From bec@haystack.mit.edu  Wed Oct 30 18:03:05 2002
Return-Path: <bec@haystack.mit.edu>
Delivered-To: pet@leo.gsfc.nasa.gov
Received: from planck.haystack.edu (planck.haystack.edu [192.52.61.1])
	by leo.gsfc.nasa.gov (Postfix) with ESMTP id ADE2361C7
	for <pet@leo.gsfc.nasa.gov>; Wed, 30 Oct 2002 18:03:04 -0500 (EST)
Received: (from bec@localhost)
	by planck.haystack.edu (8.9.3 (PHNE_22672)/8.9.3) id SAA03923;
	Wed, 30 Oct 2002 18:03:03 -0500 (EST)
From: Brian Corey <bec@haystack.mit.edu>
Message-Id: <200210302303.SAA03923@planck.haystack.edu>
Subject: zero fringe rate
To: pet@leo.gsfc.nasa.gov (Leonid Petrov)
Date: Wed, 30 Oct 2002 18:03:03 EST
Cc: mtitus@haystack.mit.edu (Mike Titus),
	rcappallo@haystack.mit.edu (Roger Cappallo),
	clonsdale@haystack.mit.edu (Colin Lonsdale),
	aniell@haystack.mit.edu (Arthur Niell)
X-Mailer: Elm [revision: 212.4]
Status: R

Leonid,

Mike mentioned to me your discussion about how to deal with zero
fringe rate cases.  In the long run, it seems to me that dealing
with it in sked, by avoiding scans with zero rate, is best.  In
the short term we can refringe affected scans with narrow SBD,
MBD, and/or rate windows; it doesn't always work, but often it
does.  (I just happened to have tried it on a grv01c scan this
morning: 241-0854 Matera/NyAlesund X-band.  Will clean that one
up, as well as any others that we can, before we export grv01c.)

Mike and I talked briefly about how fourfit might be modified to
detect and/or fix zero rate cases automatically.  Certainly one can
predict, from the model delay rate and acceleration, whether the
rate will go through, or pass sufficiently near, zero during a scan
to cause a problem, so one might use that information to flag a
scan as suspect.  But if the scan is long, or the source strong,
or the phase cal weak, the peak found by fourfit may be due to the
source and not the phase cal, so the occurrence of zero rate doesn't
necessarily mean that the fringe results are corrupted.

Better, I think, is to flag scans on the basis of something rotten
in the data rather than just a potential problem.  Whether that can
be done reliably, on the basis of the presence of peaks spaced 1 us
apart in the DRF or 1 MHz apart in the spectrum, say, is an open
question.

Fixing zero rate cases automatically in fourfit is also not so easy.
Perhaps the phase cal signals, as measured in the PCM modules of the
Mk4 correlator, could in effect be subtracted from the coherent
xcorr sums before the delay/rate search is done.  The machinery to
do such a subtraction is not in place, however.  Roger or Colin might
have other (and better) ideas.  But for now, I think we'll just have
to deal with it by identifying suspect points in plots of delay and
rate vs. time and then refringing them by hand.

Regards,
 Brian