A Possible ?13 Electronics Architecture

1
A Possible ?13Electronics Architecture

• A Strawman Proposal
• Kelby Anderson for Jim Pilcher
• 30-Apr-2004

2
The Objective

• We should discuss and examine architecture
• Do this before any detailed design
• We need to get the high level features correct
• More people can contribute ideas to high level
  planning
• Once we agree on scope the detailed planning and
  design can begin
• Lots of work for anyone interested
• This talk is intended to provide a strawman plan
  for the architecture
• Other options can be compared to it
• Readout should take advantage of modern
  electronics developments
• Can do more with given budget now compared to 10
  years ago
• This was when CHOOZ, SNO, KamLAND were designed

3
Electronics Requirements

• Digitize charge seen by each PMT
• Energy reconstruction
• Provide timing of signal from each PMT
• one component of position info (also energy
  sharing of PMTs)
• Provide trigger for DAQ
• Physics triggers
• Neutrinos (prompt EM energy, delayed neutron
  energy)
• Backgrounds (to study and subtract)
• Muons
• Electronic calibration triggers (variable test
  pulses)
• Source/laser/LED calibration triggers
• Random triggers

4
Electronics Requirements

• Provide HV to PMTs
• Provide LV for electronics
• Provide ability to control and monitor detector
  and electronics
• Temperatures
• LV, HV
• PLD firmware

5
Electronics Requirements

• Readout should not degrade intrinsic resolution
• Energy resolution
• eg. 7.5 / SqrtE(MeV) ? 2 ? 5.7 at 2 MeV
• This is KamLAND resolution. Perhaps we can do a
  bit better.
• Timing
• 1.0 ns (PMT jitter)
• ? for Hammamatsu 5912, 8 PMT
• Energy readout must cover full dynamic range
• Low end
• Single photoelectron for individual PMTs
• Assume 15 counts/pe
• High end
• Muon along diameter of detector
• Emax of 10 ? lt?Emipgt
• ? pe for closest PMTs (to be determined)

6
Attractive and Feasible Features

• Trigger separately on positron energy and neutron
  energy
• Link by recording time since last trigger
• Gives better handle on background effects
• Trigger should be able to impose loose time
  coincidence in case singles rate is too large
• In this case prescale single energy triggers
• Sample signals in time window around trigger
  event ( ? 2.5 ?s )
• Earlier times provide input on possible
  backgrounds
• Later times provide link to neutron triggers
  within the same event (cross check of event
  timing)

7
Attractive and Feasible Features

• Do zero suppression and data filtering
  off-detector
• Take advantage of modern high-speed data links
• More flexibility in PC than in front-end hardware
• Provide independent electronic calibration for
  each channel over its full dynamic range (chg.
  injection)
• Allows injection of simulated events
• Allows easy tests for cross talk
• Allows precise electronic calibration of each
  channel
• In counts/pC
• Sources give pe/MeV and pC/pe
• Allows measurement of pulse shape by varying
  timing of injected signal for successive events

8
Front-end

• Convert PMT signals to standard analog shape
• Amplitude reflects charge from PMT
• Use low-noise passive shaper
• Fully linear
• Adapts PMT to speed of commercial sampling ADCs
• eg. 12-bit, 40 megasamples/sec (every 25 ns)
• TIs ADS5130 (12-bits, 50 MSPS) or ADs AD9042
  (12-bits, 40 MSPS)

9
Front-end

• Sample signal every 25 ns
• Synchronize all ADCs using optical timing system
• Laser driven optical fiber modulated by clock
• Control and trigger can be distributed with same
  system
• LHC Timing Trigger and Control (TTC) system
• Time resolution of clock at PMT 100 ps
• Each clock pulse is numbered
• System is off the shelf with many utility modules
  and custom chips
• Provides ability to set the clock timing
  independently at each PMT
• Compensate for channel-to-channel delays
• Synchronize all PMTs using LED at center of
  detector
• Transit time variations between PMTs from
  variation in HVs
• Save 200 samples in a pipeline ( ? 2.5 ?s )
• Readout if trigger
• Overwrite if no trigger
• Use larger dual-port memory for dead-timeless
  operation

10
Front-end

• Digital pipeline systems built for LHC
  experiments
• Fewer samples but similar delay to trigger
• To obtain needed dynamic range use multiple gains
• High gain gives 15 counts/pe and 4095 counts
  (270 pe) full scale
• Low gain gives 128 counts (0.8) at high gain
  maximum and 4095 counts (8,700 pe or 14,000 pC)
  full scale
• PMT has 5 non-linearity at 1,200 pC
• Readout will cover the useful range of PMT output
• Dynamic range of readout 17 bits or 102 dB
• These figures reduced a little by pedestal
  offsets
• Useful full scale reduced by 30 counts
• Noise should be held to 1 count to preserve
  dynamic range (waste of ADC resolution)

11
Front-end

• Standard pulse shape is fitted to extract
  amplitude and time offset with respect to
  sampling clock
• Demonstrated time resolution of electronics is
  lt 100ps
• Feature extraction done off-detector
• Can be done in PC
• Digital signal processing modules used at LHC
  because of rate (100 KHz, Level 1 trigger rate)

12
Data Collection

• Attractive to use single optical or electrical
  data link to control room
• to facilitate connection/disconnection
• 500 Mbps is quite feasible with off-the-shelf
  components (doing this ourselves in ATLAS)
• Data per PMT channel per event
• 2x12 bits x 200 samples 15 overhead (CRC,
  parity, identifiers) 5.5 Kbits
• Data per detector per trigger
• 819 PMTs x 5.5 Kb 4.5 Mbits
• One data link could handle 100 ev/sec
• Only needed for calibration
• Could reduce width of time window when running
  calibrations
• Derandomizing buffer needed at link input for
  normal data

13
Detector Control System

• Attractive to have single control bus from
  counting room to each detector
• Facilitates connection/disconnection for moving
  detector
• Functions
• Set HV on PMTs
• Control electronics calibration system
• Monitor LV, HV, temperatures
• Allows bi-directional communication
• TTC system is uni-directional
• Requirements
• Must have high fanout capability ( 800 PMT
  channels served by one bus)

14
Detector Control System

• Many commercial field bus systems
• eg. CANbus
• But fan-out capability could be a problem with
  this one

15
HV System

• Single HV cable per detector
• Set HV to individual PMTs on detector
• HVs adjusted so all tubes have same gain
• Using laser ball or LED at center of detector
• Neednt have fine control of HV over full range
• Just over range of PMT gain variation
• For constant term in energy resolution of 2 and
  820 PMTs we need individual PMT gains settable to
  2/?820 0.07
• For a PMT where ? is number of
  stages
• For 10 stage tube
• Need ?V to 0.10 V out of 1500 V
• Must be able to switch off individual PMTs
• Non-trivial requirement

16
Trigger System

• Base trigger on number of photoelectrons seen in
  detector
• Useful to have sums from segmented regions of
  detector
• Protect against localized hot spots
• Tile the detector into hexagonal patches
• Generate pe sum from each patch (26 patches of 32
  PMTs 832)
• How to do this?
• Traditional option is to add analog trigger
  signals, but
• Beware of analog offsets and common mode noise
• Trigger signals have separate timing and gain
  from DAQ branch
• Must have own calibration
• Better to use digital info from ADCs
• Trigger signals then time-aligned with clock
  system
• Need to use only coarse info from ADC (high order
  bits)
• Add 32 digital signals for a patch to get local
  sum
• Trigger signal assembled every 25 ns. (with
  latency)
• May want running sum over several clock periods

17
Cost Estimate

• Cost per channel
• Analog front-end 100 (ATLAS 88)
• Digitizer and pipeline 130 (ATLAS
  94)
• Clock timing and distribution 50
• Trigger 80
• HV power and distribution 60 (ATLAS 46)
• Control system 40
• Miscel. on-detector plumbing 40 (ATLAS
  35)
• LV power 20
• ______________
• 520
• 3 detectors _at_ 812 channels each ? 1.27M
• Add 30 for EDIA, 40 contingency ? 2.31M
• Overall constn. cost of experiment 60M
  (readout is 3.8)

18
Conclusions

• These comments offer a possible architecture
• Point of comparison for other options
• Cost estimates rough but includes experience with
  LHC design (top down estimate OK at this point)
• Much work needed for detailed planning, design,
  and implementation
• Could add readout to Monte Carlo
• Evaluate granularity and trigger
• Important to agree on architecture before
  starting design
• Divide detailed planning, design, and
  implementation among interested groups

19
Conclusions

• There will be PLENTY to do