3 Observation and data processing

The observation was carried out in a 52 hour experiment on 28-30 October 1995 at 5 GHz in snapshot mode, with each scan lasting 30 minutes. Most of the scans were primarily distributed among the most interesting sources for imaging (5 - 8 scans on each source), and a few scans were devoted to the weaker sources and those with unknown milliarcsecond structure, in order to obtain their correlated flux densities for further study.

The VLBI array consisted of five antennas: Shanghai, Urumqi, Noto, Medicina and Hartebeesthoek telescopes. No fringes were detected on the baselines to Medicina. Parameters of the remaining four telescopes are listed in Table 2. The data were recorded in MK2 recording system with 1.8 MHz bandwidth in left-hand circular polarization. The correlation was performed in Medicina with the MK2 Block-0 Correlator of the IRA-CNR with 1 s integration time. Fringe fitting was done with the program PHASOR in the Caltech VLBI package. The station based residual fringe rates and the residual delays were estimated with a coherent integration time of 60 seconds. Data post-processing including editing, amplitude calibration and model fitting was carried out using the Caltech VLBI package. Amplitude calibration was done using the program CAL with the system temperatures and the gain curves of all telescopes.

  

Table 3: Model description of 11 imaged sources

The hybrid images were produced with the DIFMAP program (Shepherd et al. 1995). A point-source model was used as an initial step for iterative self-calibration procedure for all sources. The sources were initially cleaned and phase self-calibrated using uniform weighting in the (u, v) plane, then after several iterations we switched to natural weighting to reduce the image noise. Eleven sources were successfully imaged (see Figs. 1 to 11). The remaining ten sources were observed only for few scans in order to get information on their correlated flux density, and their u-v coverage was too poor to obtain images. The size of the restoring beam is shown as cross-hatched ellipse in the lower left corner and reported in the caption of each figure. The typical resolution of our images is about 1 mas, depending on the (u, v) coverage in each case. The lowest contour level in each image is 3 times of the rms noise. The rms noise in the images ranges from 1 to 6 mJy/beam. We used the MODELFIT program in the Caltech VLBI package to measure the positions and flux densities of the components in each image. Due to the small amount of data, the maximum number of Gaussian components used to fit both the closure phases and amplitudes in the calibrated data is limited to three. The results are listed in Table 3. In Table 3, Col. 1 is the IAU source name; Col. 2 is the component label; Col. 3 gives the flux density of each component; Cols. 4 and 5 list the distance and position angle of each component with respect to the strongest one, which we assumed to be at the rest frame; Cols. 6, 7 and 8 refer to the the parameters of Gaussian model: the width, the axis ratio and the orientation of the individual component, respectively. We estimate that the uncertainties of the parameters are: $\Delta S=\pm 20\%$, $\Delta r=\pm 0.1$mas, $\Delta {\rm PA}=\pm (5-10)^{\circ}$, $\Delta {\rm Major} =\pm 15\%$,$\Delta {\rm Ratio}=\pm 20\%$, $\Delta {\rm PA}=\pm (5-10)^{\circ}$.