[evlatests] Delays; Archive data truncation.

[Vivek Dhawan]

1. Delays corrections (the non-closing amplitude correction) have
    been turned off in AIPS, per Eric, and things look better:

The data I complained about last week has large delays, and gave only
~1500:1 dynamic range. It was re-processed with the latest code and
now gives >30,000:1 using only antenna-based correction. (This was
XWICloopCtr.55233.0625  taken before the phneg was done in the CBE,
and may have transposed subbands - I did not pursue it further. Better
data are already available.)

Still on the subject of delays, Ken indicated that delays changed when
the CM reset the correlator. With significant delays, and expectation
that spectral averaging will be used in the CBE and/or the archive
outlet, we need the equivalent of the VLBA CQ table saying how much
averaging was done. Those pesky delay corrections have to come back
again, and be right.


2. I tried to get a large dataset out of the archive, it was truncated:
    the UVfits file was not an integer multiple of 2880 blocks.

The archive gurus are on it - this note is just to alert others.

The file is 35GB ms, expect >50GB uvfits, but truncated at ~46GB.
sysptx.55239.077535833334

The symptom was that SDM to MS went OK, exportuvfits finished (seeming OK
I'm not sure) but FITLD died upon reading the truncated file and deleted
its partial output.

Vivek.