[evlatests] C-Band Stability Tests

[Rick Perley]

    The 'flux density' data, taken Dec 26/27, has not been fully 
calibrated for C-band. 

    The 'one-line' summary is:  These data are at least equal to, and 
probably better than, any of the C-band flux density data runs taken 
over the past 26 years, in terms of the overall quality and (amplitude) 
stability. 

    There are a few 'issues', mainly dealing with the 'switched power' 
system, detailed below. 

    There were 105 separate observations of fourteen sources taken over 
the 30 hour span.  The 'AC' IFs were tuned to 4821 MHz (center), the 
'BD' IFs to 6949 MHz.  Referenced pointing was not utilized for this 
band.  (But I probably should have).  Calibration without any temporal 
trending (i.e., a flat boxcar with 90 hour width) was used for all 
parameters except phase.  Statement on system gain stability then refer 
to the variations seen about the mean as established by this 90 hour 
average. 
    The four standard sources:  3C48, 3C138, 3C147 and 3C286 were used 
for all calibration.  The models provided by AIPS for each were utilized 
to remove partial resolution.  The flux density of 3C286 was taken as 
the fiducial standard. 

    The major problems noted are listed below. 

    1) Issues arising from basic flagging. 

    a) Antenna 17 was often slow to regain fringe power -- this was due 
to a slow subreflector.  (Has this been addressed?) 
    b) Antenna 19 was taken out partway through, as its subreflector 
became stuck. 
    c) Antenna 4 was out for nearly 9 hours due to a problem with the LO 
rack power.   But after it returned, all SY (switched power) values were 
nonsense. 
    d) Telcal got into trouble 9.5 hours into the run.  The script was 
aborted, and restarted.  All antennas came back nicely -- but many with 
a different power level.  (This was useful, as will be reported below).  
But for antenna ea03, the SY table values were essentially scrambled, 
with values at least an order of magnitude too high.  The problem was 
particularly bad for the AC IFs. 
    e) Antenna 12 suffered some sort of unreported problem (no fringes, 
but no reason known), after which its switched and total power dropped 
by factors of a few (and the bandpass changed). 

    2) Gain Stability

    Nearly all antennas show a clear diurnal variation in gain (as 
revealed by PDif), and in the amplitude of the source visibilities.  The 
max/min of the switched power variation is 1.07.  Application of this 
variation to the visibility data (and utilization of an elevation gain 
curve) results in post-correction gain flatness of better than 1%, 
pk-pk, in Stokes 'I'.  Most antennas show very little variation in the 
PDif values from the 24-hour smooth curve.  But some are notable at the 
level of a few percent -- presumably due to band switch issues.  The 
most notable of these are:  ea01 in IF 'A', ea21 in IF 'A', ea23 in IF 
'C' (this one is particularly bad!), ea24 in IF 'A', and ea28 in both A 
and C.  (I have not yet reviewed the PDif plots for IFs B and D yet).  
As noted above, a number of antennas showed a ~1dB power change after 
the resumption of observing -- for all of these, and all of the antennas 
with 'flaky' switches, the application of the switched power (PDif) 
corrected the visibilities with an accuracy better than 1%. 

     ea07, in IF 'A' shows no diurnal variation in PDif, but does show 
the variation in the visibility amplitudes.  Hence, the switched power 
is not correctly operating in this antenna-IF.  IF 'B', 'C' and 'D' are 
fine for this antenna. 

    I had reported earlier this month that there was a curious relation 
between RCP and LCP gains, wherein when one was below the norm, the 
other was higher, and speculated this might be due to a polarization 
effect, wherein the D*D, or D*P terms (the latter suggestion coming from 
George) might give the signature seen.  But the former is most doubtful 
(not time variable), and the latter should only be notable for highly 
polarized objects.  This latter signature is not present -- the 'split' 
is as large for 3C147 (unpolarized) as it is for 3C138 (highly 
polarized).  I now suggest it is due to small pointing errors -- smaller 
than the 'squint', which will, for the proper parallactic angle, perturb 
the two polarization gains in opposite direction.  Note that a 30 
arcsecond pointing error  provides a 3% change in gain -- easily 
detectable in this experiment.  This is the origin of the parenthetical 
comment made above. 
    Elevation gain curves were fitted to all subbands and all antennas.  
The effect is always small (couple of percent from horizon to zenith), 
and not the same between opposite polarizations -- hence the variations 
seen probably reflect some small elevation sensitivity in the 
electronics, and not a change in antenna efficiency. 

    2) Bandpass stability. 

    Bandpasses were generated using the four 'standard' objects -- a 
total of 32 solutions.   For most antennas that did not suffer internal 
gain jumps (i.e., no change in the attenuators), bandpass stability over 
the 30-hour period is better than 1%, pk-pk.  This is far from the 
system requirements, but anybody wanting .01% bandpass stability should 
not be observing at all eight bands, all over the sky. 
    There are plenty of antennas which are worse than 1%:  In most 
cases, these are immediately identified with the change in power levels 
following the abort and restarting of the script after the first nine 
hours had passed.  In all of these cases, the change in the bandpass 
function is about 2--4 %, pk-pk. 
    But there are some other unstable bandpasses, for which no immediate 
cause can be found.  Chief offenders are:  ea07, in both 'A' and 'C', 
with 3 solutions having high-frequency sinusoids imposed;  ea08; ea12, 
following the change in gain noted above; ea21, in the two lowest 
subbands of IF 'A'; ea23, in IF 'C', ea27, which has the same  
high-frequency oscillation on some scans in both polarizations as noted 
for antenna 7.  In all cases, the amplitude of these variations is about 
2%. 

    3) Phase stability. 

    As noted for X-band and Ku-band, the overall phase stability is much 
worse than I think it should be.  Phases within any one scan are stable, 
but 'jump' by tens of degrees between subsequent observations at this 
band.  The 'scatter' in the phases shows little relation to antenna 
separation from the center -- being about the same for each antenna.  
But each antenna jumps differently.  I can't rule out atmospheric 
effects just yet. 

    4) Polarization. 

    Absolutely beautiful polarization calibration was done, thanks to 
the new AIPS code.  Plots of calibrator and antenna polarizations were 
made over the 2 1-GHz-wide bandpasses were made, show smooth variations 
with frequency, as expected. 
    Differential 'D' term amplitudes (reference = ea08) are typically 4 
to 8 percent, occasionally higher.