[evlatests] Modcomp-free / baselines & other VLA phase problems

[Vivek Dhawan]

This was circulated to some, I thought it should go to the
list.

Ken Sowinski wrote: (I have edited this to update)

 > Vivek has examined data from the Modcompless baseline run
 > of last week.  My summary of fundamental result is that
 > EVLA antennas are about right; VLA antennas have large
 > position errors that grow to an estimated 1 meters at the
 > end of the arms.  If expressed as a rotation this is about
 > 12 arc-sec.  This is much larger than any of the antenna
 > position errors or inconsistencies we have made in the past.
 > A likely explanation is that the timetags and phases sent
 > to the L7s have slipped relative to each other by 1-1/4 sec
 > of time.  I can think of no other way to explain both the
 > magnitude and the fact that EVLA antennas are not affected.
 >
 > I had a look at VlaAntennaPhysical, saw that the OTT parameter
 > was not added to the phase, but otherwise succeeded only in
 > in confusing myself.  This is also where phase corrections
 > are applied for phased array observations which does not work
 > for VLA antennas.
 >

Repeat from my email of last week: -------

VLA and EVLA don't seem to be in the same frame by many cm;
They also seem to be looking at different a-priori positions
(and K-terms) since June 02.

Update as of today: ------------------

1. K-terms (axis offset) on VLA antennas were entered into the
    EVLA database with a X10 scale factor; Ken fixed this Wed
    Jun 20th for future data. He also found that phases during
    over-the-top (OTT) observing were inverted (see above).

2. A successful modcomp-free pointing run was done Wednesday
    night (Ken analyzed this).

    I loaded this data at Ken's suggestion to see if useful
    baseline solutions could be derived from the boresight
    pointings. Inspection of the data from the inner antennas
    shows that the large K-terms are gone. However, the fitting
    program LOCIT will not swallow this data; this is worth
    pursuing in the long term, but not today.

    This dataset also had no fringes on the following baselines:
    (Other baselines to these same antennas were fine). 9 had
    fringes to all antennas.
    6 to EVLA*, 8 to EVLA*  all fringeless; so also 1-12,
    2-3, 4-6, 4-8, 5-10, 5-25, 10-15, 15-25, 20-28, 22-28.
    Ken said something about DCS numbers - is this the same
    thing?

    Note added Monday 25: these oddities may have started after
    the CMP troubles of Wednesday June 20th.

3. Returning to the modcomp-free baselines data of Jun 19th:

    The a-priori antenna positions (in the AIPS AN table) are
    different in the modcomp-free era by upto 3 cm at the ends
    of the arms (A-array) compared to those used by the modcomps.

    According to Ken, this is because frills such as Earth tides
    and pole position are now done in CALC. The resultant phases
    should be close to the values derived by the modcomp scheme,
    if all is working well.

4. Clearly all is not well:

    After the large K-terms were fit on the data of Jun 19, the
    XYZ offsets of VLA antennas (not EVLA) show a systematic
    increase with position along the arms, out to station #32,
    beyond which the phase wraps, even using the phase difference
    between the IFs separated by 500MHz.

    The offsets are thus: (Other components are much smaller)
    Y= +20cm at N32.    X= +29cm at E32.     X= -25 at W32.

    This corresponds roughly to 12arcsec of rotation, over one
    meter of position offset at the ends of the arms.

    Ken thought that the difference between arrays might arise at
    the bifurcation point where the lobe rotation hardware is
    controlled; this update rate on the VLA is 1.25sec, or,
    for us straw-clutchers, 18.75" of angle.

    I tested this by shifting the RA of all sources by the amount
    dRA = dT*15*cos(dec) and re-fitting the data. The result is
    that all VLA antennas can be tamed within a few cm using dT
    of 0.9 seconds. 1.25s and 6.25s are clearly incorrect.

Vivek.