CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage :...

34
CAT plenary meeting VLIMANT Jean-Roch LPNHE 13 may 2003 T42 algorithm dedicated webpage : http://www-d0.fnal.gov/~vlimant/noise-suppdescr/Des cription.html T42 + Top_analyze +d0root_analysis Principle & Implementation Effect on cells Effect on EM objects NO T42 T42

Transcript of CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage :...

Page 1: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

CAT plenary meeting

VLIMANT Jean-Roch LPNHE 13 may 2003

T42 algorithmdedicated webpage :

http://www-d0.fnal.gov/~vlimant/noise-suppdescr/Description.html

T42 + Top_analyze +d0root_analysis Principle & ImplementationEffect on cellsEffect on EM objects

NO T42 T42

Page 2: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

T42Principle : reject isolated medium cellsNamed after an algorithm used in H1

Reject negative energy cells Select high signal cells (4 sigma) Keep their significant neighbours (2 sigma)

Thresholds 2 is 2.5 at the moment

What neighbours means

Page 3: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

cal_t42 Packagecreated after cal_nada classes

Classes : Calt42Struct : neighbourhood object Calt42 : container for cell lists Calt42Chunk : container for Calt42 object Calt42Reco :

– Does the cleaning and creates a Calt42Chunk with rejected cells and modify/or not the CalDataChunk (killing/shadow mode).

– Get the neighbourhood info from a Calt42Struct object which allows a linear algorithm complexity.

Calt42Combine :– When shadow mode in d0reco : produces a clean

CalDataChunk– When killing mode in d0reco : produces a un-clean

CalDataChunk And registerers RegCalt42Reco RegCalt42Combine

Page 4: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42Reco 1/2

Alogirthm framework (part1)

List of cells from the CalDatChunk :

cclist

Is the cell to be kept anyway ? YES Put into

keeplist

NO

Is the cell energy above* first threshold

YES

Is the cell energy above* second threshold YES

NO

Link Calt42Struct to the cell (DoLink) : put into

the access list

Put into wastelist

Put into keeplist and flagged as kept

Page 5: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42Reco 2/2

Alogirthm framework (part2)

Loop over the access list of Calt42Struct object

Is the neighbor linked YES

Put into wastelist

The Calt42Struct object is read for still flagged cells

(Linked but not neighbored)

Loop over the neighbours of the item of access list

(use of a Calt42Struct object)

Put into keeplist

Page 6: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42Structclass content

Objects– spine : array[81600] of structured object

Mapped to ieta, iphi, layer by :INDEX = iphi + 64( layer - 1 ) + 1088( ieta + 37 ) - 1

– Structured object fields sigma : tells the noise rms of the cell pointers to all possible neighbors within the array single pointer to be assigned to a CalCell available in CalDataChunk

Methods– Access– Fields initialiaser

The Neighbourhood information is read from the file STRUCTURE.dat

Page 7: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42Structuse in Calt42Reco

Page 8: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42StructNeighbourhood information

http://www-d0.fnal.gov/~vlimant/noise-suppdescr/Description.html

How to check ?– Trust me– WebPage : check rules and pictures– Calt42Struct::PrintNeighbours(ieta,iphi,layer)– Calorimeter display : to be written

Regular cells–normal–out of fine layers (3,4,5,6)–near fine layers (2,7)–Layer 32/33

Exceptions cells–Neighbours above or below splited into different layers–Intercryostat and nearby cells

Page 9: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Calt42CombineOnly for DST files

Only one parameter Do you want the t42 cleaned CalDataChunk ?

Automatic detection of the d0reco mode by cell matching. Modification of the CalDataChunk

Page 10: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

In the frameworkneed for post-reconstruction when shadow

mode

For DST / TMB

In the main rcpString packagename = “…unpack T42 rebuild …”

T42 = < cal_t42 Calt42> // process the t42 algo

or

T42 = < cal_t42 Calt42Combine> // only combine chunks

“rebuild” must reconstruct the object chunks you want to see the effect on

Page 11: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Study effects on dataData sets

EM stream data 10000 evts ofWZskim-emStream-*-*.raw_p13.06.01

under the project name WZstreamTMBp13.06

Unstreamed data 10000 evts in filesrecoT_all_0000174234_mrg_*-*.raw_p14.01.00

Global_CMT run

Min/Zero bias Wzskim 10000 evts ofWzskim-mzbiasStream-*-*.raw_p13.06.01

Under the project name mzbias_mvh.all1

Page 12: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Population reportunstreamed data

Page 13: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Cells effectclose to beam

EC regions More rejection than expected noise !

– Too many cells with negative energy arbitrarily increases the rejection; They are not from gaussian noise only, probably due to pile-up

– This is due to cell close to the beam 33-37

Negative energy rejection is unsuited to this regionShould be treated separately

•Keep untouched above |ieta|=33(CH) and 32 (EM, FH)•Even lower for FH ?

Page 14: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region Labels

CC_CH

CC_FH

CC_EM

EC_CH_crack

EC_CH

EC_FH E

C_E

M

Page 15: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportemstream data

Page 16: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportWZstream data with problems

Page 17: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportunstreamed data

Page 18: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportunstreamed data with problems

Page 19: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportminimum/zero bias data

Page 20: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Region reportminimum/zero bias data with problems

Page 21: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Cells effectaway from beam

CC regions

Statistically noise like rejection – Always about the same number of negative cells in the

event : probably noise and it is always rejected– Coherent number of rejected cells Region occupancies

show signal presence– ~40% on unstreamed data

Cells clustering decreases with the distance to interaction point

-/+ rejected ratio for region threshold tuning ? Although skim dependent (topology dependence)

Page 22: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Still too much rejection in FH part need to be investigate

Peaks are due to– Overpopulated cells– Cell with noise problem– strange regions ?

Cells effectclose to the beam

Page 23: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EM clustersdata/MC sets

Zee picked sample p13.05 2418 evts (not used)– Filename : /rooms/living/varnes/scratch/cab/zee-p13/*/tmbfile

Zee Monte Carlo p13.08 6000 evts– Project name ctf_p13.08_zee_sig2.5_tmb

– Filename : tmb-x13.08.00_CTF13.08.00-02-z-ee_np_mcp13_hepfm007.uta.edu_null_*

Estimators study – At high/low energy : emstream– At low energy : J/ skim from Jan Stark (not used yet)

Page 24: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EM clustersNumber of cells

Page 25: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EM clustersNumber of cells

Page 26: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EM clustersestimators pT<13 GeV

Page 27: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EMidestimators pT<13 GeV

Page 28: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on electronsestimators pT<13 GeV

Page 29: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EM clustersestimators pT>13 GeV

Page 30: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on EMidestimators pT>13 GeV

Page 31: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Effect on electronsestimators pT>13 GeV

Page 32: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

T42 effects

10% more EM cluster (10,11 ) 20% less above 13 GeV

15% more Emid clusters 80% more below 13GeV

11% more electrons 34% more below 13GeV

Page 33: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

EM Efficiency &fake rates

data set

DIEM15 stream of WZ skim 1.5MevtDIEM15-WZskim-emStream-*-*.raw_p13.06.01 (05)

under the project name Wzskim_diem

NEXT TIME

Page 34: CAT plenary meeting VLIMANT Jean-RochLPNHE13 may 2003 T42 algorithm dedicated webpage : vlimant/noise-suppdescr/Description.html.

Conclusion Code run into p14 code

– T42 chunk at DST level (checked)– D0note 4146 update soon

Noise cleaning– Electronic noise seems to be deleted– Pile-up problem near beam pipe remains

Get more electrons and photons NEXT :

– Fake and efficiency– More at low energy with J/ skim