Quantification of poor extended imaging: flux in center of image for Gaussian sources as a function of source size. Proposed ES configuration
is not as bad as SMA, but it has twice the number of antennas so should do be able to do better. It is clearly much worse than CARMA.
Dec 2010
compact.v2.cfg new config from Frederic, limiting baselines to 100m. The distribution of uv points still leaves a large inner hole, leaving doubts whether this is the best configuration for imaging extended structure.
Comparing new beams with previous Nov 2010 beams at dec=-20 and dec=-40, 6h tracks:
Analysis of Nov 2010 compact config
Compact_16.cfg, constructed by Frederic Boone w/constraints of BL > 20m, using N half of the inner cluster:
CSV-624
description of what he did is here:
alma16ant.pdf. Spreadsheet with antenna positions and baselines for the selected 16 and for all available antennas:
- ES_Arrays.xls: Spreadsheet of antennas available, baselines, etc. Ignore rows 37-52, leftovers from previous spreadsheet.
I confirm that natural weighting produces the same beams as his "unweighted" in his document, and that the sidelobes are pretty high especially for a short (1h) observation:
There's not much shadowing - none in a 6h track at dec=-20,
a little at dec +20 in a 6h track, mostly of the single antenna A015: (yes, I know plotms didn't reflect the uv points. deal. :)
a little at dec -80 in a 6h track, mostly of the single antenna A014:
I cannot reproduce Frederic's radial uv densities - I see a much narrower FWHM than in his document, unless he didn't divide by the geometric 1/r factor or something:
Next, I compare to CARMA
In general, their config is N/S elongated, so the beam is pretty noncircular at zenith, but its properties vary less dramatically with dec. The dec plotted for CARMA is (-23+37-(simulated dec)) to be compared with ALMA. CARMA's beam has a lower positive sidelobe, but a more negative bowl, so I don't know if I'd say one or the other is dramatically better, just different:
imaging an extended source:
CARMA (noiseless; log intensity scale):
ES ALMA (noiseless):
ES ALMA (noiseless, small mosaic):
ES ALMA with noise:
comparison of radial uv distribution. CARMA has been scaled to approximately the same max baseline as ALMA.
contrast with FS configuration design philosophy:
The original design
2006-06-20-ALMA-90.02.00.00-006-A-SPE
had few constraints, and specified the shortest baseline:
SCIG-00060-00/R An antenna pad separation of 15.15m shall be possible.
For the compact array of 50 antennas, the configuration was sidelobe-optimized, in contrast to the Boone ES process:
The positions of the ... 50 pads within the compact array were then
optimized to give minimum sidelobes after multiplying the beam pattern with the
autocorrelation of the primary beam (see Woody 2001, ALMA Memos 389 390). The sidelobe
optimization used a modified version of the Kogan algorithm (see Kogan 1997, ALMA
memo 171) . The algorithim was improved to minimize simultaneously a group of the
largest sidelobes. The beam optimization was done within constraints of
antenna proximity (taking into account both initial pads and reserved pad positions),
antenna transporter access and no shadowing at transit for sources up to declination
+15deg. When a moved pad violated these constraints it was returned to the original position.
Two independent studies performed by M.Wright using MEM in MIRIAD (ALMA memos 428, 430) and S.Guilloteau using CLEAN in GILDAS,using arrays
of similar design gave excellent imaging performance at all intermediate
declinations.
A particularly novel aspect of ALMA is its B9 capability. The beam is
quite small at that band and many sources will be relatively extended.
The atmosphere can be unforgiving, compromising our ability to make use
of foreshortening to provide shortest spacings. Total power capability
of ALMA has yet to be proven and in any event will be most difficult at
the highest frequencies. I think there should be a constraint for the
ES cycle 0 (ES0) compact array to provide excellent short spacing
response. I note that requirements for B9 might warrant tracks somewhat
shorter than +/- 3 hr but that is not a bad goal.
I think Frederic's design has too few short spacings. It was optimized
for 6 hours of observation around transit. Boone states that the
largest baseline available is 145m, that the uvplane may be sampled at a
rate Nyquist/1.5 and that the samples may be distributed in a gaussian
of FWHM=163m. This is a truncated gaussian, so there will be sidelobes.
I think that the ES0 design should address short spacings more than the
maximum spacings. Although beamshape was a constraint for the full
array design, it needs to be more carefully considred in theES0
configuration design I think. This graph shows the spacings
available from the pads expected to be utilizeable at ES and the
spacings in the Boone design, clearly a subset of the available spacings
which is skewed to longer baselines:
ALMA_Early_Science_MW.pdf: Mel Wright presentation on ES imaging.