[evlatests] VLBI at the EVLA test report

[Gregory Taylor]

VLBI with the EVLA Testing Report

Goals:

To test the capability of the EVLA for participation in standard VLBI
experiments both in single dish (Y1) and in phased array (Y27) modes.

Methods:

One hour of test time with the VLA and VLBA was scheduled for 12:54pm
MDT on June 18, 2007.  This time was divided into a 30 minute test
with the phased VLA, and 30 minutes in a single dish mode.  The two
experiments were designated TY003A and TY003B.  In both experiments
observations alternated between 3C84 and a nearby point source,
J0310+3,8 every 2 minutes.  The T8 (VLBI cross-bar) switch was set
manually since it was known that the executor did not have the
capability to automatically select the state of the switch.  VLA
antenna 10 was designated for the first 20 minutes of the single-dish
mode test.  EVLA antenna 13 was also cabled up to the T8 switch to
test the use of EVLA antennas for single dish mode (final 10 minutes).

Results:

Phased Array:  The VLA antennas did not phase up.
The EVLA antennas phased up reasonably well -
antennas 19 and 21 failed to operate at all, 8 others
phased up with typical offsets of 15 degrees or less.

A greater problem was no power to the BBCs.  This was noted
during the tests by examination of the T8 (VLBI cross-bar)
switch.  A square-law detector at the switch reported no
power, and this was confirmed by calling the VLBA operator
and finding that there was no power at the BBCs.

The failure appears to have been caused by delay packets heading to the
correlator which did not have the "include in analog sum" bits set.
Therefore the analog sum buffers did not operate.
A fix for this problem (according to Walter) will require two parts:
(1) the executor needs to implement the function calls that
allow these bits to be set; (2). obs2script needs to call this function.

Ken Sowinski comments:
"Some of the infrastructure is present.  At least
an equal amount is not yet.  At delay packet generation time we
probably want something like:

if(subarray.getAnalogEnable() & !antenna.isTaken() &  
antenna.getAddtoSum())
      "add in the bits for this antenna to the delay packet"

Not all those methods exist yet.  If they did, the script would have
to add subarray.setanalogEnable(bool) which might also, in turn, call
antenna.setAddToSum(bool)."

Single Dish Mode: The executor aborted after 10-20 seconds.
Several attempts were made, including some quick patches
by Walter Brisken, but nothing worked.  Walter has diagnosed
the problems as follows:

"Both failures were caused by problems in the sdvlbi.py script that I
had modified from Barry's example.  The second (more annoying one) is
thought to be caused by a comment line not starting at column 0.  This
is allowed in real python!"

Recommendations:

(R1) The executor software needs to be corrected as described
above.  This includes the ability to phase up the VLA antennas
and the ability to select the proper state of the VLBI cross-bar
switch.

(R2) Once R1 has been carried out a new test of the same basic
design should be scheduled.

(R3) Until R2 has been carried out and demonstrated to function
properly no VLBI at the VLA experiments should be scheduled
without the modcomps.

(R4) The operators did not know how to select the reference antenna
(in modcomp-free operation).  It turns out that the default
is to 24, or if 24 is not available, then 16.  This is tolerable,
but not optimal since 24 is on pad W48.  The operators should be
able to select the reference antenna.

(R5) There is no point correlating TY003A and B, so the disk
space for those can be recycled.

Acknowledgements:  Thanks to Walter Brisken and Ken Sowinski
for diagnosing the software issues.