Smart pixel sensors: towards on-sensor filtering of pixel clusters with deep learning

• DOI: 10.1088/2632-2153/ad6a00
• Publication Date: 2024-08-01T00:48:07Z
• Authors: Jieun Yoo; Jennet Dickinson; Morris Swartz; Giuseppe Di Gugli...; Alice Bean; Douglas Berry; Manuel Blanco Val...; Karri DiPetrillo; Farah Fahim; Lindsey Gray; James Hirschauer; Shruti R Kulkarni; Ron Lipton; Petar Maksimovic; Corrinne Mills; Mark S Neubauer; Benjamin Parpillon; Gauri Pradhan; Chinar Syal; Nhan Tran; Dahai Wen; Aaron Young
• Keywords: Computer engineering. Computer hardware; Physics - Instrumentation and Detectors; machine-learning; FOS: Physical sciences; QA75.5-76.95; Instrumentation and Detectors (physics.ins-det); 7. Clean energy; high energy physics; High Energy Physics - Experiment; TK7885-7895; High Energy Physics - Experiment (hep-ex); Electronic computers. Computer science; colliders; detectors

Abstract Highly granular pixel detectors allow for increasingly precise measurements of charged particle tracks. Next-generation require that sizes will be further reduced, leading to unprecedented data rates exceeding those foreseen at the High- Luminosity Large Hadron Collider. Signal processing handles incoming a rate <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:mi class="MJX-tex-calligraphic">O</mml:mi> </mml:mrow> </mml:math> (40 MHz) and intelligently reduces within pixelated region detector enhance physics performance high luminosity enable analyses are not currently possible. Using shape charge clusters deposited in an array small pixels, physical properties traversing can extracted with locally customized neural networks. In this first demonstration, we present network embedded into on-sensor readout filter out hits from low momentum tracks, reducing detector’s volume by 57.1%–75.7%. The is designed simulated as custom integrated circuit 28 nm CMOS technology expected operate less than 300 <mml:mi>μ</mml:mi> <mml:mi>W</mml:mi> area 0.2 mm 2 . temporal development investigated demonstrate possible future gains, there also discussion algorithmic technological improvements could efficiency, reduction, power per area.