hep tel aviv university lumical (pads design) simulation ronen ingbir fcal simulation meeting,...

HEP Tel Aviv University

LumiCal (pads design)Simulation

Ronen Ingbir

FCAL Simulation meeting, Zeuthen

Tel Aviv University HEP experimental Group

CollaborationHigh precision design


Presentation Outline

Reconstruction Algorithms

Detector Simulation

(& Design

optimization)

Fast Detector Simulation

(Luminosity)

Events Selection

Physics Simulation

Electronics Simulation

(Basic model)

And More …

Analyses



Electrons/Bhabhas

)(rad

Bhabha

Pure electrons

Electron energy (GeV)

Pure electrons

Bhabha


Electrons/Positrons - Geant-3 integrated generator.

Bhabha scattering - BHWIDE generator

Beamstrahlung and Beam spread

HEP Tel Aviv UniversityCollaboration

High precision design

Beamstrahlung - Circe generator.

Beam spread – Included separately (home made routine).


250GeV

26 mrad

Events selection - old approach

Gaussian fit and energy

calibration based tail cut

Acceptance cut was based

on the leakage

rec

Detector signal



33 mrad

Energy Resolution )( GeV

)(rad

The most significant event selection cut is the geometric acceptance cut.

This cut was used to get the best energy, angular resolutions and minimum biases.

Events selection



Left-Right balance RL

Simulation distribution

Distribution after acceptance and energy balance selection

Right side detector signal

Lef

t si

de

det

ecto

r si

gnal Right signal - Left signal

)(21 rad

(deg)21 Collaboration



Out

In

Eout - EinEout + EinP=

New selection cut

Ring

Signal



Eout-EinEout+EinP=

)(rad

Out

In



3 cylinders

3 Rings

2 cylinders

3 Rings

1 cylinders

3 Rings

Eout-EinEout+EinP=

)(min radCollaboration





Angular reconstruction:

• ‘Normal’ energy weight

• Logarithmic weighting with selection cutoff.

• Logarithmic weighting per layer (Bogdan’s algorithm).

• ‘Real life’ approach.

Energy reconstruction:

• Energy resolution.

• Energy calibration.


Reconstruction Algorithm

)]}ln([,0max{T

ii E

EconstW

i

ii

E

EXX

i

ii

W

WXX

Events Num.

)(radgenrec

)(radgenrec

)(radgen

We explored two reconstruction algorithms:

The log. weight fun. was designed to reduce steps in a granulated detector :

1. Selection of significant cells.

2. Log. smoothing.

Log. weight.

E weight.

T.C Awes et al. Nucl. Inst. Meth. A311 (1992) 130.



Logarithmic Constant

Constant value

Constant value

)]}ln([,0max{T

ii E

EconstW

After selecting:

We explored a more systematic approach.

The first step is finding the best constant to use under two criteria:

1. Best resolution.

2. Minimum bias.

400 GeV

))(_( radmean genrec

))(( rad



Energy dependent constant

)]}ln()([,0max{T

ibeami E

EEconstW

The goal is to find a global weight function.

Is the the log. weight constant really a constant ?

Constant value Collaboration



Shower reconstruction

Nu

m. o

f C

ells

Nu

m. o

f Se

ctor

s

Nu

m. o

f C

ylin

de r

s

En

ergy

por

tion

(%

)

En>90%

)(rad )(rad

)(rad)(rad

What happens when we select the best log. weight constant ?

Shower size

Log. Weight Selection

Most of the information is in the selected cells.



Algorithms Comparison


15 Cylinders 40 Cylinders 64 Cylinders

Pads - Full Log.

Pads - Bogdans log.

Strips - Bogdans log.

13*10e-5 5*10e-5

14*10e-5

4*10e-5

1. Two methods are equivalent - reason : in both cases Theta is reconstructed directly.

2. There is a saturation limit for the resolution (before 64 cylinders).


Real life algorithm

Working with both sides of the detector and looking at the difference between the reconstructed properties:

(In real life we don’t have generated properties)

2new



‘Pure’ electrons simulation Bhabha+Beam+BS(5e-4)

)(radgenrec

Bias study

44 103.1107.5


Two plots convinced us that the bias observed is not detector design dependent nor imperfect algorithm.


48 sectors design

(deg)genrec X (cm)

Y (cm)

Bias study



Magnetic field



Energy resolution

Ntuple

No

cuts

With

cuts

Pure electrons

31% 29%

Bhabha

42% 24%

Bhabha + Beamstrahlung

45% 24%

Bhabha + Beamstrahlung + Beam spread (0.05%)

46% 25%

Bhabha + Beamstrahlung +Beam spread (0.5%)

49% 29%



Angular resolution


Pad design

0.34 cm Tungsten

0.31 cm Silicon

Cell Size1.3cm*2cm>1.3cm*6cm<

~1 Radiation length

~1 Radius Moliere


15 cylinders * 24 sectors * 30 rings = 10800 cells

RL

8 cm

28 cm

6.10 mCollaboration


Dense0.156e-3Regular0.137e-3

Dense, 35%

Regular, 27%

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Ang

ular

res

olut

ion

(rad

)E

nerg

y re

solu

tion

(GeV

)^0.

5

Regular

Dense


0.34 cm Tungsten

0.31 cm Silicon

0.55 cm Tungsten

0.1 cm Silicon

Dense design



30 radiation length detector47 radiation length detector

Z (cm)

Detector Signal

0.8cm

1.1cm

Moliere radius & Radiation length



)(rad

Energy Resolution )( GeV

)(rad

Events

New geometric acceptance



Detector optimization



Optimization



Margins around cells

Having margins

Means

Losing InformationCollaboration



One cylinder One sector

Radius (cm) ) deg(

Det

ecto

r si

gnal

Det

ecto

r si

gnal

Loosing information



Margins in between cells

Energy resolution Polar resolution



Our basic detector is designed with

30 rings * 24 sectors * 15 cylinders = 10,800 channels

Do we use these channels in the most effective way?

Maximum peak shower design

30 rings 15 cylinders

20 cylinders

10 cylinders

24 sectors * 15 rings * (10 cylinders + 20 cylinders) = 10,800 channels

4 rings 15 rings 11 rings

10 cylinders



Maximum peak shower design

Basic Design

Angular resolution improvement

without changing the number of

channels

Other properties remain the same

))(_( radmean genrec

))(( rad

Constant value

Constant value

Polar reconstruction

0.11e-3 rad0.13e-3 rad



Electronics Simulation


Normal Cells

'Dead' Cells

Noise Cells

totalNnormalNdeadNnoiseN ____

Noise Function

5_

_

NoiseTotal

SignalTotalCase Study in which:

min max

(Signal+noise) - (signal)

70% Noise Cells example

Noise signal

N

N

(Basic model)

Random selection of cells



0%

50%

25%

10%

Signal (including Dead Cells)Signal (including Noise Cells)

100%

0% 50% NN

5_

_

NoiseTotal

SignalTotal

Noise Cells / Dead Cells

Case Study:



)%(

)( GeVHaving ‘Dead’ cells affects both the reconstruction of the energy and the angles in the same way.

Having Noise in the cells:

1. Assuming noise function with no extreme values.

2. Assuming high signal to noise ratio.

3. Using log weight function with revised cut value (that takes into account the new average ‘zero level’).

Similar resolutions

Performance with noise and dead cells

preliminary


Fast Detector Simulation


Motivation :

High statistics in required to notice precision of : 410L

L

There is an analytic calculation (and approximation) :

(Which is the precision goal of the ILC)

*2

L

L

Is it a good approximation ? This calculation takes into account only the Bhabha angular distribution, how does other factors affect (backgrounds, detector design, electronics noise) ?

We need a mechanism to actually count events, as if it was real life.


High Statistics MC

)),(( N

NL

L

)(rad)(rad

BHWIDE + fast detector simulation


min

*2

L

L 310


R&D status

))(,,),(( EENN

LL High statistics MC

410 N

NElectronics noise and ‘Dead’ Cells


Design optimization (Dense design, Margins in between cells, Maximum peak shower design)

Pure electron MC

Detector properties

Events selection

‘Real physics’ MC

Bhabha + Beamstrahlung + Beamspread



Future Steps


‘Real physics’ MCFinal

optimization

Luminosity with a crossing angle

Luminosity with polarised beams (Gideon Alexander)

Pad/Strip comparison

Geant-4 transition

Additional background studies (two photon events, beamstrahlung hitting the detector)

Additional hardware design constrains and electronics simulation (digitisation, reality noise parameters, silicon production constrains)


THE END


hep tel aviv university lumical (pads design) simulation ronen ingbir fcal simulation meeting,...

Documents

weight constant

best log

best energy

detector signalleft

energy reconstruction

granulated detector

detector design dependent

selection cutoff