Zero-Server Browser-Native Harmonic Balance Periodic Steady-State Analysis on WebGPU with Mixed-Precision Newton–Krylov Solving
The KING TSENG AI Web EDA series provides zero-server, browser-native transient simulation and WebGPU mixed-precision sparse solving, but it has lacked a periodic steady-state (PSS) engine. Reaching steady state with a long transient is wasteful for high-Q or lightly damped circuits. This paper presents the design of a browser-native harmonic balance (HB) PSS engine on WebGPU. Nonlinear devices are evaluated in the time domain on a uniform collocation grid in compute shaders, batched FFTs move between time and frequency, and a matrix-free Newton–GMRES solver with a per-harmonic block-diagonal MNA preconditioner avoids assembling the dense harmonic Jacobian. Because WebGPU shaders lack fp64, fp32 kernels are combined with mixed-precision iterative refinement of the residual; source-stepping and harmonic-count continuation aid convergence, and a WASM-SIMD CPU path handles small circuits where GPU overhead dominates. Everything runs client-side, so netlists never leave the user's machine. To the best of our knowledge, this is the first browser-native WebGPU HB PSS engine design. No measured results are reported; a reproducible validation protocol against analytic phasor solutions and ngspice is defined instead. The tool targets education, exploration, and pre-layout what-if analysis, not foundry sign-off.
Authors
- I-Wei Tseng
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-18
- DOI
- https://doi.org/10.5281/zenodo.24328837
- Primary Topic
- Electromagnetic Simulation and Numerical Methods
- Type
- preprint