I upload the circuit diagram of the box used in the original setup from the SPTS company. (they made it)
I'm still using it in the current setup, but since the InGaAs PD has a higher Current/Power responsivity, It could be that I should reduce the load resistance and send more power on the PD. If it helps to reduce the noise, I will make a new box for the InGaAs PD.

Using the Yokogawa2400 spectrum analyzer I measured the correlation between the signals coming from the two photodiodes:

- PM100D (incident power 85mW)       DC 470mV, analog out
- DET10N (incident power 15microW) DC 104mW (load resistance 6kOhm) 

The signals don't look coherent. It could be that they are really not coherent, or that the spikes at high frequencies invalidate the spectrum measurement.

After having disconnected the potentiometer of the ramp in the actuator section of the rampeauto (see previous entry # 875) we have further modified the rampeauto to reduce its noise. Following Pierre's instructions, we have soldered a 2.2K resistor in parallel to the resistor R24, in order to reduce the gain of the PA83 amplifier and then the effect of its input noise. In practice, since the resistor R24 was difficult to access, we have put the resistor in parallel with C23. 

- The attached PDF shows the electronic scheme of the rampeauto (actuator part). 

- Figure 1 and 2 show the rampeauto noise taken at the PZT output without any input signal. We see that the rampeauto noise is reduced by a factor ~ 20 in the kHz region, as expected. 

- Figure 3 shows the error signal of the filter cavity (green) with this new configuration, compared with the old ones. Note that, since now the gain of the P83 is reduced, we should increase the gain of the loop increasing the signal at the input of the P83 with the potentiometer "PZT gain". With respect to before, the error signal at high frequency is reduced by a factor ~ 8. 

- Figure 4 shows the error signal of the filter cavity (green) compared with the noise projection of the laser frequency noise (7.5 kHz/f [Hz/sqrt(Hz]) and rampeauto noise. Both the laser frequency noise and the rampeauto noise have been filtered by the pole of the cavity (2.4 kHz) and multiplied by the open loop transfer function of the loop, which assumes a 1/f behaviour at f> 3 kHz and a unity gain frequency of 10 kHz. We see that with this new configuration the error signal of the filter cavity (green) is explained very well by the laser frequency noise. 

Note that after the modification of the rampeauto it was not possible to obtain a ugf larger than 10 kHz. This should be understood better. 

I repeated the same measurements with another spectrum analyzer, the Agilent35670a

This instrument spectra are much cleaner, I did only 32 averages instead of 256 with the Yokogawa sa2400

First plot: gain 1000 in-loop pd: PM100D

Second plot: gain 1000 in-loop pd: DET10N

Third plot: gain 5000 in-loop pd: DET10N

Participiant: Matteo.B, Yuefan and Yuhang.

Yesterday, we installed a photo dector(PD) to monitor the reflection of green mode cleaner. This PD was put in the path which is the pick off of green reflection. And the monitor point has almost all the reflected power.

Then we scan the mode cleaner by sending a ramp signal to the mode cleaner. Actually here I made a mistake which was mentioned by Matteo.L before, but I didn't pay attention. It is I use a very large ramp amplitude, which is 700mV p-p. In that case, the scan is very fast. So I cannot see the reflection peak. See attached Figure one.

Then we decrease the ramp amplitude to 100mV. And we can see the reflection peak.

We can see from the figure 2, when we are in the resonance, the reflection is around 55%.

We compare the green error signal spectrum before and after the disconnection of Ramp injection point in the circuit. We can see the noise level is reduced by a factor up to 3 from 2kHz.

I mwasured the power loss due to detection point is not exactly the point of reflection and transmission.

For reflection, we lose 1%(0.12mW) of power. For transmission we lose 0.8%(0.097mW) of power.

So the loss of 2.56mW cannot be explained by this.

Manuel, Matteo B, Eleonora.

I recorded a first video of the signal of the photodiodes on the oscilloscope without control loop .
Channel1 (yellow) is the out-of-loop PD (DET10N) with a load of 6kOhm that gives 104mV of DC.
Channel2 (blue) is the in-loop PD (PM100D) analog output. DC = 470mV
The noise is mostly at high frequencies. The PM100D shows some spikes, while the DET10N doesn't.

The loop consists of a subtraction of an offset from a waveform generator, then a low pass (order one) at 100Hz and a Gain of 1000 (using the sr560)

I recorded a second video around the moment of closing the loop. On the oscilloscope there are the signal of the in-loop photodiode (blue channel), and the correction signal (yellow channel).

It shows that the in loop PD increases the noise when closing the loop. This is because the noise is mostly at high frequencies and the noise reduction at low frequencies doesn't show clearly in the time domain.

Plot1:  I made the spectrum of the 2 PDs in 3 conditions: dark, loop open, loop closed
It shows that the loop works for the inloop PD, but it doesn't for the out of loop PD.
 

Plot2: I exchanged the two PDs, so in this case the out-of-loop PD is the PM100D and the in-loop PD is the DET10N.
I copied the dark and openloop curves from the plot1 and I updated the closed loop curves. In this case I could increase the gain to 5000.

Participiant: Yuefan and Yuhang

Today we locked the MC and Mach-Zehnder(MZ). We found some problems while locking.

1. The lock of MC sometimes has drifting. It brings the locking point to a lower point.

2. The error signal for MZ locking has a fixed frequency oscillation.

3. Sometimes we can see the MC transmission goes away very large.

We tried to make sure the MC transmission has the same level and then make the measurement.

The second cloumn is the averange power. The third cloumn is the standard deviation normalized by the averange power.

Note: the end mirror loss is diffcult to measure, so I take only one measurement.

So reflection is 43.48%, transmission is 44.24%, end mirror tans is 2.96%. We loss 2.56mW (9.32%).

Please refer to wiki page.

https://gwpo.nao.ac.jp/wiki/FilterCavity/OpticalLayout

You can download if you want. Refer the instruction from the attached figure.

[Eleonora, Matteo B.]

We have measured the rampeuto PZT output noise by connecting the PZT output to the spectrum analyzer. It is about 2 uV/sqrt(Hz) at low frequency and it starts to decrease above 5 kHz, as can be seen from pic 1 (blue curve) 

We also observed with the oscilloscope that there is an offset of -10mV. (pic 2).

In order to reduce the rampeauto output noise, following the advice of Pierre, we have disconnected the ramp input (which normaly is not used for suspended cavities). (pic 3) 

The output noise has been reduced by a factor 3 and now  it is about 800 nV/sqrt(Hz),  below 5 kHz.  (see pic. 1 and 4)

In pic 5,  we have compared the rampeauto noise (actuation noise)  with the free running laser noise and found that it becomes dominant above 2 kHz.

For the conversion to Hz we used a piezo gain of 2 MHz/V, as measeured in entry  #859.

The next step is to re-measure the phase noise, as done in entry #863, with this new configuration.

Participaint: Eleonora, Matteo Barsuglia and Yuhang.

As we found before, the transmissivity of mode cleaner is not as expected. We also found for different Finesse requirement, we can have different tolerance of reflectivity error.

According to the measurement result of entry871 and entry548, we firstly get the reflectivity of mirror (r) and then we use this result and scan the reflectivity amplitude from 0 to 0.13.

We put this difference to each mirror as delta/2, for example we use r_1=r+delta/2, r_2=r-delta/2. However, when we convert this amplitude difference to power difference, we bring back a factor of 2. So they balance.

We also plot the case for Finesse as 100. As the experience of Matteo Barsuglia, this Finesse can give us 99% of clean to have TEM00. And we can increase reflectivity difference tolerance dramatically. For details, refer to attahced figure.

We also investigate the reason of higher transmission with lower Finesse. The reason is we loss some of the power through the end mirror of mode cleaner. The higher the Finesse is, the more losses we have.

Participaint: Yuhang and Eleonora

It is necessary to fit the opto-mechanical transfer function and filter design, otherwise we cannot use zpk function in matlab. We did this by using proper poles and zeros in proper frequencies. 

The zeros I used (corresponding q factor): 7250,13800,500,14600,25500,42000,31000,18100,21500(25,14,0.01,4,15,40,35,12,5)

The poles I used(corresponding q factor): 7000,12000,5000,8000,14800,16500,18900,26500,30000,34000,35900,46000,44900(12,15,0.03,0.6,6,18,10,2,10,10,2,0.8,30)

The gain is 28 at 100Hz.

I tried to put a low pass filter at 200Hz(q=0) with a gain of 20. And an intergrator at 0Hz(q=0). Finally a notch filter between 6950Hz(q=6) and 7550Hz(q=10).

Then I got unity gain frequency of 376Hz with 34 degrees of phase margin. The plot is in attahced Fig. 2 and blue line is data, red line is OLTF, green line is filter.

Manuel, Eleonora

We plotted the transfer function of the measured actuator (plant) and fitted it with a Matlab script based on the zpk function.
We used two simple poles at 40kHz, a gain of 0.16 in DC, and a delay of 2.3e-6. see the first plot

We plot the measured open loop TF obtained using a sr560 set with a first order low pass filter at 100Hz, a gain of 200.
We fitted it as the product of the modeled plant TF and a filter TF.
The filter that best fits the data is a first order low pass at 100Hz  with a gain of 240. see the second plot
The UGF is at 3.8kHz with a phase margin of 76deg.

We verify that the loop becomes unstable for a gain of 2000 (as observed experimentally).
Indeed, the UGF becomes 23kHz and the phase margin 7deg. See the third plot.

We will use this model to design a better filter in order to have more gain at low frequencies, compatibly with the possible configurations of the sr560.

I set the loop to control the intensity of the 1310nm laser.
first I measured the actuator transfer function with a random noise from the spectrum analizer of amplitude 200mV

Then I set a low pass filter at 100Hz, a DC offset from the waveform generator to keep the correction signal around 0, and a gain of 200.
I closed the loop and measured the open loop transfer function.

I measured the noise in the photodiode with the loop closed and without loop.
then I divided the noise without loop by (1 + the open loop transfer function) and compared with the closed loop pd noise.

If I increase the gain above 1000 the laser stops for exceeding the current limit.

Participant: Marc and Yuhang

Today I reveived the comment from Marc and I measured the Finesse again. This time I put the ramp signal to make sure I am looking at the correct part of signal. This is of great important to find the correct FSR(Actually here the fsr corresponding to time, so maybe not good to call it fsr)

Then I use this fsr to fit only one peak so that I can see more clearly the fit.

In the end, I got a more pleasuible result. Finesse is 381. But this result is much higher than the calculation result, which is 248.

Participant: Eleonora and Yuhang

Beccause we changed many cables and arranged the control devices space, we think the green should be misaligned. Yesterday, I checked that it is totally misaligned!

I did the procedure of standard alignment of green: check target on PR chamber, check the first iris, check the second iris and match incidence with reflection, finally end mirror. After these procedures, I got the new offset of each mirror's local control offset. Then I can lock green and infrared together. See attached Fig.1

I also found some problems and did some change during this process.

1. I connect the image of second iris to the third part of monitor. I also checked the light you can see in this second iris camera is corridor light. Because if I go to turn off the corridor light, it disappeared. See attached Fig.2

2. If you find problem in attached Fig.3, you need to reopen the program.

3. I tried to remote control of the second iris, but the network cable seems not working. 

Because we found the unexpected peak while we are scanning mode cleaner, I tried to realign it.

The mothod is to block the beam going to mirror without PZT in Mach-Zehnder. Then use the steering mirror to align and make the peak of mode-cleaner as high as possible. Then remove the block, and adjust the mirror without PZT in Mach-Zehnder.

After alignment, the unexpected peak becoms very small. So I tried to take the data and calculate Finesse. Firstly, I tried to fit with airy function. But I found the FSR here is very strange, see attached Fig.1. It's obvious that the software cannot tell us this is a good fit. So this fit is done by my hand. I don't know why this FSR can be this unstable. But maybe this can be interesting.

But I tried to fit only one peak, then I get the fit result of F=15100. See attached Fig.2. But note here, we use PD with amplification, it is 40dB means 100 times amplification. So the measurement of Finesse is only 152. If we consider R=0.992, the Finesse should be 248. So there is this discrepancy.

Beside, I checked the polarization again. This time I put a half-wave plate infront of mode cleaner, and change s to p polarization. I saw a increase of larger than 10 times of transmission on oscillscope. I talked with Matteo. From Fresnel law, p-pol has more transmissvity for the mirror now we use to dump beam. This is the main reason for this increase.

I checked the input and output mirror. From the point view of marker on the side of mirror, I am sure the mirror is installed in a correct way.

I checked also the mirror from the same box, this arrow points to the HF side of this mirror.