|M.Sc Student||Vainstein Constantine|
|Subject||Architectural and System Aspects of a SPAD-Based Muzzle|
Flash Detection System
|Department||Department of Electrical Engineering||Supervisors||Professor Yitzhak Birk|
|Professor Emeritus Yael Nemirovsky|
|Full Thesis text|
Military conflicts and terror attacks in recent years raise interest in friend-foe detection and response-aiding systems. Unlike a classical battlefield, where a ground force could rely on a soldier’s instinctive reaction to hostile gunfire, modern warfare in densely populated areas makes this approach dangerous for uninvolved civilians, requiring an immediate and accurate detection and localization of the fire source before the response.
Studies have suggested various methods to detect and localize gunfire source. The methods are based on acoustic, seismic, and electromagnetic signature, produced by gunshot event. Depending on the kind of processed signature, the detectors use different technologies. Gunfire source detection using sensors based on CMOS single photon avalanche diode (SPAD) is a promising approach because of enabling kilohertz-range sampling rates without signal-to-noise ratio (SNR) degradation, very high sensitivity to a single photon and built-in pixel-level analog-to-digital conversion (ADC) capability.
A muzzle flash detection system can be used in different operational scenarios and constellations, from being part of the personal equipment of an infantry soldier or law enforcement officer to being installed in a fixed site. Along another dimension, it can be a stand-alone detector or part of a bigger system, consisting of many detectors and control centers. The diverse usage profiles impose different, sometimes contradicting, requirements on the system, and pose different challenges for system designers.
In this work we explore architectural and system aspects of SPAD-based muzzle flash detection system, providing a framework that clarifies the design opportunities at system designers’ disposal to meet the operational needs. The framework is based on the inherent low duty cycle property of the system and the basic need to balance between the average and peak computational performance for a better cost-effectiveness and power consumption of the system. We identify the system characteristics derived from true operational needs, enabling to specify the system performance requirements. By carefully considering inherent system characteristics and true realistic requirements, we identify design guidelines that lead to various ways of optimizing the system for power consumption, latency, and throughput. We decompose the system capabilities into main building blocks, further analyzing the relations between system characteristics and design parameters. Finally, we offer enhancements to the suggested system architecture and already-existing building blocks for a better power efficiency, still meeting the system requirement dictated by the operational needs.