Document Sample
detector Powered By Docstoc
					G. Eigen, U. Bergen
The Homestead 6/17/00

 Introduction

    Issues of individual Subsystems
    Trigger
    Data Acquisition/Computing
    Silicon Vertex Tracker
    Drift Chamber
    Electromagnetic Calorimeter
    Particle Identification
    Instrumented Flux Return

 Conclusion
 Presently we have 3 B-factories operating

 Present luminosities range from L= 0.81033 cm-2 s-1
  to L=21033 cm-2 s-1

 PEP II expects to reach L= 31033 cm-2 s- 1 by fall 2000
                          L= 61033 cm-2 s- 1 by 2001
                          L=101033 cm-2 s- 1 by 2002
                          L=151033 cm-2 s- 1 by 2003
  (similar plans for BELLE, CLEO  L= 31033 cm-2 s- 1 )

 PEP II will run with L=31034 cm-2 s- 1 by 2005

 BABAR &BELLE will have accumulated ~1700 fb-1 of
  data by end of 2009  ~1.5 109 BB events assuming
  90% at Y(4S) [lower if running at Y(5S)]
  (factor 5 higher than original design ~300 fb-1 )

 This is not enough for some rare decays

 Starting a new machine in 2008 with L>11035 cm-2 s- 1
  yields >7000-8000 fb-1 after 10 years
    >4.7-5.3109 BB events [for 2/3 at Y(4S)]

 To exploit the physics potentials at the Y(4S) one needs
  general purpose detector like BABAR, BELLE, CLEO

   For measurement of CP asymmetries & tagging
     good vertex measurement
   For charged track reconstruction
     good tracking device in high B field
   For Bp 0p0, B Xse+e-
     good electromagnetic calorimeter
   For K/p separation for tagging, B0p+p-/ B0K+p-
     good particle identification system
   For m & KL identification
     instrumented flux return

 Design depends symmetric/asymmetric machine

 Need to worry about backgrounds, all subdetectors,
  triggers, data acquisition & computing






 Beam-gas interactions
  Produce particles with large oscillation amplitudes
   that may hit beam-line elements near IR
  Bremsstrahlung BG is generated local to IR
    mainly in horizontal plane
  Coulomb scattering produces BG from whole ring
  Interactions with B1 magnet & Q2 septum produce
   shower debris  detector occupancy
                   radiation damage
  BG scales like Ib PV  maintain low pressure near IR

 Synchrotron radiation
  generated in bends and focussing quads near IR
   strikes vacuum chamber and detectors
    BG scales  Ib  reduce by masking

 Two-beam background
  enhanced beam-gas due to synchrotron shine into
   other ring
  beam tails due to beam-beam disturbance
  QED physics of radiative Bhabha scattering
   (presently no issue)

 Acceptable levels of backgrounds are determined by
   Radiation hardness of subdetectors
   Trigger rate
   Detector occupancies

 Occupancy and trigger rate determine acceptable
  dynamic running conditions

 Total integrated radiation dose determines lifetime of
 Dose is accumulated under normal running conditions,
  during injection, machine studies and beam-loss events
 At PEP II dose accumulated during running dominates

 Maximal currents in PEP II
    at design luminosity: 0.8 A (HER) 2.0 A (LER)
    at L=1.51034 cm-2 s-1: 1.3 A (HER) 3.8 A (LER)
              BABAR TRIGGER

 Hardware Level 1 trigger (rate < 2 kHz)
  Decision within 12 ms delivered to fast control system
  (common front end electronics pipeline buffer depth)
  Limitation coupled to DAQ readout speed (dead time)

 L1 consists of     drift chamber trigger (DLT)
                     calorimeter trigger (EMT)
                     IFR trigger (IFT)
  each trigger generates a primitive
   sends it to global trigger (GLT) at 8 MHz

 Current L1 system logic is based on r-f projection
  sufficient for design L= 31033 cm-2 s-1

 Software Level 3 trigger (rate < 100 Hz)
  L3 operates on 32 online farms with processing speed
   of ~10 ms/event (close to 2 kHz input)
  Prescale bhabhas since L1 triggered cross section
   of ~45 nb yields 135 Hz rate
               BABAR TRIGGER

 Present L1 trigger rate is ~ 700 Hz at L= 11033 cm-2 s-1
  with recent firmware improvements expect ~550 Hz

 Incorporate DCZ trigger to reduce beam-wall BG
   ease demand on dataflow
   reduce L3 input rate

 Level 3 track z0 for L1 passthrough in typical
  colliding beam run

                                 Colliding beam events

                                 Lost particles interacted
                                 with beampipe flange
            L1 TRIGGER RATES

 Consider 2 scenarios: linear scaling, quadratic scaling

 For high L modifications are needed (DCZ trigger)

 For luminosities above L= 31034 cm-2 s-1 need to
  to reduce BG further, prescale QED events and
  need a system that accepts increased L1/L3 rates
                   L3 TRIGGER

 At L= 11033 cm-2 s-1 total rate is 60 Hz, of which
   physics rate: 6 Hz,  prescale bhabhas: ~10 Hz,
   other QED: 4 Hz,  evading bhabha veto: 13.5 Hz,
   BG: 21 Hz ,          L1 passthrough: 5 Hz

 With improved BG (11Hz) and bhabha veto (6Hz)
  yield total rate of 45 Hz  90 Hz at L= 31033 cm-2 s-1
   (Physics:Bhabha:BG + 6:6:11nb)

 At L= 1.51034 cm-2 s-1 total rate is ~370Hz
   (upper limit)  physics rate is ~ 90 Hz
   improvements in DCH tracking needed  (>CPU)
  improve tracking pattern map for pt<250 MeV tracks
  tracks must have  5 segments in 6 layers
   exclude tracks exiting below layer 5
  employ more sophisticated track segment pattern
   that is less sensitive to dead cells
  improve L3 track resolution
  prescale other QED processes?

 Other improvements
  Reduce beam-wall BG
  Improve bhabha veto
  Improve d0 & z0 cuts

 At L= 11035 cm-2 s-1 need > accepted L3 trigger rate
  L3 physics rate alone will be 560 Hz

 Set of standard Readout Module (ROM) to interface
  front-end electronics of detector subsystem

 Each ROM has two 2 modes of operation
  untriggered version to collect input from
    ≤ 3 data streams (EMC)
  triggered version to collect input of triggered L1
    events from ≤ 2 data streams (all other + L1)

 Each ROM consists of VME board + 3 custom boards

 Collected data segments are read out via VME bus by
  master ROM in each crate

 Data fragments built by master ROM from individual
  segments are sent via 100 Mbps switched ethernet to
  nodes of Online Event Processing (OEP) farm

 DAQ is designed to have performance that is scaled to
  bandwidth limits of front-end electronics scaled by
  addition of more ROM’s and OEP nodes

 DAQ system is designed to have performance that can
  be scaled to higher trigger rates and higher occupancies
  to data bandwidth (BW) limits of front-end electronics

 Presently event building rate handles 2.5-2.6kHz L1
  trigger rates with low deadtime (`~1%)

 Overall performance will be limited by portion of DAQ
  system serving detector subsystem that first encounters
  BW restriction due to occupancy or L1 rate
  Potential Restrictions on DAQ System

 BW on data links from front-end electronics to ROMs
   increase # links or decrease # data read

 ROM performance due to internal BW limitations,
  processing limitations or both:
  BW on PCI bus limit data transport rate to memory
  BW ROM CPU &memory limit feature extraction rate
  CPU performance can also limit feature extraction rate
   increase # ROMs, input data links, faster CPUs

 Rate of master ROM in each crate to perform its
  backplane build of data segments from ROMs into
  data fragments
   increase # ROM crates

 network build of data fragments from master ROMs
  into events in OEP nodes limited by master ROM
  output rates or the network (BW of network link on
  DAQ crate side, network switch itself, BW of network
  link on DAQ crate side)
   increase # master ROMs, DAQ crates &
      network links
   use higher-capacity backplane of network switch,
      increase # network links, add OEP nodes

   (# DAQ crates is limited to 32 by Fast Control System)

 For L< 61033 cm-2 s-1 present BABAR DAQ system will
  provide adequate performance w/o significant upgrade
  (L1 ~ 2.8 kHz w/o trigger upgrades )

 For L= 11034 cm-2 s-1 with L1 trigger upgrade rates
  will be reduced to levels of L= 61033 cm-2 s-1

 For L= 1.51034 cm-2 s-1 beam currents are expected
  to be similar to those for L= 11034 cm-2 s-1 and L1
  trigger rate increases by ~ 400 Hz

 For BABAR behavior of DAQ system will be studied
  Potential bottlenecks will be parameterized as function
   of L1 trigger rate and occupancy in each detector system
  Eventual problem areas will be upgraded

 For L> 31034 cm-2 s-1 (new machine + detector)
  DAQ system needs to be designed to have
  sufficient capacity to handle data flow

 Computing aspects have to consider
  software, processing, storage & networking

 These aspects are present at all levels:
  data acquisition, pattern recognition, simulation &
  physics analysis

 Scalability was built into BABAR computing system
  uncertainties in scaling of BG rates & occupancies vs Ib

 BABAR considers two scenarios:
  low BG scenario: L1  Ib
  high BG scenario: L1  Ib2

 Normalization: total L1 = 700 Hz at L= 11033 cm-2 s-1
  (cosmic m 150 Hz, physics collisions 80 Hz, BG 470 Hz)

 Machine-induced occupancy contributes to event size,
  assume event size of 35 kB at L= 11033 cm-2 s-1
  (half of which stems from occupancy)

 Physics Model: do everything or focus on core topics?
 loose L3 - tight L3 scenarios
  (“interesting” physics cross section: 2 nb)

 Computing power is assumed to follow Moore’s law

 Computing power is assumed to follow Moore’s law
  Doubling every 2 years

 Constant processing costs per CPU

 Disk storage costs evolve with Moore’s law behavior

 Primary tertiary storage utilizes STK Eagle technology
  1999 40 GB costs ~ $75 with factor of two compression
  costs per tape decrease 20% a year
  effective tape capacities 80 GB by summer 2000 &
   160 GB by 2002

 Cost of network switch remain constant

 Luminosity increases have impact on Si vertex tracker
   radiation damage to silicon detectors
   radiation damage to on detector readout electronics
    increased occupancy in silicon strips

 Presently accumulated dose in BABAR SVT at
  L= 21033 cm-2 s-1


 SVT irradiation is focussed on horizontal plane of
  layers 1-3 (layers 4-5 are essentially immune)

 SVT is designed to withstand total dose of ~2 MRad
  without significant performance degradation

 Budget 200 kRad/yr, 1999 measurements yield
   8 Rad/pb-1 outside horizontal plane
   75 Rad/pb-1 in horizontal plane (best: 25 Rad/ pb-1 )
   For total 300 fb-1 integrated luminosity horizontal
    plane receives at best 7.5 MRad (on average 22MRad)
    exceed 2 MRad by middle of 2002 (worst case)
   (based on 60 Co source and beam dump experiment)

      Projected doses in the horizontal plane


                                                Linear scaling

                         June 2002 (1034)
 P-stop short creation
   reduction in charge collection efficiency
   extra current draw (depends on HV supply)
   use modules w thicker oxide layer

 Increase in leakage current
   Calculation with coefficient 2mA/cm2/MRad yields

   Even at 20 MRad modules would operate
   need increased threshold  efficiency drop
   MIP yields 24000 e- at normal incidence dropping to
   7000 e- under large angle

 Increase in interstrip capacitance
  after 1 MRad 10-15% increase, non linear effect w dose

 Depletion voltage shift
  high Rad doses can cause change in effective doping C
   change V of module  make material more p-type
   ok till Rad-induced acceptor doping compensates
     initial donor doping  doping inversion point
     when junction &ohmic sides switch roles
   SVT sensors are not designed to work this way
      life time limit 3-4 MRad
 Radiation Damage of On-Detector Electronics

 Atom chips have been irradiated up to 2 MRad
  show 10% increase in noise and 10% decrease in gain

 Extrapolation to higher doses is difficult as most
  damage appears to happen early on

 High irradiation can cause failures in chip’s digital
  circuitry  make entire chip unusable

 Higher BG & larger detector noise from irradiation
  yield increase in occupancy

 High occupancy has drawbacks
  pattern recognition may associate wrong hit
    deteriorates track quality
   in single-hit electronics a BG hit arriving at same time
    window will shadow a real hit  cause inefficiency
   pile up may occur if BW is not sufficient to transfer
   information out of busiest module

 Presently in BABAR occupancy in worst regions
  is 3-4%

 Increases in occupancy by > factor 5 would be
  Background Effects on Drift Chamber

 Drift chamber performance can be compromised by
  wire surface contamination due to aging
  makes parts of chamber inoperable
  ability of HV power supply to provide current
   necessary at higher luminosity
  increased occupancy in chamber could push DAQ rates
   & limits offline track reconstruction to find tracks
   buried in BG

 Collective experience with DCH in high radiation
  environments suggest aging effects start to occur when
  total Q drawn by wire reaches 0.1-1.0 Cb/cm

 Chemistry of this aging process is not well-understood
   & depends on gas mix & levels of impurities leaking in
  formation of hydrocarbon deposits on anode wire
   yielding regions of reduced gain or discharge points

 Precautions:
  Use radiation hard gases
  Operating drift chamber at lower HV eg 1900 V plus
   use of low-noise amplifier IC
  Adding water (3000-4000 ppm) to stop formation of
   hydrocarbon deposits
  Extra shielding of drift chamber endplates

 At PEP II perform dedicated beam studies to measure
  dependence of DCH current on LER & HER currents
  and separate L-dependent term (beam-beam effects)

 Dependence is best fit by a quadratic form
  Use this to extrapolate to higher currents (accuracy ?)

 With improved vacuum quadratic component will
  decrease, but L-dependent term is not yet understood

 For comparison also look at a linear fit

 Integrated charge on worst wire vs t for BABAR DCH
  assuming specific luminosity profile shown

 If chamber starts to show aging effects at 0.2 Cb/cm
  this level is reached in 2008 assuming quadratic fit
  Studies of Drift Chamber Performance
 Predicted IDCH vs Ib on typical HV supply for 2 fits

 Typical chamber has 448-1024 channels per layer
  determines # channels per HV power supply
   (50 -100 nA/wire)

 Projected occupancy vs beam currents for 2 fits from
  random beam crossings during single-beam running

                                     [%]            [%]

    Track finding ok to 10%

 Occupancy impacts L1 trigger rate & DAQ (ROMs)
    Electromagnetic Calorimeter Issues

 Beam related backgrounds have two effects on EMC
  Radiation damage to CsI(Tl) crystals
  Increased tower occupancy

 Observation that ILER and IHER are largely decoupled
   in their effect on radiation dose and tower occupancy
  IHER contributes mainly in forward direction
  ILER contributes mainly in backward direction
  central region of calorimeter is only weakly affected

 Since ILER = ~3  IHER crystal closest to beam in
  backward barrel (140.8o) sees highest radiation dose
  crystal closest to beam in forward endcap is at 15.8o

 Observe nearly quadratic dependence of tower hits vs
  HER beam current (E> 5 MeV)
 Occupancy increases only by ~hundreds of hits while
  ~1000 hits/events due to 0.3 MeV incoherent noise in
  preamplifier/crystal & 1 MeV threshold cut
 BABAR EMC is designed to survive 10 kRad of dose
  over 10 years with ~20% light loss in each crystal

 115 RADFETs measure integrated dose continuously

 Doses peak in horizontal plane

 Highest doses scale ~ linearly with integrated Ib
 Lower-energy Compton & bremsstrahlung g’s dominate

 Integrated doses scale linearly with beam currents as
  backward barrel:          40 amp hr/Rad from LER
  barrel/endcap interface: 20 amp hr/Rad from HER
  endcap (lowest 3 rings): 30 amp hr/Rad from HER

         Expected integrated dose vs time

 From light-yield measurements of worst & average Xtals
  extrapolate light loss vs time

               Fractional light loss vs dose

             Expected light loss vs time

 Use 6.1 MeV g peak from neutron-activated
  fluorocarbon to calibrate crystals (stable to 0.5%)

 Crystal in endcap have ~1% larger light loss than
  average  behavior is that of typical crystals
  (need to confirm with higher dose)

 If light loss were to degrade faster than expected
  replace crystals if light loss > 30% (option for EC)

       Light loss measured with 6.1 MeV g source

 Radiation damage alters linearity of crystal response
  along crystal length

 Induced non-linearity will decrease energy resolution

 If exact nature of change is understood & predictable
  some recovery is possible

 Perform study of mapping linearity of one “worst”
  crystal vs dose and input result into MC to look at
  impact on p 0 mass resolution

 Study of improving performance in endcap with finer
  segmentation or timing information

                Change in p 0 width vs dose

 Only issue is high occupancy in outer layer due to
  beam-related backgrounds

 Presently outer RPC layer has random occupancy of
  several %

 At design currents and at higher luminosity
  this will become an unacceptably high contribution
  to p/m misidentification

 Solution: build 5 cm thick Fe shield following
  outer-most chamber

 Radiation hardness of quartz bars & glue joints is ok
Transmission properties of synthetic fused quartz are
  reasonable up to 100’s of kRad dose
 Glue joints show 1% transmission loss for 70 kRad
 Photocathode lifetime of 100 Cb at 500kHz is
  several 1000 years

Light transmission in quartz bars before / after irradiation

 Main issue is counting rate due to beam-related BG
  At L= 11033 cm-2 s-1 physics rate is 1 Hz, while
  RBG (kHz) = 0.15 + 3.5 ILER + 14.0 ILER2 (A)+
                      4.2 IHER+ 20.4 IHER2 + 20 L(1034 )
   0.5 MHz per tube at L= 1.51034 cm-2 s-1

 PMT base tolerates rate up to > 1MHz

 At 500 kHz/tube and 1ms readout window for TDC get
  get ~ 5000 TDC channels/event (50% occupancy)
  readout by 6 ROM’s running a 15MHz fiber optic link

                                         high dead times

                                         DIRC inefficiency
   .01                                   vs counting rate



 K/p separation vs inefficiency
 normalization: 3.6 s at 3 GeV for 30 pe/track
  a single photon resolution of 9.6 mr
  per track resolution varies with  Nph of photons added
  add in quadrature with systematic term 2.3 mr/track

 Effect of background is worse than it appears
 K/p separation is significantly affected by inefficiencies
  especially at low momenta where 1 relevant hypothesis
  is in veto mode

   DIRC performance for N1 (b=1) and N2 in veto mode

 DIRC counting rates are dominated by BG from LER
  install shielding of stand-off-box

 Increase data transfer rate ( ~ factor 4)

 Reduce trigger window (~ factor 2)

   Further improvements -redesign FE electronics &DAQ
   Improve timing resolution of photon detectors by 10
   Increase pixelization of detector
   Improve photon resolution by new imaging method
   Increase geometrical acceptance in forward direction

                                         DIRC readout
                                         processing time
                                         in ROM vs #TDC

 Detector operating at luminosities > L=31034 cm-2 s- 1
  needs to be well-designed
 Explicit design depends on machine parameters
  (beam currents)

 Backgrounds need to be cut down as much as possible
 very low vacuum pressure near IR, excellent masking
  & shielding

 L1 & L3 triggers need to be designed to accept larger
  rates than BABAR

 DAQ needs to be able to handle L3 w/o deadtime plus
  sufficient margin

 Computing power must be able to handle online, offline
  analysis & MC

 SiVD needs radiation hard sensor+electronics,
  operated with low occupancy (pixel detectors ?)

 Drift chamber needs radiation hard gas and reasonable
  occupancy (problematic > 3 1034 cm-2 s-1)

 EM calorimeter should use “radiation hard” crystals
  operated with reduced occupancy

 Particle ID system needs to be radiation hard and
  operated with reasonable occupancy

 IFR with appropriate shielding is basically ok

Shared By: