[mmaimcal] Re: references on how many alma antennas should do singledish work

[Al Wootten]

 We will discuss these at today's ImCal meeting 4pm EST (804) 296-7082; just
dial in.

Harvey S. Liszt writes:
 > Mel Wright asked me about singledish work with ALMA but I didn't save
 > any of the recent mail traffic that discussed how many should work as
 > singledishes after your short message from Munich.  Can you point him
 > to some of this stuff?
Well, here are those discussions, along with my notes from the System 
and Test Interferometer PDR.
First a summary of extracts of the discussions of the Imaging and Calibration
Group on this subject:
Steve Myers: It seems ludicrous to use one antenna in single-dish mode.  
Why would you add in single-dish data with much worse noise level if your 
desire is to actually improve the images?  My guess (purely a guess) is 
something like sqrt(N) = 8 for N=64 antennas is the minimum...
Butler: i don't know about 8 being the limit, but i agree that 1 is *way*
too small.  how does stephane think that the data from only 1 antenna
will be used?  if you only have 5 antennas in your array (ala IRAM
or OVRO), then adding in data from 1 dish *might* make sense.  if you
have 64 antennas, it makes no sense (to me at least).
Rupen: On the single dish end: trying to support an instrument as sensitive and
flexible as ALMA with one 12m antenna will be difficult if not impossible.
Mark & I wrote a few words on this back in 1995 (memo 128, p. 7), pointing
out that if you want to make good mosaic'd observations you would have to spend
about 1/4 your time doing total power measurements -- assuming ALL the 
then-40 antennas were useful in those measurements.  Going with only one
single dish is equivalent to the assumption that only a few per cent of
interferometric obs. will require total power data, even under the assumptions
that (1) the single dish is working in total power mode all the time, 
(2) the weather is fine for SD work at the relevant frequencies to match
those interf. obs. (this is not at all clear -- the SD after all will be
MUCH slower than the full array, by an amount which depends on the number
of dishes [is this still 64? or has inflation set in again?], so finding
good weather for high-freq. mosaic support may be difficult), and (3) there
are no problems with cross-calibration, meaning that there are no
benefits to obtaining SD and interf. data simultaneously.  This last seems
quite an assumption, given variable weather; I would think it would help
considerably to have SD and interf. data taken at the same time, given
changes in the atmosphere as well as niggling details like pointing etc.
where calibrations could neatly be applied to both data sets at once. And
of course I'm unhappy with the idea that the array will only seldom do
mosaics; at a minimum I'd like to see that stated explicitly, with a real
scientific justification.
---end ImCal discussion----     
There are a lot of good points here.  I think they are to some extent
included in Darrel's summary, which he sent to me privately and which I will
distribute if he agrees.  I summarize all of these points, I think, in
my written response to the PDR/CDR below.  We will discuss these at today's
ImCal meeting 4pm EST (804) 296-7082

This is not the official report mind you--I already have discovered that
others heard some things differently than I, and I may have misinterpreted
some remarks.  In any event, I think Darrel has done a fuller job of
trying to determine the number of total power antennas needed.

ALMA System PDR   Garching 28 February 2000
Outside reviewers: Dave Woody, Richard Hills, Y. Chikada and B. Anderson
ALMA reviewers: Napier, Payne, Sramek, Webber, Glendenning, Wootten, Wild, 
Baudry, Anderson, Raffi, Guilloteau, Baars
Summary of afternoon discussion.
There is a major problem with the schedule for the digitizer which needs to 
be addressed. The critical parameter is the 2-4 GHz passband sampling. 
Baseband (0-2 Ghz) sampling requires large changes in the IF system and does 
not advance the project, and should not be a goal.  Woody proposed having a 
1 bit digitizer for tests and fallback position; a second fallback could be 
2 bits, with of course 3 bits being the goal.  Tests occur by the end of 2001; 
close collaboration with Japan is indicated.
The photonic LO scheme looks promising but is not yet solid.  Ongoing study 
of photonic phase calibration should continue.  We may want to state that 
our goal is a fully photonic system.  We recommend that noise and stability 
be demonstrated as soon as possible.  Note that a meeting in the Netherlands 
on photonics is scheduled for 8 Mar 2000.
How many ALMA antennas will be needed for total power antennas?  Stephane 
proposed 1-2.  
I think that this is a substantial underestimate.  We didn't go into this 
at the meeting but a digression is perhaps worthwhile at this point.
     I don't think there is a single answer to this.  In part, the correct 
answer depends upon how ALMA is to be used, and of course we don't know 
that yet.  I like the concept Stephane suggested in August at URSI, that 
we target ALMA to provide images similar to those which optical astronomers 
have obtained for a century to the casual observer.  This suggests that a 
typical image will span several arcminutes at a resolution of perhaps just 
under an arcsecond
to a few arcseconds.  While some projects may be satisfied with a 30 x 30 
image, I think this standard will fail to excite ALMA users.  Most projects 
will involve mosaicing.


     Welch has just written a memo to the ASAC on this subject, concluding 
that all of the antennas should be outfitted for total power observations.  
Emerson has provided a list of factors which suggest this conclusion.  
They include:
1) a factor of two in observing time in favor of the interferometer for 
the larger collecting area and from the cross correlation process.
2) Total power switching: One needs to switch out variations in the background 
system and atmospheric noise.  Detector noise may also need cancellation, 
as is accomplished  with rapid switching between horns on the GBT Q band 
receiver.  For position switched scans, the single antenna loses by a 
factor of two; by root 2 for frequency switched data, and by a smaller 
factor for data sampled 'on the fly'.  In the latter technique, M 'on-source' 
scans are obtained for evey 'off-source' scan.  The degradation factor may be 
brought down to say 1.1 where M=100, for a penalty of 1.2 in observing time 
for the single antenna.
3) Antenna spatial frequency response: The spacings we are interested in 
sampling to stitch the single antenna data to the interferometer data are 
of the order of the antenna diameter, and attenuated by a factor of perhaps 
3 by the primary beam, depending upon how closely the array antennas are 
packed and upon what the antenna illumination taper is.  Darrel estimates 
that this is a factor of 3, for a degradation factor of 9.  He points out 
that this is an area where an array of smaller antennas can help a lot.
4) Redundant short spacings in the array: The array when closely packed 
inevitable oversamples some short spacings, requiring more total power 
integration time to match sensitivity.  This depends upon the geometry 
of the densely packed array, but Darrel estimates it to be 2-4.  Typically, 
this may perhaps be 3, suggesting a factor of 9 in observing time.
***************
NB:  We should, at least, know this number for our strawman array!  Min?
Leonia?
***************
5) Systematic effects: As Holdaway's and Emerson's simulations, it was shown 
that systematic errors are substantially reduced when one averages total 
power observations taken with many antennas, as opposed to those taken 
with only a few.  This affects both pointing errors and sidelobes, at least.  
The observing time factor is hard to estimate, but Darrel estimates it may 
range from 1 to 4.
At this point, the 'Total Power Observing Time Factor' is the product of 
these uncertain factors, but is certainly large, arguably very large.  
Darrel points out that perhaps we don't really want to equalize the noise 
on these various projects; in some projects the signal to noise ratio is 
what we want to optimize.  In this case, the factor is lessened.  Perhaps 
our goal of providing optical quality images will fail, and mosaicing will 
not be so important in array science.  In this case, the factor is lessened 
also.  But in the end, it appears to me that we will probably need to use 
all of the antennas in total power mode much of the time.
----end digression
The receiver baseline was modified.  Receivers may be either of
 - 8 GHz bandwidth, upper and lower sideband, both polarizations, separated 
SSB (old standard)
 - 8 GHz bandwidth, single upper sideband, dual polarization
 - 8 GHz bandwidth, single lower sideband, dual polarization
 - 4 GHz bandwidth, both upper and lower sideband, dual polarization, or
 - 8 GHz dual polarization, double sideband, Walsh function (for example) 
separated.
In all cases, 16 GHz of bandwidth to be processed.  Care should be taken 
to ensure that restrictions on the LO don't result in inaccessible frequencies.
The question of whether the band 3 device should be SIS or HEMT should be 
based on sensitivity and total power performance.
Amplitude calibration has not yet been addressed a two temperature load?  
To some extent this depends upon antenna design.
The implications of the mismatch between timescales in the correlator and 
in the M&C system should be re-examined.
The number of antennas will be 64.
Band 1 will be discussed by the ASAC in Leiden 10 March 2000.
We adopt a requirement to switch in 1.5 sec between any two bands, when 
receivers are already tuned.  
We adopt a requirement that 10sec be allowed for retuning, a change of more 
than 0.1% in frequency within a given band.
Frequency switching by 0.03% in 1 msec is a goal, with a 10msec specification.
There will be at least four subarrays.  Apparently I didn't interpret this 
quite correctly; I understand that a point of view that there should be 
exactly four subarrays was adopted.  I append several examples of the need 
for more than four subarrays; a good argument may be made for eight.
The current baseline is that the FIR filter should be located in the 
correlator building.  Hills noted that the site demands keeping items in 
the antenna to a minimum, and that siting this in the antenna may limit 
future correlator upgrades.  The perhaps $3M cost of the central location 
is offset by the gain in flexibility.
ALMA Test Interferometer CDR   Garching 29 February 2000
Outside reviewers: Dave Woody  and B. Anderson
ALMA reviewers: Napier, Payne, Sramek, Webber, Glendenning, Wootten, Wild, 
Baudry, Anderson, Raffi, Guilloteau, Baars
Summary of afternoon discussion.
There is no need to instrument the test interferometer with WVRs.
The 230GHz receiver is important as the highest frequency receiver usable 
at the site.  This justifies its inclusion, despite its complexity.  It 
will be useful for measurements of the non-intersection of antenna axes. 
If the 1mm OMT is a driver, it is more important than dual polarization.
Dual polarization is useful in antenna testing as it offers a reality check 
antenna misbehavior will appear in both channels.  Also, SiO masers will be 
polarized.  
There appears to be no good reason to support a 3mm SIS receiver on the 
test interferometer, and it should be dropped.  This is a 15 year old design, 
with tuners!
The HEMT receiver at 3mm will be useful to test aspects of the 8 GHz IF.
Best pointing measurements should be achieved at 30 GHz.  Note that this is a 
prototype receiver, not an evaluation receiver.
There should be a trailer an intermediate distance from the antennas, with 
easy access to them.
Funding is a concern, given the US FY2001 budget and the problems with 
manpower in Tucson.  
The antennas should be independent, as planned.  They should be tested in a 
properly unified test plan.  A group will work to establish this by early 
May, to be ready for the expected June PDR. 
Representatives of ALMA divisions should participate.  For Calibration and 
Imaging, I'd nominate Jeff Mangum as 'Prototype Antenna Scientist'.  The 
calibration system and optics needs to be planned also.  Testing may require 
other resources.
There should be an optical telescope in each antenna.
There should be nutators on each antenna.  The contractors need a dynamically 
balanced design.  CSO, SEST, SMA designs should be examined.  It was suggested 
that lacking a subreflector, SiO maser pseudo-continuum cross scans might be 
used.  I'm not sure this will work for a 12m telescope.  Nutators need not 
chop 3 beamwidths at 30 Ghz.
The compressor should run at 60 Hz.
Correlator: the time synch needs checking.  Faster (20 Hz) is better for 
SiO pointing.  Two autocorrelator modes should be supported to allow 
independent antenna tests simultaneously.
Emerson and Baars will create a shopping list of items to be shared with 
European collaborators.

Clear skies,
Al