[evlatests] More analysis of August 25, 2005 test data

[Mark Claussen]

I have taken a look at the correlation between An 14 and An 16
on the data taken on Aug 25. As a reminder, the data were taken
in PA and PB mode with a bandwidth of 1.5625 MHz, giving 127
channels across each IF. For this analysis, I have restricted
my attention to C Band, since the earlier results of L band 
seemed very strange, and the X band results would be those
most affected by pointing (and now we know that there was
some material in the X band vacuum window in An 16).  One of
the goals of this analysis was to determine if the sensitivity
of the 14-16 baseline was better (or perhaps worse) than 14 or
16 individually with a VLA antenna.  Here are some things I
learned:

1.  The birdies in An 16 could still be seen in the correlation
    with An 14 for IFs A and D (remember they could also be seen 
    in correlations with VLA antennas).

2.  There is no "spurious" correlation with time in the 14-16
    baseline as we had seen a couple of months ago (and indeed,
    thought we had fixed).

3.  I think that the sensitivity on the 14-16 baseline is 
    roughly what I would expect based on the sensitivity 
    measurements of 14-VLA and 16-VLA baselines.  What follows
    is why I think so.

The analysis that I performed is somewhat of an extension to
what I did earlier.  I selected two VLA antennas, 19 and 21,
which sit next to 14 and 16 on each arm.  I made the same
flux and bandpass calibration that I did for the earlier 
analysis, but this time I looked at a time series of rms
noises over the central part of the spectral bandpass (for IF
A and D the channel range was restricted to miss the correlated
birdie).  I took a simple arithmetic mean of this time series, for 
each of the baselines 14-16, 19-21, 14-19, and 16-21. In the
table below I report the mean rms phase in degrees over the 
time series (and over the central channels).

IF     19-21      14-19      16-21      14-16

A       2.0        1.7        4.3        3.7
B       2.0        1.7        3.5        4.1
C       2.4        5.4        4.2       12.2
D       2.3        1.7        3.9        3.6

(Note:  This pretty well agrees with my earlier analysis ---
14C was pretty screwed up, but the other 3 IFs on An 14
were slightly better than VLA antennas, and An 16 ranged
from 50 - 100 % worse than a VLA antenna).

One way to sort out how these phase noises compare for a
single antenna is to assume that the noise from the two
antennas add in quadrature to get the noise on that baseline.
If we make this assumption, and further assume that An 19
and An 21 have the same phase noise, we can work out the
phase noise for 14 and 16 independently and then add those
noises back in quadrature and compare the result with the 
last column in the above table.  That's what I did, and I get

IF    14-16 (computed)

A     4.2
B     3.3
C     6.4
D     3.6

which compares pretty favorably (again, except for IF C).  
Thus it seems to me that conclusion 3 is reasonably well 
justified.

Mark