[mmaimcal] Re: THE NECESSITY OF N > 4 SUBARRAYS FOR ALMA

[Steven T. Myers]

It is clear that this will be a difficult matter to resolve.  I will
try my best to make the issues clear...

> The fact that someone can invent a way to use many subarrays does not
> justify their construction.  Although each of the functions in Steve's
> list of 8 will sometimes be needed, most will be needed only a small
> fraction of the time.  They can therefore all be accomodated with,
> say, 3 subarrays by executing each of 6 of Steve's functions
> sequentially rather than simultaneously.  The scheduler might have to
> think a little harder, but that is the only cost.

Actually, to implement these modes, the scheduler needs to think harder.
My guess is that the current subarray capability is not fully used
on the VLA because of the difficulties involved (mea culpa, I should
be using subarrays for my calibration tests). The real cost is the fact
that you would need to waste the whole array for something that is best
done by a smaller subarray.

Again, there are a base number of subarrays that will be routinely used:

1. Examination of nearly all of the configuration plans have antennas
   being moved at far greater frequency than the VLA.  It is an
   inappropriate waste of time to spend the whole array for even an
   hour every day or two to do baselines.  It is likely that there
   will be at least one subarray at a given time for those antennas
   that have moved or are helping with baselines.  It may be that this
   will be changed if a more slavish VLA-style move schedule is adopted,
   but that does not appear to me to be likely.

   FOR THE CONFIGURATION GROUP - What are the implications of the various
   designs for array operations, and do we need routine use of a baseline
   subarray?

2. Unlike the VLA at most bands, atmospheric opacity is a significant
   contributor to calibration.  It may be that use of the WVR channels
   can provide an effective atmospheric emission/attenuation calibration
   in addition to its designed phase calibration role, but this is not
   clear.  It may be that mere monitoring of the total powers in the
   antennas will do the trick.  However, past experience has shown
   that tipping curves are the best way to calibrate the atmosphere in
   the observing band.  It seems reasonable to detach a telescope to
   do this.  On the other hand, should there be a smaller dish with
   all the same bands to do this?  As far as I know, this has not been
   considered.  I will try and look into the question of atmospheric 
   emission/absorption calibration - it may be the WVR will be sufficient.

   NOTE FOR THE ImCal GROUP - do we have an idea of what is really needed
   in overhead for proper calibration?  Do we need a good estimate of
   opacity or is total power and/or WVR OK?

3. The main observing subarray.  

4. Total power scanning.  The current favored design is that only a few
   (1-16?) antennas be outfitted (with nutators etc) for single-dish
   optimized observing.  It is ridiculous to not operate this as a
   subarray and waste the rest of the antennas!  Again, the science
   advisors for this project estimate that a significant number of
   project will require some single-dish input, due to the constrained
   field-of-view.  It may be that at some point a dedicated single dish
   is adopted, but we cannot (are will not) count on this!

   NOTE FOR THE ASAC - Is the baseline design to use a number (TBD) of
   "enhanced" ALMA array elements to do the single-dish scanning?  If
   so, then should this clearly be designated as a fundamental subarray?

5. VLBI - contrary to what some people think, the VLA participates
   (usually with a single antenna) in a large fraction of VLBA experiments
   (I will try and get this number from Ops).  This is a (mostly
   transparent) subarray on the VLA now, and would be on ALMA.  With
   the small 12m antennas, it might be more common to phase together a
   few antennas for VLBI.  This may not be a high percentage of the time,
   but must be done without disrupting the normal ALMA operations.

   NOTE TO ME - try to get the stats from VLA ops, and also estimate the
   mm usage.  I doubt if an accurate assessment can be made, but I will
   try.

> I think the VLA experience (reported by Dick Sramek at the recent ALMA
> Systems Design PDR) is telling:  the 5-subarray capability has never
> been used, and 4 subarrays are operated only on rare occaisions.  Yet
> all of Steve's functions are equally applicable to the VLA.

NO. (As far as I have heard, this was an anecdotal comment made by Dick
at the end of the meeting.  Do we have real statistics?) In fact, ALMA
as envisioned is signifcantly different than the VLA.

Because of the quarterly moving schedule, 1 is not relevant, and recently
moved antennas can be place with sufficient accuracy that they can be
instantly integrated into the array and have the baselines corrected
later.  THIS IS NOT THE CASE WITH ALMA.  Are we going to be happy with
using the entire array for baselines at every move?  (Note: the OVRO array
does this, but uses anywhere from several hours to a full day to do
baselines after any move.  Are we willing to do this?)

Most observers are happy to ignore point 2 on the VLA, even at K and Q
band, and it probably affects the data at the 5% level.  I do not think
we plan to build this shortcoming into ALMA.  It may be deemed acceptable
for the entire array to spend a few minutes (with fast moving antennas)
doing a tip curve every few hours.  We should ask the ASAC this.

There is currently no total power capability on the VLA, so point 4 is
moot.  Furthermore the field-of-view is not as limiting at the VLA.
Note that this is not a scale-independent problem, there are fundanmental
absolute angular scales associated with the science goal and target
objects, such as: arcminutes for cosmological volumes such as clusters of
galaxies, sub-mm bright galaxies, and the ISM; arcseconds for "deep
fields", distant galaxies, molecular clouds complexes in nearby galaxies,
the nearest star-forming regions; 0.1" for structure of galaxies at
cosmological distances, galactic structure in nearby galaxies, the cores
of nearby molecular clouds, and so on.  This is why the ability to mosaic
ALMA observations together is so important, especially in comparison to
the case for the VLA.

Thus, we are left with points 3 and 5 as "normal VLA-like ops".

There is a point 6 where not all antennas may have all bands functional
(likely in the early stages of ALMA, and almost certain given the
funding). This was the case with the VLA until this past year at Q-band.
Thus, a significant (>10%?) of the time the VLA operated routinely with
3 subarrays.

> If ALMA were to require 8 subarrays most of the time, it would follow
> that most observations require far less collecting area than is
> provided by the full array.  If that's true, how can we possibly
> justify to the funding agencies the enormous expense and operational
> difficulties of the huge 64-element array?

Sigh.  The justification is that it is not economical to use a LARGE
number of elements for these important ancillary tasks (eg. baselines,
mmVLBI, tip-curves, perhaps single-dish scanning) and thus the only
practical way of doing this is through subarrays. 

My case is that we use 4 subarrays in most cases.  The minimum number of
subarrays is 4.  I do not see a need for more than 8.  From the above
arguments, it may be that we could live with 4, and that 5-6 would be a
goal.  The costing of this should be presented to the ASAC.

> As a matter of design philosophy, applicable not only to this point
> but also to any suggestion of a costly enhancement to the telescope,
> we should always ask whether the proposed enhancement enables some
> science that would be impossible without it.  If so, then the value of
> that science (including whether it should be done with this telescope
> as opposed to some other) can be weighted against the cost.  If the
> enhancement merely makes some observations more convenient or faster,
> where the observations are still possible without it, then building
> the enhancement is far harder to justify.
> 
> --Larry
> 

This is true all the way through the project - the ONLY justification is
in the science!

However, the cost-benefit in the end is for the ASAC to
decide. I can only present the case as best as I
am able.  Given that there already exist operating mm and submm arrays,
what do you consider threshold for ALMA's existence?  The fact is that
these subarray-using operations do enhance (either in calibration
capability, or in the case of single-dish scanning, in image fidelity)
the science.  Different people might weight the benefits versus cost
differently however.  IMHO, poorly calibrated or understood data from
ALMA is worse than no data, given that fundamentally the justification
for ALMA is that it will do the science FAR BETTER than the current arrays
(even upgraded) will acheive.  This is the case that we present to the
decadal panel and the NSF in any event.

It appears that we disagree on the importance of these uses of subarrays.
I still feel that the VLA is a poor example in this regard (and even then
it routinely operated in 3 subarrays, though the observer was only aware
of 2).  It may be that we can engineer workarounds for some of the issues
such as opacity correction, and live with reduced efficiency for
baselines, but we will likely be also routinely operating with 3 subarrays
(VLBI, mm-Int, mm-singledish) with an additional subarray or two if not
all antennas have the full complement of recievers.  I see no good way
around this.

I strongly urge the careful consideration of what it will mean to operate
this array, and not just assume it will "be like the VLA".  It will not.

A final plea from those of us who have been advising ALMA on the
science side -  we often feel that little or nothing we say (eg. in 
Michael Rupen's Memos etc.) are taken seriously during the final design
stages, and we are left making the same arguments over and over to no
avail.  We do believe that these are compelling reasons for having at
least 4 subarrays, and that the added benefit from several more outweigh
the cost.  It is clear that we have not convinced you, nor have you
convinced us.  All we ask is that these issues be treated in an
unprejudiced manner through the design stages, and that a clear and
objective presentation of costs and benefits of the alternatives be made
to the advisory committees in order to arrive at a fully-informed
decision.

  -steve