At some point tonight we lost our CA connection to c1auxex (which is actually the computer at the X End and controls the ETMX, but has a Y sticker). We could telnet to it, but its puny RAM must have been overloaded with too many EPICS connections that bypassed the CArepeater. I went around and booted some machines and it seems to be back and allowing damping now. Along the way I keyed off the crate to c1auxex a couple of times.
When trying to close the rack door I saw that Charlie/Steve had illegally connected the power cable for the illuminator through the door so that it couldn't close, so I disconnected it so that they can run it properly and feel better about themselves.
Disclaimer: Steve had nothing to do with this connection. I rerouted the cable the correct way. 10-28-2013
** what does this coherence tell us about the noise in the arms ?
8 day minute trend of some of the IMC alignment signals.
That step ~2 days ago in the WFS2 yaw control signal shows that I didn't do such a good job on yaw.
Nic is going to come over some time and give us a new Gouy telescope that let's us have bigger beams on the WFS. At LLO, Hartmut demonstrated recently how bigger beams can reduce offsets somehow...mechanism TBD.
Also, we must angle the WFS and figure out how to dump the reflections at the same time that we rework the table for the telescope.
Steve, can you please put 2 mounted razor dumps near the WFS for this purpose??
Tuesday: Razor dumps are waiting for you.
Steve, can you please put 2 mounted razor dumps near the WFS for this purpose??
I couldn't find any dumps near the WFS. Koji looked. I looked twice. Maybe they are spooky and absorbing all of the light?
The MC alignment was bad and the WFS were making it drift. Koji aligned the beam into the PMC. I then restored the MC suspensions to where they were 8 days ago (back when the transmission and reflection were good). With the WFS OFF, this gave us a MC trans ~ 16000. With WFS ON it goes to 17500 which is about as good as its been over the last 80 days.
I centered the beam on the WFS with the MC unlocked and also centered the beam on the whole WFS path (it was near clipping between WFS 1 & 2). Also for some reason that beamsplitter which steers the beam onto WFS1 is a R=33% (!? why is this not a R=50% ??).
Steve, please swap this out to a BS1-1064-50-1025-45S if we have one sitting around. If not, we want to add this to the CVI purchase list, but not buy until we get a bigger list together.
I also centered this newly aligned beam into the IMC onto the PSL QPDs. We should now use these as a pointing reference for the beam into the IMC.
While doing this I noticed that the beam was almost clipping on the Uniblitz shutter used to block the PSL beam. That shutter is mounted too short and was also not centered horizontally. I removed it for now so that Steve can find a more adjustable mount for it and put it back into play. The beam going into the IMC is BIG, so you have to very careful when centering the shutter. Might be that we cannot leave it at 45 deg and still get a big enough aperture.
Note #3 for Steve: please also replace the mount for last steering mirror into the IMC with a Polanski or a Superman, that black Ultima is no good. Also the dogs must be steel - no aluminum dogs for our sensitive places.
Since the pointing has gone bad again, I went to the PSL to investigate. Found some bad things and removed them:
1) There was a stopped down iris AGAIN in the main beam path, after the newly installed mirror mount. I opened it. Stop closing irises in the beam path.
2) The beam dump for the IOO QPD reflection was just some black aluminum. That is not a real dump. I removed it. We need two razor blade dumps for this.
3) There was an ND filter wheel (???) after one of the PMC steering mirrors. This is not good noise / optics practice. I removed it and dumped the beam in a real dump. No elog about this ?!#?
The attached trend shows the last 20 days. The big step ~2 weeks ago is when Steve replaced the steering mirror mount with the steel one. I don't understand the drift that comes after that.
Today I also spent ~1 hour repairing the Aldabella laptop. Whoever moved it from the PSL area to the SP table seems to have corrupted the disk by improper shutdown. Please stop shutting the lid and disconnecting it from the AC power unless you want to be fixing it. Its now running in some recovery mode. Lets leave it where it is next to the PSL and MC1.
I steered the MC suspensions back to where they were on the trends before the PSL mirror mount swap and then aligned the PSL beam into it by touching the last 2 steel mounts. Once the alignment was good without WFS, I centered the beams on the IOO QPDs. If it behaves good overnight, I will center the unlocked beams on the MC WFS.
Please stay off the PSL for a couple days if you can so that we can watch the drift. This means no opening the doors, turning on the lights, or heavy work around there.
Since I saw that the trend was good, I aligned the MC refl path to the existing IMC alignment:
The reflected spots from the PD are not hitting the dump correctly. WE need to machine a shorter post to lower the dump by ~1 cm to catch the beams.
Nice camera work Steve! I will use these for publicity photos.
Now we need to get one of the video cameras hooked into the MUX so that we can see the flashing and do some image subtraction.
The trend shows a big jolt to the MC1/3 pointing this morning at 8:30.
If not, we will have to put a 'no janitor' sign on all of the 40m doors permanently to prevent mops misaligning our interferometer.
Demod boards should be at 90 deg, not 82.7 or 12 or yellow or ****. We should re-inject the RF and then set the D Phase in the filter module to make the signals orthogonal. 165 is a challenging one to get right, but its worth it since the signals are close to degenerate already.
[Radhika, Paco, Anchal]
I placed a lens in the B-beam path to focus the beam spot onto the RFPD [Attachment 1]. To align the beam spot onto the RFPD, Anchal misaligned both ETMs and ITMY so that the AS and LO beams would not interfere, and the PD output would remain at some DC level (not fringing). The RFPD response was then maximized by scanning over pitch and yaw of the final mirror in the beam path (attached to the RFPD).
Later Anchal noticed that there was no RFPD output (C1:LSC-BH55_I_ERR, C1:LSC-BH55_Q_ERR). I took out the RFPD and opened it up, and the RF OUT SMA to PCB connection wire was broken [Attachment 2]. I re-soldered the wire and closed up the box [Attachment 3]. After placing the RFPD back, we noticed spikes in C1:LSC-BH55_I_ERR and C1:LSC-BH55_Q_ERR channels on ndscope. We suspect there is still a loose connection, so I will revisit the RFPD circuit on Monday.
We proceeded with the TODO items from .
We tried to update the YARM ASS output matrix to appropriately feed back the ETM and ITM T error signals (input beam pointing) to actuate on PR2 and PR3. Using the existing matrix (used for actuating on TT1 and TT2) led to diverging error signals and big drops in transmission. We iteratively tried flipping signs on the matrix elements, but exhausting all combinations of parity was not efficent, since angular sign conventions could be arbitrary across optics.
We decided to go ahead with Yuta's suggestion of dithering on PR2 and PR3 for input beam pointing, instead of ETMY and ITMY. This would simplify the output matrix greatly since dithering and actuation would now be applied to the same optics. Anchal made the necessary model changes. We tried a diagonal identity submatrix (for input pointing) to map each error signal to the corresponding DOF. With the length (L) control loops disengaged, this configuration decreased all T error signals and increased YARM transmision. We then re-engaged the L loops: the final result is that YARM transmission reached just below 1 [Attachment 1].
[Yuta, Anchal, Radhika]
Yesterday we attempted to lock MICH and BHD using the BH55_Q_ERR signal. We adjusted the demodulation phase to send the bulk of the error signal to the Q quadrature. With the LO beam misaligned, we first locked MICH with AS55_Q_ERR. We tried handing over the feedback signal to BH55_Q_ERR, which in theory should have been equivalent to AS55_Q_ERR. But this would not reduce the error and would instead break the MICH lock. Qualitatively the BH55_Q signal looked noisier than AS55_Q.
We used the Moku:Lab to send a 55 MHz signal into the demod board, replacing the BH55 RF input [Attachment 1]. The frequency was chosen to be 10 Hz away from the demodulation frequency (5x Marconi source frequency). However, a 10Hz peak was not visible from the spectra - instead, we observed a 60 Hz peak. Tweaking the frequency offset a few times, we realized that there must be a ~50Hz offset between the Moku:Lab and the Marconi.
We generated an X-Y plot of BH55_Q vs. AS55_DC with the MICH fringe: this did not follow a circle or ellipse, but seemed to incoherently jump around. Meanwhile the X-Y plot BH55_I vs. AS55_DC looked like a coherent ellipse. This indicated that something might have been wrong with the demod board producing the I and Q quadrature signals.
We fed the BH55 RF signal into an unused demod board (previously AS165) [Attachment 2] and updated the channel routing accordingly. This step recovered elliptical I and Q signals with Moku input signal, and their relative gain was adjusted to produce a circle X-Y plot [Attachment 3]. C1:LSC-BH55_Q_GAIN was adjusted to 155.05/102.90=1.5068, and measured diff C1:LSC-BH55_PHASE_D was adjusted to 94.42 deg.
Now BH55_Q_ERR was able to be used to lock the MICH DOF. However, BH55 still appears to be noisy in both I and Q quadratures, causing the loop to feedback a lot of noise.
- Amplify the BH55 RF signal before demodulation to increase the SNR. In order to power an RF amplifier, we need to use a breakout board to divert some power from the DB15 cable currently powering BH55.
We selected a 102K (1 nF) ceramic capacitor and a 100 uF electrolytic capacitor for the RF amplifier power pins. I soldered the connections and reinstalled the amplifier [Attachments 1, 2].
1) please remember to follow the loading and power up instructions to avoid destroying our low noise RF amplifiers. Its not as easy as powering up any usual device.
2) also, please use the correct decoupling capacitors at the RF amp power pins. Its going to have problems if its powered from a distant supply over a long cable.
Here I describe efforts to cancel the AUX laser PZT mechanical resonances from ~200 kHz-400kHz. While these may not be the resonances we end up wanting to suppress, I chose this region as an exercise because it contains the most significant peaks.
The PZT transfer measurement was taken on 09/06 by myself and Anchal. The Moku:Go outputted a swept-sine (1kHz - 1MHz) I sent to the AUX laser PZT. The beat note between the AUX and frequency-doubled PSL was sent to the DFD, and the I and Q channels were routed back as input to the Moku:Go. We also took a calibration transfer function of the Moku:Go, sending output 1 to inputs 1 and 2.
Almost all of the signal was present in the I channel, so I proceeded to use the I data for fitting/next steps. After normalizing the measured frequency response by the calibration measurement (and adjusting for the calculated time delays in the loop - see ), I fit the resulting data using vectfit [Attachment 1]. I supplied the function with n_poles=16, which in reality fit for 16 complex pairs of poles. This complexity of fit was not necessary to capture the 3 prominent peaks, but would likely be needed to fit any of the more heavily-damped resonances.
I chose to invert all fitted poles between 200 kHz and 367 kHz and the corresponding fitted zeros. The result of this filter applied to the original frequency response data can be seen in Attachment 2, where the blue-shaded region contains the inverted poles/zeros. In total, 9 pairs of poles and 9 pairs of zeros were inverted.
[Yuta, JC, Radhika]
PMC input beam was aligned again, bringing transmission from 0.70 to ~0.75. To avoid blocking the PMC refl beam, I found success handling the yaw knob of the first steering mirror from below.
We set out to realign the YARM AUX laser input into the arm cavity.
- We noticed that the GTRY beam was way off the center of the screen, so we went to the vertex table to align the camera.
- The beam spot at GTRY PD was large/divergent, so we shifted the PD closer to the penultimate mirror. We also doubled the PD gain. Transmission went from ~0.3 to ~0.7 (with gain doubled).
- We returned to the YARM end table to finalize alignment with the green PZT steering mirrors. GTRY was maximized to ~0.77.
[Paco, Anchal, Radhika]
We tried to debug why the XARM green laser isn't catching lock with the arm cavity. First I tried to improve alignment:
- Aligned the arm cavity axes by maximizing IR transmission.
- Adjusted M1 and M2 steering mirrors to align the X green beam into the arm. GTRX reached ~0.3.
- At the vertex table, I adjusted the lens in the GTRX path to focus the beam onto the DCPD. This increased GTRX to ~0.7.
- Visually I confirmed that TEM00 of the green laser was flashing in the arm cavity, fairly centered. But it was not catching lock.
We suspected the XARM AUX PZT might be damaged/unresponsive. Paco, Anchal, and I fed several frequency signals to the PZT and looked for a peak in the AUX-PSL beatnote spectra at the expected frequency. We confirmed that the X-arm AUX PZT is responsive up to 12 kHz (limited by ADC samping rate). We have no reason to suspect the PZT wouldn't be responsive at the PDH modulation frequency of 231 kHz.
- Investigate PDH servo box / error signal.
I tested the mixer by feeding it a 300 kHz signal sourced from a Moku:Go. I kept the LO input the same - 231.25 kHz from the signal generator. The mixer output was a ~70 kHz waveform as expected, so demodulation is not the issue in green locking.
Next I'll align the arm cavities with IR and check to see if the green REFL signal looks as expected. If not, we'll have to invesitage the REFL PD. If the signal looks fine, and we now know it's being properly demodulated, the issue must lie further downstream.
I took a transfer function measurement of the XEND PDH servo box, from servo input to piezo output [Attachment 1]. The servo gain knob was set to 10. The swept sine input was 50 mVpp, as to not saturate the servo components. I toggled the local boost on/off for these measurements. With the boost on, coherence was lost from ~100Hz-10kHz, and the saturation light indicators were flashing. I will retake this measurement shortly.
Atachment 2 is from a previous measurement of this PDH servo TF, found here. For this measurement, boost was off and the gain knob was set to 2.0. (If there is a more recent measurement than 2010, please point me to it.)
We retook transfer function measurements of the XEND PDH servo box, this time setting the gain knob to 3.5 to avoid saturation. Once again I toggled the boost on/off. Attachment 1 shows the resulting bode plots, which now resemble the previous measurements circa 2010. This measurement along with the previous one suggest that setting the gain knob too high might affect the loop shape in an unpredictable way. With this accounted for, it seems the PDH servo box is functioning as expected.
Paco suggested that alignment could still be the primary reason why the XEND green laser is not catching lock. With the xarm cavity aligned with IR, I adjusted the M1 and M2 steering mirrors for the green laser, looking at the REFL PD output in an oscilloscope. Paco joined and was able to achieve better mode matching by adjusting mirrors and rotating the half-wave plate. At this point, we could see TEM00 consistently flashing. Green transmission also reached a value of 3, from around 0.5 that I was able to achieve previously (this channel is not normalized).
We broke the loop to make sure the demodulated signal looked as expected, and indeed it resembled a PDH error signal. After reconnecting the loop (with the gain knob set to 3.5), Paco lowered the REFL PD gain by 3 stages and I was able to raise the gain knob to 8 without the servo saturating. I turned boost on and toggled the servo inversion until the laser started to hold lock for a few seconds. The piezo output signal looked reasonable at this point, without clipping on either end.
After some final adjustments to the steering mirrors and the half-wave plate, the green laser can hold lock for around 5 seconds. However it's unclear why the loop isn't more stable, and more updates are to come.
Today I set out to align and lock the YEND green laser, and observe the expected PDH error signal and PZT control signal.
- I took note of PDH servo knobs:
- Disconnected PDH servo PZT output to break loop
- Scanned pitch and yaw of steering mirrors 1 and 2 [Attachment 1] and achieved transmission ~1.2.
- Re-engaged the loop and with TEM00 locked, and did fine adjustment of steering mirrors to maximize transmission to 1.4.
- At this point I broke the loop again to look at the PDH error signal and piezo control signal in an oscilloscope. The error signal had high frequency noise, so the SR560 was used to low pass it before sending it to the scope.
- Once I reconnected the loop and locked to TEM00, I noticed lots of noise in green transmission. Paco took spectra of GTRY and found it was line noise at multiples of 60 Hz. I checked if any BNC shields at the servo box were touching. I shifted the LO frequency from 213.12 kHz to 213.15 kHz, so that the modulation/demodulation was not an integer multiple of 60 Hz. However, these steps didn't get rid of the line noise. To be further investigated.
Next I plan to revisit the XEND AUX loop and try to reach higher lock stability.
I modeled a digital filter for adding a resonance at a desired frequency (Q~100), with a complex-conjugate pole pair and 2 real zeros (2nd order system). Paco suggested I start with a 575 Hz resonance. I loaded the digital filter onto the Moku using the Moku python API (script at labutils/moku/mokuGoPro/mokuDigitalFilter.py). I tested the filter by feeding the Moku a 2 Vpp signal around 575 Hz and looking for some noticeable gain - however the signal passed though unchanged. There might be an additional Moku command for enabling the filter - I'll look into this.
On Monday I aimed to measure the transfer function of the x-arm AUX PDH loop while momentarily locked, with a Moku:Go. I re-aligned the XEND green beam input to the arm cavity with M1 and M2 steering mirrors. I got GTRX to ~1.4 and the TEM00 mode nominally locked (back to ~5 seconds of lock, like last time). Previously Paco and I had achieved transmission of 3, so there was still a good way to go in mode matching.
However I noticed the backwards-propagating beam started to drift relative to the opening of the Faraday isolator (located after the shutter). During manual alignment the backwards beam cleared through the aperture of the FI, but around 5 minutes later it had drifted too high and the beam spot was visible against the FI body, missing the aperture. At this point transmission had dropped to 0, and I realigned the beam to clear through the opening. I tried to further increase transmission but the drift continued to occur within a few minutes of re-alignment. I double checked that there was no dithering of ITMX or ETMX. It seemed there was high residual motion of the ETM, but I was not sure how to decrease this (damping filters were on). I moved on to setting up the TF measurement and decided to return to alignment once the loop excitation was configured.
I chose to inject an excitation from the Moku at the error point of the PDH servo box. I set up the measurement from 100 kHz to 100 Hz, zoomed in around the loop UGF. I passed the mixer output / error signal (alpha) to a T-splitter and sent one copy to input A of an SR560, and routed the Moku excitation to input B. The summed output of the SR560 was sent to the PDH servo input (beta). I passed the second copy of the error signal (alpha) to the Moku, along with the servo input monitor signal (beta) from the PDH box. The Moku measured the transfer function alpha/beta to obtain G_OLG.
I returned to align the green beam and recovered flashing of the TEM00 mode. However when I closed the loop (with excitation), it didn't catch lock. I quickly reverted the loop back to its original state and confirmed that TEM00 locked for ~5 seconds. This made me think the excitation signal was too large relative to the error signal, so I reduced its amplitude to 500 mVpp. This still didn't recover the lock, and at this point the alignment had drifted again so I decided to wrap up.
I am working remotely for the next week, so I can carry out these steps in January.
[Radhika, Anchal, Paco]
AUX PDH Loop Stability
Today I tried aligning the XEND green beam into the arm cavity. Using M1 and M2 steering mirrors, I reached a max transmission ~1.2 of TEM00. In this configuration there was a "donut" mode also flashing, with transmission exceeding that of TEM00. Scanning all 4 degrees of freedom, I couldn't get TEM00 transmission to exceed 1.2, or significantly suppress the other modes. Not great mode matching. (PD gain: 20 dB; servo gain: 10.0.)
In an earlier conversation Paco had recommended I preamplify the green REFL signal with an SR560 before feeding it to the RF mixer. (For yarm this is done with an SR560 gain of 1000.) I did so and raised the gain on the SR560 until it overloaded (PD: 0 dB; SR560: 100). This didn't immediately improve the lock quality, but because alignment still needed work I wasn't surprised.
Anchal suggested the laser mode might be distorted by some lenses further upstream. We noticed some vertical spreading/distortion of the green beam by the first lens after SHG. I adjusted the pitch of an IR steering mirror until it disappeared. We then used the irises by the entrance to the arm cavity to coarsely align the input beam with M1 and M2. This time, fine alignment brought green transmission to just under 4. After slightly adjusting the half-wave plate, green transmission peaked at 4. (This is the highest I've seen it - previous max was 3.) The final combination of PD gain, SR560 gain, and servo gain that maximized transmission and duration of lock was (PD: 10 dB; SR560: 20; servo: 4.0). At its longest, lock on TEM00 was maintained for ~10 seconds.
AUX PDH Loop OLTF
In parallel with above, I was trying to take an OLTF of the loop whenever it was temporarily locked. I set up the measurement configuration like in the previous ELOG (injection at error point). Like last time, the loop would not lock when summing the PDH error signal with the excitation. I confirmed this was true even when I turned off the Moku excitation output. Checking the summed signal output, the Moku was adding an offset to the error signal. Buffering the excitation with an SR560 solved this issue.
The locked mode was switching pretty rapidly during the time I tried to measure the OLTF, and I ended up moving onto trying to improve lock. I might return today to try to take a measurement - I'll post it here.
I returned the half-wave plates on the XEND table back to their original angles, and restored the loop configuration with the PDH servo box. I returned the PD gain to 40 dB (original setting), and set the servo gain knob to 6. This was the region of highest loop stability, with the lock holding for a few seconds (as before). The control signal on the scope did not look intuitive - the peaks of the control signal corresponded with zero crossings of the error signal.
Paco encouraged me to retake transfer function measurements of the PDH servo box. The main takeaway is the PDH servo (boost on) has the expected frequency response at a gain setting of 3 or under, up to 100 mVpp of input. Attachment 1 shows the frequency response at a servo gain of 2, for varying input amplitudes.
The rest of the bode plots correspond to servo gain of 4, 6, 8, and 10 (boost on). The saturation LED would turn on above a gain value of ~3.25, so these results can't be analyzed or interpreted. But it does seem like a steep, low-frequency jump is a signature of the saturated servo. This jump doesn't appear with 10 mVpp input, at least at or above 1 Hz.
Attachments 1 and 2 are two separate lock durations, with x-axes spanning 1 second each. The trace of interest (error signal out of mixer) is Channel 1. Channel 2 contains the control signal outputted by the PDH servo box.
The error signal is contained in a slow-moving envelope at ~4.5 Hz, zoomed in with time cursors in Attachments 3 and 4.
Zooming in further, the error signal has a fast component at ~150 Hz (Attachments 5, 6).
Before taking these traces, I captured the green REFL signal and open-loop PDH error signal shapes (Attachment 7). This error signal linear range spans ~500 mVpp. From looking at this signal it seems like the closed loop contains excess noise.
From considering the above traces and loop calculations I can start to infer the closed loop shape and/or UGF, and what direction we need to move in to recover good locking.
I reconnected the green REFL monitor channel and acquired its spectra when the laser was (mostly) locked. During the collection window, TEM00 would catch lock for a few seconds, drop, and catch again. As of today this is the longest the lock will hold. I'm uploading a screenshot for now but will replace with a proper .pdf spectra image.
There is a peak ~558 Hz and at its second harmonic. Additionally there is a less sharp peak at 760 Hz.
Today while aligning optics at the xend, my knee engaged the AUX laser interlock. I spent some time trying to disable the interlock, and I'm putting the solution here for mainly my reference: pull on the red interlock button. This shorts the two pins on the back of the Mephisto controller.
While discussing xarm AUX laser locking, we noticed excess motion of ITMX. We took spectra of all ITMX sensor outputs and observed a 16 Hz peak, corresponding to the bounce mode of the optic [Attachment 1, 2 (zoomed)]. The UL sensor in particular is sensitive to the bounce DOF. A peak at the roll resonance can also be seen.
To suppress the bounce resonance, we altered the BounceRoll filter in the SUSPOS, SUSPIT, and SUSYAW filter modules. The bounce bandstop filter was widened to the range 15.25-17 Hz [Attachment 2]. The roll bandstop filter was left as is.
After seeing a 560 Hz peak in the XAUX REFL PD signal, I took spectra of the PDH error signal (post-demod) [Attachment 1]. The peak remained, warranting further investigation.
I disconnected the XAUX PDH loop (including PZT modulation) and looked at the beatnote between the PSL (locked to IMC) and the free-running XAUX laser. Attachment 2 shows the PSL-XAUX beatnote alongside the PSL-YAUX beatnote (both around 60 MHz). Note that the YAUX PDH loop was already disconnected, but I added a terminator to the PZT input BNC. Here the 560 Hz peak originating from the XAUX laser is clear. (It is also interesting that the BEATY signal has a significant comb structure compared to BEATX.)
Anchal suggested I tune the XAUX temperature for the frequency difference to switch signs (keeping magnitude at 60 MHz). The result is in Attachment 3 - the 560 Hz peak remained, showing it's not a local temperature-dependent feature.
From this is seems the 560 Hz noise is coming from the XAUX laser.
Yesterday we looked at the out-of-loop PDH error signal of the AUX laser and determined that the LO phase needed significant adjustment. Previously I suspected that the LO phase knob was not actually connected to the circuitry, and we confirmed this looking inside the PDH servo box. Instead we shifted the modulation frequency towards a large PZT resonance in order to obtain a phase shift. (Original frequency: 231.25 kHz.) On a scope it looked like the PDH error signal was improving.
Today I manually swept across modulation frequency in increments of 5 kHz. Qualitatively the PDH signal looked the cleanest between 285 and 290 kHz [Attachment 1]. Here the linear region spans 2V, although it could still be larger in amplitude relative to the side peaks. More fine tuning is still remaining, and at this frequency I'll measure spectra + time series of the err and control signals.
I retook the last spectrum measurement of ALS beatnote fluctuations, with the HEPA on and off. The top plot corresponds to BEATY, and the bottom plot corresponds to BEATX. The 560 Hz peak doesn't seem to be dependent on the status of the HEPA. The noise floor change in BEATY is probably due to drift of the beatnote frequency.
We refereced the old IFOconfigure script (/opt/rtcds/caltech/c1/burt/c1ifoconfigure/C1configurePRM_Carr.sh) to manually configure the LSC screen for PRMI. The final MICH/PRCL gain values used to achieve lock were flipped in sign:
(.snap file ---> final value)
The FM trigger levels (enable/disable) for PRCL and MICH were set to (150/50). The following filter modules were engaged:
During PRMI lock, the POPDC level reached 13130 counts, or 6.8 mW (using calibration of 1.931e6 counts/W). The POPDC counts level was ~1330 with only MICH locked, meaning the PRC gain was ~10. Attachment 1 is an image of REFL, AS, and POP monitors during lock.
Lock was maintained for close to a minute, allowing us to estimate the UGF at around 100 Hz. We used the AA_PRCL_UGF_meas.xml template to measure the loop transfer function. The GPS time (start, end) for a lock stretch with boost on is (1680547905, 1680547933 1364583160, 1364583442).
1. Achieve better angular alignment to keep MICH locked to dark fringe - ASS? Seismic FF?
We calibrated the PRM oplev.
We took calibrated spectra of oplev PIT and YAW (Attachment 1) with the oplev loops open and closed. We see noise suppression up until a few Hz, as expected. The high-frequency floor appears to be at 1e-9 rad/rtHz with this calibration.
Next steps: improving PRMI angular control
To determine the PRM angle-to-length coupling for PRCL, we want to inject pitch/yaw lines into PRM and find the corresponding peaks in the PRCL closed-loop control signal (below loop UGF). Below is a summary of PRMI locking efforts.
- Locked arm cavities, ran YARM, XARM ASS to get PR2/PR3 alignment
- Locked MICH to dark fringe
- Aligned PRM by maximizing drop in REFLDC (reaches 2 when well aligned)
*This was the hurdle when attempting to lock PRMI last week*
- Locked PRMI carrier using configurations in PRMI-AS55_REFL11.yml. There is now a "Lock PRMI (carr) using AS55/REFL11" action on the LSC screen that runs operateLSC.py with the aforementioned yaml file.
- Final DoF gains used: MICH --> 0.8; PRCL --> -0.07. At times BS was being kicked too hard, so we reduced the MICH gain from 1.2.
During PRMI lock, REFLDC was noisy with ~1 Hz fluctuations. We got PRMI to stay locked for a couple minutes at a time. Additionally it couldn't lock the AS port to dark fringe, and it stayed bright while tweaking the BS alignment.
We took spectra of ASDC during a lock stretch to quantify the DC power fluctuations at the AS port [Attachment 1]. The red trace is ASDC with PRMI locked. REF0 (black) is ASDC with MICH locked; REF1/REF2 (blue/green) are ASDC with single-bounce PRCL locked (either ITMX or ITMY misaligned). Note that the PRMI spectrum might need to be normalized by the PRCL gain / PRM transmission: ~ 10/0.057 = 175. The factor difference in ASDC fluctuations between MICH and PRMI for a single test point is ~144 --> with PRMI normalized, the ASDC fluctionations are comparable with MICH.
I measured the OLTF of the XAUX-PDH loop [Attachment 1] now that the green laser is stably locking. I injected an excitation (100mVpp) at the error point of the loop using a Moku:Go. The excitation was summed with the PDH error signal (alpha) using an SR560, and the summed signal (beta) was sent to the PDH servo. (The Moku excitation was buffered with another SR560.) The transfer function beta/alpha was measured on the Moku.
The loop has a UGF of 26.3 kHz, and a phase margin of ~25º (using 1/1-OLG convention).
- Replace PDH servo demod + controller with Moku:Go lock-in amplifier (ensure loop shape is maintained)
- Deploy digital filters to further increase loop bandwidth/phase margin
% Moku:Go Frequency Response Analyzer
% Channel 1, DC coupling, 10 Vpp range, amplitude 100 mVpp, offset 0.000 0 V, phase 0.000 deg
% Channel 2, AC coupling, 10 Vpp range, amplitude 2 mVpp, offset 0.000 0 V, phase 0.000 deg
% Logarithmic sweep from 1.000000 MHz to 99.99997 Hz with 1,024 pts, dynamic amplitude mode off, measuring fundamental, normalization off
% Averaging time 2.00 ms, 1 cycles; Settling time 100 us, 1 cycles
% Acquired 2023-04-11 T 18:45:27 -0700
% Frequency (Hz), Channel 1 Magnitude (dB), Channel 1 Phase (deg), Channel 2 Magnitude (dB), Channel 2 Phase (deg)
1.00000000e+06, 0.0000e+00, 0.0000e+00, -6.8386e+01, 5.7147e+01
9.91037142e+05, 0.0000e+00, 0.0000e+00, -6.9750e+01, 7.4405e+01
9.82154618e+05, 0.0000e+00, 0.0000e+00, -6.6775e+01, 7.2207e+01
We copied the coil balancing procedure found in /scripts/SUS/coilStrengthBalancing/AS1/CoilStrengthBalancing.ipynb to a new PRM directory. After modifying channel names for PRM, we followed the coil balancing procedure:
1. Locked PRY. This was chosen since full PRCL lock was not maintainable for the duration of measurement.
2. Injected 13 Hz line into the butterfly mode and looked for a peak in the LSC PRCL control signal (C1:LSC-PRCL_OUT_DQ). It appeared like the existing coil gains for PRM are already tuned to minimize the but-pos coupling.
3. Injected 13 Hz line into the POS mode and looked for a peak in the PRM oplev pitch and yaw signals (C1:SUS-PRM_OL_PIT_IN1_DQ / C1:SUS-PRM_OL_YAW_IN1_DQ). Like above, the existing coil gains seemed to be tuned to minimize the pos-angle coupling.
The attached spectrum was taken when POS was excited at 13 Hz using LOCKIN1. As expected the PRCL control signal sees the actuation, but we do not see a 13 Hz peak in the oplev pitch/yaw signals.
Attached are the Rayleigh spectrograms of the error/control signal channels associated with the NN nonlinear control of IMC (pitch). The 4-hour data stretch starts at 3:45pm PDT on 4/18. The spectrograms were generated with (stride=5, fftlength=2, overlap=1). PNG images are attached for reference; the generated pdf files were too large to include here or send over email.
The Rayleigh statistic measures nongaussianity of the data.
Tl;dr: Tried to replace of XEND green PDH servo controller with Moku template IIR filter, designed to match PDH servo frequency response. The green laser did not catch lock with this filter.
Attachment 1 plots the measured TF of the PDH servo controller, with boost on and the gain knob set to 7.22 (the current lock configurations). It also plots an 8th order Chebyshev type II low-pass filter, with cutoff frequency and scale chosen to best match the data. (8 was the highest order filter that could be represented by 4 second-order-sections, the maximum allowed by the Moku.) I wanted to test if the XAUX PDH lock could be maintained using this filter as the controller.
The phase of the Chebyshev II filter does not seem to be a good fit to the data, but I wanted to see how far we could get using a template filter already designed for discrete time, and with a magnitude frequency response that approximates the servo. This would bypass having to perform a bilinear transform from the s-domain to the z-domain, which can raise more complications.
The PDH error signal (mixer output) was split and sent to the Moku (input 1) and to the PDH servo input. Closing the loop with the Moku filter output, the green laser was not able to catch lock. Attachment 2 shows the Moku:Go Digital Filter Box configurations, as well as the traces comparing output of the filter and the output of the PDH servo. The red trace is the output of Moku filter, and the blue trace is the output of the PDH servo (input 2) with the loop open (nothing feeding back to laser PZT). The input gain of the filter module was chosen to match the amplitudes of the two control signals. Qualitatively, the filter output contains higher frequency components and preserves the odd polarity of the PDH error signal, compared to the servo output.
I then tried to directly fit the PDH servo TF data. I fit the (analog) poles and zeros of the TF using vectfit. In theory, using a bilinear transform can convert the analog zpk TF to digital zpk, with some frequency pre-warping required. However, vectfit did not return a "normal" transfer function, defined as having at least as many poles as zeros. This caused the bilinear transform to fail.
Next, I will need to use a different fitting package (perhaps IIRrational) to obtain a nicer TF fit, in normal form. Then I can attemp the bilinear transform, confirm it preserves the desired frequency response, and test it out with the Moku:Go.
I retook a transfer function measurement of the uPDH servo closed-loop (using the SR560 to simulate a cavity pole) [Attachment 1]. While some coherence is lost at low frequencies, the servo does not appear to be saturating. Moving forward this measurement is used to design a digital filter that can replicate the uPDH servo box response. *Note: for now the chosen sampling frequency for the discrete filters is 61.04 kHz, the lowest sampling frequency setting of the Moku:Go.
We performed a low-order fit of the TF using vectfit. Vectfit always seems to return 1 more zero than pole - this results in an "improper" transfer function that causes any transformation to the z-domain to fail. Mayank took the fitted zeros and poles from vectfit and manually removed one of the zeros. After transforming the zeros and poles to the z-domain (using control.matlab.c2d), we noticed multiple resonances around 100 kHz that reached 10-20 dB. We decided to estimate poles and zeros by eye instead of using vectfit.
2 zeros and 2 poles were selected by eye to get an estimated fit in the s-domain. Using continous-to-discrete transforms (tried scipy.signal.bilinear and control.matlab.c2d) resulted in unstable controller responses. Attachment 2 shows the original TF measurement with the designed analog filter and the resulting digital filter. The orange 'x's and 'o's mark the poles and zeros used. The digital filter contains many high-frequencies resonances, the most significant at the sampling frequency, 61.04 kHz, reaching 20 dB. Next we tried to manually load the analog ZPK coefficients into Foton. This resulted in the same digital filter as the python s-domain to z-domain functions [Attachment 3].
**UPDATE** Now looking back it's clear that the high-frequency response is limited by the sampling rate. I will redo this for the highest Moku:Go sampling rate of 3.9 MHz.
XAUX laser locked with Moku:Go controller
The analog zeros and poles used to design this filter were:
Attachment 1 shows the resulting digital SOS filter (sampling rate: 3.9 MHz) compared to the measured uPDH servo transfer function (loop closed). The filter design was loaded on the Moku:Go.
I locked the AUX laser with the uPDH servo box and maximized its transmission to ~0.8. I then fed the Moku digital filter output to the PZT and the laser was able to catch lock. However, the max green transmission I could achieve using the Moku controller was 0.5. Attachment 2 is a screenshot of the green transmission ndscope during a lock sequence.
I measured the OLTF of the loop by injecting an excitation at the error point. An SR560 was used to sum the error signal with the excitation. The Moku multi-instrument mode was configured with the Frequency Response Analyzer and Digital Filter Box; it was able to source the excitation and take a transfer function measurement of error signal / (error signal + excitation), while keeping the loop closed.
The OLTF measurement [Attachment 3] points to a loop UGF of ~4 kHz, and phase margin of ~70 deg. An optimal controller would be able to boost the gain around the UGF without changing the phase too much (lag compensator)?
[Radhika, Aaron, Mayank, Paco]
Here I'll describe the setup for T&R measurements of a PR2 replacement, using the Lightwave NPRO laser located at the S/E corner of the PSL table. Our transmittivity prior for this optic is ~20-30 ppm. Aaron and I outlined the setup for measuring the transmittivity of p- and s-polarizations using a chopper wheel for lock-in detection [Attachment 1].
JC found a Lightwave laser controller (in cabinet along YARM). Mayank and Paco helped start it up and adjust the current such that we can align with low power. I used the power meter sitting on PSL to record a quick laser calibration up to 160 mW (plot in Attachment 2 - I can go up to 500 mW in the future). Attachment 3 shows the location of the power meter when these points were collected. For alignment, I set the laser current to 0.94 A (~33 mW).
I removed most optics in the existing setup downstream of the Faraday isolator. I reused 2 PBS cubes and a HWP from the old setup (I still need a QWP). My progress as of 6/6 can be seen in Attachment 4.
After talking with Koji, the layout was revised as in Attachment 1. Instead of using reflection from the first PBS to obtain s-polarization, we'll use transmitted p-polarization and rotate to s with HWP2. This is because the reflection from a PBS isn't pure s-polarization. A second PBS will be used to verify that we have s-polarization (no light transmitted) and then removed for measurement.
I made changes to the setup as seen in Attachment 2. I found an aluminum block to raise the chopper wheel to the beam path. JC drilled a hole through the block so that the chopper wheel can be secured with a zip tie [Attachments 3-4].
Next, I'll mount one of the optics previously used by Anchal for T&R measurements. I'll see if this setup yields consistent results, and then proceed to the PR2 mirror.
The transmissivity setup layout can be seen in Attachment 1. As in Aaron's measurements in cryo lab, the DUT beam passes through the outer spokes of the chopper wheel (frequency ftest). The reference beam passes through the inner spokes of the wheel (fref). The TRANS PD receives both beams; the REF PD only receives the reference beam. The DUT can be removed from the primary beam path. The only change to the initial design was the addition of a few folding mirrors to deal with space constraints.
The formula used for calculating the transmissivity:
Here, Ptrans and Pref refer to the signals on the TRANS and REF PDs. The DUT superscript indicates the presence of the DUT; the 0 superscripts indicates the DUT has been removed from the path. Note that the naive T calculation (below) would not be sensitive to laser intensity fluctuations, or PD gain changes between the DUT and 0 measurements:
I chose to test this setup with a known 90-10 beam splitter mounted at 45º of incidence. However, it's transmissivity calculated from equation (1) came out to be 4.5% - half of the expected value. The naive transmissivity from equation (2) came out to less: 3.5%. It is possible that the angle of incidence was off from 45º, but I would be surprised if it were that significantly off.
Next I set up a simpler DC measurement of the same 90-10 BS [Attachment 2]. After a HWP and PBS, the beam transmitted through the optic to the PD. The transmission through the DUT was 359.1 mV. Removing the DUT, the PD measured 5.266 V. This results in a transmissivity of 6.9% - around double the naive T calculation from the full setup, but still not quite 10%. Looks like there's a mystery factor of 2 somewhere.
After recovering a transmissivity of ~7% for the 90-10 BS (in AC chopper configuration), I moved on to PR2. I mounted PR2 on a rotation stage to control the angle of incidence (AOI) [Attachment 1]. I did a quick check to see if transmitted light through the DUT was measureable on the TRANS PD on its existing gain setting (0 dB). It was, which means the dynamic range of the PD at that gain setting is large enough and the PD gain does not need to be adjusted when placing/removing the DUT. This simplified the transmissivity calculation, since the extra gain reference was no longer needed (the power reference from REF PD remains). The p-polarization setup can be seen in Attachment 2. For each AOI i:
During measurement, for each AOI I measured the spectra of the TRANS and REF PDs with the Moku:Go Spectrum Analyzer [Attachment 3]. I found their spectral peaks at ftest and fref respectively. Attachment 4 shows the p-polarization results for transmissivity vs. AOI. The blue points are calculated from the peak values found by Moku:Go spectrum analyzer; the orange points are calculated from manual peak finding from taking an FFT of the time-series data. Excluding the 45deg AOI, the transmissivies are below 1000 ppm and are even smaller if using the manual FFT analysis. I will tweak the setup for s-polarization measurements this afternoon.
Just realized I wrote out but failed to post this ELOG update before I left last week:
On Thursday 07/05, I simplified the transmission measurement setup since a gain reference signal was no longer needed [Attachment 1]. (This is because the reference beam previously had to hit the TRANS PD, and now this wasn’t needed).
For the s-polarization measurements, I placed a HWP after the PBS to rotate the polarization. I used another PBS to ensure that the light was being reflected and tuned the HWP angle to minimize transmission [Attachment 2].
I’ve included transmissivity plots as a function of angle of incidence, for s and p polarizations [Attachments 3, 4]. For each point, I include the result from the Moku:Go spectral peak finder and the result from performing a FFT on the time series data.
I unmounted the PR2 optic and placed it back in its original case. I’ve left it on my desk if anyone needs it [Attachment 5].
I carried out the same measurements of the Moku:Go ADC and DAC noise to compare to the results from Ando Lab. Instead of a flat filter with 50dB of gain, I used the uPDH box fitted filter shape. I recorded spectral densities with an SR785; results are in Attachment 1. These measurements are consistent with those measured in Ando Lab. I included the SR785 noise floor, measured by terminating its input.
Next I tried to measure the Moku's DAC noise using its Waveform Generator and Digital Filter Box in multi-instrument mode. I generated a single-tone digital signal and passed it to an elliptic bandstop filter (fit tightly around the tone). The filtered signal was measured by the SR785. I performed this measurement with 1 kHz and 10 kHz tones [Attachement 2]. While the fundamental peak is suppressed, we still see it and its harmonics (not DAC noise). The floor of these measurements is consistent with the DAC noise reference from the first test, and we can say that the Moku:Go's DAC noise above 100 Hz is below 1 µV/rtHz.
Picking up from where Reuben left off, I used the Moku:Go in multi-instrument mode to replace the signal generator and uPDH box entirely (Moku:Go setup shown in Attachments 1+2). The lock-in amplifier sourced the modulation for the PZT: 210.5 kHz, 1.4 Vpp amplitude (consistent with 7dBm used by the uPDH box) This LO was used to demodulate the REFL signal input. I coarsely tuned the demodulation phase to 90 degrees until the PDH error signal looked reasonable. The PDH error signal was passed to the digital filter box using the same filter as before. After slightly adjusting the gain knob in the filter module (-8 dB), the lock seemed reasonably stable - transmission screenshot in Attachment 3. I got transmission to ~0.8 with the analog loop today, so it was exciting to see this level maintained with the Moku:Go lock (ignoring oscillations from test mass motion). The system remains locked to TEM00 for 5-10 minutes before mode hopping, which is qualitatively comparable to the analog loop as well.
An OLTF of the Moku:Go loop still needs to be taken. Since the loop error point isn't outputted from the Moku (passed direclty between instruments), I'll need to inject an excitation at the control point. When I fed the control signal to input A of the SR560 and tried to lock with the direct output, the lock would repeatedly break. I noticed that the BATT light of the SR560 was on - I'll repeat this with another SR560, but the lock might be breaking due to an offset. Once this is debugged, I can inject an excitation and measure the loop OLTF.
[Radhika, Yuta, Hiroki, Paco]
During YAUX noise measurements, we also locked IR and green lasers in xarm (using Moku:Go lock-in amplifier + digital controller) to look at ALS noise in XARM. I adjusted the controller gain until green transmission looked tightly controlled (less fuzzy). We measured the out-of-loop ALS X beat note fluctuations at 2 gain levels: -12 dB and -14 dB down from the uPDH box fitted response. These are Attachments 1 and 2 respectively.
Note that there was some mirror motion at 1 Hz that is reflected in the spectral densities (coil balancing of ETMX had just been taking place). The -12dB gain adjustment causes frequency noise ~3x higher than the reference above ~70Hz. The -14dB gain adjustment has higher noise from 70-400 Hz, but has slightly suppressed noise above 1 kHz (relative to -12dB gain).
Moku:Go delay measurements will help clarify if this excess noise is a fundamental limitation of the device, or if there is room for improvement by optimizing the controller or further tweaking the gain/offset.
I measured the Moku:Go and Moku:Pro delay using a Agilent 4395A network analyzer. I considered the PID controller (0 dB gain); the IIR filter box with a 2nd-order low-pass filter; the FIR filter box with 2 coefficients and 201 coefficients (both low-pass). The current XAUX laser lock is done with a Moku:Go using the IIR filter box, so we would expect ~12 µs of delay.