[daip] Smoothing in POSSM.

[Eric Greisen]

R. Craig Walker wrote:
> When I try using POSSM on a wideband VLBA data set (8 IF, 32 MHz each full
> pol with 64 spectral channels per IF), if I use SMOOTH and restrict the
> channel range (9-56), I get droopy ends to the phases.  The ends droop for
> several channels on each end of every IF.  All go in the same direction on
> a given baseline, but can go the other direction on another baseline.  If
> I open up to all channels, I don't see the effect.  If I don't smooth, I
> don't see any justification in the unsmoothed data for such droops.  It
> happens for both gaussian and box smoothing and for both before and after
> bandpass calibration.  In fact, the effect shows up when bandpass is not
> applied.
> 
> See the attached plots.  The first shows the effect in a smoothed, channel
> restricted case.  The second shows the full band smoothed and the third
> shows the full band unsmoothed.
> 
> Below are my possm inputs followed by an image header.  The reference
> antenna in all tasks that need one is LA.
> 
> The droops seem to be some artifact of the smoothing.  Is there a reason
> that this behavior is expected, or is there some issue with POSSM?

When I stare at the unsmoothed data I do not see anything that is 
obviously wrong in the smoothed data.  Wherever the smoothed droops 
there are a fair number of unsmoothed points in the direction of the 
droop and the smoothed data occupy a smaller range of phases than the 
unsmoothed.

Have you tried the special values of SMOOTH available in POSSM alone?

I worked over this code recently and did a few little things to fix it 
up and then added the post-summation smoothing in POSSM and checked it 
all fairly carefully.

I agree that it looks curious but I am not fully convinced.

Eric