[evlatests] First full bandwidth data!

[Rick Perley]

    Seven antennas are now outfitted with the full complement of 3-bit 
samplers.  I prevailed upon Ken to generate a script to see if we are 
getting good data across the full 8.192 GHz of bandwidth now available. 

    He ran a short test script this afternoon, which worked (almost) 
flawlessly.  All 16384 channels for each baseline are present (4096 is 
each of four polarization), for all 21 baselines. 

    A few things are clearly not right.  First, some nomenclature:
   
    According to BDFLIST, the data are arranged in four 'chunks' (IFs) 
with 16 subbands each.  The four chunks are each 2.048 GHz wide, and are 
listed in reversed frequency order:  the first is the highest frequency 
(24.0 -- 26 GHz), the second is the next highest (22 -- 24 GHz), etc.  
The subbands within each of these four chunks is within proper frequency 
order, however.  In any event, BDF2AIPS will (if requested) resort the 
order of all the subbands, so all comes out as expected.    I number the 
IFS from lowest frequency
to highest. 
    Each IF (chunk) has 16 subbands of 128 MHz each, numbered from 1 
through 16. 

    1) On the RR correlations, four subbands are giving no signal on 
most of the antennas:

    subband 4 on IFs 1 and 2
    subband 13 on IFs 3 and 4. 

    2) On the LL correlations, most antennas also have four subbands 
with low signal, or unusual shape:

    subband 1 on IFs 1 and 2
    subband 16 on IFs 3 and 4. 

    3) Delays are very, very large -- as expected.  The AIPS task 
'FRING' cannot do a multi-subband delay solution for these data, since 
it doesn't (yet) know about four IFs being present within a single 
database.  (It knows about 1 or 2 IFs).  No big deal -- FRING will work 
on each subband separately. 

    4) Some IFs have powers that are wrong by up to 5 dB.  Notably:  IF 
1L, on ea 17 is low by -5 dB.  IF 3L on ea 25 low by -4 dB. 

    5) A very large spectral power slope is seen on ea21, LCP, on 
subband 2 (=#3 in the original ordering).  The slope is about 15 dB 
across the 2 GHz.  Bad slope filter value, evidently. 

    Ken's observation was on 3C84, whose flux is not exactly known.  He 
plans to run this test tonight on 3C286, after which I can derive 
sensitivities, etc.