[Gate-users] Rectangular Prism With Curved Edge

[John Apostolakis]

From the perspective of a Geant4 Geometry developer:

1) One alternative available in ‘native’ Geant4 is to use a ‘Tesselated Solid’, which allows you to create a solid using its surfaces.  Those can be either triangles or quadrilaterals. This may or not be available in GATE (the experts can advise).  If not, it should be possible to add it as a shape, within a measurable effort.

One advantage of this shape is that the implementation has recently been redone, and the CPU performance is greatly improved compared to the original version, which had to test all the surfaces.

2) Regarding the idea of using 600 boxes: the Geant4 Navigator (which GATE uses) is easily able to handle thousands of volumes placed at the same level of the geometry.  This capability is used by the geometry of large High Energy Physics experiments, and is instrumental in obtaining adequate CPU performance for them.  So having 600 boxes at the same level will not stress it.  Of course it will not be as fast as a geometry with 6 or 60 boxes - but the most important reason is the increased number of steps, since steps will stop at volume boundaries.

[ It may be possible to tweak the CPU performance using a parameter in the volume in which you place these, if you wish to obtain some improvements. Please refer to the Geant4 manual for information about this - again this option may require some code change in GATE to access . I would expect around 10% change, if any, doing this from prior experience.  Unexpected improvements would be of interest! ]

Hope this helps,
John A
===================================================
John Apostolakis,  PH Department, CERN

On 10 Jun 2015, at 01:26, Karl Spuhler <karl.spuhler at stonybrook.edu<mailto:karl.spuhler at stonybrook.edu>> wrote:

Hello,

I need to model a curved filter. I am wondering if this is possible? It will have different thicknesses at 600 points. I suppose I could use 600 different boxes, as I have calculated the thickness of each cross section in Python, but I am concerned that this might make the simulation run for too long.

--
Karl Spuhler

SUNY Stony Brook
PhD Student
(845)249-6836

_______________________________________________
Gate-users mailing list
Gate-users at lists.opengatecollaboration.org<mailto:Gate-users at lists.opengatecollaboration.org>
http://lists.opengatecollaboration.org/mailman/listinfo/gate-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengatecollaboration.org/mailman/private/gate-users/attachments/20150610/3328d88e/attachment.html>