Activities
Fast Reconstruction Algorithm for Dimuon Spectrometer of ALICE
The Dimuon Spectromrter of ALICE is designed to study the heavy quark resonances produced in high energy heavy ion collision at LHC. A typical data rate of ~500 MB/s will be generated from Pb-Pb collision in forward rapidity region. As a first step a L0 trigger will be applied to the data flowing to reduce it to 1 kHz. But this trigger cut is insufficient to produce a sharp PT cut over the particles that hit the detectors, which results the contamination of the data with huge number of background events.Dimuon High Level Triggers (DHLT) is a software cut on the data after L0 trigger to eliminate 80% of the background event to bring down the data rate to ~100 MB/s. This will be achieved by reconstructing the Muon tracks online for 4 and 5 of tracking station .
The job is divided in two parts one is the finding of the reconstructed hits and next is to draw the tracks through the hits. Saha Institute of Nuclear Physics is responsible for hit reconstruction and University of Capetown for track reconstruction and its implementation in Pub/Sub.
AliRoot (ALICE software for simulation) studies have been performed to estimate the average number of particle hits in central Pb-Pb event. It is found to be approximately 150 hits per chamber in the detectors of 4th and 5th tracking station, with a cluster size of 3 to 4 pads. 
Thus, a single event corresponds to ~500 pads firing in each chamber.
Our goal was to find a algorithm which can find out the 150 particle hits out of 500 pad hits within a time limit of ~1 ms. But the conventional scheme for clustering algorithm is too slow achieve such time limit. An algorithm was developed in SINP which identifies each pad cluster by its central pad (pad with maximum ADC value), and it is assumed that the number of geant hits it equal to number of central pads. The hit reconstruction is done in three steps first the chamber is scanned slat-wise to collect the indices of the central pads of each cluster and it is stored in two arrays one for bending and another for non bending plane. Next we apply COG method over three pads in y-direction for bending (x direction for non-bending) plane to find out the hit position with good resolution in bending (non-bending) direction. Finally, the bending and non-bending hits are merged to find out reconstructed x and y position.