Virtual Analog Bibliography

This page replaces the old local doc/research/ PDF folder. Do not commit third-party PDFs unless their redistribution license is explicit and compatible with the project. Keep public links, citations, and Greybound-specific engineering notes here instead.

Amp Modeling Review

Jyri Pakarinen and David T. Yeh, "A Review of Digital Techniques for Modeling Vacuum-Tube Guitar Amplifiers", Computer Music Journal 33(2), 85-100, 2009.

What it covers:

• taxonomy of digital tube-amp modeling,
• preamp, tone stack, power amp, output transformer, and loudspeaker/load interactions,
• black-box system identification,
• white-box circuit equations,
• nonlinear filters and Volterra-style approaches.

Greybound interest:

• Still useful as the historical map of the problem.
• Good reminder that amp modeling is not only static clipping: operating point, coupling, tone networks, power amp state, transformer, and speaker/load all matter.
• Supports our decision to keep amps as stage-level systems rather than a single waveshaper.

Current status:

• Not obsolete as a conceptual survey.
• Dated for modern neural/DDSP/SSM methods and recent WDF work.
• Use it for vocabulary and architectural decomposition, not as the state of the art.

Wave Digital Filter Foundation

Alfred Fettweis, "Wave Digital Filters: Theory and Practice", Proceedings of the IEEE 74(2), 270-327, 1986.

What it covers:

• the core WDF mapping from circuit variables to wave variables,
• passivity-preserving digital filter structures,
• port resistances and adaptor concepts,
• stability-oriented discretization thinking.

Greybound interest:

• Foundational if we promote more reusable circuit components to WDF form.
• Especially relevant for passive tone networks, coupling networks, and future physically structured nonlinear blocks.
• Explains why WDF is attractive for real-time stability compared with naive circuit solving.

Current status:

• Foundational, not obsolete.
• Too low-level and general to directly implement a Greybound amp by itself.
• Should be paired with later arbitrary-topology and nonlinear-WDF papers before implementation.

Arbitrary WDF Topologies

Kurt J. Werner, Julius O. Smith III, and Jonathan S. Abel, "Wave Digital Filter Adaptors for Arbitrary Topologies and Multiport Linear Elements", DAFx-15, Trondheim, 2015.

What it covers:

• Modified-Nodal-Analysis-derived WDF adaptors,
• complicated non-series/parallel topologies,
• multiport linear elements such as transformers and controlled sources,
• examples including Bassman tone stack and Tube Screamer tone/volume stages.

Greybound interest:

• Very relevant for making circuit diagrams executable rather than only documentary.
• A useful path for tone stacks, transformer-like linear networks, and pedal tone/volume networks that are not simple cascades.
• Matches our long-term desire to construct WDF/SPICE-style graphs from JSON circuit descriptions.

Current status:

• Still relevant.
• It expands linear/multiport topology handling; it does not solve every nonlinear multi-device case by itself.
• Use as a bridge from our circuit graph format to stable real-time linear subcircuits.

LSTM Guitar Amp Baseline

Thomas Schmitz and Jean-Jacques Embrechts, "Real Time Emulation of Parametric Guitar Tube Amplifier With Long Short Term Memory Neural Network", arXiv:1804.07145, 2018.

What it covers:

• black-box LSTM modeling of a tube guitar amplifier,
• gain-parameter conditioning,
• real-time neural inference target,
• reported low RMS error against measured amplifier output.

Greybound interest:

• Useful as a historical black-box baseline and evidence that neural sequence models can emulate nonlinear amp behavior.
• Good contrast against our preferred gray-box approach.
• Relevant for future comparison metrics and capture protocols.

Current status:

• Partly obsolete as a primary architecture choice.
• LSTM-only models are less interpretable, harder to guarantee, and less aligned with our rig/circuit state design.
• More recent SSM, DDSP, and neural gray-box work is a better fit for Greybound.

Neural WDF Nonlinearities

Oliviero Massi, Edoardo Manino, and Alberto Bernardini, "Wave Digital Modeling of Circuits with Multiple One-Port Nonlinearities Based on Lipschitz-Bounded Neural Networks", DAFx-24, Guildford, 2024.

What it covers:

• neural models inside Wave Digital Filter structures,
• multiple nonlinear one-port elements,
• Lipschitz-bounded neural networks,
• convergence conditions for fixed-point methods such as the Scattering Iterative Method.

Greybound interest:

• Important if we later model coupled nonlinear devices inside one WDF network.
• Relevant to diode networks, transistor cells, and future nonlinear WDF subcircuits.
• The Lipschitz constraint is the key idea: learned nonlinearities must preserve solver convergence, not just fit data.

Current status:

• Current and useful, but not immediately required by the present Greybound runtime.
• The paper targets multiple one-port nonlinearities; tube stages and transformers may need multiport/vector treatment.
• Use when we introduce nonlinear delay-free-loop solvers, not for every pedal or amp block.

Modulation Gray-Box Optimization

Alistair Carson, Alec Wright, and Stefan Bilbao, "Gradient-based Optimisation of Modulation Effects", arXiv:2601.04867, submitted 8 January 2026.

What it covers:

• phaser, flanger, and chorus modeling,
• differentiable digital signal processing,
• gradient-based parameter optimization,
• time-frequency training,
• zero-latency time-domain inference,
• real analog reference units.

Greybound interest:

• Directly relevant to Tron, Jetstream, and Celeste.
• Supports our live target: fitted gray-box parameters with explicit, low-latency runtime structures.
• Better fit than generic neural black-box models for modulation effects.

Current status:

• Very relevant and recent.
• Should drive our next modulation-modeling plan.
• See Modulation Gray-Box Modeling for Greybound-specific mapping.