Egamma/JetMET/Muon OpenHLT validation in 312


  • Sample: 312 W->enu RelVals: /castor/

Issue 1:

  • For non-isolated triggers, the track isolation is set to 9999 in the emulation. So...

I think this:

            if ( ohEleTiso[i] < Tiso && ohEleTiso[i] != -999.) 

should be changed to something like this:

           if ( (ohEleTiso[i] < Tiso && ohEleTiso[i] != -999.) || (Tiso == 9999.)) 

otherwise we reject events (which should be kept for non-isolated triggers) where the track isolation is not calculated and set =-999 in the ntuples. This leads to some events that pass HLT but fail OpenHLT.

  • This is implemented as of the tag RateEff-02-01-02

Issue 2:

  • After fixing issue 1, agreement's pretty good in the barrel & inner part of EE, but we're missing something in the eta ~2.5-3.0 region (inside the ECAL but outside the tracker?).
  • Esp. for photons (right plot) the OpenHLT keeps any events with photons above threshold, but the real HLT seems to have a sharp cutoff in eta (but also a few events out to eta=3.0).
  • Electrons are better (left plot), I guess because they're restricted to the ~tracker acceptance anyway? But still the OpenHLT picks up some extra events around the edges.
  • From Roberto C. - all egamma paths have an acceptance cut of |eta| < 2.65 on the supercluster. This is implemented as of the tag RateEff-02-01-02

Issue 3:

  • After fixing Issue 1 and restricting to the central region, the single and double electron triggers seem to have no events that pass HLT but fail OpenHLT, for a few hundred events. There still seems to be a small (~1%) excess of events that pass OpenHLT but fail HLT. Examining a few events it's not obvious why, but it looks like for at least one of the electrons the "pixel seeds" variable is very large (>10) in these events. Don't know if these 2 things are related - to be investigated...


  • From Florent - should have 100% agreement (or at least all differences are understood) in all JetMET paths on ~10k ttbar events as of RateEff-02-01-02
  • Issues found
    • Found 1 event with a discrepency in QuadJet30. Traced to a jet with eta = 5.02. Added an explict |eta| < 5 cut to OpenHLT jet triggers.
    • Found some disagreement in DiJetAve and SumHT paths. Traced to the fact that the real HLT has always used ET, while OpenHLT uses pT for these. JetMET really wants pT for the DijetAve triggers, so will be changed in the HLT. Still deciding on SumHT.
    • Found cuts of delta-phi > 0 and veto on a third jet in the DiJetAve HLT paths that weren't in the OpenHLT. JetMET doesn't want these - will be disabled in the real HLT.


  • 100% agreement for L1-only paths on Zmumu and Wmunu samples.
  • ~2% discrepencies in several L2 and L3 single and double muon paths.
    • Some (all?) of these traced by Jean-Roch to L2 volunteers in OpenHLT. Can turn on delta-R L1-L2 matching, but in some events this still fails.
      • Traced to events with 2 L1 muons with the same eta/phi where only 1 passes the L1 condition. L2 starts reconstructing 2 muons, but shared-segment arbitration destroys the L2 seeded by the good L1. Added an explicit L1-L2 link variable to the ntuples in the latest version of the HLTAnalyzer - still to be implemented in the emulation macros (in progress...)

-- JonathanHollar - 2009-09-01

Topic attachments
I Attachment History Action Size Date Who Comment
GIFgif r1 manage 10.2 K 2009-09-01 - 14:06 JonathanHollar  
GIFgif r1 manage 10.6 K 2009-09-01 - 14:06 JonathanHollar  
Edit | Attach | Watch | Print version | History: r5 < r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r5 - 2009-10-14 - JonathanHollar
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 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