From: Francesco Cerutti (Francesco.Cerutti@cern.ch)
Date: Sat Mar 08 2008 - 10:05:03 CET
Dear Frances,
> My question is therefore, are these high results in any way 'real' (in
> that you would expect to measure doses of this magnitude outside the
> shield wall occasionally) or are they artefacts of the biasing ? If it
> is my choice of biasing which is leading to these fluctuations, can
> anyone suggest how I might improve it to obtain 'better' results outside
> the shield wall ?
> I should also point out that the errors on the gamma dose rates outside
> the shielding are typically very high ( > 90 %) which suggests there is
> scope for improvement. Results for neutrons have lower errors but are
> still subject to fairly large fluctuations.
Leading particle biasing (activated by EMF-BIAS with SDUM=LPBEMF) is a
technique aiming to reduce CPU time (with a dramatic - essential - gain in
some cases) carrying at the same time weight fluctuations which increase the
variance, but your spikes outside the shielding to my mind are mainly due
to low statistics there, not to weight fluctuation.
In order to get acceptable errors and a more regular and reliable
pattern, you might i. [brute force!] increase significantly the total
number of run primaries (up to the order of 1e8-1e9) - the idea is that
more histories with LPB activated are more convenient than less histories
without LPB - ii. increase the bin volume (if you don't really need there
bins of 2cmx1cmx1cm) and iii. set up suitable importance biasing as
yourself envisaged in one of your previous messages.
More information on biasing is available at
http://www.fluka.org/course/Legnaro-07/Lectures/Biasing2.pdf
Hope this helps
Francesco
**************************************************
Francesco Cerutti
CERN-AB
CH-1211 Geneva 23
Switzerland
tel. ++41 22 7678962
fax ++41 22 7668854
This archive was generated by hypermail 2.1.6 : Sat Mar 08 2008 - 12:43:46 CET