1 Introduction

By definition, Gigahertz Peaked Spectrum (GPS) sources have convex spectra with a turnover frequency $\nu_{\rm max}\gt 1$ GHz. It is generally accepted that such a spectral shape results from synchrotron self-absorption due to a high compactness. Indeed, the linear sizes of GPS sources are very small (10-1000 pc) and their radio luminosities are large ($L_{\rm radio}\sim10^{45}$erg/s). An important feature which makes the GPS class particularly interesting, is that GPS objects identified with quasars often have very large redshifts. A review on the properties of GPS sources has been given by O'Dea et al. (1991) -- hereinafter OBS91 -- and O'Dea (1998). De Vries et al. (1997) compiled spectra of 72 GPS sources and constructed a canonical GPS radio spectrum. They found it to have a constant shape independent of AGN type, redshift or radio luminosity.

As Gopal-Krishna & Spoelstra (1993) pointed out, the great potential of GPS sources for discovering high-z objects continues to be the major motivation factor for enlarging the sample of GPS sources. A second motivation for increasing the number of known GPS sources comes from the discovery (Wilkinson et al. 1994) of a new class of Compact Symmetric Objects (CSO). Five archetypal CSOs (cf. Readhead et al. 1996a) are all acknowledged GPS sources. Readhead et al. (1996b) argue that CSOs are the young precursors of classical double radio sources. Increasing the list of known GPS sources offers a promising way to find more CSOs.

The search for GPS sources can be difficult because it requires both high resolution and high sensitivity observations at frequencies well below 1 GHz and high sensitivity observations at $\nu \gt\,$ 5 GHz, well above the potential peak of the spectrum. Therefore, unlike in the case of flat spectrum sources, only a modest number of GPS sources can simply be extracted from classical large catalogues e.g. the Green Bank (GB) surveys at 1.4 GHz and 4.85 GHz (White & Becker 1992 and references therein -- hereinafter WB92) and the Texas Survey at 365 MHz (Douglas et al. 1996). For example, the sample of Stanghellini et al. (1990) was derived from the Texas Survey and the 1 Jy catalogue of Kühr et al. (1981a). Out of 55 sources and candidates they listed, 41 are currently acknowledged as GPS sources (O'Dea, priv. comm. 1996) and 33 make a complete sample (Stanghellini et al. 1996).

The lists published by Gopal-Krishna et al. (1983) and Spoelstra et al. (1985) resulting from dedicated observations, e.g. with the Westerbork Synthesis Radio Telescope (WSRT) or the Ooty telescope, are not very large: each one contains 25 sources (5 and 2 out of these two samples respectively had been retracted later). The sources gathered by the authors mentioned above have been collected in OBS91 in the so called "working sample'' encompassing 95 objects. Gopal-Krishna & Spoelstra (1993) confirmed the existence of 10 GPS sources and Cersosimo et al. (1994) found 7 more. On the other hand 6 sources from the list in OBS91, namely 0218+357, 0528+134, 0902+490, 1851+488, 2053-201 and 2230+114 proved to be "not so good'' examples of the class and have been retracted. By the end of 1997 very few other GPS sources had been discovered.

There are two ways to achieve a "bulk'' increase of this number. The first one has been followed by Snellen et al. (1998) -- hereinafter SSB98 -- the second is the basis of this paper. The former one is based on the Westerbork Northern Sky Survey (WENSS) being carried out with WSRT at 325 and 609 MHz. Naturally, the sources with inverted spectra in the WENSS are GPS candidates, which are followed up with observations at higher frequencies. This survey is particularly useful since it is the most sensitive survey at low frequencies (more than an order of magnitude better than the Texas Survey).

The published part of WENSS called "mini-survey'' (Rengelink et al. 1997) is limited to $14^{\rm h} 10^{\rm m} < \alpha < 20^{\rm h} 30^{\rm m}$, $57\hbox{$^\circ$}< \delta < 72\hbox{$^\circ$}50\hbox{$^\prime$}$ and covers $\sim$570 square degrees only, so there are still large parts of the sky where this method cannot be applied. Nevertheless, SSB98 using the mini-survey plus an unpublished part of WENSS ($4^{\rm h} < \alpha < 8^{\rm h} 30^{\rm m}$, the same declination range as the mini-survey) were able to establish a list of 47 "new'' GPS sources.