Our observing sample has two main components. Firstly, we selected the
47 radio sources in the Parkes PKSCAT90 Catalog (Wright & Otrupcek
1990) which met the following criteria: (1) south of declination
; (2) possessing both 5 & 8GHz flux density measurements
in PKSCAT90; (3) accurate positions (i.e. quoted errors <2
);
and (4) expected 3mm flux density greater than 300mJy (based on a
spectral index derived from 5 & 8GHz fluxes).
The compact (milliarcsecond-scale)
structures of most of these sources are not known.
Secondly, we selected
all sources south of declination from the US Naval
Observatory (USNO) Radio-Optical Reference Frame (Johnston et
al. 1995). These are compact flat-spectrum radio sources with
highly-accurate positions (typically better than 3 milliarcseconds), and
typically have 8.4 GHz flux densities greater than 
.
Of the 60 sources identified this way, 28 were already selected as part
of the Parkes sample, leaving 32 new sources. A declination cut-off of
was used to select regions of the sky inaccessible to
northern mm telescopes. We note that the spectral indices based on
non-contemporaneous PKSCAT90 5 & 8GHz flux densities may be
significantly in error. We also note that it is possible that a
population of radio sources with curved spectra, i.e., which flatten in
spectral index or are inverted above 8GHz, could be missed by our
criteria.