer signal-to-noise on each source component.

As mentioned earlier, the Compact Array observations and our automated reduction method produced a result in approximately 82% (=6603/8068) of the sources observed. In the majority of cases we concluded that the reason no result was obtained in the other 18% of cases was that the sources were very extended and had been completely resolved by the Compact Array. This conclusion was reinforced by plotting the proportion of missing sources as a function of galactic latitude. As expected, the missing sources were located preferentially towards the galactic plane where such extended radio sources as HII regions and supernovae remnants predominate. However, we were also interested to know if more conventional reduction techniques would increase the number of detected sources. We therefore re-reduced the data for a small, representative group of objects using MIRIAD. Based on this sample, we conclude that a further 3% of objects would be detected if all the data was analysed using the more laborious method and this work may be undertaken some time in the future.

3. RESULTS

Presenting data for 6603 objects in tabular form presents a serious challenge. In addition, such large collections of data are increasingly required in "electronic" form, rather than as "hard copy". Therefore, we decided to publish the results from our work on the Australia Telescope's FTP server, which can be accessed using the account anonymous as:

ftp.atnf.csiro.au and in the area /pub/data/pmn/CA.

In "hard copy" versions of this paper we present only a sample page of the main data table. The table may be downloaded in full from the above server, in plain-text ("ASCII") format, as:

table2.txt.

This directory area also contains a READ.ME file which should be read first for the latest information.

      A typical ftp session might be as follows:

     ftp ftp.atnf.csiro.au         (connect to the Australia Telescope
					    National Facility's FTP server)

      login: anonymous              (log in as an anonymous user)

      password: your password       (supply your email address as a password)

      cd /pub/data/pmn/CA           (change to the appropriate directory)

      ls                            (list all available files as a check)

      get READ.ME                   (to download the READ.ME file)

      get table2.txt                (to download the table2 data file)

      where the commands to be typed are shown in a bold, monospaced typeface.

This paper, and the data, may also be read and downloaded from the Parkes World Wide Web HomePage, Surveys HomePage at:

http://wwwpks.atnf.csiro.au/databases/surveys/surveys.html

Table 2 lists the individual components for each PMN source observed. In this table, Column 1 contains the name of the PMN survey source from which the component is derived. Where multiple components were found for the same "parent" survey source, they have been indicated by an asterix. Note that there is no guarantee that these components are always physically associated, since there is a small, but finite, chance that two unrelated sources may close together on the sky.

Columns 2-7 refer to data at our lower observing frequency of 4800 MHz. Columns 2 and 3 list the source component right ascension and declination (equinox and equator J2000). Column 4 gives the peak flux (Speak) measured for the source component (in mJy) while Column 5 gives the corresponding integrated, or total, flux (Stot). Columns 6 and 7 give the major (Wmax) and minor (Wmin) half-power widths for the component measured from the three cuts. These have been normalised to the sizes of the telescope's synthesised beams, which were 1.5 arcsecs at 4800 MHz and 0.8 arcsecs at 8640 MHz. Thus, an unresolved source would have a width of 1.0 at either frequency.

Columns 8-13 list data similar to Columns 2-7, but refer to the higher observing frequency of 8640 MHz. Finally, in Column 14, we provide a spectral index for each component where this could be determined, defined between 4800 MHz and 8640 MHz and computed from the total fluxes, Stot.

4. DATA QUALITY

As mentioned in Section 2.1, positional calibration of the target sources was performed using 141 standard Compact Array compact calibrator sources of accurately known position. Since they are also catalogued in the PMN Southern Survey, they were also observed as ordinary target sources. Thus the positions measured for them during our programme could be compared with the more accurately known positions to provide an indication of the overall survey positional accuracy.

The results of this comparison suggest that standard errors of about 0.6 arcsec indicate the accuracy of the measured source components in each coordinate, at least for these stronger sources. The positional accuracy of the weakest sources is expected to be somewhat poorer (perhaps ~1.0 arcsecs) although we have not been able to determine this value accurately.

We have also compared the measured positions of 103 strong sources in our sample with those listed in PKSCAT90 (Wright & Otrupcek, 1990) as having accurate (s.e. Estimating the flux density accuracy is difficult because extended objects may have been resolved by our high-resolution Compact Array observations. On the other hand, compact sources may have varied between the epochs of the original PMN surveys (circa 1990.5) and the Compact Array re-measurements (circa 1994). Despite this problem we thought it worthwhile to produce a comparison of the 4800 MHz Compact Array fluxes (after combining adjacent components where appropriate) with the original PMN fluxes (measured at 4850 MHz).

The resulting flux-flux plot is shown in Fig. 1. As expected, the plot shows the effects of resolution, with the higher-resolution, Compact Array fluxes being appreciably smaller than the Parkes 64-m telescope PMN fluxes in many cases. On the other hand, a few objects have larger fluxes when measured with the Compact Array, presumably as a result of variability between the epochs of the two surveys. However, there also appears to be a well-defined upper envelope to the majority of the data, which probably results from sources which have not varied appreciably between the two epochs. This envelope has a slope close to +1.0 in the logarithmic flux-flux plot and suggests that there is no appreciable systematic flux differences between the two sets of data.

In summary, we estimate that the standard error of the flux densities is given by:

 standard error in flux2 (mJy) =  (5)^2  +  (0.05.S)^2    (4800 MHz) and
			       =  (8)^2  +  (0.07.S)^2    (8640 MHz),

5. CONCLUSIONS & FUTURE WORK

Of the 8068 sources observed, we detected at least one source component in 6603 of them. These sources revealed a total of 7177 individual source components. The 1465 sources which satisfied the PMN criteria but which were not detected by the Compact Array observations presumably have very extended structure and were totally resolved by the instrument.

The reduction of the data was made essentially automatically using two computer programs, CASNAP and POSNAP, which were developed specifically for our programme of observations. These programs were capable of producing fully-reduced data in real time.

The positional accuracy of the data reported here is about 0.6 arcsec for each component and in each coordinate at both frequencies. The flux density accuracy varies from about 5mJy for the weaker sources up to about 50mJy for a 1Jy source at 4800MHz.

The data is available as a table describing the individual source components. This table can be downloaded using "anonymous" FTP from the Australia Telescope National Facility's server, ftp.atnf.csiro.au .

Finally, we are undertaking a programme to produce high quality optical identifications for the 6603 sources listed above. This programme is now essentially complete (Tasker & Wright, 1993; Tasker, 1996). In addition, the accurate positions for the radio sources reported in this paper are being used to produce cross-correlations with catalogues compiled at other wavelengths. The results of this work will be reported in the near future.

REFERENCES

Condon, J. J., Broderick, J. J., and Seilstad, G. A. 1989, AJ, 97, 1064

Condon, J.J., Griffith, M.R., & Wright, A.E., 1993, AJ, 106, 1095 (Paper 4)

Dickel, H.R., Lortet, M.-C., & de Boer, K.S. 1987, A&AS, 68, 75

Gregory, P.C.,& Condon, J.J. 1991, ApJS, 75, 1011

Griffith, M.R., & Wright A.E. 1993, AJ, 105, 1666 (Paper 1)

Griffith, M.R., Wright, A.E., Burke, B.F. & Ekers, R.D. 1994, ApJS, 90, 179 (Paper 3)

Griffith, M.R., Wright, A.E., Burke, B.F. & Ekers, R.D. 1995, ApJS, ??, ??? (Paper 6)

Kuhr, H., Witzel, A., Pauliny-Toth, I.I.K. & Nauber, U. 1981, A&AS, 45, 367

Large, M.I., Mills, B., Little, A.G., Crawford, D.F. & Sutton, J.M. 1981, MNRAS, 194, 693

Tasker, N. & Wright, A.E., 1993, Proc. ASA, 10, 4, 320

Tasker, N., Wright, A.E., Griffith, M.R., & Condon, J.J.1994, AJ, 107, 2115 (Paper 5)

Tasker, N., Wright, A.E. & Griffith, M.R., 1996, AJ, In preparation (Paper 7)

Tasker, N., 1996, PhD thesis, In preparation

Wright, A.E., Griffith, M.R., Burke, B.F., & Ekers, R.D. 1994 ApJS, 91, 111, (Paper 2)

Wright, A.E. & Otrupcek, R., eds. 1990, ATNF, "PKSCAT90 - the southern radio database", (Available via "anonymous FTP" from ftp.atnf.csiro.au under the subdirectory /pub/data/pkscat90)

					4800 MHz                                                         8640 MHz
   NAME         RA          DEC        Speak Stot           Wmax     Wmin         RA         DEC        Speak Stot           Wmax     Wmin      SI
	       J2000       J2000        mJy         mJy     norm     norm       J2000       J2000        mJy         mJy     norm     norm
--------------------------------------------------------------------------------------------------------------------------------------------------
J0000-4722  00:00:29.85 -47:21:47.2            6      64     11.5     1.0
J0000-8539  00:00:12.04 -85:39:19.9       101        108      1.1     1.0    00:00:12.04 -85:39:19.9        69         70      1.0     1.0    -0.7
J0001-3931  00:01:43.08 -39:30:43.6            1      56      6.8     5.8
J0001-4630  00:01:26.89 -46:30:12.4        85        106      1.2     1.0
J0001-6715  00:01:20.90 -67:15:37.1            2      40     10.6     1.6    00:01:21.09 -67:15:38.3            7      16      1.9     1.2    -1.6
J0001-6849  00:01:50.63 -68:49:18.9            9      58      3.4     1.8    00:01:50.13 -68:49:22.1            2      28      9.1     1.3    -1.2
J0001-7051  00:01:01.92 -70:51:07.8            8      42      5.1     1.1
J0001-7511  00:01:01.82 -75:12:00.2        79         97      1.2     1.0
J0001-7609  00:01:23.23 -76:09:11.3        56         70      1.3     1.0    00:01:22.89 -76:09:12.8            8      29      3.4     1.0    -1.5
J0001-7702  00:01:04.21 -77:01:55.7            6      35      5.6     1.1
J0002-5025  00:02:42.30 -50:24:46.9        96         96      1.0     1.0    00:02:42.29 -50:24:46.9        38         39      1.0     1.0    -1.5
J0002-5621  00:02:53.58 -56:21:11.4       245        310      1.3     1.0    00:02:53.59 -56:21:11.3       129        135      1.0     1.0    -1.4
J0002-6122  00:02:17.22 -61:22:16.1            6     111      4.8     3.8    00:02:17.42 -61:22:15.8            1      46      9.3     4.4    -1.5
J0003-4114  00:03:53.47 -41:15:00.0        72         79      1.1     1.0
J0003-5444 *00:03:11.91 -54:45:08.0            5     223     14.4     3.2
J0003-5444 *                                                                 00:03:10.16 -54:44:55.4            1     122     21.6     6.8
J0003-5905  00:03:13.29 -59:05:48.0        91        156      1.5     1.1    00:03:13.30 -59:05:47.9        38         77      1.6     1.3    -1.2
J0004-4033  00:03:57.85 -40:32:46.5        39         58      1.5     1.0
J0004-4144  00:04:32.40 -41:43:54.1        43         84      1.7     1.1    00:04:32.39 -41:43:53.8        15         39      2.1     1.2    -1.3
J0004-4345  00:04:07.26 -43:45:09.5        77        156      1.9     1.0    00:04:07.26 -43:45:10.1       192        201      1.0     1.0     0.4
J0004-4736  00:04:35.66 -47:36:19.6       725        748      1.0     1.0    00:04:35.66 -47:36:19.7       715        736      1.0     1.0     0.0
J0004-5254  00:04:14.14 -52:54:58.4        32         52      1.6     1.0    00:04:14.15 -52:54:58.6        20         38      1.9     1.0    -0.5
J0004-5546  00:04:21.97 -55:46:56.2            7      86      6.3     1.9
J0004-6413  00:04:17.18 -64:13:38.9            7      57      3.9     2.1    00:04:17.10 -64:13:38.7            2      26      5.3     2.2    -1.3
J0005-5049  00:05:08.13 -50:49:16.0            2      30      5.2     2.6
J0005-5428  00:05:39.19 -54:28:31.7            2      74      9.4     4.3    00:05:39.07 -54:28:31.9            4      58      5.3     2.6    -0.4
J0005-5751  00:05:24.72 -57:51:53.3       107        164      1.5     1.0    00:05:24.72 -57:51:53.3        41         69      1.6     1.0    -1.5

Fig 1. Comparison of fluxes measured with the Compact Array (CA) and the original