-- XiangyangJU - 12-Nov-2009

EgammaPerformance Web Display and DQ E-gamma Flags

people involved:

* Yao Ming (Univ. Regina)

* XiangYang Ju (Univ. Regina)

* Kamal Benslama (Univ. Regina)

Contact person:

* Kamal Benslama (Univ. Regina),

General Description

Assessment of the quality of the electron and photon objects is performed in three regions: barrel, crack and endcap. So on the web display you will see three sub-directories (for the electron object) named: EIDB (barrel), EIDC (crack), and EIDE (endcap) under the main directory egamma. The corresponding sub-directories for the photon object are: PHOB, PHOC, and PHOE and they are also under the main directory egamma. In addition, in each sub-directory EIDX (where X = B, C, E) you will find two sub-sub-directories named: ElectronContainerMonitor and SWClusterMonitor. Under the subdir PHOX (X=B,C, E), you will find only one subdir named PhotonContainerMonitor. The corresponding SWClusterMonitor will be added at a later stage.

The ElectronContainerMonitor contain in turn 4 sub-directories named: all, loose, Medium, and tight. In the subdir all, we list many distributions of electron candidates in the electron Container. The description of these distribution is given below. In the sub dirs: loose, medium and tight, we list the same distributions after applying the loose, medium, and tight cuts.

The SWClusterMonitor contain 1) eta and phi distributions of the SW (sliding windows) clusters candidates, 2) eta, phi of these clusters after applying the energy check cut, 3) eta, phi of the SW clusters after applying the energy check cut and the track match cut. Please note that the energy cut check and the track much cut are the same cuts that are applied in order to select electrons candidates and fill the electron container. The goal here is to check if there is any problem with these cuts, which of course will affect the filing of the electron container.

The PhotonContainerMonitor contain three subdirs named: all, loose and tight. In the subdir all, we list distributions of photon candidates in the photon Container. The description of these variables is given below. In the subdirs loose and tight, we list the same distributions after applying the loose cuts or tight cuts.

So far the SW Cluster monitoring is done only for electron, but not for photons. It will be implemented later on. So under the sub-dirs PHOB, PHOC, PHOE, you will not find a sub-dir SWClusterMonitor, you will find only the 3 sub-dirs discussed above (all, loose, and tight).

The descriptions of the histograms in each sub-dir is given below, as well as, a typical distribution for this variable in cosmic data.

egammaPerformance:ElectronContainerMonitor

Histogram Name Description Expectations
hClusterEt
Click for image.
E_T of electron cluster ......................................
hClusterEta
Click for image.
Eta of electron cluster ......................................
hClusterPhi
Click for image.
Phi of electron cluster ......................................
hF1
Click for image.
E1/E = fraction of energy reconstructed in the first sampling, where E1 is energy in all strips belonging to the cluster and E is the total energy reconstructed in the electromagnetic calorimeter cluster ......................................
hF2
Click for image.
  ......................................
hF3
Click for image.
fraction of energy reconstructed in 3rd sampling ......................................
hEhad1
Click for image.
E leakage into 1st sampling of had calo (CaloSampling::HEC0 + CaloSampling::TileBar0 + CaloSampling::TileExt0) ......................................
hRe233e237
Click for image.
e233/e237 e233: uncalibrated energy (sum of cells) of the middle sampling in a rectangle of size 3x3 (in cell units eta X phi) e237: uncalibrated energy (sum of cells) of the middle sampling in a rectangle of size 3x7 ......................................
hEoverP
Click for image.
ratio of the cluster energy and the track momentum ......................................
hDeltaEta1
Click for image.
difference between the cluster eta (first sampling) and the eta of the track extrapolated to the first sampling ......................................
hDeltaPhi2
Click for image.
difference between the cluster phi (second sampling) and the phi of the track extrapolated to the second sampling ......................................
hCoreEM
Click for image.
core energy in em calo: E(core) = E0(3x3) + E1(15x2) + E2(5x5) + E3(3x5) ......................................
hHottestCellE
Click for image.
the Energy of Hottest Cell; Hottest cell which has the highest energy ......................................
hHottestCellEta
Click for image.
Eta of Hottest Cell ......................................
hHottestCellPhi
Click for image.
Phi of Hottest Cell ......................................
hNOfBLayerHits
Click for image.
these are the hits in the first pixel layer, i.e. b-layer ......................................
hNOfHighTRTHits
Click for image.
number of TRT hits which pass the high threshold ......................................
hNOfTrackSiHits
Click for image.
the sum of number of hits in SCT and number of the pixel hits, including the b-layer ......................................

egammaPerformance:SWContainerMonitor

Histogram Name Description Expectations
hClusterEta
Click for image.
Eta of the SWCluster ......................................
hClusterPhi
Click for image.
Phi of the SWCluster ......................................
hCluster_CheckDeposit_Eta
Click for image.
Eta of the SWCluster that passed fraction sampling check ......................................
hCluster_CheckDeposit_Phi
Click for image.
Phi of the SWCluster that passed fraction sampling check ......................................
hCluster_HasTrack_Eta
Click for image.
Eta of the SWCluster that has at least one track ......................................
hCluster_HasTrack_Phi
Click for image.
Phi of the SWCluster that has at least one track ......................................

egammaPerformance:PhotonContainerMonitor

Histogram Name Description Expectations
hClusterEta
Click for image.
Eta of the Cluster ......................................
hClusterPhi
Click for image.
Phi of the cluster ......................................
hF1
Click for image.
.E1/E = fraction of energy reconstructed in the first sampling, where E1 is energy in all strips belonging to the cluster and E is the total energy reconstructed in the electromagnetic calorimeter cluster ......................................
hF2
Click for image.
...................................... ......................................
hF3
Click for image.
fraction of energy reconstructed in 3rd sampling ......................................
hCoreEM
Click for image.
core energy in em calo: E(core) = E0(3x3) + E1(15x2) + E2(5x5) + E3(3x5) ......................................
hRe233e237
Click for image.
same as electron ......................................
hRe237e277
Click for image.
same as electron ......................................

DQ Flags Description

The criteria used to set the DQ flags will certainly improve with time and with a better understanding our detector. We are in the process of preparing several algorithms for the data period era, but for the moment, during the cosmic runs, we have defined two algorithms to set the DQ flags:

Algorithm 1: Empty-Not-Empty

Beside each subdir in the egamma web display explained above, you will see three possible colors: Red, Green, Yellow. Green means that everything is fine. Red means that you have a problem somewhere. Yellow means that you are between Red and Green and you need to do some work to understand what is going on. For cosmic runs, the criteria used so far to set these flags are simple. For each sub-dir, "X", we simply check that the histograms are not empty. If at least one histograms is empty, "X" become Yellow and consequently egamma become Yellow as well. If all histos are filled then everything is Green and egamma become Green. A given subdir "X" will become Red, only if the TRT FastOR trigger is used and this is checked by putting a cut on the mean value of the number of TRT high Threshold hits distribution. Egamma become Red, if at least one subdir is Red.

summary of algorithm 1:

1- Egamma is Red only in the case where the TRT FastOR trigger is used. If at least one subdir is Yellow, e-gamma become Yellow. If at least one histogram is yellow, a subdir become Yellow, and consequently, e-gamma become Yellow a well

2- please look at all the plots and don't rely on the "colors" on the web display. The criteria we are using so far to set the flags are simple and we reply on you to check each plot individually and report any possible problem.

Algorithm 2: Using Kolmogorov Test

Beside each subdir in the egamma web display described earlier, you will see four possible colors: Grey, Green, Yellow, Red . Grey means that we do not have enough statistics so we are unable to make a decision. Green means that everything is fine. Red means that you have a problem somewhere. Yellow means that you are between Red and Green and you need to do some work to understand what is going on.

To set the DQ flags we use the Kolmogorov test (Max Diff). For each histo in a given subdir "X", we apply a Kolmogorov test with respect to a reference histogram.

- If at least one histogram did not pass the test (Kolmogrov max difference >0.9), "X" become Red.

- If no histo is red and one or more histograms are yellow (Kolmogrov max difference between 0.7 and 0.9), then "X" is yellow.

- If there is no red and no yellow histo, and at least one histo is Green (Kolmogrov max difference is smaller than 0.7), then the subdir "X" is Green.

- If all the histograms do not have enough statistic (number of events less than 20), the subdir "X" is set to Grey.

Egamma is Red, if at least one subdir is Red. Egamma is Yellow, if no subdir is Red and there is at least one Yellow subdir. Egamma is Green, if there is no Red/Yellow subdir and there is at least one Green subdir. Egamma is Grey if all subdirs are Grey.

Summary of algorithm 2:

If at least one histo/subdir is Red, e-gamma become Red. The same rule apply also for the other colors, assuming the priority (from high to low): Red - Yellow -- Green - Grey is respected. This means that if one histogram is Red, the whole e-gamma will be red. If there is no Red histo/subdir and at least one histo/subir is Yellow, egamma will be Yellow. If there is no Red and no Yellow histo/subdir, and there is at least one Green histo/subdir, then egamma will be Green. The Grey color will be set only if all the histos/subdirs are Grey, which can happen only in the case of low stat as explained above.

Note: As indicated above, our best algorithm is "you": the colleague on shift. It is our responsability all to look at each plot individually and find problems, if any, before the start of the bulk reconstruction.

What to do during your shift?

1- Go to the DQ web display

2- Select the runs that were not covered by the previous shifter, as well as future runs that will show up during during your shifts

3- For each run, select the e-gamma web display. I mean here, select the directory egamma and look at each subdir individually.

4- Be aware of the algorithm used to set the DQ flags. You will always find this information here. The algorithm used to set the DQ flags is presently: Algorithm 1 (described above)

5- view all the plots there and even if you see Green everywhere, please take time to look at all the histograms individually and identify possible problems. Just to be on the safe side...

6- Document everything in the E-gamma elog

7- If the flags you see on the web display agree with your assessment, then it is good. Otherwise, you will need to update the DQ Flags.

8- Normally, the DQ group send everyday a list of runs to look at and you need to update the DQ flags for those specific runs.

9- If you are not familiar with updating the DQ flags in the database, please ask a colleague who can help you.

10- To upload the DQ flags go here

Edit | Attach | Watch | Print version | History: r14 < r13 < r12 < r11 < r10 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r14 - 2020-08-20 - TWikiAdminUser
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox/SandboxArchive All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2023 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback