NAOJ GW Elog Logbook 3.2
We found on the electronic board of DDS that there is port for non-filter output. So we did some measurement. Although we can detect some signals, like -78dBm or -83dBm. It really means there is no effective output. The signal we detected is the signal leak from other channels.
2.Radio Frequency(RF) signal parameter
RF for Second Harmonic Generator(SHG): Frequency is 15.2MHz, Amplitude is 799mV(pp)
RF for Filter Cavity(FC) demodulation: Frequency is 78MHz, Phase is -0.08, Amplitude is 8.5V(pp)
RF for FC modulation: Frequency is 78MHz, Phase is 109.16, Amplitude is 1V(pp)
RF for Acoustic Optical Modulation(AOM): Frequency is 110MHz, Amplitude is 11dBm.
Amplifier for AOM: Mini Circuit ZHL-2(the gain according to data sheet is around 18dB)
3.Locking measurement
We accept the suggestion of yuefan to check the beam size. And we can see the peak value of transmitted power is around 1.2V as before. The error signal is around 4.7V peak to peak. After locking, the transmitted voltage is around 390mV.
Note here: The first and third value is from the oscilloscope around PC. The second value is from the oscilloscope on the rack. They have roughly 10 times difference for the same signal. We will check later where it is from, like oscilloscope setting or cable difference.
4.Before this week's locking, I found there is also drift. But it happened while the local control is off(see attached photo). And it is really similar with the photo I tool last time. I will check after this weekend where the beam will be next week.
We made a drawing of the clean booth in Tama, in orange lines.
I addition to the current clean booths, we drew a clean boot 2000x3000mm on the squeezing table and a 2000x2000mm on the PR tank.
We can see from the figure attached that DDS has something like a low pass filter inside.
I didn't know the drift problem before. I took the picture at laser injection window of PR chamber. You can see from Fig.1 the drift after around 12 hours. I remember that it will not happen when the local control is off.
After fixing them, we still cannot fix the problem. We discussed this during the meeting and Eleonora suggested to decrease the gain. So I decreased it from 100 to 20. This is really effective. We can easily lock the cavity then.
But the problem is the locking of TEM00 is not very good. For transmitted signal, we can see from the oscilloscope that the peak is around 1.2V. But after locking, the transmitted signal can be only 0.4V in the best case. Besides, we can lock with higher-order modes and they have almost the same level with TEM00. This should come from the bad alignment. But I don't know how to make alignment better. It's really difficult to find a good way to improve this. We can see almost no change on screen the modes, on oscilloscope by changing the local control offset of mirrors.
We also measure the noise spectrum of transmitted signal and error signal with different gains. Please refer to Fig.2 and Fig.3
Members: Yuhang and Tomura
On 09 December, we found that the end mirror of the filter cavity had relatively larger variation in its pitch motion. As refer to in elog 428, the normal angular variation for the end mirror pitch with the loop closed was 3-4 urad. However, values we found there were 6-7 urad. As for the other mirrors (PR, BS, and IM), these values seemd decent or even better. We suppose this pitch noise on the end mirror can be one reason why we cannot lock the cavity although the beam seems to be well aligned. The reason of the noise itself is still unknown.
Takahashi-san showed us how to switch off the facility. Basically the vacuum system and the air compressor.
For each experiment, people working in tama must take care if turning off the instrumentation correctly.
Members: Akihiro T., Matteo L. and Yuhang W.
Some remarks about the filter cavity locking
On 07 December, we tried to lock the filter cavity at first, and then we found some problems.
When we switched on the loop, the yaw of IM mirror went to auto-oscillation.
Trying to figure out where the problem came from, we measured the meachanical transfer function and the open loop transfer function of each mirror (pictures to be posted later, sorry).
As a result, the MTF and OLTF measured had similar form. We consulted Eleonora for this problem and she advised us that it was possible that the loop filters were not applied on the LABVIEW program which we usually use. This was because the PC was wrongly shutted down the other day. When the PC is shutted down, a program which offers us the loop filters is also shutted down. In this case, one need to boot all programs again correctly. Finally, after applying the loop filters, the whole system (the filter cavity) seemed decent.
In the case of re-booting the PC, please boot the LABVIEW programs used to manipulate yaw and pitch of each mirror,and also a program named "my_Filter_Bank.vi" and load a file "filtrotot.txt" in it.
AOM miss-alignment
We also found that the first order diffraction of the green beam, which is inserted to the filter cavity, had less power. It is posssible that AOM is not alingned properly. We did rough measurement of beam power at some different points (shown below).
Right after SHG cavity: 57mW
Right after EOM: 49mW
After beam spriter before AOM: 9mW
Before iris in front of the chamber: 8mW
After the iris (the 1st order diffraction): 0.25mW
Members: Matteo L., Yu-hang Wu
On monday 6th of November several measurements has been performed. Here a list of the measurements and the results with some discussion.
HV driver TF as function of the noise amplitude:
To understand if the low pass filter present in the HV driver transfer function (logentry 582) comes from a bad slew rate of the HV internal op amp, few measurements of the TF as function of the noise amplitude has been performed. The result is presented in Fig.1. If we compare the two measurement presented with the one done previously we see that the pole of the LP filter is unchanged with respect of the noise amplitude and it is always at 77Hz. Therefore the bad slew rate of the op amp does not seem the source of the low pass filter.
High pass filter and HPfilter+HVdriver characterization:
To try to remove the pole in the HV transfer function we added an active, first order high pass filter at the input of the HV driver. The schematic of the HP filter is presented in Fig.2 and Fig.3 shows its characterization. After inserting the HP filter at the input of the HV driver the transfer function of the system has been measured again and it's presented in Fig.4.
Fig.5 shows a recap of the HVdriver TF before and after adding the HPfilter.
SHG characterization after the modification of the HVdriver:
After adding the HPfilter the SHG's open loop and cavity TFs has been measured (Fig.6 and Fig.7). In order to have a stable loop the parameters of the servo (SR560) were changed: Gain=200 (before =1000), f3dB=10Hz (unchanged), invert=ON (before =OFF). Inverting the error signal was necessary due to the 180deg phase delay introduced at low freq by the HPfilter (the op amp is in inverting configuration).
All the TFs presented are pieced together from traces with different frequencies span. Only in the SHG cavity TF (Fig.7) there seems to be some problems going from one span to the subsequent. I don't think this issue is cause by any physical effect but further investigation will be performed.
As a result of adding the HPfilter to the loop chain the SHG cavity TF is reasonably flat below the PZT resonance and the unitary gain frequency of the loop has been increased from approx. 1kHz to approx.4.5kHz. In the next days we plan to play with the servo parameters to improve the unitary gain freq. and the loop stability.
Huge dT guess:
After the previous work, the time strip-chart of the produced green and the SHG IR transmission, as seen by two phodotiode, has been recorded (Fig.8). As it's clear the low freq fluctuation is still present. To better understand what is the dominant factor in this low freq noise the thermal control of the LiNb crystal has been swithed off and the two signals has been recorded again (Fig.9). In this situation (Tcrystal = Troom) the SHG cavity does not produce any green but the transmitted IR si much more stable, therefore we suspect that the crystal thermal control while operating at the phase matching temperature introduces a dT noise that the cavity lenght control is not able to compensate.
We plan to investigate in the next days on the thermal control in order to measure the temperature stability and to improve the temperature stabilization.
The High voltage driver (first picture) is from Matsusada Precision (PZJ-0.15P-LVS2) and the datasheet can be found at the following link: https://www.matsusada.com/pdf/pzj.pdf
The piezo used for the SHG cavity (first measurement) is from Piezomechanik GmbH (HPCh 150/15-8/3) and has a capacity of 790nF.
The piezo used for the IR mode cleaner (second measurement) is from PI (P-025.20H PICA) and has a capacity of 430nF.
We measured two HVDs. One is HVD used for SHG, the other is a new SHG. Each trace data is taken with two different frequency spans and overlapped. This is to make measurement precise.
As you can see from below, the TF of HVD truly has feature like low-pass.
The High voltage driver (first picture) is from Matsusada Precision (PZJ-0.15P-LVS2) and the datasheet can be found at the following link: https://www.matsusada.com/pdf/pzj.pdf
The piezo used for the SHG cavity (first measurement) is from Piezomechanik GmbH (HPCh 150/15-8/3) and has a capacity of 790nF.
The piezo used for the IR mode cleaner (second measurement) is from PI (P-025.20H PICA) and has a capacity of 430nF.
From the experiment we had done, we had got an open loop transfer function of the SHG control loop. To optimize this control loop, we changed the parameters of the servo (corner frequency and gain) monitoring the open loop transfer function, its unity gain frequency, and phase margin, using the network analyzer (Agilent 36540A).
I show several pairs of parameters below. On the 7th column we have 224Hz unity gain freq. and 30 degree phase margin, it seems somewhat better than the others.
With a gain greater than 2000, system became unstable.
Still, there need to be more investigation.
| Corner freq.[Hz] | Gain[dB] | Unity gain freq.[Hz] | Phase margin[degree] |
| 10 | 1000 | 1000 | 1.5 |
| 10 | 200 | 512 | 11 |
| 10 | 100 | 352 | 14 |
| 10 | 50 | 224 | 23 |
| 3 | 1000 | 672 | 0.6 |
| 1 | 1000 | 384 | 15 |
| 0.3 | 1000 | 224 | 30 |
| 0.1 | 1000 | 96 | 42 |
| 0.03 | 1000 | 32 | 59 |
We used Agilent 35670A network analyzer.
All traces are pieced together from traces with different frequencies span.
Firstly, we investigated frequency response of an adder we used (1st figure).
We got a flat response in magnitude and phase.
Secondly, we took dark noise spectra of the adder and Agilent 35670A (2nd figure).
Thirdly, we measured a mechanical transfer function of SHG (3rd figure).
We guessed a peak around 25kHz indicated a resonance point of PZT.
Lastly, we measured an open loop transfer function of SHG (4th figure).
Unity gain frequency is around 1kHz.
| [urad] | |
| PR yaw | 0.78 |
| PR pitch | 1.40 |
| BS yaw | 2.21 |
| BS pitch | 4.31 |
The reference for the residual motion can be found in the previus entry n.428
The original resistor was 51 - 1W and the resistor now is 5k1 - 1/4W. The gain of that channel increased by a factor of 100 as expected.
Today, I made some measurements about the green light. The process is
Firstly, the green light was stable. Then I sent message to Matteo, and Matteo thought it's saturation. Actually this is what I and Tomura-san found last week. So I decided to use the filter to check it. And I found Matteo was right.
Secondly, the green light became unstable. I measure the noise spectrum at this moment. I want to see what will happen after a while.
Thirdly, after a while, the green light becomes stable again. I measure the noise spectrum again. This is different from the previous one certainly. Then I want to see what will happen after a while again.
Fourthly, after a long-while, the green light becomes unstable again. And I told it to Matteo, Matteo thought it could be cause by beat of two signals. So I checked it on the oscilloscope, it turns out that we can see slow variation of signal on oscilloscope(3rd picture). Indeed, it looks like a beat signal.
The main difference between unstable and stable signal is a peak around 2.7kHz. Matteo suggested to remove the thermal dissipator. After removing it, we measure noise spectrum again. It's almost the same with the stable spectrum we measured before.
Above all, we find that a reason of instability is the resonance of thermal dissipator. Actually, we can see from noise spectrum, the only difference is the peak around 2.7kHz. So there is also at least another reason of instability, it's caused by the peak around 650Hz. And we can see its harmonic peak. (But we cannot see this frequency fluctuation on oscilloscope clearly)
(Sorry for not using matlab to make plot. It's because of my saving of wrong format data.)
The first picture, it's the unstable noise spectrum.
The second picture, it's the stable noise spectrum.
The third picture, it's the long term signal in the view of oscilloscope(unstable).
The fourth picture, it's thermal dissipator.
The fifth picture, the same time scale with the 3rd picture of oscilloscope(stable). Actually, it just doesn't have something like beat or resonance. As you can see it's not stable for a long time.
I aligned for probe laser(1310 nm).
I changed the IU alignment and scanned.
I got the result like this figure.
In it, the measured peak AC signal value (~0.006 V) is lower than before(~0.02 V).
Also, secondary peak can't be seen well.
Next time, I have to align again in order to get a good result.
Pictures of the opened PR chamber.
3 mirrors have been added to obtain the IR and green references.
G1 is the green beam reflected by the dichroic
IR 1 is the IR beam after the Faraday Isolator and a folding mirror
GR2 and IR2 are reflected by the same mirror ( between the green folding mirror and the dichroic and between the 2 metallical structures).
They can be seen on the "PR references" picture.
I will try to upload a more precise optical scheme of this chamber
Using the simulations I compared the surface reference sample scan signal when changing some parameters.
First, I made the scan of the surface reference sample with the 633nm probe. The figure1 shows the signal AC/DC when I change the size of the detector.
Then I set 1mm of detector size and only change the wavelength and probe waist. Figure2
The interference a bit larger and the intensity is about half of the 633nm probe. This may be due to the different focusing alignment required for the 1310nm. However, it still doesn't explain the behavior of experimental data that have 2 large peaks instead of one.
Before moving anything we did the coil check again one by one, each of them has response to the noise and the amplitude of the motion is similar. Then we tried to adjust the picomotor in order to put the beam back at the reference we put outside the BS chamber, which is the transmission of BS mirror. Of course after moving the picomotor we should put the PSD signal back to zero, then when we tried to close the loop, we found even we put the offset as the value when the mirror is free, the correction signal is around 5 or 6(should be around zero). This means we cannot control the PR mirror anymore.
We tried to go on,by changing the local control offset, we were able to sent the beam to end and saw it on the CCD camera. Then as usual the next step should be align the reflection of the input mirror to the injection beam. But we cannot see the reflection as the nearest window of the bench, so we check the reflection beam on the 2 inch mirror to align it. Yaw of the reflection looks fine, so we only tried to adjust the pitch, the range we can move of the pitch is between -3.5 to 4. This is the normal range, but with this range we are not able to bring back the beam at the good position.The possible reason should be the beam was not well aligned, so with the PR picomotor and BS local control, we tried to align the beam better, at the same time looking at the beam at the end for not losing it. But then we reach the limit of BS local control, we are not far from well-aligned. Actually during our adjustment, the BS range was slightly get smaller, which is very strange.
V
In the attachment, it is the test we did to each magnet, when we sent a sine wave with amplitude 1V and frequency 1Hz, the motion of pitch and yaw.
The next step we are going to do :
1. check the open loop transfer function of two mirrors
2. try to bring back the beam on the BS mirror to the right height, which means we can receive the whole round beam with the two inch mirror we put at the back side of the BS mirror, and got the reflection of it outside the window.
