toa-sp 0.2.0

Multi-strategy single-pulse TOA extraction from PSRFITS search-mode data

toa_sp — Multi-Strategy Single-Pulse TOA Extraction

toa_sp extracts times of arrival (TOAs) directly from PSRFITS search-mode data, bypassing the folding step required by conventional pulsar timing pipelines. It implements nine complementary TOA estimation strategies — parametric (Gaussian, EMG, Voigt) and non-parametric (leading edge, centre of mass, peak, shapelet) — together with MCMC uncertainty estimation, automatic time-resolution optimisation, and sub-band cross-validation.

Designed for highly variable radio sources (RRATs, FRBs, mode-changing pulsars) where the integrated profile assumption breaks down.

Installation

pip install toa_sp

For MCMC support (optional):

pip install toa_sp[all]

Quick Start

toa_sp -f FRB20220529_tracking-M01_0158.fits \
    -dm 255.0 -s 1.1 -w 0.06 -bs 8 -bf 16 \
    --auto_res --compare --tim --tim_strategy best \
    -o timing_result

Or run as a script:

python toa_sp.py -f data.fits -dm 175.25 -s 2.5 -w 0.05 -bs 4 -bf 16

TOA Strategies

Strategy	Type	Description
single	Parametric	Single-component fit; TOA = μ₁
highest	Parametric	Brightest fitted component
error_weighted	Parametric	Inverse-variance weighted average
first_peak	Parametric	Earliest significant (S/N > 5) component
weighted	Parametric	Amplitude-weighted average (deprecated)
leading_edge	Non-parametric	Half-max crossing + rise-time correction
center_of_mass	Non-parametric	Flux-weighted centroid (S > 3σ_off)
peak	Non-parametric	Profile maximum, quadratic sub-bin interpolation
shapelet	Non-parametric	Hermite–Gaussian decomposition, AICc order selection

Profile Models

• Gaussian (default) — symmetric, N-component
• EMG — exponentially modified Gaussian for scattering tails
• Voigt — combined Gaussian + Lorentzian broadening
• Shapelet — flexible basis expansion, automatically regularised

Key Features

• Auto resolution optimisation (--auto_res) — maximises peak S/N via dyadic rebinning
• Sub-band cross-validation (--sub_band_fit) — tests TOA consistency across frequency
• AICc model selection (--auto_n) — data-driven choice of component number
• MCMC uncertainty (--mcmc) — robust posterior sampling via emcee
• TEMPO2 output (--tim) — writes .tim files for downstream timing analysis
• Convergence diagnostic — Δ_conv metric identifies unstable decompositions

Requirements

• Python ≥ 3.7
• numpy, scipy, astropy, matplotlib
• Optional: emcee, corner

License

MIT — see LICENSE for details.

Repository Structure

toa_sp/
├── toa_sp.py          # Main pipeline
├── pyproject.toml      # Package metadata
├── tools/
│   ├── plot_pipeline.py       # Pipeline schematic
│   └── plot_rrat_residuals.py # Residual plot generation
└── README.md