ALICE EMCal Electronics Outline PHOS Electronics review Design

ALICE EMCal Electronics Outline: • PHOS Electronics review • Design Specifications – Why PHOS readout is suitable – Necessary differences from PHOS • Shaping time / data volume problem • EMCal vs PHOS comparison summary 1

PHOS Electronics, Schematic One Channel Crystal 2 APD+Pre. Amp 8 Transition-card 4 32 Channels FEE-card w/ ALTRO

PHOS Module Assembly FEE Card 32 Channels 35 cm x 21 cm 5. 5 Watts (170 m. W/ch) 870 SF (27 SF/ch) 3

PHOS Electronics, Schematic Crystal APD+Pre. Amp 8 Transition Card 4 32 Channels FEE-card w/ ALTRO 14 448 Channels 8 OR Level 0 Level 1 TRU = Trigger Router Unit 2 896 Channels RCU = Read-out Control Unit 4 In total 5 PHOS Modules 4 RCU = 1 PHOS Module = 3584 Crystals

5

6

7

Tower/module structure: “shashlik” design Trapezoidal module: transverse size varies in depth from 63 x 63 to 63 x 67 mm 2 Going to Shashlik design allows to use thinner sampling layers to 78 layers of 1. 6 mm scint/1. 6 mm Pb improve intrinsic energy resolution. Moliere radius ~ 2 cm Pb absorber has dimensions of module Towers defined by smaller optically isolated scintillator tiles Total Pb depth = 124 mm = 22. 1 X 0 8 Comparisons: PHOS = 180 mm/8. 9 mm = 20. 2 X 0 ATLAS Liq. Ar/Pb = 25 X 0 CMS Pb. WO = 25 X 0

Use PHOS APD + Charge Sensitive Pre. Amplifier • Must operate in Magnetic Field. • Need gain (and gain adjustment for trigger) • Light yield from EMCal similar to PHOS 9

Full Scale energy… inclusive jets 10 Hz @ 50 Ge. V few x 104/year for ET>150 Ge. V EFS = 250 Ge. V (PHOS 80 Ge. V) 10 From Peter Jacobs

Light yield Light Yield (in photoelectrons) measured at WSU with Cosmic rays in prototype tower using well-calibrated PMT. For APD, with Gain M=1 expect ~2. 5 photoelectrons/Me. V Compare PHOS: 4. 4 pe/Me. V @ M=1. For same fullscale signal amplitude MEmcal = 50(MPHOS)*(4. 4*80 Ge. V)/(2. 5*250 Ge. V)=28 11

Intrinsic Energy Resolution GEANT Simulation results: • Sampling fraction 8. 1% • Intrinsic energy resolution ~12% 12 Calculations by Aleksei Pavlinov

The PHOS APD + CSP Electronic Noise from PHOS Electronics Document • PHOS measurement 625 e @ 2 ms shaping : 625/(4. 4*50)=2. 8 Me. V • If EMCal uses 100 ns shaping, expect ~1500 e : 1500/(2. 5*50)=12 Me. V (36 Me. V 3 x 3) 13

Energy Resolution: All contributions 12% intrinsic 1% calibration Digitization (full scale=250 Ge. V) PA/shaper e. NC=2000 (60 Me. V) Dual 10 -bit ADCs (high and low gain) Even with pessimistic assumptions (e. NC=2000) electronics contributions to resolution are unimportant in energy region of primary interest. Important open question: slow neutrons 14 drives choice to investigate short shaping time ~100 ns.

EMCal Resolution: The ALICE “Environment” EMCAL only 15 All ALICE material GEANT Simulations for single photons (i. e. p+p) Significant degradation of resolution A. Pavlinov

The ALICE “Environment” Central HIJING Simulations: Production point of particles with EDeposit Before 30 ns After 30 ns Large background from moderately slow neutrons. 16 Calculations by Heather Gray

Soft, Slow (neutron) Background Total EMCal EDeposit vs Time Tower neutron EDeposit Mean neutron EDeposit =36 Me. V (i. e. 3 times electronic noise!) with rms=41 Me. V Note: This is for Central HIJING (worse case, the problem is centrality dependent). 17 Calculations by Heather Gray

Bandwidth: Another shaping time argument • Propose to use tpeak = 100 ns with 20 MHz sampling • Ex: PHOS Bandwidth – Number of samples = 5*tpeak/Dtsample = 5*4 ms/100 ns = 200 – Average hit rate (>30 Me. V) = 200 Hz – GTL bus rate = (14 FEE)(32 chan)(2 Gain)(10 bit)(200 samples)(200 Hz)=44. 8 MB/s – RCU data rate = 2*GTL/RCU partition=89 MB/s (limit 100 MB/s) • EMCal Bandwidth – Number of samples = 5*tpeak/Dtsample = 5*200 ns/50 ns = 20 – Average hit rate (>30 Me. V) = 2000 Hz (from 6 x 6/2 x 2, or 80% occupancy in central Pb+Pb(GEANT) -> 25% min bias -> 2 k. Hz) – GTL bus rate = (12 FEE)(32 chan)(2 Gain)(10 bit)(20 samples)(2000 Hz)=38. 4 MB/s – RCU data rate = 2*GTL/RCU partition=77 MB/s – If tpeak = 4 ms with 200 samples then GTL bus rate=384 MB/s - Death! 18

EMCAL vs PHOS Readout Parameters 19

PHOS vs EMCal Readout comparison • Commonalities: – – Same APD + preamplifier Same GTL bus (but not identical) ~Same FEE Same RCU, TRU, etc • Differences – Different T-Card: FEE located far away, need signals driver on Tcard+twisted pair – Same FEE but with shorter shaping time, 100 ns – Numerology, FEE to GTL to RCU, TRU – New (later option) TRU’ to form larger area energy sums for jet trigger. • Other – Power consumption: 63 m. W*1152 = 73 W in SM, 450 W in FEE region of SM 20

EMCal Electronics: Numerology Tower APD+Pre. Amp 8 Transition Card 4 32 Channels FEE-card w/ ALTRO 12 3 OR per SM Level 0 384 Towers TRU = Trigger Router Unit Level 1 36 Level 1 , . . 36 1152 Towers 13824 Towers 1153(768 + 384) TRU’ = RCU = Trigger Router Unit’ Read-out Control Unit 2(1. 5) RCU/Super. Module = 1152 Towers (cf. 896 21 PHOS)

EMCal Readout Matrix per Supermodule Totals/Super. Module 36 FEE cards 3 GTL bus 3 TRU 1 RCU 22

Additional Slides 23

EMCAL Physical Parameters 24

EMCAL Readout Parameters 25

26

27

PHOS FEE • 9 Pre-production prototypes produced at Huaxiang University of science and technology. • Used in PHOS test beam period of Oct. ’ 04). 28

29

EMCAL: main jet physics capabilities S. Blyth, QM 04 1. Level 1 trigger for jets, p 0/g • essential for jet ET>50 Ge. V 2. Improved jet energy resolution • charged-only jets: poor resolution (>50%) • TPC+EMCAL: resolution ~30% • main effect: out-of-cone energy (R~0. 3 for heavy ions) • also: intrinsic resolution; missing n, K 0 L, n 3. p 0/g discrimination to p. T~30 -40 Ge. V (cross section limit for g+jet coincidences in acceptance) 30

Tower granularity (cont’d) Heather Gray, LBNL/Cape Town p 0 gg opening angle g/p 0 shower shape discrimination preliminary p 0 rejection for p. T<~30 Ge. V/c g p 0 More sophisticated SSA underway, possible large improvements Additional g+jet issues: • other backgrounds: fragmentation g, radiative decays, … • isolation cuts g+jet is important but limited measurement 31 fixed $$$: maximize acceptance for jets, granularity driven by cost

Soft, Slow (neutron) Background Kill the number of neutron hits by tower threshold or (integration) time cut. Tower Cut 1 100 Me. V 2 150 Me. V 3 200 Me. V 4 500 Me. V Time Integ. 0 20 ns 1 30 ns 2 50 ns 3 100 ns 4 200 ns 5 500 ns 6 1000 ns Tower threshold cut of ~150 Me. V is effective, but it doesn’t remove neutron energy deposit in tower with real gamma hit! Integration time cut can also reduce the number of neutron hits. Benefit also applies to tower with real hit. Note: Using PHOS cluster algorithm without splitting. 32 Calculations by Heather Gray

Soft, Slow (neutron) Background 10 -20 Ge. V/c g + HIJING (b<3 fm) Full ALICE Tower Cut 1 100 Me. V 2 150 Me. V 3 200 Me. V 4 500 Me. V Time Integ. 0 20 ns 1 30 ns 2 50 ns 3 100 ns 4 200 ns 5 500 ns 6 1000 ns Tower energy threshold and integration time cuts are correlated. Shortening integration time allows to lower tower energy resolution, which will improve performance especially at low p. T. Note: Using PHOS cluster algorithm without splitting. Feasible to use a shaping time of ~100 ns with PHOS electronics? 33 Calculations by Heather Gray

Soft, Slow (neutron) Background The Alarming Plot… due to large clusters Taking the shower core only… Conclusion: Neutrons cause large occupancy - difficulty for cluster finding. Will need to use shower core with high tower threshold. Shorter shaping time will improve the situation. Again: This is for Central HIJING (worse case, the problem is centrality dependent). 34 Calculations by Heather Gray

EMCal L 0 trigger input concerns… • Upon receipt of L 0, the ALTRO chip keeps 14 presamples: – For PHOS with 10 MHz sampling this is region of 1. 4 ms prior to L 0. – For EMCal with 20 MHz sampling this is region of 700 ns prior to L 0. – With ALICE L 0 latency of 1. 2 ms • For 10 MHz sampling this is just okay with ~no presamples • For 20 MHz sampling this is 300 ns after 200 ns peaking time - Death! • Proposed PHOS solution is to use local PHOS L 0 trigger output as ALTRO L 0 trigger input. Would “solve” problem for EMCal also, but… – This seems to be a very dangerous solution… • L 0(PHOS). ne. L 0(CTP): might have L 0(CTP) without L 0(PHOS) then L 2 request when there was no L 0… • Danger of filling ALTRO buffer with noisely local L 0’s? – Only alternative for EMCal seems to be to keep 10 MHz sampling and go to 200 ns shaping time. 35

EMCal Jet Trigger (TRU’? ) Conclusion: Increasing trigger region requires in increase trigger threshold for same trigger rejection factor (e. g. central HIJING). Not much difference in trigger efficiency (on PYTHIA jets) versus trigger region size - except for large patch sizes. 36 PHOS TRU size (4 x 4 tower) works quite well… Calculations by Bill Mayes