[evlatests] Lost Command in Holography?

[Rick Perley]

    Michiel generated an 11 x 11 script to test some new holography 
scanning features.  The observations were made at C-band on 3C84 to 
ensure both stability and high SNR.  In general, the data quality is 
superb. 

    But there is one extraordinary feature that I've not seen before -- 
although in truth, it would be very hard to catch. 

    The test comprised 11 vertical cuts through the beam, each cut with 
11 sample points.  Each sampling point is 10 seconds long.  Each 
vertical cut is offset horizontally from the last by an equal step.  The 
pattern sequence is to alternate directions:  up, shift left, down, 
shift left, up ...  For this particular test, the grid was set so the 
beam maximum was at the top edge of the grid, and offset by one vertical 
column. 

I've attached a plot showing two consecutive vertical cuts, the first 
being the cut which ends on top of the main beam, and the next cut being 
shifted over by one horizontal step. 

    The plot shows that the first cut behaves exactly as expected.  The 
stepsize is close to critical -- about one half the FWHM.  So on the 
left side of the plot, you can easily see the source moving through the 
antenna sidelobes. 
The program UVHOL reports the antenna offset coordinates for this column 
start at (0, 15.53) and end at (0,0). 

    Following the on-axis point (with amplitude = 1.0), is a time gap of 
nearly one minute -- this is the on-source calibration. 

    Following this gap, the intention was to move one horizontal step, 
then step down the beam by 11 equally spaced vertical steps.  Indeed, 
UVHOL reports the coordinates starting at (1.55, 0) and ending at (1.55, 
15.5).  But that is *NOT* what the antenna actually did. 

    Look at the plotted points immediately following the gap:  at time 
12:50:30, and amplitude 0.65.  UVHOL reports the first half of these to 
have coordinates (1.55, 0), and the second half to be (1.55, 1.55).  But 
all the plots have the same amplitude.  So either the beam forward gain 
is the same at those two positions (and this is impossible, since the 
(1.55, 1.55) position must be close to the first null), or the antenna 
didn't move, even though the SDM thinks it did. 
    Further -- and even better -- evidence of a lack of motion, despite 
the listed coordinates comes from:
    1) The first null as seen in the 2nd vertical cut is visible on the 
next position, and the 2nd sidelobe peak on the position following 
that.  Both of these should be seen one step earlier.  The entire column 
is offset by one step!
    2) The on-line flagging is very good at flagging data when the 
antenna is in motion, and not at the commanded position.  For the (1.55, 
1.55) position -- and *only* this position (out of 11 x 11 = 121 
position), there is no antenna flagging. 

    I conclude the antenna did not move, even though the recorded 
antenna positions (as reported to AIPS) say it did. 
    Such a failure to obey a command has rather serious consequences to 
accurate holography -- and is not easily correctable, as >99% of a real 
holography scan is spent on the weak sidelobes. 

    How common is this problem?  Unknown, for the reason given above.  
We got lucky here, since the problem manifested itself while we were on 
the main beam.