What question did this study set out to answer?

To validate time-frequency representations aligned with the GCSIN signal class for tokamak diagnostic analysis.

March 15, 2026Open Access

Empirical and Synthetic Validation of GCSIN-Aligned Time–Frequency Representations for Tokamak Diagnostics

Key Points

To validate time-frequency representations aligned with the GCSIN signal class for tokamak diagnostic analysis.
Constructed a synthetic benchmark suite with four signal families
Compared standard representations against GCSIN-aligned kernels
Used quantitative metrics including ridge sharpness, time-frequency entropy, and mode separability.
GCSIN-aligned representations reduced ridge dispersion and representational ambiguity
Improvements were largest for signals with frequency evolution, intermittency, and geometry overlap

Abstract

This preprint provides empirical and synthetic validation of time–frequency representations aligned with the Geometry-Constrained Scale-Invariant Nonstationary (GCSIN) signal class for tokamak diagnostic analysis. Tokamak diagnostic signals commonly exhibit strong nonstationarity, frequency chirping, intermittency, and overlapping modal activity. A parametric synthetic benchmark suite comprising four signal families was constructed to reflect recurring diagnostic phenomenology: linear-frequency chirps with geometric phase structure, multiplicative (log-frequency) chirps, intermittent burst mixtures with broadband backgrounds, and overlapping modes constrained by geometry. Standard representations (STFT, CWT, synchrosqueezed wavelets, reassignment) were systematically compared against GCSIN-aligned kernels using three quantitative metrics: ridge sharpness, time–frequency entropy, and mode separability. Results show that GCSIN-aligned representations consistently reduce ridge dispersion, representational ambiguity, and modal conflation across parameter ranges, with the largest gains for signals exhibiting relative frequency evolution, intermittency, or geometry-driven overlap. These improvements arise from structural alignment between signal properties and representational assumptions rather than from algorithmic complexity. The benchmark suite and evaluation metrics introduced here provide a neutral, reproducible framework for assessing time–frequency representations in nonstationary, multi-scale, and geometry-structured diagnostic settings. This work focuses exclusively on the representational layer and does not address physical modeling, prediction, or control. Related open-source artifacts and prior work:• Psi Universe Attractor Library v2.0: https://doi.org/10.5281/zenodo.18939068• TASD Unified Framework: https://doi.org/10.5281/zenodo.18926912• A formal class of scale-invariant, geometry-constrained nonstationary signals: https://doi.org/10.5281/zenodo.18247445• A toroidal log-chirplet transform for nonstationary, scale-invariant plasma signals: https://doi.org/10.5281/zenodo.18050116 Keywords: Time-frequency analysis, GCSIN signals, Tokamak diagnostics, Synthetic benchmarks, Ridge sharpness, Mode separability

Empirical and Synthetic Validation of GCSIN-Aligned Time–Frequency Representations for Tokamak Diagnostics

Key Points

Abstract

Cite This Study