INFIERI_poster.pdf
1
04/12/2014
16:55
TRACK FINDING ALGORITHM FOR THE L1 TRACK TRIGGER
OF THE PHASE II CMS EXPERIMENT
D. Cieri, D. Newbold, C. Shepherd, I. Tomalin, M. Pesaresi, G. Hall, A. Rose, A. Tapper, K. Uchida, I. Reid, P. Vichoudis
[email protected]
The Compact Muon Solenoid Detector
The Compact Muon Solenoid (CMS) is a general-purpose
particle physics experiment. It is located at one of the four
interaction points of the Large Hadron Collider (LHC)
at CERN in Geneva.
The goal of the CMS detector is to investigate a wide range
of physics phenomena, such as the search for the Higgs
boson, origin of electroweak symmetry breaking, or the
research of evidence for new physics.
point and it is composed of different subdetectors: the silicon tracker, which measures with high
precision the tracks left by charged particles, the electromagnetic calorimeter (ECAL), for the
detection of electrons and photons, the hadronic calorimeter (HCAL), to characterize hadronic
jets, and finally the muon chambers, to track high energy muons.
Time Multiplexed Track Trigger Proposal
Algorithm Performance
At the moment the architecture that will be used to distribute and process the tracker data is still matter of
discussion. One promising proposal is the use of a time multiplexed approach, similar to the one
implemented for the CMS calorimeter trigger of the phase I trigger upgrade (2015).
The main idea is to send all the trigger data from a given LHC bunch crossing to a single destination
(node) for processing. Processors will not be synchronous but out of phase of one LHC clock in respect
to their neighbor, so that each node will analyze only data from one single event.
The TMT architecture is ideal to be processed using FPGAs (Field Programmable Gate Array). The
. This is a board built
for the CMS upgraded L1 calorimeter trigger, based on Virtex 7 FPGA processor. It has 72 I/O optical links
that operate at 12.5 Gbps (10 Gbps in CMS), for a total bandwidth larger than 0.9 Tbps.
The HT algorithm has been tested using simulated samples of dimuon and
dielectron candidates, with different transverse momentum and pile-up content.
Those datasets have been produced using the proposed CMS phase II geometry, so
that it is possible to simulate with good accuracy the performance of the algorithm.
The efficiency of the Hough transform has been defined in relation to the number of
generated signal tracks:
The µTCA format processing board MP7
The tracker and the calorimeters are contained in a 4T magnetic field, produced by a
superconducting solenoid magnet.
ε=
n. signal tracks found by HT
n. generated signal tracks
HT Eta efficiency for 10 GeV dimuon candidates
Y
CM
Time Multiplexed Architecture
MY
Two hardware stages are required to process the data. The
data are taken directly from the front-end links by the
pre-processors (PP), which will buffer the L1 triggered data
and trigger primitive stream separately.
CY
CMY
K
Stub Filtering
Stub Clustering
Stub ordering
Buffers/time
multiplexing
The CMS Phase II Upgrade
To handle the high luminosity provided by the HL-LHC, the CMS detector will require a new
tracker because of the accumulated radiation damage which will affect the present detector.
One of the most interesting innovation of this upgrade will be the use of the tracker data for
L1 trigger purposes. This is necessary in order to contrain the trigger rate of 0.5-1 MHz with
a latency of up to 10 µs.
The proposed tracker design has a barrel-endcap layout with two different types of double sensor
modules, double strip (2S) and pixel-strip (PS), which would be able to provide a
pT measurement that could be used to reject low energy tracks.
The idea is to compare hits of the upper and lower sensor of a module, and refuse patterns
consistent whith low pT tracks. Combinations consistent with high pT particle are called “stubs”.
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1200
1.8
1000
600
0.2
h_hough
0.18
0.12
2.2
0.1
2.4
0.08
Entries
150
RMS x
RMS y
0.001993
0.04695
Mean x -1.558e-06
Mean y
0.1063
0.16
2.0
2.8
3.0
3.2
200
0
4.0
0
500
1000
1500
2000
2500
z [mm]
Track candidate fitting
Track candidate filtering
To built track candidates an approach using the
Hough transform has been proposed.
Track candidate finding
Stub mapping to global
coordinates
10
8
6
4
2.6
400
Fitted track selection
Physics data are then built, packaged and forwarded to the
data acquisition (DAQ), while the primitive data are formatted,
sorted and transmitted to the main processors (MP). In this
stage track candidates are assembled and forwarded to the
L1 Trigger.
0.06
0.04
2
0.02
0
-0.003
-0.002
min
0
-0.001
0
0.001
efficiency
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
-3
-2
-1
0
1
2
h_eff_reco_phi
Entries
175
Mean
-0.0234
1.811
RMS
1
0
3
eta
-3
-2
-1
0
1
2
3
0.002
is the lower bound of the
0.003
m
0
0
The Hough transform is a technique often used in image analysis
to identify lines in the image. The main idea is to consider a
straight line not as discrete points (x,y), but in terms of the
slope-intercept parameters (m,c). In this way a line in normal
space is converted into a point in the parameter (Hough) space.
The TMT track finding algorithm applies the Hough transform to
the stub coordinates
φ=
±0.006
r + φ0
pT
This translate into a Hough space (m,c) where m = ± 0.006/pT and c is the production
angle 0. So in the (m,c) histogram, for each m a value of c can be calculated.
c = −mr + φ − φ0min
segment under consideration. Then dividing the radial position in 16 bins, it is possible
5 compatible stubs are counted the HT cell is marked for readout.
phi
with respect to the transverse momentum of the candidates and to the pile-up
content.
This last has a crucial importance, since at the HL-LHC the algorithm will handle with
an average pile-up content of 140.
HT efficiency vs. p
HT efficiency vs. PU
T
1
0.95
1
0.95
0.9
0.9
0.85
0.85
0.8
0.8
0.75
0.75
0.7
0.7
Electrons
0.65
0.6
0.65
Muons
10
102
103
p [GeV/c]
0.6
0
Electrons
Muons
20
40
60
80
100
120
140
160
180
T
Hough Transform for a dimuon candidate event (PU=140)
0.14
800
Transmission to GT
Track Finding using Hough Transform
c
r [mm]
0.0
Main Processor (MP)
Pre-Processors (PP)
Stub Formatting
1
efficiency
M
In the coming years the Large Hadron Collider (LHC) will accelerate proton beams up to an energy
of 13 TeV, with an integrated luminosity per year of around 100 fb-1, reaching 350 fb-1 after the
phase I upgrade.
CERN is planning an upgrade program (High Luminosity LHC) that should bring the integrated
luminosity to 3000 fb-1.
A more powerful LHC would provide more accurate measurements of new particles and enable
observation of rare processes that occur below the current sensitivity level.
efficiency
C
efficiency
High Luminosity LHC
HT phi efficiency for 10 GeV dimuon candidates
h_eff_reco_eta
Entries
160
Mean -0.0009472
1.369
RMS
PU
A preliminary study shows that the Hough transform has no strong dependence on
the pT of the candidates. The muons distribution is practically flat, while in the
electrons case a slightly drop at higher momenta has been observed.
Also the pile-up distribution does not show a clear dependence of the efficiency.
Otherwise it looks quite flat with an average value around 93% for muon candidates
and of about 70% for electrons.
Furthermore a study on number of fake candidates is currently in progress. However
it will be task of addiditional stages ( e.g. filtering, fitting ) to make rid of these fakes.
References
G. Hall, D. Newbold, M. Pesaresi and A. Rose, A time-multiplexed track-trigger
architecture for CMS
M. Pesaresi, Time-Multiplexed Track Finding Proposal: Hough Transform
M. Pesaresi, G. Hall, A. Rose, A. Tapper, K. Uchida, I. Tomalin, I. Reid, D. Newbold,
D. Cieri, P. Vichoudis; Track finding using a time multiplexed architecture:
status & plans
Track Trigger Integration group, Use of tracking in the CMS L1 trigger for the phase-2
upgrade
A. Tricomi, Upgrade of the CMS tracker