We have also looked at the time constants of various channels to look into SSFS UGF decay after the lock (VIR-0390A-18).

Conclusion:
We should look into 56MHz sideband power decay in PRC after the lock. It could be from BS,CP,POP heating and/or something in the injection system. WI/NI seem to have longer time constant, and may not be the cause.

Details:
Using the 11 locks stretched for more than 1 hour between June 7 to 11, we got the following time constants (some of them fits very badly, so don't believe in the statistical errors too much).
LSC_B1_DC_mean : tau=(-4.77 +/- 7.21)e0 sec
LSC_B1p_DC_mean : tau=(-8.69 +/- 0.68)e2 sec
LSC_B1s1_DC_mean : tau=(-1.11 +/- 0.05)e3 sec
LSC_B2_DC_mean : tau=(-3.51 +/- 0.14)e2 sec
LSC_B4_DC_mean : tau=(-1.69 +/- 0.18)e2 sec
LSC_B5_DC_mean : tau=(-1.59 +/- 0.12)e2 sec
LSC_B7_DC_mean : tau=(-3.92 +/- 0.16)e2 sec
LSC_B8_DC_mean : tau=(-1.38 +/- 0.11)e2 sec
INF_TCS_PR_RH_TE : tau=(1.24 +/- 4.36)e1 sec
INF_TCS_NI_RH_TE : tau=(-2.46 +/- 0.15)e3 sec
INF_TCS_WI_RH_TE : tau=(2.96 +/- 0.36)e3 sec
ENV_TCS_NE_RH_TE : tau=(-1.40 +/- 0.00)e4 sec
ENV_TCS_WE_RH_TE : tau=(5.20 +/- 0.06)e3 sec
LSC_SSFS_UGF : tau=(-1.38 +/- 0.01)e3 sec
LSC_DARM_UGF : tau=(-7.69 +/- 0.05)e0 sec
LSC_B4_56MHz_DPHI : tau=(-1.94 +/- 0.01)e3 sec
LSC_B4_112MHz_MAG_mean : tau=(-1.54 +/- 0.01)e3 sec
SIB2_RFC_PD2_56MHz_I_DS_mean : tau=(-1.29 +/- 0.00)e2 sec
SDB2_B1p_PD2_112MHz_mag_mean : tau=(-1.26 +/- 0.01)e3 sec
DAQ_LNFS_EOM_6MHz_dPhi_100Hz_mean : tau=(-8.95 +/- 0.01)e2 sec
DAQ_LNFS_EOM_8MHz_dPhi_100Hz_mean : tau=(-1.42 +/- 0.00)e3 sec
DAQ_LNFS_EOM_56MHz_dPhi_100Hz_mean : tau=(1.58 +/- 0.28)e2 sec
INJ_IMC_TRA_DC_mean : tau=(-5.70 +/- 0.90)e2 sec
B7_DC/B4_DC : tau=(-6.06 +/- 0.13)e2 sec
B8_DC/B4_DC : tau=(-3.02 +/- 0.13)e3 sec

Attached are the typical time series data for two locks used in our analysis.
We used A*exp(t/tau)+C for the fitting.

The time constant for LSC_SSFS_UGF is consistent with LSC_B4_112MHz_MAG and LSC_B4_56MHz_DPHI, which suggests 56MHz sideband power decaying (or demodulation phase changing) is the cause of the UGF decay (B4_56MHz is used in SSFS). You can also see that SDB2_B1p_PD2_112MHz_mag has consistent time constant. We should monitor the temperature of the BS to look into this sideband power decay (this is one of the reason we added BS drum/butterfly mode tracker; logbook #41815).

Photodiode DC signals, including B4_DC, has shorter time constants, and therefore power recycling gain decay for the carrier is unlikely the cause.
The test mass (ring heater) temperatures have slightly longer time contant and this is somewhat consistent with the time constants for normalized arm transmissions (B7_DC/B4_DC, B8_DC/B4_DC). I believe this is the time constant of the NI/WI etalon effect (logbook #39767). We need a new line tracker working to monitor the real test mass temperatures to confirm this (logbook #41807) after the recovery of the interferometer.

Interestingly, DAQ_LNFS_EOM_(6MHz|8MHz)_dPhi_100Hz (created in June 7?, logbook #41748) have the similar time constant, but 56MHz one does not.
 

I have repeated the same analysis performed in 41814 using the channel of this entry and only few channels show a stable time constant, thus only few channels show a real transient after each lock see Figure 1 (I have considered only channels with time constant < 3e4s).

As it is pointed out before the 56 MHz phi has a time constant longer than the DC power but much shorted with respect to the cavity mirrors. Moreover its time constant is close to the 6 and 8MHz EOM signal decay.