RE: Presentations and Version 0.2 Wishlist

From: Bolton, Tim (bolton@phys.ksu.edu)
Date: Mon Nov 28 2005 - 15:41:09 CST


Hi Stan,
 
Very impressive as usual, and very ambitious.
 
I think it might be useful to try and imagine a little how this tool will ultimately be used, and a few things pop into my mind. The idea here is try to help make sure we don't "engineer things out" too early. Hopefully these observations are not so generic as to be meaningless:
 
Observation 1: We will run an experiment with at least two, and perhaps four, detectors. The detectors will be nominally identical, but will in fact differ in small ways; and we will surely use RAT to take residual differences out.
 

*
        How do we "run" multiple detectors? One at a time? Correlations are key. Will we want to "force" simulation jobs to run the same physics models (flux, cross section, etc) through all the detectors at once?
*
        How are multiple detectors defined? Presumably through different spec files operating on the same basic geometric objects?
*
        Can we build in enough flexibility to parameterize small differences in detectors? Naive example: maybe all the spherical surfaces should be allowed to pick up non-spherical moments via spherical harmonics? Every PMT probably has a load of parameters. There could well be individual quirks in the chimneys, source deployers, etc. What is the best way to put these in?

 
Observation 2: We are most interested in relative differences between detectors.
 

*
        Point here is that it may be most useful to always define detector parameters relative to some nominal "perfect" configuration and avoid "absolute" description (absolute gains, pedestals, etc)... if this is possible.

 
Observation 3: We are interested in modeling relative differences to at least 0.1%, and maybe even better.
 

*
        We may run bigger MC jobs than one might naively think. What is a million-event run looking like in terms of time, resources, output (ROOT!), etc., even with data set broken up.

 
Observation 4: We can never get needed accuracy from first principles calculations: our models will be driven by huge "calibration" data sets.
 

*
        There will be enormous amounts of calibration data! How do we organize it in the RAT framework? Do we use a database? Which one? Interface?
*
        How does RAT get "fed" with this data. Some of this must be done at run time so that RAT models time shifts, etc.
*
        How do we make running a simulation with the correct calibration files as "professor-proof" as possible?

Observation 5: We may be still taking data in 2015.

* Synching with Geant, ROOT, etc. may be worth the headache if it gives us longevity.
* Downgrading the cleverness level may not be a bad strategy in the long run.

 
 
TB
Tim Bolton
Professor
High Energy Physics Group
Kansas State University
tbolton@ksu.edu
785-532-1664

________________________________

From: Stan Seibert [mailto:volsung@physics.utexas.edu]
Sent: Fri 11/25/2005 4:37 AM
To: bw_sim@hep.uchicago.edu
Subject: Presentations and Version 0.2 Wishlist

Hi all,

I wanted to let people know that I've updated the RAT intro and
tutorial presentations I gave at Fermilab in October to reflect the
current state of RAT. They are now linked off a page on the wiki:

http://nu.ph.utexas.edu/bw/trac/wiki/Presentations

The intro talk has several more slides on the geometry builder and
RATDB, and the tutorial talk now includes the generator syntax updates.

The code gurus at Oxford have been very helpful and generated a lot
of suggestions for Version 0.2. I've been gathering these notes into
a RAT version 0.2 wishlist that I would encourage people to take a
look at and give feedback on:

http://nu.ph.utexas.edu/bw/trac/wiki/Version0.2Wishlist

There are some rather dramatic changes to the interface of RAT being
proposed there (including the elimination of the GEANT4 command
language). This would not impact the progress we've made on the
physics simulation, but would open up the possibility of both simpler
and more interesting job scripts.

Feel free to provide feedback on the wishlist here on the mailing list.

---
Stan Seibert


This archive was generated by hypermail 2.1.6 : Wed Nov 30 2005 - 12:01:02 CST