1. LCWS05 Stanford - P. Le Dû 1 What next  Snowmass Menu Basic parameters and constraints Detector concepts Front end electronics and read out issues Data collection architecture model SoftwareTrigger concept Snowmass program  suggestions By P. Le Dû LCWS05 Stanford pledu@cea.fr

2. LCWS05 Stanford - P. Le Dû 2Conditions 2 x 16 km superconducting Linear independant accelerators 2 interaction points  2 detectors Energy – nominale : 500 Gev –maximum : 1 Tev IP beam size ~ few  m L= 2. 10 34 cm -1 s -1 Recent International decision: « cold » machine ‘ à la Tesla’ repetitian rate 5 bunches per train2820  x 2 ? bunch separation337 ns train length950 ns train separation199 ms -> long time between trains (short between pulses) The LC is a pulsed machine Machine parameters close to  // 199 ms 1ms 2820 bunches 5 Hz

3. LCWS05 Stanford - P. Le Dû 3 Jets & leptons are the fundamental quanta at the ILC. They must be identified & measured well enough to discriminate between Z’s, W’s, H’s, Top, and new states. This requires improving jet resolution by a factor of two. Not trivial! Charged Particle tracking must precisely measure 500 GeV/c (5 x LEP!) leptons for Higgs recoil studies. …. This requires 10 x better momentum resolution than LEP/SLC detectors and 1/3 better on the Impact Paramater of SLD! To catch multi-jet final states (e.g. t tbar H has 8 jets), need real 4  solid angle coverage with full detector capability. Never been done such hermiticity & granularity! ILC vs LEP/SLD 2 Jets separation LEP-like resolution ILC goal

4. LCWS05 Stanford - P. Le Dû 4 ILC vs LHC Less demanding ….. LC Detector doesn’t have to cope with multiple minimum bias events per crossing, high rate triggering for needles in haystacks, radiation hardness…  hence many more technologies available, better performance is possible. BUT  LC Detector does have to cover full solid angle, record all the available CM energy, measure jets and charged tracks with unparalleled precision, measure beam energy and energy spread, differential luminosity, and polarization, and tag all vertices,…  hence better performance needed, more technology development needed. Complementarity with LHC  discovery vs precision The ‘Particle flow’ paradigm in calorimeters !  LC should strive to do physics with all final states. Charged particles in jets more precisely measured in tracker Good separation of charged and neutrals

5. LCWS05 Stanford - P. Le Dû 5 Detector concepts Large gaseous central tracking device (TPC) High granularity calorimeters High precision microvertex All inside 4T magnetic field LDC GLC SiD Si Strips SiW EM 5 Tesla Large Gazeous Tracker  JET,TPC W/Scint EM calor. 3 Teslas solenoid

6. LCWS05 Stanford - P. Le Dû 6 Evolution of basic parameters Evolution of basic parameters Sociology Exp. UA’s LEP 3 µsec 10-20 µsec - - 250 - 500K - - - - - - 10 Mbit/sec 5-10 MIPS 100 MIPS 150-200 300-500 Collision rate Channel count L1A rate Event building Processing. Power 1980 1989 Year LHC ILC 25 ns 330 ns 200 M* 900 M* 100 KHz 3 KHz 20-500 Gbit/s 10 Gbit/s >10 6 MIPS ~10 5 MIPS 2007 2015 ? BaBar Tevatron 4 ns 396 ns 150K ~ 800 K 2 KHz 10 - 50 KHz 400 Mbit/s 4-10 Gbit/sec 1000 MIPS 5.10 4 MIPS 400 500 1999 2002 2000 > 2000 ? * including pixels Sub-Detector Pixel Microstrips Fine grain trackers Calorimeters Muon LHC 150 M ~ 10 M ~ 400 K 200 K ~1 M ILC 800 M ~30 M 1,5 M 30 M

7. LCWS05 Stanford - P. Le Dû 7 Summary of present thinking The ILC environment poses new challenges & opportunities which will need new technical advances in VFE and RO electronics  NOT LEP/SLD, NOT LHC ! Basic scheme: The FEE integrates everything Basic scheme: The FEE integrates everything  From signal processing & digitizer to the RO BUFFER …  From signal processing & digitizer to the RO BUFFER … Very large number of channels to manage (Trakers & EM)  should exploit power pulsing to cut power usage during interburst  should exploit power pulsing to cut power usage during interburst New System aspects (boundaries..  GDN !) New System aspects (boundaries..  GDN !) Interface between detector and machine is fundamental Interface between detector and machine is fundamental  optimize the luminosity  consequence on the DAQ  optimize the luminosity  consequence on the DAQ Burst mode allows a fully software trigger ! Burst mode allows a fully software trigger !  Looks like the Ultimate Trigger: Take EVERYTHING & sort later !  Looks like the Ultimate Trigger: Take EVERYTHING & sort later !  GREAT! A sociological simplification!

8. LCWS05 Stanford - P. Le Dû 8 Technology forecast (2005-2015) Systematic use of COTS products  make decision at T0 minus 2-3 years! FPGA for signal processing and buffering –Integrates receiver links, PPC, DSPs and memory …. Processors and memories –Continuous increasing of the computing power More’s law still true until 2010! Expect x 64 by 2015! Memory size quasi illimited ! Today : 256 MB 2010 : > 1 GB … then ? Links & Networks: –Commercial telecom/computer standards –10 -30- 100 GBEthernet ! É -

9. LCWS05 Stanford - P. Le Dû 9 Software trigger concept  No hardware trigger ! Sub-Detectors FE Read-out Signal processing – digitization, no trigger interrupt Sparcification,cluster finding and/or data compression Buffering Sub-Detectors FE Read-out Signal processing – digitization, no trigger interrupt Sparcification,cluster finding and/or data compression Buffering Data Collection is triggered by every train crossing Full event building of one bunch train Trigger :Software Event Selection using full information of a complete train (equivalent to L1-L2-L3 …) with off line algorithms Select ‘Bunch Of Interest’ Event classification according to physics, calibration & machine needs Data Collection is triggered by every train crossing Full event building of one bunch train Trigger :Software Event Selection using full information of a complete train (equivalent to L1-L2-L3 …) with off line algorithms Select ‘Bunch Of Interest’ Event classification according to physics, calibration & machine needs On-line processing & monitoring S1S2S3S4Sn 1 ms 200 ms Few sec Data streams Read-Out Buffer Network Detector Front End Processor Farm(s) Storage 3000 Hz 30Hz 1 MBytes Average event size Data Flow up to 1 ms active pipeline (full train) Dead time free

10. LCWS05 Stanford - P. Le Dû 10 Advantages  all  Flexible –fully programmable –unforeseen backgrounds and physics rates easily accomodated –Machine people can adjust the beam using background events  Easy maintenance and cost effective –Commodity products : Off The Shelf products (Links, memory,switches, processors) –Commonly OS and high level languages –on-line computing ressources usable for « off-line »  Scalable : –modular system Looks like the ‘ ultimate trigger ’  satisfy everybody : no loss and fully programmable

11. LCWS05 Stanford - P. Le Dû 11 Physics Rate : –e+ e-  X 0.0002/BX –e+ e-  e+ e- X 0.7/BX e+ e- pair background : –VXD inner layer 1000 hits/BX –TPC 15tracks/BX -> Background is dominating the rates ! Estimates Rates and data volume High Level-1 Trigger (1 MHz) 10 5 10 4 10 3 10 2 LHCb KLOE HERA-B CDF/DO II CDF H1 ZEUS UA1 LEP NA49 ALICE Event Size (bytes) 10 4 10 5 10 6 ATLAS CMS 10 6 10 7 Btev Ktev ILC

12. LCWS05 Stanford - P. Le Dû 12 Tesla Architecture (TDR 2003) VTXSITFDT ECAL FCHTPC HCALMUONLCAL LAT 799 M1.5 M40 M300 K40 K75 K200 K32 M20 K 20 MB 1 MB3 MB 90 MB110 MB 2 MB1 MB PPPPPPPPPPPPPPPP Computing ressources (Storage & analysis farm) Event building Network 10 Gbit/sec Processor farm (one bunch train per processor) Event manager & Control Detector Buffering (per bunch train in Mbytes/sec) Detector Channels (LHC CMS 500 Gb/s) 30 Mbytes/sec  300TBytes/year Links Select Bunch Of Interest

13. LCWS05 Stanford - P. Le Dû 13 About systems boundaries …..moving due to !  evolution of technologies,sociology ….. Subdetectors FE Read out Signal processing Local FE Buffer Machine Synchronization Detector feedback Beam BT adjustment Data Collection (ex T/DAQ) Bunch Train Buffer RO Event Building Control - supervisor On line Processing SW trigger & algorithms Global calibration,monitoring and physics Local (temporary) Data logging (Disks) Full Integration of Machine DAQ In the data collection system NEW ! Global Detector Nerwork (worlwide) Detector Integration Remote Control Rooms (3) –Running modes Local stand alone Test Global RUN –Remote shifts Slow control Detector Monitoring Physics & data analysis (ex On- Off line) –Farms –GRID … Final Data storage Read out Node Partitionning (physical and logical) Read out Node Partitionning (physical and logical) Uniform interface

14. LCWS05 Stanford - P. Le Dû 14 Possible common RO architecture model Preampli Shaper Digitizer VTX CCD MAPS DEPFE T ….. TRK Si TPC ECAL SiW Other HCAL Digital Analog Muon RPC Scint VFD Detectors technology FPGA Receiver Signal Processing Buffer Local Data collection Node Standard links & protocol USB,Firewire ….. Laptop PC Board Intelligent mezzanine PC… NETWORKNETWORK Ethernet Services Synchro Calibration Monitoring ? Integration To be studied! On detector Very FE LOCALBUFFERLOCALBUFFER Common/uniform Interface

15. LCWS05 Stanford - P. Le Dû 15 ILC ‘today’ Data Collection Network model Run Control Monitoring Histograms Event Display DCS Databases... Analysis Farm Mass storage Data logging Config Manager  Local/ worldwide Remote (GDN) Synchro NO On line – Off line boundary Local/Global Network(s) Wordlwide! Machine Bx BT feedback Local partition Data collection Sw triggers Sub Detector Read-Out Node (COTS boards) FPGA receiver Buffer FPGA receiver Buffer FPGA receiver Buffer Proc receiver Buffer Data link(s)Services Networking Hub On detector Front End FPGA

16. LCWS05 Stanford - P. Le Dû 16 What next (1) ?  Snowmass Understanding in details Detectors Read out schemes –Data Collection (DAQ) is starting at the Detector level –By Subdetectors  Independently of detector concepts –Detector concepts SiD,LDC,GLD  what are the particularities? –Propose a common architecture ( VTX,TRACK, CALORs,Muon …) –Do not forget the Very Forward ! –Influence of technical aspects  Power cycling & beam RF pick-up …. Refine the s/w trigger concept  ILC T/DAQ model –Special triggers (Calibration,Tests,cosmics …) –Hardware Preprocesing ? Interface with machine –Which infos are needed? –Integration of Beam Train feedback Define clearly the ‘boundaries’  functional block diagram –Integration of GDN –‘Slow controls’ and monitoring –Synchronization –Partitionning …..

17. LCWS05 Stanford - P. Le Dû 17 What next (2)  Snowmass Milestone for 2005 ( CDR)  Baseline ILC document –Define a list of technical issues and challenges to be addressed –Which prototyping is needed? –Practicing state of the art technologies (FPGAs, Bussless,wireless?,networking hubs …) Toward a realistic costing model for the CDR (end 2005) Toward a realistic costing model for the CDR (end 2005) –Global to the 3 detectors concepts –Table of parameters: Estimate number of channels, bandwidth …. –Estimate quantity of hardware (interfaces, processors, links … ), software ? and manpower ( is LHC a good model?) –Cost effective, upgradable, modular ….. Build a wordwide ‘international ‘long term’ strong team –Seniors with LEP/SLD,Hera/LHC/Tevatron,Babar/Belle/KEK experience –Younger with enthousiasm! –Europ,North America and Asia  common meetings –Include long term sociology  NOT reinventing the wheel! Build ONLY what is needed ! Not competing with industry!....

18. LCWS05 Stanford - P. Le Dû 18Others Modelling using ‘simple tools’ ? –Quantitative evaluation of the size of the system Trigger criteria & algorithms  trigger scenarios What about GRID ??? LHC experience