Testing the MINOS BField Software Package
Please see my NuMI-note on these tests, NuMI-L-xxx, available here. This page holds the files used in testing, for reference purposes.
Files used in testing
- NearDetBField.default: This file is actually in the MINOS CVS repository, as BField/NearDetBField.default. This is the fundamental set of "generators" at which ANSYS computes the B field to high precision, used to generate the Voronoi diagram used by some interpolation methods.
- smalltable_3.dat: This file (and a version cut to the testing volume, st3fv.new.dat) was used as an accuracy standard; it too is generated by ANSYS, but the points are not the same as those in the above item (technically, they are the "centers of gravity" of ANSYS nodes, and the field values are found by linear interpolation of the vector potential at the vertices of the node. Although this is an interpolation rather than a direct calculation, it is still more accurate than the pure interpolation done, for example, in BFLInterpolation::PlanarInterpolation). When we refer to "true B field," we refer to the values given in this file. This file was given to us by Jeff Nelson, who made it available at this URL under "Table 3."
- bfld_142.dat: This is in the labyrinth, under ephemera/bfield. This is the standard map for the rect2d algorithm, but (as discussed in the above-mentioned NuMI-note, it probably does not correspond to the most recent ANSYS calculations of the B field.
- vorgrid8cm.dat: This is the main "pseudo-ANSYS" grid, interpolated from NearDetBField.default by the Voronoi algorithm (using the program MakeGridUsingVor, which is available in the CVS repository under BField/test), and then passed to the rect2d algorithm for the accuracy tests. Since it is interpolated, it is not as "true" as an ANSYS-calculated grid, but at least using it eliminates the systematic error from the difference between the fields used in the other maps and the old rectangular grid map.
- bouncyline.dat: This is the performance test points file. It is generated using the simple perl script makebouncyline.pl.
- genRand*: These were a series of sparser Voronoi diagrams, created by running voronjob on a randomly sparsified subset of the points in NearDetBField.default (this sparsification was done with the perl script sparsifyvor_randomly.pl). They were used to measure how the performance and accuracy of the Voronoi algorithm varies with the number of generators n (equivalently, with the memory used by the interpolator).
My home page