effghost - CERN Indico

Ghost Identification
M. Needham
EPFL
Outline
• Embedding data
• Matching study
• 2/dof study
Embedding Data or MC
Can study ghosts and detector efficiency by embedding data in data [or MC]
• This procedure can be useful for other things [e.g. measuring spillover effects]
• For this to be work:
• Mechanism to merge two file streams [re-use Boole spillover mechanism ?]
• Code to merge the streams [exists for ST and Velo, miss OT]
• Independent way to identify good tracks in one stream:
• By eye ? Standard reconstruction on clean events with wide windows ?
• Understanding how efficiency/ghost rates scale with occupancy from MC
• Tracks gained in merging - ghosts, Losses: inefficiency
• Even if we cannot extract absolute numbers can make relative MC-data
comparisions
All this exists, just a case of plugging it together
Merging Scheme I
• Re-use the TAE merging code
• Simple to do, code exists + well tested
• But:
• Don’t use all information [e.g. neighbour sum]
• For overlapping clusters algorithm selects only the best
Merging Scheme II
• Algorithms exist in ST/Velo flavours
• Breaking in digits is easy, merging uses all info [including neighbour sum]
• Re-clustering: as in Boole
• All info is in conditions database and is consistant with the data
Matching Study in MC
LHCb-2007-020
• Yield versus matching cut
• Yield versus # candidate long tracks per seed
2/dof Study
Ks
J/