Workbook for BaBar Offline Users - Event Store: Accessing Event
Information in BaBar's Databases
Description of the information stored in the BaBar Event Store
Databases and details of the event information available in the BaBar
nano and micro databases. For information on finding data, see the
Workbook section Finding
Data.
Quick Link: Quantities accessible at Nano and
Micro Level
In its most basic format, the raw data collected by the detector and seen
by the online systems is just a collection of Raw "digis" (e.g. details of
individual "hits" on crystals in
the Emc) from each subdetector. To turn this raw data into useful
information, the hits must be reconstructed into tracks (in the SVT
and DCH) and clusters (in the EMC and IFR), which in turn must be
reconstructed into particle candidates.
With the advent of the "CM2" computing model, all event
information is stored in the BaBar "Mini" database. Within
this database is stored information at different levels of detail
which are used for physics analysis jobs, including
partially-reconstructed event information corresponding to the former
"micro" and "nano" formats. While the full event information can also
be retrieved, the Mini database contains event data, such as momenta,
energies, number of hit crystals, ... - which is usually enough
information to enable a complete analysis job to be run without having
to resort to the full raw event information (which is very
time-consuming to access).
Reconstruction is a resource-intensive process, and BaBar cannot
afford to do it every time a collaborator wants to analyse an
event. Consequently, we must store the results of the reconstruction
algorithms, and provide a way of reading those stored results
quickly. The various levels of the BaBar databases represent different
balances of the level of detail and accuracy available with access and
analysis time.
The primary data content of the Mini consists of pattern-recognition
objects (tracks, clusters, rings, ...), together with the hit objects
from which they were built.
The Mini also stores a subset of the hit
objects not associated to any pattern-recognition
object. These unassigned hits can be due to background or to
real parts of the event that were missed due to
reconstruction inefficiency. It is important to store these for
use in studies to evaluate and improve the current reconstruction
alogrithms. However, there are too many background-induced hits
present to store all of them. Therefore, only a subset of the
unassigned hits, passing subsystem-specific quality cuts, are
stored in the Mini.
Together, assigned and unassigned hits cover essentially the entire
data content of a BaBar event. The Mini also stores the results of
CPU-intensive calculations such as track fit parameters at the origin.
The Mini database "persists" (stores) several levels of data which
overlap in content. The Mini is designed to that the user can, and
in fact must, decide which level of detail for their specific use
when reading back the Mini. This is usually done in the tcl "snippet" used
to run a user's analysis job. For example, in the
Quicktour, the level-of-detail (LOD)
parameter is set in the following line in "snippet.tcl":
set levelOfDetail cache
There are four options for the Level of Detail (LOD):
micro
Micro mode is intended solely to simplify validating the Mini
against a previously-used database called the "Micro". The micro mode
is less accurate and no faster than the cache mode (described below),
and is therefore not recommended for use for analysis jobs. However,
information corresponding to the old "micro" database is still
available in the Mini Most users will probably never use micro mode
for actual physics analyses, since it is no faster than cache mode
(described below) and returns less accurate values. (For example, the
micro stores only pion-hypothesis track fits.)
cache
Cache mode is probably the most useful mode for most analysis.
This mode reads the Mini so that reconstruction objects
are rebuilt directly from stored results of track-fitting, PID, and
other high-level calculations. It provides easy access to track
parameters for Kalman track fits calculated using five different
particle mass hypotheses. However, cache mode cannot
read new conditions data, and does not allow the user to access
hit-level data.
Information that was previously available in the "nano" and "micro"
databases in the previous computing model is accessible in the mini. A
listing of those quantities is here.
extend
The extended version of the cache mode extends track fits
from the origin through detector material, allowing improved precision
for studies involving particles such as Kshorts which decay outside
the beampipe.
An example of the use of extend mode is when reconstructing
long-lived particles when decay outside the beampipe, such as K0. In
cache mode (and when using the micro), these particles are vertexed
using fit results measured inside the beampipe. Extend mode will allow
the material of the detector to be taken into account to improve the
vertexing of long-lived particles when they decay in the detector material.
refit
This is a much slower mode, used mainly for detector studies,
event displays, specialized analyses that depend on hit-level data,
or analyses that need the most recent calibration constants or algorithms.
The Mini contains tracks and clusters that have already been reconstructed.
But in refit mode, the Mini's tracks and clusters are ignored, and the user
redoes the reconstruction directly from the stored hits, with the user's
choice of inputs such as conditions, particle hypotheses in track-fitting
algorithms, and PID methods.
The four Mini modes have been designed to produce equivalent
information regardless of the levelOfDetail setting. Obviously the
differnt modes will not be identical, but the same physics objects
used in a Beta job should be present in all modes, and provide
equivalent physics information.
The Conditions Database stores detector alignments, calibration
constants, and other time-dependent quantities that describe the
conditions under which data was taken. These conditions
need to be taken into account when data is reconstructed,
and when Monte Carlo simulations are produced.
Because the Mini database rebuilds much of its information from raw
detector quantities, it relies heavily on the Conditions
Database to obtain conversion factor and calibration constants.
Related documents
Micro-level information, which is also available in the mini
database contains the major variables (4-vectors, etc) and the most
important variables of each Subdetector's major purpose, namely:
| Subdetector | Micro quantities |
| Svt, Dch | Tracks which have goodness-of-fit and vertex information |
| Emc | Clusters which have various variables which depend on
the distribution of energy in a cluster and number of crystals |
| Dirc | PID information which separates Pions from Kaons |
| Ifr | Muon and neutral hadron finder |
Micro-level information is made up of a list of BetaCandidates. Their
behaviour from the point of view of kinematics is identical to that of
the full event information, but other detector information is lost at this
stage.
"Nano"-level quantities are a fast summary level that
you can query in order to select events with desirable
characteristics. Hard leptons, more than n tracks, hard
photons, thrust, Fox-Wolfram moments, etc. It is essentially used to
collect events together when some features depend on the modes of
interest.
Nano-level information is fast because you do not have to query each
object in the event for the information that you require - you just
have to look at the summary object. Nano-level quantities include event
information, such as selection bits, global variables and trigger
information. This information is also known as "tag" information, as
it has available a lot of information about "tag bits", which are set to
true/false for whether or not an event satisfies certain criteria,
such as different background filters.
Following the link below you can find several pages describing what
quantities are available in the nano(tag) and micro portions of the
BaBar event store. There is also sample code provided which can be
used to access this data. Note that this table is not currently
maintained, and has not been updated in several years, but is the best
available excellent source of information on the quantities available
in the BaBar event store.
General Related Documents:
Authors:
Jenny Williams,
Chris Roat,
Joseph Perl,
authors of BAD508
Contributors:
David Nathan Brown,
Thomas Hadig,
Fabrizio Salvatore
Last modification: 7 June 2005
Last significant update: 7 June 2005
|
