22/02/2013 I have realised this morning that models 57/58 have not solved correctly. The reason they have not solved is that without the Virostek plates in the model, there is no reduced potential region around the coils. I now need to rerun models 57 and 58 with the Virostek (& TOF) Plates.
21/02/2013 Committed comi code to Launchpad Repository via Bazaar. Revision 19. (Model 58)
21/02/2013 It has become apparent that the model is not giving reliable results at low field levels.
It is clear that any field reported below ~1 gauss is likely to be FEA noise, although the exact point at which this happens is unclear.
To try and get a better handle on this noise limit and to try and understand whether the model is elsewhere limited
by meshing resolution I am going to run a series of model tests.
To do this I have added a few variables to the code:
1) #TightenConvergence. If set to 1 this increases the convergence criteria from 1E-08 to 1E-10.
This may improve results but will increase the solve time.
2) #TightMeshResolution. If set to 1 this will halve the meshing resolution.
I'm not sure at the moment whether this will give meshing issues but if it doesn't then it is expected that this
variable will signficantly increase the solve time.
3) #x(y)(z)_scale_factor. This defines the size of the outer model body in comparison to the outer air.
Previously it was fixed at 1.02 but I'm wondering if this value was too small as there is evidence of a non-tangential field at the model border.
These variables allow this ration to be easily altered. I'm going by increasing this value to 1.2.
For simplicity it is my intention is to run a step IV solenoid 240 Mev/c model with just the NSW and SSW
(edit 22/02/2013: also need Virostek Plates) in situ to see if these variables make any noticeable difference to the results that we are seeing.
04/02/2013 Model 56 has shown a solve time of 21 minutes which is clearly in error. The error occurred because the .comi could not find the 240MeV/c Step 6 flip mode conductor file. This file was incorrectly named and so the model ran with no conductors. The model is now being re-run with a correctly named conductor file.