[evlatests] Deep Imaging!

[Rick Perley]

    I finally figured out how to generate super-deep images with the 6cm 
data. 

    Background:

    The 6-hour run is quite superb.  There are 5 sources, observed with 
12 antennas, 4 sub-bands, and full polarization.  After the obvious 
flagging was done, calibration proceeded thusly:

    1) A single delay and phase were calculated, using a short 
(2-minute) section of data.  This was applied to all sources, uniformly.

    2) SETJY was used to calculate the flux for 3C147. 

    3) BPASS was run on 0555+398, one solution for each observation.  
The solution was normalized to the center 50 channels.

    4) CALIB was used on all sources, 1 second time integration, using 
the center 50 channels, with the BP and delay solutions applied. 

    4) CLCAL was used to apply these solutions, with a 10 s average in 
phase, and 1 m smoothing in amplitude. 

    5) The X-hand delay was determined (by eye) to be 9.6 nsec (using 
antenna 28 as a reference).  This was applied with CLCOR. 

    No attempt was made to resolve out the delay steps. 

    I split out two sources for imaging experimentation:  0555+398 (5.2 
Jy point source), and 0644+392 (0.55 Jy point source).  The splitted 
databases retained the four subbands, and were done sequentially with 1, 
4, 16, 64, and 245 channels, all centered at 128.  (This was done to 
minimize the phase rocking).  Self-cal was done on each of these 
splitted databases. 

    The imaging results were notably different for the two sources:

    A) For 0555+398, the one-channel images were noise limited at 0.56 
mJy (expected 0.45, using Stokes I, and 1subband), but all deeper 
attempts showed the noise 'bottoming' out at 0.24 mJy, or a paltry DR of 
20,000. 

    B) For 0644+392, the results were much better, with noise-limited 
images for all combinations, except the widest band, where again the DR 
seems to 'bottom' out at near 20,000. 

    This encouraged use of BLCAL.  And this made all the difference:

    A) For 0555+398, the crappy pre-BLCAL image imediately cleaned up to 
reveal a number of background sources, and a noise limit (244 channels, 
Stokes I, 4 sub-bands) of 19 microJy.  The theoretical expectation is 15 
microJy.  The DR is a snappy 269,000:1. 
  
    B) For 0644+392, the BLCAL-applied image has a noise of 20 microJy, 
which is exactly what is expected based on the noise histogram.  The DR 
is 27500:1. 

    The BLCAL solutions were generated for the entire 6-hour duration -- 
one solution per baseline.  It would seem that this is all that is 
needed to reach thermal noise at this level. 

    Two baselines for 0555+398 were seen to have high residual 
amplitudes or phases (probably oscillatory).  I'll be looking into these 
now.