NAOJ GW Elog Logbook 3.2
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).
Can you do the same measurement with gain other than 10 (when error signal noise is larger)?
DDS1 | |||
---|---|---|---|
channel | function | frequency | phase |
CH0 | SHG/IRMC EOM | 15.2MHz | 0deg |
CH1 | SHG/IRMC demod | 15.2MHz | 0deg |
CH2 | OPO EOM | 87.6MHz | 0deg |
CH3 | OPO demod | 87.6MHz | 135deg |
DDS2 | |||
---|---|---|---|
channel | function | frequency | phase |
CH0 | FC/GRMC EOM | 78.0MHz | 0deg |
CH1 | FC demod | 78.0MHz | 245deg |
CH2 | GRMC demod | 78.0MHz | 35 deg |
CH3 | AA demod | 78.0MHz |
DDS3 | |||
---|---|---|---|
channel | function | frequency | phase |
CH0 | PLL CC LO | 7MHz | 0deg |
CH1 | PLL p pol LO | 0deg | |
CH2 | CC1/CCFC demod | 14MHz | |
CH3 | CC2 demod | 7MHz |
Today I increased the peal-to-peak voltage from 0.4Vpp to 0.8Vpp which is applied to the EOM for modulation.
This voltage corresponds to ~0.2[rad] modulation.
In addition, I adjusted the phase shift for demodulation.
The first picture shows before changing, and the second one shows after.
- Blue line : DC REFL
- Green line : error signal
- Red line : control signal (but not yet optimized)
On the other hand, the dip got worse than before.
In order to lock the laser, I have to improve the alignment.
Next step is to improve the alignment with two STMs, and increase the dip of reflected beam.
Then I will try to lock by feedbacking to PZT of the laser.
Eleonora and Yuhang
While I was doing this measurement, I checked the error signal on an oscilloscope. I found the error signal noise is smaller if I use gain of 10.
So I changed the gain of filter cavity locking to 10.
Following the method Matteo B and Eleonora, I characterized the filter cavity locking noise again. (Measuring sine-wave single frequency noise at a frequency higher than unity gain frequency (channel 'ramp mon' and channel 'eps1'). Calibration by considering ramp monitor factor of 100, PZT gain of 2e6 and filter cavity pole of 1+(f/f0)^2 with f0=1.45kHz )
- Take ramp mon signal, calibrate it to frequency. S_Hz = V_RMS (V) * 100 * sqrt(2) * 2e6 Hz/V = 6.477 Hz
- Take eps1 signal, compare it with ramp signal(already calibrate to frequency), to have the conversion factor from Hz to V. K(V/Hz) = Err_V/S_Hz*(1+ (f/f_0)^2) = 4.568e-3 V/Hz
Then we use this K to calibrate the measurement of error signal noise spectrum to Hz. The result is shown in the attached figure.
However, compared with the measurement of more than one year ago. This level is much lower(almost 50 times lower). We checked the calculation and measurement procedure. We couldn't find out what is the problem.
Can you do the same measurement with gain other than 10 (when error signal noise is larger)?
I installed 3 lenses in order to avoid clipping the beam at AOM.
- f=-50.0mm and f=150.0mm to reduce beam diameter at AOM
- f=100.0mm in one of double-pass AOM part
I am planning to re-install the RFPD and try to lock the laser tomorrow.
Matteo, Satoshi
We measured the transfer function of RFPD used for PDH locking.
I have thought it be a resonant RFPD, but not.
Attached picture shows the TF of the RFPD.
The gain is 33.5db at 29.1MHz which is the modulation frequency of EOM.
Here I attach the measurement of PLL phase noise. The measurement has some disconnect points although I used V/sqrt(Hz) as a unit. The result is summarized as following.
filter cavity locked | filter cavity free running | |
cc-phase noise(RMS) | 18mrad | 12mrad |
pp-phase noise(RMS) | 28mrad | 8mrad |
The lock of filter cavity make RMS phase noise increase by a factor of 1.5/3.5
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).
Aritomi and Yuhang
Today, we checked together about the polarization again with PBS plate.
- The mirror was mounted as flipped(the uncoated side was facing laser). We flipped it back to the correct side.
- The angle of the PBS plate was quite far from 45 deg. We checked today this angle. When we have all the light reflected, the angle was quite far from 45 deg(at that time, around 35 deg). When we turn the mirror to the direction of 45 deg, this reflection disappeared.
So we concluded that the beam we were using is p polarization. And we also need to pay attention next time about the surface of mirror and the angle of PBS plate.
Matteo, Simon
Since the DC value seems to be depending on the longitudinal position of the mirror-substrate (Z), we repeated taking a XY-map at Z=112 with re-calibrated DC-value. It turned out that also the absorption-values increased. We now think that we would have to readjust the DC value every time the Z-stage moves, which would make it basically impossible to receive any quantitative results on XZ and YZ maps (however, qualitative results are anyway achievable).
We run another map at the center position with adjusted DC value to have the three maps (Z=46, 79, 112) all adjusted.
At the same time we started to reconfigure the absorption bench for measuring polarization maps.
Therefore, we aquired some PBS and a 2" lens (f=100) which we could put in the outgoing beam path. In a first test, we already recognized that the main part of the pump beam is p-polarized with ~10% s-polarization, although we thought it were merely s-polarized!
So, we put another PBS in the incoming beam-path to make it almost completely p-polarized and checked the effect on the photometer. When moving the mirror into the beam-path, there seems to be some changes happening but without more precise analysis with photo-diodes, we cannot be sure.
Anyway, we let it like this and will continue tomorrow when the other absorption measurement is finished.
At last, we removed also the first-contact from the ETMY substrate which we put there to clean its surface after the laser-burning accident last week. It seemed to be successful and we could not see any dirt anymore. In this state, we put the container over the substrate to protect it.
[Aritomi, Yuhang]
First we tried DC balance of homodyne with s pol, but power unbalance of BS was large for s pol. Reflection of BS is 630 uW and transmission is 600 uW for s pol while it's almost balanced for p pol. So we decided to use p pol for the moment. We made LO p pol with HWP and confirmed its polarization with cubic PBS (Newport,10BC16PC.9). However, when we put plate PBS (thorlabs, PBSW-1064) instead of cubic PBS, most of the LO was reflected by plate PBS. Actually I didn't check the direction of plate PBS and maybe it's wrong, but even in that case we cannot explain this behavior. We'll check the direction of plate PBS and check the polarization with AMC tomorrow.
Anyway, today we managed to make LO DC balanced with "p pol".
Aritomi and Yuhang
Today, we checked together about the polarization again with PBS plate.
- The mirror was mounted as flipped(the uncoated side was facing laser). We flipped it back to the correct side.
- The angle of the PBS plate was quite far from 45 deg. We checked today this angle. When we have all the light reflected, the angle was quite far from 45 deg(at that time, around 35 deg). When we turn the mirror to the direction of 45 deg, this reflection disappeared.
So we concluded that the beam we were using is p polarization. And we also need to pay attention next time about the surface of mirror and the angle of PBS plate.
I checked the current assumption as suggested by Matteo. It is fine. The situation now is the same with entry 1426. At the same time, for -19V channel, the current assumption is 0.063A. Although we don't have reference for this negative channel, I think it is also fine.
After the optimization of locking servo parameter, we locked PLL_CC. We are interested in this PLL because this phase noise is coupled into squeezing measurement directly. However, the other PLL phase noise changes the parametric amplification and de-amplification factor. How much can they affect squeezing needs to be considered further.
Anyway, the measurement result is attached. There is two main difference:
- The high-frequency PLL noise when the filter cavity is locked is lower
- The low-frequency PLL noise when the filter cavity is locked is higher
I think this is because we are making the main laser to follow filter cavity. Filter cavity suspension makes low frequency noisier while high frequency quieter.
Yuhang and Aritomi
Last week, we found we were using FC lock mixer in an inappropriate way. Then we talked with Matteo and Matteo said that he and Marco did it on purpose.
So I checked what is the difference between amplifying LO and amplifying RF. The result is shown in the attached figure 1 and 2. It is obvious that, by amplifying RF, both signal amplitude(10 times larger) and signal to noise ratio become better.
Then I tuned FC locking servo input attenuator and gain, finally, both 7 seems a good choice(see attached figure 3 and 4). The measurement of the open-loop transfer function is shown in figure 5. Now we have a unity gain frequency of 22kHz and phase margin of 55 deg. But there is a very broad peak after unity gain frequency, maybe we should make unity gain frequency a bit lower.
Simon
Please find attached the absorption maps of GT's ETMX mirror substrate that is currently being analyzed in the absorption bench.
For now, we have three XY maps taken at different positions along the c-axis (Z = 46, 79, 112, where "79" marks the center of the mirror related to the bench-alignment; given in [mm]).
I also calculated some histograms which give statistics on the distribution of the absorption-values.
I checked the current assumption as suggested by Matteo. It is fine. The situation now is the same with entry 1426. At the same time, for -19V channel, the current assumption is 0.063A. Although we don't have reference for this negative channel, I think it is also fine.
Aritomi and Yuhang
We found we couldn't lock CC1 yesterday, and we pointed out the 600Hz and 1kHz noise. Also, we complained about the 1kHz oscillation. However, we just realized that this oscillation was not 1kHz but 600Hz. Have a look of attached figure 1. We thought it was a bit more than 1kHz because the period of oscillation was more than one block of the oscilloscope time axis. While this is a totally stupid mistake, the period of oscillation larger than a block of 1ms should be frequency smaller than 1kHz.
But anyway, we tried to find out what is the difference between we could lock before and we cannot lock now. There are two different points:
- we changed the base plate of mirror mount.
- The PZT&mirror holder was accidentally touching to fix part of mirror mount. (touching problem was not realized at that time) This way of using mirror mount make it function as only a mirror holder, which means we cannot steer it. Please refer to the attached figure 2 to see how it is touching.
First, we tried to touch again. However, after this, we still could not lock CC1. But we couldn't find ~1kHz peak which appeared yesterday. So we think maybe the appearance of that 1kHz peak on yesterday is somehow wrong. I put one of the measurements of OLTF here as attached in figure 3.
Then we tried to follow the suggestion of Matteo. Put a piece of thing on the top of the mirror. See attached figure 4 for how we put this piece of thing. After that we still have oscillation, but it is damped by this additional piece of thing. See attached figure 5.
Then we replaced back the base plate to the original one. We tried to increase the gain, at that moment, we found the oscillation appears as ~500Hz. This is in agreement with the measurement before. Then we lock it first with a bandwidth of ~200Hz. The measurement of OLTF is attached in figure 6.
Since we know that we could use this configuration to lock, so we just tried to increase the gain. When we use a gain of 5.5, we could lock our servo! Then we measured the OLTF but we found a very strange peak at 3.7kHz. This is a very large peak, in principle, this peak will make our system oscillate. However, we could lock.
After that, I also want to try to make PZT&mirror holder not touch mirror mount. But this makes the beam tilt. So, in the end, I didn't succeed. But maybe we could shift this holder a bit ahead since anyway we have a gap between the holder and mirror mount(see the attached last figure). Actually, this gap maybe is because of we are touching the mount and could not put inside anymore. Maybe we could just shift.
Matteo, Simon
Today, we got the first results of the absorption measurements which we started Yesterday. The map showing a circular area around the center of the substrate (taken in the middle of the bulk-body) can be seen in the attachement. The mean absorption coefficient of the whole map is around 75ppm/cm.
Meanwhile, I prepared a python script than can be used as an alternative to the Matlab-code written by Manuel to calculate the absorption-map from the data-file written by the Labview program. I upload it to the dropbox-folder of the PCI PC.
Aritomi, Yuhang
Today when I wanted to lock PLL of p-pol, I found p-pol and master laser beat note frequency is close to zero. Then I move this beat note frequency to 150MHz by changing p-pol laser temperature as usual. But I found that I cannot lock PLL by using the current set-up. I didn't realize at that time, actually, I have already put p-pol locking frequency to another side of 0MHz. Then I changed the PLL locking sign and in the end, although everything else was fine, I couldn't find the coherent control error signal.
So please notice next time that if you couldn't lock p-pol PLL by using the current set up, please change p-pol beat note to another side of 0MHz. (different side refers to p-pol laser frequency is higher or lower than master laser frequency)
Aritomi and Yuhang
I found beam was cut obviously at the output port of MZ. The cut position is shown in the attached figure 1.
There is also a very good point to check if the beam is cut or not. In the attached figure 2, the shown place is a round shape. But when the beam was cut, I saw clearly there is an unfilled corner.(I am sorry I didn't realize to take a picture at that time)
In the attached figure 3(cut) and 4(no cut), you can see the difference before and after solving the problem of cutting. The higher order mode becomes a lit bit lower. After that we tried to change the offset of MZ servo, then we found we could reach GRMC transmitted power of 70mW. This meets the requirement of high BAB amplification for filter cavity alignment.
MZ servo offset | green power(GRMC tra) |
---|---|
4.7 | 50mW |
5.1 | 70mW |