Efficiency Studies for CRAFT - notes

This page collects links, numbers, plots, etc., for unraveling the puzzle of the low CLCT efficiency in CRAFT data.

planned discussion

Monday, 23-March, 15:00 (GVA)
EVO: "CSC Efficiency Discussion"
http://indico.cern.ch/conferenceDisplay.py?confId=55233

presentations

Michael, DPG Meeting, 5-Mar-2009 (plots come from Stoyan)
Stoyan, DPG-PH Meeting, 16-Feb-2009
Stoyan, DPG Meeting, 12-Feb-2009
Stoyan, DPG Meeting, 4-Dec-2008
Stoyan, DPG Meeting, 8-May-2008
Yuriy, UCLA Meeting, 10-Feb-2007
Yuriy and Stoyan, DPG Meeting, 25-Jan-2007
Yuriy, CSC Analysis Meeting, 27-Nov-2006
Alexey M, EMU Meeting, 5-Dec-2006

Also, CMS NOTE 2007/031 from Yuriy.

Follow-up plots and analysis

http://grid.ihepa.ufl.edu/~kkotov/CRAFT/

http://lotus.phys.northwestern.edu/~stoyan/work/plots/04.03.09_CSCEff/

http://lotus.phys.northwestern.edu/~stoyan/work/plots/18.03.09_CSCEff/

• top half
• bottom half

http://lotus.phys.northwestern.edu/~stoyan/work/for_andrey/ (ALCT and CLCT efficiencies for run 66912)

ALCTs per event
CLCTs per event

http://lotus.phys.northwestern.edu/~stoyan/work/tmp/inefficienEvents_printOut.txt

http://www.nuhep.northwestern.edu/~schmittm/CMS/TEMP/digi_printout_69912_part1.txt

http://www.nuhep.northwestern.edu/~schmittm/CMS/TEMP/INEFF/ (event displays)

Andrey sent along some DQM plots for run 69912. See attachments.

Laria produced some powerpoint slides which are attached to http://indico.cern.ch/conferenceDisplay.py?confId=55233. Her log file is /afs/cern.ch/user/r/rredjimi/public/LCT.log

CSCValidation plots large sample.
Andy put the root file with the detailed version of the validation in ~akub19/public/valHists_run69912_detailed.root on lxplus.

Andrey pointed to "missing CLCT" DQM plots from beam halo data (run 62232) click for plot. This shows CFEBs missing at the "far" side of the muon trajectory, with no loss of CLCTs.
These plots should be compared to the same ones for a recent cosmic ray run: click for plot.

Victor ran his binary examiner on some raw data for run 69912 and found duplicated events:

Legend: <AA> - ALCT block, <TT> - TMB/CLCT blocks, <++++++++> - CFEB block

-1------------------
L1A=5492521
ME+1/3/07 Crate=003 DMB=09 1 1 01000 { <AA>  <TT>  <++++++++> } DMB=9; 312-312-312      436-436-436
ME-1/2/18 Crate=037 DMB=04 1 1 00000 { <AA>  <TT>  }            DMB=4; 600-600-600      412-412-412
ME-3/2/20 Crate=051 DMB=10 1 1 00000 { <AA>  <TT>  }            DMB=10;600-600-600      412-412-412
ME-2/2/19 Crate=045 DMB=09 1 1 00000 { <AA>  <TT>  }            DMB=9; 600-600-600      412-412-412
L1A=5492522
ME+1/3/07 Crate=003 DMB=09 1 0 00000 { <AA>  }                  DMB=9; 312-312-312      0-0-0
ME-1/2/18 Crate=037 DMB=04 1 0 00010 { <AA>  <++++++++> }       DMB=4; 600-600-600      0-0-0
ME-3/2/20 Crate=051 DMB=10 1 0 00001 { <AA>  <++++++++> }       DMB=10; 600-600-600     0-0-0
ME-2/2/19 Crate=045 DMB=09 1 0 00110 { <AA>  <++++++++> <++++++++> } DMB=9; 600-600-600      0-0-0
-2------------------
L1A=5501822
Crate=002 DMB=02 1 1 00110 { <AA>  <TT>  <++++++++> <++++++++> } DMB=2; 408-408-408      436-436-436
L1A=5501823
Crate=002 DMB=02 1 0 00000 { <AA>  }                             DMB=2; 408-408-408      0-0-0
-3------------------
L1A=5518500
Crate=016 DMB=07 1 1 00000 { <AA>  <TT>  }         DMB=7; 600-600-600      412-412-412
Crate=008 DMB=10 1 1 00000 { <AA>  <TT>  }          DMB=10; 312-312-312      436-436-436
L1A=5518501
Crate=016 DMB=07 1 0 01000 { <AA>  <++++++++> }         DMB=7; 600-600-600      0-0-0
Crate=008 DMB=10 1 0 00011 { <AA>  <++++++++> <++++++++> }    DMB=10; 312-312-312      0-0-0
-4------------------
L1A=5525387
Crate=006 DMB=01 1 1 01110 { <AA>  <TT>  <++++++++> <++++++++>  <++++++++> }   DMB=1; 408-408-408      436-436-436
L1A=5525388
Crate=006 DMB=01 1 0 00000 { <AA>  }                  DMB=1; 408-408-408      0-0-0
------------------

One can see duplicated events.

code

CSCEfficiency resides in /CMSSW/RecoLocalMuon/CSCEfficiency

Stoyan used L1Trigger/GlobalMuonTrigger/test/L1GmtTriggerSource.cc for obtaining trigger information.

Michael's digi print-out program is in ~schmittm/public/DUMP2/CMSSW_2_2_6/src/RecoLocalMuon/MyDump/

Michael's cross-check efficiency measurement is done with code in ~schmittm/public/SKIMEFF/CMSSW_2_2_6/src/DPGAnalysis/Skims/src/CSCSkim.cc

Stoyans' analysis

requirements

1. P > 25 GeV (not pT)
2. -0.8<dy/dz<-0.1
3. |dx/dz|<0.2
4. The track must be "very good".
5. Use the stepping helix propagator.

Note: dy/dz and dx/dz are computed from the cosmic track extrapolated to the given chamber. Care is taken to correct the sign for ME3 and ME4, which are oriented with their local Z axis pointing opposite to the global Z axis.

When studying the ALCT efficiency vs. dY/dZ, the cut on dY/dZ is relaxed, obviously. When studying CLCT, the cut on dX/dZ is relaxed.

Yuriy's analysis

This is taken from his presentation on 10-Feb-2007. It is based on MTCC data.

requirements

1. trigger on ME1 or ME3 in order to study ME2
2. only one ALCTxCLCT in ME1 and ME3
3. angle cuts: 0.15 < theta < 0.4 for ME2/1, 0.4 < theta < 0.8 for ME2/2
4. "efficient" means an ALCTxCLCT is present in ME2 (no cut on number of planes)

Yuri obtained 99% efficiency in MTCC data.

Michael's analysis

The idea is to measure efficiencies in ME2/2 only, using ME1 and ME3 as tags of a through-going muon.

requirements

1. exactly one good segment in ME1 and ME3.
2. no more than 2 segments in any chamber.
3. |dPhi| < 0.2 where "phi" is computed from the global (X,Y) coordinates of the segments
4. "pointing cut" : Define the radius R = sqrt(x*x+y*y). Then we expect, roughly speaking, (R1/Z1) = (R3/Z3) if the muon track points to the origin. Demand that Rproj = R1*(Z3/Z1) is between 0.7 and 1.7 times R3.

Results from run 69912 are in the attachments.

data

Stoyan's original skimmed sample: /castor/cern.ch/user/s/stoyan/data/skim/ - about 50 events, looser than standard requirements, ALCT or CLCT missing.

Run 66912 selected by Andrey

• 5 958 651 events in the run
• 4722 cases (chambers) to analyze
• Stoyan used data set /Cosmics/Commissioning08_CRAFT_ALL_V4_ReReco-v1/RECO
• Stoyan extracted 179 events to examine:
  • files
    • /castor/cern.ch/user/s/stoyan/data/inefficientEvents_run69912/run69912_inefficientEvents_ReSkim_RECO.root
    • /castor/cern.ch/user/s/stoyan/data/inefficientEvents_run69912/inefficientEvents_run69912_RAW_[01-04].root
  • http://lotus.phys.northwestern.edu/~stoyan/work/CSC_inefficiency/ - see text files w/ detailed information
  • summary:
    • CLCT inefficient (i.e. only CLCT digi is missing) chambers : 63
    • CLCT + strips inefficient chambers : 17
    • strips inefficient chambers : 49 (+9 "single layer")
    • CLCT + ALCT inefficient chambers : 22
    • ALCT + wires inefficient chambers : 24
    • ALCT inefficient chambers : 0
    • wires inefficient chambers : 0 (+ some "single layer")
  • a few chambers seem persistent - ME+2/2/18 (CLCT), ME+2/2/14 (ALCT), ME+1/3/28 (ALCT), ME-2/2/5 (ALCT+CLCT).

One event display (see attachments below) - nice event passing through both endcaps.

facts

basic facts

1. Stoyan obtains 100% for essentially all chambers in MC.
2. the CLCT inefficiency is the worst
3. there is a clear dichotomy: upper half vs. lower half
4. Events which miss CLCT do have ALCT and CFEB data. They have no TMB data at all (no comparator hits and no correlated LCT's).
5. Dayong measures very high trigger efficiencies.
6. DQM shows no problems or features in run 66912 for chambers ME+2/2/19 and ME+2/2/34.
7. The trigger used is HLT_L1MuOpen but there are no specific CSC triggers applied.
8. DT timing for upper and lower halves is shifted.
9. known bad chambers in ME+2/2 for run 66912: ME+2/2/14 (dead ALCT), 15 (CFEB5-comp dead), 16 (CFEB3-comp dead), 18 (no TMB/CFEB data).
10. Run 66912 has same CLCT inefficiency as obtained from a larger sample of CRAFT data.
11. From examination of a sample of skimmed events, events without CLCT's usually do have strip-DIGIs, and vice-versa.
12. The timing window for strip data (CFEBs) is tighter than for CLCT's and TMB's.
13. When a probe chamber is missing the CLCT, all chambers in the event have no CLCT.
14. Laria sees LCTs in the trigger path.
15. L1A is provided by CSC for all of these events. Usually there is no other L1A.
16. A chamber is read out only if both an LCT and a L1A coincide. An LCT can be a combination of an ALCT and a CLCT, but it can also be just one or the other.
17. Andrey shows that CFEBs are more sensitive to timing than CLCTs using beam halo data.

statement on how the readout works, from Andrey:

• ALCT data would be read out if there was ALCT0xL1A coincidence (7-BX window opened by ALCT0), regardless of CLCTs and TMB (unless TMB is dead and blocks ALCT data transmission)
• For TMB data to appear, there should be ALCT0xCLCT0 coincidence (6-BX window?) and then again (ALCT0xCLCT0) x L1A coincidence (7-BX window). Both windows start in synch with ALCT0 with some preset delays. The first condition ALCT0xCLCT0 also generates correlated LCT to be sent to MPC and further on...
• For CFEB to be in readout, there should be (pre-CLCT0) x L1A coincidence, now within very tight window of 3 BXx. The window is open in synch with pre-CLCT0, unlike for ALCT and TMB readout.
• Note that CLCT0 time is defined by pre-CLCT0. So if we get ALCT and and CFEB data OK, the only way not to get TMB is to have a highly inclined

track that can form pre-CLCT0 (two planes required only), but does not form CLCT0 (four planes required). Greg correctly points out that it is so much easier to lose CFEB data due to sync problems than TMB.

observations and inferences

1. For chambers in the upper half (Y>0), the muon tracks enter the CSC's and exit through the lower DT's. For chambers in the lower half (Y<0), they enter through the upper DT's and exit through the CSC's.
2. The fact that MC gives 100% efficiency means that bending in the magnetic field and multiple scattering is very unlikely to be the explanation for this effect.

questions

1. How reliable is the momentum measurement? Could we have bad tracks in the lower half, with lots of multiple scattering, etc.?
   • one should plot p for Y>0 and Y<0.
   • For a 50% inefficiency, the tracks would have to be very different, which is not plausible.
   • Momentum cuts of 10 GeV and 25 GeV show similar behavior.
2. Why were CLCT efficiencies so high in MTCC, and low now?
   • Alexei measured MTCC efficiencies. Can Stoyan try his cuts?
   • There was no handle against multiple scattering in MTCC.
3. Is there a trigger bias of some sort?
4. is multiple scattering important?
   • According to Stoyan, the rms deflection of a 25 GeV muon is 1cm after 2m of iron. He imposes a 10cm boundary around dead zones.
5. How is dY/dZ and dX/dZ measured?
   • From the cosmic muon track as extrapolated to the chamber reference surface.
   • When strip digis are present, the angle measured in the chamber can be confirmed.
6. This effect should stick out like a sore thumb in Laria's SP-DMB comparison analysis. Has it been confirmed there?
7. If top/bottom asymmetry reflects troubles with DT triggering, one might see it in anode timing plots.

---

-- MichaelSchmitt - 19 Mar 2009