Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/05 1 XFEL-PIXEL Readout and control Detector...
-
Upload
baldwin-miller -
Category
Documents
-
view
220 -
download
2
Transcript of Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/05 1 XFEL-PIXEL Readout and control Detector...
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/051
XFEL-PIXELReadout and control
Detector requirements/proposal
Basic concepts for experiments
Control electronics
Selecting/rejecting bunches
Backend-systems
Time schedule
Conclusion
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/052
Detector requirements
Call for experiments:
large pixel detectors: > 1Mega-Pixel
large number of frames: XFEL: 30000 bunches/sec
large dynamic range 104-106 Photons/pixel/frame
distinguish 0 and 1 Photon, rest Poisson-limit
With 2Bytes/pixel and 1MPixel 60GByte/sec
- but experiments will not handle frame rate
- Data reduction of "compression" is not effective:
Statements of nearly nothing to factor 2-4.
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/053
Three ProposalsAP-HPAD: analogue pipe-line Hybrid Array Pixel Detector
Si-pixel: ASIC stores signal while bunch train into capacitorsdigitization/data-transfer in pause between trains
400 frames/train
LSDD: Linear silicon drift detectorOne dimension is coded into drift-time: 200MS/s while traindata transfer in pause between trains.
600 frames/train
LPD: Large Pixel Detector analogue pipeline while train digitization between trains512 (256) frames/train
All extendable beyond 1Mpixel All around every 5-10 bunch All might run in parallel
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/054
Idea: Basic setup of all experiments
Specific to experiment, here AP_HPAD
Generic for all experiments
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/055
Idea: Concept for control electronicsGeneral task:- Boot- Collect information- Synchronize to XFEL clocks, time- Synchronize Interface electronics Backend Experimental area- Generate/distribute Clocks/data- User interface- Write to backend Monitoring/Status- Usage at other laboratories
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/056
Selecting/rejecting bunches
Aim: - Best use of limited pipeline in detector-head- Limit resources for backendStaged concept:- Store predefined bunches into pipeline of experiment- Fast reject (before next bunch): hardware detector,....- Slow reject: After bunch train
Handle slower information, information from XFEL- Transfer to backend-system
More fancy selection in CPU-farmRates: hard to guess, input from science needed- gas stream through X-ray-beam: - solid targets in beam (just first bunch)- that is not all. Who needs what?/how much?
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/057
Idea: Data stream to and concept of backend-system
Aim:- Collect complete frames into one CPU- Data in mass storage sorted for "Offline"-analysis
Allowance:
- Calculation on frames before sending to mass storage
- Decisions on frame to store/reject inside backend system.
- Rejecting frames on more fancy information from external signals.
- Write data frame-wise to mass storage - easier offline.
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/058
Interface: Experiments to backend-system
Rough:Data from Pixels written toRAM and read in other sequence
By that:One group of frames transfers to 1 Link to BE.
Still 8 or more of such modules transfer tosame backend-system
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/059
Need, controlled time sequence:So that each CPU gets data from all interface-modules without time conflict
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/0510
Idea: Concept of backend-system
Switch CPU-Farm
Inte
rfac
e el
ectr
onic
s
Sta
ge 1
: 1
00 L
inks
/1G
bit/
s
massstorage
controlelectronics
Feedback
monitoringdata
XFEL-accel.?
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/0511
Rates:
- Not well defined, what does which of experiment
- AP-HPAD: 2Bytes/pixel/frame is reasonable
- LSDD: Start with over sampling: Where to reduce to what, has to be a common diccussion- LPD: Rate expected to be similar to AP-HPAD, but I haven't seen their proposal.- First tests of compressions had been not effective Recent statement for single case was factor 2-4.Estimate to storage: 1Mpixel * 500bunches/train * 2Bytes/pixel * 10trains/s 10GBytes/secDiscussion with inputs from backend/science needed to settle rates/uptime/technical-effort/offline-effort
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/0512
Time schedule
2007: Experiments write proposal 15-April 2007
end 2007-2013: Designing, Constructing, but also
Laboratory tests and usage at other beam lines
2009(?) DAQ needed for tests - no large CPU-farm
LCLS has 120Hz and not 30000bunches/sec
2013: 1Mpixel dedectors at XFEL 3*(10-20GB)/sec
20xx: Upgrades beyond 1MPixel (e.g. 4Mpixel)
Peter Göttlicher, DESY-FEB, XFEL-DAQ-2007/03/0513
Conclusion
Concept is there, open to change and do an other wayTwo aspects: Control electronics
Backend-system
Ideas about numbers, but a lot of open questions
Discussions needed with input for - effort in backend- science requirements- offline possibilities- feasibility of interface-electronics
Common effort for all experiments is a strong wish