Acoustic Metadata Multiplexer

A low-cost, low-power system for embedding time-synchronized FSK metadata signals into audio recordings made by Autonomous Recording Units (ARUs). Designed for bioacoustic field deployments where cross-unit time synchronization and metadata recovery from audio alone are required.

The system encodes a structured metadata payload as a BFSK signal mixed into the audio recorded by any ARU model — either acoustically via a speaker, or electrically via a 3.5 mm cabled pass-through. A desktop demodulator extracts the metadata from recorded .wav files post-deployment.

---

How It Works

The system operates in a star topology: one unit acts as a base station and the others are receivers, one per ARU.

1. Base station wakes on a fixed schedule, generates a common payload (message ID, timestamp, user string), and broadcasts it over RF at 433 MHz using the CC1101 transceiver.
2. Receiver units listen on the CC1101 in continuous RX mode. On receiving a packet, each receiver generates a BFSK-modulated audio signal encoding the shared payload combined with its own local data (device ID, GPS coordinates, live temperature from DS3231).
3. The FSK signal is mixed into the audio captured by the ARU — either by playing it through a small speaker near the ARU microphone, or by mixing it into a microphone signal over a cabled 3.5 mm connection.
4. Since all receivers embed at the same time (triggered by the same RF packet), the embedded signal appears at the same absolute moment across all ARU recordings — enabling post-hoc time alignment via the message ID.
5. A recorded .wav containing the embedded signal can be decoded using the desktop demodulator to recover both the metadata and the synchronization reference point.

Standalone mode skips RF communication entirely: the unit generates and embeds its own payload on a fixed schedule, using only locally configured and measured data.

The metadata payload format is:

/MID<id>/TIM<timestamp>/STR<string>/DID<device_id>/LOC<lat,lon>/TMP<temperature>/

Each field is optional and individually togglable via user_config.h or the GUI. MID (message ID) is the synchronization anchor — recordings are aligned by matching identical message IDs across units.

---

Repository Structure

Audio_Watermark_Generator/
├── Firmware/                   # STM32G431 embedded firmware
│   ├── Core/
│   │   ├── Inc/                # Headers (drivers, config, DSP)
│   │   └── Src/                # Source (main loop, peripherals, FEC)
│   ├── Drivers/                # HAL, CMSIS, and DSP libraries
│   ├── Middlewares/            # USB device library
│   ├── USB_Device/             # USB CDC interface
│   ├── cmake/                  # CMake toolchain config
│   ├── CMakeLists.txt
│   └── CMakePresets.json
│
├── Hardware/                   # KiCad PCB design
│   ├── *.kicad_sch             # Schematics (MCU, RTC, mixer, speaker, transceiver, battery)
│   ├── *.kicad_pcb             # PCB layout
│   ├── Components/             # Custom footprints, symbols, and 3D models
│   ├── Gerber/                 # Fabrication files
│   └── JCLPCB/                 # BOM and CPL for JLCPCB assembly
│
├── Software/                   # Desktop tools (Python)
│   ├── gui/                    # Configurator and flash tool
│   │   ├── gui.py              # Main GUI application
│   │   ├── build_app.py        # PyInstaller build script
│   │   ├── scripts/            # Config generation, build automation, paths
│   │   └── tools/              # Bundled toolchain (ARM GCC, CMake, Ninja, dfu-util)
│   │
│   └── gui_demodulator/        # Metadata decoder
│       ├── demodulator_gui.py  # Decoder GUI
│       ├── demodulator.py      # FSK demodulation engine
│       └── reed_solomon.py     # Reed-Solomon decoder
│
├── Testing/                    # Test scripts and data
│   ├── scripts/                # Individual test scripts
│   │   └── helper/             # Shared utilities
│   ├── test_files/             # Input audio and measurement data
│   ├── results/                # Output plots and reports
│   └── run_all_tests.py        # Run all tests
│
└── README.md

---

Getting Started

Prerequisites

Component	Requirement
Microcontroller	STM32G431-based board (custom PCB or Nucleo)
Toolchain	GNU Arm Embedded Toolchain 13.2+
Build system	CMake + Ninja
Programmer	ST-Link V3 or DFU over USB
Python	3.8+ with tkinter

Install Python Dependencies

pip install numpy scipy matplotlib soundfile

tkinter is included with most Python installations. On Linux, you may need sudo apt install python3-tk.

---

Firmware

Key Configuration

All metadata and system parameters are defined in Firmware/Core/Inc/user_config.h:

Parameter	Default	Description
USER_STRING	"Hello World"	User-defined string payload
DEVICE_ID	42	Numeric identifier for this unit
LOCATION	"63.4190,10.4015"	GPS coordinates of this ARU
TEMPERATURE	20	Static temperature fallback (overridden by DS3231 reading at runtime)
INCLUDE_USER_STRING / INCLUDE_DEVICE_ID / etc.	true	Toggle individual payload fields on/off
FSK_LOWER_FREQUENCY	10000 Hz	FSK f0 tone (auto-aligned to DAC sample rate)
FSK_HIGHER_FREQUENCY	12000 Hz	FSK f1 tone (auto-aligned to DAC sample rate)
SIGNAL_ATTENUATION	100	Output level (0 = silence, 100 = full)
USE_REED_SOLOMON_ERROR_CORRECTION	true	Enable/disable RS error correction
RS_ERROR_CORRECTION_SYMBOLS	20	Number of RS parity symbols
OPERATION_MODE	0	0 = Receiver, 1 = Base Station, 2 = Standalone
STARTING_HOUR / END_HOUR	0 / 23	Daily schedule window (24h format)
RUN_MINUTES / SLEEP_MINUTES	3 / 0	Duty cycle within the schedule window
USE_CABLE_TRANSMISSION	false	Cable (3.5 mm) output instead of speaker

You can edit this file manually or use the GUI configurator (see below).

Operation Modes

Receiver (OPERATION_MODE=0) Listens on the CC1101 in continuous RX mode. On receiving a packet, the unit generates a metadata string combining shared fields from the RF packet (message ID, timestamp, user string) with its own local fields (device ID, location, temperature), then outputs the BFSK signal. One receiver unit is deployed at each ARU.

Base Station (OPERATION_MODE=1) Wakes on a fixed RTC-based schedule, generates a common payload (message ID, RTC timestamp, user string), and transmits it via the CC1101 RF transceiver. One unit per deployment must be configured as the base station.

Standalone (OPERATION_MODE=2) No RF communication. The unit wakes on a fixed RTC-based schedule, generates a payload from locally configured and measured values, and outputs the BFSK signal. Used when cross-unit synchronization is not needed.

Features

• BFSK metadata embedding — DMA-driven circular DAC output with pre-computed sine LUTs; frequencies are auto-aligned to integer samples-per-period for zero spectral leakage
• Reed-Solomon FEC — GF(2^8) encoder with configurable parity symbols for burst-error-resilient decoding
• CC1101 sub-GHz RF — bit-banged SPI driver for 433 MHz base station to receiver synchronization; receiver uses continuous RX mode
• DS3231 RTC — I2C temperature-compensated clock with battery backup; software-controlled power gating to save current between reads
• Scheduled operation — configurable active hours (STARTING_HOUR/END_HOUR) and run/sleep duty cycle aligned to AudioMoth duty-cycle timing; RTC wakeup timer for Base Station/Standalone, GPIO interrupt (GDO0) for Receiver (mode 0)
• Audio output paths — speaker (GPIO-switched amplifier) or cable (relay-switched gain-neutral mixing mode with op-amp signal chain)
• Battery monitoring — ADC reading through a resistor divider; halts with error code if voltage drops below 3.0 V
• USB CDC — serial interface for host communication and first-boot RTC time sync
• Flash-persistent flags — records whether time sync has been performed (survives power cycles)
• LED status codes — 4-bit binary blink patterns encoding error/status codes, repeated 3 times
• STOP1 low-power mode — full peripheral suspend between transmissions; clock and I2C re-initialized on wake

Build

cd Firmware
cmake -B build/Debug -S . -G Ninja -DCMAKE_BUILD_TYPE=Debug
cmake --build build/Debug

Flash

Via ST-Link:

openocd -f interface/stlink.cfg -f target/stm32g4x.cfg \
  -c "program build/Debug/Audio_Watermark_Generator.elf verify reset exit"

Via DFU (USB): The GUI flash tool handles DFU flashing automatically. Alternatively, use dfu-util directly.

---

Software Tools

Configurator & Flash Tool

A desktop GUI for configuring metadata parameters and flashing the firmware to the STM32.

cd Software/gui
python3 gui.py

What it does:

• Configure metadata fields per operation mode (base station, receiver, standalone)
• Set FSK frequencies, attenuation, and Reed-Solomon error correction
• Set operation schedule (active hours, run/sleep intervals)
• Generate user_config.h, build the firmware, and flash over DFU

The tool bundles its own ARM toolchain, CMake, Ninja, and dfu-util in Software/gui/tools/, so no manual toolchain setup is needed when using the GUI.

Metadata Decoder

A desktop GUI for extracting embedded metadata from recorded audio files.

cd Software/gui_demodulator
python3 demodulator_gui.py

What it does:

• Load a .wav recording containing an embedded BFSK signal
• Perform FSK demodulation and Reed-Solomon decoding
• Display the decoded metadata payload
• Visualize the demodulated spectrum and signal

Pre-built macOS Apps

Beta .app builds are available for macOS:

• Audio Watermark Flash Tool — configure and flash
• FSK Audio Demodulator — decode metadata from recordings

Download the latest release

---

Testing

The Testing/ directory contains scripts for characterizing and validating system performance.

Test Scripts

Script	Description
autotest.py	Automated demodulator correctness testing
drift.py	Frequency drift measurement over time
ecc_burst_test.py	Reed-Solomon burst error correction validation
offset_calculation.py	Timing offset between devices
localize.py	Acoustic localization from multi-receiver recordings
localize_tdoa.py	Time-difference-of-arrival (TDOA) localization
audio_passthrough.py	Audio signal chain validation (FFT analysis)
spec_leak.py	Spectral purity and leakage analysis
timing.py	Transmission timing validation
wakeup.py	Sleep/wakeup mechanism testing
plot_fsk_signals.py	FSK signal visualization
plot_bird_spectrograms.py	Spectrogram analysis of bird audio
plot_current.py	Current consumption analysis
plot_rssi.py	RF signal strength plotting
plot_temperature.py	Temperature variation analysis

Test Data

Test input data is organized in Testing/test_files/:

• demodulator_autotest/ — Pre-recorded samples for demodulator validation
• acoustic_localization/ — Multi-receiver recordings for TDOA testing
• rtc_measurements/ — RTC drift data
• range_tests/ — Distance-dependent RF signal quality
• current_measures/ — Power consumption logs
• temperature_range/ — Temperature sweep recordings

Run All Tests

cd Testing
python3 run_all_tests.py

---

Hardware

The custom 4-layer PCB is designed in KiCad and includes:

• STM32G431 microcontroller
• DS3231 RTC with battery backup
• CC1101 RF transceiver (433 MHz, sub-GHz)
• Audio mixer with gain-neutral pass-through and speaker driver
• USB connector
• Battery power management

Design files are in Hardware/:

• Schematics: individual .kicad_sch files for each subsystem
• PCB layout: Audio_watermark_generator.kicad_pcb
• 3D model: Audio_watermark_generator.step
• Fabrication: Gerber/ directory with Gerber + drill files
• Assembly: JCLPCB/ with BOM and CPL for JLCPCB ordering
• Custom components: Components/ with footprints, symbols, and 3D models

---

Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/my-feature)
3. Commit your changes
4. Push and open a pull request

---

Contributors

• Kristian Gøystdal — Developer
• Einar Bergslid — Developer

---

License

Released under the MIT License. Free to use, modify, and distribute.

---

stm32 · bioacoustics · autonomous-recording-unit · time-synchronization · fsk · metadata-embedding · reed-solomon · kicad · python