Detector DAQ Overview and Needs

27
Daresbury Aug 2005 Jean-Sébastien Graulich Detector DAQ Overview and Needs Jean-Sebastien Graulich, Univ. Genève o Introduction o Detector Systems Overview o Implementation Proposals FEE, Trigger and Architecture o Conclusions MICE DAQ and Controls Workshop

description

MICE DAQ and Controls Workshop. Detector DAQ Overview and Needs. Jean-Sebastien Graulich, Univ. Genève. Introduction Detector Systems Overview Implementation Proposals FEE, Trigger and Architecture Conclusions. Introduction. Normal Detector DAQ is synchronised with MICE Beam - PowerPoint PPT Presentation

Transcript of Detector DAQ Overview and Needs

Page 1: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich

Detector DAQ Overview and Needs

Jean-Sebastien Graulich, Univ. Genève

o Introduction

o Detector Systems Overview

o Implementation Proposals

FEE, Trigger and Architecture

o Conclusions

MICE DAQ and Controls Workshop

Page 2: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 2

IntroductionIntroduction

Normal Detector DAQ is Normal Detector DAQ is synchronised with MICE Beamsynchronised with MICE Beam

We want RF ON and RF Off DataWe want RF ON and RF Off Data (50/50 ?)

We need calibration dataWe need calibration data For each run

We want RF Noise dataWe want RF Noise data Dedicated Run

I SI S Beam MI CE Target

MI CE RF

Beam ON RF OFF

Beam ON RF ON

RF Noise Measurement

Other I SI S Users

Should not happen

Detector Calibration

Page 3: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 3

DDAQ vs CMDDAQ vs CM

Detector DAQDetector DAQSynchronised with the beam

Very fast reaction time (~s)

High transfer rate (~50 MB/s)

Read and Store, no time for on-line processing

Limited User InterfaceRun Control only

(Slow) Control and (Slow) Control and MonitoringMonitoring

Continuous and permanent

Very reliable (Safety issue)

Deal with a lot of ≠ hardware

Read and CheckCalibration, manage alarms at ≠ levels, soft interlocks, take actions, log history, etc.

Extended UISet many parameters, manage complicate initialisation procedures, etc.

Why separate Particle Detector DAQ and Why separate Particle Detector DAQ and Control Sensor DAQ ?Control Sensor DAQ ?

Page 4: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 4

Concept SchematicConcept Schematic

MICE Beam Line

MICE CoolingChannel

MICE

Detectors

Detector

DAQ

Slow Control

Data

Storage

Monitoring

Data Log

Run Control

UI

MICE

User Interface

Environment

RF Phase

Page 5: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 5

Definitions and Definitions and assumptionsassumptions

DefinitionsDefinitions Machine Cycle:

50 Hz (20 ms) Injection/Extraction cycle of ISIS MICE Target Cycle -> Spill Cycle

Beam ON / RF ON Beam ON / RF OFF

MICE RF Cycle RF duty factor: 10-3

~100 s ramping up/down time Most efficient cycle: ~1ms long RF pulse every second

Assumptions (from TRD)Assumptions (from TRD) The aim is to collect 600 good muons per spill

Instantaneous rate = 0.6 MHz (1.7 s between 2 muons) A good muon is a traversing muon (~ 1/6 in phase with

RF) 106 muons per setting is required (1700 spills, 30 min)

-> 2 spills/second (~1ms long)

Page 6: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 6

Readout CycleReadout Cycle

Particle-by-particle readout is not Particle-by-particle readout is not possible: possible:

Readout takes a few 100 s (depends on the data size)

Readout time ≠ Conversion time Readout time driven by the slowest VME crate

=> Detector Data has to be buffered in the => Detector Data has to be buffered in the FEE modulesFEE modules

Buffer size ~ 600 muon events

=> Readout performed at the end of the => Readout performed at the end of the spillspill

Normal DAQ Event triggered by Start of Spill Normal DAQ Event ≠ Particle Event

Page 7: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 7

The ADC problemThe ADC problem Even simple gated ADC is a problemEven simple gated ADC is a problem

Conversion time > ~3 s = 2 x the average time between two muons

Gated ADC => maximum 300 muons/spillGated ADC => maximum 300 muons/spill If buffer size high enough (common is 30 evts!)

Alternative: Flash ADCsAlternative: Flash ADCs The full waveform is digitised for the whole spill Sampling rate > 200 MHz (1 sample/ 5ns)

200 106 sample/channel/spill If 8 bits data -> 200 MB/ch

(or nearly 400 GB/ch/run) => Data size problem

No time to transfer No space to store

=> Need zero suppression Nobody works on that !!!Nobody works on that !!!

Page 8: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 8

Detector System Detector System OverviewOverview

5 detector systems5 detector systems 2 Tracker modules (5 Sci.Fi stations) 3 TOF stations CKOV1 Upstream CKOV2 Downstream EmCal

Page 9: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 9

Sci.Fi TrackerSci.Fi Tracker

Main requirements:Main requirements: Efficiency, Spatial Resolution, low sensitivity to RF bg

Number of channels: Number of channels: 4096 ADC per tracker = 8192 ch 8192 TDC channels under development (AFE-T)

Front End Electronic:Front End Electronic: Analog to Digital: AFE II VME (digital data buffer): VLSB (512 ch/module)

Word size: Word size: 10 bits for ADC, Probably ~12 bits for TDC 13 bits for the channel number (+ data overhead)

Average Data SizeAverage Data Size Without zero suppression (no TDC): 24 kB/μ With zero suppression and TDC: 0.25 kB/μ

Page 10: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 10

Sci.Fi Tracker Sci.Fi Tracker ConstraintsConstraints

Important limitations:Important limitations: Read out has to be synchronised Read out has to be synchronised

with the beam microstructurewith the beam microstructure Built in the AFE board, designed for D0 100 ns muon burst every 320 ns

(Compared to D0: 150 ns burst every 400 ns) Looks OK

AFE Conversion time: ~6.5 AFE Conversion time: ~6.5 microsecondsmicroseconds

600 muons/spill is NOT possibleNew assumption: Maximum 150 muons per spill

(2h/run)

Page 11: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 11

TOFTOF

Main requirements:Main requirements: Time resolution ~70 ps, high rate (2.5 MHz in TOF0)

Number of channels: Number of channels: (48 + 32 + 32) TDC = 112 ch 112 ToT (Time over threshold)

for time walk correction

Front End Electronic:Front End Electronic: CAEN 1290

Word size: Word size: 16 bits/ch

Average Data SizeAverage Data Size 0.1 kbyte/muon

TimeWalk

Page 12: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 12

TOF FEETOF FEE

Proposed FEE module:Proposed FEE module:

Multi-Events, Mutli-HitsMulti-Events, Mutli-Hits Not tested yetNot tested yet

Page 13: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 13

Event IntegrityEvent Integrity

How to retrieve particle event integrity ?How to retrieve particle event integrity ? Easy at the DAQ Event Level What about the Particle Event Level ?

Page 14: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 14

SoSEoS

DAQ

Gate

NormalDAQ Trigger

GoodMuons

(TOF0xTOF1xTOF2)

Muon Triggerto AFE II

Muon Triggerto TDC

AFE GateClock

Programmable delay Programmable delay

Progr. Width TDC EventWindow

DAQ and Muon DAQ and Muon TriggersTriggers

Page 15: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 15

Data StructureData Structure

RUN # m• Number of DAQ Events == Number of spills• Environment data

DAQ Event j-1

DAQ Event j

DAQ Event j+1

TrackerTracker

# of Events

Event k-1

Event k

Event k+1

Number of Hits

Ch_0001 ; Data_0001

Ch_8192 ; Data_8192

TOFTOF

# of Events

Event k-1

Event k

Event k+1

Number of Hits

Hit1_Ch ; Hit1_Data

Hitn_Ch ; Hitn_Data

Page 16: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 16

CKOV1CKOV1

Main requirementMain requirement Energy resolution:

Threshold between pion and muons Number of channels: Number of channels:

4 TDC ch 4 QDC ch

Front End Electronic:Front End Electronic: TDC : CAEN 767 ? QDC : ??

Word size: Word size: TDC : 20 bits/ch ADC : probably 10 bits

Average Data SizeAverage Data Size ~ 25 bytes/muon (15 kB/spill)

Page 17: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 17

CKOV FEECKOV FEE Possible FEE module:Possible FEE module:

Also Multi-Events, multi-Hits Also Multi-Events, multi-Hits -> Same Event Integrity check

Not tested yetNot tested yet

Page 18: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 18

CKOV2CKOV2

Main requirementMain requirement Low energy threshold

Number of channels: Number of channels: 12 TDC ch 12 QDC ch

Front End Electronic:Front End Electronic: TDC : CAEN 767 ? QDC : ??

Word size: Word size: TDC : 20 bits/ch ADC : probably 10 bits

Average Data SizeAverage Data Size ~ 75 bytes/muon (60 kB/spill)

Page 19: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 19

Muon Identifier Muon Identifier (EmCal)(EmCal)

Main requirementMain requirement Energy resolution

Number of channels: Number of channels: 240 QDC ch ~60 TDC ch

Front End Electronic:Front End Electronic: Not chosen (see ADC Problem)

Word size: Word size: ???

Average Data SizeAverage Data Size I guess ~ 1kB/muon

Page 20: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 20

Data Volume Data Volume SummarySummary

All detectors:All detectors: 25 kB/μ (1 kB/μ if zero suppression in the tracker) 7 MB/spill (if 2 x 150 μ/spill) 25 GB/run (if 106 μ/run)

How many runs ?How many runs ? 2 hours/run 50% overhead (setting changes) 80% efficiency 1 year, 7/7, 24/24 (just to get an idea) => ~ 2500 runs

Need storage space for ~ 60 TB/yearNeed storage space for ~ 60 TB/year

Page 21: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 21

Event typesEvent types

Normal DAQ Event ~ SpillNormal DAQ Event ~ Spill Calibration Events (Can be Single Gated !)Calibration Events (Can be Single Gated !)

Pulser events Full readout, full event building Pulses generated by the DAQ

Cosmic/source events ? Partial readout, partial event building

Pedestal events ? Only ADC readout, full event building

Special muon event with dedicated beam ?? Special DAQ EventsSpecial DAQ Events

Start of Spill (SoS) and End of Spill (EoS) can be used to check the synchronisation between subsystems.

No readout, full event building

Each different Event Type requires a Each different Event Type requires a dedicated trigger receiver IO channel in each dedicated trigger receiver IO channel in each crate !crate !

Page 22: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 22

Trigger ReceiverTrigger Receiver

Need at leastNeed at least 6 inputs

SoS and EoS Events Normal DAQ Events Calibration Events: Pedestal, pulser and

source/cosmics 7 outputs

6 Individual busies One common busy

More I/O = more flexibility One module per crate !One module per crate ! Possible Choice:Possible Choice:

CAEN V977

Page 23: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 23

Page 24: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 24

““VME” ProcessorsVME” Processors

Needed to collect data locally, at the Needed to collect data locally, at the crate level.crate level.

Could be VME processor or External Could be VME processor or External PCPC

External PC under Linux connected with a VME to PCI optical link is easy, flexible and widespread

Easier if all VME have the same Easier if all VME have the same interfaceinterface

Possible choice:Possible choice: CAEN V2718

Page 25: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 25

Page 26: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 26

DAQ ArchitectureDAQ Architecture

Trigger distribution

Optical link

Online Monitoring

Tracker EmCal TOF Trigger + Ckovs

Ethernet

Linux PCs

GigaBit Switch

Run Control Event Builder

Online Storage

VME Crates

Remote Storage

Page 27: Detector DAQ Overview and Needs

Daresbury Aug 2005

Jean-Sébastien Graulich Slide 27

ConclusionsConclusions

Detector data Readout must be Detector data Readout must be performed at the end of the spillperformed at the end of the spill

Data has to be buffered in FEE

Maximum 150 muons/spill (2h/run)Maximum 150 muons/spill (2h/run) Due to Conversion time in AFE II of Sci.Fi Tracker

Charge Measurement in the Muon Charge Measurement in the Muon Identifier (EmCal) is an issueIdentifier (EmCal) is an issue

A Detector DAQ conceptual scheme A Detector DAQ conceptual scheme already existsalready exists

Based on HARP experience

It’s Time to write Specifications !It’s Time to write Specifications !