A Spiking Neuromorphic Analytics Platform (SNAP) for Photomultiplier Signal Processing

The digitization, sampling, and processing of photomultiplier tube (PMT) signals present significant hardware-level challenges in both large scale and portable, deployable antineutrino/neutrino detector arrays, contributing to the high power consumption and suboptimal value-to-cost ratio of current commercial solutions. Additionally, the compact arrangement of PMTs generally requires integrating all essential hardware into a compact ‘base,’ adding critical physical size constraints. To address this, Phase III Physics LLC (P3P) proposes to develop a Spiking Neuromorphic Analytics Platform (SNAP) for Photomultiplier Signal Processing. SNAP includes a neuromorphic-enhanced PMT base coupled with an in-base advanced Spiking Neural Network (SNN) signal processing hardware, enabling on-device deployment of AI/ML-powered data pre-processing of raw PMT output. These networks will be deployed on energy-efficient, memristor-based application-specific integrated circuits (ASICs), enabling high-performance data processing in a compact and scalable architecture. We will leverage recent advancements in resistive RAM (ReRAM) / memristor fabrication, which now allow private firms to design and manufacture memristor application-specific integrated circuits (ASICs), eliminating the need for in-house fabrication infrastructure to enable this novel technology. Advanced data compression and reduction techniques will minimize data volume, improving scalability for large arrays. Neuromorphic-triggering algorithms will prioritize significant events, achieving adequate detection sensitivity and signal-to-noise ratio, and adaptive processing will enable real-time calibration based on environmental conditions. This effort will be led by Dr. Anuar Dorzhigulov, a neuromorphic engineer and researcher, who recently received his Ph.D. at the University of Delaware in Computer Engineering with specialization in Neuromorphic Computing Systems, where his main research focus was software-hardware co-design of neuromorphic hardware and systems.