Because CAT observes relatively small patches of sky with modest resolution, sample variance will be important. With that in mind, CAT has been observing several fields, all chosen to be relatively free from strong radio sources at frequencies up to 5 GHz (Condon et al. (1989)), and lie at high Galactic latitude (). The second of these fields, called `CAT2', is centred at (B1950). To date, CAT2 has been observed at 13.5 GHz, 15.5 GHz and 16.5 GHz for 200-400 hours per frequency (the greatest amount of data has been collected at the two extreme frequencies) in arrays with baselines scaled with frequency. The data were analysed using standard procedures, as described by O'Sullivan et al. (1995). After excluding excessively noisy regions of visibility data by eye, images were made for each frequency with a final rms sensitivity of about 9 mJy/beam at 13.5 and 16.5 GHz. Point sources were clearly visible in these images, although most were weaker on average than those in the first `CAT1' field.
Nevertheless, a total of 27 discrete sources with were removed from the CAT2 field after RT raster scanning (simultaneous with the CAT observations). Long term monitoring with the RT showed that the brightest one was very highly variable, its flux density fluctuating by as much as a factor of two in a few days. This illustrates the importance of simultaneous monitoring for source subtraction. Comparison of the final source-subtracted maps with reconstructed maps of the foreground radio sources only showed no significant correlation.
After source subtraction, excess power was seen in all three maps, mostly at 16.5 GHz, consistent with the detection of CMB signal. Figure 1 shows the 16.5 GHz image of CAT2 after the sources have been subtracted--there is excess power in the centre, although a 5- negative feature dominates the actual map (note the map has not been CLEANed, i.e. instrumental response has not been taken out). We have checked that this feature is real by splitting the data in time, frequency and polarisation, and seeing it in all subsets. Also, the feature disappears in a polarisation-difference map. Furthermore, the strength of the feature in the maps at different frequencies is well described by the spectrum of the CMB. In any case, simulations of Gaussian CMB skies show clearly that it is not necessarily unusual to see a 5- feature on its own in a map with low resolution and a relatively small field of view. Again, this convinces us of that observations of several fields are essential to obtain representative average values.