ORCID
https://orcid.org/0009-0006-1589-693X
Date of Award
Spring 5-7-2025
Degree Name
Master of Science (MS)
Degree Type
Thesis
Abstract
FPGAs are widely deployed on high-energy astroparticle physics instruments to preprocess large volumes of streaming data from various sensors. Increasingly, these deployments are finding their way to space-borne instruments, where constraints on size, weight, and power (SWaP) require careful balancing of speed and resource utilization. Although telescope designs vary widely, they often share common preprocessing elements, including channel-level readout, pedestal subtraction, waveform integration, and zero suppression from front-end ADCs, as well as identification and centroiding of signal islands across groups of multiple channels. High-Level Synthesis (HLS) tools allow these designs to be expressed at a conceptual level, which automates a significant amount of the workload in FPGA design and development; compared to traditional hardware description languages, this enables rapid prototyping and redeployment of common logic across different instruments. Nonetheless, prior work lacks sufficient generality to be applied in a broader context, requiring logic to be rewritten from scratch for a new instrument. In this work, we explore generic programming paradigms in the context of HLS, demonstrating that this allows sufficient flexibility of specification to create a pipeline that can be recompiled to automatically accommodate distinct instruments from a manifest of their unique properties and requirements. We apply this to two diverse instrument designs: scintillator-based detectors with 1D pixel arrays, and imaging atmospheric Cherenkov telescopes with 2D pixel arrays. In doing so, we also present a novel HLS-based two-pass connected-component labeling (CCL) implementation that can be easily switched between 4-way and 8-way CCL.
Language
English (en)
Chair
Roger Chamberlain, CSE
Committee Members
Michael Hall, Marion Sudvarg