SLAC Cosmic Analysis in Release 13
Last update: Jun/6/07
This is a temporary web page to
track the migration of pixel cosmic analysis to release 13 at CERN.
Reconstruction:
The
Official cosmic reconstruction instructions for release 13 seems to run out of the box, but there are some issues moving to release 13 which
are worth noting:
- The release 13.0.10 is out now and works out of the box without need
to checkout any packages. This can be selected in the setup as
source ${HOME}/cmthome/setup.csh -tag=13.0.10,32
The 32bit option is unfortunately necessary for the moment to be sure, as the 64 bit mode build seems to be missing some libraries if you
try to run on lxplus.
- To reconstruct cosmic data, you only need to open a work directory
and
- get_files InDetCosmicPixelEndRec_topOptions.py
- Make two changes to the python:
- InDetCosmicFlags.doSim = False
- FullFileName = "/castor/cern.ch/atlas/testbeam/pixel/2006/daq_Appl_PixelEB01_daqdev2_0001129_file01.data"
- Run Athena athena.py InDetCosmicPixelEndRec_topOptions.py
will output an AANT ntuple.root
- You can then change the default 100 event limit and change output
file names for the real jobs.
- The reconstruction is only outputting the AANT instead of CBNT which
is quite different in storage format so that the analysis code unfortunately
has to adjust significantly.
- Browsing the AANT directly with root also needs caution. The pixel
and even track variables can be plotted directly with the root tree browser,
while the cluster variables can show all==0 in the direct browsing plots.
However if you go ahead to analyze with the proper access code (see the
Official cosmic reconstruction instructions InDetCBNTTree example), the
cluster data is actually there fine.
- The raw pixel data itself has changed from real numbers to integer
vectors (which is a sensible thing), but there is an annoying change that
the old bunch crossing ID BCID of 0-15 ticks is now kept in a new variable
called LVL1, while the new BCID is recording a number 0-255 as some kind of
global tag.
Data samples
The raw data files are at /castor/cern.ch/atlas/testbeam/pixel/2006
in CASTOR at CERN. There is a
brief summary table of runs, but not clear the descriptions are up to date
or accurate. There is a more detailed
data
taking log describing the run conditions.
Analysis
The PixelRootUtil code can be found in the group/slac cvs repository. This
is still to be updated with David's PixelAnanlysis class then migrate to release
13.
The AANT root files are unfortunately only read in properly the 64bit SLC4
platform for root access so that annoyingly you have to switch to the 64bit mode
with source ${HOME}/cmthome/setup.csh -tag=13.0.10,64
for root analysis, or follow the 64bit root redirection in the
Official cosmic reconstruction instructions.
One nasty thing already detected from a test sample was that very
occasionally a pixel would turn up with all bits set in raw data which would say
it is layer=3, phi=63. This needs to be filtered out to prevent crashes, but the
real cause of this is clearly important to track down.
Su Dong