3 The observations

The flux densities obtained in the different observing sessions are presented in Tables 4 and 5. Selected data from the literature have been added (with the relevant references) when considered compatible with our observations and useful to define the spectral shape.

During the VLA observing sessions an additional 20 GPS candidate sources were observed (Table 2). These sources fail one or more criteria of the complete sample. Nevertheless, the majority of them show a clear GPS-like spectrum.

3.1 VLA observations

The VLA data were obtained in a number of observing sessions in the A configuration spanning several years. Eighteen candidate GPS radio sources were observed on 30 December 1984. The results from this first set of observations were already published by O'Dea et al. (1990), Stanghellini et al. (1990), and O'Dea et al. (1991). We consider here only the observed sources belonging to the complete sample.

A second set of VLA observations was obtained on 9 February 1990, and a third session on 8 June 1991. The journal of these observations in shown in Table 3. In the third session, where the majority of the sources belonging to the complete sample were observed, we split the 2 frequencies in each band to obtain the maximum possible separation with adequate sensitivity. This was used in the previous observing sessions only in the L band, here we used it also in the C, X and P bands. This technique proved to be very useful, because it permitted a better definition of the spectral shape and a more accurate selection of the sources peaking close to the limit of 0.4 GHz that we adopted for our sample.

Furthermore, the observations separated in frequency permit us to solve the $n\pi$ ambiguity that occurs in the determination of the Faraday rotation measure.

  

Table 2: Additional objects: Cols. 1 to 10, source name, redshift, optical identification, magnitude and filter/color, flux density at 5 GHz, observed and rest frame peak frequency, references for the optical information, references for the optical information as Table 1. Last column is a note indicating the radio source with extended emission

  

Table 3: Journal of the second and third VLA observing sessions. Only a few sources were observed at 22 GHz

The data reduction has been performed in a uniform way for all the VLA data sets. The errors in the flux densities are dominated by the calibration errors which are estimated to be around 3$\%$ at L, C, X bands, around 5$\%$ at P and K bands, respectively.

At VLA resolution almost all the objects observed are dominated by a single point-like component. The side lobes have been deconvolved from the images using the AIPS implementation of the Clark CLEAN algorithm (Clark 1980; Cornwell & Braun 1988). The data have been self-calibrated in phases (Schwab 1980; Cornwell & Fomalont 1988) with an initial point model, and when subsequent iterations of self-cal have been considered necessary, we used an appropriate number of non negative components obtained from the images in an interactive process until convergence to an acceptable solution was achieved. In most cases an amplitude self-calibration has been performed on the data to improve the final images and to allow a search for possible extended emission.

3.2 The WSRT data

Westerbork Synthesis Radio Telescope (WSRT) filler observations on most of the objects of the complete sample have been obtained during 1990 at 327 MHz and during 1991 at 608 MHz and 327 MHz. Each source has been observed in several snapshots a few tens of minutes long.

The data reduction for the data obtained in the 1990 has been done with the package DWARF. The data of the 1991 observations have been reduced with AIPS. These latter data in a format readable by AIPS did not have the redundant baselines, therefore the data were not self-calibrated, producing a lower dynamic range compared with the capabilities of this instrument.

3.3 Other data

A few objects were observed during the spring and the summer of 1991 with the Northern Cross, a radio telescope near Bologna operating at 408 MHz on the principle of the Mill's cross (Braccesi et al. 1969 and Ficarra et al. 1985). The radio telescope dates back to the sixties, however, recent mechanical and electronics upgrades allow completely automatic observations with increased sensitivity. The flux density values have been set to the scale of Baars (Baars et al. 1977) adopting a flux density of 37.7 Jy for the calibrator source 3C 380 (Riley 1988).

A few additional objects have been observed with the Russian radio telescope RATAN 600 (Esepkina et al. 1979), in transit mode, during various sessions in 1993 and 1994, at 11.2, 7.7, 3.9, 2.3 and 0.96 GHz (Mingaliev, private communication).