A Common-sense Test: Does removing antennas lower the image quality?

These quality measures are all nicely defined and sound reasonable, but do they actually mean anything? It is encouraging that snapshot observations yield much lower quality indicators than full syntheses, but that is a fairly coarse check. Presumably the image quality should also be better for the full arrays than it is when two antennas are missing. The ratios (C-2/C, CS-2/CS) of the various quality indicators are shown in Figures  and . The intensity-weighted and high-SNR mean fidelity indices show a great deal of scatter; particularly for the long observations, the C(S)-2 configuration in many cases is rated superior to the same configuration with no antennas missing. The peak median fidelity and the peak SNR show much less scatter, but again frequently imply that, for long observations, C(S)-2 is better than C(S). Given the nonlinearity of the deconvolution algorithms this is not impossible (though it certainly seems unlikely), and it is reassuring that images of the more complex source (Cas A) at least are consistently worse with fewer antennas. For the purposes of this memorandum I take the $\sim10\%$ apparent `improvement' of C(S)-2 compared to C(S) to show the level at which these measures of image quality cannot be trusted.

As a further check, the simulations labeled CS-4f and CS-6f in Tables 4a show how the image quality changes when 4 and 6 antennas are flagged. Although for this particular case CS-2 was measured to be better than CS, CS-4 and CS-6 show the expected (drastic, in the latter case) deterioration of the imaging as more and more antennas are dropped. Apparently the improvement from CS to CS-2 is a peculiarity of this particular source and the loss of those particular antennas.