Matteo, Eleonora, Simon

We have changed the setup of the absorption bench so to be able to measure S and P polarization of the transmitted beam.
The basic point in the new setup is that we are using two photo-sensors (PSDs, to be more specific) and a PBS together with a lens to focus and split the beam.
Since the measurements are very sensible to the setup, we needed to take several aspects into account:

• Incoming beam needs a good initial alignment in terms of polarization
  •  another PBS and a HWP were installed
• The PSDs are too sensitive for the pure transmitted beam
  • A stack of ND filters were installed between lens and PBS in the outcoming beam-path
• The setup is extremly sensitive to stray-light
  • A beam-tube is covering the PSD in the lateral beam-path from the PBS (S - polarization)
  • A beam-dump (Vantablack) is catching the reflected beams from the ND-filters
  • A razor-blade beam-dump is covering the space btw. beam-tube and ND-filters (catching scattered light)
  • A couple of obligued ND-filters is placed at the blind-side of the PBS
  • A black light-cover is placed on top of the lens to catch some ghost-beam reflections coming from the top of the sample

Especially the stray-light suppression is challenging as there are a lot of possibilities where scattered light can enter the sensors and it may be that we still find new sources.
For example, one main source of stray-light was a ghost beam created by a probe-prism which is reflected by a lens for the incoming pump-beam, and this reflection is entering the test-mass and internally reflected on the upper boundary and eventually entered the P-pol sensor from above (shielded now - see pictures)

Electronics
We are using basically the same system as already given by the absorption-bench setup.
However, we had to do some modifications:

• As we have only one channel to demodulate an AC signal on the lock-in, we have installed another lock-in to demodulate the signal from the second PSD (using the same reference signal)
• The power for the measurements is reduced to be in the mW region (2A input current)
• Of course more cables and power-sources needed for the sensors

I upload here the open-loop transfer function of the filter cavity. The situation now we are operating with.

(gain of 7, attenuation of 7)

Aritomi and Yuhang

After seeing the effect of rubber in entry 1508, we tried to put it at another position. Considering what we reported in entry 1471, we found the resonance of 500Hz/600Hz is due to different base plate we put for phase shifter. So it seems to be an overall resonance of the mirror mount.(flag pole mode) So we decided to put a piece of rubber under the whole mirror mount. We thought it will damp the resonance of 500Hz/600Hz. The rubber and the position we put for CC1 and CC2 is shown in the attached figure 1 and 2.

After put them we increase gain little by little, we couldn't find the resonance of 500Hz/600Hz. It makes a big difference. The measurement of the optomechanical transfer function is attached.

[Aritomi, Yuhang]

First we put rubber under phase shifter to damp 500Hz resonance. 500Hz resonance almost disappeared with rubber and finally we could lock CC1 with 4kHz bandwidth and CC2 with 1.5kHz bandwidth. CC2 is still a bit unstable and maybe integrator is necessary for CC2.

Then we measured squeezing spectrum, but the spectrum is similar with before... 

Tomorrow we'll measure free running and closed loop phase noise of CC1,2 to estimate phase noise.

Yuhang and Aritomi

We tried to put a rubber at the place where we think it can be related to resonance. First, we tried to put it between mirror/PZT holder and mirror mount. See attached figure 1 and 2.

In the measurement shown in entry 1503, CC1 has already been installed with this rubber while not for CC2. In this entry, I put a comparison of OLTF for CC2 with and without the rubber.

In the attached figure 3, it is shown the measurement with and without the rubber. We can see 4kHz resonance is damped.

In the attached figure 4, I add 12dB to measurement with rubber. The reason is we were locking with different unity gain frequency. In this case, we could compare better. From this comparison, we could see that the reduction of resonance at 4kHz gives energy to the peak at 600Hz and 7kHz and makes them higher.

Considering the position we put rubber, we think 4kHz corresponds to resonance along the direction of mirror/PZT holder axis.

I injected noise on lenght d.o.f of the end mirror when FC is locked and measured its TF to legth oplev and correction signal to the laser piezo.

The TFs look similar (pic 1) even if they are not very clear and the coherence is quite low. (pic 2).

The noise amplitude was 10000 count (~3V) and the driving matrix was such that only vertical coils where used. (It should couple only with pitch)

Pic 3 shows the spectra of the other d.o.fs. The damping loops were closed at that time, but pitch shows a motion larger that usual.

1. For main laser power line. The board should have a hole 11.1cm away from border and 4cm high.

2. For green injection, the beam is at the edge of the board. Height of 7.55cm

3. For IR injection, the beam is 37.2 holes(each hole 2.5cm)  away from main laser side.

4. length of homodyne power cable should be 5m

(Please look at attached figure to make it clear about where should be the holes)

[Aritomi, Yuhang]

First we measured dark noise of homodyne since we had large 50Hz noise in shot noise spectrum. The dark noise already had large 50Hz noise and we found that when we removed ground floating of spectrum analyzer we are using, these peaks disappeared (Pic 1).

For CC, we checked noise in SHG, GRMC (entry 1514). Then we tried to lock CC1 and 2. For CC1, at the beginning of today we could lock CC1 in stable region between 500Hz and 7kHz resonance, but at the end of today, we couldn't find stable region somehow. For CC2, we couldn't lock it stably so we put rubber to CC2 mirror holder and 4kHz peak is damped (entry 1508), but we still cannot lock it because of 7kHz resonance. 4kHz resonance should be resonance of mirror holder, but we don't know where 7kHz resonance comes from.

Then we measured squeezing with 50mW green (Pic 2) when both CC loops are stable. Squeezing is ~5dB and anti-squeezing is ~20dB. Squeezing spectrum was not very good due to phase noise. We also measured it with 40mW green (OPO:7.169kOhm, PLL:150MHz). Squeezing spectrum is more clear while squeezing level is almost same.

Aritomi and Yuhang

We measured the open-loop transfer functions of CC1 and CC2 yesterday.

We could see from these two figures that 0dB line cross 500Hz and 7kHz for CC1. While 0dB line cross 600Hz, 4kHz and 7kHz peaks at the same time for CC2.

These peaks make locking not easy to achieve. We are thinking about the following things to try:

• to dump resonance by putting rubber
• to use a notch filter to not excite the resonance
• to glue PZT and mirror on the phase shifter

I tried to improve the lock stability by adjusting the gain of SR560.
However, I could not improve so much.
In the end, the lock last only 0.5sec.

So we need circuits for feedback control to achieve stable lock.
Anyway, we can lock the laser to temporal silicon cavity.

To confirm if we have more green phase noise(free-running), I performed the measurement of noise spectrum of SHG and GRMC transmission again.

The comparison is with entry 1276.

We could see we have much more noise now!

Add some points,

• We observed the unstable lock of green phase. And I think it is related to the unstable of squeezing measurement.
• From the phase control behavior, the phase noise(free-running) seems to be higher than before. Need to be confirmed.
• The green phase lock has two main resonance 500Hz and 7kHz. In the beginning, we could find a gain region between these two oscillation and lock green phase. But at the end of yesterday, we couldn't find this region when we change the gain of control servo.
• The error signal for the green phase lock usually change. As pointed out by Aritomi-san, once it is due to the misalignment of green(green power was changed). Actually, we observed a second change of error signal of green phase lock. We check PLL lock, green power(also GRMC alignment), OPO temperature. But they were fine. We should use more time to investigate what makes it change and try to avoid it.

[Aritomi, Yuhang]

First we made LO DC balanced and aligned LO and BAB into AMC. DC balance of BAB is 10mV. We measured shot noise of LO (Pic 1). Although there are large 50Hz peak and its harmonics , homodyne is shot noise limited above 20Hz. To remove the 50Hz peak, we floated ground of homodyne, but the shot noise was same. We also measured shot noise spectrum when LO and CC are injected into homodyne. Shot noise with LO and CC was same as shot noise with LO only.

Then we locked green and IR phase with SR560 (bandwidth  ~ 100Hz) and found squeezing spectrum is very bad. So we locked green and IR phase with Pierre's servo (bandwidth ~ 1kHz) and finally could recover the 5dB squeezing (Pic 2). 

We thought green power is 50mW, but actually green power was smaller than 50mW due to misalignment of GRMC at that time.

I reduced the gain to 2*10^2, and tried to lock.
The oscilation was improved than higher gain one, but the lock last only 0.1sec and DC REFL dip was ~0.1mW.

Anyway, it seems that it is possible to lock TEM00 mode.
Actually, feeding back to laser PZT induces laser power fluctuation.
Therfore, ISS is needed in this experiment.

Matteo, Simon

On Friday, we started to upgrade the absorption map to be capable of taking maps on the homogeneity of mirror-substrates in terms of their polarization.
The basic idea is to use a PBS nad two photodiodes (PSDs in our case) to distinguish between S and P polarization. Actually, we recognized that the incoming beam is 90% P and 10% S polarized so that we decided to put another PBS also in the incoming beam in addition to a half-wave plate which we use to check the differences in S and P polarization and the calibration of the system.

After some first checks, we found out that the PSD designated for sensing the S-polarization is very sensible to any scattered light which is automatically an issue due to the many ND filters we have to use for not saturating the AC channel (which is fed by the S-pol PSD). Therefore, we took long time to cover the path from the PBS to the PSD. We finally took a beam-pipe which does the job quite well (see attached pictures).

Over the weekend, we took two first maps. One with incoming-beam fully P-polarized and the other one having the half-wave plate rotated by 30 degrees. Qualitative pictures can already be calculated but we still need to calibrate the DC signal properly to calculate also a quantitative number.

This morning, I again tried to lock TEM00 mode.
The attached picture shows when the laser seemed locked (the label is same as before).
It seemes something is fed back to the laser, and it last several tens of seconds.
Actually, it is oscillating, and the dip is so small, so it needs to improve the alignment, gain, and some other things for more stable lock.

[Aritomi, Yuhang]

This is work on July 20th. First we made LO and BAB overlap and more or less DC balanced at homodyne. CMRR at 1 kHz is 82 dB which is similar to before. Then we measured shot noise and squeezing spectrum (data is not saved). Shot noise is worse below 100Hz and squeezing spectrum was very bad. 

After that we found large 7 MHz peak at homodyne DC signal (attached pictures). LO and CC before homodyne BS is 1.2 mW and 8 uW respectively. This 7 MHz seems beat between LO and CC.

[Aritomi, Yuhang]

This is work on July 19th. In current setup, we only have one steering mirror and one lens for alignment of BAB going to homodyne and it's difficult to align BAB into AMC. So we changed optical layout so that we can have three steering mirrors for alignment of BAB. I updated optical layout of our bench. After this modification, we could align BAB to AMC easily. Re-alignment of IR to FC after this modification is ongoing.

Today I tried to lock TEM00 mode with 10^3 gain.
The attached pictures show DC REFL (blue), error signal (green), and feedback signal (yellow).

Actually, the lock was not stable; it last only ~10msec.
This seemes due to the fluctuation of feedback siganl which oscillated about 20Hz.
Also the error signal oscillates about this frequency, and this hinders me to lock.
I will investigate from where this oscillation comes.

The excess of high frequency noise measured last summer when the cavity is locked (see entry #903) seems disapeared. This is good but the reason is not clear to me.

Note that in order to sum the line into the PZT  for the calobration we have reconnected the ramp potentiometer which was disconnected last summer,  following Pierre's advice,  in order to reduce the rampeato noise (see entry #875).

We have left it connected for now. It would be interesting to compare the difference in the error signal spectrum when it is connect and disconnected as done last summer (entry #883).