[evlatests] Strange R-L phase symmetries
George Moellenbrock
gmoellen at nrao.edu
Mon Mar 28 15:13:26 EDT 2022
Gang-
I'm going to try to revive interest in a broadened, but still /~purely
//geometrical explanation/. In particular, I wonder if we appreciate
the true incongruities in the realized optical geometry as well as we
think we do, at least as regards the net effective rotation of the feeds
on the sky near the zenith. Apparently not quite, and more than just
simple Az axis tilts could be relevant (yet still without going behind
the feed to wires and software). I.e., where is the /primary/ boresight
actually pointing?
*Regarding Az tilts (for starters): *The properties of the symmetries
Rick has shown are entirely consistent with (differential) tilts in the
AZ axes by a few arcmin. E-W tilts cause the even symmetry (time
offsets in geometry calculation can do this to...). N-S tilts will
cause the odd symmetry. It was casually asserted early in the
conversation that the peculiar tilts (due to sag of pads, etc.) aren't
big enough for this. Is there a real quantitative basis for this claim?
The basic geometrical interpretation is really a matter of answering the
following question: "How is the antenna rotating around the direction
to the source in general, and especially nearer the zenith?" Do we
really think we know this near the zenith at levels comparable to the
scale of tilts required for the observed effect? Rick correctly stated
"parallactic angle is not a function of polarization", but
misapprehension of the realized parallactic angle evolution (in
geometric models used to correlate/calibrate) will have opposite phase
effect on the R and L polarizations. I.e., any effective rotation of
the feed on top of the assumed geometric model of the system will
advance the R phase and retard L's, or vice-versa. This is what Rick's
plots show, differentially with the refant. So, does the correlator
phase model include terms for the peculiar tilts in each antenna? If
not, then the peculiar tilts /must be /introducing these effect at some
level, i.e., at least some of the observed effect is due to the Az axis
tilts.
*But, if you need more effect than mere Az tilts can supply, my main
/new /point is /primary boresight pointing accuracy/:* Beyond the
simple peculiar tilts, we are actually also assuming that the antennas
point precisely on their /primaries' /optical boresights toward the
source. Is this true? The (joint?) optimization of pointing, focus
and collimation is presumably respectably optimizing /net forward
sensitivity/ (and somehow averaged between R and L to fall between
between the squinted beams, which we presume /is/ the primary Stokes I
boresight). This broad optimization should tend toward, but by no
means necessarily guarantee, that the pointing /of the primary /(which
is what is driven by the motors) is precisely toward the source.
Indeed, is there any way to objectively guarantee (i.e., constrain) the
primary boresight in the optimizations we perform for the optics? In
particular, I wonder if the collimation optimization does, in fact, push
us distinctly (but subtly..., and enough?) /away/ from the primary's
boresight? I.e., to what extent do we optimize collimation by
moving/pointing the feed horns themselves (at the few arcmin level), cf
just adjusting the _net_ pointing of the mechanical optics (forward of
the feed) to ~compensate and balance? A few arcmin offset only
negligibly affects the primary's forward gain, so no leverage there...
And it is interesting to note that a lot is bootstrapped between and
among bands in achieving the general combined optimization of all feeds
across the sky (i.e., everything pegged to whatever uncompensated
mechanical offsets exist in the X-band feed?). Not to mention the
likely relevance of sub-reflector rotation tricks (at higher-freq bands)
to the net geometry of the optics, which speaks to the level at which we
need to account for loss of the rigidity implicit in the simple
geometric description..... In short, optimization of the whole signal
path is almost certainly not optimization of each optical path component
in isolation, and in particular, */the drive motors are probably
effectively moving the primary for a point on the sky that is _not_
precisely the target source coordinates, such that the net rotation
around the direction to the target isn't quite the right one/**.* And
this will be most noticeable nearer the zenith, of course, in the
relative R-L phase/,/ in a manner very like the simple Az tilts. The
main drawback to this explanation is that we might have expected more
band-dependence of the effect, unless we are dominated mainly by the
bootstrapping from one band, or something else systematic (per antenna,
not band) about the effective boresight directions of (aging) VLA
antennas.....
(I pose the above based on some ongoing off-and-on (mostly off, lately)
experience studying similar questions for ALMA, where, in fact, they
have /deliberately/ chosen to translate (rather than tilt, as designed)
the subreflector to reach off-axis feeds (on 2 feed circles). This
means they are deliberately /moving off the primary boresight/. And
since we don't really know the subreflector zero points in tip/tilt and
translation, I don't think we really know how far off the boresight ALMA
antennas actually are... So, they've unintentionally compromised the
feed orientation calculation in calibration for a (measured, memo'd)
very small net loss in forward sensitivity.)
*Regarding OTF pointing updates:* Also, I think we expect blind
pointing to be poor near the zenith, by which I mean we don't expect our
ordinary optimizations to be very good.... I wonder if the amplitudes
(in particular, the /relative/ R/L amp) might give a clue about how far
from the nominal pointing we have wandered (on top of the offsets
introduced ~deliberately through nominal optimizations described
above). Also, manually tweaking up the pointing on top of the model
at, say, HA ~ -1h might actually be an effectively arbitrary
"correction" to a point decidedly off-source for the primary boresight
nearer the zenith....
*Regarding measuring cross-hand phase directly: *Examining truly
measured cross-hand phases will definitely be interesting. Note that
this will be a measurement relative to the (simple) parallactic angle
calculation used to make the sky nominally stationary in rotation. This
calculation does not include all of the inhomogeneities described above
(true axes tilts, effective optical path offsets), and is also subject
to the coordinate system chosen for the parang calculation. I think
both AIPS* and CASA have traditionally used the geocentric latitude (not
geodetic) for the parang calculation, which effectively behaves like a
~10.7 arcmin tilt to the North. This creates a several degree position
angle error near the zenith (twice this in R-L phase) of the odd
symmetry, and is conveniently nulled by the difference measurements Rick
has shown so far (owing to the fact that VLA antennas are nominally
mounted on the earth in parallel). Indeed, it is the scale of this
alone that keeps me scratching my head about just the ordinary tilts of
a few arcmin being enough to cause at least some of the effect Rick
observes. So, don't be surprised if the actual cross-hand phase
(effectively, of the refant) looks worse!
(*I'd welcome Eric's correction on this point, if I'm wrong about this.)
*Regarding 'over-the-top': *I think over-the-top might ~decouple Az
tilts from internal (feed-forward) optics, since the net primary
boresight pointing error is probably different for over-the-top, but I
haven't thought very carefully about this.... Hmmm, I think net feed
rotation is in the opposite direction for over-the-top, so I don't think
you get the same thing for the Az tilt effect--won't it reverse the sign
of your differential phases? If only a sign reversal, then that test
tends to point to Az tilt as the culprit. But there are probably also
different boresight pointing effects, so you'll sorta measure the
relative scale of those... And bending wires can also still contribute....
Cheers,
George
On 3/27/22 21:25, Rick Perley via evlatests wrote:
> Well — I certainly didn’t think I’d get so many suggestions! A healthy sign.
>
> Regarding AC/BD: Sadly, the data taken used only the AC side.
>
> The thinking seems to point to the antenna, rather than some geometrical origin. To separate these effects, perhaps tracking 3C286 through transit in two different ways may help — (a) in the normal mode, and (b) using ‘over the top’. If the effect is due to geometry (related to parallactic angle), these two should give the same results. If due to the antenna, the different elevations (86 and 94 degrees at transit) should clearly show up as giving different magnitudes.
>
> I agree that software is unlikely — but to be sure, I can generate these plots with no calibration at all (since these are differential plots, the atmosphere and most electronics effects should cancel out).
>
> I’ll plot these phase differences against elevation — if a true elevation effect, all traces should lie on the same curve. (I should have done that on Friday!).
>
> Regarding the choice of reference antenna — ea10 looks ‘reasonable’. I will use a different antenna as reference (clearly, one of the ‘odd’ ones) — but the results are easy to anticipate — the current plots will have the new reference antennas’s curve added. So I can hope that all (or most) of the ‘odd’ profiles will head to ‘zero’ (no elevation/HA effect), while the ‘even’ profiles will change in a way that I hesitate to predict … (depends on the magnitude of the ‘odd’ profile being added to the large ‘even’ profile).
>
> I probably won’t be able to do these checks until Monday afternoon.
>
> Rick
>
>
>
> Sent from my iPad
>
>> On Mar 25, 2022, at 10:23 PM, Craig Walker<cwalker at nrao.edu> wrote:
>>
>> This is an interesting puzzle. Here are a few thoughts on the problem:
>>
>> The higher dec sources have a very high rate of change of Azimuth and PA at transit. The sharp peak in the R-L phase effect makes me think it is related. The effect at the antennas with the single peak is much larger than the effect with the two peaks (one negative). If all antennas, including the reference, have a peak at transit but of random sign and with slight and maybe random offsets from actual transit, you might get what is seen. When an antenna's peak is of opposite sign from that of the reference antenna, the effects add and you get a single large peak. When they are of the same sign, so they try to cancel in the difference, the slight offsets from actual transit give the two peak character.
>>
>> The fact that the effect is scattered randomly over the array (really true?) suggests that it is some hardware effect not related to observing geometry. Also it may be important to remember that the pads are tilted so that Az, El, and PA are the same at all antennas despite the Earth curvature over the array.
>>
>> My first thought was that this all points to the azimuth cable wrap. But the fact that the values far from transit are the same on both sides doesn't match this too well.
>>
>> With the VLBA, we get an amplitude effect that looks a bit like this at the point when the source is off the end of a baseline and the fringe rate goes through zero. Then any clipper offsets, pulse cal tones or other signals that are the same at the sites correlate. Could there be something in the VLA system of the sort that acts at transit? That is definitely grasping at straws.
>>
>> Definitely a puzzle.
>>
>> Cheers,
>>
>> Craig
>>
>>
>>
>>> On 3/25/22 11:53 AM, Rick Perley via evlatests wrote:
>>> This is a long circular -- apologies to all, but the subject is a bit complex ...
>>> Many will remember a meeting called by Frank a few years ago where the subject was the very peculiar phase differences seen between the RCP and LCP phases when observing a source passing by the zenith. The general conclusion was that 'we have no idea of what is going on'.
>>> In preparation for an upcoming trip, I am reviewing my extensive observations, taken over the past decade or more, from projects with the goal of measuring, and implementing the 'absolute' D-terms. (In other words, dispensing with the usual method of measuring the antenna polarizations with respect to an assumed standard (usually zero)).
>>> One observation, taken in January 2019, is especially well suited to this task. I observed four sources, through transit, for five hours, at three bands -- L, S, and C.
>>> The four sources were:
>>> 3C286 dec = 30.5
>>> OQ208 dec = 28.5
>>> 3C287 dec = 25.2
>>> 3C273 dec = 2.0
>>> Note that OQ208 is completely unpolarized, while the others have varying degrees of polarization. All sources transit south of the zenith.
>>> The data are of exceptionally good quality. The array was in the C configuration.
>>> The attached plots show the R-L phases, using ea10 as the reference antenna. Note that these are *not* the RL or LR correlation phases -- they are the differences between the antenna phase solutions using the RR and LL data, using ea10 as the reference. This means the R-L dependence of ea10 is impressed on all the other antennas. We are looking at differentials.
>>> The plots show two antennas -- ea01 and ea12, which represent the two different symmetries seen in the data. The x-axis is HA -- plots against time and parallactic angle jumble the results -- the dependencies seen are purely a function of HA.
>>> Colors: 3C286 is red, Light green is OQ208, blue is 3C287, dark green is 3C273.
>>> ea01 is of the even symmetry type. Antennas 1 3 5 6 8 15 and 22 have this symmetry.
>>> ea12 is of the odd symmetry type. All other antennas show this, with the same sign -- positive difference before transit, negative difference after, with the possible exception of ea18. (For this antenna, the amplitude of the effect is very small, so the signature is hard to discern). Three antennas were out of the array at the time: 7, 24 and 28.
>>> Key points:
>>> 1) The phase signatures are *identical* for each band. Same width, same height, same values, same symmetry.
>>> 2) The magnitude of the effect is sharply dependent on how close the zenith the source transits. For 3C273, the effect is almost completely absent.
>>> 3) The effect is independent of source polarization. OQ 208 has less than 0.1% polarization, and shows the same symmetry signature as the strongly polarized sources 3C286 and 3C287.
>>> 4) The location of the antennas is not related to the signature -- the 'even' antennas were located all over the array: W6, W18, E14, N6, N1, E12, and W12.
>>> One conclusion is clear: The effect has nothing to do with the beam squint. And it is very hard to see how differences in the antenna pole direction can do this -- the required tilt magnitudes are just unreasonable. And in any event, the parallactic angle is not a function of polarization -- it's an antenna quantity.
>>> I have shown these data to two of our serious pundits (Barry and Steve), hoping for some insight. None was forthcoming. We are completely stumped. It seems clear that the signatures are geometric in origin -- but how does this translate into such a clear signature in the phase *difference* between polarizations?
>>> Any and all suggestions will be taken seriously!
>>> Rick
>>> _______________________________________________
>>> evlatests mailing list
>>> evlatests at listmgr.nrao.edu
>>> https://listmgr.nrao.edu/mailman/listinfo/evlatests
>> --
>> ------------------------------------------------------------------
>> R. Craig Walker Scientist Emeritus
>> 1305 Vista Dr. Array Operations Center
>> Socorro NM 87801 USA National Radio Astronomy Observatory
>> cwalker at nrao.edu P.O. Box O
>> Phone 575 835 3972 Socorro, NM 87801 USA
>> 575 835 7247
>> ------------------------------------------------------------------
>
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.nrao.edu
> https://listmgr.nrao.edu/mailman/listinfo/evlatests
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20220328/5721ff9c/attachment-0001.html>
More information about the evlatests
mailing list