[evlatests] Regarding X and Y labels for EVLA dipole feeds (<1 GHz)

[George Moellenbrock]

All-

Recently, there have been a number of ~disparate ongoing discussions about
the labeling of X and Y polarization in the <1 GHz bands on the EVLA.
I'd like to clarify what I believe the desirable end-state should be, as
I've promised this description to a number of people, and especially
because
I do _not_ think wire-swapping (as some discussions may be tending)
is necessary, in fact.

The basic discrepancy in the current labeling scheme is that the dipoles
that are parallel to the elevation axis are labeled "X"  (or "XLP").  This
is
effectively the reverse of the IAU convention[1].   This is best understood
by
considering an instantaneous observation on the meridian looking south
(i.e.,
telescope is tilted).   At that instant, when sky and telescope coordinates
are
aligned, it would be least confusing (and require the least s/w) if our
telescope
coordinate labels matched the IAU convention, which has "X" == "vertical"
(i.e., parallel to lines of R.A. in sky coords).   Instead, in the current
EVLA
labeling scheme, the "Y"-labeled dipole will be "vertical".     Arguably,
this is
mainly semantics, but it has very practical and significantly confusing
consequences that are worth avoiding.

Since the "X" and "Y" labels are ultimately used in post-processing to
trigger
formal Stokes parameter conversions in a manner that _is_ consistent with
IAU
conventions (namely, that the <XX> correlation nominally measures I+Q,
etc.), an additional 90 deg rotation is required for proper position angle
calibration[2].  Application of the 90 deg pos angle calibration amounts to
converting to the sky polarization (IAU) coordinates, and therefore
effectively
swaps visibilities w.r.t. to their correlation labels such that an "XX"
correlation
will thenceforth contain data that, in fact, originated in dipoles labeled
"Y" on the
telescopes (and similar transpositions for the other 3 correlations).  In
other
words, discussions of actual dipoles between astronomers and engineers will
be potentially confused by the particular state of the pos angle
calibration of the
dataset in question.  (Inasmuch as the nominal 90 deg rotation will be a
hard-wired correction [see [2], below], it may  often be ambiguous whether
or
not it has, in fact, been applied to the data appearing in some plot.  One
must be
explicitly cognizant of the pos angle calibration state, in any case.)

So, *I'd propose that we merely begin referring to the "vertical" (when
antenna is
tilted) dipoles as "X" or "XLP" and "horizontal" (parallel to elev axis)
dipoles
*
*as "Y" or "YLP".    This implies only _re-labeling_ (lots of stickers,
cable tags,
*
*as well as engineering documentation updates), and (I emphasize) it
_does_not_
compel any rewiring or cable-swapping*[3].   Since the nominal pos ang
calibration
will no longer be required, the confusions described above disappear, and
far
less implicit management of the orientation in s/w is required.  I'd note
also that
the mere relabeling will also result in conventionally-ordered correlations
in the output
dataset, i.e., [XX, XY, YX, YY], a mainly cosmetic feature that nonetheless
probably
simplifies data transport issues (or at least potential confusions therein).

To be clear, the new labeling scheme would be:

*"vertical" (when antenna tilted) == "X" == goes where "R" goes
*
*"horizontal" (parallel to elev axis) == "Y" == goes where "L" goes*

Currently, the correlator back-end is not distinguishing the linear feed
bands when labeling the output data.  I.e., I believe we still get data
labeled
[RR, RL, LR, LL].   In the current dipole labeling scheme, we should
be getting [YY, YX, XY, XX] for raw data (telescope frame), which will look
quirky
cf any other linear feed telescope (only cosmetic, in principle, but s/w
must cope and this has not been exercised).   In the IAU-conformant
labeling
scheme,  we would get [XX, XY, YX, YY].   These labels originate in the
Feed.xml and
Polarization.xml SDM tables.    The IAU-conformant case requires
changing R to X and L to Y (or the corresponding changes to the appropriate
enums) in the SDM-writer when the low bands are observed.   (The existing
labeling scheme requires the opposite substitutions, as well as also
recording
the 90 deg pos angle offset in Feed.xml.)

>From the science operations perspective, note that introducing the
IAU-conformant labels in existing SDMs will (trivially!) yield datasets
requiring no implicit correlation sorting or _nominal_ position angle
calibration in post-processing code.   I.e., they will work correctly in
(at least as regards polarization coordinates for science purposes)
_immediately_.   When an astronomer says "X" it will mean the vertical
dipole, so confusion remains until the h/w labels are changed.
(If the dipoles themselves are grossly and systematically mis-oriented
_mechanically_, a ~conventional  pos angle calibration using a reference
source will still be required, of course.)

(Someone should think about whether or not the relabeling compels
a swap in the Tcal value catalogs for these bands.  Since we wouldn't
be changing the order in which the physical signals are presented to the
band-independent parts of the system, it is not clear to me whether
a swap of  Tcal values is necessary.  It is perhaps possible that
the current Tcal value ordering may, in effect, be inconsistent with the
current dipole labeling!)

Cheers,
George


[1] According to Hamaker & Bregman (1996, A&AS 117, 161; see
their Fig 1), the formal IAU definitions for polarizations in _sky_
coordinates has "X" parallel to lines of constant R.A. (i.e., vertical
in images) and "Y"  parallel to lines of constant Decl (i.e.,
horizontal).
This arises from defining position angle zero to be a _vertical_
orientation in sky coordinates (coinciding with the 'X' axis), and
measured positive counter-clockwise.   With this convention,
an XX correlation measures Stokes (I+Q)  (i.e., with a plus sign), etc.

[2] Performing the 90 deg rotation is trivial, and SDM and MS have
provisions
for effectively hard-wiring the application of this rotation via information
extracted from the Feed.xml table.   This is used routinely for ALMA which
has dipole feeds at a variety of non-standard orientations  ("X" is always
the feed oriented radially on the off-axis feed circles).  However, the
specific
case of a 90 deg position angle calibration causes particular confusion
when
discussing specific dipoles as described in the text above.

[3] Swapping the cables (e.g., on the output of the low-band receiver)---
such that the "X" labeled signal follows the path "RCP" takes for ordinary
EVLA bands---will simply reverse the labeling of correlations in the
resulting
SDM:   [YY, YX, XY, XX] --> [XX, XY, YX, YY].   Cosmetically desirable, but
this
has no practically useful effect on the position angle calibration issue
described above, and so does _not_ relieve the confusions attending
discussions of actual dipoles.    As such, swapping cables amounts
to a lot of work with _only_ cosmetic (mainly) benefits.  Note that both
swapping cables _and_ relabeling dipoles (equivalently, physically rotating
the dipoles themselves by 90 deg, if that's possible) will relieve the pos
angle
confusions but yield the unconventional correlation ordering: [YY, YX, XY,
XX].
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20130926/ffae1e30/attachment.html>