ALMA Configuration Group

Date: NOV-21-2000 (Tuesday) 21 November 2000.

Time: 12:00 EST (10:00 MST, 1700UT)

Conference leader: Al Wootten

Past minutes, etc on ALMA Configuration Group Page

Minutes

------------

Attendees: Wootten, Conway, Guilloteay, Gueth, Pety, Heddle, Woody, Yun, Webster, Butler, Morita, Kogan and Radford

Summary

Pursuit of a Standard Metric - Perhaps we have a start here but this needs to be pursued.

Short Spacings - Are those presently available sufficient?

Total Power addition - These are not included now. Mosaicing needs to be simulated and the ACA defined.

Timescale - Little time is left to visit the various pad sites. Please identify questionable locations for the team leaving on 10 Dec.

Updates On strawperson configuration simulations from Steven Heddle's simulation engine. Reports by Steven can be found at his ALMA page.

Steven solved the registration problem reported to the list earlier by sampling the library images at the same scale as the final images. These were subsampled over nxn regions, averaged within those regions. Results for A, B and C configurations (to use the commonly-understood but unofficial names) and posted at his WWW site. Steven has finished the +25 degree C array simulations also and will post them. For the B array -23 degree simulations, several pages for which the grey image scaling has been matched for the Kogan and Conway arrays, to ease interpretation. How do we form a suitable metric for comparison of these simulations?

There was general agreement that the registration issue had been addressed well. The scaling addresses a problem apparent in the images--missing short spacings can cause a DC-offset in the images whose differing magnitude tends to dominate the appearance of the difference images.

Is enough weight being given to the range of baseline lengths? For the simulations which Steven has finished, the baselines are long; shorter baselines could help, and lessen the DC-offset problem. The simulations do not yet address combined arrays, but concentrate on single arrays. We anticipate that the ACA may address the short baseline problem, but the definition of the ACA is still incomplete though a subject of much work in Grenoble.

Discussion turned to the problem of quantifying differences in the simulated images. The Grenoble group has created an imaging simulation pipeline using Gildas, to be used for ACA simulations. A metric they have explored involves the usual mapping and cleaning procedures, followed by definition of beamsize and shape, resampling of the original image, creation of a difference map. This is usually dominated by large scale structure. Then a fidelity measure can be defined as:

FM = ORIG/DIFF

where ORIG is the resampled original image and DIFF is the difference image. Pixels are replaced by the rms when pixel values tend toward zero, so that the FM image is well-behaved. This measure ranges from tens to a few hundred, depending upon input image. Then the number of pixels can be plotted versus FM in a histogram. Another measure which has been used is Kogan's, which employs the ratio

FRI = PEAK/RMS

Conway brought up the importance of having this measure for both regions on the source and off the source, of which Guilloteau's measure is one variant; this depends upon the image of course. John suggested using

FMI = ORIG/DIFF with magic blanking at different levels. For one measure, one could blank with a value 10% of the peak, for example, and for another perhaps 1% and another perhaps .1%; this gives the fractional error as a function of brightness level. One must be careful not to obtain holes in the image. Conway will work with Steven to implement this in AIPS.

Stephane invited Steven to come to IRAM for a few days to see how the Grenoble group was performing simulations; this was agreed to be a good thing and will be planned.

There was some discussion of imaging with the ACA. SG: One difficult problem is the size and number of telescopes, as the need for calibration pushes one toward larger diameters and the need for good imaging pushes one to larger numbers. JC: This could be important in the design of the larger arrays. SG: The ACA will be complementary, not necessarily observing simultaneously. BB: Except for time-variable sources. SR: The ACA isn't agreed upon, and may come later so perhaps we should be careful about relying on it. It may need a fairly large number of antennas to image properly. SG: ACA may come at about the same time and normally will be used as a complement to ALMA, so images will have all data in them. BB: Cross-correlation? SG: For the time being, no, the ACA will have to survive on its own as the integration times which are appropriate to the ACA and ALMA will vary. JC: Co-sited? SF: Yes, as some cross correlation may be neede. KM: Will the ACA be at the center? SF: Reasonably close but not concentric.

MY: Steven, you do match resolution in the B&C arrays? SH: Yes, the clear restore beam is the same. SG: There will be some problems with large scale errors. The units of the clean map differ from the dirty map and this will misrepresent extended structures, for which one cannot clean deeply enough.

JC: The filling of the UV plane is very similar in the two styles. Perhaps 2-3% different in snapshots, with Kogan being more filled, but less in longer integrations. The designs have come very close in this respect. SG: Has anyone done maximum entropy simulations? Not really. These will become a focus when Holdaway joins the group on 1 Feb.

DW: How do we reach closure? Do we have the appropriate metrics in place? Beams, sidelobe levels, uv metrics, occupancy of uv cells? We will want some metric deduced from the Fourier Transform of the difference images, to help us see where the differences arise also. Various metrics are hidden in reports on array design in various places. They should be brought out in a table. Metrics involving the images are an element of the discussion today. Visually, it is hard to quantify the differences between images. Steven's calculation of the metric John and Stephane proposed will enhance our ability to quantify differences (or not!). We proceed with this step, for discussion at the next meeting. JC: The decision between the arrays will probably be operational.

Future Plans

The plans for the layout of the pads must be finalized by the beginning of construction. We hope that the first construction funds will arrive on 2001 Oct 1 when the FY2002 begins here. The project WBS has this date as 2002 Jan 1. Realistically, I think we need to inspect the pad sites before finalizing their location. We have begun this, but it is a huge task. It would be best to have some decision on the pad locations before the next austral summer, but that doesn't seem entirely practical. Radford and Butler will be on Chajnantor in mid-December. What sites should they inspect? SR: Bryan is preparing a memo, almost ready, on best centers for the arrays. His last figure shows a comparison. Chajnantor S seems best, somewhat south of the spiral center, somewhat west of the double ring center. JC: has done some egg shaped perturbations on his array and finds that perturbations of 500m to 700m can in general be accommodated but much larger than this becomes difficult. MY: One must be careful, in some experiments trying this squinting of the beam became a problem. Since the mask is designed on slope, some of the antenna placements are in the centers of quebradas in both designs. Is this a good idea? SR: We have many things to do and not much time in December. It would be good to have all the center positions for inspection, plus and questionable locations such as the quebrada antenna; soil tends to be soft there. SR and BB will be at the site in Dec, returning about 19 Dec. SR will go down in late January to early February (Chilean summer vacation; a problem at times) and April-May tends to be a good time for visits also.

The next teleconference was tentatively set for 1700UT on 20 December but some members had difficulty with that day. 21 December at 1700UT is the alternate.