This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Slide 1
Digital Filtering Performance in the ATLAS Level-1 Calorimeter
Trigger David Hadley on behalf of the ATLAS Collaboration
Slide 2
ATLAS Trigger and L1Calo Architecture Digital Filter
Implementation Digital Filter Performance Outline
Slide 3
ATLAS Trigger and L1Calo Architecture
Slide 4
ATLAS Trigger ATLAS three level trigger: Level 1 :
hardware-based, pipelined with a fixed latency
Noise Rejection Noise PDF for a single tower (left). The total
probability of noise output > 0.5GeV for all towers in the LAr
EM Barrel (right). Improved noise rejection with a matched/common
filter but little difference between them. 26/05/2010Digital Filter
Performance - David Hadley - RT201016
Slide 17
Energy Resolution from Detector Noise Energy residual
(measured-simulated) for a 25GeV pulse (left). Gaussian width of
energy residual is plotted for all towers in the LAr EM barrel
(right). Again, little difference between Matched and Common. Clear
improvement over Pass-through. 26/05/2010Digital Filter Performance
- David Hadley - RT201017
Slide 18
Optimising Usage of the Look-up Table Range Using a peak
coefficient of 15 gives the best precision for matching the filter
to the pulse shape. Precisely matching to individual towers pulse
shape does not significantly improve performance. We can change the
peak coefficient to optimise the usage of the look-up table (LUT)
range. 26/05/2010Digital Filter Performance - David Hadley -
RT201018
Slide 19
Choice of Filter Coefficients for Early LHC Running Based on
these studies the strategy decided for 2009- 2010 running: Common
filter applied across entire calorimeter layer, matched to the
average pulse shape in that layer. Coefficients scaled to optimise
usage of the LUT range. 26/05/2010Digital Filter Performance -
David Hadley - RT201019 RegionFilter EM Layer(1,8,13,10,7) Hadronic
Layer(1,9,15,11,5) FCal(0,2,13,5,0)
Slide 20
Look-up Table E T Calibration Analogue gains are applied to
scale the input pulses to 4 ADC per GeV. LUT table applies pedestal
subtraction, noise cuts and final E T calibration. At present, only
scale filter output back to the input scale. Initial calibration of
the LUT slopes were based on calibration pulses. Preliminary checks
of correlation show calibration is working well for collisions.
Nominal 4 ADC input to 1 LUT output, expect a gradient .
26/05/2010Digital Filter Performance - David Hadley - RT201020
Slide 21
Summary L1Calo Pre-processor digital filter shows good
performance in selecting the correct bunch-crossing, rejecting
noise, and improving energy resolution. The Common digital filter
configuration was implemented for 2009-2010 running. Initial cross
checks of LUT calibration with collision data are encouraging. The
process of understanding and optimising the trigger with collisions
is continuing. We are looking forward to new physics events
identified with L1Calo! 26/05/2010Digital Filter Performance -
David Hadley - RT201021
Slide 22
Backup Slides 26/05/2010Digital Filter Performance - David
Hadley - RT201022
Slide 23
Performance with Different Peak Coefficients BCID efficiency:
Energy resolution: Noise probability:
Slide 24
L1Calo Trigger (in full) Sliding-window algorithms search for
high-E T objects. In reality windows are processed in parallel.
Central cluster > threshold. Hadronic and e.m. isolation
requirements in surrounding towers. e/gamma and /hadronJet and
energy-sum ECAL+HCAL Operates on jet elements 2x2 towers. Energy in
window (EM+Had) > threshold. Variable window size. Same module
does total-E T and missing-E T triggers. Double counting is avoided
by requiring a local E T maximum. 26/05/201024Digital Filter
Performance - David Hadley - RT2010