Reports 1-1 of 1 Clear search Modify search
ISC (General)
Hiroki Fujimoto - 2:36 Tuesday 17 November 2026 (37859) Print this report
[Important] Resonance splitting issue in the PMC

[Sugioka, Hirose, T. Yamamoto, Komori, Tanaka (online), Ushiba (online), Miyoki, Fujimoto]

Abstract

After today’s work inside the mine was completed, we noticed that the PMC could not be locked.
Our investigation revealed that the PMC could not be locked because the resonance splitting prevented exceeding the locking threshold.
Although we could rule out several possible causes of the resonance splitting, as described in the Details section below, the actual cause has not yet been identified.
However, it is likely that the laser shutdown during today’s work (klog #37848) triggered this issue.
Tomorrow, we plan to restart the laser again to see if the situation improves.

Details

At around 17:05, after today’s maintenance and work inside the mine were completed, we tried to lock the IMC to resume interferometer commissioning.
However, we found that the PMC could not be locked.

We then checked the PMC cavity scan and noticed that the peak transmitted power (K1:PSL-PMC_TRANS_DC_OUT_DQ) was only about 9.7, approximately half of the value of 21.9 observed during the lock this morning, as shown in Fig. 1.
At the same time, the depth of the resonance dip in the reflected power (K1:PSL-PMC_REFL_DC_OUT_DQ) was also reduced by approximately half. Therefore, the reduction in transmitted power is unlikely to be caused by clipping of the PMC transmitted beam.

We also found that the peak, which appeared to be reduced by half, was actually split into two peaks when zoomed in, as shown in Fig. 2.
This suggests that the power, which originally reached 21.9, was distributed between the two peaks.

We investigated the cause of this splitting. Although we could not identify the actual cause, we were able to rule out the following possibilities.

(Ruled out) Just bad alignment

We adjusted the PMC input alignment using the pico motors. However, the transmitted power was already almost maximized in the initial alignment, and we could not suppress either of the two peaks.
Therefore, poor alignment is unlikely to be the cause.

(Ruled out) The input polarization was rotated by 45 degrees, causing s/p polarization resonance splitting

If the input polarization were rotated by 45 degrees and the observed double peaks were caused by s/p polarization resonance splitting in the PMC, the p-polarized component should be removed by the polarizer located after the EOM downstream of the PMC.
However, as shown in Fig. 3, both peaks remain in the PSL periscope transmitted power (K1:LAS-POW_PSL2_OUT_DQ).
Therefore, this hypothesis can be ruled out.

(Ruled out) Twisting of the PMC cavity axis, causing L/R polarization resonance splitting

If the PMC cavity axis were twisted and the observed double peaks were caused by L/R polarization resonance splitting, the power of each polarization component would be reduced by half by the polarizer downstream of the PMC.
In this case, both peaks would still be visible in the periscope transmitted power.
However, as shown in Fig. 4, the height of each split peak is approximately half of the periscope transmitted power observed during the PMC lock this morning.
If the splitting were caused by L/R polarization resonance splitting, the height of each peak should be reduced to one-quarter of the original value.
Therefore, this hypothesis can also be ruled out.

(Ruled out) Significant change in the laser output mode after the restart, causing a higher-order mode resonance comparable to TEM00

Since the laser was shut down once during today’s work in the PSL, we considered the possibility that the output beam mode had changed significantly after the restart, and that one of the double peaks corresponded to a higher-order mode resonance.
However, when we checked the higher-order mode peaks in the cavity scan, we found that these peaks were also split into two, as shown in Fig. 5.
Furthermore, when we slowly scanned the cavity and observed the transmitted beam with a CCD camera, we confirmed that the two peaks had the same spatial mode.
Therefore, we believe that both of the two largest peaks also correspond to the same TEM00 mode.

(Ruled out) Multimode oscillation near a mode hop

We considered the possibility that the laser temperature happened to be close to a mode-hopping point, causing multimode oscillation that appeared as the double peaks.
However, when we changed the laser temperature, multimode oscillation occurred at a certain temperature.
At this point, the height of the original peaks decreased by approximately half, and a new peak appeared at a different resonance frequency.
Furthermore, this new peak caused by multimode oscillation was also split into two.
Therefore, we ruled out the possibility that the observed resonance splitting was caused by the laser temperature.

Plans for Tomorrow

As described above, although we could rule out several possible causes, the actual cause of the resonance splitting remains unknown.
However, it is likely that the laser shutdown during today’s work in the mine (klog #37848) triggered this issue.
For example,  the restart may have worsened the RIN peak of the Mephisto laser at slightly below 1 MHz, causing this issue, or some problem may have occurred in the power amplifier.

Therefore, we plan to restart the laser again tomorrow to see if the situation improves.

Images attached to this report
Comments to this report:
kentaro.komori - 3:05 Tuesday 17 November 2026 (37860) Print this report

[Fujimoto, Komori]

Ushiba-san pointed out that this splitting could be caused by deep phase modulation, so we discussed this possibility.

If this hypothesis is correct, the modulation frequency can be estimated to be approximately 1.0 MHz from the peak separation of ~4 ms, the FSR corresponding to ~280 ms in the time-series data during the cavity scan, and the PMC FSR of 147 MHz.
In order for each first-order sideband to reach approximately half the height of the unmodulated resonance peak, the phase modulation depth needs to be approximately 2 rad.

Therefore, two possible scenarios are as follows:

  • A modulation of 3.1 Vpp at 1.0 MHz is applied to the laser PZT, with the laser PZT frequency actuation efficiency of 1.29 MHz/V (klog:36122).
  • A modulation of 3.2 Vpp at 1.0 MHz is applied upstream of the two HV amplifiers for the broadband EOM, with the effective EOM modulation efficiency of 1.25 rad/V (klog:35897) including the HV amplifiers (total gain: 200).
kentaro.komori - 14:42 Tuesday 17 November 2026 (37863) Print this report

[Miyakawa, Fujimoto, Sugioka, Hirose, Komori]

Abstract:

We resolved the issue related to the PMC and successfully locked it with the nominal transmitted power.

The issue was caused by phase modulation through the laser PZT.
An unexpected 5.8 V offset from the IMC CMS was applied to the SR560, causing the SR560 to overload and produce a large oscillation at approximately 1 MHz, and then applied to the laser PZT.

Details:

We entered the PSL room to investigate the PMC issue reported in the original post.
Upon entering the room, we found that the SR560 used as a 30-kHz low-pass filter immediately before the laser PZT actuator was indicating an overload condition (Fig. 1).

We then found that the SR560 output had a DC offset of 3.2 V together with an oscillation of 0.26 V at 1.1 MHz (Fig. 2), which was responsible for the observed ~1 MHz sidebands.
Please note that the voltage displayed on the oscilloscope was measured with a 50-Ω input impedance.
The oscillation amplitude was smaller than expected from the estimated modulation depth.
However, the actuator efficiency may be higher than our nominal estimate because the laser PZT has multiple mechanical resonances above 100 kHz.

Next, we found that the input to the SR560 had a DC offset exceeding 5 V, originating from the IMC CMS output.
For some reason, the slow-output offset at the final stage of the IMC CMS slow path (K1:IMC-SERVO_OFS_SLOWOUT_CALI_OFFSET) had increased to 5.8 V.
This value had remained at 0 V for a long period and suddenly went up to 5.8 V at 10:55 JST yesterday (Fig. 3).

After returning this offset to zero, the double-peak structure disappeared during the cavity scan. We then successfully locked both the PMC and the IMC.
Finally, we aligned the steering mirrors upstream of the PMC and maximized the PMC transmission (Fig. 4).
The dotted lines in the upper panel show the transmission with the frequency modulation (Y1) and that before this maximization (Y2).

In summary, the abnormal PMC behavior was originated from the unexpectedly large DC offset at the input of the SR560.
This caused the SR560 to produce the 1-MHz oscillation, which was applied to the laser PZT and resulted in frequency modulation of the laser.

The origin of the unexpected offset at the IMC CMS output is still under investigation.

Images attached to this comment
Hiroki Fujimoto - 16:04 Tuesday 17 November 2026 (37864) Print this report

I investigated other signals around the time when the IMC CMS slow-output offset changed (10:55:14 JST) and found that the IMC CMS IN1 switch had been turned on just before the offset change (Fig. 1).
Before this occurred, the IO Guardian had been down for more than an hour, and the IN1 switch had remained off.
However, it is still unclear why the IN1 switch was suddenly turned on and why the offset changed immediately afterward.

Images attached to this comment
takahiro.yamamoto - 3:09 Wednesday 18 November 2026 (37870) Print this report

> For some reason, the slow-output offset at the final stage of the IMC CMS slow path (K1:IMC-SERVO_OFS_SLOWOUT_CALI_OFFSET) had increased to 5.8 V.
Setting 5.8V to K1:IMC-SERVO_OFS_SLOWOUT_CALI_OFFSET seems default behavior of DOWN state of FSS guardian (see also attachment).
At line 25, offset value is defined as 5.825V
At line 205, offset value is set to EPICS channel.

I have no idea how to enter the run() function without executing the main() function when there is no 'return' in the main() function.
And also, it's strange that we haven't encountered this problem before.

If we won't use RefCav again, FSS guardian which is just a trouble maker should be removed.

Images attached to this comment
Search Help
×

Warning

×