This or something similar will be included in the distribution; it is
probably too much for a Web page. Any comments?
Joanne
----------------------------------------------------------------------
Jan. 20, 1999
Calorimetry for Fast MC
The overall aim is to produce, from a collection of input final-state
particles, a list of clusters (i.e., objects which have at least an
energy and a position) whose behavior is, in a statistical sense,
similar to that of the "real" clusters to be produced by a full
reconstruction, using the full Gismo simulation as input. The
cpu time needed for the Fast MC is much less the for full simulation
and reconstruction -- on the order of a tenth of a second or less per
event as opposed to 4 minutes.
Algorithm Overview
------------------
Like the full simulation, Fast MC makes some attempt to propagate particles
through the detector in a helical or linear path after throwing the dice
to choose a path length (in units of interaction length or radiation
length), as appropriate, but there are several gross simplifications.
1. Detector geometry is known only at the component (e.g., EM endcap) level.
No allowance is made for active versus inactive material within a
component or for variation in any parameter, such as interaction length
per cm, dependent on the material. Each component is treated as if
it were homogeneous, with parameters such as interaction length per cm
take to be a suitable average value.
2. Magnetic field is assumed constant throughout a component or, in the
case of endcaps, constant over each of two or three pieces making up
the component.
3. Fast MC does not simulate energy loss.
4. There is no digitization.
The first three points taken together allow Fast MC to do the propagation in
large chunks, hence quickly. The result should be the coordinates of cluster
max, but in this release of the code it's the coordinates of the start
of showering (to be fixed as soon as I'm supplied with all the parameters
I need to throw the extra path length).
Finally, the energy is smeared and the transverse (w.r.t. the momentum vector
at the "perfect" position) position is smeared. These smeared values
are the ones written to the ascii output file.
Inputs
------
In addition to the data (events of stdHep MC particles), the calorimetry
part of Fast MC needs
* a (coarse) geometry description of a detector, including dimensions
of all calorimetry components and of the coil
* a single field strength value from which, in an extremely crude fashion,
it deduces an average field value for each volume a particle might
traverse
* interaction length/cm for all components. radiation length/cm for
at least EM components
* parameters needed to compute a sigma for energy smearing (one set
for hadrons, another for e/gamma)
* similar to above but for transverse position smearing
In this release, almost all of the above are predefined in the code. The
only settable option is detector type of "Small" or "Large," which sets
detector dimensions to those of Strawman Small or Large detectors, resp.
Outputs
-------
Currently Fast MC can only be run as an independent program which produces
an ascii file as output. The calorimetry part of the file for each
event consists of the number of clusters found and some information about
each: (smeared) energy, a position (in polar coordinates), some notion
of error on position, and a list of contributing MC particles (always just
1 for this release). For this release, error on r is always 0.0 and
error on phi and theta is derived from the transverse position smearing
sigma. Cluster output also includes the number of hits a list of hits
with energies contributed, but for Fast MC "number of hits" is always 0.
Limitations and Possible Future Improvements
--------------------------------------------
1. The detector components simulated include EM barrel and endcaps, hadronic
barrel and endcaps and muon barrel. Muon endcaps are not included just
because I ran out of time. They should be in the next release.
2. The inner volume is treated as air (i.e., no showering allowed). This
could be changed if I had average material parameters for this volume.
3. The model of the field in the had and muon endcap regions may be too far
off the mark to be of any use.
4. Energy loss could be handled at least on a per-component basis without
major perturbations to the code -- I'd need some more parameters.
5. Particles which exit the tracking volume via the beam pipe rather than
the EM calorimeter are ignored.
6. Clusters are formed and reported with no filtering. In particular,
clusters may occur in the coil, and they will be recorded in the output
file exactly like clusters in the calorimeter. Analysis programs using
the output file should have access to the geometry description used by
Fast MC so that they can detect such clusters, but for this release
such a geometry description is not radily available in machine-readable
form.
7. Most of the parameters described above should ultimately be configurable.
Until that time, those who are desperate can modify the source and
remake the program.
Simply modifying values for smearing parameters should be straightforward.
See SmearFuzz.h, which defines the structure containing these parameters,
and the files GetSmall.cxx and GetLarge.cxx in which they are set for
small and large detectors. To understand how they are used, see
CalSmeared.cxx.
Interaction and radiation length/cm are defined in GetSmall.cxx and
GetLarge.cxx as is field strength.
It should also be possible to modify component dimensions as long as the
topology (what touches what) matches that of either the Strawman Small
or Strawman Large detector. Look at either GetSmall.cxx or GetLarge.cxx
as appropriate. It should only be necessary to change the definition
of one or more of the static variables at the top of the file.
|