C vs. CS Configuration

Figures - show the ratio of the image quality indicators for the CS(-2) to those for the C(-2) configurations (i.e., CS/C and CS-2/C-2). The mean fidelities are basically useless in this context, showing huge scatter. Close examination of the images for a few of the more discrepant simulations shows that the mean fidelities in all cases are completely dominated by a few very high points ( $F({\bf x})\sim10^5-10^6$), corresponding to a few very low pixels in the difference map. The peak median fidelity and the peak SNR are much less vulnerable to such outliers, and only these will be discussed further.

For the long syntheses, these indicators show the (CS, CS-2) images to be on average $10-20\%$ worse than the (C, C-2) images. Surprisingly, there is no consistent difference between the CS-2/C-2 and CS/C ratios - apparently, for these sources, the two configurations are equally robust to the loss of additional antennas. Given the 10% offset of the C(S)-2/C(S) ratios discussed above, the difference between C(-2) and CS(-2) for long observations is marginal.

The story is quite different for snapshots. Here both the peak median FI and the peak SNR are 20-40% worse for all the sources, and the imaging capabilities of CS also tend to degrade more quickly than those of C when two antennas are lost.

For all these observations, there is no obvious trend of image quality ratio with flux density - i.e. if C is better than CS at (effectively) infinite SNR, it is similarly better when the thermal noise dominates the off-source sidelobes. This is quite a useful result, since it implies that one need not run a whole set of simulations with a range of noise levels when comparing two array configurations.