cdf performance and improvements: young-kee kim (u.chicago), sept.12 2005, p5 meeting0 cdf...

CDF Performance and Improvements: Young-Kee Kim (U.Chicago), Sept.12 2005, P5 Meeting 1

CDFCDFPerformance and ImprovementsPerformance and Improvements

Detectors, Triggers, Offline, and Analysis AlgorithmsDetectors, Triggers, Offline, and Analysis Algorithms

CDF CollaborationCDF Collaboration


Data Taking EfficienciesData Taking Efficiencies

Initial Luminosity (1030 cm-1s-1) Data Taking Efficiency

Thanks to the Accelerator Div.

Detector/trigger/DAQ downtime ~5%Beam Conditions, Start/end stores ~5%

Trigger deadtime ~5%: our choice

~85% of Run IIb Upgrade Projects were commissioned with beam during this period.

Record 1.4 x 1032

2002 2003 2004 2005 2002 2003 2004 2005

83.5%


Data for PhysicsData for Physics

COT Compromised Period

QCDElectroweakBTop

Sep-Dec ‘05 Shutdown

RecordedData up to Aug. 23, 2005 Recorded: 950 pb-1

Physics: 620 ~ 880 pb-1

Data up to Aug. 2004 Recorded: 530 pb-1

Physics: 320 - 470 pb-1

Store Number (Aug.23,2005)


Spring-Summer ’05 Highlights Spring-Summer ’05 Highlights (Data up to Aug. 2004)(Data up to Aug. 2004)

hep-ex/0508051

Mtop: 5 papers in preparation

Top massl+jets

Z’: 1 paper in preparation

MSSM Higgs in ’s: paper submitted

Z’ in ee

CDF Run II Prelim. (448 pb-1)

Bs Mixing pub.: wait for better results

CDF Run II Preliminary

(Semileptonic + Hadronic)95% CL Limit - 7.9 ps-1 (355 pb-1)

MtopCDF = 172.2 ± 3.7 GeV

~2% accuracy


Run II Publication StatusRun II Publication Status

• 2005 Goal:– 40 papers submitted

• 2005 Status:– 20 papers - submitted (9 published, 2 accepted)– 10 papers - drafts under collaboration’s review

• Results approved by CDF physics groups• Drafts approved by internal reviewers

– “Godparents committee” assigned for each paper.

– 27 papers - under godparents committee’s review• Results approved by CDF physics groups

year 2001

First Collision

Commissioning

2002

First

Phys. data

2003 2004 2005

so far

# submitted 4 17 20

http://www-cdf.fnal.gov/physics/pub_run2/

http://www-cdf.fnal.gov/physics/pub_run2/


Current StatusCurrent Statusandand

Preparation for FuturePreparation for Future


CDF DetectorsCDF Detectors

Performing very well.Even Run IIb Detectors! - Operational since early 2006


CDF DetectorsCDF Detectors

• Run IIb Upgrades: – Central Preshower Detector

• Replacing with a finer segmentation system• Electron tagger, / separation• Installed fall 2004

– Electromagnetic Timing• New system for rejecting beam-halo and cosmic ray• Searches with (e.g. GMSB SUSY, long-lived particles)• Installed fall 2004

For the future, tracking systems are our main concerns.


Tracking Performance: COT and SiliconTracking Performance: COT and Silicon

• COT Aging - Fully Recovered– Aging due to hydrocarbons

coating sense wires– Fixed by adding Oxygen– Fully recovered May 2004– 99.7% working!

• Silicon detector lifetime is a complex issue involving – Component failures

• ~93% powered; ~84% working + 4% recoverable in offline• Secondary vertex trigger requires 4 layers: 21 out of 24 wedges

– Beam incidents• lost ~2% of chips: conditions improved, but still concern

– Long-term radiation damage

COT Gain vs. Time

Jan.2002 Aug.2005

May 2004


Silicon DetectorsSilicon Detectors

• Radiation damage– > 90% of total radiation is due to collisions: NIM A514, 188-193 (2003)

– Bias voltage scans as luminosity accumulates• Study collected charge (hits on tracks) and mean noise• Measurements agree with predictions up to 1 fb-1.

• Efforts to increase the Silicon lifetime– Lowered Silicon operating temp. gradually from -6oC to -10oC.– Thermally isolated SVX from COT inert regions such that the silicon

can be kept cold during COT work.

Lifetime 0 10 fb-1 20 fb-1 30 fb-1 40 fb-18 fb-1

Predicted Silicon Lifetime


Offline Software & Computing ReadinessOffline Software & Computing Readiness


Data ReconstructionData Reconstruction

• Recently achieved 6 week turn-around time between data taking and availability of physics-quality data with final calibrations.– This reduced resource needs (person and computing).

• Reconstruction algorithms are stable since January 2005.– Incorporated Run II detector upgrades.– No major changes anticipated in the future.

Ave

. in

v. m

ass

at

Z p

eak

[G

eV]

Run Number (up to July 20, 2005) M(e+e-) [GeV/c2]

CDF Run2 Prelim.L=790 pb-1

CDF Run2 Prelim.L=790 pb-1

Data up toJuly 20, 2005

yellow band: ±0.5% E scale


Data ReconstructionData Reconstruction

• CDF Production executable is fast.– ~110 million events / week

– 50 streams: high pT e, , , jet, Hadronic B, J/ , J/ ee, ….

– The ave. executable time increases by x2.5 from Lpeak = 1x1032 cm-2s-1

to 3x1032 cm-2s-1. We have already demonstrated this capacity.– Plan to process all the data until the end of Run II at Fermilab.

Number of data events written to tape by Production Farm

Aug 23 - Sep 3, 2005


Monte Carlo Simulation and Production Monte Carlo Simulation and Production

• Detector simulation reaching maturity - matching data– Incorporated detector configuration changes with time– Incorporated multiple interactions for data instantaneous luminosity

• Increasing access to global computing resources (GRID philosophy) to match physics needs.– Running on worldwide computing clusters - shared with LHC

• ~100% of MC samples are generated outside of US.• Planning data analysis centers at remote sites

– Physics analyses produced with remotely located datasets– Italian inst.s, Karlsruhe: J/ lifetime, B tagging, Single top

– Worldwide computing resources transparent to physicists.

• Aim to support more computing with fewer FTEs


Alignments, Calibrations, AlgorithmsAlignments, Calibrations, Algorithms


Tracking and High pTracking and High pTT Lepton Status Lepton Status

• COT Tracking– Alignment: wire positions aligned better than 10 m– Efficiency: 99.6% (isolated tracks), > 96% (non-isolated tracks)

• Silicon Tracking– Alignment: internal - 5 m, w.r.t. COT < 10 m– Efficiency: 94% with r-, 83% with r-and z– Misidentified: 0.5% - 1.5%

• High pT Electron Identification

– Efficiency: 82-93%, Misidentified jets: ~10-4

• High pT Muon Identification

– Efficiency: 93%, Misidentified jets: ~10-4

• Numbers are stable with time, instantaneous luminosity up to 1032 cm-1s-1.


Momentum and Energy ScaleMomentum and Energy Scale StatusStatus

• Understand passive material well:• see E/p tail

• Momentum scale:• flat over a large pT range.

• MW uncertainty due to P, E scale• Run II current (Run Ib)• : 30 (87) MeV, e: 70 (80) MeV• better than Run Ib

E / p of W electrons

p / p = - 0.0013 ± 0.0001

1 / pT(GeV-1) +- mass (GeV) near Upsilon

p / p = - 0.0010 ± 0.0001

J/+- mass vs 1/pT

pp

DataMC Simulation


Forward Tracking and Lepton ImprovementsForward Tracking and Lepton Improvements

• Forward leptons– Acceptance improvements (e.g. 30% for SM Higgs)

• Forward tracking, Forward electrons(used for many analyses)

• Forward muons:– efforts in improving triggers and algorithms

||

Electrons without tracks > 90% up to || < 3

Electrons with tracks e.g. W charge Asym.

Tracks

Eff

icie

ncy


B and Tau Tagging: Status / ImprovementsB and Tau Tagging: Status / Improvements

Ongoing efforts:• Increasing efficiency / reducing fake rate (better algorithms)• increasing acceptance (forward region)

3rd generation is important: Top, Higgs, SUSY, …

Better algorithms: Neural Network

Forward

Loose (1.8% mistag)Tight (0.6% mistag)

Sec. Vertex B Tagging

Hadronic Tau

Jet

mis

-id (

%)

e

ffic

ien

cy (

%)

Evis(GeV)

Ejet (GeV)


EEJetJet Scale & Resolution: Status / Improvements Scale & Resolution: Status / Improvements

Jet energy scale uncertainty:• precision meas.s (Mtop), searches(Higgs)• now ~2.5% uncertainty for jets in top decays • further improvements:

• generators, higher order QCD• better scale for ET > 100 GeV region• complete by end of this year

Jet energy resolution:• searches (Higgs, …)• currently 17%, goal 10-11% (jets from MW =120GeV)• further improvements:

• combine track, calorimeter Info: 2%• expand cone size: 2%• b-jet specific corrections:1-2%• better algorithms: 1-2%• complete by spring 2006 bb invariant mass (GeV)

Mhiggs = 120 GeV

Scale Corrections0 50 100 150 200 250

0 50 100 150 200 250

Resolution Improvements

No corrections17% resolution

0.2

1.0

0

0.2

1.0

0

10% resolution


Algorithms for BAlgorithms for Bss mixing: Status / Improvements mixing: Status / Improvements

• Complex meas.s involving many detector systems and analysis tools– Triggering: optimized SVT algorithms– Reconstruction modes

• Currently adding Bs Ds 3 (x1.5)• Will add Bs Ds

* by summer 2006• Currently adding Bs Dsl sample

using SVT 2-track triggers (x2)– Tagging (D2)

• e, , jet charge (1.5%)• Same-side K tagging in progress (~2%)

– Decay length resolution• Improving vertex resolution by 10~20% for larger ms • Implementing into analysis by fall 2005

– Reducing systematic uncertainties by fall 2005• Hadronic modes - better calibration in flavor tagging• Semileptonic modes - better understanding of backgrounds


High Lum. Impact on Reconstruction / PhysicsHigh Lum. Impact on Reconstruction / Physics

• Understanding Tracking, B-tagging Performance at 3 x 1032 cm-2s-1

– < # of interactions > ~ 10

– Data

• vs primary vertices

• vs bunch-by-bunch lum.

– MC + multiple interactions

– Work in progress

• Developing Analysis Techniques– W Mass

MTW

0.2 x 1032 cm-2s-1

1.0 x 1032 cm-2s-1

2.0 x 1032 cm-2s-1

3.0 x 1032 cm-2s-1

pTlepton

7 8 9 10

Tracking performance vs. # verticesZ ee events


• Instantaneous Luminosity: 2 x 1032 cm-2s-1 (IIa) 3 x 1032 cm-2s-1 (IIb)– Ave # of interactions = 10, more hits / event

• Level-1: Tracking Triggers

– low pT tracks + hits from extra interactions mimic high pT tracks

– Lower purity higher Level-1 trigger rate– Upgrade: 2D to 3D tracking high purity and lower rate

• Level-2: Decision System and Secondary Vertex Trigger– Upgrade: Lower processing time higher bandwidth, more flexible

• DAQ, Level-3 computing, Data Logging:– Upgrade: higher bandwidth + event size increase

Run IIb Trigger / DAQ UpgradesRun IIb Trigger / DAQ Upgrades


DAQ / Trigger SpecificationsDAQ / Trigger Specifications

Run IIa Specification

Run IIa Achieved (typical)

Run IIb Specification

Luminosity 0.9 x 1032 0.9 x 1032 3.0 x 1032

Level-1 Accept 45 kHz 25 kHz * 30 kHz

Level-2 Accept 300 Hz 350 Hz 1000 Hz

Event Builder 75 MB/s 75 MB/s 500 MB/s

Level-3 Accept 75 Hz 80 Hz 100 Hz

Data Logging 20 MB/s 20 MB/s 60 MB/s

Deadtime Trigger 5% 5% 5% + 5% **

•Run IIa Level-1 Accept not achieved due to •higher than specified Silicon Readout and Level-2 Trigger execution times.

** Assume ~5% from readout and ~5% from L2 processing


Run IIb Project StatusRun IIb Project Status

• Trigger and DAQ Upgrades– Level-1 Track Trigger (XFT):

• Add z (stereo) info for 3D tracking - In production– COT TDC modification to achieve L2 rate of 1000 Hz (readout time)

• 12 out of 20 crates are operational.– Level 2 decision system: faster,flexible - operational since April 2005– Level 2 Silicon Vertex Trigger (SVT)

• Faster - 3 step upgrade: the first 2 steps are operational.– Event Builder: operational since August 2005– Level-3 Computing Farm

• 1st procurement(64 PCs) in place-replace current system in Nov’05• 2nd procurement(64 PCs) - ~Jan.06

– Data Logging (20 MB/s 60 MB/s)• 1st step operational (~35MB/s), complete by end of 2005

Installation & commissioning parasitically with minimal impact on operations.


Run IIb Upgrade StatusRun IIb Upgrade Status

• Very successful so far:– 85% complete– Will finish by early 2006

• Upgrade success due to:– Highly successful Run IIa detector/trigger design &

operation– Carefully targeted to specific high luminosity needs– This allowed for incremental and parasitic implementation

and commissioning with minimal impact on operations.– Some cases (e.g. COT TDC), instead of building new

detectors, we gradually improved the systems.


Physics Triggers for 3 x 10Physics Triggers for 3 x 103232 cm cm-2-2ss-1-1


Extrapolation to 3 x 10Extrapolation to 3 x 103232cmcm-1-1ss-1-1

at 3 x 1032 cm-2s-1

~3% of Level-2 bandwidth~50% of Level-2 bandwidth.

Reduce to ~10 % with XFT upgrade

• Triggers are sensitive to multiple interactions.• Measure cross section vs # of primary interaction vertices.• Calculate cross sec vs lum. using Poisson distribution of # of primary vertices.• Good agreement with bunch-by-bunch data.

Stereo confirmationof tracking triggers

Level-2 high pT electron Level-2 high pT muon (0.6 < || < 1.1)

Average luminosity (36 bunches)Bunch-by-bunch luminosity

a highly non-linear behavior

trigger rate = cross section x L


Extrapolation to 3 x 10Extrapolation to 3 x 103232cmcm-1-1ss-1-1

~1/3 of Level-2 bandwidth at 3x1032 cm-2s-1: studying further improvements

Studied triggers for “full” high pT physics program:~2/3 of bandwidth. Aim for 50% of bandwidth

Cross sections of high pT triggers (high pT e,,,jet,ET) with Level-1 upgradeCovers W, Z, Top, WH, ZH, HWW, SUSY (partial), LED, Z’


Physics Triggers at 3 x 10Physics Triggers at 3 x 103232 cm cm-2-2 s s-1-1

• Trigger Table in current operations is good to ~1.5 x 1032 cm-2s-1

– Kept improving as luminosity increases. Significant efforts!• Trigger Table for ~3 x 1032 cm-2s-1

– high pT physics program - aim for 50%– The remaining bandwidth will be given to B physics

• Strategy: Developed high purity triggers for high luminositywhile keeping inclusive heavy flavor triggers at low luminosity.

• Level-1,2 upgrades improve purity, reduce processing time.• Other tools being implemented:

– vetoing high multiplicity events

– Final decision on the trigger composition• based on physics priority

and purity of triggers

– Straw Trigger Table ready by October 2005. 0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

0 2 4 6 8 10 12 14 16 18 20 22 24

66%34%

Lpeak = 3 x 1032

In 3.5 hours,L < 1.5 x 1032

hours

L(1032)


Concluding Remarks: CDF StrategiesConcluding Remarks: CDF Strategies


Concluding RemarksConcluding Remarks

• CDF experiment is operating well. Better than ever!– Typical data taking efficiencies in the mid 80%’s with increasing inst.

Luminosity and Run IIb commissioning– All detectors are in excellent conditions– Stable offline software– Established fast calibrations, data processing scheme– Good detector simulation– MC production at remote sites

• Challenging ahead…– x2 higher instantaneous luminosity– x8 higher integrated luminosity– Resources going down

• CDF Strategies in preparation for the future– Planning ahead: we have been identifying those areas that need

further development and are beginning to address them immediately. Goal is to complete the work by early 2006.


Concluding Remarks (cont.)Concluding Remarks (cont.)

• To be done by the end of 2005 or early 2006– Complete Run IIb upgrades (~85% currently operational)

• Expected to be done by the end of this year.

– Physics trigger table up to 3 x 1032 cm-2s-1 being prepared.• Straw Trigger Table by October 2005

– Tuning simulation• Need one more iteration for analyses with L > 1 fb-1

– Calibrations and algorithms that require large resources• Reducing Jet energy scale uncertainty

– Need one more iteration for analyses with L > 1 fb-1

• Implementing algorithms for better Jet energy resolution• Improving forward tracking and B tagging

– Preparing reconstruction algorithms for high inst. Lum.• Tracking• B tagging


Concluding Remarks (cont.)Concluding Remarks (cont.)

• Looking forward to Summer 2006 conferences• Results with x3 increase in statistics over Summer 2005• Report on > 10 x Run I Luminosity !!

• The upcoming years will be an exciting time with increasing statistics

• Discovery through searches• Discovery through precision experiments• CDF Experience:

– With ~4 pb-1, Top limits set– With ~20 pb-1, Evidence paper out!– With ~65 pb-1, Discovery paper out!

• Hoping for new discovery with ~1 fb-1 • New physics could appear with every factor of 3~4.

• CDF is committed to operating well and analyze the data through 2009.


Back-up SlidesBack-up Slides


Upgrading Silicon Vertex Trigger (SVT)Upgrading Silicon Vertex Trigger (SVT)

3 step upgrade: first two steps complete and operationalAs expected, core of timing stays the same, but tails are significantly reduced.

Bigger improvements expected later.

SVT processing time (s)

Run 203624Sep.1 2005

Run 203325Aug.26 2005


Future B TriggersFuture B Triggers

• Current B triggers– High bandwidth tracks and leptons at L1, followed by SVT at L2.

• What we’ve done so far– Developed high purity triggers for high lum, while

keeping inclusive heavy flavor triggers at low lum.• 2 tracks - vetoing high multiplicity events• Transverse mass requirement• 2 tracks +

• For the future– XFT (L1) and SVT (L2) will improve purity and reduce L2 processing time.– Multi track combination - dedicated+track trigger.– Kinematic variable requirements with stereo information at L2.

• Upgraded L2 decision is extremely flexible (firmware).

• B program will not be fully active at peak lum, but we will continue to accumulate large hadronic B samples throughout the life of CDF.

cdf performance and improvements: young-kee kim (u.chicago), sept.12 2005, p5 meeting0 cdf...

Documents