ted liu - 1 pac - 13 dec 2003 status of the run iib cdf detector upgrade project ted liu fermilab
TRANSCRIPT
Ted Liu - 2PAC - 13 Dec 2003
2b or not 2b this is no longer the question
• RunIIb spec was for 4 x 1032cm-2s-1
• All non-silicon upgrades will NOT change, the scope is still right
• a portion of silicon tasks remain to take care of current detector
New baseline accepted by DOE,endorsed by CDF internal review
Strong commitment to see all the projects through…
Goal is now for: 3 1032cm-2s-1
Ted Liu - 3PAC - 13 Dec 2003
Project Scope (I)
1.1 - Silicon Contains closeout activities, some tasks needed to
preserve the current detector (DAQ maintenance, radiation monitoring, safety maintenance)
1.2 Calorimeter 1.2.1 – Preshower Upgrade
- No scope change needed
- Significant impact on the installation plans due to silicon cancellation
1.2.2 – EM Timing- No scope change needed
Ted Liu - 4PAC - 13 Dec 2003
Project Scope (II)
1.3 Data Acquisition and Trigger 1.3.1 – TDC upgrade 1.3.2 – Level 2 Decision crate upgrade 1.3.3 – Level 1 Fast track trigger (XFT II) upgrade 1.3.4 – Event Builder upgrade 1.3.5 – Level 3 computer upgrade (buy new PCs) 1.3.6 – Silicon Vertex Trigger upgrade (details changed)
All are still needed to operate at 3 1032cm-2s-1 (design goal from the Summer 2003 DOE review).
No scope changes
Ted Liu - 5PAC - 13 Dec 2003
1.2.1: Preshower Status
Phototube status Used to be the critical path item, ending in Oct 04 Japan has accelerated orders, now delivery is March 04 All phototubes will be ready for a Summer 04 installation
Detector status All time-critical production parts ordered Ready to start production in February 04 6-month production expected based on prototype Should be ready for installation by Fall 2004
Ted Liu - 6PAC - 13 Dec 2003
Preshower Installation
Cancellation of the silicon installation and its long shutdown has implications for other projects
Preshower installation is most affected We have concluded that installation in the collision
hall is possible. Current installation schedule for both EM Timing and
Preshower requires 10 weeks Realistic manpower availability (40 hour weeks, 1 shift/day) If cannot be completed in summer 04, will be completed during 2005 summer shutdown
Ted Liu - 7PAC - 13 Dec 2003
1.2.2: EM Timing Status
Hardware is ready for installation Hardware is 4 months ahead of schedule All hardware finished & tested, ready for installation All cables, splitters, ASDs and TDCs in hand End-Plug installation completed, fully operational 4 Central Wedges completed, fully operational DAQ/online monitoring software fully working Performance excellent, < 2 ns timing resolution Installation procedures reviewed by experts Installation to be completed in the next shutdown (2004)
Ted Liu - 8PAC - 13 Dec 2003
1.3.1: TDC Status
New high speed TDC with Altera Stratix FPGAs Excellent Chicago engineer team (built many CDF boards) Design work started on Dec 2002 Core firmware design finished on June 2003 Firmware design reviewed by experts on July 2003 Layout of board close to final Backward compatible
Lots experience with TDC
testing & commissioning
within CDF
Ted Liu - 9PAC - 13 Dec 2003
1.3.2: L2 Upgrade Status (I)
Pulsar production started Hardware 10 months ahead of original schedule
One motherboard, four mezzanine and one AUX design All custom board prototypes designed&built last year Board design extensively simulated + trace analysis Have tested ALL interfaces in self-test mode (Tx Rx) Core firmware fully developed (in CVS) and tested in beam Has been used as RunIIa L2 muon interface board
for data taking (interface with legacy Alpha processor) Muon online DAQ/monitoring/offline software working Production Readiness Review done (Nov. 07, 2003) No blue wires on ALL prototypes, no revision needed
Web page: http://hep.uchicago.edu/~thliu/projects/Pulsar
Ted Liu - 10PAC - 13 Dec 2003
Mezzanine slotsAUX card
Pulsar Design modular/universal/self-testable
Pulsar
Custommezzanine
Bottom view
Pulsar design philosophy: able to interface with any user data with any link format (e.g. S-LINK or GbE) via mezzanineMany applications within & outside CDF (compatible with Atlas)
S-LINKPCIGbE
worksup to
100MHz
Top view
Ted Liu - 11PAC - 13 Dec 2003
Tested with real L2 trigger data
1.4GHz Linux PC
1.3.2: L2 Upgrade Status (II)
Pulsar PC timing
Pulsar to PC round trip timing measured (w S-Link to PCI) With Alpha L2 algorithm code running in Linux PC
Good performance (compare to that of old Alpha)
500MHzAlpha
2.4GHz Linux PC
Pulsar PC roundtrip timing
without L2algorithm
with L2algorithmrunning
PC vs old Alpha algorithm timing
Ted Liu - 12PAC - 13 Dec 2003
1.3.3: Fast Track Trigger (XFT II) Status
Lots simulation work done Have working XFT upgrade simulation
Hardware progress (in parallel with simulation) Most interfaces design do not need to change, primary changes are in the firmware (algorithm) All fully backward compatible New Linker: firmware implemented & fully simulated Finder&Linker board design begun Start production by late 2004 XFT II ready by summer 05 Three new postdocs recently joined (Sept. 2003 -)
Ted Liu - 13PAC - 13 Dec 2003
1.3.4: Event Builder Status
Technology decision made (Gigabit Ethernet) VME to switch readout: VMIVME-7805 GbE switch: Cisco 6509
Well underway: System design has been decided Final system switch ordered Readout test boards ordered Expertise on board (from both CD and CDF) Working prototype by Aug 2004 Plan to have new system ready by Summer 2005
Ted Liu - 14PAC - 13 Dec 2003
1.3.6: SVT Upgrade (SVT II)
Original plan was only to handle SVX IIb geometry additional Merger boards + new Track Fitter boards
Now main motivation is to improve SVT efficiency with good timing (critical for high Luminosity L2 latency) Redundant roads removal @ earlier stage less fits Finer roads (larger Associative Memory or AM) less fits Improve track fitting Replace obsolete boards, add additional flexibility Ensure good SVT performance all the way through 2009
Ted Liu - 15PAC - 13 Dec 2003
1.3.6: SVT Upgrade (SVT II)
Phase One: (funded by operation money) Use Pulsar as Road Warrior to remove redundant roads RW firmware successfully tested on Pulsar last month Ready for commissioning early 2004 (Pulsar production)
Phase Two: (AM++ funded by INFN R&D) Replace old AM boards with new AM++ developed for LHC Use Road Warrior to replace Hit Buffer & AM Sequencer
(mostly firmware changes, replace obsolete boards,
which also provides additional flexibility) Replace Track Fitter (RunIIb money) All backward compatible, can develop&tune standalone
Ted Liu - 16PAC - 13 Dec 2003
Conclusions
Our baseline schedule will not change We will work towards earlier completion Our target is to install as much as possible in the summer
2004 and 2005 shutdowns.
We proposed a new baseline cost for the DOE MIE of $10.4M, accepted by the DOE on 8 December.
By Fall 2005, CDF will be ready for operation at 3 1032cm-2s-1
Ted Liu - 17PAC - 13 Dec 2003
Rebaselined Cost Estimate
Contingency per subproject is from 2002 low level estimate – *scaled by use to date
New DOE MIE total cost drops - $24,987K → $10,374K All costs shown are total (M&S/Labor/Overhead), in current year $K
Baseline ($K) New Scope ($K)
Cost Cont.* Cost Cont.Silicon 12,008$ 5,145$ 2,527$ 396$ Calorimeter 342$ 335$ 342$ 335$ DAQ 3,788$ 1,678$ 3,788$ 1,678$ Admin. 1,285$ 407$ 1,006$ 302$ Total 17,422$ 7,565$ 7,663$ 2,711$
Ted Liu - 18PAC - 13 Dec 2003
New L2: Pulsar FormationPulsar pre-processors
muon
trackL1 triggerSVT
PC 0
SLINK
PC 3
PC 1
PC 2
TriggersupervisorPCISLINK
muon
cluster
electron
merger
merger
L2toTS
SLINKPCIL2 CAL
ShowMax
• Pulsar’s self-testability allows us to develop&tune the upgrade standalone• then run parasitically with CDF aim at fall 2004 • minimal impact on data taking
merger
L2 Commissioning strategy
Buffer 0 & 3
Buffer 1 & 2
Ted Liu - 19PAC - 13 Dec 2003
Pulsar will be used as Road Warrior
SVT data out
SVT data in
SVT data inSVT data in
Control FPGA
DataIO FPGA 1
DataIO FPGA 2
VME chip
SRAM
SRAM
SVT
2 SLINK
SVT input
SVT output
P2 SVT inter-commLines(5):Master&Slave
Pulsar RW will also replace Hitter Buffer and AM Sequencer
Ted Liu - 20PAC - 13 Dec 2003
Why do we want 4/5?
4/4
4/5D0
D0
source 4/4 4/5 R
D0 yield 11.2 (nb) 17.5 (nb)
1.55 ±0.06
J/psi yield 663 974 1.5
Bs MC
(Pt_b>5.5; |_b|<1.3)
1.59 ±0.05
Bs MC after offline reconstruction (Ivan F.)
1.58
Ted Liu - 21PAC - 13 Dec 2003
4/5 – upgraded 4/5 – 4/4
Upgrades:T(4/5) T(4/4) !
s
After upgrade
Current (4/5)
Ted Liu - 22PAC - 13 Dec 2003
4/5
SVT is running right now @4/5
ROAD WARRIOR + AM ++ bring us back to the old 4/4 timing, with better efficiency!
B physics Zbb
SVTconfiguration # of fits # of fits
32 kpatt 4/4 3.9 9.4
32 kpatt 4/5 (now) 39.1 94
32 kpatt 4/5 + RW 17.6 42
128 kpatt 4/5 + RW 9.3 23
Ghost removal 4.3 14.5
Ted Liu - 23PAC - 13 Dec 2003
What does the new Hardware do?
How to speed up SVT:
1. Thinner roads (larger AM) less fits.
2. Road Warrior ghosts removal
Hit Finders
Merger
Associative Memory
Hit Buffer
Track Fitter
to Level 2
COT tracks
fromXTRP
12 fibers
hits
roads
hits
x 12 phi sectors
Sequencerraw data fromSVX front end
Larger AM
Road Warrior
Single Hit
Superstrip
Road
Det
ecto
r L
ayer
s
Ted Liu - 24PAC - 13 Dec 2003
AM++
Replace old AM boards with 1 AM++/wedge Increased pattern density: standard cell chips (2K
[128] pattern/5x5 mm) Potentially larger I/O bandwidth Provide backward compatibility with older
hardware Can house potentially up to 1Mpattern!
Ted Liu - 25PAC - 13 Dec 2003
AM++ schedule
• New AM-board: summer 2004 (Pisa) during summer 2004: test with FPGA chips (Pisa)•AM-chip design: july 2004 (Ferrara-Pisa)
first chip ~2 months october• New LAMB: assemble AM-chip in october 2004 (Pisa)• test chip + board: october – december 2004 (Pisa-Ferrara)• Mass production: beginning 2005 (Pisa-Ferrara)• install: summer 2005 (Pisa-Ferrara)
Ted Liu - 26PAC - 13 Dec 2003
Impact on data taking•Boards can be completely developed and tested in test-stands
•Algorithm development & tuning may require some test runs
•Overall the experiment dead time will come from:
•Boards swapping
•Development/modification of online code
•Everything will be back-compatible: virtually no point of no-return!
Ted Liu - 27PAC - 13 Dec 2003
Flexibility
The larger AM bank allows new strategies:1. Narrower patterns to improve timing
2. Trigger bit dependent patterns
3. (L1) Lepton seeded pattern recognition
4. Standalone Si tracking
As an example we tried to merge 3. And 4. To build a forward z trigger!
Ted Liu - 28PAC - 13 Dec 2003
In a glimpse..
Selection # Z0 # L1_BMU_REAR
L1_MU 250 4678
match <2.5º
Pt>4 & 2<10
132 362
+ match 126 213
L2 Eff. 0.50
L2 Rej. 22
•Efficiency @ L2: study Z0 from data
• Back. Rejection: L1 backup from data
Ted Liu - 29PAC - 13 Dec 2003
Upgrading the Track Fitter
The current TF cannot handle 1Mpattern
(current limit is [16x]32K patterns) Current design is based on ageing components at
their fullest it would not accommodate:
Handling of a large pattern bank Handling of different patterns for different trigger strategies Handling of >4 detector layers (e.g. if we want to add lepton
ID/TOF information to the SVT fit)
Ted Liu - 30PAC - 13 Dec 2003
SVT upgradeINFN Pisa
A. Annovi, A. Bardi, P. Catastini, M. Dell’Orso, P. Giannetti, L. Ristori, F. Spinella
INFN FerraraDamiani, Sartori, R. Tripiccione, Cotta, Chiozzi
INFN TriesteS. Belforte
LBNLA. Cerri
Ted Liu - 31PAC - 13 Dec 2003
Back up slides
Man power (just in case they ask) need a full list of names for each project, but couldn’t find
the full list some projects…. Experts, please help.
Motivation (just in case we need to remind people),
although this talk is just for STATUS. misc
Backup slides are for possible questions, most of them will notshow up in the actual pdf file sent to PAC
The part will be sent to PAC is slide 1-17
Ted Liu - 32PAC - 13 Dec 2003
Calorimetry Upgrade
University of Tsukuba INFN (Pisa, Rome) JINR (Dubna) Argonne National Laboratory Michigan State University Rockefeller University FNAL
Texas A&M INFN (Frascati) University of Chicago University of Michigan Argonne National Lab FNAL
Electromagnetic TimingPreshower/Crack
Steve Kuhlmann, Level-2 Manager
Joey Huston, Level-3 Manager Preshower
Dave Toback, Level-3 Manager EM Timing
Ted Liu - 33PAC - 13 Dec 2003
TDC Upgrade: people
University of Chicago Engineers: Harold Sanders, Mircea Bogdan Physicist: Henry Frisch
Fermilab Physicist: Ting Miao Engineers:
Looking for new people to get involved
Ted Liu - 34PAC - 13 Dec 2003
XFT Upgrade: people
Ohio State University Richard Hughes, Kevin Lannon(pd), Ben
Kilminster(pd), Brain Winer University of IIIinois
Mike Kasten(eng), Suzanne Levine(gs), Ryan
Mokos (eng), Kevin Pitts, Greg Veramendi(pd)
New groups are getting involved:Purdue University (Matthew Jones et al.)Rutgers University (John Conway, Amit Lath et al.)Fermilab (engineering/technician support)
Ted Liu - 35PAC - 13 Dec 2003
Related to CDF L2 decision crate upgrade:• ANL R. Blair, J. Dawson, B. Haberichter, J. Schlereth, J. Proudfoot • FNAL R. Demaat, M. Hakala, R. Kivilahti, J. Lewis, C. Lin, T. Liu, T. Masikkala, F. Marjamaa, J. Patrick, S. Pitkanen, B. Reisert, P.Wilson • Univ. of Chicago M. Bogdan, Y. Kim, W. Fedorko, H. Frisch, S. Kwang, V. Rusu, H. Sanders, M. Shochet • Upenn K. Hahn, P. Keener, J. Kroll, C. Neu, F. Stabenau, R. Van Berg, D. Whiteson, P. Wittich
The project first started as a project (with few people) to build a test-stand tool …
Pulsar project
Ted Liu - 36PAC - 13 Dec 2003
Pulsar has attracted many good young people new generation of L2 experts !
After prototype success last year, many young people
joined the project this year: Burkard Reisert (FNAL RA, from H1): Jan. 2003 - Cheng-Ju Lin (FNAL RA, SLD): Jan. 2003 - Chris Neu (Upenn postdoc, CDF): Oct. 2003 – Vadim Rusu (Chicago postdoc, SNOW): Oct. 2003 – Dan Whiteson (Upenn postdoc, D0): Dec. 2003 – Shawn Kwang (Chicago student, 2rd year): Jan. 2003 – Wojciech Fedorko (Chicago, first year): Oct. 2003 - Kristian Hahn (Upenn, third year): Jan. 2003 – Hans Stabenau (Upenn, 2rd year): May 2003 -
Ted Liu - 37PAC - 13 Dec 2003
Core team:
Markus Klute (MIT)
Bruce Knuteson (MIT)
Ron Rechenmacher (Fermilab)
Sham Sumorok (MIT)
Steve Tether (MIT)
Event Builder Upgrade
Ted Liu - 38PAC - 13 Dec 2003
Calorimetry Upgrade Motivation
Maintain capabilities of current Preshower detector, used in over 100 papers.
Preshower expected to suffer high occupancy and aging effects in Run IIB.
Preshower and Crack detectors expected to provide 5-10% Jet Energy Resolution improvement, part of the 20-30% needed improvement for the Higgs search.
Electromagnetic timing needed to reject photon backgrounds from cosmic rays, in new physics searches such as SUSY.
Ted Liu - 40PAC - 13 Dec 2003
Electromagnetic Timing
• Virtually identical to existing system on hadron calorimeter
• Re-use electronics and well-established technologies
• Add splitters for CEM. PEM already readout-ready
• Build more ASD’s • Recycle TDC’s, crate and
tracer. Purchase new power supply and processor
Ted Liu - 41PAC - 13 Dec 2003
Motivation
The DAQ/Trigger upgrades presented here are driven exclusively by our Run IIb trigger and data acquisition needs to carry out our high-pT physics program
Our current level of understanding is based upon Run IIa data: L 2x1031 cm-2 s-1, ~1 interaction per crossing Run I data: L ~2x1031 cm-2 s-1, ~2 interactions per crossing
We are extrapolating to Run IIb L = 2x1032 cm-2 s-1 w/396ns bunch spacing (~5 int/beamX) Due to significant uncertainties in extrapolation, and a desire to be
prepared for success, we have evaluated our system for: L = 4x1032 cm-2 s-1 w/396ns bunch spacing (~10 int/beamX)
Ted Liu - 42PAC - 13 Dec 2003
Trigger Strategy
Focus on Higgs & high pT searches Know that triggers needed for these modes will allow for
many beyond Standard Model searches General requirements:
High pT electrons and muons- Associated WH/ZH modes, also tWb
Missing ET triggers- ZH with , modes with taus
b-jet triggers - , b-jets tagged by displaced tracks
Calibration triggers- , J/+, photons
Z
H bb
Z bb
Ted Liu - 43PAC - 13 Dec 2003
Run IIb Trigger Tabletrigger path
L1(nb) L2 (nb)
L3 (nb)
High ET electron 1,500 170 30Plug electron + missing ET 771 55 10High PT muon (CMUP) 1,773 200 8High PT muon (CMX) 1,773 200 82 high pT b-jets 10,840 200 10missing ET + 2jets 163 126 13jets 6,500 42 12missing ET overlap 163 3Photons overlap 50 15J/
850 38 10High PT jets 19,000 60 17hadronic top overlap 50 5di- 5,000 50 4missing ET+ overlap 50 4High ET photons 13,500 110 21dileptons, trileptons 1,000 190 45total 59,200 1904 215
rate @4E32 25kHz 750Hz 85Hz
rejection factor ~100 ~33 ~9
Ted Liu - 44PAC - 13 Dec 2003
Summary of Run IIb specifications
Level 1 Accept rate: >25kHz (spec 50kHz) deadtimeless
Level 2 Accept rate: 750 Hz bursts to 1.1kHz L2 processing deadtime < 5% readout deadtime (on L2A) < 5%
Level 3 Accept rate: 85Hz Event builder rate: 400MB/s Output data rate: 40MB/s
Reminder: trigger & bandwidth rates estimated based upon Run IIa, significant underestimate possible (assumes linear growth in fake contribution)
Ted Liu - 45PAC - 13 Dec 2003
CDF Data Acquisition System
Level 1 trigger pipelined and “deadtimeless” fully synchronous designed for 132ns operation on L1A, write data to 1 of 4
local L2 buffers
Level 2 trigger asynchronous L1 + supplemental info
Level 3 trigger full detector readout PC farm runs reconstruction output to mass storage
Ted Liu - 46PAC - 13 Dec 2003
Trigger/DAQ Upgrades for Run IIb
General considerations: upgrades “targeted” to
specific needs e.g. COT TDCs replaced, but
remaining COT readout (ASDQ, repeaters) unmodified
retain existing infrastructure cables, crates unchanged I/O protocols, timings retained upstream/downstream
components unchanged
upgrades plug compatible with existing components take advantage of knowledge &
experience will aid in commissioning
Ted Liu - 47PAC - 13 Dec 2003
TDC Replacement
Limitations of current system: TDC on-board data processing
existing system performs hit processing after L2A processing time (=deadtime) grows
with # of hits COT occupancy higher than expected Run IIa processing time too large
for Run IIb VME readout
16 TDCs per crate read out serially by VME block transfer- current VME transfer rate 14MB/s with additional overhead per board
- Run IIa, 300Hz…falls to ~150Hz (!) in Run IIb Data transfer
TRACERTAXIVRB link provides bandwidth limitation- maximum TAXI VRB is <12MB/s…Run IIb requires 14MB/s
Ted Liu - 48PAC - 13 Dec 2003
Run IIb TDC Performance
TDC (on-board) processing time [time after L2A] Now: slowest TDC >650s/event Need ~360s to achieve 1kHz
L2A rate VME readout
Currently: ~ 500s per crate Run IIb: x10 more data >1ms
Data transfer Run IIb: expect 14MB/s, TAXI link
limited to <12MB/s
Internal CDF TDC Review committee convened in June Conclusion: existing COT TDCs + VME
readout system cannot maintain necessary L2A rate in Run IIb
TDC system must be replaced OR significant modifications to the DAQ & infrastructure must take place
Specification: entire TDC readout must be completed within 600s to handle 1.1kHz rate 14MB/s.
Ted Liu - 49PAC - 13 Dec 2003
New TDC Design
Address on-board processing deadtime by moving hit processing into the L1L2 transition “hide” hit processing behind L2 trigger
Address VME and Readout problems via bypassing the VMETRACERTAXI Keep existing data path as a backup (commissioning) Maintain other pieces of DAQ chain (VRB EVB)
Design exclusive to COT system, reduces constraints Run IIa TDC will continue to work well for other systems
(muons, hadron timing, CLC)
Ted Liu - 50PAC - 13 Dec 2003
TDC Specifications Backward compatible with existing system
No change to COT front-end, cables or calibration No change to track trigger (XFT) interface Accept CDF specific signals from CDF_CLOCK/TRACER
Must handle the following rates 50kHz L1A, 1.1kHz L2A Readout time below 500s with 20kB/crate
Allow for on-board data compression Perform hit finding for track trigger “TDC Specifications” document provides details
Ted Liu - 51PAC - 13 Dec 2003
New TDC Design
Done with Altera Stratix FPGA commercially available high bandwidth differential input matches COT sufficient on-chip logic & memory to carry out all needed
functions (with room to spare) moderate price
Time-to-digital conversion performed on chip input 840MHz LVDS inputs not sensitive to routing issues remainder of FPGA functionality digital
Ted Liu - 52PAC - 13 Dec 2003
TDC Readout
A few options for TDC readout Readout by G-link: data concentrator VRB VME-PCI then to PC (commercial) VME-GbE (commercial)
TDC will support Run IIa VME readout for commissioning will be able to install new TDCs into existing system for
testing and timing resolution studies
Ted Liu - 53PAC - 13 Dec 2003
Run IIa Track Trigger System
XFT works by finding line segments in the four axial superlayers “finder” boards
Tracks are found by linking the segments into tracks “linker” boards
Stereo superlayers
(unused in XFT)
Interaction point
Ted Liu - 54PAC - 13 Dec 2003
Run IIa Track Trigger System
Lv1 trackingtrigger cable (220 ft)
168 TDCfrom COT
axial layers
24 crates
48XFT Finder
3 crates
24XFT Linker
3 crates
12XTRP
mezzanine card
on detector
1st floor counting room
Ted Liu - 55PAC - 13 Dec 2003
XFT Upgrade
Track-based triggers are responsible for >50% of the Run IIb physics program e, , , b-tags
COT occupancy at high luminosity causes significant L1 track trigger (XFT) degradation Significant growth in fake track
rate (primarily at high pT)
Degradation in pT and 0 resolution (next slide)
Minimum bias MC events
Ted Liu - 56PAC - 13 Dec 2003
XFT pT & 0 Resolution
+0 minimum bias events
Data: high pT electrons
+5 minimum bias events
+10 minimum bias events
pT 0
L = 2x1032 cm-2 s-1 @396ns
L = 4x1032 cm-2 s-1 @396ns
L = now
Ted Liu - 57PAC - 13 Dec 2003
XFT IIb Design Reduce fakes and improve resolution with
improved axial track finding & 3D information Take advantage of existing design and infrastructure
Cables, I/O, data formats unchanged
Difference between XFT & offline tracking is time binning. Segment angle match improves with finer time bins.
Upgrade: Utilize 396ns baseline to pipe more information per beam-X
from TDCXFT. Go from two time bins to six time bins in the trigger
Supplement axial tracking with stereo measurement Segment finding identical to axial XFT Stereo information provides:
- improved fake track rejection (important at high L)
- new: electron & muon matching in
Ted Liu - 58PAC - 13 Dec 2003
XFT Upgrade Performance
Additional timing information plus stereo provides high efficiency for tracks while keeping the fake rate low
Plots shown are for 10 interactions/crossing high efficiency low fake rate improved pT and resolution
Ted Liu - 59PAC - 13 Dec 2003
DAQ/Trigger Summary These are the pieces of the
CDF Front-end, Trigger and Data Acquisition system we need to upgrade/update to carry out a high pT physics program TDC replacement XFT upgrade SVT upgrade L2 replacement Event builder/L3 upgrade
The remainder of the Run Iia CDF front-end/trigger/DAQ system will perform well throughout Run IIb
Ted Liu - 60PAC - 13 Dec 2003
Event Builder/L3 Upgrade
Full detector readout occurs on Level 2 trigger accept
Event Builder subsystems send data to VRBs Event builder accepts data from
VRBs, assembles the full event
Level 3 Trigger event sent: EVBL3 PC farm single PC node per event runs
reconstruction & trigger algorithms greater rejection at L3 needed in
Run IIb
Det
ecto
r R
ead
ou
t p
ath
Ted Liu - 61PAC - 13 Dec 2003
Assumed luminosity was 4 x 1032
Justification was high-pT program
These triggers give 750 Hz
Rate spec is 1 kHz
Event size assumed to grow
Size spec is 500 kB
Throughput spec is
500 MB/sec
Event Builder Upgrade
Ted Liu - 62PAC - 13 Dec 2003
DØ has a system that runs at our spec (1 kHz, 250+ kB/event)
Put together in a year by a core team of 6 people
Made use of off the shelf, Ethernet-based hardware
Our strategy: Make use of a demonstrated, working solution to the same problem to implement a system appropriate for CDF
Ted Liu - 63PAC - 13 Dec 2003
Level 3 PC Farm Upgrade
More Level 3 PC processing power is required for Run IIb reconstruction takes longer
- higher occupancy
trigger algorithms more elaborate
- greater rejection required
sample Mean CPU time(seconds)†
t-tbar + 0 minimum bias 0.820.03
t-tbar + 5 minimum bias 2.660.10
t-tbar + 10 minimum bias 5.930.19
t-tbar + 15 minimum bias 8.320.67
† CPU time to run COT tracking on offline analysis computer (fcdfsgi2). Tracking code not optimized for higher luminosity conditions.
Purchase processors at constant rate over 3 years of project Expect processor improvements will keep up with our growth in
L3 needs throughout Run IIb increased complexity of events offset by improved processing power
Ted Liu - 64PAC - 13 Dec 2003
Nov. 12, 2002
2b or not 2b this is no longer the question
Goal is for: 3 1032cm-2s-1
• All non-silicon upgrades will NOT change, the scope is right• a portion of silicon tasks remain (take care current detector)New baseline accepted by DOE,endorsed by CDF internal review
Strong commitment to see all the projects through…