[evlatests] Continuum subtracting EVLA data with aliasing

[Barry Clark]

Special procedures are needed for the EVLA-EVLA baselines.  That below
should work.

1.  Flag (temporally) all EVLA-EVLA baselines.  Calibrate bandpass
on all antennnas.  Unflag the EVLA-EVLA baselines.

2.  For the EVLA-EVLA baselines, for a bandpass calibrator, generate a
"Continuum subtracted" set of visibilities, by subtracting a level (a
single complex amplitude) determined at frequencies unaffected by the
aliasing.  It is my understanding that uvlsf will do this.

3.  Resolve this dataset onto the antennas.  I believe a standard
bandpass solution will do this for the frequencies of interest.  But this
solution will need to be stashed somewhere so it doesn't get in the
way of other calibration applications.

4.  Correct this solution for the w (of (u,v,w)) at which the observation
was made.  The result is the aliasing function for the IF in question,
and should have very high stability - it can be used for all frequencies,
within a band, for all days of observation (as long as the antenna
has not been moved).  (There will be small delay errors if the formatter
gets resynced, but these should not be large enough to affect this
bandpass.  There may be differences from band to band, because of
differing cable lengths, but even these are expected to be fairly
small.)  There is, I think, no current way to do this in AIPS.

5.  A new version of uvlsf will be needed to fit for the continuum as
aliased, that is for 1 + BP * exp(4*i*pi*deltaf*w).  This should accomplish
continuum subtraction about as well as uvlsf works for VLA antennas now.  
The increased noise, due to the aliased noise, of course remains.

6.  If you are observing at a frequency where VLA antennas are not
available, bandpass calibration will be a problem.  This can be addressed
by creating the alias solution at some other frequency, and from this,
applying a correction of (1+exp(4*i*pi*deltaf*w)) to the data and 
then proceeding as usual.  There is currently no software to do this.