Background rejection

Physics ─ J/Ψ

Z Vertex Trigger

DAQ Compliance

GTT Architecture

The ZEUS Trigger

STT Data

MVD datapreprocessing

Functional building

Pattern Recognition

AndSeed Production

Track Fitting(Next step)

MVD Data

STT1 sector STT2 sector

HERAHERA after upgrade:Ep = 920 GeVEe = 27 GeV LDesigned=1.5*10 31 cm-2s-1

Three experiments are running:ZEUS, H1, HERMES

ZEUS

ZEUS

ZEUS Components:

1. CTD2. MVD3. STT4. F/R TD5. F/B/R CAL 6. F/B/R Muon System7. BAC8. SRTD9. VETO

The ZEUS Global Tracking Trigger.

The current design, implementation and performance of the ZEUS Global Tracking Trigger (GTT) Barrel Algorithm are described.  The ZEUS GTT integrates track information from the ZEUS central tracking chamber (CTD) and micro vertex detector (MVD) to obtain a global picture of the track topology in the barrel region (─1.5 < η < 1.5) of the ZEUS detector at the Second Level Trigger (SLT) stage.  Algorithm processing is performed on a farm of Linux PCs and, to avoid unacceptable deadtime in the ZEUS readout system, must be completed within the strict requirements of the ZEUS trigger system. The GTT plays a vital role in the selection of good physics events and the rejection of non─physics background within the very harsh trigger environment provided by the upgraded HERA collider.  The GTT Barrel Algorithm greatly improves the vertex resolution and the track finding efficiency of the ZEUS SLT while the mean event processing latency and throughput are well within the trigger requirements. Recent running experience with HERA production luminosity is briefly discussed.

The design, implementation and performance of the ZEUS Global Tracking Trigger (GTT) Barrel and Forward Algorithms are described.  The ZEUS GTT Forward Algorithm integrates track information  from the ZEUS Micro Vertex Detector (MVD) and forward Straw Tube Tracker (STT) to provide a picture of the event topology in the forward  direction (1.5 < η < 3) of the ZEUS detector at the SLT stage. This region is particularly challenging because of inhomegenieties in the solenoid magnetic field, and the high occupancies in the forward direction from beam gas interactions and secondary scatters with the ZEUS beampipe.  The Forward Algorithm is distinct from the GTT Barrel Algorithm,  but runs in parallel on the GTT CPU farm. To avoid unacceptable  deadtime in the  ZEUS readout system, the Forward Algorithm processing  must be compliant with the strict requirements of the  ZEUS trigger system. The current status of the integration with the ZEUS DAQ and trigger systems is also reviewed.

► Pair parallel STT sectors with same azimuthal angle and straw position. ► Within straw size window construct max and min slope straight lines trough all fired straw.► Build single projection envelope.► Collect information from all single projection envelopes to a functional.► Maximum of functional gives vertex prediction from STT

► Decode MVD data: cluster information.► “2D” hit construction: forward wheel module geometry.

► Hough transformation of fired straws, using vertex prediction.

► Find spikes in 2D phase space.

► Segment finding in CTD. Using drift time data to break hit ambiguity, take segment candidates pointing to the beam line. ► Axial track finding in rφ plane. Seed segment from outer CTD superlayer. Matching internal segments with beam line constrain and fit track with fast circle fit in rφ plane. Match hits from MVD to the track and refit track, using MVD hits.► Stereo track finding in sz plane ( s: path length in rφ plane). Using information from stereo layers to define z positions of hits. Match hits to found track in rφ plane and refit the track with matched hits. ► MVD hit matching. Match hits from MVD to the track and refit track, using MVD hits.

Decode MVD data: cluster information.► “2D” cluster location: wafer geometry and orientation.

► Usage of binning algorithm with overlapping bins.

Three level ZEUS Trigger System :

► First Level Trigger (FLT) based on deadtime free hardware pipeline clocked every 96ns by bunch crossing signal.

► Second Level Trigger (SLT) based on a distributed transputer network. First intercomponent trigger evaluation.

► Third Level Trigger (TLT) based on a PC farm performing event reconstruction using ~offline code.

GTT Hardware:

► MVD readout: 3 Motorola MVME2400 450MHz

► CTD, CTDZ and STT interface NIKHEF-2TP VME-Transputer Motorola MVME2400 450MHz

► PC farm: 12 DELL PowerEdge 4400 Dual 1GHz

► GTT/GSLT result interface: Motorola MVME2700 367MHz

► GSLT/EVB trigger result interface: DELL PowerEdge 4400 Dual 1GHz DELL PowerEdge 6450 Quad 700 MHz

► Network switches: 3 Intel Express 480T Fast/Giga 16 port MVME = Fast and PC = GigaEthernetPC and switch hardware provided by Intel Corp.

► The STT was installed, in a forward endplug, during 2001 upgrade.

► STT (provides rφ information) ● 11000 straw tube connected ● 48 trapezoidal sectors ● three layer of straws per sector ● resolution ~ 200 μm ● 6º–25º polar coverage ● radii inner 120 mm, outer 800 mm, length 600 mm ● ideally 24 hits per track

► The silicon MVD was installed, inside CTD, during 2001 upgrade.► Forward MVD (provides rφ information) ● 4 wheels of 28 silicon sensors ● 480 strips per sensor with ~30 μm resolution ● 7º–20º polar coverage ● radii inner 70 mm, outer 160 mm, length 410 mm ● ideally 8 hits per track ► Barrel MVD (provides rφ and rz information) ● 3 layers of silicon sensor ladders ● rφ and rz planes connect to the same readout channel ● 512 strips per sensor with ~30μm resolution ● radii inner 60mm, outer 160 mm, length 622 mm ● ideally 6 hits per track

CTD is a large cylindrical drift chamber.

► CTD ● 4608 sense wires in 9 superlayers ● odd numbered layer provides rφ information ● even numbered layer provides rz information ● radii: inner 170 mm, outer 850 mm, length 2050 mm ● field 1.43 T, resolution ~ 200 μm ● ideally 72 hits per track

e± 27 Gev p 920 Gev

Barrel MVDForward MVD

Why a Global Tracking Trigger ?

► New tracking detectors, STT and MVD, added.

► Pre upgrade: SLT tracking from CTD only. Vertex resolution ~ 9 cm.

► MVD cannot contribute to FLT; SLT possible.

► Combining MVD , CTD and STT information at hit level improves track finding, vertex resolution, etc. at SLT.

MVD datapreprocessing

Track Finding

Primary Vertex Finding

Secondary VertexFinding

(future step)

MVD Data

CTD Data

Ref.: S. Goers ” The Straw Tube Tracker of the ZEUS-Detector at HERA“ at IEEE IMTC 18-24/05/2004 Ref.: E. N. Koffeman et al., NIM A473, 26 (2001)Ref.: B. Foster et al., NIM A338, 254 (1994)

MC data: Latency ~ 4ms

► The Forward Algorithm in being development. First results are presented.► Next steps: add FMVD information to the Algorithm and develop track fit.

Real data: Latency ~ 1ms

D* MC data: Vertex sigma

~8cm

Real data:vertex

Straw Tube TrackerMicro Vertex DetectorCentral Tracking Detector

GTT components: CTD, MVD and STT

Forward AlgorithmBarrel Algorithm

GTT Barrel and Forward Algorithms

Event BuilderEvent Builder

STTSTT CTDCTD

Third Level TriggerThird Level Trigger

cpucpucpucpu cpucpu cpucpu cpucpu cpucpu

Offline Offline StorageStorage

Global Second Global Second Level TriggerLevel Trigger

GSLT Accept/RejectGSLT Accept/Reject

Global First Global First Level TriggerLevel Trigger

GFLT Accept/RejectGFLT Accept/Reject

CTDCTDFront EndFront End

STTSTTFront EndFront End

Other Other ComponentsComponents

Other Other ComponentsComponents

CTDCTDSLTSLT

STTSTTSLTSLT

STTSTTFLTFLT

CTDCTDFLTFLT

8Hz8Hz

80Hz80Hz

500Hz500Hz

101077 Hz Hz

Eve

nt

Bu

ffer

s

Eve

nt

Bu

ffer

s

55 s

pip

elin

es

pip

elin

e

55 s

pip

elin

es

pip

elin

e

~0.

7 s

~10

ms

CTD/CTDZ/STT interface MVD readout

PC farm and switches

GTT/GSLT interface EVB/GSLT result interface

Results

Summary► GTT latency at SLT within CTD─SLT envelope.

► Mean GTT latency ~ 10 msec.

► GTT now in use with physics triggers (GTT performance superior to CTD─SLT).

► Barrel Algorithm stable development.

► Forward Algorithm first results seen.

► Next steps: include MVD into Barrel and Forward Algorithms .

D.Gladkov on behalf of the ZEUS GTT group

Reconstruction Steps:Reconstruction Steps:

► Next steps: add BMVD information to the Algorithm and develop track fit.

► No data size cuts are required to control latency ► Data size cut made, but needs tuning ► Data size cut made, algorithm under development

► GTT latency at SLT within CTD─SLT envelope.

► Mean GTT latency less than CTD─SLT.

► GTT Latency stable at 500Hz FLT rate.

► D* MC sample: sigma ~ 8 cm., latency ~ 1ms.► First results from Forward Algorithm running online ► Next step: Forward GTT vertex efficiency study

► Single electron track reconstruction efficiency ~80% ► Next step: improve reconstruction for low pT tracks

Efficiency => run by run

J/Ψ → μ+ + μ- ( e+ + e-)3000 events

Dmitri.Gladkov@desy.de

GTT ─ OFFLINEReal Data

http://mvddaq.desy.de/~mvddaq/GTT/