clear-dsp

A lightweight JAX-optimized package of pure functions for digital signal processing.

clear-dsp provides a clean, functional, and high-performance set of DSP primitives built on JAX. Every function is pure, JIT-compatible, and designed for composition, autodiff, and acceleration on CPU, GPU, or TPU.

Design Principles

• Pure functions only — no classes, no mutable state, no side effects
• JAX-native — fully compatible with jax.jit, jax.vmap, jax.grad, and jax.pmap
• Lightweight — minimal dependencies (just JAX)
• Composable — small, focused building blocks that combine cleanly
• Type-safe — full type annotations and mypy-friendly

Installation

pip install clear-dsp

Or for development:

git clone https://github.com/yourusername/clear-dsp.git
cd clear-dsp
pip install -e ".[dev]"

Quick Start

import jax.numpy as jnp
from clear_dsp import filters, spectral

# Example: simple low-pass FIR filter (placeholder API)
# signal = jnp.sin(jnp.linspace(0, 10 * jnp.pi, 1000))
# filtered = filters.fir_lowpass(signal, cutoff=0.1, num_taps=51)

Note: The API is under active development. Check the source and tests for currently implemented functions.

Project Structure

clear-dsp/
├── src/
│   └── clear_dsp/
│       ├── __init__.py
│       ├── filters.py      # FIR/IIR filters, windowing
│       ├── spectral.py     # FFT-related, spectrograms, STFT
│       ├── windows.py      # Window functions
│       └── ...
├── tests/
├── pyproject.toml
└── README.md

Development

# Install with development dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Lint
ruff check src tests

# Type check
mypy src

Roadmap (initial ideas)

• Window functions (Hann, Hamming, Blackman, Kaiser, etc.)
• FIR filter design and application
• Basic IIR filters (biquads)
• STFT / ISTFT
• Spectral analysis helpers
• Convolution and correlation
• Resampling
• Envelope detection / Hilbert transform
• Signal analysis

Contributions and suggestions are welcome!

License

MIT License — see LICENSE.