OBSERVATIONS

We obtained quasi-simultaneous B, V, R, I, J, H and K_n photometry of a subset of the Parkes sample. Observations were taken during 26 nights in 1997 (Table 1) at Siding Spring Observatory. Optical images were obtained with either the 1 m telescope, or with the imager on the 2.3 m telescope. Near-IR images were obtained with the CASPIR 256 x 256 InSb array camera (McGregor et al. 1994) on the 2.3 m telescope. 157 Parkes sources were observed in some or all of the bands, as well as a small control sample of 12 optically selected QSOs randomly selected from the Large Bright QSO survey (LBQS, Morris et al. 1991); an optical QSO survey well matched in size and redshift distribution to the Parkes sample. To minimise the effects of variability, all the observations of an individual source were made within, at most, a six day period (Table 2). Flat spectrum quasars typically vary by 10% or less on these timescales, though very occasional greater variations are seen, typically in BL Lac objects (eg. Wagner et al. 1990, Heidt & Wagner 1996). Only data taken in photometric conditions were used: seeing was typically 1-2''.

Bright objects were typically observed for five minutes in each band. Fainter objects were observed for up to two hours in our most sensitive bands (R, I and H). If they were seen in these bands, we observed them in progressively bluer bands as time allowed. Four sources were not detected in any band: PKS 1535+004, PKS 1601-222, PKS 1649-062 and PKS 2047+098.

About five standard stars, spanning a range of colours, were observed each night: in the optical, the Graham E regions (Graham 1982) were used, while in the near-IR, photometric calibration was obtained using the IRIS standard stars, which have magnitudes on the Carter SAAO system (Carter & Meadows 1995). Within individual nights, the scatter in photometric zero points (without using colour corrections) was < 3% rms, so all the standards in a given band were simply averaged to give the final calibration.

All 98 Parkes sources lying in the RA. ranges 00:36 - 00:57, 01:53 - 02:40 and 14:50 - 22:52 (B1950) were observed in both the optical and the IR: these should thus form an unbiassed, complete sub-sample of the whole Parkes Half-Jansky sample. The remaining 59 sources were selected for observation mainly on the basis of prevailing weather conditions, and so should also form a reasonably unbiassed sub-sample. No selection was made against radio galaxies: sources with resolved optical or near-IR images (as classified by the COSMOS plate measuring machine from UK Schmidt plates, and checked by visual inspection of our images) are listed in Table 2. Where appropriate, they are excluded from the following analysis.

Optical images were bias- and overscan-subtracted, and then flat fielded using twilight sky flats. For the fainter sources, multiply dithered 300- or 600-second exposures were taken: these were combined using inverse variance weighting. The infrared exposures were made up of multiple dithered 60 second images, each made up of two averaged 30 sec exposures in J, six averaged 10 sec exposures in H and twelve averaged 5 sec exposures in K_n. These were bias- and dark-subtracted, and then corrected for the non-linearity of the CASPIR detector using a simple quadratic correction term (derived from plots of median counts against exposure time obtained from dome flats). Known bad pixels were replaced by the interpolated flux from neighbouring pixels. Flat fields were obtained by taking exposures of the dome with lamps on and off, and subtracting one from the other: this removes the contribution from telescope emission, and substantially improves the photometric accuracy attainable. Individual images were sky subtracted, using a median of the 10 images taken nearest in time. The dithered images were then aligned and combined, using the median to remove residual errors.

The radio sources were identified from the radio positions by using astrometry from nearby stars, bootstrapped from positions in the COSMOS/UKST and APM/POSS sky catalogues, maintained on-line at the Anglo-Australian Observatory. Magnitudes were then measured using circular apertures, with the sky level determined from the median flux in an annulus around the sky aperture. For unresolved sources, the photometric apertures were set by the seeing: typical aperture radii were ''. For resolved sources (mostly low redshift radio galaxies) larger circular apertures were used, centred on the galactic nucleus. These larger aperture radii are listed in the footnotes to Table 2. Standard stars were measured with similar aperture sizes.

Quoted errors are the sum (in quadrature) of random errors and an assumed 5% error in the photometric zero points. Random errors were determined by measuring the rms (root-mean-squared) pixel-to-pixel variation in sky regions, and scaling to the aperture size used. This will be accurate for fainter (sky or read-noise limited) sources, but will underestimate random errors for the brightest few sources. The photometric zero point errors were estimated from the scatter in zero points between different standard star measurements in an individual night: typical rms scatters are < 3%, so we adopted a conservative value of 5% as our zero point error.

For modelling and plotting purposes, we converted the magnitudes into fluxes. We assumed fluxes for zero magnitude objects as listed in Table 3. In the optical, our filter set approximate the Johnson & Cousins system, and were calibrated using the Graham standards (also approximating Johnson & Cousins). The zero magnitude star fluxes for this system were taken from Bessell Castelli & Plez (1998). In the infrared, our observations used the CASPIR filter set calibrated by the IRIS standards. Zero magnitude fluxes were calculated by P. McGregor, assuming that Vega is well represented in the near-IR by a black body of temperature 11200 K, and normalisation

(Bersanelli, Bouchet & Falomo 1991). These normalisations agree closely with those quoted for UKIRT near-IR standards (MacKenty et al. 1997). Our observations were made with the K_n filter, but were calibrated using the quoted K magnitudes of the IRIS standards without applying a colour correction term, and should thus be normalised to a K-band zero point.

Table 3: Assumed Fluxes of a Zero Magnitude Star

Filter	Mean	Flux of Zero Magnitude
	Wavelength (m)	Star (, )
B	0.440	6.32 x 10-11
V	0.550	3.64 x 10-11
R	0.700	2.18 x 10-11
I	0.880	1.13 x 10-11
J	1.239	3.11 x 10-12
H	1.649	1.15 x 10-12
K	2.132	4.10 x 10-13

© Copyright Astronomical Society of Australia 1997