[evlatests] A-B and C-D delay differences

[Michael Rupen]

We ran a quick test this morning to check for time offsets between
IFs A and B, and C and D.  For once, the result made sense!  In brief,
there is a delay difference of 40-50nsec between A and B, and between
C and D.  This difference is roughly the same for a given pol'n for all
antennas.  This is basically what we expected to see, and the ~10 nsec
offset between AB and CD implies that we've got the RL delay offset roughly
right as well.

Details for those who care:

- The original rationale was to check whether a time offset between the
   two basebands might explain the surprisingly high noise levels seen
   when forming the differences AA-BB and CC-DD, when the basebands
   (and subbands) are tuned to the same sky frequency.  Upon
   reflection this seems an implausible mechanism, given that we are
   differencing _correlations_ and averaging over a second, but we went
   ahead and did the test anyhow.

- To do the test we lied in the VCI file, claiming that IF B was LCP,
   and IF D was RCP (see the aptly names "terrify.vci" in the operations
   script area).

- After finding that PHASEMOD was not being received for IF BD we
   disabled the use of phase models on the Baseline Boards, which then
   gave nice AB and CD fringes.  This PHASEMOD difficulty is likely because we
   forgot that the generation of phase models for A and B were both set to
   "late", while C and D were both set to "early"; so we were clobbering one
   set of PHASEMODs before sending them along to the BlBs.

Cheers,

     Michael & Ken