Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://zebu.uoregon.edu/~uochep/talks/talks05/daq.pdf Äàòà èçìåíåíèÿ: Tue Mar 29 23:21:59 2005 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 09:34:14 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: tail

A Possible Approach to the SiD Data Acquisition
David Strom ­ University of Oregon

· Bunch structure at the ILC

· SiD Detector

· Exp ected detector o ccupancies

· Possible common approach to on-detector DAQ

· Implication for data rates and volumes

LCWS 05

1

19 March 05 ­ David Strom ­ UO


Bunch Structure at the ILC
~3000 to ~6000 bunches/train (~150 to ~332 ns between bunches) 5 Hz Repetition

1 ms 199 ms

1 ms

· Final bunch structure of cold machine not yet known · Bunches unlikely to b e closer than 150 ns (kickers) · Total length of bunch train unlikely to b e more than 1 ms (damping ring size)
LCWS 05

2

19 March 05 ­ David Strom ­ UO


SiD Detector · Cost optimization small tracker and calorimeter radius ( 1.2m) · Tracker with 5 layers (50µm pitch and 10cm long detectors) · ECAL segmentation 0.2 â 0.2 (0.4cm/125cm) â30 · HCAL segmenation 0.3 â 0.3 (1.0cm/200cm) â36

LCWS 05

3

19 March 05 ­ David Strom ­ UO


Except in forward region o ccupancies will b e very low

Toshi Ab e's (LCWS04-Victoria) warm o ccupancies (â

20 for cold)

A system with a limited numb er of buffers in the Front-End Electronics seems viable for most subsystems
LCWS 05

4

19 March 05 ­ David Strom ­ UO


· Small o ccupancy and 1/200 duty cycle allows for relatively low p ower front end systems. · Excessive heat would require invasive co oling systems which would compromise several detector goals:

Effective calorimeter granularity would decline due larger gap size

Additional material causes conversions and multiple­ scattering in the tracker

LCWS 05

5

19 March 05 ­ David Strom ­ UO


Can we get the heat out of the ECAL?

Back of the envelop e calculation of change in temp erature: · Thermal Conductivity of W alloy 120W/(K-m) · Thermal Conductivity of Cu 400W/(K-m) Need to reduce heat to b elow 100mW/wafer.

temperature deg. C

20 18 2.5 mm of W 100mW 16 1.0 mm of Cu 100mW 14 1.0 mm of Cu + 2.5 mm of W at 100mW 12 2.5 mm of W 40mW 10 8 6 4 2 0

0

20

40

60

80

100

120 140 Length (cm)

LCWS 05

6

19 March 05 ­ David Strom ­ UO


Main physics impact of buffer overflow: · If a missing energy event o ccurs when buffers are full it will b e imp os¯ sible to exclude background such as f f
Ru n : e v e n t 4 0 5 5 : 1 7 3 5 5 Da t e 9 3 0 5 1 9 T i me 4 3 9 3 3 C t r k ( N= 1 7 Sump = 3 4 . 2 ) E c a l ( N= 2 0 SumE= 6 5 . 8 ) Hc a l ( N= 1 1 SumE= Eb e am 4 5 . 6 5 9 E v i s 9 6 . 8 Em i s s - 5 . 5 V t x ( - 0 . 0 3 , 0 . 0 9 , - 0 . 5 6 ) Mu o n ( N= 2 ) Se c V t x ( N= 1 ) F d e t ( N= 0 SumE= B z = 4 . 3 5 0 T h r u s t = 0 . 9 5 4 4 Ap l a n = 0 . 0 0 1 8 Ob l a t = 0 . 1 3 2 6 Sp h e r = 0 . 0 2 0 4 7.7) 0.0)

Ru n : e v e n t x x x x x: a l t e r e d Da t e 9 3 0 5 1 9 T i me 4 3 9 3 3 C t r k ( N= 1 7 Sump = 3 4 . 2 ) E c a l ( N= 2 0 SumE= 6 5 . 8 ) Hc a l ( N= 1 1 SumE= Eb e am 4 5 . 6 5 9 E v i s 9 6 . 8 Em i s s - 5 . 5 V t x ( - 0 . 0 3 , 0 . 0 9 , - 0 . 5 6 ) Mu o n ( N= 2 ) Se c V t x ( N= 1 ) F d e t ( N= 0 SumE= B z = 4 . 3 5 0 T h r u s t = 0 . 9 5 4 4 Ap l a n = 0 . 0 0 1 8 Ob l a t = 0 . 1 3 2 6 Sp h e r = 0 . 0 2 0 4

7.7) 0.0)

Y

Y

Z

X
2 0 0 . cm . 5 10 20 5 0 GeV

Z

X
2 0 0 . cm . 5 10 20 5 0 GeV

Ce n t r e o f s c r e e n i s (

0 . 0000 ,

0 . 0000 ,

0 . 0000 )

Ce n t r e o f s c r e e n i s (

0 . 0000 ,

0 . 0000 ,

0 . 0000 )

Background

New Physics?

Suggested criterion for buffer size: In any region in require that 1% of buffers are full in at most 1% or all events.
LCWS 05

7

19 March 05 ­ David Strom ­ UO


Main implication for DAQ system: · Except in the forward region data should time tagged and lo cally stored during the bunch train. · Each bunch train should b e analyzed as whole. This allows buffer overflows to b e flagged. · DAQ should b e able to continuously readout detector with completely full buffers ­ e.g. temp orary bad background, but this will in general not b e necessary.

LCWS 05

8

19 March 05 ­ David Strom ­ UO


Possible Common Architecture for Tracker, ECAL, HCAL
Bunch Number Digital Memory 4 x 13 bits

Cf Detector

Shaper 1 T

+ -

Shaper 2

Buffers To Wilkinson ADC

Simplified circuit of single channel (1024 channels/chip) (See M. Breidenbach's talk in Calorimetry) · Digitize buffers during time b etween bunch trains · Adjust numb er of buffers for detector o ccupancy · Adjust charge amplifier for detector typ e

LCWS 05

9

19 March 05 ­ David Strom ­ UO


Data Concentrator ­ ECAL barrel mo dule
3.6 Meters 1.1-1.3 Meters

Data Concentrator

· Each plane of the ECAL will have 250 wafers, each with 1024/channels. · Raw rate out of the concentrator is 150 Mbits/s

LCWS 05

10

19 March 05 ­ David Strom ­ UO


Estimated maximum data volumes Detector Channels (M) 3300 132 20 85 40 Amp bits TS bits Buffers Raw Size Gbits 9.9 6.9 6.8 8.8 3.3 36 Raw Rate Gbits/s 50 34 34 42 16 176

VXD Tracker ECAL HCAL Total

3 8 13 8 13 13 13 13

1 4 16 4 4

· The total raw event size for a single bunch crossing is 5 GBytes, so it is p ossible that this could b e analyzed in a single CPU at once · In the normal case, there would b e a zero suppression factor of 20 or more, giving typical bunch train "event" sizes of 250 MBytes · Have ignored contribution from hadron tail catcher and muon system


educated (and biased) guesses



10µm â 10µm pixels 11
19 March 05 ­ David Strom ­ UO

LCWS 05


What ab out forward region of the detector? · In the forward region (b elow ab out 100 milliradians) we quickly encounter high o ccupancies and it will probably b e necessary to read everything out. · If we keep the same segmentation as in the reset of the detector (60 longitudinal layers and 4mm â 4mm pixels) and readout every b eam crossing separately: Detector Channels (M) 2 Amp bits Buffers Raw Size Gbits 77 Raw rate Gbits/s 383

Forward

13

2820

Challenge for Data Acquisition system ­ how to handle such a large data rate
LCWS 05

12

19 March 05 ­ David Strom ­ UO


Observations: · Main physics need is to veto high energy (> 100 GeV) electrons, photons and muons · Next most imp ortant physics need is the luminosity sp ectrum · Solid angle b elow 100 mrad is probably not key for hadronic jet reconstruction · Background from tails of low energy electromagnet showers will complicate reconstruction of low energy hadrons MIP sensitivity and excellent energy resolution probably not needed in this region Could use hardware pro cessing to find quantities of interest to physics Detector DAQ/Design work very much needed in forward area
LCWS 05

13

19 March 05 ­ David Strom ­ UO


Conclusion

· "Triggerless" DAQ for a granular LC detector such as SiD lo oks plausible

· Occupancy in almost all of the detector is sufficiently low that for each channel only a few buffers/bunch train will b e needed

· Front-end electronics could b e based on a common chip design

· Main challenge will b e far-forward region may require dedicated hardware reconstruction

LCWS 05

14

19 March 05 ­ David Strom ­ UO