[mmaimcal] some background on the central element question

[Harvey Liszt]

It shouldn't be too surprising that the French are considering the issue of a 
dedicated central element (CE).  It shows they're being thorough, they didn't 
partake of the debate (such as it was) here ca. 1990, and not everything is the
same as when the CE was dropped back then.  The mma has many more and somewhat
bigger antennas now (the usual 8 vs. 11+/-1m), somewhat longer maximum 
baselines, and will be at a much better sight (we were still considering 
eastern AZ or central NM back then), with a much much higher maximum frequency
(before it was probably 1mm, and even 345 GHz wasn't seen as a workhorse freq-
uency). On the other side, we now make singledish maps on the fly which 
undoubtedly produces better sampling, and the scale of the array has grown to 
the point where a single big singledish CE might be a much smaller fraction of
the overall effort.

As historical background, let me remind you that there was a meeting at the 
AOC in June 1990 at which HSL moderated or brokered between Tim Cornwell and 
Mark Holdaway one one side -- the array people arguing that no CE was needed 
-- and Darrel Emerson and Phil Jewell from the 12m -- the singledish people 
who argued for it.  Perhaps someone from the meeting has the mou which I 
cranked out as we were to break up and go our respective ways -- I don't -- 
but it was concluded that the CE was no longer wanted.  The discussion took 
several guises:

In favor of the CE were three arguments, the last of which was unvoiced 
(i) It gives you what you need in the most direct way; (2) there isn't any 
other proven way to do it; (3)  after you build the central element you have 
this really neat single dish to play with.

Against the CE we had (i) the claim that an array could be contrived to reduce
the problem of mosaicing to either a linear deconvolution or a tractable non-
linear one (ii) the fact that the singledish is really a separate instrument 
requiring a very substantial dedicated effort which only marginally overlaps 
the mission of the array.

The CE must be of order 2 - 2.5 times the size of the array elements to give 
enough overlap in the uv plane to integrate single dish and array datatsets.  
This would now seem to be a formidable problem, with a 25-30m dish required 
to work at frequencies as high as those of the array antennas, with comparable
pointing, etc.  Before, the CE could have been a 20m dish working to much 
lower frequency; the CE is now rather more difficult to achieve, if perhaps 
a smaller fraction of the much-increased scope of the whole MMA.

When it is considered how the CE works, it rapidly becomes apparent that the 
regions of the uv-plane it samples are very limited (in the limit, one cell) 
and that it can't take long to do this.  So you explore mapping wide fields 
with the CE using it like a single dish in its spare time.  But for really 
wide field mapping it can't work nearly as fast as the 40 or so individual 
elements with their larger field of view.  This is a paradox; it does its 
array-related chores too fast and its single-dish uses too slowly.  So you 
imagine putting array receivers on the CE, perhaps.

The fact that the CE works so much faster on its array-related chores presents
a conundrum.  If the CE doesn't observe simultaneously with the array, what 
about calibration?  But if it does, most of the data it takes is vastly 
redundant.  To relieve the redundancy, it was suggested to outfit the CE with 
array receivers, for example.  But the escalation in the effort required to 
maintain its usefulness is again apparent. 

The small field of view of the CE is a problem as well; if you want it to 
observe with the array for calibration purposes, you are left to consider 
just what it is that the central element sees when it points in the same 
direction as the other antennas.  The answer of course is "not much", so 
again, perhaps, you imagine it to be equipped with an array feed.

So the central element winds up being a ding an sich, if you know what I mean.
  

But beyond this, I think the arguments pro or con haven't been supported by a 
good flow of mosaiced maps either with or without single dish data added, in 
either regime of linear or non-linear deconvolution.  And if this is so, and 
if the logistical arguments do not make the central element singledish 
unattractive, the scientific case will need to be revisited in more detail.

regards, Harvey