I observed what at first sight looked like a strange behaviour by the
RubberSheet module: A regular grid of 3x3 points that are all assigned the
value 0 except the one in the middle (see data description below) is not
rendered as a four-sided pyramid, but instead as a hat-shaped surface that is
not inclined at all at two of the four corners. As the data itself is
symmetric, would one not expect symmetry with respect to all corners? If the
grid is extended to 4x3 and the two values in the centre the non-zeroes, the
result is not even symmetric to any axis.
Extending the net to Import->RubberSheet->Isosurface->Image to show a handful
of isolines does result in a symmetric appearance -- now all corner areas are
entirely flat, but I suppose that's an acceptable result. It appears that
RubberSheet does do a proper job after all (a conclusion also supported by
testing with ShowConnections), but its results are rendered wrongly.
I put online a picture that should show what I mean:
http://support.rz.rwth-aachen.de/public/DX/rubberproblem.jpg
Are there errors in my thinking? Admittedly, the problem will hardly surface
at usual meshes' density, but...is there a way to get a 'proper'
rubbersheeting without misusing Isosurface with a high number of isolines?
Here's the data:
object 1 class gridpositions counts 3 3
origin 0 0
delta 1 0
delta 0 1
object 2 class gridconnections counts 3 3
object 3 class array type float rank 0 items 9 data follows
0 0 0 0 2 0 0 0 0
attribute "dep" string "positions"
object 4 class field
component "positions" value 1
component "connections" value 2
component "data" value 3
end
Urs
--
Urs Enke <[EMAIL PROTECTED]>
High-Performance Computing
Center for Computing and Communication
RWTH Aachen University
