[daip] forward message w/o files

[Eric Greisen]

From: "Sandor Frey" <frey at jive.nl>
To: <egreisen at nrao.edu>
Cc: <lgurvits at astron.nl>, <paragi at jive.nl>, <frey at sgo.fomi.hu>
Subject: freqency averaging problem in AIPS
Date: Wed, 23 Jun 2004 20:53:22 +0200

Dear Eric,

I encountered the following problem while reducing EVN VLBI data at JIVE
using AIPS (31DEC03 version).

It is a C-band phase-referencing experiment with a compact reference
source J0836+0052 (~200 mJy). The data reduction goes smoothly until the
final fringe-fit solutions are applied. After SPLITting (averaging over
frequency), the phases on most baselines show large deviations that are
more or less constant over the 7.5-long observing time. Amplitudes are
also changing. The problem persists if I use AVSPC instead for
averaging.

I went on keeping all 8 IFs and 16 channels in SPLIT, in a hope that
IMAGR can correctly handle them. But maybe IMAGR uses the same algorithm
for spectral averaging, since the data are also degraded. (In fact I
tried imaging the calibrator using a single channel and it worked
nicely.)

I copied a short 70-sec scan on the cal source. The following files are
attached:

VPLOT-BEFORE.PS - amplitudes and phases on baselines to EF. Phases are
around 0, amplitudes are ~200 mJy. The fringe-fit solutions (CL#4) are
applied. VPLOT averages the channels and IFs for plotting - apparently
correctly. (The fringe-fit solutions look fine in POSSM as well.)

VPLOT-AFTER.PS - the same plot after AVSPC. Phases and amplitudes are
offset. The effect is the same if I use SPLIT for spectral averaging.

EF011.UVCOP - a portion of the uv-data

EF011.AVSPC - the same data, averaged over the entire freq range

I've never met such a problem before. I tried 31DEC02 version as well,
with the same result. I still hope I myself did something wrong, and not
AIPS... Please let me know if you need more information or data for
possible debugging. Thank you for your help in advance.

Best regards,
Sandor