The ESP32-Synth is a 16-key polyphonic digital synthesizer built on the ESP32 microcontroller. It leverages the chip's dual-core architecture to ensure reliable, low-latency audio generation while simultaneously managing keypad input and a comprehensive Wi-Fi Web UI for sound design.
This project demonstrates real-time audio synthesis using I2S with the built-in 8-bit DAC, software-level noise mitigation, and concurrent programming (FreeRTOS).
- Polyphonic Engine: Supports up to 16 simultaneous voices (one per key) with dedicated voices for true polyphony.
-
Dual Oscillators (DCO): Two oscillators per voice (
OSC1andOSC2) with independent gain mixing. - Waveforms: Features four classic waveforms: Sine, Square, Sawtooth, and Triangle.
- ADSR Envelope: Full Attack, Decay, Sustain, and Release control, applied per voice for expressive shaping.
-
16-Key Matrix Input: Hardware interface using a
$4 \times 4$ matrix keypad with robust software debouncing. - Wi-Fi Web UI: Provides a full control interface over Wi-Fi AP for adjusting waveforms, gains, ADSR times, and musical scales.
- I2S DAC Output: Audio output via the ESP32's internal 8-bit DAC pins (GPIO 25/26), driven by the I2S peripheral.
- Noise Reduction: Includes software dithering and scaling to significantly reduce the harsh quantization clicking noise inherent to 8-bit DAC output.
To replicate this project, you will need:
| Component | Quantity | Notes |
|---|---|---|
| ESP32 Dev Board | 1 | Any standard ESP32 model (WROOM, etc.). |
| 4x4 Matrix Keypad | 1 | Used for musical input. |
| Audio Output | 1 | Simple speaker/headphones connected via an amplifier to DAC pins. |
As defined in Control.h and the I2S configuration:
| Function | Pin(s) | Notes |
|---|---|---|
| Keypad Rows | 15, 2, 4, 5 |
Output drivers. |
| Keypad Columns | 18, 19, 21, 22 |
Input with PULLUP. |
| I2S DAC Out (Left) | GPIO 25 |
Audio signal output. |
| I2S DAC Out (Right) | GPIO 26 |
Audio signal output. |
The system utilizes FreeRTOS to guarantee real-time audio performance, isolating critical functions onto separate CPU cores:
| Core | Task | Description |
|---|---|---|
| Core 1 | AudioTask |
Real-Time Synthesis: Runs the Synth::audioGeneratorLoop(). It handles sample mixing (16 voices), envelope processing, and continuous I2S buffer writing. Pinned at high priority. |
| Core 0 | loop() |
Control/UI: Manages keypad scanning, debouncing, note ON/OFF events, and all Wi-Fi Web Server client requests. |
-
Synth.h/Synth.cpp: Contains the digital signal processing (DSP) logic, includingOscillator,Envelope, and theVoiceclasses that enable polyphony. -
Control.h/Control.cpp: Handles hardware input, specifically the$4 \times 4$ matrix keypad scan and software debouncing. -
UI.h/HTML_Content.h: Manages the Wi-Fi Access Point setup and serves the custom HTML interface for remote control.
- Ensure you have the ESP32 Board Manager and the
WebServerlibrary installed in your Arduino environment. - Wire the hardware as specified in the Pinout table above.
- Open
ESP32_Synth.inoin the Arduino IDE (all files must be in the same folder). - Upload the code to your ESP32 board.
- After booting, the ESP32 creates a Wi-Fi Access Point. Check the Serial Monitor for the IP address.
- Connect a device (phone, laptop) to the network:
- SSID:
APSIT_SYNTH - Password:
12345678
- SSID:
- Open a web browser and navigate to the assigned IP address (e.g.,
192.168.4.1). - You can now control the synth parameters (Waveforms, Gain, ADSR) from the Web UI while playing notes on the physical keypad.
- Press and hold keys on the
$4 \times 4$ matrix keypad. Due to the polyphonic engine, you can press up to 16 keys simultaneously.
The audio output relies on the ESP32's built-in 8-bit DAC. The mids-frequency clicking you might hear is caused by coarse quantization error during the 16-bit to 8-bit sample conversion.
The implemented solution is a form of dithering:
- The mixed polyphonic 16-bit sample is heavily scaled down to ensure headroom (prevent clipping).
- A small pseudo-random noise value is added to the sample before the final 8-bit quantization.
This technique converts the harsh, noticeable digital clicking into a much softer, continuous background hiss or white noise, which is far less disruptive to the musical sound.
Implementation:
