[evlatests] More results from X-band Saturday observation

[Rick Perley]

    It just gets better and better ...

    I more carefully reviewed the X-band data from Saturday.    One hour 
on our 4.3 Jy point source at declination 74 degrees. 

1) Other than the initial slew time to get on source, virtually no data 
had to be removed.  There are occasional 1-second long amplitude dips on 
all antennas -- comprising less than 0.3% of the total data.  The 'dip' 
is perhaps by a factor of two.  These were easily identified and removed. 

2) The residual delays (steady throughout the hour-long run, and 
differing by less than 1 nsec between the two IFs are:   Ant 1   -28.5nsec
                      2    reference
                      3  -9.8 nsec
                      9  +1.2
                     18  -11.3
                     19  -8.0
                     23   -2.4
                     24   -4.2
                     25   -10.2
                    28   -9.6

    Except for antenna 1, these are all very closely the same as the 
C-band delays that I found two weeks ago, except that the sign is 
reversed.  Presumably this is because I had to use FLOPM to reverse the 
spectral order so as to get FRING to work. 

3)  As reported yesterday, baseline 2 x 18 has the peculiar phase 
oscillation on both IFs.  No other baseline shows any hint, whatsover, 
of this problem.

4) I found that on IF#2 *ONLY*, baseline 19 x 28 had no correlated 
signal -- only noise. 

5) The bandpass stability is utterly superb.  I ran 5-minute BPASS 
solutions (i.e., coherent average over 5 minutes, properly normalized by 
a central group of channels), and plotted the deviations of these 
solutions about the mean solution (this is nicely done by BPLOT). 
    a) The phase residuals show a *perfectly linear* slope, pesumably 
due to residual delay, corresponding to a maximum delay residual of 65 
ps.  (The total phase difference from one end of a sub-band to the other 
is only 3 degrees).  There is no sign of any wobbles, even to the edge 
channels.
    b) The amplitude residuals, with the single exception of antenna 23, 
show no structure on any frequency scale, to a level of less than 0.1% 
-- that's the noise level. 

6) The raw numbers we are getting are supposed to be proportional to 
correlation coefficient, but they are not.  The range in amplitude 
corrections needed to get the data to a proper visibility scale ranges 
over a factor of more than 2:1 (i.e., a range of 4 in power), but the 
histogram width spreads (seen by plotting various individual baselines) 
never deviates by more than a factor of 1.2.  We know from VLA 
correlator studies (which we know to give us proper correlation 
coefficients) that the distribution of antenna sensitivities at X-band 
is very tight.  Hence, it is clear that the WIDAR amplitudes are not 
proportional to correlation coefficients.  (Ken says he knows why this 
is ...)

7) I ran BLCAL to see the range in closure corrections.  As noted 
yesterday, the typical values are from 0.01 to 0.001%.  I don't believe 
these values are significant, and that so far as we can measure, there 
are no closure corrections at all.  The correlator is perfect. 

8) I ran a two-IF image (2048 x 2048, with 1 arcsecond cellsize), using 
a 916 channel average.  The deconvolved image shows nothing but uniform 
noise, at a level of 40 microJy -- the formal DR is 108,000:1.  This 
noise level is *exactly* as predicted from the histogram widths done on 
the 125 kHz-wide data, reduced by the number of channels and number of 
visibilities. 

9) I averaged the data to a ten-second average, to see if there are any 
low-level wobbles in the time-sequence.  Nothing but noise about the 
mean visibility was seen.