v0.11.0-alpha

[v0.11.0-alpha] - 2026-10-15

Changes since v0.10.0-alpha.

Added

• Receiver DSP: feedforward (FFE), decision feedback (DFE) and Volterra equalizers (up to cubic order) for IM-DD links, with samples-per-symbol support, training mode and custom initial coefficients; adaptive filtering cores (NLMS, DD-LMS, CMA, RDE, DA-RDE, RLS, DD-RLS) with time- and frequency-domain MIMO equalization; Levinson recursion, autocorrelation and whitening filter; optic.dsp.synchronization module with syncDataSequences; frequency shifting (freqShift) and M-th power frequency offset estimation.
• Forward error correction: LDPC encoding and decoding (edge-based message passing, DVB-S2 codes, ALIST files), Hamming codes, extrinsic LLR metric and MLSE (Viterbi) routine.
• Devices and models: DAC model with quantization, ENOB and jitter; ENOB, clipping and noise model in the ADC; extinction ratio support in the MZM and IQ modulator; IQ imbalance and skew (iqMixing) and frontend parameters in the coherent receiver; NLIN perturbation models rewritten with Numba kernels.
• Metrics: theoretical GMI and mutual information (MI) support.
• Plotting: customizable eye diagram and power spectral density plots; figsize option in constellation plots; comprehensive plotting tests.
• Tests and CI: pytest-based test suite and GitHub Actions workflow running it on Python 3.10-3.13.
• Packaging: GitHub Actions workflow that publishes the package to PyPI (Trusted Publishing) when a release is published; the package version is now derived from the Git tag (setuptools-scm).
• CITATION.cff metadata and a revamped README with full project overview.

Changed

• Python 3.10 or newer is required.
• Numba caching and fastmath enabled in performance-critical DSP, FEC and model routines; FIR filtering uses fftconvolve.
• Renamed parameters for consistency (Ei/Eo to sigIn/sigOut, outEq to yEq, Ipd_sat to IpdSat, signalRotation to polRotation); awgn now takes a parameters object.
• Docstrings expanded with equations and theory notes across the optic package, now with a Returns section for all documented public functions; Sphinx documentation refreshed and optic.dsp.coreGPU added to it.
• Modern NumPy/SciPy/Matplotlib APIs replace deprecated ones.

Fixed

• AWGN noise variance for real-valued signals.
• Sign handling of LLRs for bit probabilities.
• ber2Qfactor dB conversion formula, extinction ratio dB-to-linear conversion in the MZM, and ADC anti-alias filter frequency inputs.
• PRBS-13 taps, CAZAC sequence formula and RRC filter tap scaling with the symbol period.
• Viterbi & Viterbi carrier phase recovery for PSK/APSK constellations.
• Photodiode noise seeds are now independent across receiver channels; PSD labels in single-mode plots.

Other Changes

• Fix docstring inconsistencies across optic folder (function parameters and param object attributes) by @edsonportosilva with @Copilot in #32
• Rework the test suite and add a CI workflow by @Hewei603 in #35
• Add PyPI publish workflow and tag-based versioning by @edsonportosilva in #37

New Contributors

• @edsonportosilva with @Copilot made their first contribution in #32
• @Hewei603 made their first contribution in #35

Full Changelog: v0.10.0-alpha...v0.11.0-alpha