L. Dartez, J. Kissel

We're investigating some issues with time-dependent correction factors (TDCFs) as computed by the front-end vs. GDS, and reminding ourselves of the parameters that define the "smoothing out" of the computation. In particular the uncertainty threshold against which the live uncertainty is compared.

Apparently, all front-end thresholds are set at 5% uncertainty.

    H1:CAL-CS_TDEP_CAVITY_POLE_F_C_GATE_UNC_THRESH       0.05
    H1:CAL-CS_TDEP_CAVITY_POLE_KAPPA_C_GATE_UNC_THRESH   0.05
    H1:CAL-CS_TDEP_D2N_SPRING_F_S_GATE_UNC_THRESH        0.05
    H1:CAL-CS_TDEP_D2N_SPRING_Q_S_GATE_UNC_THRESH        0.05
    H1:CAL-CS_TDEP_KAPPA_PUM_GATE_UNC_THRESH             0.05
    H1:CAL-CS_TDEP_KAPPA_TST_GATE_UNC_THRESH             0.05
    H1:CAL-CS_TDEP_KAPPA_UIM_GATE_UNC_THRESH             0.05
(Remember, the live uncertainty is defined by taking each calibration line's coherence, coh, with DARM_ERR, converting it to uncertainty via sqrt([1-coh]/[2*Navg*coh]), where Navg is determined by the FFT length -- see LHO:69175 for a recent review of this calculation.)

I could make up a story from my foggy memory as to why the front-end thresholds are set at 5% (and not 0.5%, like we expected -- and like what we *think* the GDS pipeline's threshold is), but I'll spare you. They've been this way since Apr 2021.

Attached is the time-series comparison between GDS (in red) and Front-end (in gray) computed \kappa_C for the 24 hour period after July 31 2023 00:00 UTC. The gray, front-end trace shows the characteristic ~3 hour thermalization transient we've come to know and love from our 76W IFO (see e.g. LHO:69796). The red trace, doesn't show this, and it's in fact frozen at the "last good value" by the GDS gating system -- because it's uncertainty threshold is much lower than 5%. 

We presume that this is because the detector noise was pretty DARM bad during this time period -- see 24-hour statistics on the sensitivity in second attachment.