Skip to the content
2 layers81 × 62 mmPreamp, the printed circuit board in 3D

computer engineering capstone group project

Size
81 × 62 mm
Copper layers
2
Thickness
1.6 mm
Parts
28
Nets
16
Pads
60
Tracks, vias
57, 0
Pours
1
Routed
Yes
KiCad
KiCad 10.0 (file 20260206)
Created
31 Jan 2026
Updated
20 Jun 2026

Overview

Preamp is an open-source printed circuit board by sbarry753, from the sbarry753/cse123 repository on GitHub. It is a 2-layer board of 81 × 62 mm with 28 parts on 16 nets, routed with 57 tracks and 0 vias.

Its parts include 9 resistors, 5 capacitors, 2 connectors, 1 IC and 1 diode. It belongs with the audio and music boards of this library.

README

README.md

GuSynth – Ultra-Low-Latency Polyphonic Guitar Synth Pedal

Aesthetic Prototype GuSynth is a research + hardware project building a real-time polyphonic guitar synth pedal capable of transforming a guitar into other instruments with ultra-low latency.

The project combines:

  • Transient-guided neural note detection
  • Polyphonic inference for guitar input
  • Direct neural timbre transfer
  • Embedded deployment on microcontroller hardware
  • Analog tone shaping and post-processing

The goal is a standalone guitar pedal that lets a guitarist play naturally while hearing the output as another instrument.


Overview

Most pitch detection and guitar synthesis systems rely on longer analysis windows, which adds noticeable delay. GuSynth instead focuses on extracting useful information from the earliest part of the note, especially the pick attack, so the system can react quickly enough for live performance.

The project currently explores two connected ML stages:

  1. First neural network: fast polyphonic note detection from very short windows
  2. Second neural network: direct timbre transfer from guitar to another instrument

A harmonic LUT-based pipeline and analog processing stages support the overall system.


Signal Path

This is the full signal path of the pedal, from instrument input through conversion, inference/synthesis, and output shaping.

Signal Path


Overall Pedal / System Schematic

This diagram shows the higher-level pedal architecture and how the DSP / neural / synthesis stages connect together.

Overall Schematic


424 / Analog Circuit Reference

This section documents one of the analog reference directions used in the project for tone shaping / overdrive exploration.

424 Schematic


Core System Architecture

1. First Neural Network – Polyphonic Note Detection

The first model is designed to recognize notes from very short guitar input windows, with a strong focus on the transient / pick attack region.

It is intended to:

  • detect note activity quickly
  • work under polyphonic conditions
  • provide a MIDI-style or note-based representation
  • operate under strict latency constraints suitable for embedded systems
Audio Demo – First NN

First NN Input

After (first NN output / note-detection-driven result):

First NN MIDI Out

2. Second Neural Network – Timbre Transfer

The second model takes guitar input and transforms it toward the sound of another instrument.

Current work is focused on guitar → piano-style transfer, with the long-term goal of supporting more instruments and cleaner real-time conversion.

It is intended to:

  • preserve performance timing and expression
  • reshape the raw guitar waveform into a target timbre
  • run in a streaming low-latency pipeline
  • support eventual embedded deployment
Video Demo – Second NN

Second NN Out

Harmonic LUT Support Pipeline

Alongside the neural models, the project includes a harmonic fingerprint LUT approach used for analysis and refinement.

This system helps by:

  • storing harmonic note templates
  • comparing incoming spectra to expected note structure
  • refining note hypotheses
  • supporting interpretable DSP-guided analysis

Rather than treating the project as purely neural or purely classical DSP, GuSynth explores a hybrid approach.


Embedded Hardware Target

The long-term target is a compact embedded pedal platform.

Current embedded direction

  • Daisy Seed / Cortex-M7 class hardware
  • 48 kHz real-time audio
  • small-window streaming inference
  • ADC / DAC audio pipeline
  • analog front-end and output shaping

Embedded constraints

  • low total latency
  • predictable real-time processing
  • limited memory / compute budget
  • models must be compact enough for sustained streaming

Analog Processing

The pedal is not only digital / ML based. Analog stages are also important for:

  • input conditioning
  • gain staging
  • distortion / overdrive
  • EQ shaping
  • reducing unpleasant artifacts in synthesized output

This is especially useful when blending guitar and synthetic instrument textures.


Project Status

  • ✅ First neural network prototype for transient-guided note detection
  • ✅ Second neural network prototype for timbre transfer
  • ✅ Harmonic LUT research pipeline
  • ✅ System-level pedal architecture defined
  • ⚠️ Embedded optimization still in progress
  • ⚠️ More training data and refinement needed for cleaner output quality

Repository Layout

  • Documentation/images/
    Diagrams, schematics, and system figures

  • Documentation/audio/
    WAV demos for model outputs

  • Documentation/video/
    Video demonstrations of the system

  • LUT - In Development/
    Harmonic fingerprint / template detection work

  • DAISY - In Development/
    Embedded hardware and firmware work

  • JUCE - In Development/
    Desktop or plugin prototyping


Long-Term Goal

A standalone guitar pedal that:

  • detects polyphonic playing with minimal delay
  • converts guitar into the sound of other instruments
  • runs on embedded hardware
  • combines neural models, DSP, and analog circuitry
  • remains responsive enough for real live performance

Summary

GuSynth is a hybrid guitar-synthesis system exploring how far real-time note inference and timbre transfer can be pushed under embedded, performance-critical latency constraints.

It is not just a guitar effect — it is an attempt to turn the guitar into a real-time expressive controller for other instruments.

Parts

28
QtyValueFootprintReferences
4TP_GNDTestPoint:TestPoint_Loop_D3.50mm_Drill0.9mm_BeadedTP1, TP2, TP3, TP4
2100kResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR8, R9
2AudioJack2_SwitchConnector_Audio:Jack_6.35mm_Neutrik_NMJ4HCD2_HorizontalJ1, J2
10.1uFCapacitor_THT:CP_Radial_D8.0mm_P3.80mmC1
110MResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR2
110kResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR7
115uFCapacitor_THT:CP_Radial_D8.0mm_P3.80mmC4
11N4007Diode_THT:D_DO-41_SOD81_P10.16mm_HorizontalD1
11kResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR3
11uFCapacitor_THT:CP_Radial_D8.0mm_P3.80mmC5
12.7kResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR5
122MResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR1
14.7uFCapacitor_THT:CP_Radial_D8.0mm_P3.80mmC3
1470Resistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR6
147pFCapacitor_THT:CP_Radial_D8.0mm_P3.80mmC2
156kResistor_THT:R_Axial_DIN0414_L11.9mm_D4.5mm_P15.24mm_HorizontalR4
1BatteryConnector_PinSocket_2.54mm:PinSocket_1x02_P2.54mm_VerticalBT1
1C500kPotentiometer_THT:Potentiometer_Alpha_RD901F-40-00D_Single_VerticalRV1
1LOGOLOGOG***
1TL071Package_DIP:DIP-8_W7.62mm_LongPadsU1
1TP_4.5VTestPoint:TestPoint_Loop_D3.50mm_Drill0.9mm_BeadedTP7
1TP_POSTAMPTestPoint:TestPoint_Loop_D3.50mm_Drill0.9mm_BeadedTP6
1TP_PREAMPTestPoint:TestPoint_Loop_D3.50mm_Drill0.9mm_BeadedTP5

More from sbarry753/cse123

2
  • 2 layers100 × 69 mmAnalog, the printed circuit board in 3D

    Analog

    sbarry753/cse123

    computer engineering capstone group project

    40 parts3
  • 2 layersWamp overdrive, the printed circuit board in 3D

    Wamp overdrive

    sbarry753/cse123

    computer engineering capstone group project

    21 parts3

Similar boards

  • 2 layers44.5 × 72.7 mmOSO SWAT C1-jig, the printed circuit board in 3D

    OSO SWAT C1-jig

    joeycastillo/Sensor-Watch

    A board replacement for the classic Casio F-91W wristwatch

    28 parts2.1k
  • 2 layers95 × 54 mmEinkPDA screen, the printed circuit board in 3D

    EinkPDA screen

    TalismanDesign/PocketMage_PDA

    An E-Ink PDA device using the ESP32 S3.

    28 parts2kApache-2.0
  • 2 layers45 × 45 mmSmartbuttons, the printed circuit board in 3D

    Smartbuttons

    hackclub/OnBoard

    💡 Join 1,000 teenagers and make your first PCB with a $100 grant!

    28 parts915
  • 2 layers24.1 × 36.2 mmEsp32miniBT, the printed circuit board in 3D

    Esp32miniBT

    asterics/esp32_mouse_keyboard

    ESP32 implementation for HID over GATT Keyboard and Mouse (Bluetooth Low Energy). Including serial API for external modules (similar to Adafruit EZKey HID)

    28 parts911GPL-3.0
  • 2 layers34.8 × 12.2 mmInfnoise goldfingers, the printed circuit board in 3D

    Infnoise goldfingers

    waywardgeek/infnoise

    The world's easiest TRNG to get right

    28 parts854CC0-1.0
  • 2 layers55.9 × 99.1 mmStm32f0 disco Luos shield, the printed circuit board in 3D

    Stm32f0 disco Luos shield

    Luos-io/luos_engine

    Open-source and real-time orchestrator for cyber-physical-systems, to easily design, test and deploy embedded applications and digital twins.

    28 parts546MIT
  • 2 layers88 × 25 mmEleLab v2 Frontpanel a, the printed circuit board in 3D

    EleLab v2 Frontpanel a

    Chrismettal/EleLab_v2

    Cheap DIY lab system for hobbyists and the tinkering folk

    28 parts531GPL-3.0

Questions

How big is the Preamp PCB?

It measures 81 × 62 mm and has 2 copper layers, 1.6 mm thick.

What parts does the Preamp use?

28 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.

Where can I download the Preamp KiCad files?

From the sbarry753/cse123 repository on GitHub, at the commit this page was made from: the board is Analog/preamp/preamp/preamp.kicad_pcb and its schematic Analog/424_overdrive_eq/424_overdrive_eq.kicad_sch. PCBKit opens the board in the browser, with nothing to install.

Can I use the Preamp design in my own project?

Its repository states no licence, so its author keeps all rights: ask sbarry753 before reusing the design.