C vs. CS Configuration

The main intent of CS configuration is to improve the short-spacing coverage of C configuration by moving one or two antennas from their standard C locations to stations normally occupied only in D configuration. Braun's original (1993) memo suggested, and Holdaway (1994) simulated the effects of, moving the northernmost antenna from N18 to N1. Unfortunately an antenna at N1 would prevent antennas being moved onto pads W1 and E1 (cf. Figure ). Barry Clark therefore suggested moving two antennas from intermediate stations on the east and west arms (W12 and E12) to stations E3 and W3; this gives excellent short-spacing coverage, even better than Braun's original suggestion, while allowing the switch from C to D configuration without additional antenna moves. Both Rupen's (1997) experiment and the first `official' CS configuration used this antenna distribution.

Of course this removes two stations from the normal C configuration, and the resulting uv-coverage shows gaps at intermediate baselines. The uv-coverage for C and CS configurations is compared in Figures - for a long (hour angle (HA)= $\pm4$) and a snapshot (HA=$0.0-0.1$) observation at 1.42 GHz of a source at $\delta=60^\circ$. These figures also show the coverage with two antennas missing (these are referred to hereafter as the C-2 and CS-2 configurations), to illustrate the effects of the antenna loss mentioned above. The missing antennas are E14 and W4, the first chosen to maximize the difference between C-2 and CS-2 (since E14 is adjacent to one of the antennas moved into the center in CS configuration), the second chosen at random. The C-2 and CS-2 configurations should represent what one might obtain most of the time in practice, while C and CS illustrate the best possible coverage. The corresponding beams (for a Briggs (1995) robust weighting of 0 in task IMAGR) are shown in Figures -.

The differences between the configurations are readily apparent. The additional short spacings of CS fill in the central hole rather nicely, even in the snapshots; while the missing intermediate baselines are glaringly obvious in the CS snapshots (especially CS-2), and show up as additional holes (most obvious at $u\sim\pm5\rm k\lambda$) in the long syntheses. The point-spread functions naturally reflect these differences, in the form of near-in sidelobes roughly 25 arcsec (2 beams) from the main peak which are $\sim70-80\%$ higher for CS/CS-2 than for C/C-2 configuration. For long integrations one can however adjust the robustness and taper of the CS data to give a beam which is almost identical to that given by C configuration. For example, the beam for the C-2 configuration with robustness 0 and no taper is very similar to that given by the CS-2 configuration with a robustness of $-1$ and a Gaussian taper with a $14\rm k\lambda$ full-width at half-maximum (FWHM). The cost is a corresponding increase in the thermal noise of about 12%. Of course, for maximum sensitivity one would prefer a naturally-weighted beam, which will have somewhat higher `skirts' for CS than for C configuration. Note as well that one cannot force the CS-2 `snapshot' beam to look like that of C-2 - the near-in sidelobes remain relentlessly high, for any value of the robustness, uv-taper, etc.