TRT HT middle bit tuning


Introduction

Using only the middle of the tree TRT high threshold (HT) bits reduces the effect of pileup under higher instantaneous luminosity conditions during Run 2. Discrepancies in the modelling (timing and width of the high level signal) results in a HT middle bit fraction lower in MC by 3 - 10% depending on the TRT region. Since TRT particle identification is now based on using only the middle HT bit, the goal of this study is to tune the TRT digitization to improve the agreement between data and simulation.

The definition of HT middle bit fraction, applied in any region of interest, is as follows:

HT middle bit fraction ≡ number of HT hits with middle HT bit high / total number of HT hits

Datasets

Since HT variables are sensitive to pile-up, low <mu> data (<mu>~6) was selected for the study. Although the HT bit pattern distributions for electrons and muons are very similar, both electrons and muons are studied for completeness.

  • Data
    • data11_7TeV.periodD.physics_Egamma.PhysCont.DESD_SGLEL.repro09_v01/
    • data11_7TeV.periodG.physics_Egamma.PhysCont.DESD_SGLEL.repro09_v01/
    • data11_7TeV.periodI.physics_Egamma.PhysCont.DESD_SGLEL.repro09_v01/
  • MC
    • J/ψ → μμ sample
      • mc11_7TeV.108494.Pythia_directJpsimu2p5mu2p5.merge.HITS.e835_s1310_s1300/
    • Z → ee sample
      • mc11_7TeV.147770.Sherpa_CT10_Zee.merge.HITS.e1443_s1372_s1370/

Software

The Athena software releases to process data and MC are different since data is processed with an old release and MC has to be reprocessed with later TRT digitization versions.

  • for data
    asetup 17.2.5.2,AtlasProduction,here
  • for MC
    asetup 19.2.1.2,AtlasProduction,slc6,here

Hopefully we can have another look at this when we have the Run 2 13TeV data and MC - using the same releases

Digitization

Two TRT_Digitization tags which are related to the study.

  • TRT_Digitization-01-00-14
    • The baseline TRT digitization tag for tuning study.
    • (Most) tuning parameters are overrided in
      postInclude.OverrideTRTparameters.py
    • HT shaping is tabulated in ??''?? and has to be hard-coded.
  • TRT_Digitization-01-00-24
    • The tag which all final tuning parameters are implemented.

HT middle bit tuning

List of tuning parameters

  1. High threshold shaping function
  2. High threshold value
  3. High threshold ΔT0 shift

Tuning of high threshold shaping function

The ASBDLR has separate discriminators for tracking (low threshold) and TR photon (high threshold). The high threshold discriminator has an additional shaping circuitry with a peaking time of 11ns to preserve signal uniformity. For the tuning it is seen that the HT signal has to be stretched to increase HT middle bit fraction as shown below. A new HT shaping function (HT shape 3) has slower falling time than original and a peak time also delayed about 1.5 ns.

HTshape.png

The Figure below shows the agreement for the HT middle bit fraction between data and MC as a function of HT ΔT0 shift for different TRT partitions. The HT middle bit fraction is always underestimated when using the original HT shape for all HT ΔT0 shifts. The HT shape 3 is sufficiently large to have a good data/MC agreement.

tuningHTshape.png

Tuning of high threshold value

The high threshold discriminator settings have to be tuned to restore the increased HT probability caused by the new HT shape; the increases in probabilities are 8.9% for Barrel short straws, 9.8% for Barrel long straws, 8.0% for Endcap A-wheels and 6.9% for Endcap B-wheels as shown in the figure below. High threshold values are raised accordingly to restore the original HT probability for each TRT partitions. The table below shows the original high threshold values and the new high threshold values to cope with new HT shape.

Original high threshold New high threshold
Barrel short 0.006047 0.006576
Barrel long 0.005477 0.006016
Endcap A-wheel 0.005916 0.006390
Endcap B-wheel 0.005680 0.006074

tuningHLthreshold_HTprob.png

Tuning of high threshold ΔT0 shift

Since there is an additional shaping circuitry for high threshold, high threshold T0 setting can be differ from the overall (or low threshold) T0 setting. Tuning of high threshold T0 setting is done by an additional T0 shift on top of overall T0 setting. The shift is applied after discrimination so that it has a discrete step of 0.78125 ns Discussion: 1. The step of 0.78125 ns exists before and after discrimination - I'd simply state that this is the step size. 2. Does it matter if the shift is applied before or after the call to DiscriminatorResponse() ? I guess not since it is the middle bit we are only interested in. (See Andrew's slides for the detail). The implementation can be applied for each TRT partitions: Barrel short, Barrel long, Endcap A-wheel and Endcap B-wheel.

The following figure shows data and MC agreement for the HT middle bit fraction as a function of HT ΔT0 shift. Different colors represent each TRT partitions, and empty squares have intersection point(s) with a red dashed line which means perfect data and MC agreement. Red arrows indicate possible HT ΔT0 shifts from original digitization setting. Barrel short is straightforward since it has only one point of intersection, but the other three TRT partitions have two points of intersections. The HT bit pattern distribution can be used to determine better HT ΔT0 shifts as shown below.

tuningHTdeltaT0.png

The criterion for better HT ΔT0 shift from HT bit pattern distribution is the data and MC agreement of HT bit high patterns, especially one HT bit high patterns since they mainly affect HT middle bit fraction. It is clear that Endcaps show better agreement with HT ΔT0 shift of -4.6875 ns (Blue line) for all HT bit patterns. The optimal HT ΔT0 shift for Barrl long is chosen to have a better one HT bit high pattern agreements.

tuningHTdeltaT0_HTbitpattern.png

Final results

Below figures show HT middle bit fraction vs. straw layer for Barrel (left) and Endcap (right). Simulation with baseline TRT digitization (TRT_Digitization-01-00-14) and the new TRT digitization (TRT_Digitization-01-00-24) are shown. All final tuning parameters are applied in the new TRT digitization tag of TRT_Digitization-01-00-24. The agreement of data and simulation is very good for both Barrel and Endcap and weighted mean values of MC/data are shown below.

  Weighted mean of MC/data
Barrel short 1.0006 0.0020
Barrel long 0.9952 0.0006
Endcap A-wheel 0.9997 0.0006
Endcap B-wheel 1.0013 0.0006
final_HTmiddleBitfraction.png

There are also noticeable improvements in the HT bit pattern distributions with tuning as shown below.

final_HTbitpattern.png

References

Topic attachments
I Attachment History Action Size Date Who Comment
PNGpng HTshape.png r1 manage 299.5 K 2015-06-19 - 06:24 KyungeonChoi  
PNGpng final_HTbitpattern.png r1 manage 308.2 K 2015-06-19 - 06:24 KyungeonChoi  
PNGpng final_HTmiddleBitfraction.png r1 manage 698.2 K 2015-06-19 - 06:25 KyungeonChoi  
PNGpng tuningHLthreshold_HTprob.png r1 manage 684.4 K 2015-06-19 - 06:25 KyungeonChoi  
PNGpng tuningHTdeltaT0.png r1 manage 1024.2 K 2015-06-19 - 06:25 KyungeonChoi  
PNGpng tuningHTdeltaT0_HTbitpattern.png r1 manage 557.8 K 2015-06-19 - 06:25 KyungeonChoi  
PNGpng tuningHTshape.png r1 manage 519.8 K 2015-06-19 - 06:26 KyungeonChoi  
Edit | Attach | Watch | Print version | History: r7 < r6 < r5 < r4 < r3 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r6 - 2015-06-19 - AndrewBeddall
 
    • 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