C Port of RADE V1

This repo contains a pure C implementaion of RADE V1, with all Python code and dependencies removed. It was derived from the reference Python implementation with the asistance of Claude Code. It has been reviewed, and carefully tested by the FreeDV team. It passes the same suite of automated tests as the Python version.

This has been tested on Linux and macOS.

Build

cd radae_nopy
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j$(nproc) # or -j$(sysctl -n hw.logicalcpu) on macOS

Demo tools

RADE Demod: WAV RADE → WAV Speech Audio

Take a wav file off air and produce a demodulated wav file

Usage:
rade_demod_wav [-v 0|1|2] <input.wav> <output.wav>

RADE Modulate: WAV Speech Audio → WAV RADE

Take a wav file with speech in it and produce a RADE OFDM encoded output wav file ready for transmission.

Usage:
rade_modulate_wav [-v 0|1|2] <intput.wav> <output.wav>

Encode: WAV → IQ

sox ../voice.wav -r 16000 -t .s16 -c 1 - | \
  ./src/lpcnet_demo -features /dev/stdin - | \
  ./src/radae_tx > tx.iq

Decode: IQ → WAV

cat tx.iq | \
  ./src/radae_rx | \
  ./src/lpcnet_demo -fargan-synthesis /dev/stdin - | \
  sox -t .s16 -r 16000 -c 1 - decoded.wav

Decode: WAV RADE → WAV (multiple steps)

sox ../FDV_offair.wav -r 8000 -e float -b 32 -c 1 -t raw - | \
./src/real2iq | \
./src/radae_rx > features.f32
./src/lpcnet_demo -fargan-synthesis features.f32 - | \
sox -t .s16 -r 16000 -c 1 - decoded.wav
play decoded.wav

Files

File	Purpose
src/rade_dsp.h/c	Complex math utilities, constants, pilot generation
src/rade_ofdm.h/c	OFDM modulation/demodulation with DFT matrices
src/rade_bpf.h/c	Complex bandpass filter
src/rade_tx.h/c	Transmitter (encoder + OFDM mod)
src/rade_acq.h/c	Acquisition and pilot detection
src/rade_rx.h/c	Receiver with sync state machine
src/rade_api_nopy.c	Python-free API implementation
src/radae_tx_nopy.c / src/radae_rx_nopy.c	Standalone executables

radae_nopy/
├── CMakeLists.txt
├── cmake/
│   └── BuildOpus.cmake
└── src/
    ├── CMakeLists.txt
    ├── lpcnet_demo.c          # Feature extraction
    ├── radae_tx_nopy.c        # Transmitter
    ├── radae_rx_nopy.c        # Receiver
    ├── real2iq.c              # Real → IQ converter
    ├── rade_api.h             # Public API
    ├── rade_api_nopy.c        # API implementation
    ├── rade_dsp.h/c           # DSP primitives
    ├── rade_ofdm.h/c          # OFDM mod/demod
    ├── rade_bpf.h/c           # Bandpass filter
    ├── rade_tx.h/c            # Transmitter internals
    ├── rade_rx.h/c            # Receiver internals
    ├── rade_acq.h/c           # Acquisition
    ├── rade_enc.h/c           # Neural encoder
    ├── rade_dec.h/c           # Neural decoder
    ├── rade_enc_data.c        # Encoder weights
    ├── rade_dec_data.c        # Decoder weights
    └── opus-nnet.h.diff       # Opus patch

The implementation uses built-in neural network weights (compiled from rade_enc_data.c and rade_dec_data.c), eliminating any need for Python or external model files at runtime.

Automated testing

A suite of tests will be run on every GitHub push. They can also be run locally.