NOTES ON ORGANIZATIONAL MEETING ON MC PRODUCTION, 23-Jul-02
	===========================================================


	Topics for discussion at the meeting:
	-------------------------------------

	1. What can we do to better simulation of detector response?

	   * Background

	   * Resolutions (DC)

	   * EmC simulation (microscopic level?)


	2. What tools are available for introducing MB in MC events?

	   * ACCELE (EmC)

	   * MBCKADD (DC)

	   Do sufficient reference samples of MB to insert exist?
	   How hard are they to obtain?
	   
	   What other tools are needed?


	3. How do we organize MC event production with background?

	Volunteers ?!?


	Brief meeting summary
	---------------------

	Most discussion focused on the DC, in part because most 
	of the people who were present are more centrally involved 
	in DC-based analyses.

	Points 1 and 2 above were discussed. Point 3 above was not 
	discussed both for lack off time, and because intelligent 
	solutions depend on the implementation of suggestions 
	concerning point 2.



	1) What can we do to better simulation of detector response
	----------------------------------------------------------

	Discussion focused on the DC.

	People involved (Luca, Paolo V.) feel that the factors 
	contributing to differences in momentum resolutions on the 
	order of 10--20% and residuals on the order of 0.1% between 
	MC and data may have multiple causes that are not well 
	understood. They suggest an incremental approach and urge 
	avoidance of quick fixes that may not be based on a sound 
	understanding of the causes of the discrepancies observed.

	In particular, any of the following factors alone or
	in combination could be responsible for the observed 
	discrepancies:

	- Coherent misrepresentation of detector geometry
	  (misalignment, rotation, etc.)

	- Incoherent misrepresentation of detector geometry
	  (MC events are reconstructed with exact knowledge 
	  of positions of detector elements a priori)

	- Misrepresentation of the magnetic field

	- Misrepresentation of the space-time relations

	Acting on the field map alone for example might not be 
	a good idea if a significant root cause is geometry.

	The most probable starting point for study of the detector
	geometry is the representation of the wire sag in the MC
	and in the reconstruction of MC events. Luca and Mario 
	are looking into this.

	Patrizia, after having been informed by Kulikov that 
	drift-distance errors for MC events seem to be overestimated, 
	has traced the problem to a bug in DCONVR. For MC events,
	the errors for the raw s-t relations were being used in place 
	of those for the fine s-t relations. The bug has been 
	fixed and will soon be installed in the TRK library.


	2) What tools are available for introducing MB in MC events?
	------------------------------------------------------------

	Most discussion focused on MBCKADD, but there was significant 
	discussion about the need for convergence between MBCKADD and
	ACCELE. The issues are easier to explain with an understanding 
	of, say, MBCKADD in hand.

	MBCKADD consists of two parts. The first part is a "selector" 
	which identifies gamma-gamma, Bhabha, or eta-gamma -> 3 gamma 
	events and identifies the hits in these events that are not 
	associated with the physics topology. In the case of gamma-gamma 
	or eta-gamma events this means all hits. In the case of Bhabha
	events, this means hits not associated with the Bhabha track.
	These hits are currently stored in a formatted text file.
	Typical file space is 200 bytes/event.
	The second part is MBCKADD proper, which inserts the background 
	hits into MC events during the reconstruction phase.

	In its current form, it would be possible to run the MBCKADD
	selector on all 2001 data, so as to generate a reservoir of 
	background which has a time profile which perfectly represents
	the data. 1 million events would take up 200 MB of file space, 
	which is acceptable. In principle, it should be possible to 
	insert the same background in multiple MC events, so 1 million 
	background events is sufficient. The selector can run on the 
	L3BHA stream, for example, and is thus fairly fast.

	However, it seems more efficient to combine the background 
	harvesting for MBCKADD and ACCELE. 

	In principle, it would also be desirable to output the 
	background events from the two selectors in an identical 
	file format. The standard file format for the experiment is 
	compressed YBOS.

	On a related note, it would be desirable to include both 
	calorimeter and DC background when inserting background
	into MC events, and unifying the format of the output from 
	the selectors for MBCKADD and ACCELE would seem to go in this
	direction.

	There are some technical points to address. We need to 
	investigate the compatibility of the event selection 
	criteria used to identify events with isolable background 
	in MBCKADD and ACCELE, for example.

	The consensus, however, was that an executable with both 
	the MBCKADD and ACCELE event selectors, which writes output
	from both into a unique background file written in YBOS 
	format, is a desirable goal.

	The following people (were) volunteered to discuss the 
	problem in more detail, probably in early-to-mid September:
	S. Miscetti, M. Moulson, V. Patera, A. Ventura.