3 Data reduction

The data were correlated with the NRAO processor at Socorro. Amplitude calibration was obtained using the system temperature measured at the various antennas during the observations and the gain information provided by NRAO. Consistency checks were performed with OQ208 and DA193, whose amplitudes were found to be generally within 2% of what expected.

Standard editing and fringe-fitting were applied. Most sources provided fringes with high signal-to-noise ratio on all baselines.

Radio images were obtained using AIPS, after a number of phase self-calibrations, ended by a final amplitude self-calibration. The last step was made with great care, as the gain self-calibration tends to depress the total flux density when extended components are poorly sampled in uv coverage. Generally the gain corrections from self-calibration were <3%. In those sources in which they were $\geq$3%, or in which the total flux density decreased by more than 1-2%, they were not considered.

The rms noise level (1 $\sigma$), measured on the images far from the sources, is in the range of 0.1-0.2 mJy/beam, comparable to the expected thermal noise. The dynamic range, defined as the ratio of peak brightness to 1 $\sigma$, is $\sim$700 on average, ranging from 300 to a few thousands. The typical resolution is $\sim$ $4 \times 8$mas. We must remark that the LAS determined on the VLBA images (Table 1, Col. 9) does not take into account the eventual presence of components completely resolved out by the VLBA observations. We note also that it may exceed the VLA rough estimate due to the source structure (e.g. a double edge-brightened source will appear smaller when fitted by a single Gaussian with a resolution smaller than the separation between components).

The total flux density "seen'' by VLBA (Table 1, Col. 8) has been measured by integration over the source images using the AIPS task TVSTAT. The brightest sub-structures of each source were fitted by a Gaussian model (task JMFIT) which allows to obtain total flux density, the beam-deconvolved Half Maximum Widths (HMW) and major axis position angle (PA). Sub-components are referred to as North (N), South (S), East (E), West (W) and Central (Ce). When a component is split into more pieces, a digit (1, 2, etc.) is added (e.g. N1, N2). Other symbols used are self-explanatory. The source images are shown in Fig. 3. The labels are marked on the images, next to each sub-component. The parameters of the brightest sub-components are given in Table 2, which has the following layout:

Columns 1, 2: Source name and sub-component label; an "$\ast$'' indicates that the component is extended and its flux density is obtained by TVSTAT (see text);

Column 3: VLBA flux density;

Column 4: deconvolved angular sizes of major and minor axis of a Gaussian component and position angle of the major axis as estimated using the AIPS Gaussian-fitting task JMFIT.

For more extended features, marked by an "$\ast$'' in Table 2, flux densities are evaluated by means of TVSTAT. For extended structures underlying more compact ones, the flux density is obtained as difference between the integrated flux density and the sum of the sub-component Gaussian fit flux densities.

We have checked the integrated flux densities in our images with the low resolution flux densities obtained by interpolating to our observing frequency the values from the NVSS (Condon et al. 1998) and from the VLA data at 5 GHz from Paper I. The comparison is displayed in Fig. 2.

There is clearly a systematic flux density difference, typically $\mathrel{>\kern-1.0em\lower0.9ex\hbox{$\sim$ }}$5%, which could be partially ascribed to calibration uncertainties. For the more extended objects there are larger flux density losses, clearly due to the absence of short uv-spacings (see, as an example, Fig. 4). The two sources with the largest deviations in Fig. 2 (0800+472 and 0809+404) do possess indeed another component detected by the VLA but completely resolved out by the VLBA baselines (see also next section).

Figure 3: VLBA images: the first contour (f.c) is generally three times the rms noise level on the image; contour levels increase by a factor of 2; the restoring beam is shown in the bottom left corner in each image.

Figure 3: continued.

Figure 3: continued.

Figure 3: continued.

Figure 3: continued.