DØ Level 3
Algorithms Meeting: 13th November 2002 in the Farside
Talks
-
Tracking
studies: Daniel Whiteson
-
Reminder of results presented before from Z->ee: axial 80% efficiency; becomes 95%
if 7 or 8 CFT axial hits are present within road around calorimeter cluster.
-
New study:
comparison between L3 and offline global tracks.
-
Clear match between L3 and offline global tracks in eta and phi
-
93% efficiency to find L3 track if offline track with 7 or 8 CFT axial
hits present.
N.B. events are possibly biassed by L3 track trigger!
-
Reasons for lost tracks in cases where 7/8 cft clusters are present
will be investigated further:
-
large dca of track after a few layers
-
large chisuared
-
clusters in first two layers (8 and 7) don't point near to hit in
third layer
-
currently possible to rank 7cft+2smt track higher than 8+1 track, but
the 7+2 track is currently later thrown away because 7+3 are required
- inconsistent criterion - will be fixed
-
No correlation between event multiplicity and L3 track efficiency.
What about tracks in jets?
-
Stereo finding uses a histogram method:
allows one miss only in total (cft+smt).
8% inconsistency found between stereo parameters of L3 and offline tracks which
match in r-phi. Not yet clear what fraction of these are correct in
L3 and what fraction are correct in offline.
-
What threshold in cft are we using? Follow up from after the meeting:
nominal threshold+20 channels in both L3 and offline.
Offline will be reconsidering this and we should keep in touch with
them on this.
-
smt matching efficiency, especially stereo still needs to be investigated.
-
MuoLocal and MuoCentralMatch status: Martin Wegner
Martin presented substantial progress in two areas:
-
L2/L3 muon comparison:
-
Efficiency of L2 muon in doubt.
-
Special run: taken with L2L0, so some small bias wrt. L2.
-
Events reprocessed with trigsim p12.04. Efficiency wrt offline medium muons: 95% L3, 85% L2.
L2 efficiency shows dips at |eta|=1.
No such dips observed in L3.
(It is not until p13 that muons crossing between central and forward
are taken into account in offline/L3.)
-
Origin of L3/offline differences still need to be investigated? Need
to repeat test of using identical parameters in offline/L3 to see if
we get identical
results in p13?
-
"Purity" defined as fraction of events with a trigger level muon that
also have an offline loose muon. This is higher for L3 than L2.
-
Ideally, it would be very useful to have a high enough statistics
sample of muons identified in the calorimeter for unbiassed
measurements.
However, for certification purposes demonstrating that L3 muon does
sensible things on events containing offline muons really should be adequate!
-
L3 central match:
-
Measure efficiency of L3 central match wrt events in which there is an
offline track-muon match:
-
About 11% of offline muons have L3 match failures - they are almost are all at |eta| = 1.
-
These are
mainly bad direction measurement in eta - are these good muons ?
Probably, but should
study delta_phi match for these muons to confirm.
-
Actually,
we start with the approximately the same central tracks and local
muons found in both L3 and offline. Using the
same matching gives same results! Therefore, at least as a temporary
measure we should loosen matching cuts. We then need to see
measurements of
rejection vs matching cuts.
-
With loosened matching cuts the losses are reduced to
only about 1% where there is no L3 global track found
+ a few percent that have pt<1.5 GeV for the L3 track.
-
Rejection (see plot): Compared with the rejection obtained when an L3 global
track is found, we get about a factor 1.5 (2.0) relative to L2M0 (L2L0) by requiring this track to
be matched to a muon.
-
For pt > 5 GeV the
turn on curve of a pt cut at L3 wrt offline pt is rather slow?
Daniel will do some studies comparing pt(offline) and pt(L3).
Do we need to refit the tracks at L3?
-
Online-Offline Comparator studies:
Han Do
A new comparator function has been written that takes
a tolerance as an .rcp input.
This has been used so far to compare the physics_results
for the tau tool online vs. offline.
Suggestions that arose from the audience:
-
It would probably make sense to use the fractional difference between quantities
rather than simply the difference.
-
It would be good to get plots of fractional difference for each of the
relevant quantities, so that we can decide whether one overall
tolerance is OK or we need to define tolerances for each variable individually.
-
p14
plans - opportunities for new people to get involved: Terry Wyatt
Lots of very urgent and interesting work is currently going on to
study the performance of p12/p13. Indeed, we have a number of
important features of p12 that are not yet being used online because
the effect of switching them on still needs to be studied (e.g.,
calorimeter calibrations, primary vertex Et corrections, road method electrons).
However, the "to do" list for p14 looks very worrying at the moment.
There are lots of urgent and interesting projects that have no names
associated to them. Please tell all your friends that we need them to
come and work in L3!
Scribe: Terry Wyatt.