Neural Modeling of Guitar Tone Stacks
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Sähkötekniikan korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2024-05-20
Department
Major/Subject
Acoustics and Audio Technology
Mcode
ELEC3030
Degree programme
CCIS - Master’s Programme in Computer, Communication and Information Sciences (TS2013)
Language
en
Pages
60
Series
Abstract
Guitar tone stack circuits are complicated combinations of analog filters. Their purpose is to control the tonal characteristics of guitar and bass amplifiers or effects. Many unique and desired tones belong to boutique amplifiers. However, analog hardware can be expensive and fragile. Therefore, it is desirable to have digital models that are more affordable while accurately capturing the essential characteristics of the physical devices. To address this, research in virtual analog modeling has gained significance. This thesis aims to model the tone stack circuits of guitar amplifiers and effect units using a neural network with a differentiable filter. The neural network directly learns to map input tone controls to filter parameters. A differentiable coupled-form state-space filter is proposed and compared against a baseline model based on a direct-form structure. The coupled-form state-space representation is inherently robust to quantization noise and stable with time-varying coefficients when all poles are inside the unit circle. In total, ten tone stack circuits from guitar amplifiers and distortion effects are modeled. The experimental results indicate that the proposed method outperforms the baseline, achieving a lower overall loss on the test data. Furthermore, this method provides generalizability across various tone circuits and scalability for high-order filters. The proposed model can be implemented into amplifier modelers and used in real-time applications such as audio plugins.Description
Supervisor
Välimäki, VesaThesis advisor
Damskägg, Eero-PekkaKeywords
artificial neural networks, audio system, digital filters, real-time systems, state-space methods