Akademik Veri Yönetim Sistemi
Tezin Türü: Yüksek Lisans
Tezin Yürütüldüğü Kurum: Dokuz Eylül Üniversitesi, Fen Bilimleri Enstitüsü, Türkiye
Tezin Onay Tarihi: 2025
Tezin Dili: İngilizce
Öğrenci: İLAYDA BONCUK
Danışman: Serkan Günel
Açık Arşiv Koleksiyonu: AVESİS Açık Erişim Koleksiyonu
Özet:
Accurate simulation of electronic circuits under parameter uncertainty is essential for modern design and verification. Classical SPICE simulations rely on nominal values and therefore provide deterministic solutions that may not reflect real-world variations such as manufacturing tolerances, ageing, or temperature changes. To address this challenge, two main approaches are widely used: probabilistic Monte Carlo methods and deterministic interval arithmetic. Monte Carlo simulations provide statistical insight but are computationally demanding and may overlook rare worst-case scenarios. Interval methods, in contrast, guarantee mathematically rigorous enclosures, yet often suffer from overestimation due to dependency and wrapping effects. This thesis presents the design and implementation of a Julia-based SPICE-like simulator that integrates interval arithmetic into the Modified Nodal Analysis (MNA) framework. The simulator automatically parses netlists, constructs MNA matrices, and applies multiple interval solvers, including Gaussian elimination, Jacobi, Gauss–Seidel, Krawczyk, and Oettli–Prager methods. The enclosures produced by these solvers are systematically compared with Monte Carlo results using the Hausdorff distance as a quantitative metric. Benchmark results highlight the trade-offs between accuracy, enclosure quality, and computational efficiency. While Monte Carlo remains versatile for probabilistic analysis, interval methods provide strong guarantees suitable for safety-critical applications. The outcomes of this study demonstrate the potential of interval arithmetic as a reliable complement to traditional simulation, offering practical guidelines for uncertainty-aware circuit design and verification.