March 16, 2001 by S.Guilloteau
Recommendations of the Configuration Preliminary Design Review
The Configuration PDR was held in Grenoble on 26-27 Feb, 2001.
Attendants were
ANTERRIEU Eric.Anterrieu@cerfacs.fr
BRONFMAN leo@das.uchile.cl
Butler bbutler@aoc.nrao.edu
Conway jconway@ebur.oso.chalmers.se
Cox Cox@ias.u-psud.fr
Emerson demerson@tuc.nrao.edu
Gie-Han Tan ghtan@eso.org
Gueth gueth@iram.fr
Guilloteau guilloteau@iram.fr
Heddle heddle@totalise.co.uk
HOFSTADT dhofstad@eso.org
Holdaway mholdawa@nrao.edu
Ishiguro ishiguro@nro.nao.ac.jp
Kogan lkogan@aoc.nrao.edu
Morita morita@nro.nao.ac.jp
MUNDY lgm@astro.umd.edu
Pety pety@iram.fr
Radford sradford@nrao.edu
Sakamoto seiichi@nro.nao.ac.jp
Viallefond fviallef@obspm.fr
Webster asw@roe.ac.uk
Wootten awootten@nrao.edu
WRIGHT wright@astron.berkeley.edu
YUN myun@bonito.astro.umass.edu
Reviewers were
ANTERRIEU Eric.Anterrieu@cerfacs.fr
Emerson demerson@tuc.nrao.edu
Gie-Han Tan ghtan@eso.org
Guilloteau guilloteau@iram.fr
HOFSTADT dhofstad@eso.org
Morita morita@nro.nao.ac.jp
MUNDY lgm@astro.umd.edu
Radford sradford@nrao.edu
Sakamoto seiichi@nro.nao.ac.jp
Wootten a wootten@nrao.edu
WRIGHT wright@astron.berkeley.edu
Eric.Anterrieu@cerfacs.fr;demerson@tuc.nrao.edu;ghtan@eso.org;guilloteau@
iram.fr;dhofstad@eso.org;morita@nro.nao.ac.jp;lgm@astro.umd.edu;sradford@
nrao.edu;seiichi@nro.nao.ac.jp;awootten@nrao.edu;wright@astron.berkeley.edu
Site:
-----
The site characteristics have been presented by B.Butler, S.Radford and
S.Sakamoto. The site working group has produced digital maps of the
Chascon Science Preserve. Masks defining areas where the antenna
stations could be located have been defined. They account for slope
constraints, shadowing limits by the landmarks, avoidance zones around
the gas pipelines, etc... Visual inspection of the most promising areas
for the location of the compact configuration of the ALMA array. Two
particularly flat areas were suggested as the best candidates. The
coordinate system for the maps seem under control, at least within about
half an antenna diameter.
Soil testing was performed in several areas, including the selected flat
areas. In these regions, the bed rock seems to be only 1-2 meters below
the ground. Concerns about the high soil resistivity exist, but the
measurements were made in the coldest period of the year.
Strawman Intermediate Configurations:
-------------------------------------
Two strawman configuration design were presented. The first one, by
M.Yun and L.Kogan, is based on doughnut shaped pad distributions, with 4
discrete configurations ranging from 300-m diameter to 3 km diameter.
The scale factor between adjacent configurations is about 2.1. The pad
sharing between adjacent configurations is 50 %. The pad distributions
were optimized in order to minimize the near in sidelobes, in snapshot
mode, up to 20 synthesized beam width, but taking into account the
terrain constraints. A possible configuration cycle scheme was
presented. Configuration change strategies leading naturally to
elongated hybrid configurations between adjacent nominal ones were
presented. This would allow optimizing beamshapes for extreme
declination sources.
The second design was presented by J.Conway. It is based on a
self-similar, scale free layout of the pads. Self-similar design lead to
centrally condensed pad distributions, producing near Gaussian beams.
Although the initial concept was based on spiral pattern, some
optimisation of the beam sidelobes was applied to get rid off the
regularities, and account for the terrain constraints. The design allows
for continuous configuration changes, in which you take the innermost
antenna and locate it on the nearest outer pad (or vice versa). A
configuration cycle strategy was also presented.
Although the two designs started with different concepts, they have
actually a lot of common features. They share the same inner and outer
configurations, have the same number of pads, the number of antenna
moves in both cases is identical, and recognizing the initial design
pattern in the actual layout is almost impossible! Both designs
incorporate a 10 % N/S elongation to produce a better average beam
circularity. Studies have shown that the overhead due to the
configuration changes is essentially proportional to the number of
antenna moves, and not to whether these changes occur in burst or more
continuous modes. The most apparent difference is the near-in sidelobe
level: while Conway's configuration produces near Gaussian beams,
specially in long tracks observations, Yun & Kogan design tend to
produce an Airy-like pattern with a double ring of near-in sidelobes.
The impact of the location of the compact configuration has also been
studied. Results show that unless the de-centering becomes extreme, this
is not a serious concern.
Compact Configuration:
----------------------
M.Yun presented a strawman design for the compact configuration. The
main design goal was to produce the highest possible brightness
sensitivity, but the layout also has to accomodate serious constraints:
close packing limit, shadowing, and accessibility. The design consist in
a 150-m diameter ring containing about half of the antennas, with the
other antennas inside. It provides dedicated channels to access the
inner antennas (for maintenance and/or configuration change), with a
18-m wide channel access. Beam sidelobes were optimized using Kogan's
algorithm.
An alternative concept was proposed by F.Boone, in which the compact
configuration could be optimized to produce a nearly Gaussian beam, with
significantly less sidelobe levels. It would fit into a 200-m diameter
region. No detailed design concerning the accessibility has been
performed on that concept yet. This concept would provide better imaging
performance, but at the expense of more limited brightness sensitivity.
F.Boone also presented general guidelines for array design showing that
single configuration observations with adequate uv sampling and Gaussian
beam should be possible up to the 3 km configuration.
Long Baseline Configuration:
----------------------------
The long baseline configuration presents a different problem. Terrain
topography imposes more severe constraints, since such a large
configuration has to go around Cerro Chascon. L.Kogan presented a
strawman layout of a nearly circular ring of about 14 km diameter, with
optimized sidelobes and pads located within the 10 degree slope mask.
The configuration goes beyond the Jama international road, with
avoidance region of 400-m around each of the two gas pipelines. This
strawman design shows it is possible to accomodate this high angular
resolution configuration with site constraints and good imaging quality.
However, no considerations on pad accessibility have been incorporated so
far. This would require an extensive site survey.
Simulations:
------------
A complete configuration evaluation pipeline has been implemented within
Classic AIPS by S.Heddle. The pipeline includes standard imaging and
Clean-based deconvolution (with fixed parameters). It has been
systematically applied to a library of 6 images, with different
properties (dynamic range, spatial scale, point source or extended
structures, ...), for all "standard" configurations of the proposed
designs. Simulations were performed for snapshot and also long (4h)
tracks, to allow comparison of the deconvolved images.
"Objective" image quality assessment, like fidelity index, fidelity
range, histogram of rms distribution as function of image intensity, all
of which were agreed upon by the configuration working group, have been
computed and provided for all simulations. The simulations have revealed
the importance of the short spacing information, and this has been
reflected into the proposed array designs. In general, both strawman
designs provide good imaging quality, the only noticeable difference
being the "doughnut" concept yields slightly higher image errors on the
brightest points in the image.
This basic evaluation pipeline was intended to test the ultimate imaging
performances of the proposed configurations, and as such did not include
any error (whether thermal noise, phase noise, calibration uncertainties
or pointing errors). Simulations including pointing errors and thermal
noise were presented by several people or groups. K-I. Morita used SDE
and MEM-based tools to do that, L.Kogan used AIPS and Clean-based tools,
while J.Pety & F.Gueth developed a complete simulation package in
GILDAS, with Clean-based tools. Pointing errors tend to reduce the
differences between the two proposed configurations styles, but do not
reverse the trends.
Metrics:
--------
A memo presented by D.Woody advocated the use of the beam shape, and
specially the importance of the sidelobe levels as a general estimator
of a configuration quality. F.Gueth presented a few general guidelines
concerning the science requirements on the configuration design. Except
for the few quantitative estimates agreed upon and incorporated into the
common simulations (the AIPS pipeline), it was felt that image quality
could not be represented in a general way by a simple metrics.
ALMA Compact Array
------------------
Given the importance of short spacings in constraining the image
quality, the ASAC recommended evaluation of the importance of an array of
small antennas to fill in the uv gap around 6-9 m, between the shortest
baseline and what can be recovered using total power from individual
antennas. In the presence of pointing errors, such an ACA is expected to
give better results than the homogeneous ALMA array. For these purposes,
two major simulation activities were carried out.
K-I.Morita performed an extensive set of simulations using the SDE
package and MEM-based deconvolution tools. The simulations revealed, as
expected, that the ACA improves the image quality (in this case the
Fidelity Index by a factor 2 to 3) provided the total power data is
smoothed before the joint deconvolution. An ACA with smaller dishes
provide marginally better results in the absence of noise. A single
image was used in this simulation.
J.Pety & F.Gueth developed an independent simulation package in GILDAS,
with a new Clean-based deconvolution tool for heterogeneous arrays.
Simulations with pointing errors were performed on several images,
showing a typical improvement of 60-100 % in image fidelity (Fidelity
Index was computed for several image thresholds).
In addition to these two major activities, M.Yun performed a comparison
between two variants of MEM methods, but without pointing errors. These
revealed that even in the absence of errors, the homogeneous array
provides just enough constraints to drive the deconvolution, and that
the inclusion of ACA gives robustness to the process.
In the discussion, it was pointed out that the total power should be
measured on a larger area than the mosaic itself in the homogeneous
array concept. This was included in Morita's simulation (by actually
simulating a larger mosaic), but not in Pety & Gueth. On the other hand,
Pety & Gueth simulation uses a new deconvolution method for the hybrid
(ACA + ALMA) array which is not yet optimized.
All simulations mentioned above used a separate ACA, of 8 to 19
antennas of 6 to 8-m. The amount of total power data used in complement
was different. Morita used total power data from each 12-m antenna,
while Pety & Gueth showed that the equivalent of 16 single-dish
antennas yield the best results. In this last concept, ACA plus 4 12-m
antennas would be used to provide short spacing information to the
scientific projects which require wide field imaging. A time ratio of
3-4 is required between the short spacing data and the data obtained in
the compact configuration of ALMA. This ratio can be matched if no more
than 25 % of the ALMA time is spent on projects requiring imaging in the
compact configuration.
In contrast with this concept, F. Viallefond presented a fully
heterogeneous configuration merging the ACA and ALMA antennas in a
compact array where each small antenna is surrounded by 5 large ones.
Hybrid correlations between the small and large antennas provide the uv
spacings between 6 and 12-m. This scheme provides a better weighting
function than the separate ACA concept, as well as better calibration
capabilities. However, its performances in terms of imaging in the
presence of pointing errors have not been evaluated yet. Also, the
detailed design of the configuration has not been performed, in
particular concerning antenna accessibility and shadowing constraints,
which are difficult to assess in the absence of detailed design of the
antennas.
Operations & Cost constraints
-----------------------------
D.Hoffstadt presented the status of the operation concept. ALMA is to be
remotely operated from the Operation Support Facility near San Pedro de
Atacama, where all the heavy maintenance will occur. It is foreseen to
bring back the antennas for major maintenance task at the OSF. In this
concept, the crew at the Array Observation Site (AOS) will be minimal,
with pre-planned tasks limited to simple maintenance (module exchange,
planned preventive maintenance) and configuration changes. The lack of
large crew on the site make configuration changes a significant fraction
of the work at the AOS. Regular antenna motions with few antennas per
day are preferred over burst mode requiring the maximum possible number
of antenna motions per day.
The overall operation concept is currently being elaborated, so that no
further detailed implication could be given at that time. Cost issues
were also mentioned, but no definite number can be given until a call
for tender is actually sent out.
The construction of the antenna pads is anticipated to occur in 3
stages. However, a global view is required before starting the tendering
for the first phase.
Recommendations of the Committee
--------------------------------
The committee was impressed by the quality of the work put into the
configuration design. Design difficulties have clearly been addressed,
with preliminary designs being fully developed up to the location on
site and the operational aspects. Extensive simulations have allowed a
comparison of different concepts and a better understanding of the key
parameters. The committee feels that all elements are now sufficiently
well understood to allow detailed design to proceed. The committee
wishes to thank all the participant to this collective effort for their
work over the last years.
Based on the presentations and discussions, the committee issues the
following recommendations to complete the configuration design:
1) The location known as "Chajnantor South" should be used as the place
for the most compact configurations, i.e. the ALMA array center. The
operation buildings will be located about 200-m away from the location
of the most compact configuration. The exact location of the building
will have to be derived after the configuration layout is selected.
2) The operation model should use single configurations for most
observations as much as possible. This is theoretically possible except
for the largest configuration.
3) A flexible reconfiguration scheme is preferred. ALMA will be built
for several decades, so that building flexibility in the design may
allow optimization of observing strategies at a later time. Telescope
moves from one configuration to another should gracefully change the
resolution without seriously compromising the beam sidelobes.
4) The INTERMEDIATE configurations design should attempt to produce
Gaussian UV coverage, with 7 to 10 dB edge taper, in a single
configuration. This choice of Gaussian beams minimizes sidelobes, and
result in lower reconstruction errors.
5) The COMPACT configuration should be optimized for maximum surface
brightness. This choice is justified because Gaussian beams at similar
angular resolution are expected to be provided by the nearest
intermediate configuration. In this design, "channels" of 20-m wide must
be provided to allow removal of the innermost antennas. It is acceptable
that a few (< 4) antennas can only be accessible after removing one
additional antenna. The width of the channel, 20-m, takes into account
the antenna design and the most favourable orientation of the antennas
alongside the channel. It is important that close contact be maintained
with the antenna groups, to ensure that the chosen compact configuration
remains completely compatible with the design of antenna transporter.
6) The EXTENDED configuration should be designed to be complimentary to
the smaller configurations, and to be optimized for angular resolution (uniform
uv coverage). It is recommended to avoid crossing the Jama road, and to
use the 5 degree slope mask as far as possible for acceptable locations
of the pads (an exception to this is the South-Eastern part of Cerro
Chascon where locations within the 10 degree mask may be required). In
designing this configuration, care should be taken that the hybrid
configurations between the 3 km "ring" and the extended configuration
yield reasonable beam shapes. These hybrid configurations will be used
for a significant amount of time, since it is estimated that moving
between the 3 km and 14 km configurations will take about 12 days.
7) The total number of pads should not exceed about 250 (the ACA is
excluded in this accounting)
8) The configurations should be designed for 60 antennas. The best
locations of the remaining 4 antennas among the remaining existing pads
should be indicated. It is also important to investigate the robustness
of the configuration layout against random failures by assessing the
imaging quality when only 56 antennas out of 60 are assumed to be
operational
9) A 10 % N/S stretching factor is accepted as a good rule to optimize
beam circularity. For the more extreme declinations, possible hybrid
configurations utilizing the available pads should be suggested
10) It is anticipated that the so-called ACA (ALMA Compact Array) will
be build, although it is not yet in the baseline ALMA project. The ACA
could consist of about 12 to 16 7-m antennas, with a close packing ratio
of 1.25, plus 4 12-m antennas used for calibration and single dish work.
ALMA configuration design should take into account an area of about 55-m
in each dimension, where this ACA will be located. This area should be
near the compact configuration of ALMA, and as such may interfere with
the most compact "intermediate" configurations.
11) The ACA configuration should be designed for best brightness
sensitivity. Since shadowing constraints will play a significant role,
it is allowed to have at most 50 % extra pads to provide good UV
coverage even at the extreme declinations. However, since simulations
have shown the ACA configuration plays a role in the final image
quality, proposed designs should be compared using available simulation
tools.
GOALS & DELIVERABLES
The ALMA configuration design should include
- location of the pads
- the length of roads, using a maximum 15 % grade to the road layout
- a tentative layout of the conduits, which should limit the number of road crossings
- the location of the ACA (the design of the ACA configuration is a separate package)
- a suggested location of the buildings
- considering the current ALMA planning, the deadline is September 1st, 2001.
Task division should be agreed upon with the ALMA management ASAP.
Considering the work done up to now, three groups appear in a position
to provide a design: NRAO (L.Kogan), OSO (J.Conway) and DEMIRM (F.Boone
& F.Viallefond).
The committee stresses that the design is a global process so that all
configurations should be matched together. A possible working scheme is
to define the COMPACT configuration first, and to progress continuously
to the most extended one. This requires proper collaboration between the
various participants to the effort.