NOTES ON WORKING MEETING ON MC PRODUCTION, MM 24-Sep-02
	=======================================================

	My slides from the meeting are available on the offline 
	working group web page:

	http://www.lnf.infn.it/kloe/private/mc/Presentations

	These notes briefly summarize discussion related to 
	those slides, by title.



	BVLAB scan of 2001 data
	-----------------------

	Achim showed some histograms produced by the BVLAB module.
	Two salient points were evident:

	- The various indices of background levels in the data are
	  very highly correlated. Quantitites such as the number of 
	  hits in gamma-gamma events, the number of unassociated 
	  hits in Bhabha events, and the number of accidental clusters
	  track each other remarkably well as a function of run number.

	- There is a significant, constant difference in the number of 
	  hits in gamma-gamma events and the number of unassociated 
	  (to any track) hits in Bhabha events. Clearly, the unassociated 
	  hits in Bhabha events are causually related to the tracks. 
	  Possible explanations include: a hit-association efficiency 
	  which is unusually low (missed track segments might be 
	  a possibility), and cross-talk in the chamber and/or readout
	  electronics. This needs more investigation.    


	Background tools
	----------------

	There are serious concerns about how well hits causually related 
	to tracks can be isolated. MBCKADD can selects Bhabha events for 
	insertion, as well as gamma-gamma or eta-gamma -> 3 gamma events.
	The concerns about crosstalk, etc. discussed above make it 
	clear that the Bhabha sample poses additional problems.
	ACCELE does not use this sample, and the idea is to not use
	this sample when the modules are unified.

	Neither module at the moment is capable of flagging the inserted
	hits as added background, i.e., in the equivalent of MC truth. 
	This capability will need to be added. 


	Conclusions from last meeting
	-----------------------------

	These conclusions were unanimously accepted



	Event-selection issues
	---------------------- 

	Once the decision has been made to not use the Bhabha sample
	in MBCKADD, the event selection criteria in the two modules 
	are (or can be) perfectly reconciled. Thus, essentially 
	all phases of the insertion analysis can be unified.


	Quality-control issues
	----------------------

	MC and data geometry is in fact slightly different, in 
	particular for the EmC (the description of the DC is 
	somewhat more accurate).

	Since the physical event topology is described in the MC, 
	it is clearly the MC geometry that must be used for 
	reconstruction. This raises the possibility that inserted 
	tracklets may not be reconstructed as they were in the 
	data set from which they were sampled. This raises the
	possibility of splits and false vertices and other pathologies
	in the combined event that are not present in data. 
	Among those present, however, this problem was not widely 
	regarded as significant; the feeling is that the effect is 
	too small to worry about. The separate utility of a 
	comprehensive review and resolution of discrepancies between 
	MC and data geometries was noted, however.

	Everyone agress that the macroscopic (distribution of 
	background parameters in combined data) and microscopic 
	(quality of tracks in combined data) quality tests suggested
	are useful and important; clearly, the study of the 
	quality of the combined data will be one of the more labor-
	intensive aspects of this enterprise and will require 
	collaboration from many people.


	Database and production issues
	------------------------------

	While no firm commitment was made on global vs. local 
	sampling, it was realized that there is a continuum between 
	the sampling techniques described. For local sampling, 
	at the very least, for each run group, the sampling should 
	be done as described for global sampling.


	Logistical considerations
	-------------------------

	The idea that the first reconstruction should be done 
	without background insertion may be valid. However,
	it may not be strictly necessary. To the extent that the 
	element banks written by the MC are not the banks which 
	are merged (MCEL, MDTC), keeping these banks in the 
	output data should in principle allow the removal of 
	background from an event into which background has 
	already been inserted once, and the introduction of 
	new background into that event, by nothing more complex
	than an input bank drop.


	Other issues discussed (not from slides)
	========================================

	Modules the purpose of which is to improve the quality 
	of the simulation by means other than the addition of 
	background can be added to the MC reconstruction executable
	along with the background insertion module as well. 

	In particular, Miscetti's tools for the simulation of 
	the calorimeter threshold and non-linearities at low 
	energy can and will be included.