Citation: Vanhatalo, T.; Legrand, P.;
Desainte-Catherine, M.; Hanna, P.;
Brusco, A.; Pille, G.; Bayle, Y. A
Review of Neural Network-Based
Emulation of Guitar Amplifiers. Appl.
Sci. 2022, 12, 5894. https://doi.org/
10.3390/app12125894
Academic Editors: Phivos Mylonas,
Katia Lida Kermanidis and Manolis
Maragoudakis
Received: 26 April 2022
Accepted: 8 June 2022
Published: 9 June 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Review
A Review of Neural Network-Based Emulation of Guitar Amplifiers
Tara Vanhatalo
1,2,3,
* , Pierrick Legrand
1
, Myriam Desainte-Catherine
2
, Pierre Hanna
2
, Antoine Brusco
3
,
Guillaume Pille
3
and Yann Bayle
3
1
Inria Bordeaux Sud-Ouest, Institute of Mathematics of Bordeaux, UMR 5251 CNRS, University of Bordeaux,
F-33405 Talence, France; pierrick.legrand@u-bordeaux.fr
2
University of Bordeaux, CNRS, Bordeaux INP, LaBRI, UMR 5800, F-33400 Talence, France;
myriam.desainte-catherine@labri.fr (M.D.-C.); pierre.hanna@labri.fr (P.H.)
3
Orosys, F-34980 Saint-Gély-du-Fesc, France; antoine.brusco@orosys.fr (A.B.); guillaume.pille@orosys.fr (G.P.);
yann.bayle@orosys.fr (Y.B.)
* Correspondence: tara.vanhatalo@u-bordeaux.fr
Abstract:
Vacuum tube amplifiers present sonic characteristics frequently coveted by musicians, that
are often due to the distinct nonlinearities of their circuits, and accurately modelling such effects can
be a challenging task. A recent rise in machine learning methods has lead to the ubiquity of neural
networks in all fields of study including virtual analog modelling. This has lead to the appearance of
a variety of architectures tailored to this task. This article aims to provide an overview of the current
state of the research in neural emulation of analog distortion circuits by first presenting preceding
methods in the field and then focusing on a complete review of the deep learning landscape that has
appeared in recent years, detailing each subclass of available architectures. This is done in order to
bring to light future possible avenues of work in this field.
Keywords:
audio signal processing; nonlinear modelling; deep learning; audio effects modelling;
virtual analog modelling; neural network; modelling nonlinear audio effects; distortion effects;
electric musical instruments
1. Introduction
The use of vacuum tubes in electronics has largely diminished since the advent of
semiconductor technologies, leading to their replacement in almost all fields except that
of music technology, where a return to and rise of vacuum tubes can be observed [
1
].
Musicians tend to prefer the sonic characteristics of vacuum tube amplifiers over those of
solid-state ones. Nevertheless, the shortcomings of vacuum tubes in the realm of guitar
amplification are numerous and include elevated cost and weight, poor durability, and
high power consumption. Their solid-state counterparts remedy some of the disadvantages
of vacuum tube amplifiers but remain less popular among musicians who often seek the
particular tone of the tube amplifiers used in iconic albums and guitar rigs.
Researchers have turned to Digital Signal Processing (DSP) methods for the emulation
of vacuum tube amplifiers in order to circumvent some of the downsides of this technology.
These DSP methods are not limited to guitar amplification or distortion and can consist of a
wide range of audio effects such as reverberation, delay, and pitch-shifting [
2
]. The existing
DSP methods for tube amplifier emulation can be divided into three categories depending
on the degree of prior knowledge of the target device used. The “white-box” approach is
historically the first and emulates each electronic component. The “black-box” approach
tries to match the output of the target device given a specific input using mathematical
functions uncorrelated with the internal components. Gray-box methods are similar to
black-box but utilize some information about the target device in order to fine-tune the
model. They comprise a block-oriented structure inspired by internal information of the
device but the simulation disregards the behaviour of each of the individual components.
These three modelling approaches will be discussed in further detail in Section 2.
Appl. Sci. 2022, 12, 5894. https://doi.org/10.3390/app12125894 https://www.mdpi.com/journal/applsci
