Preview

Soundboard Analysis: Variables for Automation in Lutherie is a 338-page technical study exploring soundboard behavior, measurement, and automation in modern guitar manufacturing. The book includes 88 supporting figures and a full observational study of 69 soundboards. Structured like an academic research publication, the book is organized into four primary sections: Introduction, Methods, Results, and Discussion. It includes statistical analysis, experimental methodology, and machine-learning frameworks applied to guitar soundboard design.

PART I — INTRODUCTION

Chapter 1 — About the Book
Background, survey study, and the conceptual development that led to this research.

Chapter 2 — Observational Study
Preparation and framework for the observational study of guitar soundboards.

PART II — METHODS

Chapter 1 — Research
Research order of operations, raw data collection, conversions, and scatterplot preparation.

Chapter 2 — Preparation & Overview of Brace Data Collection
Evaluation of the deflection jig, brace descriptions, and measurement considerations.

Chapter 3 — Collecting Rondak-Series Brace Data
Measurement methods, calculated variables, and water displacement testing.

Chapter 4 — Collecting Rondak-Series Soundboard Data
Measured and calculated soundboard features, velocity-of-sound testing, and documentation.

PART III — RESULTS

Chapter 1 — Summary

Chapter 2 — Sample Size

Chapter 3 — Volume Analysis
Bridge plate volume, brace volume, and soundboard volume analysis.

Chapter 4 — Statistical Analysis
Raw data analysis, brace resistance, Young’s modulus, velocity of sound, scatterplots, and distribution analysis.

Chapter 5 — Mini-Experiment
Hypothesis, experimental setup, dreadnought and orchestra model results, and analysis.

PART IV — DISCUSSION

Chapter 1 — Key Takeaways

Chapter 2 — Addressing Initial Questions

Chapter 3 — Areas for Improvement

Chapter 4 — If I Could Turn Back Time…

Chapter 5 — Variables for Automation
Input variables, output variables, and modal analysis methods.

Chapter 6 — Brace Volume Life Cycle of Rondak-Series Soundboards

Chapter 7 — Machine Learning Training Data Sheet

Chapter 8 — Machine Learning in Manufacturing

Chapter 9 — Machine Learning in Lutherie

Chapter 10 — Voicing Soundboards with Automation
The AVS-1411 Automated Voicing System.

Chapter 11 — Final Thoughts for Future Research

Chapter 12 — Conclusion

Table of Contents (abridged)

Selected Excerpts

Part I: Introduction

Chapter 1: About the Book

This book is a behind-the-scenes look at my personal journey in the world of professional lutherie. I will share a detailed narrative of my notes and observations gathered during my tenure at the prestigious boutique acoustic guitar manufacturer, Ampliùs Guitars. My role as a full-time luthier allowed me to delve into the intricacies of our production-line soundboards, the 'Rondak-Series,' and this book is the result of my research.

I will conduct a thorough analysis of my observational study of 69 soundboards, perform a small experiment using a Lucchi Meter, and discuss the insights gained from my research. Finally, I will introduce my idea to fully automate the 'voicing' process, a concept that could revolutionize acoustic guitar manufacturing. The content in this book may be dense at times, but I encourage you to stick with it as we explore the world of acoustic guitar manufacturing together.

1.1 – Background

After my first day in the brace milling department, I was convinced I would do anything I could to get taken off the task. It was horrible… But then something extraordinary happened… after a few months of brace milling, I began to enjoy the task. Brace day, for me, was a form of therapy. Because the process was repetitive, I could perform the milling operation while letting my thoughts wander freely. There was something so meditative about the hum of the table router and the clatter of wood chips getting sucked away into the dust collector.

Over time, I began to see myself as a crucial part of the guitar-making process. If carving soundboard braces, 'voicing' is the luthier's final attempt to shape the instrument's tone before boxing it up; then the selection of quality braces must also make a difference! I began to wonder if there was anything I could do to improve our instruments by picking the 'right' braces to use on our soundboards.

Chapter 2: Observational Study

Suppose I were to document different physical and behavioral features of soundboards. Would there be any correlative relationship between these features that could give us insight into how soundboards function?

Part III: Results

Chapter 1: Summary

Between July 2022 and April 2023, I conducted an observational study on 69 randomly sampled Ampliùs Guitar Rondak-Series soundboards. This included 37 dreadnought and 32 orchestra model soundboards. I collected data through a series of thorough observations, using designated tools and devices to measure soundboard mass, deflection, thickness, brace stiffness, moisture content, and the velocity of sound. All measurements were directly entered into spreadsheets known as the "Rondak-Series Dreadnought" and "Rondak-Series Orchestra Model data sheet(s)", which were programmed to convert units and calculate values such as Young's modulus and descriptive statistics for analysis. I later used scatterplots, frequency distribution tables, and bell curves to further analyze the collected data.

Part IV: Discussion

Chapter 5: Variables for Automation

I will now present my vision for a machine learning application used for soundboard construction. Imagine a process where we collect data and use it to train a machine learning model. This model, once developed, will be able to interpret key input variables and accurately map them to the output variables required to produce a quality-sounding instrument. In theory, these output variables will be predictable and can be used to program an AI-generated robotic tool path for a fully automated soundboard voicing system. This application has the potential to revolutionize soundboard construction, paving the way for a future of unprecedented quality and efficiency. I hope this vision inspires you as much as it does me.

Chapter 10: Voicing Soundboards with Automation

Once your ML model is trained and ready to deploy, the second half of this process involves robotic routing. Thus far, we have reviewed methods for collecting input, output, and secondary output data, and we have gone over different types of ML models and how to implement them. In this section, I will illustrate my designs for a device capable of physically collecting the majority of the data in a designated location, with the exception of velocity of sound and thickness. This device is called the AVS-1411 or the Automated Voicing System. The AVS-1411 is designed to collect data and fully automate the voicing process of a guitar’s soundboard using machine learning, a robotic arm, and precision sensors integrated with modal analysis and deflection measurement systems.