[Ushiba, Komori]

Abstract:

We measured the transfer functions from the BF GAS, BF damper, and F3 GAS to DARM and performed a noise budget analysis for the corresponding coil driver noises.
All were found to be negligible.

Detail:

In a previous study, we observed a significant improvement in DARM sensitivity upon enabling the dewhitening filters for the type-A tower suspensions (klog:33197).
This raised the possibility that coil driver noise from these suspensions might be limiting the current DARM sensitivity.
To assess this, we measured the transfer functions from selected type-A suspensions to DARM and evaluated the associated noise contributions.

We selected three suspensions—BF GAS, BF damper, and F3 GAS—since they are located closest to the cryopayload, and thus expected to have the largest potential impact.

First, we estimated the coil driver noises of the BF damper and F3 GAS to be negligible.
The nominal coil driver noise levels are 3.3e-5 cnt/√Hz at each SUMOUT port (klog:33428).
Even when injecting white noise of 1e-2 cnt/√Hz at 5-100 Hz, no significant coherence with DARM was observed.
An example is shown in Fig. 1, where the left bottom panel shows no coherence between DARM and IY BF_V3, despite the noise injection of 1e-2 cnt/√Hz (right middle panel).

Next, we injected white noise at the SUMOUT ports of the BF GAS actuators with amplitudes of 1e-2 cnt/√Hz for EX, EY, and IX, and 3e-2 cnt/√Hz for IY.
This resulted in observable coherence in the 15-45 Hz range (Figs. 2-5).
The projected noise contributions can be roughly estimated by scaling the excited spectra by the ratio of the nominal noise to the injection level: 3.3e-5/1e-2 = 3.3e-3 for EX, EY, IX and 3.3e-5/3e-2 = 1.1e-3 for IY.
These projections are shown in Fig. 7 and confirm that even the coil driver noise from the BF GAS is negligible.

Discussion:

Figure 6 compares the nominal DARM spectrum with those obtained during noise injection into the BF GAS actuators.
While we observed an increase in DARM noise by a factor of a few or an order of magnitude in the 15-45 Hz range, we also saw a clear degradation of sensitivity in the 60-100 Hz band.
This suggests that a nonlinear noise source—possibly related to the observed coherence—is close to the current sensitivity limit, as also suggested by klog:34571.

In Fig.8, we compare the SUMOUT noise spectra of the BF GAS actuators during nominal interferometer operation.
The noise in IY GAS is consistently below the coil driver noise level of 3e-5 cnt/√Hz, whereas other suspensions exceed this threshold at certain frequencies.
The broad 2e-4 cnt/√Hz noise observed at IX may be explained by numerical errors in diaggui due to large DC offsets.
The excess noise below 20 Hz at EX and EY might be reduced by implementing more aggressive low-pass filter at the control stage.

Finally, this result can be inconsistent with the prior observation that enabling the dewhitening filters improved the DARM sensitivity.
The DAC noise without dewhitening is at most 4e-4 cnt/√Hz, which still seems too small to explain the observed sensitivity limitation.
Further investigation is needed—such as turning off the dewhitening filters one by one and monitoring the impact on sensitivity—to resolve this discrepancy.