evo-gw-matched-filter-engine

Core GW data processing and matched filtering engine for PyCBC-based gravitational wave detection.

Functions

read_and_condition_data(gwf_path, channel)

Reads a GWF frame file and conditions the data:

1. Read entire frame (no time bounds needed)
2. Highpass filter at 15 Hz
3. Crop 2s from both ends (FIR wraparound removal)
4. Resample to 2048 Hz

estimate_psd(conditioned_data)

Estimates PSD using Welch's method (4s segments), interpolates to data resolution, applies inverse spectrum truncation (4s max filter, 15 Hz cutoff).

generate_template(approximant, mass1, mass2, delta_t, f_lower=20.0)

Generates a waveform template using get_td_waveform. Returns hp (plus polarization). Supported approximants: SEOBNRv4_opt, IMRPhenomD, TaylorT4.

compute_matched_filter_snr(conditioned_data, psd, hp, f_lower=20.0)

Performs matched filtering:

1. Resizes template to data length
2. Cyclic shifts template (hp.start_time)
3. Runs matched_filter with PSD weighting
4. Takes abs() of complex SNR
5. Crops edges (8s start, 4s end)
6. Returns (peak_snr, peak_time)

find_peak_snr(gwf_path, channel, mass1, mass2, approximant, conditioned_data=None, psd=None)

High-level function combining all steps. Accepts pre-conditioned data/PSD to avoid recomputation.

Usage

import sys
sys.path.insert(0, '/app/environment/skills/evo-gw-matched-filter-engine/scripts')
from utils import read_and_condition_data, estimate_psd, generate_template, compute_matched_filter_snr, find_peak_snr

# Option 1: All-in-one
snr, time = find_peak_snr('data.gwf', 'H1:TEST-STRAIN', 30, 30, 'SEOBNRv4_opt')

# Option 2: Reuse conditioned data across many templates
data = read_and_condition_data('data.gwf', 'H1:TEST-STRAIN')
psd = estimate_psd(data)
hp = generate_template('SEOBNRv4_opt', 30, 30, data.delta_t)
snr, time = compute_matched_filter_snr(data, psd, hp)

Key Parameters

• Highpass cutoff: 15 Hz
• Resample rate: 2048 Hz
• Data crop: 2s both ends
• PSD Welch segment: 4s
• Inverse spectrum truncation: 4s max filter, 15 Hz cutoff
• Template f_lower: 20 Hz
• SNR crop: 8s start, 4s end
• Only hp polarization used for single-detector filtering