VTESS vs. CLEAN

CLEAN was used for most of these simulations because (1) most observers in practice use CLEAN for relatively small sources; (2) the ``truth images'' were made with VTESS, and using VTESS in the simulations might have been able to model those VTESS'd models unusually well; (3) VTESS would probably make better use of (or rely more heavily on) the short spacings provided by CS, so using CLEAN seemed the conservative choice when checking whether C was ever markedly superior to CS. However, if VTESS makes much better images than CLEAN, that's presumably what the observer would use; and even if it does not, conclusions based purely on CLEAN would be worrisome if VTESS suggested very different answers.

To check these points I ran one suite of simulations for each source using VTESS rather than CLEAN. In each case I used a flat default model with a roughly correct initial total flux density, and deconvolved the same region used for the CLEAN. VTESS uses a combination of the gradient and the noise to decide whether it has converged; for Cas A using the $\sigma$ from the CLEAN image gave convergence within 400 iterations, while for Cyg A I was forced to increase this by a factor 5-15, as listed in the tables, to ensure convergence with fewer than 500 iterations. VTESS should be more appropriate than UTESS for these data, unlike those discussed in Rupen (1997), because here the SNR is high enough for positivity to be a reasonable and helpful constraint.

Both visual inspection and the measures of image quality given in the tables show that VTESS gives images comparable to CLEAN for Cas A, but does a much worse job than CLEAN for Cyg A. VTESS tends to do a slightly better job for CS than for C configuration (cf. Figures -), and in some cases implies that CS is actually better than C, contrary to the CLEAN results presented below. However, this is not true in all cases, and VTESS' poor deconvolution of sources like Cyg A implies that CLEAN will certainly be used for those objects. VTESS may rescue CS imaging in some cases, but not consistently, and not for all of the sources which will actually be observed.