Five pulse filters find no signal above the 99 percent control threshold 2 to 10 seconds before GW170814

We search public H1 and L1 strain for short pulses before GW170814. Five filters use frequencies from 60 to 300 Hz and a 20 ms Gaussian envelope. The test searches time, frequency, and detector lags within 9.77 ms. No pulse passes the 99 percent threshold from 229 control windows. Two noise-power estimates give full-search ranks of 0.035 and 0.091. Equal-amplitude pulses with strain hrss of 5e-23 sqrt(s) per detector pass the main threshold in 45 to 57 of 57 reserved trials, depending on frequency. Direct waveform addition checks the filter response. This is a restricted null result, not a source detection or a limit on eccentric binaries. Only one pulse width and five frequencies are tested. Noise changes, hardware injection lines, detector response, and untested waveforms limit the result. The package contains code, aggregate results, figures, input hashes, and retrieval steps.

astronomy · GW170814 · precursor pulses · finite filters · null result · injection recovery