Merge branch 'master' of https://git.cern.ch/reps/AliRoot

[u/mrichter/AliRoot.git] / HMPID / api.txt

How to open session:
	use static method  AliRunLoader::Open("galice.root","AlicE","update") or just AliRunLoader::Open() for defaults. 
        Returns pointer to AliRunLoader on success or fatal termination on error
How to get total number of events in galice.root:
	AliRunLoader::GetNumberOfEvents()  (or AliRun::GetEventsPerRun() using f.e. gAlice deprecated)
How to get pointer to HMPID:
        AliRunLoader()->GetAliRun()->GetDetector("HMPID") but before one needs to AliRunLoade()->Set
How to avoid using gAlice:
	detector->GetLoader()->GetRunLoader()->GetAliRun() returns gAlice global pointer.
How to retrieve pointer to alice run loader:
        use pHMPID->GetLoader()->GetRunLoader() (all detector classes inherit from AliDetector which has GetLoader())
	use method AliRun::GetRunLoader for gAlice (deprecated)
How to get pointers to different root trees:
	TreeE belongs to AliRunLoader, available after AliRunLoader::LoadHeader()
	TreeK belongs to AliRunLoader, available after AliRunLoader::LoadKinematics()
	TreeH belongs to AliLoader   , available after AliLoader::LoadHits()
	TreeS belongs to AliLoader   , available after AliLoader::LoadSDigits()
	TreeD belongs to AliLoader   , available after AliLoader::LoadDigits()
	TreeR belongs to AliLoader   , available after AliLoader::LoadRecPoints()
        all methods return 0 on success. 
How to get event of interest:
	AliRunLoader::GetEvent(event_number) returns 0 on success
How to deal with the stack of particles?
        - first of all, the stack includes primary as well as secondary particles
	- pointer to the stack is taken:
          	AliRunLoader::Stack() but before one needs to load event header by AliRunLoader::LoadHeader() otherwise both methods return 0.
          	Moreover loading header gives the information about number of particles only. 
          	To retrieve the list of particle one also needs to load kinematics by AliRunLoader::LoadKinematics()        
	- total amount of particles in stack for a given event:        
         	AliStack::GetNtrack() 
	- total amount of primary particles in stack for a given event (after LoadHeader()): 
        	AliStack::GetNprimary() 
How to retrieve hits:
	Hits are stored on primary by primary basis (hits corresponding to primary particles stored in a single entry; total number of entries in hits tree equals
        to number of primaries; order of primaries is not preserved). Hits for the given primary is TClonesArray.
   	To retrieve all hits one needs to do:
     	-initialize the root tree and containers:   pRich->GetLoader()->LoadHits();  (AliLoader::LoadHits() returns 0 on success)
     	-read number of entries in TreeH:           pRich->GetLoader()->TreeH()->GetEntries()
     	-then for each entry:                       pRich->GetLoader()->TreeH()->GetEntry(i)	
How to retrieve sdigits? 
	Sdigits stored in tree S with the branch of TClonesArray, all sdigits in a single TClonesArray
	So the tree has only one entry.
	One needs to say:
	-pRich->GetLoader()->LoadSDigits(); this one open file, get the tree and invoke AliHMPID::SetTreeAddress()    
How to retrieve digits? 
	Digits stored in tree D with the 7 branches of TClonesArray, one per chamber, all digits of a given chamber in a single TClonesArray
	So the tree has only one entry.
	-One needs to say:
	  pRich->GetLoader()->LoadDigits(); this one opens file, gets the tree and invoke AliHMPID::SetTreeAddress() which in turn corresponds 
          branches of the tree to the digits containers in memory. There are 7 containers, one per chamber, all of them belong to AliHMPID.
	-Then one needs to take the tree entry (only one) to the memory:
          pRich->GetLoader()->TreeD()->GetEntry(0) 
	-Finally pRich->Digits(chamber_number) returns the pointer to TClonesArray of AliHMPIDdigit          
What are the debug methods avail:
        AliLog::SetGlobalDebugLevel(AliLog::kDebug)
How to get info for a given particle number:
	Header and Kinematics trees must be loaded, then possible to retrieve pointer to Stack of particles
	Int_t AliRunLoader::LoadHeader(); Int_t AliRunLoader::LoadKinematics()
	AliStack *AliRunLoader::Stack()
	TParticle *AliStack::Particle(tid)
	TParticle::Print()
How to deal with AliRunDigitizer:
	AliRunDigitizer::Exec() just call AliRunDigitizer::Digitize()	
What are the meanings of different VMC flags:         
	gMC->IsTrackAlive()
	gMC->IsTrackStop()
	gMC->IsTrackDisappeared()
How to get pad number for a local position:
	use static TVector AliHMPIDParam::Loc2Pad(TVector2 position);

How to check if a given stack particle is primary:
	Stack is TClonesArray of TParticle. TParticle::GetMother(0) returns -1 if it's primary (no mother)         
How to loop over all possible object:         
  for(Int_t iEventN=0;iEventN < GetLoader()->GetRunLoader()->GetAliRun()->GetEventsPerRun();iEventN++){//events loop
    for(Int_t iEntryN=0;iEntryN < GetLoader()->TreeH()->GetEntries();iEntryN++){//TreeH loop
      GetLoader()->TreeH()->GetEntry(iEntryN);//get current entry (prim)  
      for(Int_t iHitN=0;iHitN<Hits()->GetEntries();iHitN++){//hits loop
        AliHMPIDHit *pHit=(AliHMPIDHit*)Hits()->At(iHitN);//get current hit 
        
      }//hits loop
    }//TreeH loop
  }//events loop


HMPID full simulation-reconstruction sequence

hits creation:
        HMPID uses 2 types fo hits: MIP hit and photon hit. Both types are implemented in the same class AliHMPIDHit.
        The difference in ctor patterns is only in energy variable: for photon it is full energy, whereas for MIP it's energy lost in amplification gap gas.
        Hit for photon is created when photon enters the volume representing a single pad of segmeneted photocathode and survives QE test.
        Hit for MIP is created when MIP exits amplification gap (or disappired for whatever reason). The responsible code is AliHMPIDv1::StepManager().
        Internally in ctor, the energy is converted to the total charge accamulated for this hit expressed in QCD channels. This is done in a honest manner for photon:
        photon always produces a single electron, and the response of the chamber to a single electron pulse is parametrized out of test beam data for few HV sets.
        For MIP the same procedure is generally wrong: the total energy lost by particle is devided by ionization potential, this value is interpreted as number of electrons
        created, then each electon contribution is sampled according to puasonian distribution with the same single electron pulse mean as fro photons. The final charge is
        then a sum of all single electon's contributions, it substitutes the value of energy.
hits->sdigit:
	Due to segmented photocathode, single hit might affect few neighboring pads (up to 9 in our default parametrization). The total charge collected by a single pad
        is the answer to electrostatic problem solved in article by Mathieson (see ref ???). Actual disintegration is implemented in AliHMPIDHit::Hit2Sdi().
        The implementation creates by default sdigits only for closest neighbours of a pad containing hit positon (further neighbours might also be checked subject to
        parameterization flag ?????? but the contribution to them from single hit is such a tiny that only large nubmer of close hits may provide something not negligable). 
sdigits->digits:
	The necessety of sdigits is dictated by the fact that transport engine transports tracks in a continuous sequence track by track. 
	It means that it may happen that the same pad is affected by few tracks. But this might be known only after the transport of full event is finished. 
	So the task of this convertion is to collect all the sdigits for the same pad and sum them up. This is done in AliHMPIDDigitizer::Exec(). This technique also provides
        the possibility to mix events from seperate simulations facilitating studies of rare signals on top of huge background.
digits->clusters
	The aim of this conversion is to try to reconstruct the initial position of hits.
        This it done by 2 steps procedure. On first step so called raw cluster is formed as a composition of all neighboring pads (diagonal pads are not allowed).
        Then center of gravity of the cluster is calculated and used as a naive estimate of hit position. 
        On second step, the code tries to improve the hit position  knowledge by fitting by local maxima number of Mathieson shapes. If the procedure failes, the cog
        position is preserved as hit position.
clusters+tracks->theta cerenkov 

	
	
	
How to get correct magnetic field:
	mag field is needed for simulation as well as reconstruction
        






Generalized structure of AliReconstruction:

Run()
{
  if(there is galice.root)                                        <-|
    AliRunLoader::Open(....)                                        | 
  else                                                              | this is done in InitRunLoader()
    if(raw data process requested)                                  |
      create galice.root on the base of AliRawReader::NextEvent   <-|   
  
  for(all detectors){                                             <-|
   if(detector not selected to run) skip this detector              | this is done in RunLocalReconstruction()
   reconstructor=get detector's reconstructor                       | 
                                                                    |   
   if(detector HasLocalReconstruction) skip this detector           | IMPORTANT! if HasLocalReconstruction() returns YES use RunLocalEventReconstruction instead
   if(run upon raw data)                                            |  
     reconstructor->Reconstruct(fRunLoader, fRawReader);            |   
   else                                                             | <- this approach is currently used by HMPID as all branches are mounted in AliHMPID.cxx
     reconstructor->Reconstruct(fRunLoader);                        |
  }                                                               <-|  

  for(all events){                                                      
  
    for(all detectors){                                                 |            
      if(detector not selected to run) skip this detector               |
      reconstructor=get detector's reconstructor                        | 
      loader=get detector's loader                                      | this is done in RunLocalEventReconstruction()
      if(raw data process requested and detector HasDigitConversion){   | 
        loader->LoadDigits("update");                                   | open file and invoke  detector->SetTreeAddress();
        loader->CleanDigits();                                          |   
        loader->MakeDigitsContainer();                                  | create tree
        reconstructor->Reconstruct(fRawReader,loader->TreeD());         | expected to fill TreeD out of raw reader
        loader->WriteDigits("overwrite");                               | 
        loader->UnloadDigits();                                         |  
      }                                                                 |
      if(detector do not HasLocalReconstruction) skip this detector     | IMPORTANT! assumed that this detector is already processed in RunLocalReconstruction()
      loader->LoadRecPoints("update");                                  |
      loader->CleanRecPoints();                                         |
      loader->MakeRecPointsContainer();                                 | 
      if(fRawReader && reconstructor do not HasDigitConversion()){      | 
        reconstructor->Reconstruct(fRawReader, loader->TreeR());        | expected to fill TreeR out of raw reader
      }else{                                                            |
        loader->LoadDigits("read");                                     |
        reconstructor->Reconstruct(loader->TreeD(),loader->TreeR());    | the only operations inside are pDigTree->GetEntry(0) and pCluTree->Fill();
        loader->UnloadDigits();                                         | 
      }                                                                 |
      loader->WriteRecPoints("OVERWRITE");                              |
      loader->UnloadRecPoints();                                        |
    }//detectors loop                                                   |
    
  }//events loop 
}


HMPID calibration and alignment.

Abstract
HMPID calibration and alignment strategy is described with emphasis put on those aspects of the procedure which are relevant for reconstruction and thus the final detector
figure of merit. In particular, the refractive index calibration technique based on mass plot shifts analysis and chamber alignment with respect to core detectors 
are explained in details. External sources of calibration and alignment data are also mentioned as well as the way HMPID intends to handle those data, including initial CDB
creator. 

Calibration. 
Looking on HMPID chamber structure, full description of which is available elsewhere (ref RichTDR), easy to compile the table of all possible parameters affecting reconstruction.
The first one of major importance is a freon refractive index. Although the full optical path visible by photons includes freon vessel, proximity and amplification gaps filled 
with methane and quartz window separating above mentioned volumes, only freon refractive index is subject for calibration.  Refractive index of SiO2 window is not practically 
affected by any external parameters, while influence of methane temperature to it's refractive index is negligible. So it's enough to measure there optical curves just once.
In the rest, the only changeable parameter is refractive index of freon. Temperature influence on freon refractive index was measured experimentally. The parametrization 
found to be:
  n=n0-0.0005(T-20)                        where T is freon temperature in degrees Celsius
  n0=Sqrt(1+ 0.554*lamda^2/(lamda^2-5769)) where lamda is photon wavelength in nm taken at 20 degrees Celsius
Preliminary, the parametrization of refractive index versus temperature and photon energy is considered to be permanent. 
As the reconstruction itself is only interested in mean refractive index Nmean C6F14, the most elegant solution is to store in OCDB this value, prcalculated in 
AliHMPIDPrecprocessor. 2 measurements of temperature is avaiable from DCS: for inlet and outlet. They come in form of TObjArray of AliDCSValue, where AliDCSValue
holds the value of temperature plus a time stamp at wich the value was taken. Due to organization of DCS, it's not possible to implay that all the points are taken at the same 
time, hence marked with the same time stamp. So the mean temperature are not calculable from inlet-outlet pair. Instead each sperate temperature data points are fitted.
The resulting functions provide mean temperature function.



In local CDB storage (default directory is $ALICE_ROOT) two versions of freon refractive index are written by external macro RichCdb.C :
Run0_0_v0_s0.root contains DiMauro's parametrization and the temperature is set to 20 degrees. To be used as default for simulation and reconstruction.
Run0_0_v0_s1.root contains DiMauro's parametrization and the temperature is set to 50 degrees. To be used in special uncalibrated reconstruction to test calibration procedure. 
Both of them are valid in run range from run number 0 to run number 0, thus in no way affecting any normal operations.   

Refractive index of freon (C6F14) is taken in AliHMPIDRecon for 3 different photon energies by means of 2 methods: Set



Alignment.
Information about detector position and orientation is needed during reconstruction phase. This information affects track-cluster matching procedure, the relevant peace of 
code comes to AliHMPIDTracker::PropogateBack(). Matching procedure consists in prolongation of the track reconstructed in core detectors up to each HMPID chamber plane in 
a sequence. The plane used is the entrance to HMPID radiators. If the intersection exists and inside the sensitive area, the point of intersection is to be transformed to HMPID
local reference system. Note, that in this check, the dead zones in-between radiators are not taken into account. This operation requiring MARS to LORS transformations is done
in AliHMPIDHelix::RichIntersection(). Plane to be intersected is defined by a point belonging to that plane served by AliHMPIDParam::Center(ChamberNumber) and a vector normal 
to the plane served by AliHMPIDParam::Norm(ChamberNumber). Transformations itself are done in AliHMPIDParam::Mars2Lors() and AliHMPIDParam::Lors2Mars(). Internaly in AliHMPIDParam, 
each chamber is represented by TGeoHMatrix. It's worth to stress again that geometry related operations are needed to be done for 3 different planes per chamber, namely entrance 
to radiator, anode wires plane and photocathode plane. So AliHMPIDParam sustains 7*3=21 planes. Also important to say, that direct usage of TGeoHMatrix::MasterToLocal() 
and vice versa is not possible due to special nature of HMPID LORS. According to the decision made about 3 years ago, HMPID local reference system is centered in low left
hand corner of the chamber if one looks from outside to direction pointing to intersection point.
So the most obvious candidate for alignable objects to be stored are these 21 TGeoHMatrix objects. 
The approach suggested in AliAlignObj is not quite feasible mainly due to the fact it relays on incrementing procedure using import from geometry.root. HMPID geometry is defined 
in a way that there is no volumes exactly corresponding to the HMPID planes. 

Geometry of HMPID chambers.
After the decision to rotate the whole HMPID setup from 12 o'clock position to 2 o'clock position we have the following situation:

Theta = 109.5 degrees for chambers 1,3
Theta =  90.0 degrees for chambers 2,4,6
Theta =  70.5 degrees for chambers 5,7

Phi   =  50.0 degrees for chambers 6,7 
Phi   =  30.0 degrees for chambers 3,4,5 
Phi   =  10.0 degrees for chambers 1,2 


Old parametrization by AliHMPIDChamber:
HMPID chamber 1 (454.877118 ,  80.207109 , -163.565361)(rho,theta,phi)=(490.0,109.5,10.0)
HMPID chamber 2 (482.555799 ,  85.087607 ,    0.000000)(rho,theta,phi)=(490.0, 90.0,10.0)
HMPID chamber 3 (400.012224 , 230.947165 , -163.565361)(rho,theta,phi)=(490.0,109.5,30.0)
HMPID chamber 4 (424.352448 , 245.000000 ,    0.000000)(rho,theta,phi)=(490.0, 90.0,30.0)
HMPID chamber 5 (400.012224 , 230.947165 ,  163.565361)(rho,theta,phi)=(490.0, 70.5,30.0)
HMPID chamber 6 (314.965929 , 375.361777 ,    0.000000)(rho,theta,phi)=(490.0, 90.0,50.0)
HMPID chamber 7 (296.899953 , 353.831585 ,  163.565361)(rho,theta,phi)=(490.0, 70.5,50.0)

New parametrization by TGeoHMatrix:  perfect geometry, no misalignment
HMPID 0
 -0.328736   -0.173648    0.928321    Tx = 454.877118
 -0.057965    0.984808    0.163688    Ty =  80.207109
 -0.942641    0.000000   -0.333807    Tz = -163.565361
HMPID 1
  0.000000   -0.173648    0.984808    Tx = 482.555799
  0.000000    0.984808    0.173648    Ty =  85.087607
 -1.000000    0.000000    0.000000    Tz =   0.000000
HMPID 2
 -0.289085   -0.500000    0.816351    Tx = 400.012224
 -0.166903    0.866025    0.471321    Ty = 230.947165
 -0.942641    0.000000   -0.333807    Tz = -163.565361
HMPID 3
  0.000000   -0.500000    0.866025    Tx = 424.352448
  0.000000    0.866025    0.500000    Ty = 245.000000
 -1.000000    0.000000    0.000000    Tz =   0.000000
HMPID 4
  0.289085   -0.500000    0.816351    Tx = 400.012224
  0.166903    0.866025    0.471321    Ty = 230.947165
 -0.942641    0.000000    0.333807    Tz = 163.565361
HMPID 5
  0.000000   -0.766044    0.642788    Tx = 314.965929
  0.000000    0.642788    0.766044    Ty = 375.361777
 -1.000000    0.000000    0.000000    Tz =   0.000000
HMPID 6
  0.214567   -0.766044    0.605918    Tx = 296.899953
  0.255711    0.642788    0.722105    Ty = 353.831585
 -0.942641    0.000000    0.333807    Tz = 163.565361



Manual:
Diclaimer: We call the operator any persons who wants "to operate" HMPID that is to do something reasonalbe to understand current perfomance coming from the detector. 
This person is not necesarelly an expert in HMPID hardware and or software. 
Be also aware that analysis of HMPID data can only be perfomed from AliESD (taking also into acount some info which is not specific to HMPID). The task to plot something from
hits, digits or clusters is called quality assesment.

From operator point of view, one needs to do 3 different tasks with HMPID: simulation, reconstruction and QA.