tiPlan.par

A typical tiPlan.par file has parameters as described in this document. The parameters specifying the particulars a tiling run are:

• targetInputFile: Input FITS file with targets, in form of a tiInputTargets.fit file
• targetOutputFile: Output FITS file with targets, in form of a tiOutputTargets.fit file
• tileOutputFile: Input FITS file with targets, in form of a tileOut.par file
• seed: Random seed used to randomize the order of the galaxies. The random number generator is the same as used by SM. The randomization is used to cure the stripiness and gappiness which occurred in the old tiling code. It does not affect the order of the targets in the output file, which remains the same as those in the input file.
• firstTile: Number at which to start the tile indexing, so that each tile in the survey has a unique number.
• tileRun: Index identifying this tiling chunk.

The parameters which determine the conditions to be satisfied are:

• desCoverage: This is the fraction of decollided targets which are required to get fibers (see sdss-target/243 for the definition of a decollided target).
• desCoverageResolved: This is the fraction of collided targets which are required to be recovered in overlaps of tiles (sorry for the poor nomenclature). There is no requirement on this fraction in the survey.

The parameters which determine the initial set of tile positions are:

• uniformTiling: This is the method tiling should use to lay down the original uniform grid of tiles. There are three options at the moment:
  • sloane: This uses the uniform covering of a sphere provided by a program written by Neil Sloane. It will include in the solution all tiles within a tileRadius or so of the edge of your rectangular region (in coordinate, not proper, distance). To trim the edge tiles, use the "edgeTrim" parameter below. The covering it uses will be based on the "nTilesSphere" parameter, unless it is zero, in which case it bases the number on the density.
  • equatorial: This assumes that you are on or near the equator in either ra-dec space (if toSurveyCoords is 0) or in eta-lambda space (if toSurveyCoords is 1). It lays out tiles in a 1d strip centered on the mean declination of the targets.
  • rectangle: This assumes you are approximately in a rectangular region in either ra-dec space (if toSurveyCoords is 0) or in eta-lambda space (if toSurveyCoords is 1). Judging from the area, dimensions, and number of targets, it tries to pick a sensible grid (with each row in the declination direction offset!) to start from. It will space tiles equally in sin(lambda); ie. it assumes the stripes get narrower near the survey edges.
  • rectangleES: This is like "rectangle", except it spaces tiles equally in lambda, rather than sin(lambda); ie. it assumes the stripes do not get narrower near the survey edges (appropriate for the south).
  If the value does not match any of these three, the code looks for an ASCII file of that name, with the first column holding ra values, the second column holding dec values (and no other columns!), and uses those positions. (Note that in this case tileSearchFudge must be set to -n, where n is the number of tiles in the file).
• nTilesSphere: In the case sloane-type initial conditions, this forces the number of tiles for the full sphere tiling to be this number (unless the value is 0). For the SDSS, 7682 is the appropriate number.

Other strategic parameters are:

• tileSearchFudge: Fudge factor used to start the binary search for the optimal number of tiles. So if it is 1.05, and the number of tiles needed to put fibers on all targets is 100, it starts the binary search with limits of 95 and 105 tiles. Note, it will expand the search if it has to, so this doesn't affect the final answer at all! If this is negative but greater than -10000., it is taken to be the negative of the number of tiles to use (i.e. no binary search is performed). If it is less even than -10000., number of tiles is set to be that appropriate to the density in the case that one is using sloane uniform tiling.
• maxTileIter: Maximum number of iterations to cost-minimize for.
• moveLimit: Stop iterating when the maximum change in the position of any tile is less than this value.
• toSurveyCoords: Although tiling is given targets in ra and dec, the geometry of the chunks is usually a rectangle in eta-lambda space. If this is the case, set this to 1, so that tiling will convert to eta and lambda and take advantage of the nicer geometry.
• edgeTrim: Throw away (at the very end of the analysis) tiles this close to the edge of a chunk (in degrees and in coordinate, not proper, units). If negative, reassigns fibers after throwing away tiles. The conditions on how many targets can be missed are imposed before the trimming.
• fractionTrim: Throw away (at the very end of the analysis) tiles with fewer than this fraction of their fibers assigned to targets. If negative, reassigns fibers after throwing away tiles. The conditions on how many targets can be missed are imposed before the trimming.
• retainDecollided: If a target has been called decollided in the past, guarantee that it will still be called decollided.

There are also some parameters which opdb needs for historical reasons.