Roma, 22/11/01 CMS Software & Computing Workshop - E. Longo
1
Calibrazioni del calorimetro:esperienze (L3) e prospettive
Egidio Longo
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
2
a ~ 2.5%/√Eb < 200 MeV c ~ 0.5%
and an angular resolution ~ 50 mrad/E
CMS ECAL benchmark
m = 2 E1 E2 (1 - cos)
σm
m=
12
σ1
E1
⎛
⎝ ⎜ ⎞
⎠ ⎟
2
+σ 2
E2
⎛
⎝ ⎜ ⎞
⎠ ⎟
2
+σθ
tgθ /2
⎛ ⎝ ⎜ ⎞
⎠
2
+⎡
⎣ ⎢
⎤
⎦ ⎥
1/ 2
σ E( )E
=aE
⊕bE
⊕ c
H (mH 100 GeV) ~ 2 – 100 MeV H /mH ≤ 10-3
target
low mass Higgs discovery:
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
3
resolution: stocastic term a
• photostatistics contribution, including– LY– light collection efficiency– geometrical efficiency of the photodetector– photocatode quantum efficiency
Npe/GeV = 4000 for 0.5 cm2 APD 1.6%
• electron current multiplication in APD, contributinga square root of excess noise factor, F = 2
1.61.4 = 2.25%
• Lateral containment (55 matrix) 1.5%
Total stochastic term a = 2.7 %
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
4
resolution: noise term b40 ns shaping time, summed over 5x5 channels• Serial noise (p.d. capacitance) 1/t
– 150 MeV• Parallel noise (dark current) t, mostly radiation
induced– negligible at the start of the experiment– 30 MeV after one year at low luminosity– 100 MeV after one year at high luminosity
• Physics pile-up (simulated, with big uncertainties)– low luminosity 30 MeV– high luminosity 100 MeV
Total contribution– low luminosity 155 MeV– high luminosity 210 MeV
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
5
resolution: constant term c
• leakage (front, rear, blind material)CMS full shower simulation < 0.2 %
• system instabilities designed to be at the permill level– temperature stabilization < 0.1 ˚C ( LY = -1.9 % per ˚C)– APD bias stable at 20 mV (dM/dV = 3%/V)
• light collection uniformity,Specifications to stay < 0.3% reached by
single face depolishing
• Key issue to have c 0.5 % intercalibration by monitor and physics signals at 0.5 % including the radiation damage effect
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
6
resolution at test beam on 1999 prototype30 preproduction crystals and APDs
fit as a function of E:
(few crystal intercalibration)
σΕ
=2.74%
E⊕0.40%⊕
142MeVE
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
7
• Is it realistic to require a target0.5 % intercalibration
for such a large scale calorimeter?
• Look at L3 experience
intercalibration
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
8
L3 BGO calorimeter
• 7680 (barrel) + 3840 (EC) BGO Crystals
• PIN diode read-out (no gain)
• 12 bit ADC for 6 gains (1 to 512)
• noise level around 1 MeV
• sparse scan (zero suppression) read-out
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
9
L3 calibration tools
• Individual Test beam calibration (2-50 GeV e-)
• Xenon monitor of the response once per day
• Bhabha events at Z pole (45 GeV e)
• at LEP200, RFQ calibration (17 MeV )
Standard in situ procedure:• Individual intercalibration with Xe lamps• Absolute calibration with bhabha electrons
for groups of crystals sharing the same Xe lamp
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
10
test beam calibrations
• Individual calibration for 7680 crystals– two half barrels in two years (1987,1988)– first half barrel recalibrated in 1988– 4 energies (2, 10, 20 and 50 GeV) – temperature control (nominal ± 0.5 ºC)– temperature monitor (1300 sensors) at 0.2
ºC– fully automated procedure:
sequence, positioning, monitor and quality check
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
11
intercalibration in 1990
after Xe corrections 1.25 %- 0.9 % 0.87 %
measured resolution MC predictedintercalibration spread
(average response decreased by 5 %, due to ageing of some optical component)
1990 Bhabha spectrum
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
12
10 years of follow-up
1990
1991
1992
1993
1994
Barrel
Time (days)
0 200 400 600 800 1000 1200 1400 1600 1800
Electron energy/Beam energy
0.88
0.9
0.92
0.94
0.96
0.98
1
1.02
RB26 (Hb 1)
RB24 (Hb 2)
• system able to track this response decrease (few %/year)
• porting of previous year calibration: 1.3%
• spread after Xe+Bhabha corrections: 0.8%
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
13
follow-up at LEP200• at LEP200 (too few Bhabha) add RFQ
gunsend protons on Li target, producing17 MeV photons, million events in few days
• year after year, algorithms improved – several months of work to get final
calibrations every year
• latest result: 1.06% resolution measured on 45-100 GeV Bhabha electrons intercalibration contribution 0.5 %
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
14
conclusions for L3• L3 BGO calibration combined test-beam
precalibration, monitor and physics events• Porting from test-beam to the experiment was
performed at < 1.5%• After all corrections, first year intercalibration
was known at 0.9%• After ten years of improvements,
intercalibration known at 0.5%
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
15
CMS-ECAL calibration scheme
• individual crystal precalibration at beam test at several energies
• intercalibration porting to the experiment via laser monitor system
• follow-up of radiation damage by laser monitor system
• calibration by physics events every few months
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
17
ECAL installation in
V31
SM assembly
electronics
calibration
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
18
precalibrationWe probably will not be able to precalibrate all
ECAL supermodulesWe may be forced to choose between • the “production serial precalibration” of the
maximum number of modules, or• the precalibration of a smaller number of
modules with deeper understanding of a few modules to be recalibrated several times
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
19
intercalibration• Issue is crystal intercalibration
(the energy scale can be fixed relatively simplyreconstructing the Z invariant mass with just a few Z ee)
– Raw intercalibration from lab measurements < 6%
– Target test beam pre-calibration < 2%
– Target final intercalibration using electrons 0.4
• Is precalibration crucial?– Not clear how well HLT system will
work with 6 % intercalibration
– Much easier to understand what is going on with 2 % intercalibration
– Precalibration ultimate system test before LHC data taking
(F. Cavallari)
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
20
in situ calibrationsBaseline intercalibration uses W electrons measured in tracker• Calibration electron selection and its efficiency
What efficiency to select ‘good’ (not much brem) electrons that can be well reconstructed in 5x5 crystals?
• Calibration algorithmHow many electrons per crystal needed; what is best algorithm to unscramble
the calibration constants?
• To do: simulate the whole procedure (milestone 3/02):Introduce inter-calibration errors and show how well they can be recovered
Use of Z ee invariant mass• 5 times less events
• “internal” ECAL calibration (minimal tracker information needed)
• can be combined with monitor for absolute calibration
Large samples of W’s and Z’s to study these issues
( P. Meriadiani talk)
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
21
alternative intercalibrations
What to do if missing pre-calibration and poor electron capability (tracker) ?
• look at the energy flow in minimum bias– dependence, simmetry (?)– trigger bias
• study on sample of500k fully simulatedmb events available
• cut out noise
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
22
energy flow calibration (1)Total energy after truncation cuts (200 MeV). • Each entry in histogram represents total
energy in a single crystal. • 360 crystals at the same Since MC crystals are perfectly calibrated,
the width represents the error on intercalibration achievable with 500k
events
2% requires 11 M min-bias events
6.6 M crossings
183 hours (at 10Hz)
(< 2 months for 30% duty cycle)
2 hours (at 1kHz —to do in HLT farm)
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
23
energy flow calibration (2)• limitations:
– intercalibration per raws only (combine with Z absolute calibration ?)
– tracker inhomogeneitiesneeds for realistic tracker services simulation
(and a combined tracker-ECAL test beam?)
– sentitivity to channel-to-channel noise variations
• better algorithms? – slope of ET rather than truncated mean
Roma, 22/11/01 CMS Software & Computing Workshop - E.
Longo
24
conclusioni• intercalibrazioni @ 0.5% difficili ma
possibili• baseline calibrazioni in situ con W
milestone per 03/02• Roma partecipa attivamente agli studi
di calibrazioni con W e Z• precalibrazione completa impossibile;
studi su possibili alternative iniziati da poco
• simulazione realistica del tracker essenziale per le prossime produzioni
• test beam con sezioni di tracker e calorimetro auspicabili