A central place for Adam Gibson, and others if they like, to store information about cosmics commissioning, especially in the LAr. This information is spread about several different wikis currently, some of it's not publicly available at all, and it's difficult to maintain. Perhaps this should be moved some place more central than my personal area.


For some time, cosmics reconstruction has been facilitated by the RecExCommission framework and the related offline commissioning and joint reconstruction efforts described here.

Lately, the most complete releases for such reconstruction have been so-called point 1 (P1) releases. For M4 data and after releases such as and were helpful.

For the period just before M5 the P1 nightlies have been helpful, for example rel_2 from two different weeks has performed well. Details about these nightlies can be found elsewhere. Unfortunately, the definition of a nightly (e.g. rel_2) changes once a week. Now numbered P1 releases are also available, renamed versions of particular nightlies - they offer the advantage of durability. The offline commissioning page includes details on how to use the P1 nightlies.

I generally run with a release directory e.g. AtlasOffline-13.0.30_rel_2-OfflineCommissioningOct10 which contains a "runarea" directory. In the release directory I have an INSTALL.NOTES file, designed to be sourcable, which installs necessary packages and copies job options which will be used or require changing. In the runarea, I keep my top level job option file (lately and in subdirectories have all other job option files that I'd like to change. For example, RecExCommission/ and LArConditionsCommon/ The INSTALL.NOTES file also described necessary or potentially useful job option changes. Many steering options for the job, such as whether to make ntuples, ESD, or event display XML files, what files to process, etc. can be set in Of course, one could instead set these in a custom top-level job option. The release directory also contains a setup file setup_AtlasOffline-13.0.30_rel_2-OfflineCommissioningOct10 which I source each time I log in, before using the release. With some modifications (e.g. to your particular directory structure, the location of your cmthome, etc.) these scripts could be used by others as well. Examples are the setup script here and the install notes/script here.

UPDATE:, a fixed, numbered, release should perform fairly well for LAr and tile reconstruction. Its use is described in a setup script and install notes/script. It includes additional tile bad channels for masking, and improvements to xml-file creation for LAr.


Not all of the LAr has a full set of conditions data, needed for the best reconstruction of a single. For example, none of endcap C, none of the FCAL, and none of the barrel presampler has full conditions. Many other FT's, FEB's, and smaller groups of channels are also missing some conditions.

To have some confidence in what I mean when I took about a cells "energy" I generally require that all conditions were available to reconstruction, so that reconstruction could be proceed by iterative application of the OFC's, with pedestal and ramp values taken from the database.

This information is written to the upper bits of the cell quality factor. So, in CBNT we can enforce this requirement with a cut e.g. ((QCellsLAr & 0xFF0000) == 0xA50000).


Attempts at centralized reconstruction. Where to find RDO's and ntuples.

The location of the raw data has changed from time to time. For M5 they can be found at /castor/ For M4 and after they can be found at /castor/ For M3 they were at /castor/

Under recent version of reconstruction, file access is via the "FullFileName" scheme.

A possibly useful new mechanism for access to RDO's etc. is discussed in the castor slides at this phase 2 commissioning meeting.

Bin Guo has been running reconstruction with 25 sample OFC's over most recent data. Ntuples are copied to e.g. /castor/ and /castor/

/castor/ is a fairly good quality, 38 hour, run that Adam's looked at in some detail.

The M4 data was re-reconstructed by the tier 0 with a version of the LAr reconstruction that should be useful. With e.g. 25 sample OFC's. Details can be found on the joint reconstruction page.

Since M4 the RDO's have contained a string indicating the trigger type. e.g. b00000001

001 is the tile trigger

010 is the RPC trigger

100 is the TGC trigger

events that pass multiple triggers are written once to each file, and also to combined files e.g. 011 for an event passing the tile and RPC triggers.

For early M5 data the pattern was broken,,

= Correct Reality=


Tile      00000001  00000101

RPC     00000010  00000110

TGC    00000100  00000110

CTP     00001000  00000011

But, it was later fixed, with the CTP (clock) trigger added to the M4 pattern, as 1000.

Analyzing Ntuples

Descriptions of some of the CBNT variables are collected on another twiki page.

Adam generally uses MakeClass() from ROOT for a framework. Attached are snapshots of a script that masks some bad channels, makes some plots, and prints some information to stdout about cosmic candidates. The script is attached here with header file and can be run from the ROOT command line with, e.g. { gROOT->ProcessLine(".L APG.C+"); APG t; t.Loop(); gROOT->ProcessLine(".q"); }

The bad channel file should be renamed, by default, badChannels.txt. The location of the ntuples is set in the header file.

It's not been cleaned up for public consumption, so, caveat emptor.

Bad Channels

Problematic LAr channels can of course affect cosmic data. It may be that cosmic data taking and analysis is sensitive to some problems not seen by other analysis. Sometimes this just means that we're looking at cell energies, instead of more basic quantities. So, we're effected by calibrations/conditions as well as properties of the cells and electronics. And, if you want to produce a pure sample of cosmics, it's important to eliminate pathologies that will allow channels to satisfy your event selection reuqirements.


Links to maps from Pavol.

Adam and Jessica have produced lists from cosmic runs so far. Jessica more automatically, and with different methodology, but only for EMB2 and EMEC2 so far. Jessica's list includes all cells in these layersthat have a mean or RMS energy more than 3 sigma away from comparison cells.

Adam includes EMB1-3, EMEC0-3, HEC0-3, but done a bit more by hand.

Adam's and Jessica's lists have been merged, including a few channels found independently by Mark Cooke.

More details are available on a LAr Bad Channel wiki.

Bad Events

In addition to the issue of bad channels, some pathological events have also been observed in the LAr during cosmic runs. In a typical cosmic event, a few LAr cells may have significant energy deposits (e.g. six times the pedestal RMS using 25-sample OFC's). In these pathological events, instead, thousands of cells have significant positive and negative energy deposits.

In one run, 24847, there are 13 such events during the 38 hour run. No other event has more than 111 LAr hits by Adam's current definition. These events have at least 1200, and as many as 5800. Some In one numbering scheme, the pathological events in run 24847 are 16779076, 16781203, 16782531, 16784418, 16785261, 16786042, 16787836, 16789250, 16792121, 16792169, 16793674, 16796801, 16808161, and 16808920. (In more modern notation, these should be events1860, 3987, 5315, 7202, 8045, 8826, 10620, 12034, 14905, 14953, 16458, 19585, 30945, and 31704). Some of these events have dramatic effects on barrel and endcap layers, others mostly on the barrel. No significant effect has been observed in the HEC.

It's not yet clear if these events are the same "noise bursts" under study by Benjamin Trocme and Martin Aleksa.

Naively, these events would account for 2/3 of the LAr hits in the entire run! So, it's important that any plots made at the cell level be masked against such events, or they can dominate the distribution. Only 4 events in 10,000 are pathological in this fashion, so event level plots are not significantly polluted.

Similar bad events have been seen in run 28923 and correlated with "Sca out of range" errors in FEBMon. We should go back to 24847 and check FEBMon for the problematic events.


One useful page for understanding tile PMT's, cells, and their eta coordinates is

Links to maps from Sasha.

Links to lists from Ulla.

For M5 the tile will be masking cells at the reconstruction level, the code having gone into the P1 nightlies as of rel_5 or so, on or about October 18. I suspect that or may be the first useful numbered release including the masking.

Apparently there are two versions of the bad channel list so far. The M4 version includes the EBA, while the M5 does not. The EBA is not expected to be used for M5, and may come back afterwards with a different set of good and bad channels. I'm not sure if there are other differences between the M4 and M5 lists (e.g. reflecting different coverage, or with newly appeared bad channels, or channels that have been repaired, etc.)

In preliminary checks of M5 data, the default tile masking does a pretty good job. One quite bad cell was seen in run 28923: Layer A, eta 0.75 (A8) phi 3.0925054 (LBA32?).

Another cell sometimes has very negative energies in 28923, but isn't nearly as bad as the first: Layer A, eta 0.5500000 (A6) phi 1.8162332 (LBA19?).

These have been reported to the tile folks, but it's not clear that they'll be added to the automated masking. They prefer to verify problems in calibration runs.

Beyond individual very bad channels, the tile seems to have a largish number of channels (~10 groups of ~3 channels each) that contribute to non-Gaussian tails in the energy distribution.

-- AdamGibson - 22 Oct 2007

Topic attachments
I Attachment History Action Size Date Who Comment
PowerPointppt 2007-10-27-Run28923.ppt r1 manage 899.0 K 2007-10-27 - 18:04 AdamGibson Quick Look at M5 run 28923
PowerPointppt 2008-03-17-Run43868-L1Calo-badChannels.ppt r1 manage 568.5 K 2008-03-17 - 10:57 AdamGibson A look at some bad channels and a bad event in Run 43868 L1Calo triggers
C source code filec APG.C r1 manage 49.9 K 2007-10-22 - 22:01 AdamGibson Snapshot of an ntuple analysis
Header fileh APG.h r1 manage 34.1 K 2007-10-22 - 22:08 AdamGibson  
Unknown file formatnotes INSTALL.NOTES r1 manage 4.6 K 2007-10-22 - 22:29 AdamGibson Installation notes for a release based on a 13.0.30 P1 nightly release
Unknown file format7 INSTALL.NOTES_13.0.30.7 r1 manage 3.7 K 2007-10-26 - 17:28 AdamGibson Installation notes for a 13.0.30 P1 numbered release
Unknown file formatout allButCrash_BadChan_v0_1_3.out r1 manage 4075.8 K 2007-10-22 - 22:22 AdamGibson Analysis output run 24847, with bad channel list v0.1.3
Unknown file formatroot allButCrash_BadChan_v0_1_3.root r1 manage 1313.9 K 2007-10-22 - 22:26 AdamGibson Analysis output run 24847, with bad channel list v0.1.3, including cell energy plots of each channel.
Texttxt badChannelNotes.txt r1 manage 0.8 K 2007-10-25 - 15:27 AdamGibson Description of bad channel text files.
Texttxt badChannels_v0_1_3.txt r1 manage 151.6 K 2007-10-22 - 22:13 AdamGibson Cosmics bad channel list v0.1.3
Texttxt badChannels_v0_1_3_WalterFormat.txt r1 manage 12.9 K 2007-10-22 - 22:15 AdamGibson Cosmics bad channel list v0.1.3, Walter's format
Unknown file format7-offlinecommissioningoct26 setup_AtlasOffline- r1 manage 0.8 K 2007-10-26 - 17:29 AdamGibson Setup script for a release based on a 13.0.30 P1 numbered release
Unknown file format30_rel_2-offlinecommissioningoct10 setup_AtlasOffline-13.0.30_rel_2-OfflineCommissioningOct10 r1 manage 0.9 K 2007-10-22 - 22:30 AdamGibson Setup script for a release based on a 13.0.30 P1 nightly release
Edit | Attach | Watch | Print version | History: r14 < r13 < r12 < r11 < r10 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r14 - 2008-03-17 - AdamGibson
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback