[Fujimoto, Tanaka, Takano]

Summary

We diagonalised the SRM oplev. Now, the coupling from other DoFs is below 1% at the resonance frequencies.

Detail

We noticed that when we shake SRM TM in Pitch we observed motion in Yaw as well. At first, we suspected the actuator couplings, but the health check results obviously told us the coupling of the sensor (the tilt oplev), see Fig. 1 (Pitch) and Fig. 2 (Yaw) (blue: measurement in 2022, green: before the diagonalisation, red: after the diagonalisation). The 2x2 matrix about the coupling between pitch and yaw is as follows:

After normalisation:

Using this matrix, we calculated its inverse and multiplied it by the current SENSALIGN matrix from the left. The health check results using the new SENSALIGN matrix are shown in the figure: in both Pitch and Yaw, the coupling is below 1% at the resonance peak frequencies. 

[Fujimoto, Tanaka, Takano]

Summary

We continued the DRMI optimisation. We could stably lock DRMI with POP17Q for MICH, instead of REFL51Q. REFL51 signals looked strange, probably due to a bad demodulation phase. ADS implementation was tried, but not succeeded yet.

Detail

We continued the DRMI locking trial. This time, we investigated the REFL51 signals and discussed whether they could be used for locking. As reported here, there is an offset on both REFL51I and REFL51Q, but we don't understand its mechanism. In addition, the locking with REFL51Q for MICH looked unstable, so we decided to change the MICH control signal from REFL51Q to POP17Q for better stability. With POP17Q,  stability seemed to be improved so much. The lock can now be kept for more than 10 minutes.

The measured open loop transfer function for each loop are shown in Fig. 1 - 3. They looked stable enough, but sometimes we observed that MICH control got unstable and started oscillation at apparently a random frequency in 20 - 40 Hz. It may be that because of the alignment, MICH sensing gain increased, and the phase margin got smaller, or another control loop sucked the phase margin in the MICH loop via unknown couplings. Anyway, MICH control seems the key to better stability of DRMI control.

Once DRMI got stable, we injected the dithering signal in SRM and tried the phasing of SRM ADS with AS34I signal. The phase of the transfer function from the dithering LO to the signal (AS34I) is shown in Fig. 4. The measured phase was put in ADS, and tried to engage the control for SRM. It was not obvious how it worked, because the buidup level of AS34I looked almost the same due to its large fluctuation. Then, ADS of other mirrors, PRM, BS and IMMT2 were engaged using POP90I signal, but it seemed that the feedback signal made the alignment worse. We should tune the phases of the other mirrors as well for full ADS in DRMI.

With DRMI locked with POP17I&Q and REFL135I, the REFL51 signals were quickly examined. Obviously, both REFL51I and REFL51Q had a large offset. When the demodulation phase was rotated, the offset value also changed in sine curve for I phase and in cosine curve for Q phase (Fig.5). We don't know the source of these offsets, probably either MICH or SRCL has this offset.

Now that we can lock DRMI without REFL51 signals, the investigation is easier than before. I hope these issues will be solved soon and DRMI will be locked using fully 3f signals.

Optical table keeps the temperature around 297K even when HPBD temperature reaches around 365K.

[Tanaka, Fujimoto, Saito]

Following klog:36995, a high-pass filter with a cutoff frequency of approximately 12 MHz was added to suppress the 900 kHz noise originating from the sub-laser. In addition, the beat signal was found to be approximately 10 dB lower than in the previous measurement (klog:37020), although the cause remains unclear. Lock acquisition was then attempted, but no lock was achieved. An RF amplifier was also inserted after the high-pass filter; however, although the beat signal amplitude increased, the signal waveform became distorted.
 

• To suppress the 900 kHz noise from the sub-laser identified in klog:36995, a high-pass filter with a cutoff frequency of approximately 12 MHz was constructed using a 270 pF capacitor (Photo 1). The measured transfer function of the high-pass filter is shown in Photo 2. The high-pass filter was then installed immediately after the output of the RFPD, and the spectra with and without the filter were compared. When the RFPD output was passed through the high-pass filter, the signal level decreased by approximately 20 dBm compared with the unfiltered case (Photo 3). The red trace corresponds to the case without the high-pass filter, the light red trace corresponds to the case with the high-pass filter, and the blue trace corresponds to the measurement with nothing connected. Next, the high-pass filter was tested with the power splitter and mixer used in the PLL setup connected. In this configuration, the signal level decreased by approximately 11 dBm when the high-pass filter was inserted (Photo 4).
   
• To determine the conversion between dBm and Vpp in the Moku:Lab spectrum analyzer, a signal generated by the function generator of one Moku:Lab was measured using the spectrum analyzer of another Moku:Lab. When a 3 mVpp signal was applied, the measured level was -46.35 dBm. Therefore, the beat signal observed in the previous experiment (klog:37020) corresponds to approximately 3 mVpp.
   

• The beat signal was then re-examined and found to be smaller than in the previous measurement (klog:37020). To investigate this, the alignment was readjusted, the RFPD was reinstalled, and the polarization was varied. After reinstalling the RFPD, the DC output became saturated, so the sub-laser power was reduced to 0.825 mW. Ultimately, the beat signal level reached approximately -55 dBm (Photo 5). This is about 10 dB lower than in the previous measurement (klog:37020). The reduction cannot be explained solely by the decrease in sub-laser power, and the cause remains unclear. However, very little fluctuation in the beat signal amplitude was observed.

• Next, lock acquisition was attempted using Moku:Lab with a 10 kHz low-pass filter while varying the gain and adding an integrator. However, no behavior indicating that the beat frequency was being pulled toward the LO frequency was observed. Lock acquisition was also attempted using the low-pass filter of an SR560 instead of Moku:Lab while varying the gain, but no lock was achieved. In addition, when the sub-laser temperature was lowered from 31.51 ℃, a beat signal was still observed at 29.74 ℃. Based on the absolute frequency measurement reported in klog:36760, this is believed to be due to a mode hop. PLL operation was also tested at this temperature, but lock acquisition was again unsuccessful.
   

• Finally, an RF amplifier was inserted after the high-pass filter in an attempt to increase the beat signal amplitude. Although the beat signal became larger, its waveform became significantly distorted (Photo 6). When the mixer was removed, the waveform returned to a clean shape, indicating that the mixer was responsible for the distortion. Therefore, this RF amplifier is not suitable for the present application.

One of possibilties why SRCL build up is lower than our expected is detuned from the resonance point. Actually, When DRMI was locked, RF51 I was not zero even though SRCL error signal by 1f was zero (fig.1). This offset is often caused by RFAM. So I checked current RFAM status of each demodulated signal from each REFL RFPD when the single bounced beam from PRM is on each PD.

Fig.2 shows that REFL demodulated signals and DC powers on REFL PDs. As you can see, RF56, RF45 seems to have some RFAMs. On the other hands, 3f PDs seems to be zero. Therefore, The offset on SRCL error signal cannot be explained by RFAM.

[Ushiba, Hirose]

We checked their Open-loop Transfer Function. We adjusted their gain and confirmed the phase margin.

And the adjusted filter is replaced. We also confirmed that the control stabilizes when switching to AC in the “ALIGNED state”.
We confirmed that switching to DC in the lock acquisition state stabilizes the control.
If the TM doesn't seem to stabilize in Lock acquisition state, it's better to set it to ALIGNED state and apply IM control.

By the way, there are several thermometers attached around the high power beam dump. One of them would be on the top surface of the optical table in the IFI chamber, probably. How about their values?

[Ushiba, Hirose]

Because the SRM was highly unstable in the 0.1–1 Hz band, we switched off the integrator in the OLDAMP of the IM stage and switched on a “DC” filter acting as an integrator into the TM stage.

After that, the integrator was further adjusted to include a gain of 10 dB or more in the 0.1–1 Hz range. 
Furthermore, regarding the Damp filter, since the phase does not rotate in the high-frequency band, the high-frequency section of the current filter was replaced with a Chebyshev filter. 
We also adjusted the gain to ensure phase margin stability in the UGF. 
(FIG1, FIG2: PIT and YAW filter with sus model. Blue shows before adjustment, and red shows after adjustment.)
Currently, I have added modified filters to FM3 and FM5 of the PITYAW control filters for the SRM and TM OLDAMPs. After PLL work is completed, I will run tests, and if stability appears to be sufficient, I will replace the current filters in FM1 and FM9 with these modified ones.

Note：
・The suspension model was already included in OLDAMP's FM10. We have confirmed that this is a filter, just like the following files that will be updated in the latest health check.
/users/VISsvn/HealthCheck/PLANT_SRM_PAY_FLOAT_TM_TEST_{P,Y}.xml

・The Chebyshev filter is normalized such that the peak gain in the passband is 0 dB. In the 1 dB ripple Chebyshev filter used here, the DC gain is −1 dB. Therefore, a gain of +1 dB was applied to the entire filter to set the DC gain to 0 dB.

Recently, high power IMC output with misaligning PRM is often used and sometimes IMC is kept locking with high power over night, which results in making HPBD hot,
So, I discussed with uchiyama-san and decided to add the safety function in the IMC guardian.

If hitting 10W beam for 8 and 24 hours, HPBD temperature becomes 350K and 370K, respectively.
So, I set the threshold at 365K as shown in fig1.

From now, if the HPBD temperature exceeds 365K, IMC lock will be lost once to make IMC output smaller.

Note:

During the work, I noticed we have two ZERO_SERVOOFFSET state with the same index.
Since this state is not used, I didn't change them but it would be better to renew the IMC guardian.

Since the satellite box was replaced, we haven't observed any glitches. Very good. We'll keep our eyes on it until the next maintanance day.

[Takano]

We tuned the demodulation phase of REFL51 in the same way as klog 36990.

We injected the excitation at 50 Hz in SRCL1_EXC when SRMI was locked using 1f signals and measured the ASD. The demodulaiton was tuned so that the peak at 50 Hz in REFL51Q was minimised. The phase was changed from -60 to -85.

Yes, that's one typo; I removed 0 unintentionally.

Another typo exists here:

> Next, we changed the PRCL error signal from REFL45I to POP17I, because in RSE, REFL45I is used for CARM and POP17I for PRCL.

Not POP17I but POP45I.

>~7 deg rotated from the optimal angle, according to Ushiba-san

Not ~7 degs but 70 degs if I remembered correctly.

[Takano, Tanaka, Fujimoto, Saito]

To improve the mode matching between the main laser beam and the sub-laser beam incident on the RFPD, the optical path length of the sub-laser was adjusted. The resulting mode-matching ratio was approximately 76%. The alignment was performed using two irises, and the RFPD was installed. The alignment of the main laser was then finely adjusted to maximize the beat signal amplitude, and the polarization of the sub-laser was also optimized. As a result, the beat signal level increased from -58.67 dBm in the previous measurement (klog:36974) to -45.59 dBm.
 

• In klog:37004, the waist positions of the two beams differed by 352 mm. Therefore, part of the optical setup was modified to increase the optical path length of the sub-laser by approximately 350 mm. The beam profiles of both the main laser and the sub-laser were then measured at locations far from the beam waist on both sides of the waist, and fitting was performed (Figure 1). The green and orange curves represent the fitting results for the main laser, while the red and blue curves represent those for the sub-laser. The origin of the coordinate system is defined at the beam sampler where the two beams are combined.

  The waist positions and waist radii obtained from the fitting are summarized below.

  Main Laser

  x-direction: Waist position = -220.1 ± 3.7 mm, Waist radius = 0.0577 ± 0.0006 mm
  y-direction: Waist position = -170.7 ± 7.6 mm, Waist radius = 0.0674 ± 0.0014 mm
  →Average: Waist position = -195.4 mm, Waist radius = 0.0626 mm

  Sub-Laser

  x-direction: Waist position = -522.6 ± 1.5 mm, Waist radius = 0.0929 ± 0.0003 mm
  y-direction: Waist position = -522.6 ± 6.3 mm, Waist radius = 0.1077 ± 0.0015 mm
  →Average: Waist position = -522.6 mm, Waist radius = 0.1003 mm

  The waist position and waist radius of the main laser, which should not have changed, were found to differ significantly from the values reported previously. In klog:37004, the fitting had been performed using measurements taken only after the waist, which likely resulted in a large fitting error. If the sub-laser waist position is assumed to be -172.6 mm while all other parameters are taken from the present measurement, the calculated mode-matching ratio is 62%. Therefore, the mode-matching ratio of the optical setup in klog:37004 was likely around 62%, rather than the previously estimated 5%.

  Since the waist positions were still offset by 327.2 mm, part of the optical setup was modified again to reduce the sub-laser optical path length by approximately 325 mm. The beam profiles of the main laser and sub-laser were then measured only at locations far beyond the waist. The waist radii were fixed to the values obtained in Figure 1, and fitting was performed only for the waist positions (Figure 2). The green and orange curves represent the fitting results for the main laser, while the red and blue curves represent those for the sub-laser. The origin is again defined at the beam sampler where the two beams are combined.

  The fitted waist positions are summarized below.

  Main Laser

  x-direction: Waist position = -199.0 ± 13.1 mm
  y-direction: Waist position = -148.7 ± 14.8 mm
  →Average: Waist position = -173.9 mm

  Sub-Laser

  x-direction: Waist position = -133.0 ± 10.4 mm
  y-direction: Waist position = -194.1 ± 3.9 mm
  →Average: Waist position = -163.6 mm

  Using these results, the mode-matching ratio was calculated to be approximately 76%.
   

• Next, two irises were installed, and the alignments of both the main laser and sub-laser were adjusted so that the beams passed through both irises. A 50 mm focal-length lens used to focus the beams onto the RFPD, the steering mirror located immediately before the RFPD, and the RFPD itself were then installed. The current optical layout is shown in Figure 3.
   

• When both beams were directed onto the RFPD using the steering mirror, the DC output was -4.825 V. When only the main laser beam was incident on the RFPD, the DC output was -54.88 mV. The optical power incident on the RFPD was approximately 19 μW for the main laser and 1.638 mW for the sub-laser. After confirming the presence of the beat signal, the alignment of the main laser was finely adjusted to maximize the beat signal amplitude. In addition, the polarization of the sub-laser was adjusted using a HWP to further maximize the beat signal (Figure 4). As a result, the beat signal level increased from -58.67 dBm in the previous measurement (klog:36974) to -45.59 dBm. Furthermore, the beat signal amplitude exhibited very little fluctuation over time.

I intentionally changed the Pcal-X beam position by moving the pico motors and checked the response in the QPD signals.
(ETMX was in the LOCK_ACQUISITION state during this work.)

TCam images were taken before and after the pico-motor movement.

A detailed analysis of the QPD response will be performed later.

I modified the guardian to change the nominal PSL HWP angle from 148 to 152 so that IMC output becomes 1W when HWP is at nominal angle.

(Remote) R. Takahashi, Nakagaki, Ikeda

This work is related to K-Log #37013.
We replaced the PR3 satellite box.

[Details]
1. We measured the forward voltage at the end of the conversion cable connected to the feedthrough.
   There was no significant change in the values, and all measurements were normal.

IM H3

(Reference value)IM H1

2. Next, we replaced the satellite box.
We placed the new satellite box (S1604893) on top of the old satellite box (S1504430) and reconnected the DSUB cables from the old box to the new one.
H2 and H3 were connected to the replacement satellite box.Fig.1
 

Detail

The first trial

First, we started using 1f signals for locking the DRMI. The signals were as follows:

• PRCL: REFL 45I
• MICHL POP 17Q
• SRCL: POP 17I

The gain settings of them were decided as follows:

• PRCL: Measured the relative gain between REFL 45I and REFL 135I when PRMI was locked using 3f signals. Multiplied the factor by the gain for PRMI 3f locking.
• MICH and SRCL: Estimated the relative gain from SRMI to DRMI using Finesse simulation as shown in Fig. 1 (15 for MICH and -17 for SRCL). Multiplied the factors by the gains for SRMI 1f locking.

The initial guess of the appropriate gains and the actual gains after tuning are as follows:

We built these gains beforehand and set triggers to trigger the output when POP90 and AS34 buildups exceeded a certain level. With these configurations, we could hold the interferometer in a good position for ~1 sec (Fig. 1).

Second trial

Next, we changed the PRCL error signal from REFL45I to POP17I, because in RSE, REFL45I is used for CARM and POP17I for PRCL. We tried w/o and w/ triggers, but we could not lock the DRMI, even tweaking the gains. The situation got worse. 

We noticed that the PRCL feedback was incredibly small, more than 2 orders of magnitude smaller than the others, so we increased the gain of PRCL, but the stability didn't change significantly. We decided to use REFL135I, which is used for PRMI3f lock, to control PRCL instead of REFL45I. The PRCL gain for the new signal was chosen to maintain the overall gain from POP17I to REFL135I. Soon after changing the signal, we could lock the DRMI in 26 seconds (Fig. 2). Something was wrong with POP17I.

Tuning gains, changing signals

Then we started further tuning of the gains and selecting the signals. We tried replacing POP17Q (1f) with REFL51Q (3f) for MICH control, and it seemed that 3f was better than 1f.  We also tried replacing POP17I (1f) with REFL51I (3f) for SRCL control, but in this case, 1f seemed better than 3f.

The final configuration of the selected signal, the control gain, and the upper and lower thresholds for each DOF are shown below:

With these conditions, we could lock DRMI in> 5 minutes (Fig. 3). We could barely measure the open-loop transfer function of each DoF, as attached (Fig.4 for PRCL, Fig. 5 for MICH, Fig. 6 for SRCL). Probably because of control instability, the measurement data have lower coherence, so it is not clear exactly what they looked like.

Issues

As reported here, the buildup of AS34 would be more than 5, whereas it is currently 0.5-0.6. It could be that the current locking point is not what we want due to the incorrect signs of the gains and the offset in the error signals. However, the POP90 buildup looked good (it's almost 1, as expected), suggesting the SRCL might be controlled at a weird point. On the other hand, the buildup during DRMI swings reached ~2 and never reached 5. It could be due to poor alignment of SRMI, but I'm not sure.

Another issue is the coupling between the DoFs. The open-loop transfer functions looked very weird in shape, especially in the 10s of Hz, and we suspect the couplings are the cause. At least, we know the demodulation phase of REFL51I is far from optimal (~7 deg rotated from the optimal angle, according to Ushiba-san). We have to tune the phase in SRMI or DRMI, and also tune the control gain again.