Tracking in the High Level Trigger

Tracking in 2015

Running the tracking only HLT tests

There is a suite of qmtests available at Hlt/Moore/tests/qmtest/moore/physics/tracking

To run them standalone go to Hlt/Moore/tests/options and do:

gaudirun.py Moore_Hlt1And2.py Input_Default.py Moore_Hlt1Only.py hlt1_trackingonly-threshold.py --option="from Configurables import Moore; Moore().OutputLevel=VERBOSE"

you can check if the TrackReports RawBank has been created by running

gaudirun.py --option="from Gaudi.Configuration import *; from Configurables import RawEventDump; ApplicationMgr().TopAlg=[RawEventDump()]; from GaudiConf import IOExtension; IOExtension().inputFiles(['hlt1_reqhlt1.raw']); from Configurables import LHCbApp; LHCbApp(); " 

more details you can get by actually running a decoder and dumping the tracks:

gaudirun.py --option="from Gaudi.Configuration import *; from Configurables import RawEventDump, HltTrackReportsDecoder, DumpTracks; ApplicationMgr().TopAlg=[HltTrackReportsDecoder('VeloDecoder',OutputLocation='Event/Track/Velo',SourceID=1), HltTrackReportsDecoder('ForwardDecoder',OutputLocation='Event/Track/Forward',SourceID=3), DumpTracks(TracksLocation='Event/Track/Forward')]; from GaudiConf import IOExtension; IOExtension().inputFiles(['hlt1_reqhlt1.raw']); from Configurables import LHCbApp; LHCbApp(); " | tee decode.log

HLT2 is run like this:

gaudirun.py Moore_Hlt1And2.py Input_Default.py Moore_Hlt2Only.py hlt2_trackingonly-threshold.py --option="from GaudiConf import IOHelper; IOHelper().inputFiles([],clear=True); from Configurables import Moore; Moore().inputFiles = [ 'hlt1_reqhlt1.raw' ]; from Configurables import HltConf ; HltConf().PruneHltANNSvc = False
Moore().OutputLevel=VERBOSE;Moore().EnableTimer=True"

You can switch between decoding the Hlt1 tracks in HLT2 or redoing the tracking from scratch with these options:

from Configurables import Hlt2Conf; Hlt2Conf().Hlt1TrackOption = 'Decode' 
from Configurables import Hlt2Conf; Hlt2Conf().Hlt1TrackOption = 'Rerun' 

2015 Tracking Sequence

Discussion: HLT Tracking beyond LS1

Here is a collection of ideas for possible improvements/changes to the tracking after LS1. Developments have to further he goals of bringing online and offline reco closer together to reduce systematic biases, as well as improving the selectivity of the trigger. Ideas are fairly unstructured at the moment and there clearly are overlaps. Feel free to add (but not remove!) stuff.

Topic Description Material Contacts
Tracking Performance Monitoring Add HLT test running tracking sequence into JENKINS Savanna Task   Sebastian, Rob, Stefan
Hlt timing and rates on 2012 and 2015 MC Look at "2012 Hlt" tracking performances on 2012 and 2015 minimum bias MC   Simone, others?
Seeding performance What about IT only? or IT and OT separately (e.g., excluding tracks shared between both regions)? Getting initial feedback from the tracking group Simone, Michael, others?
Forward tracking performance Review forward tracking code for performance improvements.   Sebastian, Manuel(?)
Review online/offline differences Compile a list of differences Talk by Sascha showing comparisons online/offline for Ks reconstruction: Gerhard(?), Sebastian, Simone
Review tracking sequence Upgrade all tracks? If not, can the full Kalman be run on upgraded tracks? Can we run the same Kalman in Hlt1 and Hlt2? See also this TASK Savannah Task   Gerhard, Vava
Simone, Michael
Partial IP cut Can we live without IP cuts for ery high PT tracks?   Sebastian, Simone
Add FixMomentum Kalman filter right after FastVelo in HLT1 Test if this makes it feasible to use IPCHI2 in HLT1 alreday   Tim, Sebastian
VeloTT momentum estimate Backporting developments from UT studies to TT. Adapt forward pattern reco to asymmetric momentum estimate from TT. Talk by Espen Espen, Sebastian
D candidates for Alignment The alignment team needs D candidates for alignment    
Use HLT1 tracks for RICH mirror alignement can we use a special HLT1 track selection for the RICH alignment. Put this into the calibration stream. Talk by Andrew  
Downstream tracking hit flagging Hits are lost to ghost tracks due to tagging as "used". Cut on quality of long tracks before tagging? Talk by Paul on tracking with 25ns bunch spacing: Sascha, Mika, Paul,
Ghosts in HLT1 Can we cut on ghost probability in HLT1? See TASK   Sascha
How can we reuse tracks fitted in HLT1 in HLT2? How much information can be stored (and resurrected) between Hlt1 and Hlt2? Can Hlt2 profit from better seeds? Savannah Task   Sebastian, Simone
Propagating info from HLT1 to HLT2 Which constraints to the amount of information is set by online data rates? What is the coding/decoding overhead? TrackReports implemented Sebastian, Gerhard, Tomasz

Scenarios

Given: FullVelo; reuse tracks in HLT2;

We do upgrade all tracks above pt=0.5...0.8 GeV

We want lifetime unbiased --> Golden exclusive Modes in HLT1 (mass cuts, lifetime cut)

Rest of the rate with single track with IPCHI2 cut

Possibility: For lower pt cut a displaced vertex trigger would be awesome (Inclusive D* in HLT1??)

PV after Forward

FullVelo + ForwardUpgrade + PV on long tracks (no Kalman fit)

is compatible with full foward pattern recognition --> seeding HLT2 with HLT1

Seeding

FullVelo + FixMomentum +PV on VeloTracks + SimpleSeeding (IT,OT separately without shared regions) + Matching (no Kalman fit)

seeding HLT2 with HLT1

PV with Improved VeloTracks

FullVelo + FixMomentum + PV on velo tracks + ForwardUpgrade + Kalman fit

makes only sense if you use the improved PV to reduce the number of tracks that are forwarded --> only partial seeding of HLT2

Budget

What is CPU / Disk budget?

2.0\pm0.2 x current CPU / 8PB ????

Differences Online / Offline Tracking -- Status

Online Tracking Offline Tracking
HLT1 Velo tracking Full Velo
HLT2 Velo tracking  
Track quality cuts: TrackLenght....  
PV from Velo tracks PV from long tracks (clone killed)
VeloTT optional / confirmation mode VeloTT included
Forward thresholds: ... Forward thresholds: ...

Second forward loop optional

Second forward loop always

Kalman iterations: Kalman iterations:

HLT Tracking Benchmark

The following quantities are used to benchmark the HLT tracking (Brainstorming)

Please add your suggestions to the table below:

Quantity Tag Algorithm Comment
IP resolution IPchi2 ?? Good for overall performance measure
"Radial" impact parameter DOCAZ Example Inefficiency for "off-axis" tracks from secondary vertices
Edit | Attach | Watch | Print version | History: r29 < r28 < r27 < r26 < r25 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r29 - 2014-09-07 - SebastianNeubert
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    LHCb 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