## Abstract

The IFO lock duration is too short to conduct detailed investigations.  
The main cause of the lock loss appears to be sudden kicks to Type-A suspensions, especially ETMX—possibly due to cooling effects.  
We compensated for the gain discrepancy between IR and ALS CARM by adjusting IMC IN2GAIN and LSC-MCL gain.  
PR3 YAW control behavior appears to be abnormal.

---

## What We Did

### Lock Stability and Initial Observation

Currently, the lock duration is limited to ~30 minutes, even with arm ASCs engaged via WFS and PRMI ASCs via ADS.  
Since yesterday, PRMI ASC via WFS cannot be engaged. To improve the overall IFO stability, we began investigating the cause of this PRMI ASC instability.

The GR-to-IR HANDOVER process appears smooth today.  
However, the lock duration became shorter than before.  
The lock loss guardian frequently reports "ISC_WD_L" as the cause.  
Prior to lock loss, ETMX exhibits sudden oscillations at specific resonance frequencies:

• Fig. 1: The DARM error signal oscillates at 15.9 Hz before lock loss.  This matches the 1st chain vertical resonance of the test mass (TM), according to Okutomi-san’s document (JGWdoc).  However, no motion was detected by local vertical sensors.
• Fig. 2: DARM error signal oscillates at 172.7 Hz, likely a violin mode.  Subsequently, ETMX slowly starts to oscillate.  (Note: after RF lock, violin peaks can be seen at ~10⁻¹³ m/√Hz!)

These kicks caused large DARM feedback excursions, pushing the RMS beyond the LSC_LOCK guardian threshold in the `ENGAGE_WFS_DC` state.  
These effects may be related to cooling-induced kicks, similar to those observed during previous cooldowns (e.g., klog33013).

---

### ALS and IR CARM Gain Discrepancy

During lock loss investigation, we measured the ALS_CARM transfer function with Moku.  

• Fig. 3: The current UGF is ~1 kHz, about 8 dB lower than expected.  
  Of this:
  • ~2 dB is attributed to an increase in IMC IN1GAIN (14 → 16 dB)
  • ~2–3 dB may be due to a lower IMC UGF  
  • The remaining 3–4 dB is unexplained

To compensate:

• Increased `IMC IN2GAIN` from –10 dB → 0 dB
• Reduced `CARM IN2GAIN` (ALS path) from 25 dB → 22 dB to prevent oscillation
• Increased MCL loop gain by +10 dB via FM5 (BST) of `K1:LSC-MCL` to maintain cross-over frequency with the PZT loop

As a result:

• Fig. 4: ALS CARM UGF recovered to ~1.9 kHz  
• Fig. 5: IR CARM UGF recovered to ~50 kHz

---

### HARD Loops and PRMI ASC Investigation

Before engaging PRMI ASC, we measured the OLTFs of each HARD loop ({C,D}HARD {P,Y}).

Figs. 6–9: All loops except CHARD Y show a gain decrease of ~4–5 dB.

We then tested PRMI ASC components individually.  
In each case, ADS was disabled before engaging WFS control.

#### BS ASC

Fig. 10: Shows BS ASC error/feedback signals and POP90 during control engagement.  Setpoints are overlaid as solid lines.   Controls engaged, but alignment appeared suboptimal in terms of POP90 power.

#### IMMT2 ASC

Fig. 11: Error/feedback signals and POP90 when only IMMT2 controls engaged (setpoints = 0).  Controls engaged normally, and POP90 was unaffected.

#### PR3 PIT ASC

Fig. 12: Error/feedback signals and POP90 when only PR3 PIT engaged.  Control appears stable, but alignment appeared suboptimal in terms of POP90 power.

#### PR3 YAW ASC

Fig. 13: Error/feedback signals and POP90 when only PR3 YAW engaged (setpoint = 0).  Upon engagement, the PR3 alignment shifted **away** from the setpoint.  This suggests a problem—possibly a **sign error** in the control.

---

### Note

Around 10:20 UTC, IMC became suddenly unstable.  Fig. 14: PMC error signal spectrum becomes noisy significantly.  IMC could not lock for ~40 minutes due to increased frequency noise.  Afterward, the frequency noise settled, and IMC lock was restored.