X-UIDL: 818085698.003
Date: Mon, 4 Dec 95 00:32:25 -0800
From: Xiaorong Shi <shi@hep1.llnl.gov>
To: Bob_Jacobsen@lbl.gov (Bob Jacobsen)
Cc: John.Allison@man.ac.uk, N.I.Geddes@rl.ac.uk, DAUNCEY@v2.rl.ac.uk,
        SHI@hep1.llnl.gov, MARSISKE@slac.stanford.edu
Subject: EMC reconstruction software meeting

Dear all,
Here is the ASCII version of my transparency. Will Neil please forward
this to UK members whom I don't have their email addresses.

xiaorong



              What is in the CVS, what will be in soon.

                        Xiaorong Shi
                Lawrence Livermore National Lab

Page 1:

Definition of some terms:

Hits: as defined in EmcGHit. Energy deposited in the Xtal by a certain track.

Digi: Energy deposited in a xtal, sum up all tracks. In BaBar, it is hold
      in EmcDigi.

Local Maxima: A Digi with energy above a certain cut1 AND is greater than
              all its side and corner neighbors.

Cluster: There are two versions of this,
         a) Geometrically connected region where all digi's are above
            a certain cut2. It can have more than one Local Maximum.
         b) see Bump.

Bump: A few Digis around the Local Maxima caused by the one particle(we think).
      Some Digis have to split its energy between two Bumps.

Page 2:

Digi from DAQ and MC:

In the data stream, we get digis after the feature extraction and multiplied
by calibration constance. So it is in GeV. In the case of Monte Carlo,
GEANT uses GeV.

Page 3:

Data members of class EmcGHit(EmcData/EmcGHit.hh):

  protected: float _x;
  protected: float _y;
  protected: float _z;
  protected: float _t;
  protected: float _edew;
  protected: float _edep;
  protected: int _idnt;
  protected: int _igtr;

Page 4:

Data members of the class EmcDigi, file (EmcData/EmcDigi.hh):

class EmcDigi {
....
private:
  
  float _energy;
  int _theta;
  int _phi;
  float _time;
  HepAList< EmcGHit > *_memberAList;

};

This is just a minimum necessity. 
Possible additions:
1) A status word for hot crystals, 
2) etc.

Page 5:

Without outside information, there are two stages in the Emc reconstruction:

1) Clustering.

2) Find the position and energy of the photon or pi0.

With outside information, (I think this is not the topic of this meeting).

1) separate MIP;
2) K0L candidate;
3) Subtract hadron shower bumps in a multibump cluster;
4) etc.

Page 6:

Methods used to go from Cluster information to Energy and position:

1) position:    COG + position correction.
   energy:      sum of digis in the Bump.

2) position:    non-linear weighted COG.
   energy:      sum of digis in the Bump.

3) position:    shower shape fit.
   energy:      sum of digis in the Bump. If there is dead xtal in the
                Bump, the energy can be recovered.

Speed: (1) and (2) are fast, (3) is slow.

If properly done, all three give similar position resolution.

Page 7:

Numbering scheme of the digis:
In the EmcGHit, there is a hit element _idnt = 128*theta + phi.
Theta and phi are decoded from NUMBV array from GEANT HITS bank and
they are made to follow the convention below,

theta: from 1 to 59, where 1 to 9 are for endcap and 10 to 59 are for barrel.

phi: It is theta dependent,
     For theta = 1 - 3, phi goes from 1 to 80;
         theta = 4 - 6, phi goes from 1 to 100;
         theta = 7 - 59, phi goes from 1 to 120.

phi start from 3 o'clock counter clockwise, looking from +z to -z.

There are helper functions to decode _idnt to theta and phi.

Page 8:

What exist in the repository:

EmcData/EmcGHit.hh(.cc), the .dhh and .dcc include files therein.

EmcData/EmcDigi.hh(.cc), hand coded with variable and function names
identicle to the ones would have been generated by dbio. Checked in.

EmcReco/EmcMakeDigi.hh(.cc), checked in.

EmcData/EmcCluster.hh(.cc), hand coded, not checked in yet.

EmcReco/EmcMakeCluster.hh(.cc), not yet checked in.

EmcData/EmcBump.hh(.cc), defined, not yet coded.
EmcReco/EmcMakeBump.hh(.cc), defined, not yet coded.


The classes in EmcReco/EmcMakeXxxxx, are Framework Modules.
e.g.
EmcMakeDigi takes a list of EmcGHit as input and outputs is a list of 
EmcDigi. 

Page 9:

EmcData/EmcCluster.hh(.cc) and EmcReco/EmcMakeCluster.hh(.cc) will be
checked in within a week.