rfproto

Python library for RF and SDR prototyping. Provides reusable methods for RF measurements, communication systems, radar, antenna arrays, and related signal processing tasks. The library supports both floating-point and fixed-point arithmetic for modeling hardware implementations (FPGA/ASIC).

Features & Modules

The library is organized into focused modules for different aspects of RF/SDR signal processing:

Core Signal Processing

• nco.py: Numerically Controlled Oscillator (NCO) with both floating-point and fixed-point support, including frequency-to-FCW conversion and Taylor series approximations for sine/cosine generation
• filter.py: Filter design including Raised Cosine, Root Raised Cosine, and polyphase filtering
• multirate.py: Multirate signal processing (interpolation, decimation, rational resampling with polyphase filters)
• sig_gen.py: Signal generation utilities for test waveforms and modulated signals
• modulation.py: Modulation/demodulation schemes for digital communications

Measurements & Analysis

• measurements.py: Signal quality metrics (EVM, PSD, SFDR, SNR, and spectral analysis)
• plot.py: Visualization utilities (time-domain, frequency-domain, constellation diagrams, eye diagrams)
• impairments.py: Channel impairment modeling (noise, phase offset, timing errors)

Fixed-Point Arithmetic

• fxp_int.py: Fixed-point integer class modeling hardware bit-width constraints with overflow/wrapping behavior
• utils.py: Fixed-point/floating-point conversion, binary representation, rounding methods, and dB/magnitude conversions

Control & Processing

• agc_magest.py: Automatic Gain Control (AGC) using magnitude estimation
• pi_filter.py: Proportional-Integral filter implementations
• signal.py: Basic signal operations and transformations
• frequency.py: Frequency-related utilities and conversions

Architecture

Fixed-Point Arithmetic Support

A key feature of rfproto is dual floating-point and fixed-point implementations for hardware prototyping:

1. fxp_int class: Models fixed bit-width integers with 2's complement overflow/wrapping behavior
2. Conversion utilities: dbl_to_fxp() / fxp_to_dbl() for converting between representations
3. Rounding operations: fxp_round_halfup() / fxp_truncate() for proper fixed-point arithmetic
4. Q-format notation: Uses standard Q-format (e.g., Q0.15 for 16-bit signed with 15 fractional bits)
5. Switchable implementations: Some modules like nco.py have use_fxp flags to toggle between floating-point and fixed-point

Design Patterns

• Normalized filters: Filter functions return normalized coefficients (sum to 1.0 for unity passband gain)
• Rational resampling: Use multirate.get_rational_resampling_factors() for optimal interpolation/decimation ratios
• Type hints: All functions include type hints for parameters and return values
• Doctests: Inline documentation examples that serve as both docs and tests

Developing

Building & CI

• Install editable local version (preferably within a venv) with all optional packages for testing with $ pip install --upgrade -e .[docs,test] (add --user if not in venv).
• Install pre-commit checks with $ ln -sf ../../scripts/pre-commit ./.git/hooks/pre-commit
• Trigger GitHub action to publish to PyPI with a tagged commit (e.x. git tag -am "test auto versioning" 0.0.2) on main branch. Note versioning is also inferred from the git tag value, and this will only run on push on tag.

Testing

Run test suite with $ ./scripts/run-tests.sh

Additional testing options:

# Run tests with plots displayed (default suppresses plots)
./scripts/run-tests.sh --show_plots

# Run specific test file
pytest tests/test_filter.py

# Run with coverage report
pytest --doctest-modules --cov-report=term-missing:skip-covered --cov=rfproto/ tests/

# Run mypy static type checking
mypy rfproto/

The test suite includes:

• pytest with doctest support for inline documentation examples
• Tests in both tests/ directory and doctests embedded in module docstrings
• Coverage reporting via pytest-cov
• Full mypy type checking (enforced in CI)
• Pre-commit hook that runs full test suite and auto-generates TODO.md

Documentation

Documentation uses mkdocs-material, preview with $ mkdocs serve -a localhost:8888. Publishes with GitHub action as well.

Code Formatting

The project uses black for code formatting. The pre-commit hook automatically runs the full test suite (including formatting checks) before commits are allowed.

TODO

See auto-generated TODO.md.

• Look at https://github.com/veeresht/CommPy since its unmaintained
• Filtering/convolution kernels like https://joht.github.io/johtizen/algorithm/2022/10/22/a-different-approach-to-convolution.html and https://dsp.stackexchange.com/questions/15412/fir-filters-direct-form-transposed-fir