[evlatests] UX converter tests

[Ken Sowinski]

It is time to think about how to carry out the long threatened
T303 conversion path tests.  We tried this in the Spring of last
year with much confusion.  This time we might have two antennas
available with M301 control of the T303 switches which will at 
least double the potential for confusion.  The purpose of this 
note is to urge people to tell me what I have wrong and to 
suggest solutions to the problems that really exist.  evlatests
is a broader distribution than needed, but I don't want to miss
anyone.

Assuming that we want to make an interferometric test I list below
the many things which must work considering both one and two test
antennas.  The executor was not organized to use this signal path
for the high frequency bands so getting this to work will have to
involve modifying, tricking, or bypassing the executor.  I will
assume that the test we want to make is to observe the same sky
frequency with both IF pairs, one using the direct path and the
other using the conversion path of the T303.

The last time we did this the results were far from clear.  This time
we should proceed more cautiously.  The first step should be to
measure Y factor at the downconverter input under various T303 setups.
If we don't get the same answer for the two paths at the same sky
frequency then we fix that or understand it before going further.
Only the L301s and the T303 need to be setup for this test.

The interesting frequency dependent phenomena that Rick has been 
reporting at L and C band may be another source of confusion in 
these tests.

L301-2
We need to have the executor set this to a desired value or be able
to set it ourselves and not have it mucked with by the executor.
I assert, without having considered the issue, that we can choose
an L301-2 setting to tune the conversion path to the same frequency
as the direct path consistent with the default K band settings for
the L301-1 and the L302.  If not the the observations will have to be
serial in time rather than simultaneous, ot the observe script will
be more complicated and error prone.

M301
We must verify that the M301 really controls all those switches in 
the UX converter as we expect.

T304
The big problem here is fringe rotation.  The sense of the fringe
frequency will be reversed when the conversion path is used and I know
of no provision in the software to allow for this.  If there is only
one test antenna we artificially place it at the center of the array
and punt on the issue.  If we want to use two test antennas to provide
an ELVA only baseline, the issue will have to be faced.

Transition Module
It is here that we will have to account for the additional spectral 
inversion and produce a bandpass consistent with VLA antennas to
present to the old correlator.  To do this we will have to use the
deformatter 'mode' command and be sure that the executor does not
change it out from under us.  This is needed if we expect to 
correlate against VLA antennas.  If we have two EVLA test antennas
they should produce a correlated signal even if the wrong sideband
is selected.

Delays
The delay will almost certainly change if we choose the conversion
path.  I suggest we make the observations in line mode so this is
not important and we can measure it.  If we need to make continuum
observations and the delay change is large then a non-standard K band
IF file can be created to suit the purpose.

Levels
The signal levels presented to the downonverter will likely change if
we choose the conversion path.  I argue that the best way to deal with
this is to restrict the conversion path to the AC IF pair and exercise 
the automatic level setting feature of the executor.  18C will need a 
downconverter with power detectors before this can work.