[evlatests] Modcomp-Free Test, Friday night.

[Rick Perley]

    One hour of dynamic time, 06 - 07 LST (approximately 11 PM -- 
midnight) were used to stress the Modcomp-free system. 

    Most of the time went to checking each band's performance, in 
continuum, with 0.418 seconds integration/dump time. 

A reference pointing was done at X-band, and the results applied to 
subsequent X, K, and Q band scans.  To check if referenced pointing is 
working, observations of the same source were made without referenced 
pointing applied.  The results of this exercise will be reported later. 

    The last part of the run was used to observing a number of 
closely-spaced sources at L-band, to potentially help resolve the VLA 
Baseline coordinate frame rotation problem.  These results will be 
reported later. 

    Here I report on the basic data quality results from the continuum 
observations, for each band.  This report is lengthy, because of the 
many issues uncovered.  I have organized it by order of seriousness, 
with the most worrisome/critical issues first. 

    A)  Pulsing Data. 

    At all bands and all IFs and all sources, at all times, data from 
most VLA antennas is 'pulsing'.  There are three very clear behaviors:

    1) Antennas which pulse 'upwards'.  These give single (0.417 second) 
records of strong visibility, with either 1 or 3 following records of 
weak, or dead, visibility.  The antennas which alternate high and low 
are:  1, 12, and 27.  The antennas which have 3 downs for every up are:  
2, 7, 9, 20, and 22. 
    2) Antennas which pulse 'downwards' -- i.e., they drop out.  These 
are 3 and 28.  For both, there are three good records, followed by one 
bad (dead, or weak) record.    
    3) Antennas with don't pulse notably:  These are 4, 5, 10, 15, and 25. 

    It should be noted that all EVLA antennas behave normally -- no 
pulses of any kind. 

    This is (obviously) a serious problem, which requires quick 
resolution. 

    B)  VLA antennas 'late to get started, and early to leave'. 

    Gustaaf has reported that the initial scan of a Modcomp-free file 
has no VLA antennas on source, while subsequent observations get them, 
as expected.  Because of this, I constructed my observation with a 
5-minute 'dummy' observation, followed by a 2-minute observation of the 
same source at the same frequency.  (The significance of this will be 
explained later).  The results are in complete accord with Gustaaf's 
result.  A time log of events is useful here:

    05:07   Operator reports run begins.
    05:10:30  Operator reports switchover to Modcomp-free status complete.
    05:10:42   First correlator data found in database -- it has not 
been flagged, but is only noise. 
    05:11:33   First good data appears in database -- EVLA antennas 
only!  All VLA antennas give only  noise, and are not flagged.
    05:12   Operator reports all antennas on source  (close enough!)
    05:15:00   VLA data abruptly ends (has been flagged as bad) , 2 
seconds before end of scan.
    05:15:10   End of first scan. 
    05:15:11   Good EVLA data only on 2nd scan (same source, same 
band).  VLA data present, is bad, but has not been flagged.
    05:15:22   Good VLA data appears.   
    05:17:20   VLA data abruptly ends, 10 seconds before end of scan -- 
flagged as bad, but are actually good. 
    05:17:30   End of scan.  Next scan is at a different frequency. 

    For all subsequent observations, the VLA antennas appeared normally, 
in synchronism with the EVLA antennas. 

    All that Gustaaf has reported is repeated here. 

    My hypothesis is that the VLA antennas are actually at the wrong 
frequency in the initial scan.  Is this possible? 

    C)  VLA antennas flagged bad 10 seconds before the end of a scan.

    This happened on a number of observations, but not on others.  In 
all cases where it happened, the next scan was at a different 
frequency.  In all cases where there was no frequency change, there was 
no early flagging.  In all cases, the data flagged as bad were not. 

    D)  Every scan had a start time, and a single record, belonging to 
the previous scan.  In all cases, the datum was bad, but was not flagged. 

    E)  Referenced Pointing turns off flags? 

    Antennas 19 and 21 have no Q-band receivers.  As noted above, two 
observations were made at this band, the first without the pointing 
corrections applied, the second with the corrections applied.  
Remarkably -- for this first scan, the data were properly flagged for 
the first scan.  But they were not flagged for the second scan.  (This 
test cannot be applied at X and K bands, since all EVLA antennas are 
outfitted with receivers). 

    The following items are of a lesser concern (I hope).

    Antennas 6 and 8 gave no useable data at any band -- but were not 
flagged.  Could be delays, but the error will have to be really large! 

    Antenna 13 only worked well in IF A.  13B was dead, 13 C was weak, 
and 13D was very, very weak.  None of these were flagged at any band. 

    Antenna 5 was dead at K and Q bands -- and not flagged.  Delays?  
Pointing? 

    At P-band, only EVLA antennas 13 and 16 failed to fringe (and 
neither should).  The data were not flagged. 
    Also at P-band, a number of VLA antennas gave very poor SNR.

    In general, flagging at the beginning of scans is nearly right 
(ignoring that 'lost' first record problem, and the troubles with the 
first two scans reported at length above). 

*************************************************************************
    Conclusion:

    We've got a lot of work to do before we're ready to turn off the 
Modcomps.