HLT Kalman Filter

1
HLT Kalman Filter

• Implementation of a Kalman Filter in the ALICE
  High Level Trigger.
• Thomas Vik, UiO

2
The ALICE detector
3
Data rates

• Event rates
  -Central Pb-Pb lt 200 Hz -Min. bias
  pp lt 1000 Hz
• Event size (zero suppressed) -PbPb
  75Mbyte -pp
  2.5Mbyte
• Data rates
  -PbPb lt 15 Gbyte/sec -pp
  2.5 Gbyte/sec
• Forseen DAQ bandwidth 1.2
  GByte/sec

4
What is the Kalman Filter

• Iterative data processing algorithm.
• Advantages as a tracker
  -Track finding
  and fitting simultanously
  -No large matrices needs to be inverted.
  Number of computations
  increases linearly with the number of
  measurements.
  Therefore fast in situations with
  multiple scattering and many measurements.
  
  -Estimated track parameters follow closely
  the physical track.

5
Why do HLT need Kalman Filter

• Merge tracks from the different ALICE
  sub-detectors.
• Get good estimates of track parameters.
• Reduce amount of fake tracks from the HLT tracker.

6
Kalman in the TPC

• Track parameters (y, z, D
  , tan ?, C)
• Three steps -Make
  seeds initialize track parameters.
  -Propagate to next padrow. Look
  for clusters in window around predicted
  position. -Update with hit which gives
  minimum ?2.

7
Track following
8
Kalman in the HLT

• Calculate seeds from track parameters of the HLT
  fast tracker.
• Use clusters already assigned to tracks.
• Only needs to propagate tracks according to track
  model, and update parameters with already found
  clusters.
• Gives a fast global tracker.

9
Status

• Timing -HLT tracker
  setup 50 ms/sector
  tracking 70 ms/sector -HLT Kalman Filter
  load tracks 1000 ms tracking
  1000 ms

10
Example heavy charm decay

• TRD triggers on possible dielectron candidates.
• Large background
  pions misidentified as
  electrons electron
  from conversions.
• Merge with TPC tracks, add TPC dE/dx PID
  remove pions misidentified as electrons.
• Merge with ITS tracks, impact parameter cut.
• Background reduction by factor 10.

11
Conclusion

• Merge tracks from the different sub-detectors.
• Need HLT tracking for ITS and TRD.
• Only implemented for the TPC, however some
  debugging is still needed. Must reproduce offline
  results.