npDSP

A small, composable digital signal processing library built on NumPy.

npDSP provides simple DSP blocks that can be combined into pipelines using Python's >> operator.

Installation

Install from PyPI:

pip install npDSP

Or with uv:

uv add npDSP

Quick start

import numpy as np
import npdsp

pipeline = npdsp.Add(1) >> npdsp.Multiply(2)

x = np.array([1, 2, 3])

y = pipeline(x)

print(y)
# [ 4  6  8]

Blocks can also be given names and accessed through a pipeline:

pipeline = npdsp.Add(1, name="offset") >> npdsp.Multiply(2, name="gain")

print(pipeline["gain"])

Features

• NumPy-based signal processing
• Composable processing blocks
• Pipeline construction with >>
• Stateful processing blocks such as Delay
• Mathematical operations
• Conversion and utility blocks
• Type hints
• Block profiling
• Sphinx API documentation

Documentation

Documentation is available on Read the Docs.

The documentation includes:

• Getting Started
• Concepts
• API Reference
• Block reference
• Core API reference
• Examples

Development

npDSP uses uv for development and dependency management.

Clone the repository:

git clone https://github.com/mrhiemstra/npDSP.git
cd npDSP

Install the development dependencies:

uv sync

Run the test suite:

uv run pytest

Build the documentation:

uv run sphinx-build -b html docs/source docs/build/html

Build the package:

uv build

Requirements

• Python 3.10+
• NumPy 2.2.6+
• si-prefix 1.3.3+
• typing-extensions 4.16.0+

License

npDSP is released under the MIT License.