Hardware Detection and Localization of Single-event Upsets Based on the Dual-code Weight Difference Invariant in Sbns and Nfns
Year: 2026
Type: preprint
Abstract: This paper presents a method for the hardware detection, localization, and automatic correction of single-event upsets (SEUs) in arithmetic logic units and cache memory operating in the Scaled Bergman Number System (SBNS) and the New Fibonacci Number System (NFNS). As an alternative to classical Hamming codes, data integrity verification is proposed based on the Hamming weight (PopCount) difference invariant for a dual pair of valid code states: the working code “A” and the check code “B”. It is shown that a deviation of the weight delta from its reference value allows not only detecting a fault without computing complex syndromes, but also instantaneously localizing the corrupted code and determining the physical nature of the error (a bit flip from 1 to 0 or from 0 to 1). A symmetry edge effect for the number 1 in NFNS is examined, and a microarchitectural solution in the form of a logic guard is proposed. An architecture of mirrored cache memory (SRAM) banks with virtual bit-grid shifting is substantiated to protect against multiple-bit upsets (MBUs) and enable background cell self-healing.
Author: Voron, Andrey
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236421436
