We first took data of a simple low pass filter, and attempted to perform a fit to both the magnitude and phase in order to find the Z of the components. Once we felt confident in our ability to measure tranfer functions, we took data and plotted the transfer function of the existing control loop of the AUX laser. What we found generally followed the trend of, but was lower than, 10^4/f, which is what we hoped to match, and also had a strange unexplained notch ~1.3 kHz. The magnitude and phase data both got worse after around 40-50 kHz, which we believe is because the laser came out of lock near the end of the run.
[Attachment 2 and 3] are the frequency response of the low pass filter, curves fitted using least squares in python.
[Attachment 1 and 4] is the same measurement of OLTF of the actual AUX circuit, and the control diagram pointing out the location of excitation and test point.
On Friday Cici and I set up the Mokulab to take readings of our loop. The aim is to characterise the PZT, in a similar manner as before, by exciting the circuit using our input noise (a swept sine) and recording the corresponding changes in the output. We used the MokuLab to observe the beat note created by the signals of the AUX and PSL, as well as the ASD of the output signal. The MokuLab simplifies the entire process.
Pictured : The beat note as observed by Cici
EDITED by YM on 22:11 June 27, 2022 to correct for a factor of two in the modulation index
Since we have measured optical gain in MICH to be an order of magnitude less compared with Yehonathan's FINESSE model (40m/16923), we measured the power at AS55 RF PD, and measured the modulation depths using Yarm cavity scan.
We found that 50/50 beam splitter which splits AS55 path into RF PD and RF QPD was not included in the FINESSE model. Measured modulation index were as follows:
TEM00 peak height: 0.6226 +/- 0.0237
RF11 peak height: 0.0067 +/- 0.0007
RF55 peak height: 0.0081 +/- 0.0014
RF11 modulation index: 0.208 +/- 0.012
RF55 modulation index: 0.229 +/- 0.020
RF11 modulation index: 0.104 +/- 0.006
RF55 modulation index: 0.114 +/- 0.010
Here, modulation depth m is defined in E=E_0*exp(i*(w*t+m*sin(w_m*t))), and m m/2 equals to square of the intensity ratio between sidebands and TEM00.
Power measurement at AS55 RF PD:
- ITMY and ITMX single bounce reflection was measured to be 50-60 uW at the front of AS55 RFPD.
- In the FINESSE model, it was expected to be ~110 uW with 0.8 W input to PRM (0.8 W * 5%(PRM) * 50%(BS) * 50%(BS) * 10%(SRM) * 10%(AS2) gives 100 uW)
- In AP table, AS55 beam was split into two paths with 50/50 beam splitter, one for AS55 RF PD and one for AS WFS and AS110. This will be included in the FINESSE model.
Modulation depth measurement using Yarm cavity scan:
- Aligned Yarm using ASS, and unlocked Yarm to get the 2sec scan data of C1:LSC-TRY_OUT_DQ, C1:LSC-POY11_I_ERR_DQ, C1:LSC-AS55_I_ERR_DQ.
- TRY data was used to get TEM00 peak heights
- POY11/AS55 data was used to find RF11/RF55 sideband peaks, and height was measured at TRY (see attached).
- If we define m to be E=E_0*exp(i*(w*t+m*sin(w_m*t))), the amplitude of TEM00 I_00 is proportional to J_0(m) and the amplitude of upper/lower sideband I_f1 is proportional to J_1(m), where J_n(m) is the bessel function of the first kind.
- m can be calculated using 2*sqrt(I_f1 / I_00).
- Results were shown above. Error is calculated from the standard deviation of multiple measurements with multiple peaks,
- The code for doing this lives in https://git.ligo.org/40m/measurements/-/blob/main/LSC/YARM/modulationIndex.ipynb
- Power at AS55 account for the factor of 2, In the FINESSE model, modulation index of 0.3 was used (could be m=0.3/2 or m=0.3; needs check). These combined can explain a factor of 3 at least (or 6).
- Gautam's measurement in Jan 2021 (40m/15769) gives almost double modulation index, but I'm not sure what is the definition Gautam used. It agrees with Gautam's measurement in Jan 2021.
The plan for the tools in 40m
As of right now, there are 4 tool boxes. X-end, Y-end, Vertex, and the main tool box along the X-arm. The plan is the give each toolbox a set of their own tools. The tools of X-end, Y-end, and Vertex toolboxes will be very similar containing the basic tools such as pliers, screwdrivers, allen ball drivers. Along with this, each tool box will have a tape measure, caliper, level, and other measuring tools we find convinient.
As for the new toolbox, I have done research and found a few good selections. The only problem I have ran into with this is the width of the tool box corresponding with the prices. The tool cabinet we have now is 41" wide. The issue I have is not in finding another toolbox of the same width, but for a similar price we can find a 54" wide tool cabinet. Would anyone be objected to making a bit more space for this?
How the tools will stay organized.
I the original idea I had was to use a specified color of electrical tape for each tool box. Then to wrap the corresponding tools tools with the same color tape. But it was brought to my attention that the electrical tape would become sticky over time. So, I think the using the label maker would be the best idea. with the labels being 'X' for X-end, 'Y' for Y-end, 'V' for vertex, and 'M' for main toolboxes.
An idea for the optical tables:
Anchal brought it up to me that it is a hassle to go back and forth searching for the correct sizes of Hex Keys and Allen Wrenches. The idea of a pouch on the outside of each optical table was mentioned so I brought this up to Paco. Paco also gave me the idea of a 3D printed stand we could make for allen ball drives. Does anyone have a preference or an idea of what would be the best choice and why?
A few sidenotes:
Anchal mentioned to me a while back that there are many cables that are laying on the racks that are not being used. Is there a way we could identify which ones are being used?
I noticed that when we were vented that a few of the chamber doors were leaning up against the wall and not on a wooden stand like others. Although, the seats for the chamber doors are pretty spacious and do not give us much clearance. For the future ones, could we make something more sleek and put the wider seats at the end chambers?
The cabinets along the Y-Arm are labelled, but do not correspond with all the materials inside or are too full to take in more items. Could I organize these?
To circumvent IPC error sending BHD DC PD signals from c1sus2 to c1lsc, DB9 cable from BHD DC PD box sent to c1sus2 is now split and sent also to c1lsc.
They are now available in both
C1:X07-MADC1_EPICS_CH16 (DC PD A) and CH17 (DC PD B)
C1:X04-MADC1_EPICS_CH4 (DC PD A) and CH5 (DC PD B)
- Add battery powered SR560 to decouple c1sus2 and c1lsc to avoid the ground loop
We attempted to use vectfit to fit our earlier transfer function data, and were generally unsuccessful (see vectfit_firstattempt.png), but are much closer to understanding vectfit than before. Couple of problems to address - finding the right set of initial poles to start with has been very hard, and also however vectfit is plotting the phase data is unwrapping it, which makes it generally unreadable. Still working on how to mess with the vectfit automatically-generated plots. In general, our data is very messy (this is old data of the transfer function from last week), so we took more data today to see if our coherence was the problem (see TFSR785_28-06-2022_161937.pdf). As is visible from the graph, our coherence is terrible, and above 1kHz is almost entirely below 0.5 (or 0.2) on both channels. Figuring out why this is and fixing it is our first priority.
In the process of taking new data, we also found out that the optical table enclosure at the end of the X-arm does a decent job of sound isolation (see enclosure_open.mp4 and enclosure_closed.mp4). The clicking from the shutter is visible on a spectrogram at high frequencies when the enclosure is open, but not when it is closed. We also discovered that the script to toggle the shutter can run indefinitely, which can break the shutter, so we need to fix that problem!
I took ~ 7 minutes of XALS beatnote data with the XAUX laser locked to the XARM cavity, and the XARM locked to PSL to develop an allan deviation estimator. The resulting timeseries for the channel C1:ALS-BEATX_FINE_PHASE_OUT_HZ_DQ (decimated timeseries in Attachment #1) was turned into an allan variance using the "overlapped variable tau estimator":
Where represents the k-th data point in the raw timeseries, and are the variable integration intervals under which two point variances are computed (the allan variance is a special case of M-point variance, where M=2). Then, the allan deviation is just the square root of that. Attachment #2 shows the fractional deviation (normalized by the mean beat frequency ~ 3 MHz for this measurement) for 100 integration times spanning the full duration (~ 7 min = 420 s).
The code used for this lives in Git/40m/labutils/measuremens/ALS/
If this estimate is any good, wherever the fractional beatnote deviation reaches a minimum value can be used as a proxy for the longest averaging time that give a statistical increase in SNR. After this timescale, the frequency comparison is usually taken over by "environmental instabilities" which I don't think I can comment further on. In our particular estimate, the 100 second integration gives a fractional deviation of ~ 0.44 %, or absolute deviation of 12.925 kHz.
I added today demodulation of C1:LSC-BEATX/Y_FINE_I/Q in the c1cal demodulation where different degrees of freedom can be dithered. For McCal (formerly soCal), we'll dither the arm cavity for which we can use any of the DOFs (like DARM) to send the dither to ETMX/ETMY. Then with green laser locked as well, we'll get the calibration signal from the beatnotes in the demodulaed channels. We can also read right after the mixing in c1cal model and try differnt poles for integration .
I've also added medm screens in the sensing matrix part of LSC screen. These let you see demodulation of beatnote frequency signals.
I measured electronics noise of WFSs and QPD (of the WFS/QPD, whitening, ADC...) by closing PSL and measuring the error signal. It was needed to put the offset in C1:IOO-MC_TRANS_SUMFILT_OFFSET to 14000 cts (without offset the sum of quadrants would give zero, and 14000 cts is the value when the cavity is locked). For WFS that are RF, if there is intensity noise at low frequencies, it is not affecting the measurement.
In the attachment please find the power spectrum of the error signal when the PSL shutter is on and off.
what's the reasoning behind using df/f_beat instead of df/f_laser ?
I took ~ 7 minutes of XALS beatnote data with the XAUX laser locked to the XARM cavity, and the XARM locked to PSL to develop an allan deviation estimator.
For MICH noise budgeting we measure the input electronics noise which includes the AS55 RFPD, preamp, demod board, the whitening, and the AA filters, and the ADC noises. To do so we simply close the laser shutter and take the spectrum of C1:LSC-AS55_I_ERR_DQ and C1:LSC-AS55_Q_ERR_DQ shown in attachment 1.
Next, we measured the output electronics noise which includes the DAC, dewhitening and AI filters, and coil driver noises. We disabled the BS watchdog and went to 1X4 rack. We measured the spectrum of one of the lemo outputs on the BS coil driver module using an SR785. Attachment 2 shows the spectrum together with the SR785 dark noise.
We measured the AS55 demod board conversion from the amplitude of a 55MHz signal to a demodulated signal. We hooked the unused REFL55 LO into the PD input port on the AS55 demod board.
The REFL55 LO was measured to be 1.84 Vpp. The IQ outputs were: I = 0.86 Vpp, Q = 2.03 Vpp giving an amplitude of 2.205 Vpp. The overall conversion factor is sqrt(0.86**2+2.03**2)/(1.82/2)=2.422.
We also set to measure the loss in the RF cable from AS55 PD to the demod board on 1Y2. REFL55 was connected with a long BNC cable to the input of the cable under test. REFL55 at the input was measured to be 1.466 Vpp and 1.28 Vpp at the output signifying a transmission of 87.6%.
I guess it didn't make sense since f_beat can be arbitrarily moved, but the beat is taken around the PSL freq ~ 281.73 THz. Attachment #1 shows the overlapping tau allan deviation for the exact same dataset but using the python package allantools, where this time I used the PSL freq as the base frequency. This time, I can see the minimum fractional deviation of 1.33e-13 happening at ~ 20 seconds.
The allan variance is related to the beatnote spectral density as a mean-square integral (the deviation is then like the rms) with a sinc window.
this is just the CDS error signal, but is not the electronics noise. You have to go into the lab and measure the noise at several points. It can't be done from the control room. You must measure before and afte the whitening.
We were able to greatly improve the quality of our readings by changing the parameters in the config file (particularly increasing the integration and settle cycles, as well as gradually increasing our excitation signals' amplitude). Attached are the readings taken from the same (the files directly printed by ssh'ing the SR785 (apologies)) - Attachment 1 depicts the graph w/ 30 data points and attachment 2 depicts the graph with 300 data points.
Cici successfully vectfit to the data, as included in Attachment 3. (This is the vectfit of the entire control loop's OLTF). There are two main concerns that need to be looked into, firstly, the manner in which to get the poles and zeros to input into the vectfit program. Similarly, the program works best when the option to enforce stable poles is disabled, once again it may be worth looking into how the program works on a deeper level in order to understand how to proceed.
Just as the servo's individual transfer function was taken, we also came up with a plan to measure the PZT's individual transfer function (using the MokuLab). The connections for the same have been made and the Moku is at the Xend (disconnected). We may also have to build a highpass filter (similar to the one whose signal enters the PZT) to facilitate taking readings at high frequencies using the Moku.
In the morning I took some time to align the AUX beams in the XEND table. Later in the afternoon, I did the same on the YEND table. I then locked the AUX beams to the arm cavities while they were stabilized using POX/POY and turned off the PSL hepa off temporarily (this should be turned on after today's work).
After checking the the temperature slider sign on the spectrum analyzer of the control room I took some out-of-loop measurements of both ALS beatnotes (Attachment #1) by running diaggui /users/Templates/ALS/ALS_outOfLoop_Ref_DQ.xml and by comparing them against their old references (red vs magenta and blue vs cyan); it seems that YAUX is not doing too bad, but XAUX has increased residual noise around and above 100 Hz; perhaps as a result of the ongoing ALS SURF loop investigations? It does look like the OLTF UGF has dropped by half from ~ 11 kHz to ~ 5.5 kHz.
Anyways let this be a reference measurement for current locking tasks, as well as for ongoing SURF projects.
For the PSL HEPA, we wanted it to remain at full speed during the vent, when anyone is working on the PSL, or when there is a lot of dust due to outside conditions or cleaning in the lab.
For NORMAL conditions, the policy is to turn it to 30% for some flow, but low noise.
I think we ought to lock one of the arms on IR PDH and change the HEPA flow settings and plot the arm error signal, and transmitted power for each flow speed to see what's important. Record the times of each setting so that we can make a specgram later
For the proposed construction in the NW corner of the CES building (near the 40m BS chamber), they did a simulated construction activity on Wednesday from 12-1.
In the attached image, you can see the effect as seen in our seismometers:
this image is calculated by the 40m summary pages codes that Tega has been shepherding back to life, luckily just in time for this test.
Since our local time PDT = UTC - 7 hours, 1900 UTC = noon local. So most of the disturbance happens from 1130-1200, presumably while they are setting up the heavy equipment. If you look in the summary pages for that day, you can also see the IM lost lock. Unclear if this was due to their work or if it was coincidence. Thoughts?
[Anchal, Paco, Yuta]
We tried to lock FPMI with REFL55 and AS55 this week, but no success yet.
FPMI locks with POX11, POY11 and ASDC for MICH stably, but handing over to 55's couldn't be done yet.
What we did:
- REFL55: Increased the whitening gain to 24dB. Demodulation phase tuned to minimize MICH signal in I when both arms are locked with POX and POY. REFL55 is noisier than AS55. Demodulation phase and amplitude of the signal seem to drift a lot also. Might need investigation.
- AS55: Demodulation phase tuned to minimize MICH signal in I when both arms are locked with POX and POY. Whitening gain is 24dB.
- Script for demodulation phase tuning lives in https://git.ligo.org/40m/scripts/-/blob/main/RFPD/getPhaseAngle.py
- Locking MICH with REFL55 Q: Kicks BS much and not so stable probably because of noisy REFL55. Offtet also needs to be adjusted to lock MICH to dark fringe.
- BS coil balancing: When MICH is "locked" with REFL55 Q, TRX drops rapidly and AS fringe gets worse, indicating BS coil balancing is not good. We balanced the coils by dithering POS with different coil output matrix gains to minimize oplev PIT and YAW output manually using LOCKINs.
- Locking MICH with ASDC: Works nicely. Offset is set to -0.1 in MICH filter and reduced to -0.03 after lock acquisition.
- ETMX/ETMY actuation balancing: We found that feedback signal to ETMX and ETMY at LSC output is unbalanced when locking with POX and POY. We dithered MC2 at 71 Hz, and checked feedback signals when Xarm/Yarm are locked to find out actuation efficiency imbalance. A gain of 2.987 is put into C1:LSC-ETMX filter to balance ETMX/ETMY. I think we need to check this factor carefully again.
- TRX and TRY: We normalized TRX and TRY to give 1 when arms are aligned. Before doing this, we also checked the alignment of TRX and TRY DC PDs (also reduced green scattering for TRY). Together with ETMX/ETMY balancing, this helped making filter gains the same for POX and POY lock to be 0.02 (See, also 40m/16888).
- Single arm with REFL55/AS55: We checked that single arm locking with both REFL55_I and AS55_Q works. Single arm locking feeding back to MC2 also worked.
- Handing over to REFL55/AS55: After locking Xarm and Yarm using POX to ETMX and POY to ETMY, MICH is locked with ASDC to BS. Handing over to REFL55_I for CARM using ETMX+ETMY and AS55_Q for DARM using -ETMX+ETMY was not successful. Changing an actuator for CARM to MC2 also didn't work. There might be an unstable point when turning off XARM/YARM filter modules and switching on DARM/CARM filter modules with a ramp time. We also need to re-investigate correct gains and signs for DARM and CARM. (Right now, gains are 0.02 for POX and POY, -0.02 for DARM with AS55_Q (-ETMX+ETMY), -0.02 for CARM with REFL55_I with MC2 are the best we found so far)
- Measure ETMX and ETMY actuation efficiencies with Xarm/Yarm to balance the output matrix for DARM.
- Measure optical gains of POX11, POY11, AS55 and REFL55 when FPMI is locked with POX/POY/ASDC to find out correct filter gains for them.
- Make sure to measure OLTFs when doing above to correct for loop gains.
- Lock CARM with POY11 to MC2, DARM with POX11 to ETMX. Use input matrix to hand over instead of changing filter modules from XARM/YARM to DARM/CARM.
- Try using ALS to lock FPMI.
I found the IMC was largely misaligned and was not locking. The WFS feedback signals were saturated and MC2 was still largely misaligned in yaw after resetting the saturation.
It seemed that the MC WFS started to put the large offset at 6:30AM~7:00AM (local).
MC2 was aligned and the lock was recovered then the MC WFS seems working for ~10min now.