[evlatests] how often do we observe in narrow bandwidths?

[Michael Rupen]

Hello all,

   a couple comments on Bryan's numbers, and his conclusions.

* The July 1 cutoff for discussing recent observations is a rather odd
   one.
   - This corresponds to sometime in A configuration.  The distribution of
     projects (esp. line vs. continuum) varies enormously with the array
     configuration, since surface brightness sensitivity prohibits
     most emission line work in the larger configurations, and emission line
     projects are much more time consuming in general than absorption work.
     Bryan's July 1 cutoff will likely show many more continuum observations
     than an average over all configurations.  In this context I'm surprised
     that we don't see a substantially *smaller* fraction of observing hours
     in the narrow-band modes.

   - The two large proposals affected by aliasing are those of Ott and
     Hunter.  Both projects center on HI in dwarf galaxies, and use the 0.78
     and 1.56 MHz bandwidths exclusively.  Ott's observations began in late
     November (BnA); those of Hunter began on Dec. 19th.  The fraction of
     observing hours devoted to those projects will become even higher as
     B configuration continues, with the bulk of Hunter's observations at
     least scheduled for January.

* Comparing fractions of *total* observing hours mostly reveals the
   distinction between continuum and line observations, and shows that
   most VLA observations -- esp. in the A and B configs. -- are continuum
   ones.

   - BWcodes 0 and 1 (50 and 25 MHz) are primarily continuum experiments.

     2003-now:
     > BWcode   all observing        science observing
     > 0        787.5  (62.4%)        433.3  (50.9%)
     > 1        138.4  (11.0%)        122.7  (14.4%)

     July 1-now:
     > BWcode   all observing        science observing
     > 0         90.6  (58.2%)         41.5  (42.3%)
     > 1         30.9  (19.9%)         27.4  (27.9%)

     The increase in BWcode 1 corresponds to the use of pseudo-continuum line
     mode to take care of BW smearing in the A & B configurations.

     These continuum observations account for about 65% of all observations,
     and about 70% of the recent ones.

   - Taking BWC 3 and up (6.25 MHz and narrower) as line experiments, I've
     added columns corresponding to the fraction of line data taken in the
     narrower BWCs.  The tables are then as follows:

     2003-now:
     > BWcode   all observing        science observing   % of line expt
     > 3         68.4   (5.4%)         59.5   (7.0%)         21.4%
     > 4        101.7   (8.1%)         91.5  (10.7%)         32.9%
     > 5         99.0   (7.8%)         95.2  (11.2%)         34.2%
     > 6         29.3   (2.3%)         28.0   (3.3%)         10.1%
     > 7          0.1   (0.0%)          0.1   (0.0%)          0.0%
     > 8          0.8   (0.1%)          0.5   (0.1%)          0.2%
     > 9          4.0   (0.3%)          3.2   (0.4%)          1.2%

     July 1-now:
     > BWcode   all observing        science observing    % of line expt
     > 3          7.9   (5.1%)          6.6   (6.7%)          23.6%
     > 4          2.7   (1.7%)          1.7   (1.7%)           6.1%
     > 5         11.6   (7.5%)         11.1  (11.3%)          39.6%
     > 6          8.8   (5.6%)          8.4   (8.6%)          30.0%
     > 7          0.0   (0.0%)          0.0   (0.0%)           0.0%
     > 8          0.0   (0.0%)          0.0   (0.0%)           0.0%
     > 9          0.3   (0.2%)          0.2   (0.2%)           0.7%

     I read this as a major increase in the fraction of time spent at BWC 6
     (0.78 MHz), at the expense of BWC 4 (3.125 MHz).  BWC 5 (1.56 MHz)
     also increases somewhat.


* Conclusions
   Bryan's conclusion was as follows:
     > so, the question is, do we want to go to much effort, including a
     > potential many-months-long delay in EVLA, to make a change to fix
     > something that might affect something like 10%-20% of observing in the
     > transition system (i'm counting modes 5 and above here)?  i leave it to
     > those more erudite than i to decide.

   I have several issues with this.

   1- The "potential many-months-delay" seems more a boogeyman than a
     reality.
     * I have not seen the cost analysis of this, and my impression
       from speaking with Bob and Mark yesterday is that such an analysis has
       not been done yet.  There are several possible avenues being explored,
       from hardware to software; some of these would likely be rather easy
       to implement (cf. Eric's ridiculously rapid modification of UVLSF).

       I presume this very long delay estimate comes from Mike Revnell's
       early off-the-cuff guess as to the time required for a hardware fix.
       He had revised that downwards by an order of magnitude when last we
       spoke.

     * Clearly we cannot tolerate a huge delay in the EVLA, and that is
       being factored into the discussions of possible approaches.
       Significant delays in EVLA developments are NOT on the table.

     * The nature and severity of the problem are still being analyzed. It seems
       early to panic about a fix, when we don't have a clear understanding
       of the impact of the problem on scientific observations. Taking the
       worst case, which has NOT been demonstrated, we might decide that the
       problem isn't fixable in post-processing.  [Note that the primary
       responsibility for gathering the data leading to that decision has
       been given to the *observers*, not NRAO staff.]  One -- extreme! --
       approach would simply be to allocate more observing time to make up
       for the increase in the noise.  This would have a major impact on the
       array configuration schedule, for instance, but not on the EVLA
       project.  I do not advocate this, but even this worst case need not
       lead to an EVLA meltdown.

     This is a serious problem, and we should take it seriously.  We cannot
     allow a transition issue to dominate our thinking; but neither should we
     stop thinking, on the off chance that all possible approaches might lead
     to significant delays in other projects.

   2- It's not obvious that the major impact of even the most pessimistic
     version of a fix would be on the EVLA.  Mike for instance is primarily
     involved in VLBA developments at the moment.  Those developments are
     of course also very important -- the point is that we should not focus
     exclusively on the EVLA when considering what to do about the aliasing
     problem.

   3- Counting the impact purely in overall observing hours is rather naive.
     * The extreme of this approach would be to say the transition system
       is purely a continuum instrument: spectral line observations are not
       allowed.  That might be a fine policy, but it is not the current
       one.

     * We have made a special commitment to the Large Proposals, to the
       extent of (for instance) changing the array configuration schedule.
       Those proposals have been subjected to much more extensive scientific
       and technical reviews.  Further, the PIs have committed to spending
       much of their time over the next several years on these projects,
       have gotten NSF and other grants to support them, and have hired
       graduate students and postdocs with those moneys.  We have a duty to
       at least try to mitigate major, unanticipated, unadvertised
       instrumental problems which have a major impact on these Large
       Proposals.

Cheers,

            Michael