1. M.Biglietti (Univ. Naples and INFN Naples)
Muon Slice Plans
M.Biglietti (Univ. Naples and INFN Naples)
for
the Muon HLT Communiy
Pesa Meeting 9/07/2003

2. LVL1 (barrel)
Code completed: only detailed Sector Logic needs a second iteration, but it is not essential: simplified version is adequate
Physics validation very important now; check the trigger efficiency curves for pt threshold = 6 and 20 GeV

3. Muon HLT- on going Continue the work still on going with:
Store the muFast output as Trigger Element to allow Moore seeding
using the seeding mechanism of the Pesa Core Software (G. Comune)
Physics performances of muFast and Moore with single muons: check momentum resolution and trigger efficiencies for some points of PTs; use for muFast “old” LUTs (no feet regions)
Check the reconstruction robustness and timing with single muons plus cavern background
Debug and update -(e.g. TrigMoore updated to work with Compact Identifiers and IDHelpers as for the latest offline versions. Offline/Online versions fully aligned)

4. Muon HLT – possible near terms targets
Evaluation of the LVL1-LVL2-EF single muons trigger rates using the full chain
Combined reconstruction
muComb postponed for the moment (no time/man power)
Combination at the EF stage using Moore and iPatRec tracks MuIDComb: hopefully will be done.
Evaluate the physics degradation and the probability of reconstructing fakes muons as a function of background intensity
Very preliminary studies in the full offline environment with Moore/MUID on the pileup and fakes from K/p done
Implement the HLT chain in the testbed
First successfully implementation of muFast done
Full Moore online implementation (postponed for the moment)
needs RPCs “prepared RDO” – bytestream converters
... If no big technical problem (software stability)
and enough time/man power

5. Muon HLT : long term effort
Caculation of muFast LUTs for layout P03
Extension of muFast to the endcap
Needs LVL1 endcap simulation
muComb and its extension to the endcap
Trigger rates studies (including dimuons) in the full rapidity regions and with different cavern background intensities
Muon isolation
Muons with Tilecal

6. mFast physic performances for pt = 20 GeV
Implementation of the TP code in the HLT software
Physics validation : same results of the TP
Perfect match!
HLT TP, layout M
HLT TDR, layout P

7. mFast algorithm timing
Little effect of the cavern background
(L=1034cm-2s-1, safety factor 2) on mFast

8. Total Elapsed time Most of the cpu time is needed by the
MDT converter.
The Cavern Background rises the
MDT converter time about a factor
of 4
To be coungruent with the the TP
background estimations we should
boost the MDT occupancy at least of a factor of 5. This raises the MDT converter time from 8 ms up to 40 ms

10. Moore/MUID
TrigMoore: Implementation of MOORE/MuID in the framework of High Level Trigger at the Event Filter stage.
Wrapped Strategy- Allows to execute the offline reconstruction chain from the online environment, reconstructing the event by accessing at all the Spectrometer data. The implemented chain is composed by Moore and MuidStandAlone.
Seeded Strategy – Uses the hypotesys formed at the previous stage in the trigger process, accessing only to subdetectors data that are of interest for the selected trigger regions. (By now only regions selected by the LVL1 trigger have been used).
When we speak about Moore in the High Level Trigger, we essentially speaks about 3 packages.
1- Moore- is an offline package performing the reconstruction in the Muon Spectrometer. It is able to reconstruct in the full eta range
(barrel and endcaps are included). It considers the MuonSpectrometer as a standalone detector (i.e. the tracks reconstructed by Moore are expressed at the first measured point inside the Spectrometer.
2- MuonIdentification- is an offline package dealing with reconstruction and Identification of Muons. It is divided in two components:
MuidStandalone - where the tracks reconstructed by Moore are extrapolated to the interaction point taking into account the energy loss and the
multiplo scattering in the Calorimeters.
MuidComb – where the Moore-tracks are associated with tracks reconstructed in the inner detector (iPatRec), and a combined track is
reconstructed
Note-MuonIdentification always performs a refit.
3 – TrigMoore – is the implementation of Moore/MuID in the framework of High Level Trigger at the Event Filter stage.
Two main strategies have been implemented:
Wrapped – Allows to execute the offline reconstruction chain in the online environment (only Moore+MuidStandalone has been
implemented.) It allows to reconstruct the event in his completeness accessing all the Spectrometer data.
Seeded – uses the Trigger hypothesys from the previous stages (Region Of Interest of the Trigger).

11. Moore/MUID timing tests
Atlas pcatr machine (2.4GHz)
OPT build
Timing package: TrigTools/TrigTime
Strategies Seeded/wrapped
Conservative approach (i.e. data access and data preparation are included)
Timing test performed on release.
(The use of Compact Identifiers from version can influence the timing)
Seeded: Access only the RoI from the LVL1
Wrapped, Full event
Reconstruction
DC1 single Muons execution times
pT (GeV/c)
<T> (msec)
rms (msec) 8 73 30 68 20 59 15 58 21 50 61 19 25
100 75 64 26
300 23 32
100+X1
763 37
2680
450
100+X2
1218
5900
1100
In the table the average and rms are calculated on the events left after rejecting 5% of the events included in the high execution time tail
The timing in the graph include everything and show that very few is spent in the
extrapolation to the vertex. The timing in the table are on the 95 % of the event.
The timings are well below the 1 sec requested to process at the EF stage.
Single Muons + Pile up

12. Moore timing efficiency
A time efficiency has been defined as the ratio between the number of reconstructed
tracks in one second and the
total number of reconstructed tracks
Average on events with time<1sec
Mean time almost constant over Pt
Wrapped mode times rather similar
Time efficiency.

13. Moore Performances Wrapped strategy |h| < 1.05
Efficiency for the reconstruction of single Muons sample in wrapped strategy (event scan). This is equivalent to the execution of the offline version of the code. The analysis has been restricted to the barrel region in order to be directly compared with the seeded strategy. (LVL1/LVL2 restricted to the barrel)
Resolution in wrapped and seeded strategy.
Strategy wrapped |h| < 1.05
Strategy seeded |h| < 1.05
Here we have resolution for seeded and wrapped mode, and efficiency for wrapped mode.
In the resolution since there is an ad hoc implementation only for the seeding (search in a limited region and then access to the data) we expect to reproduce the result reported in the offline. The difference on the point at 8GeV/c is related to the trigger threshold used. In this case we have analyzed samples passing the trigger with a configuration threshold of 10 GeV/c therefore we are on a queue.