-
Fast Neural Network Inference on FPGAs for Triggering on Long-Lived Particles at Colliders
Authors:
Andrea Coccaro,
Francesco Armando Di Bello,
Stefano Giagu,
Lucrezia Rambelli,
Nicola Stocchetti
Abstract:
Experimental particle physics demands a sophisticated trigger and acquisition system capable to efficiently retain the collisions of interest for further investigation. Heterogeneous computing with the employment of FPGA cards may emerge as a trending technology for the triggering strategy of the upcoming high-luminosity program of the Large Hadron Collider at CERN. In this context, we present two…
▽ More
Experimental particle physics demands a sophisticated trigger and acquisition system capable to efficiently retain the collisions of interest for further investigation. Heterogeneous computing with the employment of FPGA cards may emerge as a trending technology for the triggering strategy of the upcoming high-luminosity program of the Large Hadron Collider at CERN. In this context, we present two machine-learning algorithms for selecting events where neutral long-lived particles decay within the detector volume studying their accuracy and inference time when accelerated on commercially available Xilinx FPGA accelerator cards. The inference time is also confronted with a CPU- and GPU-based hardware setup. The proposed new algorithms are proven efficient for the considered benchmark physics scenario and their accuracy is found to not degrade when accelerated on the FPGA cards. The results indicate that all tested architectures fit within the latency requirements of a second-level trigger farm and that exploiting accelerator technologies for real-time processing of particle-physics collisions is a promising research field that deserves additional investigations, in particular with machine-learning models with a large number of trainable parameters.
△ Less
Submitted 19 December, 2023; v1 submitted 11 July, 2023;
originally announced July 2023.
-
Configurable calorimeter simulation for AI applications
Authors:
Francesco Armando Di Bello,
Anton Charkin-Gorbulin,
Kyle Cranmer,
Etienne Dreyer,
Sanmay Ganguly,
Eilam Gross,
Lukas Heinrich,
Lorenzo Santi,
Marumi Kado,
Nilotpal Kakati,
Patrick Rieck,
Matteo Tusoni
Abstract:
A configurable calorimeter simulation for AI (COCOA) applications is presented, based on the Geant4 toolkit and interfaced with the Pythia event generator. This open-source project is aimed to support the development of machine learning algorithms in high energy physics that rely on realistic particle shower descriptions, such as reconstruction, fast simulation, and low-level analysis. Specificati…
▽ More
A configurable calorimeter simulation for AI (COCOA) applications is presented, based on the Geant4 toolkit and interfaced with the Pythia event generator. This open-source project is aimed to support the development of machine learning algorithms in high energy physics that rely on realistic particle shower descriptions, such as reconstruction, fast simulation, and low-level analysis. Specifications such as the granularity and material of its nearly hermetic geometry are user-configurable. The tool is supplemented with simple event processing including topological clustering, jet algorithms, and a nearest-neighbors graph construction. Formatting is also provided to visualise events using the Phoenix event display software.
△ Less
Submitted 8 March, 2023; v1 submitted 3 March, 2023;
originally announced March 2023.
-
Reconstructing particles in jets using set transformer and hypergraph prediction networks
Authors:
Francesco Armando Di Bello,
Etienne Dreyer,
Sanmay Ganguly,
Eilam Gross,
Lukas Heinrich,
Anna Ivina,
Marumi Kado,
Nilotpal Kakati,
Lorenzo Santi,
Jonathan Shlomi,
Matteo Tusoni
Abstract:
The task of reconstructing particles from low-level detector response data to predict the set of final state particles in collision events represents a set-to-set prediction task requiring the use of multiple features and their correlations in the input data. We deploy three separate set-to-set neural network architectures to reconstruct particles in events containing a single jet in a fully-simul…
▽ More
The task of reconstructing particles from low-level detector response data to predict the set of final state particles in collision events represents a set-to-set prediction task requiring the use of multiple features and their correlations in the input data. We deploy three separate set-to-set neural network architectures to reconstruct particles in events containing a single jet in a fully-simulated calorimeter. Performance is evaluated in terms of particle reconstruction quality, properties regression, and jet-level metrics. The results demonstrate that such a high dimensional end-to-end approach succeeds in surpassing basic parametric approaches in disentangling individual neutral particles inside of jets and optimizing the use of complementary detector information. In particular, the performance comparison favors a novel architecture based on learning hypergraph structure, HGPflow, which benefits from a physically-interpretable approach to particle reconstruction.
△ Less
Submitted 2 August, 2023; v1 submitted 2 December, 2022;
originally announced December 2022.
-
Towards a Computer Vision Particle Flow
Authors:
Francesco Armando Di Bello,
Sanmay Ganguly,
Eilam Gross,
Marumi Kado,
Michael Pitt,
Lorenzo Santi,
Jonathan Shlomi
Abstract:
In High Energy Physics experiments Particle Flow (PFlow) algorithms are designed to provide an optimal reconstruction of the nature and kinematic properties of the particles produced within the detector acceptance during collisions. At the heart of PFlow algorithms is the ability to distinguish the calorimeter energy deposits of neutral particles from those of charged particles, using the compleme…
▽ More
In High Energy Physics experiments Particle Flow (PFlow) algorithms are designed to provide an optimal reconstruction of the nature and kinematic properties of the particles produced within the detector acceptance during collisions. At the heart of PFlow algorithms is the ability to distinguish the calorimeter energy deposits of neutral particles from those of charged particles, using the complementary measurements of charged particle tracking devices, to provide a superior measurement of the particle content and kinematics. In this paper, a computer vision approach to this fundamental aspect of PFlow algorithms, based on calorimeter images, is proposed. A comparative study of the state of the art deep learning techniques is performed. A significantly improved reconstruction of the neutral particle calorimeter energy deposits is obtained in a context of large overlaps with the deposits from charged particles. Calorimeter images with augmented finer granularity are also obtained using super-resolution techniques.
△ Less
Submitted 3 February, 2021; v1 submitted 19 March, 2020;
originally announced March 2020.
-
A 50 ps resolution monolithic active pixel sensor without internal gain in SiGe BiCMOS technology
Authors:
G. Iacobucci,
R. Cardarelli,
S. Débieux,
F. A. Di Bello,
Y. Favre,
D. Hayakawa,
M. Kaynak,
M. Nessi,
L. Paolozzi,
H. Rücker,
DMS Sultan,
P. Valerio
Abstract:
A monolithic pixelated silicon detector designed for high time resolution has been produced in the SG13G2 130 nm SiGe BiCMOS technology of IHP Mikroelektronik. This proof-of-concept chip contains hexagonal pixels of 65 μm and 130 μm side. The SiGe front-end electronics implemented provides an equivalent noise charge of 90 and 160 electrons for a pixel capacitance of 70 and 220 fF, respectively, an…
▽ More
A monolithic pixelated silicon detector designed for high time resolution has been produced in the SG13G2 130 nm SiGe BiCMOS technology of IHP Mikroelektronik. This proof-of-concept chip contains hexagonal pixels of 65 μm and 130 μm side. The SiGe front-end electronics implemented provides an equivalent noise charge of 90 and 160 electrons for a pixel capacitance of 70 and 220 fF, respectively, and a total time walk of less than 1 ns. Lab measurements with a 90Sr source show a time resolution of the order of 50 ps. This result is competitive with silicon technologies that integrate an avalanche gain mechanism.
△ Less
Submitted 26 August, 2019;
originally announced August 2019.
-
Characterisation of AMS H35 HV-CMOS monolithic active pixel sensor prototypes for HEP applications
Authors:
S. Terzo,
M. Benoit,
E. Cavallaro,
R. Casanova,
F. A. Di Bello,
F. Förster,
S. Grinstein,
G. Iacobucci,
I. Perić,
C. Puigdengoles,
M. Vicente Barrero Pinto,
E. Vilella Figueras
Abstract:
Monolithic active pixel sensors produced in High Voltage CMOS (HV-CMOS) technology are being considered for High Energy Physics applications due to the ease of production and the reduced costs. Such technology is especially appealing when large areas to be covered and material budget are concerned. This is the case of the outermost pixel layers of the future ATLAS tracking detector for the HL-LHC.…
▽ More
Monolithic active pixel sensors produced in High Voltage CMOS (HV-CMOS) technology are being considered for High Energy Physics applications due to the ease of production and the reduced costs. Such technology is especially appealing when large areas to be covered and material budget are concerned. This is the case of the outermost pixel layers of the future ATLAS tracking detector for the HL-LHC. For experiments at hadron colliders, radiation hardness is a key requirement which is not fulfilled by standard CMOS sensor designs that collect charge by diffusion. This issue has been addressed by depleted active pixel sensors in which electronics are embedded into a large deep implantation ensuring uniform charge collection by drift. Very first small prototypes of hybrid depleted active pixel sensors have already shown a radiation hardness compatible with the ATLAS requirements. Nevertheless, to compete with the present hybrid solutions a further reduction in costs achievable by a fully monolithic design is desirable. The H35DEMO is a large electrode full reticle demonstrator chip produced in AMS 350 nm HV-CMOS technology by the collaboration of Karlsruher Institut für Technologie (KIT), Institut de Física d'Altes Energies (IFAE), University of Liverpool and University of Geneva. It includes two large monolithic pixel matrices which can be operated standalone. One of these two matrices has been characterised at beam test before and after irradiation with protons and neutrons. Results demonstrated the feasibility of producing radiation hard large area fully monolithic pixel sensors in HV-CMOS technology. H35DEMO chips with a substrate resistivity of 200$Ω$ cm irradiated with neutrons showed a radiation hardness up to a fluence of $10^{15}$n$_{eq}$cm$^{-2}$ with a hit efficiency of about 99% and a noise occupancy lower than $10^{-6}$ hits in a LHC bunch crossing of 25ns at 150V.
△ Less
Submitted 30 January, 2019; v1 submitted 19 November, 2018;
originally announced November 2018.
-
Charge collection characterisation with the Transient Current Technique of the ams H35DEMO CMOS detector after proton irradiation
Authors:
John Anders,
Mathieu Benoit,
Saverio Braccini,
Raimon Casanova,
Hucheng Chen,
Kai Chen,
Francesco Armando di Bello,
Armin Fehr,
Didier Ferrere,
Dean Forshaw,
Tobias Golling,
Sergio Gonzalez-Sevilla,
Giuseppe Iacobucci,
Moritz Kiehn,
Francesco Lanni,
Hongbin Liu,
Lingxin Meng,
Claudia Merlassino,
Antonio Miucci,
Marzio Nessi,
Ivan Perić,
Marco Rimoldi,
D M S Sultan,
Mateus Vincente Barreto Pinto,
Eva Vilella
, et al. (4 additional authors not shown)
Abstract:
This paper reports on the characterisation with Transient Current Technique measurements of the charge collection and depletion depth of a radiation-hard high-voltage CMOS pixel sensor produced at ams AG. Several substrate resistivities were tested before and after proton irradiation with two different sources: the 24 GeV Proton Synchrotron at CERN and the 16.7 MeV Cyclotron at Bern Inselspital.
This paper reports on the characterisation with Transient Current Technique measurements of the charge collection and depletion depth of a radiation-hard high-voltage CMOS pixel sensor produced at ams AG. Several substrate resistivities were tested before and after proton irradiation with two different sources: the 24 GeV Proton Synchrotron at CERN and the 16.7 MeV Cyclotron at Bern Inselspital.
△ Less
Submitted 25 July, 2018;
originally announced July 2018.
-
Performance of CMOS pixel sensor prototypes in ams H35 and aH18 technology for the ATLAS ITk upgrade
Authors:
Moritz Kiehn,
Francesco Armando Di Bello,
Mathieu Benoit,
Raimon Casanova Mohr,
Hucheng Chen,
Kai Chen,
Sultan D. M. S.,
Felix Ehrler,
Didier Ferrere,
Dylan Frizell,
Sergio Gonzalez Sevilla,
Giuseppe Iacobucci,
Francesco Lanni,
Hongbin Liu,
Claudia Merlassino,
Jessica Metcalfe,
Antonio Miucci,
Ivan Peric,
Mridula Prathapan,
Rudolf Schimassek,
Mateus Vicente Barreto,
Thomas Weston,
Eva Vilella Figueras,
Alena Weber,
Michele Weber
, et al. (5 additional authors not shown)
Abstract:
Pixel sensors based on commercial high-voltage CMOS processes are an exciting technology that is considered as an option for the outer layer of the ATLAS inner tracker upgrade at the High Luminosity LHC. Here, charged particles are detected using deep n-wells as sensor diodes with the depleted region extending into the silicon bulk. Both analog and digital readout electronics can be added to achie…
▽ More
Pixel sensors based on commercial high-voltage CMOS processes are an exciting technology that is considered as an option for the outer layer of the ATLAS inner tracker upgrade at the High Luminosity LHC. Here, charged particles are detected using deep n-wells as sensor diodes with the depleted region extending into the silicon bulk. Both analog and digital readout electronics can be added to achieve different levels of integration up to a fully monolithic sensor. Small scale prototypes using the ams CMOS technology have previously demonstrated that it can achieve the required radiation tolerance of $10^{15}~\text{n}_\text{eq}/\text{cm}^2$ and detection efficiencies above $99.5~\%$. Recently, large area prototypes, comparable in size to a full sensor, have been produced that include most features required towards a final design: the H35demo prototype produced in ams H35 technology that supports both external and integrated readout and the monolithic ATLASPix1 pre-production design produced in ams aH18 technology. Both chips are based on large fill-factor pixel designs, but differ in readout structure. Performance results for H35DEMO with capacitively-coupled external readout and first results for the monolithic ATLASPix1 are shown.
△ Less
Submitted 8 June, 2020; v1 submitted 16 July, 2018;
originally announced July 2018.
-
Test beam measurement of ams H35 HV-CMOS capacitively coupled pixel sensor prototypes with high-resistivity substrate
Authors:
M. Benoit,
S. Braccini,
R. Casanova,
E. Cavallaro,
H. Chen,
K. Chen,
F. A. Di Bello,
D. Ferrere,
D. Frizzell,
T. Golling,
S. Gonzalez-Sevilla,
S. Grinstein,
G. Iacobucci,
M. Kiehn,
F. Lanni,
H. Liu,
J. Metcalfe,
L. Meng,
C. Merlassino,
A. Miucci,
D. Muenstermann,
M. Nessi,
H. Okawa,
I. Perić,
M. Rimoldi
, et al. (12 additional authors not shown)
Abstract:
In the context of the studies of the ATLAS High Luminosity LHC programme, radiation tolerant pixel detectors in CMOS technologies are investigated. To evaluate the effects of substrate resistivity on CMOS sensor performance, the H35DEMO demonstrator, containing different diode and amplifier designs, was produced in ams H35 HV-CMOS technology using four different substrate resistivities spanning fr…
▽ More
In the context of the studies of the ATLAS High Luminosity LHC programme, radiation tolerant pixel detectors in CMOS technologies are investigated. To evaluate the effects of substrate resistivity on CMOS sensor performance, the H35DEMO demonstrator, containing different diode and amplifier designs, was produced in ams H35 HV-CMOS technology using four different substrate resistivities spanning from $\mathrm{80}$ to $\mathrm{1000~Ω\cdot cm}$. A glueing process using a high-precision flip-chip machine was developed in order to capacitively couple the sensors to FE-I4 Readout ASIC using a thin layer of epoxy glue with good uniformity over a large surface. The resulting assemblies were measured in beam test at the Fermilab Test Beam Facilities with 120 GeV protons and CERN SPS H8 beamline using 80 GeV pions. The in-time efficiency and tracking properties measured for the different sensor types are shown to be compatible with the ATLAS ITk requirements for its pixel sensors.
△ Less
Submitted 3 December, 2018; v1 submitted 22 December, 2017;
originally announced December 2017.
-
Testbeam results of irradiated ams H18 HV-CMOS pixel sensor prototypes
Authors:
M. Benoit,
S. Braccini,
G. Casse,
H. Chen,
K. Chen,
F. A. Di Bello,
D. Ferrere,
T. Golling,
S. Gonzalez-Sevilla,
G. Iacobucci,
M. Kiehn,
F. Lanni,
H. Liu,
L. Meng,
C. Merlassino,
A. Miucci,
D. Muenstermann,
M. Nessi,
H. Okawa,
I. Peric,
M. Rimoldi,
B. Ristic,
M. Vicente Barrero Pinto,
J. Vossebeld,
M. Weber
, et al. (4 additional authors not shown)
Abstract:
HV-CMOS pixel sensors are a promising option for the tracker upgrade of the ATLAS experiment at the LHC, as well as for other future tracking applications in which large areas are to be instrumented with radiation-tolerant silicon pixel sensors. We present results of testbeam characterisations of the $4^{\mathrm{th}}$ generation of Capacitively Coupled Pixel Detectors (CCPDv4) produced with the am…
▽ More
HV-CMOS pixel sensors are a promising option for the tracker upgrade of the ATLAS experiment at the LHC, as well as for other future tracking applications in which large areas are to be instrumented with radiation-tolerant silicon pixel sensors. We present results of testbeam characterisations of the $4^{\mathrm{th}}$ generation of Capacitively Coupled Pixel Detectors (CCPDv4) produced with the ams H18 HV-CMOS process that have been irradiated with different particles (reactor neutrons and 18 MeV protons) to fluences between $1\cdot 10^{14}$ and $5\cdot 10^{15}$ 1-MeV-n$_\textrm{eq}$/cm$^2$. The sensors were glued to ATLAS FE-I4 pixel readout chips and measured at the CERN SPS H8 beamline using the FE-I4 beam telescope. Results for all fluences are very encouraging with all hit efficiencies being better than 97% for bias voltages of $85\,$V. The sample irradiated to a fluence of $1\cdot 10^{15}$ n$_\textrm{eq}$/cm$^2$ - a relevant value for a large volume of the upgraded tracker - exhibited 99.7% average hit efficiency. The results give strong evidence for the radiation tolerance of HV-CMOS sensors and their suitability as sensors for the experimental HL-LHC upgrades and future large-area silicon-based tracking detectors in high-radiation environments.
△ Less
Submitted 28 November, 2017; v1 submitted 8 November, 2016;
originally announced November 2016.
-
Development of a modular test system for the silicon sensor R&D of the ATLAS Upgrade
Authors:
H. Liu,
M. Benoit,
H. Chen,
K. Chen,
F. A. Di Bello,
G. Iacobucci,
F. Lanni,
I. Peric,
B. Ristic,
M. Vicente Barreto Pinto,
W. Wu,
L. Xu,
G. Jin
Abstract:
High Voltage CMOS sensors are a promising technology for tracking detectors in collider experiments. Extensive R&D studies are being carried out by the ATLAS Collaboration for a possible use of HV-CMOS in the High Luminosity LHC upgrade of the Inner Tracker detector. CaRIBOu (Control and Readout Itk BOard) is a modular test system developed to test Silicon based detectors. It currently includes fi…
▽ More
High Voltage CMOS sensors are a promising technology for tracking detectors in collider experiments. Extensive R&D studies are being carried out by the ATLAS Collaboration for a possible use of HV-CMOS in the High Luminosity LHC upgrade of the Inner Tracker detector. CaRIBOu (Control and Readout Itk BOard) is a modular test system developed to test Silicon based detectors. It currently includes five custom designed boards, a Xilinx ZC706 development board, FELIX (Front-End LInk eXchange) PCIe card and a host computer. A software program has been developed in Python to control the CaRIBOu hardware. CaRIBOu has been used in the testbeam of the HV-CMOS sensor CCPDv4 at CERN. Preliminary results have shown that the test system is very versatile. Further development is ongoing to adapt to different sensors, and to make it available to various lab test stands.
△ Less
Submitted 17 October, 2016; v1 submitted 25 March, 2016;
originally announced March 2016.
-
Results of the 2015 testbeam of a 180 nm AMS High-Voltage CMOS sensor prototype
Authors:
M. Benoit,
J. Bilbao de Mendizabal,
G. Casse,
H. Chen,
K. Chen,
F. A. Di Bello,
D. Ferrere,
T. Golling,
S. Gonzalez-Sevilla,
G. Iacobucci,
F. Lanni,
H. Liu,
F. Meloni,
L. Meng,
A. Miucci,
D. Muenstermann,
M. Nessi,
I. Peric,
M. Rimoldi,
B. Ristic,
M. Vicente Barrero Pinto,
J. Vossebeld,
M. Weber,
W. Wu,
L. Xu
Abstract:
Active pixel sensors based on the High-Voltage CMOS technology are being investigated as a viable option for the future pixel tracker of the ATLAS experiment at the High-Luminosity LHC. This paper reports on the testbeam measurements performed at the H8 beamline of the CERN Super Proton Synchrotron on a High-Voltage CMOS sensor prototype produced in 180 nm AMS technology. Results in terms of track…
▽ More
Active pixel sensors based on the High-Voltage CMOS technology are being investigated as a viable option for the future pixel tracker of the ATLAS experiment at the High-Luminosity LHC. This paper reports on the testbeam measurements performed at the H8 beamline of the CERN Super Proton Synchrotron on a High-Voltage CMOS sensor prototype produced in 180 nm AMS technology. Results in terms of tracking efficiency and timing performance, for different threshold and bias conditions, are shown.
△ Less
Submitted 30 June, 2016; v1 submitted 24 March, 2016;
originally announced March 2016.
-
The FE-I4 Telescope for particle tracking in testbeam experiments
Authors:
M. Benoit,
J. Bilbao De Mendizabal,
F. A. Di Bello,
D. Ferrere,
T. Golling,
S. Gonzalez-Sevilla,
G. Iacobucci,
M. Kocian,
D. Muenstermann,
B. Ristic,
A. Sciuccati
Abstract:
A testbeam telescope, based on ATLAS IBL silicon pixel modules, has been built. It comprises six planes of planar silicon sensors with 250 x 50 um^2 pitch, read out by ATLAS FE-I4 chips. In the CERN SPS H8 beamline (180 GeV pi+) a resolution of better than 8 x 12 um^2 at the position of the device under test was achieved. The telescope reached a trigger rate of 6kHz with two measured devices. It i…
▽ More
A testbeam telescope, based on ATLAS IBL silicon pixel modules, has been built. It comprises six planes of planar silicon sensors with 250 x 50 um^2 pitch, read out by ATLAS FE-I4 chips. In the CERN SPS H8 beamline (180 GeV pi+) a resolution of better than 8 x 12 um^2 at the position of the device under test was achieved. The telescope reached a trigger rate of 6kHz with two measured devices. It is mainly designed for studies using FE-I4 based prototypes, but has also been successfully run with independent DAQ systems. Specialised trigger schemes ensure data synchronisation between these external devices and the telescope. A region-of-interest trigger can be formed by setting masks on the first and the last pixel sensor planes. The setup infrastructure provides centrally controlled and monitored high and low voltage power supplies, silicon oil cooling, temperature and humidity sensors and movable stages.
△ Less
Submitted 20 June, 2016; v1 submitted 24 March, 2016;
originally announced March 2016.