2.12 Configuration Studies

SCHED has capabilities, added in 2002, that are useful in array configuration studies. These are tied to the plotting capabilities. If OBSTYPE = CONFIG is specified and a UV plot is requested from the interactive plotting menu with more than one requested source, a map of the station locations is plotted first. It is possible to adjust station selection by clicking on this map, and it is even possible to move stations. Also, the station selection can be used to set up a configuration optimization that uses, as a quality measure, some statistic about the sampled cells in a grid. Alternatively, a contour plot of the quality measure around a selected station can be made. The effect of multifrequency synthesis can be examined with the help of the UVMFS parameter. This all makes for a rather interactive way of examining possible configurations. The style is much like that used for the VLBA configuration search, but is far more modern and interactive. The optimizer is fast enough that a 1.8 GHz Pentium IV under Linux, optimizing arrays of 20 stations using 12 hour tracks of 10 minute scans on 4 sources can process over 50 different configurations per second. To describe the capabilities, the steps involved in using them will be described.

First, start with a planning type schedule such as is described in the schedule planning section. The scan characteristics can be adjusted at will. The examples are good for “hunt and peck” configuration examinations. For the optimizer, which looks at the uv points at the beginning and end of scans, it is useful to use something like DUR = 10:00 and GAP = 10:00 to evenly space these points. Recall that setting the frequency to 0.3 MHz (the default for uv plots when no setup is given) gives a one kilometer wavelength so the UV coverage is in km.

There are two ways to look at alternative arrays. One is to specify a lot of stations up to the maximum that SCHED will accept. Then the ones to include in the uv plots can be selected from the sources menu or interactively from the map. This allows rapid trials of many options. As and alternative, or in addition, any station can be moved interactively. Note that you can use OPTMODE=UPTIME to get continuous tracks on each source, or you can give an actual schedule. The author has found it useful to invoke two instances of SCHED and carefully align the plot windows (most easily by expanding them to full screen). Then the window manager can be used to blink back and forth to see differences easily. Of course, selecting the station of interest to highlight (plot uv points in red) helps distinguish differences too.

Note that, you can adjust the maximum number of stations by setting the FORTRAN parameters MAXSTA in sched.inc, and PSTMAX in plot.inc. You might also have to increase MK in schin.f to increase the allowed number of input parameters. If you just change MAXSTA, SCHED should complain about other changes that are needed. Be warned that, as these parameters are increased, the size of SCHED increases significantly since they affect several 2D arrays with MCHAN, which is rather large, on the other axis. Average users are not expected to do this sort of thing, but if you are doing serious configuration studies, you are probably not an average user.

The latitude and longitude range to set for the map should be set using MAPLIM. SCHED will examine these ranges and look for vector files that provide useful geographic information. If the region is large, country and ocean boundaries are plotted. If the US is included, state boundaries are also plotted. Finally, if New Mexico is of significant size, the roads in New Mexico will be plotted (this facility is being used for New Mexico Array - a part of the EVLA - configuration studies). For now, the map file names are hardwired to be $PLANET_DATA/wdbtemp4.e00 for the world map, $PLANET_DATA/states.e00 for the US states map, and $PLANET_DATA/tra0003-2.dat for the NM roads map. $PLANET_DATA is an environment variable used locally to point to the directory where the planetary ephemeris is kept. The maps will not be distributed with SCHED (they are much larger than all the rest of SCHED put together), but can be made available to interested users.

Start SCHED in the manner for plotting (specify PLOT and the schedule file). Once the menu comes up, set various things. Under Options, it is useful to decrease the default line widths. Under AXIS, select UV Plot and set the desired scale. It is useful to reverse the signs of the X axis so that the UV tracks can be easily associated with stations on the map. Under files, select Both stations selected so that any unselected stations are left out of the plot. Also select the sources you wish to use. I have defined several sources at different declinations and at ra=0 for these studies and like to use the ones at 44, 18, -6 and -30 degrees as a representative sample. Then it is a good idea to save all this in a parameter file from the FILES page. On the next run, loading that parameter file can save all the other setups. Finally, click on PLOT.

At this point, a map will appear with selected stations in blue or red and unselected stations in yellow or yellow with a red central dot. You can right click on the map and the uv coverage for your sources will be plotted. You can drag the corners of the pgplot window to make it larger, if you wish, or let the window manager expand it to full screen. The next plot (click PLOT again) will then be larger. In the uv plots, any baseline involving a highlighted (red) station will be plotted in red. Others will be plotted in blue. Left clicking on the map will toggle the selection status of the nearest station to the cursor. Middle clicking on the map will flag the nearest station to be moved. The next left click (maybe others too) will put that station at the cursor position. Note that you cannot return a station to its original position without restarting the program. Once a station has been moved, its symbol will be surrounded by a green ring and, in the station list printed with the uv plots (right click), the position will be in green. Checking configurations involves clicking stations on and off and running the uv plots.

The map can be zoomed in and out with ’Z’ and ’z’ respectively. The zoomed image will be centered at the location of the cursor when the letter was typed. The zoom out is by a factor of 1.5, in by a factor of 1.0/1.5.

The optimizer is triggered by typing the letter O (not case sensitive) while the cursor is on the map. What it does is count all the highlighted stations, then it tries all combinations of that many stations, choosing from all stations shown either in red or yellow with a red dot. Thus you can use the toggle operation to select that stations that will be tried with the optimizer. Be warned that the number of arrays to try can get very large if you try to optimize too many stations at once. SCHED will inform the user of the number it plans to try and then tells about its progress so he/she can decide whether to wait and watch, go get a cup of coffee, go to bed, or abort the process (kill the job - I suppose some day this should be made friendlier).

The optimizer calculates a quality measure for each array and ranks the resulting arrays accordingly. The quality measure is based on statistics of sampled cells in a special grid. There are two options, one is to simply count the number of sampled cells. The other is to look at the rms scatter of the number of samples per cell. Other quality measure options could be added without a lot of trouble. Some details are written to the nnnn.opt file (nnnn is the project code), including the results for the top 200 arrays.

The optimizing grid is a polar with a radial cell size that scales with the root sum of the radius squared plus a constant squared. If the constant is small, this ends up scaling with radius (a logarithmic grid) which is useful for large arrays that cannot be reconfigured to adjust the resolution. For a large constant, the cells are evenly spaced in radius, which places more importance on the longer spacings. With an intermediate constant (the constant is in the units of the UV plot), the cells have even spacing in the middle and go toward scaling with radius on the outside. This has been found to be the most useful case, which is why this complexity is included. If the constant is too small, arrays that are excessively centrally condensed are generated. If too large, there can be poor short spacing coverage. You will need to experiment with the values. Also, you will need to experiment with the number of cells to use in the radial and azimuthal directions. If the cells are too small, you will not distinguish between arrays very well (each track is effectively isolated). If too large, all arrays will fill most or all cells so again this will not be a good test. After an optimization, there will be a faint representation of the grid plotted under the UV coverage plots.

The parameters of the grid are set with GRIDNR (the number of radial cells), GRIDNT (the number of azimuthal cells), GRIDMIN (the inner radius of the grid), GRIDMAX (the outer radius of the grid - which can bias the resulting size of the array), GRIDW0 (the constant discussed above), GRIDMEAS (select the quality measure), GRIDSTEP (select the spacing of latitude and longitude points calculated for making the contour plot of quality — see below), and GRIDVLA (restrict the quality measure calculation to only baselines involving a VLA station).

In addition to optimization, SCHED can plot contours of the quality measure around a selected station. To invoke this, toggle the station selections until only one is red (if more are red, only one will be used), then type S. Sched will calculate the quality measure with the selected station at points on a latitude and longitude grid, spaced by GRIDSTEP arc minutes, and then present a contour plot of the results. This can be useful for finding directions to move a station to improve coverage and for finding the area over which a station can be moved to account for logistical considerations. The contours will remain until the map is replotted, so if you repeat the process, multiple versions can end up on the plot. On replot, the contours will be replotted unless some other action, like station selection or moving, has been taken. This allows hard copies including the contours invoked from the files menu to be made for printing or whatever.

When the station map first appears, it is in full screen mode. But when you right click to get the uv plots, it is replotted in one of the panels.

If one clicks ’K’ on the map, two postscript files are produced that are useful for VLA centered SKA configuration studies. These are highly specific outputs with many parameters hardwired, so they are not expected to be useful to the general user. Someday, they may be generalized.

Note that it would not be hard for a user willing to do a bit of FORTRAN programming to cause other maps to be plotted or to add other optimization quality measures. If you want to do so, please let Craig Walker know (cwalker@nrao.edu) so we can integrate a way to choose the modes and input parameters and keep from having multiple versions of SCHED. It is also just a matter of changing a few parameter statements to increase the number of allowed stations. This is kept down for the released version of SCHED because that dimension is combined with other large ones, like the maximum number of scans, to set the size of some arrays and not everyone has huge memory computers.

These facilities were added by Craig Walker in early 2002. Putting them in SCHED was the easiest way to obtain facilities long wanted for UV studies. This scheme for UV optimization is especially useful for arrays of relatively small numbers of antennas with strong geographic constraints, as opposed to large arrays laid out on an open plain.