[evlatests] Switched Power Tests -- overview

[Rick Perley]

     Given here are the top-level results of the band switch test.  
Subsequent circulars will details some new (and unusual) behavior found.

     The test consisted of observing a single source (J0359+5057 -- 
about 10 Jy at all bands), and alternating between all pairs of bands 
(excepting 4-P).  Each observation was one minute long. Each band pair 
sequence consisted of 5 observations at each band (10 minutes total per 
pair).

     Standard system flags were employed.  In a very few cases, minor 
additional flagging was required to remove bad data missed by the system 
flags.

     The PSum and PDif values, for each band, were examined for 
behavior.  Gain calibration was done for both the 'raw' data (not 
adjusted by PDif) and the properly calibrated (switched power applied) 
data, to judge the value of the corrections.  In this, I used two clean 
spectral windows -- those with no RFI visible -- one from the 'AC' side, 
the other from 'BD'.

     The observations were made with the 8-bit samplers, to avoid the 
additional issues associated with the 3-bit system.

     Of particular interest was ea10 -- which has had the switches 
within the T303 (UX converter) replaced with a new type.  This antenna 
was identified earlier as a 'bad actor' with respect to band changes.

     Summary Results:

     A) Low Frequency Bands (L, S, C, X)

     The quality of the PDif values is outstanding for all antennas, 
with only a couple minor exceptions.  In all cases, application of the 
switched power stabilized the gain to typically 1% or better.  (Without 
the switched power, the stability is still very good -- nearly always 
better than 4%).

     I previously sent around a spectacular example of how switched 
power can correct data -- ea02, on the RCP side, at C-band, suddenly 
developed large and rapid gain variations (sinusoidal, 20% peak-peak), 
which were completely removed by using the switched power.

     The only exceptions to this rosy picture are:

     * At L-band, ea22, the PDif values for RCP undergo fluctuations on 
~1min timescale, with up to 10% power amplitude variations.  These are 
not seen in the antenna gains.

     * At X-band, ea05 had ~8% PDif variation not seen in the visibility 
amplitudes, for a single 10 minute cycle (with L-band).

     B) High Frequency Bands (Ku, K, Ka, Q).

     As expected, the 'AD' side showed all manner of rapid gain 
instabilities.  These come in three flavors:  (a) 'Shredded -- random 
gain changes of up to tens of percent, varying on second timescales, 
lasting the entire 1 minute for each observation, (b) Two-states, where 
the gain changes by up to tens of percent, in the middle of the scan, 
and (c) Steady changes, where the gain continuously varies, by up to 
tens of percent, over the length of the scan.  Most antennas are 
affected, and there is no correlation between behaviors on the opposite 
polarizations.  Of the four high frequency bands, Ku band is the most 
affected.  Some antenna-IFs are good, others very bad.  By far the worst 
is ea07.  The good news is that ea10 is clean -- it is likely that the 
changes made to the T303 are effective.

     In contrast, the 'BD' side is much more quiescent, with only a few 
examples of this behavior.

     In all cases, the variable values of PDif are seen in the actual 
antenna gains, and application of the switched power reduced the level 
of the variations, usually to within ~2 -- 3%. But -- the cost of this 
is in the noise -- the post-correction visibility gains are noticeably 
noisier than pre-correction, as no smoothing of the switched power can 
be done due to the rapid timescale of the variations.  This effect 
degrades sensitivity -- although if the target source is very strong, 
this degradation can be removed by self-calibration (since the noisy 
PDif values are applied on an antenna-basis).

     On the basis of this, I conclude:

     1) Use of the switched power will result in a much improved 
stability of the antenna gains.  (This presumes the values employed are 
free of RFI-induced variations -- a Big IF!).

     2) The switch change in ea10 has made this antenna immune to the 
'shredded/variable' gain phenomenon, and should be retrofitted on all 
antennas.  If another test is deemed necessary, I recommend ea07 to get 
the next change.

-------------------------------------------------------------------------

     But of course, it's not as easy as this.  These tests have 
demonstrated two new phenomena, which are sufficiently complicated (and 
potentially serious) that they will be the subject of subsequent 
circulars.  Stay tuned ...