[evlatests] Antsol oddities when spectra are bad.

[Vivek Dhawan]

These data (that Michael referred to below) were all fine.
The data which _did_ have problems are: X_osro_003.55497.768572534726

In this data, antsol/telcal produces no solution (not just weak or failed
solution, but no output at all) for half the scans on all the antennas
on IF C only. The rest of the IFS were better than 95%.

Loading the same data into AIPS, and doing a scan-average solution to get
close to what antsol is doing, showed that all the data were present and
all solutions were found without any trouble. So there is no missing data.

However, this is false comfort: Antenna 5 had weak amplitudes on IFs
B,C,D by a factor 2 (IFA was OK), and the real story is in the spectra.
The DTS modules (it could also be the station boards i guess) on IFs BCD
were screwy, and IFA had a large delay but was otherwise OK. Later in the
day ant5 was removed from service to be fixed. With 5 out of the array, a
re-run of X_osro showed near-perfect solutions from antsol.

So, a feature of antsol is that corrupted data for one bad antenna/IF can
mangle (or disappear) the solution for all antennas for a given solution
interval for that ant/IF. It produced weak solutions at other times
for the same antenna/IF, though the data were just as corrupted. It also
found solutions for amplitude, phase and delay for 95% of the time on the
other 2 corrupted IFs on the same antenna, which were not as bad as IF C.

Except for the first feature, AIPS (calib) processing shows the same, but
it better isolates the failures to the antenna/IF where they are present.


mrupen at aoc.nrao.edu wrote:
> Hi Vivek --
> 
>   would you mind doing one further check on the C_oddint stuff? actually
> the C_osro just after C_oddint would be the best.  We would like to move
> to the new CBE which allows "oddint", but would like first to be sure
> that the TelCal problems found at the same time are due to TelCal rather
> than to bad data.  So the basic test would be to load the data, apply
> on-line flags, then do a scan-based self-cal, preferably with refant 24.
> Our hope is that the self-cal works & gives reasonable results for all
> antennas, subbands (IFs), and pol'ns; TelCal was often failing to find
> *any* solutions for one or more of the {subband, pol'n} pairs.
> 
>   Thanks very much --
> 
>       Michael
>