nx, dx, ny, dy Request a new map grid of given width and cell size EXAMPLES -------- 1. To request a square map with square pixels: 0>mapsize 256,0.5 Map grid = 256x256 pixels with 0.5x0.5 milli-arcsec cellsize. 0> 2. To request a rectangular grid with rectangular pixels. 0>maps 256,0.5,128,1.0 New map grid = 256x128 pixels with 0.5x1.0 milli-arcsec cellsize. 0> 3. To see what the current map size is, omit all arguments: 0>maps Map grid = 256x128 pixels with 0.500x1.00 milli-arcsec cellsize. 0> The second line of each example shows the response from the 'mapsize' command, confirming what was typed. PARAMETERS ---------- nx - The number of pixels along the X-axis map and beam grids. This number must be positive and be an integer power of 2, such as 256 or 512 or 1024. dx - This is the width of each pixel along the X-axis. Unless changed with the mapunits command, the units are milli-arcsec. ny - Optional: If not provided, ny is given the value of nx. The number of pixels along the Y-axis map and beam grids. This number must be positive and be an integer power of 2, such as 256 or 512 or 1024. dy - Optional: If not provided, dy is given the value of dx. This is the width of each pixel along the Y-axis. Unless changed with the mapunits command, the units are milli-arcsec. SIDE EFFECTS ------------ Any existing map will be destroyed, and the new map will be initialized to contain 0.0 in each element. CONTEXT ------- This command specifies the size of the grid to use to sample the map and beam in the image plane. All commands that operate on maps, such as 'clean', 'restore', 'mapplot' and 'invert', require that such a grid be created before they are invoked. 'mapsize' may be called at any time thereafter to change the size of the map or beam arrays. The maxmimum size of the map and beam grids is limited only by the amount of memory available on your computer. Note that the amount of memory required by the two arrays is 2.(nx.ny).sizeof(float) bytes. If sizeof(float) is 4 bytes, then the following numbers apply for square grids: nx=ny Memory ----- ------ 128 131 Kb 256 524 Kb 512 2.0 Mb 1024 8.4 Mb 2048 34.0 Mb Clearly, large grids require large amounts of memory, take a much longer time to process and may cause your computer to page heavily. For this reason it is beneficial to use the smallest grid that can accomodate the region of interest. In particular, if there are many sources to be mapped, scattered over a wide area, it may be faster to partially map each source separately, using the 'shift' command to temporarily shift each source into the center of a smaller map area. CHOOSING THE CELL SIZE ---------------------- The appropriate choice of cellsize along the X and Y axes of a map grid depends on the sampling of the UV plane by visibilities. To sample the finer detail in the image-plane implied by larger U or V coordinates, requires smaller cell sizes in the map. In particular, to avoid undersampling the image plane, the highest spatial frequencies sampled by the UV data must have a period of at least two pixels. This contraint applies separately to spatial frequencies along the X and Y axes of the map. In theory, if the maximum absolute U and V coordinates sampled are umax and vmax (wavelengths), then the corresponding maxmimum cellsizes along the corresponding X and Y axes of the map and beam, are given by: dx = 1/(4.umax + 2) radians => 2.063e8/(4.umax + 2) milli-arcsec dy = 1/(4.vmax + 2) radians => 2.063e8/(4.vmax + 2) milli-arcsec where the "+ 2" component is due to the 2 pixel fuzz introduced by the convolution function that is used to interpolate visibilities onto a regular grid. This constraint is enforced by the difmap 'invert' command, whose job it is to create a dirty map and beam from the visibility data. Visibilities that represent higher spatial frequencies than can be Nyquist sampled by the chosen pixel size are excluded from the maps that it creates. When it does this, the 'invert' command tells you what percentage of the visibilities had to be excluded and displays the above limiting values of dx and dy so that you can, if you want, change the map grid to avoid omitting any data from subsequent maps. Note that it is not necessarily an error to choose a cell size that is too big to sample all of the visibilities. Sometimes this is the only practicle way to create a map that is large enough to view one's field of interest. Also note that because the standard CLEAN algorithm only cleans the inner nx/2 by ny/2 pixels of the map, the maximum map area cleanable by 'clean' and displayed by 'mapplot' is actually the area: X range = -nx/(4.umax+2)/2 -> +nx/(4.umax+2)/2 Y range = -ny/(4.umax+2)/2 -> +ny/(4.umax+2)/2 RELATED COMMANDS ---------------- invert - Invert the UV visibility residuals to form a residual map and dirty beam. restore - Combine model components and a residual map to produce a clean map. mapplot - Display the residual or clean map, or dirty beam.