[Washimi, Ushiba, Takahashi]
We measured the transfer functions of the IP from the virtual actuator (L, T) to each virtual sensor and blended sensors. The inertial sensors and the LVDTs were blended at 50mHz. The blended TF of L has a small dip around 60mHz. The phase correction for the inertial sensors should be adjusted.
I measured the IP spectra in the following cases;
I have not adjusted the phase correction filters for the inertial sensors yet. In the case of the 50mHz blending, 40mHz oscillation was glowing. The noise of the FLDACCs looks large.
I adjusted the phase correction filters for the inertial sensors and tried the control using the 50mHz blending. 40mHz oscillation was not improved. So I measured the L and T TFs to each sensor (LVDT, ACC(Geophone), and FLDACC) without any phase corrections to confirm the phase deviations.
I remade the phase correction filters as FM4 'Cor' in ACCBLEND for the inertial sensors and tried the control using the 50mHz blending. Though the control loop for L only, T only, L&Y, and T&Y could work without 40mHz oscillation, the L&T loop increased the oscillation. The spurious pass due to the coupling of L and T seems to make the oscillation.
I checked the TF from the L actuator to the L or T FLDACC. The coupling L to T was large (86% at 50mHz). I found the present matrix had a large non-diagonalized component (0.3658) for L to T. I remeasured the TF using the original (identity) matrix. The coupling L to T was degreased (50% at 50mHz). I tried to control the IP using both L&T loops with the original matrix. It could work without 40mHz oscillation.
I tried to control the IP using the L+T+Y loops in the 50mHz blending. When I added the Y loop to the L+T loops, 40mHz oscillation increased. I modified the matrix for the FLDACC as shown in Fig.1. The oscillation was settled, but the noise level of the inertial sensors was larger than the spectra measured yesterday.
I compared the IP spectra in the following cases.
I checked the performance of the FLDACCs. The plot shows the spectra of the IP in the case of the control using the LVDT only. The noise floor of the H2 FLDACC was larger than the H1/H3 FLDACCs by several times. According to the klog in the tuning of the folded pendulums, the TF of the H2 FLDACC (made of Al) showed a hysteresis damping due to friction. It might be the reason for the bad performance of the H2 FLDACC.
I compared the IP spectra in the following cases.
In these cases, the performance of the #1 case was the highest. There is plenty of room for improvement in the present sensor correction.
I compared the IP spectra measured by the geophones in the following cases. The IP was controlled with the LVDTs and the geophones blended at 50mHz (110mHz for Yaw).
The filter "AC2" reduced the noise due to the LVDT around 100mHz.
I modified the Gurdian for Type-A in EY. The inertial damping of IP is applied in the LOCK_ACQUISITION mode. The LVDT and geophone signals are blended at 50mHz. The sensor correction using the "AC2" filter is also applied.
I compared the spectra of each FLDACC (H1, H2, H3) with the spectra before replacing the coil driver from the HPCD to the LPCD in 2023. Though the control status (LVDT only in 2024 and free run in 2023) was different, the noise level in 2024 is much larger than the spectra in 2023. We should check the performance of the present LPCD.
(Log on the 7th.)
I measured the TFs of the folded pendulums in the FLDACCs. The TFs measured before the installation in 2022 show resonant frequencies of 0.2, 0.3, and 0.3Hz for H1, H2, and H3 respectively (Fig.1). The resonant frequency of H1 did not change, but that of H2 and H3 shifted to 0.8 and 0.5Hz respectively (Fig.2). The condition of the counterweights on the pendulums might have changed due to the Noto earthquake. The Q of H1 and H3 became slightly higher. It may be due to the vacuum environment.
