[evlatests] Collimation errors, and a problem with the OPT ...

[Rick Perley]

    A script to determine the antenna gains (efficiency) as a function 
of elevation was run Saturday morning.  The concept is to follow a 
strong source, from zenith to transit (or from transit to zenith), from 
near midnight to just before dawn.  The time requirement is to minimize 
any temperature-dependent effects, primarily to the noise diodes.   The 
need to sample a wide range of elevations means the target source must 
be in a declination range from about 25 to 42 degrees.  All these 
requirements imply a rather small number (about six) potential target 
sources. 

    To remove any electronic gain changes (which will look like antenna 
gain changes), we utilize our switched power calibration system.

    For the Saturday run, the source chosen was J2007+4029.   
Elevation-dependent gains are only significant at the higher frequency 
bands, so the observation used X, Ku, K, Ka, and Q bands.  I observed 
them in that order.   I selected two separate tuning pairs for each 
band  (hence, four frequencies per band) to permit a possible gain 
formulation as a function of frequency within any one band.

    To remove apparent gain changes which are caused by bad pointing, 
'referenced pointing' was run at each band.  Because of the possibility 
of large collimation errors (see below!), I ran the referenced pointing 
on the source for each band, and for each observation, and utilized the 
result only for that band.  (Hence, x-ref for X, Ku-ref for Ku, etc.).  

    The run lasted about 7 hours, and each frequency tuning pair was 
observed twenty times.  For each observation, following correction for 
electronic gain variations, a calibration solution was determined.  The 
elevation dependency was found using the AIPS program 'ELINT'. 

    In general, very good results were obtained.  But there are some issues:

    1) My parenthetical  note above about possible large collimation 
errors proved true.  Ku-band, in particular, has a number of antennas 
with large collimation offsets -- nearly all in elevation.  Here is a 
short list (elevation offsets, unless otherwise noted):
         ea01         +0.4 arcminutes in elevation, -0.2 arcminutes in 
azimuth
         ea02         -0.7 arcminutes in elevation, +0.3 arcminutes in 
azimuth
         ea10         -0.4                                     also, 
-0.3 arcminutes in azimuth.
         ea11         +0.5
         ea14         +0.3
         ea27         -0.35
          Other bands are much better.  K-band has two apparent large 
offsets, all in azimuth: 
         ea10         -0.4 arcminutes
         ea15         +0.5 arcminutes

    2) Unfortunately, most Q-band determinations failed -- a surprise to 
me, since the source is over 2 Jy.  The failures all appear to be SNR 
related.  However, I expect the collimations at Q-band to be good w.r.t. 
Ka (which preceded Q-band -- see next item for explanation), so the 
Q-band gain curves determined should be o.k. 

    3) An OPT issue:  The OPT has a very useful capability when 
determining a referenced pointing solution:  One can either determine it 
'blind' (i.e., without applying a previous solution), or one can 
determine the solution after applying the (not 'a') previous solution, 
and hence determine a 'differential' solution, which is then properly 
added to the earlier offsets.  This, for example, permits a Q-band 
referenced pointing differential offset from another band (such as 
X-band).  This is a very useful capability when one is trying to get the 
very best pointing, and is worried about residual collimation offsets. 
       However, the problem is that this useful feature can *only* apply 
the pre-existing solution, and does not allow one to  use a single 
reference pointing band, such as X-band, as reference for all others.   
So, for example, in my case, the X-band solution is used for Ku-band, 
the Ku-band solution was used for K band, K-band solution for Ka-band, 
and Ka-band for Q band.  Not a problem is there is not a  large 
collimation offset between those pairs of bands -- but a *big* problem 
if there is.  (And there was!). 
    Is it possible to have the OPT remember a particular solution (not 
just the existing one), and apply that one to subsequent observations?  
That is, develop a true 'secondary referenced pointing' capability, like 
we used to have?