It was pointed out that the measurement of oplev p/y spectrums are not meaningful. So I went to check again what can be the problem. In the end, I found the laser head of the oplev is totally loose. After I fixed it, I took spectrums again. Figure 1 shows these spectrums.

Besides, I also put some pictures of the KAGRA QPD set-up. (note that we could close the black box, pictures are just taken with box open) (we will use a better box in the future if we really use this QPD)

I tried to close BS local control with KAGRA QPD.

The comparison of pitch/yaw spectrum when loop is open or closed is in the attached figure.

Michael and Yuhang

We checked the QPD signal from each segment, which is consistent with the matrix we set for channels.

We also compared with TFs of BS pitch/yaw with old measurements. This comparison is in figure 1 and 2. From this measurement, it seems that QPD is working well.

The detuning frequency has been set in elog2296 by Aritomi-san. According to that value, I calculated several different numbers which could be used to set even higher detuning.

According to elog2350, the detuning values in this elog should be 105Hz, 115Hz, 125Hz, 135Hz, 145Hz, 155Hz, 255Hz.

Michael and Yuhang

Since we always have TAMA old PSDs broken and suspect AA noise coming from PR/BS oplev, we decide to buy new PSD/QPDs. Akutsu-san provided us two QPDs from KAGRA, we report the test of one of them in this elog.

The tested QPD is JGW-D1402411-v2.

We powered it with a DC voltage supply, the current consumpation is 0.069A for -15V while 0.084 for 15V.

We set this PD for BS oplev and compared with the measurement from TAMA PSD. Figure 1 shows this comparison. We could see that a better SNR can be achieved from KAGRA QPD. Besides, if we used a higher gain in KAGRA QPD, we gain even better SNR below ~50Hz. However, one strange thing is that the yaw spectrums show different peak structure for QPD or PSD.

Marc and Yuhang

On Monday, we found CC PLL couldn't be locked. We checked the RF signals, LO signals, locking loop set-up and correction signals.

LO signal is about -2dBm. Locking loop set-up was reloaded. Correction signal seems to be saturated. In the end, we found the RF signal was very small.

Although this may be related with just alignment, we decide to replace fiber since I remember there is a damage of CC PLL fiber (see fig 1).

The new fiber is fixed on top of AA breadboard, which makes it easier to be checked in the future. (see fig 2)

The RF signal reaches PLL lock loop is about -8.5dBm with new fiber and amplifier. (see fig 3)

The monitor channel shows -33dBm on the lower spectrum amplifier (see fig 4). Note that the fiber PD used for monitor is battery powered, which seems to be able to provide only RF signal.

Michael and Yuhang

Yesterday, we locked AA, Zcorr, pointing and CCFC and performed a FDS measurment. The lock last for almost one hour. 6 measurements were performed during this time (with a separation of almost 10min).

The result is shown in the attached figure 1.

Within this 1 hour, all measurements overlap very well. Even the low frequency back scattering seems to be more stable.

(50:50 BS was used for CCSB pick-off. 30mW pump power is used for SQZ production.The detuning was expected to be 104Hz according to elog2353.)

(Figure 2 shows a FDS measurement of detuning smaller than 100Hz, it seems backscattering makes FDS not reaching shot noise level. This is also why we took several measurement with higher detuning.)

(Figure 3 shows a fit for higher detuning. Since the detected detuning is 145Hz, we expect all the frequencies to be increased by 40Hz in elog2353. )

Marc and Yuhang

We checked OPO CC power in transmission with different green power injected.

The injected CC power is 1.52mW. In transmission, 3.9uW of offset has been removed (coming from p-pol).

Reflected CC power (no green) is picked out by a 25:75 (R:T) BS. The pick-off power reaching CC1 PD is about 133uW, which is strange. Since we know that OPO makes 99.7% CC reflected and 25% of pick-off, we expect 379uW reaching CC1 PD. 

Marc, Yuhang

Yesterday, when we just entered the cleanroom, we realized that the main laser head was quite hot (by hand). However, we found the laser was fine (the laser was on, and we could lock SHG). So we didn't care so much about that.

However, after we used the main laser for a short period, the main laser suddenly went off. Almost one week ago, we reported the main laser sudden off. We didn't know the reason. In total, we found three times of laser off in the recent month.

After the laser suddenly off, we found that we couldn't turn it on by pressing the switch. So we put a thermal meter on top of the laser head and wait for a while. Figure 1 shows the laser head temperature change during this waiting time. Then we turned the main laser key off. After turning it on, the laser starts to work well.

Recently we found another issue related to the main laser. More than two years ago, we did an investigation of our main laser stability (elog). After three hours of operation, laser power becomes very stable. But now, even after one day, the power stability is not very ideal. Figure 2 shows this laser power change. Although this is not a big issue for squeezing production since its frequency is quite low, this may mean we are having small problems.

After switching the laser on, we monitored laser power again and found laser head temperature increases back. Figure 3 shows this laser power change. This temperature change is related to the laser power change since they have almost the same frequency. We will check this coherence in the future.

Marc, Michael, Yuhang

Matteo ordered BSF10-C to pick off CCFC error signal from filter cavity reflected squeezing. This mirror should take about 2% power, which means small optical losses for not degrading squeezing field. Today, we replaced the old 50:50 BS with this BSF10-C. We report several check we did here.

1. By using BAB and power meter, we checked power splitting ratio of BSF10-C. Incident power is 325uW, reflected power is 7uW, transmitted power is 321uW. From this measurement, the power loss is about 1~2%.

2. We checked CCFC sideband. When 40mW green is incident inside OPO, we have CCFC sideband as figure 1.

3. We compared CCFC error signal before and after replacing beam splitter. Figure 2 (after) and 3 (before) show them. We could see that error signal becomes not usable after this replacement.

4. We tried to amplify CCFC signal from photodetector, we got new sideband as figure 4. But, as figure 5,  the demodulated error signal just became overall larger.

I also attach here, as the last figure, how much FDS level we expect when we use different pick-off beam samplers. But since we just want to demonstrate this technique, maybe the current 50% BS is enough to see squeezing and stabilized detuning.

Marc (remote), Eleonora (remote), Yuhang

In elog2341, we reported the INPUT oplev was not set-up properly. I checked signal connection and found that input and output of SR560 were swapped. After solving this problem, INPUT oplev was recovered. (SR560 used gain as 100 and no filters) In the end, I compared INPUT oplev signals. Figure 1 shows this comparison, which shows a lower noise level.

Michael, Yuhang

After installing RF amplifiers and USB switches, we worked on the recovery today. 

1. We reboot standalone to solve the problem of timing

2. We delete some second trend and released 10% space (this corresponds to about 1 month data)

3. We checked PLL p-pol frequency changed to 265MHz (53) with OPO 7.162 temperature and no green. It was 240MHz (48). In this case, BAB power before PBS is 0.46mW.

4. The mode matching from BAB into filter cavity was checked to be better than 93% (520 counts was observed for FC BAB tra)

5. Then we found main laser turned off by itself. We didn't notice how it happened. Then we checked dataviewer. The PR/BS/INPUT/END oplevs, FC GR/IR tra, FC GR correction/error were checked. Figure 1 shows this check. From this check, we could find

• The first change happend for FC GR/IR tra, FC GR error suddenly.
• PR oplev didn't have any change. BS oplev changed gradually due to pointing loop. INPUT oplev didn't change because of its wrong set-up. END oplev didn't have sudden change, either. So this main laser sudden off should not come from suspension.
• FC GR correction looks not that sudden compared with GR tra. So this problem seems also not come from a sudden large laser frequency or cavity length change.

So it is still not for sure why we see this change.

Michael and Yuhang

In elog2336, we checked RF signal generated from DDS2, we found some sidebands around the generated RF signal.

Today, we checked RF signal from DDS3. Attached figure shows this check.

We could see that this peak is much cleaner.

Marc, Michael and Yuhang

After yesterday's investigation, we found although RF signal is not very clean, but it doesn't have large noise. So we were thinking the 20kHz noise should be just at low frequency and it goes to many other places. For example, elog2330 shows this noise from DDS filter-out even when there is no signal. Besides, elog2331 shows this noise from power supply of DDS board. An important fact was ~20kHz noise is only related with the connection of DDS board.

So we tried to connect individual voltage to DDS board one by one. Then we found out that ~20kHz noise shows up only when voltage is provided to USB. So we could infer that ~20kHz noise comes from USB voltage supply. Although we still don't quite understand what is the exact reason of introducing this noise, we could avoid having this noise by disconnecting USB voltage supply. This is also feasible because we don't need to connect USB so often.

Then we found a solution. We decide to use a switch, which decide whether the voltage will be provided to USB or not. But note that don't connect all USB if the DDS software is open. This is also good because this makes it easier to operate DDS boards. Before this modification, we need to change USB connection by hand if we want to control different DDS board. Now it becomes easier, we use switch to decide which DDS board to be controlled. If we don't need control DDS, we need to switch USB voltage off for avoiding noise.

After applying switches to DDS boards, we did comparison with USB voltage off and on. Figure 1 shows this comparison. We could see that noise frequency is changed. But anyway, if USB voltage is off, we will have 'clean' SHG error signal.

For reference and future similar investigations :

     The power supply was tested using a DC block (here up to 50V).

     The ground was tested with the power supply disconnected.

Marc, Michael, Yuhang

We checked the resistor and noise spectrum of several points on DDS board.

Brief introduction of DDS board configuration: Power supply (24V) > voltage regulator (1.8V goes to generate signal, 3.3V goes to power up chip/USB) > 1.8V > AVDD point> DDS process > filter-in point > 200MHz low pass filter > filter-out point > to be used

Figure 1~3 show resistors from ground to:

AVDD: 0.83kOhm

filter-in: very high (seems to be not connected)

filter-out: 2.7Ohm

Figure 4 shows the noise spectrum of those points and power supply.

We could see that power supply is clean. But once DDS board is connected, almost every point inside DDS board show ~20kHz peaks.

Marc and Yuhang

To understand if ~20kHz noise is present as sidebands of RF signals, for DDS2, we checked RF signals coming from CH0 and CH1 with spectrum analyzer.

Figure 1 shows 78MHz signal coming out from CH1 on spectrum analyzer. For this singal, we could see:

1. It has noise shoulder, which is between 0~3kHz.

2. We don't see any noise peak around 20kHz.

Then we checked also CH0, which shows the same behavior.

Since we don't see 20kHz peak around RF signal, we tried to use CH0 to demodulate CH1 and we did't expect to see ~20kHz peak.  (Note that RF amplifier was used for CH0 in this measurement) Figure 2 shows the demodulated signal spectrum. Peaks around 20kHz show up clearly in this spectrum. This means that these peaks doesn't come from RF signals but they were just there (actually everywhere).

Marc and Yuhang

We checked again the connection from ground to power supply or signal output today. And we realize that we were not checking in a good way last week.

This time, we measured the resistor between ground and many other parts. We found the resistor is as high as from 0.3kOhm to 0.9kOhm. In this case, it means the ground is well isolated with those channels. So there should not be ground issue.

Participants : Marc, Yuhang

To test the effect of the amplifier we checked the spectrum of CH0 when disconnecting the amplifier :

    FIg1 shows the CH0 spectrum when the voltage supply of the amplifier is disconnected but amplifier is still connected to CH2.

     Fig2 shows the same with the ground also disconnected.

     Fig3 shows the CH0 without the amplifier at all.

In all these configurations the peak around 20kHz is present : The 20kHz peak does not arise from the amplifier

We used the spectrum analyzer of the elec shop to check if the 20kHz peak is present in the voltage supply of this board.

     Fig4 shows in yellow CH0 and blue the 1.8V supply : same peak in both at 17.4 kHz

     Fig5 shows in yellow CH0 and blue before the regulator of the 1.8V supply : same as previous

     Fig6 shows in yellow CH0 and blue the 24V supply : same as previous

Finally, we found out that the ground may be connected to all voltage supply of this board as well as DDS output channels (see connection.mp4)

Participants : Marc, Yuhang

We locked the SHG and connect only DDS3 to its power supply. The peaks are still present on the SHG error signal.

We checked each output channel of this DDS3 with the spectrum analyzer :

     No peaks present from CH0 nor CH1 (Fig1).

     CH2 exhibits a peak at 21.957 kHz (Fig2) at exact same frequency as the peak on SHG error spectrum ! (note that the peak freauency changed with respect to yesterday measurement. Also we checked that it changes by few hundred Hz within ~10mn)

     As the CH2 output is connected to a splitter, we also checked the splitter second output which also exhibits a similar peak (Fig3).

    We disconnected all outputs from DDS3 but the peaks are still present on the SHG error signal (Fig4).

Remind that we added an amplifier to each channel of DDS3 which could mean that the problem does not arise from the amplifier itself. Only difference between these channels output is that CH0,1 and 3 have attenuators at their output but not CH2.

     We removed attenuator from CH0 and a peak at around 20 kHz appeared (fig5).

We then tested DDS2 (as it has only 1 amplifier).

SHG ERR signal exhibits peak at 17.4kHz (Fig6)

    CH1 (amplifier, no attenuator at its output) : only harmonic visible in Fig7

     DDCH0 ( no amplifier, no attenuator) : clear peaks in Fig8

The conclusion seems to be that the amplifier is not the culprit.